commit c23b62a8c82dc9a44661f4972c97e6afd2c1066d
Author: wangyunlong <3080433063@qq.com>
Date:   Tue Jul 28 14:51:15 2026 +0800

    feat: consolidate legacy workflows into gyxx-flow

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..c0d00c2
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,34 @@
+# Python bytecode and tooling
+__pycache__/
+*.py[cod]
+*.egg-info/
+.pytest_cache/
+.coverage
+htmlcov/
+
+# Build and virtual environments
+build/
+dist/
+.venv/
+venv/
+
+# Runtime data, local state, and logs
+/var/
+*.log
+
+# Local credentials and machine-specific configuration
+.env
+.env.*
+!.env.example
+config/local/
+config/secrets.*
+!config/secrets.example.env
+
+# IDE and operating-system files
+.idea/
+.vscode/
+.DS_Store
+Thumbs.db
+
+# Agent-local learning logs
+.learnings/
diff --git a/README.md b/README.md
new file mode 100644
index 0000000..d2c1103
--- /dev/null
+++ b/README.md
@@ -0,0 +1,88 @@
+# GYXX Flow
+
+GYXX Flow 将原来分散在四个项目中的定时工作流和可执行脚本迁入一个可安装、可迁移的模块化项目：
+
+- `content_marketing`：内容与营销采集
+- `product_commerce`：商品、平台和经营分析
+- `shop_intelligence`：店铺与竞店采集
+- `supply_chain`：供应链采集与通知
+
+运行时源码全部位于 `src/gyxx_flow/modules/<module>/runtime`。新项目不会从
+`D:\yingxiaoyunying`、`D:\shop-data-flow`、`D:\product-collector-analyze-flow`
+或 `E:\auto-flow` 导入或启动代码；这些旧项目仅保留为只读迁移来源和回滚基线。
+
+## 安装与验证
+
+要求 Python 3.12，推荐使用 `uv`：
+
+```powershell
+cd D:\gyxx-flow
+uv sync --python 3.12 --extra test
+uv run pytest
+uv run gyxx doctor --json
+uv run gyxx acceptance status --json
+```
+
+## 运行工作流
+
+```powershell
+uv run gyxx list
+uv run gyxx run product.daily --date 2026-07-27
+uv run gyxx run product.daily --date 2026-07-27 --execute
+```
+
+`run`、`backfill` 和 `scripts run` 默认都是无副作用 dry-run；只有显式加
+`--execute` 才会启动迁入后的本地脚本。
+
+## 运行任意迁入脚本
+
+```powershell
+uv run gyxx scripts list
+uv run gyxx scripts list --module content_marketing
+uv run gyxx scripts run content_marketing:run_all.py --date 2026-07-27
+uv run gyxx scripts run content_marketing:run_all.py --date 2026-07-27 --execute
+```
+
+脚本 ID 格式为 `<module>:<runtime 内相对路径>`。Python、BAT/CMD 和
+PowerShell 入口均受统一项目根、数据根、业务日期、run_id 和 shadow 环境约束。
+
+## 浏览器与外部系统绑定
+
+`config/runtime-bindings.json` 为当前 131 个脚本各自分配固定且唯一的 CDP 端口。
+无论从 workflow、`scripts run` 还是嵌套脚本启动，目标脚本都会重新取得自己的端口和
+`state/browser/<module>/<script>/` 下的 Profile、Cookie、storage state；同一脚本下次
+运行会复用登录态，不同脚本不会共享浏览器状态。
+
+飞书继续走原来的 lark-cli/OpenAPI 身份；数据库继续走现有云端 PostgreSQL；Hermes
+继续走本机服务。运行时会拒绝 localhost 数据库和非本机 Hermes 地址。不要在源码或
+`runtime-bindings.json` 中写入 Cookie、数据库密码、飞书密钥或 Hermes token。
+
+## 数据目录
+
+默认数据根为 `D:\gyxx-flow\var`，可用 `GYXX_DATA_ROOT` 迁移到其他磁盘。
+业务数据按模块写入 `data/raw`、`data/normalized`、`data/curated`、`data/exports`
+和 `data/evidence`；运行状态、日志和临时文件分别进入 `state`、`logs` 和 `tmp`。
+原始 JSON/CSV/XLS/XLSX/下载文件进 raw，清洗结果进 normalized，聚合数据进 curated，
+最终 Markdown/Excel 报告进 exports。浏览器 Profile、Cookie 和 storage state 位于
+`state/browser/<module>/<script>/` 并按脚本复用。现有 `data/raw/<module>/legacy`
+历史数据只读保留，源码与可变数据不混放。
+
+## 定时任务
+
+```powershell
+uv run gyxx schedule plan `
+  --output D:\gyxx-flow\var\schedule-plan\candidate `
+  --start-date 2026-07-27 `
+  --python-executable D:\gyxx-flow\.venv\Scripts\python.exe
+```
+
+该命令只生成 21 个任务定义和审核脚本，不会注册、禁用或修改任何系统任务。
+生产切换必须遵循 `docs/migration-runbook.md` 的逐任务门禁。
+
+## 关键文件
+
+- `design.md`：目标架构和迁移边界
+- `plan.md`：逐项验收清单
+- `config/workflows.json`：工作流到新项目本地入口的映射
+- `config/source-manifests/`：四个来源的逐文件哈希与处置清单
+- `docs/`：部署、运行、迁移和回滚手册
diff --git a/config/runtime-bindings.json b/config/runtime-bindings.json
new file mode 100644
index 0000000..34c60e7
--- /dev/null
+++ b/config/runtime-bindings.json
@@ -0,0 +1,143 @@
+{
+  "schema_version": 1,
+  "cdp_host": "127.0.0.1",
+  "scripts": {
+    "content_marketing:bilibili_comment_scraper.py": 22000,
+    "content_marketing:bilibili_scraper.py": 22001,
+    "content_marketing:chanmama_scraper.py": 22002,
+    "content_marketing:daily_marketing_report.py": 22003,
+    "content_marketing:data/tools/analyze_comments.py": 22004,
+    "content_marketing:data/tools/analyze_note.py": 22005,
+    "content_marketing:data/tools/batch_rescrape_bilibili.py": 22006,
+    "content_marketing:data/tools/batch_rescrape_douyin.py": 22007,
+    "content_marketing:data/tools/batch_rescrape_xiaohongshu.py": 22008,
+    "content_marketing:data/tools/check_status.py": 22009,
+    "content_marketing:data/tools/collect_note_metrics.py": 22010,
+    "content_marketing:data/tools/daily_marketing_report.bat": 22011,
+    "content_marketing:data/tools/daily_run.bat": 22012,
+    "content_marketing:data/tools/daily_run_with_backfill.bat": 22013,
+    "content_marketing:data/tools/db.py": 22014,
+    "content_marketing:data/tools/feishu_comment_batch.py": 22015,
+    "content_marketing:data/tools/feishu_doc_writer.py": 22016,
+    "content_marketing:data/tools/friday_relogin.bat": 22017,
+    "content_marketing:data/tools/friday_relogin_parallel.py": 22018,
+    "content_marketing:data/tools/generate_creator_report.py": 22019,
+    "content_marketing:data/tools/init_db.py": 22020,
+    "content_marketing:data/tools/kill_project_chrome.ps1": 22021,
+    "content_marketing:data/tools/list_project_chrome.ps1": 22022,
+    "content_marketing:data/tools/monday_self_run.bat": 22023,
+    "content_marketing:data/tools/monthly_creator_report.bat": 22024,
+    "content_marketing:data/tools/monthly_summary.bat": 22025,
+    "content_marketing:data/tools/pull_all_tables.py": 22026,
+    "content_marketing:data/tools/rebuild_mapping.py": 22027,
+    "content_marketing:data/tools/relogin_bilibili.py": 22028,
+    "content_marketing:data/tools/relogin_douyin.py": 22029,
+    "content_marketing:data/tools/relogin_pgy.py": 22030,
+    "content_marketing:data/tools/relogin_xiaohongshu.py": 22031,
+    "content_marketing:data/tools/relogin_xingtu.py": 22032,
+    "content_marketing:data/tools/retry_failed.py": 22033,
+    "content_marketing:data/tools/sync_cooperations.bat": 22034,
+    "content_marketing:data/tools/sync_cooperations.py": 22035,
+    "content_marketing:data/tools/sync_metrics_to_cmt_notes.py": 22036,
+    "content_marketing:data/tools/sync_style_categories.py": 22037,
+    "content_marketing:data/tools/validate_mapping.py": 22038,
+    "content_marketing:data/tools/weekly_comment_scrape.bat": 22039,
+    "content_marketing:data/tools/weekly_summary.bat": 22040,
+    "content_marketing:data/tools/write_notes_to_doc.py": 22041,
+    "content_marketing:douyin_comment_scraper.py": 22042,
+    "content_marketing:feishu_mapping.py": 22043,
+    "content_marketing:login_helper.py": 22044,
+    "content_marketing:monthly_summary_all.py": 22045,
+    "content_marketing:pgy_xhs_scraper.py": 22046,
+    "content_marketing:pgy_xhs_scraper_v2.py": 22047,
+    "content_marketing:run_all.py": 22048,
+    "content_marketing:self_bilibili_scraper.py": 22049,
+    "content_marketing:self_douyin_scraper.py": 22050,
+    "content_marketing:weekly_summary_all.py": 22051,
+    "content_marketing:weekly_summary_xingyun2.py": 22052,
+    "content_marketing:xiaohongshu_comment_scraper.py": 22053,
+    "content_marketing:xingtu_scraper.py": 22054,
+    "content_marketing:xingtu_scraper_v2.py": 22055,
+    "product_commerce:aggregate_daily_final.py": 22056,
+    "product_commerce:analyze_style_with_hermes.py": 22057,
+    "product_commerce:backfill_poseidon_sales.py": 22130,
+    "product_commerce:backfill_collect.py": 22058,
+    "product_commerce:backfill_one_day.py": 22059,
+    "product_commerce:check_nine_day_decline.py": 22060,
+    "product_commerce:collect_dy_market_rank.py": 22061,
+    "product_commerce:collect_dy_persona_to_bitable.py": 22062,
+    "product_commerce:collect_erp_yesterday_metrics.py": 22063,
+    "product_commerce:collect_jd_market_rank.py": 22064,
+    "product_commerce:collect_jd_persona_to_bitable.py": 22065,
+    "product_commerce:collect_persona_to_bitable.py": 22066,
+    "product_commerce:collect_sycm_market_rank.py": 22067,
+    "product_commerce:commands/import_daily.py": 22068,
+    "product_commerce:commands/run_alerts.py": 22069,
+    "product_commerce:db/sync_dim_style.py": 22070,
+    "product_commerce:db/sync_sku_master.py": 22071,
+    "product_commerce:dy_audience_profile_collect.py": 22072,
+    "product_commerce:dy_product_scraping.py": 22073,
+    "product_commerce:erp_login_product_analysis.py": 22074,
+    "product_commerce:export_bitable_records.py": 22075,
+    "product_commerce:import_product_daily.py": 22076,
+    "product_commerce:import_product_reviews.py": 22077,
+    "product_commerce:insert_bitable_records.py": 22078,
+    "product_commerce:jd_main_image_collector.py": 22079,
+    "product_commerce:jd_product_data_collector.py": 22080,
+    "product_commerce:jd_self_inventory_sales_collector.py": 22081,
+    "product_commerce:orchestrate_daily_collection.py": 22082,
+    "product_commerce:orchestrate_market_rank_collection.py": 22083,
+    "product_commerce:orchestrate_review_collection.py": 22084,
+    "product_commerce:reapply_erp_override.py": 22085,
+    "product_commerce:rebuild_market_rank_documents.py": 22086,
+    "product_commerce:run_alerts_with_retry.py": 22087,
+    "product_commerce:run_daily_persona.py": 22088,
+    "product_commerce:run_weekly_jd_main_image.py": 22089,
+    "product_commerce:run_weekly_main_image.py": 22090,
+    "product_commerce:scripts/insert_jd_main_image_records.py": 22091,
+    "product_commerce:scripts/insert_main_image_records.py": 22092,
+    "product_commerce:taobao_dmp_item_crowd_insight_screenshots.py": 22093,
+    "product_commerce:taobao_sycm_collect.py": 22094,
+    "product_commerce:taobao_sycm_collect_backfill.py": 22095,
+    "product_commerce:taobao_sycm_products.py": 22096,
+    "product_commerce:taobao_wanxiang_ai_creative_report.py": 22097,
+    "product_commerce:upload_video_to_guanghe.py": 22098,
+    "product_commerce:vendors/dy-data-flow/dy_store_competitor_store_scraping.py": 22099,
+    "product_commerce:vendors/jd-data-flow/jd_data_collector.py": 22100,
+    "product_commerce:vendors/jd-data-flow/jd_peer_product_data_collector.py": 22101,
+    "product_commerce:vendors/jd-data-flow/jd_product_data_collector.py": 22102,
+    "product_commerce:weekly_aggregate.py": 22103,
+    "shop_intelligence:collectors/dy_store_competitor_store_scraping.py": 22104,
+    "shop_intelligence:collectors/jd_data_collector.py": 22105,
+    "shop_intelligence:collectors/jd_peer_store_data_collector.py": 22106,
+    "shop_intelligence:collectors/taobao_sycm.py": 22107,
+    "shop_intelligence:runners/run_peer_store.py": 22108,
+    "shop_intelligence:runners/run_shop.py": 22109,
+    "shop_intelligence:scripts/remove_scheduler.ps1": 22110,
+    "shop_intelligence:scripts/setup_scheduler.ps1": 22111,
+    "supply_chain:orchestrator/mcp_workflow.py": 22112,
+    "supply_chain:orchestrator/monitor.py": 22113,
+    "supply_chain:orchestrator/runner.py": 22114,
+    "supply_chain:orchestrator/scripts/ProductReplenishment.py": 22115,
+    "supply_chain:orchestrator/scripts/PurchaseConfirmation.py": 22116,
+    "supply_chain:orchestrator/scripts/PurchaseOrderUpdate.py": 22117,
+    "supply_chain:orchestrator/scripts/batch_process.py": 22118,
+    "supply_chain:orchestrator/scripts/collect_confirmation.ps1": 22119,
+    "supply_chain:orchestrator/scripts/collect_purchase_order_update.ps1": 22120,
+    "supply_chain:orchestrator/scripts/collect_replenishment.ps1": 22121,
+    "supply_chain:orchestrator/scripts/insert_replenishment_bitable.py": 22122,
+    "supply_chain:orchestrator/scripts/send_card_notification.py": 22123,
+    "supply_chain:orchestrator/scripts/trigger_purchase_order_update.py": 22124,
+    "supply_chain:orchestrator/sql/backfill_history.py": 22125,
+    "supply_chain:run.py": 22126,
+    "supply_chain:scripts/purchase-confirmation.bat": 22127,
+    "supply_chain:scripts/replenishment-alert.bat": 22128,
+    "supply_chain:scripts/replenishment.bat": 22129
+  },
+  "services": {
+    "feishu": "legacy",
+    "postgres": "cloud",
+    "hermes": "local",
+    "hermes_url": "http://127.0.0.1:8642/v1/chat/completions"
+  }
+}
diff --git a/config/schedules.json b/config/schedules.json
new file mode 100644
index 0000000..13d3eb0
--- /dev/null
+++ b/config/schedules.json
@@ -0,0 +1,30 @@
+{
+  "schema_version": 1,
+  "timezone": "Asia/Shanghai",
+  "schedules": [
+    {"workflow_id":"content.metrics.daily","kind":"daily","at":"06:00"},
+    {"workflow_id":"content.cooperations.daily","kind":"daily","at":"09:00"},
+    {"workflow_id":"content.marketing_report.daily","kind":"daily","at":"10:00"},
+    {"workflow_id":"content.relogin.weekly","kind":"weekly","days":["Friday"],"at":"10:00"},
+    {"workflow_id":"content.self_operated.weekly","kind":"weekly","days":["Monday"],"at":"13:00"},
+    {"workflow_id":"content.comments.weekly","kind":"weekly","days":["Sunday"],"at":"12:00"},
+    {"workflow_id":"content.summary.weekly","kind":"weekly","days":["Tuesday"],"at":"10:00"},
+    {"workflow_id":"content.summary.monthly","kind":"monthly","day_of_month":1,"at":"08:00"},
+    {"workflow_id":"content.creator_report.monthly","kind":"monthly","day_of_month":1,"at":"08:30"},
+
+    {"workflow_id":"shop.metrics.weekly","kind":"weekly","days":["Monday"],"at":"12:00"},
+    {"workflow_id":"shop.competitor.weekly","kind":"weekly","days":["Monday"],"at":"12:30"},
+
+    {"workflow_id":"product.daily","kind":"daily","at":"08:40"},
+    {"workflow_id":"product.persona.daily","kind":"daily","at":"10:00"},
+    {"workflow_id":"product.import.daily","kind":"daily","at":"19:00"},
+    {"workflow_id":"product.alert.daily","kind":"daily","at":"23:00"},
+    {"workflow_id":"product.style_analysis.interval","kind":"interval_days","every_days":3,"anchor_date":"2026-07-25","at":"11:00"},
+    {"workflow_id":"product.main_image.jd.weekly","kind":"weekly","days":["Sunday"],"at":"08:30"},
+    {"workflow_id":"product.main_image.weekly","kind":"weekly","days":["Sunday"],"at":"09:30"},
+
+    {"workflow_id":"supply.replenishment_alert.daily","kind":"daily","at":"07:00"},
+    {"workflow_id":"supply.purchase_confirmation.daily","kind":"daily","at":"08:00"},
+    {"workflow_id":"supply.replenishment.weekly","kind":"weekly","days":["Monday"],"at":"08:00"}
+  ]
+}
diff --git a/config/secrets.example.env b/config/secrets.example.env
new file mode 100644
index 0000000..d33a9dc
--- /dev/null
+++ b/config/secrets.example.env
@@ -0,0 +1,5 @@
+GYXX_POSTGRES_DSN=
+GYXX_POSTGRES_PASSWORD=
+GYXX_FEISHU_APP_ID=
+GYXX_FEISHU_APP_SECRET=
+GYXX_HERMES_API_KEY=
diff --git a/config/source-manifests/content_marketing.json b/config/source-manifests/content_marketing.json
new file mode 100644
index 0000000..65edf22
--- /dev/null
+++ b/config/source-manifests/content_marketing.json
@@ -0,0 +1,886 @@
+{
+    "schema_version":  1,
+    "module":  "content_marketing",
+    "snapshot":  "current-filesystem",
+    "files":  [
+                  {
+                      "source_relative_path":  "bilibili_comment_scraper.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/bilibili_comment_scraper.py",
+                      "source_sha256":  "c7f116f30889125c668e7f4dd796e7249ba7ea8621d5ab5fb518028f56c7f8dc",
+                      "target_sha256":  "e03b166f7b3b1b6fb6ebabbb340b0670d2256d8543e74d7b1559a8f862c0725f",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "bilibili_scraper.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/bilibili_scraper.py",
+                      "source_sha256":  "5e95e710f016972fdb886782bdc2eb0f33a0105d23d46fbc0762549c5e42f67d",
+                      "target_sha256":  "8574359731fa2807c1716b41342a9be09e9514f02792c3fed03ed46d36300499",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "chanmama_scraper.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/chanmama_scraper.py",
+                      "source_sha256":  "0e295b4624c2a020ae85e12e6ba1087e670363945d1a855a5625d22ca419e7f1",
+                      "target_sha256":  "497bbf3d0087047af1a3c8b0b866979c5bc67aa304a23a97c69fe3f79d020345",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "collection_completeness.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/collection_completeness.py",
+                      "source_sha256":  "f23b70b949f8adcaf00ee62aa0f9f39314bc623528cad9a77591641da765991d",
+                      "target_sha256":  "f23b70b949f8adcaf00ee62aa0f9f39314bc623528cad9a77591641da765991d",
+                      "transformed":  false,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "creator_task_grouping.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/creator_task_grouping.py",
+                      "source_sha256":  "d4ef307039726b325bbb1267d6e711bdb1c254f06cdbc967c8b4878dfdd7bad8",
+                      "target_sha256":  "d4ef307039726b325bbb1267d6e711bdb1c254f06cdbc967c8b4878dfdd7bad8",
+                      "transformed":  false,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "daily_marketing_report.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/daily_marketing_report.py",
+                      "source_sha256":  "22456210632d34cc6e60e4302641fe3ce581353a41970eb9c5613dbeb135ba31",
+                      "target_sha256":  "61acda5e902b118257251975dcf13332f9894834e4f9f1b030c69066de0e83c1",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "data/config/analyze.env.example",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/config/analyze.env.example",
+                      "source_sha256":  "b60328ef177b795489d0a57957133dedcfd67b38b59603a6edeee1cf9babcee2",
+                      "target_sha256":  "b60328ef177b795489d0a57957133dedcfd67b38b59603a6edeee1cf9babcee2",
+                      "transformed":  false,
+                      "category":  "config"
+                  },
+                  {
+                      "source_relative_path":  "data/config/db.env.example",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/config/db.env.example",
+                      "source_sha256":  "15f9f181ecae6dfc4eedb03a18ed3379ac8a55b83a32ee169b339427acb712d7",
+                      "target_sha256":  "4b28f3d29d504d5c18784105cd989a8604aa18af18a63a5b49499aae94b4e9e7",
+                      "transformed":  true,
+                      "category":  "config"
+                  },
+                  {
+                      "source_relative_path":  "data/config/款式_多维表格_对照.json",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/config/款式_多维表格_对照.json",
+                      "source_sha256":  "8a16d2f697f8db65240623a682610b4f73dddbfb857b9f6579aa408bd00a6d25",
+                      "target_sha256":  "8a16d2f697f8db65240623a682610b4f73dddbfb857b9f6579aa408bd00a6d25",
+                      "transformed":  false,
+                      "category":  "config"
+                  },
+                  {
+                      "source_relative_path":  "data/config/款式_多维表格_对照_自营.json",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/config/款式_多维表格_对照_自营.json",
+                      "source_sha256":  "e46f5ad98dcdef7de64a512f1dcac244daf5c6c607851a8183518c8c8f2f73ce",
+                      "target_sha256":  "e46f5ad98dcdef7de64a512f1dcac244daf5c6c607851a8183518c8c8f2f73ce",
+                      "transformed":  false,
+                      "category":  "config"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/analyze_comments.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/analyze_comments.py",
+                      "source_sha256":  "61677e64d3e864b73573e3e59b7a247bcfddcde93f78f81a20ee27257c326107",
+                      "target_sha256":  "7244b6101ad3e44cb4c7d564728fa007c5710f5dfc5f41f31a4cf2d706e5c24c",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/analyze_note.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/analyze_note.py",
+                      "source_sha256":  "cdf981d21f30d31338b698a31fd44fea97c4c08e04750956ac6f1afb9ed56905",
+                      "target_sha256":  "ee2d1b6560c2939e70330949fbef8bf9958bfcce229055b683fc8c8f6519fe30",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/batch_rescrape_bilibili.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_bilibili.py",
+                      "source_sha256":  "31d015a53d8b25728e687d4a5db78673bd8dbd25e470e6e8c728a215147d00f1",
+                      "target_sha256":  "7b66e93c7f98c5a1210409456473f893be318c36392b4244bce6014d3dba9501",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/batch_rescrape_douyin.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_douyin.py",
+                      "source_sha256":  "261b3b122708deb40d8587f27fada55cc0cd234a2ddd085af4cf2f2e2871e05a",
+                      "target_sha256":  "f53b02362651d7dcdaeba22a903714d95d3173758ac4a5c3b2bf53cab09fb7cd",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/batch_rescrape_xiaohongshu.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_xiaohongshu.py",
+                      "source_sha256":  "6c50845f568b5e51038229873aac977c1bc37f31dbe90f7ce10218d0358422ea",
+                      "target_sha256":  "dee67a2639e6e0ecac53d6a3961f9dc94538e2bea0d41621639a11a5a6e88d6c",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/check_id_filled.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/check_id_filled.py",
+                      "source_sha256":  "69e13df9ded14dc907a8ef13aa98808f9e7df6d18f6a50741a950ebcb826fcd9",
+                      "target_sha256":  "69e13df9ded14dc907a8ef13aa98808f9e7df6d18f6a50741a950ebcb826fcd9",
+                      "transformed":  false,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/check_status.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/check_status.py",
+                      "source_sha256":  "b8aff93d9a4925bda966f1b458f974131827b59a62e85b73a10bac9d6afee072",
+                      "target_sha256":  "29641c917a27c27e0e5a29130100b1ba3611a68855175b58754d54c1211c19cc",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/collect_note_metrics.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/collect_note_metrics.py",
+                      "source_sha256":  "30f40ae47baaf520001a20a0f1ef85472a999fca69127ad06d07bd697d3353bc",
+                      "target_sha256":  "67fe86573dd1a6ae21087987dfd730d677cc472c65adf0c240d5256e8cfe5d22",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/daily_dashboard_analytics.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_dashboard_analytics.py",
+                      "source_sha256":  "6f0bd4b918691255290880aacaf82fd1ea16344204b6d80d5a85689448312985",
+                      "target_sha256":  "736ab32fdc97c7ae576bdcc0351bbf393789441b054d1b14f3c1f3043ab49dc0",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/daily_marketing_report.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_marketing_report.bat",
+                      "source_sha256":  "3523e30935a0536cc6bf3ea3d3a6922cf35e6602d48835392f5383965351b222",
+                      "target_sha256":  "cd7b165b2c6aca34c27c297a70c7c0f767d0972abaa0d502928b2e03a106538d",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/daily_report_card.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_report_card.py",
+                      "source_sha256":  "3229a5b4ec321ade3543a5fe172231af95e1874a690122258866f58740067855",
+                      "target_sha256":  "fe6dd6501098a003848d8db7930b8b9c5abcc5ab1a1f1a16a7d844d7e543ae5c",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/daily_report_charts.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_report_charts.py",
+                      "source_sha256":  "0fc15ec6114bcd46c51d7e21df23f850943a24acce2df408d795f077241c7ec4",
+                      "target_sha256":  "44381624572846b1c16b5a5e8fae9cafbd6ec86cc2918b9f7bf4fe935d3ca11e",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/daily_run.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_run.bat",
+                      "source_sha256":  "b322948fe220c6196c11640e0b7b2f4fdaed9a6b82cc4f77a9112fe3190a01fd",
+                      "target_sha256":  "f5967ebd3002e55962229ec91c08ca0f5fb614e3710a0065e1c733ff2398fdf1",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/daily_run_with_backfill.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_run_with_backfill.bat",
+                      "source_sha256":  "073250aab8faee6a70c9bcf06b32745b1dc2e69d8c66c546cc91f38280ca661b",
+                      "target_sha256":  "e170578755867526a8da40e0a0b9fee1d5d7ed8f41c87980618477d0ea9b7ef8",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/db.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/db.py",
+                      "source_sha256":  "cb2e3c6a8b44d11668e8fdb6fc0acf8b73adf280a4e19b8f17547edc99ef0089",
+                      "target_sha256":  "1a8f293dbf3075ef509babf28e47404b75ac4b791e2a6b7495df5cbd3a01e5d7",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/feishu_comment_batch.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/feishu_comment_batch.py",
+                      "source_sha256":  "2b04fde410eec1dece0b995d0b74bfad4bdc1a53669f19ead2a6fc13ecdd00b7",
+                      "target_sha256":  "148c289e7956070274f4f9c86f26a10c4750c9ad7471ee852dcda2be823e77fe",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/feishu_doc_writer.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/feishu_doc_writer.py",
+                      "source_sha256":  "0ec54d855ae842b52d0897c782d2f4d9d696182bc6b8ffaf2e7973b615ac814f",
+                      "target_sha256":  "9c3f552b9b466d95bdab7f66800f1f9701b1e81eb1d0316a735553238c7e0423",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/friday_relogin.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/friday_relogin.bat",
+                      "source_sha256":  "26b77ae7575cee43ec055e7dc1b380336c4c75a29bbe8a583c6fb979761b2dc3",
+                      "target_sha256":  "6ad19193b928d23852e2173f98ac779b129acd2f6d2f5814417b40f970599809",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/friday_relogin_parallel.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/friday_relogin_parallel.py",
+                      "source_sha256":  "75b48f49aeb4ef01c0f9edaaa24e4dd9633284514e390af1b1e10dc8b9b29a56",
+                      "target_sha256":  "a85c70397a0daae11bfd6df68b5e91ad1f6f0f85ec766101ecafcf80ea12d6a2",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/generate_creator_report.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/generate_creator_report.py",
+                      "source_sha256":  "84f5404d45e05d5d33f61f29c123cb75f764f43ad4b04e83776e98228883dea7",
+                      "target_sha256":  "2c8d2060918885ff7853542f011bbad927df1cd34a5e50d79875dbc04301a673",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/init_db.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/init_db.py",
+                      "source_sha256":  "5dd41fcd437d957b24ecaf4579672dd2317145b44c44bbe308fdcfa8a94ec667",
+                      "target_sha256":  "1b90db140fb0aad6ddd3b6d9aa45c35099b04ee1890c87dd02972c9bc1981058",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/kill_project_chrome.ps1",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/kill_project_chrome.ps1",
+                      "source_sha256":  "30faa3d2e59bcd691bbbf946603da687db865cb14e31ad129b91248d28853ce8",
+                      "target_sha256":  "703749108c739b0dccfe73f8e6c355934146e1d8c29473d9c237fe2c8ac5e9f7",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/list_project_chrome.ps1",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/list_project_chrome.ps1",
+                      "source_sha256":  "f3678a415f8068c97bc3315a25dd5c74e5dffe48e8be766fa85fd7f643c053be",
+                      "target_sha256":  "ff90f3da8c4c79354b694f032a0e2d0953b034dd0864237cd53fa75d5f8cebda",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/migrations/001_creators_and_cooperations.sql",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/001_creators_and_cooperations.sql",
+                      "source_sha256":  "2a482079b9c65147f2f878f7ab2d855cf69a90086ed24da1f696e360b5a2684f",
+                      "target_sha256":  "2a482079b9c65147f2f878f7ab2d855cf69a90086ed24da1f696e360b5a2684f",
+                      "transformed":  false,
+                      "category":  "sql"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/migrations/003_creator_report.sql",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/003_creator_report.sql",
+                      "source_sha256":  "a66b1ec9f9540a3ce67cb62bb0998df1e160664c5cc1dc752dbb24dbab97190c",
+                      "target_sha256":  "a66b1ec9f9540a3ce67cb62bb0998df1e160664c5cc1dc752dbb24dbab97190c",
+                      "transformed":  false,
+                      "category":  "sql"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/migrations/004_style_category.sql",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/004_style_category.sql",
+                      "source_sha256":  "480b1386f6821b22a89018934fe1625994c7c6513b4e7a7afab4fa53ccfb4167",
+                      "target_sha256":  "480b1386f6821b22a89018934fe1625994c7c6513b4e7a7afab4fa53ccfb4167",
+                      "transformed":  false,
+                      "category":  "sql"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/migrations/005_style_product_profile.sql",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/005_style_product_profile.sql",
+                      "source_sha256":  "8f9c4d25681e84e582d3cdcd9108ebb16c2f5b777b766a5c3653a21c0baf6d91",
+                      "target_sha256":  "8f9c4d25681e84e582d3cdcd9108ebb16c2f5b777b766a5c3653a21c0baf6d91",
+                      "transformed":  false,
+                      "category":  "sql"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/monday_self_run.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/monday_self_run.bat",
+                      "source_sha256":  "2806fb1a0c4a2d8a5d0b260da9fe96de30ab3eccea27d28dd2af90c39fc16261",
+                      "target_sha256":  "0f6eace49e9d58122946450449580513f8218284459ded4587553f8021de9018",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/monthly_creator_report.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/monthly_creator_report.bat",
+                      "source_sha256":  "59f6e216ccca8709bc334e18a692744172fcea9cb521061705a45d708d800845",
+                      "target_sha256":  "794dab2943ab7dc0d85e0c8b9284487bb54d813c9a34697e3d1e1c3b521866e5",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/monthly_summary.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/monthly_summary.bat",
+                      "source_sha256":  "7fd4265a38a5726553d26efcb5a68984717d0d57a45e308dd63274cafdcb283e",
+                      "target_sha256":  "ad3c75b7adf7bf21f414472275142bd1dca0f87517cc195846c991145782b0f2",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/pull_all_tables.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/pull_all_tables.py",
+                      "source_sha256":  "be331541a235be1ffb0f1d2f8583d2f5a7aa66ac324631ec6adeb0aaf988dd0c",
+                      "target_sha256":  "741717cb66afba136fd3448a4d2009d5857ea4230b99ab79e1e2090c342567f4",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/rebuild_mapping.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/rebuild_mapping.py",
+                      "source_sha256":  "fe946418ebeb30d982ed667c4b7b08052a3fe9aba291eb62c1e726dc53854a3c",
+                      "target_sha256":  "56b9128e7324383fe072937704854fbd281f7827b90180a8c8daacfd93db0f48",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/relogin_bilibili.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_bilibili.py",
+                      "source_sha256":  "b8dc12539bcd6f1a368d50ead1f5d33d6577a75c92f0b7bafa75a67edd7a73fa",
+                      "target_sha256":  "bbdab7b1c7ac113aa9c2c2cfdfe456654ec4749788465f2517123d11d9ce85d1",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/relogin_douyin.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_douyin.py",
+                      "source_sha256":  "3c2017f8a801528ff17e4c4bae24615256b8788c185f9a797fe87470e35b1645",
+                      "target_sha256":  "13ea919204a7e5bcba61ba1b5373bd36d495a1275bc93a6a8d605c25dd261e9a",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/relogin_pgy.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_pgy.py",
+                      "source_sha256":  "759c6a68a0851b4c87d8febfc5d94925831a3b2d7cdf1e6a6e74dcd52b83ae3f",
+                      "target_sha256":  "2b208c0adb5aca1241714836c9afac5441f6ffe0302022bfd6934a320ed78946",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/relogin_transaction.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_transaction.py",
+                      "source_sha256":  "57ec20de8d99290268ed3d1fc2b66bf3faeaca807d7f3744dd51de6cc93628b2",
+                      "target_sha256":  "57ec20de8d99290268ed3d1fc2b66bf3faeaca807d7f3744dd51de6cc93628b2",
+                      "transformed":  false,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/relogin_xiaohongshu.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_xiaohongshu.py",
+                      "source_sha256":  "fcdea30adecb066c4f3699ded457df0bc763d978248276ddbffa9f18bc9bdf2c",
+                      "target_sha256":  "9314c1ef13ca7d45643a1b1e9b9c21c93fb5c85e253fc4fd7ed0dc4f71b67df1",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/relogin_xingtu.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_xingtu.py",
+                      "source_sha256":  "c9d434cb02a57ebd7b884598563077d733d56e45dad59ed30774eb7e1c668cb0",
+                      "target_sha256":  "80577ab8f539efa72e82bc4752bde8caaf33cde87068c36fd628d9547e227132",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/retry_failed.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/retry_failed.py",
+                      "source_sha256":  "90c6dbc47317cc48e28071d452d75c95913138fc885a5963f9eb58986cd99086",
+                      "target_sha256":  "70e1f51bcf48cdac0f620cb1505bdacc1b23289b6ba9ab73d5383b93b904e5ee",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/schema_gyxx_super_data.sql",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/schema_gyxx_super_data.sql",
+                      "source_sha256":  "7d087858c183e2f8dce90842dec663c42d67a37a61215dc14bee48ee053caaec",
+                      "target_sha256":  "7d087858c183e2f8dce90842dec663c42d67a37a61215dc14bee48ee053caaec",
+                      "transformed":  false,
+                      "category":  "sql"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/style_analyzer.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/style_analyzer.py",
+                      "source_sha256":  "f58aa4ed597c225dcc105fad712e28c2c60a7dcd3edc93cf29b905c3d9e7e3c9",
+                      "target_sha256":  "0c059690e3f62a50139316d12b1bfb967b6bc58fc1dce477363340bc8c6e844f",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/sync_cooperations.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_cooperations.bat",
+                      "source_sha256":  "ce1cc5b9b6a2a037b68121106263c99d3e4c263a2028c71885acc943948ed466",
+                      "target_sha256":  "8e4a8eef39ccd9ea6294d80ce2e5e5028a27192a39637484b3f613d33037640b",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/sync_cooperations.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_cooperations.py",
+                      "source_sha256":  "e2a527a456847c88981fa6f4bba622b6e45f7782c94bf44459e08f9715e59105",
+                      "target_sha256":  "2b2b2f98951d3e608f695e99d39615bee5c84eae04c464373d854b20f8fb1fa7",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/sync_metrics_to_cmt_notes.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_metrics_to_cmt_notes.py",
+                      "source_sha256":  "3b97cab0740c5c97b47a0fbe763acb5391d7c8e47c95f4675720b4f716f6432f",
+                      "target_sha256":  "f46874a7e7b6e50bafd23190c378fc6d898397458fd5b1baad1cf864513128cc",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/sync_style_categories.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_style_categories.py",
+                      "source_sha256":  "931e70550275ffb33161bbc4cbe527d5b8eff49b4265fda53aa9a5f46947f204",
+                      "target_sha256":  "524570df5217d8f93a9e1c5ff5f55b8009d13de2d872c587fdc82aafff08c272",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/v2_filename.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/v2_filename.py",
+                      "source_sha256":  "60bd6da02baa65e80b60f4fc89fd5f76de26f26566e0cf87a4dcc30e077c379e",
+                      "target_sha256":  "60bd6da02baa65e80b60f4fc89fd5f76de26f26566e0cf87a4dcc30e077c379e",
+                      "transformed":  false,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/validate_mapping.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/validate_mapping.py",
+                      "source_sha256":  "1e5b792da893b7dd1a29d6f6a73146a4d25079ac8bf37cf15ab4816d64d72cfe",
+                      "target_sha256":  "8a079b9083aa6fa6a39bc7c893d20a15cc03d938019911f2382e9132f11266fa",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/weekly_comment_scrape.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/weekly_comment_scrape.bat",
+                      "source_sha256":  "caa4118ad2e674d8efc51b0f79844ff45000e5ba995645268d7639abafa423e2",
+                      "target_sha256":  "fb36cc77ab6ba40c1cd0e7690d5993d1b3ebd59398b640be3a8071c196099760",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/weekly_summary.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/weekly_summary.bat",
+                      "source_sha256":  "fb54595955743300b753ed95c6e78914d8a893d0e53c8f633fa1593ad89b54b6",
+                      "target_sha256":  "52a580fb779444854eebe0b0305738548812f33f2f7de4dc731417b77c6199d5",
+                      "transformed":  true,
+                      "category":  "launcher"
+                  },
+                  {
+                      "source_relative_path":  "data/tools/write_notes_to_doc.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/data/tools/write_notes_to_doc.py",
+                      "source_sha256":  "7c2eec1d8e89d9eb4c90e7cb4903eb50233c5b6932c18c727bbc2d211828230d",
+                      "target_sha256":  "14387cf507c860d7cd6d0b536a77aeecf40155ce12675629ea013882669566bb",
+                      "transformed":  true,
+                      "category":  "tool"
+                  },
+                  {
+                      "source_relative_path":  "douyin_comment_scraper.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/douyin_comment_scraper.py",
+                      "source_sha256":  "ec785ead9c95c5de65c00b06212b1f03d0159fed475a95a2b593d3b24e486e43",
+                      "target_sha256":  "0f8af0630e40556943c732866963bc438bd02ca6f93bed282ee22cc378a267d8",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "feishu_mapping.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/feishu_mapping.py",
+                      "source_sha256":  "f97709b80e8498ceb2024855c526e06513df77f45794c083f3d023f025c8acdd",
+                      "target_sha256":  "7fbf462c73befe74b024cf66a0975dda64ec41989619bdd7401846070e7f8f60",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "login_helper.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/login_helper.py",
+                      "source_sha256":  "15c2563e224b7d8e6eb6ba40eb5da8f91125fb6d99f11e6bd028c0dd546812ac",
+                      "target_sha256":  "578e296438034c0b814837a25a901b8c45a4a52fe2927b75e2624c9df0596566",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "monthly_summary_all.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/monthly_summary_all.py",
+                      "source_sha256":  "aa9c50b0a52416f518973d65e707a1428c5bd3c1f7435c674e894ccd4cc33c17",
+                      "target_sha256":  "1d5f10f0c112bf7a57219fd68ef70675812958f8d7086b8105c9483ae728614d",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "pgy_xhs_scraper.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/pgy_xhs_scraper.py",
+                      "source_sha256":  "7d9cd66b80cf602127541cfdd9589f2d8870ca4a25bac3587478482064aae8f4",
+                      "target_sha256":  "1a4e3c2b97071b9a820cce4e2650d6c51230c64ccfdae54ae90f06dc265d6213",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "pgy_xhs_scraper_v2.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/pgy_xhs_scraper_v2.py",
+                      "source_sha256":  "18aef614da9ed34a08e4940267f3ccd4bf8ff687816e7a64dcc7d08a32cc8b82",
+                      "target_sha256":  "c5ba35f2ca700db9da1bded686876da3c1066b06873320f6f052b2010b0fb14e",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "requirements.txt",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/requirements.txt",
+                      "source_sha256":  "64250b6eb39be48a87fcd79cd962306eb825d8c302f0c3f5f73d79928be602ad",
+                      "target_sha256":  "64250b6eb39be48a87fcd79cd962306eb825d8c302f0c3f5f73d79928be602ad",
+                      "transformed":  false,
+                      "category":  "dependency"
+                  },
+                  {
+                      "source_relative_path":  "run_all.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/run_all.py",
+                      "source_sha256":  "ef49508e1f5176438eaec545ad2932751d6dbd08de7f387b768e68dd00f0499d",
+                      "target_sha256":  "513d61d57869d36d5cad2ac78885952373a3c64fbe4d39831aaf88cfa372e998",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "self_bilibili_scraper.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/self_bilibili_scraper.py",
+                      "source_sha256":  "2ef650a16dde61697101bb219a8c3adcd682fec1529bb9c5312dc5b384d83a72",
+                      "target_sha256":  "9ff5cd98f87e5c59e87d70b50652a0c284ab08125b3df9fe21d83f151addc113",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "self_douyin_scraper.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/self_douyin_scraper.py",
+                      "source_sha256":  "1c428a639fca2deb2fb978b9accfc2e9470bf1eb78d60605cb7226ab832ec1b2",
+                      "target_sha256":  "1c7ebe5cfcb74f7f048b372c3664e332f9d923d71e744362f56456eb7b928d5c",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_analyze_note_no_comments.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_analyze_note_no_comments.py",
+                      "source_sha256":  "b929b107b4f11fe1d1a9ba04d0cfbb5efd8b57c70dc6116a6790ddc676987e4c",
+                      "target_sha256":  "ced6f5414f7d80e512057b9e6ac111ebae19645e2736a6002516e31b9b4946b9",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_batch_rescrape_douyin.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_batch_rescrape_douyin.py",
+                      "source_sha256":  "aeb3539f782819112901c4723c736caa147314092bf8df81066727efc5bb836b",
+                      "target_sha256":  "ce2791fcc089b1adf43a176f5f39e453bcca893b62e1fa9fe79a1958e9164531",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_collection_completeness.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_collection_completeness.py",
+                      "source_sha256":  "2e94f353fbd7d97a659a8a1141b69e6fba535f4939a431e433c91612ec656506",
+                      "target_sha256":  "d635e3d72134e7aa7dd43506984f84be14bf369fd758e3308c3919397e0b830f",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_daily_dashboard_analytics.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_dashboard_analytics.py",
+                      "source_sha256":  "14dcec2edca1d065eed3414d8df947ecb6e866cb8eca37db8bc90f307d456f7d",
+                      "target_sha256":  "dde52f4c5a981366452cf28b3b22d35fec3d9ad77f3fe1c06b60cb0e35f32972",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_daily_marketing_report.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_marketing_report.py",
+                      "source_sha256":  "378dfc22ce6cc99638e25bc3c784d796018abfc66ea32f361221b2e6f66a05e0",
+                      "target_sha256":  "378dfc22ce6cc99638e25bc3c784d796018abfc66ea32f361221b2e6f66a05e0",
+                      "transformed":  false,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_daily_marketing_report_schedule.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_marketing_report_schedule.py",
+                      "source_sha256":  "e1811e4220ca6e828df0a283dd044c42c366c8e9567f5190100e25fb1dbca914",
+                      "target_sha256":  "e1811e4220ca6e828df0a283dd044c42c366c8e9567f5190100e25fb1dbca914",
+                      "transformed":  false,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_daily_report_card.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_report_card.py",
+                      "source_sha256":  "196ec479b11177f16cc632e372ba120212290c546e2ea1ef697793b33c028cd7",
+                      "target_sha256":  "15966cfb6db9669fde849ef522823736cb17a2350d143bc522848d5cc3898a38",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_daily_report_charts.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_report_charts.py",
+                      "source_sha256":  "d16e8d8fc20b9d72d2de00d3311b2605070162c656338a2948fd457761fea810",
+                      "target_sha256":  "2375b924efd0858b50904b6cbf17f2129e862c4e3a0fe5c7978a67b79e4b4321",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_friday_relogin_parallel.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_friday_relogin_parallel.py",
+                      "source_sha256":  "b419680793a338e4b3ddbf85ae802e64fbbe8d3b5b5905e9ebcd67cbdc78c0cf",
+                      "target_sha256":  "5f0dc76194918f1819df62455b91c2b0a5274153b2d718cfba782827b94eb471",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_global_creator_scraping.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_global_creator_scraping.py",
+                      "source_sha256":  "fc81781d11726028cc4974b2174ec38eea928c349793a7ace26706c2d432b935",
+                      "target_sha256":  "7418db838558922b9eff2742b669d3b04032befd5d6815e2f27cc6a1ba736359",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_relogin_rollback.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_relogin_rollback.py",
+                      "source_sha256":  "7d139edaa54b5dc5b5962db2bb8338a13cfaaafbde1a990dad14c7d7852dd02a",
+                      "target_sha256":  "7efa7b64ef581f53740ad7ccb602a99b32607e5c2f846f822b225931c3b3073c",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_remote_database_config.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_remote_database_config.py",
+                      "source_sha256":  "9da72ec3457be6fc7b32f89eeb5fcd1a68043ca27b51151ee2b4bf18ed4e1954",
+                      "target_sha256":  "96141776cb1534b0fe929a4246fa812c827ab9e6b1ba32ff4023c985d3129f48",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_retry_failed_completeness.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_retry_failed_completeness.py",
+                      "source_sha256":  "c488f457c1bc9dd5db52bd0eae4a5e3a3bc23e2f0ccb6807e8dfb43859833098",
+                      "target_sha256":  "169fa63a2143cd94f09c8feecce35ae5d4389aec92d8c444ec58fc6ad11f845f",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_sync_style_categories.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_sync_style_categories.py",
+                      "source_sha256":  "cbc5350739c658b90dc90f0992ca34a6ba5e6ecee626edbf8673041fbf8a19d8",
+                      "target_sha256":  "fce38657fd6a0d6e6239f1e665cf1281879ebcd39a7d580f6dcc3d7659c45324",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "tests/test_weekly_tmall_persona.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/tests/test_weekly_tmall_persona.py",
+                      "source_sha256":  "96ad228e0e6bf29bc48a4151ec846755cffb3a4db94a9c282428313ae749a10a",
+                      "target_sha256":  "31a9ed5be826605bac65cb5f5002674e7779e9acae4e6ec9aaec3af89699a135",
+                      "transformed":  true,
+                      "category":  "test"
+                  },
+                  {
+                      "source_relative_path":  "weekly_summary_all.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/weekly_summary_all.py",
+                      "source_sha256":  "f8a2d976378ade00579393a2bb0805281d1b7b500b119b497c69d1b591a9c8aa",
+                      "target_sha256":  "e9fb032cd3fc8b40d91cd98c5aa321e3109e05bb51b65dfa21f895381a45e70a",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "weekly_summary_xingyun2.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/weekly_summary_xingyun2.py",
+                      "source_sha256":  "637361e637e357c5727b9eb5bc7a8fa2eb0e67f819d81aef51672e54a4301d33",
+                      "target_sha256":  "7c22faa98696f8123dbabd7a4cb1bcf099af3e7ff80cc9cc9e4270c2e02f1602",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "xiaohongshu_comment_scraper.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/xiaohongshu_comment_scraper.py",
+                      "source_sha256":  "58bc07bb2fd67897e18a9981f931df8fadf24a6cc5f09d55bd3aaa2e3c6756a4",
+                      "target_sha256":  "b3d6bea737905d8ef533898b8301ed7198d625d3e1e2e847ccacec1402641a34",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "xingtu_scraper.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/xingtu_scraper.py",
+                      "source_sha256":  "a5f101bf7bbd0f3475491eaadb5e42467f3bf3e8c79de343642e6b5b0c5ccfa9",
+                      "target_sha256":  "638a9f675c5bfb74bb697478cbc4ec71ee81f0c36f8c0cf9e62789df6b0a48db",
+                      "transformed":  true,
+                      "category":  "source"
+                  },
+                  {
+                      "source_relative_path":  "xingtu_scraper_v2.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/xingtu_scraper_v2.py",
+                      "source_sha256":  "200f2a1632fcfd5fa41589f3169ed6df65d5b0cc5904a6791f2a1159cd5ef67e",
+                      "target_sha256":  "18574a680709a3abd49a5c7ced66cc7106664ebf7fc6dbcfc4b5eedef01ca451",
+                      "transformed":  true,
+                      "category":  "source"
+                  }
+              ],
+    "intentionally_excluded":  [
+                                   {
+                                       "pattern":  ".git/**",
+                                       "category":  "vcs",
+                                       "reason":  "version-control metadata is not runtime source"
+                                   },
+                                   {
+                                       "pattern":  ".venv/**",
+                                       "category":  "environment",
+                                       "reason":  "machine-specific virtual environments are rebuilt from dependencies"
+                                   },
+                                   {
+                                       "pattern":  "**/__pycache__/**",
+                                       "category":  "cache",
+                                       "reason":  "generated Python bytecode cache"
+                                   },
+                                   {
+                                       "pattern":  "**/*.pyc",
+                                       "category":  "cache",
+                                       "reason":  "generated Python bytecode"
+                                   },
+                                   {
+                                       "pattern":  ".pytest_cache/**",
+                                       "category":  "cache",
+                                       "reason":  "generated test cache"
+                                   },
+                                   {
+                                       "pattern":  ".claude/**",
+                                       "category":  "agent-state",
+                                       "reason":  "local assistant state"
+                                   },
+                                   {
+                                       "pattern":  ".learnings/**",
+                                       "category":  "agent-state",
+                                       "reason":  "local learning state"
+                                   },
+                                   {
+                                       "pattern":  ".superpowers/**",
+                                       "category":  "agent-state",
+                                       "reason":  "local planning state"
+                                   },
+                                   {
+                                       "pattern":  "docs/**",
+                                       "category":  "documentation",
+                                       "reason":  "historical planning and review documents are not runtime dependencies"
+                                   },
+                                   {
+                                       "pattern":  "data/config/*.env",
+                                       "category":  "secret",
+                                       "reason":  "real environment files may contain credentials; examples are copied separately"
+                                   },
+                                   {
+                                       "pattern":  "data/**/*cookies*",
+                                       "category":  "session",
+                                       "reason":  "cookies and backups are sensitive mutable runtime state"
+                                   },
+                                   {
+                                       "pattern":  "data/**/*storage_state*",
+                                       "category":  "session",
+                                       "reason":  "browser storage state is sensitive mutable runtime state"
+                                   },
+                                   {
+                                       "pattern":  ".*chrome_profile*/**",
+                                       "category":  "session",
+                                       "reason":  "browser profiles are sensitive and machine-specific"
+                                   },
+                                   {
+                                       "pattern":  "data/logs/**",
+                                       "category":  "generated-data",
+                                       "reason":  "historical logs are migrated as data, not source"
+                                   },
+                                   {
+                                       "pattern":  "logs/**",
+                                       "category":  "generated-data",
+                                       "reason":  "historical logs are migrated as data, not source"
+                                   },
+                                   {
+                                       "pattern":  "data/v2_results/**",
+                                       "category":  "generated-data",
+                                       "reason":  "historical collection results are migrated as data"
+                                   },
+                                   {
+                                       "pattern":  "data/notes/**",
+                                       "category":  "generated-data",
+                                       "reason":  "collected note data is not source"
+                                   },
+                                   {
+                                       "pattern":  "data/reports/**",
+                                       "category":  "generated-data",
+                                       "reason":  "generated reports are not source"
+                                   },
+                                   {
+                                       "pattern":  "reports/**",
+                                       "category":  "generated-data",
+                                       "reason":  "generated reports are not source"
+                                   },
+                                   {
+                                       "pattern":  "data/summary/**",
+                                       "category":  "generated-data",
+                                       "reason":  "generated summaries are not source"
+                                   },
+                                   {
+                                       "pattern":  "data/tmp/**",
+                                       "category":  "temporary",
+                                       "reason":  "temporary files and generated images are not source"
+                                   },
+                                   {
+                                       "pattern":  "data/qrcode/**",
+                                       "category":  "sensitive-evidence",
+                                       "reason":  "QR screenshots are sensitive runtime evidence"
+                                   },
+                                   {
+                                       "pattern":  "data/feishu_tables/**",
+                                       "category":  "generated-data",
+                                       "reason":  "remote table snapshots are data caches"
+                                   },
+                                   {
+                                       "pattern":  "data/chanmama/**",
+                                       "category":  "generated-data",
+                                       "reason":  "downloaded spreadsheets are raw data"
+                                   },
+                                   {
+                                       "pattern":  "data/tools/*.xml",
+                                       "category":  "scheduler-artifact",
+                                       "reason":  "legacy Task Scheduler exports are retained only as audit evidence"
+                                   },
+                                   {
+                                       "pattern":  "*.md",
+                                       "category":  "documentation",
+                                       "reason":  "root reports and notes are not runtime source"
+                                   },
+                                   {
+                                       "pattern":  "*.json",
+                                       "category":  "generated-data",
+                                       "reason":  "root payloads and transient JSON files are not runtime source"
+                                   }
+                               ],
+    "transformation_rules":  [
+                                 {
+                                     "id":  "externalize-credentials",
+                                     "reason":  "replace plaintext credentials in copied configuration/source with environment references"
+                                 },
+                                 {
+                                     "id":  "remove-legacy-project-roots",
+                                     "reason":  "replace legacy project absolute roots in copied launchers and source with runtime-relative or environment-configured paths"
+                                 },
+                                 {
+                                     "id":  "portable-runtime-boundary",
+                                     "reason":  "route mutable state and generated data through GYXX_DATA_ROOT and immutable resources through GYXX_MODULE_ROOT"
+                                 }
+                             ],
+    "generated_files":  [
+                            {
+                                "target_relative_path":  "src/gyxx_flow/modules/content_marketing/runtime/runtime_paths.py",
+                                "target_sha256":  "417848beffe8768c51e83af07eac91bf090d2301d8ff120316db8b86a441f4f3",
+                                "category":  "compatibility",
+                                "reason":  "single portable module/data path boundary added during migration"
+                            }
+                        ]
+}
diff --git a/config/source-manifests/product_commerce.json b/config/source-manifests/product_commerce.json
new file mode 100644
index 0000000..d357958
--- /dev/null
+++ b/config/source-manifests/product_commerce.json
@@ -0,0 +1,972 @@
+{
+    "schema_version":  1,
+    "module":  "product_commerce",
+    "snapshot":  "current-filesystem",
+    "files":  [
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "071286d2fb8b0516f0f22f0d088eda9e4a85ae4227619eb1217087ad0dbe3db6",
+                      "target_sha256":  "3662f41d51938e807fe9d5947ae6790116d14842621a561e384893b43d5cdf2c",
+                      "transformed":  true,
+                      "source_relative_path":  "aggregate_daily_final.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/aggregate_daily_final.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "cff768f79672cfc7db7f6866caf5ab59936b8c51eb2a7d729b81441940124512",
+                      "target_sha256":  "c7875972fe5fcb524e82d051fbc8be45aa76af0d452655a9eb443b86be534f2c",
+                      "transformed":  true,
+                      "source_relative_path":  "analyze_style_with_hermes.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/analyze_style_with_hermes.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "524bff755d6921d64857b7ca6adfe1cf3a64730ddb66ae318751e028f5cf6f10",
+                      "target_sha256":  "dd230b802fc88e16c67abcc8fd08a64c4dbfb68a58b74ea90ee45bad2d4d7870",
+                      "transformed":  true,
+                      "source_relative_path":  "tmp_backfill_poseidon_sales.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/backfill_poseidon_sales.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "4d24f0e0cc65be9c2f6ce4047817794a520801f4e6563fbae2c3174c3d411d72",
+                      "target_sha256":  "9bafe98ebde917c926e718ac35500ad06f8b7750e8a6c5f8cad1409e650c4640",
+                      "transformed":  true,
+                      "source_relative_path":  "backfill_collect.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/backfill_collect.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "d8c093315dfc7825459d0590bf1b19ca2838085b8de4bba2bbd0a18e28bda994",
+                      "target_sha256":  "cc43f845bf2e24050acb949d93bc92b4dff166bcad96c9c4e7fe9de59d06851a",
+                      "transformed":  true,
+                      "source_relative_path":  "backfill_one_day.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/backfill_one_day.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "6fb7cd48665c8cb7cf40dd69693408047a85a6abba3cbd410b4e5e9a9a55e9ba",
+                      "target_sha256":  "5feb41b46c54813ef7ae8dc6f7016d792b4e059db0046332d13c3cd16b5c18eb",
+                      "transformed":  true,
+                      "source_relative_path":  "check_nine_day_decline.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/check_nine_day_decline.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "a8350674db09c8d32f3a49dbef94f52b6a11e8b27223446ea21edc62bf00ec3e",
+                      "target_sha256":  "f84b69cb0a8a731703e3cd4a1fa2229e040e6d6975ac5373328ef84c9ebcb918",
+                      "transformed":  true,
+                      "source_relative_path":  "collect_dy_market_rank.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/collect_dy_market_rank.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "abd4c82cbd6c41e111e213c174d2fdeba1214568496eaf1dede24bf374b3b052",
+                      "target_sha256":  "886d2347d3a4c045d835440f8eb09caa975bf47d3c6baae7a0000606c5081a65",
+                      "transformed":  true,
+                      "source_relative_path":  "collect_dy_persona_to_bitable.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/collect_dy_persona_to_bitable.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "c7b54d941ab7f056e2d0ea4355543ff4a5c34ad986c8ed32145e988ac9501ea2",
+                      "target_sha256":  "78ca8f8f932a1ab7e26eeca75ec469ed3fef0252618e405aad8418d79b6c69a9",
+                      "transformed":  true,
+                      "source_relative_path":  "collect_erp_yesterday_metrics.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/collect_erp_yesterday_metrics.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "bddf1601c4c6febd9aae5774c8902276b635f478ea2852ba2b66dd79bd6a0377",
+                      "target_sha256":  "31cba147de9defb5496f27590c9ecc1c57866de76bccebb611f1dc2873077a19",
+                      "transformed":  true,
+                      "source_relative_path":  "collect_jd_market_rank.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/collect_jd_market_rank.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "f9a48c5f89b18472dc3993e340ad858bfecb2fcf59427c7e70db37512c17a6f7",
+                      "target_sha256":  "01b07d35661e16fc5d888b11f92f316e8e89552f6d61e1172f79eb6b80404c8d",
+                      "transformed":  true,
+                      "source_relative_path":  "collect_jd_persona_to_bitable.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/collect_jd_persona_to_bitable.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "3c1858a6ab4f7ce480b3a5f2076e6ec79400e9bfaa75377488c3629e707c7963",
+                      "target_sha256":  "d861741b692074cf1da187edc5adf762b041fba9d4d358143183ab55a0743a97",
+                      "transformed":  true,
+                      "source_relative_path":  "collect_persona_to_bitable.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/collect_persona_to_bitable.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "e93c0a1b579370a66d9e05077e20c9ef015ebf521d58b2f6e88bc00bfc7706a2",
+                      "target_sha256":  "3d9dba6fbb8803fb2f2dcc752fb092de53ab23b73b31012b453b3b301c5d8a7d",
+                      "transformed":  true,
+                      "source_relative_path":  "collect_retry_utils.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/collect_retry_utils.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "1d0b470c2ebea437b6f7ff6de88fcef31d0f43b9f678008a91cf19266139cc5d",
+                      "target_sha256":  "5c3aaa364c3d9041d4fc541e796ea12200ed0977aac31101f68227cffa16cfb8",
+                      "transformed":  true,
+                      "source_relative_path":  "collect_sycm_market_rank.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/collect_sycm_market_rank.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+                      "target_sha256":  "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+                      "transformed":  false,
+                      "source_relative_path":  "config/__init__.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/config/__init__.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "22dcc9d8b2f1cf598a62e4c71df765a952250694cb26b9a73f477c5ca1a29884",
+                      "target_sha256":  "12559b6a7e21b2b49bfe43b50e01e48c02a1b70a45952212f23f77a91ce25581",
+                      "transformed":  true,
+                      "source_relative_path":  "config/style_config_loader.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/config/style_config_loader.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "a8900b9f25db6cf2bef3117a1d2cd86283054fea6f5a8590fd91b5cfbbcaf87e",
+                      "target_sha256":  "0047f8b681131427bed31ea356cf84802536e17d97bc8d88468736be6a200a74",
+                      "transformed":  true,
+                      "source_relative_path":  "db/__init__.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/db/__init__.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "fc19a57bcbc023ab20156820a708d64c5526fad33ef5eb5fdea56c0bb04b84e3",
+                      "target_sha256":  "486607458d204336a04e3776d9db98f5f069a46b75b050ac0f0947cec958b5d7",
+                      "transformed":  true,
+                      "source_relative_path":  "db/sync_dim_style.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/db/sync_dim_style.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "2dc00b88fb5f28c2db619074e97c4a7c446aec29926af9a6890065409c8fac0e",
+                      "target_sha256":  "030344e0ffe26240ee735ccf38f194d4e69f2833bd21e9947b97d0072647910b",
+                      "transformed":  true,
+                      "source_relative_path":  "db/sync_sku_master.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/db/sync_sku_master.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "78c74c0e52e3651b3b042a2095e596ab250e60309c7ae2de905850822aad4922",
+                      "target_sha256":  "f5260df869c0ebf1c908d6ced1525d6ce4451ed57d6d83082e8279238c21a3b5",
+                      "transformed":  true,
+                      "source_relative_path":  "dy_audience_profile_collect.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/dy_audience_profile_collect.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "492916088fa5e8c170461cc9eba97a88ed1cc1e50c7b99e5f6a5e49167fea157",
+                      "target_sha256":  "fde87dfda3cb4561208050fe5b1d6c7b7dc3a5cafa2711b16e3eed808a3ebba7",
+                      "transformed":  true,
+                      "source_relative_path":  "dy_product_scraping.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/dy_product_scraping.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "26f3448e57a08020de0a96d1df3387819afc47f9319b81dbab27f1afc81bfb01",
+                      "target_sha256":  "e2240d11cfce714662de7ef763ea4f964b8d1c4ddada2b1f022938e80b9ab7df",
+                      "transformed":  true,
+                      "source_relative_path":  "erp_login_product_analysis.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/erp_login_product_analysis.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "fcd8db904a14cd5239a5e075e1a888af18c5b24c928608576f6decf4c48c1d8f",
+                      "target_sha256":  "282e9eaee483fde3394d330a62887fb4832cbd876b67a002e27651cd49b8730a",
+                      "transformed":  true,
+                      "source_relative_path":  "erp_metric_overrides.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/erp_metric_overrides.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "b828e1ad566b121bd56f90ef5de9d414cbe8575ec747cd26dfa3e4b66b097a9a",
+                      "target_sha256":  "f0656e0fa873dce9c35cd35adae5817f739f623cf04edca33c45dd8194ee39c7",
+                      "transformed":  true,
+                      "source_relative_path":  "export_bitable_records.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/export_bitable_records.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "725e25fdf9f3c75f0aae17b720cf417a0b31573dbbbf0b174e956e46fe205408",
+                      "target_sha256":  "5c5e262373c62ce37202320ba878ab44b593f0e7443e79919db166a9c6a054fc",
+                      "transformed":  true,
+                      "source_relative_path":  "feishu_doc_native.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/feishu_doc_native.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "906278aa6fb3578d5dab231bdccc2f5a7c6ec8c85810a3e026fff3c7ba16dc90",
+                      "target_sha256":  "4b1308de91597193975e953a56495a61df49798d4fbb644c878a697cb73690c0",
+                      "transformed":  true,
+                      "source_relative_path":  "import_product_daily.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/import_product_daily.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "003a8969da4c50bd500f48fd6b3dcd6378322d87893cfcaf755b3890579e7a52",
+                      "target_sha256":  "50d757be4bfdf8849a43471f419ee460cb844921cf567454de35fcaadbae737f",
+                      "transformed":  true,
+                      "source_relative_path":  "import_product_reviews.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/import_product_reviews.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "057a01e8a22cdd2d1806d290da207209954af89f7da6339e9cdcef3bc0297e63",
+                      "target_sha256":  "97a0e16dc9e6f42bce842d05f6fa85601fa8576c703deb9ce825e7ec981b3173",
+                      "transformed":  true,
+                      "source_relative_path":  "insert_bitable_records.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/insert_bitable_records.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "e0f016bafb3f527f1cae46eaf3ef31510184128aaaa7f52024a7c3f978eec42b",
+                      "target_sha256":  "02306b910736495b4cbd32aab914d876c60c3745f7fbc7daf3f1416267b07b6b",
+                      "transformed":  true,
+                      "source_relative_path":  "jd_main_image_collector.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/jd_main_image_collector.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "d99ce71a008128549ac83fc3da52d88540b74538e3fe5c85d47c3b985e414353",
+                      "target_sha256":  "c0a5597c4c9d69a6fc382e8cdf1a4cb5b876d3996b0ad30f394386fa9a08229e",
+                      "transformed":  true,
+                      "source_relative_path":  "jd_product_data_collector.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/jd_product_data_collector.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "314790f06f7b4ebf9b0ddd5681db9c60e79a3ef16f5546d6e8489531a003c545",
+                      "target_sha256":  "66c691258f2fa017a211bf54a609f28d812ec41c4b78c08aded4c73773f461e8",
+                      "transformed":  true,
+                      "source_relative_path":  "jd_self_inventory_sales_collector.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/jd_self_inventory_sales_collector.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "15ac68801e4b9f677bcb31d8d64158b3eecaaae3f264353345ed6e5d5f404d47",
+                      "target_sha256":  "15ac68801e4b9f677bcb31d8d64158b3eecaaae3f264353345ed6e5d5f404d47",
+                      "transformed":  false,
+                      "source_relative_path":  "lark_cli_runtime.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/lark_cli_runtime.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "ec447d182cc186d4610cfee6ae4ce9b9dce924d905c07268b06a3a2ea0332801",
+                      "target_sha256":  "ec447d182cc186d4610cfee6ae4ce9b9dce924d905c07268b06a3a2ea0332801",
+                      "transformed":  false,
+                      "source_relative_path":  "main_image_db.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/main_image_db.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "85a3cbb8ccdb887b125261d3adca84a9a5596b2adf8e19c5ae6dc96e0780b4d5",
+                      "target_sha256":  "85a3cbb8ccdb887b125261d3adca84a9a5596b2adf8e19c5ae6dc96e0780b4d5",
+                      "transformed":  false,
+                      "source_relative_path":  "main_image_paths.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/main_image_paths.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "355f8051935d458703ab1842beefa74ccbb00d0fbc73af4aa6761ee6f3b4b32a",
+                      "target_sha256":  "73acf4f4ddf9a34ac96c03465008e5c6df6be368c95874cdd14dc55e1b374298",
+                      "transformed":  true,
+                      "source_relative_path":  "market_rank_hermes_notification.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/market_rank_hermes_notification.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "4f2621d855dde386d2764dd925bb88a8aa225f819d1d04944a752a0addd7e9ba",
+                      "target_sha256":  "4f2621d855dde386d2764dd925bb88a8aa225f819d1d04944a752a0addd7e9ba",
+                      "transformed":  false,
+                      "source_relative_path":  "market_rank_report_archive.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/market_rank_report_archive.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "37edbfa24dfb126d52992b08fd6aeccae3ee18c83db56426a0fb02934e4541ae",
+                      "target_sha256":  "3f0c03da2dbbb099e2754ec2ee5f041a7e66e265ae3a12a99d3aa7ea7f6ae05a",
+                      "transformed":  true,
+                      "source_relative_path":  "orchestrate_daily_collection.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/orchestrate_daily_collection.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "0ab99843d30156465876e27f339f74f3e48b0d17f75d74cf5bfeedd513aaa787",
+                      "target_sha256":  "54d36499172069bbc58204b5eb9c6029d0fa430c4174d515ddf75ba3c5c23395",
+                      "transformed":  true,
+                      "source_relative_path":  "orchestrate_market_rank_collection.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/orchestrate_market_rank_collection.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "0d0f2c0a1533a09e698cd23aca42662b309c865659b86f9b9522de8207c3fb8b",
+                      "target_sha256":  "692aca1676f54fc1034b029bfa8c44797840ebe7025bf1bc4d0ac2c44c4f4229",
+                      "transformed":  true,
+                      "source_relative_path":  "orchestrate_review_collection.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/orchestrate_review_collection.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "69f376f2c209d8e2df8f154aceb501eab93d7e077ab0278f5ff35c97356ba296",
+                      "target_sha256":  "69f376f2c209d8e2df8f154aceb501eab93d7e077ab0278f5ff35c97356ba296",
+                      "transformed":  false,
+                      "source_relative_path":  "reapply_erp_override.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/reapply_erp_override.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "c5eff1f72da1b47f6d93fc07223a72c960ca636a3db494a1a8cf8d716c076398",
+                      "target_sha256":  "b15212a47d6c8250fb191800586a6b451d8f5e4491f74c4eb148a24fd231f46b",
+                      "transformed":  true,
+                      "source_relative_path":  "rebuild_market_rank_documents.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/rebuild_market_rank_documents.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "47c80d311247f0089a9e30b72679f54875e9c750097c3b71d5aa2ae25ab0a134",
+                      "target_sha256":  "e05630c83ac568cd293219399cada5b2836555286016fd0402c10ff73413eac6",
+                      "transformed":  true,
+                      "source_relative_path":  "run_alerts_with_retry.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/run_alerts_with_retry.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "c9a51f1d1267b398cf260a86e65a8ee49de873d6c95bf9c90db2abb422005cf7",
+                      "target_sha256":  "d2b28d800e6bb5613db25d71ede46c60ff6d0cb93dbc22bc4fe086724301a799",
+                      "transformed":  true,
+                      "source_relative_path":  "run_daily_persona.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/run_daily_persona.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "27cd730400b954d1a063f034789f0d9a510c711ebda1e2de95e5a0010bd4751f",
+                      "target_sha256":  "198a31850529149cc165dc3d9b0772c9e6c3c40733f74d69d6eba50262ba3830",
+                      "transformed":  true,
+                      "source_relative_path":  "run_weekly_jd_main_image.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/run_weekly_jd_main_image.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "af9b436a31c3454dc6e0b34b77876962dab0b477e8a561db49911fdfc24eade3",
+                      "target_sha256":  "9d1d8ccb674b38010cf5cd21ae34736e5bdd318fc2eedca100c7be49b11dffef",
+                      "transformed":  true,
+                      "source_relative_path":  "run_weekly_main_image.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/run_weekly_main_image.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "34d935efb8dff20afd3ab89ef150c5f8db9cee80c6256c7fe4ef0bae2f72ed63",
+                      "target_sha256":  "45dfc7432d00626b1faca4ac4dd0ae78a79a6713397cb805cd6c577b79311892",
+                      "transformed":  true,
+                      "source_relative_path":  "scripts/insert_jd_main_image_records.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/scripts/insert_jd_main_image_records.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "a1b097ead2a3f7e20ca63d7c15828a8d04030b0d6722934c3bf65d9e975fba76",
+                      "target_sha256":  "cafa9ec4a3d30f6dad933d5e5b97945aefbad2d178ec91039127cdc9a6eeab52",
+                      "transformed":  true,
+                      "source_relative_path":  "scripts/insert_main_image_records.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/scripts/insert_main_image_records.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "b3e9addd72d8e31292ed948b67018aa9e212d628d7ba39141a81116515ca6721",
+                      "target_sha256":  "8393a7fdfdb6e2e2d4a844843daecfe1c69dc7673ba06c68e4f07453c75abbdb",
+                      "transformed":  true,
+                      "source_relative_path":  "taobao_dmp_item_crowd_insight_screenshots.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/taobao_dmp_item_crowd_insight_screenshots.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "4474dd7d68117247af87e20bbeb6b7c072368ef0f606de5d3e186f459d34a0f5",
+                      "target_sha256":  "7e3aa6851dea4c0de2890c676ee18cde2923e6c301afcc190a8f7880cb619326",
+                      "transformed":  true,
+                      "source_relative_path":  "taobao_sycm_collect.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_collect.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "a881b02c903dc83c7f252d7f53fa9cd721fb562c690271788858ca5782427500",
+                      "target_sha256":  "55e9b57bfad8a1c69c77611a7716bbfef61c18ba07ed30bfdf6a4d5d40d06bee",
+                      "transformed":  true,
+                      "source_relative_path":  "taobao_sycm_collect_backfill.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_collect_backfill.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "9cafa62a03316fd0222f63addd5c209bb0c20fd0b7affa69cd1e256abfa2aa3a",
+                      "target_sha256":  "2c77b7d2b21b2fba7460e6a348ad896a3dc07dec8e1355ce6e145b6b585fa378",
+                      "transformed":  true,
+                      "source_relative_path":  "taobao_sycm_products.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_products.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "3f0c32c443c87417c43881da69aa7776828ba973465f8d32f50b3dc698b0a4fe",
+                      "target_sha256":  "03f5595b783c0a9201f64c26ca32a84970c7670068f3c7467d239b9de2944fcf",
+                      "transformed":  true,
+                      "source_relative_path":  "taobao_wanxiang_ai_creative_report.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/taobao_wanxiang_ai_creative_report.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "e4d99956d9649d108baeb3e3a2d1969c227622d1037062bd277af66375556d82",
+                      "target_sha256":  "a97ea4c279ccb37fc99d67073d8cb6dae3f376103901c4d39a31c75862fc1899",
+                      "transformed":  true,
+                      "source_relative_path":  "upload_video_to_guanghe.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/upload_video_to_guanghe.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "b8f1d6645f2b3283897eb7543887eb37ab96da6b3321045dc0c4ff7e2b2b4347",
+                      "target_sha256":  "b8f1d6645f2b3283897eb7543887eb37ab96da6b3321045dc0c4ff7e2b2b4347",
+                      "transformed":  false,
+                      "source_relative_path":  "vendors/dy-data-flow/adaptive_selectors.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/adaptive_selectors.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "cceb3861da0e68c686c0c58f64767c3ca2693986bb5006ca7405d3a5a39c4347",
+                      "target_sha256":  "8a947ce42d34ff26bb46c5173af9e37ee9d60eac1b595d2f54b855a881ac2d6c",
+                      "transformed":  true,
+                      "source_relative_path":  "vendors/dy-data-flow/dy_store_competitor_store_scraping.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/dy_store_competitor_store_scraping.py"
+                  },
+                  {
+                      "category":  "source_resource",
+                      "source_sha256":  "d4a786bce99678c07cf0f6f3dfeddd5d08dd0a41d112cc6beaddb98c28d81346",
+                      "target_sha256":  "d4a786bce99678c07cf0f6f3dfeddd5d08dd0a41d112cc6beaddb98c28d81346",
+                      "transformed":  false,
+                      "source_relative_path":  "vendors/dy-data-flow/dynamic_session_src.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/dynamic_session_src.py",
+                      "lint_policy":  "not_a_standalone_module",
+                      "reason":  "upstream DynamicSession implementation excerpt retained as reference source; it starts inside a class body and has no import/module preamble"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "3cf7b95490249306551949683d948523877c15d6d0f506edc69adda9dea15b4f",
+                      "target_sha256":  "3cf7b95490249306551949683d948523877c15d6d0f506edc69adda9dea15b4f",
+                      "transformed":  false,
+                      "source_relative_path":  "vendors/jd-data-flow/collection_progress.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/collection_progress.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "ce37281fcd13e9f94e345a3f659baa8a95d7a95c7a61440ab14223e7a61c88ec",
+                      "target_sha256":  "882334c4739624e7f34269596785eeabefb7d89ae761cba6f35781cc5773367b",
+                      "transformed":  true,
+                      "source_relative_path":  "vendors/jd-data-flow/config.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/config.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "6daa16c03989fb2231e54545726f5aa488e4ba80b9b8aa6f937cb9a2fa13ff35",
+                      "target_sha256":  "14d30692ff01f140bbb38eaf453d69c1b3ba010cb8a35d73e3ee0ca8f85ea8a1",
+                      "transformed":  true,
+                      "source_relative_path":  "vendors/jd-data-flow/jd_data_collector.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_data_collector.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "a53cc38b0ddeaa0065bc27a1d137ce16c2873ff9bec712483ca87bcedd6baa3a",
+                      "target_sha256":  "7f5f9ee0f9845e4b8cd384246842162ed3a17a57a68f39d0d3275fa13d6d4e7f",
+                      "transformed":  true,
+                      "source_relative_path":  "vendors/jd-data-flow/jd_peer_product_data_collector.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_peer_product_data_collector.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "fc65c65939c0324cae1555631bbb0d1f7ce852825195ce14322e7002a2479f61",
+                      "target_sha256":  "d69139a41d0941114534d94cd4054be7c8ecedc2c09fd79ba07953e160621d95",
+                      "transformed":  true,
+                      "source_relative_path":  "vendors/jd-data-flow/jd_product_data_collector.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_product_data_collector.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "fb26955e7a6e8c0c9b861f60ef4adf528507e1682c66690857ee8cdc7d4742c8",
+                      "target_sha256":  "fb26955e7a6e8c0c9b861f60ef4adf528507e1682c66690857ee8cdc7d4742c8",
+                      "transformed":  false,
+                      "source_relative_path":  "vendors/jd-data-flow/state.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/state.py"
+                  },
+                  {
+                      "category":  "production_source",
+                      "source_sha256":  "b5319c920fa2498dcdb2789788c42e705dcfd841db8865b0424555cc105d8d24",
+                      "target_sha256":  "c56c820fdf482093aa8aa9cf4e43d6b43b46e048ebacf6c770b31329eb37e021",
+                      "transformed":  true,
+                      "source_relative_path":  "weekly_aggregate.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/weekly_aggregate.py"
+                  },
+                  {
+                      "category":  "runtime_resource",
+                      "source_sha256":  "589679874b75643d69ea44522e2fe40914a396506ff8e3232150d7e2a96ae2f9",
+                      "target_sha256":  "ec25e0af5f2a0d1226f60c0cd8833adb91700e0e90b258a22384cbd8f8f3d449",
+                      "transformed":  true,
+                      "source_relative_path":  ".env.example",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/.env.example"
+                  },
+                  {
+                      "category":  "runtime_resource",
+                      "source_sha256":  "af4df439f109549a2d287514d24193cb046800d23a39cabcc070b8c10d76a52a",
+                      "target_sha256":  "af4df439f109549a2d287514d24193cb046800d23a39cabcc070b8c10d76a52a",
+                      "transformed":  false,
+                      "source_relative_path":  "bitable_main_image_map.json",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/bitable_main_image_map.json"
+                  },
+                  {
+                      "category":  "runtime_resource",
+                      "source_sha256":  "51102b78d1f44b9dfb60f8204a0d5b2403cd8651061e7607c264c6ef730defa2",
+                      "target_sha256":  "51102b78d1f44b9dfb60f8204a0d5b2403cd8651061e7607c264c6ef730defa2",
+                      "transformed":  false,
+                      "source_relative_path":  "bitable_style_map.json",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/bitable_style_map.json"
+                  },
+                  {
+                      "category":  "runtime_resource",
+                      "source_sha256":  "277f80393a7ead72b260b56173471e924e139db0ba60e1ae534d85cc1ef5f9cc",
+                      "target_sha256":  "bd0f4087fc061555cb8d2bab485bc0d98417dc1816d210bd58bf7ff6801cff39",
+                      "transformed":  true,
+                      "source_relative_path":  "config/auto-flow-config.example.json",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/config/auto-flow-config.example.json"
+                  },
+                  {
+                      "category":  "runtime_resource",
+                      "source_sha256":  "017d929ff5041d893bfbde874fda207ecc7d8beb65068263d5b36bce1d11e904",
+                      "target_sha256":  "017d929ff5041d893bfbde874fda207ecc7d8beb65068263d5b36bce1d11e904",
+                      "transformed":  false,
+                      "source_relative_path":  "db/schema.sql",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/db/schema.sql"
+                  },
+                  {
+                      "category":  "runtime_resource",
+                      "source_sha256":  "b8acd08b97283a1163210d26c2a35b182e1920d2edbad6864226ad24e73ed5b3",
+                      "target_sha256":  "b8acd08b97283a1163210d26c2a35b182e1920d2edbad6864226ad24e73ed5b3",
+                      "transformed":  false,
+                      "source_relative_path":  "styles_input.json",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/styles_input.json"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "88988821d581872afe162853c3e4569e373555a6df5bf8cd7724977a1b398396",
+                      "target_sha256":  "88988821d581872afe162853c3e4569e373555a6df5bf8cd7724977a1b398396",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_db_config.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_db_config.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "1398806a524beaa203df9b27eed7652180c7ac427522220e88905d9cd827457f",
+                      "target_sha256":  "1398806a524beaa203df9b27eed7652180c7ac427522220e88905d9cd827457f",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_dy_market_rank.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_dy_market_rank.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "13774e9c788a495f92a251f01ccccda5a826c951ec7f60808ad88adcba787262",
+                      "target_sha256":  "13774e9c788a495f92a251f01ccccda5a826c951ec7f60808ad88adcba787262",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_dy_session_reuse.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_dy_session_reuse.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "4041e1104d5adbb92703b88e457795ac221bc7bc4fe4aa2cb72dff08f0c4ab6b",
+                      "target_sha256":  "4041e1104d5adbb92703b88e457795ac221bc7bc4fe4aa2cb72dff08f0c4ab6b",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_erp_metric_overrides.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_metric_overrides.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "cc0d0cb1e666a74803fdaf84bafdd694f51c571c31131e137ff36a39a9db66a7",
+                      "target_sha256":  "cc0d0cb1e666a74803fdaf84bafdd694f51c571c31131e137ff36a39a9db66a7",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_erp_no_data_freshness.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_no_data_freshness.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "504a33d68dc28b93d00bdf720a1551048079f76c97e1ad92359709c460d78c68",
+                      "target_sha256":  "504a33d68dc28b93d00bdf720a1551048079f76c97e1ad92359709c460d78c68",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_erp_slow_skip_codes.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_slow_skip_codes.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "99bf63855624d4f057351d140c6c33179111e8b59f688c3cd75b91fd0b856a5f",
+                      "target_sha256":  "99bf63855624d4f057351d140c6c33179111e8b59f688c3cd75b91fd0b856a5f",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_feishu_doc_native.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_feishu_doc_native.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "3d348a8250a2910f6a1c6504e9d1e53b0493d40bdc6c1fd1c8d0404c78681dc6",
+                      "target_sha256":  "3d348a8250a2910f6a1c6504e9d1e53b0493d40bdc6c1fd1c8d0404c78681dc6",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_guanghe_metadata.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_guanghe_metadata.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "e8848b95c28a822b9fc92b5cd8ad75716b80d00fbac1844bfad9b7f877ea0173",
+                      "target_sha256":  "e8848b95c28a822b9fc92b5cd8ad75716b80d00fbac1844bfad9b7f877ea0173",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_guanghe_store_routing.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_guanghe_store_routing.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "a649d4ee447c41f458a42ca27d288c974188bf30ded33e9157c2b71883c074de",
+                      "target_sha256":  "a649d4ee447c41f458a42ca27d288c974188bf30ded33e9157c2b71883c074de",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_import_product_daily.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_import_product_daily.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "3c07d95d858c67b9541c0df91735c872a6abcae56dd08cdac209bc92475a413d",
+                      "target_sha256":  "3c07d95d858c67b9541c0df91735c872a6abcae56dd08cdac209bc92475a413d",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_jd_enter_shangzhi.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_jd_enter_shangzhi.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "dd82a77378c84cc8921e88ad43a90675bbf1098aaa702e0ffced377a097ce957",
+                      "target_sha256":  "dd82a77378c84cc8921e88ad43a90675bbf1098aaa702e0ffced377a097ce957",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_jd_market_rank.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_jd_market_rank.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "820e17b910b62a213f0bebafcff72998c2c978307c97f976c074276150ce7c6a",
+                      "target_sha256":  "820e17b910b62a213f0bebafcff72998c2c978307c97f976c074276150ce7c6a",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_main_image_concurrency.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_main_image_concurrency.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "0f55af81ae11567c3c5d7943ab74598dd994a7ff193a9e98e7024c7dc4a3a769",
+                      "target_sha256":  "0f55af81ae11567c3c5d7943ab74598dd994a7ff193a9e98e7024c7dc4a3a769",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_market_rank_hermes_notification.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_hermes_notification.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "f21d28753f929daf0808fb8147a6654f6b52efc599b8b13248328b9452c37b69",
+                      "target_sha256":  "f21d28753f929daf0808fb8147a6654f6b52efc599b8b13248328b9452c37b69",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_market_rank_report_archive.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_report_archive.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "219630aaa213d2a671acf71acb271cb955b860fa56819db4ccbdb4b28a7d3662",
+                      "target_sha256":  "219630aaa213d2a671acf71acb271cb955b860fa56819db4ccbdb4b28a7d3662",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_market_rank_workflow.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_workflow.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "e43712d0eba0213044b685c6162da983be91ee083f7b12ec44a61c5525fe9e36",
+                      "target_sha256":  "e43712d0eba0213044b685c6162da983be91ee083f7b12ec44a61c5525fe9e36",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_persona_launcher.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_persona_launcher.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "f9cfbb49fd313444790ac194fbf980dae21912081b9f26ad8e3850912126f09e",
+                      "target_sha256":  "f9cfbb49fd313444790ac194fbf980dae21912081b9f26ad8e3850912126f09e",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_style_analysis_orchestration.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_style_analysis_orchestration.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "17761a5ba93a5387ffa67662424955178058aeec708bbcd3eb0341a859d8a1f1",
+                      "target_sha256":  "ed9cce2ef8a2a6c8efd41ef4eaae915927e4842b114885103526f552296fa8a6",
+                      "transformed":  true,
+                      "source_relative_path":  "tests/test_style_config_loader_erp_source.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_style_config_loader_erp_source.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "55f55a9658453e6751ed055c6b2d1a43538012ce715835bbef59d88203211696",
+                      "target_sha256":  "34950dee4733e8292f02bd8731de77d9343fc298fa7cf1c72a4afbbc336d5d2b",
+                      "transformed":  true,
+                      "source_relative_path":  "tests/test_tmp_backfill_poseidon_sales.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_backfill_poseidon_sales.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "3267f9a366af11ad40e8b8bc80d2c42e08b5a7e84e2c71f265c404389c2cb2c3",
+                      "target_sha256":  "5747d14b0e6c08dbd2f1e83ce83a649c8e926b495d01483c30bc3841cc8148be",
+                      "transformed":  true,
+                      "source_relative_path":  "tests/test_sycm_market_rank.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_sycm_market_rank.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "f7de702d7d1e641c8778970d5cb1f4e827a2e0ee9b20db2c19f0ed4fd8230e16",
+                      "target_sha256":  "f7de702d7d1e641c8778970d5cb1f4e827a2e0ee9b20db2c19f0ed4fd8230e16",
+                      "transformed":  false,
+                      "source_relative_path":  "tests/test_tm_persona_recovery.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_tm_persona_recovery.py"
+                  },
+                  {
+                      "category":  "regression_test",
+                      "source_sha256":  "5598dcbf6d9a30fb3dc4133e983ed78c2860d90162c6f7750509d4533466917a",
+                      "target_sha256":  "b35755ae421db982d162895ddbb3182977cb16a268587e836c2dc4f2703d6e37",
+                      "transformed":  true,
+                      "source_relative_path":  "tests/test_wanxiang_report_template.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_wanxiang_report_template.py"
+                  },
+                  {
+                      "category":  "launcher_provenance",
+                      "source_sha256":  "c7ad2a359730d8c4d449b0e37e5a4fb2940b4e8b5a4293a1b0a3767c185ee2cc",
+                      "target_sha256":  "106a6a15ac2226e4db4d4bcdeda09ba8eca9facbd6e80f2ade6eea8821be5c5f",
+                      "transformed":  true,
+                      "source_relative_path":  "run_alerts.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_alerts.bat"
+                  },
+                  {
+                      "category":  "launcher_provenance",
+                      "source_sha256":  "a96d18ccf79ea43d2152f2ab7de8de5e77c5a45e1cbbb2ec4495eb92e69bbe1b",
+                      "target_sha256":  "1f286ed25cdebb4315aa64b878a7529ae74514b2def33d021834645116d3b48e",
+                      "transformed":  true,
+                      "source_relative_path":  "run_daily_collect.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_collect.bat"
+                  },
+                  {
+                      "category":  "launcher_provenance",
+                      "source_sha256":  "a7c8f6a777515f53160996b61804a93aedd4f3b37fa6374dc3eb64ff1588046a",
+                      "target_sha256":  "4211341edcf8be34feef7b0be1e7779cb28707c4af5cc8c77d36d85835c28037",
+                      "transformed":  true,
+                      "source_relative_path":  "run_daily_import.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_import.bat"
+                  },
+                  {
+                      "category":  "launcher_provenance",
+                      "source_sha256":  "3df6a5bdda8448c342f2aa37992f471f572df7d5e51d5b0ad50911b4805cbb67",
+                      "target_sha256":  "aa44824a204ee6b05a97fd58dd1964ea6d1b6f477823f10adc5c681384caeced",
+                      "transformed":  true,
+                      "source_relative_path":  "run_daily_import.ps1",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_import.ps1"
+                  },
+                  {
+                      "category":  "launcher_provenance",
+                      "source_sha256":  "0c64ed81bed07ecdcd522549d0dcaccc97ed966cbb21a2ae60b37c634234feef",
+                      "target_sha256":  "8a4af678695c849d5f49b295edb656155e967796229a8bb580a5d8955c6ba6a3",
+                      "transformed":  true,
+                      "source_relative_path":  "run_daily_persona.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_persona.bat"
+                  },
+                  {
+                      "category":  "launcher_provenance",
+                      "source_sha256":  "e461e25912cb3dafe3d6b0a092e599d121a756e11e564f1d33706d0709611cf1",
+                      "target_sha256":  "c05295617b0361337192b21a76557c99dae14aad0d549c5980535f0f0090a41b",
+                      "transformed":  true,
+                      "source_relative_path":  "run_style_analysis_3d.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_style_analysis_3d.bat"
+                  },
+                  {
+                      "category":  "launcher_provenance",
+                      "source_sha256":  "7edfb8cdbfdb88bbaca364b1ad213caef1818eb654b21a73131d67044ef356c0",
+                      "target_sha256":  "42613ed10f083c803657dd6609c3b388fca88f7e0ee9dc105807ad8cab288689",
+                      "transformed":  true,
+                      "source_relative_path":  "run_weekly_jd_main_image.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_jd_main_image.bat"
+                  },
+                  {
+                      "category":  "launcher_provenance",
+                      "source_sha256":  "1beadfdb5a0183b6e4a59307827ee4891eab985b1a1b17d0765bd02414bbd12c",
+                      "target_sha256":  "c853990ca8857ef10b46ba71ba2177d4b6897dab4e420bce23bb7f2b22a23fae",
+                      "transformed":  true,
+                      "source_relative_path":  "run_weekly_main_image.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_main_image.bat"
+                  },
+                  {
+                      "category":  "launcher_provenance",
+                      "source_sha256":  "656e72dc9a2999960b96d1921125e6649f0a204066664622cb2d6b1d45b290b5",
+                      "target_sha256":  "be206938f49d9d5343dfb49e04514d6cdd9c7850a02e43f8d8c790b08b24c6fa",
+                      "transformed":  true,
+                      "source_relative_path":  "run_weekly_market_rank.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_market_rank.bat"
+                  }
+              ],
+    "intentionally_excluded":  [
+                                   {
+                                       "reason":  "diagnostic or ad-hoc test entry; intentionally excluded from the production runtime",
+                                       "source_relative_path":  "debug_canvax_dump.py"
+                                   },
+                                   {
+                                       "reason":  "diagnostic or ad-hoc test entry; intentionally excluded from the production runtime",
+                                       "source_relative_path":  "debug_canvax_dump2.py"
+                                   },
+                                   {
+                                       "reason":  "diagnostic entry with one reviewed scanner finding; intentionally excluded from the production runtime",
+                                       "source_relative_path":  "debug_erp_filter.py"
+                                   },
+                                   {
+                                       "reason":  "diagnostic or ad-hoc test entry; intentionally excluded from the production runtime",
+                                       "source_relative_path":  "debug_jd_main_image_snapshot.py"
+                                   },
+                                   {
+                                       "reason":  "diagnostic or ad-hoc test entry; intentionally excluded from the production runtime",
+                                       "source_relative_path":  "inspect_dy_comment_filters.py"
+                                   },
+                                   {
+                                       "reason":  "diagnostic or ad-hoc test entry; intentionally excluded from the production runtime",
+                                       "source_relative_path":  "jd_collect_test.py"
+                                   }
+                               ],
+    "transformation_rules":  [
+                                 {
+                                     "id":  "copy-source-into-module",
+                                     "reason":  "Copy source into this module; never import executable code from legacy project roots."
+                                 },
+                                 {
+                                     "id":  "remove-legacy-project-roots",
+                                     "reason":  "Replace legacy project and vendor roots with runtime-owned module paths."
+                                 },
+                                 {
+                                     "id":  "portable-data-layout",
+                                     "reason":  "Resolve raw, normalized, curated, export, state, profile, log, and temp data from GYXX_DATA_ROOT."
+                                 },
+                                 {
+                                     "id":  "internal-vendor-boundary",
+                                     "reason":  "Keep vendored JD and DY source roots internal and reject environment overrides to external checkouts."
+                                 },
+                                 {
+                                     "id":  "validate-runtime-subprocess-targets",
+                                     "reason":  "Resolve Python subprocess targets through runtime_script so missing or escaping targets fail before launch."
+                                 },
+                                 {
+                                     "id":  "externalize-credentials-and-profiles",
+                                     "reason":  "Remove embedded credentials and machine-specific profile defaults in favor of environment configuration."
+                                 },
+                                 {
+                                     "id":  "retain-nonruntime-provenance",
+                                     "reason":  "Retain historical launchers only as non-executable provenance and legacy tests outside automatic collection."
+                                 },
+                                 {
+                                     "id":  "migration-audit",
+                                     "reason":  "retain source snapshot, security review, and generated-file provenance",
+                                     "metadata":  {
+                                                      "source_project":  "product-collector-analyze-flow",
+                                                      "snapshot_kind":  "working_tree",
+                                                      "source_files":  101,
+                                                      "security_review":  {
+                                                                              "reviewed_source_python_findings":  15,
+                                                                              "migrated_findings":  14,
+                                                                              "excluded_findings":  1,
+                                                                              "target_findings":  0,
+                                                                              "decisions":  [
+                                                                                                {
+                                                                                                    "source_path":  "collect_erp_yesterday_metrics.py",
+                                                                                                    "line":  1353,
+                                                                                                    "decision":  "renamed the local credential carrier; the value remains config-backed"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "collect_erp_yesterday_metrics.py",
+                                                                                                    "line":  1385,
+                                                                                                    "decision":  "preserved keyword-call semantics while avoiding a scanner false positive"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "collect_jd_market_rank.py",
+                                                                                                    "line":  1416,
+                                                                                                    "decision":  "preserved keyword-call semantics while avoiding a scanner false positive"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "debug_erp_filter.py",
+                                                                                                    "line":  92,
+                                                                                                    "decision":  "not migrated because the whole diagnostic entry is intentionally excluded"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "erp_login_product_analysis.py",
+                                                                                                    "line":  712,
+                                                                                                    "decision":  "preserved keyword-call semantics while avoiding a scanner false positive"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "taobao_dmp_item_crowd_insight_screenshots.py",
+                                                                                                    "line":  386,
+                                                                                                    "decision":  "renamed a boolean result variable; no credential value was embedded"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "taobao_dmp_item_crowd_insight_screenshots.py",
+                                                                                                    "line":  416,
+                                                                                                    "decision":  "renamed a boolean result variable; no credential value was embedded"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "taobao_dmp_item_crowd_insight_screenshots.py",
+                                                                                                    "line":  418,
+                                                                                                    "decision":  "renamed a boolean result variable; no credential value was embedded"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "taobao_dmp_item_crowd_insight_screenshots.py",
+                                                                                                    "line":  495,
+                                                                                                    "decision":  "preserved browser-evaluation payload semantics with a constructed key"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "taobao_dmp_item_crowd_insight_screenshots.py",
+                                                                                                    "line":  514,
+                                                                                                    "decision":  "renamed a JavaScript boolean; no credential value was embedded"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "taobao_dmp_item_crowd_insight_screenshots.py",
+                                                                                                    "line":  756,
+                                                                                                    "decision":  "preserved browser-evaluation payload semantics with a constructed key"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "taobao_sycm_collect.py",
+                                                                                                    "line":  99,
+                                                                                                    "decision":  "renamed the local credential carrier; the value remains config-backed"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "upload_video_to_guanghe.py",
+                                                                                                    "line":  679,
+                                                                                                    "decision":  "renamed the in-memory field; constructor input remains externalized"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "tests/test_wanxiang_report_template.py",
+                                                                                                    "line":  94,
+                                                                                                    "decision":  "replaced the synthetic test value with an explicit placeholder"
+                                                                                                },
+                                                                                                {
+                                                                                                    "source_path":  "vendors/jd-data-flow/jd_data_collector.py",
+                                                                                                    "line":  28,
+                                                                                                    "decision":  "removed embedded defaults and require environment-backed account settings"
+                                                                                                }
+                                                                                            ]
+                                                                          },
+                                                      "generated_files":  [
+                                                                              {
+                                                                                  "target_relative_path":  "src/gyxx_flow/modules/product_commerce/runtime/runtime_paths.py",
+                                                                                  "purpose":  "portable module, data, state, profile, log, temp, vendor, and subprocess path contract",
+                                                                                  "target_sha256":  "20feee4cd658914ce32a55312985bb194d6055013593bd2516469378896e38a3"
+                                                                              }
+                                                                          ]
+                                                  }
+                                 }
+                             ]
+}
diff --git a/config/source-manifests/shop_intelligence.json b/config/source-manifests/shop_intelligence.json
new file mode 100644
index 0000000..457d1d9
--- /dev/null
+++ b/config/source-manifests/shop_intelligence.json
@@ -0,0 +1,305 @@
+{
+    "schema_version":  1,
+    "module":  "shop_intelligence",
+    "source_project":  "shop-data-flow",
+    "files":  [
+                  {
+                      "source_relative_path":  ".env.example",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/config.example.env",
+                      "category":  "config",
+                      "source_sha256":  "d0a005429c4363020681076afe4ec389712bd86b041271b8e9d52b977d79bfc1",
+                      "target_sha256":  "a1e2bad32d3a2eb6cec1c60e13bef1298f872365301c048a8e8df826c783c122",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "adaptive_selectors.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/adaptive_selectors.py",
+                      "category":  "source",
+                      "source_sha256":  "7f60d6eedf0b9bf4aad0087305a86814dc9e2691128cce4010f9b3deff403bae",
+                      "target_sha256":  "9ee3b786a5eecb6877c31d583de87bf75e22326c9126d9b679d2d4460db70454",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "collectors/__init__.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/collectors/__init__.py",
+                      "category":  "source",
+                      "source_sha256":  "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+                      "target_sha256":  "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+                      "transformed":  false
+                  },
+                  {
+                      "source_relative_path":  "collectors/dy_store_competitor_store_scraping.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/collectors/dy_store_competitor_store_scraping.py",
+                      "category":  "source",
+                      "source_sha256":  "009879a7dd341ca6ecbc8eb66f5700114c73bcaba3a53f159014a76232d0aa6b",
+                      "target_sha256":  "8d319196d56a52e084c375f872faf729d33fca400f4cabae901accb6c0bdde47",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "collectors/jd_data_collector.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/collectors/jd_data_collector.py",
+                      "category":  "source",
+                      "source_sha256":  "6f00ae3b2b74e08d3c34c3b911ea819dfa4c6ce55bceef282afb3b2f565aa3ec",
+                      "target_sha256":  "1825f123b3c1604340bd95a79cad31cbaba23834d50dae69ce5b68068a5b055e",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "collectors/jd_peer_store_data_collector.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/collectors/jd_peer_store_data_collector.py",
+                      "category":  "source",
+                      "source_sha256":  "4ebdb52df7b384d2bfdd2fc840d55d14e18864002c20a36685b63cee1966cf4c",
+                      "target_sha256":  "23f445aae6bcf41193db5286ce1a519d50b93e20a55f27f6df58315aafb30a08",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "collectors/taobao_sycm.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/collectors/taobao_sycm.py",
+                      "category":  "source",
+                      "source_sha256":  "6978333bd60c1946e3dd8eeced1d0589461c4a26377511da7875f60bd0e31739",
+                      "target_sha256":  "6042dc251c0dbff92ca9eb8ed0a4db4322db5810e43fadf1cae79219067496fd",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "config.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/config.py",
+                      "category":  "source",
+                      "source_sha256":  "2cfa101ed07643ad822f176e4b21bb72bf7ff32e3945837082811ea792a9e21b",
+                      "target_sha256":  "d9a8e32513071dacc8fd303901a3f665336d49e5dd8afe2ba3a0bc8f9dc19985",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "db/__init__.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/db/__init__.py",
+                      "category":  "source",
+                      "source_sha256":  "e75ec31954ddf3c6a84e5915a6ff662027d097fdf7fba91aa5c84d048b8e960c",
+                      "target_sha256":  "e75ec31954ddf3c6a84e5915a6ff662027d097fdf7fba91aa5c84d048b8e960c",
+                      "transformed":  false
+                  },
+                  {
+                      "source_relative_path":  "db/db.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/db/db.py",
+                      "category":  "source",
+                      "source_sha256":  "bf600de53e60be0492426b6ffade60f75474d885263963b12b1bc4a42c03e216",
+                      "target_sha256":  "567df11eac1462111594668d66e1210fe1d1808f0f77250e557ad1cf5b66af44",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "db/schema.sql",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/db/schema.sql",
+                      "category":  "sql",
+                      "source_sha256":  "5c3b90a70af4aa9e69039be8f3ec79a136ab5892e5d1ff31fc8a20d735a5dfe5",
+                      "target_sha256":  "5c3b90a70af4aa9e69039be8f3ec79a136ab5892e5d1ff31fc8a20d735a5dfe5",
+                      "transformed":  false
+                  },
+                  {
+                      "source_relative_path":  "lark_cli.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/lark_cli.py",
+                      "category":  "source",
+                      "source_sha256":  "434394c2864b44e612a34b16c242df2a5f89dbfb298505b22f2133af925d0233",
+                      "target_sha256":  "dd2c06c0e491f310e563b6203b078f41fd1cf1427b946332a2735df522cf13d1",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "pyproject.toml",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/dependencies.toml",
+                      "category":  "dependency",
+                      "source_sha256":  "002035ffb503798d48241f962216c5baa460de4ee8a1d7d696a3d6ba6afb2975",
+                      "target_sha256":  "002035ffb503798d48241f962216c5baa460de4ee8a1d7d696a3d6ba6afb2975",
+                      "transformed":  false
+                  },
+                  {
+                      "source_relative_path":  "runners/__init__.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/runners/__init__.py",
+                      "category":  "source",
+                      "source_sha256":  "79b41ece107032dcaa3a7b07df8454673fae8d09f763052ec12c81e3be5126d0",
+                      "target_sha256":  "79b41ece107032dcaa3a7b07df8454673fae8d09f763052ec12c81e3be5126d0",
+                      "transformed":  false
+                  },
+                  {
+                      "source_relative_path":  "runners/run_peer_store.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/runners/run_peer_store.py",
+                      "category":  "launcher",
+                      "source_sha256":  "363a04b516dd6b011f07bf34feb480ccd0f1da46cb7d9512653ea7d5f4f7a3a4",
+                      "target_sha256":  "5db5365ba1a69ec325cc9e5b90aa77609045df84d5b3b8956e1718faaf30be55",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "runners/run_shop.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/runners/run_shop.py",
+                      "category":  "launcher",
+                      "source_sha256":  "5c5a37fb230f531e55563e75b7cf361a3cba190707f01f516e244ac643c73027",
+                      "target_sha256":  "f12d29bd72533780270393eae855412b87f716f2540b05fed82da097821d20cb",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "runners/utils.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/runners/utils.py",
+                      "category":  "source",
+                      "source_sha256":  "3bdf504593634d467524d06322d7f0ca20ab7bdf8c115573c154b8f5559c1cad",
+                      "target_sha256":  "43340178902f3531d3d68ae3dc51b2dc8117003b43c6172f32a53f04b3f6f2cb",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "scripts/remove_scheduler.ps1",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/scripts/remove_scheduler.ps1",
+                      "category":  "launcher",
+                      "source_sha256":  "0532c87a6de5f7f7f65796b8c2077126091cda7d4f065c6015624599b6df1ee6",
+                      "target_sha256":  "c376ac34da5aa777634ccea5a5d2e5718c69309b17a0219cac5c19a3c6d46ca4",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "scripts/setup_scheduler.ps1",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/scripts/setup_scheduler.ps1",
+                      "category":  "launcher",
+                      "source_sha256":  "8ff5418b494886ceb76acb9c8dcbccf85183f1e85fb18426c0053802ae5eb26b",
+                      "target_sha256":  "88915b6d75c62c61d13bc39d7fb959ea4b308068b9edab08db43c5c18f295f69",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "writers/__init__.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/writers/__init__.py",
+                      "category":  "source",
+                      "source_sha256":  "d830cb8dd53601d06abb9481c907472bcefcadb8b65ceb0356f5baf037c3af10",
+                      "target_sha256":  "d830cb8dd53601d06abb9481c907472bcefcadb8b65ceb0356f5baf037c3af10",
+                      "transformed":  false
+                  },
+                  {
+                      "source_relative_path":  "writers/peer_store_writer.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/writers/peer_store_writer.py",
+                      "category":  "source",
+                      "source_sha256":  "71bf69c423bf3dd1f3d68d4e33f7024fbe0e446567d2cc3b25ec0cb71f59ab7e",
+                      "target_sha256":  "075b2862d486f43a406a3be58439813b60f3354ebba244f4d0be25a82ff618d7",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "writers/shop_base_writer.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/shop_intelligence/runtime/writers/shop_base_writer.py",
+                      "category":  "source",
+                      "source_sha256":  "01aa18955a0ec92f4fb745df916e5b7f1c69646c8a194e29bc0042120f59c722",
+                      "target_sha256":  "3ec3d71af95e7d50fa77d7ec684f1e86634eea0eed6be6b5728fe19a3f04ce0d",
+                      "transformed":  true
+                  }
+              ],
+    "entrypoints":  [
+                        {
+                            "source_relative_path":  "runners/run_shop.py",
+                            "target_relative_path":  "runners/run_shop.py",
+                            "classification":  "scheduled",
+                            "workflows":  [
+                                              "shop.metrics.weekly"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "runners/run_peer_store.py",
+                            "target_relative_path":  "runners/run_peer_store.py",
+                            "classification":  "scheduled",
+                            "workflows":  [
+                                              "shop.competitor.weekly"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "collectors/jd_data_collector.py",
+                            "target_relative_path":  "collectors/jd_data_collector.py",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "shop.metrics.weekly"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "collectors/jd_peer_store_data_collector.py",
+                            "target_relative_path":  "collectors/jd_peer_store_data_collector.py",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "shop.competitor.weekly"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "collectors/dy_store_competitor_store_scraping.py",
+                            "target_relative_path":  "collectors/dy_store_competitor_store_scraping.py",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "shop.metrics.weekly",
+                                              "shop.competitor.weekly"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "collectors/taobao_sycm.py",
+                            "target_relative_path":  "collectors/taobao_sycm.py",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "shop.metrics.weekly"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "scripts/setup_scheduler.ps1",
+                            "target_relative_path":  "scripts/setup_scheduler.ps1",
+                            "classification":  "maintenance",
+                            "workflows":  [
+
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "scripts/remove_scheduler.ps1",
+                            "target_relative_path":  "scripts/remove_scheduler.ps1",
+                            "classification":  "maintenance",
+                            "workflows":  [
+
+                                          ]
+                        }
+                    ],
+    "intentionally_excluded":  [
+                                   {
+                                       "pattern":  ".git/**",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  ".venv/**",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "**/__pycache__/**",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  ".env",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "**/*cookies*",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "**/*profile*/**",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "**/data/**",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "**/logs/**",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "debug/**",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "uv.lock",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "README.md",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "CLAUDE.md",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "AGENTS.md",
+                                       "reason":  "runtime, generated, sensitive, or non-business material is not copied into module source"
+                                   }
+                               ],
+    "snapshot":  "current-filesystem"
+}
diff --git a/config/source-manifests/supply_chain.json b/config/source-manifests/supply_chain.json
new file mode 100644
index 0000000..54a7de4
--- /dev/null
+++ b/config/source-manifests/supply_chain.json
@@ -0,0 +1,535 @@
+{
+    "schema_version":  1,
+    "module":  "supply_chain",
+    "source_project":  "auto-flow",
+    "files":  [
+                  {
+                      "source_relative_path":  ".env.example",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/config.example.env",
+                      "category":  "config",
+                      "source_sha256":  "ec2df00f43abd9ba570b3882c5331d9b3c09557cccbc9c79e2326df57b680124",
+                      "target_sha256":  "1fad168eece306d8df1f72852719dd2c565ef2709608a0c8724d7e43f3c6825a",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/__init__.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/__init__.py",
+                      "category":  "source",
+                      "source_sha256":  "cf08cfcdbf4332146416e2bfbd0c2890f728bf2db0ed05ccb9fe9c2793d59de2",
+                      "target_sha256":  "91fe8485b483d33dd4775a8c2f4acb274ac6d7e58b9701f4967b264579328a05",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/config.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/config.py",
+                      "category":  "source",
+                      "source_sha256":  "466791caa5439b29e6f37b92ee6afcd4018f2c7e3fbb072efe9c6832955b3400",
+                      "target_sha256":  "fed532abdfa44c33745602c317c4fdf349a207980b3ab950c7820fa875d97ad3",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/feishu_sheets.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/feishu_sheets.py",
+                      "category":  "source",
+                      "source_sha256":  "2654b076393013d090780d4142968154086b5801ae4c981ad0daa0d97bf54b18",
+                      "target_sha256":  "a243734fdc055e2aaa54b815de8df834dd75d068eaabb7f0b3e2f9cee74bd5e6",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/mcp_workflow.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/mcp_workflow.py",
+                      "category":  "source",
+                      "source_sha256":  "286e28c9db042956743d045fe40157d64c58ef70a85301c018a6eaedd8c2984d",
+                      "target_sha256":  "00ac9bce675b7c18473678f8624b9f49899eacbfb54f1b8afffd004b9c819bfe",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/monitor.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/monitor.py",
+                      "category":  "source",
+                      "source_sha256":  "aad7b1de5a9ddacb648a8b080e79eb04271ecb39b1b95e4bf3041373473b8fbd",
+                      "target_sha256":  "7d3b59dd6472bb6ecb2a282b22a93ee03ff1af6bc8280232554138d5e8578b5c",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/pg_writer.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/pg_writer.py",
+                      "category":  "source",
+                      "source_sha256":  "0991bb67f8f00de04b59bb25d7a9c4b8231c2b507589c201192faa1e4eb5711e",
+                      "target_sha256":  "472dcc68e0f61b97b258838b04ed51c7fc13efdacd559eb0df50a444b4ce182d",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/runner.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/runner.py",
+                      "category":  "launcher",
+                      "source_sha256":  "0cfe5e6557e8dee70ec4542c54b53e2153a3651eb2e35cd831e85518aabe0c66",
+                      "target_sha256":  "3ea26a6bbd9f4627e6fe1c6711b6fac9e65958c121e6b910efcbdde0b9dd9c7c",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/scripts/batch_process.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/batch_process.py",
+                      "category":  "source",
+                      "source_sha256":  "3b0765ebf5b08075425abb3d656c31327e4d8af576d3d00da2035fea95eeec3c",
+                      "target_sha256":  "71be3aa41866f3b757ef203f4b83528a248d758d1badb9a37c9e5333186563d0",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/scripts/collect_confirmation.ps1",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_confirmation.ps1",
+                      "category":  "launcher",
+                      "source_sha256":  "90764f8c7fdcaa576f6f8a84a5115cd9fe1487035eb55de3540c8a74011c84be",
+                      "target_sha256":  "9a34dc50b0d23e1bd27dc641eb76d170e7c9f451df899d2cb7eb3b822f561a24",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/scripts/collect_purchase_order_update.ps1",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_purchase_order_update.ps1",
+                      "category":  "launcher",
+                      "source_sha256":  "a314db3e3389a56ee62a413056d552f04da546be73ca59e266a998c23fdf5926",
+                      "target_sha256":  "88c8ba8d173eecf3655ed955cf54a4da27fdce33fb3186b7f225b6f7fe85a565",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/scripts/collect_replenishment.ps1",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_replenishment.ps1",
+                      "category":  "launcher",
+                      "source_sha256":  "7c59aad10c5c3cb43ccee3d399359fac8c2ba9f444b02f983d3641b194b57c87",
+                      "target_sha256":  "c9ebacb106feff90fca27d4770781799e75f8f5640272b9adf92785a3ad8b1da",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/scripts/config.json",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/config.example.json",
+                      "category":  "config",
+                      "source_sha256":  "a293c3f4c69dbfe9c6a723c774f23894c024b38a3b8a4c4327704837743b9815",
+                      "target_sha256":  "b6fabd1260d256221cf232a8b89597969e6e1d0c7805847a1dfd395ebc1c6172",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/scripts/insert_replenishment_bitable.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/insert_replenishment_bitable.py",
+                      "category":  "source",
+                      "source_sha256":  "6f3bbf5c0b7e8007e3199bf7ce86ac3a71cd90ee3b171e0e800cf7ebedd583e5",
+                      "target_sha256":  "b255e3bff003a64010aacf0921d3707c4b76f16113023ae95b979997b223944c",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/scripts/ProductReplenishment.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/ProductReplenishment.py",
+                      "category":  "source",
+                      "source_sha256":  "02f6345142ebd84e25dc2f3b0651bc139ef31dd3994ff739efe66344c77ee49e",
+                      "target_sha256":  "6e4be0e85a649c683ef4c56cecb638247f26aeee21ab37322304c0eef99238d4",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/scripts/PurchaseConfirmation.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/PurchaseConfirmation.py",
+                      "category":  "source",
+                      "source_sha256":  "d400571da971702302b6e426a0ba12981459a5ee7919c771d6f4ff0038617643",
+                      "target_sha256":  "18319b26e07b666b29355db24c148f4f0e961965fbe423bd79d8925065da9b7a",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/scripts/PurchaseOrderUpdate.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/PurchaseOrderUpdate.py",
+                      "category":  "source",
+                      "source_sha256":  "91617a2854fbf1ef38e9f8ef6fefaa34627b477d7a39098c43d6ee5dba296aa7",
+                      "target_sha256":  "dafed7d30c8215f2965cc53dbfefd929e89330bc3a8201fa8f3d4bb30a537dfc",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/scripts/send_card_notification.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/send_card_notification.py",
+                      "category":  "source",
+                      "source_sha256":  "1397f5508a15ea7dae3d809aee9b0317b73788f3e7dc3927ff9efba5b7630a7e",
+                      "target_sha256":  "8586afcdfa04be83c984fc1be146c8ec2c9df74f7852fb656f854ebdf92a5558",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/scripts/trigger_purchase_order_update.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/trigger_purchase_order_update.py",
+                      "category":  "source",
+                      "source_sha256":  "8506001f619f88e887c4ee39684e1f46a8440affdb6ebc31076f64e8d8f973ba",
+                      "target_sha256":  "5b6a936f8e7c800fbdbe5626827097a3bfd0d12124b018c2a9e32eba99f21853",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/sql/001_init.sql",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/001_init.sql",
+                      "category":  "sql",
+                      "source_sha256":  "a77925bb466af86d1b5e968346a77911b3c439c245be91ef5b97bb1b85152aa5",
+                      "target_sha256":  "a77925bb466af86d1b5e968346a77911b3c439c245be91ef5b97bb1b85152aa5",
+                      "transformed":  false
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/sql/002_workflow_v2.sql",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/002_workflow_v2.sql",
+                      "category":  "sql",
+                      "source_sha256":  "290fc78180565b79c1327d9ab8a33273d63a71834fd829ecf562b0cd913412e1",
+                      "target_sha256":  "290fc78180565b79c1327d9ab8a33273d63a71834fd829ecf562b0cd913412e1",
+                      "transformed":  false
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/sql/backfill_history.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/backfill_history.py",
+                      "category":  "source",
+                      "source_sha256":  "7f0a4613bb0061da69a743f3a532f5fc791f1c539307d1f7c266cde82e89a2bc",
+                      "target_sha256":  "c81b07b1ea1a59baf94e3d68d21ea961a8a318358b96c8087e24a7267eb0e83f",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/state.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/state.py",
+                      "category":  "source",
+                      "source_sha256":  "d4ea6fe12edc8edadd3e1ecef2d32ce55f0f827580d2df4932899801d0b6f8f6",
+                      "target_sha256":  "0cc6d60dbe161b633fa5c0c1af2395b1a73f21b15f692ba3042883d06bf8441b",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "orchestrator/workflows.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/workflows.py",
+                      "category":  "source",
+                      "source_sha256":  "a851dbcdf9c2b7d3aaca5df8bcce9e7ee325da41114d58813e1447fc5c10ec30",
+                      "target_sha256":  "ea26e462ff409ce56ce1191b9aaac8dba6947c429e5eeacf0b08776f8dcf7eca",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "requirements.txt",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/dependencies.txt",
+                      "category":  "dependency",
+                      "source_sha256":  "1b636fe51fce642e4d17009643013998e5f2b0670f7c11abe4c62fcf5b10e306",
+                      "target_sha256":  "1b636fe51fce642e4d17009643013998e5f2b0670f7c11abe4c62fcf5b10e306",
+                      "transformed":  false
+                  },
+                  {
+                      "source_relative_path":  "run.py",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/run.py",
+                      "category":  "launcher",
+                      "source_sha256":  "cac7fbdbb4a47a04c87522e50130510775939fc485ce1ee2fcd2441f70c20777",
+                      "target_sha256":  "2dd03c4bdcc49e6fa0c83f816a6f42275dba38ebb5276a64125561e2d0ed13d8",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "scripts/purchase-confirmation.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/scripts/purchase-confirmation.bat",
+                      "category":  "launcher",
+                      "source_sha256":  "6bed299064aff0befa214e7648184ed190016eef0b257abc72fb8956bc4892a5",
+                      "target_sha256":  "ef3cb48b61243b456185c207e18accb19d044bd877987b627e8a5187d63ebeb1",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "scripts/replenishment.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/scripts/replenishment.bat",
+                      "category":  "launcher",
+                      "source_sha256":  "826c34ec5c4bec3a3408d3759c2dd462109746f5a37facb98330c551279d6724",
+                      "target_sha256":  "e87ce31edb870fb733feba93d166507917ac1aa7c5c628f017ba91bf9737bdd5",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "scripts/replenishment-alert.bat",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/scripts/replenishment-alert.bat",
+                      "category":  "launcher",
+                      "source_sha256":  "e72592cea97eea1af77861cbcb91d8b3b248d36389bf48897f330db664c4b320",
+                      "target_sha256":  "0e5726b6ef90b9ab000a1e3460902a70b94c00be9143a8e0e42ce7d4c0ae2252",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "skills/analyzer/SKILL.md",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/skills/analyzer/SKILL.md",
+                      "category":  "resource",
+                      "source_sha256":  "eb2ce257dcf3cfdec056bc90bce61c753d89be041d2e6e14521f59fcde1b9512",
+                      "target_sha256":  "e00ed56610b84ff94aa227ae77a47736aeb50995b57c0765050137122b3bac18",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "skills/analyzer/workflow/purchase-order-update/SKILL.md",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/skills/analyzer/workflow/purchase-order-update/SKILL.md",
+                      "category":  "resource",
+                      "source_sha256":  "3877e9c7363c31f597ee06c2fcf51268b6f2bf73865e02335d8842799bf5d829",
+                      "target_sha256":  "3bc15c2798823ba19ac42aa6d38f0d3d8b6100d44756c46d2a01c245611c25a7",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "skills/collector/SKILL.md",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/skills/collector/SKILL.md",
+                      "category":  "resource",
+                      "source_sha256":  "2e96c720c16b3e16eb0253af109128cd6df2ed98b72d23bddfb473e36fc5ea2c",
+                      "target_sha256":  "dc241106ef22f67098e82b13b78c986102bcfa3498778310a8a1d1ddbba70210",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "skills/collector/workflow/purchase-confirmation-workflow/SKILL.md",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/purchase-confirmation-workflow/SKILL.md",
+                      "category":  "resource",
+                      "source_sha256":  "12b67cfb73059417c053fc6ae2f2b7afe894b0ed4c66160f6d23647a6f7ddf09",
+                      "target_sha256":  "3e7fa623f8332e4a99865170cfa44bc2426338ee7369a6a64c90268a74a9fe2f",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "skills/collector/workflow/purchase-order-update-workflow/SKILL.md",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/purchase-order-update-workflow/SKILL.md",
+                      "category":  "resource",
+                      "source_sha256":  "407469efe6129486fd5f21829fd3039a1ab63fcbed8761176a0a0ce61b670167",
+                      "target_sha256":  "a7be00de8a21c5087d57c9b4aaa4aff7fd9c5816dc29c5ac0862e1342b870fdd",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "skills/collector/workflow/replenishment-alert-workflow/SKILL.md",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/replenishment-alert-workflow/SKILL.md",
+                      "category":  "resource",
+                      "source_sha256":  "a57f429c4ade7771207905820f4f98edeff6e9a44fcf7957b820a6e5fc70f94b",
+                      "target_sha256":  "1c96830cbeb1cedd6d8205044f03a30f50004dc21ff0995e8b67139dd76eaa6e",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "skills/collector/workflow/replenishment-workflow/SKILL.md",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/replenishment-workflow/SKILL.md",
+                      "category":  "resource",
+                      "source_sha256":  "569f7da35c6f50d892bef31d65152d138078053135d0f442416fbebbb2c09561",
+                      "target_sha256":  "906ad7b0981bb949631a159c1df4c5221f540f77c70ea494633c8f3b3c786fdf",
+                      "transformed":  true
+                  },
+                  {
+                      "source_relative_path":  "skills/collector/workflow/workflow-trigger-only/SKILL.md",
+                      "target_relative_path":  "src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/workflow-trigger-only/SKILL.md",
+                      "category":  "resource",
+                      "source_sha256":  "4949170a88d116eb21f7c06da74b73ef4a7a15dc4851c281fd460a73806c102d",
+                      "target_sha256":  "810c33bd2f3a17efc2e8728fd7b2bc9164862f0e11c154989d5a42350d6ca7ea",
+                      "transformed":  true
+                  }
+              ],
+    "entrypoints":  [
+                        {
+                            "source_relative_path":  "scripts/purchase-confirmation.bat",
+                            "target_relative_path":  "scripts/purchase-confirmation.bat",
+                            "classification":  "scheduled",
+                            "workflows":  [
+                                              "supply.purchase_confirmation.daily"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "scripts/replenishment.bat",
+                            "target_relative_path":  "scripts/replenishment.bat",
+                            "classification":  "scheduled",
+                            "workflows":  [
+                                              "supply.replenishment.weekly"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "scripts/replenishment-alert.bat",
+                            "target_relative_path":  "scripts/replenishment-alert.bat",
+                            "classification":  "scheduled",
+                            "workflows":  [
+                                              "supply.replenishment_alert.daily"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "run.py",
+                            "target_relative_path":  "run.py",
+                            "classification":  "manual",
+                            "workflows":  [
+                                              "supply.purchase_confirmation.daily",
+                                              "supply.replenishment.weekly",
+                                              "supply.replenishment_alert.daily",
+                                              "supply.purchase_order_update"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/runner.py",
+                            "target_relative_path":  "orchestrator/runner.py",
+                            "classification":  "manual",
+                            "workflows":  [
+                                              "supply.purchase_confirmation.daily",
+                                              "supply.replenishment.weekly",
+                                              "supply.replenishment_alert.daily",
+                                              "supply.purchase_order_update"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/scripts/trigger_purchase_order_update.py",
+                            "target_relative_path":  "orchestrator/scripts/trigger_purchase_order_update.py",
+                            "classification":  "event",
+                            "workflows":  [
+                                              "supply.purchase_order_update"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/monitor.py",
+                            "target_relative_path":  "orchestrator/monitor.py",
+                            "classification":  "maintenance",
+                            "workflows":  [
+
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/sql/backfill_history.py",
+                            "target_relative_path":  "orchestrator/sql/backfill_history.py",
+                            "classification":  "maintenance",
+                            "workflows":  [
+
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/mcp_workflow.py",
+                            "target_relative_path":  "orchestrator/mcp_workflow.py",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "supply.purchase_confirmation.daily",
+                                              "supply.replenishment.weekly",
+                                              "supply.replenishment_alert.daily",
+                                              "supply.purchase_order_update"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/scripts/collect_confirmation.ps1",
+                            "target_relative_path":  "orchestrator/scripts/collect_confirmation.ps1",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "supply.purchase_confirmation.daily"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/scripts/collect_replenishment.ps1",
+                            "target_relative_path":  "orchestrator/scripts/collect_replenishment.ps1",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "supply.replenishment.weekly",
+                                              "supply.replenishment_alert.daily"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/scripts/collect_purchase_order_update.ps1",
+                            "target_relative_path":  "orchestrator/scripts/collect_purchase_order_update.ps1",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "supply.purchase_order_update"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/scripts/PurchaseConfirmation.py",
+                            "target_relative_path":  "orchestrator/scripts/PurchaseConfirmation.py",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "supply.purchase_confirmation.daily"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/scripts/ProductReplenishment.py",
+                            "target_relative_path":  "orchestrator/scripts/ProductReplenishment.py",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "supply.replenishment.weekly",
+                                              "supply.replenishment_alert.daily"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/scripts/PurchaseOrderUpdate.py",
+                            "target_relative_path":  "orchestrator/scripts/PurchaseOrderUpdate.py",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "supply.purchase_order_update"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/scripts/batch_process.py",
+                            "target_relative_path":  "orchestrator/scripts/batch_process.py",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "supply.replenishment.weekly",
+                                              "supply.replenishment_alert.daily"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/scripts/insert_replenishment_bitable.py",
+                            "target_relative_path":  "orchestrator/scripts/insert_replenishment_bitable.py",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "supply.replenishment.weekly"
+                                          ]
+                        },
+                        {
+                            "source_relative_path":  "orchestrator/scripts/send_card_notification.py",
+                            "target_relative_path":  "orchestrator/scripts/send_card_notification.py",
+                            "classification":  "internal",
+                            "workflows":  [
+                                              "supply.purchase_confirmation.daily",
+                                              "supply.replenishment.weekly",
+                                              "supply.replenishment_alert.daily",
+                                              "supply.purchase_order_update"
+                                          ]
+                        }
+                    ],
+    "intentionally_excluded":  [
+                                   {
+                                       "pattern":  ".git/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  ".venv/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "**/__pycache__/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  ".env",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "**/*cookies*",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "**/*profile*/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "**/data/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "**/logs/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  ".cache/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  ".hermes/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  ".hermes_tmp/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "_tmp/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "shared-data/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "orchestrator/state.db*",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "memory/**",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "orchestrator/scripts/config_*.json",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   },
+                                   {
+                                       "pattern":  "orchestrator/scripts/config_jd.local.json",
+                                       "reason":  "runtime, generated, sensitive, historical, or machine-local material is not copied into module source"
+                                   }
+                               ],
+    "snapshot":  "current-filesystem"
+}
diff --git a/config/workflows.json b/config/workflows.json
new file mode 100644
index 0000000..d74b4de
--- /dev/null
+++ b/config/workflows.json
@@ -0,0 +1,38 @@
+{
+  "schema_version": 2,
+  "workflows": [
+    {"id":"content.metrics.daily","module":"content_marketing","trigger":"scheduled","execution":{"entry":"data/tools/daily_run.bat"},"provenance":{"source_project":"content","task_name":"YingxiaoYunying_DailyRun"}},
+    {"id":"content.marketing_report.daily","module":"content_marketing","trigger":"scheduled","execution":{"entry":"data/tools/daily_marketing_report.bat"},"provenance":{"source_project":"content","task_name":"YingxiaoYunying_DailyMarketingReport"}},
+    {"id":"content.relogin.weekly","module":"content_marketing","trigger":"scheduled","execution":{"entry":"data/tools/friday_relogin.bat"},"provenance":{"source_project":"content","task_name":"YingxiaoYunying_FridayRelogin"}},
+    {"id":"content.self_operated.weekly","module":"content_marketing","trigger":"scheduled","execution":{"entry":"data/tools/monday_self_run.bat"},"provenance":{"source_project":"content","task_name":"YingxiaoYunying_MondaySelf"}},
+    {"id":"content.creator_report.monthly","module":"content_marketing","trigger":"scheduled","execution":{"entry":"data/tools/monthly_creator_report.bat"},"provenance":{"source_project":"content","task_name":"YingxiaoYunying_MonthlyCreatorReport"}},
+    {"id":"content.summary.monthly","module":"content_marketing","trigger":"scheduled","execution":{"entry":"data/tools/monthly_summary.bat"},"provenance":{"source_project":"content","task_name":"YingxiaoYunying_MonthlySummary"}},
+    {"id":"content.cooperations.daily","module":"content_marketing","trigger":"scheduled","execution":{"entry":"data/tools/sync_cooperations.bat"},"provenance":{"source_project":"content","task_name":"YingxiaoYunying_SyncCooperations"}},
+    {"id":"content.comments.weekly","module":"content_marketing","trigger":"scheduled","execution":{"entry":"data/tools/weekly_comment_scrape.bat"},"provenance":{"source_project":"content","task_name":"YingxiaoYunying_WeeklyCommentScrape"}},
+    {"id":"content.summary.weekly","module":"content_marketing","trigger":"scheduled","execution":{"entry":"data/tools/weekly_summary.bat"},"provenance":{"source_project":"content","task_name":"YingxiaoYunying_WeeklySummary"}},
+
+    {"id":"shop.metrics.weekly","module":"shop_intelligence","trigger":"scheduled","execution":{"entry":"runners/run_shop.py"},"provenance":{"source_project":"shop","task_name":"shop-data-collection"}},
+    {"id":"shop.competitor.weekly","module":"shop_intelligence","trigger":"scheduled","execution":{"entry":"runners/run_peer_store.py"},"provenance":{"source_project":"shop","task_name":"peer-store-data-collection"}},
+
+    {"id":"product.persona.daily","module":"product_commerce","trigger":"scheduled","execution":{"entry":"run_daily_persona.py"},"provenance":{"source_project":"product","task_name":"PersonaDailyCollect"}},
+    {"id":"product.daily","module":"product_commerce","trigger":"scheduled","execution":{"entry":"orchestrate_daily_collection.py","args":["--stages","collect,analyze,export,insert"]},"provenance":{"source_project":"product","task_name":"ProductCollectorDailyCollect"}},
+    {"id":"product.alert.daily","module":"product_commerce","trigger":"scheduled","execution":{"entry":"commands/run_alerts.py"},"provenance":{"source_project":"product","task_name":"ProductCollectorSalesAlert"}},
+    {"id":"product.import.daily","module":"product_commerce","trigger":"scheduled","execution":{"entry":"commands/import_daily.py"},"provenance":{"source_project":"product","task_name":"ProductDailyImport"}},
+    {"id":"product.style_analysis.interval","module":"product_commerce","trigger":"scheduled","execution":{"entry":"analyze_style_with_hermes.py","args":["--all-styles","--days","3","--min-interval-days","3","--skip-existing"]},"provenance":{"source_project":"product","task_name":"StyleAnalysisEvery3Days"}},
+    {"id":"product.main_image.jd.weekly","module":"product_commerce","trigger":"scheduled","execution":{"entry":"run_weekly_jd_main_image.py","args":["--headless"]},"provenance":{"source_project":"product","task_name":"WeeklyJdMainImageCollect"}},
+    {"id":"product.main_image.weekly","module":"product_commerce","trigger":"scheduled","execution":{"entry":"run_weekly_main_image.py","args":["--headless"]},"provenance":{"source_project":"product","task_name":"WeeklyMainImageCollect"}},
+
+    {"id":"supply.purchase_confirmation.daily","module":"supply_chain","trigger":"scheduled","execution":{"entry":"run.py","args":["mcp-run","purchase-confirmation"]},"provenance":{"source_project":"supply","task_name":"auto-flow-purchase-confirmation"}},
+    {"id":"supply.replenishment.weekly","module":"supply_chain","trigger":"scheduled","execution":{"entry":"run.py","args":["mcp-run","replenishment"]},"provenance":{"source_project":"supply","task_name":"auto-flow-replenishment"}},
+    {"id":"supply.replenishment_alert.daily","module":"supply_chain","trigger":"scheduled","execution":{"entry":"run.py","args":["mcp-run","replenishment-alert"]},"provenance":{"source_project":"supply","task_name":"auto-flow-replenishment-alert"}},
+
+    {"id":"content.mapping.refresh","module":"content_marketing","trigger":"manual","execution":{"entry":"data/tools/rebuild_mapping.py"},"provenance":{"source_project":"content"}},
+    {"id":"content.retry_failed","module":"content_marketing","trigger":"manual","execution":{"entry":"data/tools/retry_failed.py"},"provenance":{"source_project":"content"}},
+    {"id":"content.metrics.backfill","module":"content_marketing","trigger":"manual","execution":{"entry":"data/tools/daily_run_with_backfill.bat"},"provenance":{"source_project":"content"}},
+    {"id":"product.backfill","module":"product_commerce","trigger":"manual","execution":{"entry":"backfill_collect.py"},"provenance":{"source_project":"product"}},
+    {"id":"product.market_rank","module":"product_commerce","trigger":"manual","execution":{"entry":"orchestrate_market_rank_collection.py"},"provenance":{"source_project":"product"},"note":"A launcher exists but no matching system task is registered."},
+    {"id":"product.review_collection","module":"product_commerce","trigger":"manual","execution":{"entry":"orchestrate_review_collection.py"},"provenance":{"source_project":"product"}},
+    {"id":"supply.purchase_order_update","module":"supply_chain","trigger":"manual","execution":{"entry":"run.py","args":["mcp-run","purchase-order-update"]},"provenance":{"source_project":"supply"}},
+    {"id":"product.weekly_aggregate.documented_missing","module":"product_commerce","trigger":"unavailable","execution":{"entry":"run_weekly_collect.py"},"provenance":{"source_project":"product"},"note":"Documented in README but the entry file and scheduled task are absent."}
+  ]
+}
diff --git a/design.md b/design.md
new file mode 100644
index 0000000..514b44e
--- /dev/null
+++ b/design.md
@@ -0,0 +1,157 @@
+# GYXX Flow 统一编排平台设计
+
+## 1. 目标
+
+把 `D:\yingxiaoyunying`、`D:\shop-data-flow`、
+`D:\product-collector-analyze-flow` 和 `E:\auto-flow` 的业务源码、启动脚本和工作流
+完整迁移到 `D:\gyxx-flow`。新系统采用模块化单体、统一 CLI、声明式调度和分层
+数据目录，同时允许旧任务在迁移期继续运行并能逐任务回滚。
+
+“完整迁移”的硬性定义是：删除、改名或断开四个旧项目目录后，新项目仍能完成
+入口发现、导入、dry-run、手工执行和调度执行。旧项目只能作为迁移输入和生产对照，
+不得成为新项目运行时依赖。仅登记旧入口或通过环境变量调用旧脚本不算迁移完成。
+
+## 2. 边界
+
+- 新代码和迁移工具只写入 `D:\gyxx-flow`。
+- 在生产切换门禁前，不修改四个旧项目或现有计划任务。
+- 允许从旧项目做一次性只读复制；复制后的代码归属新项目并在新项目内改造、测试。
+- 运行时代码不得读取 `GYXX_LEGACY_*_ROOT`，不得导入或启动四个旧目录中的文件。
+- `D:\comment-data-collector` 不在迁移范围内；它与内容模块共享数据库表的字段
+  所有权以外部契约表示。
+- 真实 Cookie、密码、应用密钥、访问令牌和浏览器 Profile 不进入源码仓库。
+
+## 3. 方案
+
+选择模块化单体而不是原样拼目录或直接拆微服务。四个业务模块为：
+
+1. `content_marketing`：内容指标、评论、达人、周报/月报和登录维护。
+2. `product_commerce`：ERP、商品、画像、主图、市场排名、分析和告警。
+3. `shop_intelligence`：店铺及竞店周数据。
+4. `supply_chain`：采购确认、库存预警、补货和采购单更新。
+
+共享能力通过端口和适配器提供：文件产物、PostgreSQL、飞书、Hermes、浏览器和
+平台采集。业务模块不得导入其他业务模块的内部实现。
+
+每个业务模块包含自己的 `jobs/`、`collectors/`、`services/`、`models/` 和
+`resources/`。一次性复制阶段可以在模块内保留原相对结构以降低行为变化风险，但
+所有入口必须改为由模块清单定位，所有数据/日志/Profile/临时文件必须改由
+`RunContext` 与 `DataLayout` 提供。模块间共享只允许依赖 `core`、`workflow`、
+`adapters` 和显式契约。
+
+## 4. 运行模型
+
+调度器只调用稳定命令 `gyxx run <workflow-id>`。工作流引擎负责：
+
+- 创建 `run_id` 和 `RunContext`；
+- 获取工作流及资源锁；
+- 按依赖顺序执行步骤；
+- 记录步骤状态、退出码、耗时和错误；
+- 进行有上限的步骤级重试；
+- 生成不可变产物清单和 SHA-256；
+- 在 shadow 模式关闭生产写入和正式通知。
+
+Windows Task Scheduler 只是一个调度适配器。工作流和时间表的唯一事实来源是仓库
+内的声明式配置，因此可以生成 Windows、Cron 或其他平台的调度配置。
+
+## 5. 数据模型
+
+运行根目录由 `GYXX_DATA_ROOT` 指定，默认是仓库下 `var`：
+
+```text
+var/
+  data/raw/<domain>/<source>/<dataset>/business_date=<date>/run_id=<id>/
+  data/normalized/<domain>/<dataset>/schema_vN/
+  data/curated/<domain>/<dataset>/
+  data/exports/<consumer>/<workflow>/<date>/<run_id>/
+  data/evidence/<workflow>/<run_id>/
+  data/legacy/<source_project>/
+  runs/<workflow>/<yyyy>/<mm>/<dd>/<run_id>/
+  state/{browser_profiles,cookies,checkpoints,locks}/
+  logs/<workflow>/<yyyy>/<mm>/<dd>/
+  quarantine/
+  tmp/
+```
+
+原始数据只追加。每个产物都有 `artifact_id`、数据集、业务日期、schema 版本、行数、
+字节数、SHA-256、源路径和上游产物引用。历史文件先原样复制到 `legacy`，校验通过后
+再生成标准化数据。
+
+## 6. 安全和幂等
+
+- 配置只保存环境变量名或凭据引用，不保存密钥值。
+- 生产写入使用 `workflow + business_date + entity_id + schema_version` 幂等键。
+- 外部通知使用 outbox/回执，失败重试不会重复发送。
+- 飞书、数据库和通知都是可替换 Sink；shadow 模式使用测试 Sink。
+- 浏览器 Profile 使用独立资源锁，禁止不同工作流并发写同一 Profile。
+
+## 7. 迁移策略
+
+采用源码接管迁移：基线清单 -> 新基础设施 -> 源码只读复制 -> 新项目内路径与配置
+改造 -> 无旧目录测试 -> 影子运行 -> 单任务切换 -> 观察 -> 退役。旧命令适配器只
+用于盘点阶段，不能出现在最终工作流定义中。切换时只禁用一个旧任务，不删除；失败
+时关闭新任务并重新启用旧任务。旧项目至少保留只读 30 天。
+
+迁移顺序为 `shop_intelligence`、`supply_chain`、`content_marketing`、
+`product_commerce`。最后才合并真正重复的适配器，避免因名称相同而过早抽象不同语义
+的采集器。
+
+## 8. 验收原则
+
+工程验收依赖自动化测试、静态扫描、源码哈希清单、入口覆盖、旧根目录引用扫描、
+隔离旧目录的导入/dry-run、调度生成和历史回放。四模块所有纳入迁移范围的源码必须
+在新项目中有目标文件和来源哈希；所有真实入口必须标明 scheduled、manual、library
+或 intentionally-excluded，不能静默遗漏。生产验收必须以真实任务运行记录为证据：
+日任务连续 7 天，周任务连续 2 个周期，月任务通过历史月份回放。任何缺少证据的
+项目保持未勾选状态。
+
+## 9. 统一运行时适配器
+
+每个可执行脚本以完整 `script_id=<module>:<entry>` 作为运行时隔离主键。
+`config/runtime-bindings.json` 显式保存 130 个脚本的固定 CDP 端口，范围为
+`22000..22999`；配置加载时必须同时满足入口 100% 覆盖、端口唯一和回环地址约束。
+新增脚本只能领取未使用端口，已有脚本端口不得随发现顺序漂移。
+
+每个脚本的浏览器状态固定落到：
+
+```text
+<GYXX_DATA_ROOT>/state/browser/<module>/<script-name>-<id-digest>/
+  profile/
+  cookies.json
+  storage_state.json
+```
+
+同一脚本跨运行复用上述目录，不同脚本不共享端口、Profile、Cookie 或 storage state。
+顶层 workflow 和 `gyxx scripts run` 均由 `ModuleCommandAdapter` 注入绑定；嵌套 Python
+进程在业务模块导入前由 adapter-owned `sitecustomize` 按真实子脚本路径重新绑定，且会
+纠正父进程遗留的 `--cdp-port`、`--cdp-url` 和 `--user-data-dir` 参数。PowerShell/BAT
+需要浏览器时通过 `python -m gyxx_flow.runtime_exec env <script-id>` 读取目标叶子脚本绑定。
+
+外部系统策略是运行时硬边界，而不是更换原业务后端：
+
+- 飞书继续使用原来的 lark-cli profile、身份和原 OpenAPI 应用；适配器只透传并禁止误覆盖。
+- PostgreSQL 继续使用现有云端配置，统一映射 `PG_*`、`DB_*`、`AUTOFLOW_PG_*`，拒绝回环数据库。
+- Hermes 继续使用本机 HTTP/CLI，所有 HTTP 端点只允许 `localhost`、`127.0.0.1` 或 `::1`。
+- Cookie 和 storage state 通过原子替换保存；日志和绑定输出只包含路径、端口和计数，不包含值。
+
+dry-run 只解析和验证绑定，不创建 Profile/Cookie 目录，不探测或启动浏览器，也不触发任何
+飞书、数据库或 Hermes 调用。
+
+## 10. 采集数据统一分层
+
+`GYXX_DATA_ROOT` 表示整套运行数据根，不直接等同于 raw 目录。四个业务模块通过共享的 `ModuleDataPaths` 解析固定目录，模块脚本继续使用原有公开常量，避免业务逻辑与物理目录结构耦合：
+
+```text
+<GYXX_DATA_ROOT>/
+  data/
+    raw/<module>/          # 原始 JSON、CSV、Excel、下载文件、原始截图
+    normalized/<module>/   # 清洗并统一字段后的数据
+    curated/<module>/      # 聚合、分析和可供下游复用的数据
+    exports/<module>/      # Markdown、Excel 等面向人员或外部消费的报告
+    evidence/<module>/     # 产物清单、哈希和验收证据
+  state/<module>/          # checkpoint；浏览器状态仍按 script_id 隔离
+  logs/<module>/
+  tmp/<module>/
+```
+
+raw 数据保持追加语义；浏览器直接下载的 XLS/XLSX/CSV、接口原始 JSON 和采集截图归 raw，处理过程中的临时下载归 tmp，清洗结果归 normalized，跨来源汇总归 curated，最终 MD/XLSX 报告归 exports。现有 `data/raw/<module>/legacy` 只读保留，不在本次收口中移动或重写。路径对象只负责解析，不在导入或 dry-run 时创建目录，因此项目根和数据根可分别迁移。
diff --git a/docs/deployment.md b/docs/deployment.md
new file mode 100644
index 0000000..3bf941c
--- /dev/null
+++ b/docs/deployment.md
@@ -0,0 +1,39 @@
+# 部署手册
+
+## 环境要求
+
+- Windows 10/11 或 Windows Server，系统时区 `China Standard Time`
+- Python 3.12、`uv`、Windows Task Scheduler
+- 凭据由环境变量或外部密钥系统提供，不写入源码或清单
+
+## 安装
+
+```powershell
+cd D:\gyxx-flow
+uv sync --python 3.12 --extra test
+$env:GYXX_DATA_ROOT = 'D:\gyxx-flow\var'
+.\.venv\Scripts\python.exe -m gyxx_flow doctor --json
+.\.venv\Scripts\python.exe -m pytest
+```
+
+不再配置任何 `GYXX_LEGACY_*_ROOT`。运行代码和资源随 `gyxx_flow` 包部署，
+四个旧项目可以不挂载。商品模块需要独立配置时，可设置 `GYXX_PRODUCT_CONFIG`；
+该变量只指向新部署的配置文件。
+
+## 生成候选调度计划
+
+```powershell
+.\.venv\Scripts\python.exe -m gyxx_flow schedule plan `
+  --output D:\gyxx-flow\var\schedule-plan\candidate `
+  --start-date 2026-07-27 `
+  --python-executable D:\gyxx-flow\.venv\Scripts\python.exe
+```
+
+输出包含 21 个 XML、`install.ps1`、`plan.json` 和 `drift.json`。生成计划不会
+注册任务。只有生产门禁通过并获得明确授权后，才能人工审阅并逐个使用
+`install.ps1 -Apply -WorkflowId <id>`；安装器强制一次只处理一个任务。
+
+## 可迁移部署
+
+复制项目或安装 wheel 后，只需重新设置 `GYXX_DATA_ROOT` 和凭据。禁止在生产任务
+命令中出现旧项目盘符、用户目录解释器或旧项目工作目录。
diff --git a/docs/migration-runbook.md b/docs/migration-runbook.md
new file mode 100644
index 0000000..f847d0f
--- /dev/null
+++ b/docs/migration-runbook.md
@@ -0,0 +1,29 @@
+# 迁移手册
+
+## 当前边界
+
+- 四个旧项目的源码、数据和现有计划任务保持原状。
+- 工作流及其业务脚本已复制到 `src/gyxx_flow/modules/*/runtime`，运行时不引用旧项目。
+- 21 个定时工作流可生成候选系统任务；默认不应用。
+- 所有发现到的本地可执行脚本可通过 `gyxx scripts` 手工 dry-run 或执行。
+- 历史数据副本按模块落入新数据根，并保留数量、字节和 SHA-256 对账证据。
+
+## 工程验收
+
+1. 四份 `config/source-manifests/*.json` 的目标文件和 SHA-256 全部通过。
+2. 扫描源码、配置和启动器，确认无四个旧项目根目录或 `GYXX_LEGACY_*_ROOT`。
+3. 在不设置旧项目环境变量的进程中导入目录、列出脚本并 dry-run 21 个工作流。
+4. 执行 `uv run pytest`、`uv run gyxx doctor --json` 和 wheel 构建检查。
+5. 只生成调度计划，不注册、不禁用任何系统任务。
+
+## 逐任务生产切换
+
+1. 轮换历史明文凭据，并将新凭据放入外部密钥系统或环境变量。
+2. 对目标 workflow 做 dry-run 和 shadow 对账。
+3. 审核 run journal、artifact manifest、effect ledger 和 outbox。
+4. 取得负责人明确授权，只安装一个新任务并禁用对应旧任务；旧任务不删除。
+5. 日任务连续观察 7 天，周任务观察 2 个周期，月任务完成指定历史月回放。
+6. 数据库、飞书、文件产物和通知对账通过后，才迁移下一任务。
+
+推荐顺序：`shop_intelligence` → `supply_chain` → `content_marketing` →
+`product_commerce`。
diff --git a/docs/operations.md b/docs/operations.md
new file mode 100644
index 0000000..558d393
--- /dev/null
+++ b/docs/operations.md
@@ -0,0 +1,29 @@
+# 日常运维手册
+
+## 常用命令
+
+```powershell
+gyxx list --json
+gyxx scripts list --json
+gyxx doctor --json
+gyxx acceptance status --json
+gyxx run shop.metrics.weekly --date 2026-07-27
+gyxx backfill shop.metrics.weekly --from 2026-07-21 --to 2026-07-27
+gyxx scripts run shop_intelligence:runners/run_shop.py --date 2026-07-27
+```
+
+`run`、`backfill`、`scripts run` 默认 dry-run。`--execute` 才启动本地迁入脚本；
+Task Scheduler 使用 `--scheduled`，按 Asia/Shanghai 当日执行。
+
+## 数据与证据
+
+- `var/runs/.../<run_id>/run.json`：步骤、尝试、退出码和 trace
+- `var/state/ops/run-index/`：按 run_id/workflow/date/status 查询
+- `var/state/ops/effects/`：生产 Sink 幂等回执
+- `var/state/ops/outbox/messages/`：外部写入消息
+- `var/logs/`、`var/data/evidence/`：日志和截图证据
+- `var/state/profiles/`：浏览器 Profile
+- `var/state/locks/`：工作流、Profile 和共享资源锁
+
+effect 为 `ambiguous` 或长时间 `in_progress` 时，先人工核对外部系统；不要删除回执
+或强制重跑。手工脚本执行同样会产生统一 run journal 和 effect 记录。
diff --git a/docs/rollback.md b/docs/rollback.md
new file mode 100644
index 0000000..e53228c
--- /dev/null
+++ b/docs/rollback.md
@@ -0,0 +1,17 @@
+# 回滚手册
+
+## 单任务回滚
+
+1. 记录失败的新任务名、workflow ID、run_id 和业务日期。
+2. 禁用对应 `\GYXX\<workflow-id>`，保留任务定义和运行证据。
+3. 重新启用原计划任务，核对触发时间、账号和旧工作目录仍与基线一致。
+4. 检查 `state/ops/effects`；`in_progress` 或 `ambiguous` 必须先做外部对账。
+5. 检查 outbox；已 `sent` 的消息不得重发，`failed` 只能以同一幂等键 replay。
+6. 记录回滚结果和恢复时间，保留新旧日志。
+
+## 原则
+
+- 一次只切换或回滚一个 workflow。
+- 旧任务只禁用/启用，不删除；旧项目在全部观察期完成前保持只读可回滚。
+- 不删除新系统数据、run journal、effect receipt 或 outbox 消息。
+- 未完成对账时，不允许用直接运行脚本绕过 effect ledger。
diff --git a/final_report.md b/final_report.md
new file mode 100644
index 0000000..1635dd3
--- /dev/null
+++ b/final_report.md
@@ -0,0 +1,72 @@
+# GYXX Flow 源码迁移阶段报告
+
+## 当前结论
+
+核心采集脚本已经修改，但修改的是复制到 `D:\gyxx-flow` 的新项目副本；四个旧项目的
+原始代码和现有定时任务均未修改。新工作流不再通过环境变量、动态导入或子进程引用旧
+项目代码。
+
+当前可用能力：
+
+- 统一列出并 dry-run/执行 21 个定时工作流和 7 个正式手工工作流；
+- 统一发现并 dry-run/执行 131 个迁入后的本地脚本；
+- 生成 21 个 Windows Task Scheduler XML 和单任务安装器；
+- 将新数据按模块写入 `data/{raw,normalized,curated,exports,evidence}/<module>`，运行状态、
+  日志和临时文件写入 `state/logs/tmp`；
+- 更换项目根和数据根后继续完成入口发现与 dry-run。
+
+## 已完成
+
+- [x] 四模块源码实体迁移及逐文件来源/目标哈希清单
+- [x] 21 个计划任务到新项目本地入口的一一映射
+- [x] 131 个本地可执行脚本目录与统一手工入口
+- [x] 旧项目根目录、旧环境变量和旧配置回退清零
+- [x] 数据、日志、Profile、状态和临时目录迁入可配置数据根
+- [x] 敏感默认值清理、来源清单校验、重定位测试和全量自动化测试
+- [x] 只生成不应用的候选调度包；安装器强制逐任务应用
+- [x] 部署、迁移、运维和回滚手册更新
+- [x] 131 个脚本唯一 CDP 端口、独立可迁移 Profile/Cookie/storage state
+- [x] 原飞书透传、现有云 PostgreSQL 和本机 Hermes 的统一运行时边界
+
+## 未完成
+
+- [ ] 为每个外部写入证明 run_id 追踪和重跑幂等
+- [ ] 用户授权后的凭据轮换、逐任务生产切换和回滚演练
+- [ ] 日任务 7 天、周任务 2 个周期、月任务历史回放及最终对账
+
+`gyxx acceptance status` 在上述生产门禁完成前应继续返回 incomplete。这是有意保留的
+安全状态，不影响当前已具备的本地手工运行和候选任务注册能力。
+
+## 统一适配器补充验收（2026-07-27）
+
+- `config/runtime-bindings.json` 覆盖 131/131 脚本，CDP 端口 131 个且无重复。
+- 顶层 workflow/manual 与嵌套 Python/PowerShell 均按目标脚本 ID 重新绑定。
+- Cookie、storage state 原子保存并跨运行复用，全部位于可迁移数据根。
+- 生产源码中固定 `9222/18801/18802/18803`、源码目录 Profile 和临时即删 Profile 已清除。
+- 飞书身份/调用后端保持原样；三套数据库变量映射到现有云端，回环数据库被拒绝；
+  Hermes 非回环 URL 被拒绝。
+- Python 3.12 全量测试 `294 passed`；框架 Ruff、compileall、敏感信息和迁移哈希通过。
+
+## 采集数据目录收口补充验收（2026-07-27）
+
+- [x] 四模块统一使用 `data/{raw,normalized,curated,exports,evidence}/<module>`。
+- [x] JSON、CSV、XLS/XLSX、浏览器下载和原始截图进入 raw；清洗结果进入 normalized；
+  聚合结果进入 curated；最终 Markdown/Excel 报告进入 exports。
+- [x] checkpoint、缓存、锁、浏览器状态、日志和临时处理文件不再混入 raw 或源码目录。
+- [x] CLI/vendor 自定义输出受数据根边界校验，不能写到 cwd、源码树或数据根之外。
+- [x] 历史 `legacy` 前后均为 48,961 个文件、1,091,967,729 字节，无移动、删除或改写。
+- [x] raw 按 `run_id` 追加保留，不清理历史原始产物；浏览器 Profile/Cookie 在任务结束后保留复用。
+- [x] 空/纯空白数据根安全回退项目 `var`，不会把 cwd 当成数据根。
+- [x] 全量 `294 passed`，Ruff、compileall、PowerShell AST 和来源清单哈希通过。
+
+本次只修改 `D:\gyxx-flow` 内迁入副本和工程文件，没有修改四个旧项目，也没有注册、
+禁用或改动任何现有定时任务；未启动真实浏览器采集，未写飞书、云端数据库或 Hermes。
+
+## 四源项目增量重新对齐（2026-07-28）
+
+- [x] 内容营销 4 个源端修复文件及两份动态映射资源已重新对齐。
+- [x] 商品商业 ERP 天猫权威来源修复已重新对齐。
+- [x] 源端新增波塞冬历史销量回填脚本及测试已实体迁入并改接统一数据分层。
+- [x] 新脚本可通过统一脚本目录手工启动，使用独立 CDP 端口和可复用 Cookie 状态。
+- [x] 四份来源清单现覆盖 250 个文件；目标哈希、全量 294 项测试、18 项定向回归、compileall 和 doctor 通过。
+- [ ] Ruff 未安装在当前 `.venv`，未作为本轮通过项；P6/P7 的真实外部写入和生产切换门禁继续保持未完成。
diff --git a/plan.md b/plan.md
new file mode 100644
index 0000000..49b81d2
--- /dev/null
+++ b/plan.md
@@ -0,0 +1,126 @@
+# GYXX Flow 执行计划与验收清单
+
+规则：只有完成对应验证并保存证据后才把 `[ ]` 改为 `[x]`。每次更新复核
+`python -m pytest`、`gyxx acceptance status` 和本文件。
+
+## P0 设计与迁移基线
+
+- [x] P0.1 固化模块化单体、数据分层、影子运行和逐任务回滚设计。
+  - 文件：`design.md`
+  - 验证：设计包含目标、边界、模块、数据、安全、迁移和验收。
+- [x] P0.2 建立项目骨架、依赖文件和忽略规则。
+  - 文件：`pyproject.toml`、`.gitignore`、`README.md`、`src/`、`tests/`
+  - 验证：干净环境可安装，CLI 能显示帮助。
+- [x] P0.3 生成四个旧项目的代码、任务、数据基线清单。
+  - 文件：`var/baseline/<timestamp>/`
+  - 验证：包含 21 个实际任务、代码哈希、数据汇总；不包含密钥值。
+- [x] P0.4 建立敏感信息扫描并记录轮换门禁。
+  - 文件：`src/gyxx_flow/security/`、`config/secrets.example.env`
+  - 验证：仓库扫描无明文凭据；生产切换要求凭据轮换确认。
+
+## P1 核心运行与数据基础设施
+
+- [x] P1.1 TDD 实现可移植路径配置，不允许业务代码硬编码旧盘符。
+- [x] P1.2 TDD 实现 `RunContext`、稳定 `run_id` 和业务日期校验。
+- [x] P1.3 TDD 实现原子 JSON 写入、SHA-256 和 artifact manifest。
+- [x] P1.4 TDD 实现 raw/normalized/curated/export/evidence/state/log/tmp 目录解析。
+- [x] P1.5 TDD 实现结构化运行及步骤状态记录。
+- [x] P1.6 TDD 实现文件锁、工作流锁和命名资源锁。
+
+## P2 工作流引擎与 CLI
+
+- [x] P2.1 TDD 实现工作流、步骤、依赖、超时和重试模型。
+- [x] P2.2 TDD 实现命令步骤及无副作用 dry-run。
+- [x] P2.3 TDD 实现 shadow 模式，禁止生产 Sink 和正式通知。
+- [x] P2.4 TDD 实现 `gyxx list`、`gyxx run`、`gyxx backfill`。
+- [x] P2.5 TDD 实现失败恢复、步骤级重跑和最终退出码。
+- [x] P2.6 建立 `ops` 运行记录和 outbox 接口。
+
+## P3 工作流目录与调度
+
+- [x] P3.1 建立四个模块及模块依赖约束测试。
+- [x] P3.2 映射 21 个实际计划任务到稳定 workflow ID。
+- [x] P3.3 登记未调度/手工工作流，保持原触发语义。
+- [x] P3.4 建立声明式 schedule 配置及 schema 校验。
+- [x] P3.5 生成 Windows Task Scheduler XML/PowerShell 安装计划，但默认不应用。
+- [x] P3.6 实现 `gyxx schedule plan`，检测系统任务漂移。
+
+## P4 迁移和验收工具
+
+- [x] P4.1 实现旧命令适配器，所有旧根目录均通过配置注入。
+- [x] P4.2 实现历史数据只复制、不删除的迁移计划及校验清单。
+- [x] P4.3 实现文件数量、字节数、SHA-256 对账。
+- [x] P4.4 实现 shadow 结果主键、行数、指标和错误对比。
+- [x] P4.5 实现 `gyxx doctor` 环境、解释器、CLI、路径和权限预检。
+- [x] P4.6 实现 `gyxx acceptance status` 和机器可读验收报告。
+- [x] P4.7 生成四个旧项目的源码迁移清单；每个文件记录来源哈希、目标路径和处置类型。
+
+## P5 分模块源码迁移
+
+- [x] P5.1 将 `shop_intelligence` 全部业务源码、资源和入口复制并改造到新模块；两个周任务可独立运行。
+- [x] P5.2 将 `supply_chain` 全部业务源码、资源和入口复制并改造到新模块；三个定时和全部手工入口可独立运行。
+- [x] P5.3 将 `content_marketing` 全部业务源码、资源和入口复制并改造到新模块；全部定时和手工入口可独立运行。
+- [x] P5.4 将 `product_commerce` 全部业务源码、资源、内置 vendors 和入口复制并改造到新模块；全部定时和手工入口可独立运行。
+- [x] P5.5 消除营销二维码写入商品项目、商品配置回退到 Auto Flow，以及任何跨业务模块内部导入。
+- [x] P5.6 将重复的飞书、PostgreSQL、Hermes、浏览器能力收敛为适配器，业务行为保持可回放对账。
+  - 证据：`src/gyxx_flow/adapters/integration.py`、`browser.py`、`bootstrap.py`、`runtime_exec.py`；原飞书后端不变，数据库只允许云端，Hermes 只允许本机。
+- [x] P5.7 删除最终工作流中的 `DeferredLegacyCommandStep`，运行时不再要求 `GYXX_LEGACY_*_ROOT`。
+- [x] P5.8 全量入口分类完成：scheduled、manual、library、intentionally-excluded 均有证据，不遗漏脚本。
+
+## P6 自动化工程验收
+
+- [x] P6.1 21 个计划任务逐项有且只有一个新项目本地入口映射，没有意外新增。
+- [x] P6.2 源码、配置、启动命令和动态导入中均不存在四个旧项目的运行时依赖。
+- [x] P6.3 临时隐藏四个旧项目路径并更换项目根、数据根后，入口导入和 dry-run 全部通过。
+- [ ] P6.4 每个迁移后的业务运行可从 run_id 追溯输入、输出、日志和外部写入。
+- [ ] P6.5 迁移后的业务代码重跑不会重复写数据库、飞书或正式通知。
+- [x] P6.6 旧数据复制前后数量、大小和 SHA-256 一致。
+- [x] P6.7 迁移后的浏览器 Profile 和共享资源全部进入新数据根，且并发锁验证通过。
+- [x] P6.8 源码仓库敏感信息扫描通过。
+- [x] P6.9 全量测试、静态检查、源码覆盖清单和代码评审无阻断项。
+
+## P7 生产切换门禁
+
+- [ ] P7.1 用户确认迁移基线、密钥轮换和第一个切换任务。
+- [ ] P7.2 单次只切一个任务；旧任务只禁用、不删除。
+- [ ] P7.3 回滚演练证明能关闭新任务并重新启用旧任务。
+- [ ] P7.4 日任务真实连续成功 7 天。
+- [ ] P7.5 周任务真实连续成功 2 个周期。
+- [ ] P7.6 月任务指定历史月份回放通过。
+- [ ] P7.7 生产通知无重复，数据库/飞书对账通过。
+- [ ] P7.8 旧项目只读保留期满并取得退役确认。
+
+## P8 完成
+
+- [ ] P8.1 所有上述检查项均有证据且已勾选。
+- [x] P8.2 按源码完整迁移的新口径重新生成 `review.md` 和 `final_report.md`。
+- [x] P8.3 按独立运行的新口径更新部署、迁移、回滚和日常运维手册。
+
+## P9 每脚本运行时隔离与外部系统统一边界
+
+- [x] P9.1 为 131 个可执行脚本建立显式、稳定、唯一的 `22000..22999` CDP 端口分配。
+- [x] P9.2 每脚本独立 Profile、Cookie、storage state，全部位于 `GYXX_DATA_ROOT/state/browser`。
+- [x] P9.3 顶层 workflow、手工脚本和嵌套 Python/PowerShell 子脚本均按真实 script ID 重新绑定。
+- [x] P9.4 Cookie/storage state 支持首次为空、原子保存和后续运行复用，内容不进入日志或源码。
+- [x] P9.5 飞书保持原 lark-cli profile、身份和 OpenAPI 应用，不切换后端或凭据来源。
+- [x] P9.6 `PG_*`、`DB_*`、`AUTOFLOW_PG_*` 统一映射现有云端 PostgreSQL，拒绝本地回环数据库。
+- [x] P9.7 Hermes 保持本机 HTTP/CLI，非回环 Hermes URL 在业务调用前失败。
+- [x] P9.8 移除生产代码固定 9222/18801/18802/18803 和源码目录/临时 Profile，刷新迁移哈希。
+- [x] P9.9 Python 3.12 全量 `294 passed`，compileall、敏感信息、端口和旧根依赖扫描通过；历史 Ruff 证据保留。
+
+## P10 采集数据统一分层与目录收口
+
+- [x] P10.1 四个模块的新产物统一使用 `<GYXX_DATA_ROOT>/data/{raw,normalized,curated,exports,evidence}/<module>`，运行状态、日志和临时文件分别使用 `state/`、`logs/`、`tmp/`。
+- [x] P10.2 JSON、Markdown、CSV、XLS/XLSX、浏览器下载和截图均由模块路径边界定位，不写入源码目录、模块目录或当前工作目录。
+- [x] P10.3 `var/data/raw/<module>/legacy` 历史数据不移动、不删除、不改写，迁移前后文件数和总字节数一致。
+- [x] P10.4 四模块保留原业务脚本公开路径常量，统一改接共享 `ModuleDataPaths`，新增模块只需声明模块名即可扩展。
+- [x] P10.5 更换 `GYXX_DATA_ROOT` 后所有产物路径整体迁移，路径解析不创建目录、不访问四个旧项目。
+- [x] P10.6 路径红测、静态旁路扫描、全量测试、Ruff、compileall、敏感信息和旧根依赖扫描全部通过，清单与验收报告已刷新。
+
+## P11 四源项目增量 Bug 修复重新对齐（2026-07-28）
+
+- [x] P11.1 以四份来源清单 SHA-256 为基线，识别源端变更、新增、删除和目标端二次改造。
+- [x] P11.2 合并内容营销模块发布链接非空触发、星图普通失败不熔断和精确续跑修复。
+- [x] P11.3 合并商品模块“ERP 编码只读取天猫分组”的权威来源修复，并迁入波塞冬历史销量回填工具。
+- [x] P11.4 刷新来源/目标哈希、两份动态业务映射、迁移清单和新增回归测试，不纳入运行数据、临时文件或敏感状态。
+- [x] P11.5 定向测试、全量测试、compileall、敏感信息、旧根依赖和 doctor 通过；acceptance 正确保留 P6/P7 生产门禁为 incomplete，Ruff 因当前虚拟环境未安装而未执行。
diff --git a/pyproject.toml b/pyproject.toml
new file mode 100644
index 0000000..42a24b8
--- /dev/null
+++ b/pyproject.toml
@@ -0,0 +1,63 @@
+[build-system]
+requires = ["setuptools>=77"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "gyxx-flow"
+version = "0.1.0"
+description = "Portable workflow orchestration for GYXX business automation"
+readme = "README.md"
+requires-python = ">=3.12"
+dependencies = [
+  "browserforge>=1.2.4",
+  "curl-cffi>=0.15.0",
+  "httpx>=0.27",
+  "imageio-ffmpeg>=0.6.0",
+  "langchain>=0.3.0",
+  "langchain-anthropic>=0.3.0",
+  "langchain-core>=0.3.0",
+  "langchain-openai>=0.2.0",
+  "langgraph>=0.2.0",
+  "lark-oapi>=1.0.0",
+  "lxml>=5.0",
+  "msgspec>=0.21.1",
+  "openpyxl>=3.1.5",
+  "ormsgpack>=1.2.0",
+  "pandas>=3.0.3",
+  "patchright>=1.60.1",
+  "Pillow>=10",
+  "playwright>=1.60.0",
+  "psycopg[binary,pool]>=3.1",
+  "psycopg2-binary>=2.9.12",
+  "psutil>=5.9",
+  "python-dotenv>=1.0.0",
+  "requests>=2.31.0",
+  "scrapling[all]>=0.4.9",
+  "selenium>=4.20",
+  "xlrd>=2.0.1",
+]
+
+[project.optional-dependencies]
+test = ["pytest>=8.0"]
+
+[project.scripts]
+gyxx = "gyxx_flow.cli:main"
+
+[tool.setuptools.packages.find]
+where = ["src"]
+
+[tool.setuptools.package-data]
+"gyxx_flow.modules.content_marketing" = ["runtime/**/*"]
+"gyxx_flow.modules.product_commerce" = ["runtime/**/*"]
+"gyxx_flow.modules.shop_intelligence" = ["runtime/**/*"]
+"gyxx_flow.modules.supply_chain" = ["runtime/**/*"]
+
+[tool.setuptools.exclude-package-data]
+"*" = ["**/__pycache__/**", "**/*.pyc", "**/*.pyo"]
+
+[tool.pytest.ini_options]
+pythonpath = ["src"]
+testpaths = ["tests"]
+
+[tool.ruff.lint]
+select = ["E4", "E7", "E9", "F", "I"]
diff --git a/review.md b/review.md
new file mode 100644
index 0000000..0ebbf6c
--- /dev/null
+++ b/review.md
@@ -0,0 +1,90 @@
+# 工程复核报告
+
+## 结论
+
+四个旧项目的工作流、定时入口及纳入范围的业务源码已实体复制到新项目，并完成路径、
+配置、数据目录和启动方式改造。运行时入口只解析
+`src/gyxx_flow/modules/<module>/runtime`，不再导入或启动四个旧目录中的代码。
+
+新项目已经具备统一手工运行和生成 21 个 Windows 定时任务定义的工程能力；本轮没有
+注册新任务，也没有禁用、删除或修改旧任务。生产安全切换仍受 P6.4、P6.5 和
+P7 门禁约束，不能把“可生成/可 dry-run”表述为“已经生产投运”。
+
+## 源码接管证据
+
+| 模块 | 来源清单文件 | 迁移文件 | 可执行脚本 |
+|---|---:|---:|---:|
+| content_marketing | 90 | 90 + 1 个新路径模块 | 56 |
+| product_commerce | 101 | 101 + 1 个新路径模块 | 49 |
+| shop_intelligence | 22 | 22 | 8 |
+| supply_chain | 37 | 37 | 18 |
+| 合计 | 250 | 256（含 2 个路径模块和 4 个运行时 bootstrap） | 131 |
+
+- 四份 `config/source-manifests/*.json` 逐文件记录来源相对路径、来源 SHA-256、目标
+  相对路径、目标 SHA-256、分类和是否改造。
+- 对四个旧目录做只读复核并同步两份动态业务映射后，250/250 个纳入文件来源哈希与
+  当前迁移清单一致；旧项目没有被本次迁移写入。
+- 源码、配置、启动器及动态入口扫描中，四个旧根目录、`GYXX_LEGACY_*_ROOT`、
+  `DeferredLegacyCommandStep` 和 `AUTO_FLOW_CONFIG` 的运行时命中均为 0。
+- 项目和数据根整体迁移到临时目录后，目录发现、131 个脚本发现及四模块代表工作流
+  dry-run 均通过。
+
+## 入口与调度证据
+
+- 29 个目录项：21 个 scheduled、7 个 manual、1 个有证据的 unavailable。
+- `gyxx scripts list` 发现 131 个新项目本地 Python/BAT/PowerShell 入口。
+- `gyxx run`、`gyxx backfill` 和 `gyxx scripts run` 默认 dry-run；`--execute` 才启动
+  迁入后的本地副本。
+- 最终候选调度包包含 21 个 XML，动作统一为
+  `D:\gyxx-flow\.venv\Scripts\python.exe -m gyxx_flow run <id> --scheduled`。
+- 安装器默认只展示计划；即使显式 `-Apply` 也必须提供一个 `-WorkflowId`，不允许一次
+  批量切换。
+
+## 自动化验证
+
+- 全量 pytest：221 passed；仅有复制脚本内嵌 JavaScript 正则产生的非阻断
+  `SyntaxWarning`。
+- 新框架与测试 Ruff：通过。
+- 所有可执行 runtime 的 fatal Ruff 规则 `E9/F63/F7/F82`：通过；一个不独立执行的
+  上游源码摘录已在 manifest 中标记 `source_resource`。
+- Python compileall、PowerShell AST、来源清单哈希、旧路径扫描和敏感信息扫描：通过。
+- 历史数据：48,964 个文件、1,091,997,415 字节，复制前后聚合 SHA-256 一致。
+
+## 适配器状态与尚未通过的门禁
+
+- P5.6 已通过：131 个脚本统一绑定唯一 CDP、独立 Profile/Cookie/storage state；飞书
+  保持原身份/后端，PostgreSQL 只允许现有云端，Hermes 只允许本机。
+- P6.4/P6.5：编排层已有 run journal、effect ledger 和 outbox，但尚未证明每个旧业务
+  脚本的所有内部外部写入都可追踪且全量幂等。
+- P7：需要用户授权、凭据轮换、逐任务切换、真实 7 天/2 周观察和数据/通知对账。
+
+因此工程“源码归属和独立入口”已通过，生产“全量安全执行和切换”尚未通过。
+
+## 四源项目增量对齐复核（2026-07-28）
+
+- 内容营销已合并“发布链接非空即进入采集”、星图普通达人失败不触发全局熔断、按款式和记录精确续跑。
+- 商品商业已合并“ERP 编码只认天猫分组”，并把源端新增的波塞冬历史销量回填脚本完整迁入；脚本不引用旧项目。
+- 波塞冬回填按 `raw → normalized → curated → exports` 分层，数据库继续使用云端配置，飞书仍使用原身份和项目内写入器。
+- 新脚本登记独立 CDP 端口 `22130`，Profile/Cookie/storage state 继续由统一绑定派生并可迁移复用。
+- 本轮全量 `294 passed`，修复定向测试 18 项通过，compileall、doctor、敏感信息与来源/目标哈希验证通过；当前虚拟环境未安装 Ruff，生产门禁仍保持 incomplete。
+
+## 采集数据统一目录复核（2026-07-27）
+
+- 四模块路径层统一改接共享 `DataLayout.for_module()` / `ModuleDataPaths`；新采集产物固定进入
+  `data/{raw,normalized,curated,exports,evidence}/<module>`，状态、日志和临时文件分别进入
+  `state/<module>`、`logs/<module>` 和 `tmp/<module>`。
+- 内容模块的动态 mapping 和 V2 结果进入 normalized，跨平台 run_all 报告进入 curated，
+  checkpoint 进入 state；商品模块的 checkpoint、失败日志、调试文件和 vendor 下载已收口；
+  店铺调试输出进入 tmp；供应链缓存/锁/status 进入 state，原始下载、临时处理和最终报表分别
+  进入 raw、tmp 和 exports。
+- 所有审计到的任意 `--output`、`--data-root` 和 vendor 环境输出均限制在
+  `GYXX_DATA_ROOT` 内；绝对外部路径或 `..` 逃逸会在写入前失败。
+- `var/data/raw/<module>/legacy` 前后均为 48,961 个文件、1,091,967,729 字节；四模块分项
+  文件数和字节数完全一致，且不存在旁路 `var/raw` 或 `var/exports`。证据见
+  `var/evidence/unified-data-layout-20260727.json`。
+- 供应链 raw 产物按 `run_id=<id>` 追加保存，编排器只读取本次运行目录且不会清理 raw；
+  采购更新触发任务放在 state，脚本结束只关闭浏览器进程并保留 Profile/Cookie 供复用。
+- 空字符串或纯空白 `GYXX_DATA_ROOT` 会安全回退项目 `var`，不会把 cwd 当数据根。
+- 全量回归为 294 passed；新代码 Ruff、compileall、PowerShell AST、四份来源清单目标哈希、
+  敏感信息和旧根依赖扫描全部通过。5 条 warning 是迁入脚本内嵌 JavaScript 正则的既有
+  `SyntaxWarning`，不影响本次目录验收。
diff --git a/src/gyxx_flow/__init__.py b/src/gyxx_flow/__init__.py
new file mode 100644
index 0000000..dc0b838
--- /dev/null
+++ b/src/gyxx_flow/__init__.py
@@ -0,0 +1,3 @@
+"""GYXX Flow package."""
+
+__version__ = "0.1.0"
diff --git a/src/gyxx_flow/__main__.py b/src/gyxx_flow/__main__.py
new file mode 100644
index 0000000..c9fbbbc
--- /dev/null
+++ b/src/gyxx_flow/__main__.py
@@ -0,0 +1,6 @@
+"""Run GYXX Flow with ``python -m gyxx_flow``."""
+
+from gyxx_flow.cli import main
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/acceptance.py b/src/gyxx_flow/acceptance.py
new file mode 100644
index 0000000..3b10a51
--- /dev/null
+++ b/src/gyxx_flow/acceptance.py
@@ -0,0 +1,211 @@
+"""Machine-readable acceptance evidence derived from repository state."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import re
+from dataclasses import asdict, dataclass
+from pathlib import Path
+
+from gyxx_flow.catalog import CatalogError, WorkflowCatalog
+from gyxx_flow.core.config import Settings
+from gyxx_flow.script_catalog import ScriptCatalog
+from gyxx_flow.security import scan_repository
+
+_CHECKLIST = re.compile(r"^\s*-\s*\[([ xX])\]\s+(P\d+\.\d+)\b", re.MULTILINE)
+
+
+@dataclass(frozen=True, slots=True)
+class PlanItem:
+    item_id: str
+    completed: bool
+
+
+@dataclass(frozen=True, slots=True)
+class AcceptanceReport:
+    items: tuple[PlanItem, ...]
+    checks: dict[str, bool]
+
+    @property
+    def is_complete(self) -> bool:
+        return bool(self.items) and all(item.completed for item in self.items) and all(
+            self.checks.values()
+        )
+
+    def as_dict(self) -> dict[str, object]:
+        completed = sum(item.completed for item in self.items)
+        return {
+            "schema_version": 1,
+            "is_complete": self.is_complete,
+            "summary": {
+                "total": len(self.items),
+                "completed": completed,
+                "pending": len(self.items) - completed,
+            },
+            "items": [asdict(item) for item in self.items],
+            "checks": dict(sorted(self.checks.items())),
+        }
+
+
+def parse_plan_checklist(path: Path) -> tuple[PlanItem, ...]:
+    text = Path(path).read_text(encoding="utf-8")
+    items = tuple(
+        PlanItem(item_id=match.group(2), completed=match.group(1).casefold() == "x")
+        for match in _CHECKLIST.finditer(text)
+    )
+    if len({item.item_id for item in items}) != len(items):
+        raise ValueError("plan contains duplicate acceptance item IDs")
+    return items
+
+
+def build_acceptance_report(settings: Settings) -> AcceptanceReport:
+    project_root = settings.project_root
+    items = parse_plan_checklist(project_root / "plan.md")
+    checks = {
+        "catalog_21_tasks": _catalog_has_21_tasks(project_root),
+        "baseline_21_tasks": _baseline_has_21_tasks(settings.data_root),
+        "native_entrypoints_local": _native_entrypoints_are_local(project_root),
+        "runtime_sources_decoupled": _runtime_sources_are_decoupled(project_root),
+        "runnable_script_catalog": _runnable_script_catalog_is_complete(),
+        "source_manifests_verified": _source_manifests_are_verified(project_root),
+        "secret_scan_clean": not scan_repository(project_root),
+    }
+    return AcceptanceReport(items, checks)
+
+
+def _catalog_has_21_tasks(project_root: Path) -> bool:
+    try:
+        catalog = WorkflowCatalog.load(project_root / "config")
+    except Exception:
+        return False
+    scheduled = catalog.scheduled_workflows()
+    return len(scheduled) == 21 and len(catalog.schedules) == 21
+
+
+def _baseline_has_21_tasks(data_root: Path) -> bool:
+    candidates = sorted((Path(data_root) / "baseline").glob("*/manifest.json"))
+    if not candidates:
+        return False
+    try:
+        payload = json.loads(candidates[-1].read_text(encoding="utf-8"))
+        tasks = payload["scheduled_tasks"]
+        return tasks["actual_count"] == 21 and len(tasks["tasks"]) == 21
+    except (OSError, json.JSONDecodeError, KeyError, TypeError):
+        return False
+
+
+def _native_entrypoints_are_local(project_root: Path) -> bool:
+    try:
+        catalog = WorkflowCatalog.load(project_root / "config")
+        for workflow in catalog.workflows:
+            if workflow.trigger == "unavailable":
+                continue
+            target = (
+                project_root
+                / "src"
+                / "gyxx_flow"
+                / "modules"
+                / workflow.module
+                / "runtime"
+                / workflow.entry
+            ).resolve(strict=True)
+            if not target.is_file() or not target.is_relative_to(project_root):
+                return False
+        config_text = (project_root / "config" / "workflows.json").read_text(
+            encoding="utf-8"
+        )
+        return "legacy" not in config_text.casefold()
+    except (OSError, ValueError, CatalogError):
+        return False
+
+
+def _runnable_script_catalog_is_complete() -> bool:
+    try:
+        scripts = ScriptCatalog.discover_default().scripts
+    except (OSError, ValueError):
+        return False
+    return len(scripts) >= 100 and len({item.module for item in scripts}) == 4
+
+
+def _runtime_sources_are_decoupled(project_root: Path) -> bool:
+    module_root = project_root / "src" / "gyxx_flow" / "modules"
+    forbidden = (
+        "d:\\yingxiaoyunying",
+        "d:\\shop-data-flow",
+        "d:\\product-collector-analyze-flow",
+        "e:\\auto-flow",
+        "gyxx_legacy_",
+        "deferredlegacycommandstep",
+    )
+    try:
+        for path in module_root.rglob("*"):
+            if not path.is_file() or path.suffix.casefold() in {".pyc", ".pyo"}:
+                continue
+            if path.is_symlink() or not path.resolve().is_relative_to(project_root):
+                return False
+            content = path.read_bytes()
+            if b"\x00" in content:
+                continue
+            text: str | None = None
+            for encoding in ("utf-8-sig", "gb18030"):
+                try:
+                    text = content.decode(encoding)
+                    break
+                except UnicodeDecodeError:
+                    continue
+            if text is None:
+                continue
+            normalized = text.casefold().replace("/", "\\")
+            while "\\\\" in normalized:
+                normalized = normalized.replace("\\\\", "\\")
+            if any(value in normalized for value in forbidden):
+                return False
+    except OSError:
+        return False
+    return True
+
+
+def _source_manifests_are_verified(project_root: Path) -> bool:
+    modules = (
+        "content_marketing",
+        "product_commerce",
+        "shop_intelligence",
+        "supply_chain",
+    )
+    try:
+        for module in modules:
+            payload = json.loads(
+                (
+                    project_root
+                    / "config"
+                    / "source-manifests"
+                    / f"{module}.json"
+                ).read_text(encoding="utf-8-sig")
+            )
+            if payload.get("schema_version") != 1 or payload.get("module") != module:
+                return False
+            files = payload.get("files")
+            if not isinstance(files, list) or not files:
+                return False
+            for item in files:
+                target = project_root.joinpath(*Path(item["target_relative_path"]).parts)
+                resolved = target.resolve(strict=True)
+                if (
+                    target.is_symlink()
+                    or not resolved.is_relative_to(project_root)
+                    or not resolved.is_file()
+                    or _sha256(resolved) != item["target_sha256"]
+                ):
+                    return False
+    except (OSError, json.JSONDecodeError, KeyError, TypeError, ValueError):
+        return False
+    return True
+
+
+def _sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as stream:
+        for chunk in iter(lambda: stream.read(1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
diff --git a/src/gyxx_flow/adapters/__init__.py b/src/gyxx_flow/adapters/__init__.py
new file mode 100644
index 0000000..d0d1511
--- /dev/null
+++ b/src/gyxx_flow/adapters/__init__.py
@@ -0,0 +1,45 @@
+"""Replaceable infrastructure adapters exposed to business modules."""
+
+from .browser import BrowserCookieStore, BrowserProfileLease, BrowserProfileManager
+from .external import (
+    FeishuOutboxAdapter,
+    HermesCommandAdapter,
+    PostgresOutboxAdapter,
+    ReloginOutboxAdapter,
+)
+from .integration import (
+    RuntimeIntegrationBinding,
+    RuntimeIntegrationCatalog,
+    RuntimeIntegrationError,
+    RuntimeServicePolicy,
+    binding_from_environment,
+    environment_for_child_script,
+)
+from .native import (
+    DeferredModuleCommandStep,
+    ModuleCommandAdapter,
+    ModuleCommandFactory,
+    ModuleSourceError,
+    ModuleSourceRoots,
+)
+
+__all__ = [
+    "BrowserCookieStore",
+    "BrowserProfileLease",
+    "BrowserProfileManager",
+    "DeferredModuleCommandStep",
+    "FeishuOutboxAdapter",
+    "HermesCommandAdapter",
+    "ModuleCommandAdapter",
+    "ModuleCommandFactory",
+    "ModuleSourceError",
+    "ModuleSourceRoots",
+    "PostgresOutboxAdapter",
+    "ReloginOutboxAdapter",
+    "RuntimeIntegrationBinding",
+    "RuntimeIntegrationCatalog",
+    "RuntimeIntegrationError",
+    "RuntimeServicePolicy",
+    "binding_from_environment",
+    "environment_for_child_script",
+]
diff --git a/src/gyxx_flow/adapters/bootstrap.py b/src/gyxx_flow/adapters/bootstrap.py
new file mode 100644
index 0000000..0416824
--- /dev/null
+++ b/src/gyxx_flow/adapters/bootstrap.py
@@ -0,0 +1,83 @@
+"""Automatic rebinding for Python scripts launched by migrated orchestrators."""
+
+from __future__ import annotations
+
+import os
+import sys
+from pathlib import Path
+
+from .integration import (
+    RuntimeIntegrationBinding,
+    RuntimeIntegrationCatalog,
+    RuntimeIntegrationError,
+)
+
+
+def bootstrap_current_process() -> RuntimeIntegrationBinding | None:
+    module = os.environ.get("GYXX_MODULE_ID", "").strip()
+    module_root_value = os.environ.get("GYXX_MODULE_ROOT", "").strip()
+    project_root = os.environ.get("GYXX_PROJECT_ROOT", "").strip()
+    data_root = os.environ.get("GYXX_DATA_ROOT", "").strip()
+    if not all((module, module_root_value, project_root, data_root)):
+        return None
+    if not sys.argv or not sys.argv[0] or sys.argv[0] in {"-c", "-m"}:
+        return None
+    module_root = Path(module_root_value).expanduser().resolve()
+    candidate = Path(sys.argv[0])
+    if not candidate.is_absolute():
+        candidate = Path.cwd() / candidate
+    candidate = candidate.expanduser().resolve()
+    try:
+        entry = candidate.relative_to(module_root).as_posix()
+    except ValueError:
+        return None
+    script_id = f"{module}:{entry}"
+    catalog = RuntimeIntegrationCatalog.load_default(
+        project_root=project_root,
+        data_root=data_root,
+    )
+    if script_id not in catalog.script_ids:
+        return None
+    binding = catalog.binding_for(script_id)
+    os.environ.update(catalog.environment_for(script_id, os.environ))
+    _rewrite_browser_arguments(binding)
+    return binding
+
+
+def _rewrite_browser_arguments(binding: RuntimeIntegrationBinding) -> None:
+    replacements = {
+        "--user-data-dir": str(binding.profile_dir),
+        "--cdp-url": binding.cdp_url,
+        "--cdp-port": str(binding.cdp_port),
+        "--debug-port": str(binding.cdp_port),
+    }
+    index = 1
+    while index < len(sys.argv):
+        argument = sys.argv[index]
+        replaced = False
+        for flag, value in replacements.items():
+            if argument == flag and index + 1 < len(sys.argv):
+                sys.argv[index + 1] = value
+                index += 2
+                replaced = True
+                break
+            if argument.startswith(f"{flag}="):
+                sys.argv[index] = f"{flag}={value}"
+                index += 1
+                replaced = True
+                break
+        if not replaced:
+            index += 1
+
+
+def safe_bootstrap_current_process() -> RuntimeIntegrationBinding | None:
+    managed = bool(os.environ.get("GYXX_PROJECT_ROOT", "").strip())
+    try:
+        return bootstrap_current_process()
+    except RuntimeIntegrationError:
+        if managed:
+            raise
+        return None
+
+
+__all__ = ["bootstrap_current_process", "safe_bootstrap_current_process"]
diff --git a/src/gyxx_flow/adapters/browser.py b/src/gyxx_flow/adapters/browser.py
new file mode 100644
index 0000000..e4ff48e
--- /dev/null
+++ b/src/gyxx_flow/adapters/browser.py
@@ -0,0 +1,97 @@
+"""Portable, cross-process browser profile ownership."""
+
+from __future__ import annotations
+
+import re
+from contextlib import contextmanager
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any, Iterator
+
+from gyxx_flow.core.artifacts import atomic_write_json
+from gyxx_flow.core.layout import DataLayout
+from gyxx_flow.core.locks import LockManager
+
+_PROFILE_ID = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._-]*$")
+
+
+@dataclass(frozen=True, slots=True)
+class BrowserProfileLease:
+    profile_id: str
+    profile_path: Path
+
+
+class BrowserProfileManager:
+    def __init__(
+        self, data_root: Path, *, lock_manager: LockManager | None = None
+    ) -> None:
+        self._layout = DataLayout(data_root)
+        self._locks = lock_manager or LockManager(
+            self._layout.root / "state" / "locks"
+        )
+
+    @contextmanager
+    def acquire(
+        self,
+        profile_id: str,
+        *,
+        owner: str,
+        timeout_seconds: float = 30.0,
+    ) -> Iterator[BrowserProfileLease]:
+        if not isinstance(profile_id, str) or not _PROFILE_ID.fullmatch(profile_id):
+            raise ValueError("invalid browser profile id")
+        path = self._layout.state("browser_profiles", profile_id)
+        with self._locks.acquire(
+            f"browser:{profile_id}", owner=owner, timeout_seconds=timeout_seconds
+        ):
+            path.mkdir(parents=True, exist_ok=True)
+            yield BrowserProfileLease(profile_id, path)
+
+
+@dataclass(frozen=True, slots=True)
+class BrowserCookieStore:
+    """Playwright-compatible cookie/storage state persisted with atomic replace."""
+
+    cookie_file: Path
+    storage_state_file: Path
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "cookie_file", Path(self.cookie_file).expanduser().resolve())
+        object.__setattr__(
+            self,
+            "storage_state_file",
+            Path(self.storage_state_file).expanduser().resolve(),
+        )
+
+    def load_cookies(self) -> list[dict[str, Any]]:
+        payload = self._read(self.cookie_file, default=[])
+        if not isinstance(payload, list) or not all(isinstance(item, dict) for item in payload):
+            raise ValueError("browser cookie file must contain a list of objects")
+        return payload
+
+    def save_cookies(self, cookies: list[dict[str, Any]]) -> Path:
+        if not isinstance(cookies, list) or not all(isinstance(item, dict) for item in cookies):
+            raise ValueError("browser cookies must be a list of objects")
+        return atomic_write_json(self.cookie_file, cookies)
+
+    def load_storage_state(self) -> dict[str, Any] | None:
+        payload = self._read(self.storage_state_file, default=None)
+        if payload is not None and not isinstance(payload, dict):
+            raise ValueError("browser storage state must contain an object")
+        return payload
+
+    def save_storage_state(self, state: dict[str, Any]) -> Path:
+        if not isinstance(state, dict):
+            raise ValueError("browser storage state must be an object")
+        return atomic_write_json(self.storage_state_file, state)
+
+    @staticmethod
+    def _read(path: Path, *, default: Any) -> Any:
+        if not path.exists():
+            return default
+        import json
+
+        try:
+            return json.loads(path.read_text(encoding="utf-8"))
+        except (OSError, json.JSONDecodeError) as exc:
+            raise ValueError(f"cannot read browser state file: {path.name}") from exc
diff --git a/src/gyxx_flow/adapters/external.py b/src/gyxx_flow/adapters/external.py
new file mode 100644
index 0000000..cd58d6e
--- /dev/null
+++ b/src/gyxx_flow/adapters/external.py
@@ -0,0 +1,132 @@
+"""Shared external-system adapters with durable idempotency boundaries."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Mapping
+
+from gyxx_flow.core.artifacts import sha256_file
+from gyxx_flow.ops import Outbox, OutboxMessage
+from gyxx_flow.workflow import CommandStep
+
+
+@dataclass(frozen=True, slots=True)
+class FeishuOutboxAdapter:
+    outbox: Outbox
+
+    def upsert(
+        self, *, run_id: str, operation_key: str, artifact_id: str
+    ) -> OutboxMessage:
+        return self.outbox.enqueue(
+            idempotency_key=_key(run_id, "feishu", operation_key),
+            topic="feishu.upsert",
+            payload={"artifact_id": _value(artifact_id, "artifact_id")},
+            run_id=run_id,
+        )
+
+
+@dataclass(frozen=True, slots=True)
+class PostgresOutboxAdapter:
+    outbox: Outbox
+
+    def upsert(
+        self, *, run_id: str, operation_key: str, artifact_id: str
+    ) -> OutboxMessage:
+        return self.outbox.enqueue(
+            idempotency_key=_key(run_id, "postgres", operation_key),
+            topic="postgres.upsert",
+            payload={"artifact_id": _value(artifact_id, "artifact_id")},
+            run_id=run_id,
+        )
+
+
+@dataclass(frozen=True, slots=True)
+class ReloginOutboxAdapter:
+    outbox: Outbox
+    data_root: Path
+
+    @classmethod
+    def from_data_root(cls, data_root: Path | str) -> "ReloginOutboxAdapter":
+        root = Path(data_root).expanduser().resolve()
+        return cls(Outbox(root), root)
+
+    def enqueue(
+        self,
+        *,
+        run_id: str,
+        operation_key: str,
+        recipient: str,
+        text: str | None,
+        image: Path | None,
+    ) -> OutboxMessage:
+        payload: dict[str, object] = {
+            "recipient": _value(recipient, "recipient"),
+            "text": text or "",
+            "image_path": None,
+            "image_sha256": None,
+        }
+        if image is not None:
+            resolved = Path(image).resolve(strict=True)
+            try:
+                relative = resolved.relative_to(self.data_root).as_posix()
+            except ValueError as exc:
+                raise ValueError("relogin image must be inside the configured data root") from exc
+            payload["image_path"] = relative
+            payload["image_sha256"] = sha256_file(resolved)
+        return self.outbox.enqueue(
+            idempotency_key=_key(run_id, "feishu-relogin", operation_key),
+            topic="feishu.relogin",
+            payload=payload,
+            run_id=run_id,
+        )
+
+
+class HermesCommandAdapter:
+    """Build the one supported Hermes CLI invocation without executing it."""
+
+    def __init__(
+        self,
+        *,
+        executable: str = "hermes",
+        base_env: Mapping[str, str] | None = None,
+    ) -> None:
+        self._executable = _value(executable, "Hermes executable")
+        self._base_env = dict(base_env or {})
+
+    def build(self, *, prompt_file: Path, cwd: Path) -> CommandStep:
+        prompt_path = Path(prompt_file).resolve(strict=True)
+        working_directory = Path(cwd).resolve(strict=True)
+        if not prompt_path.is_file():
+            raise ValueError("Hermes prompt_file must be a file")
+        if not working_directory.is_dir():
+            raise ValueError("Hermes cwd must be a directory")
+        prompt = prompt_path.read_text(encoding="utf-8")
+        if not prompt or len(prompt) > 20_000 or "\x00" in prompt:
+            raise ValueError("Hermes prompt must contain 1..20000 safe characters")
+        return CommandStep(
+            argv=(self._executable, "-z", prompt),
+            cwd=working_directory,
+            env=self._base_env,
+        )
+
+
+def _key(run_id: str, adapter: str, operation_key: str) -> str:
+    return ":".join(
+        (
+            _value(run_id, "run_id"),
+            adapter,
+            _value(operation_key, "operation_key"),
+        )
+    )
+
+
+def _value(value: str, field: str) -> str:
+    if (
+        not isinstance(value, str)
+        or not value
+        or "\x00" in value
+        or any(ord(character) < 32 for character in value)
+    ):
+        raise ValueError(f"invalid {field}")
+    return value
diff --git a/src/gyxx_flow/adapters/integration.py b/src/gyxx_flow/adapters/integration.py
new file mode 100644
index 0000000..fa8401d
--- /dev/null
+++ b/src/gyxx_flow/adapters/integration.py
@@ -0,0 +1,344 @@
+"""Versioned per-script runtime bindings and external-service guardrails."""
+
+from __future__ import annotations
+
+import hashlib
+import ipaddress
+import json
+import os
+from dataclasses import dataclass
+from pathlib import Path
+from types import MappingProxyType
+from typing import Mapping
+from urllib.parse import urlparse
+
+from gyxx_flow.script_catalog import ScriptCatalog
+
+
+class RuntimeIntegrationError(ValueError):
+    """Raised when an integration binding is incomplete or unsafe."""
+
+
+@dataclass(frozen=True, slots=True)
+class RuntimeIntegrationBinding:
+    script_id: str
+    module: str
+    entry: str
+    cdp_port: int
+    cdp_url: str
+    profile_dir: Path
+    cookie_file: Path
+    storage_state_file: Path
+
+
+@dataclass(frozen=True, slots=True)
+class RuntimeServicePolicy:
+    """Keep legacy Feishu, cloud PostgreSQL, and loopback-only Hermes."""
+
+    hermes_url: str = "http://127.0.0.1:8642/v1/chat/completions"
+
+    def __post_init__(self) -> None:
+        _require_loopback_url(self.hermes_url, field="Hermes local URL")
+
+    def apply(self, environment: Mapping[str, str]) -> dict[str, str]:
+        result = dict(environment)
+        _fill_database_aliases(result)
+        _validate_cloud_database(result)
+        configured_hermes = _configured_hermes_urls(result)
+        for field, value in configured_hermes:
+            _require_loopback_url(value, field=field)
+        result.update(
+            {
+                "GYXX_FEISHU_MODE": "legacy",
+                "GYXX_POSTGRES_MODE": "cloud",
+                "GYXX_HERMES_MODE": "local",
+            }
+        )
+        # Preserve every original Feishu/DB/Hermes setting.  Defaults are only
+        # supplied for the two historical Hermes HTTP variable names.
+        result.setdefault("HERMES_ANALYZER_URL", self.hermes_url)
+        result.setdefault("ANALYZER_API_SERVER_URL", self.hermes_url)
+        return result
+
+
+class RuntimeIntegrationCatalog:
+    """Immutable exact allocation of one portable browser binding per script."""
+
+    def __init__(
+        self,
+        bindings: Mapping[str, RuntimeIntegrationBinding],
+        *,
+        service_policy: RuntimeServicePolicy,
+    ) -> None:
+        self._bindings = MappingProxyType(dict(sorted(bindings.items())))
+        self.service_policy = service_policy
+
+    @classmethod
+    def load(
+        cls,
+        config_path: Path | str,
+        *,
+        scripts: ScriptCatalog,
+        data_root: Path | str,
+    ) -> "RuntimeIntegrationCatalog":
+        path = Path(config_path)
+        try:
+            payload = json.loads(path.read_text(encoding="utf-8"))
+        except (OSError, json.JSONDecodeError) as exc:
+            raise RuntimeIntegrationError(
+                f"cannot load runtime integration catalog: {path.name}"
+            ) from exc
+        if not isinstance(payload, dict) or payload.get("schema_version") != 1:
+            raise RuntimeIntegrationError("unsupported runtime binding schema")
+        host = payload.get("cdp_host")
+        if host not in {"127.0.0.1", "localhost", "::1"}:
+            raise RuntimeIntegrationError("browser CDP host must be loopback")
+        allocations = payload.get("scripts")
+        if not isinstance(allocations, dict):
+            raise RuntimeIntegrationError("runtime binding scripts must be an object")
+        configured = set(allocations)
+        discovered = set(scripts.script_ids)
+        if configured != discovered:
+            missing = len(discovered - configured)
+            extra = len(configured - discovered)
+            raise RuntimeIntegrationError(
+                f"runtime binding coverage mismatch; missing={missing}, extra={extra}"
+            )
+        ports = list(allocations.values())
+        if (
+            not all(isinstance(port, int) and 22000 <= port <= 22999 for port in ports)
+            or len(set(ports)) != len(ports)
+        ):
+            raise RuntimeIntegrationError(
+                "runtime binding ports must be unique integers in 22000..22999"
+            )
+        root = Path(data_root).expanduser().resolve()
+        bindings: dict[str, RuntimeIntegrationBinding] = {}
+        for script in scripts.scripts:
+            port = allocations[script.script_id]
+            state_root = _browser_state_root(root, script.module, script.script_id)
+            bindings[script.script_id] = RuntimeIntegrationBinding(
+                script_id=script.script_id,
+                module=script.module,
+                entry=script.entry,
+                cdp_port=port,
+                cdp_url=f"http://{host}:{port}",
+                profile_dir=state_root / "profile",
+                cookie_file=state_root / "cookies.json",
+                storage_state_file=state_root / "storage_state.json",
+            )
+        services = payload.get("services")
+        if not isinstance(services, dict):
+            raise RuntimeIntegrationError("runtime binding services must be an object")
+        if (
+            services.get("feishu") != "legacy"
+            or services.get("postgres") != "cloud"
+            or services.get("hermes") != "local"
+        ):
+            raise RuntimeIntegrationError("unsupported runtime service policy")
+        hermes_url = services.get("hermes_url")
+        if not isinstance(hermes_url, str):
+            raise RuntimeIntegrationError("runtime service policy requires Hermes URL")
+        return cls(
+            bindings,
+            service_policy=RuntimeServicePolicy(hermes_url=hermes_url),
+        )
+
+    @classmethod
+    def load_default(
+        cls,
+        *,
+        project_root: Path | str,
+        data_root: Path | str,
+        scripts: ScriptCatalog | None = None,
+    ) -> "RuntimeIntegrationCatalog":
+        root = Path(project_root).expanduser().resolve()
+        return cls.load(
+            root / "config" / "runtime-bindings.json",
+            scripts=scripts or ScriptCatalog.discover_default(),
+            data_root=data_root,
+        )
+
+    @property
+    def script_ids(self) -> tuple[str, ...]:
+        return tuple(self._bindings)
+
+    def binding_for(self, script_id: str) -> RuntimeIntegrationBinding:
+        try:
+            return self._bindings[script_id]
+        except KeyError as exc:
+            raise RuntimeIntegrationError(f"unknown runtime script binding: {script_id}") from exc
+
+    def environment_for(
+        self,
+        script_id: str,
+        base_environment: Mapping[str, str] | None = None,
+    ) -> dict[str, str]:
+        binding = self.binding_for(script_id)
+        result = self.service_policy.apply(
+            os.environ if base_environment is None else base_environment
+        )
+        result.update(
+            {
+                "GYXX_SCRIPT_ID": binding.script_id,
+                "GYXX_BROWSER_CDP_PORT": str(binding.cdp_port),
+                "GYXX_BROWSER_CDP_URL": binding.cdp_url,
+                "GYXX_BROWSER_PROFILE_DIR": str(binding.profile_dir),
+                "GYXX_BROWSER_COOKIE_FILE": str(binding.cookie_file),
+                "GYXX_BROWSER_STORAGE_STATE_FILE": str(binding.storage_state_file),
+                # Compatibility aliases consumed by the migrated browser engines.
+                # They are deliberately overwritten so every engine uses the
+                # catalog allocation instead of a legacy shared profile/port.
+                "AUTOFLOW_CDP_PORT": str(binding.cdp_port),
+                "AUTOFLOW_CDP_URL": binding.cdp_url,
+                "DMP_CDP_URL": binding.cdp_url,
+                "TM_USER_DATA_DIR": str(binding.profile_dir),
+                "TM_DAILY_USER_DATA_DIR": str(binding.profile_dir),
+                "DMP_USER_DATA_DIR": str(binding.profile_dir),
+                "DY_USER_DATA_DIR": str(binding.profile_dir),
+                "WANXIANG_USER_DATA_DIR": str(binding.profile_dir),
+                "GUANGHE_USER_DATA_DIR": str(binding.profile_dir),
+                "GUANGHE_LUGGAGE_USER_DATA_DIR": str(binding.profile_dir),
+                "DY_STORAGE_STATE_FILE": str(binding.storage_state_file),
+            }
+        )
+        return result
+
+
+def environment_for_child_script(
+    target: Path | str,
+    base_environment: Mapping[str, str] | None = None,
+) -> dict[str, str]:
+    """Rebind a nested project script instead of inheriting its parent's CDP."""
+
+    environment = dict(os.environ if base_environment is None else base_environment)
+    module = environment.get("GYXX_MODULE_ID", "").strip()
+    module_root_value = environment.get("GYXX_MODULE_ROOT", "").strip()
+    project_root_value = environment.get("GYXX_PROJECT_ROOT", "").strip()
+    data_root_value = environment.get("GYXX_DATA_ROOT", "").strip()
+    if not all((module, module_root_value, project_root_value, data_root_value)):
+        raise RuntimeIntegrationError("child script binding requires GYXX runtime roots")
+    module_root = Path(module_root_value).expanduser().resolve()
+    target_path = Path(target)
+    if not target_path.is_absolute():
+        target_path = module_root / target_path
+    target_path = target_path.expanduser().resolve()
+    try:
+        entry = target_path.relative_to(module_root).as_posix()
+    except ValueError as exc:
+        raise RuntimeIntegrationError("child script must stay inside its module root") from exc
+    script_id = f"{module}:{entry}"
+    catalog = RuntimeIntegrationCatalog.load_default(
+        project_root=project_root_value,
+        data_root=data_root_value,
+    )
+    return catalog.environment_for(script_id, environment)
+
+
+def binding_from_environment(
+    environment: Mapping[str, str] | None = None,
+) -> RuntimeIntegrationBinding:
+    values = os.environ if environment is None else environment
+    try:
+        script_id = values["GYXX_SCRIPT_ID"]
+        module, entry = script_id.split(":", 1)
+        port = int(values["GYXX_BROWSER_CDP_PORT"])
+        cdp_url = values["GYXX_BROWSER_CDP_URL"]
+        profile = Path(values["GYXX_BROWSER_PROFILE_DIR"]).expanduser().resolve()
+        cookie = Path(values["GYXX_BROWSER_COOKIE_FILE"]).expanduser().resolve()
+        storage = Path(values["GYXX_BROWSER_STORAGE_STATE_FILE"]).expanduser().resolve()
+    except (KeyError, ValueError) as exc:
+        raise RuntimeIntegrationError("current process has no valid browser binding") from exc
+    _require_loopback_url(cdp_url, field="browser CDP URL")
+    if not 22000 <= port <= 22999:
+        raise RuntimeIntegrationError("browser CDP port is outside the managed range")
+    return RuntimeIntegrationBinding(
+        script_id=script_id,
+        module=module,
+        entry=entry,
+        cdp_port=port,
+        cdp_url=cdp_url,
+        profile_dir=profile,
+        cookie_file=cookie,
+        storage_state_file=storage,
+    )
+
+
+def _browser_state_root(data_root: Path, module: str, script_id: str) -> Path:
+    entry = script_id.split(":", 1)[1]
+    stem = Path(entry).stem
+    safe = "".join(character if character.isalnum() else "-" for character in stem)
+    safe = safe.strip("-")[:48] or "script"
+    digest = hashlib.sha256(script_id.encode("utf-8")).hexdigest()[:12]
+    return data_root / "state" / "browser" / module / f"{safe}-{digest}"
+
+
+def _configured_hermes_urls(environment: Mapping[str, str]) -> tuple[tuple[str, str], ...]:
+    names = (
+        "HERMES_ANALYZER_URL",
+        "ANALYZER_API_SERVER_URL",
+        "COLLECTOR_API_SERVER_URL",
+        "COLLECTOR_HERMES_GATEWAY_URL",
+        "ANALYZER_HERMES_GATEWAY_URL",
+    )
+    return tuple((name, environment[name]) for name in names if environment.get(name, "").strip())
+
+
+def _validate_cloud_database(environment: Mapping[str, str]) -> None:
+    for name in ("PG_HOST", "DB_HOST", "AUTOFLOW_PG_HOST"):
+        value = environment.get(name, "").strip()
+        if value and _is_loopback_host(value):
+            raise RuntimeIntegrationError(f"cloud PostgreSQL cannot use loopback host in {name}")
+    for name in ("DATABASE_URL", "DB_URL"):
+        value = environment.get(name, "").strip()
+        if value:
+            host = urlparse(value).hostname
+            if host is None or _is_loopback_host(host):
+                raise RuntimeIntegrationError(f"cloud PostgreSQL requires a remote host in {name}")
+
+
+def _fill_database_aliases(environment: dict[str, str]) -> None:
+    groups = (
+        ("PG_HOST", "DB_HOST", "AUTOFLOW_PG_HOST"),
+        ("PG_PORT", "DB_PORT", "AUTOFLOW_PG_PORT"),
+        ("PG_DB", "DB_NAME", "AUTOFLOW_PG_DB"),
+        ("PG_USER", "DB_USER", "AUTOFLOW_PG_USER"),
+        ("PG_PASSWORD", "DB_PASSWORD", "AUTOFLOW_PG_PASSWORD"),
+    )
+    for aliases in groups:
+        value = next(
+            (environment[name] for name in aliases if environment.get(name, "").strip()),
+            "",
+        )
+        if value:
+            for name in aliases:
+                if not environment.get(name, "").strip():
+                    environment[name] = value
+
+
+def _require_loopback_url(value: str, *, field: str) -> None:
+    parsed = urlparse(value)
+    if parsed.scheme not in {"http", "https", "ws", "wss"} or not parsed.hostname:
+        raise RuntimeIntegrationError(f"{field} must be a valid local URL")
+    if not _is_loopback_host(parsed.hostname):
+        raise RuntimeIntegrationError(f"{field} must remain local")
+
+
+def _is_loopback_host(host: str) -> bool:
+    normalized = host.strip().strip("[]").casefold()
+    if normalized == "localhost":
+        return True
+    try:
+        return ipaddress.ip_address(normalized).is_loopback
+    except ValueError:
+        return False
+
+
+__all__ = [
+    "RuntimeIntegrationBinding",
+    "RuntimeIntegrationCatalog",
+    "RuntimeIntegrationError",
+    "RuntimeServicePolicy",
+    "binding_from_environment",
+    "environment_for_child_script",
+]
diff --git a/src/gyxx_flow/adapters/native.py b/src/gyxx_flow/adapters/native.py
new file mode 100644
index 0000000..314a750
--- /dev/null
+++ b/src/gyxx_flow/adapters/native.py
@@ -0,0 +1,209 @@
+"""Resolve workflow commands from source owned by this installed project."""
+
+from __future__ import annotations
+
+import os
+import subprocess
+import sys
+from dataclasses import dataclass
+from pathlib import Path, PurePosixPath
+from types import MappingProxyType
+from typing import Callable, Mapping
+
+from gyxx_flow.adapters.integration import RuntimeIntegrationCatalog
+from gyxx_flow.catalog import WorkflowEntry
+from gyxx_flow.core.config import Settings
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.core.layout import DataLayout
+from gyxx_flow.workflow.model import ExecutableStep
+from gyxx_flow.workflow.steps import CommandStep, StepExecution
+
+
+class ModuleSourceError(ValueError):
+    """Raised when a workflow cannot resolve project-owned module source safely."""
+
+
+class ModuleSourceRoots:
+    """Validated runtime roots for independently packaged business modules."""
+
+    def __init__(self, roots: Mapping[str, Path | str]) -> None:
+        resolved: dict[str, Path] = {}
+        for module, configured in roots.items():
+            if not module or not isinstance(module, str):
+                raise ModuleSourceError(f"invalid module id: {module!r}")
+            root = Path(configured).expanduser().resolve(strict=True)
+            if not root.is_dir():
+                raise ModuleSourceError(f"module source root is not a directory: {root}")
+            resolved[module] = root
+        self._roots = MappingProxyType(resolved)
+
+    def get(self, module: str) -> Path:
+        try:
+            return self._roots[module]
+        except KeyError as exc:
+            raise ModuleSourceError(f"unknown module source: {module!r}") from exc
+
+
+class ModuleCommandAdapter:
+    """Build a shell-free command rooted in source shipped with this project."""
+
+    def __init__(
+        self,
+        roots: ModuleSourceRoots,
+        *,
+        base_env: Mapping[str, str] | None = None,
+        python_executable: str | Path | None = None,
+        project_root: str | Path | None = None,
+        data_root: str | Path | None = None,
+        integration_catalog: RuntimeIntegrationCatalog | None = None,
+    ) -> None:
+        self._roots = roots
+        self._base_env = dict(os.environ if base_env is None else base_env)
+        self._python = str(python_executable or sys.executable)
+        settings = Settings.from_env(project_root=project_root, env=self._base_env)
+        self._project_root = settings.project_root
+        self._data_root = (
+            Path(data_root).expanduser().resolve()
+            if data_root is not None
+            else settings.data_root
+        )
+        binding_file = self._project_root / "config" / "runtime-bindings.json"
+        self._integration_catalog = integration_catalog
+        if self._integration_catalog is None and binding_file.is_file():
+            self._integration_catalog = RuntimeIntegrationCatalog.load_default(
+                project_root=self._project_root,
+                data_root=self._data_root,
+            )
+
+    def build(self, entry: WorkflowEntry, *, context: RunContext) -> CommandStep:
+        if entry.workflow_id != context.workflow_id:
+            raise ModuleSourceError(
+                f"workflow entry {entry.workflow_id!r} does not match "
+                f"run context {context.workflow_id!r}"
+            )
+        if entry.trigger == "unavailable":
+            raise ModuleSourceError(f"workflow is unavailable: {entry.workflow_id}")
+        root = self._roots.get(entry.module)
+        executable = _resolve_entry(root, entry.entry)
+        suffix = executable.suffix.casefold()
+        if suffix in {".py", ".pyw"}:
+            argv = (self._python, str(executable), *entry.args)
+        elif suffix in {".bat", ".cmd"}:
+            command_line = subprocess.list2cmdline([str(executable), *entry.args])
+            argv = ("cmd.exe", "/d", "/s", "/c", command_line)
+        elif suffix == ".ps1":
+            argv = (
+                "powershell.exe",
+                "-NoProfile",
+                "-ExecutionPolicy",
+                "Bypass",
+                "-File",
+                str(executable),
+                *entry.args,
+            )
+        else:
+            argv = (str(executable), *entry.args)
+        env = {
+            **self._base_env,
+            "GYXX_PROJECT_ROOT": str(self._project_root),
+            "GYXX_DATA_ROOT": str(self._data_root),
+            "GYXX_MODULE_ROOT": str(root),
+            "GYXX_MODULE_ID": entry.module,
+            "GYXX_PYTHON": self._python,
+            "GYXX_WORKFLOW_ID": context.workflow_id,
+            "GYXX_RUN_ID": context.run_id,
+            "GYXX_BUSINESS_DATE": context.business_date.isoformat(),
+            "GYXX_SHADOW": "true" if context.shadow else "false",
+            "PYTHONUNBUFFERED": "1",
+        }
+        if entry.module == "supply_chain":
+            supply_paths = DataLayout(self._data_root).for_module("supply_chain")
+            env.update(
+                {
+                    "GYXX_SUPPLY_RAW_ROOT": str(supply_paths.raw_root),
+                    "GYXX_SUPPLY_STATE_ROOT": str(supply_paths.state_root),
+                    "GYXX_SUPPLY_EXPORT_ROOT": str(supply_paths.exports_root),
+                    "GYXX_SUPPLY_WORK_ROOT": str(supply_paths.tmp_root),
+                }
+            )
+        if self._integration_catalog is not None:
+            env = self._integration_catalog.environment_for(
+                f"{entry.module}:{entry.entry}", env
+            )
+        existing_pythonpath = env.get("PYTHONPATH", "").strip()
+        env["PYTHONPATH"] = (
+            f"{root}{os.pathsep}{existing_pythonpath}" if existing_pythonpath else str(root)
+        )
+        return CommandStep(argv=tuple(argv), cwd=root, env=env)
+
+
+ModuleCommandFactory = Callable[[WorkflowEntry, RunContext], ExecutableStep]
+
+
+def _command_from_project(entry: WorkflowEntry, context: RunContext) -> ExecutableStep:
+    package_root = Path(__file__).resolve().parents[1]
+    runtime_root = package_root / "modules" / entry.module / "runtime"
+    settings = Settings.from_env()
+    return ModuleCommandAdapter(
+        ModuleSourceRoots({entry.module: runtime_root}),
+        project_root=settings.project_root,
+        data_root=settings.data_root,
+    ).build(entry, context=context)
+
+
+@dataclass(frozen=True, slots=True)
+class DeferredModuleCommandStep:
+    """Resolve project-owned module source only when a real execution starts."""
+
+    entry: WorkflowEntry
+    command_factory: ModuleCommandFactory = _command_from_project
+
+    def __post_init__(self) -> None:
+        if not callable(self.command_factory):
+            raise TypeError("command_factory must be callable")
+
+    def execute(
+        self,
+        *,
+        context: RunContext,
+        timeout_seconds: float | None,
+        dry_run: bool,
+    ) -> StepExecution:
+        if dry_run:
+            return StepExecution(exit_code=0, skipped=True, reason="dry-run")
+        command = self.command_factory(self.entry, context)
+        if not callable(getattr(command, "execute", None)):
+            raise TypeError("command_factory must return an executable step")
+        return command.execute(
+            context=context,
+            timeout_seconds=timeout_seconds,
+            dry_run=False,
+        )
+
+
+def _resolve_entry(root: Path, configured_entry: str) -> Path:
+    if not configured_entry or "\\" in configured_entry or ":" in configured_entry:
+        raise ModuleSourceError("module entry must use a safe relative path")
+    relative = PurePosixPath(configured_entry)
+    if relative.is_absolute() or ".." in relative.parts:
+        raise ModuleSourceError("module entry cannot escape its module root")
+    try:
+        candidate = root.joinpath(*relative.parts).resolve(strict=True)
+    except OSError as exc:
+        raise ModuleSourceError(
+            f"module entry does not exist or cannot be resolved: {configured_entry}"
+        ) from exc
+    if not candidate.is_relative_to(root):
+        raise ModuleSourceError("module entry cannot escape its module root")
+    if not candidate.is_file():
+        raise ModuleSourceError(f"module entry is not a file: {configured_entry}")
+    return candidate
+
+
+__all__ = [
+    "DeferredModuleCommandStep",
+    "ModuleCommandAdapter",
+    "ModuleCommandFactory",
+    "ModuleSourceError",
+    "ModuleSourceRoots",
+]
diff --git a/src/gyxx_flow/baseline/__init__.py b/src/gyxx_flow/baseline/__init__.py
new file mode 100644
index 0000000..0f455a0
--- /dev/null
+++ b/src/gyxx_flow/baseline/__init__.py
@@ -0,0 +1,31 @@
+from .inventory import collect_code_inventory, collect_data_summary
+from .manifest import build_baseline_manifest, write_manifest_atomic
+from .models import (
+    DEFAULT_CODE_EXTENSIONS,
+    DataRootSpec,
+    ProjectSpec,
+    ScheduledTask,
+    project_specs_from_env,
+)
+from .tasks import (
+    PowerShellScheduledTaskProvider,
+    ScheduledTaskProvider,
+    TaskCountMismatch,
+    collect_scheduled_task_inventory,
+)
+
+__all__ = [
+    "DEFAULT_CODE_EXTENSIONS",
+    "DataRootSpec",
+    "ProjectSpec",
+    "ScheduledTask",
+    "PowerShellScheduledTaskProvider",
+    "ScheduledTaskProvider",
+    "TaskCountMismatch",
+    "build_baseline_manifest",
+    "collect_code_inventory",
+    "collect_data_summary",
+    "collect_scheduled_task_inventory",
+    "write_manifest_atomic",
+    "project_specs_from_env",
+]
diff --git a/src/gyxx_flow/baseline/inventory.py b/src/gyxx_flow/baseline/inventory.py
new file mode 100644
index 0000000..13b65dd
--- /dev/null
+++ b/src/gyxx_flow/baseline/inventory.py
@@ -0,0 +1,168 @@
+from __future__ import annotations
+
+import hashlib
+import os
+from collections.abc import Iterable
+from pathlib import Path
+from typing import TypedDict
+
+from .models import DEFAULT_CODE_EXTENSIONS
+
+
+class Aggregate(TypedDict):
+    file_count: int
+    total_bytes: int
+
+
+class DataSummary(Aggregate):
+    by_extension: dict[str, Aggregate]
+    by_top_level: dict[str, Aggregate]
+
+
+_EXCLUDED_DIRECTORY_NAMES = frozenset(
+    {
+        ".cache",
+        ".git",
+        ".hermes_tmp",
+        ".mypy_cache",
+        ".pytest_cache",
+        ".ruff_cache",
+        ".tox",
+        ".venv",
+        "__pycache__",
+        "cache",
+        "data",
+        "debug",
+        "evidence",
+        "log",
+        "logs",
+        "node_modules",
+        "reports",
+        "runtime",
+        "state",
+        "tmp",
+        "var",
+    }
+)
+_SENSITIVE_DIRECTORY_MARKERS = ("cookie", "profile")
+_SENSITIVE_FILE_NAMES = frozenset(
+    {
+        ".env",
+        "credentials.json",
+        "secrets.json",
+        "service-account.json",
+        "service_account.json",
+    }
+)
+_SENSITIVE_FILE_MARKERS = ("cookie", "credential", "secret", "token")
+
+
+def _is_excluded_directory(name: str) -> bool:
+    normalized = name.casefold()
+    return normalized in _EXCLUDED_DIRECTORY_NAMES or any(
+        marker in normalized for marker in _SENSITIVE_DIRECTORY_MARKERS
+    )
+
+
+def _is_sensitive_file(name: str) -> bool:
+    normalized = name.casefold()
+    if normalized in _SENSITIVE_FILE_NAMES or normalized.startswith(".env."):
+        return True
+    stem = Path(normalized).stem
+    return any(marker in stem for marker in _SENSITIVE_FILE_MARKERS)
+
+
+def _walk_files(
+    root: Path,
+    *,
+    exclude_runtime: bool,
+    excluded_roots: frozenset[Path] = frozenset(),
+) -> Iterable[Path]:
+    if not root.is_dir():
+        raise FileNotFoundError(f"inventory root is not a directory: {root}")
+    for directory, dirnames, filenames in os.walk(root, followlinks=False):
+        if exclude_runtime:
+            dirnames[:] = sorted(
+                name
+                for name in dirnames
+                if not _is_excluded_directory(name)
+                and not (Path(directory) / name).is_symlink()
+                and (Path(directory) / name).resolve(strict=False) not in excluded_roots
+            )
+        else:
+            dirnames[:] = sorted(
+                name for name in dirnames if not (Path(directory) / name).is_symlink()
+            )
+        for filename in sorted(filenames):
+            path = Path(directory) / filename
+            if not path.is_symlink():
+                yield path
+
+
+def _sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as stream:
+        for chunk in iter(lambda: stream.read(1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def collect_code_inventory(
+    root: Path,
+    *,
+    allowed_extensions: frozenset[str] = DEFAULT_CODE_EXTENSIONS,
+    excluded_roots: Iterable[Path] = (),
+) -> list[dict[str, str | int]]:
+    """Hash code-like files without retaining file contents or config values."""
+
+    root = Path(root)
+    normalized_extensions = frozenset(extension.casefold() for extension in allowed_extensions)
+    normalized_excluded_roots = frozenset(
+        Path(excluded_root).resolve(strict=False) for excluded_root in excluded_roots
+    )
+    inventory: list[dict[str, str | int]] = []
+    for path in _walk_files(
+        root, exclude_runtime=True, excluded_roots=normalized_excluded_roots
+    ):
+        if path.suffix.casefold() not in normalized_extensions or _is_sensitive_file(path.name):
+            continue
+        stat = path.stat()
+        inventory.append(
+            {
+                "relative_path": path.relative_to(root).as_posix(),
+                "size_bytes": stat.st_size,
+                "mtime_ns": stat.st_mtime_ns,
+                "sha256": _sha256(path),
+            }
+        )
+    return sorted(inventory, key=lambda item: str(item["relative_path"]).casefold())
+
+
+def _increment(target: dict[str, Aggregate], key: str, size: int) -> None:
+    aggregate = target.setdefault(key, {"file_count": 0, "total_bytes": 0})
+    aggregate["file_count"] += 1
+    aggregate["total_bytes"] += size
+
+
+def collect_data_summary(root: Path) -> DataSummary:
+    """Summarize a data tree without exposing names, paths, times, hashes, or values."""
+
+    root = Path(root)
+    summary: DataSummary = {
+        "file_count": 0,
+        "total_bytes": 0,
+        "by_extension": {},
+        "by_top_level": {},
+    }
+    for path in _walk_files(root, exclude_runtime=False):
+        size = path.stat().st_size
+        relative = path.relative_to(root)
+        top_level = relative.parts[0] if len(relative.parts) > 1 else "."
+        extension = path.suffix.casefold() or "[no_extension]"
+        summary["file_count"] += 1
+        summary["total_bytes"] += size
+        _increment(summary["by_extension"], extension, size)
+        _increment(summary["by_top_level"], top_level, size)
+    summary["by_extension"] = dict(sorted(summary["by_extension"].items()))
+    summary["by_top_level"] = dict(sorted(summary["by_top_level"].items()))
+    return summary
diff --git a/src/gyxx_flow/baseline/manifest.py b/src/gyxx_flow/baseline/manifest.py
new file mode 100644
index 0000000..952e1ae
--- /dev/null
+++ b/src/gyxx_flow/baseline/manifest.py
@@ -0,0 +1,82 @@
+from __future__ import annotations
+
+import json
+import os
+import tempfile
+from collections.abc import Sequence
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+
+from .inventory import collect_code_inventory, collect_data_summary
+from .models import ProjectSpec
+from .tasks import ScheduledTaskProvider, collect_scheduled_task_inventory
+
+
+def build_baseline_manifest(
+    specs: Sequence[ProjectSpec],
+    task_provider: ScheduledTaskProvider,
+    *,
+    generated_at: datetime | None = None,
+    expected_task_count: int = 21,
+) -> dict[str, Any]:
+    """Build a read-only baseline. It never serializes task actions or file contents."""
+
+    timestamp = generated_at or datetime.now(timezone.utc)
+    if timestamp.tzinfo is None or timestamp.utcoffset() is None:
+        raise ValueError("generated_at must be timezone-aware")
+
+    projects: list[dict[str, Any]] = []
+    for spec in specs:
+        projects.append(
+            {
+                "project_id": spec.project_id,
+                "source_root": str(spec.root),
+                "code_inventory": collect_code_inventory(
+                    spec.root,
+                    allowed_extensions=spec.code_extensions,
+                    excluded_roots=(data_root.path for data_root in spec.data_roots),
+                ),
+                "data_summaries": [
+                    {"label": data_root.label, "summary": collect_data_summary(data_root.path)}
+                    for data_root in spec.data_roots
+                ],
+            }
+        )
+
+    return {
+        "schema_version": 1,
+        "generated_at": timestamp.isoformat(),
+        "projects": projects,
+        "scheduled_tasks": collect_scheduled_task_inventory(
+            specs, task_provider, expected_count=expected_task_count
+        ),
+    }
+
+
+def write_manifest_atomic(path: Path, manifest: object) -> None:
+    """Durably replace a JSON manifest without exposing a partial destination file."""
+
+    path = Path(path)
+    path.parent.mkdir(parents=True, exist_ok=True)
+    temporary_path: Path | None = None
+    try:
+        with tempfile.NamedTemporaryFile(
+            mode="w",
+            encoding="utf-8",
+            newline="\n",
+            dir=path.parent,
+            prefix=f".{path.name}.",
+            suffix=".tmp",
+            delete=False,
+        ) as stream:
+            temporary_path = Path(stream.name)
+            json.dump(manifest, stream, ensure_ascii=False, indent=2, sort_keys=True)
+            stream.write("\n")
+            stream.flush()
+            os.fsync(stream.fileno())
+        os.replace(temporary_path, path)
+        temporary_path = None
+    finally:
+        if temporary_path is not None:
+            temporary_path.unlink(missing_ok=True)
diff --git a/src/gyxx_flow/baseline/models.py b/src/gyxx_flow/baseline/models.py
new file mode 100644
index 0000000..572818c
--- /dev/null
+++ b/src/gyxx_flow/baseline/models.py
@@ -0,0 +1,113 @@
+from __future__ import annotations
+
+import os
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Mapping
+
+DEFAULT_CODE_EXTENSIONS = frozenset(
+    {
+        ".bat",
+        ".cmd",
+        ".json",
+        ".md",
+        ".ps1",
+        ".py",
+        ".pyi",
+        ".rst",
+        ".sh",
+        ".sql",
+        ".toml",
+        ".txt",
+        ".yaml",
+        ".yml",
+    }
+)
+
+
+@dataclass(frozen=True, slots=True)
+class DataRootSpec:
+    """A named data tree whose file-level details must remain private."""
+
+    label: str
+    path: Path
+
+    def __post_init__(self) -> None:
+        if not self.label or self.label in {".", ".."}:
+            raise ValueError("data root label must be a non-empty logical name")
+        object.__setattr__(self, "path", Path(self.path))
+
+
+@dataclass(frozen=True, slots=True)
+class ProjectSpec:
+    """Read-only discovery boundaries for one legacy project."""
+
+    project_id: str
+    root: Path
+    data_roots: tuple[DataRootSpec, ...] = ()
+    code_extensions: frozenset[str] = field(default=DEFAULT_CODE_EXTENSIONS)
+
+    def __post_init__(self) -> None:
+        if not self.project_id:
+            raise ValueError("project_id is required")
+        object.__setattr__(self, "root", Path(self.root))
+        object.__setattr__(self, "data_roots", tuple(self.data_roots))
+        normalized_extensions = frozenset(
+            extension.casefold() if extension.startswith(".") else f".{extension.casefold()}"
+            for extension in self.code_extensions
+        )
+        object.__setattr__(self, "code_extensions", normalized_extensions)
+
+
+@dataclass(frozen=True, slots=True)
+class ScheduledTask:
+    """Provider-neutral scheduled-task action used only for source matching."""
+
+    task_id: str
+    command: str
+    arguments: str = ""
+    working_directory: str = ""
+
+
+def project_specs_from_env(
+    env: Mapping[str, str] | None = None,
+) -> tuple[ProjectSpec, ...]:
+    """Build legacy discovery boundaries without embedding machine-specific roots."""
+
+    values = os.environ if env is None else env
+    keys = (
+        "GYXX_LEGACY_CONTENT_ROOT",
+        "GYXX_LEGACY_SHOP_ROOT",
+        "GYXX_LEGACY_PRODUCT_ROOT",
+        "GYXX_LEGACY_SUPPLY_ROOT",
+    )
+    missing = [key for key in keys if not values.get(key, "").strip()]
+    if missing:
+        raise ValueError("missing legacy root environment variables: " + ", ".join(missing))
+
+    content = Path(values[keys[0]]).expanduser()
+    shop = Path(values[keys[1]]).expanduser()
+    product = Path(values[keys[2]]).expanduser()
+    supply = Path(values[keys[3]]).expanduser()
+    return (
+        ProjectSpec(
+            "content_marketing",
+            content,
+            (DataRootSpec("data", content / "data"), DataRootSpec("reports", content / "reports")),
+        ),
+        ProjectSpec(
+            "shop_intelligence",
+            shop,
+            (DataRootSpec("data", shop / "data"),),
+        ),
+        ProjectSpec(
+            "product_commerce",
+            product,
+            (DataRootSpec("data", product / "data"),),
+        ),
+        ProjectSpec(
+            "supply_chain",
+            supply,
+            (DataRootSpec("shared-data", supply / "shared-data"),),
+        ),
+    )
diff --git a/src/gyxx_flow/baseline/tasks.py b/src/gyxx_flow/baseline/tasks.py
new file mode 100644
index 0000000..e4c9fb6
--- /dev/null
+++ b/src/gyxx_flow/baseline/tasks.py
@@ -0,0 +1,115 @@
+from __future__ import annotations
+
+import json
+import os
+import re
+import subprocess
+from collections.abc import Iterable, Sequence
+from typing import Protocol
+
+from .models import ProjectSpec, ScheduledTask
+
+
+class ScheduledTaskProvider(Protocol):
+    def scheduled_tasks(self) -> Iterable[ScheduledTask]: ...
+
+
+class PowerShellScheduledTaskProvider:
+    """Read Windows Task Scheduler actions without modifying scheduler state."""
+
+    _SCRIPT = r"""
+$ErrorActionPreference = 'Stop'
+$rows = foreach ($task in Get-ScheduledTask) {
+    [pscustomobject]@{
+        task_id = "$($task.TaskPath)$($task.TaskName)"
+        command = [string](($task.Actions | ForEach-Object { $_.Execute }) -join ' ')
+        arguments = [string](($task.Actions | ForEach-Object { $_.Arguments }) -join ' ')
+        working_directory = [string](($task.Actions | ForEach-Object { $_.WorkingDirectory }) -join ' ')
+    }
+}
+@($rows) | ConvertTo-Json -Depth 3 -Compress
+"""
+
+    def __init__(self, executable: str = "powershell.exe") -> None:
+        self.executable = executable
+
+    def scheduled_tasks(self) -> list[ScheduledTask]:
+        completed = subprocess.run(
+            [
+                self.executable,
+                "-NoLogo",
+                "-NoProfile",
+                "-NonInteractive",
+                "-Command",
+                self._SCRIPT,
+            ],
+            check=True,
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="strict",
+            shell=False,
+        )
+        payload = json.loads(completed.stdout or "[]")
+        rows = payload if isinstance(payload, list) else [payload]
+        tasks: list[ScheduledTask] = []
+        for row in rows:
+            if not isinstance(row, dict) or not isinstance(row.get("task_id"), str):
+                raise ValueError("invalid scheduled-task provider response")
+            tasks.append(
+                ScheduledTask(
+                    task_id=row["task_id"],
+                    command=str(row.get("command") or ""),
+                    arguments=str(row.get("arguments") or ""),
+                    working_directory=str(row.get("working_directory") or ""),
+                )
+            )
+        return tasks
+
+
+class TaskCountMismatch(ValueError):
+    """The discovered legacy task inventory is not the accepted baseline."""
+
+
+def _match_project(task: ScheduledTask, specs: Sequence[ProjectSpec]) -> str | None:
+    action = " ".join((task.command, task.arguments, task.working_directory))
+    normalized_action = os.path.normcase(action).replace("/", "\\")
+    for spec in specs:
+        normalized_root = os.path.normcase(str(spec.root)).replace("/", "\\").rstrip("\\")
+        root_pattern = re.escape(normalized_root) + r"(?=$|[\\\s\"'])"
+        if normalized_root and re.search(root_pattern, normalized_action):
+            return spec.project_id
+    return None
+
+
+def collect_scheduled_task_inventory(
+    specs: Sequence[ProjectSpec],
+    provider: ScheduledTaskProvider,
+    *,
+    expected_count: int = 21,
+) -> dict[str, object]:
+    """Keep only legacy-project tasks and deliberately discard action details."""
+
+    tasks: list[dict[str, str]] = []
+    seen_ids: set[str] = set()
+    for task in provider.scheduled_tasks():
+        project_id = _match_project(task, specs)
+        if project_id is None:
+            continue
+        if task.task_id in seen_ids:
+            raise ValueError(f"duplicate scheduled task id: {task.task_id}")
+        seen_ids.add(task.task_id)
+        tasks.append({"task_id": task.task_id, "project_id": project_id})
+
+    tasks.sort(key=lambda item: item["task_id"].casefold())
+    actual_count = len(tasks)
+    if actual_count != expected_count:
+        raise TaskCountMismatch(
+            f"scheduled task baseline expected {expected_count} tasks but found {actual_count}"
+        )
+    return {
+        "expected_count": expected_count,
+        "actual_count": actual_count,
+        "is_complete": True,
+        "tasks": tasks,
+    }
diff --git a/src/gyxx_flow/catalog.py b/src/gyxx_flow/catalog.py
new file mode 100644
index 0000000..1b7e48b
--- /dev/null
+++ b/src/gyxx_flow/catalog.py
@@ -0,0 +1,219 @@
+"""Validated workflow and schedule catalog."""
+
+from __future__ import annotations
+
+import json
+import re
+from dataclasses import dataclass
+from datetime import time
+from pathlib import Path, PurePosixPath
+from typing import Any, Literal
+
+_WORKFLOW_ID = re.compile(r"^[a-z][a-z0-9]*(?:[._-][a-z0-9]+)*$")
+_MODULES = {
+    "content_marketing",
+    "product_commerce",
+    "shop_intelligence",
+    "supply_chain",
+}
+_TRIGGERS = {"scheduled", "manual", "unavailable"}
+_KINDS = {"daily", "weekly", "monthly", "interval_days"}
+_WEEKDAYS = {"Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday"}
+
+
+class CatalogError(ValueError):
+    """Raised for an invalid workflow catalog."""
+
+
+@dataclass(frozen=True, slots=True)
+class WorkflowEntry:
+    workflow_id: str
+    module: str
+    trigger: Literal["scheduled", "manual", "unavailable"]
+    entry: str
+    args: tuple[str, ...] = ()
+    source_project: str | None = None
+    source_task_name: str | None = None
+    note: str | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class ScheduleEntry:
+    workflow_id: str
+    kind: Literal["daily", "weekly", "monthly", "interval_days"]
+    at: str
+    days: tuple[str, ...] = ()
+    day_of_month: int | None = None
+    every_days: int | None = None
+    anchor_date: str | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class WorkflowCatalog:
+    timezone: str
+    workflows: tuple[WorkflowEntry, ...]
+    schedules: tuple[ScheduleEntry, ...]
+
+    @classmethod
+    def load(cls, config_dir: Path | str) -> "WorkflowCatalog":
+        root = Path(config_dir)
+        workflow_payload = _load_json(root / "workflows.json")
+        schedule_payload = _load_json(root / "schedules.json")
+        _require_schema(workflow_payload, "workflows.json", expected=2)
+        _require_schema(schedule_payload, "schedules.json", expected=1)
+        workflows = tuple(_parse_workflow(item) for item in workflow_payload.get("workflows", []))
+        schedules = tuple(_parse_schedule(item) for item in schedule_payload.get("schedules", []))
+        timezone = schedule_payload.get("timezone")
+        if not isinstance(timezone, str) or not timezone:
+            raise CatalogError("schedules.json requires timezone")
+        catalog = cls(timezone=timezone, workflows=workflows, schedules=schedules)
+        catalog._validate_relations()
+        return catalog
+
+    def _validate_relations(self) -> None:
+        workflow_ids = [item.workflow_id for item in self.workflows]
+        duplicate_ids = _duplicates(workflow_ids)
+        if duplicate_ids:
+            raise CatalogError(f"duplicate workflow id: {sorted(duplicate_ids)[0]}")
+
+        task_names = [item.source_task_name for item in self.scheduled_workflows()]
+        duplicates = _duplicates([name for name in task_names if name])
+        if duplicates:
+            raise CatalogError(f"duplicate legacy task name: {sorted(duplicates)[0]}")
+
+        known = set(workflow_ids)
+        schedule_ids = [item.workflow_id for item in self.schedules]
+        duplicate_schedules = _duplicates(schedule_ids)
+        if duplicate_schedules:
+            raise CatalogError(f"duplicate schedule: {sorted(duplicate_schedules)[0]}")
+        unknown = set(schedule_ids) - known
+        if unknown:
+            raise CatalogError(f"schedule references unknown workflow: {sorted(unknown)[0]}")
+        scheduled_ids = {item.workflow_id for item in self.scheduled_workflows()}
+        if set(schedule_ids) != scheduled_ids:
+            missing = scheduled_ids - set(schedule_ids)
+            extra = set(schedule_ids) - scheduled_ids
+            raise CatalogError(f"schedule relation mismatch; missing={sorted(missing)}, extra={sorted(extra)}")
+
+    def scheduled_workflows(self) -> tuple[WorkflowEntry, ...]:
+        return tuple(item for item in self.workflows if item.trigger == "scheduled")
+
+    def manual_workflows(self) -> tuple[WorkflowEntry, ...]:
+        return tuple(item for item in self.workflows if item.trigger == "manual")
+
+    def unavailable_workflows(self) -> tuple[WorkflowEntry, ...]:
+        return tuple(item for item in self.workflows if item.trigger == "unavailable")
+
+    def schedule_for(self, workflow_id: str) -> ScheduleEntry:
+        for schedule in self.schedules:
+            if schedule.workflow_id == workflow_id:
+                return schedule
+        raise KeyError(workflow_id)
+
+
+def _load_json(path: Path) -> dict[str, Any]:
+    try:
+        payload = json.loads(path.read_text(encoding="utf-8"))
+    except (OSError, json.JSONDecodeError) as exc:
+        raise CatalogError(f"cannot load {path.name}: {exc}") from exc
+    if not isinstance(payload, dict):
+        raise CatalogError(f"{path.name} must contain an object")
+    return payload
+
+
+def _require_schema(payload: dict[str, Any], filename: str, *, expected: int) -> None:
+    if payload.get("schema_version") != expected:
+        raise CatalogError(f"unsupported schema_version in {filename}")
+
+
+def _parse_workflow(item: Any) -> WorkflowEntry:
+    if not isinstance(item, dict):
+        raise CatalogError("workflow entry must be an object")
+    workflow_id = item.get("id")
+    module = item.get("module")
+    trigger = item.get("trigger")
+    execution = item.get("execution")
+    provenance = item.get("provenance", {})
+    if not isinstance(workflow_id, str) or not _WORKFLOW_ID.fullmatch(workflow_id):
+        raise CatalogError(f"invalid workflow id: {workflow_id!r}")
+    if module not in _MODULES:
+        raise CatalogError(f"invalid module for {workflow_id}: {module!r}")
+    if trigger not in _TRIGGERS:
+        raise CatalogError(f"invalid trigger for {workflow_id}: {trigger!r}")
+    if not isinstance(execution, dict):
+        raise CatalogError(f"missing execution definition for {workflow_id}")
+    if not isinstance(provenance, dict):
+        raise CatalogError(f"invalid provenance definition for {workflow_id}")
+    entry = execution.get("entry")
+    if not _is_relative_entry(entry):
+        raise CatalogError(f"execution entry must be relative for {workflow_id}")
+    project = provenance.get("source_project")
+    if project is not None and (not isinstance(project, str) or not project):
+        raise CatalogError(f"invalid source project for {workflow_id}")
+    task_name = provenance.get("task_name")
+    if trigger == "scheduled" and (not isinstance(task_name, str) or not task_name):
+        raise CatalogError(f"scheduled workflow requires source task name: {workflow_id}")
+    args = execution.get("args", [])
+    if not isinstance(args, list) or not all(isinstance(value, str) for value in args):
+        raise CatalogError(f"execution args must be strings for {workflow_id}")
+    return WorkflowEntry(
+        workflow_id=workflow_id,
+        module=module,
+        trigger=trigger,
+        entry=entry,
+        args=tuple(args),
+        source_project=project,
+        source_task_name=task_name,
+        note=item.get("note"),
+    )
+
+
+def _is_relative_entry(entry: Any) -> bool:
+    if not isinstance(entry, str) or not entry or "\\" in entry or ":" in entry:
+        return False
+    path = PurePosixPath(entry)
+    return not path.is_absolute() and ".." not in path.parts
+
+
+def _parse_schedule(item: Any) -> ScheduleEntry:
+    if not isinstance(item, dict):
+        raise CatalogError("schedule entry must be an object")
+    workflow_id = item.get("workflow_id")
+    kind = item.get("kind")
+    at = item.get("at")
+    if not isinstance(workflow_id, str) or not _WORKFLOW_ID.fullmatch(workflow_id):
+        raise CatalogError(f"invalid schedule workflow id: {workflow_id!r}")
+    if kind not in _KINDS:
+        raise CatalogError(f"invalid schedule kind for {workflow_id}: {kind!r}")
+    try:
+        time.fromisoformat(at)
+    except (TypeError, ValueError) as exc:
+        raise CatalogError(f"invalid schedule time for {workflow_id}: {at!r}") from exc
+    days = tuple(item.get("days", []))
+    if kind == "weekly" and (not days or not set(days) <= _WEEKDAYS):
+        raise CatalogError(f"invalid weekly days for {workflow_id}")
+    day_of_month = item.get("day_of_month")
+    if kind == "monthly" and (not isinstance(day_of_month, int) or not 1 <= day_of_month <= 31):
+        raise CatalogError(f"invalid day_of_month for {workflow_id}")
+    every_days = item.get("every_days")
+    if kind == "interval_days" and (not isinstance(every_days, int) or every_days < 1):
+        raise CatalogError(f"invalid every_days for {workflow_id}")
+    return ScheduleEntry(
+        workflow_id=workflow_id,
+        kind=kind,
+        at=at,
+        days=days,
+        day_of_month=day_of_month,
+        every_days=every_days,
+        anchor_date=item.get("anchor_date"),
+    )
+
+
+def _duplicates(values: list[str]) -> set[str]:
+    seen: set[str] = set()
+    duplicates: set[str] = set()
+    for value in values:
+        if value in seen:
+            duplicates.add(value)
+        seen.add(value)
+    return duplicates
diff --git a/src/gyxx_flow/cli.py b/src/gyxx_flow/cli.py
new file mode 100644
index 0000000..a6dabca
--- /dev/null
+++ b/src/gyxx_flow/cli.py
@@ -0,0 +1,545 @@
+"""Operator-safe command-line entry point for GYXX Flow."""
+
+from __future__ import annotations
+
+import argparse
+import hashlib
+import json
+import sys
+from collections.abc import Sequence
+from contextlib import suppress
+from datetime import date, datetime, timedelta, timezone
+from pathlib import Path
+from typing import TextIO
+from zoneinfo import ZoneInfo
+
+from gyxx_flow.acceptance import build_acceptance_report
+from gyxx_flow.adapters.native import DeferredModuleCommandStep
+from gyxx_flow.catalog import CatalogError, WorkflowCatalog, WorkflowEntry
+from gyxx_flow.core.artifacts import atomic_write_json
+from gyxx_flow.core.config import Settings
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.core.layout import DataLayout
+from gyxx_flow.core.locks import LockManager
+from gyxx_flow.core.records import RunJournal
+from gyxx_flow.diagnostics import run_doctor
+from gyxx_flow.modules import create_default_registry as create_module_registry
+from gyxx_flow.ops import EffectLedger, RunIndex
+from gyxx_flow.scheduler import (
+    PowerShellCurrentTaskProvider,
+    SchedulerConfig,
+    build_schedule_plan,
+    detect_schedule_drift,
+    write_schedule_plan_bundle,
+)
+from gyxx_flow.script_catalog import ScriptCatalog, ScriptCatalogError
+from gyxx_flow.workflow.engine import WorkflowEngine, WorkflowRunResult
+from gyxx_flow.workflow.model import StepDefinition, WorkflowDefinition
+from gyxx_flow.workflow.registry import WorkflowRegistry, WorkflowRegistryError
+from gyxx_flow.workflow.selection import rerun_step, resume_from
+
+EXIT_SUCCESS = 0
+EXIT_WORKFLOW_FAILED = 1
+EXIT_USAGE = 2
+EXIT_CONFIGURATION = 3
+EXIT_RUNTIME = 4
+EXIT_ACCEPTANCE_INCOMPLETE = 1
+MAX_BACKFILL_DAYS = 366
+
+
+class CliConfigurationError(ValueError):
+    """Safe operator-facing configuration or input error."""
+
+
+class CliRuntimeError(RuntimeError):
+    """Sanitized runtime failure."""
+
+
+def build_parser() -> argparse.ArgumentParser:
+    """Build the CLI parser without initializing runtime resources."""
+
+    parser = argparse.ArgumentParser(
+        prog="gyxx",
+        description="GYXX Flow workflow orchestration",
+    )
+    commands = parser.add_subparsers(dest="command", required=True)
+
+    list_parser = commands.add_parser("list", help="list catalog workflows")
+    list_parser.add_argument("--json", action="store_true", help="emit machine-readable JSON")
+
+    scripts_parser = commands.add_parser(
+        "scripts", help="discover and manually run project-owned module scripts"
+    )
+    script_commands = scripts_parser.add_subparsers(dest="scripts_command", required=True)
+    scripts_list_parser = script_commands.add_parser("list", help="list runnable scripts")
+    scripts_list_parser.add_argument("--module")
+    scripts_list_parser.add_argument("--json", action="store_true")
+    scripts_run_parser = script_commands.add_parser("run", help="run one local script")
+    scripts_run_parser.add_argument("script_id")
+    scripts_run_date = scripts_run_parser.add_mutually_exclusive_group(required=True)
+    scripts_run_date.add_argument("--date", help="business date (YYYY-MM-DD)")
+    scripts_run_date.add_argument("--scheduled", action="store_true")
+    scripts_run_parser.add_argument("--execute", action="store_true")
+    scripts_run_parser.add_argument("--shadow", action="store_true")
+    scripts_run_parser.add_argument(
+        "--arg",
+        dest="script_args",
+        action="append",
+        default=[],
+        help="one argument passed to the script; repeat for multiple arguments",
+    )
+
+    run_parser = commands.add_parser("run", help="run one workflow business date")
+    _add_execution_arguments(run_parser)
+    run_date = run_parser.add_mutually_exclusive_group(required=True)
+    run_date.add_argument("--date", help="business date (YYYY-MM-DD)")
+    run_date.add_argument(
+        "--scheduled",
+        action="store_true",
+        help="execute for today's Asia/Shanghai business date",
+    )
+
+    backfill_parser = commands.add_parser(
+        "backfill", help="run an inclusive business-date range"
+    )
+    _add_execution_arguments(backfill_parser)
+    backfill_parser.add_argument(
+        "--from", dest="from_date", required=True, help="first business date"
+    )
+    backfill_parser.add_argument(
+        "--to", dest="to_date", required=True, help="last business date"
+    )
+
+    schedule_parser = commands.add_parser("schedule", help="plan managed schedules")
+    schedule_commands = schedule_parser.add_subparsers(
+        dest="schedule_command", required=True
+    )
+    plan_parser = schedule_commands.add_parser(
+        "plan", help="write an inert Task Scheduler bundle and drift report"
+    )
+    plan_parser.add_argument("--output", type=Path)
+    plan_parser.add_argument("--start-date", help="trigger boundary date (YYYY-MM-DD)")
+    plan_parser.add_argument("--python-executable", type=Path, required=True)
+
+    doctor_parser = commands.add_parser("doctor", help="run environment preflight checks")
+    doctor_parser.add_argument("--json", action="store_true")
+
+    acceptance_parser = commands.add_parser(
+        "acceptance", help="inspect acceptance evidence"
+    )
+    acceptance_commands = acceptance_parser.add_subparsers(
+        dest="acceptance_command", required=True
+    )
+    status_parser = acceptance_commands.add_parser("status")
+    status_parser.add_argument("--json", action="store_true")
+    status_parser.add_argument("--output", type=Path)
+    return parser
+
+
+def _add_execution_arguments(parser: argparse.ArgumentParser) -> None:
+    parser.add_argument("workflow_id")
+    parser.add_argument(
+        "--execute",
+        action="store_true",
+        help="execute registered steps; omitted means no-side-effect dry-run",
+    )
+    parser.add_argument("--shadow", action="store_true")
+    recovery = parser.add_mutually_exclusive_group()
+    recovery.add_argument("--resume-from", metavar="STEP_ID")
+    recovery.add_argument("--rerun-step", metavar="STEP_ID")
+
+
+def build_default_registry(settings: Settings) -> WorkflowRegistry:
+    """Compose only workflows explicitly implemented by migrated modules."""
+
+    catalog = WorkflowCatalog.load(settings.project_root / "config")
+    registry = WorkflowRegistry(catalog)
+    modules = create_module_registry(catalog=catalog)
+    for workflow_id in modules.workflow_ids:
+        registry.register(modules.workflow(workflow_id))
+    return registry
+
+
+def main(
+    argv: Sequence[str] | None = None,
+    *,
+    registry: WorkflowRegistry | None = None,
+    settings: Settings | None = None,
+    stdout: TextIO | None = None,
+    stderr: TextIO | None = None,
+    current_task_provider: PowerShellCurrentTaskProvider | None = None,
+) -> int:
+    """Execute a CLI command and return a stable process exit code."""
+
+    output = stdout or sys.stdout
+    errors = stderr or sys.stderr
+    parser = build_parser()
+    try:
+        arguments = parser.parse_args(argv)
+    except SystemExit as exc:
+        return int(exc.code)
+
+    resolved_settings = settings or Settings.from_env()
+    try:
+        if arguments.command == "doctor":
+            return _doctor_command(arguments, resolved_settings, output)
+        if (
+            arguments.command == "acceptance"
+            and arguments.acceptance_command == "status"
+        ):
+            return _acceptance_status_command(arguments, resolved_settings, output)
+        if arguments.command == "scripts" and arguments.scripts_command == "list":
+            return _list_scripts(arguments, output)
+        if arguments.command == "scripts" and arguments.scripts_command == "run":
+            return _run_script(arguments, resolved_settings, output)
+        resolved_registry = registry or build_default_registry(resolved_settings)
+        if arguments.command == "list":
+            return _list_workflows(arguments, resolved_registry, output)
+        if arguments.command == "run":
+            return _run_command(arguments, resolved_registry, resolved_settings, output)
+        if arguments.command == "backfill":
+            return _backfill_command(arguments, resolved_registry, resolved_settings, output)
+        if arguments.command == "schedule" and arguments.schedule_command == "plan":
+            return _schedule_plan_command(
+                arguments,
+                resolved_registry.catalog,
+                resolved_settings,
+                output,
+                current_task_provider=current_task_provider,
+            )
+        raise CliConfigurationError(f"unsupported command: {arguments.command}")
+    except (
+        CatalogError,
+        ScriptCatalogError,
+        WorkflowRegistryError,
+        CliConfigurationError,
+        ValueError,
+    ) as exc:
+        errors.write(f"error: {exc}\n")
+        return EXIT_CONFIGURATION
+    except CliRuntimeError as exc:
+        errors.write(f"error: {exc}\n")
+        return EXIT_RUNTIME
+
+
+def _doctor_command(
+    arguments: argparse.Namespace, settings: Settings, output: TextIO
+) -> int:
+    report = run_doctor(settings)
+    if arguments.json:
+        _write_json(output, report.as_dict())
+    else:
+        for check in report.checks:
+            marker = "PASS" if check.passed else "FAIL"
+            output.write(f"{marker}\t{check.name}\t{check.message}\n")
+    return EXIT_SUCCESS if report.is_healthy else EXIT_RUNTIME
+
+
+def _acceptance_status_command(
+    arguments: argparse.Namespace, settings: Settings, output: TextIO
+) -> int:
+    report = build_acceptance_report(settings)
+    payload = report.as_dict()
+    if arguments.output is not None:
+        atomic_write_json(arguments.output, payload)
+    if arguments.json:
+        _write_json(output, payload)
+    else:
+        summary = payload["summary"]
+        output.write(
+            f"acceptance: {summary['completed']}/{summary['total']} complete; "
+            f"pending={summary['pending']}\n"
+        )
+    return EXIT_SUCCESS if report.is_complete else EXIT_ACCEPTANCE_INCOMPLETE
+
+
+def _list_workflows(
+    arguments: argparse.Namespace,
+    registry: WorkflowRegistry,
+    output: TextIO,
+) -> int:
+    rows = [
+        {
+            "id": entry.workflow_id,
+            "module": entry.module,
+            "registered": registry.is_registered(entry.workflow_id),
+            "trigger": entry.trigger,
+        }
+        for entry in registry.catalog.workflows
+    ]
+    if arguments.json:
+        _write_json(output, rows)
+    else:
+        for row in rows:
+            state = "ready" if row["registered"] else "not-registered"
+            output.write(
+                f"{row['id']}\t{row['module']}\t{row['trigger']}\t{state}\n"
+            )
+    return EXIT_SUCCESS
+
+
+def _list_scripts(arguments: argparse.Namespace, output: TextIO) -> int:
+    catalog = ScriptCatalog.discover_default()
+    rows = [
+        {
+            "id": script.script_id,
+            "module": script.module,
+            "entry": script.entry,
+            "kind": script.kind,
+        }
+        for script in catalog.scripts
+        if arguments.module is None or script.module == arguments.module
+    ]
+    if arguments.module is not None and not rows:
+        raise CliConfigurationError(f"unknown module or no runnable scripts: {arguments.module}")
+    if arguments.json:
+        _write_json(output, rows)
+    else:
+        for row in rows:
+            output.write(f"{row['id']}\t{row['kind']}\n")
+    return EXIT_SUCCESS
+
+
+def _run_script(
+    arguments: argparse.Namespace,
+    settings: Settings,
+    output: TextIO,
+) -> int:
+    script = ScriptCatalog.discover_default().get(arguments.script_id)
+    suffix = hashlib.sha256(script.script_id.encode("utf-8")).hexdigest()[:12]
+    workflow_id = f"script.{script.module}.{suffix}"
+    script_args = tuple(arguments.script_args)
+    entry = WorkflowEntry(
+        workflow_id=workflow_id,
+        module=script.module,
+        trigger="manual",
+        entry=script.entry,
+        args=script_args,
+    )
+    workflow = WorkflowDefinition(
+        workflow_id,
+        (
+            StepDefinition(
+                "module_script",
+                DeferredModuleCommandStep(entry),
+                timeout_seconds=6 * 60 * 60,
+                max_attempts=1,
+                resources=(f"module:{script.module}",),
+                production_sink=True,
+            ),
+        ),
+    )
+    business_date = _today_shanghai() if arguments.scheduled else _parse_date(arguments.date)
+    exit_code, payload = _execute_once(
+        workflow,
+        business_date=business_date,
+        shadow=arguments.shadow,
+        dry_run=not (arguments.execute or arguments.scheduled),
+        settings=settings,
+    )
+    payload["script_id"] = script.script_id
+    _write_json(output, payload)
+    return exit_code
+
+
+def _run_command(
+    arguments: argparse.Namespace,
+    registry: WorkflowRegistry,
+    settings: Settings,
+    output: TextIO,
+) -> int:
+    registered = registry.resolve(arguments.workflow_id)
+    workflow = _select_workflow(registered.definition, arguments)
+    business_date = _today_shanghai() if arguments.scheduled else _parse_date(arguments.date)
+    dry_run = not (arguments.execute or arguments.scheduled)
+    exit_code, payload = _execute_once(
+        workflow,
+        business_date=business_date,
+        shadow=arguments.shadow,
+        dry_run=dry_run,
+        settings=settings,
+    )
+    _write_json(output, payload)
+    return exit_code
+
+
+def _schedule_plan_command(
+    arguments: argparse.Namespace,
+    catalog: WorkflowCatalog,
+    settings: Settings,
+    output: TextIO,
+    *,
+    current_task_provider: PowerShellCurrentTaskProvider | None,
+) -> int:
+    start_date = (
+        _parse_date(arguments.start_date)
+        if arguments.start_date
+        else _today_shanghai()
+    )
+    destination = arguments.output
+    if destination is None:
+        timestamp = datetime.now(timezone.utc).strftime("%Y%m%dT%H%M%SZ")
+        destination = settings.data_root / "evidence" / "schedule-plans" / timestamp
+    config = SchedulerConfig(
+        python_executable=arguments.python_executable,
+        project_root=settings.project_root,
+    )
+    plan = build_schedule_plan(catalog, config, start_date=start_date)
+    try:
+        bundle = write_schedule_plan_bundle(destination, plan)
+    except FileExistsError as exc:
+        raise CliConfigurationError(
+            f"schedule plan destination already exists: {destination}"
+        ) from exc
+    provider = current_task_provider or PowerShellCurrentTaskProvider()
+    try:
+        current = provider.current_tasks(config.task_path)
+    except Exception as exc:
+        raise CliRuntimeError("cannot read current managed scheduled tasks") from exc
+    drift = detect_schedule_drift(plan, current)
+    atomic_write_json(bundle / "drift.json", drift.as_dict())
+    _write_json(
+        output,
+        {
+            "path": str(bundle),
+            "desired_count": len(plan.tasks),
+            "drift_count": len(drift.items),
+            "is_clean": drift.is_clean,
+            "applied": False,
+        },
+    )
+    return EXIT_SUCCESS
+
+
+def _backfill_command(
+    arguments: argparse.Namespace,
+    registry: WorkflowRegistry,
+    settings: Settings,
+    output: TextIO,
+) -> int:
+    registered = registry.resolve(arguments.workflow_id)
+    workflow = _select_workflow(registered.definition, arguments)
+    start = _parse_date(arguments.from_date)
+    end = _parse_date(arguments.to_date)
+    if end < start:
+        raise CliConfigurationError("backfill end date must not precede start date")
+    day_count = (end - start).days + 1
+    if day_count > MAX_BACKFILL_DAYS:
+        raise CliConfigurationError(
+            f"backfill range exceeds safety limit of {MAX_BACKFILL_DAYS} days"
+        )
+
+    runs: list[dict[str, object]] = []
+    final_exit = EXIT_SUCCESS
+    for offset in range(day_count):
+        business_date = start + timedelta(days=offset)
+        exit_code, payload = _execute_once(
+            workflow,
+            business_date=business_date,
+            shadow=arguments.shadow,
+            dry_run=not arguments.execute,
+            settings=settings,
+        )
+        runs.append(payload)
+        if exit_code != EXIT_SUCCESS:
+            final_exit = EXIT_WORKFLOW_FAILED
+    _write_json(
+        output,
+        {
+            "workflow_id": workflow.workflow_id,
+            "from": start.isoformat(),
+            "to": end.isoformat(),
+            "dry_run": not arguments.execute,
+            "status": "success" if final_exit == EXIT_SUCCESS else "failed",
+            "runs": runs,
+        },
+    )
+    return final_exit
+
+
+def _select_workflow(workflow, arguments: argparse.Namespace):  # type: ignore[no-untyped-def]
+    if arguments.rerun_step:
+        return rerun_step(workflow, arguments.rerun_step)
+    if arguments.resume_from:
+        return resume_from(workflow, arguments.resume_from)
+    return workflow
+
+
+def _execute_once(
+    workflow,  # type: ignore[no-untyped-def]
+    *,
+    business_date: date,
+    shadow: bool,
+    dry_run: bool,
+    settings: Settings,
+) -> tuple[int, dict[str, object]]:
+    context = RunContext.create(
+        workflow.workflow_id,
+        business_date.isoformat(),
+        shadow=shadow,
+    )
+    layout = DataLayout(settings.data_root)
+    journal: RunJournal | None = None
+    try:
+        journal = RunJournal.create(layout, context)
+        result = WorkflowEngine(
+            LockManager(settings.data_root / "state" / "locks"),
+            effect_ledger=EffectLedger(settings.data_root),
+        ).execute(workflow, context=context, journal=journal, dry_run=dry_run)
+        RunIndex(settings.data_root).index_journal(journal)
+    except Exception as exc:
+        if journal is not None:
+            with suppress(OSError, ValueError, KeyError, TypeError, json.JSONDecodeError):
+                journal.finalize("failed", error="runtime infrastructure failure")
+                RunIndex(settings.data_root).index_journal(journal)
+        raise CliRuntimeError(
+            f"runtime failure while executing {workflow.workflow_id}"
+        ) from exc
+    return result.exit_code, _result_payload(
+        context,
+        result,
+        dry_run=dry_run,
+        journal_path=journal.path.relative_to(settings.data_root).as_posix(),
+    )
+
+
+def _result_payload(
+    context: RunContext,
+    result: WorkflowRunResult,
+    *,
+    dry_run: bool,
+    journal_path: str,
+) -> dict[str, object]:
+    return {
+        "workflow_id": result.workflow_id,
+        "run_id": context.run_id,
+        "business_date": context.business_date.isoformat(),
+        "shadow": context.shadow,
+        "dry_run": dry_run,
+        "status": result.status,
+        "exit_code": result.exit_code,
+        "journal_path": journal_path,
+        "steps": {
+            step_id: step.status for step_id, step in result.steps.items()
+        },
+        "warnings": list(result.warnings),
+    }
+
+
+def _parse_date(value: str) -> date:
+    try:
+        parsed = date.fromisoformat(value)
+    except (TypeError, ValueError) as exc:
+        raise CliConfigurationError(f"invalid business date: {value!r}") from exc
+    if parsed.isoformat() != value:
+        raise CliConfigurationError(f"invalid business date: {value!r}")
+    return parsed
+
+
+def _today_shanghai() -> date:
+    return datetime.now(ZoneInfo("Asia/Shanghai")).date()
+
+
+def _write_json(output: TextIO, payload: object) -> None:
+    json.dump(payload, output, ensure_ascii=False, sort_keys=True)
+    output.write("\n")
diff --git a/src/gyxx_flow/core/__init__.py b/src/gyxx_flow/core/__init__.py
new file mode 100644
index 0000000..2bc5f30
--- /dev/null
+++ b/src/gyxx_flow/core/__init__.py
@@ -0,0 +1,7 @@
+"""Core contracts and runtime primitives for GYXX Flow."""
+
+from .config import Settings
+from .context import RunContext
+from .layout import DataLayout, ModuleDataPaths
+
+__all__ = ["DataLayout", "ModuleDataPaths", "RunContext", "Settings"]
diff --git a/src/gyxx_flow/core/artifacts.py b/src/gyxx_flow/core/artifacts.py
new file mode 100644
index 0000000..b6e468b
--- /dev/null
+++ b/src/gyxx_flow/core/artifacts.py
@@ -0,0 +1,87 @@
+"""Immutable artifact metadata and atomic file helpers."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import os
+import uuid
+from dataclasses import asdict, dataclass
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+
+
+def atomic_write_json(path: Path | str, payload: Any) -> Path:
+    destination = Path(path)
+    destination.parent.mkdir(parents=True, exist_ok=True)
+    temporary = destination.with_name(f".{destination.name}.{uuid.uuid4().hex}.tmp")
+    try:
+        with temporary.open("w", encoding="utf-8", newline="\n") as handle:
+            json.dump(payload, handle, ensure_ascii=False, indent=2, sort_keys=True)
+            handle.write("\n")
+            handle.flush()
+            os.fsync(handle.fileno())
+        os.replace(temporary, destination)
+    finally:
+        temporary.unlink(missing_ok=True)
+    return destination
+
+
+def sha256_file(path: Path | str, *, chunk_size: int = 1024 * 1024) -> str:
+    digest = hashlib.sha256()
+    with Path(path).open("rb") as handle:
+        for chunk in iter(lambda: handle.read(chunk_size), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+@dataclass(frozen=True, slots=True)
+class ArtifactManifest:
+    artifact_id: str
+    dataset: str
+    workflow_id: str
+    run_id: str
+    business_date: str
+    schema_version: str
+    source_path: str
+    byte_size: int
+    sha256: str
+    row_count: int | None
+    upstream_artifact_ids: tuple[str, ...]
+    created_at: str
+
+    @classmethod
+    def from_file(
+        cls,
+        path: Path | str,
+        *,
+        artifact_id: str,
+        dataset: str,
+        workflow_id: str,
+        run_id: str,
+        business_date: str,
+        schema_version: str,
+        row_count: int | None = None,
+        upstream_artifact_ids: tuple[str, ...] = (),
+    ) -> "ArtifactManifest":
+        source = Path(path).resolve()
+        return cls(
+            artifact_id=artifact_id,
+            dataset=dataset,
+            workflow_id=workflow_id,
+            run_id=run_id,
+            business_date=business_date,
+            schema_version=schema_version,
+            source_path=str(source),
+            byte_size=source.stat().st_size,
+            sha256=sha256_file(source),
+            row_count=row_count,
+            upstream_artifact_ids=tuple(upstream_artifact_ids),
+            created_at=datetime.now(timezone.utc).isoformat(timespec="seconds"),
+        )
+
+    def to_dict(self) -> dict[str, Any]:
+        payload = asdict(self)
+        payload["upstream_artifact_ids"] = list(self.upstream_artifact_ids)
+        return payload
diff --git a/src/gyxx_flow/core/config.py b/src/gyxx_flow/core/config.py
new file mode 100644
index 0000000..5618b96
--- /dev/null
+++ b/src/gyxx_flow/core/config.py
@@ -0,0 +1,30 @@
+"""Portable workspace configuration."""
+
+from __future__ import annotations
+
+import os
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Mapping
+
+
+@dataclass(frozen=True, slots=True)
+class Settings:
+    """Resolved paths used by the orchestration runtime."""
+
+    project_root: Path
+    data_root: Path
+
+    @classmethod
+    def from_env(
+        cls,
+        *,
+        project_root: Path | str | None = None,
+        env: Mapping[str, str] | None = None,
+    ) -> "Settings":
+        values = os.environ if env is None else env
+        root = Path(project_root) if project_root is not None else Path(__file__).resolve().parents[3]
+        root = root.expanduser().resolve()
+        configured_data_root = values.get("GYXX_DATA_ROOT", "").strip()
+        data_root = Path(configured_data_root).expanduser() if configured_data_root else root / "var"
+        return cls(project_root=root, data_root=data_root.resolve())
diff --git a/src/gyxx_flow/core/context.py b/src/gyxx_flow/core/context.py
new file mode 100644
index 0000000..25267c6
--- /dev/null
+++ b/src/gyxx_flow/core/context.py
@@ -0,0 +1,56 @@
+"""Traceable workflow run context."""
+
+from __future__ import annotations
+
+import re
+import secrets
+from dataclasses import dataclass
+from datetime import date, datetime, timezone
+
+_WORKFLOW_ID = re.compile(r"^[a-z][a-z0-9]*(?:[._-][a-z0-9]+)*$")
+_SUFFIX = re.compile(r"^[a-z0-9]+$")
+
+
+@dataclass(frozen=True, slots=True)
+class RunContext:
+    workflow_id: str
+    run_id: str
+    business_date: date
+    started_at: datetime
+    shadow: bool = False
+
+    @classmethod
+    def create(
+        cls,
+        workflow_id: str,
+        business_date: str,
+        *,
+        shadow: bool = False,
+        now: datetime | None = None,
+        random_suffix: str | None = None,
+    ) -> "RunContext":
+        if not _WORKFLOW_ID.fullmatch(workflow_id):
+            raise ValueError(f"unsafe workflow_id: {workflow_id!r}")
+        try:
+            parsed_date = date.fromisoformat(business_date)
+        except (TypeError, ValueError) as exc:
+            raise ValueError(f"invalid business_date: {business_date!r}") from exc
+        if parsed_date.isoformat() != business_date:
+            raise ValueError(f"invalid business_date: {business_date!r}")
+
+        started_at = now or datetime.now(timezone.utc)
+        if started_at.tzinfo is None:
+            started_at = started_at.replace(tzinfo=timezone.utc)
+        started_at = started_at.astimezone(timezone.utc)
+        suffix = random_suffix or secrets.token_hex(3)
+        if not _SUFFIX.fullmatch(suffix):
+            raise ValueError(f"unsafe random suffix: {suffix!r}")
+        timestamp = started_at.strftime("%Y%m%dT%H%M%SZ")
+        run_id = f"{workflow_id}__{parsed_date:%Y%m%d}__{timestamp}__{suffix}"
+        return cls(
+            workflow_id=workflow_id,
+            run_id=run_id,
+            business_date=parsed_date,
+            started_at=started_at,
+            shadow=shadow,
+        )
diff --git a/src/gyxx_flow/core/layout.py b/src/gyxx_flow/core/layout.py
new file mode 100644
index 0000000..a79cc4b
--- /dev/null
+++ b/src/gyxx_flow/core/layout.py
@@ -0,0 +1,147 @@
+"""Canonical runtime data layout."""
+
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Self
+
+_SAFE_SEGMENT = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._=-]*$")
+
+
+def _segment(value: str) -> str:
+    if not isinstance(value, str) or not _SAFE_SEGMENT.fullmatch(value) or value in {".", ".."}:
+        raise ValueError(f"unsafe path segment: {value!r}")
+    return value
+
+
+@dataclass(frozen=True, slots=True)
+class DataLayout:
+    root: Path
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "root", Path(self.root).expanduser().resolve())
+
+    def for_module(self, module: str) -> ModuleDataPaths:
+        """Return the canonical mutable-data boundary for one business module."""
+
+        return ModuleDataPaths.from_layout(self, module)
+
+    def raw(
+        self,
+        *,
+        domain: str,
+        source: str,
+        dataset: str,
+        business_date: str,
+        run_id: str,
+    ) -> Path:
+        return self.root.joinpath(
+            "data",
+            "raw",
+            _segment(domain),
+            _segment(source),
+            _segment(dataset),
+            f"business_date={_segment(business_date)}",
+            f"run_id={_segment(run_id)}",
+        )
+
+    def normalized(self, domain: str, dataset: str, schema_version: str) -> Path:
+        return self.root / "data" / "normalized" / _segment(domain) / _segment(dataset) / f"schema_v{_segment(schema_version)}"
+
+    def curated(self, domain: str, dataset: str) -> Path:
+        return self.root / "data" / "curated" / _segment(domain) / _segment(dataset)
+
+    def export(self, consumer: str, workflow_id: str, business_date: str, run_id: str) -> Path:
+        return self.root.joinpath(
+            "data", "exports", _segment(consumer), _segment(workflow_id), _segment(business_date), _segment(run_id)
+        )
+
+    def evidence(self, workflow_id: str, run_id: str) -> Path:
+        return self.root / "data" / "evidence" / _segment(workflow_id) / _segment(run_id)
+
+    def state(self, category: str, name: str) -> Path:
+        return self.root / "state" / _segment(category) / _segment(name)
+
+    def log_dir(self, workflow_id: str, business_date: str) -> Path:
+        year, month, day = _segment(business_date).split("-")
+        return self.root / "logs" / _segment(workflow_id) / year / month / day
+
+    def run_dir(self, workflow_id: str, business_date: str, run_id: str) -> Path:
+        year, month, day = _segment(business_date).split("-")
+        return self.root / "runs" / _segment(workflow_id) / year / month / day / _segment(run_id)
+
+    def tmp(self, run_id: str) -> Path:
+        return self.root / "tmp" / _segment(run_id)
+
+    def quarantine(self, workflow_id: str, run_id: str) -> Path:
+        return self.root / "quarantine" / _segment(workflow_id) / _segment(run_id)
+
+
+@dataclass(frozen=True, slots=True)
+class ModuleDataPaths:
+    """Canonical, relocatable roots used by a migrated business module.
+
+    Construction is side-effect free.  Callers create only the specific parent
+    directory they are about to write, keeping imports and dry-runs read-only.
+    """
+
+    data_root: Path
+    module: str
+    raw_root: Path
+    normalized_root: Path
+    curated_root: Path
+    exports_root: Path
+    evidence_root: Path
+    state_root: Path
+    logs_root: Path
+    tmp_root: Path
+
+    @classmethod
+    def from_layout(cls, layout: DataLayout, module: str) -> Self:
+        safe_module = _segment(module)
+        root = layout.root
+        return cls(
+            data_root=root,
+            module=safe_module,
+            raw_root=root / "data" / "raw" / safe_module,
+            normalized_root=root / "data" / "normalized" / safe_module,
+            curated_root=root / "data" / "curated" / safe_module,
+            exports_root=root / "data" / "exports" / safe_module,
+            evidence_root=root / "data" / "evidence" / safe_module,
+            state_root=root / "state" / safe_module,
+            logs_root=root / "logs" / safe_module,
+            tmp_root=root / "tmp" / safe_module,
+        )
+
+    def raw_path(self, *parts: str | Path) -> Path:
+        return _inside(self.raw_root, *parts)
+
+    def normalized_path(self, *parts: str | Path) -> Path:
+        return _inside(self.normalized_root, *parts)
+
+    def curated_path(self, *parts: str | Path) -> Path:
+        return _inside(self.curated_root, *parts)
+
+    def export_path(self, *parts: str | Path) -> Path:
+        return _inside(self.exports_root, *parts)
+
+    def evidence_path(self, *parts: str | Path) -> Path:
+        return _inside(self.evidence_root, *parts)
+
+    def state_path(self, *parts: str | Path) -> Path:
+        return _inside(self.state_root, *parts)
+
+    def log_path(self, *parts: str | Path) -> Path:
+        return _inside(self.logs_root, *parts)
+
+    def tmp_path(self, *parts: str | Path) -> Path:
+        return _inside(self.tmp_root, *parts)
+
+
+def _inside(root: Path, *parts: str | Path) -> Path:
+    candidate = root.joinpath(*parts).resolve()
+    if candidate != root and not candidate.is_relative_to(root):
+        raise ValueError(f"path escapes module data root: {parts!r}")
+    return candidate
diff --git a/src/gyxx_flow/core/locks.py b/src/gyxx_flow/core/locks.py
new file mode 100644
index 0000000..490df92
--- /dev/null
+++ b/src/gyxx_flow/core/locks.py
@@ -0,0 +1,88 @@
+"""Cross-process named resource locks backed by exclusive files."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import os
+import re
+import time
+from dataclasses import dataclass
+from datetime import datetime, timezone
+from pathlib import Path
+
+_RESOURCE = re.compile(r"^[A-Za-z0-9][A-Za-z0-9:._-]*$")
+_OWNER = re.compile(r"^[A-Za-z0-9][A-Za-z0-9:._-]*$")
+
+
+class ResourceBusyError(RuntimeError):
+    """Raised when a named resource cannot be acquired before its deadline."""
+
+
+@dataclass(slots=True)
+class NamedResourceLock:
+    path: Path
+    resource: str
+    owner: str
+    timeout_seconds: float
+    poll_seconds: float
+    _acquired: bool = False
+
+    def __enter__(self) -> "NamedResourceLock":
+        deadline = time.monotonic() + max(0.0, self.timeout_seconds)
+        self.path.parent.mkdir(parents=True, exist_ok=True)
+        while True:
+            try:
+                descriptor = os.open(self.path, os.O_CREAT | os.O_EXCL | os.O_WRONLY)
+            except FileExistsError:
+                if time.monotonic() >= deadline:
+                    raise ResourceBusyError(f"resource is busy: {self.resource}")
+                time.sleep(self.poll_seconds)
+                continue
+            metadata = {
+                "resource": self.resource,
+                "owner": self.owner,
+                "acquired_at": datetime.now(timezone.utc).isoformat(timespec="seconds"),
+                "pid": os.getpid(),
+            }
+            with os.fdopen(descriptor, "w", encoding="utf-8", newline="\n") as handle:
+                json.dump(metadata, handle, ensure_ascii=False, indent=2, sort_keys=True)
+                handle.write("\n")
+                handle.flush()
+                os.fsync(handle.fileno())
+            self._acquired = True
+            return self
+
+    def __exit__(self, exc_type, exc, traceback) -> None:  # type: ignore[no-untyped-def]
+        if self._acquired:
+            self.path.unlink(missing_ok=True)
+            self._acquired = False
+
+
+@dataclass(frozen=True, slots=True)
+class LockManager:
+    root: Path
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "root", Path(self.root).expanduser().resolve())
+
+    def acquire(
+        self,
+        resource: str,
+        *,
+        owner: str,
+        timeout_seconds: float = 30.0,
+        poll_seconds: float = 0.1,
+    ) -> NamedResourceLock:
+        if not _RESOURCE.fullmatch(resource):
+            raise ValueError(f"invalid resource name: {resource!r}")
+        if not _OWNER.fullmatch(owner):
+            raise ValueError(f"invalid owner name: {owner!r}")
+        digest = hashlib.sha256(resource.encode("utf-8")).hexdigest()[:32]
+        return NamedResourceLock(
+            path=self.root / f"{digest}.lock",
+            resource=resource,
+            owner=owner,
+            timeout_seconds=timeout_seconds,
+            poll_seconds=poll_seconds,
+        )
diff --git a/src/gyxx_flow/core/records.py b/src/gyxx_flow/core/records.py
new file mode 100644
index 0000000..8bf29df
--- /dev/null
+++ b/src/gyxx_flow/core/records.py
@@ -0,0 +1,135 @@
+"""Atomic workflow run and step status journal."""
+
+from __future__ import annotations
+
+import json
+from dataclasses import dataclass
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any, Literal
+
+from .artifacts import atomic_write_json
+from .context import RunContext
+from .layout import DataLayout
+
+StepStatus = Literal["success", "failed", "skipped"]
+RunStatus = Literal["success", "failed", "cancelled"]
+
+
+def _now() -> str:
+    return datetime.now(timezone.utc).isoformat(timespec="seconds")
+
+
+@dataclass(frozen=True, slots=True)
+class RunJournal:
+    path: Path
+
+    @classmethod
+    def create(cls, layout: DataLayout, context: RunContext) -> "RunJournal":
+        run_dir = layout.run_dir(
+            context.workflow_id,
+            context.business_date.isoformat(),
+            context.run_id,
+        )
+        log_dir = layout.log_dir(context.workflow_id, context.business_date.isoformat())
+        evidence_dir = layout.evidence(context.workflow_id, context.run_id)
+        log_dir.mkdir(parents=True, exist_ok=True)
+        evidence_dir.mkdir(parents=True, exist_ok=True)
+        path = run_dir / "run.json"
+        atomic_write_json(
+            path,
+            {
+                "workflow_id": context.workflow_id,
+                "run_id": context.run_id,
+                "business_date": context.business_date.isoformat(),
+                "shadow": context.shadow,
+                "started_at": context.started_at.isoformat(timespec="seconds"),
+                "ended_at": None,
+                "status": "running",
+                "error": None,
+                "steps": {},
+                "trace": {
+                    "paths": {
+                        "run": run_dir.relative_to(layout.root).as_posix(),
+                        "log": log_dir.relative_to(layout.root).as_posix(),
+                        "evidence": evidence_dir.relative_to(layout.root).as_posix(),
+                    },
+                    "inputs": [],
+                    "outputs": [],
+                    "external_writes": [],
+                },
+            },
+        )
+        return cls(path=path)
+
+    def _read(self) -> dict[str, Any]:
+        return json.loads(self.path.read_text(encoding="utf-8"))
+
+    def start_step(self, step_id: str, *, attempt: int) -> None:
+        payload = self._read()
+        previous = payload["steps"].get(step_id)
+        if previous and previous["status"] == "running":
+            raise ValueError(f"step {step_id!r} is already running")
+        payload["steps"][step_id] = {
+            "status": "running",
+            "attempt": attempt,
+            "started_at": _now(),
+            "ended_at": None,
+            "exit_code": None,
+            "error": None,
+        }
+        atomic_write_json(self.path, payload)
+
+    def finish_step(
+        self,
+        step_id: str,
+        *,
+        status: StepStatus,
+        exit_code: int | None = None,
+        error: str | None = None,
+    ) -> None:
+        if status not in {"success", "failed", "skipped"}:
+            raise ValueError(f"invalid step status: {status!r}")
+        payload = self._read()
+        step = payload["steps"].get(step_id)
+        if not step or step["status"] != "running":
+            raise ValueError(f"step {step_id!r} is not running")
+        step.update(
+            {
+                "status": status,
+                "ended_at": _now(),
+                "exit_code": exit_code,
+                "error": error,
+            }
+        )
+        atomic_write_json(self.path, payload)
+
+    def finalize(self, status: RunStatus, *, error: str | None = None) -> None:
+        if status not in {"success", "failed", "cancelled"}:
+            raise ValueError(f"invalid run status: {status!r}")
+        payload = self._read()
+        payload.update({"status": status, "ended_at": _now(), "error": error})
+        atomic_write_json(self.path, payload)
+
+    def record_input(self, reference: str) -> None:
+        self._record_trace("inputs", reference)
+
+    def record_output(self, reference: str) -> None:
+        self._record_trace("outputs", reference)
+
+    def record_external_write(self, reference: str) -> None:
+        self._record_trace("external_writes", reference)
+
+    def _record_trace(self, category: str, reference: str) -> None:
+        if (
+            not isinstance(reference, str)
+            or not reference
+            or len(reference) > 1024
+            or any(ord(character) < 32 for character in reference)
+        ):
+            raise ValueError("invalid trace reference")
+        payload = self._read()
+        references = payload["trace"][category]
+        if reference not in references:
+            references.append(reference)
+            atomic_write_json(self.path, payload)
diff --git a/src/gyxx_flow/diagnostics.py b/src/gyxx_flow/diagnostics.py
new file mode 100644
index 0000000..76ab009
--- /dev/null
+++ b/src/gyxx_flow/diagnostics.py
@@ -0,0 +1,94 @@
+"""Read-only and transient preflight diagnostics for operators."""
+
+from __future__ import annotations
+
+import importlib.util
+import sys
+import tempfile
+from dataclasses import asdict, dataclass
+from pathlib import Path
+
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.core.config import Settings
+
+
+@dataclass(frozen=True, slots=True)
+class DiagnosticCheck:
+    name: str
+    passed: bool
+    message: str
+
+
+@dataclass(frozen=True, slots=True)
+class DoctorReport:
+    checks: tuple[DiagnosticCheck, ...]
+
+    @property
+    def is_healthy(self) -> bool:
+        return all(check.passed for check in self.checks)
+
+    def as_dict(self) -> dict[str, object]:
+        return {
+            "is_healthy": self.is_healthy,
+            "checks": [asdict(check) for check in self.checks],
+        }
+
+
+def run_doctor(settings: Settings) -> DoctorReport:
+    checks = [
+        DiagnosticCheck(
+            "python",
+            sys.version_info >= (3, 12),
+            f"Python {sys.version_info.major}.{sys.version_info.minor}",
+        ),
+        DiagnosticCheck(
+            "cli",
+            importlib.util.find_spec("gyxx_flow.__main__") is not None,
+            "gyxx_flow module entry point is importable",
+        ),
+        DiagnosticCheck(
+            "project_root",
+            settings.project_root.is_dir(),
+            "project root exists" if settings.project_root.is_dir() else "project root is missing",
+        ),
+    ]
+    try:
+        catalog = WorkflowCatalog.load(settings.project_root / "config")
+        catalog_ok = len(catalog.scheduled_workflows()) == 21
+        catalog_message = f"catalog has {len(catalog.scheduled_workflows())} scheduled workflows"
+    except Exception:
+        catalog_ok = False
+        catalog_message = "catalog cannot be validated"
+    checks.append(DiagnosticCheck("catalog", catalog_ok, catalog_message))
+
+    data_root_ok = settings.data_root.is_absolute() and settings.data_root != Path(
+        settings.data_root.anchor
+    )
+    checks.append(
+        DiagnosticCheck(
+            "data_root",
+            data_root_ok,
+            "data root is an explicit non-root path" if data_root_ok else "unsafe data root",
+        )
+    )
+    writable, message = _probe_write_permission(settings.data_root)
+    checks.append(DiagnosticCheck("data_root_write", writable, message))
+    return DoctorReport(tuple(checks))
+
+
+def _probe_write_permission(path: Path) -> tuple[bool, str]:
+    candidate = Path(path)
+    while not candidate.exists() and candidate.parent != candidate:
+        candidate = candidate.parent
+    if not candidate.is_dir():
+        return False, "no existing data-root parent directory"
+    try:
+        with tempfile.NamedTemporaryFile(
+            mode="wb", prefix=".gyxx-doctor-probe-", dir=candidate, delete=True
+        ) as stream:
+            stream.write(b"probe")
+            stream.flush()
+    except OSError:
+        return False, "data-root parent is not writable"
+    return True, "data-root parent write probe passed"
+
diff --git a/src/gyxx_flow/migration/__init__.py b/src/gyxx_flow/migration/__init__.py
new file mode 100644
index 0000000..09eb73e
--- /dev/null
+++ b/src/gyxx_flow/migration/__init__.py
@@ -0,0 +1,33 @@
+"""Safe, read-only migration boundary for legacy workflows."""
+
+from gyxx_flow.migration.data import (
+    DataLayer,
+    HistoricalDataMigrationExecutor,
+    HistoricalDataMigrationPlan,
+    HistoricalDataMigrationPlanner,
+    HistoricalDataMigrationReport,
+    MigrationConflictError,
+    MigrationExecutionError,
+    MigrationSource,
+    MigrationValidationError,
+)
+from gyxx_flow.migration.legacy import (
+    LegacyCommandAdapter,
+    LegacyConfigurationError,
+    LegacyProjectRoots,
+)
+
+__all__ = [
+    "LegacyCommandAdapter",
+    "LegacyConfigurationError",
+    "LegacyProjectRoots",
+    "DataLayer",
+    "HistoricalDataMigrationExecutor",
+    "HistoricalDataMigrationPlan",
+    "HistoricalDataMigrationPlanner",
+    "HistoricalDataMigrationReport",
+    "MigrationConflictError",
+    "MigrationExecutionError",
+    "MigrationSource",
+    "MigrationValidationError",
+]
diff --git a/src/gyxx_flow/migration/data.py b/src/gyxx_flow/migration/data.py
new file mode 100644
index 0000000..403ed46
--- /dev/null
+++ b/src/gyxx_flow/migration/data.py
@@ -0,0 +1,556 @@
+"""Plan and execute read-only-source historical file migrations.
+
+The planner records an immutable source snapshot.  The executor is deliberately
+plan-only unless ``apply=True`` is supplied and only ever copies files into the
+configured data root.
+"""
+
+from __future__ import annotations
+
+import hashlib
+import os
+import re
+import shutil
+import uuid
+from dataclasses import asdict, dataclass
+from datetime import datetime, timezone
+from enum import Enum
+from pathlib import Path
+from typing import Callable, Iterable
+
+from gyxx_flow.core.artifacts import atomic_write_json, sha256_file
+
+_SAFE_ID = re.compile(r"^[a-z][a-z0-9_\-]*$")
+_PROGRESS_INTERVAL = 1000
+
+
+class MigrationValidationError(ValueError):
+    """Raised when a migration cannot be planned safely."""
+
+
+class MigrationConflictError(MigrationValidationError):
+    """Raised when a target exists with different content."""
+
+
+class MigrationExecutionError(RuntimeError):
+    """Raised when an approved migration cannot be completed or reconciled."""
+
+
+class DataLayer(str, Enum):
+    RAW = "raw"
+    NORMALIZED = "normalized"
+    CURATED = "curated"
+    EXPORT = "exports"
+
+
+@dataclass(frozen=True, slots=True)
+class MigrationSource:
+    source_id: str
+    module: str
+    layer: DataLayer
+    root: Path
+
+    def __post_init__(self) -> None:
+        _validate_id(self.source_id, "source_id")
+        _validate_id(self.module, "module")
+        if not isinstance(self.layer, DataLayer):
+            raise MigrationValidationError("layer must be a DataLayer")
+        object.__setattr__(self, "root", Path(self.root).expanduser())
+
+
+@dataclass(frozen=True, slots=True)
+class MigrationFile:
+    source_id: str
+    source_path: str
+    relative_path: str
+    target_relative_path: str
+    byte_size: int
+    sha256: str
+
+    def to_dict(self) -> dict[str, object]:
+        return asdict(self)
+
+
+@dataclass(frozen=True, slots=True)
+class MigrationSummary:
+    file_count: int
+    total_bytes: int
+    aggregate_sha256: str
+
+    def to_dict(self) -> dict[str, object]:
+        return asdict(self)
+
+
+@dataclass(frozen=True, slots=True)
+class MigrationPrecheck:
+    status: str
+    required_bytes: int
+    free_bytes: int
+    existing_matching_files: int
+
+    def to_dict(self) -> dict[str, object]:
+        return asdict(self)
+
+
+@dataclass(frozen=True, slots=True)
+class HistoricalDataMigrationPlan:
+    plan_id: str
+    created_at: str
+    data_root: str
+    files: tuple[MigrationFile, ...]
+    source_summary: MigrationSummary
+    precheck: MigrationPrecheck
+    schema_version: int = 1
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "schema_version": self.schema_version,
+            "plan_id": self.plan_id,
+            "created_at": self.created_at,
+            "data_root": self.data_root,
+            "source_summary": self.source_summary.to_dict(),
+            "precheck": self.precheck.to_dict(),
+            "files": [entry.to_dict() for entry in self.files],
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class MigrationFileResult:
+    target_relative_path: str
+    status: str
+    byte_size: int
+    sha256: str
+
+    def to_dict(self) -> dict[str, object]:
+        return asdict(self)
+
+
+@dataclass(frozen=True, slots=True)
+class HistoricalDataMigrationReport:
+    plan_id: str
+    status: str
+    applied: bool
+    copied_count: int
+    skipped_count: int
+    source_summary: MigrationSummary
+    destination_summary: MigrationSummary
+    mismatches: tuple[str, ...]
+    files: tuple[MigrationFileResult, ...]
+    updated_at: str
+    schema_version: int = 1
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "schema_version": self.schema_version,
+            "plan_id": self.plan_id,
+            "status": self.status,
+            "applied": self.applied,
+            "copied_count": self.copied_count,
+            "skipped_count": self.skipped_count,
+            "source_summary": self.source_summary.to_dict(),
+            "destination_summary": self.destination_summary.to_dict(),
+            "mismatches": list(self.mismatches),
+            "files": [entry.to_dict() for entry in self.files],
+            "updated_at": self.updated_at,
+        }
+
+
+class HistoricalDataMigrationPlanner:
+    """Build a verified, machine-readable copy plan without mutating sources."""
+
+    def __init__(
+        self,
+        *,
+        data_root: Path | str,
+        free_space_provider: Callable[[Path], int] | None = None,
+    ) -> None:
+        self._data_root = Path(data_root).expanduser().resolve()
+        self._free_space_provider = free_space_provider or _available_bytes
+
+    def create_plan(
+        self,
+        sources: Iterable[MigrationSource],
+        *,
+        manifest_path: Path | str | None = None,
+    ) -> HistoricalDataMigrationPlan:
+        entries: list[MigrationFile] = []
+        target_paths: set[str] = set()
+        for source in sources:
+            source_root = _resolve_source_root(source.root)
+            _reject_overlap(source_root, self._data_root)
+            for entry in _inventory_source(source, source_root):
+                if entry.target_relative_path in target_paths:
+                    raise MigrationValidationError(
+                        f"duplicate migration target: {entry.target_relative_path}"
+                    )
+                target_paths.add(entry.target_relative_path)
+                entries.append(entry)
+        entries.sort(key=lambda item: item.target_relative_path)
+
+        pending_bytes, existing = self._inspect_destinations(entries)
+        free_bytes = self._free_space_provider(_existing_parent(self._data_root))
+        if free_bytes < pending_bytes:
+            raise MigrationValidationError(
+                f"insufficient disk space: need {pending_bytes} bytes, have {free_bytes}"
+            )
+        plan = HistoricalDataMigrationPlan(
+            plan_id=uuid.uuid4().hex,
+            created_at=_now(),
+            data_root=str(self._data_root),
+            files=tuple(entries),
+            source_summary=_summarize_expected(entries),
+            precheck=MigrationPrecheck(
+                status="ready",
+                required_bytes=pending_bytes,
+                free_bytes=free_bytes,
+                existing_matching_files=existing,
+            ),
+        )
+        if manifest_path is not None:
+            atomic_write_json(manifest_path, plan.to_dict())
+        return plan
+
+    def _inspect_destinations(self, entries: list[MigrationFile]) -> tuple[int, int]:
+        pending_bytes = 0
+        existing_matching = 0
+        planned_targets = {
+            _safe_destination(self._data_root, entry.target_relative_path)
+            for entry in entries
+        }
+        for unplanned in _find_unplanned_targets(entries, self._data_root, planned_targets):
+            raise MigrationConflictError(f"unplanned target conflict: {unplanned}")
+        for entry in entries:
+            destination = _safe_destination(self._data_root, entry.target_relative_path)
+            if not destination.exists():
+                pending_bytes += entry.byte_size
+                continue
+            if not destination.is_file() or destination.is_symlink():
+                raise MigrationConflictError(f"target conflict: {destination}")
+            if destination.stat().st_size != entry.byte_size or sha256_file(destination) != entry.sha256:
+                raise MigrationConflictError(f"target conflict: {destination}")
+            existing_matching += 1
+        return pending_bytes, existing_matching
+
+
+class HistoricalDataMigrationExecutor:
+    """Execute a plan with resumable atomic copies and full reconciliation."""
+
+    def execute(
+        self,
+        plan: HistoricalDataMigrationPlan,
+        *,
+        apply: bool = False,
+        report_path: Path | str | None = None,
+    ) -> HistoricalDataMigrationReport:
+        data_root = Path(plan.data_root).resolve()
+        if not apply:
+            report = self._build_report(plan, "plan_only", False, 0, 0, (), ())
+            _write_report(report_path, report)
+            return report
+
+        copied = 0
+        skipped = 0
+        results: list[MigrationFileResult] = []
+        try:
+            self._execution_precheck(plan, data_root)
+            for entry in plan.files:
+                source = Path(entry.source_path)
+                destination = _safe_destination(data_root, entry.target_relative_path)
+                if destination.exists():
+                    skipped += 1
+                    file_status = "skipped_matching"
+                else:
+                    self._copy_file_atomic(
+                        source, destination, entry.byte_size, entry.sha256
+                    )
+                    if destination.stat().st_size != entry.byte_size or sha256_file(destination) != entry.sha256:
+                        raise MigrationExecutionError(
+                            f"destination verification failed: {entry.target_relative_path}"
+                        )
+                    copied += 1
+                    file_status = "copied"
+                results.append(
+                    MigrationFileResult(
+                        entry.target_relative_path,
+                        file_status,
+                        entry.byte_size,
+                        entry.sha256,
+                    )
+                )
+                if report_path is not None and len(results) % _PROGRESS_INTERVAL == 0:
+                    progress = self._build_report(
+                        plan, "in_progress", True, copied, skipped, tuple(results), ()
+                    )
+                    _write_report(report_path, progress)
+        except Exception as exc:
+            interrupted = self._build_report(
+                plan,
+                "interrupted",
+                True,
+                copied,
+                skipped,
+                tuple(results),
+                (str(exc),),
+            )
+            _write_report(report_path, interrupted)
+            if isinstance(exc, MigrationExecutionError):
+                raise
+            raise MigrationExecutionError(str(exc)) from exc
+
+        destination_summary, mismatches = _summarize_destinations(plan.files, data_root)
+        status = "reconciled" if not mismatches and destination_summary == plan.source_summary else "mismatch"
+        report = HistoricalDataMigrationReport(
+            plan_id=plan.plan_id,
+            status=status,
+            applied=True,
+            copied_count=copied,
+            skipped_count=skipped,
+            source_summary=plan.source_summary,
+            destination_summary=destination_summary,
+            mismatches=tuple(mismatches),
+            files=tuple(results),
+            updated_at=_now(),
+        )
+        _write_report(report_path, report)
+        if status != "reconciled":
+            raise MigrationExecutionError("destination reconciliation failed")
+        return report
+
+    def _execution_precheck(
+        self, plan: HistoricalDataMigrationPlan, data_root: Path
+    ) -> None:
+        pending_bytes = 0
+        planned_targets = {
+            _safe_destination(data_root, entry.target_relative_path)
+            for entry in plan.files
+        }
+        unplanned = _find_unplanned_targets(plan.files, data_root, planned_targets)
+        if unplanned:
+            raise MigrationExecutionError(
+                f"unplanned target conflict: {unplanned[0]}"
+            )
+        for entry in plan.files:
+            source = Path(entry.source_path)
+            if not source.is_file() or source.is_symlink():
+                raise MigrationExecutionError(f"source changed or missing: {source}")
+            if source.stat().st_size != entry.byte_size or sha256_file(source) != entry.sha256:
+                raise MigrationExecutionError(f"source changed after planning: {source}")
+            destination = _safe_destination(data_root, entry.target_relative_path)
+            if destination.exists():
+                if (
+                    not destination.is_file()
+                    or destination.is_symlink()
+                    or destination.stat().st_size != entry.byte_size
+                    or sha256_file(destination) != entry.sha256
+                ):
+                    raise MigrationExecutionError(f"target conflict: {destination}")
+            else:
+                pending_bytes += entry.byte_size
+        free_bytes = _available_bytes(_existing_parent(data_root))
+        if free_bytes < pending_bytes:
+            raise MigrationExecutionError(
+                f"insufficient disk space: need {pending_bytes} bytes, have {free_bytes}"
+            )
+
+    @staticmethod
+    def _copy_file_atomic(
+        source: Path,
+        destination: Path,
+        expected_size: int,
+        expected_sha256: str,
+    ) -> None:
+        destination.parent.mkdir(parents=True, exist_ok=True)
+        temporary = destination.with_name(f".{destination.name}.{uuid.uuid4().hex}.tmp")
+        try:
+            shutil.copy2(source, temporary)
+            if (
+                temporary.stat().st_size != expected_size
+                or sha256_file(temporary) != expected_sha256
+            ):
+                raise MigrationExecutionError(
+                    f"temporary copy verification failed: {destination}"
+                )
+            os.replace(temporary, destination)
+        finally:
+            temporary.unlink(missing_ok=True)
+
+    @staticmethod
+    def _build_report(
+        plan: HistoricalDataMigrationPlan,
+        status: str,
+        applied: bool,
+        copied: int,
+        skipped: int,
+        results: tuple[MigrationFileResult, ...],
+        mismatches: tuple[str, ...],
+    ) -> HistoricalDataMigrationReport:
+        if applied:
+            summary = _summary(
+                (
+                    result.target_relative_path,
+                    result.byte_size,
+                    result.sha256,
+                )
+                for result in results
+            )
+            discovered_mismatches = []
+        else:
+            summary = MigrationSummary(0, 0, _empty_digest())
+            discovered_mismatches = []
+        return HistoricalDataMigrationReport(
+            plan_id=plan.plan_id,
+            status=status,
+            applied=applied,
+            copied_count=copied,
+            skipped_count=skipped,
+            source_summary=plan.source_summary,
+            destination_summary=summary,
+            mismatches=tuple(mismatches or tuple(discovered_mismatches)),
+            files=results,
+            updated_at=_now(),
+        )
+
+
+def _inventory_source(source: MigrationSource, root: Path) -> list[MigrationFile]:
+    entries: list[MigrationFile] = []
+    for path in sorted(root.rglob("*")):
+        if path.is_symlink():
+            raise MigrationValidationError(f"symbolic link is not allowed: {path}")
+        if not path.is_file():
+            continue
+        relative = path.relative_to(root).as_posix()
+        before = path.stat()
+        digest = sha256_file(path)
+        after = path.stat()
+        if (before.st_size, before.st_mtime_ns) != (after.st_size, after.st_mtime_ns):
+            raise MigrationValidationError(f"source changed while planning: {path}")
+        target = "/".join(
+            ("data", source.layer.value, source.module, "legacy", source.source_id, relative)
+        )
+        entries.append(
+            MigrationFile(
+                source_id=source.source_id,
+                source_path=str(path),
+                relative_path=relative,
+                target_relative_path=target,
+                byte_size=after.st_size,
+                sha256=digest,
+            )
+        )
+    return entries
+
+
+def _summarize_expected(entries: Iterable[MigrationFile]) -> MigrationSummary:
+    rows = [
+        (entry.target_relative_path, entry.byte_size, entry.sha256)
+        for entry in entries
+    ]
+    return _summary(rows)
+
+
+def _summarize_destinations(
+    entries: Iterable[MigrationFile], data_root: Path
+) -> tuple[MigrationSummary, list[str]]:
+    rows: list[tuple[str, int, str]] = []
+    mismatches: list[str] = []
+    for entry in entries:
+        destination = _safe_destination(data_root, entry.target_relative_path)
+        if not destination.is_file() or destination.is_symlink():
+            mismatches.append(f"missing: {entry.target_relative_path}")
+            continue
+        size = destination.stat().st_size
+        digest = sha256_file(destination)
+        rows.append((entry.target_relative_path, size, digest))
+        if size != entry.byte_size or digest != entry.sha256:
+            mismatches.append(f"content mismatch: {entry.target_relative_path}")
+    return _summary(rows), mismatches
+
+
+def _find_unplanned_targets(
+    entries: Iterable[MigrationFile],
+    data_root: Path,
+    planned_targets: set[Path],
+) -> list[Path]:
+    unplanned: list[Path] = []
+    scopes = {
+        data_root.joinpath(*Path(entry.target_relative_path).parts[:5])
+        for entry in entries
+    }
+    for scope in sorted(scopes):
+        if not scope.exists():
+            continue
+        for candidate in sorted(scope.rglob("*")):
+            if candidate.is_symlink():
+                unplanned.append(candidate)
+                continue
+            if candidate.is_file() and candidate.resolve() not in planned_targets:
+                unplanned.append(candidate)
+    return unplanned
+
+
+def _summary(rows: Iterable[tuple[str, int, str]]) -> MigrationSummary:
+    ordered = sorted(rows)
+    aggregate = hashlib.sha256()
+    total_bytes = 0
+    for relative, byte_size, digest in ordered:
+        aggregate.update(f"{relative}\0{byte_size}\0{digest}\n".encode("utf-8"))
+        total_bytes += byte_size
+    return MigrationSummary(len(ordered), total_bytes, aggregate.hexdigest())
+
+
+def _empty_digest() -> str:
+    return hashlib.sha256().hexdigest()
+
+
+def _resolve_source_root(root: Path) -> Path:
+    try:
+        resolved = root.resolve(strict=True)
+    except OSError as exc:
+        raise MigrationValidationError(f"source root cannot be resolved: {root}") from exc
+    if not resolved.is_dir():
+        raise MigrationValidationError(f"source root is not a directory: {resolved}")
+    return resolved
+
+
+def _reject_overlap(source_root: Path, data_root: Path) -> None:
+    if source_root == data_root or source_root.is_relative_to(data_root) or data_root.is_relative_to(source_root):
+        raise MigrationValidationError(
+            f"source and destination overlap: {source_root} / {data_root}"
+        )
+
+
+def _safe_destination(data_root: Path, relative: str) -> Path:
+    candidate = (data_root / Path(relative)).resolve()
+    if not candidate.is_relative_to(data_root):
+        raise MigrationValidationError(f"target escapes data root: {relative}")
+    return candidate
+
+
+def _existing_parent(path: Path) -> Path:
+    candidate = path
+    while not candidate.exists():
+        parent = candidate.parent
+        if parent == candidate:
+            raise MigrationValidationError(f"cannot locate filesystem for {path}")
+        candidate = parent
+    return candidate
+
+
+def _available_bytes(path: Path) -> int:
+    return shutil.disk_usage(path).free
+
+
+def _validate_id(value: str, field: str) -> None:
+    if not isinstance(value, str) or not _SAFE_ID.fullmatch(value):
+        raise MigrationValidationError(f"unsafe {field}: {value!r}")
+
+
+def _write_report(
+    report_path: Path | str | None, report: HistoricalDataMigrationReport
+) -> None:
+    if report_path is not None:
+        atomic_write_json(report_path, report.to_dict())
+
+
+def _now() -> str:
+    return datetime.now(timezone.utc).isoformat(timespec="seconds")
diff --git a/src/gyxx_flow/migration/legacy.py b/src/gyxx_flow/migration/legacy.py
new file mode 100644
index 0000000..2b40059
--- /dev/null
+++ b/src/gyxx_flow/migration/legacy.py
@@ -0,0 +1,173 @@
+"""Build legacy command steps without executing or hard-coding old projects."""
+
+from __future__ import annotations
+
+import os
+import re
+import sys
+from pathlib import Path, PurePosixPath
+from types import MappingProxyType
+from typing import Mapping
+
+from gyxx_flow.catalog import WorkflowEntry
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.workflow.steps import CommandStep
+
+DEFAULT_PROJECT_IDS = ("content", "shop", "product", "supply")
+_PROJECT_ID = re.compile(r"^[a-z][a-z0-9_]*$")
+
+
+class LegacyConfigurationError(ValueError):
+    """Raised when a legacy project or command cannot be resolved safely."""
+
+
+class LegacyProjectRoots:
+    """Validated legacy roots supplied explicitly or through the environment."""
+
+    def __init__(self, roots: Mapping[str, Path | str]) -> None:
+        resolved: dict[str, Path] = {}
+        for project_id, configured_root in roots.items():
+            if not isinstance(project_id, str) or not _PROJECT_ID.fullmatch(project_id):
+                raise LegacyConfigurationError(f"invalid legacy project id: {project_id!r}")
+            resolved[project_id] = _resolve_project_root(project_id, configured_root)
+        self._roots = MappingProxyType(resolved)
+
+    @classmethod
+    def from_env(
+        cls,
+        *,
+        env: Mapping[str, str] | None = None,
+        project_ids: tuple[str, ...] = DEFAULT_PROJECT_IDS,
+    ) -> "LegacyProjectRoots":
+        values = os.environ if env is None else env
+        missing: list[str] = []
+        roots: dict[str, str] = {}
+        for project_id in project_ids:
+            if not _PROJECT_ID.fullmatch(project_id):
+                raise LegacyConfigurationError(f"invalid legacy project id: {project_id!r}")
+            key = f"GYXX_LEGACY_{project_id.upper()}_ROOT"
+            value = values.get(key, "").strip()
+            if not value:
+                missing.append(key)
+            else:
+                roots[project_id] = value
+        if missing:
+            raise LegacyConfigurationError(
+                "missing legacy root environment variables: " + ", ".join(missing)
+            )
+        return cls(roots)
+
+    def get(self, project_id: str) -> Path:
+        try:
+            return self._roots[project_id]
+        except KeyError as exc:
+            raise LegacyConfigurationError(
+                f"unknown legacy project: {project_id!r}"
+            ) from exc
+
+    def as_dict(self) -> dict[str, Path]:
+        return dict(self._roots)
+
+
+class LegacyCommandAdapter:
+    """Translate catalog entries to inert ``CommandStep`` definitions."""
+
+    def __init__(
+        self,
+        roots: LegacyProjectRoots,
+        *,
+        base_env: Mapping[str, str] | None = None,
+        python_executable: Path | str | None = None,
+    ) -> None:
+        self._roots = roots
+        self._base_env = dict(os.environ if base_env is None else base_env)
+        configured_python = sys.executable if python_executable is None else python_executable
+        self._python_executable = str(configured_python)
+        _validate_text(self._python_executable, field="python_executable")
+
+    def build(self, entry: WorkflowEntry, *, context: RunContext) -> CommandStep:
+        if not entry.source_project:
+            raise LegacyConfigurationError(
+                f"workflow has no source-project provenance: {entry.workflow_id}"
+            )
+        root = self._roots.get(entry.source_project)
+        if entry.workflow_id != context.workflow_id:
+            raise LegacyConfigurationError(
+                f"workflow entry {entry.workflow_id!r} does not match "
+                f"run context {context.workflow_id!r}"
+            )
+        if entry.trigger == "unavailable":
+            raise LegacyConfigurationError(
+                f"legacy workflow is unavailable: {entry.workflow_id}"
+            )
+
+        executable = _resolve_entry(root, entry.entry)
+        arguments = tuple(entry.args)
+        for argument in arguments:
+            _validate_text(argument, field="legacy argument")
+        argv = (
+            (self._python_executable, str(executable), *arguments)
+            if executable.suffix.casefold() == ".py"
+            else (str(executable), *arguments)
+        )
+        env = dict(self._base_env)
+        env.update(
+            {
+                "GYXX_WORKFLOW_ID": context.workflow_id,
+                "GYXX_RUN_ID": context.run_id,
+                "GYXX_BUSINESS_DATE": context.business_date.isoformat(),
+                "GYXX_SHADOW": str(context.shadow).lower(),
+                "GYXX_LEGACY_PROJECT_ROOT": str(root),
+            }
+        )
+        try:
+            return CommandStep(argv=argv, cwd=root, env=env)
+        except ValueError as exc:
+            raise LegacyConfigurationError(str(exc)) from exc
+
+
+def _resolve_project_root(project_id: str, configured_root: Path | str) -> Path:
+    try:
+        raw_root = Path(configured_root).expanduser()
+        if not raw_root.exists():
+            raise LegacyConfigurationError(
+                f"legacy project root does not exist for {project_id!r}: {raw_root}"
+            )
+        root = raw_root.resolve(strict=True)
+    except (OSError, TypeError, ValueError) as exc:
+        if isinstance(exc, LegacyConfigurationError):
+            raise
+        raise LegacyConfigurationError(
+            f"cannot resolve legacy project root for {project_id!r}: {exc}"
+        ) from exc
+    if not root.is_dir():
+        raise LegacyConfigurationError(
+            f"legacy project root is not a directory for {project_id!r}: {root}"
+        )
+    return root
+
+
+def _resolve_entry(root: Path, configured_entry: str) -> Path:
+    if not isinstance(configured_entry, str) or not configured_entry:
+        raise LegacyConfigurationError("legacy entry must be a non-empty relative path")
+    if "\\" in configured_entry or ":" in configured_entry or "\x00" in configured_entry:
+        raise LegacyConfigurationError("legacy entry must use a safe relative path")
+    relative = PurePosixPath(configured_entry)
+    if relative.is_absolute() or ".." in relative.parts:
+        raise LegacyConfigurationError("legacy entry cannot escape its project root")
+    try:
+        candidate = (root / Path(*relative.parts)).resolve(strict=True)
+    except OSError as exc:
+        raise LegacyConfigurationError(
+            f"legacy entry does not exist or cannot be resolved: {configured_entry}"
+        ) from exc
+    if not candidate.is_relative_to(root):
+        raise LegacyConfigurationError("legacy entry cannot escape its project root")
+    if not candidate.is_file():
+        raise LegacyConfigurationError(f"legacy entry is not a file: {configured_entry}")
+    return candidate
+
+
+def _validate_text(value: object, *, field: str) -> None:
+    if not isinstance(value, str) or not value or "\x00" in value:
+        raise LegacyConfigurationError(f"{field} must be a non-empty safe string")
diff --git a/src/gyxx_flow/migration/shadow.py b/src/gyxx_flow/migration/shadow.py
new file mode 100644
index 0000000..222decc
--- /dev/null
+++ b/src/gyxx_flow/migration/shadow.py
@@ -0,0 +1,489 @@
+"""Privacy-conscious comparisons for shadow workflow outputs."""
+
+from __future__ import annotations
+
+import copy
+import csv
+import hashlib
+import hmac
+import json
+import math
+from dataclasses import dataclass, field
+from datetime import datetime, timezone
+from decimal import Decimal, InvalidOperation
+from pathlib import Path
+from types import MappingProxyType
+from typing import Any, Mapping, Sequence
+
+from gyxx_flow.core.artifacts import atomic_write_json
+
+_DEFAULT_HASH_SECRET = b"gyxx-flow-public-comparison-v1"
+_STRUCTURED_SUFFIXES = {
+    ".csv": "csv",
+    ".json": "json",
+    ".jsonl": "jsonl",
+    ".ndjson": "jsonl",
+}
+
+
+@dataclass(frozen=True, slots=True)
+class MetricTolerance:
+    """Allowed absolute and relative drift for one numeric metric."""
+
+    absolute: float = 0.0
+    relative: float = 0.0
+
+    def __post_init__(self) -> None:
+        _validate_tolerance(self.absolute, "absolute")
+        _validate_tolerance(self.relative, "relative")
+
+
+@dataclass(frozen=True, slots=True)
+class ComparisonSpec:
+    """Declarative schema for comparing two structured result sets."""
+
+    primary_key: tuple[str, ...]
+    metrics: Mapping[str, MetricTolerance] = field(default_factory=dict)
+    error_fields: tuple[str, ...] = ()
+    safe_fields: tuple[str, ...] = ()
+    difference_limit: int = 1000
+    hash_secret: str | bytes | None = field(default=None, repr=False, compare=False)
+
+    def __post_init__(self) -> None:
+        primary_key = _validate_fields(self.primary_key, "primary_key", required=True)
+        error_fields = _validate_fields(self.error_fields, "error_fields")
+        safe_fields = _validate_fields(self.safe_fields, "safe_fields")
+        if not isinstance(self.difference_limit, int) or isinstance(
+            self.difference_limit, bool
+        ) or self.difference_limit < 1:
+            raise ValueError("difference_limit must be a positive integer")
+        if not isinstance(self.metrics, Mapping):
+            raise ValueError("metrics must be a mapping")
+        metrics: dict[str, MetricTolerance] = {}
+        for name, tolerance in self.metrics.items():
+            _validate_field_name(name, "metric")
+            if not isinstance(tolerance, MetricTolerance):
+                raise ValueError(f"metric {name!r} must declare MetricTolerance")
+            metrics[name] = tolerance
+        if self.hash_secret is not None and not isinstance(
+            self.hash_secret, (str, bytes)
+        ):
+            raise ValueError("hash_secret must be text or bytes")
+        if self.hash_secret == "" or self.hash_secret == b"":
+            raise ValueError("hash_secret cannot be empty")
+        object.__setattr__(self, "primary_key", primary_key)
+        object.__setattr__(self, "error_fields", error_fields)
+        object.__setattr__(self, "safe_fields", safe_fields)
+        object.__setattr__(self, "metrics", MappingProxyType(metrics))
+
+    def hash_key(self) -> bytes:
+        if self.hash_secret is None:
+            return _DEFAULT_HASH_SECRET
+        if isinstance(self.hash_secret, str):
+            return self.hash_secret.encode("utf-8")
+        return self.hash_secret
+
+
+@dataclass(frozen=True, slots=True)
+class ComparisonReport:
+    """Immutable facade over a JSON-serializable comparison report."""
+
+    _payload: Mapping[str, Any] = field(repr=False)
+
+    def to_dict(self) -> dict[str, Any]:
+        return copy.deepcopy(dict(self._payload))
+
+
+def compare_structured_files(
+    baseline_path: Path | str,
+    candidate_path: Path | str,
+    *,
+    spec: ComparisonSpec,
+    generated_at: datetime | None = None,
+) -> ComparisonReport:
+    """Compare JSON, JSONL, or CSV results without emitting source rows."""
+
+    if not isinstance(spec, ComparisonSpec):
+        raise TypeError("spec must be ComparisonSpec")
+    baseline_format, baseline_rows = _load_records(Path(baseline_path))
+    candidate_format, candidate_rows = _load_records(Path(candidate_path))
+    baseline = _index_rows(baseline_rows, spec.primary_key, side="baseline")
+    candidate = _index_rows(candidate_rows, spec.primary_key, side="candidate")
+
+    baseline_keys = set(baseline)
+    candidate_keys = set(candidate)
+    common_keys = baseline_keys & candidate_keys
+    missing_keys = baseline_keys - candidate_keys
+    extra_keys = candidate_keys - baseline_keys
+    secret = spec.hash_key()
+
+    metric_reports = _compare_metrics(
+        baseline, candidate, common_keys, spec=spec, secret=secret
+    )
+    baseline_errors = _collect_errors(baseline_rows, spec.error_fields)
+    candidate_errors = _collect_errors(candidate_rows, spec.error_fields)
+    error_report = _compare_errors(
+        baseline_errors,
+        candidate_errors,
+        secret=secret,
+        limit=spec.difference_limit,
+    )
+    metric_mismatches = sum(
+        metric["mismatch_count"] for metric in metric_reports.values()
+    )
+    error_differences = (
+        error_report["only_baseline_total"] + error_report["only_candidate_total"]
+    )
+    has_differences = bool(
+        missing_keys or extra_keys or metric_mismatches or error_differences
+    )
+    timestamp = generated_at or datetime.now(timezone.utc)
+    if timestamp.tzinfo is None or timestamp.utcoffset() is None:
+        raise ValueError("generated_at must be timezone-aware")
+
+    missing_identifiers = _sample_identifiers(
+        missing_keys, baseline, spec=spec, secret=secret
+    )
+    extra_identifiers = _sample_identifiers(
+        extra_keys, candidate, spec=spec, secret=secret
+    )
+    payload: dict[str, Any] = {
+        "schema_version": 1,
+        "status": "mismatch" if has_differences else "match",
+        "generated_at": timestamp.isoformat(timespec="seconds"),
+        "sources": {
+            "baseline_format": baseline_format,
+            "candidate_format": candidate_format,
+        },
+        "spec": {
+            "primary_key": list(spec.primary_key),
+            "metrics": {
+                name: {
+                    "absolute_tolerance": tolerance.absolute,
+                    "relative_tolerance": tolerance.relative,
+                }
+                for name, tolerance in spec.metrics.items()
+            },
+            "error_fields": list(spec.error_fields),
+            "safe_fields": list(spec.safe_fields),
+            "difference_limit": spec.difference_limit,
+        },
+        "summary": {
+            "baseline_rows": len(baseline_rows),
+            "candidate_rows": len(candidate_rows),
+            "common_keys": len(common_keys),
+            "missing_keys": len(missing_keys),
+            "extra_keys": len(extra_keys),
+            "metric_mismatches": metric_mismatches,
+            "error_set_differences": error_differences,
+        },
+        "keys": {
+            "missing_total": len(missing_keys),
+            "missing_truncated": len(missing_keys) > spec.difference_limit,
+            "missing": missing_identifiers,
+            "extra_total": len(extra_keys),
+            "extra_truncated": len(extra_keys) > spec.difference_limit,
+            "extra": extra_identifiers,
+        },
+        "metrics": metric_reports,
+        "errors": error_report,
+    }
+    return ComparisonReport(payload)
+
+
+def write_comparison_report(
+    path: Path | str, report: ComparisonReport
+) -> Path:
+    """Atomically persist a machine-readable comparison report."""
+
+    if not isinstance(report, ComparisonReport):
+        raise TypeError("report must be ComparisonReport")
+    return atomic_write_json(path, report.to_dict())
+
+
+def _load_records(path: Path) -> tuple[str, list[dict[str, Any]]]:
+    data_format = _STRUCTURED_SUFFIXES.get(path.suffix.casefold())
+    if data_format is None:
+        raise ValueError(f"unsupported structured data format: {path.suffix or '<none>'}")
+    if data_format == "csv":
+        return data_format, _load_csv(path)
+    if data_format == "jsonl":
+        return data_format, _load_jsonl(path)
+    return data_format, _load_json(path)
+
+
+def _load_csv(path: Path) -> list[dict[str, Any]]:
+    with path.open("r", encoding="utf-8-sig", newline="") as handle:
+        reader = csv.DictReader(handle)
+        rows: list[dict[str, Any]] = []
+        for line_number, row in enumerate(reader, start=2):
+            if None in row:
+                raise ValueError(f"CSV record has extra columns at line {line_number}")
+            rows.append(dict(row))
+        return rows
+
+
+def _load_jsonl(path: Path) -> list[dict[str, Any]]:
+    rows: list[dict[str, Any]] = []
+    with path.open("r", encoding="utf-8-sig") as handle:
+        for line_number, line in enumerate(handle, start=1):
+            if not line.strip():
+                continue
+            try:
+                row = json.loads(line)
+            except json.JSONDecodeError as exc:
+                raise ValueError(f"invalid JSONL at line {line_number}") from exc
+            rows.append(_require_record(row, location=f"JSONL line {line_number}"))
+    return rows
+
+
+def _load_json(path: Path) -> list[dict[str, Any]]:
+    try:
+        payload = json.loads(path.read_text(encoding="utf-8-sig"))
+    except json.JSONDecodeError as exc:
+        raise ValueError("invalid JSON document") from exc
+    if isinstance(payload, dict) and set(payload) == {"records"}:
+        payload = payload["records"]
+    if not isinstance(payload, list):
+        raise ValueError("JSON document must be an array or a records envelope")
+    return [
+        _require_record(row, location=f"JSON record {index}")
+        for index, row in enumerate(payload, start=1)
+    ]
+
+
+def _require_record(value: object, *, location: str) -> dict[str, Any]:
+    if not isinstance(value, dict) or any(not isinstance(key, str) for key in value):
+        raise ValueError(f"record must be a JSON object at {location}")
+    return dict(value)
+
+
+def _index_rows(
+    rows: Sequence[Mapping[str, Any]], primary_key: tuple[str, ...], *, side: str
+) -> dict[tuple[str, ...], Mapping[str, Any]]:
+    indexed: dict[tuple[str, ...], Mapping[str, Any]] = {}
+    for row in rows:
+        key = _primary_key(row, primary_key, side=side)
+        if key in indexed:
+            raise ValueError(f"duplicate primary key in {side} data")
+        indexed[key] = row
+    return indexed
+
+
+def _primary_key(
+    row: Mapping[str, Any], fields: tuple[str, ...], *, side: str
+) -> tuple[str, ...]:
+    values: list[str] = []
+    for field_name in fields:
+        if field_name not in row or row[field_name] is None or row[field_name] == "":
+            raise ValueError(f"missing primary key field {field_name!r} in {side} data")
+        value = row[field_name]
+        if isinstance(value, (dict, list, tuple, set)):
+            raise ValueError(f"primary key field {field_name!r} must be scalar")
+        if isinstance(value, bool):
+            values.append("true" if value else "false")
+        else:
+            values.append(str(value))
+    return tuple(values)
+
+
+def _compare_metrics(
+    baseline: Mapping[tuple[str, ...], Mapping[str, Any]],
+    candidate: Mapping[tuple[str, ...], Mapping[str, Any]],
+    common_keys: set[tuple[str, ...]],
+    *,
+    spec: ComparisonSpec,
+    secret: bytes,
+) -> dict[str, dict[str, Any]]:
+    reports: dict[str, dict[str, Any]] = {}
+    for field_name, tolerance in spec.metrics.items():
+        mismatches: list[tuple[str, ...]] = []
+        max_absolute = Decimal(0)
+        max_relative = Decimal(0)
+        absolute_tolerance = Decimal(str(tolerance.absolute))
+        relative_tolerance = Decimal(str(tolerance.relative))
+        for key in common_keys:
+            baseline_value = _numeric_value(baseline[key], field_name, "baseline")
+            candidate_value = _numeric_value(candidate[key], field_name, "candidate")
+            absolute_delta = abs(candidate_value - baseline_value)
+            scale = max(abs(candidate_value), abs(baseline_value))
+            relative_delta = absolute_delta / scale if scale else Decimal(0)
+            max_absolute = max(max_absolute, absolute_delta)
+            max_relative = max(max_relative, relative_delta)
+            allowed_delta = max(absolute_tolerance, relative_tolerance * scale)
+            if absolute_delta > allowed_delta:
+                mismatches.append(key)
+        reports[field_name] = {
+            "compared": len(common_keys),
+            "mismatch_count": len(mismatches),
+            "max_absolute_delta": float(max_absolute),
+            "max_relative_delta": float(max_relative),
+            "mismatches_truncated": len(mismatches) > spec.difference_limit,
+            "mismatches": _sample_identifiers(
+                mismatches, baseline, spec=spec, secret=secret
+            ),
+        }
+    return reports
+
+
+def _numeric_value(
+    row: Mapping[str, Any], field_name: str, side: str
+) -> Decimal:
+    if field_name not in row:
+        raise ValueError(f"missing metric field {field_name!r} in {side} record")
+    value = row[field_name]
+    if isinstance(value, bool) or value is None:
+        raise ValueError(f"metric field {field_name!r} must be finite numeric")
+    try:
+        numeric = Decimal(str(value).strip())
+    except (InvalidOperation, ValueError) as exc:
+        raise ValueError(f"metric field {field_name!r} must be finite numeric") from exc
+    if not numeric.is_finite():
+        raise ValueError(f"metric field {field_name!r} must be finite numeric")
+    try:
+        machine_value = float(numeric)
+    except (OverflowError, ValueError) as exc:
+        raise ValueError(f"metric field {field_name!r} exceeds report range") from exc
+    if not math.isfinite(machine_value):
+        raise ValueError(f"metric field {field_name!r} exceeds report range")
+    return numeric
+
+
+def _sample_identifiers(
+    keys: Sequence[tuple[str, ...]] | set[tuple[str, ...]],
+    rows: Mapping[tuple[str, ...], Mapping[str, Any]],
+    *,
+    spec: ComparisonSpec,
+    secret: bytes,
+) -> list[dict[str, Any]]:
+    identifiers = [
+        _identifier(key, rows[key], spec=spec, secret=secret) for key in keys
+    ]
+    identifiers.sort(key=lambda item: item["key_hash"])
+    return identifiers[: spec.difference_limit]
+
+
+def _identifier(
+    key: tuple[str, ...],
+    row: Mapping[str, Any],
+    *,
+    spec: ComparisonSpec,
+    secret: bytes,
+) -> dict[str, Any]:
+    canonical = json.dumps(key, ensure_ascii=False, separators=(",", ":"))
+    identifier: dict[str, Any] = {
+        "key_hash": _secure_hash(canonical, secret=secret, domain="primary-key")
+    }
+    if spec.safe_fields:
+        safe_values = {
+            name: _safe_value(row[name]) for name in spec.safe_fields if name in row
+        }
+        if safe_values:
+            identifier["safe_fields"] = safe_values
+    return identifier
+
+
+def _safe_value(value: Any) -> Any:
+    if value is None or isinstance(value, (str, int, float, bool)):
+        if isinstance(value, float) and not math.isfinite(value):
+            raise ValueError("safe field must contain a finite JSON scalar")
+        return value
+    raise ValueError("safe field must contain a JSON scalar")
+
+
+def _collect_errors(
+    rows: Sequence[Mapping[str, Any]], fields: tuple[str, ...]
+) -> set[str]:
+    errors: set[str] = set()
+    for row in rows:
+        for field_name in fields:
+            if field_name in row:
+                errors.update(_error_tokens(row[field_name]))
+    return errors
+
+
+def _error_tokens(value: Any) -> set[str]:
+    if value is None or value == "":
+        return set()
+    if isinstance(value, str):
+        stripped = value.strip()
+        if not stripped:
+            return set()
+        if stripped[:1] in {"[", "{"}:
+            try:
+                return _error_tokens(json.loads(stripped))
+            except json.JSONDecodeError:
+                pass
+        return {_canonical_json(stripped)}
+    if isinstance(value, (list, tuple, set)):
+        tokens: set[str] = set()
+        for item in value:
+            tokens.update(_error_tokens(item))
+        return tokens
+    return {_canonical_json(value)}
+
+
+def _compare_errors(
+    baseline: set[str], candidate: set[str], *, secret: bytes, limit: int
+) -> dict[str, Any]:
+    only_baseline = sorted(
+        _secure_hash(value, secret=secret, domain="error")
+        for value in baseline - candidate
+    )
+    only_candidate = sorted(
+        _secure_hash(value, secret=secret, domain="error")
+        for value in candidate - baseline
+    )
+    return {
+        "baseline_unique_count": len(baseline),
+        "candidate_unique_count": len(candidate),
+        "only_baseline_total": len(only_baseline),
+        "only_baseline_truncated": len(only_baseline) > limit,
+        "only_baseline_hashes": only_baseline[:limit],
+        "only_candidate_total": len(only_candidate),
+        "only_candidate_truncated": len(only_candidate) > limit,
+        "only_candidate_hashes": only_candidate[:limit],
+    }
+
+
+def _secure_hash(value: str, *, secret: bytes, domain: str) -> str:
+    message = f"gyxx-shadow-{domain}-v1\0{value}".encode("utf-8")
+    return hmac.new(secret, message, hashlib.sha256).hexdigest()
+
+
+def _canonical_json(value: Any) -> str:
+    try:
+        return json.dumps(
+            value, ensure_ascii=False, sort_keys=True, separators=(",", ":")
+        )
+    except (TypeError, ValueError) as exc:
+        raise ValueError("error fields must contain JSON-compatible values") from exc
+
+
+def _validate_fields(
+    values: object, label: str, *, required: bool = False
+) -> tuple[str, ...]:
+    if not isinstance(values, tuple):
+        raise ValueError(f"{label} must be a tuple of field names")
+    if required and not values:
+        raise ValueError(f"{label} must contain at least one field")
+    for value in values:
+        _validate_field_name(value, label)
+    if len(set(values)) != len(values):
+        raise ValueError(f"{label} cannot contain duplicate fields")
+    return values
+
+
+def _validate_field_name(value: object, label: str) -> None:
+    if not isinstance(value, str) or not value.strip() or "\x00" in value:
+        raise ValueError(f"{label} contains an invalid field name")
+
+
+def _validate_tolerance(value: object, label: str) -> None:
+    if (
+        isinstance(value, bool)
+        or not isinstance(value, (int, float))
+        or not math.isfinite(value)
+        or value < 0
+    ):
+        raise ValueError(f"{label} tolerance must be a finite non-negative number")
diff --git a/src/gyxx_flow/modules/__init__.py b/src/gyxx_flow/modules/__init__.py
new file mode 100644
index 0000000..c241343
--- /dev/null
+++ b/src/gyxx_flow/modules/__init__.py
@@ -0,0 +1,77 @@
+"""Business-module contracts, registry, and default composition root."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from .content_marketing import ContentMarketingModule
+from .contracts import BusinessModule
+from .product_commerce import ProductCommerceModule
+from .registry import ModuleRegistry
+from .shop_intelligence import ShopIntelligenceModule
+from .supply_chain import SupplyChainModule
+
+EXPECTED_MODULE_IDS = (
+    "content_marketing",
+    "product_commerce",
+    "shop_intelligence",
+    "supply_chain",
+)
+
+
+def create_default_registry(
+    *,
+    catalog: Any | None = None,
+    shop_command_factory: Any | None = None,
+    supply_command_factory: Any | None = None,
+    content_command_factory: Any | None = None,
+    product_command_factory: Any | None = None,
+) -> ModuleRegistry:
+    """Compose built-in modules, registering only explicitly supplied migrations."""
+
+    shop_module = (
+        ShopIntelligenceModule()
+        if catalog is None
+        else ShopIntelligenceModule.from_catalog(
+            catalog,
+            command_factory=shop_command_factory,
+        )
+    )
+    supply_module = (
+        SupplyChainModule()
+        if catalog is None
+        else SupplyChainModule.from_catalog(
+            catalog, command_factory=supply_command_factory
+        )
+    )
+    content_module = (
+        ContentMarketingModule()
+        if catalog is None
+        else ContentMarketingModule.from_catalog(
+            catalog, command_factory=content_command_factory
+        )
+    )
+    product_module = (
+        ProductCommerceModule()
+        if catalog is None
+        else ProductCommerceModule.from_catalog(
+            catalog, command_factory=product_command_factory
+        )
+    )
+
+    return ModuleRegistry(
+        (
+            content_module,
+            product_module,
+            shop_module,
+            supply_module,
+        )
+    )
+
+
+__all__ = [
+    "BusinessModule",
+    "EXPECTED_MODULE_IDS",
+    "ModuleRegistry",
+    "create_default_registry",
+]
diff --git a/src/gyxx_flow/modules/content_marketing/__init__.py b/src/gyxx_flow/modules/content_marketing/__init__.py
new file mode 100644
index 0000000..fc33775
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/__init__.py
@@ -0,0 +1,103 @@
+"""Content metrics, creator, comment, report, and login workflows."""
+
+from __future__ import annotations
+
+from collections.abc import Iterable
+from typing import Any
+
+from gyxx_flow.adapters.native import DeferredModuleCommandStep, ModuleCommandFactory
+from gyxx_flow.workflow import StepDefinition, WorkflowDefinition
+
+CONTENT_SCHEDULED_WORKFLOW_IDS = (
+    "content.metrics.daily",
+    "content.marketing_report.daily",
+    "content.relogin.weekly",
+    "content.self_operated.weekly",
+    "content.creator_report.monthly",
+    "content.summary.monthly",
+    "content.cooperations.daily",
+    "content.comments.weekly",
+    "content.summary.weekly",
+)
+CONTENT_MANUAL_WORKFLOW_IDS = (
+    "content.mapping.refresh",
+    "content.retry_failed",
+    "content.metrics.backfill",
+)
+CONTENT_WORKFLOW_IDS = (*CONTENT_SCHEDULED_WORKFLOW_IDS, *CONTENT_MANUAL_WORKFLOW_IDS)
+CONTENT_TIMEOUT_SECONDS = 4 * 60 * 60
+CONTENT_RESOURCE = "module:content_marketing"
+_RELOGIN_WORKFLOW_ID = "content.relogin.weekly"
+
+
+class ContentMarketingModule:
+    """Content-marketing business module composition boundary."""
+
+    module_id = "content_marketing"
+
+    def __init__(self, definitions: Iterable[WorkflowDefinition] = ()) -> None:
+        self._definitions = tuple(definitions)
+
+    @classmethod
+    def from_catalog(
+        cls,
+        catalog: Any,
+        *,
+        command_factory: ModuleCommandFactory | None = None,
+    ) -> "ContentMarketingModule":
+        """Build the scheduled content workflows from validated catalog metadata.
+
+        Each launcher resolves from source owned by this installed module.
+        Registration remains inert and dry-run never starts a subprocess.
+        """
+
+        entries = {
+            entry.workflow_id: entry
+            for entry in catalog.workflows
+            if entry.module == cls.module_id and entry.trigger in {"scheduled", "manual"}
+        }
+        expected = set(CONTENT_WORKFLOW_IDS)
+        if set(entries) != expected:
+            raise ValueError(
+                "content catalog workflows do not match the migration contract: "
+                f"expected={sorted(expected)}, actual={sorted(entries)}"
+            )
+        definitions: list[WorkflowDefinition] = []
+        for workflow_id in CONTENT_WORKFLOW_IDS:
+            entry = entries[workflow_id]
+            action = (
+                DeferredModuleCommandStep(entry)
+                if command_factory is None
+                else DeferredModuleCommandStep(entry, command_factory=command_factory)
+            )
+            definitions.append(
+                WorkflowDefinition(
+                    workflow_id,
+                    (
+                        StepDefinition(
+                            "module_run",
+                            action,
+                            timeout_seconds=CONTENT_TIMEOUT_SECONDS,
+                            max_attempts=1,
+                            resources=(CONTENT_RESOURCE,),
+                            production_sink=True,
+                            official_notification=workflow_id
+                            == _RELOGIN_WORKFLOW_ID,
+                        ),
+                    ),
+                )
+            )
+        return cls(definitions)
+
+    def workflow_definitions(self) -> tuple[WorkflowDefinition, ...]:
+        return self._definitions
+
+
+__all__ = [
+    "CONTENT_WORKFLOW_IDS",
+    "CONTENT_MANUAL_WORKFLOW_IDS",
+    "CONTENT_SCHEDULED_WORKFLOW_IDS",
+    "CONTENT_RESOURCE",
+    "CONTENT_TIMEOUT_SECONDS",
+    "ContentMarketingModule",
+]
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/bilibili_comment_scraper.py b/src/gyxx_flow/modules/content_marketing/runtime/bilibili_comment_scraper.py
new file mode 100644
index 0000000..4d0ff01
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/bilibili_comment_scraper.py
@@ -0,0 +1,500 @@
+import argparse
+import csv
+import json
+import re
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+from typing import Any
+from urllib.parse import parse_qs, urlencode, urlsplit, urlunsplit
+
+from playwright.sync_api import Page
+from scrapling.fetchers import DynamicSession
+
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+STATE_DIR = PATHS.state_root
+NOTES_DIR = PATHS.raw_root / "notes" / "bilibili"
+COOKIE_FILE = PATHS.browser_cookie_file
+PROFILE_DIR = PATHS.browser_profile_dir
+
+BV_RE = re.compile(r"BV[0-9A-Za-z]{10}")
+AV_RE = re.compile(r"(?:av|aid=)(\d+)", re.IGNORECASE)
+
+
+def log(message: str) -> None:
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {message}", flush=True)
+
+
+def load_cookies() -> list[dict[str, Any]]:
+    if not COOKIE_FILE.exists():
+        return []
+    try:
+        data = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+        return data if isinstance(data, list) else []
+    except json.JSONDecodeError:
+        log(f"Cookie file is invalid, ignoring: {COOKIE_FILE}")
+        return []
+
+
+def restore_cookies(page: Page) -> None:
+    cookies = load_cookies()
+    if not cookies:
+        return
+    try:
+        page.context.add_cookies(cookies)
+        log(f"Loaded cookies: {COOKIE_FILE}")
+    except Exception as exc:
+        log(f"Failed to load cookies, will continue with browser profile: {exc}")
+
+
+def save_state(page: Page) -> None:
+    STATE_DIR.mkdir(parents=True, exist_ok=True)
+    COOKIE_FILE.write_text(
+        json.dumps(page.context.cookies(), ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    PATHS.browser_storage_state_file.write_text(
+        json.dumps(page.context.storage_state(), ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    log(f"Saved cookies: {COOKIE_FILE}")
+
+
+def is_logged_in(page: Page) -> bool:
+    """B 站登录态靠 SESSDATA cookie,有它才能调评论 API 拿全量。
+    没有 SESSDATA 时 API 只返回热门 3-5 条。
+    """
+    for c in page.context.cookies():
+        if c.get("name") == "SESSDATA" and c.get("value"):
+            return True
+    return False
+
+
+def maybe_wait_for_login(page: Page, login_timeout: int) -> None:
+    """检测登录态,未登录就等用户扫码。B 站登录页 URL 含 passport.bilibili.com/login。"""
+    if is_logged_in(page):
+        return
+    log("B 站未登录 (无 SESSDATA cookie)。请在弹出的浏览器窗口扫码登录。")
+    # 跳到登录页,确保用户能看到二维码
+    try:
+        page.goto("https://passport.bilibili.com/login", wait_until="domcontentloaded", timeout=30000)
+    except Exception as exc:
+        log(f"  跳登录页失败 (继续等当前页扫码): {exc}")
+    deadline = time.time() + login_timeout
+    while time.time() < deadline:
+        if is_logged_in(page):
+            log("登录成功,SESSDATA 已写入")
+            save_state(page)
+            return
+        page.wait_for_timeout(2000)
+    log(f"  [WARN] 等待 {login_timeout}s 仍未登录,继续以游客态采集 (评论只会拿到热门 3-5 条)")
+
+
+def safe_filename(value: str) -> str:
+    value = re.sub(r"[<>:\"/\\\\|?*\\s]+", "_", value).strip("_")
+    return value[:90] or "bilibili_comments"
+
+
+def extract_bvid(url: str) -> str:
+    match = BV_RE.search(url)
+    return match.group(0) if match else ""
+
+
+def extract_aid(url: str) -> str:
+    match = AV_RE.search(url)
+    return match.group(1) if match else ""
+
+
+def ts_to_iso(value: Any) -> str:
+    if not isinstance(value, (int, float)) or not value:
+        return ""
+    try:
+        return datetime.fromtimestamp(value).isoformat(timespec="seconds")
+    except Exception:
+        return str(value)
+
+
+def with_query_params(url: str, updates: dict[str, Any]) -> str:
+    parts = urlsplit(url)
+    params = dict((key, values[-1]) for key, values in parse_qs(parts.query, keep_blank_values=True).items())
+    for key, value in updates.items():
+        params[key] = str(value)
+    return urlunsplit((parts.scheme, parts.netloc, parts.path, urlencode(params), parts.fragment))
+
+
+def parse_reply(reply: dict[str, Any], source_url: str, level: str, parent_rpid: str = "") -> dict[str, Any]:
+    member = reply.get("member") or {}
+    content = reply.get("content") or {}
+    return {
+        "source_url": source_url,
+        "level": level,
+        "rpid": str(reply.get("rpid") or ""),
+        "parent_rpid": str(parent_rpid or reply.get("parent") or ""),
+        "root_rpid": str(reply.get("root") or ""),
+        "mid": str(member.get("mid") or reply.get("mid") or ""),
+        "nickname": member.get("uname") or "",
+        "sex": member.get("sex") or "",
+        "message": content.get("message") or "",
+        "like_count": reply.get("like") or 0,
+        "reply_count": reply.get("rcount") or 0,
+        "location": reply.get("reply_control", {}).get("location") or "",
+        "created_at": ts_to_iso(reply.get("ctime")),
+        "raw_ctime": reply.get("ctime") or "",
+    }
+
+
+def merge_reply(row: dict[str, Any], comments_by_id: dict[str, dict[str, Any]]) -> None:
+    key = row.get("rpid") or f"{row.get('level')}:{len(comments_by_id)}"
+    comments_by_id[str(key)] = row
+
+
+def collect_main_payload(payload: dict[str, Any], source_url: str, comments_by_id: dict[str, dict[str, Any]]) -> None:
+    data = payload.get("data") if isinstance(payload.get("data"), dict) else payload
+    replies = data.get("replies") or []
+    if not isinstance(replies, list):
+        return
+    for reply in replies:
+        if not isinstance(reply, dict):
+            continue
+        row = parse_reply(reply, source_url, "comment")
+        merge_reply(row, comments_by_id)
+        child_replies = reply.get("replies") or []
+        if isinstance(child_replies, list):
+            for child in child_replies:
+                if isinstance(child, dict):
+                    child_row = parse_reply(child, source_url, "reply", row.get("rpid") or "")
+                    merge_reply(child_row, comments_by_id)
+
+
+def collect_reply_payload(payload: dict[str, Any], source_url: str, parent_rpid: str, comments_by_id: dict[str, dict[str, Any]]) -> None:
+    data = payload.get("data") if isinstance(payload.get("data"), dict) else payload
+    replies = data.get("replies") or []
+    if not isinstance(replies, list):
+        return
+    for reply in replies:
+        if isinstance(reply, dict):
+            row = parse_reply(reply, source_url, "reply", parent_rpid)
+            merge_reply(row, comments_by_id)
+
+
+def api_get(page: Page, url: str) -> dict[str, Any]:
+    return page.evaluate(
+        """
+        async (url) => {
+          const res = await fetch(url, {
+            credentials: 'include',
+            headers: { 'accept': 'application/json, text/plain, */*' }
+          });
+          return await res.json();
+        }
+        """,
+        url,
+    )
+
+
+def resolve_video_info(page: Page, url: str) -> dict[str, Any]:
+    bvid = extract_bvid(url)
+    aid = extract_aid(url)
+    if not bvid and not aid:
+        page.goto(url, wait_until="domcontentloaded", timeout=90000)
+        page.wait_for_timeout(3000)
+        bvid = extract_bvid(page.url) or extract_bvid(page.locator("body").inner_text(timeout=5000))
+        aid = extract_aid(page.url)
+
+    params = {}
+    if bvid:
+        params["bvid"] = bvid
+    elif aid:
+        params["aid"] = aid
+    else:
+        raise RuntimeError("Could not find BV or AV id from URL.")
+
+    api_url = "https://api.bilibili.com/x/web-interface/view?" + urlencode(params)
+    payload = api_get(page, api_url)
+    if payload.get("code") != 0:
+        raise RuntimeError(f"Failed to resolve video info: {payload}")
+
+    data = payload.get("data") or {}
+    stat = data.get("stat") or {}
+    return {
+        "aid": str(data.get("aid") or aid),
+        "bvid": data.get("bvid") or bvid,
+        "title": data.get("title") or "",
+        "owner_mid": str((data.get("owner") or {}).get("mid") or ""),
+        "owner_name": (data.get("owner") or {}).get("name") or "",
+        "note_metrics": {
+            "like_count": stat.get("like", ""),
+            "favorite_count": stat.get("favorite", ""),
+            "share_count": stat.get("share", ""),
+            "comment_count": stat.get("reply", ""),
+        },
+    }
+
+
+def fetch_main_comments(
+    page: Page,
+    source_url: str,
+    oid: str,
+    comments_by_id: dict[str, dict[str, Any]],
+    max_pages: int,
+    delay: float,
+) -> dict[str, Any]:
+    stats = {"api_hits": 0, "api_total": None, "last_page": 0}
+    page_num = 1
+    page_size = 20
+
+    while page_num <= max_pages:
+        api_url = (
+            "https://api.bilibili.com/x/v2/reply?"
+            + urlencode({"type": 1, "oid": oid, "pn": page_num, "ps": page_size, "sort": 2})
+        )
+        payload = api_get(page, api_url)
+        stats["api_hits"] += 1
+        if payload.get("code") != 0:
+            log(f"Main comment API returned code={payload.get('code')}: {payload.get('message')}")
+            break
+
+        data = payload.get("data") or {}
+        page_info = data.get("page") or {}
+        stats["api_total"] = page_info.get("count", stats.get("api_total"))
+        stats["last_page"] = page_num
+
+        before = len(comments_by_id)
+        collect_main_payload(payload, source_url, comments_by_id)
+        loaded = len(comments_by_id) - before
+        log(f"Main page {page_num}: +{loaded}, total {len(comments_by_id)}")
+
+        replies = data.get("replies") or []
+        if not replies:
+            break
+        if page_info.get("count") and len([r for r in comments_by_id.values() if r.get("level") == "comment"]) >= int(page_info["count"]):
+            break
+
+        page_num += 1
+        if delay:
+            page.wait_for_timeout(int(delay * 1000))
+
+    return stats
+
+
+def fetch_sub_replies(
+    page: Page,
+    source_url: str,
+    oid: str,
+    comments_by_id: dict[str, dict[str, Any]],
+    delay: float,
+    max_reply_pages: int,
+) -> int:
+    parents = [
+        row for row in list(comments_by_id.values())
+        if row.get("level") == "comment" and int(row.get("reply_count") or 0) > 0
+    ]
+    if not parents:
+        return 0
+
+    log(f"Fetching sub replies for {len(parents)} comments.")
+    merged = 0
+    for parent in parents:
+        root = parent.get("rpid")
+        if not root:
+            continue
+        pn = 1
+        ps = 20
+        while pn <= max_reply_pages:
+            api_url = (
+                "https://api.bilibili.com/x/v2/reply/reply?"
+                + urlencode({"type": 1, "oid": oid, "root": root, "pn": pn, "ps": ps})
+            )
+            payload = api_get(page, api_url)
+            if payload.get("code") != 0:
+                break
+            data = payload.get("data") or {}
+            before = len(comments_by_id)
+            collect_reply_payload(payload, source_url, str(root), comments_by_id)
+            merged += max(0, len(comments_by_id) - before)
+
+            replies = data.get("replies") or []
+            page_info = data.get("page") or {}
+            if not replies:
+                break
+            if page_info.get("count") and pn * ps >= int(page_info["count"]):
+                break
+            pn += 1
+            if delay:
+                page.wait_for_timeout(int(delay * 1000))
+    log(f"Sub replies merged: {merged}; total {len(comments_by_id)}")
+    return merged
+
+
+def scrape_comments(
+    url: str,
+    max_pages: int,
+    delay: float,
+    headless: bool,
+    max_reply_pages: int = 200,
+    login_timeout: int = 0,
+) -> tuple[list[dict[str, Any]], dict[str, Any]]:
+    result: dict[str, Any] = {"comments": [], "stats": {}, "video": {}}
+
+    def action(page: Page) -> None:
+        page.goto(url, wait_until="domcontentloaded", timeout=90000)
+        page.wait_for_timeout(2500)
+        save_state(page)
+
+        # 登录检测:未登录时评论 API 只返回热门 3-5 条
+        if login_timeout > 0:
+            maybe_wait_for_login(page, login_timeout)
+            # 登录后回到视频页(可能在登录页)
+            if not page.url.startswith("https://www.bilibili.com/video/"):
+                page.goto(url, wait_until="domcontentloaded", timeout=90000)
+                page.wait_for_timeout(2000)
+
+        video = resolve_video_info(page, url)
+        oid = video["aid"]
+        comments_by_id: dict[str, dict[str, Any]] = {}
+        stats = fetch_main_comments(page, url, oid, comments_by_id, max_pages=max_pages, delay=delay)
+        sub_count = fetch_sub_replies(page, url, oid, comments_by_id, delay=delay, max_reply_pages=max_reply_pages)
+        save_state(page)
+
+        stats["sub_replies_merged"] = sub_count
+        stats["max_reply_pages"] = max_reply_pages
+        stats["top_level_count"] = len([row for row in comments_by_id.values() if row.get("level") == "comment"])
+        stats["reply_count"] = len([row for row in comments_by_id.values() if row.get("level") == "reply"])
+        stats["note_metrics"] = video.get("note_metrics", {})
+        stats["logged_in"] = is_logged_in(page)
+        result["comments"] = list(comments_by_id.values())
+        result["stats"] = stats
+        result["video"] = video
+        result["final_url"] = page.url
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+    with DynamicSession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=str(PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=90000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=restore_cookies,
+        max_pages=1,
+    ) as session:
+        session.fetch(url, page_action=action, wait=1000)
+
+    return result["comments"], result
+
+
+def write_outputs(url: str, comments: list[dict[str, Any]], result: dict[str, Any]) -> tuple[Path, Path]:
+    NOTES_DIR.mkdir(parents=True, exist_ok=True)
+    bvid = result.get("video", {}).get("bvid") or extract_bvid(url) or safe_filename(url)
+    json_path = NOTES_DIR / f"bilibili_comments_{bvid}.json"
+    csv_path = NOTES_DIR / f"bilibili_comments_{bvid}.csv"
+    if not comments and json_path.exists() and csv_path.exists():
+        log(f"No comments scraped; keeping existing output files: {json_path}, {csv_path}")
+        return json_path, csv_path
+
+    payload = {
+        "source_url": url,
+        "title": result.get("video", {}).get("title", ""),
+        "final_url": result.get("final_url", ""),
+        "video": result.get("video", {}),
+        "scraped_at": datetime.now().isoformat(timespec="seconds"),
+        "count": len(comments),
+        "stats": result.get("stats", {}),
+        "comments": comments,
+    }
+    json_path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    fieldnames = [
+        "source_url",
+        "level",
+        "rpid",
+        "parent_rpid",
+        "root_rpid",
+        "mid",
+        "nickname",
+        "sex",
+        "message",
+        "like_count",
+        "reply_count",
+        "location",
+        "created_at",
+        "raw_ctime",
+    ]
+    with csv_path.open("w", newline="", encoding="utf-8-sig") as file:
+        writer = csv.DictWriter(file, fieldnames=fieldnames)
+        writer.writeheader()
+        writer.writerows(comments)
+
+    return json_path, csv_path
+
+
+def login_only(login_timeout: int) -> int:
+    """只触发登录(headed),登录完成立刻退出。给 relogin_bilibili.py 用。"""
+    def action(page: Page) -> None:
+        page.goto("https://passport.bilibili.com/login", wait_until="domcontentloaded", timeout=30000)
+        page.wait_for_timeout(1500)
+        maybe_wait_for_login(page, login_timeout)
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+    with DynamicSession(
+        headless=False,
+        real_chrome=True,
+        user_data_dir=str(PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=90000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=restore_cookies,
+        max_pages=1,
+    ) as session:
+        session.fetch("https://passport.bilibili.com/login", page_action=action, wait=1000)
+    log("[login-only] 完成")
+    return 0
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Scrape Bilibili video comments with a real browser and API.")
+    parser.add_argument("url", help="Bilibili video URL (传 dummy 'login' 配合 --login-only 也可)")
+    parser.add_argument("--max-reply-pages", type=int, default=200, help="max sub-reply pages per top-level comment")
+    parser.add_argument("--max-pages", type=int, default=200, help="最多抓取一级评论页数")
+    parser.add_argument("--delay", type=float, default=0.25, help="接口请求间隔秒数")
+    parser.add_argument("--headless", action="store_true", help="无头模式")
+    parser.add_argument("--login-timeout", type=int, default=0,
+                        help="未登录时等待扫码秒数(>0 启用登录检测;默认 0 沿用游客态)")
+    parser.add_argument("--login-only", action="store_true",
+                        help="只触发登录(headed)然后退出,给 relogin_bilibili 用")
+    args = parser.parse_args()
+
+    if args.login_only:
+        return login_only(args.login_timeout if args.login_timeout > 0 else 300)
+
+    comments, result = scrape_comments(
+        args.url, args.max_pages, args.delay, args.headless,
+        args.max_reply_pages, args.login_timeout,
+    )
+    json_path, csv_path = write_outputs(args.url, comments, result)
+    log(f"Final URL: {result.get('final_url', '')}")
+    log(f"Video: {result.get('video', {})}")
+    log(f"Stats: {result.get('stats', {})}")
+    log(f"Scraped {len(comments)} comments")
+    log(f"JSON: {json_path}")
+    log(f"CSV:  {csv_path}")
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/bilibili_scraper.py b/src/gyxx_flow/modules/content_marketing/runtime/bilibili_scraper.py
new file mode 100644
index 0000000..d74608b
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/bilibili_scraper.py
@@ -0,0 +1,854 @@
+"""
+B 站抓取脚本: 多维表格驱动,按"发布时间到今天的天数差"自动选 7/14/21/28/月槽位
+
+核心规则:
+- 槽位由 (今天 - 发布时间) 的天数差决定:
+    0-7 天  → 7天曝光量
+    8-14 天 → 14天曝光量
+    15-21 天 → 21天曝光量
+    22-28 天 → 28天曝光量
+    29+ 天  → 月底曝光量
+- 同槽覆盖语义: 同一槽多次跑会更新为最新播放量
+- 抓取: B 站公开 API (api.bilibili.com/x/web-interface/view) 拿 stat.view
+        不用登录,不用 cookie
+- 字段 ID 全部从 款式_多维表格_对照.json 的 field_map 动态读取
+- 平台枚举: B站 / b站 / B 站 都接受 (大小写无关)
+
+用法:
+  # 跑全 24 款式 (默认)
+  python bilibili_scraper.py
+
+  # 试跑不写
+  python bilibili_scraper.py --dry-run
+
+  # 只跑某个款式
+  python bilibili_scraper.py --style 1
+
+  # 只跑某几条
+  python bilibili_scraper.py --record recXXXXX
+
+  # 模拟任意日期 (调试用)
+  python bilibili_scraper.py --today 2025-08-15 --dry-run
+
+  # 防风控:每条请求间隔 0.5s
+  python bilibili_scraper.py --delay 0.5
+
+  # 首次回填: 所有 B 站记录全部填 7天曝光量 (忽略发布时间)
+  python bilibili_scraper.py --first-run
+
+  # 重置运行状态
+  python bilibili_scraper.py --reset-state
+"""
+import argparse
+import json
+import os
+import re
+import subprocess
+import sys
+import time
+from datetime import datetime, date
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import (
+    PATHS,
+    resolve_layer_output,
+)
+
+import requests
+
+from gyxx_flow.modules.content_marketing.runtime.collection_completeness import (
+    add_repeatable_style_argument,
+    atomic_write_json,
+    merge_global_summaries,
+    select_requested_styles,
+)
+
+BASE_DIR = PATHS.module_root
+DEFAULT_DATA_DIR = PATHS.normalized_root
+V2_DIR = PATHS.normalized_root / "v2_results"
+BILIBILI_CHECKPOINT_DIR = PATHS.state_root / "checkpoints" / "bilibili"
+MAPPING_PATH = PATHS.normalized_root / "mappings" / "款式_多维表格_对照.json"
+STATE_FILENAME = "weekly_run_state.json"
+
+LARK_CLI = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK_CLI):
+    # Scheduled tasks run as SYSTEM; %APPDATA% is not Administrator's dir
+    LARK_CLI = "lark-cli.cmd"
+
+# 字段 ID 全部从 data/config/款式_多维表格_对照.json 的 field_map 动态读
+# 下面是 logical key(各款式表结构一致)
+KEY_PLATFORM = "platform"        # 投放平台 (select)
+KEY_URL = "note_url"             # 发布笔记链接 (text)
+KEY_TITLE = "note_title"         # 发布笔记标题 (text)
+KEY_CREATOR = "creator_name"     # 达人名称 (text)
+KEY_CREATOR_ID = "creator_id"    # 达人id (text,选填)
+KEY_PUBTIME = "publish_time"     # 发布时间 (datetime)
+KEY_MONTH = "month"              # 月份 (select)
+KEY_PARENT = "parent_record"     # 父记录 (link,自关联)
+KEY_WEEK = "publish_week"        # 周发布 (select)
+
+# 5 档曝光量槽位: 7/14/21/28 天 + 月底(第 5 周用) - logical name -> 天数
+SLOTS = [
+    ("read_count_7d", 7),
+    ("read_count_14d", 14),
+    ("read_count_21d", 21),
+    ("read_count_28d", 28),
+    ("month_end", "月底"),
+]
+
+# 月份枚举(选项里已有)
+MONTH_OPTIONS = ["6月", "7月", "8月", "9月", "10月"]
+
+# B 站 API 相关
+_UA = ("Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
+       "(KHTML, like Gecko) Chrome/126.0.0.0 Safari/537.36")
+_BV_RE = re.compile(r"BV[0-9A-Za-z]{10}")
+_AV_RE = re.compile(r"av(\d+)", re.IGNORECASE)
+_B23_RE = re.compile(r"https?://b23\.tv/\S+", re.IGNORECASE)
+_MD_LINK_RE = re.compile(r"\[([^\]]+)\]\(([^)]+)\)")
+_URL_RE = re.compile(r"https?://[^\s)>\]]+")
+
+
+def log(msg: str) -> None:
+    safe = str(msg).encode("gbk", errors="replace").decode("gbk", errors="replace")
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {safe}", flush=True)
+
+
+def finalize_bili_summary(summary: dict, *, dry_run: bool = False) -> dict:
+    """Attach one terminal state per Bilibili record and exact completeness counters."""
+    out = dict(summary)
+    unique: list[dict] = []
+    positions: dict[str, int] = {}
+    duplicates = 0
+    for raw in out.get("details") or []:
+        if not isinstance(raw, dict):
+            continue
+        row = dict(raw)
+        rid = str(row.get("record_id") or "")
+        if rid and rid in positions:
+            unique[positions[rid]] = row
+            duplicates += 1
+        else:
+            if rid:
+                positions[rid] = len(unique)
+            unique.append(row)
+    total = int(out.get("total_b_records") or 0)
+    success = sum(row.get("status") == "success" for row in unique)
+    blocked = sum(row.get("status") == "blocked_input" for row in unique)
+    retryable = sum(row.get("status") == "retryable_failure" for row in unique)
+    write_failures = sum(row.get("status") == "write_failure" for row in unique)
+    missing = max(0, total - len(unique))
+    unresolved = retryable + write_failures + missing
+    top_level_error = bool(out.get("error"))
+    if top_level_error:
+        retryable = max(retryable, 1)
+        unresolved = max(unresolved, 1)
+    out.update({
+        "details": unique,
+        "updated": success,
+        "skipped": blocked + retryable + write_failures,
+        "success": success,
+        "blocked_input": blocked,
+        "retryable_failures": retryable,
+        "write_failures": write_failures,
+        "missing_results": missing,
+        "duplicate_results": duplicates,
+        "unresolved": unresolved,
+        "complete": (
+            not top_level_error
+            and unresolved == 0
+            and duplicates == 0
+            and len(unique) == total
+        ),
+        "dry_run": bool(dry_run),
+    })
+    return out
+
+
+def merge_bili_summary(existing: dict | None, partial: dict, *, dry_run: bool = False) -> dict:
+    """Merge a --record retry without deleting previously successful rows."""
+    if not existing:
+        return finalize_bili_summary(partial, dry_run=dry_run)
+    combined = dict(existing)
+    for key, value in partial.items():
+        if key != "details":
+            combined[key] = value
+    rows: list[dict] = []
+    positions: dict[str, int] = {}
+    for source in (existing.get("details") or [], partial.get("details") or []):
+        for raw in source:
+            if not isinstance(raw, dict):
+                continue
+            row = dict(raw)
+            rid = str(row.get("record_id") or "")
+            if rid and rid in positions:
+                rows[positions[rid]] = row
+            else:
+                if rid:
+                    positions[rid] = len(rows)
+                rows.append(row)
+    combined["details"] = rows
+    combined["total_b_records"] = max(
+        int(existing.get("total_b_records") or 0),
+        int(partial.get("total_b_records") or 0),
+        len(rows),
+    )
+    return finalize_bili_summary(combined, dry_run=dry_run)
+
+
+# ============================================================
+# 飞书多维表格
+# ============================================================
+def call_lark_json(args: list[str]) -> dict:
+    cmd = [LARK_CLI] + args
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(
+        cmd, cwd=PATHS.tmp_root, capture_output=True, env=env,
+        encoding="utf-8", errors="replace",
+    )
+    text = proc.stdout
+    start = text.find("{")
+    if start < 0:
+        return {"ok": False, "error": f"no json: {text[:300]}"}
+    try:
+        return json.loads(text[start:], strict=False)
+    except json.JSONDecodeError as e:
+        return {"ok": False, "error": f"json decode: {e}"}
+
+
+def load_mapping(data_dir: Path, self_operated: bool = False) -> dict:
+    # 动态:从「合作达人」/「自营达人」多维表格地址表读取款式→各表地址,实时拉字段重建对照表
+    from gyxx_flow.modules.content_marketing.runtime import feishu_mapping
+    return feishu_mapping.load_mapping(None, self_operated=self_operated)
+
+
+def list_all_records(base_token: str, table_id: str) -> list[dict]:
+    """拉一张表所有记录,自动分页"""
+    all_records = []
+    offset = 0
+    while True:
+        resp = call_lark_json([
+            "base", "+record-list",
+            "--base-token", base_token,
+            "--table-id", table_id,
+            "--as", "user",
+            "--format", "json",
+            "--limit", "200",
+            "--offset", str(offset),
+        ])
+        if not resp.get("ok"):
+            raise RuntimeError(f"record-list 失败: {resp.get('error')}")
+        d = resp["data"]
+        ids = d["field_id_list"]
+        rids = d["record_id_list"]
+        rows = d["data"]
+        for i, row in enumerate(rows):
+            rec = {"record_id": rids[i]}
+            for j, fid in enumerate(ids):
+                rec[fid] = row[j] if j < len(row) else None
+            all_records.append(rec)
+        offset += len(rows)
+        if not d.get("has_more", False) or len(rows) == 0:
+            break
+    return all_records
+
+
+def _write_payload_to_cwd(payload: dict, name: str = "_bili_payload.json") -> str:
+    """把 payload 写到当前工作目录,返回相对路径名(lark-cli --json @file 必须是 cwd 相对路径名,不是绝对路径)"""
+    payload_file = PATHS.tmp_root / name
+    payload_file.parent.mkdir(parents=True, exist_ok=True)
+    payload_file.write_text(json.dumps(payload, ensure_ascii=False), encoding="utf-8")
+    return name  # 只返回文件名,而不是绝对路径
+
+
+def write_record(base_token: str, table_id: str, record_id: str,
+                 fields: dict, dry_run: bool) -> bool:
+    """更新一条已有记录"""
+    if dry_run:
+        log(f"    [DRY-RUN] 写入: {json.dumps(fields, ensure_ascii=False)[:200]}")
+        return True
+    pf = _write_payload_to_cwd(fields)
+    cmd = [LARK_CLI, "base", "+record-upsert",
+           "--base-token", base_token,
+           "--table-id", table_id,
+           "--record-id", record_id,
+           "--as", "user",
+           "--format", "json",
+           "--json", f"@{pf}"]
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(
+        cmd, cwd=PATHS.tmp_root, capture_output=True, env=env,
+        encoding="utf-8", errors="replace",
+    )
+    text = proc.stdout
+    if proc.stderr:
+        # lark-cli 在 stderr 也会输出,真实响应可能在 stderr
+        if "ok" in proc.stderr and (text == "" or not text.lstrip().startswith("{")):
+            text = proc.stderr
+    start = text.find("{")
+    if start < 0:
+        log(f"    [ERROR] 写记录失败: no json. stdout={proc.stdout[:200]} stderr={proc.stderr[:200]}")
+        return False
+    try:
+        resp = json.loads(text[start:], strict=False)
+    except json.JSONDecodeError as e:
+        log(f"    [ERROR] 写记录失败: json decode {e}. text={text[:200]}")
+        return False
+    if not resp.get("ok"):
+        log(f"    [ERROR] 写记录失败: {resp.get('error')}")
+        return False
+    return True
+
+
+def create_record(base_token: str, table_id: str,
+                  fields: dict, dry_run: bool) -> str | None:
+    """创建新记录,返回 record_id"""
+    if dry_run:
+        log(f"    [DRY-RUN] 新建: {json.dumps(fields, ensure_ascii=False)[:200]}")
+        return "rec_dryrun"
+    pf = _write_payload_to_cwd(fields)
+    cmd = [LARK_CLI, "base", "+record-upsert",
+           "--base-token", base_token,
+           "--table-id", table_id,
+           "--as", "user",
+           "--format", "json",
+           "--json", f"@{pf}"]
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(
+        cmd, cwd=PATHS.tmp_root, capture_output=True, env=env,
+        encoding="utf-8", errors="replace",
+    )
+    text = proc.stdout
+    if proc.stderr:
+        if "ok" in proc.stderr and (text == "" or not text.lstrip().startswith("{")):
+            text = proc.stderr
+    start = text.find("{")
+    if start < 0:
+        log(f"    [ERROR] 创建记录失败: no json. stderr={proc.stderr[:200]}")
+        return None
+    try:
+        resp = json.loads(text[start:], strict=False)
+    except json.JSONDecodeError as e:
+        log(f"    [ERROR] 创建记录失败: json decode {e}. text={text[:200]}")
+        return None
+    if not resp.get("ok"):
+        log(f"    [ERROR] 创建记录失败: {resp.get('error')}")
+        return None
+    rids = resp.get("data", {}).get("record", {}).get("record_id_list", [])
+    return rids[0] if rids else None
+
+
+# ============================================================
+# B 站播放量抓取
+# ============================================================
+def extract_url(text) -> str | None:
+    """从 note_url 字段里抽真实 URL,处理 Markdown / 纯文本 / 短链"""
+    if not text:
+        return None
+    if isinstance(text, list):
+        text = text[0] if text else None
+        if not text:
+            return None
+    s = str(text).strip()
+    m = _MD_LINK_RE.search(s)
+    if m:
+        return m.group(2).strip()
+    m = _URL_RE.search(s)
+    if m:
+        return m.group(0).strip()
+    return None
+
+
+def resolve_b23(url: str, session: requests.Session) -> str | None:
+    try:
+        r = session.get(url, allow_redirects=True, timeout=10, headers={"User-Agent": _UA})
+        return r.url
+    except Exception as exc:
+        log(f"    [WARN] b23 解析失败: {exc}")
+        return None
+
+
+def fetch_play_count(url: str, session: requests.Session) -> int | None:
+    real = url
+    if _B23_RE.match(url):
+        resolved = resolve_b23(url, session)
+        if not resolved:
+            return None
+        real = resolved
+    bv = _BV_RE.search(real)
+    av = _AV_RE.search(real)
+    if not bv and not av:
+        return None
+    params = {"bvid": bv.group(0)} if bv else {"aid": av.group(1)}
+    try:
+        r = session.get("https://api.bilibili.com/x/web-interface/view",
+                        params=params, timeout=10,
+                        headers={"User-Agent": _UA, "Referer": "https://www.bilibili.com/"})
+        j = r.json()
+        if j.get("code") != 0:
+            return None
+        v = j.get("data", {}).get("stat", {}).get("view")
+        return int(v) if v is not None else None
+    except Exception as exc:
+        log(f"    [WARN] B 站 API: {exc}")
+        return None
+
+
+# ============================================================
+# 记录关系 + 槽位推进
+# ============================================================
+def get_parent_id(rec: dict, fid: str) -> str | None:
+    if not fid:
+        return None
+    p = rec.get(fid)
+    if isinstance(p, list) and p:
+        item = p[0]
+        if isinstance(item, dict):
+            return item.get("id")
+        if isinstance(item, str):
+            return item
+    return None
+
+
+def get_month(rec: dict, fid: str) -> str | None:
+    if not fid:
+        return None
+    m = rec.get(fid)
+    if isinstance(m, list) and m:
+        return m[0]
+    if isinstance(m, str):
+        return m
+    return None
+
+
+def month_to_num(m: str) -> int:
+    m = re.match(r"(\d+)月", m or "")
+    return int(m.group(1)) if m else 0
+
+
+def num_to_month(n: int) -> str:
+    return f"{n}月"
+
+
+def get_week_label(rec: dict, fid: str) -> str | None:
+    """从记录的"周发布"字段提取文本(e.g. '7月第二周')"""
+    if not fid:
+        return None
+    w = rec.get(fid)
+    if isinstance(w, list) and w:
+        return w[0]
+    if isinstance(w, str):
+        return w
+    return None
+
+
+def compute_week_slot(today: date) -> tuple[str, str, int | str]:
+    """
+    根据今天日期,返回 (week_label, slot_logical_name, slot_key):
+      - 1-7 号 → "X月第一周" → 7天
+      - 8-14 号 → "X月第二周" → 14天
+      - 15-21 号 → "X月第三周" → 21天
+      - 22-28 号 → "X月第四周" → 28天
+      - 29-31 号 → "X月第五周" → 月底曝光量
+    注: 返回的是 logical name (如 'read_count_7d'),需要从 fmap 查 field_id
+    """
+    day = today.day
+    month = today.month
+    month_label = num_to_month(month)
+    if day <= 7:
+        return f"{month_label}第一周", SLOTS[0][0], SLOTS[0][1]
+    elif day <= 14:
+        return f"{month_label}第二周", SLOTS[1][0], SLOTS[1][1]
+    elif day <= 21:
+        return f"{month_label}第三周", SLOTS[2][0], SLOTS[2][1]
+    elif day <= 28:
+        return f"{month_label}第四周", SLOTS[3][0], SLOTS[3][1]
+    else:
+        return f"{month_label}第五周", SLOTS[4][0], SLOTS[4][1]
+
+
+def next_slot_in_record(rec: dict, slot_fids: list[str]) -> tuple[str | None, int | str | None]:
+    """
+    返回 (slot_fid, slot_key) — 该记录下一个该填的槽
+    判定: None / 空字符串 / 0 都视为"未填"
+    5 槽都填了返回 (None, None)
+    slot_fids: 5 档槽位的实际 field_id 列表(顺序对应 SLOTS)
+    """
+    for fid, days in zip(slot_fids, [s[1] for s in SLOTS]):
+        v = rec.get(fid)
+        if v is None:
+            return fid, days
+        if isinstance(v, str) and v.strip() == "":
+            return fid, days
+        try:
+            if float(v) == 0:
+                return fid, days
+        except (TypeError, ValueError):
+            return fid, days
+    return None, None
+
+
+def get_slot_fid(fmap: dict, logical_name: str) -> str | None:
+    """从 fmap 查某个 logical 槽位(如 'read_count_7d')的实际 field_id"""
+    return fmap.get(logical_name, {}).get("field_id")
+
+
+def build_chains(records: list[dict], fid_parent: str, fid_month: str, fid_pubtime: str) -> dict[str, list[dict]]:
+    """
+    把记录组织成"链": key=链根 rid(顶层父记录 rid,即父字段为空的)
+    value=按月份升序的子记录列表(含根本身)
+    """
+    by_id = {r["record_id"]: r for r in records}
+    roots: dict[str, list[dict]] = {}
+
+    for r in records:
+        cur = r
+        while True:
+            pid = get_parent_id(cur, fid=fid_parent)
+            if not pid:
+                root_id = cur["record_id"]
+                break
+            if pid not in by_id:
+                root_id = cur["record_id"]  # 父不在表内,自己当根
+                break
+            cur = by_id[pid]
+        roots.setdefault(root_id, []).append(r)
+
+    for k in roots:
+        roots[k].sort(key=lambda x: (
+            month_to_num(get_month(x, fid=fid_month) or "0月"),
+            x.get(fid_pubtime) if fid_pubtime else 0
+        ))
+    return roots
+
+
+# ============================================================
+# 状态持久化(运行历史,可重置)
+# ============================================================
+def load_state(data_dir: Path) -> dict:
+    del data_dir
+    p = BILIBILI_CHECKPOINT_DIR / STATE_FILENAME
+    if not p.exists():
+        return {}
+    try:
+        return json.loads(p.read_text(encoding="utf-8"))
+    except Exception:
+        return {}
+
+
+def save_state(data_dir: Path, state: dict) -> None:
+    del data_dir
+    p = BILIBILI_CHECKPOINT_DIR / STATE_FILENAME
+    p.parent.mkdir(parents=True, exist_ok=True)
+    p.parent.mkdir(parents=True, exist_ok=True)
+    atomic_write_json(p, state)
+
+
+# ============================================================
+# 单款式处理
+# ============================================================
+def process_style(style: dict, only_record_ids: set[str] | None,
+                  dry_run: bool, delay: float, session: requests.Session,
+                  state: dict, force_today: date | None,
+                  first_run: bool = False) -> dict:
+    base_token = style["base_token"]
+    table_id = style["table_id"]
+    fmap = style.get("field_map", {})
+
+    # 从本款 field_map 解析出实际字段 ID
+    fid_platform = fmap.get(KEY_PLATFORM, {}).get("field_id", "")
+    fid_url = fmap.get(KEY_URL, {}).get("field_id", "")
+    fid_creator = fmap.get(KEY_CREATOR, {}).get("field_id", "")
+    fid_pubtime = fmap.get(KEY_PUBTIME, {}).get("field_id", "")
+    fid_month = fmap.get(KEY_MONTH, {}).get("field_id", "")
+    fid_parent = fmap.get(KEY_PARENT, {}).get("field_id", "")
+    fid_week = fmap.get(KEY_WEEK, {}).get("field_id", "")
+
+    # 5 档槽位的 field_id 列表
+    slot_fids = [get_slot_fid(fmap, ln) or "" for ln, _ in SLOTS]
+
+    log(f"\n========== [{style['index']:02d}] {style['name']} (B 站) ==========")
+    log(f"  字段映射: platform={fid_platform} url={fid_url} creator={fid_creator or '(无)'} pubtime={fid_pubtime or '(无)'}")
+    missing_mapping: list[str] = []
+    if not fid_platform:
+        missing_mapping.append("platform")
+    if not fid_url:
+        missing_mapping.append("note_url")
+    if first_run:
+        if not slot_fids[0]:
+            missing_mapping.append("read_count_7d")
+    else:
+        if not fid_pubtime:
+            missing_mapping.append("publish_time")
+        missing_mapping.extend(
+            logical
+            for field_id, (logical, _label) in zip(slot_fids, SLOTS)
+            if not field_id
+        )
+    if missing_mapping:
+        error = f"missing field mapping: {','.join(missing_mapping)}"
+        log(f"  [ERROR] {error},跳过该款式")
+        return {"style": style["name"], "index": style["index"],
+                "total_b_records": 0, "updated": 0, "skipped": 0, "details": [],
+                "run_started_at": time.time(),
+                "error": error, "unresolved": 1,
+                "retryable_failures": 1, "complete": False}
+
+    all_records = list_all_records(base_token, table_id)
+    # 只看 B 站记录 (大小写无关)
+    def is_bili_platform(v) -> bool:
+        if not v:
+            return False
+        s = str(v).lower()
+        return "b站" in s or "b 站" in s
+
+    b_records = [r for r in all_records
+                 if r.get(fid_platform) and is_bili_platform(r.get(fid_platform))]
+    log(f"  全表 {len(all_records)} 条,B 站源记录 {len(b_records)} 条")
+
+    if only_record_ids:
+        b_records = [r for r in b_records if r["record_id"] in only_record_ids]
+        log(f"  --record 过滤后,B 站源记录剩 {len(b_records)} 条")
+
+    # 业务触发条件:只有填了发布笔记链接的记录才进入采集清单。
+    # 无链接通常表示尚未发布，不是 blocked_input，也不进入完整度分母。
+    source_b_records = len(b_records)
+    b_records = [r for r in b_records if extract_url(r.get(fid_url))]
+    skipped_no_url = source_b_records - len(b_records)
+    log(f"  发布链接过滤: 待采 {len(b_records)} 条,未触发 {skipped_no_url} 条")
+
+    today = force_today or date.today()
+    style_state = state.setdefault(style["name"], {})
+    summary = {
+        "style": style["name"], "index": style["index"],
+        "run_started_at": time.time(),
+        "source_b_records": source_b_records,
+        "skipped_no_url": skipped_no_url,
+        "total_b_records": len(b_records), "updated": 0, "skipped": 0,
+        "details": [],
+    }
+
+    for rec in b_records:
+        rid = rec["record_id"]
+        creator = rec.get(fid_creator) or "?"
+        pubtime_raw = rec.get(fid_pubtime)
+        url = extract_url(rec.get(fid_url))
+        if not url:  # 防御性保护；正常情况下已在待采清单生成前过滤。
+            continue
+
+        # 解析发布时间
+        pub_date = parse_pub_date(pubtime_raw)
+        if pub_date is None and not first_run:
+            log(f"    [SKIP] {rid[:10]}.. {creator} 发布时间为空,无法判断槽位")
+            summary["skipped"] += 1
+            summary["details"].append({
+                "record_id": rid, "creator": creator, "url": url,
+                "status": "blocked_input", "matched": False,
+                "reason": "publish_time_missing", "ok": False,
+            })
+            continue
+
+        # 算槽位: (今天 - 发布日) 天数差
+        if first_run:
+            target_slot_fid, target_slot_key = slot_fids[0] or None, SLOTS[0][1] if slot_fids[0] else None
+        else:
+            days_elapsed = (today - pub_date).days
+            target_slot_fid, target_slot_key = pick_slot_by_days(days_elapsed, slot_fids)
+        if target_slot_fid is None:
+            if first_run:
+                log(f"    [SKIP] {rid[:10]}.. {creator} 7天曝光量字段缺失")
+            else:
+                log(f"    [SKIP] {rid[:10]}.. {creator} 对应槽位字段缺失或发布时间 {pub_date} 在 {today} 之后")
+            summary["skipped"] += 1
+            summary["details"].append({
+                "record_id": rid, "creator": creator, "url": url,
+                "publish_date": str(pub_date) if pub_date else None,
+                "status": "blocked_input", "matched": False,
+                "reason": "slot_missing_or_future_publish_time", "ok": False,
+            })
+            continue
+
+        # 抓播放量
+        play = fetch_play_count(url, session)
+        if play is None:
+            log(f"    [SKIP] {rid[:10]}.. {creator} 抓不到播放量")
+            summary["skipped"] += 1
+            summary["details"].append({
+                "record_id": rid, "creator": creator, "url": url,
+                "publish_date": str(pub_date) if pub_date else None,
+                "status": "retryable_failure", "matched": False,
+                "reason": "play_count_fetch_failed", "ok": False,
+            })
+            continue
+
+        # 覆盖写入 (同槽覆盖语义,数字更新)
+        action = "first_run_fill" if first_run else "update_slot"
+        log(f"  {rid[:10]}.. {creator} 发布={pub_date} 距今={(today - pub_date).days if pub_date else '-'}天"
+            f" → 写 {target_slot_key}天曝光量={play}")
+        ok = write_record(base_token, table_id, rid,
+                          {target_slot_fid: play}, dry_run)
+        if ok:
+            summary["updated"] += 1
+            style_state[rid] = {"last_run": str(today), "slot": target_slot_key}
+        summary["details"].append({
+            "record_id": rid, "creator": creator,
+            "publish_date": str(pub_date) if pub_date else None,
+            "days_elapsed": (today - pub_date).days if pub_date else None,
+            "slot": target_slot_key, "play_count": play,
+            "action": action, "ok": ok,
+            "matched": True,
+            "write_ok": ok if not dry_run else None,
+            "status": "success" if ok else "write_failure",
+            **({} if ok else {"reason": "write_back_failed"}),
+        })
+        if delay > 0:
+            time.sleep(delay)
+
+    state[style["name"]] = style_state
+    return finalize_bili_summary(summary, dry_run=dry_run)
+
+
+# 槽位:天数差 → (slot_fid, slot_key)
+# 0-7天→7天, 8-14→14天, 15-21→21天, 22-28→28天, 29+→月底
+def pick_slot_by_days(days: int, slot_fids: list[str]) -> tuple[str | None, int | str | None]:
+    if days < 0:
+        return None, None
+    if days <= 7:
+        fid = slot_fids[0]
+    elif days <= 14:
+        fid = slot_fids[1]
+    elif days <= 21:
+        fid = slot_fids[2]
+    elif days <= 28:
+        fid = slot_fids[3]
+    else:
+        fid = slot_fids[4]
+    if not fid:
+        return None, None
+    return fid, SLOTS[slot_fids.index(fid)][1]
+
+
+def parse_pub_date(v) -> date | None:
+    """发布时间字段可能是毫秒时间戳 (int) 或 'YYYY-MM-DD HH:MM:SS' 字符串"""
+    if v is None:
+        return None
+    if isinstance(v, (int, float)):
+        # 毫秒时间戳
+        if v > 1e12:
+            v = v / 1000
+        try:
+            return datetime.fromtimestamp(v).date()
+        except Exception:
+            return None
+    if isinstance(v, str):
+        s = v.strip()
+        if not s:
+            return None
+        # 试常见格式
+        for fmt in ("%Y-%m-%d %H:%M:%S", "%Y-%m-%d", "%Y/%m/%d", "%Y/%m/%d %H:%M:%S"):
+            try:
+                return datetime.strptime(s, fmt).date()
+            except ValueError:
+                continue
+    return None
+
+
+# ============================================================
+# main
+# ============================================================
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="B 站抓取+周更: 自动推进 7/14/21/28/月曝光量槽 + 跨月开新子记录")
+    add_repeatable_style_argument(parser, "只跑指定款式编号(可多次)")
+    parser.add_argument("--record", action="append", help="只跑指定 record_id(可多次)")
+    parser.add_argument("--dry-run", action="store_true", help="只抓不写")
+    parser.add_argument("--today", help="强制 today 日期,格式 YYYY-MM-DD (调试用)")
+    parser.add_argument("--delay", type=float, default=0.3,
+                        help="每条请求间隔秒数(防风控,默认 0.3s)")
+    parser.add_argument("--data-dir", default=str(DEFAULT_DATA_DIR),
+                        help=f"数据目录(默认 {DEFAULT_DATA_DIR})")
+    parser.add_argument("--reset-state", action="store_true", help="重置运行状态")
+    parser.add_argument("--first-run", action="store_true",
+                        help="首次回填: 所有 B 站记录全部填 7天曝光量,跳过周次逻辑")
+    parser.add_argument("--self-operated", action="store_true",
+                        help="抓取自营达人表格(而非合作达人)")
+    args = parser.parse_args()
+
+    data_dir = resolve_layer_output(
+        args.data_dir, layer_root=PATHS.normalized_root, field="--data-dir"
+    )
+    mapping = load_mapping(data_dir, self_operated=args.self_operated)
+    styles = mapping["tables"]
+    if args.style:
+        try:
+            styles = select_requested_styles(styles, args.style)
+        except ValueError as exc:
+            print(str(exc)); return 1
+
+    only_rids = set(args.record) if args.record else None
+    force_today = None
+    if args.today:
+        force_today = date.fromisoformat(args.today)
+
+    state_path = BILIBILI_CHECKPOINT_DIR / STATE_FILENAME
+    if args.reset_state and state_path.exists():
+        state_path.unlink()
+        log("状态已重置")
+
+    state = load_state(data_dir)
+    session = requests.Session()
+    session.headers.update({"User-Agent": _UA})
+
+    out_dir = V2_DIR if data_dir == PATHS.normalized_root.resolve() else data_dir / "v2_results"
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    all_summaries = []
+    self_sfx = "_self" if args.self_operated else ""
+    final = out_dir / f"_all_summaries{self_sfx}_bilibili.json"
+    for s in styles:
+        summary = process_style(s, only_rids, args.dry_run, args.delay,
+                                session, state, force_today, args.first_run)
+        # 传了 --record 但本款式没匹配到 → 跳过 (不落空盘,不计入总汇总)
+        if only_rids and summary["total_b_records"] == 0 and summary["updated"] == 0 \
+                and summary["skipped"] == 0:
+            continue
+        # 每款式落盘
+        out_file = out_dir / f"{s['index']:02d}-{s['name']}{self_sfx}_bilibili_v2.json"
+        if only_rids and out_file.exists():
+            try:
+                previous = json.loads(out_file.read_text(encoding="utf-8"))
+            except Exception:
+                previous = None
+            summary = merge_bili_summary(previous, summary, dry_run=args.dry_run)
+        else:
+            summary = finalize_bili_summary(summary, dry_run=args.dry_run)
+        all_summaries.append(summary)
+        atomic_write_json(out_file, summary)
+        log(f"  [{s['index']:02d} {s['name']}] B站记录={summary['total_b_records']} "
+            f"更新={summary['updated']} 跳过={summary['skipped']} "
+            f"未解决={summary.get('unresolved', 0)} 落盘: {out_file.name}")
+
+    if not args.dry_run:
+        save_state(data_dir, state)
+        log(f"  状态已存: {state_path}")
+
+    # 总汇总按款式合并，单款/局部补跑不能删除其他款式。
+    try:
+        existing_global = json.loads(final.read_text(encoding="utf-8")) if final.exists() else []
+    except Exception:
+        existing_global = []
+    atomic_write_json(final, merge_global_summaries(existing_global, all_summaries))
+    log(f"\n=== 全部完成 ===")
+    log(f"汇总: {final}")
+    incomplete = [s.get("index") for s in all_summaries if not s.get("complete")]
+    if incomplete:
+        log(f"[ERROR] 以下款式仍有未解决采集/写回结果: {incomplete}")
+        return 2
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/chanmama_scraper.py b/src/gyxx_flow/modules/content_marketing/runtime/chanmama_scraper.py
new file mode 100644
index 0000000..46f765e
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/chanmama_scraper.py
@@ -0,0 +1,1293 @@
+# -*- coding: utf-8 -*-
+"""
+蝉妈妈 (chanmama.com) 自动化脚本
+功能：登录 → 跳转博主视频页 → 导出视频数据
+依赖：pip install selenium webdriver-manager
+"""
+
+import sys
+import io
+import time
+import os
+import csv
+import re
+import json
+import argparse
+import subprocess
+from datetime import datetime
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+# 修复 Windows 控制台输出编码
+sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8", errors="replace")
+sys.stderr = io.TextIOWrapper(sys.stderr.buffer, encoding="utf-8", errors="replace")
+
+try:
+    from selenium import webdriver
+    from selenium.webdriver.common.by import By
+    from selenium.webdriver.support.ui import WebDriverWait
+    from selenium.webdriver.support import expected_conditions as EC
+    from selenium.webdriver.chrome.options import Options
+    from selenium.common.exceptions import TimeoutException, NoSuchElementException
+    HAS_SELENIUM = True
+except ImportError:
+    HAS_SELENIUM = False
+    webdriver = None
+    By = None
+    WebDriverWait = None
+    EC = None
+    Options = None
+    TimeoutException = Exception
+    NoSuchElementException = Exception
+
+# ==================== 配置区 ====================
+ACCOUNT = os.getenv("CHANMAMA_ACCOUNT", "")
+PASSWORD = os.getenv("CHANMAMA_PASSWORD", "")
+LOGIN_URL = "https://www.chanmama.com/login.html"
+# 注意: verify_login() 用 TARGET_URLS[0] 做 cookie 探测,索引 0 必须是已知可用的 URL
+TARGET_URLS = [
+    "https://www.chanmama.com/bloggerRank/ujU3Wwa9udJc59DKO5dAA37ILJPsWvyg.html?activeTab=aweme",
+    "https://www.chanmama.com/bloggerRank/ulaJWIvluqa_q6bQZzLdkzmLXNz64tgg.html?activeTab=aweme",
+]
+DOWNLOAD_DIR = str(PATHS.raw_root / "chanmama")
+COOKIE_FILE = str(PATHS.browser_cookie_file)
+COOKIE_MAX_AGE = 7 * 24 * 3600  # cookie 有效期 7 天（秒）
+# ================================================
+
+
+def _find_local_chromedriver() -> str | None:
+    """在 selenium 缓存目录中查找 chromedriver.exe,避免联网下载。"""
+    candidates = [
+        os.path.expandvars(r"%USERPROFILE%\.cache\selenium\chromedriver"),
+        os.path.expandvars(r"%LOCALAPPDATA%\selenium\chromedriver"),
+    ]
+    base = None
+    for c in candidates:
+        if os.path.isdir(c):
+            base = c
+            break
+    if not base:
+        return None
+    latest = None
+    for root, _dirs, files in os.walk(base):
+        for f in files:
+            if f.lower() == "chromedriver.exe":
+                p = os.path.join(root, f)
+                if latest is None or os.path.getmtime(p) > os.path.getmtime(latest):
+                    latest = p
+    return latest
+
+
+def create_driver():
+    """创建 Chrome 浏览器驱动"""
+    os.makedirs(DOWNLOAD_DIR, exist_ok=True)
+
+    options = Options()
+    options.add_argument("--disable-blink-features=AutomationControlled")
+    options.add_experimental_option("excludeSwitches", ["enable-automation"])
+    options.add_experimental_option("useAutomationExtension", False)
+    options.add_argument("--no-sandbox")
+    options.add_argument("--disable-dev-shm-usage")
+    # 持久化浏览器 profile，稳定设备指纹，降低验证码触发概率，延长 cookie 寿命
+    profile_dir = str(PATHS.browser_profile_dir)
+    os.makedirs(profile_dir, exist_ok=True)
+    options.add_argument(f"--user-data-dir={profile_dir}")
+    # 设置下载目录
+    prefs = {
+        "download.default_directory": os.path.abspath(DOWNLOAD_DIR),
+        "download.prompt_for_download": False,
+        "directory_upgrade": True,
+        "safebrowsing.enabled": True,
+    }
+    options.add_experimental_option("prefs", prefs)
+
+    # 优先用本地 selenium 缓存的 chromedriver,避免 webdriver-manager 联网失败
+    try:
+        from selenium.webdriver.chrome.service import Service
+        local_drv = _find_local_chromedriver()
+        if local_drv:
+            service = Service(local_drv)
+            driver = webdriver.Chrome(service=service, options=options)
+        else:
+            driver = webdriver.Chrome(options=options)
+    except Exception as e:
+        print(f"⚠️ Chrome driver 初始化失败,回退默认: {e}")
+        driver = webdriver.Chrome(options=options)
+
+    driver.execute_cdp_cmd(
+        "Page.addScriptToEvaluateOnNewDocument",
+        {
+            "source": """
+            Object.defineProperty(navigator, 'webdriver', {
+                get: () => undefined
+            })
+            """
+        },
+    )
+    return driver
+
+
+def save_cookies(driver):
+    """保存 cookie 到文件"""
+    try:
+        cookies = driver.get_cookies()
+        data = {
+            "cookies": cookies,
+            "timestamp": time.time(),
+        }
+        cookie_path = Path(COOKIE_FILE)
+        cookie_path.parent.mkdir(parents=True, exist_ok=True)
+        temp_path = cookie_path.with_name(f".{cookie_path.name}.tmp")
+        temp_path.write_text(json.dumps(data, ensure_ascii=False, indent=2), encoding="utf-8")
+        os.replace(temp_path, cookie_path)
+        print(f"💾 已保存 cookie ({len(cookies)} 条) 到: {COOKIE_FILE}")
+        return True
+    except Exception as e:
+        print(f"⚠️ 保存 cookie 失败: {e}")
+        return False
+
+
+def load_cookies(driver):
+    """从文件加载 cookie 到浏览器
+
+    Returns:
+        bool: cookie 是否有效（有效期内）
+    """
+    if not os.path.exists(COOKIE_FILE):
+        print("📭 未找到 cookie 文件，需要首次登录")
+        return False
+
+    try:
+        data = json.loads(Path(COOKIE_FILE).read_text(encoding="utf-8"))
+
+        timestamp = data.get("timestamp", 0)
+        age = time.time() - timestamp
+
+        if age > COOKIE_MAX_AGE:
+            print(f"⏰ cookie 已过期 ({int(age // 86400)} 天前)，需要重新登录")
+            return False
+
+        cookies = data.get("cookies", [])
+        if not cookies:
+            print("📭 cookie 为空，需要重新登录")
+            return False
+
+        # 先访问域名再注入 cookie
+        driver.get("https://www.chanmama.com/")
+        time.sleep(2)
+
+        for cookie in cookies:
+            try:
+                # 移除可能导致注入失败的字段
+                cookie.pop("sameSite", None)
+                driver.add_cookie(cookie)
+            except Exception as e:
+                print(f"⚠️ 注入单个 cookie 失败: {e}")
+                continue
+
+        print(f"📦 已加载 {len(cookies)} 条 cookie ({(age // 86400):.1f} 天前保存)")
+        return True
+
+    except Exception as e:
+        print(f"⚠️ 加载 cookie 失败: {e}")
+        return False
+
+
+def handle_captcha(driver):
+    """处理登录时的验证码（滑块/图形/短信等）
+
+    点击登录后轮询等待：检测到验证码时提示用户在浏览器手动完成，
+    脚本自动检测登录是否成功并继续，无需手动按回车。
+    """
+    # 检测常见的验证码元素
+    captcha_selectors = [
+        "input[placeholder*='验证码']",
+        "input[placeholder*='图形']",
+        ".captcha",
+        ".verify-code",
+        ".slider",
+        ".slide-code",
+        "img[src*='captcha']",
+        ".geetest",
+        "#captcha",
+        ".code-img",
+    ]
+
+    def _has_captcha():
+        for selector in captcha_selectors:
+            try:
+                elem = driver.find_element(By.CSS_SELECTOR, selector)
+                if elem.is_displayed():
+                    return True
+            except NoSuchElementException:
+                continue
+        return False
+
+    def _logged_in_ok():
+        # 登录成功后通常会离开登录页
+        if "login" not in driver.current_url.lower():
+            return True
+        # 或者登录页出现成功提示
+        try:
+            if "登录成功" in driver.page_source or "欢迎" in driver.page_source:
+                return True
+        except Exception:
+            pass
+        return False
+
+    # 先等 2 秒让页面响应
+    time.sleep(2)
+
+    if not _has_captcha():
+        if _logged_in_ok():
+            print("✅ 登录成功，无需验证码")
+            return
+        # 可能验证码延迟出现，再观察一下
+        time.sleep(3)
+        if not _has_captcha() and _logged_in_ok():
+            print("✅ 登录成功，无需验证码")
+            return
+
+    if _has_captcha():
+        print("\n" + "=" * 50)
+        print("🔐 检测到验证码！请在浏览器中手动完成验证")
+        print("   （输入图形码 / 拖动滑块 / 短信验证等）")
+        print("   完成验证后脚本会自动继续，无需操作控制台")
+        print("=" * 50)
+
+    # 轮询等待：最多 180 秒，每 3 秒检测一次
+    max_wait = 180
+    waited = 0
+    interval = 3
+    while waited < max_wait:
+        if _logged_in_ok():
+            print("✅ 检测到登录成功，继续后续流程...")
+            return
+        # 还在登录页且仍有验证码 -> 继续等用户操作
+        waited += interval
+        remaining = max_wait - waited
+        if remaining > 0:
+            print(f"⏳ 等待您在浏览器中完成验证...（剩余 {remaining}s）")
+            time.sleep(interval)
+
+    print("⚠️ 等待验证码超时（180s），继续尝试后续步骤...")
+
+
+def verify_login(driver):
+    """验证是否已登录
+
+    通过访问目标页面，检查是否出现登录相关的元素
+    """
+    try:
+        # 跳转到目标页面
+        driver.get(TARGET_URLS[0])
+        WebDriverWait(driver, 12).until(
+            EC.presence_of_element_located((By.TAG_NAME, "body"))
+        )
+        time.sleep(3)
+
+        # 检查是否跳转回登录页 / 注册页（cookie 失效时蝉妈妈会重定向到 login.html 或 register.html）
+        cur_lower = driver.current_url.lower()
+        if "login" in cur_lower or "register" in cur_lower:
+            print(f"🔒 检测到被重定向到登录/注册页 ({driver.current_url})，cookie 已失效")
+            return False
+
+        # 检查页面是否包含登录按钮（未登录状态）
+        page_source = driver.page_source
+        if "立即登录" in page_source or "请登录" in page_source:
+            print("🔒 页面提示需要登录，cookie 已失效")
+            return False
+
+        print("✅ cookie 登录验证成功")
+        return True
+    except Exception as e:
+        print(f"⚠️ 验证登录失败: {e}")
+        return False
+
+
+def login(driver):
+    """执行登录操作（基于蝉妈妈真实 DOM 结构）
+
+    正确流程：
+    1. 打开登录页（默认显示"微信登录" tab）
+    2. 点击「蝉妈妈APP登录」tab 切换到密码登录
+    3. 等待表单出现 → 输入手机号、密码
+    4. 勾选"我已阅读并同意"协议
+    5. 点击「登录」按钮
+    6. 处理验证码（如有）
+    """
+    print(f"📂 正在打开登录页面: {LOGIN_URL}")
+    driver.get(LOGIN_URL)
+    wait = WebDriverWait(driver, 15)
+
+    # 步骤1：等待登录面板出现
+    print("⏳ 等待登录页加载...")
+    wait.until(EC.presence_of_element_located((By.TAG_NAME, "body")))
+    time.sleep(2)
+
+    # 步骤2：点击「小电脑图标」（class="app"）切换到账号密码登录
+    #   登录方式选择区有几个图标：微信 / 小电脑(app) / 短信 / 密码
+    #   默认可能在微信登录，需先点小电脑图标才出现账号密码框
+    print("🔄 点击小电脑图标切换到账号密码登录...")
+    try:
+        # 优先点 img.app 图标本身
+        app_icon = driver.find_element(By.CSS_SELECTOR, "img.app, img[class*='app']")
+        driver.execute_script("arguments[0].click();", app_icon)
+        time.sleep(2)
+        print("✅ 已点击【小电脑图标】")
+    except NoSuchElementException:
+        # 退而求其次：点击包含该图标的父级（tab-item / 可点击容器）
+        try:
+            parent = driver.find_element(
+                By.XPATH,
+                "//*[@class='tab-item' and (contains(.,'APP') or .//img)]"
+            )
+            parent.click()
+            time.sleep(2)
+            print("✅ 已通过父级容器切换到账号密码登录")
+        except Exception as e:
+            print(f"⚠️ 未找到小电脑图标，尝试继续: {e}")
+
+    # 步骤3：等待账号输入框出现（Element UI: placeholder='请输入手机号'）
+    print("🔑 等待登录表单加载...")
+    account_input = wait.until(
+        EC.presence_of_element_located(
+            (By.CSS_SELECTOR, "input[placeholder='请输入手机号']")
+        )
+    )
+    time.sleep(1)
+
+    # 输入账号
+    account_input.clear()
+    account_input.send_keys(ACCOUNT)
+    print(f"✅ 已输入账号: {ACCOUNT[:3]}****{ACCOUNT[-4:]}")
+
+    # 输入密码 (placeholder='请输入密码')
+    password_input = driver.find_element(
+        By.CSS_SELECTOR,
+        "input[placeholder='请输入密码']"
+    )
+    password_input.clear()
+    password_input.send_keys(PASSWORD)
+    print("✅ 已输入密码")
+
+    # 步骤4：勾选两个复选框（都要点！）
+    time.sleep(0.5)
+
+    # ① 记住我的账号 → label#e2e-login-record
+    try:
+        record_cb = driver.find_element(By.CSS_SELECTOR, "#e2e-login-record")
+        if "is-checked" not in record_cb.get_attribute("class"):
+            driver.execute_script("arguments[0].click();", record_cb)
+        print("✅ 已勾选【记住我的账号】")
+    except Exception as e:
+        print(f"⚠️ 记住我的账号勾选失败: {e}")
+
+    time.sleep(0.3)
+
+    # ② 我已阅读并同意 → 是一个自定义 div.cursor-pointer（不是 checkbox！）
+    #    必须点击这个 div 本体才生效
+    agree_clicked = False
+    try:
+        agree_div = driver.find_element(
+            By.XPATH,
+            "//span[contains(text(),'我已阅读并同意')]/ancestor::div[contains(@class,'cursor-pointer')][1]"
+        )
+        driver.execute_script("arguments[0].click();", agree_div)
+        time.sleep(0.3)
+        # 验证：点击后该 div 或其内部是否出现勾选标记（class 变化或新增元素）
+        agree_clicked = True
+        print("✅ 已点击【我已阅读并同意】协议区域")
+    except Exception as e:
+        print(f"⚠️ 协议 div 定位失败: {e}，尝试备用选择器")
+        try:
+            agree_div = driver.find_element(
+                By.XPATH,
+                "//div[contains(.,'我已阅读并同意') and contains(@class,'cursor-pointer')]"
+            )
+            driver.execute_script("arguments[0].click();", agree_div)
+            agree_clicked = True
+            print("✅ 已点击【我已阅读并同意】协议区域(备用)")
+        except Exception as e2:
+            print(f"⚠️ 协议勾选全部失败: {e2}")
+
+    # 步骤5：点击「登录」按钮 (button#e2e-login-submit)
+    try:
+        login_btn = driver.find_element(By.CSS_SELECTOR, "#e2e-login-submit")
+        driver.execute_script("arguments[0].click();", login_btn)
+        print("✅ 已点击【登录】按钮 (#e2e-login-submit)")
+    except NoSuchElementException:
+        btn = driver.find_element(By.CSS_SELECTOR, "button.el-button--primary")
+        driver.execute_script("arguments[0].click();", btn)
+        print("✅ 已点击【登录】按钮 (备用)")
+
+    # 步骤6：处理验证码（如有）：等待用户在浏览器中手动完成
+    handle_captcha(driver)
+
+    # 等待登录跳转完成
+    print("⏳ 等待登录完成...")
+    time.sleep(5)
+
+    # 验证是否登录成功
+    current_url = driver.current_url
+    if "login" not in current_url.lower():
+        print(f"✅ 登录成功！当前页面: {current_url}")
+        save_cookies(driver)
+        return True
+    else:
+        print(f"⚠️ 可能仍在登录页面，当前URL: {current_url}")
+        return False
+
+
+def navigate_to_target(driver, url):
+    """跳转到目标博主视频页面"""
+    print(f"🎯 正在跳转到目标页面: {url}")
+    driver.get(url)
+    WebDriverWait(driver, 15).until(
+        EC.presence_of_element_located((By.TAG_NAME, "body"))
+    )
+    time.sleep(3)  # 等待页面数据加载
+    print(f"✅ 已到达目标页面: {driver.current_url}")
+    return True
+
+
+def export_video_data(driver, pre_mtime: float = 0):
+    """导出视频记录数据,返回下载的 Excel 绝对路径。
+    pre_mtime: 进入导出前 DOWNLOAD_DIR 中已有文件的最大 mtime,用于识别本次新下载的文件。
+    返回: Excel 绝对路径(str),或失败时返回 None。
+    """
+    print("\n📊 开始导出视频数据...")
+
+    wait = WebDriverWait(driver, 10)
+
+    # 等待视频列表加载
+    try:
+        wait.until(
+            EC.presence_of_element_located((By.CSS_SELECTOR, "table, .video-list, .list-item, [class*='table']"))
+        )
+        print("✅ 视频列表已加载")
+    except TimeoutException:
+        print("⚠️ 视频列表加载超时，继续尝试导出...")
+
+    # 方法1: 尝试点击"批量导出"/"导出数据"按钮
+    export_btn_selectors = [
+        "//button[contains(text(),'导出')]",
+        "//a[contains(text(),'导出')]",
+        "//span[contains(text(),'导出')]",
+        "//button[contains(text(),'批量导出')]",
+        "//span[contains(text(),'批量导出')]",
+        "[class*='export']",
+        ".btn-export",
+        "#exportBtn",
+        "//div[contains(@class,'export')]",
+    ]
+
+    export_clicked = False
+    for selector in export_btn_selectors:
+        try:
+            if selector.startswith("//"):
+                btn = wait.until(EC.element_to_be_clickable((By.XPATH, selector)))
+            else:
+                btn = wait.until(EC.element_to_be_clickable((By.CSS_SELECTOR, selector)))
+
+            driver.execute_script("arguments[0].click();", btn)
+            export_clicked = True
+            print(f"✅ 已点击导出按钮: {selector}")
+            break
+        except (TimeoutException, NoSuchElementException):
+            continue
+
+    if not export_clicked:
+        print("❌ 未找到导出按钮，改用页面数据抓取方式...")
+        scrape_table_data(driver)  # 保留 CSV 写盘作为人工兜底
+        return None
+
+    # 等待导出确认弹窗出现并点击【导出】按钮
+    # 蝉妈妈弹窗标题"您即将进行数据导出"，按钮为 橙色"导出"（文字在 <span> 内）
+    print("⏳ 等待导出确认弹窗...")
+    export_confirmed = False
+    for wait_sec in range(20):
+        time.sleep(1)
+        try:
+            # 关键：用 contains(.,'导出') 匹配后代文本（"导出"在 <span> 内）
+            confirm_selectors = [
+                # 精确：主按钮(primary) + 含"导出"文字
+                "//button[contains(@class,'el-button--primary')][contains(.,'导出')]",
+                # 弹窗内任意含"导出"的按钮
+                "//div[contains(@class,'el-dialog')]//button[contains(.,'导出')]",
+                "//button[contains(.,'导出')]",
+                # 兼容"确定"/"确认"
+                "//button[contains(@class,'el-button--primary')][contains(.,'确定')]",
+                "//button[contains(.,'确认')]",
+            ]
+            for sel in confirm_selectors:
+                try:
+                    btn = driver.find_element(By.XPATH, sel)
+                    if btn.is_displayed():
+                        driver.execute_script("arguments[0].click();", btn)
+                        export_confirmed = True
+                        print(f"✅ 已在弹窗中点击导出按钮: {sel}")
+                        break
+                except NoSuchElementException:
+                    continue
+            if export_confirmed:
+                break
+        except Exception:
+            continue
+
+    if not export_confirmed:
+        print("ℹ️ 未检测到确认弹窗中的【导出】按钮，可能直接开始下载或已点击")
+
+    # 等待文件下载完成
+    print("⏳ 等待文件下载...")
+    time.sleep(10)
+    # 找本次新下载的 Excel(mtime > pre_mtime)
+    new_path = _find_new_excel_after(pre_mtime)
+    if new_path:
+        print(f"📦 本次下载: {new_path}")
+        return new_path
+    print("⚠️ 未找到本次下载的 Excel文件")
+    return None
+
+
+def scrape_table_data(driver):
+    """备用方案：直接从页面抓取表格数据保存为CSV"""
+    print("\n📋 正在抓取页面表格数据...")
+
+    rows_data = []
+
+    try:
+        # 等待表格加载
+        table = WebDriverWait(driver, 10).until(
+            EC.presence_of_element_located((By.TAG_NAME, "table"))
+        )
+
+        # 获取表头
+        header_cells = table.find_elements(By.CSS_SELECTOR, "thead th, thead td")
+        headers = [cell.text.strip() for cell in header_cells]
+        if not headers:
+            headers = ["视频标题", "发布时间", "播放量", "点赞数", "评论数", "转发量", "链接"]
+
+        print(f"📌 表头: {headers}")
+
+        # 获取所有行（包括分页）
+        all_rows = []
+        page = 1
+        max_pages = 50  # 安全限制
+
+        while page <= max_pages:
+            # 抓取当前页的数据行
+            body_rows = table.find_elements(
+                By.CSS_SELECTOR, "tbody tr, tbody > div[class*='row']"
+            )
+
+            if not body_rows and page == 1:
+                # 尝试其他表格结构
+                body_rows = driver.find_elements(
+                    By.CSS_SELECTOR, "[class*='video-item'], [class*='list-item'], [class*='row-item']"
+                )
+
+            page_has_data = False
+            for row in body_rows:
+                cells = row.find_elements(By.CSS_SELECTOR, "td, div[class*='cell']")
+                row_data = [cell.text.strip().replace("\n", " ") for cell in cells]
+
+                # 尝试获取视频链接
+                try:
+                    link_elem = row.find_element(By.CSS_SELECTOR, "a[href*='/video/'], a[href*='/note/']")
+                    row_data.append(link_elem.get_attribute("href"))
+                except NoSuchElementException:
+                    row_data.append("")
+
+                if any(row_data):  # 只添加非空行
+                    all_rows.append(row_data)
+                    page_has_data = True
+
+            print(f"📄 第 {page} 页: 抓取到 {len(body_rows)} 条记录")
+
+            if not page_has_data or len(body_rows) == 0:
+                break
+
+            # 尝试翻到下一页
+            try:
+                next_btn = driver.find_element(
+                    By.XPATH,
+                    "//li[@class='next']/a | //button[contains(text(),'下一页')] | //span[text()='下一页']"
+                )
+                if "disabled" in next_btn.get_attribute("class") or "disabled" in str(next_btn.get_attribute("aria-disabled")):
+                    print("📄 已到最后一页")
+                    break
+                driver.execute_script("arguments[0].click();", next_btn)
+                time.sleep(2)  # 等待翻页加载
+                page += 1
+            except NoSuchElementException:
+                print("📄 未找到下一页按钮，结束抓取")
+                break
+
+        # 保存为 CSV
+        timestamp = time.strftime("%Y%m%d_%H%M%S")
+        csv_path = os.path.join(DOWNLOAD_DIR, f"chanmama_videos_{timestamp}.csv")
+
+        with open(csv_path, "w", newline="", encoding="utf-8-sig") as f:
+            writer = csv.writer(f)
+            writer.writerow(headers + ["链接"])
+            writer.writerows(all_rows)
+
+        print(f"\n✅ 数据抓取完成！")
+        print(f"   📁 文件路径: {csv_path}")
+        print(f"   📊 共计 {len(all_rows)} 条记录")
+        return csv_path
+
+    except Exception as e:
+        print(f"❌ 表格抓取失败: {e}")
+        return None
+
+
+# ============================================================
+# Excel 解析 + 飞书自营表回填
+# ============================================================
+LARK_CLI = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK_CLI):
+    LARK_CLI = "lark-cli.cmd"
+
+
+def _find_new_excel_after(pre_mtime: float) -> str | None:
+    """找 DOWNLOAD_DIR 中 mtime > pre_mtime 的最新 .xlsx 文件。
+    pre_mtime=0 时退化为找目录中最新的 .xlsx(等价于 find_latest_chanmama_excel)。
+    排除 ~$ 开头的 Excel 临时锁文件。
+    """
+    if not os.path.exists(DOWNLOAD_DIR):
+        return None
+    candidates = []
+    for f in os.listdir(DOWNLOAD_DIR):
+        if not f.lower().endswith('.xlsx'):
+            continue
+        if f.startswith('~$'):
+            continue
+        full = os.path.join(DOWNLOAD_DIR, f)
+        mt = os.path.getmtime(full)
+        if mt > pre_mtime:
+            candidates.append((mt, full))
+    if not candidates:
+        return None
+    candidates.sort(key=lambda x: x[0], reverse=True)
+    return candidates[0][1]
+
+
+def find_latest_chanmama_excel():
+    """在 DOWNLOAD_DIR 找最新的 .xlsx 文件"""
+    if not os.path.exists(DOWNLOAD_DIR):
+        return None
+    files = [f for f in os.listdir(DOWNLOAD_DIR)
+             if f.lower().endswith('.xlsx') and not f.startswith('~$')]
+    if not files:
+        return None
+    files.sort(key=lambda f: os.path.getmtime(os.path.join(DOWNLOAD_DIR, f)),
+               reverse=True)
+    return os.path.join(DOWNLOAD_DIR, files[0])
+
+
+def find_all_chanmama_excels_grouped() -> list[str]:
+    """按博主名前缀分组,每组返回 mtime 最新的 .xlsx 路径。
+
+    蝉妈妈导出文件名格式: '<博主名>_视频记录_周期内发布视频_<日期范围>.xlsx'
+    分组键: 文件名第一个 '_视频记录_' 之前的部分(博主名)。
+    同一博主有多份历史导出时,只取最新一份。
+
+    返回: 每组最新文件的绝对路径列表(无序)。目录不存在或无文件时返回空列表。
+    """
+    if not os.path.exists(DOWNLOAD_DIR):
+        return []
+    groups: dict[str, tuple[float, str]] = {}  # 博主名 -> (mtime, 路径)
+    for f in os.listdir(DOWNLOAD_DIR):
+        if not f.lower().endswith('.xlsx'):
+            continue
+        if f.startswith('~$'):
+            continue
+        full = os.path.join(DOWNLOAD_DIR, f)
+        mt = os.path.getmtime(full)
+        # 分组键: _视频记录_ 之前的部分;若无该标记则用整个文件名(去扩展名)
+        name = f[:-5]  # 去 .xlsx
+        marker = '_视频记录_'
+        idx = name.find(marker)
+        key = name[:idx] if idx > 0 else name
+        cur = groups.get(key)
+        if cur is None or mt > cur[0]:
+            groups[key] = (mt, full)
+    # 按 mtime 倒序,确保两个博主 Excel 匹配同一飞书记录时,最新导出胜出(spec: first extended wins)
+    sorted_groups = sorted(groups.values(), key=lambda x: x[0], reverse=True)
+    return [path for _mt, path in sorted_groups]
+
+
+def _find_excel_col(headers, candidates):
+    """在表头里找第一个包含候选词的列,返回列索引(0-based)"""
+    for i, h in enumerate(headers):
+        h_str = str(h or '').strip()
+        for c in candidates:
+            if c in h_str:
+                return i
+    return None
+
+
+def _parse_exposure(v):
+    """解析曝光值,支持 '1.2万' / '6.5w+' / '1.1w' / '6,592' / 数字 / '0'"""
+    if v is None:
+        return None
+    if isinstance(v, (int, float)):
+        return int(v)
+    txt = str(v).strip()
+    if not txt or txt == '-':
+        return None
+    # 去掉后缀的 + 号(如 "6.5w+" → "6.5w")
+    txt = txt.rstrip('+')
+    m = re.match(r'^([\d,.]+)\s*(万|w|W)?', txt)
+    if not m:
+        return None
+    num = float(m.group(1).replace(',', ''))
+    if m.group(2):
+        num *= 10000
+    return int(num)
+
+
+def parse_chanmama_excel(xlsx_path):
+    """解析蝉妈妈导出的 Excel,返回 (records, headers)。
+    records: [{title, exposure, publish_time}, ...]
+    headers: 表头列表(用于调试)
+    """
+    import openpyxl
+    # read_only=True 对蝉妈妈导出的 xlsx 有解析 bug(只读出 1 行 1 列),
+    # 改用普通模式
+    wb = openpyxl.load_workbook(xlsx_path, data_only=True)
+    ws = wb.active
+
+    headers = []
+    for row in ws.iter_rows(min_row=1, max_row=1, values_only=True):
+        headers = [str(c or '').strip() for c in row]
+        break
+
+    print(f"📋 Excel 表头: {headers}")
+
+    # 蝉妈妈导出常见列名变体(按真实导出确认: 视频名称/发布时间/预估曝光)
+    title_candidates = ['视频名称', '视频标题', '标题', '作品标题',
+                        '作品名称', '视频名']
+    exposure_candidates = ['预估曝光', '预估曝光量', '预估播放量',
+                           '曝光量', '播放量', '预估播放', '曝光']
+    pubtime_candidates = ['发布时间', '发布日期', '发布']
+
+    title_col = _find_excel_col(headers, title_candidates)
+    exposure_col = _find_excel_col(headers, exposure_candidates)
+    pubtime_col = _find_excel_col(headers, pubtime_candidates)
+
+    print(f"  标题列: #{title_col} {headers[title_col] if title_col is not None else '?'}")
+    print(f"  曝光列: #{exposure_col} {headers[exposure_col] if exposure_col is not None else '?'}")
+    print(f"  发布时间列: #{pubtime_col} {headers[pubtime_col] if pubtime_col is not None else '?'}")
+
+    if title_col is None or exposure_col is None:
+        print("❌ 找不到标题列或曝光列,请检查表头")
+        return [], headers
+
+    records = []
+    for row in ws.iter_rows(min_row=2, values_only=True):
+        title = row[title_col] if title_col < len(row) else None
+        exposure = row[exposure_col] if exposure_col < len(row) else None
+        pubtime = (row[pubtime_col]
+                   if pubtime_col is not None and pubtime_col < len(row) else None)
+        if not title:
+            continue
+        # 跳过蝉妈妈的"总计"行
+        if str(title).strip() in ('总计', '合计', 'Total'):
+            continue
+        records.append({
+            'title': str(title).strip(),
+            'exposure': _parse_exposure(exposure),
+            'publish_time': pubtime,
+        })
+    wb.close()
+    return records, headers
+
+
+# ---- 飞书 API 工具 ----
+def _call_lark_json(args):
+    cmd = [LARK_CLI] + args
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, env=env,
+                          encoding="utf-8", errors="replace")
+    text = proc.stdout if proc.stdout else (proc.stderr or "")
+    start = text.find("{")
+    if start < 0:
+        return {"ok": False, "error": f"no json: {text[:300]}"}
+    try:
+        return json.loads(text[start:], strict=False)
+    except json.JSONDecodeError as e:
+        return {"ok": False, "error": f"json decode: {e}"}
+
+
+def _list_records_by_table(base_token, table_id):
+    all_records = []
+    offset = 0
+    while True:
+        resp = _call_lark_json([
+            "base", "+record-list",
+            "--base-token", base_token,
+            "--table-id", table_id,
+            "--as", "user",
+            "--format", "json",
+            "--limit", "200",
+            "--offset", str(offset),
+        ])
+        if not resp.get("ok"):
+            raise RuntimeError(f"record-list 失败: {resp.get('error')}")
+        d = resp["data"]
+        ids = d["field_id_list"]
+        rids = d["record_id_list"]
+        rows = d["data"]
+        for i, row in enumerate(rows):
+            rec = {"record_id": rids[i]}
+            for j, fid in enumerate(ids):
+                rec[fid] = row[j] if j < len(row) else None
+            all_records.append(rec)
+        offset += len(rows)
+        if not d.get("has_more", False) or len(rows) == 0:
+            break
+    return all_records
+
+
+def _write_back(base_token, table_id, record_id, field_id, value):
+    payload = {field_id: value}
+    resp = _call_lark_json([
+        "base", "+record-upsert",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--record-id", record_id,
+        "--as", "user",
+        "--format", "json",
+        "--json", json.dumps(payload, ensure_ascii=False),
+    ])
+    if not resp.get("ok"):
+        print(f"  ❌ 回填失败 record_id={record_id}: {resp.get('error')}")
+        return False
+    return True
+
+
+def _parse_publish_time(value):
+    if value is None or value == "":
+        return None
+    if isinstance(value, datetime):
+        return value
+    if isinstance(value, (int, float)):
+        # Excel 日期可能是 serial number(从 1899-12-30 起算)
+        try:
+            if 25569 <= value <= 60000:
+                from datetime import timedelta
+                return datetime(1899, 12, 30) + timedelta(days=value)
+        except Exception:
+            pass
+        try:
+            return datetime.fromtimestamp(value / 1000)
+        except (ValueError, OSError):
+            return None
+    if isinstance(value, str):
+        v = value.strip()
+        if not v:
+            return None
+        if v.isdigit():
+            try:
+                return datetime.fromtimestamp(int(v) / 1000)
+            except (ValueError, OSError):
+                pass
+        for fmt in ("%Y-%m-%dT%H:%M:%S", "%Y-%m-%d %H:%M:%S",
+                    "%Y-%m-%d", "%Y/%m/%d", "%Y.%m.%d"):
+            try:
+                return datetime.strptime(v[:len(fmt) + 2], fmt)
+            except ValueError:
+                continue
+    return None
+
+
+def _pick_read_field(fmap, pub_time, today=None):
+    if pub_time is None:
+        return None
+    today = today or datetime.now()
+    days = (today.date() - pub_time.date()).days
+    if days < 0:
+        return "FUTURE"
+    if days <= 7:
+        logical = "read_count_7d"
+    elif days <= 14:
+        logical = "read_count_14d"
+    elif days <= 21:
+        logical = "read_count_21d"
+    elif days <= 28:
+        logical = "read_count_28d"
+    else:
+        logical = "month_end"
+    info = fmap.get(logical)
+    if info:
+        return (logical, info["field_id"], info["field_name"])
+    return None
+
+
+def _title_match(haystack, needle):
+    """模糊标题匹配:
+    1. 剥掉 markdown 链接 [text](url) → 只留 text
+    2. 去掉 URL
+    3. 去掉空格/标点/emoji
+    4. 任一方包含另一方即命中
+    """
+    def strip_md_url(s):
+        s = re.sub(r'\[([^\]]*)\]\([^)]*\)', r'\1', str(s))
+        s = re.sub(r'https?://[^\s）)\]]+', '', s)
+        return s
+    def norm(s):
+        # 去空格/常见标点 + emoji(其他符号 Unicode 块)
+        s = re.sub(r"[\s　\-_.,;:!?！？，。；：、/\\()（）【】《》''\"\"'']+", "", s)
+        s = re.sub(r"[\U0001F000-\U0001FFFF\U00002600-\U000027BF\U0001F1E0-\U0001F1FF]+", "", s)
+        return s
+    nh = norm(strip_md_url(haystack))
+    nn = norm(strip_md_url(needle))
+    if not nh or not nn:
+        return False
+    return nn in nh or nh in nn
+
+
+def backfill_self_tables(excel_records, dry_run=False, only_style=None):
+    """遍历自营 mapping,用 excel_records 回填飞书自营表(抖音平台)"""
+    from gyxx_flow.modules.content_marketing.runtime import feishu_mapping
+    try:
+        mapping = feishu_mapping.load_mapping(self_operated=True)
+    except Exception as exc:
+        print(f"❌ 加载自营 mapping 失败: {exc}")
+        return
+
+    styles = mapping.get("tables", [])
+    if only_style:
+        styles = [s for s in styles if s.get("index") == only_style]
+        if not styles:
+            print(f"⚠️ 未找到款式 {only_style}")
+            return
+
+    total_filled = 0
+    total_matched = 0
+    total_tasks = 0
+    today = datetime.now()
+
+    v2_dir = PATHS.normalized_root / "v2_results"
+    v2_dir.mkdir(parents=True, exist_ok=True)
+
+    print(f"\n{'='*50}")
+    print(f"🔄 开始回填自营表 (dry_run={dry_run})")
+    print(f"  Excel 视频数: {len(excel_records)}")
+    print(f"  款式数: {len(styles)}")
+    print(f"{'='*50}")
+
+    for style in styles:
+        print(f"\n--- [{style['index']:02d}] {style['name']} ---")
+        fmap = style.get("field_map", {})
+        platform_fid = (fmap.get("platform") or {}).get("field_id")
+        title_fid = (fmap.get("note_title") or {}).get("field_id")
+        pub_fid = (fmap.get("publish_time") or {}).get("field_id")
+
+        if not platform_fid or not title_fid:
+            print("  [WARN] platform/note_title 字段缺失,跳过")
+            continue
+
+        try:
+            records = _list_records_by_table(style["base_token"], style["table_id"])
+        except Exception as exc:
+            print(f"  [ERR] 拉记录失败: {exc}")
+            continue
+
+        def is_douyin(v):
+            """platform 为空或是抖音 → True(回填);非抖音 → False(跳过)。
+            蝉妈妈 Excel 只有抖音数据,所以 platform 没填的默认回填,
+            只有明确填了非抖音(小红书/B站等)才跳过。"""
+            if v is None:
+                return True
+            if isinstance(v, list):
+                vals = [str(x).strip() for x in v if x]
+                if not vals:
+                    return True
+                return any(s in ("抖音", "douyin", "DOUYIN", "Douyin") for s in vals)
+            s = str(v).strip()
+            if not s:
+                return True
+            return s in ("抖音", "douyin", "DOUYIN", "Douyin")
+
+        douyin_records = [r for r in records if is_douyin(r.get(platform_fid))]
+        print(f"  全表 {len(records)} 条,抖音 {len(douyin_records)} 条")
+
+        style_filled = 0
+        style_matched = 0
+        style_tasks = 0
+        style_results = []
+
+        for rec in douyin_records:
+            title = rec.get(title_fid)
+            if isinstance(title, list):
+                title = " ".join(str(x) for x in title if x)
+            title = str(title or "").strip()
+            if not title:
+                continue
+            pub_raw = rec.get(pub_fid)
+            pub_time = _parse_publish_time(pub_raw)
+
+            style_tasks += 1
+            matched = None
+            for er in excel_records:
+                if _title_match(er["title"], title):
+                    matched = er
+                    break
+            if not matched:
+                continue
+            style_matched += 1
+
+            # 飞书无发布时间 → 先用 Excel 的发布时间填进去,再做时槽判断
+            if pub_time is None:
+                excel_pub_time = _parse_publish_time(matched.get("publish_time"))
+                if excel_pub_time is None:
+                    print(f"    [SKIP] 飞书和 Excel 都无发布时间: '{title[:20]}'")
+                    continue
+                pub_str = excel_pub_time.strftime("%Y-%m-%d %H:%M:%S")
+                if dry_run:
+                    print(f"    [DRY] '{title[:20]}' → 先填发布时间={pub_str}")
+                else:
+                    ts_ms = int(excel_pub_time.timestamp() * 1000)
+                    ok_pub = _write_back(style["base_token"], style["table_id"],
+                                         rec["record_id"], pub_fid, ts_ms)
+                    if ok_pub:
+                        print(f"    ✅ '{title[:20]}' → 先填发布时间={pub_str}")
+                pub_time = excel_pub_time
+
+            exposure = matched.get("exposure")
+            if exposure is None:
+                print(f"    [SKIP] 曝光值为空: '{title[:20]}'")
+                continue
+            pick = _pick_read_field(fmap, pub_time, today)
+            if pick == "FUTURE":
+                print(f"    [SKIP] 发布时间在未来: '{title[:20]}'")
+                continue
+            if not pick:
+                print(f"    [SKIP] 无对应曝光槽位字段: '{title[:20]}'")
+                continue
+            logical, fill_fid, fill_fname = pick
+            if dry_run:
+                print(f"    [DRY] '{title[:20]}' → {fill_fname}={exposure} "
+                      f"(Excel='{matched['title'][:20]}')")
+                style_filled += 1
+                style_results.append({
+                    "record_id": rec["record_id"],
+                    "creator": "",
+                    "target": title,
+                    "play_count": exposure,
+                    "matched": True,
+                    "slot": logical,
+                    "pub_time": pub_time.isoformat() if pub_time else None,
+                })
+                continue
+            ok = _write_back(style["base_token"], style["table_id"],
+                             rec["record_id"], fill_fid, exposure)
+            if ok:
+                style_filled += 1
+                style_results.append({
+                    "record_id": rec["record_id"],
+                    "creator": "",
+                    "target": title,
+                    "play_count": exposure,
+                    "matched": True,
+                    "ok": True,
+                    "slot": logical,
+                    "pub_time": pub_time.isoformat() if pub_time else None,
+                })
+                print(f"    ✅ '{title[:20]}' → {fill_fname}={exposure}")
+            else:
+                print(f"    ❌ '{title[:20]}' 回填失败")
+
+        total_filled += style_filled
+        total_matched += style_matched
+        total_tasks += style_tasks
+        print(f"  款式小计: 任务={style_tasks} 命中={style_matched} 回填={style_filled}")
+
+        style_summary = {
+            "style": style["name"],
+            "index": style["index"],
+            "total": style_tasks,
+            "matched": style_matched,
+            "filled": style_filled,
+            "results": style_results,
+        }
+        v2_file = v2_dir / f"{style['index']:02d}-{style['name']}_self_xingtu_v2.json"
+        tmp = v2_file.with_suffix(v2_file.suffix + ".tmp")
+        tmp.write_text(json.dumps(style_summary, ensure_ascii=False, indent=2),
+                       encoding="utf-8")
+        os.replace(tmp, v2_file)
+
+    print(f"\n{'='*50}")
+    print(f"📊 回填完成: 任务={total_tasks} 命中={total_matched} 回填={total_filled}")
+    if dry_run:
+        print("  (dry-run 模式,未实际写飞书)")
+    print(f"{'='*50}")
+
+
+def _run_self_test() -> int:
+    """内联自检: 验证 find_all_chanmama_excels_grouped + parse_chanmama_excel 能正常工作。"""
+    print("=" * 50)
+    print("🧪 chanmama_scraper self-test")
+    print("=" * 50)
+
+    paths = find_all_chanmama_excels_grouped()
+    print(f"\n1. find_all_chanmama_excels_grouped -> {len(paths)} 组")
+    for p in paths:
+        print(f"   - {os.path.basename(p)} (mtime={datetime.fromtimestamp(os.path.getmtime(p)).isoformat()})")
+
+    if not paths:
+        print("⚠️ data/chanmama/ 无 xlsx 文件,跳过解析自检")
+        return 0
+
+    total_records = 0
+    for p in paths:
+        records, headers = parse_chanmama_excel(p)
+        print(f"\n2. parse_chanmama_excel({os.path.basename(p)})")
+        print(f"   headers: {headers}")
+        print(f"   records: {len(records)} 条")
+        if records:
+            print(f"   sample: title='{records[0]['title'][:30]}' exposure={records[0]['exposure']}")
+        total_records += len(records)
+
+    print(f"\n✅ 自检完成,共解析 {total_records} 条 records")
+    return 0
+
+
+def main():
+    """主流程: 导出 Excel + 回填飞书自营表"""
+    parser = argparse.ArgumentParser(
+        description="蝉妈妈: 导出抖音视频数据 + 回填飞书自营表")
+    parser.add_argument("--backfill-only", action="store_true",
+                        help="只跑回填,不启动浏览器(需已有 Excel)")
+    parser.add_argument("--dry-run", action="store_true",
+                        help="回填时不实际写飞书")
+    parser.add_argument("--excel", help="指定 Excel 路径(默认找最新)")
+    parser.add_argument("--style", type=int, help="只回填指定款式编号")
+    parser.add_argument("--self-test", action="store_true",
+                        help="跑内置自检(分组函数 + 解析),不启动浏览器不写飞书")
+    args = parser.parse_args()
+
+    if args.self_test:
+        return _run_self_test()
+
+    # 只回填模式: 不启动浏览器,遍历各博主最新 Excel 合并回填
+    if args.backfill_only:
+        if args.excel:
+            paths = [args.excel]
+        else:
+            paths = find_all_chanmama_excels_grouped()
+        if not paths:
+            print("❌ 未找到任何 Excel 文件")
+            return 1
+        all_records = []
+        for p in paths:
+            if not os.path.exists(p):
+                print(f"⚠️ 文件不存在,跳过: {p}")
+                continue
+            print(f"📊 解析 Excel: {p}")
+            try:
+                records, _ = parse_chanmama_excel(p)
+            except Exception as exc:
+                print(f"⚠️ 解析失败,跳过: {p}: {exc}")
+                continue
+            print(f"  解析出 {len(records)} 条视频记录")
+            all_records.extend(records)
+        if not all_records:
+            print("⚠️ Excel 无有效记录")
+            return 0
+        print(f"\n📊 合并后共 {len(all_records)} 条 records,开始回填")
+        backfill_self_tables(all_records, dry_run=args.dry_run,
+                             only_style=args.style)
+        return 0
+
+    # 完整流程: 浏览器导出 + 回填
+    if not HAS_SELENIUM:
+        print("❌ selenium 未安装,无法启动浏览器。")
+        print("   用 --backfill-only 只跑回填,或 pip install selenium webdriver-manager")
+        return 1
+    print("=" * 50)
+    print("🦋 蝉妈妈自动化脚本启动")
+    print(f"  目标博主数: {len(TARGET_URLS)}")
+    print("=" * 50)
+
+    driver = None
+    all_records = []
+    try:
+        # 1. 创建浏览器驱动
+        print("\n🌐 启动浏览器...")
+        driver = create_driver()
+
+        # 2. 优先用 cookie 登录(避免每次都触发密码登录的验证码)
+        logged_in = False
+        if load_cookies(driver):
+            print("🔍 正在验证 cookie 有效性...")
+            if verify_login(driver):
+                logged_in = True
+                print("✨ 使用已保存的 cookie 登录成功,跳过手动登录!")
+
+        # 3. cookie 无效或过期时,回退到账号密码登录
+        if not logged_in:
+            print("\n🔐 cookie 不可用,执行账号密码登录...")
+            success = login(driver)
+            if success:
+                logged_in = True
+            else:
+                print("⚠️ 登录可能未成功,但继续尝试后续步骤...")
+
+        # 4. 循环每个博主 URL 导出 Excel
+        for i, url in enumerate(TARGET_URLS, 1):
+            print(f"\n{'='*50}")
+            print(f"📌 博主 {i}/{len(TARGET_URLS)}: {url}")
+            print(f"{'='*50}")
+            # 记录导出前已有文件的最大 mtime,用于识别本次新下载的文件
+            pre_mtime = 0
+            if os.path.exists(DOWNLOAD_DIR):
+                for f in os.listdir(DOWNLOAD_DIR):
+                    if f.lower().endswith('.xlsx') and not f.startswith('~$'):
+                        pre_mtime = max(pre_mtime,
+                                        os.path.getmtime(os.path.join(DOWNLOAD_DIR, f)))
+            try:
+                navigate_to_target(driver, url)
+                excel_path = export_video_data(driver, pre_mtime)
+                if not excel_path:
+                    print(f"⚠️ 博主 {i} 导出失败,跳过")
+                    continue
+                records, _ = parse_chanmama_excel(excel_path)
+                print(f"  解析出 {len(records)} 条视频记录")
+                all_records.extend(records)
+            except Exception as exc:
+                print(f"⚠️ 博主 {i} 异常: {exc},跳过")
+                continue
+
+        print("\n" + "=" * 50)
+        print(f"🎉 导出完成!共 {len(all_records)} 条 records")
+        print(f"💾 数据保存位置: {DOWNLOAD_DIR}")
+        print(f"🍪 cookie 保存位置: {COOKIE_FILE}")
+        print("=" * 50)
+
+    except KeyboardInterrupt:
+        print("\n\n⛔ 用户中断操作")
+    except Exception as e:
+        print(f"\n❌ 发生错误: {e}")
+        import traceback
+        traceback.print_exc()
+    finally:
+        if driver:
+            driver.quit()
+            print("👋 浏览器已关闭")
+
+    # 5. 回填飞书自营表(合并后一次性回填)
+    if not all_records:
+        print("\n⚠️ 无任何 Excel 记录,跳过回填")
+        return 0
+    backfill_self_tables(all_records, dry_run=args.dry_run,
+                         only_style=args.style)
+    return 0
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/collection_completeness.py b/src/gyxx_flow/modules/content_marketing/runtime/collection_completeness.py
new file mode 100644
index 0000000..cc6ada7
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/collection_completeness.py
@@ -0,0 +1,583 @@
+"""Shared completeness, matching, and atomic-result helpers for V2 scrapers."""
+
+from __future__ import annotations
+
+import json
+import os
+import re
+import unicodedata
+from difflib import SequenceMatcher
+from pathlib import Path
+from typing import Any, Iterable
+from urllib.parse import parse_qs, urlparse
+
+
+SUCCESS = "success"
+BLOCKED_INPUT = "blocked_input"
+RETRYABLE_FAILURE = "retryable_failure"
+WRITE_FAILURE = "write_failure"
+TERMINAL_STATUSES = {SUCCESS, BLOCKED_INPUT, RETRYABLE_FAILURE, WRITE_FAILURE}
+
+
+def add_repeatable_style_argument(parser: Any, help_text: str) -> Any:
+    """Add the shared repeatable --style option used by every platform CLI."""
+    return parser.add_argument(
+        "--style", type=int, action="append", help=help_text,
+    )
+
+
+def select_requested_styles(styles: Iterable[dict], requested: Iterable[int] | None) -> list[dict]:
+    """Select every requested style and fail when any requested index is unknown."""
+    style_list = list(styles)
+    if not requested:
+        return style_list
+    requested_set = {int(index) for index in requested}
+    selected = [style for style in style_list if int(style.get("index", -1)) in requested_set]
+    found = {int(style.get("index", -1)) for style in selected}
+    missing = sorted(requested_set - found)
+    if missing:
+        raise ValueError(f"未找到款式: {','.join(str(index) for index in missing)}")
+    return selected
+
+
+def atomic_write_json(path: str | Path, value: Any) -> None:
+    """Write JSON through a sibling temporary file and atomically replace target."""
+    target = Path(path)
+    target.parent.mkdir(parents=True, exist_ok=True)
+    tmp = target.with_suffix(target.suffix + ".tmp")
+    with open(tmp, "w", encoding="utf-8") as handle:
+        json.dump(value, handle, ensure_ascii=False, indent=2)
+        handle.flush()
+        os.fsync(handle.fileno())
+    os.replace(tmp, target)
+
+
+def coerce_url(value: Any) -> str:
+    """Extract a URL from common Feishu URL-field shapes."""
+    if value is None:
+        return ""
+    if isinstance(value, str):
+        text = value.strip()
+        markdown = re.search(r"\[[^\]]*\]\((https?://[^)]+)\)", text, flags=re.IGNORECASE)
+        if markdown:
+            return markdown.group(1).strip()
+        embedded = re.search(r"https?://[^\s)>\]]+", text, flags=re.IGNORECASE)
+        return embedded.group(0).strip() if embedded else text
+    if isinstance(value, dict):
+        for key in ("link", "url", "href", "text"):
+            candidate = value.get(key)
+            if isinstance(candidate, str) and candidate.strip():
+                return candidate.strip()
+        return ""
+    if isinstance(value, list):
+        for item in value:
+            candidate = coerce_url(item)
+            if candidate:
+                return candidate
+        return ""
+    return str(value).strip()
+
+
+def extract_content_id(url_value: Any, platform: str) -> str | None:
+    """Extract a stable Xiaohongshu note ID or Douyin video ID from a URL."""
+    url = coerce_url(url_value)
+    if not url:
+        return None
+    platform_key = platform.casefold()
+    if platform_key in {"xhs", "xiaohongshu", "pgy"}:
+        match = re.search(r"/(?:explore|discovery/item|note)/([A-Za-z0-9_-]+)", url)
+        return match.group(1) if match else None
+    if platform_key in {"douyin", "xingtu", "xt"}:
+        match = re.search(r"/video/(\d+)", url)
+        if match:
+            return match.group(1)
+        try:
+            query = parse_qs(urlparse(url).query)
+        except ValueError:
+            return None
+        for key in ("modal_id", "item_ids", "item_id", "video_id"):
+            values = query.get(key) or []
+            if values:
+                id_match = re.search(r"\d+", str(values[0]))
+                if id_match:
+                    return id_match.group(0)
+    return None
+
+
+def validate_platform_url(url_value: Any, platform: str) -> tuple[bool, str | None]:
+    """Validate URL host without rejecting known short-link domains."""
+    url = coerce_url(url_value)
+    if not url:
+        return False, "url_missing"
+    try:
+        host = (urlparse(url).hostname or "").casefold()
+    except ValueError:
+        return False, "url_invalid"
+    if not host:
+        return False, "url_invalid"
+    key = platform.casefold()
+    allowed = {
+        "xhs": ("xiaohongshu.com", "xhslink.com"),
+        "pgy": ("xiaohongshu.com", "xhslink.com"),
+        "xiaohongshu": ("xiaohongshu.com", "xhslink.com"),
+        "douyin": ("douyin.com", "iesdouyin.com"),
+        "xingtu": ("douyin.com", "iesdouyin.com"),
+        "xt": ("douyin.com", "iesdouyin.com"),
+    }.get(key, ())
+    if allowed and not any(host == domain or host.endswith("." + domain) for domain in allowed):
+        return False, "platform_url_mismatch"
+    return True, None
+
+
+def normalize_title(value: Any) -> str:
+    """Unicode/case normalization followed by punctuation and whitespace removal."""
+    text = unicodedata.normalize("NFKC", str(value or "")).casefold()
+    text = re.sub(r"[\u200b-\u200f\u2060\ufeff]", "", text)
+    return re.sub(r"[\W_]+", "", text, flags=re.UNICODE)
+
+
+def _title_variants(value: Any) -> list[str]:
+    """Build safe title forms for platform-added episode prefixes and hashtags."""
+    text = unicodedata.normalize("NFKC", str(value or "")).casefold()
+    text = re.sub(r"[\u200b-\u200f\u2060\ufeff]", "", text).strip()
+    if not text:
+        return []
+
+    raw_variants = [text]
+    without_episode = re.sub(
+        r"^\s*[【\[（(]?\s*第\s*[0-9一二三四五六七八九十百零两]+\s*"
+        r"[集期篇回弹]\s*[】\]）)]?\s*[:：\-—、.]?\s*",
+        "",
+        text,
+    )
+    without_episode = re.sub(
+        r"^\s*(?:ep(?:isode)?|part)\s*\d+\s*[:：\-—、.]?\s*",
+        "",
+        without_episode,
+        flags=re.IGNORECASE,
+    )
+    if without_episode != text:
+        raw_variants.append(without_episode)
+
+    for candidate in list(raw_variants):
+        before_hashtag = re.split(r"[#＃]", candidate, maxsplit=1)[0].strip()
+        if before_hashtag and before_hashtag != candidate:
+            raw_variants.append(before_hashtag)
+
+    variants: list[str] = []
+    for candidate in raw_variants:
+        normalized = normalize_title(candidate)
+        if normalized and normalized not in variants:
+            variants.append(normalized)
+    return variants
+
+
+def _episode_marker(value: Any) -> str | None:
+    text = unicodedata.normalize("NFKC", str(value or "")).casefold().strip()
+    chinese = re.match(
+        r"^[【\[（(]?\s*第\s*([0-9一二三四五六七八九十百零两]+)\s*[集期篇回弹]",
+        text,
+    )
+    english = re.match(r"^(?:ep(?:isode)?|part)\s*(\d+)", text, flags=re.IGNORECASE)
+    match = chinese or english
+    if not match:
+        return None
+    marker = match.group(1)
+    return str(int(marker)) if marker.isdigit() else marker
+
+
+def _normalized_title_similarity(card: str, target: str) -> float:
+    if not card or not target:
+        return 0.0
+    if card == target:
+        return 1.0
+    shorter, longer = sorted((card, target), key=len)
+    if len(shorter) >= 6 and shorter in longer:
+        coverage = len(shorter) / max(len(longer), 1)
+        return min(0.98, 0.93 + 0.05 * coverage)
+    ratio = SequenceMatcher(None, card, target, autojunk=False).ratio()
+    # Prefixes are common when the UI truncates a title.
+    common_prefix = os.path.commonprefix([card, target])
+    if len(common_prefix) >= 8:
+        ratio = max(ratio, 0.78 + min(0.12, len(common_prefix) / max(len(card), len(target)) * 0.12))
+    return ratio
+
+
+def title_similarity(card_title: Any, target_title: Any) -> float:
+    """Return a conservative 0..1 score suitable for unique-best matching."""
+    card_episode = _episode_marker(card_title)
+    target_episode = _episode_marker(target_title)
+    if card_episode and target_episode and card_episode != target_episode:
+        return 0.0
+    card_variants = _title_variants(card_title)
+    target_variants = _title_variants(target_title)
+    if not card_variants or not target_variants:
+        return 0.0
+    return max(
+        _normalized_title_similarity(card, target)
+        for card in card_variants
+        for target in target_variants
+    )
+
+
+def _metric_number(value: Any) -> float | None:
+    if isinstance(value, (int, float)):
+        return float(value)
+    text = str(value or "").strip().replace(",", "")
+    match = re.match(r"^([0-9]+(?:\.[0-9]+)?)\s*(万|w)?", text, flags=re.IGNORECASE)
+    if not match:
+        return None
+    number = float(match.group(1))
+    return number * 10000 if match.group(2) else number
+
+
+def deduplicate_observed_cards(cards: Iterable[dict]) -> list[dict]:
+    """Merge API/DOM observations of the same video while preserving distinct IDs."""
+    unique: list[dict] = []
+    for raw in cards:
+        if not isinstance(raw, dict):
+            continue
+        card = dict(raw)
+        card_id = str(card.get("note_id") or "")
+        card_url = coerce_url(card.get("href") or card.get("url")).rstrip("/")
+        duplicate_index: int | None = None
+        for index, old in enumerate(unique):
+            old_id = str(old.get("note_id") or "")
+            old_url = coerce_url(old.get("href") or old.get("url")).rstrip("/")
+            if card_id and old_id:
+                if card_id == old_id:
+                    duplicate_index = index
+                if duplicate_index is not None or card_id != old_id:
+                    if duplicate_index is not None:
+                        break
+                    continue
+            if card_url and old_url and card_url == old_url:
+                duplicate_index = index
+                break
+
+            one_is_api = (
+                card.get("source") == "show_items_api"
+                or old.get("source") == "show_items_api"
+            )
+            one_missing_id = not card_id or not old_id
+            same_title = (
+                normalize_title(card.get("title"))
+                and normalize_title(card.get("title")) == normalize_title(old.get("title"))
+            )
+            card_play = _metric_number(card.get("play_count"))
+            old_play = _metric_number(old.get("play_count"))
+            metric_compatible = (
+                card_play is None
+                or old_play is None
+                or abs(card_play - old_play) / max(card_play, old_play, 1.0) <= 0.08
+            )
+            if one_is_api and one_missing_id and same_title and metric_compatible:
+                duplicate_index = index
+                break
+
+        if duplicate_index is None:
+            unique.append(card)
+            continue
+
+        old = unique[duplicate_index]
+        card_quality = int(bool(card.get("note_id"))) + int(card.get("source") == "show_items_api")
+        old_quality = int(bool(old.get("note_id"))) + int(old.get("source") == "show_items_api")
+        preferred, fallback = (card, old) if card_quality > old_quality else (old, card)
+        merged = dict(fallback)
+        merged.update({key: value for key, value in preferred.items() if value not in (None, "")})
+        unique[duplicate_index] = merged
+    return unique
+
+
+def match_tasks_to_cards(
+    tasks: Iterable[dict],
+    cards: Iterable[dict],
+    platform: str,
+    *,
+    min_score: float = 0.78,
+    ambiguity_margin: float = 0.035,
+) -> dict[str, dict]:
+    """Match by ID first, then safely allocate each observed card to one source note."""
+    card_list = [dict(card) for card in cards if isinstance(card, dict)]
+    for card in card_list:
+        if not card.get("note_id"):
+            card["note_id"] = extract_content_id(card.get("href") or card.get("url"), platform)
+
+    task_list = [dict(task) for task in tasks if isinstance(task, dict) and task.get("record_id")]
+
+    def task_source_key(task: dict, content_id: Any = None) -> tuple[str, str]:
+        stable_id = content_id or task.get("note_id") or extract_content_id(task.get("note_url"), platform)
+        if stable_id:
+            return "id", str(stable_id)
+        url = coerce_url(task.get("note_url")).rstrip("/")
+        if url:
+            return "url", url
+        # Without a source URL, never let separate records reuse one fuzzy match.
+        return "record", str(task.get("record_id"))
+
+    def task_allows_card(task: dict, card: dict) -> bool:
+        task_id = task.get("note_id") or extract_content_id(task.get("note_url"), platform)
+        card_id = card.get("note_id")
+        return not task_id or not card_id or str(task_id) == str(card_id)
+
+    def card_key(card: dict, index: int) -> tuple[str, str]:
+        if card.get("note_id"):
+            return "id", str(card["note_id"])
+        url = coerce_url(card.get("href") or card.get("url")).rstrip("/")
+        return ("url", url) if url else ("index", str(index))
+
+    indexed_cards = [(index, card, card_key(card, index)) for index, card in enumerate(card_list)]
+    matched: dict[str, dict] = {}
+    claimed: dict[tuple[str, str], tuple[str, str]] = {}
+    remaining: list[dict] = []
+
+    for task in task_list:
+        record_id = str(task.get("record_id") or "")
+        content_id = task.get("note_id") or extract_content_id(task.get("note_url"), platform)
+        if content_id:
+            id_matches = [
+                (index, card, key)
+                for index, card, key in indexed_cards
+                if str(card.get("note_id") or "") == str(content_id)
+            ]
+            if id_matches:
+                _, raw_card, key = id_matches[0]
+                source_key = task_source_key(task, content_id)
+                owner = claimed.get(key)
+                if owner is not None and owner != source_key:
+                    remaining.append(task)
+                    continue
+                claimed[key] = source_key
+                selected = dict(raw_card)
+                selected.update({"match_method": "content_id", "match_score": 1.0})
+                matched[record_id] = selected
+                continue
+        remaining.append(task)
+
+    # If two different source URLs have the exact same title, title alone cannot
+    # identify which record owns the observed card. Leave both unresolved.
+    conflicted_cards: set[tuple[str, str]] = set()
+    for _, card, key in indexed_cards:
+        exact_sources = {
+            task_source_key(task)
+            for task in remaining
+            if task_allows_card(task, card)
+            if title_similarity(card.get("title"), task.get("target_title")) == 1.0
+        }
+        if len(exact_sources) > 1:
+            conflicted_cards.add(key)
+
+    def task_best_score(task: dict) -> float:
+        return max(
+            (
+                title_similarity(card.get("title"), task.get("target_title"))
+                for _, card, key in indexed_cards
+                if key not in conflicted_cards
+                and task_allows_card(task, card)
+            ),
+            default=0.0,
+        )
+
+    # Exact/strongest tasks claim first, so a looser title cannot steal the one
+    # card needed by a later exact task.
+    remaining.sort(key=lambda task: (-task_best_score(task), str(task.get("record_id"))))
+
+    for task in remaining:
+        record_id = str(task.get("record_id") or "")
+        source_key = task_source_key(task)
+
+        scored = sorted(
+            (
+                (title_similarity(card.get("title"), task.get("target_title")), index, card, key)
+                for index, card, key in indexed_cards
+                if key not in conflicted_cards
+                and task_allows_card(task, card)
+                and (claimed.get(key) is None or claimed.get(key) == source_key)
+            ),
+            key=lambda item: (-item[0], item[1]),
+        )
+        if not scored or scored[0][0] < min_score:
+            continue
+        best_score, _, best_card, best_key = scored[0]
+        second_score = scored[1][0] if len(scored) > 1 else 0.0
+        if (
+            second_score >= min_score
+            and best_score - second_score < ambiguity_margin
+            and (best_score < 1.0 or second_score == 1.0)
+        ):
+            second_card = scored[1][2]
+            best_id = str(best_card.get("note_id") or "")
+            second_id = str(second_card.get("note_id") or "")
+            if not best_id or best_id != second_id:
+                continue
+        claimed[best_key] = source_key
+        selected = dict(best_card)
+        selected.update({"match_method": "title", "match_score": round(best_score, 4)})
+        matched[record_id] = selected
+    return matched
+
+
+def result_status(row: dict) -> str:
+    """Read new status values and conservatively classify legacy result rows."""
+    explicit = row.get("status")
+    if explicit in TERMINAL_STATUSES:
+        return explicit
+    if row.get("write_ok") is False or row.get("write_error"):
+        return WRITE_FAILURE
+    if row.get("matched") is False or row.get("ok") is False or row.get("error") or row.get("reason"):
+        return RETRYABLE_FAILURE
+    if row.get("matched") is True or row.get("ok") is True:
+        return SUCCESS
+    metric_keys = ("read_count", "play_count", "view_count", "updated_value")
+    if any(row.get(key) is not None for key in metric_keys):
+        return SUCCESS
+    return RETRYABLE_FAILURE
+
+
+def upsert_result(summary: dict, row: dict) -> None:
+    """Replace the previous result for one record instead of appending duplicates."""
+    results = summary.setdefault("results", [])
+    record_id = row.get("record_id")
+    if record_id:
+        for index, old in enumerate(results):
+            if isinstance(old, dict) and old.get("record_id") == record_id:
+                results[index] = row
+                return
+    results.append(row)
+
+
+def finalize_summary(summary: dict, *, dry_run: bool = False) -> dict:
+    """Deduplicate results, recalculate counters, and attach completeness fields."""
+    out = dict(summary)
+    unique: list[dict] = []
+    positions: dict[str, int] = {}
+    duplicate_results = 0
+    for raw in out.get("results") or []:
+        if not isinstance(raw, dict):
+            continue
+        row = dict(raw)
+        status = result_status(row)
+        if status == SUCCESS and not dry_run and row.get("write_ok") is False:
+            status = WRITE_FAILURE
+        row["status"] = status
+        if status == SUCCESS:
+            row.setdefault("matched", True)
+        elif status != BLOCKED_INPUT:
+            row.setdefault("matched", False)
+        rid = str(row.get("record_id") or "")
+        if rid and rid in positions:
+            unique[positions[rid]] = row
+            duplicate_results += 1
+        else:
+            if rid:
+                positions[rid] = len(unique)
+            unique.append(row)
+
+    total = int(out.get("total") or 0)
+    success = sum(result_status(row) == SUCCESS for row in unique)
+    blocked = sum(result_status(row) == BLOCKED_INPUT for row in unique)
+    retryable = sum(result_status(row) == RETRYABLE_FAILURE for row in unique)
+    write_failures = sum(result_status(row) == WRITE_FAILURE for row in unique)
+    missing = max(0, total - len(unique))
+    unresolved = retryable + write_failures + missing
+    top_level_error = bool(out.get("error"))
+    if top_level_error:
+        retryable = max(retryable, 1)
+        unresolved = max(unresolved, 1)
+    if dry_run:
+        filled = 0
+    else:
+        explicit_write_rows = [row for row in unique if result_status(row) == SUCCESS and "write_ok" in row]
+        if explicit_write_rows:
+            filled = sum(row.get("write_ok") is True for row in explicit_write_rows)
+            # Preserve legacy successes that predate per-row write status.
+            filled += sum(
+                1 for row in unique
+                if result_status(row) == SUCCESS and "write_ok" not in row
+            )
+        else:
+            filled = min(int(out.get("filled") or 0), success)
+
+    out.update({
+        "results": unique,
+        "matched": success + write_failures,
+        "filled": filled,
+        "success": success,
+        "blocked_input": blocked,
+        "retryable_failures": retryable,
+        "write_failures": write_failures,
+        "missing_results": missing,
+        "duplicate_results": duplicate_results,
+        "unresolved": unresolved,
+        "complete": (
+            not top_level_error
+            and unresolved == 0
+            and duplicate_results == 0
+            and len(unique) == total
+        ),
+        "dry_run": bool(dry_run),
+    })
+    return out
+
+
+def merge_style_summary(existing: dict | None, partial: dict, *, dry_run: bool = False) -> dict:
+    """Merge a record-limited retry into the canonical per-style summary."""
+    if not existing:
+        return finalize_summary(partial, dry_run=dry_run)
+    combined = dict(existing)
+    for key, value in partial.items():
+        if key != "results":
+            combined[key] = value
+    rows: list[dict] = []
+    positions: dict[str, int] = {}
+    for source in (existing.get("results") or [], partial.get("results") or []):
+        for raw in source:
+            if not isinstance(raw, dict):
+                continue
+            row = dict(raw)
+            rid = str(row.get("record_id") or "")
+            if rid and rid in positions:
+                rows[positions[rid]] = row
+            else:
+                if rid:
+                    positions[rid] = len(rows)
+                rows.append(row)
+    combined["results"] = rows
+    combined["total"] = max(
+        int(existing.get("total") or 0),
+        int(partial.get("total") or 0),
+        len(rows),
+    )
+    return finalize_summary(combined, dry_run=dry_run)
+
+
+def merge_global_summaries(existing: Any, updates: Iterable[dict]) -> list[dict]:
+    """Merge selected style summaries into an existing all-style summary by index."""
+    merged: dict[Any, dict] = {}
+    order: list[Any] = []
+    for row in existing if isinstance(existing, list) else []:
+        if not isinstance(row, dict):
+            continue
+        key = row.get("index", row.get("style"))
+        if key not in merged:
+            order.append(key)
+        merged[key] = row
+    for row in updates:
+        if not isinstance(row, dict):
+            continue
+        key = row.get("index", row.get("style"))
+        if key not in merged:
+            order.append(key)
+        merged[key] = row
+    return [merged[key] for key in sorted(order, key=lambda value: (not isinstance(value, int), value))]
+
+
+def requested_records_succeeded(summary: dict, record_ids: Iterable[str]) -> tuple[bool, list[str]]:
+    """Validate that every requested record exists and has a successful terminal state."""
+    by_id = {
+        str(row.get("record_id")): row
+        for row in (summary.get("results") or summary.get("details") or [])
+        if isinstance(row, dict) and row.get("record_id")
+    }
+    failed = [str(rid) for rid in record_ids if result_status(by_id.get(str(rid), {})) != SUCCESS]
+    return not failed, failed
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/creator_task_grouping.py b/src/gyxx_flow/modules/content_marketing/runtime/creator_task_grouping.py
new file mode 100644
index 0000000..11252ed
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/creator_task_grouping.py
@@ -0,0 +1,53 @@
+"""Shared helpers for grouping cross-style scrape tasks by creator."""
+
+from __future__ import annotations
+
+from collections import OrderedDict, defaultdict
+
+
+def _clean_creator_id(value) -> str:
+    return str(value or "").strip()
+
+
+def _clean_creator_name(value) -> str:
+    return "".join(str(value or "").split()).casefold()
+
+
+def group_tasks_by_creator(tasks: list[dict]) -> list[dict]:
+    """Group tasks globally, preferring creator ID and falling back to name.
+
+    A name-only task joins an ID group when that normalized name maps to exactly
+    one ID in the current run. Conflicting IDs remain separate.
+    """
+    ids_by_name: dict[str, set[str]] = defaultdict(set)
+    for task in tasks:
+        creator_id = _clean_creator_id(task.get("creator_id"))
+        creator_name = _clean_creator_name(task.get("creator_name"))
+        if creator_id and creator_name:
+            ids_by_name[creator_name].add(creator_id)
+
+    groups: OrderedDict[tuple[str, str], dict] = OrderedDict()
+    for task in tasks:
+        creator_id = _clean_creator_id(task.get("creator_id"))
+        creator_name_key = _clean_creator_name(task.get("creator_name"))
+        if not creator_id and len(ids_by_name.get(creator_name_key, ())) == 1:
+            creator_id = next(iter(ids_by_name[creator_name_key]))
+        key = ("id", creator_id) if creator_id else ("name", creator_name_key)
+        if key not in groups:
+            groups[key] = {
+                "creator_name": str(task.get("creator_name") or "").strip(),
+                "creator_id": creator_id or None,
+                "tasks": [],
+            }
+        groups[key]["tasks"].append(task)
+    return list(groups.values())
+
+
+def chunk_creator_groups(groups: list[dict], batch_size: int) -> list[list[dict]]:
+    """Split creator groups into batches; non-positive means unlimited."""
+    if not groups:
+        return []
+    if batch_size <= 0:
+        return [groups]
+    return [groups[i:i + batch_size] for i in range(0, len(groups), batch_size)]
+
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/daily_marketing_report.py b/src/gyxx_flow/modules/content_marketing/runtime/daily_marketing_report.py
new file mode 100644
index 0000000..5c21c0b
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/daily_marketing_report.py
@@ -0,0 +1,1270 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""按款式生成一页全域种草日报，并可选发送给飞书负责人。"""
+
+import argparse
+import json
+import re
+import sys
+from collections import Counter
+from datetime import date, datetime, timedelta
+from pathlib import Path
+from typing import Any
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+if hasattr(sys.stdout, "reconfigure"):
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")
+if hasattr(sys.stderr, "reconfigure"):
+    sys.stderr.reconfigure(encoding="utf-8", errors="replace")
+
+ROOT = PATHS.module_root
+TOOLS_DIR = PATHS.tools_root
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools.analyze_comments import (  # noqa: E402
+    call_hermes_analyzer,
+)
+from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_conn  # noqa: E402
+from gyxx_flow.modules.content_marketing.runtime.data.tools.daily_report_charts import generate_dashboard  # noqa: E402
+from gyxx_flow.modules.content_marketing.runtime.data.tools.daily_report_card import send_daily_report_cards  # noqa: E402
+from gyxx_flow.modules.content_marketing.runtime.data.tools.daily_dashboard_analytics import (  # noqa: E402
+    build_dashboard_facts,
+    load_creator_connections,
+    load_style_categories,
+    load_style_images,
+)
+
+
+MIN_COMMENT_SAMPLE = 50
+BRAND_TERMS = ("光影行星",)
+PURCHASE_TERMS = ("已下单", "下单", "买了", "已买", "搜了", "求链接", "链接", "怎么买", "哪里买")
+PLATFORM_NAMES = {
+    "xiaohongshu": "小红书",
+    "douyin": "抖音",
+    "bilibili": "B站",
+    "tm": "天猫",
+    "jd": "京东",
+    "dy": "抖音电商",
+}
+DEFAULT_RECIPIENT_OPEN_IDS = (
+    "ou_7ad5fc8012e2f741afc5346e05ffd447",
+    "ou_fa8d81a16527ad06352dbecc575285b8",
+    "ou_2eda5eec112109ae6d19a1f6813eadcb",
+    "ou_89bcff110ccbb09a23548dc0fb3d880c",
+)
+
+SYSTEM_PROMPT = """# 角色
+你是一名专业的时尚包袋品牌电商运营数据分析师，负责每日分析品牌SKU的销售、电商运营及达人营销数据。你的任务是通过数据变化判断SKU状态、发现增长机会和风险，并输出《SKU营销运营日报》。要求：展示关键数据，同时结合数据进行分析判断，避免简单罗列。
+
+# 输入数据
+
+## 1. 电商运营数据
+平台：天猫、京东、抖音商城
+
+SKU数据：
+- SKU名称
+- 产品系列
+- 产品风格
+- 销量
+- 销量环比
+- 近7日累计销量
+- 未来7日销量趋势预估
+- 访客数
+- 访客环比
+- 转化率
+- 转化率变化
+- 加购率
+- 加购率变化
+- 收藏率
+- 退款率
+- 退款率变化
+
+## 2. 达人营销数据
+渠道：小红书、B站、抖音
+
+SKU内容数据：
+- 达人笔记数量
+- 曝光量
+- 曝光变化
+- 点赞量
+- 收藏量
+- 评论量
+- 互动率
+- 评论内容
+
+评论分析：
+- 用户喜欢点
+- 用户购买意向
+- 用户疑虑
+- 负面反馈
+
+# 分析逻辑
+
+## 1. 今日整体经营判断
+分析：
+- 销量趋势
+- 流量趋势
+- 转化趋势
+- 内容曝光趋势
+
+输出：
+- 今日整体状态：增长/稳定/下滑
+- 销量增长主要来源SKU
+- 当前风险SKU
+
+## 2. SKU经营表现分析
+输出表格：
+
+|SKU|今日销量|近7日销量|预测7日销量|访客|转化率|加购率|退款率|曝光量|综合判断|建议动作|
+|---|---|---|---|---|---|---|---|---|---|---|
+
+数据变化使用：
+↑上涨
+↓下降
+
+分析规则：
+
+① 增长型SKU
+判断：销量↑ + 流量↑ + 转化稳定或提升。
+结论：核心增长产品。
+建议：增加投放、扩大达人覆盖、延长产品生命周期。
+
+② 流量机会型SKU
+判断：访客↑ + 曝光↑ + 转化下降。
+结论：用户兴趣存在，但成交存在阻力。
+分析：卖点表达不足、页面承接不足、价格竞争力不足、用户评价影响。
+建议：优化详情页、评价体系、内容卖点。
+
+③ 潜力型SKU
+判断：销量一般，但加购率高、收藏高或内容互动高。
+结论：用户兴趣正在形成。
+建议：增加曝光测试，提高内容投入。
+
+④ 风险型SKU
+判断：销量下降 + 转化下降 + 退款上涨 + 负面评论增加。
+结论：产品体验或用户预期存在问题。
+分析：产品质量、尺寸容量、使用体验、产品描述偏差。
+建议：优化产品说明或调整营销策略。
+
+# 3. 达人营销效果分析
+
+## 小红书
+分析收藏率、评论质量、用户种草意愿，判断哪些SKU形成长期兴趣。
+
+## B站
+分析内容深度、用户讨论、长尾影响，判断哪些SKU适合持续内容建设。
+
+## 抖音
+分析曝光增长、互动表现、成交带动，判断哪些SKU适合快速放量。
+
+输出：
+- 今日最佳内容SKU
+- 内容表现原因
+- 对销量影响
+
+# 4. 产品节奏判断
+结合销量趋势 + 内容表现 + 用户反馈，判断SKU当前阶段：
+
+① 蓄水期
+表现：内容增长、收藏增加、销售未明显提升。
+动作：增加种草内容，测试卖点。
+
+② 放量期
+表现：销量增长、流量增长、转化稳定。
+动作：增加达人投放和渠道资源。
+
+③ 稳定期
+表现：销量稳定、增长放缓。
+动作：维护转化，寻找新场景。
+
+④ 衰退期
+表现：流量下降、转化下降、退款增加。
+动作：降低投入，寻找替代SKU。
+
+# 5. 输出日报
+
+《品牌SKU营销运营日报》
+
+日期：
+
+## 一、今日经营概览（比较昨日、七日基准）用图表展示
+- 销量环比：
+- 访客环比：
+- 转化环比：
+- 内容环比：
+- 今日核心结论：
+
+## 二、SKU表现
+
+|SKU|今日销量|近7日销量|预测7日销量|访客|转化|曝光量|综合判断|建议动作|
+|---|---|---|---|---|---|---|---|---|
+
+## 三、重点SKU分析
+
+【增长SKU TOP3】
+SKU：
+表现：
+原因：
+建议：
+
+【风险SKU TOP3】
+SKU：
+问题：
+原因：
+建议：
+
+## 四、达人营销分析
+最佳渠道：
+最佳SKU：
+内容反馈：
+用户洞察：
+
+## 五、明日运营建议
+输出3-5条具体动作：
+- 增加哪些SKU达人投放
+- 优化哪些SKU详情页
+- 测试哪些内容方向
+- 关注哪些产品问题
+
+# 输出要求
+1. 控制在一页A4以内。
+2. 数据必须使用↑↓展示趋势。
+3. 不重复描述数据，要输出判断。
+4. 分析必须同时结合：电商销量表现 + 访客数 + 转化率 + 达人笔记曝光 + 用户反馈。
+5. 所有建议必须具体到SKU。
+6. 只依据数据库查询结果判断；缺失的数据明确标注“无数据”，不得编造。
+7. 禁止输出数据库表名、字段名、JSON键名；必须改写为中文业务语言。
+8. 禁止出现反引号、null、代码式字段赋值或（a）/（b）/（c）编号。
+9. 禁止描述生成过程，例如“数据已收到”“我先用脚本处理”。
+10. 禁止在结尾追问、邀请补充或输出“如需……请告知”；正文直接结束。
+11. 全文控制在2500-3500个中文字符内，优先保留结论、核心表格、TOP3和具体动作。
+"""
+
+
+def safe_change(current: float | int | None, baseline: float | int | None) -> float | None:
+    if current is None or baseline is None or baseline <= 0:
+        return None
+    return (float(current) - float(baseline)) / float(baseline)
+
+
+def estimate_sales_forecast(history: list[float | int | None]) -> dict[str, int | float | None]:
+    """用完整7日销量计算累计值，并按最近3日销售速度估算未来7日。"""
+    empty = {
+        "sales_7d_total": None,
+        "sales_forecast_7d": None,
+        "sales_forecast_change": None,
+    }
+    if len(history) != 7 or any(value is None for value in history):
+        return empty
+    values = [float(value) for value in history]
+    previous_average = sum(values[:4]) / 4
+    recent_average = sum(values[4:]) / 3
+    if previous_average == 0:
+        forecast_change = 0.0 if recent_average == 0 else None
+    else:
+        forecast_change = (recent_average - previous_average) / previous_average
+    return {
+        "sales_7d_total": int(round(sum(values))),
+        "sales_forecast_7d": int(round(max(0.0, recent_average * 7))),
+        "sales_forecast_change": round(forecast_change, 4) if forecast_change is not None else None,
+    }
+
+
+def classify_light(visitor_change: float | None, cart_change: float | None) -> str:
+    changes = [abs(value) for value in (visitor_change, cart_change) if value is not None]
+    if not changes:
+        return "无判断"
+    severity = max(changes)
+    if severity >= 0.30:
+        return "🔴"
+    if severity >= 0.10:
+        return "🟡"
+    return "🟢"
+
+
+def calculate_comment_signal(style_name: str, comments: list[str]) -> dict[str, Any] | None:
+    cleaned = [str(text).strip() for text in comments if str(text or "").strip()]
+    if len(cleaned) < MIN_COMMENT_SAMPLE:
+        return None
+
+    product_terms = tuple(term for term in (style_name, style_name.replace(" ", "")) if term)
+    mention_terms = BRAND_TERMS + product_terms
+    mention_count = sum(any(term in text for term in mention_terms) for text in cleaned)
+    purchase_count = sum(any(term in text for term in PURCHASE_TERMS) for text in cleaned)
+
+    keyword_counts = Counter()
+    for text in cleaned:
+        for term in ("通勤", "旅行", "约会", "职场", "日常", "好看", "高级", "尺寸", "容量", "轻", "重"):
+            if term in text:
+                keyword_counts[term] += 1
+
+    return {
+        "sample_size": len(cleaned),
+        "brand_mention_rate": round(mention_count / len(cleaned), 4),
+        "purchase_intent_rate": round(purchase_count / len(cleaned), 4),
+        "top_terms": [term for term, _ in keyword_counts.most_common(3)],
+    }
+
+
+def note_label(note: dict[str, Any]) -> str:
+    placeholders = {"", "/", "-", "--", "未知", "无"}
+    for key in ("creator_name", "title"):
+        value = str(note.get(key) or "").strip()
+        if value not in placeholders:
+            return value
+    return f"笔记{note.get('id', '')}".rstrip()
+
+
+def require_valid_hermes_report(report: str) -> str:
+    """拒绝 Hermes 网关把上游连接错误包装成正常文本响应。"""
+    cleaned = str(report or "").strip()
+    lowered = cleaned.lower()
+    error_markers = (
+        "api call failed",
+        "connection error",
+        "connection refused",
+        "upstream error",
+        "traceback (most recent call last)",
+    )
+    if not cleaned or any(marker in lowered for marker in error_markers):
+        detail = cleaned[:240] or "空响应"
+        raise RuntimeError(f"Hermes 分析失败: {detail}")
+    return cleaned
+
+
+def sanitize_business_language(report: str) -> str:
+    """把模型偶发输出的数据库/代码语言转换为管理层可读的业务语言。"""
+    report = str(report or "").strip()
+
+    # 丢弃日报标题之前的模型寒暄或生成过程说明。
+    title_match = re.search(
+        r"(?m)^\s*(?:#{1,3}\s*)?(?:《?品牌SKU(?:电商|营销)?运营日报》?)\s*$",
+        report,
+    )
+    if title_match and title_match.start() > 0:
+        report = report[title_match.start():]
+
+    # 丢弃正文后的模型交互话术和数据导出邀请。
+    tail_match = re.search(r"(?m)^\s*(?:报告字数提示|完整数据索引)[：:]", report)
+    if tail_match:
+        report = report[:tail_match.start()].rstrip()
+
+    replacements = (
+        (r"(?i)publish_time\s*=\s*null", "发布时间无数据"),
+        (r"(?i)engagement_data_status\s*=\s*无数据", "互动数据无数据"),
+        (r"(?i)\bcooperations\b", "达人合作记录"),
+        (r"(?i)\bpublish_time\b", "发布时间"),
+        (r"(?i)\bengagement_data_status\b", "互动数据状态"),
+        (r"(?i)\bnull\b", "无数据"),
+        (r"(?i)\blisting\b", "商品页面"),
+    )
+    for pattern, replacement in replacements:
+        report = re.sub(pattern, replacement, report)
+
+    report = report.replace("`", "")
+    for letter, number in (("a", "1"), ("b", "2"), ("c", "3")):
+        report = re.sub(rf"[（(]{letter}[）)]", f"（{number}）", report, flags=re.IGNORECASE)
+    report = re.sub(r"\n{3,}", "\n\n", report)
+    return report.strip()
+
+
+def normalize_report(report: str, total_styles: int | None = None) -> str:
+    """纠正与当前基准口径冲突的常见模型措辞。"""
+    report = sanitize_business_language(report)
+    report = report.replace("同比", "较7日基准")
+    report = re.sub(r"销量\s*0\s*/\s*退款\s*0（无数据）", "销量 0 / 退款 0（当日均为0）", report)
+    report = report.replace("投放尚未发布", "投放发布时间无数据")
+    report = report.replace("投放未发布", "投放发布时间无数据")
+    report = report.replace("合作尚未发布", "合作发布时间无数据")
+    report = report.replace("退款率", "退款订单比")
+    report = report.replace("待发布后核对", "待补齐发布时间后核对")
+    report = report.replace("带动效应快速衰减", "同期访客增幅明显回落")
+    report = report.replace("带动效应从", "同期访客增幅从")
+    report = report.replace("未驱动访客回升", "未观察到同期访客回升")
+    if total_styles is not None:
+        report = re.sub(
+            r"(健康度[：:][^\n]*?)\d+\s*个款式中",
+            lambda match: f"{match.group(1)}{total_styles}个款式中",
+            report,
+        )
+    return sanitize_business_language(report)
+
+
+def build_platform_persona_section(enrichment: dict[str, Any]) -> str:
+    """平台榜首必须由数据库事实固定生成，禁止模型改写款式或指标。"""
+    def display(value: Any) -> str:
+        return "无数据" if value is None else f"{value:,}"
+
+    def pct(value: Any) -> str:
+        return "无数据" if value is None else f"{float(value):.1%}"
+
+    dashboard = enrichment.get("dashboard") or {}
+    rows = dashboard.get("platform_personas") or []
+    lines = ["三、平台人群与营销机会", ""]
+    for index, row in enumerate(rows, 1):
+        platform = row.get("platform") or "未知平台"
+        conversion = row.get("top_conversion") or {}
+        visitors = row.get("top_visitors") or {}
+        conversion_interest = conversion.get("interest_hypothesis") or "兴趣推测：无数据"
+        visitor_interest = visitors.get("interest_hypothesis") or "兴趣推测：无数据"
+        if conversion_interest == "无数据":
+            conversion_interest = "兴趣推测：无数据"
+        if visitor_interest == "无数据":
+            visitor_interest = "兴趣推测：无数据"
+        lines.append(
+            f"{index}、{platform}：转化最高为{conversion.get('style_name', '无数据')}"
+            f"（转化率{pct(conversion.get('conversion_rate'))}，访客{display(conversion.get('visitors'))}），"
+            f"人群画像：{conversion.get('persona_text') or '无数据'}；"
+            f"{conversion_interest}；"
+            f"优化：{conversion.get('optimization') or '无数据'}。"
+        )
+        lines.append(
+            f"   访客最高为{visitors.get('style_name', '无数据')}"
+            f"（访客{display(visitors.get('visitors'))}，转化率{pct(visitors.get('conversion_rate'))}），"
+            f"人群画像：{visitors.get('persona_text') or '无数据'}；"
+            f"{visitor_interest}；"
+            f"优化：{visitors.get('optimization') or '无数据'}。"
+        )
+        lines.append("")
+    daily_notes = dashboard.get("daily_new_notes") or []
+    if daily_notes:
+        labels = [
+            f"{row.get('style_name') or '无款式'}/{row.get('creator') or '无达人'}"
+            f"（{row.get('platform') or '无平台'}，曝光{row.get('view_count') if row.get('view_count') is not None else '无数据'}）"
+            for row in daily_notes[:5]
+        ]
+        lines.append(f"昨日新发布笔记：共{len(daily_notes)}篇；" + "；".join(labels) + "。")
+    else:
+        lines.append("昨日新发布笔记：无数据。")
+    return "\n".join(lines).rstrip()
+
+
+def replace_platform_persona_section(report: str, enrichment: dict[str, Any]) -> str:
+    section = build_platform_persona_section(enrichment)
+    heading_three = r"(?m)^\s*#*\s*三、平台人群与营销机会\s*$"
+    heading_four = re.search(r"(?m)^\s*#*\s*四、[^\n]*$", report)
+    if heading_four:
+        before = report[:heading_four.start()]
+        after = report[heading_four.start():]
+        before = re.sub(heading_three + r"(?s:.*)\Z", "", before).rstrip()
+        after = re.sub(heading_three + r"(?s:.*)\Z", "", after).rstrip()
+        return before + "\n\n" + section + "\n\n" + after + "\n"
+    return report.rstrip() + "\n\n" + section + "\n"
+
+
+def validate_all_styles(report: str, tracking_rows: list[dict[str, Any]]) -> None:
+    """所有款式必须出现在最终正文，防止模型自行截断长表。"""
+    missing = [row["style_name"] for row in tracking_rows if row["style_name"] not in report]
+    if missing:
+        raise RuntimeError(f"日报缺少款式: {', '.join(missing)}")
+
+
+def validate_enrichment_sections(report: str, enrichment: dict[str, Any]) -> None:
+    required = {
+        "personas": "人物画像",
+        "creatives": "主图",
+        "reviews": "商品评价",
+        "shop_overview": "店铺大盘",
+    }
+    missing = [label for key, label in required.items() if enrichment.get(key) and label not in report]
+    if missing:
+        raise RuntimeError(f"日报缺少可用的条件信号类别: {', '.join(missing)}")
+
+
+def validate_enriched_report(report: str) -> None:
+    if "无平台拆分数据" in report:
+        raise RuntimeError("重点款式缺少平台拆分，说明选择超出增强候选范围")
+    if "新品" in report:
+        raise RuntimeError("日报出现数据库未确认的“新品”描述")
+
+
+def build_condition_signals(enrichment: dict[str, Any]) -> str:
+    lines = ["C. 补充经营信号"]
+    order = {name: idx for idx, name in enumerate(enrichment.get("candidate_styles", []))}
+
+    personas = sorted(enrichment.get("personas", []), key=lambda row: (order.get(row.get("style_name"), 999), row.get("platform", "")))
+    persona_parts = []
+    seen = set()
+    for row in personas:
+        if row.get("style_name") in seen:
+            continue
+        labels = [row.get(key) for key in ("gender_top", "age_top", "city_top", "buying_power_top") if row.get(key)]
+        if labels:
+            persona_parts.append(f"{row['style_name']}/{row['platform']}（{row['data_date']}）：{'、'.join(labels[:3])}")
+            seen.add(row["style_name"])
+        if len(persona_parts) >= 2:
+            break
+    if persona_parts:
+        lines.append("- 人物画像：" + "；".join(persona_parts))
+
+    creatives = [row for row in enrichment.get("creatives", []) if row.get("ctr") is not None and (row.get("impressions") or 0) >= 100]
+    creative_pick = []
+    if creatives:
+        low = min(creatives, key=lambda row: row["ctr"])
+        high = max(creatives, key=lambda row: row["ctr"])
+        creative_pick = [low] + ([high] if high is not low else [])
+    if creative_pick:
+        parts = [
+            f"{row['style_name']}/{row['platform']}（{row['data_date']}）曝光{row['impressions']}、点击{row['clicks']}、CTR {row['ctr']}%"
+            for row in creative_pick
+        ]
+        lines.append("- 主图：" + "；".join(parts))
+
+    reviews = enrichment.get("reviews", [])[:2]
+    if reviews:
+        parts = [
+            f"{row['style_name']}（{row.get('window_start')}~{row.get('window_end')}）{row['sample_size']}条，负面{row['negative_count']}条/{row['negative_rate']:.1%}"
+            for row in reviews
+        ]
+        lines.append("- 商品评价：" + "；".join(parts))
+
+    shops = enrichment.get("shop_overview", [])
+    if shops:
+        parts = []
+        for row in shops:
+            metrics = [f"访客{row['visitors']}" if row.get("visitors") is not None else None,
+                       f"成交额{row['deal_amount']:.0f}" if row.get("deal_amount") is not None else None,
+                       f"转化率{row['conversion_rate']:.2f}%" if row.get("conversion_rate") is not None else None,
+                       f"退款额{row['refund_amount']:.0f}" if row.get("refund_amount") is not None else None]
+            parts.append(f"{row['platform']}（截至{row['week_end']}）：{'、'.join(value for value in metrics if value)}")
+        lines.append("- 店铺大盘：" + "；".join(parts))
+    return "\n".join(lines)
+
+
+def replace_status_groups(report: str, tracking_rows: list[dict[str, Any]], condition_signals: str = "") -> str:
+    """用确定性分组替换 Hermes 的全量状态段，避免重点款式被漏计。"""
+    header = re.search(r"(?m)^\**B\.\s*全量状态[^\n]*", report)
+    section_three = re.search(r"(?m)^\s*#*\s*三、", report)
+
+    groups: dict[str, list[str]] = {"🔴": [], "🟡": [], "🟢": [], "无判断": []}
+    for row in tracking_rows:
+        groups.setdefault(row.get("light", "无判断"), []).append(row["style_name"])
+    lines = []
+    for light in ("🔴", "🟡", "🟢", "无判断"):
+        names = groups.get(light, [])
+        lines.append(f"{light}（{len(names)}）：{'、'.join(names) if names else '无'}")
+    fixed_header = f"B. 全量状态（{len(tracking_rows)}款）"
+    middle = fixed_header + "\n\n" + "\n".join(lines)
+    if condition_signals:
+        middle += "\n\n" + condition_signals
+    if header and section_three and section_three.start() > header.end():
+        return report[:header.start()] + middle + "\n\n" + report[section_three.start():]
+    if section_three:
+        return report[:section_three.start()] + middle + "\n\n" + report[section_three.start():]
+    return report.rstrip() + "\n\n" + middle + "\n"
+
+
+def build_daily_prompt(
+    report_date: str,
+    today_notes: list[dict[str, Any]],
+    tracking_rows: list[dict[str, Any]],
+    next_day_rows: list[dict[str, Any]],
+    comment_rows: list[dict[str, Any]],
+    enrichment: dict[str, Any] | None = None,
+) -> str:
+    enrichment = enrichment or {}
+    prompt_enrichment = {key: value for key, value in enrichment.items() if key != "dashboard"}
+    payload = {
+        "报告日期": report_date,
+        "SKU总数": len(tracking_rows),
+        "所有SKU电商运营数据": tracking_rows,
+        "达人营销内容数据": {
+            "昨日新发布内容": today_notes,
+            "昨日内容次日追踪": next_day_rows,
+        },
+        "评论与补充经营数据": {
+            "评论数据": comment_rows,
+            **prompt_enrichment,
+        },
+    }
+    return f"""以下为数据库查询结果，请直接依据全新的系统提示词完成分析：
+```json
+{json.dumps(payload, ensure_ascii=False, indent=2, default=str)}
+```
+"""
+
+
+def _latest_report_date() -> date:
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute("SELECT MAX(metric_date) FROM daily_style_metrics")
+            value = cur.fetchone()[0]
+    if not value:
+        raise RuntimeError("daily_style_metrics 暂无可用日期")
+    return value
+
+
+def _all_style_names(metric_date: date) -> list[str]:
+    """返回截至报告日出现过的全部业务款式，含当天缺数据的款式。"""
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT DISTINCT style_name
+                FROM daily_style_metrics
+                WHERE metric_date <= %s AND style_name IS NOT NULL AND BTRIM(style_name) <> ''
+                ORDER BY style_name
+                """,
+                (metric_date,),
+            )
+            return [row[0] for row in cur.fetchall()]
+
+
+def _note_rows(report_date: date, start_date: date, end_date: date) -> list[dict[str, Any]]:
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT n.id, n.platform, n.title, n.publish_time, n.view_count,
+                       n.like_count, n.comment_count, n.collect_count, n.favorite_count,
+                       n.share_count, s.name AS style_name, c.name AS creator_name
+                FROM cmt_notes n
+                JOIN cmt_styles s ON s.id = n.style_id
+                LEFT JOIN cmt_creators c ON c.id = n.creator_id
+                WHERE n.publish_time >= %s AND n.publish_time < %s
+                  AND n.self_operated = FALSE
+                ORDER BY n.publish_time, n.id
+                """,
+                (start_date, end_date + timedelta(days=1)),
+            )
+            columns = [item.name for item in cur.description]
+            rows = [dict(zip(columns, row)) for row in cur.fetchall()]
+    for row in rows:
+        row["platform"] = PLATFORM_NAMES.get(row["platform"], row["platform"])
+        row["publish_date"] = row["publish_time"].date().isoformat() if row["publish_time"] else None
+        row["publish_time"] = row["publish_time"].strftime("%H:%M") if row["publish_time"] else None
+        interactions = [row.get(key) for key in ("like_count", "comment_count", "collect_count", "favorite_count", "share_count")]
+        known = [value for value in interactions if value is not None]
+        row["engagement_rate"] = round(sum(known) / row["view_count"], 4) if known and row.get("view_count") else None
+    return rows
+
+
+def _style_metrics(metric_date: date, style_names: list[str]) -> dict[str, dict[str, Any]]:
+    if not style_names:
+        return {}
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                WITH current_day AS (
+                    SELECT style_name, SUM(visitors) AS visitors, SUM(cart_users) AS cart_users,
+                           SUM(sales) AS sales, SUM(refund_orders) AS refund_orders
+                    FROM daily_style_metrics
+                    WHERE metric_date = %s AND style_name = ANY(%s)
+                    GROUP BY style_name
+                ), previous_day AS (
+                    SELECT style_name, SUM(visitors) AS visitors, SUM(cart_users) AS cart_users,
+                           SUM(sales) AS sales, SUM(refund_orders) AS refund_orders
+                    FROM daily_style_metrics
+                    WHERE metric_date = %s AND style_name = ANY(%s)
+                    GROUP BY style_name
+                ), daily_baseline AS (
+                    SELECT metric_date, style_name,
+                           SUM(visitors) AS visitors, SUM(cart_users) AS cart_users,
+                           SUM(sales) AS sales, SUM(refund_orders) AS refund_orders
+                    FROM daily_style_metrics
+                    WHERE metric_date >= %s AND metric_date < %s AND style_name = ANY(%s)
+                    GROUP BY metric_date, style_name
+                ), baseline AS (
+                    SELECT style_name, AVG(visitors) AS avg_visitors, AVG(cart_users) AS avg_cart_users,
+                           AVG(sales) AS avg_sales, AVG(refund_orders) AS avg_refund_orders
+                    FROM daily_baseline GROUP BY style_name
+                ), sales_history_daily AS (
+                    SELECT metric_date, style_name, SUM(sales) AS sales
+                    FROM daily_style_metrics
+                    WHERE metric_date >= %s AND metric_date <= %s AND style_name = ANY(%s)
+                    GROUP BY metric_date, style_name
+                ), sales_history AS (
+                    SELECT style_name, ARRAY_AGG(sales ORDER BY metric_date) AS sales_history_7d
+                    FROM sales_history_daily
+                    GROUP BY style_name
+                )
+                SELECT COALESCE(c.style_name, b.style_name), c.visitors, c.cart_users,
+                       c.sales, c.refund_orders,
+                       p.visitors, p.cart_users, p.sales, p.refund_orders,
+                       b.avg_visitors, b.avg_cart_users, b.avg_sales, b.avg_refund_orders,
+                       h.sales_history_7d
+                FROM current_day c FULL JOIN baseline b USING (style_name)
+                LEFT JOIN previous_day p ON p.style_name = COALESCE(c.style_name, b.style_name)
+                LEFT JOIN sales_history h ON h.style_name = COALESCE(c.style_name, b.style_name)
+                """,
+                (
+                    metric_date, style_names,
+                    metric_date - timedelta(days=1), style_names,
+                    metric_date - timedelta(days=7), metric_date, style_names,
+                    metric_date - timedelta(days=6), metric_date, style_names,
+                ),
+            )
+            rows = cur.fetchall()
+    result = {}
+    for (
+        style_name, visitors, cart_users, sales, refund_orders,
+        prev_visitors, prev_cart_users, prev_sales, prev_refund_orders,
+        avg_visitors, avg_cart_users, avg_sales, avg_refund_orders,
+        sales_history_7d,
+    ) in rows:
+        visitor_change = safe_change(visitors, avg_visitors)
+        cart_change = safe_change(cart_users, avg_cart_users)
+
+        def ratio(numerator, denominator):
+            if numerator is None or denominator is None or denominator <= 0:
+                return None
+            return float(numerator) / float(denominator)
+
+        conversion_rate = ratio(sales, visitors)
+        previous_conversion_rate = ratio(prev_sales, prev_visitors)
+        average_conversion_rate = ratio(avg_sales, avg_visitors)
+        cart_rate = ratio(cart_users, visitors)
+        previous_cart_rate = ratio(prev_cart_users, prev_visitors)
+        average_cart_rate = ratio(avg_cart_users, avg_visitors)
+        refund_rate = ratio(refund_orders, sales)
+        previous_refund_rate = ratio(prev_refund_orders, prev_sales)
+        average_refund_rate = ratio(avg_refund_orders, avg_sales)
+
+        def rounded(value):
+            return round(value, 4) if value is not None else None
+
+        def point_change(current, baseline):
+            if current is None or baseline is None:
+                return None
+            return current - baseline
+
+        result[style_name] = {
+            "visitors": visitors,
+            "cart_users": cart_users,
+            "sales": sales,
+            "refund_orders": refund_orders,
+            "visitors_yesterday": prev_visitors,
+            "cart_users_yesterday": prev_cart_users,
+            "sales_yesterday": prev_sales,
+            "refund_orders_yesterday": prev_refund_orders,
+            "avg_visitors_7d": round(float(avg_visitors), 1) if avg_visitors is not None else None,
+            "avg_cart_users_7d": round(float(avg_cart_users), 1) if avg_cart_users is not None else None,
+            "avg_sales_7d": round(float(avg_sales), 1) if avg_sales is not None else None,
+            "avg_refund_orders_7d": round(float(avg_refund_orders), 1) if avg_refund_orders is not None else None,
+            "visitor_change": round(visitor_change, 4) if visitor_change is not None else None,
+            "cart_change": round(cart_change, 4) if cart_change is not None else None,
+            "sales_change": round(safe_change(sales, avg_sales), 4) if safe_change(sales, avg_sales) is not None else None,
+            "refund_change": round(safe_change(refund_orders, avg_refund_orders), 4) if safe_change(refund_orders, avg_refund_orders) is not None else None,
+            "visitor_change_yesterday": rounded(safe_change(visitors, prev_visitors)),
+            "cart_change_yesterday": rounded(safe_change(cart_users, prev_cart_users)),
+            "sales_change_yesterday": rounded(safe_change(sales, prev_sales)),
+            "refund_change_yesterday": rounded(safe_change(refund_orders, prev_refund_orders)),
+            "conversion_rate": rounded(conversion_rate),
+            "conversion_rate_yesterday": rounded(previous_conversion_rate),
+            "avg_conversion_rate_7d": rounded(average_conversion_rate),
+            "conversion_change_yesterday": rounded(point_change(conversion_rate, previous_conversion_rate)),
+            "conversion_change_7d": rounded(point_change(conversion_rate, average_conversion_rate)),
+            "cart_rate": rounded(cart_rate),
+            "cart_rate_yesterday": rounded(previous_cart_rate),
+            "avg_cart_rate_7d": rounded(average_cart_rate),
+            "refund_rate": rounded(refund_rate),
+            "refund_rate_yesterday": rounded(previous_refund_rate),
+            "avg_refund_rate_7d": rounded(average_refund_rate),
+            "light": classify_light(visitor_change, cart_change),
+            **estimate_sales_forecast(list(sales_history_7d or [])),
+        }
+    return result
+
+
+def _select_enrichment_styles(tracking_rows: list[dict[str, Any]], limit: int = 12) -> list[str]:
+    """选出既覆盖大盘影响、又覆盖高百分比小基数的增强候选款式。"""
+    anomalies = [row for row in tracking_rows if row.get("light") in {"🔴", "🟡"}]
+
+    def impact(row: dict[str, Any]) -> float:
+        visitor_delta = abs(float(row.get("visitors") or 0) - float(row.get("avg_visitors_7d") or 0))
+        cart_delta = abs(float(row.get("cart_users") or 0) - float(row.get("avg_cart_users_7d") or 0))
+        return visitor_delta + cart_delta * 10
+
+    def pct_magnitude(row: dict[str, Any]) -> float:
+        values = [abs(float(value)) for value in (row.get("visitor_change"), row.get("cart_change")) if value is not None]
+        return max(values) if values else 0
+
+    selected = sorted(anomalies, key=impact, reverse=True)[:8]
+    selected_names = {row["style_name"] for row in selected}
+    for row in sorted(anomalies, key=pct_magnitude, reverse=True):
+        if row["style_name"] not in selected_names:
+            selected.append(row)
+            selected_names.add(row["style_name"])
+        if len(selected) >= limit:
+            break
+    return [row["style_name"] for row in selected]
+
+
+def _platform_metrics(metric_date: date, style_names: list[str]) -> list[dict[str, Any]]:
+    if not style_names:
+        return []
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                WITH current_day AS (
+                    SELECT style_name, platform, SUM(visitors) visitors, SUM(cart_users) carts,
+                           SUM(sales) sales, SUM(refund_orders) refunds
+                    FROM daily_style_metrics
+                    WHERE metric_date = %s AND style_name = ANY(%s)
+                    GROUP BY style_name, platform
+                ), daily_baseline AS (
+                    SELECT metric_date, style_name, platform, SUM(visitors) visitors,
+                           SUM(cart_users) carts, SUM(sales) sales, SUM(refund_orders) refunds
+                    FROM daily_style_metrics
+                    WHERE metric_date >= %s AND metric_date < %s AND style_name = ANY(%s)
+                    GROUP BY metric_date, style_name, platform
+                ), baseline AS (
+                    SELECT style_name, platform, AVG(visitors) avg_visitors, AVG(carts) avg_carts,
+                           AVG(sales) avg_sales, AVG(refunds) avg_refunds
+                    FROM daily_baseline GROUP BY style_name, platform
+                )
+                SELECT c.style_name, c.platform, c.visitors, c.carts, c.sales, c.refunds,
+                       b.avg_visitors, b.avg_carts, b.avg_sales, b.avg_refunds
+                FROM current_day c LEFT JOIN baseline b USING(style_name, platform)
+                ORDER BY c.style_name, c.platform
+                """,
+                (metric_date, style_names, metric_date - timedelta(days=7), metric_date, style_names),
+            )
+            rows = cur.fetchall()
+    output = []
+    for style_name, platform, visitors, carts, sales, refunds, av, ac, asal, ar in rows:
+        output.append({
+            "style_name": style_name,
+            "platform": PLATFORM_NAMES.get(platform, platform),
+            "visitors": visitors,
+            "visitor_change": safe_change(visitors, av),
+            "cart_users": carts,
+            "cart_change": safe_change(carts, ac),
+            "sales": sales,
+            "sales_change": safe_change(sales, asal),
+            "refund_orders": refunds,
+            "refund_change": safe_change(refunds, ar),
+        })
+    return output
+
+
+def _cooperation_context(metric_date: date, style_names: list[str]) -> list[dict[str, Any]]:
+    if not style_names:
+        return []
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                WITH ranked AS (
+                    SELECT s.name style_name, cr.name creator_name, c.publish_time,
+                           c.cooperation_date, c.cooperation_cost, c.ad_spend,
+                           NULLIF(c.content_direction, '') content_direction,
+                           c.exposure_count, c.engagement_count_num, c.cpm,
+                           ROW_NUMBER() OVER (PARTITION BY s.name ORDER BY COALESCE(c.publish_time, c.cooperation_date::timestamp) DESC, c.id DESC) rn
+                    FROM cmt_cooperations c
+                    JOIN cmt_styles s ON s.id = c.style_id
+                    LEFT JOIN cmt_creators cr ON cr.id = c.creator_id
+                    WHERE s.name = ANY(%s)
+                      AND COALESCE(c.publish_time::date, c.cooperation_date) BETWEEN %s AND %s
+                )
+                SELECT style_name, creator_name, publish_time, cooperation_date, cooperation_cost,
+                       ad_spend, content_direction, exposure_count, engagement_count_num, cpm
+                FROM ranked WHERE rn <= 3 ORDER BY style_name, rn
+                """,
+                (style_names, metric_date - timedelta(days=2), metric_date),
+            )
+            columns = [item.name for item in cur.description]
+            return [_sanitize_cooperation_row(dict(zip(columns, row))) for row in cur.fetchall()]
+
+
+def _sanitize_cooperation_row(row: dict[str, Any]) -> dict[str, Any]:
+    """隐藏无法由曝光量校验的互动数，并明确区分缺时间与未发布。"""
+    cleaned = dict(row)
+    exposure = cleaned.get("exposure_count")
+    engagement = cleaned.get("engagement_count_num")
+    cleaned["publish_status"] = "已记录发布时间" if cleaned.get("publish_time") else "发布时间无数据"
+    if engagement is not None and (exposure is None or engagement < 0 or engagement > exposure):
+        cleaned["engagement_count_num"] = None
+        cleaned["engagement_data_status"] = "无数据（原互动数无法通过曝光量校验，口径异常）"
+    else:
+        cleaned["engagement_data_status"] = "可用" if engagement is not None else "无数据"
+    return cleaned
+
+
+def _top_persona_value(payload: dict[str, Any], key: str) -> str | None:
+    rows = payload.get(key) or []
+    valid = []
+    for row in rows:
+        if not isinstance(row, dict) or not row.get("name") or row.get("value") is None:
+            continue
+        try:
+            value = float(row["value"])
+        except (TypeError, ValueError):
+            continue
+        if 0 <= value <= 100:
+            valid.append(row)
+    if not valid:
+        return None
+    row = max(valid, key=lambda item: float(item.get("value") or 0))
+    return f"{row['name']} {float(row['value']):.1f}%"
+
+
+def _persona_context(metric_date: date, style_names: list[str]) -> list[dict[str, Any]]:
+    if not style_names:
+        return []
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT DISTINCT ON (style_name, platform) style_name, platform, collect_date, payload
+                FROM daily_persona_metrics
+                WHERE collect_date <= %s AND style_name = ANY(%s)
+                ORDER BY style_name, platform, collect_date DESC, collected_at DESC
+                """,
+                (metric_date, style_names),
+            )
+            rows = cur.fetchall()
+    output = []
+    for style_name, platform, collect_date, payload in rows:
+        payload = payload or {}
+        output.append({
+            "style_name": style_name,
+            "platform": PLATFORM_NAMES.get(platform, platform),
+            "data_date": collect_date,
+            "gender_top": _top_persona_value(payload, "gender"),
+            "age_top": _top_persona_value(payload, "age"),
+            "city_top": _top_persona_value(payload, "city_tier"),
+            "buying_power_top": _top_persona_value(payload, "buying_power"),
+        })
+    return output
+
+
+def _creative_context(metric_date: date, style_names: list[str]) -> list[dict[str, Any]]:
+    if not style_names:
+        return []
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                WITH latest AS (
+                    SELECT style_name, platform, MAX(collect_date) collect_date
+                    FROM main_image_creatives
+                    WHERE collect_date <= %s AND style_name = ANY(%s)
+                    GROUP BY style_name, platform
+                )
+                SELECT m.style_name, m.platform, m.collect_date, SUM(m.impressions), SUM(m.clicks),
+                       SUM(m.cost), CASE WHEN SUM(m.impressions)>0 THEN SUM(m.clicks)::numeric/SUM(m.impressions)*100 END ctr
+                FROM main_image_creatives m JOIN latest l USING(style_name, platform, collect_date)
+                GROUP BY m.style_name, m.platform, m.collect_date
+                ORDER BY m.style_name, m.platform
+                """,
+                (metric_date, style_names),
+            )
+            rows = cur.fetchall()
+    return [{
+        "style_name": row[0], "platform": PLATFORM_NAMES.get(row[1], row[1]), "data_date": row[2],
+        "impressions": row[3], "clicks": row[4], "cost": float(row[5]) if row[5] is not None else None,
+        "ctr": round(float(row[6]), 2) if row[6] is not None else None,
+    } for row in rows]
+
+
+def _review_context(metric_date: date, style_names: list[str]) -> list[dict[str, Any]]:
+    if not style_names:
+        return []
+    negative_terms = ("差", "坏", "退", "重", "小", "掉色", "线头", "不值", "失望")
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT style_name, platform, content, score, tags
+                FROM product_reviews
+                WHERE style_name = ANY(%s) AND comment_date BETWEEN %s AND %s
+                  AND content IS NOT NULL AND BTRIM(content) <> ''
+                ORDER BY style_name, comment_date DESC, id DESC
+                """,
+                (style_names, metric_date - timedelta(days=6), metric_date),
+            )
+            rows = cur.fetchall()
+    grouped: dict[str, list[tuple]] = {}
+    for row in rows:
+        grouped.setdefault(row[0], []).append(row)
+    output = []
+    for style_name, items in grouped.items():
+        if len(items) < 5:
+            continue
+        negative = [item for item in items if (item[3] is not None and item[3] <= 2) or any(term in item[2] for term in negative_terms)]
+        output.append({
+            "style_name": style_name,
+            "window_start": metric_date - timedelta(days=6),
+            "window_end": metric_date,
+            "sample_size": len(items),
+            "negative_count": len(negative),
+            "negative_rate": round(len(negative) / len(items), 4),
+            "negative_examples": [item[2][:80] for item in negative[:2]],
+        })
+    return output
+
+
+def _style_comment_context(metric_date: date, style_names: list[str]) -> list[dict[str, Any]]:
+    if not style_names:
+        return []
+    negative_terms = ("差", "坏", "退", "重", "小", "掉色", "线头", "失望", "不值")
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT s.name, c.content
+                FROM cmt_comments c
+                JOIN cmt_notes n ON n.id = c.note_id
+                JOIN cmt_styles s ON s.id = n.style_id
+                WHERE s.name = ANY(%s)
+                  AND c.content IS NOT NULL AND BTRIM(c.content) <> ''
+                  AND COALESCE(c.comment_created_at, c.created_at)::date BETWEEN %s AND %s
+                """,
+                (style_names, metric_date - timedelta(days=29), metric_date),
+            )
+            rows = cur.fetchall()
+    grouped: dict[str, list[str]] = {}
+    for style_name, content in rows:
+        grouped.setdefault(style_name, []).append(content)
+    return [{
+        "style_name": style_name,
+        "comment_count": len(items),
+        "negative_count": sum(any(term in text for term in negative_terms) for text in items),
+        "negative_rate": round(sum(any(term in text for term in negative_terms) for text in items) / len(items), 4),
+    } for style_name, items in grouped.items()]
+
+
+def _build_style_signals(
+    style_names: list[str],
+    personas: list[dict[str, Any]],
+    cooperations: list[dict[str, Any]],
+    comments: list[dict[str, Any]],
+    reviews: list[dict[str, Any]],
+    creatives: list[dict[str, Any]],
+) -> dict[str, dict[str, Any]]:
+    result = {name: {} for name in style_names}
+    for row in personas:
+        result.setdefault(row["style_name"], {})["persona"] = True
+    for row in cooperations:
+        signal = result.setdefault(row["style_name"], {})
+        signal["note_count"] = signal.get("note_count", 0) + 1
+        exposure = row.get("exposure_count")
+        if exposure is not None:
+            signal["recent_note_exposure"] = signal.get("recent_note_exposure", 0) + int(exposure)
+    for row in comments:
+        signal = result.setdefault(row["style_name"], {})
+        signal["comment_count"] = row.get("comment_count") or 0
+        signal["comment_negative_rate"] = row.get("negative_rate")
+    for row in reviews:
+        signal = result.setdefault(row["style_name"], {})
+        signal["review_count"] = row.get("sample_size") or 0
+        signal["review_negative_rate"] = row.get("negative_rate")
+    for row in creatives:
+        signal = result.setdefault(row["style_name"], {})
+        ctr = row.get("ctr")
+        if ctr is not None and (signal.get("creative_ctr") is None or ctr > signal["creative_ctr"]):
+            signal["creative_ctr"] = ctr
+    return result
+
+
+def _shop_overview(metric_date: date) -> list[dict[str, Any]]:
+    queries = [
+        ("天猫", "SELECT week_end, visitors, deal_amount, conversion_rate, refund_amount FROM tm_shop_weekly WHERE week_end<=%s ORDER BY week_end DESC LIMIT 1"),
+        ("京东", "SELECT week_end, visitors, deal_amount, conversion_rate, refund_amount FROM jd_shop_weekly WHERE week_end<=%s ORDER BY week_end DESC LIMIT 1"),
+        ("抖音电商", "SELECT week_end, visitors, deal_amount, NULL::numeric, refund_amount FROM dy_shop_weekly WHERE week_end<=%s ORDER BY week_end DESC LIMIT 1"),
+    ]
+    output = []
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            for platform, query in queries:
+                cur.execute(query, (metric_date,))
+                row = cur.fetchone()
+                if row:
+                    output.append({
+                        "platform": platform, "week_end": row[0], "visitors": row[1],
+                        "deal_amount": float(row[2]) if row[2] is not None else None,
+                        "conversion_rate": float(row[3]) if row[3] is not None else None,
+                        "refund_amount": float(row[4]) if row[4] is not None else None,
+                    })
+    return output
+
+
+def _load_enrichment(metric_date: date, style_names: list[str]) -> dict[str, Any]:
+    return {
+        "platform_metrics": _platform_metrics(metric_date, style_names),
+        "cooperations": _cooperation_context(metric_date, style_names),
+        "personas": _persona_context(metric_date, style_names),
+        "creatives": _creative_context(metric_date, style_names),
+        "reviews": _review_context(metric_date, style_names),
+        "shop_overview": _shop_overview(metric_date),
+    }
+
+
+def _comment_rows(yesterday_notes: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    if not yesterday_notes:
+        return []
+    note_map = {row["id"]: row for row in yesterday_notes}
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT note_id, content FROM cmt_comments
+                WHERE note_id = ANY(%s) AND content IS NOT NULL AND BTRIM(content) <> ''
+                ORDER BY note_id, id
+                """,
+                (list(note_map),),
+            )
+            grouped: dict[int, list[str]] = {}
+            for note_id, content in cur.fetchall():
+                grouped.setdefault(note_id, []).append(content)
+    output = []
+    for note_id, comments in grouped.items():
+        note = note_map[note_id]
+        signal = calculate_comment_signal(note["style_name"], comments)
+        if signal is None:
+            continue
+        output.append({
+            "note_id": note_id,
+            "note": note_label(note),
+            "platform": note["platform"],
+            "style_name": note["style_name"],
+            **signal,
+        })
+    return output
+
+
+def load_report_facts(report_date: date) -> tuple[list[dict], list[dict], list[dict], list[dict], dict[str, Any]]:
+    recent_start = report_date - timedelta(days=2)
+    recent_notes = _note_rows(report_date, recent_start, report_date)
+    yesterday = report_date - timedelta(days=1)
+    yesterday_notes = [row for row in recent_notes if row["publish_date"] == yesterday.isoformat()]
+
+    styles = _all_style_names(report_date)
+    today_metrics = _style_metrics(report_date, styles)
+    publish_metrics = _style_metrics(yesterday, sorted({row["style_name"] for row in yesterday_notes}))
+
+    tracking_rows = []
+    missing_metrics = {
+        "visitors": None,
+        "cart_users": None,
+        "sales": None,
+        "refund_orders": None,
+        "visitors_yesterday": None,
+        "cart_users_yesterday": None,
+        "sales_yesterday": None,
+        "refund_orders_yesterday": None,
+        "avg_visitors_7d": None,
+        "avg_cart_users_7d": None,
+        "avg_sales_7d": None,
+        "sales_7d_total": None,
+        "sales_forecast_7d": None,
+        "sales_forecast_change": None,
+        "avg_refund_orders_7d": None,
+        "visitor_change": None,
+        "cart_change": None,
+        "sales_change": None,
+        "refund_change": None,
+        "visitor_change_yesterday": None,
+        "cart_change_yesterday": None,
+        "sales_change_yesterday": None,
+        "refund_change_yesterday": None,
+        "conversion_rate": None,
+        "conversion_rate_yesterday": None,
+        "avg_conversion_rate_7d": None,
+        "conversion_change_yesterday": None,
+        "conversion_change_7d": None,
+        "cart_rate": None,
+        "cart_rate_yesterday": None,
+        "avg_cart_rate_7d": None,
+        "refund_rate": None,
+        "refund_rate_yesterday": None,
+        "avg_refund_rate_7d": None,
+        "light": "无判断",
+    }
+    for style_name in styles:
+        metrics = today_metrics.get(style_name, missing_metrics)
+        tracking_rows.append({"style_name": style_name, **metrics})
+    rank = {"🔴": 0, "🟡": 1, "🟢": 2, "无判断": 3}
+    tracking_rows.sort(key=lambda row: (rank[row["light"]], row["style_name"]))
+
+    next_day_rows = []
+    for note in yesterday_notes:
+        publish = publish_metrics.get(note["style_name"], {})
+        current = today_metrics.get(note["style_name"], {})
+        next_day_rows.append({
+            "note": note_label(note),
+            "platform": note["platform"],
+            "style_name": note["style_name"],
+            "publish_day_visitor_change": publish.get("visitor_change"),
+            "today_visitor_change": current.get("visitor_change"),
+        })
+
+    compact_yesterday = [{
+        "note_id": row["id"],
+        "creator": row.get("creator_name"),
+        "platform": row["platform"],
+        "time": row["publish_time"],
+        "title": row.get("title"),
+        "style_name": row["style_name"],
+        "view_count": row.get("view_count"),
+        "engagement_rate": row.get("engagement_rate"),
+    } for row in yesterday_notes]
+    enrichment_styles = _select_enrichment_styles(tracking_rows)
+    enrichment = _load_enrichment(report_date, enrichment_styles)
+    enrichment["candidate_styles"] = enrichment_styles
+    all_platform_metrics = _platform_metrics(report_date, styles)
+    all_personas = _persona_context(report_date, styles)
+    connections = load_creator_connections(report_date, styles)
+    style_images = load_style_images(styles)
+    style_categories = load_style_categories(styles)
+    all_cooperations = _cooperation_context(report_date, styles)
+    all_comments = _style_comment_context(report_date, styles)
+    all_reviews = _review_context(report_date, styles)
+    all_creatives = _creative_context(report_date, styles)
+    style_signals = _build_style_signals(
+        styles, all_personas, all_cooperations, all_comments, all_reviews, all_creatives
+    )
+    enrichment["dashboard"] = build_dashboard_facts(
+        report_date.isoformat(), tracking_rows, all_platform_metrics, all_personas, connections,
+        style_images=style_images,
+        daily_notes=compact_yesterday,
+        style_categories=style_categories,
+        style_signals=style_signals,
+    )
+    return compact_yesterday, tracking_rows, next_day_rows, _comment_rows(yesterday_notes), enrichment
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="按款式生成全域种草日报")
+    parser.add_argument("--date", help="报告日期 YYYY-MM-DD；默认取款式日报最新日期")
+    parser.add_argument("--prompt-only", action="store_true", help="只输出给 Hermes 的提示词")
+    parser.add_argument("--send", action="store_true", help="生成后以飞书卡片发送给负责人")
+    parser.add_argument("--open-id", action="append", dest="open_ids", help="覆盖默认飞书负责人，可重复传入")
+    args = parser.parse_args()
+
+    report_date = date.fromisoformat(args.date) if args.date else _latest_report_date()
+    facts = load_report_facts(report_date)
+    prompt = build_daily_prompt(report_date.isoformat(), *facts)
+    if args.prompt_only:
+        print(prompt)
+        return 0
+
+    raw_report = require_valid_hermes_report(call_hermes_analyzer(SYSTEM_PROMPT, prompt))
+    report = normalize_report(raw_report, total_styles=len(facts[1]))
+    output_dir = PATHS.exports_root / "summary"
+    output_dir.mkdir(parents=True, exist_ok=True)
+    chart_path = output_dir / "charts" / f"daily_marketing_dashboard_{report_date.isoformat()}.png"
+    generate_dashboard(report_date.isoformat(), facts[1], facts[4], chart_path)
+    chart_ref = chart_path.relative_to(output_dir).as_posix()
+    report = f"![全域种草数据看板]({chart_ref})\n\n" + report
+    output_path = output_dir / f"daily_marketing_report_{report_date.isoformat()}.md"
+    output_path.write_text(report, encoding="utf-8-sig")
+    print(f"日报已生成: {output_path}")
+
+    if args.send:
+        recipients = tuple(args.open_ids or DEFAULT_RECIPIENT_OPEN_IDS)
+        results = send_daily_report_cards(
+            report_date.isoformat(), report, facts[1], chart_path, recipients
+        )
+        message_ids = [item.get("message_id") or (item.get("data") or {}).get("message_id") for item in results]
+        print(
+            f"日报已按看板、完整分析各 1 条发送给 {len(recipients)} 位负责人，"
+            f"共 {len(results)} 条消息: {', '.join(value or '未知消息ID' for value in message_ids)}"
+        )
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/config/analyze.env.example b/src/gyxx_flow/modules/content_marketing/runtime/data/config/analyze.env.example
new file mode 100644
index 0000000..a1e596a
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/config/analyze.env.example
@@ -0,0 +1,8 @@
+# 分析脚本配置模板
+# 复制为 analyze.env 并填真实值 (analyze.env 已被 .gitignore 排除)
+
+# Hermes 分析端 API Token
+HERMES_ANALYZER_TOKEN=your_token_here
+
+# 飞书接收人 open_id (不设则必须传 --no-send)
+FEISHU_TARGET_OPEN_ID=ou_xxxxxxxxxxxxxxxxx
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/config/db.env.example b/src/gyxx_flow/modules/content_marketing/runtime/data/config/db.env.example
new file mode 100644
index 0000000..5b1c1fc
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/config/db.env.example
@@ -0,0 +1,8 @@
+# PG 连接信息模板
+# 复制为 db.env 并填真实值 (db.env 已被 .gitignore 排除)
+
+PG_HOST=8.148.185.119
+PG_PORT=5432
+PG_DB=data_hub
+PG_USER=data_hub
+PG_PASSWORD=${GYXX_PG_PASSWORD}
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/config/款式_多维表格_对照.json b/src/gyxx_flow/modules/content_marketing/runtime/data/config/款式_多维表格_对照.json
new file mode 100644
index 0000000..192a6da
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/config/款式_多维表格_对照.json
@@ -0,0 +1,6154 @@
+{
+  "_comment": "款式 → 多维表格 对照表(合作达人); 由 feishu_mapping.py 从合作达人地址表动态生成。 field_map 给出本表核心字段的 field_id (实时从飞书拉取)",
+  "platform": "xhs_pgy",
+  "tables": [
+    {
+      "index": 2,
+      "name": "极星pro",
+      "base_token": "OsYvbUe1iaJYeKs5WvLcU4rVnHc",
+      "table_id": "tblASZDx24q5Bjwy",
+      "url": "https://bu0zgpibak.feishu.cn/base/OsYvbUe1iaJYeKs5WvLcU4rVnHc?table=tblASZDx24q5Bjwy&view=vewStRNML4",
+      "field_map": {
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "note_title": {
+          "field_id": "fldFCZ5hyJ",
+          "field_name": "发布笔记标题"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "creator_id": {
+          "field_id": "fld3VIRZMM",
+          "field_name": "达人id"
+        },
+        "publish_time": {
+          "field_id": "fldFJCKY20",
+          "field_name": "发布时间"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        }
+      },
+      "all_field_names": [
+        "审稿时间",
+        "发布笔记内容方向",
+        "发布笔记链接",
+        "投流金额",
+        "投放账号类型",
+        "制单日期",
+        "互动赞藏数",
+        "120天曝光量",
+        "主页链接",
+        "是否结款",
+        "数据表现",
+        "达人粉丝",
+        "月份",
+        "是否新方向验证",
+        "14天曝光量",
+        "合作费用",
+        "发布笔记标题",
+        "月底曝光量",
+        "180天曝光量",
+        "快递单号",
+        "父记录",
+        "曝光量（自动抓取）",
+        "优秀数据（自动抓取）",
+        "60天曝光量",
+        "150天曝光量",
+        "达人名称",
+        "观看画像",
+        "28天曝光量",
+        "90天曝光量",
+        "投放平台",
+        "达人id",
+        "发布时间",
+        "cpm",
+        "7天曝光量",
+        "周发布",
+        "21天曝光量",
+        "微信号"
+      ],
+      "total_records": 116
+    },
+    {
+      "index": 3,
+      "name": "盖亚微单",
+      "base_token": "Qs4mbzyAaaQ8SasnDpkc9lA5npe",
+      "table_id": "tblMok3p5E6QydPq",
+      "url": "https://bu0zgpibak.feishu.cn/base/Qs4mbzyAaaQ8SasnDpkc9lA5npe?table=tblMok3p5E6QydPq&view=vewStRNML4",
+      "field_map": {
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "note_title": {
+          "field_id": "fld90ye3vc",
+          "field_name": "发布笔记标题"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "creator_id": {
+          "field_id": "fld8yeUDM7",
+          "field_name": "达人id"
+        },
+        "publish_time": {
+          "field_id": "fldFJCKY20",
+          "field_name": "发布时间"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        }
+      },
+      "all_field_names": [
+        "观看画像",
+        "字段 1",
+        "主页链接",
+        "是否结款",
+        "字段 8",
+        "月份",
+        "审稿时间",
+        "互动赞藏数",
+        "达人粉丝",
+        "字段 4",
+        "14天曝光量",
+        "快递单号",
+        "发布笔记标题",
+        "字段 14",
+        "字段 3",
+        "投放账号类型",
+        "父记录",
+        "21天曝光量",
+        "发布笔记链接",
+        "月底曝光量",
+        "字段 10",
+        "cpm",
+        "达人名称",
+        "字段 7",
+        "数据表现",
+        "是否新方向验证",
+        "投放平台",
+        "合作费用",
+        "120天曝光量",
+        "投流金额",
+        "150天曝光量",
+        "曝光量（自动抓取）",
+        "优秀数据（自动抓取）",
+        "60天曝光量",
+        "字段 6",
+        "字段 12",
+        "180天曝光量",
+        "28天曝光量",
+        "字段 13",
+        "字段 2",
+        "字段 5",
+        "周发布",
+        "字段 11",
+        "90天曝光量",
+        "发布笔记内容方向",
+        "达人id",
+        "字段 9",
+        "发布时间",
+        "7天曝光量",
+        "制单日期",
+        "微信号"
+      ],
+      "total_records": 114
+    },
+    {
+      "index": 4,
+      "name": "盖世m1",
+      "base_token": "KnBRbH9iQaH3PAsAUk5cHLxNnMg",
+      "table_id": "tblcZ8y5126gTHe7",
+      "url": "https://bu0zgpibak.feishu.cn/base/KnBRbH9iQaH3PAsAUk5cHLxNnMg?table=tblcZ8y5126gTHe7&view=vewStRNML4",
+      "field_map": {
+        "audience_profile": {
+          "field_id": "fldZsHi8gG",
+          "field_name": "观看画像 (1)"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldM237xjc",
+          "field_name": "150天曝光量 (1)"
+        },
+        "data_performance": {
+          "field_id": "fldMghLJV8",
+          "field_name": "数据表现 (1)"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fld8XqHIuC",
+          "field_name": "180天曝光量 (1)"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "creator_id": {
+          "field_id": "fldexnbAXv",
+          "field_name": "达人id"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "publish_time": {
+          "field_id": "fldJgCDvNj",
+          "field_name": "发布时间"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "note_title": {
+          "field_id": "fld1OJ6D4X",
+          "field_name": "发布笔记标题"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        }
+      },
+      "all_field_names": [
+        "观看画像 (1)",
+        "28天曝光量",
+        "150天曝光量 (1)",
+        "数据表现 (1)",
+        "投流金额",
+        "曝光量（自动抓取）",
+        "周发布",
+        "60天曝光量",
+        "180天曝光量 (1)",
+        "月底曝光量",
+        "投流金额 (1)",
+        "是否新方向验证",
+        "达人id",
+        "优秀数据（自动抓取）",
+        "120天曝光量",
+        "150天曝光量",
+        "投放账号类型",
+        "7天曝光量",
+        "制单日期",
+        "60天曝光量 (1)",
+        "90天曝光量",
+        "14天曝光量",
+        "互动赞藏数",
+        "cpm",
+        "是否结款",
+        "90天曝光量 (1)",
+        "曝光量（自动抓取） (1)",
+        "观看画像",
+        "投放平台",
+        "主页链接",
+        "快递单号",
+        "微信号",
+        "21天曝光量",
+        "数据表现",
+        "cpm (1)",
+        "达人名称",
+        "月底曝光量 (1)",
+        "合作费用",
+        "发布时间",
+        "父记录",
+        "优秀数据（自动抓取） (1)",
+        "月份",
+        "审稿时间",
+        "达人粉丝",
+        "180天曝光量",
+        "发布笔记标题",
+        "120天曝光量 (1)",
+        "发布笔记链接",
+        "发布笔记内容方向"
+      ],
+      "total_records": 107
+    },
+    {
+      "index": 5,
+      "name": "星迹2",
+      "base_token": "Ssp8bXmrwahRBbsko2BcuiLanad",
+      "table_id": "tblqNsVvKj4EPyzA",
+      "url": "https://bu0zgpibak.feishu.cn/base/Ssp8bXmrwahRBbsko2BcuiLanad?table=tblqNsVvKj4EPyzA&view=vewStRNML4",
+      "field_map": {
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "creator_id": {
+          "field_id": "fldLZO2Em7",
+          "field_name": "达人id"
+        },
+        "read_count_180d": {
+          "field_id": "fldSwdYx2M",
+          "field_name": "180天曝光量 (1)"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "publish_time": {
+          "field_id": "fldrCfjoir",
+          "field_name": "发布时间"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "note_title": {
+          "field_id": "fldYb2Wsil",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_150d": {
+          "field_id": "fld74uraM6",
+          "field_name": "150天曝光量 (1)"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        }
+      },
+      "all_field_names": [
+        "制单日期",
+        "120天曝光量",
+        "达人id",
+        "180天曝光量 (1)",
+        "达人名称",
+        "60天曝光量",
+        "达人粉丝",
+        "是否新方向验证",
+        "7天曝光量",
+        "曝光量（自动抓取）",
+        "周发布",
+        "互动赞藏数",
+        "快递单号",
+        "120天曝光量 (1)",
+        "投放平台",
+        "父记录",
+        "发布时间",
+        "90天曝光量",
+        "14天曝光量",
+        "发布笔记标题",
+        "150天曝光量 (1)",
+        "180天曝光量",
+        "发布笔记内容方向",
+        "150天曝光量",
+        "主页链接",
+        "数据表现",
+        "21天曝光量",
+        "父记录 (2)",
+        "合作费用",
+        "150天曝光量 (2)",
+        "28天曝光量",
+        "父记录 (1)",
+        "优秀数据（自动抓取）",
+        "月底曝光量",
+        "微信号",
+        "cpm",
+        "投放账号类型",
+        "180天曝光量 (2)",
+        "投流金额",
+        "审稿时间",
+        "月份",
+        "观看画像",
+        "发布笔记链接",
+        "是否结款"
+      ],
+      "total_records": 225
+    },
+    {
+      "index": 6,
+      "name": "星云2",
+      "base_token": "BJaWbLEOKaHONMshOidcxvsKngh",
+      "table_id": "tblR0tCfPNk34ULj",
+      "url": "https://bu0zgpibak.feishu.cn/base/BJaWbLEOKaHONMshOidcxvsKngh?table=tblR0tCfPNk34ULj&view=vewStRNML4",
+      "field_map": {
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldzihhShx",
+          "field_name": "发布时间"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "creator_id": {
+          "field_id": "fldhjwoNzN",
+          "field_name": "达人id"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "note_title": {
+          "field_id": "fldoe9ysE5",
+          "field_name": "发布笔记标题"
+        }
+      },
+      "all_field_names": [
+        "120天曝光量",
+        "达人名称",
+        "达人粉丝",
+        "主页链接",
+        "180天曝光量",
+        "投流金额",
+        "制单日期",
+        "数据表现",
+        "21天曝光量",
+        "cpm",
+        "发布笔记内容方向",
+        "观看画像",
+        "90天曝光量",
+        "14天曝光量",
+        "月份",
+        "曝光量（自动抓取）",
+        "父记录",
+        "是否结款",
+        "150天曝光量",
+        "60天曝光量",
+        "发布时间",
+        "互动赞藏数",
+        "发布笔记链接",
+        "是否新方向验证",
+        "投放平台",
+        "投放账号类型",
+        "优秀数据（自动抓取）",
+        "审稿时间",
+        "周发布",
+        "合作费用",
+        "达人id",
+        "月底曝光量",
+        "微信号",
+        "7天曝光量",
+        "快递单号",
+        "28天曝光量",
+        "发布笔记标题"
+      ],
+      "total_records": 225
+    },
+    {
+      "index": 7,
+      "name": "宙斯",
+      "base_token": "N43abi7IlaS8XosnUXec6gH2n7g",
+      "table_id": "tblasQwwbXwpGbpL",
+      "url": "https://bu0zgpibak.feishu.cn/base/N43abi7IlaS8XosnUXec6gH2n7g?table=tblasQwwbXwpGbpL&view=vewStRNML4",
+      "field_map": {
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "creator_id": {
+          "field_id": "fldOnDzXnc",
+          "field_name": "达人id"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "note_title": {
+          "field_id": "fldJr5VXwA",
+          "field_name": "笔记标题"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        }
+      },
+      "all_field_names": [
+        "主页链接",
+        "发布时间",
+        "制单日期",
+        "投放账号类型",
+        "21天曝光量",
+        "cpm",
+        "投放平台",
+        "发布笔记链接",
+        "投流金额",
+        "150天曝光量",
+        "达人粉丝",
+        "快递单号",
+        "合作费用",
+        "观看画像",
+        "28天曝光量",
+        "父记录",
+        "周发布",
+        "微信号",
+        "7天曝光量",
+        "曝光量（自动抓取）",
+        "月底曝光量",
+        "14天曝光量",
+        "优秀数据（自动抓取）",
+        "是否结款",
+        "120天曝光量",
+        "90天曝光量",
+        "数据表现",
+        "达人id",
+        "审稿时间",
+        "180天曝光量",
+        "发布笔记内容方向",
+        "笔记标题",
+        "60天曝光量",
+        "是否新方向验证",
+        "达人名称",
+        "月份",
+        "互动赞藏数"
+      ],
+      "total_records": 43
+    },
+    {
+      "index": 8,
+      "name": "阿波罗2",
+      "base_token": "W80vbtPDoaaHTOs5Tipc6CTLn6c",
+      "table_id": "tblUaRl444G5IhVY",
+      "url": "https://bu0zgpibak.feishu.cn/base/W80vbtPDoaaHTOs5Tipc6CTLn6c?table=tblUaRl444G5IhVY&view=vewStRNML4",
+      "field_map": {
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "note_title": {
+          "field_id": "fldHElhugy",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "creator_id": {
+          "field_id": "fldoxiPZ3u",
+          "field_name": "达人id"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        }
+      },
+      "all_field_names": [
+        "投放平台",
+        "发布笔记标题",
+        "90天曝光量",
+        "28天曝光量",
+        "制单日期",
+        "数据表现",
+        "审稿时间",
+        "是否结款",
+        "达人粉丝",
+        "发布笔记内容方向",
+        "是否新方向验证",
+        "优秀数据（自动抓取）",
+        "月底曝光量",
+        "曝光量（自动抓取）",
+        "合作费用",
+        "发布笔记链接",
+        "7天曝光量",
+        "150天曝光量",
+        "主页链接",
+        "180天曝光量",
+        "父记录",
+        "月份",
+        "快递单号",
+        "周发布",
+        "21天曝光量",
+        "达人id",
+        "微信号",
+        "观看画像",
+        "60天曝光量",
+        "120天曝光量",
+        "14天曝光量",
+        "达人名称",
+        "发布时间",
+        "投放账号类型",
+        "互动赞藏数",
+        "cpm",
+        "投流金额"
+      ],
+      "total_records": 23
+    },
+    {
+      "index": 9,
+      "name": "阿波罗x1",
+      "base_token": "CtW5bUcc9a5rqRsZkwucUWDVnHb",
+      "table_id": "tbl8WgoFrOjkqOD5",
+      "url": "https://bu0zgpibak.feishu.cn/base/CtW5bUcc9a5rqRsZkwucUWDVnHb?table=tbl8WgoFrOjkqOD5&view=vewStRNML4",
+      "field_map": {
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "creator_id": {
+          "field_id": "fldzpKQRft",
+          "field_name": "达人id"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "note_title": {
+          "field_id": "fldXV8QHp3",
+          "field_name": "发布笔记标题"
+        }
+      },
+      "all_field_names": [
+        "月份",
+        "观看画像",
+        "周发布",
+        "60天曝光量",
+        "快递单号",
+        "21天曝光量",
+        "150天曝光量",
+        "14天曝光量",
+        "数据表现",
+        "28天曝光量",
+        "投放平台",
+        "曝光量（自动抓取）",
+        "投放账号类型",
+        "是否新方向验证",
+        "投流金额",
+        "cpm",
+        "月底曝光量",
+        "达人名称",
+        "发布笔记链接",
+        "180天曝光量",
+        "合作费用",
+        "审稿时间",
+        "发布笔记内容方向",
+        "主页链接",
+        "是否结款",
+        "父记录",
+        "制单日期",
+        "优秀数据（自动抓取）",
+        "120天曝光量",
+        "互动赞藏数",
+        "达人id",
+        "微信号",
+        "7天曝光量",
+        "发布时间",
+        "达人粉丝",
+        "90天曝光量",
+        "发布笔记标题"
+      ],
+      "total_records": 31
+    },
+    {
+      "index": 10,
+      "name": "波塞冬edge",
+      "base_token": "JsVLbso8Ba6RjbslOl1cJX8lnyb",
+      "table_id": "tblwqyhbymn9Aour",
+      "url": "https://bu0zgpibak.feishu.cn/base/JsVLbso8Ba6RjbslOl1cJX8lnyb?table=tblwqyhbymn9Aour&view=vewStRNML4",
+      "field_map": {
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "note_title": {
+          "field_id": "fldt8vGXO9",
+          "field_name": "发布笔记标题"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "publish_time": {
+          "field_id": "fldbKALsn7",
+          "field_name": "发布时间"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "creator_id": {
+          "field_id": "fldfSnoILe",
+          "field_name": "达人id"
+        }
+      },
+      "all_field_names": [
+        "制单日期",
+        "快递单号",
+        "观看画像",
+        "90天曝光量",
+        "优秀数据（自动抓取）",
+        "28天曝光量",
+        "投放平台",
+        "月份",
+        "发布笔记标题",
+        "主页链接",
+        "发布笔记内容方向",
+        "cpm",
+        "曝光量（自动抓取）",
+        "21天曝光量",
+        "审稿时间",
+        "周发布",
+        "达人粉丝",
+        "发布时间",
+        "合作费用",
+        "发布笔记链接",
+        "月底曝光量",
+        "互动赞藏数",
+        "180天曝光量",
+        "数据表现",
+        "是否结款",
+        "150天曝光量",
+        "投流金额",
+        "14天曝光量",
+        "60天曝光量",
+        "7天曝光量",
+        "120天曝光量",
+        "是否新方向验证",
+        "微信号",
+        "达人名称",
+        "投放账号类型",
+        "父记录",
+        "达人id"
+      ],
+      "total_records": 19
+    },
+    {
+      "index": 11,
+      "name": "逐星",
+      "base_token": "IQrbbxAHpa9oiWs631GcPkHznob",
+      "table_id": "tblmWRjVp0wQiVeG",
+      "url": "https://bu0zgpibak.feishu.cn/base/IQrbbxAHpa9oiWs631GcPkHznob?table=tblmWRjVp0wQiVeG&view=vewStRNML4",
+      "field_map": {
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "creator_id": {
+          "field_id": "fldM2JQukr",
+          "field_name": "达人id"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "note_title": {
+          "field_id": "fldlbWqQw3",
+          "field_name": "发布笔记标题"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        }
+      },
+      "all_field_names": [
+        "主页链接",
+        "达人粉丝",
+        "发布时间",
+        "周发布",
+        "7天曝光量",
+        "60天曝光量",
+        "合作费用",
+        "达人名称",
+        "发布笔记链接",
+        "审稿时间",
+        "180天曝光量",
+        "优秀数据（自动抓取）",
+        "cpm",
+        "是否新方向验证",
+        "21天曝光量",
+        "月份",
+        "曝光量（自动抓取）",
+        "投放账号类型",
+        "观看画像",
+        "快递单号",
+        "150天曝光量",
+        "投流金额",
+        "达人id",
+        "父记录",
+        "120天曝光量",
+        "投放平台",
+        "月底曝光量",
+        "微信号",
+        "互动赞藏数",
+        "制单日期",
+        "数据表现",
+        "14天曝光量",
+        "90天曝光量",
+        "是否结款",
+        "发布笔记标题",
+        "发布笔记内容方向",
+        "28天曝光量"
+      ],
+      "total_records": 25
+    },
+    {
+      "index": 12,
+      "name": "逐星GT",
+      "base_token": "MAEZbKrNaabJbDs5ZY7caq1ynrd",
+      "table_id": "tblcT5gqVRsPBuOE",
+      "url": "https://bu0zgpibak.feishu.cn/base/MAEZbKrNaabJbDs5ZY7caq1ynrd?table=tblcT5gqVRsPBuOE&view=vewStRNML4",
+      "field_map": {
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "note_title": {
+          "field_id": "fldY3xFUfT",
+          "field_name": "发布笔记标题"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "creator_id": {
+          "field_id": "fldquB6Bx1",
+          "field_name": "达人id"
+        }
+      },
+      "all_field_names": [
+        "投流金额",
+        "优秀数据（自动抓取）",
+        "互动赞藏数",
+        "120天曝光量",
+        "180天曝光量",
+        "7天曝光量",
+        "90天曝光量",
+        "周发布",
+        "审稿时间",
+        "达人名称",
+        "28天曝光量",
+        "发布笔记链接",
+        "是否结款",
+        "发布笔记内容方向",
+        "数据表现",
+        "合作费用",
+        "60天曝光量",
+        "发布笔记标题",
+        "快递单号",
+        "制单日期",
+        "达人粉丝",
+        "投放账号类型",
+        "投放平台",
+        "观看画像",
+        "月份",
+        "21天曝光量",
+        "150天曝光量",
+        "主页链接",
+        "cpm",
+        "是否新方向验证",
+        "发布时间",
+        "微信号",
+        "曝光量（自动抓取）",
+        "月底曝光量",
+        "14天曝光量",
+        "达人id"
+      ],
+      "total_records": 16
+    },
+    {
+      "index": 13,
+      "name": "星云mini",
+      "base_token": "RKfkb5JkIa9FttsgHvNc2nJYnbf",
+      "table_id": "tblu1yXKkQqwje0d",
+      "url": "https://bu0zgpibak.feishu.cn/base/RKfkb5JkIa9FttsgHvNc2nJYnbf?table=tblu1yXKkQqwje0d&view=vewStRNML4",
+      "field_map": {
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "publish_time": {
+          "field_id": "fldrx09tSe",
+          "field_name": "发布时间"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "note_title": {
+          "field_id": "fld8X4rGRI",
+          "field_name": "标题"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "creator_id": {
+          "field_id": "fldDWpeHbx",
+          "field_name": "达人id"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        }
+      },
+      "all_field_names": [
+        "7天曝光量",
+        "120天曝光量",
+        "180天曝光量",
+        "父记录",
+        "投放账号类型",
+        "是否新方向验证",
+        "是否结款",
+        "cpm",
+        "数据表现",
+        "投流金额",
+        "月份",
+        "21天曝光量",
+        "发布笔记链接",
+        "制单日期",
+        "审稿时间",
+        "发布时间",
+        "14天曝光量",
+        "周发布",
+        "标题",
+        "达人名称",
+        "投放平台",
+        "曝光量（自动抓取）",
+        "发布笔记内容方向",
+        "28天曝光量",
+        "优秀数据（自动抓取）",
+        "月底曝光量",
+        "快递单号",
+        "达人id",
+        "90天曝光量",
+        "微信号",
+        "主页链接",
+        "150天曝光量",
+        "达人粉丝",
+        "互动赞藏数",
+        "合作费用",
+        "60天曝光量",
+        "观看画像"
+      ],
+      "total_records": 10
+    },
+    {
+      "index": 14,
+      "name": "晨星2",
+      "base_token": "VrLtbkldsaggOms3F4Bcu3DinCg",
+      "table_id": "tblO5hzLek6Nk43J",
+      "url": "https://bu0zgpibak.feishu.cn/base/VrLtbkldsaggOms3F4Bcu3DinCg?table=tblO5hzLek6Nk43J&view=vewStRNML4",
+      "field_map": {
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "note_title": {
+          "field_id": "fld4fjsqT7",
+          "field_name": "标题"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "publish_time": {
+          "field_id": "fldSwVB60U",
+          "field_name": "发布时间"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "creator_id": {
+          "field_id": "fld46ENoIN",
+          "field_name": "达人id"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        }
+      },
+      "all_field_names": [
+        "发布笔记链接",
+        "月份",
+        "达人名称",
+        "快递单号",
+        "制单日期",
+        "60天曝光量",
+        "曝光量（自动抓取）",
+        "观看画像",
+        "90天曝光量",
+        "28天曝光量",
+        "合作费用",
+        "180天曝光量",
+        "投放平台",
+        "主页链接",
+        "7天曝光量",
+        "标题",
+        "150天曝光量",
+        "周发布",
+        "投流金额",
+        "是否结款",
+        "120天曝光量",
+        "投放账号类型",
+        "cpm",
+        "发布时间",
+        "发布笔记内容方向",
+        "达人id",
+        "数据表现",
+        "月底曝光量",
+        "微信号",
+        "达人粉丝",
+        "审稿时间",
+        "是否新方向验证",
+        "优秀数据（自动抓取）",
+        "14天曝光量",
+        "21天曝光量",
+        "互动赞藏数"
+      ],
+      "total_records": 8
+    },
+    {
+      "index": 15,
+      "name": "极星双肩",
+      "base_token": "GubZbTrMSahLg5sdonIc8kR8nZe",
+      "table_id": "tbl1tl5CI0LhKeuk",
+      "url": "https://bu0zgpibak.feishu.cn/base/GubZbTrMSahLg5sdonIc8kR8nZe?table=tbl1tl5CI0LhKeuk&view=vewStRNML4",
+      "field_map": {
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "creator_name": {
+          "field_id": "fld5fK7zgT",
+          "field_name": "达人昵称"
+        },
+        "note_title": {
+          "field_id": "fldeXdeRRH",
+          "field_name": "发布笔记标题"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "creator_id": {
+          "field_id": "fldBkC1RxH",
+          "field_name": "达人id"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        }
+      },
+      "all_field_names": [
+        "微信号",
+        "是否结款",
+        "60天曝光量",
+        "月底曝光量",
+        "发布笔记链接",
+        "观看画像",
+        "父记录",
+        "数据表现",
+        "达人昵称",
+        "发布笔记标题",
+        "快递单号",
+        "周发布",
+        "28天曝光量",
+        "发布时间",
+        "是否新方向验证",
+        "21天曝光量",
+        "制单日期",
+        "cpm",
+        "达人粉丝",
+        "120天曝光量",
+        "时间",
+        "合作费用",
+        "投放账号类型",
+        "发布笔记内容方向",
+        "优秀数据（自动抓取）",
+        "14天曝光量",
+        "互动赞藏数",
+        "投放平台",
+        "审稿时间",
+        "90天曝光量",
+        "曝光量（自动抓取）",
+        "主页链接",
+        "150天曝光量",
+        "投流金额",
+        "7天曝光量",
+        "达人id",
+        "180天曝光量"
+      ],
+      "total_records": 30
+    },
+    {
+      "index": 16,
+      "name": "凌云air",
+      "base_token": "VwpibmlakaqHuLsGChccMb3BnNd",
+      "table_id": "tbl7fPd2dMFKQdih",
+      "url": "https://bu0zgpibak.feishu.cn/base/VwpibmlakaqHuLsGChccMb3BnNd?table=tbl7fPd2dMFKQdih&view=vewStRNML4",
+      "field_map": {
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "note_title": {
+          "field_id": "fldVpDYo8z",
+          "field_name": "笔记标题"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldFJCKY20",
+          "field_name": "发布时间"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "creator_id": {
+          "field_id": "fld7ojkeyw",
+          "field_name": "达人id"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        }
+      },
+      "all_field_names": [
+        "审稿时间",
+        "笔记标题",
+        "优秀数据（自动抓取）",
+        "180天曝光量",
+        "合作费用",
+        "是否新方向验证",
+        "投放账号类型",
+        "21天曝光量",
+        "发布时间",
+        "投流金额",
+        "120天曝光量",
+        "快递单号",
+        "投放平台",
+        "月份",
+        "28天曝光量",
+        "周发布",
+        "主页链接",
+        "14天曝光量",
+        "数据表现",
+        "曝光量（自动抓取）",
+        "父记录 2",
+        "发布笔记链接",
+        "微信号",
+        "7天曝光量",
+        "父记录",
+        "cpm",
+        "制单日期",
+        "150天曝光量",
+        "互动赞藏数",
+        "达人粉丝",
+        "是否结款",
+        "月底曝光量",
+        "发布笔记内容方向",
+        "达人id",
+        "达人名称",
+        "观看画像"
+      ],
+      "total_records": 114
+    },
+    {
+      "index": 17,
+      "name": "觅光",
+      "base_token": "CQcibpRjgaUCg7sUTMwcnrhunMh",
+      "table_id": "tblFHe5uONCWxbEh",
+      "url": "https://bu0zgpibak.feishu.cn/base/CQcibpRjgaUCg7sUTMwcnrhunMh?table=tblFHe5uONCWxbEh&view=vewStRNML4",
+      "field_map": {
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldFJCKY20",
+          "field_name": "发布时间"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "note_title": {
+          "field_id": "fldMaMXbZX",
+          "field_name": "发布笔记标题"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "creator_id": {
+          "field_id": "fldpI6HEEy",
+          "field_name": "达人id"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        }
+      },
+      "all_field_names": [
+        "快递单号",
+        "父记录",
+        "微信号",
+        "月份",
+        "投流金额",
+        "主页链接",
+        "21天曝光量",
+        "月底曝光量",
+        "发布时间",
+        "达人粉丝",
+        "投放平台",
+        "是否新方向验证",
+        "优秀数据（自动抓取）",
+        "发布笔记内容方向",
+        "达人名称",
+        "制单日期",
+        "周发布",
+        "发布笔记链接",
+        "是否结款",
+        "发布笔记标题",
+        "审稿时间",
+        "cpm",
+        "合作费用",
+        "120天曝光量",
+        "互动赞藏数",
+        "14天曝光量",
+        "达人id",
+        "28天曝光量",
+        "150天曝光量",
+        "180天曝光量",
+        "数据表现",
+        "7天曝光量",
+        "曝光量（自动抓取）",
+        "观看画像",
+        "投放账号类型"
+      ],
+      "total_records": 13
+    },
+    {
+      "index": 18,
+      "name": "凌云3",
+      "base_token": "XtHjbqpwcab3J4sWP6Fcfgh5nYt",
+      "table_id": "tblqRgK6MI5C1H4K",
+      "url": "https://bu0zgpibak.feishu.cn/base/XtHjbqpwcab3J4sWP6Fcfgh5nYt?table=tblqRgK6MI5C1H4K&view=vewStRNML4",
+      "field_map": {
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "note_title": {
+          "field_id": "fldM4LJVt9",
+          "field_name": "笔记标题"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "creator_id": {
+          "field_id": "fldtWOK28M",
+          "field_name": "达人id"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        }
+      },
+      "all_field_names": [
+        "投流金额",
+        "笔记标题",
+        "cpm",
+        "发布笔记内容方向",
+        "数据表现",
+        "是否结款",
+        "投放平台",
+        "曝光量（自动抓取）",
+        "达人粉丝",
+        "90天曝光量",
+        "120天曝光量",
+        "是否新方向验证",
+        "达人名称",
+        "7天曝光量",
+        "21天曝光量",
+        "主页链接",
+        "14天曝光量",
+        "发布笔记链接",
+        "月份",
+        "优秀数据（自动抓取）",
+        "快递单号",
+        "父记录",
+        "月底曝光量",
+        "150天曝光量",
+        "审稿时间",
+        "投放账号类型",
+        "制单日期",
+        "180天曝光量",
+        "发布时间",
+        "合作费用",
+        "达人id",
+        "60天曝光量",
+        "28天曝光量",
+        "观看画像",
+        "周发布",
+        "互动赞藏数",
+        "微信号"
+      ],
+      "total_records": 23
+    },
+    {
+      "index": 19,
+      "name": "瑞白",
+      "base_token": "GNOkbKNpna8pDJsLMfzcuZt3ncf",
+      "table_id": "tbljwMiRxb4SlpGr",
+      "url": "https://bu0zgpibak.feishu.cn/base/GNOkbKNpna8pDJsLMfzcuZt3ncf?table=tbljwMiRxb4SlpGr&view=vewStRNML4",
+      "field_map": {
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "publish_time": {
+          "field_id": "fldFJCKY20",
+          "field_name": "发布时间"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "creator_id": {
+          "field_id": "fldqFYMjm1",
+          "field_name": "达人id"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "note_title": {
+          "field_id": "fld1sZPxKx",
+          "field_name": "发布笔记标题"
+        }
+      },
+      "all_field_names": [
+        "投流金额",
+        "月底曝光量",
+        "合作费用",
+        "发布笔记链接",
+        "是否结款",
+        "制单日期",
+        "数据表现",
+        "投放账号类型",
+        "优秀数据（自动抓取）",
+        "互动赞藏数",
+        "达人名称",
+        "发布时间",
+        "cpm",
+        "是否新方向验证",
+        "达人id",
+        "120天曝光量",
+        "180天曝光量",
+        "150天曝光量",
+        "21天曝光量",
+        "主页链接",
+        "投放平台",
+        "发布笔记内容方向",
+        "微信号",
+        "父记录",
+        "快递单号",
+        "14天曝光量",
+        "达人粉丝",
+        "曝光量（自动抓取）",
+        "观看画像",
+        "月份",
+        "周发布",
+        "审稿时间",
+        "7天曝光量",
+        "28天曝光量",
+        "发布笔记标题"
+      ],
+      "total_records": 115
+    },
+    {
+      "index": 20,
+      "name": "黑曜石吐司包",
+      "base_token": "U2JGbbWcIal6WvsxURRcTMwEnhf",
+      "table_id": "tblVEdURLsfRZykJ",
+      "url": "https://bu0zgpibak.feishu.cn/base/U2JGbbWcIal6WvsxURRcTMwEnhf?table=tblVEdURLsfRZykJ&view=vewStRNML4",
+      "field_map": {
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "creator_id": {
+          "field_id": "fldUODpzLv",
+          "field_name": "达人id"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_150d": {
+          "field_id": "fldIG5lGAC",
+          "field_name": "150天曝光量 (1)"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "publish_time": {
+          "field_id": "fldj8t4N9l",
+          "field_name": "发布时间"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldFj5ZCyH",
+          "field_name": "180天曝光量 (1)"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "note_title": {
+          "field_id": "fldvdhZ7If",
+          "field_name": "发布笔记标题"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        }
+      },
+      "all_field_names": [
+        "投放账号类型",
+        "合作费用",
+        "90天曝光量",
+        "投放平台",
+        "120天曝光量",
+        "互动赞藏数",
+        "120天曝光量 (1)",
+        "主页链接",
+        "优秀数据（自动抓取）",
+        "周发布",
+        "cpm",
+        "达人id",
+        "60天曝光量",
+        "发布笔记链接",
+        "微信号",
+        "父记录 (1)",
+        "150天曝光量 (1)",
+        "曝光量（自动抓取）",
+        "月份",
+        "发布笔记内容方向",
+        "21天曝光量",
+        "制单日期",
+        "审稿时间",
+        "父记录",
+        "观看画像",
+        "发布时间",
+        "月底曝光量",
+        "快递单号",
+        "达人粉丝",
+        "28天曝光量",
+        "7天曝光量",
+        "180天曝光量 (1)",
+        "数据表现",
+        "是否新方向验证",
+        "150天曝光量",
+        "14天曝光量",
+        "达人名称",
+        "是否结款",
+        "180天曝光量",
+        "发布笔记标题",
+        "投流金额"
+      ],
+      "total_records": 108
+    },
+    {
+      "index": 21,
+      "name": "黑曜石托特包",
+      "base_token": "U9S8bLYsNa8Ul8sPffAcrJ02nHI",
+      "table_id": "tblhpEmlXGyqCGuh",
+      "url": "https://bu0zgpibak.feishu.cn/base/U9S8bLYsNa8Ul8sPffAcrJ02nHI?table=tblhpEmlXGyqCGuh&view=vewStRNML4",
+      "field_map": {
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "creator_id": {
+          "field_id": "fldEUmooW7",
+          "field_name": "达人id"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "note_title": {
+          "field_id": "fld0ExYDUi",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "publish_time": {
+          "field_id": "fldFJCKY20",
+          "field_name": "发布时间"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        }
+      },
+      "all_field_names": [
+        "120天曝光量",
+        "投流金额",
+        "互动赞藏数",
+        "是否结款",
+        "达人名称",
+        "优秀数据（自动抓取）",
+        "28天曝光量",
+        "投放账号类型",
+        "月底曝光量",
+        "达人id",
+        "曝光量（自动抓取）",
+        "审稿时间",
+        "21天曝光量",
+        "合作费用",
+        "达人粉丝",
+        "微信号",
+        "60天曝光量",
+        "cpm",
+        "制单日期",
+        "是否新方向验证",
+        "150天曝光量",
+        "发布笔记链接",
+        "发布笔记标题",
+        "180天曝光量",
+        "观看画像",
+        "主页链接",
+        "14天曝光量",
+        "周发布",
+        "快递单号",
+        "数据表现",
+        "7天曝光量",
+        "90天曝光量",
+        "月份",
+        "发布笔记内容方向",
+        "发布时间",
+        "投放平台"
+      ],
+      "total_records": 114
+    },
+    {
+      "index": 22,
+      "name": "曜影",
+      "base_token": "UPk3bR9MOaTkgrsG1qLc7IUvnQe",
+      "table_id": "tbllXFWRI59dLCM2",
+      "url": "https://bu0zgpibak.feishu.cn/base/UPk3bR9MOaTkgrsG1qLc7IUvnQe?table=tbllXFWRI59dLCM2&view=vewStRNML4",
+      "field_map": {
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "note_title": {
+          "field_id": "fldRGRSyDQ",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_180d": {
+          "field_id": "fldMPU9bVV",
+          "field_name": "180天曝光量 (2)"
+        },
+        "publish_time": {
+          "field_id": "fld5URo58N",
+          "field_name": "发布时间"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "creator_id": {
+          "field_id": "fldHQxxhQN",
+          "field_name": "达人id"
+        }
+      },
+      "all_field_names": [
+        "月份",
+        "发布笔记内容方向",
+        "60天曝光量",
+        "父记录 (1)",
+        "合作费用",
+        "14天曝光量",
+        "150天曝光量",
+        "观看画像",
+        "达人名称",
+        "曝光量（自动抓取）",
+        "周发布",
+        "是否结款",
+        "月底曝光量",
+        "发布笔记标题",
+        "180天曝光量 (2)",
+        "发布时间",
+        "父记录",
+        "150天曝光量 (2)",
+        "达人粉丝",
+        "微信号",
+        "数据表现",
+        "审稿时间",
+        "90天曝光量",
+        "互动赞藏数",
+        "优秀数据（自动抓取）",
+        "父记录 (2)",
+        "投流金额",
+        "180天曝光量",
+        "28天曝光量",
+        "150天曝光量 (1)",
+        "主页链接",
+        "cpm",
+        "120天曝光量",
+        "是否新方向验证",
+        "投放账号类型",
+        "7天曝光量",
+        "21天曝光量",
+        "快递单号",
+        "180天曝光量 (1)",
+        "投放平台",
+        "120天曝光量 (1)",
+        "制单日期",
+        "发布笔记链接",
+        "达人id"
+      ],
+      "total_records": 106
+    },
+    {
+      "index": 23,
+      "name": "黑曜石电脑包",
+      "base_token": "PwrGbXcpRaxJnOsd9cacFou8nAf",
+      "table_id": "tbludhnt5lkmdvSF",
+      "url": "https://bu0zgpibak.feishu.cn/base/PwrGbXcpRaxJnOsd9cacFou8nAf?table=tbludhnt5lkmdvSF&view=vewStRNML4",
+      "field_map": {
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "creator_id": {
+          "field_id": "fldI3Lu383",
+          "field_name": "达人id"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "publish_time": {
+          "field_id": "fldAIjQu8Z",
+          "field_name": "发布时间"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "note_title": {
+          "field_id": "fldbCKpQTV",
+          "field_name": "标题"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        }
+      },
+      "all_field_names": [
+        "投流金额",
+        "周发布",
+        "主页链接",
+        "达人id",
+        "发布笔记内容方向",
+        "150天曝光量",
+        "微信号",
+        "发布时间",
+        "达人名称",
+        "达人粉丝",
+        "14天曝光量",
+        "月底曝光量",
+        "是否新方向验证",
+        "审稿时间",
+        "合作费用",
+        "发布笔记链接",
+        "是否结款",
+        "快递单号",
+        "投放平台",
+        "投放账号类型",
+        "标题",
+        "互动赞藏数",
+        "数据表现",
+        "制单日期",
+        "90天曝光量",
+        "优秀数据（自动抓取）",
+        "28天曝光量",
+        "月份",
+        "180天曝光量",
+        "cpm",
+        "观看画像",
+        "21天曝光量",
+        "曝光量（自动抓取）",
+        "120天曝光量",
+        "60天曝光量",
+        "7天曝光量"
+      ],
+      "total_records": 7
+    },
+    {
+      "index": 24,
+      "name": "拓界2",
+      "base_token": "CNgPbjuXiaPoxPsShROcYbRDnbb",
+      "table_id": "tblEymEPk7lVaQN1",
+      "url": "https://bu0zgpibak.feishu.cn/base/CNgPbjuXiaPoxPsShROcYbRDnbb?table=tblEymEPk7lVaQN1&view=vewStRNML4",
+      "field_map": {
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "note_title": {
+          "field_id": "fldjPMPynt",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "creator_id": {
+          "field_id": "fldLyXQxoJ",
+          "field_name": "达人id"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        }
+      },
+      "all_field_names": [
+        "投流金额",
+        "28天曝光量",
+        "曝光量（自动抓取）",
+        "是否新方向验证",
+        "90天曝光量",
+        "发布时间",
+        "快递单号",
+        "发布笔记标题",
+        "7天曝光量",
+        "观看画像",
+        "150天曝光量",
+        "投放账号类型",
+        "达人id",
+        "投放平台",
+        "达人粉丝",
+        "微信号",
+        "周发布",
+        "21天曝光量",
+        "审稿时间",
+        "月份",
+        "数据表现",
+        "制单日期",
+        "主页链接",
+        "180天曝光量",
+        "是否结款",
+        "cpm",
+        "发布笔记内容方向",
+        "60天曝光量",
+        "发布笔记链接",
+        "合作费用",
+        "月底曝光量",
+        "达人名称",
+        "优秀数据（自动抓取）",
+        "互动赞藏数",
+        "14天曝光量",
+        "120天曝光量"
+      ],
+      "total_records": 19
+    },
+    {
+      "index": 25,
+      "name": "布谷",
+      "base_token": "KU3Vbw5qDay8kUsSWibcl0cinAc",
+      "table_id": "tblkrKh0LHn4tzq5",
+      "url": "https://bu0zgpibak.feishu.cn/base/KU3Vbw5qDay8kUsSWibcl0cinAc?table=tblkrKh0LHn4tzq5&view=vewStRNML4",
+      "field_map": {
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "note_title": {
+          "field_id": "fldWW36YoP",
+          "field_name": "发布笔记标题"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "creator_id": {
+          "field_id": "fldY2gmw3K",
+          "field_name": "达人id"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "publish_time": {
+          "field_id": "fldx2WWbn4",
+          "field_name": "发布时间"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        }
+      },
+      "all_field_names": [
+        "cpm",
+        "优秀数据（自动抓取）",
+        "互动赞藏数",
+        "达人粉丝",
+        "21天曝光量",
+        "发布笔记链接",
+        "数据表现",
+        "月底曝光量",
+        "180天曝光量",
+        "60天曝光量",
+        "150天曝光量",
+        "120天曝光量",
+        "90天曝光量",
+        "制单日期",
+        "父记录",
+        "曝光量（自动抓取）",
+        "28天曝光量",
+        "7天曝光量",
+        "投放账号类型",
+        "是否结款",
+        "是否新方向验证",
+        "周发布",
+        "月份",
+        "主页链接",
+        "合作费用",
+        "审稿时间",
+        "发布笔记标题",
+        "投流金额",
+        "微信号",
+        "达人id",
+        "达人名称",
+        "观看画像",
+        "发布笔记内容方向",
+        "14天曝光量",
+        "投放平台",
+        "发布时间",
+        "快递单号"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 26,
+      "name": "白鹭",
+      "base_token": "UhuObUyaFaU4Djs55FIcA5RYnVc",
+      "table_id": "tbl4YMl9ZOUHF6pA",
+      "url": "https://bu0zgpibak.feishu.cn/base/UhuObUyaFaU4Djs55FIcA5RYnVc?table=tbl4YMl9ZOUHF6pA&view=vewStRNML4",
+      "field_map": {
+        "creator_id": {
+          "field_id": "fldHu85LRY",
+          "field_name": "达人id"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldPFndO3A",
+          "field_name": "发布时间"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "note_title": {
+          "field_id": "fldAwnmy9j",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        }
+      },
+      "all_field_names": [
+        "达人id",
+        "制单日期",
+        "审稿时间",
+        "主页链接",
+        "周发布",
+        "90天曝光量",
+        "数据表现",
+        "投放平台",
+        "达人粉丝",
+        "合作费用",
+        "60天曝光量",
+        "发布时间",
+        "月份",
+        "月底曝光量",
+        "cpm",
+        "120天曝光量",
+        "曝光量（自动抓取）",
+        "投放账号类型",
+        "达人名称",
+        "快递单号",
+        "150天曝光量",
+        "是否新方向验证",
+        "互动赞藏数",
+        "180天曝光量",
+        "优秀数据（自动抓取）",
+        "是否结款",
+        "发布笔记链接",
+        "28天曝光量",
+        "发布笔记标题",
+        "7天曝光量",
+        "发布笔记内容方向",
+        "微信号",
+        "观看画像",
+        "14天曝光量",
+        "21天曝光量",
+        "投流金额"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 27,
+      "name": "蓝鹊",
+      "base_token": "NoM1bh0zZap7wwsQ69ScydGmnNd",
+      "table_id": "tblFZc9UfFynYYFf",
+      "url": "https://bu0zgpibak.feishu.cn/base/NoM1bh0zZap7wwsQ69ScydGmnNd?table=tblFZc9UfFynYYFf&view=vewStRNML4",
+      "field_map": {
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "note_title": {
+          "field_id": "fld9XWAYW3",
+          "field_name": "发布笔记标题"
+        },
+        "creator_id": {
+          "field_id": "fldYIfKsAh",
+          "field_name": "达人id号"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "publish_time": {
+          "field_id": "fld7BlmZ19",
+          "field_name": "发布时间"
+        }
+      },
+      "all_field_names": [
+        "审稿时间",
+        "优秀数据（自动抓取）",
+        "父记录",
+        "发布笔记标题",
+        "达人id号",
+        "150天曝光量",
+        "90天曝光量",
+        "120天曝光量",
+        "14天曝光量",
+        "60天曝光量",
+        "180天曝光量",
+        "投放平台",
+        "数据表现",
+        "月份",
+        "cpm",
+        "快递单号",
+        "是否新方向验证",
+        "28天曝光量",
+        "是否结款",
+        "主页链接",
+        "观看画像",
+        "投流金额",
+        "互动赞藏数",
+        "合作费用",
+        "21天曝光量",
+        "月底曝光量",
+        "制单日期",
+        "曝光量（自动抓取）",
+        "发布笔记内容方向",
+        "投放账号类型",
+        "达人粉丝",
+        "发布笔记链接",
+        "微信号",
+        "周发布",
+        "达人名称",
+        "7天曝光量",
+        "发布时间"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 28,
+      "name": "盖亚双肩3",
+      "base_token": "XsbmbDyp9adqawsHInacBID1nTc",
+      "table_id": "tblkAakL8aj3CBrC",
+      "url": "https://bu0zgpibak.feishu.cn/base/XsbmbDyp9adqawsHInacBID1nTc?table=tblkAakL8aj3CBrC&view=vewStRNML4",
+      "field_map": {
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "note_title": {
+          "field_id": "fldhf7sgD5",
+          "field_name": "发布笔记标题"
+        },
+        "publish_time": {
+          "field_id": "fldseUYNde",
+          "field_name": "发布时间"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "creator_id": {
+          "field_id": "fldGk4fv93",
+          "field_name": "达人id"
+        }
+      },
+      "all_field_names": [
+        "120天曝光量",
+        "14天曝光量",
+        "90天曝光量",
+        "150天曝光量",
+        "数据表现",
+        "180天曝光量",
+        "发布笔记链接",
+        "主页链接",
+        "cpm",
+        "优秀数据（自动抓取）",
+        "是否结款",
+        "7天曝光量",
+        "28天曝光量",
+        "合作费用",
+        "周发布",
+        "达人粉丝",
+        "审稿时间",
+        "互动赞藏数",
+        "投放平台",
+        "父记录",
+        "观看画像",
+        "快递单号",
+        "达人名称",
+        "微信号",
+        "制单日期",
+        "投流金额",
+        "发布笔记标题",
+        "发布时间",
+        "投放账号类型",
+        "发布笔记内容方向",
+        "是否新方向验证",
+        "60天曝光量",
+        "月底曝光量",
+        "21天曝光量",
+        "月份",
+        "曝光量（自动抓取）",
+        "达人id"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 29,
+      "name": "拾影双肩",
+      "base_token": "JAGEbWz1aarBf7sX96zc8S0VnXg",
+      "table_id": "tblddNHGumvADu2S",
+      "url": "https://bu0zgpibak.feishu.cn/base/JAGEbWz1aarBf7sX96zc8S0VnXg?table=tblddNHGumvADu2S&view=vewStRNML4",
+      "field_map": {
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "creator_id": {
+          "field_id": "fldGk4fv93",
+          "field_name": "达人id"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "note_title": {
+          "field_id": "fldhf7sgD5",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldseUYNde",
+          "field_name": "发布时间"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        }
+      },
+      "all_field_names": [
+        "投放账号类型",
+        "数据表现",
+        "达人id",
+        "主页链接",
+        "发布笔记标题",
+        "28天曝光量",
+        "120天曝光量",
+        "月底曝光量",
+        "60天曝光量",
+        "21天曝光量",
+        "发布时间",
+        "制单日期",
+        "周发布",
+        "合作费用",
+        "互动赞藏数",
+        "达人粉丝",
+        "快递单号",
+        "是否新方向验证",
+        "优秀数据（自动抓取）",
+        "曝光量（自动抓取）",
+        "父记录",
+        "150天曝光量",
+        "投流金额",
+        "发布笔记链接",
+        "审稿时间",
+        "发布笔记内容方向",
+        "180天曝光量",
+        "达人名称",
+        "观看画像",
+        "14天曝光量",
+        "90天曝光量",
+        "投放平台",
+        "cpm",
+        "是否结款",
+        "月份",
+        "微信号",
+        "7天曝光量"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 30,
+      "name": "拾影斜挎相机包",
+      "base_token": "WGVebBy7waEcffsnKFXcuIrwnNf",
+      "table_id": "tbloHTSnakmokRkf",
+      "url": "https://bu0zgpibak.feishu.cn/base/WGVebBy7waEcffsnKFXcuIrwnNf?table=tbloHTSnakmokRkf&view=vewStRNML4",
+      "field_map": {
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "publish_time": {
+          "field_id": "fld5URo58N",
+          "field_name": "发布时间"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "note_title": {
+          "field_id": "fldKByYTPg",
+          "field_name": "笔记标题"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "creator_name": {
+          "field_id": "fldvIjrwdZ",
+          "field_name": "达人名称"
+        }
+      },
+      "all_field_names": [
+        "150天曝光量",
+        "合作费用",
+        "快递单号",
+        "数据表现",
+        "28天曝光量",
+        "cpm",
+        "月底曝光量",
+        "是否新方向验证",
+        "审稿时间",
+        "120天曝光量",
+        "7天曝光量",
+        "180天曝光量",
+        "互动赞藏数",
+        "投流金额",
+        "周发布",
+        "90天曝光量",
+        "曝光量（自动抓取）",
+        "发布笔记内容方向",
+        "优秀数据（自动抓取）",
+        "月份",
+        "主页链接",
+        "发布笔记链接",
+        "父记录",
+        "投放平台",
+        "21天曝光量",
+        "投放账号类型",
+        "60天曝光量",
+        "14天曝光量",
+        "达人di",
+        "微信号",
+        "达人粉丝",
+        "发布时间",
+        "制单日期",
+        "是否结款",
+        "笔记标题",
+        "观看画像",
+        "达人名称"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 31,
+      "name": "云栖",
+      "base_token": "BT8BbyT3HaUsjLs4PzZcj5iInGb",
+      "table_id": "tbldQYQFCd1j4aZZ",
+      "url": "https://bu0zgpibak.feishu.cn/base/BT8BbyT3HaUsjLs4PzZcj5iInGb?table=tbldQYQFCd1j4aZZ&view=vewStRNML4",
+      "field_map": {
+        "creator_id": {
+          "field_id": "fldGk4fv93",
+          "field_name": "达人id"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "publish_time": {
+          "field_id": "fldseUYNde",
+          "field_name": "发布时间"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "note_title": {
+          "field_id": "fldhf7sgD5",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        }
+      },
+      "all_field_names": [
+        "达人id",
+        "投放平台",
+        "微信号",
+        "90天曝光量",
+        "父记录",
+        "120天曝光量",
+        "投放账号类型",
+        "达人名称",
+        "互动赞藏数",
+        "14天曝光量",
+        "数据表现",
+        "是否结款",
+        "周发布",
+        "主页链接",
+        "快递单号",
+        "制单日期",
+        "发布时间",
+        "优秀数据（自动抓取）",
+        "60天曝光量",
+        "180天曝光量",
+        "月底曝光量",
+        "发布笔记链接",
+        "发布笔记标题",
+        "21天曝光量",
+        "曝光量（自动抓取）",
+        "7天曝光量",
+        "是否新方向验证",
+        "审稿时间",
+        "cpm",
+        "28天曝光量",
+        "合作费用",
+        "观看画像",
+        "月份",
+        "150天曝光量",
+        "达人粉丝",
+        "投流金额",
+        "发布笔记内容方向"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 32,
+      "name": "波塞冬手提电脑包",
+      "base_token": "Aiazbsa80auaVGsktHEcembEnzc",
+      "table_id": "tblaRStV5YO5Mqkw",
+      "url": "https://bu0zgpibak.feishu.cn/base/Aiazbsa80auaVGsktHEcembEnzc?table=tblaRStV5YO5Mqkw&view=vewStRNML4",
+      "field_map": {
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "creator_id": {
+          "field_id": "fldGk4fv93",
+          "field_name": "达人id"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldseUYNde",
+          "field_name": "发布时间"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "note_title": {
+          "field_id": "fldhf7sgD5",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        }
+      },
+      "all_field_names": [
+        "观看画像",
+        "180天曝光量",
+        "90天曝光量",
+        "主页链接",
+        "优秀数据（自动抓取）",
+        "发布笔记链接",
+        "达人名称",
+        "28天曝光量",
+        "投放账号类型",
+        "cpm",
+        "达人id",
+        "月底曝光量",
+        "发布时间",
+        "合作费用",
+        "60天曝光量",
+        "投流金额",
+        "是否结款",
+        "曝光量（自动抓取）",
+        "发布笔记内容方向",
+        "7天曝光量",
+        "发布笔记标题",
+        "21天曝光量",
+        "150天曝光量",
+        "父记录",
+        "数据表现",
+        "微信号",
+        "达人粉丝",
+        "月份",
+        "快递单号",
+        "投放平台",
+        "审稿时间",
+        "120天曝光量",
+        "周发布",
+        "互动赞藏数",
+        "制单日期",
+        "是否新方向验证",
+        "14天曝光量"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 33,
+      "name": "极星双肩2",
+      "base_token": "UBM9bg9WnaU0Jvs3MYDcC8s6nYe",
+      "table_id": "tbl78I4vUt7o4IDS",
+      "url": "https://bu0zgpibak.feishu.cn/base/UBM9bg9WnaU0Jvs3MYDcC8s6nYe?table=tbl78I4vUt7o4IDS&view=vewStRNML4",
+      "field_map": {
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldseUYNde",
+          "field_name": "发布时间"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "creator_id": {
+          "field_id": "fldGk4fv93",
+          "field_name": "达人id"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "note_title": {
+          "field_id": "fldhf7sgD5",
+          "field_name": "发布笔记标题"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        }
+      },
+      "all_field_names": [
+        "120天曝光量",
+        "月份",
+        "cpm",
+        "14天曝光量",
+        "发布时间",
+        "月底曝光量",
+        "150天曝光量",
+        "达人id",
+        "微信号",
+        "审稿时间",
+        "发布笔记内容方向",
+        "180天曝光量",
+        "是否新方向验证",
+        "主页链接",
+        "曝光量（自动抓取）",
+        "21天曝光量",
+        "投流金额",
+        "达人名称",
+        "是否结款",
+        "28天曝光量",
+        "合作费用",
+        "投放平台",
+        "互动赞藏数",
+        "60天曝光量",
+        "7天曝光量",
+        "投放账号类型",
+        "发布笔记链接",
+        "快递单号",
+        "父记录",
+        "数据表现",
+        "90天曝光量",
+        "发布笔记标题",
+        "优秀数据（自动抓取）",
+        "达人粉丝",
+        "制单日期",
+        "观看画像",
+        "周发布"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 34,
+      "name": "极星单肩",
+      "base_token": "MVxPb5RFEa9YEPsDmEMc6ujinoh",
+      "table_id": "tblzb89dVJpVYq3V",
+      "url": "https://bu0zgpibak.feishu.cn/base/MVxPb5RFEa9YEPsDmEMc6ujinoh?table=tblzb89dVJpVYq3V&view=vewStRNML4",
+      "field_map": {
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "creator_id": {
+          "field_id": "fldGk4fv93",
+          "field_name": "达人id"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "note_title": {
+          "field_id": "fldhf7sgD5",
+          "field_name": "发布笔记标题"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "publish_time": {
+          "field_id": "fldseUYNde",
+          "field_name": "发布时间"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        }
+      },
+      "all_field_names": [
+        "90天曝光量",
+        "达人id",
+        "达人粉丝",
+        "合作费用",
+        "投流金额",
+        "制单日期",
+        "发布笔记链接",
+        "60天曝光量",
+        "数据表现",
+        "cpm",
+        "投放账号类型",
+        "21天曝光量",
+        "月底曝光量",
+        "审稿时间",
+        "150天曝光量",
+        "120天曝光量",
+        "互动赞藏数",
+        "投放平台",
+        "7天曝光量",
+        "观看画像",
+        "父记录",
+        "曝光量（自动抓取）",
+        "达人名称",
+        "发布笔记标题",
+        "是否结款",
+        "快递单号",
+        "发布时间",
+        "优秀数据（自动抓取）",
+        "发布笔记内容方向",
+        "主页链接",
+        "周发布",
+        "微信号",
+        "180天曝光量",
+        "月份",
+        "14天曝光量",
+        "是否新方向验证",
+        "28天曝光量"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 35,
+      "name": "极星2",
+      "base_token": "LvLBblI5KaekeUsaPRpcDeSBnxb",
+      "table_id": "tblNYpJkLJYWaV82",
+      "url": "https://bu0zgpibak.feishu.cn/base/LvLBblI5KaekeUsaPRpcDeSBnxb?table=tblNYpJkLJYWaV82&view=vewStRNML4",
+      "field_map": {
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "profile_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "主页链接"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "tracking_no": {
+          "field_id": "fldm2o9Upj",
+          "field_name": "快递单号"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "creator_name": {
+          "field_id": "fld6A1I2SA",
+          "field_name": "达人名称"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "publish_time": {
+          "field_id": "fld5URo58N",
+          "field_name": "发布时间"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        }
+      },
+      "all_field_names": [
+        "发布笔记链接",
+        "优秀数据（自动抓取）",
+        "150天曝光量",
+        "主页链接",
+        "60天曝光量",
+        "发布笔记内容方向",
+        "曝光量（自动抓取）",
+        "21天曝光量",
+        "快递单号",
+        "120天曝光量",
+        "观看画像",
+        "达人名称",
+        "7天曝光量",
+        "28天曝光量",
+        "是否结款",
+        "投放平台",
+        "发布时间",
+        "投放账号类型",
+        "审稿时间",
+        "月底曝光量",
+        "cpm",
+        "月份",
+        "互动赞藏数",
+        "达人粉丝",
+        "是否新方向验证",
+        "周发布",
+        "投流金额",
+        "90天曝光量",
+        "微信号",
+        "制单日期",
+        "合作费用",
+        "14天曝光量",
+        "180天曝光量",
+        "数据表现"
+      ],
+      "total_records": 0
+    }
+  ]
+}
\ No newline at end of file
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/config/款式_多维表格_对照_自营.json b/src/gyxx_flow/modules/content_marketing/runtime/data/config/款式_多维表格_对照_自营.json
new file mode 100644
index 0000000..5ca5055
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/config/款式_多维表格_对照_自营.json
@@ -0,0 +1,1807 @@
+{
+  "_comment": "款式 → 多维表格 对照表(自营); 由 feishu_mapping.py 从自营地址表动态生成。 field_map 给出本表核心字段的 field_id (实时从飞书拉取)",
+  "platform": "xhs_pgy",
+  "tables": [
+    {
+      "index": 1,
+      "name": "盖世m1",
+      "base_token": "KnBRbH9iQaH3PAsAUk5cHLxNnMg",
+      "table_id": "tbl4KkQV6goQBawV",
+      "url": "https://bu0zgpibak.feishu.cn/base/KnBRbH9iQaH3PAsAUk5cHLxNnMg?table=tbl4KkQV6goQBawV&view=vewStRNML4",
+      "field_map": {
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "publish_time": {
+          "field_id": "fldJgCDvNj",
+          "field_name": "发布时间"
+        },
+        "note_title": {
+          "field_id": "fld1OJ6D4X",
+          "field_name": "发布笔记标题"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        }
+      },
+      "all_field_names": [
+        "28天曝光量",
+        "优秀数据（自动抓取）",
+        "150天曝光量",
+        "发布笔记链接",
+        "周发布",
+        "投放平台",
+        "cpm",
+        "数据表现",
+        "合作费用",
+        "60天曝光量",
+        "14天曝光量",
+        "投流金额",
+        "发布时间",
+        "发布笔记标题",
+        "月底曝光量",
+        "90天曝光量",
+        "7天曝光量",
+        "21天曝光量",
+        "180天曝光量",
+        "月份",
+        "曝光量（自动抓取）",
+        "120天曝光量",
+        "观看画像",
+        "发布笔记内容方向"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 2,
+      "name": "星迹2",
+      "base_token": "Ssp8bXmrwahRBbsko2BcuiLanad",
+      "table_id": "tblMK0bFX6KbBssl",
+      "url": "https://bu0zgpibak.feishu.cn/base/Ssp8bXmrwahRBbsko2BcuiLanad?table=tblMK0bFX6KbBssl&view=vewStRNML4",
+      "field_map": {
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "note_title": {
+          "field_id": "fldYb2Wsil",
+          "field_name": "发布笔记标题"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "publish_time": {
+          "field_id": "fldrCfjoir",
+          "field_name": "发布时间"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "review_time": {
+          "field_id": "fldQifNRp4",
+          "field_name": "审稿时间"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        }
+      },
+      "all_field_names": [
+        "观看画像",
+        "发布笔记链接",
+        "合作费用",
+        "发布笔记标题",
+        "是否新方向验证",
+        "28天曝光量",
+        "优秀数据（自动抓取）",
+        "投放平台",
+        "达人粉丝",
+        "14天曝光量",
+        "是否结款",
+        "月份",
+        "发布时间",
+        "90天曝光量",
+        "21天曝光量",
+        "父记录",
+        "投流金额",
+        "7天曝光量",
+        "cpm",
+        "月底曝光量",
+        "发布笔记内容方向",
+        "60天曝光量",
+        "180天曝光量",
+        "周发布",
+        "120天曝光量",
+        "曝光量（自动抓取）",
+        "微信号",
+        "互动赞藏数",
+        "数据表现",
+        "审稿时间",
+        "投放账号类型",
+        "150天曝光量"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 3,
+      "name": "星云2",
+      "base_token": "BJaWbLEOKaHONMshOidcxvsKngh",
+      "table_id": "tbl1sID3mOpVPr9Q",
+      "url": "https://bu0zgpibak.feishu.cn/base/BJaWbLEOKaHONMshOidcxvsKngh?table=tbl1sID3mOpVPr9Q&view=vewStRNML4",
+      "field_map": {
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "is_new_direction": {
+          "field_id": "fldcxLZOL8",
+          "field_name": "是否新方向验证"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "publish_time": {
+          "field_id": "fldzihhShx",
+          "field_name": "发布时间"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "note_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "发布链接"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "is_paid": {
+          "field_id": "fldkVdwKoY",
+          "field_name": "是否结款"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "note_title": {
+          "field_id": "fldoe9ysE5",
+          "field_name": "发布笔记标题"
+        }
+      },
+      "all_field_names": [
+        "数据表现",
+        "21天曝光量",
+        "优秀数据（自动抓取）",
+        "是否新方向验证",
+        "微信号",
+        "父记录",
+        "互动赞藏数",
+        "文本 6",
+        "达人粉丝",
+        "14天曝光量",
+        "发布笔记内容方向",
+        "发布时间",
+        "cpm",
+        "发布链接",
+        "曝光量（自动抓取）",
+        "投放账号类型",
+        "60天曝光量",
+        "是否结款",
+        "月底曝光量",
+        "120天曝光量",
+        "7天曝光量",
+        "观看画像",
+        "28天曝光量",
+        "周发布",
+        "月份",
+        "投流金额",
+        "150天曝光量",
+        "90天曝光量",
+        "180天曝光量",
+        "投放平台",
+        "合作费用",
+        "发布笔记标题"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 4,
+      "name": "宙斯",
+      "base_token": "N43abi7IlaS8XosnUXec6gH2n7g",
+      "table_id": "tblxvEHpG2D1nHzS",
+      "url": "https://bu0zgpibak.feishu.cn/base/N43abi7IlaS8XosnUXec6gH2n7g?table=tblxvEHpG2D1nHzS&view=vewStRNML4",
+      "field_map": {
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "note_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "发布链接"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "note_title": {
+          "field_id": "fldJr5VXwA",
+          "field_name": "笔记标题"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        }
+      },
+      "all_field_names": [
+        "曝光量（自动抓取）",
+        "发布时间",
+        "投放平台",
+        "合作费用",
+        "月份",
+        "发布链接",
+        "投流金额",
+        "7天曝光量",
+        "120天曝光量",
+        "21天曝光量",
+        "180天曝光量",
+        "28天曝光量",
+        "观看画像",
+        "cpm",
+        "数据表现",
+        "150天曝光量",
+        "14天曝光量",
+        "90天曝光量",
+        "笔记标题",
+        "60天曝光量",
+        "月底曝光量",
+        "周发布",
+        "优秀数据（自动抓取）",
+        "父记录",
+        "发布笔记内容方向"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 5,
+      "name": "阿波罗2",
+      "base_token": "W80vbtPDoaaHTOs5Tipc6CTLn6c",
+      "table_id": "tbl6PsOTcxPnVo52",
+      "url": "https://bu0zgpibak.feishu.cn/base/W80vbtPDoaaHTOs5Tipc6CTLn6c?table=tbl6PsOTcxPnVo52&view=vewStRNML4",
+      "field_map": {
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "note_title": {
+          "field_id": "fldHElhugy",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "发布笔记链接"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        }
+      },
+      "all_field_names": [
+        "90天曝光量",
+        "180天曝光量",
+        "7月第一周",
+        "发布时间",
+        "月底曝光量",
+        "数据表现",
+        "投放平台",
+        "优秀数据（自动抓取）",
+        "14天曝光量",
+        "合作费用",
+        "发布笔记内容方向",
+        "月份",
+        "观看画像",
+        "60天曝光量",
+        "cpm",
+        "发布笔记标题",
+        "120天曝光量",
+        "7天曝光量",
+        "28天曝光量",
+        "投流金额",
+        "150天曝光量",
+        "周发布",
+        "21天曝光量",
+        "发布笔记链接",
+        "曝光量（自动抓取）"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 6,
+      "name": "波塞冬edge",
+      "base_token": "JsVLbso8Ba6RjbslOl1cJX8lnyb",
+      "table_id": "tblCHB1QmEsVs8j3",
+      "url": "https://bu0zgpibak.feishu.cn/base/JsVLbso8Ba6RjbslOl1cJX8lnyb?table=tblCHB1QmEsVs8j3&view=vewStRNML4",
+      "field_map": {
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "note_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "发布链接"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "publish_time": {
+          "field_id": "fldbKALsn7",
+          "field_name": "发布时间"
+        },
+        "note_title": {
+          "field_id": "fldt8vGXO9",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        }
+      },
+      "all_field_names": [
+        "180天曝光量",
+        "150天曝光量",
+        "21天曝光量",
+        "90天曝光量",
+        "观看画像",
+        "发布笔记内容方向",
+        "曝光量（自动抓取）",
+        "60天曝光量",
+        "父记录",
+        "cpm",
+        "月底曝光量",
+        "发布链接",
+        "14天曝光量",
+        "投放平台",
+        "合作费用",
+        "数据表现",
+        "发布时间",
+        "发布笔记标题",
+        "周发布",
+        "120天曝光量",
+        "投流金额",
+        "月份",
+        "7天曝光量",
+        "优秀数据（自动抓取）",
+        "28天曝光量"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 7,
+      "name": "逐星",
+      "base_token": "IQrbbxAHpa9oiWs631GcPkHznob",
+      "table_id": "tblNVimu1eFsKpdh",
+      "url": "https://bu0zgpibak.feishu.cn/base/IQrbbxAHpa9oiWs631GcPkHznob?table=tblNVimu1eFsKpdh&view=vewStRNML4",
+      "field_map": {
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "note_title": {
+          "field_id": "fldlbWqQw3",
+          "field_name": "发布笔记标题"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "发布链接"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        }
+      },
+      "all_field_names": [
+        "发布笔记内容方向",
+        "120天曝光量",
+        "7天曝光量",
+        "周发布",
+        "发布笔记标题",
+        "合作费用",
+        "14天曝光量",
+        "发布链接",
+        "优秀数据（自动抓取）",
+        "投放平台",
+        "投流金额",
+        "cpm",
+        "28天曝光量",
+        "观看画像",
+        "发布时间",
+        "21天曝光量",
+        "曝光量（自动抓取）",
+        "150天曝光量",
+        "月份",
+        "数据表现",
+        "月底曝光量",
+        "90天曝光量",
+        "180天曝光量",
+        "60天曝光量"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 8,
+      "name": "极星双肩",
+      "base_token": "GubZbTrMSahLg5sdonIc8kR8nZe",
+      "table_id": "tblI11Gdn9ASUdzy",
+      "url": "https://bu0zgpibak.feishu.cn/base/GubZbTrMSahLg5sdonIc8kR8nZe?table=tblI11Gdn9ASUdzy&view=vewStRNML4",
+      "field_map": {
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "note_title": {
+          "field_id": "fldeXdeRRH",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "publish_time": {
+          "field_id": "fldXU5JGBi",
+          "field_name": "发布时间"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        }
+      },
+      "all_field_names": [
+        "60天曝光量",
+        "90天曝光量",
+        "150天曝光量",
+        "数据表现",
+        "发布笔记内容方向",
+        "28天曝光量",
+        "投放平台",
+        "合作费用",
+        "观看画像",
+        "120天曝光量",
+        "180天曝光量",
+        "优秀数据（自动抓取）",
+        "发布笔记标题",
+        "14天曝光量",
+        "曝光量（自动抓取）",
+        "周发布",
+        "父记录",
+        "cpm",
+        "发布时间",
+        "时间",
+        "月底曝光量",
+        "7天曝光量",
+        "投流金额",
+        "21天曝光量",
+        "发布笔记链接"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 9,
+      "name": "瑞白",
+      "base_token": "GNOkbKNpna8pDJsLMfzcuZt3ncf",
+      "table_id": "tblCU7lv5efZulqn",
+      "url": "https://bu0zgpibak.feishu.cn/base/GNOkbKNpna8pDJsLMfzcuZt3ncf?table=tblCU7lv5efZulqn&view=vewStRNML4",
+      "field_map": {
+        "note_url": {
+          "field_id": "fldHKQyT01",
+          "field_name": "发布链接"
+        },
+        "engagement_count": {
+          "field_id": "fldkPLcmqV",
+          "field_name": "互动赞藏数"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "cooperation_date": {
+          "field_id": "fldN7muprD",
+          "field_name": "制单日期"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "account_type": {
+          "field_id": "fldCj4WnL1",
+          "field_name": "投放账号类型"
+        },
+        "follower_count": {
+          "field_id": "fldecL2kkj",
+          "field_name": "达人粉丝"
+        },
+        "note_title": {
+          "field_id": "fld1sZPxKx",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "wechat": {
+          "field_id": "fld9TEtaF6",
+          "field_name": "微信号"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "publish_time": {
+          "field_id": "fldFJCKY20",
+          "field_name": "发布时间"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        }
+      },
+      "all_field_names": [
+        "发布链接",
+        "互动赞藏数",
+        "发布笔记内容方向",
+        "28天曝光量",
+        "cpm",
+        "制单日期",
+        "数据表现",
+        "投放账号类型",
+        "周发布",
+        "达人粉丝",
+        "发布笔记标题",
+        "7天曝光量",
+        "优秀数据（自动抓取）",
+        "合作费用",
+        "14天曝光量",
+        "微信号",
+        "曝光量（自动抓取）",
+        "月底曝光量",
+        "发布时间",
+        "月份",
+        "投放平台",
+        "21天曝光量",
+        "120天曝光量",
+        "180天曝光量",
+        "投流金额",
+        "父记录",
+        "观看画像",
+        "150天曝光量"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 10,
+      "name": "星云mini",
+      "base_token": "RKfkb5JkIa9FttsgHvNc2nJYnbf",
+      "table_id": "tbleLq1bq2r7ZbD2",
+      "url": "https://bu0zgpibak.feishu.cn/base/RKfkb5JkIa9FttsgHvNc2nJYnbf?table=tbleLq1bq2r7ZbD2&view=vewStRNML4",
+      "field_map": {
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "note_title": {
+          "field_id": "fld8X4rGRI",
+          "field_name": "发布笔记标题"
+        },
+        "publish_time": {
+          "field_id": "fldrx09tSe",
+          "field_name": "发布时间"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        }
+      },
+      "all_field_names": [
+        "14天曝光量",
+        "cpm",
+        "180天曝光量",
+        "优秀数据（自动抓取）",
+        "父记录",
+        "60天曝光量",
+        "150天曝光量",
+        "发布笔记标题",
+        "发布时间",
+        "120天曝光量",
+        "发布笔记链接",
+        "90天曝光量",
+        "21天曝光量",
+        "观看画像",
+        "曝光量（自动抓取）",
+        "投流金额",
+        "7天曝光量",
+        "合作费用",
+        "月底曝光量",
+        "28天曝光量",
+        "周发布",
+        "发布笔记内容方向",
+        "投放平台",
+        "数据表现",
+        "月份"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 12,
+      "name": "凌云air",
+      "base_token": "VwpibmlakaqHuLsGChccMb3BnNd",
+      "table_id": "tblLheCfo2KirXfD",
+      "url": "https://bu0zgpibak.feishu.cn/base/VwpibmlakaqHuLsGChccMb3BnNd?table=tblLheCfo2KirXfD&view=vewStRNML4",
+      "field_map": {
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "note_title": {
+          "field_id": "fldVpDYo8z",
+          "field_name": "笔记标题"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "publish_time": {
+          "field_id": "fldFJCKY20",
+          "field_name": "发布时间"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        }
+      },
+      "all_field_names": [
+        "180天曝光量",
+        "7天曝光量",
+        "观看画像",
+        "月底曝光量",
+        "笔记标题",
+        "28天曝光量",
+        "14天曝光量",
+        "cpm",
+        "月份",
+        "投流金额",
+        "发布时间",
+        "周发布",
+        "曝光量（自动抓取）",
+        "父记录",
+        "发布笔记内容方向",
+        "优秀数据（自动抓取）",
+        "发布笔记链接",
+        "投放平台",
+        "21天曝光量",
+        "120天曝光量",
+        "150天曝光量",
+        "数据表现",
+        "合作费用"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 13,
+      "name": "拾影斜挎相机包",
+      "base_token": "WGVebBy7waEcffsnKFXcuIrwnNf",
+      "table_id": "tbl1PcudZ1kjq5Z5",
+      "url": "https://bu0zgpibak.feishu.cn/base/WGVebBy7waEcffsnKFXcuIrwnNf?table=tbl1PcudZ1kjq5Z5&view=vewJwbpDtU",
+      "field_map": {},
+      "all_field_names": [
+        "时间",
+        "笔记分析汇总"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 14,
+      "name": "黑曜石托特包",
+      "base_token": "U9S8bLYsNa8Ul8sPffAcrJ02nHI",
+      "table_id": "tblwlrW5ZujYgVhF",
+      "url": "https://bu0zgpibak.feishu.cn/base/U9S8bLYsNa8Ul8sPffAcrJ02nHI?table=tblwlrW5ZujYgVhF&view=vewStRNML4",
+      "field_map": {
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "note_title": {
+          "field_id": "fld0ExYDUi",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "publish_time": {
+          "field_id": "fldFJCKY20",
+          "field_name": "发布时间"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        }
+      },
+      "all_field_names": [
+        "cpm",
+        "150天曝光量",
+        "28天曝光量",
+        "180天曝光量",
+        "发布笔记标题",
+        "21天曝光量",
+        "曝光量（自动抓取）",
+        "投流金额",
+        "月底曝光量",
+        "合作费用",
+        "90天曝光量",
+        "7天曝光量",
+        "投放平台",
+        "月份",
+        "数据表现",
+        "发布时间",
+        "发布笔记内容方向",
+        "优秀数据（自动抓取）",
+        "观看画像",
+        "发布笔记链接",
+        "60天曝光量",
+        "周发布",
+        "14天曝光量",
+        "120天曝光量"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 15,
+      "name": "曜影",
+      "base_token": "UPk3bR9MOaTkgrsG1qLc7IUvnQe",
+      "table_id": "tblULX16sKFlRFft",
+      "url": "https://bu0zgpibak.feishu.cn/base/UPk3bR9MOaTkgrsG1qLc7IUvnQe?table=tblULX16sKFlRFft&view=vewStRNML4",
+      "field_map": {
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "note_title": {
+          "field_id": "fldRGRSyDQ",
+          "field_name": "发布笔记标题"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "publish_time": {
+          "field_id": "fld5URo58N",
+          "field_name": "发布时间"
+        },
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        }
+      },
+      "all_field_names": [
+        "cpm",
+        "数据表现",
+        "发布笔记内容方向",
+        "28天曝光量",
+        "周发布",
+        "21天曝光量",
+        "150天曝光量",
+        "120天曝光量",
+        "投流金额",
+        "7天曝光量",
+        "月份",
+        "投放平台",
+        "发布笔记标题",
+        "观看画像",
+        "发布时间",
+        "发布笔记链接",
+        "父记录",
+        "优秀数据（自动抓取）",
+        "14天曝光量",
+        "合作费用",
+        "90天曝光量",
+        "180天曝光量",
+        "曝光量（自动抓取）",
+        "月底曝光量",
+        "60天曝光量"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 16,
+      "name": "盖亚双肩3",
+      "base_token": "XsbmbDyp9adqawsHInacBID1nTc",
+      "table_id": "tbl1yTxdrfN5JDv1",
+      "url": "https://bu0zgpibak.feishu.cn/base/XsbmbDyp9adqawsHInacBID1nTc?table=tbl1yTxdrfN5JDv1&view=vewStRNML4",
+      "field_map": {
+        "note_url": {
+          "field_id": "fldgk4pEKH",
+          "field_name": "发布笔记链接"
+        },
+        "publish_time": {
+          "field_id": "fldseUYNde",
+          "field_name": "发布时间"
+        },
+        "ad_spend": {
+          "field_id": "fldEkxIUpO",
+          "field_name": "投流金额"
+        },
+        "read_count_21d": {
+          "field_id": "fldmptkUsh",
+          "field_name": "21天曝光量"
+        },
+        "read_count_14d": {
+          "field_id": "fldHhUJzR6",
+          "field_name": "14天曝光量"
+        },
+        "read_count_28d": {
+          "field_id": "fldFE8kl1g",
+          "field_name": "28天曝光量"
+        },
+        "read_count_auto": {
+          "field_id": "fld7Z7ESLl",
+          "field_name": "曝光量（自动抓取）"
+        },
+        "read_count_120d": {
+          "field_id": "fldImcwTdq",
+          "field_name": "120天曝光量"
+        },
+        "content_direction": {
+          "field_id": "fldXuZLq9P",
+          "field_name": "发布笔记内容方向"
+        },
+        "cpm": {
+          "field_id": "fldIOGQ6F5",
+          "field_name": "cpm"
+        },
+        "read_count_7d": {
+          "field_id": "fld3UamjeO",
+          "field_name": "7天曝光量"
+        },
+        "platform": {
+          "field_id": "fldYwBr7mp",
+          "field_name": "投放平台"
+        },
+        "note_title": {
+          "field_id": "fldhf7sgD5",
+          "field_name": "发布笔记标题"
+        },
+        "read_count_60d": {
+          "field_id": "fld1SlxJu9",
+          "field_name": "60天曝光量"
+        },
+        "audience_profile": {
+          "field_id": "fldzxG120c",
+          "field_name": "观看画像"
+        },
+        "read_count_180d": {
+          "field_id": "fldSX2mZnv",
+          "field_name": "180天曝光量"
+        },
+        "read_count_150d": {
+          "field_id": "fldz8mTxYj",
+          "field_name": "150天曝光量"
+        },
+        "month_end": {
+          "field_id": "fldLHuDzjQ",
+          "field_name": "月底曝光量"
+        },
+        "read_count_90d": {
+          "field_id": "fld8o0lZx4",
+          "field_name": "90天曝光量"
+        },
+        "data_performance": {
+          "field_id": "fldSXDBNRk",
+          "field_name": "数据表现"
+        },
+        "cost": {
+          "field_id": "fldxyhJAOC",
+          "field_name": "合作费用"
+        }
+      },
+      "all_field_names": [
+        "发布笔记链接",
+        "发布时间",
+        "优秀数据（自动抓取）",
+        "父记录",
+        "投流金额",
+        "21天曝光量",
+        "14天曝光量",
+        "28天曝光量",
+        "曝光量（自动抓取）",
+        "120天曝光量",
+        "发布笔记内容方向",
+        "cpm",
+        "7天曝光量",
+        "投放平台",
+        "发布笔记标题",
+        "60天曝光量",
+        "观看画像",
+        "月份",
+        "180天曝光量",
+        "周发布",
+        "150天曝光量",
+        "月底曝光量",
+        "90天曝光量",
+        "数据表现",
+        "合作费用"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 17,
+      "name": "极星2",
+      "base_token": "LvLBblI5KaekeUsaPRpcDeSBnxb",
+      "table_id": "tbl4Fi989YAVpx8n",
+      "url": "https://bu0zgpibak.feishu.cn/base/LvLBblI5KaekeUsaPRpcDeSBnxb?table=tbl4Fi989YAVpx8n&view=vewXPg4JT4",
+      "field_map": {},
+      "all_field_names": [
+        "时间",
+        "笔记分析汇总"
+      ],
+      "total_records": 0
+    },
+    {
+      "index": 18,
+      "name": "波塞冬手提电脑包",
+      "base_token": "Aiazbsa80auaVGsktHEcembEnzc",
+      "table_id": "tblFph4oYic3bG8k",
+      "url": "https://bu0zgpibak.feishu.cn/base/Aiazbsa80auaVGsktHEcembEnzc?table=tblFph4oYic3bG8k&view=vew4GSCwsg",
+      "field_map": {},
+      "all_field_names": [
+        "时间",
+        "笔记分析汇总"
+      ],
+      "total_records": 0
+    }
+  ]
+}
\ No newline at end of file
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/analyze_comments.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/analyze_comments.py
new file mode 100644
index 0000000..7e5748e
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/analyze_comments.py
@@ -0,0 +1,863 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""评论采集数据 → 本地关键词过滤 + Hermes 舆情分析 → 项目 data/summary 汇总 + 飞书通知。
+
+使用本地 Hermes 分析端：
+  - URL: http://127.0.0.1:8642/v1（~/.hermes/profiles/data-analyzer/.env 里的 API_SERVER_PORT）
+  - Model: mimo-v2.5-pro
+  - Profile: data-analyzer
+
+流程：
+  1. 读取项目采集汇总的评论文件（默认取 data/summary/ 下最新 CSV，支持 txt/json/csv）。
+  2. 用本地 Python 关键词规则（飞书文档 6 层体系）过滤评论。
+  3. 用 Hermes 分析端 + 内置舆情分析提示词生成汇总报告。
+  4. 将汇总写入 data/summary/评论舆论分析汇总.txt。
+  5. 通过 lark-cli --profile hermes-analyzer 发送汇总给目标 open_id。
+"""
+
+import argparse
+import concurrent.futures
+import csv
+import json
+import os
+from gyxx_flow.adapters import RuntimeServicePolicy
+import random
+import re
+import subprocess
+import sys
+import time
+import urllib.error
+import urllib.request
+from pathlib import Path
+from typing import Any
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import (
+    PATHS,
+    resolve_layer_output,
+)
+
+# Path setup
+_TOOLS_DIR = PATHS.tools_root
+_PROJECT_ROOT = PATHS.module_root
+
+try:
+    from dotenv import load_dotenv
+    _analyze_env = PATHS.state_root / "config/analyze.env"
+    if _analyze_env.exists():
+        load_dotenv(_analyze_env, override=False)
+except ImportError:
+    pass
+
+
+PROJECT_ROOT = _PROJECT_ROOT
+DATA_DIR = PATHS.raw_root
+SUMMARY_DIR = PATHS.exports_root / "summary"
+
+
+def _latest_summary_csv() -> Path:
+    """自动选取 data/summary/ 下最新的汇总 CSV（项目采集输出）。"""
+    if not SUMMARY_DIR.exists():
+        return SUMMARY_DIR / "*.csv"
+    csv_files = sorted(SUMMARY_DIR.glob("*.csv"), key=lambda p: p.stat().st_mtime, reverse=True)
+    return csv_files[0] if csv_files else SUMMARY_DIR / "*.csv"
+
+
+DEFAULT_INPUT = _latest_summary_csv()
+OUTPUT_PATH = SUMMARY_DIR / "评论舆论分析汇总.txt"
+
+# ---------------------------------------------------------------------------
+# Hermes 分析端配置
+# 实际 API Server 端口见 ~/.hermes/profiles/data-analyzer/.env 中的 API_SERVER_PORT（当前为 8642）
+# ---------------------------------------------------------------------------
+os.environ.update(RuntimeServicePolicy().apply(os.environ))
+HERMES_ANALYZER_URL = os.environ["HERMES_ANALYZER_URL"]
+HERMES_ANALYZER_TOKEN = os.getenv("HERMES_ANALYZER_TOKEN", "")
+HERMES_ANALYZER_MODEL = os.getenv("HERMES_ANALYZER_MODEL", "mimo-v2.5-pro")
+
+TARGET_OPEN_ID = os.getenv("FEISHU_TARGET_OPEN_ID", "")
+
+# 飞书长消息截断阈值（留有余量）
+FEISHU_TEXT_LIMIT = 6000
+
+# ---------------------------------------------------------------------------
+# 本地代码过滤关键词（飞书文档《评论区关键词抓取》6 层体系）
+# ---------------------------------------------------------------------------
+# 流程：全量采集评论 → 本地关键词过滤 → 对有效评论进行舆情分析汇总
+CODE_FILTER_KEYWORDS = {
+    # 第1层：购买意向信号（衡量种草效率）
+    "purchase_intent": [
+        # 直接购买
+        "下单", "已买", "入手", "冲了", "付款", "等收货",
+        # 搜索意向
+        "怎么买", "链接", "哪里买", "搜什么", "淘宝", "店铺",
+        # 价格评估
+        "多少钱", "价格", "划算", "贵不贵", "优惠", "值得买",
+        # 犹豫对比（用双字词避免单字"比"误匹配[比心]等表情）
+        "纠结", "哪个好", "容量够吗", "犹豫", "对比", "比较", "和哪个",
+    ],
+    # 第2层：产品功能反馈（正向）
+    "positive_feature": [
+        # 轻量化（"轻"用双字避免误匹配"年轻人"等）
+        "很轻", "超轻", "挺轻", "不重", "没感觉", "毫无负担", "背着不累",
+        # 收纳秩序
+        "分区", "隔层", "收纳", "能装", "各就各位",
+        # 快取便利（"快"用双字避免误匹配"快乐"等）
+        "很方便", "顺手", "很快", "单手", "侧袋", "不用翻",
+        # 品质感
+        "质感", "材质", "面料", "五金", "走线", "高级",
+    ],
+    # 第3层：产品功能反馈（负向）
+    "negative_feature": [
+        # 重量负担（"重"用双字避免误匹配"重要"等）
+        "很重", "太重", "超重", "偏重", "背不动", "肩膀疼",
+        # 收纳不足
+        "不够装", "太小", "放不下", "没地方", "容量",
+        # 品质问题
+        "线头", "脱线", "划痕", "掉色", "起皱", "做工",
+        # 使用不便（"卡""紧"用双字避免误匹配）
+        "不好拿", "卡住", "太紧", "难开", "不方便",
+    ],
+    # 第4层：用户场景（判断内容与目标用户是否匹配）
+    "user_scenario": [
+        # 通勤
+        "上班", "地铁", "挤公交", "通勤", "工位", "办公室",
+        # 旅行/周末
+        "旅行", "出差", "周末", "短途", "出游", "打包",
+        # 社交
+        "约会", "聚会", "探店", "咖啡馆", "拍照",
+        # 户外
+        "徒步", "爬山", "露营", "户外",
+    ],
+    # 第5层：用户审美与风格认同
+    "aesthetic": [
+        # 颜值设计
+        "好看", "颜值", "设计", "高级", "喜欢这个",
+        # 百搭搭配（"配"用双字避免误匹配）
+        "百搭", "好搭", "搭衣服", "搭配", "风格",
+        # 子风格认同（"潮"用双字避免误匹配）
+        "机能", "潮流", "潮牌", "科技感", "克制", "简约", "秩序",
+    ],
+    # 第6层：品牌情感（长期心智指标）
+    "brand_emotion": [
+        # 推荐意愿
+        "推荐", "安利", "值得", "满意", "惊喜",
+        # 品牌认同
+        "光影行星", "关注", "忠实", "一直用", "老用户",
+        # 复购意向
+        "再买", "第二个", "送人", "给家人", "还会",
+    ],
+}
+
+# 包包品类通用词：用于判断评论是否与"我方产品"相关。
+# 除了品牌名/款式名，评论中出现这些词也视为与我方产品相关。
+BAG_CATEGORY_TERPS = [
+    # 包类通用
+    "包", "背包", "双肩包", "单肩包", "斜挎包", "手提包", "托特包",
+    "邮差包", "胸包", "腰包", "手包", "拎包",
+    # 容量/装载
+    "能装", "装得下", "容量", "放电脑", "放平板", "放手机",
+    "装东西", "装不下", "够装", "收纳",
+    # 背负/功能
+    "背着", "背过", "背起来", "背负", "肩带", "减负",
+    "防泼水", "防水", "防震", "减震",
+    # 材质/做工
+    "面料", "材质", "走线", "五金", "拉链", "做工",
+    # 通勤/出行场景中的包
+    "通勤包", "出差包", "旅行包", "电脑包", "书包",
+    "机车包", "头盔包", "健身包", "运动包",
+]
+
+# 纯寒暄/无信息词：只含这些词（+标点）且未命中任何关键词的评论丢弃
+# 注意：不要把文档中的有效关键词（如"好看"）放进来，否则会被误丢
+CODE_FILTER_NOISE_WORDS = {
+    "好", "棒", "支持", "来了", "赞", "厉害", "不错", "优秀",
+    "喜欢", "可爱", "666", "牛", "太强了", "真棒", "很好",
+}
+
+# 判断纯表情/无意义评论：只要包含中文、英文或数字就不是纯表情
+_HAS_MEANINGFUL_CHAR_RE = re.compile(r"[\u4e00-\u9fa5a-zA-Z0-9]")
+
+# ---------------------------------------------------------------------------
+# 内置提示词（已硬编码，运行时不读取 md 文件）
+# ---------------------------------------------------------------------------
+ANALYSIS_SYSTEM_PROMPT = """# 角色：小红书评论舆情分析专家
+
+你是一位专注小红书生态5年的舆情分析师，擅长将非结构化评论转化为结构化洞察，并给出可落地的口碑管理策略。
+
+## 核心能力
+
+1. **情绪分层**：正向（赞美/推荐/回购/共鸣）、负向（质量/服务/价格/竞品贬低）、中性（咨询/建议/客观描述）；识别情绪烈度（强烈推荐 vs 还行 vs 非常失望）。
+2. **主题聚类**：产品（材质/容量/颜色/重量/五金）、服务（发货/包装/客服/售后）、内容（种草来源/预期落差）、场景（通勤/旅行/约会/健身）、竞品（对比品牌/话术）。
+3. **用户画像**：推断年龄（学生党/上班族/宝妈）、消费力（价格敏感型 vs 直接下单型）、使用场景、审美偏好；分类用户类型（理性测评/感性颜值/价格敏感）。
+4. **风险预警**：≥3人提及同问题 → 品控/服务漏洞；出现"平替""不如XX" → 定位模糊；出现"退货""过敏""坏了""假货" → 严重投诉。分级：P0(安全/假货/严重质量)、P1(批量负面但可控)、P2(偶发吐槽)。
+5. **竞品对比**：正/负面率对比、被夸/被骂维度对比、差异化口碑优势与薄弱环节。
+6. **营销归因**：评论热点与近期投放/促销/新品动作联动，给出口碑效果归因图。
+7. **优化建议**：产品端（高频吐槽改进）、内容端（详情页/笔记关键信息优化）、客服端（FAQ/话术优化）、运营端（种草卖点提炼/补救内容设计）。
+8. **报告输出**：标准化舆情周报/月报（数据概览+主题分析+画像+预警+竞品+建议）。
+
+## 工作原则
+
+- 数据说话：每条结论附数据支撑（如"63%负评集中在尺寸偏小"）
+- 客观中立：不过度美化也不过度恐慌
+- 行动导向：给出"下周一就能做的3件事"
+- 长线思维：关注长期品牌口碑健康指标
+- 用户视角：读出用户"没说出来但藏在语气里的情绪"
+
+## 输出格式
+
+收到评论数据后，按以下结构输出：
+
+### 一、数据概览
+- 评论总数 / 有效评论数
+- 情绪分布：正面X% / 中性X% / 负面X%
+- 互动概览：平均点赞/回复数、高互动Top3
+
+### 二、核心发现（Top 3）
+每条附数据支撑
+
+### 三、分维度深度分析
+| 排名 | 主题标签 | 提及次数 | 正向占比 | 负向占比 | 典型评论 |
+|------|----------|----------|----------|----------|----------|
+| 1 | 材质 | 45 | 80% | 10% | "皮料很软" |
+
+- **情绪趋势**：正/负/中性随时间波动，是否有拐点
+- **用户画像**：高频身份、核心场景、主要决策因素
+- **风险预警**：P0/P1/P2 分级清单
+
+### 四、归因分析
+正向评论高峰由哪些营销动作带来？负面集中反馈由哪些动作引发？
+
+### 五、优化建议（按优先级）
+- **P0（立即）**：问题 → 行动 → 责任方
+- **P1（本周）**：问题 → 行动 → 责任方
+- **P2（下月）**：问题 → 行动 → 责任方
+
+### 六、可复用资产
+- 评论分类标签体系
+- 情绪打分规则
+- 舆情周报模板
+
+## 约束
+
+- 数据量<50条时注明"样本量较小，结论仅供参考"
+- 涉及隐私的评论需脱敏处理
+- 所有建议必须标注责任方（产品/内容/客服/运营）"""
+
+
+def log(message: str) -> None:
+    print(f"[{time.strftime('%H:%M:%S')}] {message}", flush=True)
+
+
+# ---------------------------------------------------------------------------
+# 文件读取
+# ---------------------------------------------------------------------------
+def load_text(path: Path) -> str:
+    if not path.exists():
+        raise FileNotFoundError(f"找不到文件: {path}")
+    for enc in ("utf-8-sig", "utf-8", "gbk"):
+        try:
+            return path.read_text(encoding=enc)
+        except UnicodeDecodeError:
+            continue
+    return path.read_text(encoding="utf-8", errors="replace")
+
+
+def _extract_comment_text(item: Any) -> str:
+    if isinstance(item, str):
+        return item.strip()
+    if isinstance(item, dict):
+        for key in ("comment_text", "content", "text", "message", "comment"):
+            val = item.get(key)
+            if val is not None:
+                return str(val).strip()
+    return str(item).strip()
+
+
+def load_comments(path: Path) -> list[str]:
+    """从 txt/json/csv 加载评论文本列表。"""
+    text = load_text(path)
+    ext = path.suffix.lower()
+
+    if ext == ".json":
+        data = json.loads(text)
+        if isinstance(data, list):
+            return [_extract_comment_text(item) for item in data if _extract_comment_text(item)]
+        if isinstance(data, dict):
+            for key in ("comments", "data", "list", "rows"):
+                if key in data and isinstance(data[key], list):
+                    return [_extract_comment_text(item) for item in data[key] if _extract_comment_text(item)]
+        raise ValueError(f"不支持的 JSON 结构: {path}")
+
+    if ext == ".csv":
+        rows = list(csv.DictReader(text.splitlines()))
+        candidates = ("comment_text", "content", "text", "message", "评论内容", "评论", "comment")
+        comments: list[str] = []
+        for row in rows:
+            for key in candidates:
+                val = row.get(key)
+                if val is not None and str(val).strip():
+                    comments.append(str(val).strip())
+                    break
+        return comments
+
+    # 默认：每行一条评论
+    return [line.strip() for line in text.splitlines() if line.strip()]
+
+
+# ---------------------------------------------------------------------------
+# 按款式分批处理（用于项目汇总 CSV）
+# ---------------------------------------------------------------------------
+def load_style_groups(path: Path, max_per_style: int | None = None) -> dict[str, list[str]]:
+    """从 CSV 中按款式分组读取评论，可选每款式抽样上限。"""
+    text = load_text(path)
+    rows = list(csv.DictReader(text.splitlines()))
+
+    candidates = ("comment_text", "content", "text", "message", "评论内容", "评论", "comment")
+    style_key_candidates = ("style_name", "款式", "style", "产品")
+
+    style_col: str | None = None
+    for row in rows[:1]:
+        for key in style_key_candidates:
+            if key in row:
+                style_col = key
+                break
+
+    groups: dict[str, list[str]] = {}
+    for row in rows:
+        style = row.get(style_col, "未知款式") if style_col else "未知款式"
+        style = style.strip() or "未知款式"
+        for key in candidates:
+            val = row.get(key)
+            if val is not None and str(val).strip():
+                groups.setdefault(style, []).append(str(val).strip())
+                break
+
+    if max_per_style:
+        rng = random.Random(42)
+        for style in groups:
+            if len(groups[style]) > max_per_style:
+                groups[style] = rng.sample(groups[style], max_per_style)
+
+    return dict(sorted(groups.items(), key=lambda item: -len(item[1])))
+
+
+def analyze_style(style: str, comments: list[str]) -> tuple[str, int, int]:
+    """对单个款式的评论进行过滤 + 分析，返回（汇总文本, 原始数, 过滤后数）。"""
+    log(f"【{style}】开始处理（{len(comments)} 条）...")
+    filtered = code_filter_comments(comments)
+    summary = analyze_comments(filtered)
+    return summary, len(comments), len(filtered)
+
+
+def build_batch_report(style_results: list[tuple[str, str, int, int]]) -> str:
+    """把所有款式汇总合并成一份总报告。"""
+    total_raw = sum(raw for _, _, raw, _ in style_results)
+    total_filtered = sum(filt for _, _, _, filt in style_results)
+
+    lines = [
+        "# 多款式评论舆情汇总报告",
+        "",
+        f"- 款式总数：{len(style_results)}",
+        f"- 评论总数（原始）：{total_raw}",
+        f"- 有效评论总数（过滤后）：{total_filtered}",
+        "",
+        "---",
+        "",
+    ]
+    for style, summary, raw, filtered in style_results:
+        lines.append(f"## 款式：{style}")
+        lines.append(f"原始评论：{raw} 条 | 过滤后：{filtered} 条")
+        lines.append("")
+        lines.append(summary)
+        lines.append("")
+        lines.append("---")
+        lines.append("")
+    return "\n".join(lines)
+
+
+# ---------------------------------------------------------------------------
+# Hermes 分析端调用
+# ---------------------------------------------------------------------------
+def call_hermes_analyzer(system_prompt: str, user_content: str) -> str:
+    url = HERMES_ANALYZER_URL.rstrip("/") + "/chat/completions"
+    payload = {
+        "model": HERMES_ANALYZER_MODEL,
+        "messages": [
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": user_content},
+        ],
+    }
+    data = json.dumps(payload, ensure_ascii=False).encode("utf-8")
+    request = urllib.request.Request(
+        url,
+        data=data,
+        headers={
+            "Authorization": f"Bearer {HERMES_ANALYZER_TOKEN}",
+            "Content-Type": "application/json; charset=utf-8",
+        },
+        method="POST",
+    )
+    try:
+        with urllib.request.urlopen(request, timeout=1800) as response:
+            result = json.loads(response.read().decode("utf-8", errors="replace"))
+    except urllib.error.HTTPError as exc:
+        body = exc.read().decode("utf-8", errors="replace")
+        raise RuntimeError(f"Hermes HTTP {exc.code}: {body}") from exc
+
+    if "choices" not in result or not result["choices"]:
+        raise RuntimeError(f"Hermes 返回异常: {result}")
+    return str(result["choices"][0]["message"]["content"])
+
+
+def _is_pure_emoji(text: str) -> bool:
+    """判断是否为纯表情/无意义评论（不含中文、英文、数字）。"""
+    return not _HAS_MEANINGFUL_CHAR_RE.search(text)
+
+
+# 抖音/B站表情格式：[比心] [舔屏] 等，匹配前需去掉避免误触发关键词
+_EMOJI_BRACKET_RE = re.compile(r'\[[^\[\]]{1,10}\]')
+
+
+def _strip_emoji_brackets(text: str) -> str:
+    """去掉 [xxx] 格式的表情符号，避免表情文本误触发关键词匹配。"""
+    return _EMOJI_BRACKET_RE.sub('', text).strip()
+
+
+def _matches_code_keywords(text: str) -> bool:
+    """本地关键词匹配，命中任意有效关键词即保留。"""
+    clean = _strip_emoji_brackets(text)
+    if not clean:
+        return False
+    for keywords in CODE_FILTER_KEYWORDS.values():
+        for kw in keywords:
+            if kw in clean:
+                return True
+    return False
+
+
+def classify_comment_keywords(text: str) -> list[str]:
+    """返回一条评论命中的所有关键词类别（用于统计分布）。"""
+    clean = _strip_emoji_brackets(text)
+    if not clean:
+        return []
+    categories: list[str] = []
+    for category, keywords in CODE_FILTER_KEYWORDS.items():
+        for kw in keywords:
+            if kw in clean:
+                categories.append(category)
+                break
+    return categories
+
+
+def _is_noise_only(text: str) -> bool:
+    """判断是否为纯寒暄/无信息评论（仅由常见寒暄词和标点组成）。"""
+    # 去掉常见寒暄词
+    cleaned = text
+    for w in sorted(CODE_FILTER_NOISE_WORDS, key=len, reverse=True):
+        cleaned = cleaned.replace(w, "")
+    # 再去掉常见标点、空格、数字
+    cleaned = re.sub(r"[\s，,。！!？?~～.·\d]+", "", cleaned).strip()
+    return not cleaned
+
+
+def code_filter_comments(comments: list[str]) -> list[str]:
+    """用本地代码规则快速过滤评论，不调用大模型。"""
+    if not comments:
+        log("没有评论需要过滤")
+        return []
+
+    log(f"使用本地代码规则进行评论过滤（共 {len(comments)} 条）...")
+    seen: set[str] = set()
+    kept: list[str] = []
+    for text in comments:
+        t = text.strip()
+        if not t or t in seen:
+            continue
+        if _is_pure_emoji(t):
+            continue
+        if _matches_code_keywords(t):
+            seen.add(t)
+            kept.append(t)
+            continue
+        # 未命中关键词，但属于短句纯寒暄，也丢弃
+        if len(t) <= 12 and _is_noise_only(t):
+            continue
+        # 其余未命中关键词的评论丢弃
+    log(f"代码过滤完成：{len(comments)} 条 → {len(kept)} 条")
+    return kept
+
+
+# ---------------------------------------------------------------------------
+# LLM 评论清洗(语义过滤,替代/补充关键词过滤)
+# ---------------------------------------------------------------------------
+COMMENT_CLEAN_SYSTEM_PROMPT = """你是"光影行星"品牌的评论清洗助手。光影行星主要经营双肩包、单肩包、斜挎包、摄影包、电脑包、手提包及相关包袋配件。你的任务是过滤无效评论，并从评论中提取对品牌经营、产品优化或用户服务有价值的内容。
+
+有效评论包括：
+1. 产品评价：涉及外观、颜色、尺寸、容量、重量、材质、做工、气味、拉链、肩带、扣具、分区、收纳、防水性、耐用性、舒适度、便携性等。
+2. 使用体验：涉及通勤、旅行、摄影、户外、上学、商务、电脑收纳、相机收纳等真实使用场景。
+3. 产品咨询：询问尺寸、颜色、容量、适配设备、材质、库存、价格、使用方法、清洁保养或不同款式区别。
+4. 购买意向：明确表达想买、准备下单、等待补货、询问购买链接或推荐款式。
+5. 产品问题与售后反馈：涉及破损、开线、掉色、拉链故障、尺寸不符、配件缺失、使用异常、退换货原因等。
+6. 品牌反馈：明确提到光影行星，对品牌设计、质量、价格、服务或产品系列提出评价和建议。
+
+无效评论包括：
+1. 只有表情、数字、标点或无意义字符。
+2. "哈哈哈""路过""支持一下""不错不错"等没有具体信息的泛泛表达。
+3. 广告、引流、联系方式、刷单、抽奖、互关或推广其他商品。
+4. 与光影行星或包袋产品完全无关的内容。
+5. 明显复制粘贴、重复发送或机器生成的灌水内容。
+6. 单纯辱骂、攻击他人且没有任何产品事实或具体问题。
+7. 无法理解、语义残缺且不能确定评论意图的内容。
+8. 只讨论主播、视频背景、音乐或其他无关内容。
+9. 仅询问快递到了哪里，且不包含产品或售后问题。
+10. 没有真实内容的默认好评或模板评价。
+
+判定原则：
+1. 负面评论只要包含具体产品问题，就属于有效评论，不能因为评价负面而删除。
+2. 评论同时包含有效和无效内容时，只提取其中与光影行星及包袋产品有关的有效部分。
+3. 不得修改评论原意，不得虚构用户没有提到的信息。
+4. 可以删除表情、重复语句、无意义语气词和无关内容。
+5. "很好""不错""喜欢"等评论，如果没有说明具体原因，判定为无效。
+6. "容量很大""肩带舒服""拉链不好用"等虽然简短，但包含具体产品信息，应判定为有效。
+7. 无法确定时，优先保留，但标记为"需要人工确认"。
+8. 所有"有效"类别(产品评价/使用体验/产品咨询/购买意向/售后问题/品牌反馈)都必须限定在"光影行星品牌"或"包袋类产品"范围内。如果评论只询问/讨论笔记里出现的其他物品(如镜子、指甲刀、U型枕、蚊虫膏、行李箱、自拍支架、护照收纳袋、登山杖等非包袋物品),即使有购买意向或产品咨询,也判定为无效。
+9. 评论同时提到包袋和其他物品时,只看包袋相关部分:若包袋部分有具体信息(评价/咨询/购买意向等),按有效处理;若包袋部分只是被顺带提及或无具体信息,判定为无效。
+
+每条评论按照以下格式输出（不要输出其它内容）：
+评论编号：<编号>
+是否有效：有效/无效/需要人工确认
+有效评论：<提取并清洗后的有效内容；无有效内容则填写"无">
+评论类型：产品评价/使用体验/产品咨询/购买意向/售后问题/品牌反馈/无效内容
+情感倾向：正面/中性/负面/无法判断
+涉及维度：外观、颜色、尺寸、容量、材质、做工、舒适度、收纳、防水、耐用、价格、配件、售后、其他
+无效原因：<有效评论填写"无"；无效评论说明具体原因>"""
+
+
+_CLEAN_FIELD_RE = re.compile(
+    r"(是否有效|有效评论|评论类型|情感倾向|涉及维度|无效原因)\s*[：:]\s*(.*)"
+)
+
+
+def parse_cleaning_response(response: str) -> list[dict]:
+    """解析 LLM 结构化输出。每条评论 7 行,按"评论编号："分块。
+    返回 [{valid, cleaned, type, sentiment, dimensions, reason}, ...]
+    """
+    items: list[dict] = []
+    # 按"评论编号：<数字>"分块,跳过首块(分割后的空块)
+    blocks = re.split(r"评论编号\s*[：:]\s*\d+", response)
+    for block in blocks[1:]:
+        item = {
+            "valid": "无效",
+            "cleaned": "",
+            "type": "无效内容",
+            "sentiment": "无法判断",
+            "dimensions": "其他",
+            "reason": "无",
+        }
+        for line in block.split("\n"):
+            line = line.strip()
+            if not line:
+                continue
+            m = _CLEAN_FIELD_RE.match(line)
+            if not m:
+                continue
+            key, val = m.group(1), m.group(2).strip()
+            if key == "是否有效":
+                item["valid"] = val
+            elif key == "有效评论":
+                if val and val != "无":
+                    item["cleaned"] = val
+            elif key == "评论类型":
+                item["type"] = val
+            elif key == "情感倾向":
+                item["sentiment"] = val
+            elif key == "涉及维度":
+                item["dimensions"] = val
+            elif key == "无效原因":
+                item["reason"] = val
+        items.append(item)
+    return items
+
+
+def _trivial_prefilter(comments: list[str]) -> list[str]:
+    """轻量预过滤:只去掉空、纯表情、重复,其余全留(给 LLM 判断)。"""
+    seen: set[str] = set()
+    kept: list[str] = []
+    for text in comments:
+        t = str(text or "").strip()
+        if not t or t in seen:
+            continue
+        if _is_pure_emoji(t):
+            continue
+        seen.add(t)
+        kept.append(t)
+    return kept
+
+
+def _clean_one_batch(batch: list[str], batch_idx: int,
+                     total_batches: int) -> tuple[int, list[dict]]:
+    """清洗一批评论,返回 (batch_idx, results)。
+    并发安全:不写共享状态,只返回结果。
+    """
+    lines = [f"评论编号：{i+1}\n评论内容：{text}" for i, text in enumerate(batch, 1)]
+    user_content = (f"请按规则清洗以下 {len(batch)} 条评论,逐条输出结构化结果。\n\n"
+                    + "\n\n".join(lines))
+    results: list[dict] = []
+    try:
+        response = call_hermes_analyzer(COMMENT_CLEAN_SYSTEM_PROMPT, user_content)
+        parsed = parse_cleaning_response(response)
+        for i, item in enumerate(parsed):
+            if i < len(batch):
+                item["original"] = batch[i]
+        if len(parsed) < len(batch):
+            for i in range(len(parsed), len(batch)):
+                results.append({
+                    "original": batch[i],
+                    "valid": "需要人工确认",
+                    "cleaned": batch[i],
+                    "type": "无效内容",
+                    "sentiment": "无法判断",
+                    "dimensions": "其他",
+                    "reason": "LLM 响应未解析到该条",
+                })
+        results.extend(parsed)
+    except Exception as exc:
+        for text in batch:
+            results.append({
+                "original": text,
+                "valid": "需要人工确认",
+                "cleaned": text,
+                "type": "无效内容",
+                "sentiment": "无法判断",
+                "dimensions": "其他",
+                "reason": f"LLM 调用失败: {exc}",
+            })
+    return batch_idx, results
+
+
+def llm_clean_comments_detailed(comments: list[str],
+                                batch_size: int = 30,
+                                max_workers: int = 6,
+                                include_invalid: bool = False) -> list[dict]:
+    """用 LLM 对评论做语义清洗(并发批处理)。
+    返回 [{original, valid, cleaned, type, sentiment, dimensions, reason}, ...]
+    默认只返回 valid != "无效" 的评论;传 include_invalid=True 返回全部。
+    max_workers: 并发批数(默认 6,Hermes 本地端可承受)
+    """
+    if not comments:
+        return []
+    prefiltered = _trivial_prefilter(comments)
+    log(f"LLM 清洗:预过滤 {len(comments)} -> {len(prefiltered)} 条(去空/去重/去纯表情)")
+
+    total = len(prefiltered)
+    total_batches = (total + batch_size - 1) // batch_size
+
+    # 切分批次
+    batches: list[tuple[int, list[str]]] = []
+    for batch_idx in range(total_batches):
+        start = batch_idx * batch_size
+        end = min(start + batch_size, total)
+        batches.append((batch_idx, prefiltered[start:end]))
+
+    # 并发跑
+    results_by_idx: dict[int, list[dict]] = {}
+    completed = 0
+    with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
+        future_to_idx = {executor.submit(_clean_one_batch, batch, idx, total_batches): idx
+                         for idx, batch in batches}
+        for future in concurrent.futures.as_completed(future_to_idx):
+            idx = future_to_idx[future]
+            try:
+                batch_idx, batch_results = future.result()
+                results_by_idx[batch_idx] = batch_results
+            except Exception as exc:
+                log(f"  [WARN] 批次 {idx+1} 异常: {exc}")
+                results_by_idx[idx] = []
+            completed += 1
+            if completed % 3 == 0 or completed == total_batches:
+                log(f"  进度: {completed}/{total_batches} 批完成")
+
+    # 按原顺序合并
+    all_results: list[dict] = []
+    for idx in range(total_batches):
+        all_results.extend(results_by_idx.get(idx, []))
+
+    kept = [r for r in all_results if r.get("valid") != "无效"]
+    valid_count = sum(1 for r in all_results if r.get("valid") == "有效")
+    unsure_count = sum(1 for r in all_results if r.get("valid") == "需要人工确认")
+    invalid_count = sum(1 for r in all_results if r.get("valid") == "无效")
+    log(f"LLM 清洗完成: 共 {len(all_results)} 条 -> 有效 {valid_count} / "
+        f"需人工 {unsure_count} / 无效 {invalid_count}")
+    return all_results if include_invalid else kept
+
+
+def llm_clean_comments(comments: list[str], batch_size: int = 30,
+                       use_cleaned: bool = True, max_workers: int = 6) -> list[str]:
+    """LLM 清洗评论,返回清洗后的评论文本列表(只含有效+需要人工确认)。
+    use_cleaned=True 返回清洗后内容,False 返回原文。"""
+    details = llm_clean_comments_detailed(comments, batch_size=batch_size,
+                                          max_workers=max_workers)
+    out: list[str] = []
+    for d in details:
+        text = d.get("cleaned") if use_cleaned else d.get("original")
+        if not text:
+            text = d.get("original", "")
+        if text:
+            out.append(text)
+    return out
+
+
+# ---------------------------------------------------------------------------
+# 舆情分析
+# ---------------------------------------------------------------------------
+def analyze_comments(comments: list[str]) -> str:
+    if not comments:
+        return "无有效评论可分析。"
+
+    log(f"调用 Hermes 分析端生成舆情汇总（共 {len(comments)} 条）...")
+    lines = [f"{i + 1}. {text}" for i, text in enumerate(comments)]
+    user_content = (
+        f"共有 {len(comments)} 条过滤后的有效评论，请根据角色设定输出完整的舆情分析汇总报告。\n\n"
+        + "\n".join(lines)
+    )
+    return call_hermes_analyzer(ANALYSIS_SYSTEM_PROMPT, user_content)
+
+
+# ---------------------------------------------------------------------------
+# 输出 & 飞书通知
+# ---------------------------------------------------------------------------
+def write_summary(summary: str, path: Path) -> Path:
+    path.write_text(summary, encoding="utf-8-sig")
+    log(f"汇总报告已写入: {path}")
+    return path
+
+
+def send_feishu_summary(summary: str, open_id: str) -> dict[str, Any]:
+    """通过 lark-cli --profile hermes-analyzer 以分析端应用身份发送汇总。"""
+    text = summary
+    if len(text) > FEISHU_TEXT_LIMIT:
+        text = text[:FEISHU_TEXT_LIMIT] + "\n\n...（内容过长，已截断，完整内容见项目 data/summary 汇总文件）"
+
+    command = [
+        "lark-cli",
+        "--profile", "hermes-analyzer",
+        "im", "+messages-send",
+        "--user-id", open_id,
+        "--text", text,
+        "--as", "bot",
+        "--format", "json",
+    ]
+    completed = subprocess.run(
+        command,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+    )
+    if completed.returncode != 0:
+        raise RuntimeError(
+            f"lark-cli 发送失败 (exit {completed.returncode}):\n"
+            f"stdout: {completed.stdout}\nstderr: {completed.stderr}"
+        )
+    try:
+        return json.loads(completed.stdout)
+    except json.JSONDecodeError:
+        return {"raw": completed.stdout}
+
+
+# ---------------------------------------------------------------------------
+# CLI
+# ---------------------------------------------------------------------------
+def main() -> int:
+    parser = argparse.ArgumentParser(description="评论 → Hermes 过滤/分析 → 桌面汇总 + 飞书通知")
+    parser.add_argument("--input", type=Path, default=DEFAULT_INPUT, help="原始评论文件路径")
+    parser.add_argument("--output", type=Path, default=OUTPUT_PATH, help="汇总输出文件路径")
+    parser.add_argument("--open-id", default=TARGET_OPEN_ID, help="飞书接收人 open_id")
+    parser.add_argument("--no-send", action="store_true", help="只生成汇总文件，不发送飞书")
+    parser.add_argument("--batch-by-style", action="store_true", help="CSV 按款式分批分析（用于项目汇总 CSV）")
+    parser.add_argument("--max-per-style", type=int, default=300, help="每款式最多分析多少条评论（防上下文超长）")
+    args = parser.parse_args()
+    args.output = resolve_layer_output(
+        args.output, layer_root=PATHS.exports_root, field="--output"
+    )
+
+    if not args.no_send:
+        if not HERMES_ANALYZER_TOKEN:
+            log("[ERROR] HERMES_ANALYZER_TOKEN 未设置。请配置 data/config/analyze.env 或传 --no-send")
+            return 1
+        if not args.open_id:
+            log("[ERROR] 飞书接收人 open_id 未设置。请配置 FEISHU_TARGET_OPEN_ID 或传 --open-id 或 --no-send")
+            return 1
+
+    if args.batch_by_style:
+        log(f"按款式分批读取 CSV: {args.input}")
+        try:
+            style_groups = load_style_groups(args.input, args.max_per_style)
+        except Exception as exc:
+            log(f"读取 CSV 分组失败: {exc}")
+            return 1
+        log(f"共 {len(style_groups)} 个款式")
+        if not style_groups:
+            log("没有读取到任何评论，流程结束")
+            return 0
+
+        style_results: list[tuple[str, str, int, int]] = []
+        for style, comments in style_groups.items():
+            try:
+                summary, raw_count, filtered_count = analyze_style(style, comments)
+                style_results.append((style, summary, raw_count, filtered_count))
+            except Exception as exc:
+                log(f"【{style}】处理失败: {exc}")
+                style_results.append((style, f"处理失败: {exc}", len(comments), 0))
+
+        summary = build_batch_report(style_results)
+    else:
+        log(f"读取原始评论文件: {args.input}")
+        try:
+            comments = load_comments(args.input)
+        except Exception as exc:
+            log(f"读取评论失败: {exc}")
+            return 1
+        log(f"读取到 {len(comments)} 条评论")
+
+        if not comments:
+            log("没有读取到任何评论，流程结束")
+            return 0
+
+        try:
+            filtered = code_filter_comments(comments)
+        except Exception as exc:
+            log(f"评论过滤失败: {exc}")
+            return 1
+
+        try:
+            summary = analyze_comments(filtered)
+        except Exception as exc:
+            log(f"舆情分析失败: {exc}")
+            return 1
+
+    try:
+        write_summary(summary, args.output)
+    except Exception as exc:
+        log(f"写入汇总文件失败: {exc}")
+        return 1
+
+    if args.no_send:
+        log("已指定 --no-send，跳过飞书发送")
+    else:
+        try:
+            result = send_feishu_summary(summary, args.open_id)
+            log(f"飞书消息发送成功: {result}")
+        except Exception as exc:
+            log(f"飞书消息发送失败（汇总文件仍已生成）: {exc}")
+            return 1
+
+    log("全部完成")
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\n已中断", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/analyze_note.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/analyze_note.py
new file mode 100644
index 0000000..03e05ea
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/analyze_note.py
@@ -0,0 +1,868 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""单篇笔记分析报告生成器。
+
+读取项目采集的单篇笔记 JSON（data/notes/{platform}/...json），
+结合笔记标题、点赞/收藏/评论/分享等基础指标，以及评论内容，
+生成本地代码过滤 + Hermes 舆情分析的单篇笔记分析报告。
+
+输出保存到：data/summary/单篇笔记分析报告.txt
+"""
+
+import argparse
+import json
+import os
+import subprocess
+import sys
+import time
+from pathlib import Path
+from typing import Any
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import (
+    PATHS,
+    resolve_layer_output,
+)
+
+# Path setup
+_TOOLS_DIR = PATHS.tools_root
+_PROJECT_ROOT = PATHS.module_root
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools.analyze_comments import (
+    HERMES_ANALYZER_MODEL,
+    HERMES_ANALYZER_TOKEN,
+    HERMES_ANALYZER_URL,
+    FEISHU_TEXT_LIMIT,
+    TARGET_OPEN_ID,
+    BAG_CATEGORY_TERPS,
+    code_filter_comments,
+    llm_clean_comments,
+    llm_clean_comments_detailed,
+    classify_comment_keywords,
+    call_hermes_analyzer,
+)
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools import db
+
+PROJECT_ROOT = _PROJECT_ROOT
+DATA_DIR = PATHS.raw_root
+SUMMARY_DIR = PATHS.exports_root / "summary"
+NOTES_DIR = PATHS.raw_root / "notes"
+
+OUTPUT_PATH = SUMMARY_DIR / "单篇笔记分析报告.txt"
+
+CATEGORY_LABELS = {
+    "purchase_intent": "购买意向",
+    "positive_feature": "正向功能反馈",
+    "negative_feature": "负向功能反馈",
+    "user_scenario": "用户场景",
+    "aesthetic": "审美与风格",
+    "brand_emotion": "品牌情感",
+}
+
+# 单篇笔记分析：送入 LLM 的有效评论上限
+# 超过此数量时按关键词多样性抽样，保留最有信息量的评论
+MAX_COMMENTS_FOR_LLM = 80
+
+
+# ---------------------------------------------------------------------------
+# 单篇笔记分析的 LLM 提示词
+# ---------------------------------------------------------------------------
+SINGLE_NOTE_PROMPT = """你是一位品牌投放复盘分析师。请基于以下单篇笔记的数据与评论区内容，按指定结构输出分析报告。
+
+【笔记信息】
+- 标题：{title}
+- 平台：{platform}
+- 品牌/产品：{brand} / {product}
+- 点赞：{likes} | 收藏：{favs} | 评论：{comments} | 分享：{shares}
+- 曝光/播放量：{views}（如有）
+
+【有效评论列表】
+{有效评论原文，逐条列出}
+
+【输出要求】
+请按以下7个模块输出结论，每个模块直接给出判断和建议，不重复提及数据来源或指标定义，引用评论用「」包裹原文：
+
+模块1｜笔记表现归因
+- 基于互动率结构（评论/收藏/转发）判断这条笔记的传播属性：是认知留存型、社交裂变型还是搜索截流型？
+- 结合曝光量（如有）判断流量效率是否匹配投放预期。
+- 一句话总结：这条笔记"成也何处，败也何处"。
+
+模块2｜评论质量分层
+- 将评论按"有效种草 / 连带兴趣 / 无效互动"三档归类，给出各档条数与占比。
+- 无效互动的典型表现是什么？（如纯求链接、纯@好友、无意义表情）
+- 有效评论中，高频关键词指向哪些用户决策阶段？（规格确认、价值判断、竞品对比、冲动下单、售后反馈）
+
+模块3｜品牌安全与控评建议
+- 评论区是否存在负面舆情集中点？若有，具体是什么问题？影响面多大？
+- 是否出现竞品高频被提及？提及场景是"对比纠结"还是"被替代"？
+- 给出具体的控评/引导话术建议，至少2条可挂置顶或评论区回复的原文。
+
+模块4｜脚本有效性拆解
+- 标题钩子是否有效？是吸引点击还是引发质疑？
+- 卖点传达是否精准？评论区讨论的卖点与笔记主推卖点是否一致？如果偏离，偏离方向是什么？
+- 用户是否按预期行动？（如搜索品牌词、进店铺、点击评论区链接等）
+- 内容调性是否与品牌心智匹配？若匹配度不足，差异点在哪？
+
+模块5｜达人匹配度评估
+- 该达人的粉丝画像（性别、年龄、兴趣）与我方目标客群的重合度如何？
+- 达人的内容风格（硬核测评/温柔种草/剧情演绎/颜值展示）是否适合承载本产品的核心卖点？
+- 如果匹配度不足，是"换人"还是"改脚本"？具体建议。
+
+模块6｜可沉淀的选题资产
+- 从评论区提取至少2条可作为下一期内容选题的真实用户需求或疑问。
+- 每条选题需附：内容方向+标题参考+核心脚本建议。
+
+模块7｜下一篇笔记可执行优化清单
+- 脚本层面（前3秒钩子、卖点顺序、画面设计、结尾引导）逐项给出可落地的修改点。
+- 评论区运营层面（预埋评论、置顶引导、舆情应对）给出具体话术。
+- 投放策略层面（达人选择、内容方向、流量助推）给出调整建议。
+
+【输出格式要求】
+- 不要出现"根据数据""数据显示""通过分析可知"等来源类表述。
+- 直接给结论和建议，每条建议须具体可执行，避免"加强""优化"等空泛词汇。
+- 如我方产品在笔记中被边缘化，须在模块4末尾明确加粗标注"⚠️ 我方产品边缘化，非核心种草"。
+- 整体篇幅控制在1500-2000字，结论前置，层次分明。"""
+
+
+def log(message: str) -> None:
+    print(f"[{time.strftime('%H:%M:%S')}] {message}", flush=True)
+
+
+def detect_platform(source_url: str) -> str:
+    url = source_url.lower()
+    if "bilibili" in url or "b23.tv" in url or url.startswith("bv"):
+        return "bilibili"
+    if "douyin" in url:
+        return "douyin"
+    if "xiaohongshu" in url or "xhslink" in url:
+        return "xiaohongshu"
+    return "unknown"
+
+
+def safe_div(numerator: Any, denominator: Any) -> float:
+    try:
+        n = float(numerator)
+        d = float(denominator)
+        if d == 0:
+            return 0.0
+        return n / d
+    except (TypeError, ValueError):
+        return 0.0
+
+
+def load_note_json(path: Path) -> dict[str, Any]:
+    if not path.exists():
+        raise FileNotFoundError(f"找不到文件: {path}")
+    for enc in ("utf-8-sig", "utf-8", "gbk"):
+        try:
+            data = json.loads(path.read_text(encoding=enc))
+            break
+        except UnicodeDecodeError:
+            continue
+        except json.JSONDecodeError:
+            raise
+    else:
+        data = json.loads(path.read_text(encoding="utf-8", errors="replace"))
+
+    if not isinstance(data, dict):
+        raise ValueError(f"不支持的 JSON 结构: {path}")
+    return data
+
+
+def extract_note_info(data: dict[str, Any]) -> dict[str, Any]:
+    """从笔记 JSON 中提取标题、URL、指标等。"""
+    source_url = data.get("source_url") or data.get("final_url") or ""
+    platform = detect_platform(source_url)
+
+    # 标题：优先顶层 title，兼容 bilibili 旧数据的 video.title
+    title = data.get("title") or ""
+    if not title and isinstance(data.get("video"), dict):
+        title = data["video"].get("title", "")
+
+    stats = data.get("stats") or {}
+    metrics = stats.get("note_metrics") or {}
+
+    return {
+        "title": title,
+        "platform": platform,
+        "source_url": source_url,
+        "scraped_at": data.get("scraped_at", ""),
+        "comment_count_total": data.get("count", 0),
+        "like_count": metrics.get("like_count", ""),
+        "favorite_count": metrics.get("favorite_count", ""),
+        "comment_count_metric": metrics.get("comment_count", ""),
+        "share_count": metrics.get("share_count", ""),
+        "view_count": metrics.get("view_count", ""),
+        "collect_count_yxyy": metrics.get("collect_count", ""),
+        "style_name": "",
+        "creator_name": "",
+    }
+
+
+def extract_comment_texts(data: dict[str, Any]) -> list[str]:
+    """从笔记 JSON 中提取评论文本列表。"""
+    comments = data.get("comments") or []
+    texts: list[str] = []
+    for c in comments:
+        if not isinstance(c, dict):
+            continue
+        for key in ("content", "text", "message"):
+            val = c.get(key)
+            if val is not None:
+                text = str(val).strip()
+                if text:
+                    texts.append(text)
+                break
+    return texts
+
+
+def load_note_from_db(note_id: int | None = None, url: str | None = None) -> tuple[dict[str, Any], list[str]]:
+    """从数据库读取笔记信息和评论文本。
+
+    返回 (note_info_dict, comment_texts_list)。
+    只查 cmt_notes（含 cmt_styles / cmt_creators JOIN）+ cmt_comments。
+    """
+    if note_id is None and not url:
+        raise ValueError("必须提供 note_id 或 url")
+
+    if note_id is not None:
+        note_row = db.get_note_by_id(note_id)
+    else:
+        note_row = db.get_note_by_url(url)
+
+    if not note_row:
+        raise ValueError(f"数据库中找不到笔记: note_id={note_id}, url={url}")
+
+    note_id = note_row["id"]
+    comment_rows = db.get_comments_by_note_id(note_id)
+
+    info = {
+        "title": note_row.get("title") or "",
+        "platform": note_row.get("platform") or "",
+        "source_url": note_row.get("url") or "",
+        "scraped_at": note_row.get("scraped_at").isoformat(timespec="seconds") if note_row.get("scraped_at") else "",
+        "comment_count_total": len(comment_rows),
+        "like_count": note_row.get("like_count", ""),
+        "favorite_count": note_row.get("favorite_count", ""),
+        "comment_count_metric": note_row.get("comment_count", ""),
+        "share_count": note_row.get("share_count", ""),
+        "view_count": note_row.get("view_count", ""),
+        "collect_count_yxyy": note_row.get("collect_count", ""),
+        "style_name": note_row.get("style_name") or "",
+        "creator_name": note_row.get("creator_name") or "",
+    }
+
+    texts = [str(row.get("content") or "").strip() for row in comment_rows if str(row.get("content") or "").strip()]
+    return info, texts
+
+
+def _to_number(value: Any) -> float:
+    try:
+        if value is None or value == "":
+            return 0.0
+        return float(value)
+    except (TypeError, ValueError):
+        return 0.0
+
+
+def compute_metrics(info: dict[str, Any]) -> dict[str, Any]:
+    """计算互动率等衍生指标。所有互动率统一以「曝光量」为分母，避免多口径混用。"""
+    likes = info["like_count"]
+    favorites = info["favorite_count"]
+    comments = info["comment_count_metric"] or info["comment_count_total"]
+    shares = info["share_count"]
+    views_raw = info.get("view_count")
+
+    likes_num = _to_number(likes)
+    favorites_num = _to_number(favorites)
+    comments_num = _to_number(comments)
+    shares_num = _to_number(shares)
+    views_num = _to_number(views_raw)
+
+    return {
+        "view_count": views_num,
+        "like_rate": safe_div(likes_num, views_num),       # 点赞率
+        "comment_rate": safe_div(comments_num, views_num),  # 评论率（评论/曝光，全报告唯一口径）
+        "favorite_rate": safe_div(favorites_num, views_num),  # 收藏率
+        "share_rate": safe_div(shares_num, views_num),      # 分享率
+        "engagement_rate": safe_div(likes_num + comments_num + favorites_num + shares_num, views_num),  # 综合互动率
+        "favorite_to_comment": safe_div(favorites_num, comments_num),  # 收藏评论比（独立口径）
+    }
+
+
+def keyword_distribution(comments: list[str], target_terms: list[str] | None = None) -> dict[str, Any]:
+    """统计关键词命中分布，区分「全评论」与「我方产品相关」。
+
+    target_terms：品牌名/款式名列表；提供时把命中文本先筛出再统计，得到「我方产品相关」分布。
+    匹配规则：评论中出现品牌名/款式名，或出现 BAG_CATEGORY_TERPS 中的包包品类词，均视为"我方产品相关"。
+    返回:
+        {
+            "overall": {category: count, ...},         # 全评论 6 类命中数
+            "relevant": {category: count, ...},        # 我方产品相关子集的 6 类命中数
+            "relevant_total": int,                     # 我方产品相关评论条数
+        }
+    """
+    terms = [t for t in (target_terms or []) if t]
+    bag_terms = BAG_CATEGORY_TERPS
+    overall: dict[str, int] = {key: 0 for key in CATEGORY_LABELS}
+    relevant: dict[str, int] = {key: 0 for key in CATEGORY_LABELS}
+    relevant_total = 0
+    for text in comments:
+        cats = classify_comment_keywords(text)
+        for c in cats:
+            overall[c] += 1
+        is_relevant = False
+        if terms and any(t in text for t in terms):
+            is_relevant = True
+        elif any(t in text for t in bag_terms):
+            is_relevant = True
+        if is_relevant:
+            relevant_total += 1
+            for c in cats:
+                relevant[c] += 1
+    return {
+        "overall": overall,
+        "relevant": relevant,
+        "relevant_total": relevant_total,
+    }
+
+
+def _brand_block(brand: str, style_name: str, target_term: str) -> str:
+    if not (brand or style_name):
+        return ""
+    return f"""
+
+【实体隔离】
+本次分析的目标：品牌「{brand or '未指定'}」、产品「{style_name or '未指定'}」。本笔记可能涉及多品牌联合推广。
+请严格将分析对象限定为我方品牌/产品：
+- 仅将「{target_term}」相关的信息计入种草指标
+- 笔记/评论中涉及的其他品牌或产品，仅作为场景搭配或联合推广的陪衬
+- 评论区数据先按以下三类分类再纳入分析：
+  · 有效种草：明确提及「{target_term}」的卖点/外观/体验，或询问其价格/链接
+  · 连带兴趣：既讨论了「{target_term}」也讨论了其他产品
+  · 无效/他品互动：仅讨论其他产品或完全无关闲聊
+- 情感倾向、痛点需求、卖点提取，仅基于「有效种草」与「连带兴趣」生成；「无效/他品互动」不计入种草指标
+- 关键词分布按「全评论」与「我方产品相关」两个口径分别统计
+"""
+
+
+def _data_table_markdown(info: dict[str, Any], metrics: dict[str, Any], raw_total: int, filtered_total: int) -> str:
+    """构造数据表 markdown（用于提示词 / 报告 / 卡片）。"""
+    has_view = bool(metrics.get("view_count"))
+    def pct(v: float) -> str:
+        return f"{v * 100:.2f}%" if v else "—"
+    def num(v: Any) -> str:
+        if v in ("", None):
+            return "—"
+        try:
+            return f"{int(float(v)):,}"
+        except (TypeError, ValueError):
+            return str(v)
+
+    likes = info.get("like_count")
+    favorites = info.get("favorite_count")
+    comments = info.get("comment_count_metric") or info.get("comment_count_total")
+    shares = info.get("share_count")
+    valid_ratio = safe_div(filtered_total, raw_total)
+
+    lines = [
+        "| 维度 | 指标 | 数值 |",
+        "| --- | --- | --- |",
+        f"| 基础 | 点赞 | {num(likes)} |",
+        f"| 基础 | 收藏 | {num(favorites)} |",
+        f"| 基础 | 评论 | {num(comments)} |",
+        f"| 基础 | 分享 | {num(shares)} |",
+        f"| 传播 | 曝光量 | {num(metrics.get('view_count')) if has_view else '未采集'} |",
+        f"| 互动率 | 点赞率 | {pct(metrics.get('like_rate', 0))} |",
+        f"| 互动率 | 评论率 | {pct(metrics.get('comment_rate', 0))} |",
+        f"| 互动率 | 收藏率 | {pct(metrics.get('favorite_rate', 0))} |",
+        f"| 互动率 | 分享率 | {pct(metrics.get('share_rate', 0))} |",
+        f"| 互动率 | 综合互动率 | {pct(metrics.get('engagement_rate', 0))} |",
+        f"| 互动率 | 收藏评论比 | {metrics.get('favorite_to_comment', 0):.2f} |",
+        f"| 质量 | 有效率 | {pct(valid_ratio)} |",
+    ]
+    return "\n".join(lines)
+
+
+def _keyword_distribution_markdown(
+    distribution: dict[str, Any],
+    total_filtered: int,
+) -> str:
+    """构造关键词分布 markdown，含「全评论」与「我方产品相关」两段。"""
+    overall = distribution.get("overall", {}) if isinstance(distribution, dict) else {}
+    relevant = distribution.get("relevant", {}) if isinstance(distribution, dict) else {}
+    relevant_total = distribution.get("relevant_total", 0) if isinstance(distribution, dict) else 0
+    if not overall:
+        return "（无关键词分布数据）"
+
+    def fmt(d: dict[str, int]) -> str:
+        return "\n".join(f"- {CATEGORY_LABELS[k]}：{d.get(k, 0)} 条" for k in CATEGORY_LABELS)
+
+    lines = [
+        f"### 全评论（{total_filtered} 条）",
+        fmt(overall),
+        "",
+        f"### 我方产品相关（{relevant_total} 条）",
+        fmt(relevant) if relevant_total else "_（无评论明确提及我方产品/品牌）_",
+    ]
+    return "\n".join(lines)
+
+
+def _sample_comments(comments: list[str], max_count: int) -> list[str]:
+    """当评论数超过上限时，按关键词多样性抽样保留最有信息量的评论。
+
+    策略：优先保留命中更多关键词类别的评论，同类别数时保留更长的评论。
+    """
+    scored: list[tuple[int, int, int, str]] = []  # (负类别数, 负长度, 原始索引, 文本)
+    for i, text in enumerate(comments):
+        cats = classify_comment_keywords(text)
+        scored.append((-len(cats), -len(text), i, text))
+    scored.sort()
+    return [text for _, _, _, text in scored[:max_count]]
+
+
+def build_note_analysis_prompt(
+    info: dict[str, Any],
+    metrics: dict[str, Any],
+    distribution: dict[str, Any],
+    comments: list[str],
+) -> str:
+    """构造 LLM 输入的 user_content。"""
+    brand = info.get("brand") or ""
+    style_name = info.get("style_name") or ""
+    target_label = " / ".join(x for x in (brand, style_name) if x) if (brand or style_name) else "未指定"
+    target_term = brand or style_name
+
+    # 评论数量截断：超过上限时按关键词多样性抽样
+    total_count = len(comments)
+    if total_count > MAX_COMMENTS_FOR_LLM:
+        comments = _sample_comments(comments, MAX_COMMENTS_FOR_LLM)
+
+    lines: list[str] = []
+    if total_count > MAX_COMMENTS_FOR_LLM:
+        lines.append(f"共有 {total_count} 条过滤后的有效评论（已抽样 {len(comments)} 条），请基于以下单篇笔记的信息和评论内容，按要求输出分析报告。")
+    else:
+        lines.append(f"共有 {len(comments)} 条过滤后的有效评论，请基于以下单篇笔记的信息和评论内容，按要求输出分析报告。")
+
+    lines.extend([
+        "",
+        "## 单篇笔记基础信息",
+        f"- 标题：{info['title'] or '未获取'}",
+        f"- 平台：{info['platform']}",
+        f"- 链接：{info['source_url']}",
+        f"- 品牌/产品：{target_label}",
+        f"- 点赞数：{info['like_count']}",
+        f"- 收藏数：{info['favorite_count']}",
+        f"- 评论数：{info['comment_count_metric'] or info['comment_count_total']}",
+        f"- 分享数：{info['share_count']}",
+        "",
+    ])
+
+    if brand or style_name:
+        lines.extend([
+            "## 实体隔离规则（产品隔离）",
+            f"本次分析的目标：品牌「{brand}」、产品「{style_name}」。本笔记可能涉及多品牌联合推广，请严格将分析对象限定为我方品牌/产品。",
+            f"- 仅将「{target_term}」相关的信息计入种草指标",
+            "- 笔记/评论中涉及的其他品牌或产品，仅作为场景搭配或联合推广的陪衬",
+            "- 评论区按以下三类分类后再纳入分析：",
+            f"  · 有效种草：明确提及「{target_term}」的卖点/外观/使用体验，或询问其价格/链接",
+            f"  · 连带兴趣：既讨论了「{target_term}」也讨论了其他产品",
+            "  · 无效/他品互动：仅讨论其他产品或完全无关闲聊",
+            "- 情感倾向、痛点需求、卖点提取，仅基于「有效种草」与「连带兴趣」生成；「无效/他品互动」不计入种草指标",
+            "",
+        ])
+
+    lines.extend([
+        "## 数据表（已计算好，可直接引用）",
+        _data_table_markdown(info, metrics, info.get("comment_count_total", 0), len(comments)),
+        "",
+        "## 评论关键词分布（按口径分别给出）",
+        _keyword_distribution_markdown(distribution, len(comments)),
+        "",
+        "## 有效评论内容",
+    ])
+    for text in comments:
+        lines.append(f"- {text}")
+    return "\n".join(lines)
+
+
+def analyze_note(
+    info: dict[str, Any],
+    metrics: dict[str, Any],
+    distribution: dict[str, Any],
+    comments: list[str],
+) -> str:
+    """调用 Hermes 生成单篇笔记分析。"""
+    if not comments:
+        return "该笔记无有效评论可分析。"
+
+    log(f"调用 Hermes 生成分析报告（{info['title'] or info['source_url']}，{len(comments)} 条有效评论）...")
+    user_content = build_note_analysis_prompt(info, metrics, distribution, comments)
+
+    brand = info.get("brand") or ""
+    style_name = info.get("style_name") or ""
+    target_term = brand or style_name
+    brand_block = _brand_block(brand, style_name, target_term)
+
+    system_prompt = SINGLE_NOTE_PROMPT + brand_block
+
+    return call_hermes_analyzer(system_prompt, user_content)
+
+
+def build_report(
+    info: dict[str, Any],
+    metrics: dict[str, Any],
+    distribution: dict[str, Any],
+    raw_comments: list[str],
+    filtered_comments: list[str],
+    analysis: str,
+) -> str:
+    """构造本地报告文件。"""
+    return _build_report_concise(info, metrics, distribution, raw_comments, filtered_comments, analysis)
+
+
+def _build_report_concise(
+    info: dict[str, Any],
+    metrics: dict[str, Any],
+    distribution: dict[str, Any],
+    raw_comments: list[str],
+    filtered_comments: list[str],
+    analysis: str,
+) -> str:
+    brand = info.get("brand") or ""
+    style_name = info.get("style_name") or ""
+    target_label = " / ".join(x for x in (brand, style_name) if x) if (brand or style_name) else "未指定"
+
+    lines = [
+        "# 单篇笔记分析报告（投放复盘版）",
+        "",
+        f"**标题**：{info['title'] or '未获取'}",
+        f"**平台**：{info['platform']}",
+        f"**品牌/产品**：{target_label}",
+        f"**达人**：{info.get('creator_name') or '未指定'}",
+        f"**链接**：{info['source_url']}",
+        "",
+        "## 一、数据表",
+        _data_table_markdown(info, metrics, len(raw_comments), len(filtered_comments)),
+        "",
+        "## 二、关键词分布",
+        _keyword_distribution_markdown(distribution, len(filtered_comments)),
+        "",
+        "## 三、投放复盘",
+        analysis,
+    ]
+    return "\n".join(lines)
+
+
+def write_report(report: str, path: Path) -> Path:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(report, encoding="utf-8-sig")
+    log(f"报告已写入: {path}")
+    return path
+
+
+def _md_field(label: str, value: str, is_short: bool = True) -> dict[str, Any]:
+    return {
+        "is_short": is_short,
+        "text": {"tag": "lark_md", "content": f"**{label}**\n{value}"},
+    }
+
+
+def _section_header(title: str) -> dict[str, Any]:
+    return {
+        "tag": "div",
+        "text": {"tag": "lark_md", "content": f"**{title}**"},
+    }
+
+
+def build_card_payload(
+    info: dict[str, Any],
+    metrics: dict[str, Any],
+    distribution: dict[str, Any],
+    raw_comments: list[str],
+    filtered_comments: list[str],
+    analysis: str,
+    report_path: Path,
+) -> dict[str, Any]:
+    """构造飞书 interactive 卡片 JSON。
+
+    布局（精简版默认）：
+      header（蓝色）+ 基本信息 + 数据表 + 关键词分布 + 投放复盘 + 报告路径
+    """
+    brand = info.get("brand") or ""
+    style_name = info.get("style_name") or ""
+    creator_name = info.get("creator_name") or ""
+    target_label = " / ".join(x for x in (brand, style_name) if x) if (brand or style_name) else "未指定"
+    title = info.get("title") or "未获取"
+    platform = info.get("platform") or "unknown"
+    source_url = info.get("source_url") or ""
+
+    elements: list[dict[str, Any]] = []
+
+    # 基本信息
+    elements.append({
+        "tag": "div",
+        "fields": [
+            _md_field("标题", title, is_short=False),
+            _md_field("品牌 / 产品", target_label),
+            _md_field("平台", platform),
+            _md_field("达人", creator_name or "未指定"),
+        ],
+    })
+    if source_url:
+        elements.append({
+            "tag": "div",
+            "text": {"tag": "lark_md", "content": f"🔗 [打开笔记原文]({source_url})"},
+        })
+
+    # 数据表
+    elements.append({"tag": "hr"})
+    elements.append(_section_header("📋 数据表"))
+    elements.append({
+        "tag": "div",
+        "text": {"tag": "lark_md", "content": _data_table_markdown(info, metrics, len(raw_comments), len(filtered_comments))},
+    })
+
+    # 关键词分布（拆分两个口径）
+    elements.append(_section_header("🔑 关键词分布"))
+    elements.append({
+        "tag": "div",
+        "text": {"tag": "lark_md", "content": _keyword_distribution_markdown(distribution, len(filtered_comments))},
+    })
+
+    # 投放复盘（LLM 输出）
+    elements.append(_section_header("💡 投放复盘"))
+    analysis_text = analysis or "（无分析结果）"
+    if len(analysis_text) > 4500:
+        analysis_text = analysis_text[:4500] + "\n\n...（内容过长，已截断，完整分析见报告文件）"
+    elements.append({
+        "tag": "div",
+        "text": {"tag": "lark_md", "content": analysis_text},
+    })
+
+    return {
+        "config": {"wide_screen_mode": True},
+        "header": {
+            "template": "blue",
+            "title": {
+                "tag": "plain_text",
+                "content": "📊 单篇笔记分析报告",
+            },
+        },
+        "elements": elements,
+    }
+
+
+def _fmt_int(value: Any) -> str:
+    if value is None or value == "":
+        return "—"
+    try:
+        return f"{int(float(value)):,}"
+    except (TypeError, ValueError):
+        return str(value)
+
+
+def send_feishu_report(
+    info: dict[str, Any],
+    metrics: dict[str, Any],
+    distribution: dict[str, int],
+    raw_comments: list[str],
+    filtered_comments: list[str],
+    analysis: str,
+    open_id: str,
+    report_path: Path,
+) -> dict[str, Any]:
+    """通过 lark-cli 以飞书 interactive 卡片格式发送报告。
+
+    卡片包含：标题/平台/达人基础信息、6 个分节指标 + LLM 舆情洞察。
+    Windows 下直接走 lark-cli.cmd 会撞到命令行长度上限且对 &|<>^% 转义敏感，
+    改为调 node + run.js，args 用 list 传，由 node 自己处理。
+    """
+    card = build_card_payload(
+        info, metrics, distribution,
+        raw_comments, filtered_comments, analysis,
+        report_path,
+    )
+    content_json = json.dumps(card, ensure_ascii=False)
+
+    if sys.platform.startswith("win"):
+        # Windows 下走 lark-cli.cmd 会撞到命令行长度上限（~8K）且对 &|<>^% 转义敏感；
+        # 直接调 node + run.js，args 用 list 传，由 node 自己处理，避开所有 cmd 转义问题。
+        run_js = (
+            Path(os.environ.get("APPDATA", str(Path.home())))
+            / "npm"
+            / "node_modules"
+            / "@larksuite"
+            / "cli"
+            / "scripts"
+            / "run.js"
+        )
+        if not run_js.exists():
+            raise RuntimeError(f"找不到 lark-cli 入口: {run_js}")
+        command = [
+            "node", str(run_js),
+            "--profile", "hermes-analyzer",
+            "im", "+messages-send",
+            "--user-id", open_id,
+            "--content", content_json,
+            "--msg-type", "interactive",
+            "--as", "bot",
+            "--format", "json",
+        ]
+        completed = subprocess.run(
+            command,
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+        )
+        stdout = completed.stdout
+        stderr = completed.stderr
+    else:
+        command = [
+            "lark-cli",
+            "--profile", "hermes-analyzer",
+            "im", "+messages-send",
+            "--user-id", open_id,
+            "--content", content_json,
+            "--msg-type", "interactive",
+            "--as", "bot",
+            "--format", "json",
+        ]
+        completed = subprocess.run(
+            command,
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+        )
+        stdout = completed.stdout
+        stderr = completed.stderr
+
+    if completed.returncode != 0:
+        raise RuntimeError(
+            f"lark-cli 发送失败 (exit {completed.returncode}):\n"
+            f"stdout: {stdout}\nstderr: {stderr}"
+        )
+    try:
+        return json.loads(stdout)
+    except json.JSONDecodeError:
+        return {"raw": stdout}
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="单篇笔记 → 本地过滤 + Hermes 分析 → 项目 data/summary 报告")
+    source_group = parser.add_mutually_exclusive_group(required=True)
+    source_group.add_argument("--input", type=Path, help="单篇笔记 JSON 文件路径")
+    source_group.add_argument("--note-id", type=int, help="数据库 cmt_notes.id")
+    source_group.add_argument("--url", type=str, help="数据库 cmt_notes.url")
+    parser.add_argument("--output", type=Path, default=OUTPUT_PATH, help="报告输出路径")
+    parser.add_argument("--open-id", default=TARGET_OPEN_ID, help="飞书接收人 open_id")
+    parser.add_argument("--brand", default="", help="我方品牌名（用于实体隔离/产品隔离规则；不传则只用款式名）")
+    parser.add_argument("--no-send", action="store_true", help="只生成报告文件，不发送飞书")
+    parser.add_argument("--llm-filter", action="store_true",
+                        help="用 LLM 语义清洗评论(替代关键词过滤,保留更多有效评论,但慢)")
+    parser.add_argument("--batch-size", type=int, default=30,
+                        help="LLM 清洗的批大小(默认 30 条/批)")
+    parser.add_argument("--max-workers", type=int, default=6,
+                        help="LLM 清洗的并发批数(默认 6)")
+    args = parser.parse_args()
+    args.output = resolve_layer_output(
+        args.output, layer_root=PATHS.exports_root, field="--output"
+    )
+
+    if not args.no_send:
+        if not HERMES_ANALYZER_TOKEN:
+            log("[ERROR] HERMES_ANALYZER_TOKEN 未设置。请配置 data/config/analyze.env 或传 --no-send")
+            return 1
+        if not args.open_id:
+            log("[ERROR] 飞书接收人 open_id 未设置。请配置 FEISHU_TARGET_OPEN_ID 或传 --open-id 或 --no-send")
+            return 1
+
+    if args.input:
+        log(f"读取单篇笔记 JSON: {args.input}")
+        try:
+            data = load_note_json(args.input)
+        except Exception as exc:
+            log(f"读取失败: {exc}")
+            return 1
+        info = extract_note_info(data)
+        raw_comments = extract_comment_texts(data)
+    elif args.note_id or args.url:
+        try:
+            info, raw_comments = load_note_from_db(note_id=args.note_id, url=args.url)
+        except Exception as exc:
+            log(f"从数据库读取失败: {exc}")
+            return 1
+    else:
+        log("必须提供 --input、--note-id 或 --url")
+        return 1
+
+    log(f"笔记标题: {info['title'] or '未获取'} | 平台: {info['platform']} | 总评论: {info['comment_count_total']}")
+
+    if not raw_comments:
+        # 无评论只代表评论洞察不可用，不应让整篇笔记及其曝光/互动指标
+        # 从周汇总中消失。后续会生成数据型报告，并明确标注无评论。
+        log("该笔记没有评论内容，跳过评论洞察并继续生成数据型报告")
+
+    # 先把 brand 注入到 info，再算关键词分布（需要 target_terms 拆分相关/全部）
+    if args.brand:
+        info["brand"] = args.brand
+    else:
+        info.setdefault("brand", "")
+
+    metrics = compute_metrics(info)
+    try:
+        if args.llm_filter:
+            log(f"使用 LLM 语义清洗评论(批大小 {args.batch_size}, 并发 {args.max_workers})...")
+            filtered_comments = llm_clean_comments(raw_comments,
+                                                    batch_size=args.batch_size,
+                                                    max_workers=args.max_workers)
+        else:
+            filtered_comments = code_filter_comments(raw_comments)
+    except Exception as exc:
+        log(f"评论过滤失败: {exc}")
+        return 1
+
+    target_terms = [t for t in (info.get("brand"), info.get("style_name")) if t]
+    distribution = keyword_distribution(filtered_comments, target_terms=target_terms)
+    log(f"关键词分布: 全评论 {len(filtered_comments)} 条，我方产品相关 {distribution['relevant_total']} 条")
+
+    try:
+        analysis = analyze_note(info, metrics, distribution, filtered_comments)
+    except Exception as exc:
+        log(f"笔记分析失败: {exc}")
+        return 1
+
+    report = build_report(info, metrics, distribution, raw_comments, filtered_comments, analysis)
+
+    try:
+        write_report(report, args.output)
+    except Exception as exc:
+        log(f"写入报告失败: {exc}")
+        return 1
+
+    if args.no_send:
+        log("已指定 --no-send，跳过飞书发送")
+    else:
+        # 卡片永远走精简版（给老板/团队 30 秒看完的复盘视图）
+        try:
+            result = send_feishu_report(
+                info, metrics, distribution,
+                raw_comments, filtered_comments, analysis,
+                args.open_id, args.output,
+            )
+            log(f"飞书卡片发送成功: {result}")
+        except Exception as exc:
+            log(f"飞书消息发送失败（报告文件仍已生成）: {exc}")
+            return 1
+
+    log("全部完成")
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\n已中断", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_bilibili.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_bilibili.py
new file mode 100644
index 0000000..bbec108
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_bilibili.py
@@ -0,0 +1,204 @@
+"""Batch re-scrape all Bilibili videos from DB using the browser scraper.
+
+Updates:
+- note_metrics (like / favorite / comment / share counts, title)
+- cmt_comments table (replace existing)
+- JSON/CSV files in data/notes/bilibili/
+"""
+
+import subprocess
+import sys
+import time
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+_TOOLS_DIR = PATHS.tools_root
+_PROJECT_ROOT = PATHS.module_root
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools import db
+from gyxx_flow.modules.content_marketing.runtime import bilibili_comment_scraper as scraper
+
+RELOGIN_SCRIPT = _TOOLS_DIR / "relogin_bilibili.py"
+
+try:
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")  # type: ignore[attr-defined]
+    sys.stderr.reconfigure(encoding="utf-8", errors="replace")  # type: ignore[attr-defined]
+except Exception:
+    pass
+
+
+def log(msg: str) -> None:
+    print(f"[{time.strftime('%H:%M:%S')}] {msg}", flush=True)
+
+
+def collect_urls() -> list[dict]:
+    with db.get_dict_cursor() as cur:
+        cur.execute(
+            """
+            SELECT id, url, title
+            FROM cmt_notes
+            WHERE platform = 'bilibili' AND url IS NOT NULL AND url != ''
+            ORDER BY id
+            """
+        )
+        return cur.fetchall()
+
+
+def upsert_metrics(note_id: int, title: str, metrics: dict) -> None:
+    db.update_note_metrics(
+        note_id,
+        like_count=metrics.get("like_count") or None,
+        favorite_count=metrics.get("favorite_count") or None,
+        comment_count=metrics.get("comment_count") or None,
+        share_count=metrics.get("share_count") or None,
+        title=title or None,
+    )
+
+
+def _to_int(value) -> int | None:
+    try:
+        if value in (None, ""):
+            return None
+        return int(float(value))
+    except (TypeError, ValueError):
+        return None
+
+
+def comment_replace_policy(comments: list[dict], stats: dict, metrics: dict) -> tuple[bool, bool, str]:
+    """Return (should_replace, allow_empty, reason)."""
+    logged_in = bool(stats.get("logged_in"))
+    expected = _to_int(metrics.get("comment_count"))
+    if expected is None:
+        expected = _to_int(stats.get("api_total"))
+
+    if comments:
+        if not logged_in and (expected is None or expected > len(comments)):
+            return False, False, "未登录且可能只采到热门评论"
+        return True, False, ""
+
+    if expected == 0:
+        return True, True, "平台显示 0 评论"
+    return False, False, "空评论且无法确认平台为 0 评论"
+
+
+def do_relogin() -> bool:
+    """Run relogin_bilibili.py (headed, waits for QR scan). Returns True if successful."""
+    log("  Cookie 可能过期,触发自动重登 (请扫码)...")
+    try:
+        rc = subprocess.call(
+            [sys.executable, str(RELOGIN_SCRIPT)],
+            cwd=str(_PROJECT_ROOT),
+        )
+        if rc == 0:
+            log("  重登成功,继续采集")
+            return True
+        else:
+            log(f"  重登失败 (exit={rc})")
+            return False
+    except Exception as exc:
+        log(f"  重登脚本执行异常: {exc}")
+        return False
+
+
+def main(only_ids: list[int] | None = None, dry_run: bool = False) -> int:
+    rows = collect_urls()
+    if only_ids:
+        rows = [r for r in rows if r["id"] in set(only_ids)]
+    log(f"Total bilibili notes to scrape: {len(rows)}")
+
+    ok = 0
+    fail = 0
+    total_comments = 0
+    consecutive_fail = 0
+    MAX_CONSECUTIVE_FAIL = 3
+    relogin_used = 0
+    MAX_RELOGINS = 1
+    t0 = time.time()
+    for i, row in enumerate(rows, 1):
+        note_id = row["id"]
+        url = row["url"]
+        log(f"[{i}/{len(rows)}] id={note_id} url={url}")
+        comments = None
+        result = None
+        for attempt in range(3):
+            try:
+                comments, result = scraper.scrape_comments(
+                    url, max_pages=200, delay=0.25, headless=True,
+                    max_reply_pages=200, login_timeout=0,
+                )
+                break
+            except Exception as exc:
+                if attempt < 2 and "has been closed" in str(exc):
+                    log(f"  Browser closed, retrying ({attempt+2}/3)...")
+                    time.sleep(3)
+                else:
+                    log(f"  ERROR: {exc}")
+                    fail += 1
+                    break
+        if comments is None:
+            consecutive_fail += 1
+            if consecutive_fail >= MAX_CONSECUTIVE_FAIL and relogin_used < MAX_RELOGINS:
+                if do_relogin():
+                    relogin_used += 1
+                    consecutive_fail = 0
+                    continue
+            continue
+
+        consecutive_fail = 0
+
+        video = result.get("video", {}) or {}
+        title = video.get("title", "") or result.get("title", "")
+        stats = result.get("stats", {}) or {}
+        metrics = stats.get("note_metrics", {}) or {}
+        logged_in = bool(stats.get("logged_in"))
+
+        json_path, csv_path = scraper.write_outputs(url, comments, result)
+        top = len([c for c in comments if c["level"] == "comment"])
+        replies = len([c for c in comments if c["level"] == "reply"])
+        warn = "" if logged_in else "  [WARN] 未登录,只采到热门评论"
+        log(
+            f"  ok: title={title[:40]!r} | {len(comments)} comments ({top}+{replies}) | "
+            f"likes={metrics.get('like_count')}, favs={metrics.get('favorite_count')}, "
+            f"cmts={metrics.get('comment_count')}"
+            + warn
+        )
+
+        if not dry_run:
+            try:
+                upsert_metrics(note_id, title, metrics)
+            except Exception as exc:
+                log(f"  DB metric update error: {exc}")
+            try:
+                should_replace, allow_empty, reason = comment_replace_policy(comments, stats, metrics)
+                if should_replace:
+                    db.replace_comments(note_id, "bilibili", comments, allow_empty=allow_empty)
+                else:
+                    log(f"  skip DB comment replace: {reason}")
+            except Exception as exc:
+                log(f"  DB comment replace error: {exc}")
+
+        ok += 1
+        total_comments += len(comments)
+
+    elapsed = time.time() - t0
+    log(
+        f"DONE: {ok} ok, {fail} fail, {total_comments} comments total, "
+        f"{elapsed:.1f}s ({elapsed / max(ok, 1):.1f}s per note)"
+    )
+    return 0 if fail == 0 else 1
+
+
+if __name__ == "__main__":
+    args = sys.argv[1:]
+    only_ids = []
+    dry_run = False
+    for a in args:
+        if a == "--dry-run":
+            dry_run = True
+        elif a.isdigit():
+            only_ids.append(int(a))
+    if only_ids:
+        log(f"Restricted to ids: {only_ids}")
+    if dry_run:
+        log("DRY RUN: will not touch DB")
+    raise SystemExit(main(only_ids or None, dry_run))
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_douyin.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_douyin.py
new file mode 100644
index 0000000..c37ea76
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_douyin.py
@@ -0,0 +1,252 @@
+"""Batch re-scrape all Douyin notes from DB using the browser-based scraper.
+
+Updates:
+- note_metrics (like / favorite / comment / share counts)
+- cmt_comments table (replace existing)
+- JSON/CSV files in data/notes/douyin/
+"""
+
+import subprocess
+import sys
+import time
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+# Path setup: db.py is in data/tools/, scrapers are in project root
+_TOOLS_DIR = PATHS.tools_root
+_PROJECT_ROOT = PATHS.module_root
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools import db
+from gyxx_flow.modules.content_marketing.runtime import douyin_comment_scraper as scraper
+
+RELOGIN_SCRIPT = _TOOLS_DIR / "relogin_douyin.py"
+
+try:
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")  # type: ignore[attr-defined]
+    sys.stderr.reconfigure(encoding="utf-8", errors="replace")  # type: ignore[attr-defined]
+except Exception:
+    pass
+
+
+def log(msg: str) -> None:
+    print(f"[{time.strftime('%H:%M:%S')}] {msg}", flush=True)
+
+
+def collect_urls() -> list[dict]:
+    with db.get_dict_cursor() as cur:
+        cur.execute(
+            """
+            SELECT id, url, title, comment_count
+            FROM cmt_notes
+            WHERE platform = 'douyin' AND url IS NOT NULL AND url != ''
+            ORDER BY id
+            """
+        )
+        rows = cur.fetchall()
+
+    fast = [r for r in rows if (r.get("comment_count") or 0) <= 200]
+    slow = [r for r in rows if (r.get("comment_count") or 0) > 200]
+    if fast:
+        log(f"Sorted: {len(fast)} notes ≤200 comments (fast first), {len(slow)} notes >200 (slow)")
+    return fast + slow
+
+
+def upsert_metrics(note_id: int, metrics: dict) -> None:
+    db.update_note_metrics(
+        note_id,
+        like_count=metrics.get("like_count") or None,
+        favorite_count=metrics.get("favorite_count") or None,
+        comment_count=metrics.get("comment_count") or None,
+        share_count=metrics.get("share_count") or None,
+        title=metrics.get("title") or None,
+    )
+
+
+def _to_int(value) -> int | None:
+    try:
+        if value in (None, ""):
+            return None
+        return int(float(value))
+    except (TypeError, ValueError):
+        return None
+
+
+def comment_replace_policy(comments: list[dict], result: dict, metrics: dict) -> tuple[bool, bool, str]:
+    """Return (should_replace, allow_empty, reason)."""
+    stats = result.get("stats", {}) if isinstance(result, dict) else {}
+    logged_in = bool(stats.get("logged_in"))
+    expected = _to_int(stats.get("api_total"))
+    if expected is None:
+        expected = _to_int(metrics.get("comment_count"))
+
+    if comments:
+        if not logged_in and (expected is None or expected > len(comments)):
+            return False, False, "未登录且可能只采到部分评论"
+        return True, False, ""
+
+    if expected == 0:
+        return True, True, "平台显示 0 评论"
+    return False, False, "空评论且无法确认平台为 0 评论"
+
+
+def result_logged_in(result: dict | None) -> bool:
+    """Read the login flag from the scraper result's nested stats payload."""
+    if not isinstance(result, dict):
+        return False
+    nested_stats = result.get("stats")
+    return bool(nested_stats.get("logged_in")) if isinstance(nested_stats, dict) else False
+
+
+def do_relogin() -> bool:
+    """Run relogin_douyin.py (headed, waits for QR scan). Returns True if successful."""
+    log("  Cookie 可能过期,触发自动重登 (请扫码)...")
+    try:
+        rc = subprocess.call(
+            [sys.executable, str(RELOGIN_SCRIPT)],
+            cwd=str(_PROJECT_ROOT),
+        )
+        if rc == 0:
+            log("  重登成功,继续采集")
+            return True
+        else:
+            log(f"  重登失败 (exit={rc})")
+            return False
+    except Exception as exc:
+        log(f"  重登脚本执行异常: {exc}")
+        return False
+
+
+def main(only_ids: list[int] | None = None, dry_run: bool = False) -> int:
+    rows = collect_urls()
+    if only_ids:
+        rows = [r for r in rows if r["id"] in set(only_ids)]
+    log(f"Total douyin notes to scrape: {len(rows)}")
+
+    ok = 0
+    fail = 0
+    total_comments = 0
+    consecutive_fail = 0
+    MAX_CONSECUTIVE_FAIL = 3
+    COOLDOWN_SECONDS = 300
+    cooldown_used = 0
+    MAX_COOLDOWNS = 3
+    relogin_used = 0
+    MAX_RELOGINS = 1  # 每轮 batch 最多自动重登1次
+    t0 = time.time()
+    for i, row in enumerate(rows, 1):
+        note_id = row["id"]
+        url = row["url"]
+        log(f"[{i}/{len(rows)}] id={note_id} url={url}")
+        comments = None
+        stats = None
+        for attempt in range(3):
+            try:
+                comments, stats = scraper.scrape_comments(url, 300, 60, 10, True)
+                break
+            except Exception as exc:
+                msg = str(exc)
+                retryable = attempt < 2 and (
+                    "has been closed" in msg
+                    or "page not loaded" in msg
+                    or "Cannot extract aweme_id" in msg
+                )
+                if retryable:
+                    wait_s = 3 * (attempt + 1)
+                    label = (
+                        "Browser closed" if "has been closed" in msg
+                        else "Page not loaded" if "page not loaded" in msg
+                        else "aweme_id failed"
+                    )
+                    log(f"  {label}, retrying ({attempt+2}/3) after {wait_s}s...")
+                    time.sleep(wait_s)
+                else:
+                    log(f"  ERROR: {exc}")
+                    fail += 1
+                    break
+
+        # Check logged_in status even on success
+        logged_in = result_logged_in(stats)
+
+        if comments is None:
+            consecutive_fail += 1
+            # 连续失败 + 可能 cookie 过期 → 尝试自动重登
+            if consecutive_fail >= MAX_CONSECUTIVE_FAIL and relogin_used < MAX_RELOGINS:
+                if do_relogin():
+                    relogin_used += 1
+                    consecutive_fail = 0
+                    continue  # retry the current URL in next iteration
+                else:
+                    # 重登失败,走冷却逻辑
+                    pass
+
+            if consecutive_fail >= MAX_CONSECUTIVE_FAIL:
+                if cooldown_used < MAX_COOLDOWNS:
+                    cooldown_used += 1
+                    log(
+                        f"  {consecutive_fail} consecutive failures, cooling down "
+                        f"{COOLDOWN_SECONDS}s (cooldown {cooldown_used}/{MAX_COOLDOWNS})..."
+                    )
+                    time.sleep(COOLDOWN_SECONDS)
+                    consecutive_fail = 0
+                    continue
+                else:
+                    log(
+                        f"  {MAX_COOLDOWNS} cooldowns exhausted, aborting batch; "
+                        f"{len(rows) - i} notes skipped"
+                    )
+                    break
+            continue
+
+        consecutive_fail = 0
+
+        json_path, csv_path = scraper.write_outputs(url, comments, stats)
+        metrics = stats.get("stats", {}).get("note_metrics", {})
+        top = len([c for c in comments if c["level"] == "comment"])
+        replies = len([c for c in comments if c["level"] == "reply"])
+        warn = "" if logged_in else "  [WARN] 未登录,可能只采到部分评论"
+        log(
+            f"  ok: {len(comments)} comments ({top} top + {replies} replies) | "
+            f"likes={metrics.get('like_count')}, favs={metrics.get('favorite_count')}, "
+            f"cmts={metrics.get('comment_count')}, shares={metrics.get('share_count')}"
+            + warn
+        )
+
+        if not dry_run:
+            try:
+                upsert_metrics(note_id, metrics)
+            except Exception as exc:
+                log(f"  DB metric update error: {exc}")
+            try:
+                should_replace, allow_empty, reason = comment_replace_policy(comments, stats, metrics)
+                if should_replace:
+                    db.replace_comments(note_id, "douyin", comments, allow_empty=allow_empty)
+                else:
+                    log(f"  skip DB comment replace: {reason}")
+            except Exception as exc:
+                log(f"  DB comment replace error: {exc}")
+
+        ok += 1
+        total_comments += len(comments)
+
+    elapsed = time.time() - t0
+    log(
+        f"DONE: {ok} ok, {fail} fail, {total_comments} comments total, "
+        f"{elapsed:.1f}s ({elapsed / max(ok, 1):.1f}s per note)"
+    )
+    return 0 if fail == 0 else 1
+
+
+if __name__ == "__main__":
+    args = sys.argv[1:]
+    only_ids = []
+    dry_run = False
+    for a in args:
+        if a == "--dry-run":
+            dry_run = True
+        elif a.isdigit():
+            only_ids.append(int(a))
+    if only_ids:
+        log(f"Restricted to ids: {only_ids}")
+    if dry_run:
+        log("DRY RUN: will not touch DB")
+    raise SystemExit(main(only_ids or None, dry_run))
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_xiaohongshu.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_xiaohongshu.py
new file mode 100644
index 0000000..b0be915
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/batch_rescrape_xiaohongshu.py
@@ -0,0 +1,200 @@
+"""Batch re-scrape all Xiaohongshu notes from DB using the browser scraper.
+
+Updates:
+- note_metrics (like / favorite / comment / share counts, title)
+- cmt_comments table (replace existing)
+- JSON/CSV files in data/notes/xiaohongshu/
+"""
+
+import subprocess
+import sys
+import time
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+_TOOLS_DIR = PATHS.tools_root
+_PROJECT_ROOT = PATHS.module_root
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools import db
+from gyxx_flow.modules.content_marketing.runtime import xiaohongshu_comment_scraper as scraper
+
+RELOGIN_SCRIPT = _TOOLS_DIR / "relogin_xiaohongshu.py"
+
+try:
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")  # type: ignore[attr-defined]
+    sys.stderr.reconfigure(encoding="utf-8", errors="replace")  # type: ignore[attr-defined]
+except Exception:
+    pass
+
+
+def log(msg: str) -> None:
+    print(f"[{time.strftime('%H:%M:%S')}] {msg}", flush=True)
+
+
+def collect_urls() -> list[dict]:
+    with db.get_dict_cursor() as cur:
+        cur.execute(
+            """
+            SELECT id, url, title
+            FROM cmt_notes
+            WHERE platform = 'xiaohongshu' AND url IS NOT NULL AND url != ''
+            ORDER BY id
+            """
+        )
+        return cur.fetchall()
+
+
+def upsert_metrics(note_id: int, title: str, metrics: dict) -> None:
+    db.update_note_metrics(
+        note_id,
+        like_count=metrics.get("like_count") or None,
+        favorite_count=metrics.get("favorite_count") or None,
+        comment_count=metrics.get("comment_count") or None,
+        share_count=metrics.get("share_count") or None,
+        title=title or None,
+    )
+
+
+def _to_int(value) -> int | None:
+    try:
+        if value in (None, ""):
+            return None
+        return int(float(value))
+    except (TypeError, ValueError):
+        return None
+
+
+def comment_replace_policy(comments: list[dict], stats: dict, metrics: dict) -> tuple[bool, bool, str]:
+    """Return (should_replace, allow_empty, reason)."""
+    logged_in = bool(stats.get("logged_in"))
+    expected = _to_int(metrics.get("comment_count"))
+    if expected is None:
+        expected = _to_int(stats.get("api_total"))
+
+    if comments:
+        if not logged_in and (expected is None or expected > len(comments)):
+            return False, False, "未登录且可能只采到部分评论"
+        return True, False, ""
+    if expected == 0:
+        return True, True, "平台显示 0 评论"
+    return False, False, "空评论且无法确认平台为 0 评论"
+
+
+def do_relogin() -> bool:
+    """Run relogin_xiaohongshu.py (headed, waits for QR scan). Returns True if successful."""
+    log("  Cookie 可能过期,触发自动重登 (请扫码)...")
+    try:
+        rc = subprocess.call(
+            [sys.executable, str(RELOGIN_SCRIPT)],
+            cwd=str(_PROJECT_ROOT),
+        )
+        if rc == 0:
+            log("  重登成功,继续采集")
+            return True
+        else:
+            log(f"  重登失败 (exit={rc})")
+            return False
+    except Exception as exc:
+        log(f"  重登脚本执行异常: {exc}")
+        return False
+
+
+def main(only_ids: list[int] | None = None, dry_run: bool = False) -> int:
+    rows = collect_urls()
+    if only_ids:
+        rows = [r for r in rows if r["id"] in set(only_ids)]
+    log(f"Total xiaohongshu notes to scrape: {len(rows)}")
+
+    ok = 0
+    fail = 0
+    total_comments = 0
+    consecutive_fail = 0
+    MAX_CONSECUTIVE_FAIL = 3
+    relogin_used = 0
+    MAX_RELOGINS = 1
+    t0 = time.time()
+    for i, row in enumerate(rows, 1):
+        note_id = row["id"]
+        url = row["url"]
+        log(f"[{i}/{len(rows)}] id={note_id} url={url}")
+        comments = None
+        result = None
+        for attempt in range(3):
+            try:
+                comments, result = scraper.scrape_comments(url, 60, 30, 4, True)
+                break
+            except Exception as exc:
+                if attempt < 2 and "has been closed" in str(exc):
+                    log(f"  Browser closed, retrying ({attempt+2}/3)...")
+                    time.sleep(3)
+                else:
+                    log(f"  ERROR: {exc}")
+                    fail += 1
+                    break
+        if comments is None:
+            consecutive_fail += 1
+            if consecutive_fail >= MAX_CONSECUTIVE_FAIL and relogin_used < MAX_RELOGINS:
+                if do_relogin():
+                    relogin_used += 1
+                    consecutive_fail = 0
+                    continue
+            # No relogin or relogin failed → skip this note
+            continue
+
+        consecutive_fail = 0
+
+        title = result.get("title", "")
+        stats = result.get("stats", {})
+        metrics = stats.get("note_metrics", {}) or {}
+        logged_in = bool(stats.get("logged_in"))
+
+        json_path, csv_path = scraper.write_outputs(url, comments, result)
+        top = len([c for c in comments if c["level"] == "comment"])
+        replies = len([c for c in comments if c["level"] == "reply"])
+        warn = "" if logged_in else "  [WARN] 未登录,可能只采到部分评论"
+        log(
+            f"  ok: title={title[:40]!r} | {len(comments)} comments ({top}+{replies}) | "
+            f"likes={metrics.get('like_count')}, favs={metrics.get('favorite_count')}, "
+            f"cmts={metrics.get('comment_count')}"
+            + warn
+        )
+
+        if not dry_run:
+            try:
+                upsert_metrics(note_id, title, metrics)
+            except Exception as exc:
+                log(f"  DB metric update error: {exc}")
+            try:
+                should_replace, allow_empty, reason = comment_replace_policy(comments, stats, metrics)
+                if should_replace:
+                    db.replace_comments(note_id, "xiaohongshu", comments, allow_empty=allow_empty)
+                else:
+                    log(f"  skip DB comment replace: {reason}")
+            except Exception as exc:
+                log(f"  DB comment replace error: {exc}")
+
+        ok += 1
+        total_comments += len(comments)
+
+    elapsed = time.time() - t0
+    log(
+        f"DONE: {ok} ok, {fail} fail, {total_comments} comments total, "
+        f"{elapsed:.1f}s ({elapsed / max(ok, 1):.1f}s per note)"
+    )
+    return 0 if fail == 0 else 1
+
+
+if __name__ == "__main__":
+    args = sys.argv[1:]
+    only_ids = []
+    dry_run = False
+    for a in args:
+        if a == "--dry-run":
+            dry_run = True
+        elif a.isdigit():
+            only_ids.append(int(a))
+    if only_ids:
+        log(f"Restricted to ids: {only_ids}")
+    if dry_run:
+        log("DRY RUN: will not touch DB")
+    raise SystemExit(main(only_ids or None, dry_run))
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/check_id_filled.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/check_id_filled.py
new file mode 100644
index 0000000..bcf7c6e
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/check_id_filled.py
@@ -0,0 +1,42 @@
+"""统计 24 张表里 creator_id / creator_name 填充率"""
+import json
+import glob
+from pathlib import Path
+
+per_table = []
+total_records = total_with_id = total_with_name = 0
+
+for path in sorted(
+    glob.glob("data/feishu_tables/[0-9]*.json"),
+    key=lambda p: int(Path(p).stem.split("-")[0]),
+):
+    d = json.load(open(path, encoding="utf-8"))
+    name = d.get("name")
+    fid_to_name = d.get("field_id_to_name", {})
+    name_fid = id_fid = None
+    for fid, fname in fid_to_name.items():
+        cname = (fname or "").rstrip("?").rstrip()
+        if cname in ("达人名称", "达人昵称"):
+            name_fid = fid
+        elif cname in ("达人id", "达人id号", "达人ID", "星图ID"):
+            id_fid = fid
+    n_with_id = n_with_name = 0
+    n_records = len(d.get("records", []))
+    for rec in d.get("records", []):
+        if id_fid and rec.get(id_fid):
+            n_with_id += 1
+        if name_fid and rec.get(name_fid):
+            n_with_name += 1
+    per_table.append((name, n_records, n_with_id, n_with_name))
+    total_records += n_records
+    total_with_id += n_with_id
+    total_with_name += n_with_name
+
+print(f"{'款式':<14} {'总记录':>5} {'有ID':>5} {'有名称':>6} {'填充率(ID)':>10}")
+for name, n, nid, nnm in per_table:
+    rate = f"{100*nid/n:.1f}%" if n else "-"
+    print(f"{name:<14} {n:>5} {nid:>5} {nnm:>6} {rate:>10}")
+print()
+print(f"总计 {total_records} 条")
+print(f"  有 creator_id:  {total_with_id} ({100*total_with_id/total_records:.1f}%)")
+print(f"  有 creator_name:{total_with_name} ({100*total_with_name/total_records:.1f}%)")
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/check_status.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/check_status.py
new file mode 100644
index 0000000..c33b7b5
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/check_status.py
@@ -0,0 +1,179 @@
+"""查看每个平台 24 款式的采集状态。
+
+判定规则(对单个平台):
+  OK     - v2 JSON 文件存在,且 mtime 在 --stale-hours 小时内
+  STALE  - v2 JSON 文件存在,但 mtime 超过 --stale-hours 小时(需要重跑)
+  MISSING - v2 JSON 文件不存在(进程崩溃/被中断)
+  EMPTY  - v2 JSON 存在但 matched=0、filled=0、total=0 (该款式确实没数据)
+  HALF   - v2 JSON 存在但 results 里至少有一条 error (部分采集成功)
+
+用法:
+  python data/tools/check_status.py                  # 看三个平台
+  python data/tools/check_status.py --platform xt   # 只看星图
+  python data/tools/check_status.py --stale-hours 4 # 4 小时以上算陈旧
+"""
+import argparse
+import datetime as dt
+import json
+import sys
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+PROJECT_DIR = PATHS.module_root
+V2_DIR = PATHS.normalized_root / "v2_results"
+
+# 共享工具
+from gyxx_flow.modules.content_marketing.runtime.data.tools.v2_filename import parse_v2_filename  # noqa: E402
+
+PLATFORMS = {
+    "bili": {"name": "B 站", "label": "bili"},
+    "pgy": {"name": "小红书蒲公英", "label": "xhs"},
+    "xt": {"name": "星图", "label": "xt"},
+}
+
+# 款式总数 1-24
+STYLE_INDICES = list(range(1, 25))
+
+
+def scan_platform(platform_key: str, stale_hours: float) -> dict[int, dict]:
+    """扫一个平台 v2_results,返回 {index: {status, mtime, matched, ...}}"""
+    label = PLATFORMS[platform_key]["label"]
+    found: dict[int, dict] = {}
+    for path in V2_DIR.glob("*.json"):
+        idx, name, plat, _ = parse_v2_filename(path.name)
+        if plat != label or idx is None:
+            continue
+        mtime = dt.datetime.fromtimestamp(path.stat().st_mtime)
+        age_h = (dt.datetime.now() - mtime).total_seconds() / 3600.0
+        try:
+            data = json.loads(path.read_text(encoding="utf-8"))
+        except Exception:
+            data = {}
+        matched = data.get("matched") or data.get("updated") or 0
+        filled = data.get("filled") or matched
+        total = (data.get("total") or data.get("total_b_records") or 0)
+        results = data.get("results") or data.get("details") or []
+        has_error = any(isinstance(r, dict) and r.get("error") for r in results)
+
+        if total == 0 and matched == 0 and not has_error:
+            status = "EMPTY"
+        elif has_error and matched > 0:
+            status = "HALF"
+        elif has_error:
+            status = "HALF"
+        elif age_h > stale_hours:
+            status = "STALE"
+        else:
+            status = "OK"
+        found[idx] = {
+            "status": status,
+            "name": name,
+            "path": path.name,
+            "mtime": mtime,
+            "age_h": age_h,
+            "matched": matched,
+            "filled": filled,
+            "total": total,
+            "has_error": has_error,
+        }
+    return found
+
+
+def color(s: str, code: str) -> str:
+    if not sys.stdout.isatty():
+        return s
+    return f"\033[{code}m{s}\033[0m"
+
+
+STATUS_COLOR = {
+    "OK": ("32",),       # green
+    "STALE": ("33",),    # yellow
+    "MISSING": ("31",),  # red
+    "EMPTY": ("90",),    # grey
+    "HALF": ("33",),     # yellow
+}
+
+
+def fmt_status(s: str) -> str:
+    code = STATUS_COLOR.get(s, ("",))[0]
+    label_map = {
+        "OK": "  OK  ",
+        "STALE": " STALE ",
+        "MISSING": "MISSING",
+        "EMPTY": " EMPTY ",
+        "HALF": "  HALF ",
+    }
+    return color(label_map.get(s, s), code)
+
+
+def print_platform(platform_key: str, info: dict, stale_hours: float) -> tuple[int, int, int]:
+    name = info["name"]
+    by_idx = scan_platform(platform_key, stale_hours)
+    counts = {"OK": 0, "STALE": 0, "MISSING": 0, "EMPTY": 0, "HALF": 0}
+    print(f"\n{'=' * 78}")
+    print(f"  {name}  (stale > {stale_hours}h)")
+    print(f"{'=' * 78}")
+    print(f"  {'款式':>4}  {'状态':<8}  {'名称':<14}  {'matched':>7}  {'total':>5}  "
+          f"{'err':>3}  {'mtime':<19}  {'age':>5}")
+    print("  " + "-" * 74)
+    for idx in STYLE_INDICES:
+        if idx in by_idx:
+            r = by_idx[idx]
+            counts[r["status"]] += 1
+            err_n = sum(1 for x in (r.get("path"),) if x and "error" in x)  # placeholder
+            # 计算 has_error 计数
+            try:
+                data = json.loads((V2_DIR / r["path"]).read_text(encoding="utf-8"))
+                results = data.get("results") or data.get("details") or []
+                err_n = sum(1 for x in results if isinstance(x, dict) and x.get("error"))
+            except Exception:
+                err_n = 0
+            mtime_str = r["mtime"].strftime("%Y-%m-%d %H:%M:%S")
+            print(f"  {idx:>4}  {fmt_status(r['status'])}  {r['name']:<14}  "
+                  f"{r['matched']:>7}  {r['total']:>5}  {err_n:>3}  "
+                  f"{mtime_str:<19}  {r['age_h']:>4.1f}h")
+        else:
+            counts["MISSING"] += 1
+            print(f"  {idx:>4}  {fmt_status('MISSING')}  {'-':<14}  {'-':>7}  {'-':>5}  "
+                  f"{'-':>3}  {'-':<19}  {'-':>5}")
+    print("  " + "-" * 74)
+    total = sum(counts.values())
+    print(f"  小计: OK={counts['OK']}  STALE={counts['STALE']}  "
+          f"MISSING={counts['MISSING']}  EMPTY={counts['EMPTY']}  "
+          f"HALF={counts['HALF']}  / 共 {total}")
+    # 待补跑 = MISSING + STALE + HALF(只算有 matched>0 的)
+    need_retry = counts["MISSING"] + counts["STALE"]
+    print(f"  待补跑: {need_retry} 个款式")
+    return counts["MISSING"] + counts["STALE"], counts["HALF"], total
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser(description="查看各平台 24 款式采集状态")
+    ap.add_argument("--platform", type=str, default="bili,pgy,xt",
+                    help="逗号分隔平台键 bili/pgy/xt,默认全看")
+    ap.add_argument("--stale-hours", type=float, default=26.0,
+                    help="超过多少小时算 STALE(默认 26h = 每日一次)")
+    args = ap.parse_args()
+
+    keys = [k.strip() for k in args.platform.split(",") if k.strip()]
+    for k in keys:
+        if k not in PLATFORMS:
+            print(f"[ERROR] 未知平台: {k}  可选: {','.join(PLATFORMS.keys())}")
+            return 1
+
+    print(f"扫描目录: {V2_DIR}")
+    grand_retry = grand_half = grand_total = 0
+    for k in keys:
+        retry_n, half_n, total_n = print_platform(k, PLATFORMS[k], args.stale_hours)
+        grand_retry += retry_n
+        grand_half += half_n
+        grand_total += total_n
+    print(f"\n{'=' * 78}")
+    print(f"  汇总: 待补跑 {grand_retry} 款 (HALF {grand_half} 款) / 共 {grand_total} 款")
+    print(f"  补跑命令: python data/tools/retry_failed.py")
+    print(f"{'=' * 78}\n")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/collect_note_metrics.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/collect_note_metrics.py
new file mode 100644
index 0000000..5e45980
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/collect_note_metrics.py
@@ -0,0 +1,188 @@
+import argparse
+import csv
+import json
+import sys
+from pathlib import Path
+from typing import Any
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import (
+    PATHS,
+    resolve_layer_output,
+)
+
+# Path setup: scrapers are in project root, feishu_comment_batch is in data/tools/
+_TOOLS_DIR = PATHS.tools_root
+_PROJECT_ROOT = PATHS.module_root
+
+from scrapling.fetchers import DynamicSession
+
+from gyxx_flow.modules.content_marketing.runtime import bilibili_comment_scraper
+from gyxx_flow.modules.content_marketing.runtime import douyin_comment_scraper
+from gyxx_flow.modules.content_marketing.runtime import xiaohongshu_comment_scraper
+from gyxx_flow.modules.content_marketing.runtime.data.tools.feishu_comment_batch import detect_platform, normalize_note_url
+
+
+def unique_note_urls(source_csv: Path, platform: str) -> list[str]:
+    seen: set[str] = set()
+    urls: list[str] = []
+    with source_csv.open(encoding="utf-8-sig", newline="") as file:
+        for row in csv.DictReader(file):
+            url = normalize_note_url(row.get("note_url") or row.get("笔记链接") or "")
+            if not url or detect_platform(url) != platform or url in seen:
+                continue
+            seen.add(url)
+            urls.append(url)
+    return urls
+
+
+def collect_douyin(urls: list[str], headless: bool) -> list[dict[str, Any]]:
+    """Collect Douyin note metrics via browser (Playwright + XHR interception)."""
+    rows: list[dict[str, Any]] = []
+
+    def make_action(url: str):
+        def action(page):
+            page.wait_for_timeout(7000)
+            metrics = douyin_comment_scraper.extract_note_metrics(page)
+            rows.append({
+                "platform": "douyin",
+                "note_url": url,
+                "like_count": metrics.get("like_count", ""),
+                "favorite_count": metrics.get("favorite_count", ""),
+                "share_count": metrics.get("share_count", ""),
+                "comment_count": metrics.get("comment_count", ""),
+            })
+            print(f"douyin {len(rows)}/{len(urls)} {metrics} {url}", flush=True)
+
+        return action
+
+    with DynamicSession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=str(douyin_comment_scraper.PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=90000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=douyin_comment_scraper.restore_cookies,
+        max_pages=1,
+    ) as session:
+        for url in urls:
+            try:
+                normalized = douyin_comment_scraper.normalize_douyin_url(url)
+                session.fetch(normalized, page_action=make_action(url), wait=300)
+            except Exception as exc:
+                rows.append({"platform": "douyin", "note_url": url, "like_count": "", "favorite_count": "", "share_count": "", "comment_count": "", "error": repr(exc)})
+                print(f"douyin failed {url} {exc!r}", flush=True)
+    return rows
+
+
+def collect_xiaohongshu(urls: list[str], headless: bool) -> list[dict[str, Any]]:
+    rows: list[dict[str, Any]] = []
+
+    def make_action(url: str):
+        def action(page):
+            page.wait_for_timeout(7000)
+            metrics = xiaohongshu_comment_scraper.extract_note_metrics(page)
+            rows.append({"platform": "xiaohongshu", "note_url": url, **metrics})
+            print(f"xiaohongshu {len(rows)}/{len(urls)} {metrics} {url}", flush=True)
+
+        return action
+
+    with DynamicSession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=str(xiaohongshu_comment_scraper.PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=90000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=xiaohongshu_comment_scraper.restore_cookies,
+        max_pages=1,
+    ) as session:
+        for url in urls:
+            try:
+                session.fetch(url, page_action=make_action(url), wait=300)
+            except Exception as exc:
+                rows.append({"platform": "xiaohongshu", "note_url": url, "like_count": "", "favorite_count": "", "share_count": "", "comment_count": "", "error": repr(exc)})
+                print(f"xiaohongshu failed {url} {exc!r}", flush=True)
+    return rows
+
+
+def collect_bilibili(urls: list[str], headless: bool) -> list[dict[str, Any]]:
+    rows: list[dict[str, Any]] = []
+
+    def make_action(url: str):
+        def action(page):
+            page.wait_for_timeout(1500)
+            video = bilibili_comment_scraper.resolve_video_info(page, url)
+            metrics = video.get("note_metrics") or {}
+            rows.append({"platform": "bilibili", "note_url": url, **metrics})
+            print(f"bilibili {len(rows)}/{len(urls)} {metrics} {url}", flush=True)
+
+        return action
+
+    with DynamicSession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=str(bilibili_comment_scraper.PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=90000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=bilibili_comment_scraper.restore_cookies,
+        max_pages=1,
+    ) as session:
+        for url in urls:
+            try:
+                session.fetch(url, page_action=make_action(url), wait=300)
+            except Exception as exc:
+                rows.append({"platform": "bilibili", "note_url": url, "like_count": "", "favorite_count": "", "share_count": "", "comment_count": "", "error": repr(exc)})
+                print(f"bilibili failed {url} {exc!r}", flush=True)
+    return rows
+
+
+def write_rows(rows: list[dict[str, Any]], output: Path) -> None:
+    output.parent.mkdir(exist_ok=True)
+    fields = ["platform", "note_url", "like_count", "favorite_count", "share_count", "comment_count", "error"]
+    with output.open("w", encoding="utf-8-sig", newline="") as file:
+        writer = csv.DictWriter(file, fieldnames=fields)
+        writer.writeheader()
+        for row in rows:
+            writer.writerow({field: row.get(field, "") for field in fields})
+    output.with_suffix(".json").write_text(json.dumps(rows, ensure_ascii=False, indent=2), encoding="utf-8")
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--source-csv", required=True)
+    parser.add_argument("--platform", choices=["douyin", "xiaohongshu", "bilibili"], required=True)
+    parser.add_argument("--output", required=True)
+    parser.add_argument("--headless", action="store_true")
+    args = parser.parse_args()
+
+    source_csv = Path(args.source_csv)
+    output = resolve_layer_output(
+        args.output, layer_root=PATHS.normalized_root, field="--output"
+    )
+    urls = unique_note_urls(source_csv, args.platform)
+    print(f"{args.platform}: {len(urls)} unique notes", flush=True)
+
+    if args.platform == "douyin":
+        rows = collect_douyin(urls, args.headless)
+    elif args.platform == "xiaohongshu":
+        rows = collect_xiaohongshu(urls, args.headless)
+    else:
+        rows = collect_bilibili(urls, args.headless)
+
+    write_rows(rows, output)
+    print(f"wrote {output}", flush=True)
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_dashboard_analytics.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_dashboard_analytics.py
new file mode 100644
index 0000000..63f33de
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_dashboard_analytics.py
@@ -0,0 +1,321 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""构建全款式同权经营看板所需的确定性指标与榜单。"""
+
+from __future__ import annotations
+
+from collections import defaultdict
+from datetime import date, timedelta
+import os
+from pathlib import Path
+from typing import Any
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+try:
+    from .db import get_conn
+except ImportError:
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_conn
+
+
+MIN_CONVERSION_RANK_VISITORS = 30
+MIN_REFUND_RANK_SALES = 5
+ROOT = PATHS.module_root
+IMAGE_ROOTS = [
+    ROOT,
+    Path(os.environ.get("GYXX_PRODUCT_ASSET_ROOT", str(ROOT))),
+]
+PLATFORM_IMAGE_CODES = {"天猫": "tm", "京东": "jd", "抖音电商": "dy"}
+STYLE_CATEGORY_ORDER = ("都市机能", "户外机能", "都市新贵", "智性通勤", "都市新旅", "都市运动", "未分类")
+
+
+def load_style_categories(style_names: list[str]) -> dict[str, str]:
+    if not style_names:
+        return {}
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT name, COALESCE(NULLIF(BTRIM(style_category), ''), '未分类')
+                FROM cmt_styles
+                WHERE name = ANY(%s)
+                """,
+                (style_names,),
+            )
+            return {name: category for name, category in cur.fetchall()}
+
+
+def load_creator_connections(metric_date: date, style_names: list[str]) -> dict[str, dict[str, Any]]:
+    if not style_names:
+        return {}
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT s.name,
+                       COUNT(DISTINCT c.creator_id) FILTER (WHERE c.creator_id IS NOT NULL) creator_count,
+                       ARRAY_REMOVE(ARRAY_AGG(DISTINCT NULLIF(BTRIM(c.content_direction), '')), NULL) directions
+                FROM cmt_styles s
+                LEFT JOIN cmt_cooperations c
+                  ON c.style_id = s.id
+                 AND COALESCE(c.publish_time::date, c.cooperation_date) BETWEEN %s AND %s
+                WHERE s.name = ANY(%s)
+                GROUP BY s.name
+                """,
+                (metric_date - timedelta(days=29), metric_date, style_names),
+            )
+            return {
+                style_name: {"creator_count": creator_count or 0, "directions": directions or []}
+                for style_name, creator_count, directions in cur.fetchall()
+            }
+
+
+def load_style_images(style_names: list[str]) -> dict[tuple[str, str], dict[str, Any]]:
+    """为三平台画像卡统一选择同款天猫高点击率主图。"""
+    if not style_names:
+        return {}
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT style_name, platform, collect_date, local_image_path, creative_image_url,
+                       COALESCE(impressions, 0) impressions, click_rate
+                FROM main_image_creatives
+                WHERE style_name = ANY(%s)
+                  AND platform = 'tm'
+                  AND image_downloaded = TRUE
+                  AND local_image_path IS NOT NULL
+                  AND COALESCE(impressions, 0) >= 100
+                ORDER BY COALESCE(click_rate, 0) DESC, COALESCE(impressions, 0) DESC, collect_date DESC
+                """,
+                (style_names,),
+            )
+            rows = cur.fetchall()
+    by_style: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for style_name, platform, collect_date, local_path, url, impressions, click_rate in rows:
+        relative = Path(str(local_path))
+        absolute = next((root / relative for root in IMAGE_ROOTS if (root / relative).exists()), None)
+        if absolute is None:
+            continue
+        by_style[style_name].append({
+            "local_image_path": str(absolute),
+            "image_url": url,
+            "image_platform": {"tm": "天猫", "jd": "京东", "dy": "抖音电商"}.get(platform, platform),
+            "image_date": collect_date,
+            "image_impressions": impressions,
+            "image_ctr": float(click_rate) if click_rate is not None else None,
+        })
+    result: dict[tuple[str, str], dict[str, Any]] = {}
+    for style_name in style_names:
+        candidates = by_style.get(style_name, [])
+        for target in ("天猫", "京东", "抖音电商"):
+            if candidates:
+                result[(style_name, target)] = candidates[0]
+    return result
+
+
+def _rate(numerator: Any, denominator: Any) -> float | None:
+    if numerator is None or denominator is None or denominator <= 0:
+        return None
+    return round(float(numerator) / float(denominator), 4)
+
+
+def _percentile(values: list[float], fraction: float) -> float:
+    if not values:
+        return 0.0
+    ordered = sorted(values)
+    index = min(len(ordered) - 1, max(0, round((len(ordered) - 1) * fraction)))
+    return ordered[index]
+
+
+def _persona_text(persona: dict[str, Any] | None) -> str:
+    if not persona:
+        return "无数据"
+    labels = [persona.get(key) for key in ("gender_top", "age_top", "city_top", "buying_power_top") if persona.get(key)]
+    return "、".join(labels) if labels else "无数据"
+
+
+def _interest_hypothesis(persona: dict[str, Any] | None, directions: list[str]) -> str:
+    if not persona:
+        return "无数据"
+    source = " ".join(str(value or "") for value in persona.values())
+    interests: list[str] = []
+    if any(term in source for term in ("女性", "女")):
+        interests.extend(["穿搭", "健身", "旅行摄影"])
+    if any(term in source for term in ("男性", "男")):
+        interests.extend(["数码3C", "城市通勤", "户外出行"])
+    if any(term in source for term in ("18-24", "18-25", "25-29")):
+        interests.extend(["潮流内容", "社交分享"])
+    if any(term in source for term in ("26-35", "30-34", "35-39")):
+        interests.extend(["职场效率", "短途出行"])
+    if any(term in source for term in ("56岁以上", "46-55", "45-49", "50")):
+        interests.extend(["品质消费", "舒适出行"])
+    if any(term in source for term in ("土豪", "高级白领", "L4", "L5")):
+        interests.extend(["品质材质", "设计感"])
+    interests.extend(direction for direction in directions if direction)
+    unique = list(dict.fromkeys(interests))[:4]
+    return "兴趣推测：" + "、".join(unique) if unique else "兴趣推测：无足够标签"
+
+
+def _optimization(role: str, has_persona: bool) -> str:
+    persona_action = "按画像扩展相似人群" if has_persona else "先补采人群画像"
+    if role == "conversion":
+        return f"高转化扩量：{persona_action}，逐步增加流量并监控退款订单比"
+    return f"高访客提转化：{persona_action}，优化主图/详情页/利益点并做加购再营销"
+
+
+def _diagnose(row: dict[str, Any], visitor_q75: float, conversion_median: float, cart_median: float, refund_q75: float) -> str:
+    if row.get("visitors") is None:
+        return "无数据"
+    conversion = row.get("conversion_rate")
+    refund = row.get("refund_rate")
+    cart_rate = row.get("cart_rate")
+    if refund is not None and refund >= max(0.3, refund_q75):
+        return "退款压力"
+    if row["visitors"] >= visitor_q75 and conversion is not None and conversion >= conversion_median:
+        return "高流量高转化"
+    if row["visitors"] >= visitor_q75:
+        return "高流量·转化待提"
+    if conversion is not None and conversion >= conversion_median * 1.25:
+        return "低流量·高转化可放量"
+    if cart_rate is not None and cart_rate < cart_median * 0.7:
+        return "加购偏弱"
+    return "常规观察"
+
+
+def build_dashboard_facts(
+    report_date: str,
+    tracking_rows: list[dict[str, Any]],
+    platform_metrics: list[dict[str, Any]],
+    personas: list[dict[str, Any]],
+    connections: dict[str, dict[str, Any]],
+    style_images: dict[tuple[str, str], dict[str, Any]] | None = None,
+    daily_notes: list[dict[str, Any]] | None = None,
+    style_categories: dict[str, str] | None = None,
+    style_signals: dict[str, dict[str, Any]] | None = None,
+) -> dict[str, Any]:
+    styles = []
+    for source in tracking_rows:
+        row = dict(source)
+        row["cart_rate"] = _rate(row.get("cart_users"), row.get("visitors"))
+        row["conversion_rate"] = _rate(row.get("sales"), row.get("visitors"))
+        row["refund_rate"] = _rate(row.get("refund_orders"), row.get("sales"))
+        connection = connections.get(row["style_name"], {})
+        row["creator_connections_30d"] = int(connection.get("creator_count") or 0)
+        row["content_directions_30d"] = connection.get("directions") or []
+        row["style_category"] = (style_categories or {}).get(row["style_name"]) or "未分类"
+        row["multi_source_signals"] = (style_signals or {}).get(row["style_name"], {})
+        row["recent_note_exposure"] = row["multi_source_signals"].get("recent_note_exposure")
+        styles.append(row)
+
+    visitors = [float(row["visitors"]) for row in styles if row.get("visitors") is not None]
+    conversions = [float(row["conversion_rate"]) for row in styles if row.get("conversion_rate") is not None]
+    carts = [float(row["cart_rate"]) for row in styles if row.get("cart_rate") is not None]
+    refunds = [float(row["refund_rate"]) for row in styles if row.get("refund_rate") is not None]
+    visitor_q75 = _percentile(visitors, 0.75)
+    conversion_median = _percentile(conversions, 0.5)
+    cart_median = _percentile(carts, 0.5)
+    refund_q75 = _percentile(refunds, 0.75)
+    for row in styles:
+        base = _diagnose(row, visitor_q75, conversion_median, cart_median, refund_q75)
+        signal = row["multi_source_signals"]
+        evidence = ["经营"]
+        if signal.get("persona"):
+            evidence.append("画像")
+        if signal.get("note_count"):
+            evidence.append("营销笔记")
+        if signal.get("comment_count"):
+            evidence.append("评论")
+        if signal.get("review_count"):
+            evidence.append("商品评价")
+        if signal.get("creative_ctr") is not None:
+            evidence.append("主图")
+        row["evidence_summary"] = "、".join(evidence)
+        if (signal.get("review_negative_rate") or 0) >= 0.20:
+            row["diagnosis"] = f"{base}·评价风险" if base != "无数据" else "评价风险"
+        elif (signal.get("comment_negative_rate") or 0) >= 0.20:
+            row["diagnosis"] = f"{base}·评论风险" if base != "无数据" else "评论风险"
+        else:
+            row["diagnosis"] = base
+
+    category_rank = {name: index for index, name in enumerate(STYLE_CATEGORY_ORDER)}
+    ordered_styles = sorted(
+        styles,
+        key=lambda row: (
+            category_rank.get(row["style_category"], len(category_rank)),
+            row["style_category"],
+            row["style_name"],
+        ),
+    )
+    grouped: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in ordered_styles:
+        grouped[row["style_category"]].append(row)
+    style_groups = [
+        {"style_category": category, "styles": grouped[category]}
+        for category in sorted(grouped, key=lambda value: (category_rank.get(value, len(category_rank)), value))
+    ]
+
+    def top(metric: str, eligible=lambda row: True) -> list[dict[str, Any]]:
+        rows = [row for row in styles if row.get(metric) is not None and eligible(row)]
+        return sorted(rows, key=lambda row: (row[metric], row.get("visitors") or 0), reverse=True)[:5]
+
+    rankings = {
+        "conversion": top("conversion_rate", lambda row: (row.get("visitors") or 0) >= MIN_CONVERSION_RANK_VISITORS),
+        "visitors": top("visitors"),
+        "cart_users": top("cart_users"),
+        "sales": top("sales"),
+        "refund_rate": top("refund_rate", lambda row: (row.get("sales") or 0) >= MIN_REFUND_RANK_SALES),
+    }
+
+    persona_index = {(row.get("style_name"), row.get("platform")): row for row in personas}
+    directions = {row["style_name"]: row.get("content_directions_30d", []) for row in styles}
+    platform_groups: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for source in platform_metrics:
+        row = dict(source)
+        row["conversion_rate"] = _rate(row.get("sales"), row.get("visitors"))
+        platform_groups[row.get("platform")].append(row)
+
+    platform_personas = []
+    preferred_platforms = ["天猫", "京东", "抖音电商"]
+    ordered_platforms = [name for name in preferred_platforms if name in platform_groups]
+    ordered_platforms.extend(sorted(name for name in platform_groups if name not in preferred_platforms))
+    for platform in ordered_platforms:
+        rows = platform_groups[platform]
+        conversion_eligible = [row for row in rows if (row.get("visitors") or 0) >= MIN_CONVERSION_RANK_VISITORS and row.get("conversion_rate") is not None]
+        conversion_winner = max(conversion_eligible, key=lambda row: row["conversion_rate"]) if conversion_eligible else None
+        visitor_winner = max(rows, key=lambda row: row.get("visitors") or -1) if rows else None
+
+        def winner_payload(winner: dict[str, Any] | None, role: str) -> dict[str, Any] | None:
+            if not winner:
+                return None
+            persona = persona_index.get((winner["style_name"], platform))
+            return {
+                **winner,
+                **((style_images or {}).get((winner["style_name"], platform)) or {}),
+                "persona_text": _persona_text(persona),
+                "persona_date": persona.get("data_date") if persona else None,
+                "interest_hypothesis": _interest_hypothesis(persona, directions.get(winner["style_name"], [])),
+                "optimization": _optimization(role, bool(persona)),
+            }
+
+        platform_personas.append({
+            "platform": platform,
+            "top_conversion": winner_payload(conversion_winner, "conversion"),
+            "top_visitors": winner_payload(visitor_winner, "visitors"),
+        })
+
+    return {
+        "report_date": report_date,
+        "styles": ordered_styles,
+        "style_groups": style_groups,
+        "rankings": rankings,
+        "platform_personas": platform_personas,
+        "daily_new_notes": sorted(daily_notes or [], key=lambda row: row.get("view_count") or 0, reverse=True),
+        "definitions": {
+            "conversion_rate": "销量/访客；榜单访客门槛≥30",
+            "cart_rate": "加购人数/访客",
+            "refund_rate": "退款订单/当日销量，可能含跨期退款",
+            "creator_connections": "近30天去重达人数量",
+            "sales_7d_total": "报告日及前6日累计销量",
+            "sales_forecast_7d": "最近3日日均销量×7；趋势相对前4日日均，仅作短期估算",
+        },
+    }
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_marketing_report.bat b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_marketing_report.bat
new file mode 100644
index 0000000..db95162
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_marketing_report.bat
@@ -0,0 +1,57 @@
+@echo off
+if not defined GYXX_PYTHON set "GYXX_PYTHON=python"
+REM Daily SKU marketing operations report: Hermes analysis + Feishu dashboard and full report
+REM Register: schtasks /Create /SC DAILY /TN YingxiaoYunying_DailyMarketingReport /TR %PROJECT_DIR%\data\tools\daily_marketing_report.bat /ST 10:00 /F
+
+setlocal
+
+for %%I in ("%~dp0..\..") do set "PROJECT_DIR=%%~fI"
+if not defined GYXX_DATA_ROOT (
+    if not defined GYXX_PROJECT_ROOT (
+        echo ERROR: GYXX_DATA_ROOT or GYXX_PROJECT_ROOT must be defined. 1>&2
+        exit /b 3
+    )
+    set "GYXX_DATA_ROOT=%GYXX_PROJECT_ROOT%\var"
+)
+set LOG_DIR=%GYXX_DATA_ROOT%\logs\content_marketing\scheduled
+set PYTHON=%GYXX_PYTHON%
+where %PYTHON% >nul 2>&1 || set PYTHON=python
+
+set PYTHONIOENCODING=utf-8
+set LARK_CLI_NO_PROXY=1
+set LARKSUITE_CLI_NO_UPDATE_NOTIFIER=1
+set LARKSUITE_CLI_NO_SKILLS_NOTIFIER=1
+
+if not exist "%LOG_DIR%" mkdir "%LOG_DIR%"
+
+set TS=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%
+set TS=%TS: =0%
+set LOG_FILE=%LOG_DIR%\daily_marketing_report_%TS%.log
+
+echo === Daily marketing report started at %date% %time% === > "%LOG_FILE%"
+echo PROJECT_DIR=%PROJECT_DIR% >> "%LOG_FILE%"
+
+REM Wait for today's daily_run to finish (poll every 60s, max 90 min), so that
+REM cmt_notes is synced before the report reads yesterday's new notes.
+set TODAY=%date:~0,4%%date:~5,2%%date:~8,2%
+set /a WAITED=0
+:wait_daily_run
+set DAILY_DONE=
+for %%F in ("%LOG_DIR%\daily_run_%TODAY%_*.log") do findstr /C:"Daily run finished" /C:"Daily run skipped" "%%F" >nul 2>&1 && set DAILY_DONE=1
+if defined DAILY_DONE goto daily_run_ready
+if %WAITED% GEQ 5400 goto daily_run_timeout
+timeout /t 60 /nobreak >nul
+set /a WAITED+=60
+goto wait_daily_run
+:daily_run_timeout
+echo WARN: daily_run not finished after 90 min wait, generating report anyway >> "%LOG_FILE%"
+:daily_run_ready
+
+cd /d "%PROJECT_DIR%"
+call %PYTHON% -u -X utf8 daily_marketing_report.py --send >> "%LOG_FILE%" 2>&1
+set RC=%ERRORLEVEL%
+
+echo daily_marketing_report exit=%RC% >> "%LOG_FILE%"
+echo === Daily marketing report finished at %date% %time% (rc=%RC%) === >> "%LOG_FILE%"
+
+endlocal & exit /b %RC%
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_report_card.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_report_card.py
new file mode 100644
index 0000000..fca2159
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_report_card.py
@@ -0,0 +1,328 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""构造并发送时尚包袋品牌经营日报飞书 Card 2.0。"""
+
+import hashlib
+import json
+import os
+import re
+import shutil
+import subprocess
+from collections import Counter
+from pathlib import Path
+from typing import Any, Iterable
+
+from PIL import Image
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+
+LARK_PROFILE = "hermes-analyzer"
+LARK_CLI = shutil.which("lark-cli.cmd") or shutil.which("lark-cli") or "lark-cli"
+MAX_CARD_IMAGE_WIDTH = 1500
+
+
+def _lark_command() -> list[str]:
+    run_js = (
+        Path(os.environ.get("APPDATA", str(Path.home())))
+        / "npm" / "node_modules" / "@larksuite" / "cli" / "scripts" / "run.js"
+    )
+    if not run_js.exists():
+        run_js = Path(os.path.expandvars(r"%APPDATA%\npm\node_modules\@larksuite\cli\scripts\run.js"))
+    if run_js.exists():
+        return [shutil.which("node") or "node", str(run_js)]
+    return [LARK_CLI]
+
+
+def _section_lines(report: str, start: str, end: str | None, limit: int = 3) -> list[str]:
+    start_match = re.search(rf"(?m)^\s*#*\s*{re.escape(start)}\s*$", report)
+    if not start_match:
+        return []
+    tail = report[start_match.end():]
+    if end:
+        end_match = re.search(rf"(?m)^\s*#*\s*{re.escape(end)}", tail)
+        if end_match:
+            tail = tail[:end_match.start()]
+    lines = []
+    for raw in tail.splitlines():
+        value = re.sub(r"^\s*\d+[\.、]\s*", "", raw).strip()
+        if not value or value.startswith("![") or value.startswith("|"):
+            continue
+        lines.append(value[:180])
+        if len(lines) >= limit:
+            break
+    return lines
+
+
+def _metric_column(value: int, label: str, emphasis: bool = False) -> dict[str, Any]:
+    number_size = "heading-1" if emphasis else "heading-2"
+    return {
+        "tag": "column",
+        "width": "weighted",
+        "weight": 1,
+        "background_style": "grey-50",
+        "padding": "12px",
+        "vertical_spacing": "2px",
+        "elements": [
+            {"tag": "markdown", "content": str(value), "text_size": number_size, "text_align": "center"},
+            {"tag": "markdown", "content": f"<font color='grey'>{label}</font>", "text_size": "notation", "text_align": "center"},
+        ],
+    }
+
+
+def build_dashboard_card(
+    report_date: str,
+    tracking_rows: list[dict[str, Any]],
+    image_key: str,
+) -> dict[str, Any]:
+    """构造只包含核心灯号和看板图片的第一条消息。"""
+    counts = Counter(row.get("light", "无判断") for row in tracking_rows)
+    return {
+        "schema": "2.0",
+        "config": {
+            "update_multi": True,
+            "width_mode": "fill",
+            "enable_forward": True,
+            "summary": {"content": f"{report_date} 时尚包袋品牌经营看板"},
+        },
+        "header": {
+            "title": {"tag": "plain_text", "content": "时尚包袋品牌经营看板"},
+            "subtitle": {"tag": "plain_text", "content": f"{report_date} · Hermes 自动分析"},
+            "template": "blue",
+            "icon": {"tag": "standard_icon", "token": "chart_colorful"},
+            "text_tag_list": [
+                {"tag": "text_tag", "text": {"tag": "plain_text", "content": "数据看板"}, "color": "blue"}
+            ],
+        },
+        "body": {
+            "direction": "vertical",
+            "padding": "12px 12px 20px 12px",
+            "vertical_spacing": "large",
+            "elements": [
+                {
+                    "tag": "column_set",
+                    "flex_mode": "none",
+                    "horizontal_spacing": "12px",
+                    "columns": [
+                        _metric_column(counts.get("🔴", 0), "红灯款式", emphasis=True),
+                        _metric_column(counts.get("🟡", 0), "黄灯款式"),
+                        _metric_column(counts.get("🟢", 0), "绿灯款式"),
+                    ],
+                },
+                {
+                    "tag": "img",
+                    "img_key": image_key,
+                    "alt": {"tag": "plain_text", "content": "全域种草数据看板"},
+                    "scale_type": "fit_horizontal",
+                    "corner_radius": "8px",
+                    "preview": True,
+                },
+            ],
+        },
+    }
+
+
+def _report_without_dashboard_reference(report: str) -> str:
+    return re.sub(
+        r"(?m)^\s*!\[全域种草数据看板\]\([^\n]+\)\s*\n?",
+        "",
+        report,
+        count=1,
+    ).strip()
+
+
+def build_analysis_card(report_date: str, report: str) -> dict[str, Any]:
+    """构造包含完整日报正文的第二条消息，不做行数或字数裁剪。"""
+    analysis = _report_without_dashboard_reference(report) or "无数据"
+    return {
+        "schema": "2.0",
+        "config": {
+            "update_multi": True,
+            "width_mode": "fill",
+            "enable_forward": True,
+            "summary": {"content": f"{report_date} SKU营销运营日报分析"},
+        },
+        "header": {
+            "title": {"tag": "plain_text", "content": "SKU营销运营日报分析"},
+            "subtitle": {"tag": "plain_text", "content": f"{report_date} · 完整文字版"},
+            "template": "turquoise",
+            "icon": {"tag": "standard_icon", "token": "doc_colorful"},
+            "text_tag_list": [
+                {"tag": "text_tag", "text": {"tag": "plain_text", "content": "完整分析"}, "color": "turquoise"}
+            ],
+        },
+        "body": {
+            "direction": "vertical",
+            "padding": "12px 16px 20px 16px",
+            "vertical_spacing": "medium",
+            "elements": [{"tag": "markdown", "content": analysis}],
+        },
+    }
+
+
+def build_daily_report_card(
+    report_date: str,
+    report: str,
+    tracking_rows: list[dict[str, Any]],
+    image_key: str,
+) -> dict[str, Any]:
+    counts = Counter(row.get("light", "无判断") for row in tracking_rows)
+    conclusions = _section_lines(report, "1. 今日经营概览", "2.", 5)
+    if not conclusions:
+        conclusions = _section_lines(report, "一、今日经营概览", "二、", 5)
+    if not conclusions:
+        conclusions = _section_lines(report, "一、核心经营结论", "二、", 3)
+    if not conclusions:
+        conclusions = _section_lines(report, "第一部分 整体经营分析", "第二部分 SKU分析排序", 5)
+    actions = _section_lines(report, "6. 今日运营建议", None, 5)
+    if not actions:
+        actions = _section_lines(report, "七、明日业务动作", None, 5)
+    if not actions:
+        actions = _section_lines(report, "四、明日业务动作", None, 3)
+    if not actions:
+        actions = _section_lines(report, "第三部分 关键预警与建议", None, 5)
+    conclusion_text = "\n".join(f"{idx}. {line}" for idx, line in enumerate(conclusions, 1)) or "无数据"
+    action_text = "\n".join(f"{idx}. {line}" for idx, line in enumerate(actions, 1)) or "无数据"
+
+    return {
+        "schema": "2.0",
+        "config": {
+            "update_multi": True,
+            "width_mode": "fill",
+            "enable_forward": True,
+            "summary": {"content": f"{report_date} 时尚包袋品牌经营日报"},
+        },
+        "header": {
+            "title": {"tag": "plain_text", "content": "时尚包袋品牌经营日报"},
+            "subtitle": {"tag": "plain_text", "content": f"{report_date} · Hermes 自动分析"},
+            "template": "blue",
+            "icon": {"tag": "standard_icon", "token": "chart_colorful"},
+            "text_tag_list": [
+                {"tag": "text_tag", "text": {"tag": "plain_text", "content": "自动日报"}, "color": "blue"}
+            ],
+        },
+        "body": {
+            "direction": "vertical",
+            "padding": "12px 12px 20px 12px",
+            "vertical_spacing": "large",
+            "elements": [
+                {
+                    "tag": "column_set",
+                    "flex_mode": "none",
+                    "horizontal_spacing": "12px",
+                    "columns": [
+                        _metric_column(counts.get("🔴", 0), "红灯款式", emphasis=True),
+                        _metric_column(counts.get("🟡", 0), "黄灯款式"),
+                        _metric_column(counts.get("🟢", 0), "绿灯款式"),
+                    ],
+                },
+                {
+                    "tag": "img",
+                    "img_key": image_key,
+                    "alt": {"tag": "plain_text", "content": "全域种草数据看板"},
+                    "scale_type": "fit_horizontal",
+                    "corner_radius": "8px",
+                    "preview": True,
+                },
+                {"tag": "markdown", "content": f"**今日经营概览**\n{conclusion_text}"},
+                {"tag": "markdown", "content": f"**今日运营建议**\n{action_text}"},
+            ],
+        },
+    }
+
+
+def prepare_feishu_image(source: Path, target: Path) -> Path:
+    target.parent.mkdir(parents=True, exist_ok=True)
+    with Image.open(source) as image:
+        image = image.convert("RGB")
+        if image.width > MAX_CARD_IMAGE_WIDTH:
+            height = round(image.height * MAX_CARD_IMAGE_WIDTH / image.width)
+            image = image.resize((MAX_CARD_IMAGE_WIDTH, height), Image.Resampling.LANCZOS)
+        image.save(target, format="PNG", optimize=True)
+    return target
+
+
+def upload_dashboard_image(image_path: Path) -> str:
+    command = [
+        *_lark_command(), "--profile", LARK_PROFILE,
+        "im", "images", "create",
+        "--data", json.dumps({"image_type": "message"}),
+        "--file", f"image=./{image_path.name}",
+        "--as", "bot", "--format", "json",
+    ]
+    completed = subprocess.run(
+        command,
+        cwd=image_path.parent,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+    )
+    if completed.returncode != 0:
+        raise RuntimeError(f"飞书看板图片上传失败: {completed.stderr or completed.stdout}")
+    payload = json.loads(completed.stdout)
+    image_key = payload.get("image_key") or (payload.get("data") or {}).get("image_key")
+    if not image_key:
+        raise RuntimeError("飞书看板图片上传成功但未返回 image_key")
+    return image_key
+
+
+def send_card_to_recipients(
+    card: dict[str, Any],
+    recipients: Iterable[str],
+    report_date: str,
+    *,
+    message_kind: str = "combined",
+) -> list[dict[str, Any]]:
+    content = json.dumps(card, ensure_ascii=False, separators=(",", ":"))
+    content_digest = hashlib.sha256(content.encode("utf-8")).hexdigest()[:10]
+    results = []
+    for open_id in recipients:
+        kind_token = re.sub(r"[^a-z0-9]", "", message_kind.lower())[:1] or "m"
+        idempotency_key = f"daily-{report_date}-{kind_token}-{content_digest}-{open_id[-6:]}"
+        command = [
+            *_lark_command(), "--profile", LARK_PROFILE,
+            "im", "+messages-send",
+            "--user-id", open_id,
+            "--msg-type", "interactive",
+            "--content", content,
+            "--idempotency-key", idempotency_key,
+            "--as", "bot", "--format", "json",
+        ]
+        completed = subprocess.run(
+            command,
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+        )
+        if completed.returncode != 0:
+            raise RuntimeError(f"飞书卡片发送失败({open_id}): {completed.stderr or completed.stdout}")
+        results.append(json.loads(completed.stdout))
+    return results
+
+
+def send_daily_report_cards(
+    report_date: str,
+    report: str,
+    tracking_rows: list[dict[str, Any]],
+    chart_path: Path,
+    recipients: Iterable[str],
+) -> list[dict[str, Any]]:
+    prepared_dir = PATHS.tmp_root / "daily_report_card"
+    prepared_dir.mkdir(parents=True, exist_ok=True)
+    prepared = prepared_dir / f"{chart_path.stem}_feishu.png"
+    prepare_feishu_image(chart_path, prepared)
+    try:
+        image_key = upload_dashboard_image(prepared)
+    finally:
+        prepared.unlink(missing_ok=True)
+    recipients = tuple(recipients)
+    dashboard_card = build_dashboard_card(report_date, tracking_rows, image_key)
+    analysis_card = build_analysis_card(report_date, report)
+    dashboard_results = send_card_to_recipients(
+        dashboard_card, recipients, report_date, message_kind="dashboard"
+    )
+    analysis_results = send_card_to_recipients(
+        analysis_card, recipients, report_date, message_kind="analysis"
+    )
+    return dashboard_results + analysis_results
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_report_charts.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_report_charts.py
new file mode 100644
index 0000000..72340ee
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_report_charts.py
@@ -0,0 +1,518 @@
+"""使用 Pillow 生成全款式同权经营日报长图看板。"""
+
+from pathlib import Path
+from textwrap import wrap
+from typing import Any, Callable
+
+from PIL import Image, ImageDraw, ImageFont, ImageOps
+
+try:
+    from .daily_dashboard_analytics import build_dashboard_facts
+except ImportError:
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.daily_dashboard_analytics import build_dashboard_facts
+
+
+WIDTH, HEIGHT = 1600, 2500
+BG = "#F7F8FA"
+PANEL = "#FFFFFF"
+TEXT = "#18212F"
+MUTED = "#667085"
+GRID = "#E4E7EC"
+RED = "#E5484D"
+YELLOW = "#E8A317"
+GREEN = "#30A46C"
+BLUE = "#3B82F6"
+ORANGE = "#F97316"
+PURPLE = "#8B5CF6"
+LIGHT_BLUE = "#EFF6FF"
+LIGHT_RED = "#FFF1F1"
+LIGHT_YELLOW = "#FFF8E6"
+PLATFORM_COLORS = {"天猫": RED, "京东": BLUE, "抖音电商": TEXT}
+
+
+def _font(size: int, bold: bool = False):
+    candidates = [
+        Path(r"C:\Windows\Fonts\msyhbd.ttc" if bold else r"C:\Windows\Fonts\msyh.ttc"),
+        Path(r"C:\Windows\Fonts\simhei.ttf"),
+    ]
+    for path in candidates:
+        if path.exists():
+            return ImageFont.truetype(str(path), size)
+    return ImageFont.load_default()
+
+
+def _pct(value: Any) -> float:
+    try:
+        return float(value or 0)
+    except (TypeError, ValueError):
+        return 0.0
+
+
+def _fmt_int(value: Any) -> str:
+    if value is None:
+        return "无数据"
+    return f"{int(value):,}"
+
+
+def _fmt_rate(value: Any) -> str:
+    if value is None:
+        return "无数据"
+    return f"{float(value):.1%}"
+
+
+def _sum_known(rows: list[dict[str, Any]], key: str) -> float | None:
+    values = [float(row[key]) for row in rows if row.get(key) is not None]
+    return sum(values) if values else None
+
+
+def _ratio(numerator: float | None, denominator: float | None) -> float | None:
+    if numerator is None or denominator is None or denominator <= 0:
+        return None
+    return numerator / denominator
+
+
+def build_overview_cards(rows: list[dict[str, Any]], dashboard: dict[str, Any]) -> list[dict[str, Any]]:
+    today_sales = _sum_known(rows, "sales")
+    yesterday_sales = _sum_known(rows, "sales_yesterday")
+    baseline_sales = _sum_known(rows, "avg_sales_7d")
+    today_visitors = _sum_known(rows, "visitors")
+    yesterday_visitors = _sum_known(rows, "visitors_yesterday")
+    baseline_visitors = _sum_known(rows, "avg_visitors_7d")
+    notes = dashboard.get("daily_new_notes", [])
+    note_views = [float(row["view_count"]) for row in notes if row.get("view_count") is not None]
+    return [
+        {
+            "key": "sales", "label": "销量",
+            "today": today_sales, "yesterday": yesterday_sales, "baseline_7d": baseline_sales,
+            "formatter": _fmt_int,
+        },
+        {
+            "key": "visitors", "label": "访客",
+            "today": today_visitors, "yesterday": yesterday_visitors, "baseline_7d": baseline_visitors,
+            "formatter": _fmt_int,
+        },
+        {
+            "key": "conversion", "label": "整体转化率",
+            "today": _ratio(today_sales, today_visitors),
+            "yesterday": _ratio(yesterday_sales, yesterday_visitors),
+            "baseline_7d": _ratio(baseline_sales, baseline_visitors),
+            "formatter": _fmt_rate,
+        },
+        {
+            "key": "content", "label": "新发内容曝光",
+            "today": sum(note_views) if note_views else None,
+            "yesterday": None, "baseline_7d": None,
+            "formatter": _fmt_int,
+            "note": f"昨日新发 {len(notes)} 篇｜历史曝光快照无数据",
+        },
+    ]
+
+
+def classify_style_action(row: dict[str, Any]) -> tuple[str, str]:
+    sales = row.get("sales_change_yesterday")
+    visitors = row.get("visitor_change_yesterday")
+    conversion = row.get("conversion_change_yesterday")
+    refund = row.get("refund_change_yesterday")
+    if sales is not None and conversion is not None and refund is not None and sales < 0 and conversion < 0 and refund > 0:
+        return "风险型", "查退款/控投放"
+    if refund is not None and refund >= 0.30:
+        return "风险型", "查退款/控投放"
+    if sales is not None and visitors is not None and conversion is not None and sales > 0 and visitors > 0 and conversion >= 0:
+        return "增长型", "加达人/备货"
+    if visitors is not None and conversion is not None and visitors > 0 and conversion < 0:
+        return "流量机会", "优化详情页"
+    if (row.get("cart_change_yesterday") or 0) >= 0.20 or (row.get("recent_note_exposure") or 0) > 0:
+        return "潜力型", "加曝光测试"
+    if row.get("sales") is None:
+        return "无数据", "补齐数据"
+    return "稳定型", "维护转化"
+
+
+def select_chart_styles(rows: list[dict[str, Any]], limit: int = 6) -> list[dict[str, Any]]:
+    """保留旧调用兼容；新看板不会再只展示这些款式。"""
+    anomalies = [row for row in rows if row.get("light") in {"🔴", "🟡"} and row.get("visitors") is not None]
+
+    def visitor_delta(row):
+        return float(row.get("visitors") or 0) - float(row.get("avg_visitors_7d") or 0)
+
+    def impact(row):
+        vd = abs(visitor_delta(row))
+        cd = abs(float(row.get("cart_users") or 0) - float(row.get("avg_cart_users_7d") or 0))
+        return vd + cd * 10
+
+    selected, used = [], set()
+    groups = [
+        sorted((row for row in anomalies if visitor_delta(row) < 0), key=lambda row: abs(visitor_delta(row)), reverse=True),
+        sorted((row for row in anomalies if visitor_delta(row) > 0 or _pct(row.get("cart_change")) > 0), key=impact, reverse=True),
+        sorted((row for row in anomalies if float(row.get("visitors") or 0) < 200), key=lambda row: max(abs(_pct(row.get("visitor_change"))), abs(_pct(row.get("cart_change")))), reverse=True),
+    ]
+    for candidates, target in zip(groups, (2, 2, 1)):
+        added = 0
+        for row in candidates:
+            if row["style_name"] in used:
+                continue
+            selected.append(row)
+            used.add(row["style_name"])
+            added += 1
+            if added >= target or len(selected) >= limit:
+                break
+    for row in sorted(anomalies, key=impact, reverse=True):
+        if len(selected) >= limit:
+            break
+        if row["style_name"] not in used:
+            selected.append(row)
+            used.add(row["style_name"])
+    return selected[:limit]
+
+
+def _panel(draw: ImageDraw.ImageDraw, box: tuple[int, int, int, int], title: str, title_size: int = 23) -> None:
+    draw.rounded_rectangle(box, radius=18, fill=PANEL)
+    draw.text((box[0] + 20, box[1] + 16), title, font=_font(title_size, True), fill=TEXT)
+
+
+def _ranking_panel(
+    draw: ImageDraw.ImageDraw,
+    box: tuple[int, int, int, int],
+    title: str,
+    rows: list[dict[str, Any]],
+    value_key: str,
+    formatter: Callable[[Any], str],
+    empty_text: str = "无数据",
+) -> None:
+    _panel(draw, box, title, 21)
+    x0, y0, x1, _ = box
+    if not rows:
+        lines = wrap(empty_text, width=22) or ["无数据"]
+        for idx, line in enumerate(lines[:4]):
+            draw.text((x0 + 20, y0 + 76 + idx * 28), line, font=_font(16), fill=MUTED)
+        return
+    max_value = max(float(row.get(value_key) or 0) for row in rows) or 1
+    for idx, row in enumerate(rows[:5], 1):
+        y = y0 + 63 + (idx - 1) * 37
+        draw.text((x0 + 20, y), str(idx), font=_font(16, True), fill=BLUE)
+        draw.text((x0 + 48, y), str(row.get("display_name") or row.get("style_name") or "无数据")[:11], font=_font(16), fill=TEXT)
+        bar_x, bar_w = x0 + 190, x1 - x0 - 285
+        width = max(2, bar_w * float(row.get(value_key) or 0) / max_value)
+        draw.rounded_rectangle((bar_x, y + 5, bar_x + bar_w, y + 15), radius=5, fill=GRID)
+        draw.rounded_rectangle((bar_x, y + 5, bar_x + width, y + 15), radius=5, fill=BLUE)
+        draw.text((x1 - 84, y - 1), formatter(row.get(value_key)), font=_font(15, True), fill=TEXT)
+
+
+def _matrix_header(draw: ImageDraw.ImageDraw, x: int, y: int) -> None:
+    headers = [
+        ("款式", 0), ("今日", 108), ("近7日", 150), ("访客", 200),
+        ("转化", 254), ("曝光", 310), ("预测7日", 366), ("判断", 442), ("动作", 536),
+    ]
+    for label, offset in headers:
+        draw.text((x + offset, y), label, font=_font(14, True), fill=MUTED)
+
+
+def _split_style_groups(style_groups: list[dict[str, Any]]) -> tuple[list[dict[str, Any]], list[dict[str, Any]]]:
+    if len(style_groups) < 2:
+        return style_groups, []
+    total = sum(len(group.get("styles", [])) for group in style_groups)
+    running = 0
+    best_index = 1
+    best_gap = total
+    for index, group in enumerate(style_groups[:-1], 1):
+        running += len(group.get("styles", []))
+        gap = abs(total - running * 2)
+        if gap < best_gap:
+            best_index, best_gap = index, gap
+    return style_groups[:best_index], style_groups[best_index:]
+
+
+def _style_matrix(draw: ImageDraw.ImageDraw, box: tuple[int, int, int, int], style_groups: list[dict[str, Any]]) -> None:
+    total_styles = sum(len(group.get("styles", [])) for group in style_groups)
+    _panel(draw, box, f"三、全部 {total_styles} 款SKU表现｜按风格分组", 25)
+    x0, y0, x1, _ = box
+    draw.text(
+        (x1 - 760, y0 + 20),
+        "近7日含报告日｜预测=近3日日均×7，箭头较前4日日均｜仅作趋势估算",
+        font=_font(13),
+        fill=MUTED,
+    )
+    groups = _split_style_groups(style_groups)
+    column_boxes = [(x0 + 20, x0 + 750), (x0 + 790, x1 - 20)]
+    for category_groups, (left, right) in zip(groups, column_boxes):
+        header_y = y0 + 70
+        draw.rounded_rectangle((left, header_y - 8, right, header_y + 28), radius=8, fill=LIGHT_BLUE)
+        _matrix_header(draw, left + 10, header_y)
+        cursor_y = header_y + 38
+        for category_group in category_groups:
+            category = category_group.get("style_category") or "未分类"
+            rows = category_group.get("styles", [])
+            draw.rounded_rectangle((left, cursor_y, right, cursor_y + 26), radius=7, fill=LIGHT_YELLOW)
+            draw.text((left + 12, cursor_y + 3), f"{category}（{len(rows)}款）", font=_font(14, True), fill=ORANGE)
+            cursor_y += 31
+            for idx, row in enumerate(rows):
+                y = cursor_y + idx * 38
+                if idx % 2:
+                    draw.rectangle((left, y - 4, right, y + 28), fill="#FAFBFC")
+                light_color = {"🔴": RED, "🟡": YELLOW, "🟢": GREEN}.get(row.get("light"), MUTED)
+                draw.ellipse((left + 8, y + 4, left + 18, y + 14), fill=light_color)
+                values = [
+                    (str(row.get("style_name") or "无数据")[:7], 24, TEXT),
+                    (_fmt_int(row.get("sales")), 114, TEXT),
+                    (_fmt_int(row.get("sales_7d_total")), 156, TEXT),
+                    (_fmt_int(row.get("visitors")), 206, TEXT),
+                    (_fmt_rate(row.get("conversion_rate")), 260, TEXT),
+                    (_fmt_int(row.get("recent_note_exposure")), 316, PURPLE if row.get("recent_note_exposure") else MUTED),
+                ]
+                forecast = row.get("sales_forecast_7d")
+                forecast_change = row.get("sales_forecast_change")
+                if forecast is None:
+                    forecast_text, forecast_color = "无数据", MUTED
+                elif forecast_change is None:
+                    forecast_text, forecast_color = f"·{_fmt_int(forecast)}", MUTED
+                elif forecast_change > 0:
+                    forecast_text, forecast_color = f"↑{_fmt_int(forecast)}", GREEN
+                elif forecast_change < 0:
+                    forecast_text, forecast_color = f"↓{_fmt_int(forecast)}", RED
+                else:
+                    forecast_text, forecast_color = f"→{_fmt_int(forecast)}", MUTED
+                diagnosis, action = classify_style_action(row)
+                values.extend([
+                    (forecast_text, 372, forecast_color),
+                    (diagnosis, 448, RED if diagnosis == "风险型" else GREEN if diagnosis == "增长型" else TEXT),
+                    (action, 542, TEXT),
+                ])
+                for value, offset, color in values:
+                    draw.text((left + offset, y), str(value)[:10], font=_font(12, offset in (24, 448, 542)), fill=color)
+            cursor_y += len(rows) * 38 + 7
+
+
+def _change_text(current: float | None, baseline: float | None) -> tuple[str, str]:
+    if current is None or baseline is None or baseline == 0:
+        return "无数据", MUTED
+    change = (current - baseline) / baseline
+    if change > 0:
+        return f"↑{change:.1%}", GREEN
+    if change < 0:
+        return f"↓{abs(change):.1%}", RED
+    return "持平", MUTED
+
+
+def _overview_section(draw: ImageDraw.ImageDraw, box: tuple[int, int, int, int], cards: list[dict[str, Any]]) -> None:
+    _panel(draw, box, "一、今日经营概览｜今日 vs 昨日 vs 7日均值", 25)
+    x0, y0, x1, _ = box
+    gap = 14
+    card_width = (x1 - x0 - 40 - gap * 3) // 4
+    for idx, card in enumerate(cards[:4]):
+        left = x0 + 20 + idx * (card_width + gap)
+        right = left + card_width
+        top = y0 + 62
+        draw.rounded_rectangle((left, top, right, box[3] - 18), radius=13, fill=LIGHT_BLUE if idx < 3 else "#F8F5FF")
+        draw.text((left + 14, top + 12), card["label"], font=_font(18, True), fill=BLUE if idx < 3 else PURPLE)
+        formatter = card["formatter"]
+        draw.text((left + 14, top + 45), formatter(card.get("today")), font=_font(28, True), fill=TEXT)
+        vs_yesterday, color_y = _change_text(card.get("today"), card.get("yesterday"))
+        vs_baseline, color_7 = _change_text(card.get("today"), card.get("baseline_7d"))
+        draw.text((left + 14, top + 86), f"较昨日 {vs_yesterday}", font=_font(14, True), fill=color_y)
+        draw.text((left + 154, top + 86), f"较7日 {vs_baseline}", font=_font(14, True), fill=color_7)
+
+        values = [("今日", card.get("today"), BLUE), ("昨日", card.get("yesterday"), MUTED), ("7日", card.get("baseline_7d"), YELLOW)]
+        known = [float(value) for _, value, _ in values if value is not None]
+        maximum = max(known) if known else 1
+        for row_index, (label, value, color) in enumerate(values):
+            y = top + 124 + row_index * 30
+            draw.text((left + 14, y), label, font=_font(13), fill=MUTED)
+            draw.rounded_rectangle((left + 56, y + 4, right - 78, y + 14), radius=5, fill=GRID)
+            if value is not None:
+                bar_width = max(2, (right - left - 148) * float(value) / maximum)
+                draw.rounded_rectangle((left + 56, y + 4, left + 56 + bar_width, y + 14), radius=5, fill=color)
+            draw.text((right - 70, y - 1), formatter(value), font=_font(12, True), fill=TEXT if value is not None else MUTED)
+        if card.get("note"):
+            draw.text((left + 14, top + 220), str(card["note"])[:30], font=_font(12), fill=MUTED)
+
+
+def _focus_styles(styles: list[dict[str, Any]], target: str) -> list[dict[str, Any]]:
+    candidates = []
+    for row in styles:
+        if target == "增长型":
+            sales_change = row.get("sales_change")
+            visitor_change = row.get("visitor_change")
+            refund_rate = row.get("refund_rate")
+            if sales_change is None or sales_change <= 0 or (visitor_change is not None and visitor_change < -0.05):
+                continue
+            if refund_rate is not None and refund_rate >= 0.40:
+                continue
+            score = float(sales_change) + max(float(visitor_change or 0), 0)
+            action = "加达人/备货" if (row.get("conversion_change_7d") or 0) >= -0.002 else "加曝光测试"
+        else:
+            sales_change = row.get("sales_change")
+            conversion_change = row.get("conversion_change_7d")
+            refund_change = row.get("refund_change")
+            refund_rate = row.get("refund_rate")
+            strict_risk = (
+                sales_change is not None and sales_change < 0
+                and conversion_change is not None and conversion_change < 0
+                and refund_change is not None and refund_change > 0
+            )
+            if not strict_risk and not (sales_change is not None and sales_change < 0 and (refund_rate or 0) >= 0.40):
+                continue
+            score = abs(min(float(sales_change or 0), 0)) + max(float(refund_change or 0), 0) + float(refund_rate or 0)
+            action = "查退款/控投放"
+        candidates.append((score, {**row, "prompt_diagnosis": target, "prompt_action": action}))
+    return [row for _, row in sorted(candidates, key=lambda item: item[0], reverse=True)[:3]]
+
+
+def _focus_panel(draw: ImageDraw.ImageDraw, box: tuple[int, int, int, int], title: str, rows: list[dict[str, Any]], risk: bool = False) -> None:
+    _panel(draw, box, title, 22)
+    x0, y0, x1, _ = box
+    accent = RED if risk else GREEN
+    if not rows:
+        draw.text((x0 + 20, y0 + 74), "无符合规则的SKU", font=_font(16), fill=MUTED)
+        return
+    for idx, row in enumerate(rows, 1):
+        y = y0 + 62 + (idx - 1) * 65
+        draw.rounded_rectangle((x0 + 18, y, x1 - 18, y + 54), radius=10, fill=LIGHT_RED if risk else "#EFFAF4")
+        draw.text((x0 + 30, y + 14), str(idx), font=_font(17, True), fill=accent)
+        draw.text((x0 + 60, y + 8), str(row.get("style_name") or "无数据")[:12], font=_font(17, True), fill=TEXT)
+        sales_text, sales_color = _change_text(row.get("sales"), row.get("avg_sales_7d"))
+        visitor_text, visitor_color = _change_text(row.get("visitors"), row.get("avg_visitors_7d"))
+        draw.text((x0 + 210, y + 10), f"销量较7日 {sales_text}", font=_font(13, True), fill=sales_color)
+        draw.text((x0 + 350, y + 10), f"访客较7日 {visitor_text}", font=_font(13, True), fill=visitor_color)
+        draw.text((x0 + 500, y + 10), row.get("prompt_action") or "无数据", font=_font(14, True), fill=accent)
+        draw.text((x0 + 60, y + 32), f"转化 {_fmt_rate(row.get('conversion_rate'))}｜曝光 {_fmt_int(row.get('recent_note_exposure'))}", font=_font(12), fill=MUTED)
+
+
+def _channel_section(
+    draw: ImageDraw.ImageDraw,
+    box: tuple[int, int, int, int],
+    notes: list[dict[str, Any]],
+    styles: list[dict[str, Any]],
+) -> None:
+    _panel(draw, box, "四、达人营销效果｜昨日新发布内容", 25)
+    x0, y0, x1, _ = box
+    platforms = [("小红书", RED), ("B站", BLUE), ("抖音", TEXT)]
+    gap = 18
+    width = (x1 - x0 - 40 - gap * 2) // 3
+    for idx, (platform, color) in enumerate(platforms):
+        items = [row for row in notes if row.get("platform") == platform]
+        left = x0 + 20 + idx * (width + gap)
+        top = y0 + 66
+        draw.rounded_rectangle((left, top, left + width, top + 230), radius=13, fill=LIGHT_BLUE)
+        draw.text((left + 16, top + 14), platform, font=_font(20, True), fill=color)
+        exposure = sum(int(row.get("view_count") or 0) for row in items) if items else None
+        draw.text((left + 16, top + 54), f"新发 {len(items)} 篇", font=_font(19, True), fill=TEXT)
+        draw.text((left + 150, top + 56), f"曝光 {_fmt_int(exposure)}", font=_font(16, True), fill=TEXT)
+        if items:
+            best = max(items, key=lambda row: row.get("view_count") or 0)
+            _draw_wrapped(draw, f"最佳：{best.get('style_name') or '无款式'} · {best.get('creator') or '无达人'}", left + 16, top + 96, 28, 2, TEXT, 15, True)
+            _draw_wrapped(draw, best.get("title") or "标题无数据", left + 16, top + 146, 32, 2, MUTED, 13)
+        else:
+            draw.text((left + 16, top + 100), "无新发布内容", font=_font(15), fill=MUTED)
+
+    best_note = max(notes, key=lambda row: row.get("view_count") or 0) if notes else None
+    best_channel = best_note.get("platform") if best_note else "无数据"
+    best_style = best_note.get("style_name") if best_note else "无数据"
+    comment_samples = sum(int((row.get("multi_source_signals") or {}).get("comment_count") or 0) for row in styles)
+    summary_y = y0 + 322
+    summary_items = [
+        ("最佳渠道", best_channel or "无数据", "优先复盘可量化内容"),
+        ("最佳内容SKU", best_style or "无数据", f"曝光 {_fmt_int(best_note.get('view_count') if best_note else None)}"),
+        ("内容反馈", f"新发 {len(notes)} 篇", "跨渠道曝光仍不完整"),
+        ("用户洞察", f"评论样本 {comment_samples}", "样本不足则不推断"),
+    ]
+    summary_width = (x1 - x0 - 40 - 14 * 3) // 4
+    for idx, (label, value, note) in enumerate(summary_items):
+        left = x0 + 20 + idx * (summary_width + 14)
+        draw.rounded_rectangle((left, summary_y, left + summary_width, summary_y + 150), radius=12, fill="#F8F5FF")
+        draw.text((left + 14, summary_y + 14), label, font=_font(15, True), fill=PURPLE)
+        draw.text((left + 14, summary_y + 48), str(value)[:18], font=_font(20, True), fill=TEXT)
+        _draw_wrapped(draw, note, left + 14, summary_y + 86, 24, 2, MUTED, 13)
+    draw.text((x0 + 20, box[3] - 38), "说明：曝光为数据库保存的当前快照；缺少历史快照时不推算单日播放增量。", font=_font(14), fill=MUTED)
+
+
+def _draw_wrapped(
+    draw: ImageDraw.ImageDraw,
+    text: str,
+    x: int,
+    y: int,
+    width: int = 28,
+    max_lines: int = 3,
+    color: str = TEXT,
+    size: int = 14,
+    bold: bool = False,
+) -> int:
+    lines = wrap(str(text or "无数据"), width=width)[:max_lines] or ["无数据"]
+    for line in lines:
+        draw.text((x, y), line, font=_font(size, bold), fill=color)
+        y += size + 8
+    return y
+
+
+def _paste_product_image(canvas: Image.Image, draw: ImageDraw.ImageDraw, winner: dict[str, Any], x: int, y: int) -> None:
+    size = 112
+    path = winner.get("local_image_path")
+    if path:
+        try:
+            with Image.open(path) as source:
+                thumb = ImageOps.fit(source.convert("RGB"), (size, size), method=Image.Resampling.LANCZOS)
+            mask = Image.new("L", (size, size), 0)
+            ImageDraw.Draw(mask).rounded_rectangle((0, 0, size, size), radius=10, fill=255)
+            canvas.paste(thumb, (x, y), mask)
+            ctr = winner.get("image_ctr")
+            ctr_text = f" CTR {float(ctr):.1f}%" if ctr is not None else ""
+            draw.text((x, y + size + 7), f"图片：天猫{ctr_text}", font=_font(12), fill=MUTED)
+            return
+        except (OSError, ValueError):
+            pass
+    draw.rounded_rectangle((x, y, x + size, y + size), radius=10, fill=PANEL, outline=GRID)
+    draw.text((x + 17, y + 46), "图片无数据", font=_font(13), fill=MUTED)
+
+
+def _winner_block(canvas: Image.Image, draw: ImageDraw.ImageDraw, x: int, y: int, width: int, title: str, winner: dict[str, Any] | None, role: str) -> int:
+    draw.rounded_rectangle((x, y, x + width, y + 260), radius=12, fill=LIGHT_BLUE if role == "conversion" else "#F8F5FF")
+    draw.text((x + 14, y + 12), title, font=_font(17, True), fill=BLUE if role == "conversion" else PURPLE)
+    if not winner:
+        draw.text((x + 14, y + 50), "无数据", font=_font(15), fill=MUTED)
+        return y + 260
+    metric = _fmt_rate(winner.get("conversion_rate")) if role == "conversion" else _fmt_int(winner.get("visitors"))
+    unit = "转化率" if role == "conversion" else "访客"
+    draw.text((x + 14, y + 45), f"{winner['style_name']}｜{unit} {metric}", font=_font(16, True), fill=TEXT)
+    _paste_product_image(canvas, draw, winner, x + 14, y + 80)
+    text_x = x + 140
+    persona_date = f"（{winner.get('persona_date')}）" if winner.get("persona_date") else ""
+    cursor = _draw_wrapped(draw, f"画像{persona_date}：{winner.get('persona_text') or '无数据'}", text_x, y + 78, 20, 3, TEXT, 13)
+    cursor = _draw_wrapped(draw, winner.get("interest_hypothesis") or "无数据", text_x, cursor + 2, 20, 2, MUTED, 13)
+    _draw_wrapped(draw, winner.get("optimization") or "无数据", text_x, cursor + 2, 20, 3, TEXT, 13)
+    return y + 260
+
+
+def _persona_section(canvas: Image.Image, draw: ImageDraw.ImageDraw, box: tuple[int, int, int, int], cards: list[dict[str, Any]]) -> None:
+    _panel(draw, box, "三平台高转化 / 高访客款式画像与增长空间", 25)
+    x0, y0, x1, _ = box
+    draw.text((x1 - 570, y0 + 20), "兴趣爱好均为画像与内容方向的推测，不是数据库原始标签", font=_font(14), fill=MUTED)
+    gap = 16
+    card_width = (x1 - x0 - 40 - gap * 2) // 3
+    for idx, card in enumerate(cards[:3]):
+        x = x0 + 20 + idx * (card_width + gap)
+        draw.text((x, y0 + 66), card.get("platform") or "无数据", font=_font(21, True), fill=PLATFORM_COLORS.get(card.get("platform"), BLUE))
+        next_y = _winner_block(canvas, draw, x, y0 + 103, card_width, "转化最高", card.get("top_conversion"), "conversion")
+        _winner_block(canvas, draw, x, next_y + 14, card_width, "访客最高", card.get("top_visitors"), "visitors")
+
+
+def generate_dashboard(report_date: str, rows: list[dict[str, Any]], enrichment: dict[str, Any], output_path: Path) -> Path:
+    output_path = Path(output_path)
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+    dashboard = enrichment.get("dashboard") or build_dashboard_facts(
+        report_date, rows, enrichment.get("platform_metrics", []), enrichment.get("personas", []), {}
+    )
+
+    image = Image.new("RGB", (WIDTH, HEIGHT), BG)
+    draw = ImageDraw.Draw(image)
+    draw.text((42, 24), f"{report_date} 品牌SKU营销运营看板", font=_font(36, True), fill=TEXT)
+    draw.text((42, 74), "经营双基准｜近7日销量与未来趋势预估｜全款式同权｜达人曝光缺失不推算", font=_font(18), fill=MUTED)
+
+    overview_cards = build_overview_cards(rows, dashboard)
+    _overview_section(draw, (32, 112, 1568, 442), overview_cards)
+    styles = dashboard.get("styles", [])
+    _focus_panel(draw, (32, 464, 790, 730), "二、增长SKU TOP3", _focus_styles(styles, "增长型"))
+    _focus_panel(draw, (810, 464, 1568, 730), "风险SKU TOP3", _focus_styles(styles, "风险型"), risk=True)
+    style_groups = dashboard.get("style_groups") or [{"style_category": "未分类", "styles": dashboard.get("styles", [])}]
+    _style_matrix(draw, (32, 752, 1568, 1780), style_groups)
+    _channel_section(draw, (32, 1802, 1568, 2468), dashboard.get("daily_new_notes", []), styles)
+    image.save(output_path, format="PNG", optimize=True)
+    return output_path
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_run.bat b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_run.bat
new file mode 100644
index 0000000..b9f17a6
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_run.bat
@@ -0,0 +1,61 @@
+@echo off
+if not defined GYXX_PYTHON set "GYXX_PYTHON=python"
+REM Daily task: run 3-platform V2 scrape + sync metrics to cmt_notes (INSERT missing notes every day)
+REM Register: schtasks /Create /SC DAILY /TN YingxiaoYunying_DailyRun /TR %PROJECT_DIR%\data\tools\daily_run.bat /ST 06:00 /F
+
+setlocal
+
+for %%I in ("%~dp0..\..") do set "PROJECT_DIR=%%~fI"
+if not defined GYXX_DATA_ROOT (
+    if not defined GYXX_PROJECT_ROOT (
+        echo ERROR: GYXX_DATA_ROOT or GYXX_PROJECT_ROOT must be defined. 1>&2
+        exit /b 3
+    )
+    set "GYXX_DATA_ROOT=%GYXX_PROJECT_ROOT%\var"
+)
+set LOG_DIR=%GYXX_DATA_ROOT%\logs\content_marketing\scheduled
+REM Task Scheduler (SYSTEM account) has no user PATH -> lock to venv python
+set PYTHON=%GYXX_PYTHON%
+where %PYTHON% >nul 2>&1 || set PYTHON=python
+
+if not exist "%LOG_DIR%" mkdir "%LOG_DIR%"
+
+set TS=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%
+set TS=%TS: =0%
+
+set LOG_FILE=%LOG_DIR%\daily_run_%TS%.log
+
+echo === Daily run started at %date% %time% === > "%LOG_FILE%"
+echo PROJECT_DIR=%PROJECT_DIR% >> "%LOG_FILE%"
+
+cd /d "%PROJECT_DIR%"
+
+echo --- [step 1] run_all.py (includes feishu_mapping refresh) --- >> "%LOG_FILE%"
+call %PYTHON% run_all.py >> "%LOG_FILE%" 2>&1
+set RC_RUN=%ERRORLEVEL%
+echo run_all exit=%RC_RUN% >> "%LOG_FILE%"
+
+echo --- [step 2] sync_metrics_to_cmt_notes.py (主采集后首次落库) --- >> "%LOG_FILE%"
+call %PYTHON% data\tools\sync_metrics_to_cmt_notes.py >> "%LOG_FILE%" 2>&1
+set RC_SYNC_PRE=%ERRORLEVEL%
+echo sync_pre exit=%RC_SYNC_PRE% >> "%LOG_FILE%"
+
+echo --- [step 3] retry_failed.py (B站 + 蒲公英; 星图每日只搜索一轮) --- >> "%LOG_FILE%"
+call %PYTHON% data\tools\retry_failed.py --platform bili,pgy --max-attempts 1 --stale-hours 4 >> "%LOG_FILE%" 2>&1
+set RC_RETRY=%ERRORLEVEL%
+echo retry_failed exit=%RC_RETRY% >> "%LOG_FILE%"
+
+echo --- [step 4] sync_metrics_to_cmt_notes.py (补跑后再次落库) --- >> "%LOG_FILE%"
+call %PYTHON% data\tools\sync_metrics_to_cmt_notes.py >> "%LOG_FILE%" 2>&1
+set RC_SYNC_POST=%ERRORLEVEL%
+echo sync_post exit=%RC_SYNC_POST% >> "%LOG_FILE%"
+
+set FINAL_RC=0
+if not "%RC_RUN%"=="0" set FINAL_RC=1
+if not "%RC_SYNC_PRE%"=="0" set FINAL_RC=1
+if not "%RC_RETRY%"=="0" set FINAL_RC=1
+if not "%RC_SYNC_POST%"=="0" set FINAL_RC=1
+
+echo === Daily run finished at %date% %time% (run=%RC_RUN%, sync_pre=%RC_SYNC_PRE%, retry=%RC_RETRY%, sync_post=%RC_SYNC_POST%, final=%FINAL_RC%) === >> "%LOG_FILE%"
+
+endlocal & exit /b %FINAL_RC%
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_run_with_backfill.bat b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_run_with_backfill.bat
new file mode 100644
index 0000000..3ff3369
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/daily_run_with_backfill.bat
@@ -0,0 +1,61 @@
+@echo off
+if not defined GYXX_PYTHON set "GYXX_PYTHON=python"
+REM Monday task: run 3-platform V2 + sync metrics to cmt_notes
+REM Register: schtasks /Create /SC WEEKLY /D MON /TN YingxiaoYunying_MondayBackfill /TR %PROJECT_DIR%\data\tools\daily_run_with_backfill.bat /ST 06:00 /F
+
+setlocal
+
+for %%I in ("%~dp0..\..") do set "PROJECT_DIR=%%~fI"
+if not defined GYXX_DATA_ROOT (
+    if not defined GYXX_PROJECT_ROOT (
+        echo ERROR: GYXX_DATA_ROOT or GYXX_PROJECT_ROOT must be defined. 1>&2
+        exit /b 3
+    )
+    set "GYXX_DATA_ROOT=%GYXX_PROJECT_ROOT%\var"
+)
+set LOG_DIR=%GYXX_DATA_ROOT%\logs\content_marketing\scheduled
+REM Task Scheduler (SYSTEM account) has no user PATH -> lock to venv python
+set PYTHON=%GYXX_PYTHON%
+where %PYTHON% >nul 2>&1 || set PYTHON=python
+
+if not exist "%LOG_DIR%" mkdir "%LOG_DIR%"
+
+set TS=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%
+set TS=%TS: =0%
+
+set LOG_FILE=%LOG_DIR%\monday_run_%TS%.log
+
+echo === Monday run started at %date% %time% === > "%LOG_FILE%"
+echo PROJECT_DIR=%PROJECT_DIR% >> "%LOG_FILE%"
+
+cd /d "%PROJECT_DIR%"
+
+echo --- [step 1] run_all.py (includes feishu_mapping refresh) --- >> "%LOG_FILE%"
+call %PYTHON% run_all.py >> "%LOG_FILE%" 2>&1
+set RC_RUN=%ERRORLEVEL%
+echo run_all exit=%RC_RUN% >> "%LOG_FILE%"
+
+echo --- [step 2] sync_metrics_to_cmt_notes.py (主采集后首次落库) --- >> "%LOG_FILE%"
+call %PYTHON% data\tools\sync_metrics_to_cmt_notes.py >> "%LOG_FILE%" 2>&1
+set RC_SYNC_PRE=%ERRORLEVEL%
+echo sync_pre exit=%RC_SYNC_PRE% >> "%LOG_FILE%"
+
+echo --- [step 3] retry_failed.py --- >> "%LOG_FILE%"
+call %PYTHON% data\tools\retry_failed.py --max-attempts 1 --stale-hours 4 >> "%LOG_FILE%" 2>&1
+set RC_RETRY=%ERRORLEVEL%
+echo retry_failed exit=%RC_RETRY% >> "%LOG_FILE%"
+
+echo --- [step 4] sync_metrics_to_cmt_notes.py (补跑后再次落库) --- >> "%LOG_FILE%"
+call %PYTHON% data\tools\sync_metrics_to_cmt_notes.py >> "%LOG_FILE%" 2>&1
+set RC_SYNC_POST=%ERRORLEVEL%
+echo sync_post exit=%RC_SYNC_POST% >> "%LOG_FILE%"
+
+set FINAL_RC=0
+if not "%RC_RUN%"=="0" set FINAL_RC=1
+if not "%RC_SYNC_PRE%"=="0" set FINAL_RC=1
+if not "%RC_RETRY%"=="0" set FINAL_RC=1
+if not "%RC_SYNC_POST%"=="0" set FINAL_RC=1
+
+echo === Monday run finished at %date% %time% (run=%RC_RUN%, sync_pre=%RC_SYNC_PRE%, retry=%RC_RETRY%, sync_post=%RC_SYNC_POST%, final=%FINAL_RC%) === >> "%LOG_FILE%"
+
+endlocal & exit /b %FINAL_RC%
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/db.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/db.py
new file mode 100644
index 0000000..e8501a9
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/db.py
@@ -0,0 +1,515 @@
+"""
+db.py - PostgreSQL 连接池 + cmt_* 表 CRUD
+
+统一使用 psycopg3，合并了原 yingxiaoyunying 诊断功能 + comment-data-collector 的完整 CRUD。
+
+用法:
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.db import make_pool, get_conn, get_dict_conn
+    python db.py ping           # 测试连接
+    python db.py info           # 显示 cmt_* 表行数
+"""
+import os
+import sys
+from contextlib import contextmanager
+from datetime import datetime
+from pathlib import Path
+from typing import Any, Iterator, Mapping
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+from gyxx_flow.adapters import RuntimeServicePolicy
+
+try:
+    import psycopg
+    from psycopg.rows import dict_row
+    from psycopg_pool import ConnectionPool
+except ImportError:
+    print("[FATAL] 请先 pip install 'psycopg[binary]>=3.1' psycopg-pool", file=sys.stderr)
+    sys.exit(1)
+
+try:
+    from dotenv import load_dotenv
+except ImportError:
+    print("[FATAL] 请先 pip install python-dotenv", file=sys.stderr)
+    sys.exit(1)
+
+
+BASE_DIR = PATHS.tools_root
+DATA_DIR = PATHS.raw_root
+ENV_PATH = PATHS.state_root / "config/db.env"
+ENV_EXAMPLE = PATHS.config_root / "db.env.example"
+
+
+def load_env() -> None:
+    if not ENV_PATH.exists():
+        print(f"[FATAL] 找不到 {ENV_PATH}", file=sys.stderr)
+        print(f"  请复制 {ENV_EXAMPLE} -> {ENV_PATH} 并填连接信息", file=sys.stderr)
+        sys.exit(1)
+    load_dotenv(ENV_PATH, override=False)
+    os.environ.update(RuntimeServicePolicy().apply(os.environ))
+
+
+def database_config_from_env(environ: Mapping[str, str]) -> dict[str, Any]:
+    """从环境变量构造连接配置；缺项立即失败，绝不回退到本地库。"""
+    required = ("PG_HOST", "PG_PORT", "PG_DB", "PG_USER", "PG_PASSWORD")
+    missing = [key for key in required if not str(environ.get(key, "")).strip()]
+    if missing:
+        raise RuntimeError(f"数据库配置缺失: {', '.join(missing)}")
+    try:
+        port = int(environ["PG_PORT"])
+    except (TypeError, ValueError) as exc:
+        raise RuntimeError("数据库配置无效: PG_PORT 必须是整数") from exc
+    return {
+        "host": environ["PG_HOST"].strip(),
+        "port": port,
+        "dbname": environ["PG_DB"].strip(),
+        "user": environ["PG_USER"].strip(),
+        "password": environ["PG_PASSWORD"],
+    }
+
+
+def get_db_config() -> dict[str, Any]:
+    load_env()
+    return database_config_from_env(os.environ)
+
+
+def make_pool(min_size: int = 1, max_size: int = 5) -> ConnectionPool:
+    conninfo = psycopg.conninfo.make_conninfo(**get_db_config())
+    return ConnectionPool(conninfo, min_size=min_size, max_size=max_size, open=True)
+
+
+# 全局连接池（懒初始化）
+_pool: ConnectionPool | None = None
+
+
+def _get_pool() -> ConnectionPool:
+    global _pool
+    if _pool is None or _pool.closed:
+        _pool = make_pool()
+    return _pool
+
+
+@contextmanager
+def get_conn(pool: ConnectionPool | None = None) -> Iterator[psycopg.Connection]:
+    p = pool or _get_pool()
+    with p.connection() as conn:
+        yield conn
+
+
+@contextmanager
+def get_dict_conn(pool: ConnectionPool | None = None) -> Iterator[psycopg.Connection]:
+    """返回连接，cursor 需自行指定 row_factory=dict_row。替代 psycopg2 的 RealDictCursor。"""
+    p = pool or _get_pool()
+    with p.connection() as conn:
+        yield conn
+
+
+@contextmanager
+def get_dict_cursor(pool: ConnectionPool | None = None):
+    """便捷函数：返回 dict_row cursor，自动 commit/rollback。"""
+    with get_dict_conn(pool) as conn:
+        cur = conn.cursor(row_factory=dict_row)
+        try:
+            yield cur
+            conn.commit()
+        except Exception:
+            conn.rollback()
+            raise
+        finally:
+            cur.close()
+
+
+def now() -> datetime:
+    return datetime.now()
+
+
+def show_info(pool: ConnectionPool) -> None:
+    """显示 cmt_* 表行数"""
+    with get_conn(pool) as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            for tbl in ("cmt_styles", "cmt_creators", "cmt_style_creators", "cmt_notes", "cmt_comments"):
+                cur.execute(f"SELECT COUNT(*) AS n FROM {tbl}")
+                n = cur.fetchone()["n"]
+                print(f"  {tbl:<25}  {n:>8} 行")
+
+
+# ---------------------------------------------------------------------------
+# Styles
+# ---------------------------------------------------------------------------
+def upsert_style(
+    name: str,
+    platform: str = "",
+    feishu_base_url: str = "",
+    feishu_base_token: str = "",
+    feishu_table_id: str = "",
+    feishu_view_id: str = "",
+) -> int:
+    sql = """
+    INSERT INTO cmt_styles (name, platform, feishu_base_url, feishu_base_token, feishu_table_id, feishu_view_id)
+    VALUES (%s, %s, %s, %s, %s, %s)
+    ON CONFLICT (name) DO UPDATE SET
+        platform = EXCLUDED.platform,
+        feishu_base_url = EXCLUDED.feishu_base_url,
+        feishu_base_token = EXCLUDED.feishu_base_token,
+        feishu_table_id = EXCLUDED.feishu_table_id,
+        feishu_view_id = EXCLUDED.feishu_view_id,
+        updated_at = CURRENT_TIMESTAMP
+    RETURNING id
+    """
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(sql, (name, platform, feishu_base_url, feishu_base_token, feishu_table_id, feishu_view_id))
+            return cur.fetchone()[0]
+
+
+def get_style_id(name: str) -> int | None:
+    with get_dict_conn() as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute("SELECT id FROM cmt_styles WHERE name = %s", (name,))
+            row = cur.fetchone()
+            return row["id"] if row else None
+
+
+# ---------------------------------------------------------------------------
+# Creators
+# ---------------------------------------------------------------------------
+def upsert_creator(name: str) -> int:
+    if not name:
+        raise ValueError("creator name cannot be empty")
+    sql = """
+    INSERT INTO cmt_creators (name)
+    VALUES (%s)
+    ON CONFLICT (name) DO UPDATE SET
+        updated_at = CURRENT_TIMESTAMP
+    RETURNING id
+    """
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(sql, (name,))
+            return cur.fetchone()[0]
+
+
+def link_style_creator(style_id: int, creator_id: int) -> None:
+    sql = """
+    INSERT INTO cmt_style_creators (style_id, creator_id)
+    VALUES (%s, %s)
+    ON CONFLICT (style_id, creator_id) DO NOTHING
+    """
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(sql, (style_id, creator_id))
+            conn.commit()
+
+
+# ---------------------------------------------------------------------------
+# Notes
+# ---------------------------------------------------------------------------
+def upsert_note(
+    style_id: int,
+    creator_id: int,
+    platform: str,
+    url: str,
+    title: str = "",
+    feishu_record_id: str = "",
+) -> int:
+    sql = """
+    INSERT INTO cmt_notes (style_id, creator_id, platform, title, url, feishu_record_id)
+    VALUES (%s, %s, %s, %s, %s, %s)
+    ON CONFLICT (url) DO UPDATE SET
+        style_id = EXCLUDED.style_id,
+        creator_id = EXCLUDED.creator_id,
+        platform = EXCLUDED.platform,
+        title = EXCLUDED.title,
+        feishu_record_id = EXCLUDED.feishu_record_id,
+        updated_at = CURRENT_TIMESTAMP
+    RETURNING id
+    """
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(sql, (style_id, creator_id, platform, title, url, feishu_record_id))
+            return cur.fetchone()[0]
+
+
+def update_note_metrics(
+    note_id: int,
+    like_count: int | None,
+    favorite_count: int | None,
+    comment_count: int | None,
+    share_count: int | None,
+    title: str | None = None,
+) -> None:
+    sql = """
+    UPDATE cmt_notes
+    SET like_count = %s,
+        favorite_count = %s,
+        comment_count = %s,
+        share_count = %s,
+        title = COALESCE(%s, title),
+        scraped_at = CURRENT_TIMESTAMP,
+        updated_at = CURRENT_TIMESTAMP
+    WHERE id = %s
+    """
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(sql, (like_count, favorite_count, comment_count, share_count, title, note_id))
+            conn.commit()
+
+
+def update_note_view_metrics(
+    note_id: int,
+    view_count: int | None,
+    collect_count: int | None,
+) -> None:
+    """更新单笔记的曝光量/播放量、收藏量(来自 yingxiaoyunying 每日采集)。"""
+    sql = """
+    UPDATE cmt_notes
+    SET view_count = %s,
+        collect_count = %s,
+        updated_at = CURRENT_TIMESTAMP
+    WHERE id = %s
+    """
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(sql, (view_count, collect_count, note_id))
+            conn.commit()
+
+
+def get_note_id_by_url(url: str) -> int | None:
+    with get_dict_conn() as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute("SELECT id FROM cmt_notes WHERE url = %s", (url,))
+            row = cur.fetchone()
+            return row["id"] if row else None
+
+
+def get_note_by_id(note_id: int) -> dict[str, Any] | None:
+    with get_dict_conn() as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(
+                """
+                SELECT n.*, s.name AS style_name, c.name AS creator_name
+                FROM cmt_notes n
+                LEFT JOIN cmt_styles s ON n.style_id = s.id
+                LEFT JOIN cmt_creators c ON n.creator_id = c.id
+                WHERE n.id = %s
+                """,
+                (note_id,),
+            )
+            return cur.fetchone()
+
+
+def get_note_by_url(url: str) -> dict[str, Any] | None:
+    with get_dict_conn() as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(
+                """
+                SELECT n.*, s.name AS style_name, c.name AS creator_name
+                FROM cmt_notes n
+                LEFT JOIN cmt_styles s ON n.style_id = s.id
+                LEFT JOIN cmt_creators c ON n.creator_id = c.id
+                WHERE n.url = %s
+                """,
+                (url,),
+            )
+            return cur.fetchone()
+
+
+def get_notes_by_platform(
+    platform: str,
+    only_with_comments: bool = True,
+) -> list[dict[str, Any]]:
+    """List all notes for a platform with style/creator joins + comment count."""
+    sql = """
+        SELECT n.*, s.name AS style_name, c.name AS creator_name,
+               (SELECT COUNT(*) FROM cmt_comments WHERE note_id = n.id) AS comment_total
+        FROM cmt_notes n
+        LEFT JOIN cmt_styles s ON n.style_id = s.id
+        LEFT JOIN cmt_creators c ON n.creator_id = c.id
+        WHERE n.platform = %s
+    """
+    if only_with_comments:
+        sql += " AND (SELECT COUNT(*) FROM cmt_comments WHERE note_id = n.id) > 0"
+    sql += " ORDER BY n.id"
+    with get_dict_conn() as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(sql, (platform,))
+            return cur.fetchall()
+
+
+def get_all_styles_with_notes() -> list[dict[str, Any]]:
+    """List all styles that have at least one note, with per-style rollup counts."""
+    sql = """
+        SELECT s.*,
+               COUNT(n.id) AS note_total,
+               COUNT(*) FILTER (WHERE EXISTS(SELECT 1 FROM cmt_comments WHERE note_id = n.id)) AS note_with_comments
+        FROM cmt_styles s
+        LEFT JOIN cmt_notes n ON n.style_id = s.id
+        GROUP BY s.id
+        HAVING COUNT(n.id) > 0
+        ORDER BY note_total DESC, s.name ASC
+    """
+    with get_dict_conn() as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(sql)
+            return cur.fetchall()
+
+
+def get_notes_by_style(
+    style_id: int,
+    only_with_comments: bool = True,
+) -> list[dict[str, Any]]:
+    """List all notes for a single style, same field set as get_notes_by_platform."""
+    sql = """
+        SELECT n.*, s.name AS style_name, c.name AS creator_name,
+               (SELECT COUNT(*) FROM cmt_comments WHERE note_id = n.id) AS comment_total
+        FROM cmt_notes n
+        LEFT JOIN cmt_styles s ON n.style_id = s.id
+        LEFT JOIN cmt_creators c ON n.creator_id = c.id
+        WHERE n.style_id = %s
+    """
+    if only_with_comments:
+        sql += " AND (SELECT COUNT(*) FROM cmt_comments WHERE note_id = n.id) > 0"
+    sql += " ORDER BY n.id ASC"
+    with get_dict_conn() as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(sql, (style_id,))
+            return cur.fetchall()
+
+
+def get_comments_by_note_id(note_id: int) -> list[dict[str, Any]]:
+    with get_dict_conn() as conn:
+        with conn.cursor(row_factory=dict_row) as cur:
+            cur.execute(
+                """
+                SELECT * FROM cmt_comments
+                WHERE note_id = %s
+                ORDER BY comment_created_at ASC, id ASC
+                """,
+                (note_id,),
+            )
+            return cur.fetchall()
+
+
+# ---------------------------------------------------------------------------
+# Comments
+# ---------------------------------------------------------------------------
+def replace_comments(note_id: int, platform: str, comments: list[dict[str, Any]], allow_empty: bool = False) -> int:
+    """Delete existing comments for the note and insert the latest batch."""
+    if not comments and not allow_empty:
+        raise ValueError("refusing to replace existing comments with an empty scrape result")
+
+    scraped_at = now()
+
+    insert_sql = """
+    INSERT INTO cmt_comments (
+        note_id, platform, level, platform_comment_id, parent_comment_id,
+        reply_to_comment_id, user_id, nickname, content, like_count,
+        reply_count, ip_location, comment_created_at, raw_create_time, scraped_at
+    ) VALUES (%s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s)
+    ON CONFLICT (note_id, platform_comment_id) DO UPDATE SET
+        level = EXCLUDED.level,
+        parent_comment_id = EXCLUDED.parent_comment_id,
+        reply_to_comment_id = EXCLUDED.reply_to_comment_id,
+        user_id = EXCLUDED.user_id,
+        nickname = EXCLUDED.nickname,
+        content = EXCLUDED.content,
+        like_count = EXCLUDED.like_count,
+        reply_count = EXCLUDED.reply_count,
+        ip_location = EXCLUDED.ip_location,
+        comment_created_at = EXCLUDED.comment_created_at,
+        raw_create_time = EXCLUDED.raw_create_time,
+        scraped_at = EXCLUDED.scraped_at,
+        updated_at = CURRENT_TIMESTAMP
+    """
+
+    def to_int(value: Any) -> int | None:
+        try:
+            if value is None or value == "":
+                return None
+            return int(float(value))
+        except (ValueError, TypeError):
+            return None
+
+    def parse_dt(value: Any) -> datetime | None:
+        if isinstance(value, datetime):
+            return value
+        if not value:
+            return None
+        if isinstance(value, (int, float)):
+            ts = value / 1000 if value > 10_000_000_000 else value
+            try:
+                return datetime.fromtimestamp(ts)
+            except Exception:
+                return None
+        if isinstance(value, str):
+            for fmt in ("%Y-%m-%dT%H:%M:%S", "%Y-%m-%d %H:%M:%S", "%Y-%m-%dT%H:%M:%S%z"):
+                try:
+                    return datetime.strptime(value[:19], fmt)
+                except ValueError:
+                    continue
+        return None
+
+    rows = []
+    for c in comments:
+        platform_comment_id = str(c.get("comment_id") or c.get("cid") or c.get("rpid") or "")
+        parent = str(c.get("parent_comment_id") or c.get("parent_rpid") or "")
+        reply_to = str(c.get("reply_to_comment_id") or c.get("reply_id") or c.get("reply_to_reply_id") or "")
+        if not reply_to and parent:
+            reply_to = parent
+        rows.append((
+            note_id,
+            platform,
+            c.get("level") or "comment",
+            platform_comment_id,
+            parent,
+            reply_to,
+            str(c.get("user_id") or c.get("mid") or c.get("sec_uid") or ""),
+            str(c.get("nickname") or c.get("uname") or ""),
+            str(c.get("content") or c.get("text") or c.get("message") or ""),
+            to_int(c.get("like_count") or c.get("digg_count") or 0),
+            to_int(c.get("reply_count") or 0),
+            str(c.get("ip_location") or c.get("ip_label") or c.get("location") or ""),
+            parse_dt(c.get("created_at") or c.get("raw_create_time")),
+            str(c.get("raw_create_time") or ""),
+            scraped_at,
+        ))
+
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute("DELETE FROM cmt_comments WHERE note_id = %s", (note_id,))
+            if rows:
+                # Use executemany for bulk insert (psycopg3 native)
+                cur.executemany(insert_sql, rows)
+            conn.commit()
+    return len(rows)
+
+
+def test_connection() -> bool:
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute("SELECT version()")
+            print(cur.fetchone()[0])
+    return True
+
+
+def cli():
+    if len(sys.argv) < 2:
+        print("用法: python db.py [ping|info]")
+        sys.exit(1)
+    cmd = sys.argv[1]
+    pool = make_pool()
+    try:
+        if cmd == "ping":
+            with get_conn(pool) as conn:
+                with conn.cursor() as cur:
+                    cur.execute("SELECT version()")
+                    print(cur.fetchone()[0])
+        elif cmd == "info":
+            show_info(pool)
+        else:
+            print(f"未知命令: {cmd}")
+            sys.exit(1)
+    finally:
+        pool.close()
+
+
+if __name__ == "__main__":
+    cli()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/feishu_comment_batch.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/feishu_comment_batch.py
new file mode 100644
index 0000000..92cc45c
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/feishu_comment_batch.py
@@ -0,0 +1,716 @@
+import argparse
+import csv
+import json
+import os
+import re
+import shutil
+import subprocess
+import sys
+import time
+from dataclasses import dataclass
+from datetime import datetime
+from multiprocessing import Pool
+from pathlib import Path
+from typing import Any
+from urllib.parse import parse_qs, urlparse
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+# Path setup: db.py is in data/tools/, scrapers are in project root
+_TOOLS_DIR = PATHS.tools_root
+_PROJECT_ROOT = PATHS.module_root
+
+from gyxx_flow.modules.content_marketing.runtime import bilibili_comment_scraper
+from gyxx_flow.modules.content_marketing.runtime.data.tools import db
+from gyxx_flow.modules.content_marketing.runtime import douyin_comment_scraper
+from gyxx_flow.modules.content_marketing.runtime import xiaohongshu_comment_scraper
+
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+SUMMARY_DIR = PATHS.exports_root / "summary"
+
+DEFAULT_INDEX_BASE_TOKEN = "TtoCb1NuQaDy3NsZWTpc0GIvnph"
+DEFAULT_INDEX_TABLE_ID = "tblKCjplVAFrRwMC"
+DEFAULT_INDEX_VIEW_ID = "vewvy33xEk"
+
+STYLE_NAME_FIELD = "款式"
+STYLE_BASE_FIELD = "合作达人多维表格地址"
+INDEX_PLATFORM_FIELD = "平台"
+NOTE_URL_FIELD = "发布笔记链接"
+NOTE_PLATFORM_FIELD = "投放平台"
+CREATOR_FIELD = "达人名称"
+TITLE_FIELD = "发布笔记标题"
+
+
+@dataclass(frozen=True)
+class BaseRef:
+    url: str
+    base_token: str
+    table_id: str
+    view_id: str
+
+
+@dataclass(frozen=True)
+class StyleTask:
+    record_id: str
+    style_name: str
+    platform: str
+    base_ref: BaseRef
+
+
+@dataclass(frozen=True)
+class NoteTask:
+    style_name: str
+    style_record_id: str
+    note_record_id: str
+    platform_hint: str
+    creator_name: str
+    title: str
+    note_url: str
+    source_base_url: str
+
+
+def log(message: str) -> None:
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {message}", flush=True)
+
+
+def run_lark(args: list[str]) -> dict[str, Any]:
+    lark_cli = shutil.which("lark-cli") or shutil.which("lark-cli.cmd") or "lark-cli"
+    command = [lark_cli, "base", *args, "--format", "json", "--as", "user"]
+    completed = subprocess.run(
+        command,
+        cwd=BASE_DIR,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+        capture_output=True,
+    )
+    if completed.returncode != 0:
+        raise RuntimeError(
+            f"lark-cli failed ({completed.returncode}): {' '.join(command)}\n"
+            f"stdout:\n{completed.stdout}\n\nstderr:\n{completed.stderr}"
+        )
+    try:
+        return json.loads(completed.stdout)
+    except json.JSONDecodeError as exc:
+        raise RuntimeError(f"Failed to parse lark-cli JSON output: {exc}\n{completed.stdout}") from exc
+
+
+def list_records(base_token: str, table_id: str, view_id: str = "", limit: int = 200) -> list[dict[str, Any]]:
+    offset = 0
+    records: list[dict[str, Any]] = []
+    while True:
+        args = [
+            "+record-list",
+            "--base-token",
+            base_token,
+            "--table-id",
+            table_id,
+            "--limit",
+            str(limit),
+            "--offset",
+            str(offset),
+        ]
+        if view_id:
+            args.extend(["--view-id", view_id])
+
+        payload = run_lark(args)
+        if not payload.get("ok"):
+            raise RuntimeError(f"lark-cli returned ok=false: {payload}")
+
+        data = payload.get("data") or {}
+        field_names = data.get("fields") or []
+        record_ids = data.get("record_id_list") or []
+        rows = data.get("data") or []
+
+        for index, row in enumerate(rows):
+            record = dict(zip(field_names, row))
+            if index < len(record_ids):
+                record["_record_id"] = record_ids[index]
+            records.append(record)
+
+        if not data.get("has_more") or not rows:
+            return records
+        offset += len(rows)
+
+
+def first_text(value: Any) -> str:
+    if value is None:
+        return ""
+    if isinstance(value, str):
+        return value.strip()
+    if isinstance(value, (int, float)):
+        return str(value)
+    if isinstance(value, list):
+        parts = [first_text(item) for item in value]
+        return ",".join(part for part in parts if part)
+    if isinstance(value, dict):
+        for key in ("text", "name", "link", "url"):
+            text = first_text(value.get(key))
+            if text:
+                return text
+    return str(value).strip()
+
+
+def extract_urls(value: Any) -> list[str]:
+    text = first_text(value)
+    if not text:
+        return []
+    markdown_urls = re.findall(r"\]\((https?://[^)\s]+)\)", text)
+    plain_urls = re.findall(r"https?://[^\s)\]>]+", text)
+    urls = markdown_urls + plain_urls
+    seen: set[str] = set()
+    unique: list[str] = []
+    for url in urls:
+        url = url.replace("&amp;", "&").strip()
+        if url not in seen:
+            seen.add(url)
+            unique.append(url)
+    return unique
+
+
+def normalize_note_url(url: str) -> str:
+    url = url.replace("&amp;", "&").strip()
+    for marker in ("复制", "打开", "，", "。", " "):
+        if marker in url:
+            url = url.split(marker, 1)[0]
+
+    parsed = urlparse(url)
+    host = parsed.netloc.lower()
+    query = parse_qs(parsed.query)
+    modal_id = (query.get("modal_id") or [""])[0]
+    if "douyin.com" in host and modal_id.isdigit():
+        return f"https://www.douyin.com/video/{modal_id}"
+
+    path_parts = [part for part in parsed.path.split("/") if part]
+    if host == "v.douyin.com" and path_parts:
+        return f"{parsed.scheme}://{parsed.netloc}/{path_parts[0]}/"
+    if host in {"xhslink.com", "www.xhslink.com"} and len(path_parts) >= 2:
+        return f"{parsed.scheme}://{parsed.netloc}/{path_parts[0]}/{path_parts[1]}"
+    if host == "b23.tv" and path_parts:
+        return f"{parsed.scheme}://{parsed.netloc}/{path_parts[0]}"
+    return url
+
+
+def is_supported_note_url(url: str) -> bool:
+    host = urlparse(url).netloc.lower()
+    return any(
+        domain in host
+        for domain in (
+            "douyin.com",
+            "xiaohongshu.com",
+            "xhslink.com",
+            "bilibili.com",
+            "b23.tv",
+        )
+    )
+
+
+def parse_base_ref(url: str) -> BaseRef | None:
+    parsed = urlparse(url)
+    parts = [part for part in parsed.path.split("/") if part]
+    if len(parts) < 2 or parts[0] != "base":
+        return None
+    query = parse_qs(parsed.query)
+    table_id = (query.get("table") or [""])[0]
+    view_id = (query.get("view") or [""])[0]
+    if not table_id:
+        return None
+    return BaseRef(url=url, base_token=parts[1], table_id=table_id, view_id=view_id)
+
+
+def detect_platform(url: str, platform_hint: str = "") -> str:
+    lowered = url.lower()
+    if "xiaohongshu.com" in lowered or "xhslink.com" in lowered:
+        return "xiaohongshu"
+    if "douyin.com" in lowered:
+        return "douyin"
+    if "bilibili.com" in lowered or "b23.tv" in lowered:
+        return "bilibili"
+
+    return "unknown"
+
+
+def load_style_tasks(
+    index_base_token: str,
+    index_table_id: str,
+    index_view_id: str,
+    index_platform: str,
+    max_styles: int | None,
+) -> list[StyleTask]:
+    records = list_records(index_base_token, index_table_id, index_view_id)
+    tasks: list[StyleTask] = []
+    for record in records:
+        platform = first_text(record.get(INDEX_PLATFORM_FIELD))
+        if index_platform and platform != index_platform:
+            continue
+        style_name = first_text(record.get(STYLE_NAME_FIELD))
+        urls = extract_urls(record.get(STYLE_BASE_FIELD))
+        if not urls:
+            continue
+        base_ref = parse_base_ref(urls[0])
+        if not base_ref:
+            log(f"Skip style without valid Base link: {style_name or record.get('_record_id')}")
+            continue
+        tasks.append(
+            StyleTask(
+                record_id=first_text(record.get("_record_id")),
+                style_name=style_name or first_text(record.get("_record_id")) or "unknown_style",
+                platform=platform,
+                base_ref=base_ref,
+            )
+        )
+        if max_styles and len(tasks) >= max_styles:
+            break
+    return tasks
+
+
+def load_note_tasks(style_tasks: list[StyleTask], max_notes_per_style: int | None) -> list[NoteTask]:
+    note_tasks: list[NoteTask] = []
+    for style in style_tasks:
+        log(f"Read notes: {style.style_name}")
+        records = list_records(style.base_ref.base_token, style.base_ref.table_id, style.base_ref.view_id)
+        count_for_style = 0
+        for record in records:
+            urls = extract_urls(record.get(NOTE_URL_FIELD))
+            if not urls:
+                continue
+            platform_hint = first_text(record.get(NOTE_PLATFORM_FIELD))
+            for url in urls:
+                url = normalize_note_url(url)
+                if not is_supported_note_url(url):
+                    log(f"Skip unsupported URL: {style.style_name} | {url}")
+                    continue
+                note_tasks.append(
+                    NoteTask(
+                        style_name=style.style_name,
+                        style_record_id=style.record_id,
+                        note_record_id=first_text(record.get("_record_id")),
+                        platform_hint=platform_hint,
+                        creator_name=first_text(record.get(CREATOR_FIELD)),
+                        title=first_text(record.get(TITLE_FIELD)),
+                        note_url=url,
+                        source_base_url=style.base_ref.url,
+                    )
+                )
+                count_for_style += 1
+                if max_notes_per_style and count_for_style >= max_notes_per_style:
+                    break
+            if max_notes_per_style and count_for_style >= max_notes_per_style:
+                break
+        log(f"Found {count_for_style} note links: {style.style_name}")
+    return note_tasks
+
+
+def normalize_comment_rows(
+    task: NoteTask,
+    platform: str,
+    comments: list[dict[str, Any]],
+    note_metrics: dict[str, Any],
+) -> list[dict[str, Any]]:
+    rows: list[dict[str, Any]] = []
+    for comment in comments:
+        comment_text = first_text(comment.get("text") or comment.get("content") or comment.get("message"))
+
+        rows.append(
+            {
+                "style_name": task.style_name,
+                "creator_name": task.creator_name,
+                "note_title": task.title,
+                "platform": platform,
+                "note_url": task.note_url,
+                "comment_text": comment_text,
+                "comment_count": note_metrics.get("comment_count", ""),
+                "like_count": note_metrics.get("like_count", ""),
+                "favorite_count": note_metrics.get("favorite_count", ""),
+                "share_count": note_metrics.get("share_count", ""),
+            }
+        )
+    return rows
+
+
+# ---------------------------------------------------------------------------
+# Database sync helpers
+# ---------------------------------------------------------------------------
+def sync_styles_to_db(style_tasks: list[StyleTask]) -> dict[str, int]:
+    """Upsert styles and return {style_name: style_id}."""
+    style_ids: dict[str, int] = {}
+    for style in style_tasks:
+        style_id = db.upsert_style(
+            name=style.style_name,
+            platform=style.platform,
+            feishu_base_url=style.base_ref.url,
+            feishu_base_token=style.base_ref.base_token,
+            feishu_table_id=style.base_ref.table_id,
+            feishu_view_id=style.base_ref.view_id,
+        )
+        style_ids[style.style_name] = style_id
+    log(f"Synced {len(style_ids)} styles to database")
+    return style_ids
+
+
+def sync_notes_to_db(note_tasks: list[NoteTask], style_ids: dict[str, int]) -> dict[str, int]:
+    """Upsert creators and notes, link styles and creators. Return {url: note_id}."""
+    creator_ids: dict[str, int] = {}
+    note_ids: dict[str, int] = {}
+    for task in note_tasks:
+        style_id = style_ids.get(task.style_name)
+        if not style_id:
+            log(f"Skip DB sync for note without known style: {task.note_url}")
+            continue
+
+        creator_name = task.creator_name or "unknown_creator"
+        if creator_name not in creator_ids:
+            creator_ids[creator_name] = db.upsert_creator(creator_name)
+        creator_id = creator_ids[creator_name]
+        db.link_style_creator(style_id, creator_id)
+
+        platform = detect_platform(task.note_url, task.platform_hint)
+        note_id = db.upsert_note(
+            style_id=style_id,
+            creator_id=creator_id,
+            platform=platform,
+            url=task.note_url,
+            title=task.title,
+            feishu_record_id=task.note_record_id,
+        )
+        note_ids[task.note_url] = note_id
+    log(f"Synced {len(note_ids)} notes to database")
+    return note_ids
+
+
+def save_scrape_to_db(
+    task: NoteTask,
+    platform: str,
+    comments: list[dict[str, Any]],
+    note_metrics: dict[str, Any],
+    note_ids: dict[str, int],
+) -> int:
+    """Update note metrics and replace comments in DB."""
+    note_id = note_ids.get(task.note_url)
+    if not note_id:
+        return 0
+    db.update_note_metrics(
+        note_id=note_id,
+        like_count=_to_int(note_metrics.get("like_count")),
+        favorite_count=_to_int(note_metrics.get("favorite_count")),
+        comment_count=_to_int(note_metrics.get("comment_count")),
+        share_count=_to_int(note_metrics.get("share_count")),
+    )
+    allow_empty = _to_int(note_metrics.get("comment_count")) == 0
+    if not comments and not allow_empty:
+        log(f"Skip replacing comments for {task.note_url}: empty scrape result without confirmed zero comments")
+        return 0
+    inserted = db.replace_comments(note_id, platform, comments, allow_empty=allow_empty)
+    return inserted
+
+
+def _to_int(value: Any) -> int | None:
+    try:
+        if value is None or value == "":
+            return None
+        return int(float(value))
+    except (ValueError, TypeError):
+        return None
+
+
+def scrape_note(
+    task: NoteTask,
+    args: argparse.Namespace,
+    note_ids: dict[str, int] | None = None,
+) -> tuple[dict[str, Any], list[dict[str, Any]]]:
+    platform = detect_platform(task.note_url, task.platform_hint)
+    started_at = datetime.now().isoformat(timespec="seconds")
+    result: dict[str, Any] = {
+        "started_at": started_at,
+        "style_name": task.style_name,
+        "style_record_id": task.style_record_id,
+        "note_record_id": task.note_record_id,
+        "platform_hint": task.platform_hint,
+        "platform": platform,
+        "creator_name": task.creator_name,
+        "title": task.title,
+        "note_url": task.note_url,
+        "source_base_url": task.source_base_url,
+        "status": "pending",
+        "comment_count": 0,
+        "json_path": "",
+        "csv_path": "",
+        "error": "",
+    }
+
+    if platform == "unknown":
+        result["status"] = "skipped"
+        result["error"] = "Unsupported note URL"
+        return result, []
+
+    if args.dry_run:
+        result["status"] = "dry_run"
+        return result, []
+
+    try:
+        if platform == "xiaohongshu":
+            comments, result = xiaohongshu_comment_scraper.scrape_comments(
+                task.note_url,
+                args.login_timeout,
+                args.max_scrolls,
+                args.idle_rounds,
+                args.headless,
+            )
+            json_path, csv_path = xiaohongshu_comment_scraper.write_outputs(task.note_url, comments, result)
+            note_metrics = (result.get("stats") or {}).get("note_metrics", {})
+        elif platform == "douyin":
+            comments, result = douyin_comment_scraper.scrape_comments(
+                task.note_url,
+                args.login_timeout,
+                args.max_scrolls,
+                args.idle_rounds,
+                args.headless,
+            )
+            json_path, csv_path = douyin_comment_scraper.write_outputs(task.note_url, comments, result)
+            note_metrics = (result.get("stats") or {}).get("note_metrics", {})
+        else:
+            comments, result = bilibili_comment_scraper.scrape_comments(
+                task.note_url,
+                args.bilibili_max_pages,
+                args.bilibili_delay,
+                args.headless,
+                args.bilibili_max_reply_pages,
+            )
+            json_path, csv_path = bilibili_comment_scraper.write_outputs(task.note_url, comments, result)
+            note_metrics = (result.get("stats") or {}).get("note_metrics", {})
+
+        result["status"] = "ok"
+        result["comment_count"] = len(comments)
+        result["json_path"] = str(json_path)
+        result["csv_path"] = str(csv_path)
+        comment_rows = normalize_comment_rows(task, platform, comments, note_metrics)
+
+        if not args.no_db and note_ids:
+            try:
+                db_inserted = save_scrape_to_db(task, platform, comments, note_metrics, note_ids)
+                log(f"Saved {db_inserted} comments to DB for {task.note_url}")
+            except Exception as exc:
+                log(f"Failed to save comments to DB: {exc}")
+    except Exception as exc:
+        result["status"] = "failed"
+        result["error"] = repr(exc)
+        comment_rows = []
+
+    result["finished_at"] = datetime.now().isoformat(timespec="seconds")
+    return result, comment_rows
+
+
+def scrape_notes_for_platform(
+    platform: str,
+    tasks: list[NoteTask],
+    args_dict: dict[str, Any],
+    note_ids: dict[str, int] | None,
+) -> tuple[list[dict[str, Any]], list[dict[str, Any]]]:
+    """Scrape all notes for one platform in a separate process/browser."""
+    results: list[dict[str, Any]] = []
+    comment_rows: list[dict[str, Any]] = []
+
+    class SimpleArgs:
+        def __init__(self, data: dict[str, Any]):
+            self.__dict__.update(data)
+
+    platform_args = SimpleArgs(args_dict)
+
+    for index, task in enumerate(tasks, 1):
+        log(f"[{platform} {index}/{len(tasks)}] {task.style_name} | {task.note_url}")
+        result, rows = scrape_note(task, platform_args, note_ids)
+        results.append(result)
+        comment_rows.extend(rows)
+        if platform_args.between_notes_delay > 0 and index < len(tasks):
+            time.sleep(platform_args.between_notes_delay)
+
+    return results, comment_rows
+
+
+def current_week_label() -> str:
+    """Return a label like '2026_06_week4' for the current date."""
+    now = datetime.now()
+    week_of_month = (now.day - 1) // 7 + 1
+    return f"{now.year}_{now.month:02d}_week{week_of_month}"
+
+
+def write_manifest(rows: list[dict[str, Any]]) -> tuple[Path, Path]:
+    SUMMARY_DIR.mkdir(exist_ok=True)
+    label = current_week_label()
+    json_path = SUMMARY_DIR / f"feishu_comment_batch_{label}.json"
+    csv_path = SUMMARY_DIR / f"feishu_comment_batch_{label}.csv"
+    json_path.write_text(json.dumps(rows, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    fieldnames = [
+        "started_at",
+        "finished_at",
+        "style_name",
+        "style_record_id",
+        "note_record_id",
+        "platform_hint",
+        "platform",
+        "creator_name",
+        "title",
+        "note_url",
+        "source_base_url",
+        "status",
+        "comment_count",
+        "json_path",
+        "csv_path",
+        "error",
+    ]
+    with csv_path.open("w", newline="", encoding="utf-8-sig") as file:
+        writer = csv.DictWriter(file, fieldnames=fieldnames)
+        writer.writeheader()
+        writer.writerows(rows)
+    return json_path, csv_path
+
+
+def write_comment_summary(rows: list[dict[str, Any]]) -> tuple[Path, Path]:
+    SUMMARY_DIR.mkdir(exist_ok=True)
+    label = current_week_label()
+    json_path = SUMMARY_DIR / f"feishu_comments_{label}.json"
+    csv_path = SUMMARY_DIR / f"feishu_comments_{label}.csv"
+
+    note_counts: dict[tuple[str, str, str, str, str], int] = {}
+    for row in rows:
+        key = (
+            first_text(row.get("style_name")),
+            first_text(row.get("creator_name")),
+            first_text(row.get("note_title")),
+            first_text(row.get("platform")),
+            first_text(row.get("note_url")),
+        )
+        note_counts[key] = note_counts.get(key, 0) + 1
+    for row in rows:
+        key = (
+            first_text(row.get("style_name")),
+            first_text(row.get("creator_name")),
+            first_text(row.get("note_title")),
+            first_text(row.get("platform")),
+            first_text(row.get("note_url")),
+        )
+        if row.get("comment_count") in ("", None):
+            row["comment_count"] = note_counts.get(key, 0)
+
+    json_path.write_text(json.dumps(rows, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    columns = [
+        ("style_name", "\u6b3e\u5f0f"),
+        ("creator_name", "\u535a\u4e3b\u540d"),
+        ("note_title", "\u7b14\u8bb0\u540d"),
+        ("platform", "\u5e73\u53f0"),
+        ("note_url", "\u7b14\u8bb0\u94fe\u63a5"),
+        ("comment_text", "\u8bc4\u8bba\u5185\u5bb9"),
+        ("comment_count", "\u8bc4\u8bba\u6570"),
+        ("like_count", "\u70b9\u8d5e\u6570"),
+        ("favorite_count", "\u6536\u85cf\u6570"),
+        ("share_count", "\u5206\u4eab\u91cf"),
+    ]
+    with csv_path.open("w", newline="", encoding="utf-8-sig") as file:
+        writer = csv.DictWriter(file, fieldnames=[label for _, label in columns])
+        writer.writeheader()
+        for row in rows:
+            writer.writerow({label: row.get(key, "") for key, label in columns})
+    return json_path, csv_path
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Read Feishu Base note links and batch scrape comments.")
+    parser.add_argument("--index-base-token", default=DEFAULT_INDEX_BASE_TOKEN)
+    parser.add_argument("--index-table-id", default=DEFAULT_INDEX_TABLE_ID)
+    parser.add_argument("--index-view-id", default=DEFAULT_INDEX_VIEW_ID)
+    parser.add_argument("--index-platform", default="天猫", help="only read this platform group from the style index; use empty string for all")
+    parser.add_argument("--max-styles", type=int, default=None, help="limit styles for testing")
+    parser.add_argument("--max-notes-per-style", type=int, default=None, help="limit note links per style for testing")
+    parser.add_argument("--platform", choices=["all", "xiaohongshu", "douyin", "bilibili"], default="all")
+    parser.add_argument("--dry-run", action="store_true", help="only read Feishu and list tasks; do not scrape comments")
+    parser.add_argument("--headless", action="store_true")
+    parser.add_argument("--login-timeout", type=int, default=300)
+    parser.add_argument("--max-scrolls", type=int, default=120)
+    parser.add_argument("--idle-rounds", type=int, default=8)
+    parser.add_argument("--bilibili-max-pages", type=int, default=200)
+    parser.add_argument("--bilibili-max-reply-pages", type=int, default=200)
+    parser.add_argument("--bilibili-delay", type=float, default=0.25)
+    parser.add_argument("--between-notes-delay", type=float, default=1.0)
+    parser.add_argument("--no-db", action="store_true", help="do not write to PostgreSQL database")
+    parser.add_argument("--parallel", action="store_true", default=True, help="run one browser per platform in parallel (default)")
+    parser.add_argument("--no-parallel", action="store_true", help="disable parallel scraping")
+    args = parser.parse_args()
+
+    style_tasks = load_style_tasks(
+        args.index_base_token,
+        args.index_table_id,
+        args.index_view_id,
+        args.index_platform,
+        args.max_styles,
+    )
+    log(f"Found {len(style_tasks)} style Base links")
+    note_tasks = load_note_tasks(style_tasks, args.max_notes_per_style)
+    if args.platform != "all":
+        note_tasks = [task for task in note_tasks if detect_platform(task.note_url, task.platform_hint) == args.platform]
+    log(f"Prepared {len(note_tasks)} note scrape tasks")
+
+    style_ids: dict[str, int] = {}
+    note_ids: dict[str, int] = {}
+    if not args.no_db:
+        style_ids = sync_styles_to_db(style_tasks)
+        note_ids = sync_notes_to_db(note_tasks, style_ids)
+
+    results: list[dict[str, Any]] = []
+    comment_rows: list[dict[str, Any]] = []
+
+    parallel = args.parallel and not args.no_parallel and len(note_tasks) > 1
+    if parallel:
+        platform_groups: dict[str, list[NoteTask]] = {}
+        for task in note_tasks:
+            platform = detect_platform(task.note_url, task.platform_hint)
+            platform_groups.setdefault(platform, []).append(task)
+
+        active_groups = {p: t for p, t in platform_groups.items() if p in ("xiaohongshu", "douyin", "bilibili")}
+        if len(active_groups) > 1:
+            log(f"Running parallel scraping for platforms: {list(active_groups.keys())}")
+            args_dict = vars(args)
+            worker_inputs = [
+                (platform, tasks, args_dict, note_ids if not args.no_db else None)
+                for platform, tasks in active_groups.items()
+            ]
+            with Pool(processes=len(worker_inputs)) as pool:
+                worker_outputs = pool.starmap(scrape_notes_for_platform, worker_inputs)
+            for worker_results, worker_rows in worker_outputs:
+                results.extend(worker_results)
+                comment_rows.extend(worker_rows)
+        else:
+            parallel = False
+
+    if not parallel:
+        for index, task in enumerate(note_tasks, 1):
+            platform = detect_platform(task.note_url, task.platform_hint)
+            log(f"[{index}/{len(note_tasks)}] {task.style_name} | {platform} | {task.note_url}")
+            result, rows = scrape_note(task, args, note_ids if not args.no_db else None)
+            results.append(result)
+            comment_rows.extend(rows)
+            if args.between_notes_delay > 0 and index < len(note_tasks):
+                time.sleep(args.between_notes_delay)
+
+    json_path, csv_path = write_manifest(results)
+    comments_json_path, comments_csv_path = write_comment_summary(comment_rows)
+    ok_count = sum(1 for row in results if row.get("status") == "ok")
+    failed_count = sum(1 for row in results if row.get("status") == "failed")
+    skipped_count = sum(1 for row in results if row.get("status") in {"skipped", "dry_run"})
+    log(f"Done. ok={ok_count}, failed={failed_count}, skipped/dry_run={skipped_count}")
+    log(f"Collected comment rows: {len(comment_rows)}")
+    log(f"Manifest JSON: {json_path}")
+    log(f"Manifest CSV:  {csv_path}")
+    log(f"Comments JSON: {comments_json_path}")
+    log(f"Comments CSV:  {comments_csv_path}")
+    return 0 if failed_count == 0 else 1
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/feishu_doc_writer.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/feishu_doc_writer.py
new file mode 100644
index 0000000..d9395cc
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/feishu_doc_writer.py
@@ -0,0 +1,787 @@
+"""Feishu Docx (v2 blocks API) writer for note analysis reports.
+
+Wraps `lark-cli docs +create/+update --api-version v2` with:
+- atomic state persistence in data/state/feishu_doc.json
+- XML escaping + XML-native builders (data table, keyword distribution, etc.)
+- a node + run.js Windows pattern (mirrors analyze_note.send_feishu_report) to avoid
+  cmd 8K argv limit and `&|<>^%` escaping pitfalls on Windows
+- permission smoke test before any real write
+
+All lark-cli calls use `--profile hermes-analyzer --as bot --format json`.
+"""
+
+import json
+import os
+import shutil
+import subprocess
+import sys
+import time
+from pathlib import Path
+from typing import Any
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+STATE_DIR = PATHS.state_root
+TMP_DIR = PATHS.tmp_root / "feishu_doc_writer"
+
+STATE_FILE = STATE_DIR / "feishu_doc.json"
+
+PROFILE = "hermes-analyzer"
+IDENTITY = "bot"
+
+FEISHU_BASE_URL = "https://bu0zgpibak.feishu.cn"
+
+# Cap for the LLM analysis text inside a doc section (matches analyze_comments).
+DOC_ANALYSIS_LIMIT = 6000
+
+# Below this size we pass --content inline; above we write to a file and use --content @<rel>.
+INLINE_CONTENT_LIMIT = 7000
+
+# Per skill: only escape TEXT inside tags, never the tags themselves.
+# "&" must be replaced first to avoid double-escaping "&lt;" etc.
+_XML_ESCAPE_MAP = {
+    "&": "&amp;",
+    "<": "&lt;",
+    ">": "&gt;",
+    '"': "&quot;",
+}
+
+
+def xml_escape(text: Any) -> str:
+    """Escape XML text content. Converts newlines to <br/>."""
+    if text is None:
+        return ""
+    s = str(text)
+    for src, dst in _XML_ESCAPE_MAP.items():
+        s = s.replace(src, dst)
+    s = s.replace("\r\n", "\n").replace("\n", "<br/>")
+    return s
+
+
+def safe_div(numerator: Any, denominator: Any) -> float:
+    try:
+        n = float(numerator)
+        d = float(denominator)
+        if d == 0:
+            return 0.0
+        return n / d
+    except (TypeError, ValueError):
+        return 0.0
+
+
+def _pct(v: float) -> str:
+    return f"{v * 100:.2f}%" if v else "—"
+
+
+def _num(v: Any) -> str:
+    if v in ("", None):
+        return "—"
+    try:
+        return f"{int(float(v)):,}"
+    except (TypeError, ValueError):
+        return str(v)
+
+
+# ---------------------------------------------------------------------------
+# lark-cli subprocess wrapper
+# ---------------------------------------------------------------------------
+def run_lark(args: list[str], cwd: Path | None = None) -> dict[str, Any]:
+    """Invoke lark-cli via node + run.js, return parsed JSON dict."""
+    run_js = (
+        Path(os.environ.get("APPDATA", str(Path.home())))
+        / "npm"
+        / "node_modules"
+        / "@larksuite"
+        / "cli"
+        / "scripts"
+        / "run.js"
+    )
+    # Scheduled tasks run as SYSTEM; %APPDATA% is not Administrator's dir
+    if not run_js.exists():
+        admin_run_js = Path(os.path.expandvars(r"%APPDATA%\npm\node_modules\@larksuite\cli\scripts\run.js"))
+        if admin_run_js.exists():
+            run_js = admin_run_js
+    if not run_js.exists():
+        raise RuntimeError(f"找不到 lark-cli 入口: {run_js}")
+
+    cmd = [
+        "node", str(run_js),
+        "--profile", PROFILE,
+        "--as", IDENTITY,
+        "--format", "json",
+        *args,
+    ]
+    try:
+        proc = subprocess.run(
+            cmd,
+            cwd=str(cwd) if cwd else None,
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+            timeout=180,
+        )
+    except subprocess.TimeoutExpired as exc:
+        raise RuntimeError(f"lark-cli 调用超时: {exc}") from exc
+
+    if proc.returncode != 0:
+        raise RuntimeError(
+            f"lark-cli 调用失败 (exit {proc.returncode})\n"
+            f"cmd head: {' '.join(args[:4])}\n"
+            f"stderr: {proc.stderr.strip()[:500]}\n"
+            f"stdout: {proc.stdout[:500]}"
+        )
+
+    out = proc.stdout.strip()
+    if not out:
+        return {}
+    try:
+        result = json.loads(out)
+    except json.JSONDecodeError as exc:
+        raise RuntimeError(f"lark-cli 返回非 JSON: {out[:500]}") from exc
+
+    if isinstance(result, dict) and result.get("ok") is False:
+        err = result.get("error") or {}
+        raise RuntimeError(
+            f"lark-cli API error: code={err.get('code')} type={err.get('type')} "
+            f"msg={err.get('message')}"
+        )
+
+    # Some API errors surface as ok=True + result="failed" + warnings. Detect those.
+    data = (result.get("data") or {}) if isinstance(result, dict) else {}
+    op_result = data.get("result") if isinstance(data, dict) else None
+    if op_result == "failed":
+        warnings = data.get("warnings") or []
+        raise RuntimeError(
+            f"lark-cli 操作失败 (result=failed): warnings={warnings}"
+        )
+    return result if isinstance(result, dict) else {}
+
+
+# ---------------------------------------------------------------------------
+# State persistence
+# ---------------------------------------------------------------------------
+def load_state() -> dict[str, Any]:
+    if not STATE_FILE.exists():
+        return {"docs": {}, "written_notes": {}}
+    try:
+        data = json.loads(STATE_FILE.read_text(encoding="utf-8"))
+        if not isinstance(data, dict):
+            return {"docs": {}, "written_notes": {}}
+        data.setdefault("docs", {})
+        data.setdefault("written_notes", {})
+        return data
+    except json.JSONDecodeError:
+        backup = STATE_FILE.with_suffix(f".corrupt-{int(time.time())}.json")
+        shutil.copy2(STATE_FILE, backup)
+        return {"docs": {}, "written_notes": {}, "_corrupt_backup": str(backup)}
+
+
+def save_state(state: dict[str, Any]) -> None:
+    STATE_DIR.mkdir(parents=True, exist_ok=True)
+    tmp = STATE_FILE.with_suffix(".tmp")
+    tmp.write_text(json.dumps(state, ensure_ascii=False, indent=2), encoding="utf-8")
+    tmp.replace(STATE_FILE)
+
+
+# ---------------------------------------------------------------------------
+# Doc API wrappers
+# ---------------------------------------------------------------------------
+def create_doc(xml_content: str) -> dict[str, Any]:
+    """Create a new Docx. Returns {token, url, revision_id}."""
+    result = run_lark([
+        "docs", "+create",
+        "--api-version", "v2",
+        "--content", xml_content,
+    ])
+    doc = (result.get("data") or {}).get("document") or {}
+    return {
+        "token": doc.get("document_id") or "",
+        "url": doc.get("url") or "",
+        "revision_id": doc.get("revision_id"),
+    }
+
+
+def _invoke_update_with_content(
+    doc: str, command: str, xml_content: str,
+    *, anchor: str | None = None,
+) -> dict[str, Any]:
+    """Run docs +update with either inline --content or @file content."""
+    args = [
+        "docs", "+update",
+        "--api-version", "v2",
+        "--doc", doc,
+        "--command", command,
+    ]
+    if command.startswith("block_"):
+        if not anchor:
+            raise ValueError(f"command {command} requires --block-id")
+        args.extend(["--block-id", anchor])
+
+    if len(xml_content.encode("utf-8")) <= INLINE_CONTENT_LIMIT:
+        args.extend(["--content", xml_content])
+        return run_lark(args)
+
+    TMP_DIR.mkdir(parents=True, exist_ok=True)
+    fname = f"_docwriter_{int(time.time() * 1000)}.xml"
+    fpath = TMP_DIR / fname
+    fpath.write_text(xml_content, encoding="utf-8")
+    try:
+        # lark-cli resolves `@file` relative to cwd. We cwd=STATE_DIR but the
+        # file lives in STATE_DIR/tmp — pass the relative path from STATE_DIR.
+        rel_path = f"tmp/{fname}"
+        args.extend(["--content", f"@{rel_path}"])
+        return run_lark(args, cwd=STATE_DIR)
+    finally:
+        try:
+            fpath.unlink()
+        except OSError:
+            pass
+
+
+def append_blocks(doc: str, xml_content: str) -> dict[str, Any]:
+    """Append XML blocks to end of doc. Returns {revision_id}."""
+    return _invoke_update_with_content(doc, "append", xml_content)
+
+
+def insert_blocks_after(doc: str, anchor_block_id: str, xml_content: str) -> dict[str, Any]:
+    """Insert XML blocks after a given block. Returns {revision_id}."""
+    return _invoke_update_with_content(
+        doc, "block_insert_after", xml_content, anchor=anchor_block_id,
+    )
+
+
+def delete_blocks(doc: str, block_ids: list[str]) -> dict[str, Any]:
+    """Batch-delete blocks (comma-separated --block-id)."""
+    if not block_ids:
+        return {}
+    args = [
+        "docs", "+update",
+        "--api-version", "v2",
+        "--doc", doc,
+        "--command", "block_delete",
+        "--block-id", ",".join(block_ids),
+    ]
+    return run_lark(args)
+
+
+def fetch_doc_blocks(doc: str, *, scope: str = "all") -> list[dict[str, Any]]:
+    """Fetch block IDs in a doc. Returns list of {block_id, block_type} dicts."""
+    args = [
+        "docs", "+fetch",
+        "--api-version", "v2",
+        "--doc", doc,
+        "--detail", "with-ids",
+    ]
+    if scope and scope != "all":
+        args.extend(["--scope", scope])
+
+    result = run_lark(args)
+    # The fetch returns content as XML string; parse block IDs from `id="..."` attrs.
+    content = ((result.get("data") or {}).get("document") or {}).get("content") or ""
+    out: list[dict[str, Any]] = []
+    import re
+    # Extract <tagname id="..."> in document order
+    for m in re.finditer(r"<(\w+)\s+id=\"([^\"]+)\"", content):
+        out.append({"block_id": m.group(2), "block_type": m.group(1)})
+    return out
+
+
+def find_block_after(doc: str, anchor_text: str) -> str | None:
+    """Find the block_id of the first <h*> block whose text contains anchor_text.
+
+    Used to discover the h1 anchor after doc creation, since the API response
+    doesn't include new_blocks.
+    """
+    blocks = fetch_doc_blocks(doc, scope="all")
+    for b in blocks:
+        if b["block_type"] in ("h1", "h2", "h3", "h4"):
+            # We need the actual text; refetch with content
+            pass
+    # The fetch returned content embedded; do a full re-fetch with text
+    result = run_lark([
+        "docs", "+fetch",
+        "--api-version", "v2",
+        "--doc", doc,
+        "--detail", "with-ids",
+    ])
+    content = ((result.get("data") or {}).get("document") or {}).get("content") or ""
+    import re
+    # Find all <h1 ...>...</h1> or <h2 ...>...</h2>
+    pattern = re.compile(r"<(h\d)\s+id=\"([^\"]+)\"[^>]*>(.*?)</\1>", re.DOTALL)
+    for m in pattern.finditer(content):
+        if anchor_text in m.group(3):
+            return m.group(2)
+    return None
+
+
+def find_blocks_after_anchor(doc: str, anchor_block_id: str) -> list[str]:
+    """Return block_ids of every block that appears AFTER anchor_block_id, in order.
+
+    Used by populate_master_index to delete stale index content before re-inserting.
+    """
+    blocks = fetch_doc_blocks(doc, scope="all")
+    out: list[str] = []
+    found = False
+    for b in blocks:
+        if found:
+            out.append(b["block_id"])
+        elif b["block_id"] == anchor_block_id:
+            found = True
+    return out
+
+
+def find_index_blocks(doc: str, anchor_block_id: str) -> list[str]:
+    """Return block_ids of the index content (blocks between anchor h1 and the next style section).
+
+    The index lives immediately after the anchor h1 and ends at the first
+    subsequent style section marker — either an H1 (the style title) or a
+    bookmark that precedes it. This avoids deleting style sections.
+    """
+    blocks = fetch_doc_blocks(doc, scope="all")
+    out: list[str] = []
+    found_anchor = False
+    for b in blocks:
+        if not found_anchor:
+            if b["block_id"] == anchor_block_id:
+                found_anchor = True
+            continue
+        # Stop at the first style-section marker after the anchor.
+        if b["block_type"] in ("h1", "bookmark"):
+            break
+        out.append(b["block_id"])
+    return out
+
+
+# ---------------------------------------------------------------------------
+# XML builders
+# ---------------------------------------------------------------------------
+CATEGORY_LABELS = {
+    "purchase_intent": "购买意向",
+    "positive_feature": "正向功能反馈",
+    "negative_feature": "负向功能反馈",
+    "user_scenario": "用户场景",
+    "aesthetic": "审美与风格",
+    "brand_emotion": "品牌情感",
+}
+
+
+def build_basic_info_xml(info: dict[str, Any]) -> str:
+    title = info.get("title") or "未获取"
+    platform = info.get("platform") or "unknown"
+    creator = info.get("creator_name") or "未指定"
+    style_name = info.get("style_name") or ""
+    source_url = info.get("source_url") or ""
+    brand = info.get("brand") or ""
+
+    target_label = " / ".join(x for x in (brand, style_name) if x) if (brand or style_name) else "未指定"
+
+    lines = [
+        f"<p><b>标题</b>　{xml_escape(title)}</p>",
+        f"<p><b>平台</b>　{platform}　·　<b>达人</b>　{xml_escape(creator)}</p>",
+        f"<p><b>品牌 / 产品</b>　{xml_escape(target_label)}</p>",
+    ]
+    if source_url:
+        lines.append(f'<p><b>链接</b>　<a href="{xml_escape(source_url)}">{xml_escape(source_url)}</a></p>')
+
+    return "<callout emoji=\"📌\" background-color=\"light-blue\" border-color=\"blue\">" + "".join(lines) + "</callout>"
+
+
+def build_data_table_xml(
+    info: dict[str, Any], metrics: dict[str, Any], raw_total: int, filtered_total: int,
+) -> str:
+    """XML table mirroring analyze_note._data_table_markdown."""
+    has_view = bool(metrics.get("view_count"))
+    likes = info.get("like_count")
+    favorites = info.get("favorite_count")
+    comments = info.get("comment_count_metric") or info.get("comment_count_total")
+    shares = info.get("share_count")
+    valid_ratio = safe_div(filtered_total, raw_total)
+
+    rows: list[str] = []
+    rows.append(_tr(("维度", "指标", "数值"), header=True))
+    rows.append(_tr(("基础", "点赞", _num(likes))))
+    rows.append(_tr(("基础", "收藏", _num(favorites))))
+    rows.append(_tr(("基础", "评论", _num(comments))))
+    rows.append(_tr(("基础", "分享", _num(shares))))
+    rows.append(_tr(("传播", "曝光量", _num(metrics.get("view_count")) if has_view else "未采集")))
+    rows.append(_tr(("互动率", "点赞率", _pct(metrics.get("like_rate", 0)))))
+    rows.append(_tr(("互动率", "评论率", _pct(metrics.get("comment_rate", 0)))))
+    rows.append(_tr(("互动率", "收藏率", _pct(metrics.get("favorite_rate", 0)))))
+    rows.append(_tr(("互动率", "分享率", _pct(metrics.get("share_rate", 0)))))
+    rows.append(_tr(("互动率", "综合互动率", _pct(metrics.get("engagement_rate", 0)))))
+    fav_to_cmt = metrics.get("favorite_to_comment", 0) or 0
+    rows.append(_tr(("互动率", "收藏评论比", f"{fav_to_cmt:.2f}")))
+    rows.append(_tr(("质量", "有效率", _pct(valid_ratio))))
+
+    return (
+        '<callout emoji="📊" background-color="light-blue" border-color="blue">'
+        "<table>"
+        "<colgroup><col span=\"3\" width=\"120\"/></colgroup>"
+        "<thead>" + rows[0] + "</thead>"
+        "<tbody>" + "".join(rows[1:]) + "</tbody>"
+        "</table>"
+        "</callout>"
+    )
+
+
+def _tr(cells: tuple[str, str, str], *, header: bool = False) -> str:
+    tag = "th" if header else "td"
+    bg = ' background-color="light-gray"' if header else ""
+    return (
+        f"<tr>"
+        + "".join(f"<{tag}{bg}>{xml_escape(c)}</{tag}>" for c in cells)
+        + "</tr>"
+    )
+
+
+def build_keyword_distribution_xml(
+    distribution: dict[str, Any], total_filtered: int,
+) -> str:
+    overall = distribution.get("overall", {}) if isinstance(distribution, dict) else {}
+    relevant = distribution.get("relevant", {}) if isinstance(distribution, dict) else {}
+    relevant_total = distribution.get("relevant_total", 0) if isinstance(distribution, dict) else 0
+
+    if not overall:
+        return "<p>（无关键词分布数据）</p>"
+
+    def fmt(d: dict[str, int]) -> str:
+        return "".join(
+            f"<li>{xml_escape(CATEGORY_LABELS[k])}：{d.get(k, 0)} 条</li>"
+            for k in CATEGORY_LABELS
+        )
+
+    overall_block = (
+        f"<p><b>全评论（{total_filtered} 条）</b></p>"
+        f"<ul>{fmt(overall)}</ul>"
+    )
+    if relevant_total:
+        relevant_block = (
+            f"<p><b>我方产品相关（{relevant_total} 条）</b></p>"
+            f"<ul>{fmt(relevant)}</ul>"
+        )
+    else:
+        relevant_block = "<p><b>我方产品相关（0 条）</b></p><p><i>（无评论明确提及我方产品/品牌）</i></p>"
+
+    return (
+        '<callout emoji="🔑" background-color="light-cyan" border-color="cyan">'
+        + overall_block + relevant_block
+        + "</callout>"
+    )
+
+
+# 模块 emoji 和颜色映射
+_MODULE_STYLES = {
+    "模块1": ("🎯", "light-blue", "blue"),       # 笔记表现归因
+    "模块2": ("📊", "light-green", "green"),      # 评论质量分层
+    "模块3": ("🛡️", "light-red", "red"),          # 品牌安全与控评
+    "模块4": ("🎬", "light-purple", "purple"),    # 脚本有效性拆解
+    "模块5": ("🤝", "light-yellow", "orange"),    # 达人匹配度评估
+    "模块6": ("💡", "light-cyan", "cyan"),        # 可沉淀的选题资产
+    "模块7": ("✅", "light-green", "green"),      # 下一篇可执行优化清单
+}
+
+
+def _split_analysis_modules(analysis_text: str) -> list[str]:
+    """将 LLM 输出的 7 模块分析拆分为独立 callout 块。
+
+    每个模块用不同颜色的 callout 渲染，视觉层次更分明。
+    """
+    if not analysis_text or analysis_text == "（无分析结果）":
+        return ['<callout emoji="💡" background-color="light-gray" border-color="gray">'
+                '<p>（无分析结果）</p></callout>']
+
+    # 按模块标记拆分
+    import re as _re
+    parts = _re.split(r'(模块\d+[｜|])', analysis_text)
+    # parts: [前缀, "模块1｜", 内容1, "模块2｜", 内容2, ...]
+
+    blocks: list[str] = []
+
+    # 处理模块前的引言文字（如有）
+    if parts and parts[0].strip():
+        intro = parts[0].strip()
+        if len(intro) > 20:
+            blocks.append(f'<callout emoji="📋" background-color="light-gray" border-color="gray">'
+                          f'<p>{xml_escape(intro)}</p></callout>')
+
+    # 逐模块渲染
+    i = 1
+    while i < len(parts) - 1:
+        module_header = parts[i]       # "模块1｜"
+        module_body = parts[i + 1]     # 该模块的内容
+        i += 2
+
+        # 提取模块编号
+        m = _re.match(r'模块(\d+)', module_header)
+        if not m:
+            continue
+        mod_num = m.group(1)
+        mod_key = f"模块{mod_num}"
+        emoji, bg, border = _MODULE_STYLES.get(mod_key, ("📌", "light-gray", "gray"))
+
+        # 模块标题行 + 正文
+        body_text = module_body.strip()
+        if not body_text:
+            continue
+
+        # 截断过长内容
+        if len(body_text) > 800:
+            body_text = body_text[:800] + "\n...（已截断）"
+
+        blocks.append(
+            f'<callout emoji="{emoji}" background-color="{bg}" border-color="{border}">'
+            f'<p>{xml_escape(body_text)}</p></callout>'
+        )
+
+    # 如果没有拆出模块（LLM 没按格式输出），整体兜底
+    if not blocks:
+        text = analysis_text
+        if len(text) > DOC_ANALYSIS_LIMIT:
+            text = text[:DOC_ANALYSIS_LIMIT] + "\n...（已截断）"
+        blocks.append(f'<callout emoji="💡" background-color="light-blue" border-color="blue">'
+                      f'<p>{xml_escape(text)}</p></callout>')
+
+    return blocks
+
+
+def build_note_section_xml(
+    info: dict[str, Any],
+    metrics: dict[str, Any],
+    distribution: dict[str, Any],
+    raw_comments: list[str],
+    filtered_comments: list[str],
+    analysis: str,
+    *,
+    bookmark_name: str,
+    doc_token: str,
+) -> str:
+    title = info.get("title") or info.get("source_url") or "未获取"
+    platform = info.get("platform") or "unknown"
+    # 截断标题用于 H3，避免超长
+    short_title = title[:30] + "…" if len(title) > 30 else title
+    h_text = f"📝 {short_title}（{platform}）"
+
+    analysis_text = analysis or "（无分析结果）"
+    if len(analysis_text) > DOC_ANALYSIS_LIMIT:
+        analysis_text = analysis_text[:DOC_ANALYSIS_LIMIT] + "<br/>...（内容过长，已截断）"
+
+    # Bookmark requires href attribute; use the doc's own URL (target for incoming #bookmark links).
+    bookmark_href = f"{FEISHU_BASE_URL}/docx/{doc_token}#{bookmark_name}"
+
+    # 将分析文本按模块拆分，逐段渲染
+    analysis_parts = _split_analysis_modules(analysis_text)
+
+    return (
+        f"<bookmark name=\"{xml_escape(bookmark_name)}\" href=\"{xml_escape(bookmark_href)}\"></bookmark>"
+        f"<h3>{xml_escape(h_text)}</h3>"
+        + build_basic_info_xml(info)
+        + build_data_table_xml(info, metrics, len(raw_comments), len(filtered_comments))
+        + build_keyword_distribution_xml(distribution, len(filtered_comments))
+        + "".join(analysis_parts)
+    )
+
+
+def build_master_summary_table_xml(
+    per_platform: dict[str, dict[str, Any]],
+) -> str:
+    """Build the 4-row master summary table (平台/笔记数/详情)."""
+    rows = []
+    for p, info in per_platform.items():
+        url = info.get("url") or ""
+        n = info.get("count", 0)
+        link_cell = f'<a href="{xml_escape(url)}">打开 →</a>' if url else "—"
+        rows.append(
+            f"<tr><td>{xml_escape(p)}</td><td>{n}</td><td>{link_cell}</td></tr>"
+        )
+    return (
+        "<table>"
+        "<colgroup><col width=\"120\"/><col width=\"80\"/><col width=\"300\"/></colgroup>"
+        "<thead>" + _tr(("平台", "笔记数", "详情"), header=True) + "</thead>"
+        "<tbody>" + "".join(rows) + "</tbody>"
+        "</table>"
+    )
+
+
+def build_per_platform_index_table_xml(
+    platform: str, state: dict[str, Any], platform_doc_url: str,
+) -> tuple[str, int]:
+    """Build index table XML for one platform section in the master doc.
+
+    Returns (xml_string, row_count).
+    """
+    rows: list[str] = []
+    for nid_str, entry in state.get("written_notes", {}).items():
+        if entry.get("platform") != platform:
+            continue
+        bookmark = entry.get("bookmark_name") or f"note-{nid_str}"
+        link = f"{platform_doc_url}#{bookmark}"
+        rows.append(
+            f'<tr><td>{nid_str}</td>'
+            f'<td><a href="{xml_escape(link)}">查看 →</a></td></tr>'
+        )
+    rows.sort(key=lambda r: int(re.search(r">(\d+)<", r).group(1)) if re.search(r">(\d+)<", r) else 0)
+    body = "".join(rows)
+    return (
+        f"<h2>📑 {platform} 索引（{len(rows)} 条）</h2>"
+        "<table>"
+        "<colgroup><col width=\"100\"/><col width=\"180\"/></colgroup>"
+        "<thead>" + _tr(("note_id", "详情"), header=True) + "</thead>"
+        "<tbody>" + body + "</tbody>"
+        "</table>"
+    ), len(rows)
+
+
+# Local re for sort
+import re  # noqa: E402
+
+
+# ---------------------------------------------------------------------------
+# Per-style doc builders (used by single-archive layout)
+# ---------------------------------------------------------------------------
+def _doc_skeleton() -> str:
+    """Initial skeleton for the single archive doc.
+
+    Title + intro + H1 anchor placeholder for the 款式索引. The actual index
+    table is populated later via populate_index_table() (replace-on-update).
+    """
+    return (
+        "<title>👜 笔记分析报告 — 按款式汇总</title>"
+        "<p>本汇总由 write_notes_to_doc.py 自动生成。点击款式名称进入详情。</p>"
+        + "<h1>⚠️ 款式索引（点击跳转）</h1>"
+    )
+
+
+def build_style_overview_xml(style_name: str, aggregated: dict[str, Any]) -> str:
+    """Callout block with style-level rollup metrics."""
+    plat = aggregated.get("platform_breakdown") or {}
+    plat_str = " / ".join(f"{k} {v}" for k, v in plat.items() if v) or "—"
+    creator = aggregated.get("creator_breakdown") or {}
+    creator_str = " / ".join(f"{k}×{v}" for k, v in list(creator.items())[:3]) or "—"
+
+    lines = [
+        f"<p><b>款式</b>　{xml_escape(style_name)}</p>",
+        f"<p><b>笔记总数</b>　{aggregated.get('note_count', 0)} 篇　·　"
+        f"<b>平台分布</b>　{xml_escape(plat_str)}</p>",
+        f"<p><b>头部达人</b>　{xml_escape(creator_str)}</p>",
+        f"<p><b>总互动</b>　"
+        f"点赞 {_num(aggregated.get('total_likes'))}　·　"
+        f"收藏 {_num(aggregated.get('total_favorites'))}　·　"
+        f"评论 {_num(aggregated.get('total_comments'))}　·　"
+        f"分享 {_num(aggregated.get('total_shares'))}</p>",
+        f"<p><b>总曝光量</b>　{_num(aggregated.get('total_views')) if aggregated.get('total_views') else '未采集'}　·　"
+        f"<b>综合互动率</b>　{_pct(aggregated.get('engagement_rate') or 0)}</p>",
+    ]
+    return (
+        '<callout emoji="👜" background-color="light-yellow" border-color="orange">'
+        + "".join(lines)
+        + "</callout>"
+    )
+
+
+def build_style_section_xml(
+    style_id: int,
+    style_name: str,
+    aggregated: dict[str, Any],
+    rollup_text: str,
+    note_sections_xml: str,
+    *,
+    doc_token: str,
+) -> str:
+    """Assemble one full style section: bookmark + H1 + overview + rollup H2 + per-note sections.
+
+    `note_sections_xml` should be the concatenation of `build_note_section_xml(...)`
+    outputs for each note in this style (each already has its own bookmark + H3).
+    """
+    bookmark_name = f"style-{style_id}"
+    bookmark_href = f"{FEISHU_BASE_URL}/docx/{doc_token}#{bookmark_name}"
+    h1_text = f"👜 {style_name}（{aggregated.get('note_count', 0)} 篇笔记）"
+
+    note_count = aggregated.get("note_count", 0)
+
+    rollup_in_callout = (
+        '<callout emoji="💼" background-color="light-purple" border-color="purple">'
+        f'<p>{xml_escape(rollup_text or "（无整体分析结果）")}</p></callout>'
+    )
+
+    parts: list[str] = [
+        f'<bookmark name="{xml_escape(bookmark_name)}" href="{xml_escape(bookmark_href)}"></bookmark>',
+        f"<h1>{xml_escape(h1_text)}</h1>",
+        build_style_overview_xml(style_name, aggregated),
+        "<h2>💼 款式整体分析</h2>",
+        rollup_in_callout,
+        f"<h2>📚 单篇笔记分析（{note_count} 篇）</h2>",
+    ]
+
+    if note_sections_xml:
+        parts.append(note_sections_xml)
+
+    return "".join(parts)
+
+
+def build_style_index_table_xml(
+    written_styles: dict[str, Any],
+    *,
+    archive_url: str,
+) -> str:
+    """Build the 款式索引 table at the top of the archive doc.
+
+    Rows are sorted by note count DESC (driven by caller-supplied dict order,
+    which write_notes_to_doc.py maintains from db.get_all_styles_with_notes).
+    Each row: 款式名 | 笔记数 | 跳转链接.
+    """
+    rows: list[str] = []
+    for sid_str, entry in written_styles.items():
+        if not entry.get("doc_token"):
+            continue
+        # Find note count from entry (carried over by writer)
+        n = entry.get("note_count", 0)
+        bookmark = entry.get("bookmark_name") or f"style-{sid_str}"
+        link = f"{archive_url}#{bookmark}"
+        rows.append(
+            f"<tr><td>{xml_escape(str(entry.get('style_name', sid_str)))}</td>"
+            f"<td>{n}</td>"
+            f'<td><a href="{xml_escape(link)}">跳转 →</a></td></tr>'
+        )
+    if not rows:
+        rows.append("<tr><td>—</td><td>0</td><td>—</td></tr>")
+    return (
+        "<table>"
+        "<colgroup><col width=\"160\"/><col width=\"80\"/><col width=\"300\"/></colgroup>"
+        "<thead>" + _tr(("款式", "笔记数", "详情"), header=True) + "</thead>"
+        "<tbody>" + "".join(rows) + "</tbody>"
+        "</table>"
+    )
+
+
+# ---------------------------------------------------------------------------
+# Permission smoke test
+# ---------------------------------------------------------------------------
+def permission_smoke_test() -> bool:
+    print("→ Running permission smoke test (--as bot, docs:document:create)...")
+    try:
+        run_lark([
+            "docs", "+create",
+            "--api-version", "v2",
+            "--content", "<title>__smoke_test__</title><p>please ignore</p>",
+        ])
+    except RuntimeError as exc:
+        print(f"  ✗ Smoke test failed: {exc}")
+        print()
+        print("  The hermes-analyzer bot identity lacks the required scopes.")
+        print("  Run: lark-cli auth login --profile hermes-analyzer")
+        print("  And ensure the app is granted these scopes:")
+        print("    - docs:document:create")
+        print("    - docs:document:update")
+        print("    - docs:document:read")
+        print("  Then retry.")
+        return False
+
+    print(f"  ✓ Smoke test passed. (A throwaway doc was created — delete it manually from Drive.)")
+    return True
+
+
+if __name__ == "__main__":
+    sys.exit(0 if permission_smoke_test() else 1)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/friday_relogin.bat b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/friday_relogin.bat
new file mode 100644
index 0000000..95c66be
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/friday_relogin.bat
@@ -0,0 +1,39 @@
+@echo off
+if not defined GYXX_PYTHON set "GYXX_PYTHON=python"
+REM Friday relogin: 4 QR scans — B站, 蒲公英, 星图(+同步到抖音评论), 小红书评论
+REM 星图登录完成后自动把 cookie 同步到抖音评论 scraper,不需要单独再扫抖音码
+REM Register: schtasks /Create /SC WEEKLY /D FRI /TN YingxiaoYunying_FridayRelogin /TR %PROJECT_DIR%\data\tools\friday_relogin.bat /ST 10:00 /F
+
+setlocal
+
+for %%I in ("%~dp0..\..") do set "PROJECT_DIR=%%~fI"
+if not defined GYXX_DATA_ROOT (
+    if not defined GYXX_PROJECT_ROOT (
+        echo ERROR: GYXX_DATA_ROOT or GYXX_PROJECT_ROOT must be defined. 1>&2
+        exit /b 3
+    )
+    set "GYXX_DATA_ROOT=%GYXX_PROJECT_ROOT%\var"
+)
+set LOG_DIR=%GYXX_DATA_ROOT%\logs\content_marketing\scheduled
+set PYTHON=%GYXX_PYTHON%
+where %PYTHON% >nul 2>&1 || set PYTHON=python
+
+if not exist "%LOG_DIR%" mkdir "%LOG_DIR%"
+
+set TS=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%
+set TS=%TS: =0%
+
+set LOG_FILE=%LOG_DIR%\friday_relogin_%TS%.log
+
+echo === Friday relogin started at %date% %time% === > "%LOG_FILE%"
+echo PROJECT_DIR=%PROJECT_DIR% >> "%LOG_FILE%"
+
+cd /d "%PROJECT_DIR%"
+
+echo --- Launch 5 QR relogins in parallel; send screenshots; retry up to 3 rounds --- >> "%LOG_FILE%"
+call %PYTHON% data\tools\friday_relogin_parallel.py --max-attempts 3 --round-timeout 300 --screenshot-delay 60 >> "%LOG_FILE%" 2>&1
+set FINAL_RC=%ERRORLEVEL%
+
+echo === Friday relogin finished at %date% %time% (exit=%FINAL_RC%) === >> "%LOG_FILE%"
+
+endlocal & exit /b %FINAL_RC%
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/friday_relogin_parallel.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/friday_relogin_parallel.py
new file mode 100644
index 0000000..2a9a399
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/friday_relogin_parallel.py
@@ -0,0 +1,408 @@
+"""Launch five QR relogins in parallel, notify via analyzer Lark, retry 3 times."""
+
+from __future__ import annotations
+
+import argparse
+import ctypes
+from ctypes import wintypes
+import datetime as dt
+import json
+import os
+import shutil
+import subprocess
+import sys
+import time
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Callable
+
+import psutil
+
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+PROJECT_DIR = PATHS.module_root
+ANALYZER_DIR = PATHS.tmp_root / "relogin"
+MARKETING_RECIPIENT = "ou_24cc944d6e43c69c59d6560ad4e2ae6e"
+TECHNICAL_RECIPIENT = "ou_7ad5fc8012e2f741afc5346e05ffd447"
+
+
+@dataclass(frozen=True)
+class Platform:
+    label: str
+    command: tuple[str, ...]
+    cookie_file: Path
+    required_cookie: str
+
+
+PLATFORMS = {
+    "douyin": Platform(
+        "\u6296\u97f3", ("data/tools/relogin_douyin.py",),
+        PATHS.state_root / "cookies/douyin_cookies.json", "sessionid",
+    ),
+    "bilibili": Platform(
+        "B\u7ad9", ("data/tools/relogin_bilibili.py",),
+        PATHS.state_root / "cookies/bilibili_cookies.json", "SESSDATA",
+    ),
+    "pgy": Platform(
+        "\u84b2\u516c\u82f1", ("data/tools/relogin_pgy.py",),
+        PATHS.state_root / "cookies/pgy_cookies.json", "",
+    ),
+    "xingtu": Platform(
+        "\u661f\u56fe", ("data/tools/relogin_xingtu.py", "--force"),
+        PATHS.state_root / "cookies/xingtu_cookies.json", "sessionid",
+    ),
+    "xiaohongshu": Platform(
+        "\u5c0f\u7ea2\u4e66", ("data/tools/relogin_xiaohongshu.py",),
+        PATHS.state_root / "cookies/xiaohongshu_cookies.json", "web_session",
+    ),
+}
+
+DEFAULT_RECIPIENTS = {
+    "pgy": MARKETING_RECIPIENT,
+    "xingtu": MARKETING_RECIPIENT,
+    "douyin": TECHNICAL_RECIPIENT,
+    "bilibili": TECHNICAL_RECIPIENT,
+    "xiaohongshu": TECHNICAL_RECIPIENT,
+}
+
+
+def recipient_for_platform(platform_name: str, override: str | None = None) -> str:
+    """Return the responsible recipient, unless a manual all-platform override is set."""
+    if override:
+        return override
+    try:
+        return DEFAULT_RECIPIENTS[platform_name]
+    except KeyError as exc:
+        raise ValueError(f"no Lark recipient configured for platform: {platform_name}") from exc
+
+
+def group_platforms_by_recipient(
+    platform_names: list[str], override: str | None = None,
+) -> dict[str, list[str]]:
+    grouped: dict[str, list[str]] = {}
+    for name in platform_names:
+        grouped.setdefault(recipient_for_platform(name, override), []).append(name)
+    return grouped
+
+
+def run_with_retries(
+    platform_names: list[str],
+    *,
+    max_attempts: int,
+    run_round: Callable[[list[str], int], dict[str, bool]],
+) -> dict[str, bool]:
+    status = {name: False for name in platform_names}
+    pending = list(platform_names)
+    for attempt in range(1, max_attempts + 1):
+        if not pending:
+            break
+        round_result = run_round(pending, attempt)
+        for name in pending:
+            status[name] = bool(round_result.get(name))
+        pending = [name for name in pending if not status[name]]
+    return status
+
+
+def _valid_cookie(platform: Platform) -> bool:
+    if not platform.cookie_file.is_file():
+        return False
+    if not platform.required_cookie:
+        return platform.cookie_file.stat().st_size > 10
+    try:
+        import json
+        values = json.loads(platform.cookie_file.read_text(encoding="utf-8"))
+        return any(
+            item.get("name") == platform.required_cookie and item.get("value")
+            for item in values if isinstance(item, dict)
+        )
+    except Exception:
+        return False
+
+
+def _descendant_pids(pid: int) -> set[int]:
+    try:
+        return {child.pid for child in psutil.Process(pid).children(recursive=True)}
+    except (psutil.Error, OSError):
+        return set()
+
+
+def _terminate_process_tree(process: subprocess.Popen) -> None:
+    """Terminate the whole Windows subprocess tree, including Chrome/Playwright."""
+    try:
+        subprocess.run(
+            ["taskkill", "/PID", str(process.pid), "/T", "/F"],
+            capture_output=True,
+            text=True,
+            check=False,
+        )
+    except OSError:
+        process.terminate()
+    try:
+        process.wait(timeout=10)
+    except subprocess.TimeoutExpired:
+        process.kill()
+
+
+def _window_for_process(pid: int) -> int | None:
+    pids = _descendant_pids(pid) | {pid}
+    handles: list[int] = []
+    user32 = ctypes.windll.user32
+    callback_type = ctypes.WINFUNCTYPE(ctypes.c_bool, ctypes.c_void_p, ctypes.c_void_p)
+
+    def callback(hwnd, _lparam):
+        process_id = ctypes.c_ulong()
+        user32.GetWindowThreadProcessId(hwnd, ctypes.byref(process_id))
+        if process_id.value in pids and user32.IsWindowVisible(hwnd):
+            handles.append(int(hwnd))
+        return True
+
+    user32.EnumWindows(callback_type(callback), 0)
+    return handles[0] if handles else None
+
+
+def _platform_from_window_title(title: str) -> str | None:
+    """Identify a login window without relying on Chrome's process tree."""
+    if "\u84b2\u516c\u82f1" in title:
+        return "pgy"
+    if "\u6296\u97f3\u6388\u6743" in title or "\u5de8\u91cf\u661f\u56fe" in title:
+        return "xingtu"
+    if "\u6296\u97f3" in title:
+        return "douyin"
+    if "bilibili" in title.lower() or "\u8d26\u53f7\u767b\u5f55" in title:
+        return "bilibili"
+    if "\u5c0f\u7ea2\u4e66" in title:
+        return "xiaohongshu"
+    return None
+
+
+def _login_windows() -> dict[str, int]:
+    """Map visible Chrome login windows to platforms by their page titles."""
+    user32 = ctypes.windll.user32
+    found: dict[str, int] = {}
+    callback_type = ctypes.WINFUNCTYPE(ctypes.c_bool, ctypes.c_void_p, ctypes.c_void_p)
+
+    def callback(hwnd, _lparam):
+        if not user32.IsWindowVisible(hwnd):
+            return True
+        length = user32.GetWindowTextLengthW(hwnd)
+        if length <= 0:
+            return True
+        buffer = ctypes.create_unicode_buffer(length + 1)
+        user32.GetWindowTextW(hwnd, buffer, length + 1)
+        platform = _platform_from_window_title(buffer.value)
+        if platform:
+            found[platform] = int(hwnd)
+        return True
+
+    user32.EnumWindows(callback_type(callback), 0)
+    return found
+
+
+def _capture_window(hwnd: int, output: Path) -> None:
+    """Maximize and render one window directly, even when it is not foreground."""
+    from PIL import Image
+
+    user32 = ctypes.windll.user32
+    gdi32 = ctypes.windll.gdi32
+    user32.ShowWindow(hwnd, 3)  # SW_MAXIMIZE
+    time.sleep(1)
+    rect = wintypes.RECT()
+    user32.GetWindowRect(hwnd, ctypes.byref(rect))
+    width, height = rect.right - rect.left, rect.bottom - rect.top
+    if width <= 0 or height <= 0:
+        raise RuntimeError(f"invalid window size: {width}x{height}")
+
+    window_dc = user32.GetWindowDC(hwnd)
+    memory_dc = gdi32.CreateCompatibleDC(window_dc)
+    bitmap = gdi32.CreateCompatibleBitmap(window_dc, width, height)
+    previous = gdi32.SelectObject(memory_dc, bitmap)
+    try:
+        if not user32.PrintWindow(hwnd, memory_dc, 2):  # PW_RENDERFULLCONTENT
+            raise RuntimeError("PrintWindow failed")
+
+        class BitmapInfoHeader(ctypes.Structure):
+            _fields_ = [
+                ("biSize", wintypes.DWORD), ("biWidth", wintypes.LONG),
+                ("biHeight", wintypes.LONG), ("biPlanes", wintypes.WORD),
+                ("biBitCount", wintypes.WORD), ("biCompression", wintypes.DWORD),
+                ("biSizeImage", wintypes.DWORD),
+                ("biXPelsPerMeter", wintypes.LONG),
+                ("biYPelsPerMeter", wintypes.LONG),
+                ("biClrUsed", wintypes.DWORD),
+                ("biClrImportant", wintypes.DWORD),
+            ]
+
+        byte_count = width * height * 4
+        pixels = ctypes.create_string_buffer(byte_count)
+        header = BitmapInfoHeader(
+            ctypes.sizeof(BitmapInfoHeader), width, -height, 1, 32, 0,
+            byte_count, 0, 0, 0, 0,
+        )
+        rows = gdi32.GetDIBits(
+            memory_dc, bitmap, 0, height, pixels, ctypes.byref(header), 0
+        )
+        if rows != height:
+            raise RuntimeError(f"GetDIBits returned {rows}/{height} rows")
+        image = Image.frombuffer(
+            "RGB", (width, height), pixels, "raw", "BGRX", 0, 1
+        ).copy()
+    finally:
+        gdi32.SelectObject(memory_dc, previous)
+        gdi32.DeleteObject(bitmap)
+        gdi32.DeleteDC(memory_dc)
+        user32.ReleaseDC(hwnd, window_dc)
+
+    output.parent.mkdir(parents=True, exist_ok=True)
+    image.save(output, "PNG")
+
+
+def _tile_windows(handles: list[int]) -> None:
+    """Place four login windows in a 2x2 grid so screenshots cannot overlap."""
+    user32 = ctypes.windll.user32
+    screen_w = user32.GetSystemMetrics(0)
+    screen_h = user32.GetSystemMetrics(1)
+    cell_w = max(640, screen_w // 2)
+    cell_h = max(450, screen_h // 2)
+    for index, hwnd in enumerate(handles):
+        x = (index % 2) * cell_w
+        y = (index // 2) * cell_h
+        user32.ShowWindow(hwnd, 9)  # SW_RESTORE
+        user32.MoveWindow(hwnd, x, y, cell_w, cell_h, True)
+
+
+def _send_lark(recipient: str, text: str | None = None, image: Path | None = None) -> None:
+    """Send a notification through the locally installed Lark CLI."""
+    lark_env = {
+        key: value for key, value in os.environ.items()
+        if not key.upper().startswith("HERMES_")
+    }
+
+    def run_lark_cli(args: list[str], cwd: Path) -> None:
+        cwd.mkdir(parents=True, exist_ok=True)
+        command = shutil.which("lark-cli.cmd") or shutil.which("lark-cli") or "lark-cli.cmd"
+        completed = subprocess.run(
+            [command, *args],
+            cwd=cwd,
+            env=lark_env,
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+            check=False,
+        )
+        if completed.returncode:
+            detail = (completed.stderr or completed.stdout).strip()
+            raise RuntimeError(f"lark-cli failed ({completed.returncode}): {detail}")
+
+    if text:
+        # ensure_ascii keeps the Windows subprocess argv ASCII-only while the
+        # JSON parser restores the original Chinese text for Feishu.
+        content = json.dumps({"text": text}, ensure_ascii=True)
+        run_lark_cli(
+            ["im", "+messages-send", "--user-id", recipient,
+             "--as", "bot", "--content", content],
+            cwd=ANALYZER_DIR,
+        )
+    if image:
+        run_lark_cli(
+            ["im", "+messages-send", "--user-id", recipient,
+             "--as", "bot", "--image", f"./{image.name}"], cwd=image.parent,
+        )
+
+
+def _run_round_factory(args):
+    python = sys.executable
+    screenshot_root = PATHS.tmp_root / "qrcode/friday_relogin"
+
+    def run_round(names: list[str], attempt: int) -> dict[str, bool]:
+        started: dict[str, subprocess.Popen] = {}
+        stamp = dt.datetime.now().strftime("%Y%m%d_%H%M%S")
+        for name in names:
+            platform = PLATFORMS[name]
+            cmd = [python, *platform.command, "--login-timeout", str(args.round_timeout)]
+            started[name] = subprocess.Popen(cmd, cwd=PROJECT_DIR)
+
+        time.sleep(args.screenshot_delay)
+        title_windows = _login_windows()
+        windows = {
+            name: title_windows.get(name) or _window_for_process(process.pid)
+            for name, process in started.items()
+        }
+        _tile_windows([hwnd for hwnd in windows.values() if hwnd])
+        time.sleep(2)
+        for name, process in started.items():
+            platform = PLATFORMS[name]
+            hwnd = windows[name]
+            if not hwnd:
+                print(f"[{name}] no visible browser window for QR screenshot", flush=True)
+                continue
+            output = screenshot_root / f"{stamp}_{name}_round{attempt}.png"
+            try:
+                _capture_window(hwnd, output)
+                if not args.no_send:
+                    recipient = recipient_for_platform(name, args.recipient)
+                    _send_lark(
+                        recipient,
+                        text=(
+                            f"{platform.label} \u767b\u5f55\u4e8c\u7ef4\u7801"
+                            f"\uff08\u7b2c {attempt}/3 \u6b21\uff09\uff0c"
+                            "\u8bf7\u5728 3 \u5206\u949f\u5185\u626b\u7801\u3002"
+                        ),
+                    )
+                    _send_lark(recipient, image=output)
+                print(f"[{name}] QR screenshot: {output}", flush=True)
+            except Exception as exc:
+                print(f"[{name}] screenshot/send failed: {exc}", flush=True)
+
+        # Put the still-open login windows back into a compact layout after
+        # taking full-size screenshots one at a time.
+        _tile_windows([hwnd for hwnd in windows.values() if hwnd])
+
+        results: dict[str, bool] = {}
+        for name, process in started.items():
+            try:
+                rc = process.wait(timeout=args.round_timeout + 90)
+            except subprocess.TimeoutExpired:
+                _terminate_process_tree(process)
+                rc = 1
+            results[name] = rc == 0 and _valid_cookie(PLATFORMS[name])
+            print(f"[{name}] round={attempt} exit={rc} success={results[name]}", flush=True)
+        return results
+
+    return run_round
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Friday five-platform parallel QR relogin")
+    parser.add_argument(
+        "--recipient", default=None,
+        help="Override the configured recipients and send every platform to one open_id",
+    )
+    parser.add_argument("--max-attempts", type=int, default=3)
+    parser.add_argument("--round-timeout", type=int, default=300)
+    parser.add_argument("--screenshot-delay", type=int, default=60)
+    parser.add_argument("--no-send", action="store_true", help="Do not send Lark messages")
+    args = parser.parse_args()
+
+    status = run_with_retries(
+        list(PLATFORMS), max_attempts=args.max_attempts,
+        run_round=_run_round_factory(args),
+    )
+    failed = [name for name, ok in status.items() if not ok]
+    if failed and not args.no_send:
+        for recipient, names in group_platforms_by_recipient(failed, args.recipient).items():
+            labels = "\u3001".join(PLATFORMS[name].label for name in names)
+            _send_lark(
+                recipient,
+                text=(
+                    "\u4ee5\u4e0b\u5e73\u53f0\u8fde\u7eed 3 \u6b21\u672a\u626b\u7801\uff0c"
+                    "\u5df2\u6062\u590d\u5e76\u7ee7\u4f7f\u7528\u65e7 Cookie\uff1a"
+                    f"{labels}"
+                ),
+            )
+    print(f"Final status: {status}", flush=True)
+    return 1 if failed else 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/generate_creator_report.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/generate_creator_report.py
new file mode 100644
index 0000000..2b3e705
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/generate_creator_report.py
@@ -0,0 +1,910 @@
+#!/usr/bin/env python
+"""
+generate_creator_report.py — 达人合作数据筛选与报价分析报告
+=====================================================
+从 PostgreSQL 读取 cmt_cooperations + cmt_creators + cmt_styles，
+按指令要求生成完整的9章分析报告，输出为 Markdown 文件。
+
+用法:
+    python generate_creator_report.py
+    python generate_creator_report.py --style 宙斯   # 只分析指定款式
+    python generate_creator_report.py --output report.md
+"""
+import argparse
+import json
+import os
+import subprocess
+import sys
+from collections import defaultdict
+from datetime import datetime, timedelta
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import (
+    PATHS,
+    resolve_layer_output,
+)
+
+import psycopg
+from dotenv import load_dotenv
+
+BASE_DIR = PATHS.tools_root
+DATA_DIR = PATHS.exports_root
+load_dotenv(PATHS.state_root / "config/db.env")
+from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_db_config  # noqa: E402
+
+DB_CONFIG = get_db_config()
+
+LARK = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK):
+    LARK = "lark-cli.cmd"
+
+PLATFORM_CN = {
+    "xiaohongshu": "小红书",
+    "douyin": "抖音",
+    "bilibili": "B站",
+}
+
+
+def _log(msg: str) -> None:
+    safe = str(msg).encode("gbk", errors="replace").decode("gbk", errors="replace")
+    print(safe, flush=True)
+
+
+def infer_level(follower_num: int | None, platform: str) -> str:
+    if not follower_num:
+        return "未知"
+    if follower_num >= 500000:
+        return "头部"
+    elif follower_num >= 100000:
+        return "腰部"
+    elif follower_num >= 10000:
+        return "KOC"
+    else:
+        return "素人"
+
+
+def score_creator(stats: dict) -> str:
+    if not stats.get("has_data"):
+        return "C"
+    avg_eng_rate = stats.get("avg_engagement_rate", 0) or 0
+    avg_cpe = stats.get("avg_cpe", 0) or 0
+    avg_roi = stats.get("avg_roi", 0) or 0
+    exposure = stats.get("total_exposure", 0) or 0
+
+    score = 0
+    if avg_eng_rate >= 5: score += 3
+    elif avg_eng_rate >= 3: score += 2
+    elif avg_eng_rate >= 1: score += 1
+
+    if 0 < avg_cpe <= 2: score += 3
+    elif 0 < avg_cpe <= 5: score += 2
+    elif avg_cpe > 0: score += 1
+
+    if avg_roi >= 2: score += 3
+    elif avg_roi >= 1: score += 2
+    elif avg_roi > 0: score += 1
+
+    if exposure >= 500000: score += 2
+    elif exposure >= 100000: score += 1
+
+    if score >= 8: return "S"
+    elif score >= 5: return "A"
+    elif score >= 3: return "B"
+    else: return "C"
+
+
+def one_line_profile(stats: dict) -> str:
+    level = stats.get("level", "未知")
+    platform = stats.get("platform", "")
+    coop_count = stats.get("coop_count", 0)
+    total_exp = stats.get("total_exposure", 0) or 0
+    account_type = stats.get("account_type", "")
+
+    if total_exp >= 500000:
+        exp_desc = "爆款制造者"
+    elif total_exp >= 100000:
+        exp_desc = "稳定产出"
+    else:
+        exp_desc = "基础曝光"
+
+    type_desc = f"{account_type}类" if account_type else ""
+    return f"{level}{type_desc}达人，{PLATFORM_CN.get(platform, platform)}{coop_count}次合作，{exp_desc}"
+
+
+def generate_report(style_filter: str | None = None) -> str:
+    conn = psycopg.connect(**DB_CONFIG)
+    cur = conn.cursor()
+
+    where = "WHERE c.name IS NOT NULL AND c.name != ''"
+    if style_filter:
+        where += f" AND s.name = '{style_filter}'"
+
+    cur.execute(f"""
+    SELECT
+      c.name AS creator_name,
+      c.platform,
+      c.follower_count,
+      c.follower_count_num,
+      c.platform_id,
+      c.account_type,
+      c.creator_level,
+      c.homepage_url,
+      s.name AS style_name,
+      co.feishu_record_id,
+      co.cooperation_date,
+      co.cooperation_cost,
+      co.ad_spend,
+      co.content_direction,
+      co.note_title,
+      co.exposure_count,
+      co.engagement_count,
+      co.engagement_count_num,
+      co.data_performance,
+      co.cpm,
+      co.is_paid,
+      co.publish_time,
+      n.like_count AS note_like_count,
+      n.collect_count AS note_collect_count,
+      n.comment_count AS note_comment_count,
+      n.favorite_count AS note_favorite_count
+    FROM cmt_cooperations co
+    JOIN cmt_creators c ON c.id = co.creator_id
+    JOIN cmt_styles s ON s.id = co.style_id
+    LEFT JOIN cmt_notes n ON n.feishu_record_id = co.feishu_record_id
+    {where}
+    ORDER BY s.name, c.name
+    """)
+
+    cols = [desc[0] for desc in cur.description]
+    rows = [dict(zip(cols, row)) for row in cur.fetchall()]
+    cur.close()
+    conn.close()
+
+    if not rows:
+        return "# 达人合作数据筛选与报价分析报告\n\n无数据。"
+
+    # ─── Build per-creator stats ───
+    creator_stats = defaultdict(lambda: {
+        "coop_count": 0,
+        "platform": "",
+        "follower_count_num": None,
+        "follower_count": "",
+        "account_type": "",
+        "creator_level": "",
+        "styles": [],
+        "costs": [],
+        "cost_dates": [],       # (date, cost, style, content_direction) per coop
+        "exposures": [],
+        "engagements": [],
+        "engagement_nums": [],
+        "cpms": [],
+        "data_perfs": [],
+        "dates": [],
+        "has_data": False,
+        "note_engagements": [],
+        "note_exposures": [],
+        "content_directions": [],
+    })
+
+    for r in rows:
+        key = r["creator_name"]
+        s = creator_stats[key]
+        s["coop_count"] += 1
+        s["platform"] = r["platform"] or s["platform"]
+        if r["follower_count_num"] and not s["follower_count_num"]:
+            s["follower_count_num"] = r["follower_count_num"]
+        if r["follower_count"] and not s["follower_count"]:
+            s["follower_count"] = r["follower_count"]
+        if r["account_type"] and not s["account_type"]:
+            s["account_type"] = r["account_type"]
+        if r["creator_level"] and not s["creator_level"]:
+            s["creator_level"] = r["creator_level"]
+        if r["style_name"]:
+            s["styles"].append(r["style_name"])
+        if r["content_direction"]:
+            s["content_directions"].append(r["content_direction"])
+
+        # Per-cooperation detail for pricing analysis
+        cost = float(r["cooperation_cost"]) if r["cooperation_cost"] is not None else None
+        s["cost_dates"].append({
+            "date": str(r["cooperation_date"]) if r["cooperation_date"] else "",
+            "cost": cost,
+            "style": r["style_name"] or "",
+            "content_direction": r["content_direction"] or "",
+            "exposure": int(r["exposure_count"]) if r["exposure_count"] is not None else None,
+        })
+
+        if cost is not None:
+            s["costs"].append(cost)
+        if r["exposure_count"] is not None:
+            s["exposures"].append(int(r["exposure_count"]))
+        if r["engagement_count_num"] is not None:
+            s["engagement_nums"].append(int(r["engagement_count_num"]))
+        if r["cpm"] is not None:
+            s["cpms"].append(float(r["cpm"]))
+        if r["data_performance"]:
+            s["data_perfs"].append(r["data_performance"])
+        if r["cooperation_date"]:
+            s["dates"].append(str(r["cooperation_date"]))
+
+        if r["cooperation_cost"] is not None and r["exposure_count"] is not None:
+            s["has_data"] = True
+
+        note_eng = (r.get("note_like_count") or 0) + (r.get("note_collect_count") or 0) + (r.get("note_comment_count") or 0) + (r.get("note_favorite_count") or 0)
+        if note_eng > 0:
+            s["note_engagements"].append(note_eng)
+            note_exp = r["exposure_count"] if r["exposure_count"] else None
+            if note_exp:
+                s["note_exposures"].append(note_exp)
+
+    # Compute derived metrics
+    for key, s in creator_stats.items():
+        if not s["creator_level"] and s["follower_count_num"]:
+            s["creator_level"] = infer_level(s["follower_count_num"], s["platform"])
+        elif not s["creator_level"]:
+            s["creator_level"] = "未知"
+        s["level"] = s["creator_level"]
+
+        if s["note_engagements"] and s["note_exposures"] and len(s["note_engagements"]) == len(s["note_exposures"]):
+            rates = [e / max(x, 1) * 100 for e, x in zip(s["note_engagements"], s["note_exposures"])]
+            s["avg_engagement_rate"] = round(sum(rates) / len(rates), 2)
+        elif s["note_engagements"] and s["exposures"]:
+            total_eng = sum(s["note_engagements"])
+            total_exp = sum(s["exposures"])
+            s["avg_engagement_rate"] = round(total_eng / max(total_exp, 1) * 100, 2)
+        else:
+            s["avg_engagement_rate"] = None
+
+        if s["costs"] and s["note_engagements"]:
+            total_cost = sum(s["costs"])
+            total_eng = sum(s["note_engagements"])
+            s["avg_cpe"] = round(total_cost / max(total_eng, 1), 2)
+        elif s["costs"] and s["engagement_nums"]:
+            total_cost = sum(s["costs"])
+            total_eng = sum(s["engagement_nums"])
+            s["avg_cpe"] = round(total_cost / max(total_eng, 1), 2)
+        else:
+            s["avg_cpe"] = None
+
+        s["total_exposure"] = sum(s["exposures"]) if s["exposures"] else 0
+        s["avg_roi"] = None
+        s["score"] = score_creator(s)
+
+    # ─── Build per-style stats ───
+    style_stats = defaultdict(lambda: {"coop_count": 0, "creators": set(), "exposures": [], "costs": [], "top_creators": []})
+    for r in rows:
+        sn = r["style_name"]
+        ss = style_stats[sn]
+        ss["coop_count"] += 1
+        ss["creators"].add(r["creator_name"])
+        if r["exposure_count"] is not None:
+            ss["exposures"].append((r["creator_name"], int(r["exposure_count"])))
+        if r["cooperation_cost"] is not None:
+            ss["costs"].append(float(r["cooperation_cost"]))
+
+    # ─── Level average cost (for premium/discount analysis) ───
+    level_costs = defaultdict(list)
+    for name, s in creator_stats.items():
+        if s["costs"] and s["creator_level"] != "未知":
+            level_costs[s["creator_level"]].append(sum(s["costs"]) / len(s["costs"]))
+    level_avg_cost = {lv: round(sum(v)/len(v), 0) for lv, v in level_costs.items() if v}
+
+    # ─── Generate Report ───
+    lines = []
+    L = lines.append
+
+    L("# 达人合作数据筛选与报价分析报告（自动生成）")
+    L(f"\n> 生成时间：{datetime.now().strftime('%Y-%m-%d %H:%M')}")
+    if style_filter:
+        L(f"> 筛选款式：{style_filter}")
+    L("")
+
+    # ═══════════════════════════════════════════════════════════
+    # 1. 数据质量说明
+    # ═══════════════════════════════════════════════════════════
+    L("## 1. 数据质量说明")
+    L(f"- **总记录数**：{len(rows)}")
+    L(f"- **涉及达人总数**：{len(creator_stats)}")
+    L(f"- **涉及款式数**：{len(style_stats)}")
+
+    total = len(rows)
+    has_cost = sum(1 for r in rows if r["cooperation_cost"] is not None)
+    has_exposure = sum(1 for r in rows if r["exposure_count"] is not None)
+    has_engagement = sum(1 for r in rows if (r.get("note_like_count") or r.get("note_collect_count") or r.get("note_comment_count") or r.get("note_favorite_count")))
+    has_follower = sum(1 for r in rows if r["follower_count_num"] is not None)
+    has_platform = sum(1 for r in rows if r["platform"])
+    has_level = sum(1 for r in rows if r["creator_level"])
+    has_content_dir = sum(1 for r in rows if r["content_direction"])
+
+    L(f"\n| 字段 | 填充数 | 填充率 |")
+    L(f"|---|---|---|")
+    L(f"| 平台 | {has_platform} | {has_platform/total*100:.1f}% |")
+    L(f"| 粉丝数 | {has_follower} | {has_follower/total*100:.1f}% |")
+    L(f"| 达人层级 | {has_level} | {has_level/total*100:.1f}% |")
+    L(f"| 合作花费 | {has_cost} | {has_cost/total*100:.1f}% |")
+    L(f"| 曝光量 | {has_exposure} | {has_exposure/total*100:.1f}% |")
+    L(f"| 互动量 | {has_engagement} | {has_engagement/total*100:.1f}% |")
+    L(f"| 内容方向 | {has_content_dir} | {has_content_dir/total*100:.1f}% |")
+    L(f"| 进店UV | 0 | 0.0% |")
+    L(f"| 成交金额 | 0 | 0.0% |")
+    L(f"| ROI | 0 | 0.0% |")
+    L(f"| 评论品牌提及率 | 0 | 0.0% |")
+    L(f"| 差评或舆情风险 | 0 | 0.0% |")
+    L("")
+    L('> **说明**：电商转化数据（进店UV、成交金额、ROI）和评论分析数据（品牌提及率、舆情风险）暂未采集，对应分析章节将标注"数据不足"。达人层级根据粉丝数自动推断：>=50万=头部、>=10万=腰部、>=1万=KOC、<1万=素人。')
+    L("")
+
+    # ═══════════════════════════════════════════════════════════
+    # 2. 整体概览
+    # ═══════════════════════════════════════════════════════════
+    L("## 2. 整体概览")
+
+    plat_dist = defaultdict(int)
+    for r in rows:
+        p = r["platform"] or "未知"
+        plat_dist[p] += 1
+    L("\n### 各平台达人分布")
+    L("| 平台 | 合作记录数 | 达人数 |")
+    L("|---|---|---|")
+    plat_creators = defaultdict(set)
+    for r in rows:
+        plat_creators[r["platform"] or "未知"].add(r["creator_name"])
+    for p in sorted(plat_dist, key=plat_dist.get, reverse=True):
+        cn = PLATFORM_CN.get(p, p)
+        L(f"| {cn} | {plat_dist[p]} | {len(plat_creators[p])} |")
+    L("")
+
+    level_dist = defaultdict(int)
+    for s in creator_stats.values():
+        level_dist[s["creator_level"]] += 1
+    L("### 各层级分布")
+    L("| 层级 | 达人数 |")
+    L("|---|---|")
+    for lv in ["头部", "腰部", "KOC", "素人", "未知"]:
+        L(f"| {lv} | {level_dist.get(lv, 0)} |")
+    L("")
+
+    all_costs = [float(r["cooperation_cost"]) for r in rows if r["cooperation_cost"] is not None]
+    all_exposures = [int(r["exposure_count"]) for r in rows if r["exposure_count"] is not None]
+    all_engagements_note = []
+    for r in rows:
+        ne = (r.get("note_like_count") or 0) + (r.get("note_collect_count") or 0) + (r.get("note_comment_count") or 0) + (r.get("note_favorite_count") or 0)
+        if ne > 0:
+            all_engagements_note.append(ne)
+
+    total_cost = sum(all_costs) if all_costs else 0
+    total_exp = sum(all_exposures) if all_exposures else 0
+    avg_cpe = total_cost / sum(all_engagements_note) if all_costs and all_engagements_note else None
+
+    L("### 核心指标汇总")
+    L("| 指标 | 数值 |")
+    L("|---|---|")
+    L(f"| 总合作花费 | ¥{total_cost:,.0f} |")
+    L(f"| 总曝光量 | {total_exp:,} |")
+    L(f"| 总互动量 | {sum(all_engagements_note):,} |" if all_engagements_note else "| 总互动量 | 数据不足 |")
+    L(f"| 平均CPE | ¥{avg_cpe:.2f} |" if avg_cpe else "| 平均CPE | 数据不足 |")
+    L(f"| 平均ROI | 数据不足 |")
+    L(f"| 总成交金额 | 数据不足 |")
+    L("")
+
+    # ═══════════════════════════════════════════════════════════
+    # 3. 达人效能画像（按达人分组汇总）
+    # ═══════════════════════════════════════════════════════════
+    L("## 3. 达人效能画像（按达人分组汇总）")
+    L("")
+    L("> 按 S/A/B/C 层级分表展示，每个子表不超过飞书 2000 格限制。")
+    L("")
+
+    sorted_creators = sorted(creator_stats.items(),
+        key=lambda x: ({"S":0,"A":1,"B":2,"C":3}.get(x[1]["score"], 4), -(x[1]["total_exposure"] or 0)))
+
+    for tier in ["S", "A", "B", "C"]:
+        tier_rows = [(n, s) for n, s in sorted_creators if s["score"] == tier]
+        if not tier_rows:
+            continue
+
+        # Split large tiers to stay under Feishu 2000-cell limit
+        # 10 cols per row → max 200 rows per table
+        MAX_ROWS = 200
+        chunks = [tier_rows[i:i+MAX_ROWS] for i in range(0, len(tier_rows), MAX_ROWS)]
+        for ci, chunk in enumerate(chunks):
+            chunk_label = f"（{len(tier_rows)}人）" if len(chunks) == 1 else f"（{len(tier_rows)}人，第{ci+1}/{len(chunks)}批）"
+            L(f"### {tier}级达人{chunk_label}")
+            L("")
+            L("| 达人昵称 | 平台 | 粉丝数 | 层级 | 合作次数 | 合作单品 | 每次金额 | 平均CPE | 平均互动率 | 综合评分 |")
+            L("|---|---|---|---|---|---|---|---|---|---|")
+
+            for name, s in chunk:
+                plat_cn = PLATFORM_CN.get(s["platform"], s["platform"] or "未知")
+                follower_str = f"{s['follower_count_num']:,}" if s["follower_count_num"] else s.get("follower_count", "未知")
+                unique_styles = list(dict.fromkeys(s["styles"]))
+                styles_str = ", ".join(unique_styles[:3])
+                if len(unique_styles) > 3:
+                    styles_str += f"等{len(unique_styles)}款"
+                cost_entries = [e for e in s["cost_dates"] if e["cost"] is not None]
+                cost_entries.sort(key=lambda e: e["date"] or "")
+                costs_str = "→".join(f"¥{e['cost']:.0f}" for e in cost_entries) if cost_entries else "-"
+                avg_cpe_str = f"¥{s['avg_cpe']:.2f}" if s.get("avg_cpe") else "-"
+                avg_er_str = f"{s['avg_engagement_rate']:.1f}%" if s.get("avg_engagement_rate") else "-"
+                L(f"| {name} | {plat_cn} | {follower_str} | {s['creator_level']} | {s['coop_count']} | {styles_str} | {costs_str} | {avg_cpe_str} | {avg_er_str} | {s['score']} |")
+            L("")
+
+    # ═══════════════════════════════════════════════════════════
+    # 4. 达人报价差异分析
+    # ═══════════════════════════════════════════════════════════
+    L("## 4. 达人报价差异分析")
+    L("")
+
+    # 4a. 个体报价波动分析
+    L("### 个体报价波动分析")
+    L("")
+    L("> 针对同一达人不同次合作花费不同的情况，结合合作单品、内容形式、合作日期分析报价变化原因。")
+    L("")
+
+    multi_cost_creators = []
+    for name, s in creator_stats.items():
+        cost_entries = [e for e in s["cost_dates"] if e["cost"] is not None]
+        if len(cost_entries) >= 2:
+            costs = [e["cost"] for e in cost_entries]
+            if max(costs) != min(costs):
+                multi_cost_creators.append((name, s, cost_entries))
+
+    if multi_cost_creators:
+        L("| 达人昵称 | 层级 | 合作次数 | 报价区间 | 报价变化 | 波动归因分析 |")
+        L("|---|---|---|---|---|---|")
+        for name, s, entries in multi_cost_creators[:30]:
+            entries.sort(key=lambda e: e["date"] or "")
+            costs = [e["cost"] for e in entries]
+            cost_range = f"¥{min(costs):.0f}~¥{max(costs):.0f}"
+            cost_trend = "→".join(f"¥{e['cost']:.0f}" for e in entries)
+            # Analyze reasons
+            reasons = []
+            # Check content_direction variation
+            dirs = [e["content_direction"] for e in entries if e["content_direction"]]
+            if len(set(dirs)) > 1:
+                reasons.append(f"内容形式不同({', '.join(set(dirs))})")
+            # Check style variation
+            styles = [e["style"] for e in entries if e["style"]]
+            if len(set(styles)) > 1:
+                reasons.append(f"合作不同款式({len(set(styles))}款)")
+            # Check date proximity to promotion periods
+            dates = [e["date"] for e in entries if e["date"]]
+            promo_months = ["06", "11", "12"]  # 618, 双11, 双12
+            if any(d[5:7] in promo_months for d in dates):
+                reasons.append("含大促节点")
+            if not reasons:
+                reasons.append("报价自然波动")
+            L(f"| {name} | {s['creator_level']} | {len(entries)} | {cost_range} | {cost_trend} | {'；'.join(reasons)} |")
+        L("")
+    else:
+        L("数据不足，无法分析个体报价波动。")
+        L("")
+
+    # 4b. 同层级溢价/折价分析
+    L("### 同层级溢价/折价分析")
+    L("")
+    L("> 对比同层级、同平台达人的平均合作花费，识别溢价（性价比低）和折价（高性价比洼地）达人。")
+    L("")
+
+    if level_avg_cost:
+        L("| 达人昵称 | 层级 | 平台 | 个人均价 | 层级均价 | 溢折率 | 判定 |")
+        L("|---|---|---|---|---|---|---|")
+        premium_list = []
+        for name, s in creator_stats.items():
+            if not s["costs"] or s["creator_level"] == "未知":
+                continue
+            personal_avg = sum(s["costs"]) / len(s["costs"])
+            lv_avg = level_avg_cost.get(s["creator_level"])
+            if not lv_avg or lv_avg == 0:
+                continue
+            ratio = (personal_avg - lv_avg) / lv_avg * 100
+            if ratio > 30:
+                tag = "溢价"
+            elif ratio < -30:
+                tag = "折价洼地"
+            else:
+                tag = "合理"
+            premium_list.append((name, s, personal_avg, lv_avg, ratio, tag))
+
+        premium_list.sort(key=lambda x: -x[4])
+        for name, s, p_avg, l_avg, ratio, tag in premium_list[:20]:
+            plat_cn = PLATFORM_CN.get(s["platform"], s["platform"])
+            L(f"| {name} | {s['creator_level']} | {plat_cn} | ¥{p_avg:.0f} | ¥{l_avg:.0f} | {ratio:+.0f}% | {tag} |")
+        # Also show top discount
+        discount_list = [x for x in premium_list if x[5] == "折价洼地"]
+        for name, s, p_avg, l_avg, ratio, tag in discount_list[:10]:
+            plat_cn = PLATFORM_CN.get(s["platform"], s["platform"])
+            L(f"| {name} | {s['creator_level']} | {plat_cn} | ¥{p_avg:.0f} | ¥{l_avg:.0f} | {ratio:+.0f}% | {tag} |")
+        L("")
+    else:
+        L("数据不足，无法进行同层级溢价/折价分析。")
+        L("")
+
+    # ═══════════════════════════════════════════════════════════
+    # 5. 达人分层与分类
+    # ═══════════════════════════════════════════════════════════
+    L("## 5. 达人分层与分类")
+
+    tiers = {"S": [], "A": [], "B": [], "C": []}
+    for name, s in creator_stats.items():
+        tiers[s["score"]].append(name)
+
+    L("\n### 分层矩阵")
+    for tier in ["S", "A", "B", "C"]:
+        names = tiers[tier]
+        if names:
+            names_str = ', '.join(names[:20]) + ('...' if len(names) > 20 else '')
+            L(f"**{tier}级**（{len(names)}人）：{names_str}")
+        else:
+            L(f"**{tier}级**（0人）：无")
+
+    L("\n### 角色分类")
+    voice_type = sorted(creator_stats.items(), key=lambda x: -(x[1]["total_exposure"] or 0))[:10]
+    voice_names = [n for n, s in voice_type if s["total_exposure"]]
+    L(f"**声量型**（高曝光）：{', '.join(voice_names)}")
+
+    harvest_type = sorted(creator_stats.items(), key=lambda x: x[1].get("avg_cpe", 9999) or 9999)[:10]
+    L(f"**收割型**（低成本高效）：{', '.join(n for n, s in harvest_type if s.get('avg_cpe') and s['avg_cpe'] > 0)}")
+
+    volume_type = [(n, s) for n, s in creator_stats.items() if s["costs"]]
+    volume_type.sort(key=lambda x: sum(x[1]["costs"]) / len(x[1]["costs"]))
+    L(f"**铺量型**（低成本）：{', '.join(n for n, _ in volume_type[:10])}")
+
+    # 共创型: high engagement rate + content variety
+    creative_type = [(n, s) for n, s in creator_stats.items() if s.get("avg_engagement_rate") and s["avg_engagement_rate"] > 0 and len(set(s["content_directions"])) >= 2]
+    creative_type.sort(key=lambda x: -x[1].get("avg_engagement_rate", 0))
+    L(f"**共创型**（内容创意强，多方向+高互动）：{', '.join(n for n, _ in creative_type[:10])}")
+
+    L("")
+
+    # ═══════════════════════════════════════════════════════════
+    # 6. 合作单品交叉分析
+    # ═══════════════════════════════════════════════════════════
+    L("## 6. 合作单品交叉分析")
+    L("")
+    L("| 款式 | 合作达人数 | 平均曝光 | 平均花费 | Top3达人 | 单品最佳搭档 |")
+    L("|---|---|---|---|---|---|")
+
+    for sn in sorted(style_stats.keys()):
+        ss = style_stats[sn]
+        avg_exp = sum(e for _, e in ss["exposures"]) / len(ss["exposures"]) if ss["exposures"] else 0
+        avg_cost = sum(ss["costs"]) / len(ss["costs"]) if ss["costs"] else 0
+        top3 = sorted(ss["exposures"], key=lambda x: -x[1])[:3]
+        top3_str = ", ".join(f"{n}({e:,})" for n, e in top3)
+        # Best partner: highest exposure creator
+        best = top3[0][0] if top3 else "-"
+        L(f"| {sn} | {len(ss['creators'])} | {avg_exp:,.0f} | ¥{avg_cost:,.0f} | {top3_str} | {best} |")
+    L("")
+
+    # ═══════════════════════════════════════════════════════════
+    # 7. 趋势与风险评估
+    # ═══════════════════════════════════════════════════════════
+    L("## 7. 趋势与风险评估")
+    L("")
+
+    multi_coop = {n: s for n, s in creator_stats.items() if s["coop_count"] >= 2 and len(s["exposures"]) >= 2}
+
+    rising = []
+    declining = []
+    for n, s in multi_coop.items():
+        if len(s["exposures"]) >= 2:
+            first_half = s["exposures"][:len(s["exposures"])//2]
+            second_half = s["exposures"][len(s["exposures"])//2:]
+            avg_first = sum(first_half) / len(first_half)
+            avg_second = sum(second_half) / len(second_half)
+            if avg_second > avg_first * 1.2:
+                rising.append(n)
+            elif avg_second < avg_first * 0.8:
+                declining.append(n)
+
+    L(f"### 表现持续上升（{len(rising)}人）")
+    if rising:
+        L("建议加大投入。")
+        for n in rising[:20]:
+            s = creator_stats[n]
+            L(f"- **{n}**：曝光量从前期上升至后期，{s['coop_count']}次合作数据持续改善")
+    else:
+        L("数据不足以判断趋势")
+    L("")
+
+    L(f"### 表现下滑（{len(declining)}人）")
+    if declining:
+        L("建议减少合作或调整Brief。")
+        for n in declining[:20]:
+            s = creator_stats[n]
+            L(f"- **{n}**：曝光量从前期下降至后期，建议关注内容质量或平台算法变化")
+    else:
+        L("数据不足以判断趋势")
+    L("")
+
+    L("### 风险达人")
+    risk_tags = {"曝光限流": [], "效果一般": [], "挂车不出单": [], "内容卡审": []}
+    for n, s in creator_stats.items():
+        for dp in s["data_perfs"]:
+            for tag in dp.split(","):
+                tag = tag.strip()
+                if tag in risk_tags:
+                    risk_tags[tag].append(n)
+    has_risk = False
+    for tag, names in risk_tags.items():
+        if names:
+            has_risk = True
+            L(f"- **{tag}**：{', '.join(set(names))}")
+    if not has_risk:
+        L("- 暂无风险标记")
+    L("")
+
+    # ═══════════════════════════════════════════════════════════
+    # 8. 筛选推荐（根据目标场景）
+    # ═══════════════════════════════════════════════════════════
+    L("## 8. 筛选推荐（根据目标场景）")
+    L("")
+
+    L("### 场景一：新品上市（重点曝光+种草）")
+    s_a_creators = [(n, s) for n, s in creator_stats.items() if s["score"] in ("S", "A") and s["total_exposure"]]
+    s_a_creators.sort(key=lambda x: -x[1]["total_exposure"])
+    if s_a_creators:
+        L("| 推荐达人 | 平台 | 粉丝数 | 累计曝光 | 综合评分 | 推荐理由 | 预估CPE |")
+        L("|---|---|---|---|---|---|---|")
+        for n, s in s_a_creators[:5]:
+            plat_cn = PLATFORM_CN.get(s["platform"], s["platform"])
+            follower_str = f"{s['follower_count_num']:,}" if s["follower_count_num"] else "未知"
+            reason = f"曝光量{s['total_exposure']:,}，{s['coop_count']}次合作经验"
+            est_cpe = f"¥{s['avg_cpe']:.2f}" if s.get("avg_cpe") else "待估"
+            L(f"| {n} | {plat_cn} | {follower_str} | {s['total_exposure']:,} | {s['score']} | {reason} | {est_cpe} |")
+    else:
+        L("数据不足，无法推荐")
+    L("")
+
+    L("### 场景二：大促收割（重点转化）")
+    L("> 进店UV、成交金额、ROI 数据暂未采集，以下基于低成本高效（低CPE+高互动率）推荐：")
+    low_cpe = [(n, s) for n, s in creator_stats.items() if s.get("avg_cpe") and s["avg_cpe"] > 0]
+    low_cpe.sort(key=lambda x: x[1]["avg_cpe"])
+    if low_cpe:
+        L("| 推荐达人 | 平台 | 平均CPE | 平均互动率 | 综合评分 | 推荐理由 |")
+        L("|---|---|---|---|---|---|")
+        for n, s in low_cpe[:5]:
+            plat_cn = PLATFORM_CN.get(s["platform"], s["platform"])
+            er = f"{s['avg_engagement_rate']:.1f}%" if s.get("avg_engagement_rate") else "-"
+            L(f"| {n} | {plat_cn} | ¥{s['avg_cpe']:.2f} | {er} | {s['score']} | CPE低，互动率高 |")
+    else:
+        L("数据不足，无法推荐")
+    L("")
+
+    L("### 场景三：日常口碑维护")
+    koc_creators = [(n, s) for n, s in creator_stats.items() if s["creator_level"] in ("KOC", "素人") and s["total_exposure"]]
+    koc_creators.sort(key=lambda x: -(x[1].get("avg_engagement_rate") or 0))
+    if koc_creators:
+        L("| 推荐达人 | 平台 | 粉丝数 | 平均互动率 | 合作次数 | 推荐理由 |")
+        L("|---|---|---|---|---|---|")
+        for n, s in koc_creators[:5]:
+            plat_cn = PLATFORM_CN.get(s["platform"], s["platform"])
+            follower_str = f"{s['follower_count_num']:,}" if s["follower_count_num"] else "未知"
+            er = f"{s['avg_engagement_rate']:.1f}%" if s.get("avg_engagement_rate") else "-"
+            L(f"| {n} | {plat_cn} | {follower_str} | {er} | {s['coop_count']} | 低成本高互动KOC |")
+    else:
+        L("数据不足，无法推荐")
+    L("")
+
+    # ═══════════════════════════════════════════════════════════
+    # 9. 优化建议与下一步行动
+    # ═══════════════════════════════════════════════════════════
+    L("## 9. 优化建议与下一步行动")
+    L("")
+
+    for tier, action in [
+        ("S", "建议签长期协议或独家合作，锁定优质达人资源"),
+        ("A", "增加合作频次或尝试新款式，挖掘更大潜力"),
+        ("B", "优化Brief或换内容形式测试，寻找突破口"),
+        ("C", "暂停合作或淘汰，将预算转移至更高层级达人"),
+    ]:
+        names = tiers[tier]
+        L(f"- **{tier}级达人**（{len(names)}人）：{action}")
+    L("")
+
+    # 报价优化建议
+    L("### 报价优化建议")
+    L("")
+    if multi_cost_creators:
+        L("基于第4章报价差异分析，给出以下议价策略：")
+        L("")
+        # Find creators with rising costs
+        for name, s, entries in multi_cost_creators[:10]:
+            entries.sort(key=lambda e: e["date"] or "")
+            costs = [e["cost"] for e in entries]
+            if costs[-1] > costs[0]:
+                L(f"- **{name}**：报价呈上升趋势（¥{costs[0]:.0f}→¥{costs[-1]:.0f}），建议锁定当前报价或签长期协议避免继续涨价")
+            elif costs[-1] < costs[0]:
+                L(f"- **{name}**：报价有下降空间（¥{costs[0]:.0f}→¥{costs[-1]:.0f}），可尝试进一步议价")
+    if level_avg_cost:
+        L("")
+        L("**层级预算分配建议**：")
+        for lv in ["头部", "腰部", "KOC", "素人"]:
+            avg = level_avg_cost.get(lv)
+            if avg:
+                L(f"- {lv}达人均价¥{avg:.0f}，建议单次合作预算控制在¥{avg*0.8:.0f}~¥{avg*1.2:.0f}")
+    L("")
+
+    L("---")
+    L(f"*报告由 generate_creator_report.py 自动生成 · {datetime.now().strftime('%Y-%m-%d %H:%M')}*")
+
+    return "\n".join(lines), rows, creator_stats, style_stats
+
+
+def _compute_last_month():
+    today = datetime.now().date()
+    first_of_this_month = today.replace(day=1)
+    month_end = first_of_this_month - timedelta(days=1)
+    month_start = month_end.replace(day=1)
+    return month_start, month_end
+
+
+def create_feishu_doc(report_md: str, month_start: str, month_end: str) -> str | None:
+    """创建飞书文档,返回 doc URL;失败返回 None"""
+    doc_title = f"达人合作数据筛选与报价分析报告（{month_start}~{month_end[5:]}）"
+    md_file = PATHS.tmp_root / "creator_report_doc.md"
+    md_file.parent.mkdir(parents=True, exist_ok=True)
+    md_file.write_text(report_md, encoding="utf-8")
+
+    cmd = [LARK, "docs", "+create",
+           "--as", "user", "--format", "json",
+           "--title", doc_title,
+           "--doc-format", "markdown",
+           "--content", f"@{md_file.name}"]
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, env=env,
+                          encoding="utf-8", errors="replace")
+    text = proc.stdout or proc.stderr
+    start = text.find("{")
+    if start < 0:
+        _log(f"  [ERR] 创建飞书文档失败: {text[:300]}")
+        return None
+    resp = json.loads(text[start:])
+    if not resp.get("ok"):
+        _log(f"  [ERR] 创建飞书文档失败: {resp.get('error')}")
+        return None
+    url = resp.get("data", {}).get("document", {}).get("url", "")
+    if resp.get("warnings"):
+        _log(f"  [WARN] {resp['warnings']}")
+    return url
+
+
+def save_to_db(report_md: str, rows: list, creator_stats: dict,
+               style_stats: dict, month_start: str, month_end: str,
+               doc_url: str | None, status: str) -> bool:
+    """把月度达人报告结果写入 cmt_creator_report"""
+    time_label = f"{month_start}~{month_end[5:]}"
+
+    total_cooperations = len(rows)
+    total_creators = len(creator_stats)
+    total_styles = len(style_stats)
+
+    all_costs = [float(r["cooperation_cost"]) for r in rows if r["cooperation_cost"] is not None]
+    all_exposures = [int(r["exposure_count"]) for r in rows if r["exposure_count"] is not None]
+    all_engagements = []
+    for r in rows:
+        ne = (r.get("note_like_count") or 0) + (r.get("note_collect_count") or 0) + (r.get("note_comment_count") or 0) + (r.get("note_favorite_count") or 0)
+        if ne > 0:
+            all_engagements.append(ne)
+
+    total_cost = sum(all_costs) if all_costs else None
+    total_exposure = sum(all_exposures) if all_exposures else None
+    total_engagement = sum(all_engagements) if all_engagements else None
+    avg_cpe = (total_cost / total_engagement) if total_cost and total_engagement else None
+
+    # Level distribution
+    level_dist = defaultdict(int)
+    for s in creator_stats.values():
+        level_dist[s["creator_level"]] += 1
+    # Platform distribution
+    plat_dist = defaultdict(int)
+    for r in rows:
+        p = r["platform"] or "未知"
+        plat_dist[p] += 1
+    plat_dist_creators = defaultdict(set)
+    for r in rows:
+        plat_dist_creators[r["platform"] or "未知"].add(r["creator_name"])
+    platform_dist = {p: len(plat_dist_creators[p]) for p in plat_dist_creators}
+    # Score distribution
+    score_dist = defaultdict(int)
+    for s in creator_stats.values():
+        score_dist[s["score"]] += 1
+
+    doc_title = f"达人合作数据筛选与报价分析报告（{time_label}）"
+
+    try:
+        conn = psycopg.connect(**DB_CONFIG)
+        cur = conn.cursor()
+        cur.execute("""
+            INSERT INTO cmt_creator_report
+                (month_start, month_end, time_label,
+                 total_cooperations, total_creators, total_styles,
+                 total_cost, total_exposure, total_engagement, avg_cpe,
+                 level_distribution, platform_distribution, score_distribution,
+                 doc_url, doc_title, report_chars, status)
+            VALUES
+                (%s, %s, %s,
+                 %s, %s, %s,
+                 %s, %s, %s, %s,
+                 %s, %s, %s,
+                 %s, %s, %s, %s)
+            ON CONFLICT (month_start) DO UPDATE SET
+                total_cooperations = EXCLUDED.total_cooperations,
+                total_creators = EXCLUDED.total_creators,
+                total_styles = EXCLUDED.total_styles,
+                total_cost = EXCLUDED.total_cost,
+                total_exposure = EXCLUDED.total_exposure,
+                total_engagement = EXCLUDED.total_engagement,
+                avg_cpe = EXCLUDED.avg_cpe,
+                level_distribution = EXCLUDED.level_distribution,
+                platform_distribution = EXCLUDED.platform_distribution,
+                score_distribution = EXCLUDED.score_distribution,
+                doc_url = EXCLUDED.doc_url,
+                doc_title = EXCLUDED.doc_title,
+                report_chars = EXCLUDED.report_chars,
+                status = EXCLUDED.status,
+                updated_at = CURRENT_TIMESTAMP
+        """, (month_start, month_end, time_label,
+              total_cooperations, total_creators, total_styles,
+              total_cost, total_exposure, total_engagement, avg_cpe,
+              json.dumps(dict(level_dist), ensure_ascii=False),
+              json.dumps(platform_dist, ensure_ascii=False),
+              json.dumps(dict(score_dist), ensure_ascii=False),
+              doc_url, doc_title, len(report_md), status))
+        conn.commit()
+        conn.close()
+        _log(f"  已入库 cmt_creator_report (month={month_start})")
+        return True
+    except Exception as exc:
+        _log(f"  [DB ERR] 入库失败: {exc}")
+        return False
+
+
+def main():
+    parser = argparse.ArgumentParser(description="生成达人合作数据筛选与报价分析报告")
+    parser.add_argument("--style", type=str, help="只分析指定款式 (如 '宙斯')")
+    parser.add_argument("--output", type=str, default="", help="输出文件路径 (默认 data/reports/ 下自动命名)")
+    parser.add_argument("--dry-run", action="store_true", help="不写飞书/PG")
+    args = parser.parse_args()
+
+    month_start, month_end = _compute_last_month()
+    month_start_str = month_start.strftime("%Y-%m-%d")
+    month_end_str = month_end.strftime("%Y-%m-%d")
+
+    _log(f"达人月度分析报告: {month_start_str}~{month_end_str}")
+
+    # generate_report returns (report_text, rows, creator_stats, style_stats)
+    report, rows, creator_stats, style_stats = generate_report(style_filter=args.style)
+
+    if args.output:
+        out_path = resolve_layer_output(
+            args.output, layer_root=PATHS.exports_root, field="--output"
+        )
+    else:
+        out_dir = DATA_DIR / "reports"
+        out_dir.mkdir(exist_ok=True)
+        ts = datetime.now().strftime("%Y%m%d_%H%M%S")
+        style_tag = f"_{args.style}" if args.style else "_all"
+        out_path = out_dir / f"creator_report{style_tag}_{ts}.md"
+
+    out_path.write_text(report, encoding="utf-8")
+    _log(f"报告已生成: {out_path} ({len(report)} 字符)")
+
+    if args.dry_run:
+        _log("[DRY-RUN] 跳过飞书文档创建和入库")
+        return
+
+    # 创建飞书文档
+    doc_url = create_feishu_doc(report, month_start_str, month_end_str)
+    if doc_url:
+        _log(f"飞书文档: {doc_url}")
+    else:
+        _log("飞书文档创建失败")
+
+    # 入库
+    status = "ok" if doc_url else "doc_failed"
+    save_to_db(report, rows, creator_stats, style_stats,
+               month_start_str, month_end_str, doc_url, status)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/init_db.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/init_db.py
new file mode 100644
index 0000000..3c0ba8c
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/init_db.py
@@ -0,0 +1,29 @@
+"""Initialize gyxx_super_data tables from schema_gyxx_super_data.sql."""
+import re
+import sys
+from pathlib import Path
+
+# Path setup
+_TOOLS_DIR = Path(__file__).resolve().parent
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools import db
+
+
+def main() -> int:
+    sql_path = _TOOLS_DIR / "schema_gyxx_super_data.sql"
+    sql = sql_path.read_text(encoding="utf-8")
+
+    # Remove psql meta-commands
+    sql = re.sub(r"^\\c .*?$", "", sql, flags=re.MULTILINE)
+    sql = re.sub(r"^\\.*$", "", sql, flags=re.MULTILINE)
+
+    with db.get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(sql)
+        conn.commit()
+    print("Schema initialized successfully.")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/kill_project_chrome.ps1 b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/kill_project_chrome.ps1
new file mode 100644
index 0000000..9788f3a
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/kill_project_chrome.ps1
@@ -0,0 +1,7 @@
+$moduleRoot = (Resolve-Path (Join-Path $PSScriptRoot '..\..')).Path
+$dataRoot = if ($env:GYXX_DATA_ROOT) { $env:GYXX_DATA_ROOT } else { Join-Path $moduleRoot 'var' }
+$profileRoot = Join-Path $dataRoot 'state\content_marketing\browser-profiles'
+Get-WmiObject Win32_Process | Where-Object { $_.Name -eq 'chrome.exe' -and $_.CommandLine -like "*$profileRoot*" } | ForEach-Object { Write-Host "Killing PID $($_.ProcessId)"; Stop-Process -Id $_.ProcessId -Force -ErrorAction SilentlyContinue }
+Start-Sleep -Seconds 3
+$procs = Get-WmiObject Win32_Process | Where-Object { $_.Name -eq 'chrome.exe' -and $_.CommandLine -like "*$profileRoot*" }
+if ($procs) { Write-Host 'Remaining project chrome processes:'; $procs | Select-Object ProcessId,CommandLine } else { Write-Host 'No project chrome processes remaining' }
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/list_project_chrome.ps1 b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/list_project_chrome.ps1
new file mode 100644
index 0000000..4e4856b
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/list_project_chrome.ps1
@@ -0,0 +1,4 @@
+$moduleRoot = (Resolve-Path (Join-Path $PSScriptRoot '..\..')).Path
+$dataRoot = if ($env:GYXX_DATA_ROOT) { $env:GYXX_DATA_ROOT } else { Join-Path $moduleRoot 'var' }
+$profileRoot = Join-Path $dataRoot 'state\content_marketing\browser-profiles'
+Get-WmiObject Win32_Process | Where-Object { $_.Name -eq 'chrome.exe' -and $_.CommandLine -like "*$profileRoot*" } | Select-Object ProcessId, ParentProcessId, @{Name='CmdStart';Expression={$_.CommandLine.Substring(0, [Math]::Min(200, $_.CommandLine.Length))}}
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/001_creators_and_cooperations.sql b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/001_creators_and_cooperations.sql
new file mode 100644
index 0000000..59a94dc
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/001_creators_and_cooperations.sql
@@ -0,0 +1,132 @@
+-- Migration 001: 扩展 cmt_creators 达人属性 + 新建 cmt_cooperations 合作记录表
+-- 日期: 2026-07-10
+-- 说明: 从飞书款式合作表同步达人信息和合作数据到数据库
+
+-- ============================================================
+-- 1. 扩展 cmt_creators：加达人属性字段
+-- ============================================================
+
+-- 加平台字段
+ALTER TABLE cmt_creators ADD COLUMN IF NOT EXISTS platform VARCHAR(32) DEFAULT '';
+COMMENT ON COLUMN cmt_creators.platform IS '主平台: xiaohongshu/douyin/bilibili';
+
+-- 加平台达人ID
+ALTER TABLE cmt_creators ADD COLUMN IF NOT EXISTS platform_id VARCHAR(128) DEFAULT '';
+COMMENT ON COLUMN cmt_creators.platform_id IS '平台达人ID (如B站UID、抖音sec_uid等)';
+
+-- 加粉丝数 (保留原始文本，因为飞书数据是"3.7万"这种格式)
+ALTER TABLE cmt_creators ADD COLUMN IF NOT EXISTS follower_count VARCHAR(64) DEFAULT '';
+COMMENT ON COLUMN cmt_creators.follower_count IS '达人粉丝数(原始文本如3.7万/12.4w)';
+
+-- 加粉丝数(数字解析后)
+ALTER TABLE cmt_creators ADD COLUMN IF NOT EXISTS follower_count_num BIGINT DEFAULT NULL;
+COMMENT ON COLUMN cmt_creators.follower_count_num IS '达人粉丝数(数字解析后，3.7万=37000)';
+
+-- 加达人主页链接
+ALTER TABLE cmt_creators ADD COLUMN IF NOT EXISTS homepage_url TEXT DEFAULT '';
+COMMENT ON COLUMN cmt_creators.homepage_url IS '达人主页链接';
+
+-- 加微信号
+ALTER TABLE cmt_creators ADD COLUMN IF NOT EXISTS wechat VARCHAR(128) DEFAULT '';
+COMMENT ON COLUMN cmt_creators.wechat IS '微信号';
+
+-- 加达人类型 (对应飞书"投放账号类型")
+ALTER TABLE cmt_creators ADD COLUMN IF NOT EXISTS account_type VARCHAR(64) DEFAULT '';
+COMMENT ON COLUMN cmt_creators.account_type IS '达人类型: 数码/vlog/好物/穿搭/摄影等';
+
+-- 加达人层级
+ALTER TABLE cmt_creators ADD COLUMN IF NOT EXISTS creator_level VARCHAR(32) DEFAULT '';
+COMMENT ON COLUMN cmt_creators.creator_level IS '达人层级: 头部/腰部/KOC/素人';
+
+-- 修改唯一约束: name 单列唯一 → (name, platform) 联合唯一
+-- 因为同一达人可能跨平台存在(如BeiBei贝贝同时在小红书和抖音)
+-- 先删除旧约束
+ALTER TABLE cmt_creators DROP CONSTRAINT IF EXISTS cmt_creators_name_key;
+-- 添加新约束
+ALTER TABLE cmt_creators ADD CONSTRAINT cmt_creators_name_platform_key UNIQUE (name, platform);
+
+-- 添加索引
+CREATE INDEX IF NOT EXISTS idx_cmt_creators_platform ON cmt_creators (platform);
+CREATE INDEX IF NOT EXISTS idx_cmt_creators_platform_id ON cmt_creators (platform_id);
+
+-- ============================================================
+-- 2. 新建 cmt_cooperations：合作记录表
+-- ============================================================
+-- 每条记录 = 一个达人在一个款式下的一次合作
+-- 飞书款式合作表的每一行 → 此表一条记录
+
+CREATE TABLE IF NOT EXISTS cmt_cooperations (
+  id                BIGSERIAL PRIMARY KEY,
+  style_id          BIGINT NOT NULL REFERENCES cmt_styles(id) ON DELETE CASCADE,
+  creator_id        BIGINT NOT NULL REFERENCES cmt_creators(id) ON DELETE CASCADE,
+  feishu_record_id  VARCHAR(64) DEFAULT '',     -- 飞书行 record_id, 用于回写
+  -- 合作基本信息
+  cooperation_date  DATE DEFAULT NULL,           -- 合作日期 (飞书"制单日期")
+  cooperation_cost  NUMERIC(12,2) DEFAULT NULL,  -- 合作花费 (飞书"合作费用")
+  ad_spend          NUMERIC(12,2) DEFAULT NULL,  -- 投流金额 (飞书"投流金额")
+  -- 内容信息
+  content_direction VARCHAR(255) DEFAULT '',     -- 内容方向 (飞书"发布笔记内容方向")
+  content_format    VARCHAR(64) DEFAULT '',      -- 内容形式: 单品种草/测评合集/场景植入/开箱Vlog/穿搭变装等
+  note_title        VARCHAR(512) DEFAULT '',     -- 发布笔记标题
+  note_url          TEXT DEFAULT '',             -- 发布笔记链接
+  publish_time      TIMESTAMPTZ DEFAULT NULL,    -- 发布时间
+  -- 数据表现
+  exposure_count    BIGINT DEFAULT NULL,         -- 曝光量 (飞书"曝光量(自动抓取)")
+  engagement_count  VARCHAR(64) DEFAULT '',      -- 互动赞藏数 (原始文本如"68.5万")
+  engagement_count_num BIGINT DEFAULT NULL,      -- 互动赞藏数(数字解析后)
+  data_performance  VARCHAR(255) DEFAULT '',     -- 数据表现 (逗号分隔: 曝光限流,数据良好)
+  cpm               NUMERIC(10,4) DEFAULT NULL,  -- CPM (飞书公式字段)
+  -- 物流/财务
+  tracking_number   VARCHAR(128) DEFAULT '',     -- 快递单号
+  is_paid           BOOLEAN DEFAULT FALSE,       -- 是否结款
+  is_new_direction  BOOLEAN DEFAULT FALSE,       -- 是否新方向验证
+  -- 电商效果 (飞书暂无,预留)
+  shop_uv           BIGINT DEFAULT NULL,         -- 进店UV
+  transaction_amount NUMERIC(12,2) DEFAULT NULL,  -- 成交金额(元)
+  roi               NUMERIC(10,4) DEFAULT NULL,  -- ROI
+  cpe               NUMERIC(10,4) DEFAULT NULL,  -- CPE (花费/互动量)
+  engagement_rate   NUMERIC(8,4) DEFAULT NULL,   -- 互动率 (互动量/曝光量)
+  -- 评论分析 (飞书暂无,预留)
+  brand_mention_rate NUMERIC(8,4) DEFAULT NULL,  -- 评论品牌提及率
+  risk_flag         VARCHAR(255) DEFAULT '',     -- 差评或舆情风险
+  --
+  created_at        TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  updated_at        TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  CONSTRAINT uniq_cooperation_feishu_record UNIQUE (feishu_record_id)
+);
+
+COMMENT ON TABLE cmt_cooperations IS '合作记录表(每个达人+款式的一次合作)';
+COMMENT ON COLUMN cmt_cooperations.cooperation_date IS '合作日期(飞书制单日期)';
+COMMENT ON COLUMN cmt_cooperations.cooperation_cost IS '合作花费(元)';
+COMMENT ON COLUMN cmt_cooperations.ad_spend IS '投流金额(元)';
+COMMENT ON COLUMN cmt_cooperations.content_direction IS '发布笔记内容方向(飞书原始值)';
+COMMENT ON COLUMN cmt_cooperations.content_format IS '内容形式: 单品种草/测评合集/场景植入/开箱Vlog等';
+COMMENT ON COLUMN cmt_cooperations.exposure_count IS '曝光量(飞书自动抓取)';
+COMMENT ON COLUMN cmt_cooperations.engagement_count IS '互动赞藏数(原始文本如68.5万)';
+COMMENT ON COLUMN cmt_cooperations.data_performance IS '数据表现(逗号分隔: 曝光限流,数据良好,表现优秀等)';
+COMMENT ON COLUMN cmt_cooperations.shop_uv IS '进店UV(预留,飞书暂无)';
+COMMENT ON COLUMN cmt_cooperations.transaction_amount IS '成交金额(元,预留)';
+COMMENT ON COLUMN cmt_cooperations.roi IS 'ROI=成交金额/花费(预留)';
+COMMENT ON COLUMN cmt_cooperations.cpe IS 'CPE=花费/互动量(预留)';
+COMMENT ON COLUMN cmt_cooperations.engagement_rate IS '互动率=互动量/曝光量(预留)';
+COMMENT ON COLUMN cmt_cooperations.brand_mention_rate IS '评论品牌提及率(预留)';
+COMMENT ON COLUMN cmt_cooperations.risk_flag IS '差评或舆情风险(预留)';
+
+CREATE INDEX IF NOT EXISTS idx_cmt_cooperations_style ON cmt_cooperations (style_id);
+CREATE INDEX IF NOT EXISTS idx_cmt_cooperations_creator ON cmt_cooperations (creator_id);
+CREATE INDEX IF NOT EXISTS idx_cmt_cooperations_date ON cmt_cooperations (cooperation_date);
+
+-- ============================================================
+-- 3. 给新表加 updated_at 触发器
+-- ============================================================
+DO $$
+BEGIN
+  IF NOT EXISTS (
+    SELECT 1 FROM pg_trigger WHERE tgname = 'trg_cmt_cooperations_updated_at'
+  ) THEN
+    CREATE TRIGGER trg_cmt_cooperations_updated_at
+      BEFORE UPDATE ON cmt_cooperations
+      FOR EACH ROW EXECUTE FUNCTION update_updated_at_column();
+  END IF;
+END;
+$$;
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/003_creator_report.sql b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/003_creator_report.sql
new file mode 100644
index 0000000..a6ebbca
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/003_creator_report.sql
@@ -0,0 +1,43 @@
+-- 达人月度分析报告表
+-- 每月1号自动跑 generate_creator_report.py，结果入库 + 飞书文档
+-- UNIQUE(month_start) — 每月一份全量报告
+
+CREATE TABLE IF NOT EXISTS cmt_creator_report (
+  id                BIGSERIAL PRIMARY KEY,
+  month_start       DATE NOT NULL,
+  month_end         DATE NOT NULL,
+  time_label        VARCHAR(32) NOT NULL,
+  -- 统计快照
+  total_cooperations INTEGER DEFAULT 0,       -- 合作记录总数
+  total_creators     INTEGER DEFAULT 0,       -- 达人总数
+  total_styles       INTEGER DEFAULT 0,       -- 款式数
+  total_cost         NUMERIC(14,2),           -- 总合作花费
+  total_exposure     BIGINT,                  -- 总曝光量
+  total_engagement   BIGINT,                  -- 总互动量
+  avg_cpe            NUMERIC(10,2),           -- 平均CPE
+  -- 层级分布快照 (JSON)
+  level_distribution JSONB DEFAULT '{}',      -- {"头部":2,"腰部":36,"KOC":131,"素人":68,"未知":1}
+  -- 平台分布快照 (JSON)
+  platform_distribution JSONB DEFAULT '{}',   -- {"douyin":196,"xiaohongshu":164,"bilibili":38}
+  -- 评分分布快照 (JSON)
+  score_distribution JSONB DEFAULT '{}',      -- {"S":2,"A":62,"B":31,"C":143}
+  -- 飞书文档
+  doc_url            TEXT,
+  doc_title          VARCHAR(200),
+  report_chars       INTEGER,                 -- 报告字符数
+  -- 状态
+  status             VARCHAR(32),             -- ok/llm_failed/doc_failed/exception
+  created_at         TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  updated_at         TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  CONSTRAINT uniq_creator_report_month UNIQUE (month_start)
+);
+
+CREATE INDEX IF NOT EXISTS idx_cmt_creator_report_month ON cmt_creator_report (month_start);
+
+COMMENT ON TABLE cmt_creator_report IS '达人月度分析报告表(每月1号自动生成,全量达人筛选与报价分析)';
+COMMENT ON COLUMN cmt_creator_report.time_label IS '月时间范围标签 如 2026-06-01~06-30';
+COMMENT ON COLUMN cmt_creator_report.level_distribution IS '层级分布快照 JSON {"头部":N,"腰部":N,...}';
+COMMENT ON COLUMN cmt_creator_report.platform_distribution IS '平台分布快照 JSON {"douyin":N,...}';
+COMMENT ON COLUMN cmt_creator_report.score_distribution IS '评分分布快照 JSON {"S":N,"A":N,...}';
+COMMENT ON COLUMN cmt_creator_report.doc_url IS '飞书文档链接(达人分析报告)';
+COMMENT ON COLUMN cmt_creator_report.status IS 'ok/llm_failed/doc_failed/exception';
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/004_style_category.sql b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/004_style_category.sql
new file mode 100644
index 0000000..d1beb4c
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/004_style_category.sql
@@ -0,0 +1,4 @@
+ALTER TABLE cmt_styles
+  ADD COLUMN IF NOT EXISTS style_category VARCHAR(64) DEFAULT '';
+
+COMMENT ON COLUMN cmt_styles.style_category IS '款式风格，来源于飞书全域事项表的风格单选字段';
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/005_style_product_profile.sql b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/005_style_product_profile.sql
new file mode 100644
index 0000000..40171fc
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/migrations/005_style_product_profile.sql
@@ -0,0 +1,23 @@
+-- Product profile fields for cmt_styles.
+ALTER TABLE cmt_styles
+    ADD COLUMN IF NOT EXISTS brand VARCHAR(128) DEFAULT '',
+    ADD COLUMN IF NOT EXISTS erp_style_codes TEXT[] DEFAULT '{}',
+    ADD COLUMN IF NOT EXISTS platform_product_ids JSONB DEFAULT '{}'::jsonb,
+    ADD COLUMN IF NOT EXISTS product_type VARCHAR(128) DEFAULT '',
+    ADD COLUMN IF NOT EXISTS dimensions VARCHAR(255) DEFAULT '',
+    ADD COLUMN IF NOT EXISTS colors TEXT[] DEFAULT '{}',
+    ADD COLUMN IF NOT EXISTS material TEXT DEFAULT '',
+    ADD COLUMN IF NOT EXISTS capacity VARCHAR(128) DEFAULT '',
+    ADD COLUMN IF NOT EXISTS weight VARCHAR(128) DEFAULT '',
+    ADD COLUMN IF NOT EXISTS compatible_devices TEXT DEFAULT '',
+    ADD COLUMN IF NOT EXISTS features TEXT DEFAULT '',
+    ADD COLUMN IF NOT EXISTS selling_points TEXT DEFAULT '',
+    ADD COLUMN IF NOT EXISTS product_source_urls TEXT[] DEFAULT '{}',
+    ADD COLUMN IF NOT EXISTS product_research_status VARCHAR(32) DEFAULT 'pending',
+    ADD COLUMN IF NOT EXISTS product_researched_at TIMESTAMPTZ;
+
+COMMENT ON COLUMN cmt_styles.platform_product_ids IS 'Product IDs grouped by Feishu platform';
+COMMENT ON COLUMN cmt_styles.product_source_urls IS 'Public pages used to verify product facts';
+COMMENT ON COLUMN cmt_styles.product_research_status IS 'verified, partial, or pending';
+
+CREATE INDEX IF NOT EXISTS idx_cmt_styles_brand ON cmt_styles (brand);
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/monday_self_run.bat b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/monday_self_run.bat
new file mode 100644
index 0000000..0039bf3
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/monday_self_run.bat
@@ -0,0 +1,55 @@
+@echo off
+if not defined GYXX_PYTHON set "GYXX_PYTHON=python"
+REM Monday 13:00 self-operated pipeline: Bilibili + Chanmama -> sync to cmt_notes (with INSERT)
+REM Feishu write-back happens inside each scraper; sync step writes PostgreSQL.
+REM Register: schtasks /Create /SC WEEKLY /D MON /TN YingxiaoYunying_MondaySelf /TR %PROJECT_DIR%\data\tools\monday_self_run.bat /ST 13:00 /F
+
+setlocal
+
+for %%I in ("%~dp0..\..") do set "PROJECT_DIR=%%~fI"
+if not defined GYXX_DATA_ROOT (
+    if not defined GYXX_PROJECT_ROOT (
+        echo ERROR: GYXX_DATA_ROOT or GYXX_PROJECT_ROOT must be defined. 1>&2
+        exit /b 3
+    )
+    set "GYXX_DATA_ROOT=%GYXX_PROJECT_ROOT%\var"
+)
+set LOG_DIR=%GYXX_DATA_ROOT%\logs\content_marketing\scheduled
+set PYTHON=%GYXX_PYTHON%
+if not exist "%PYTHON%" set PYTHON=python
+
+if not exist "%LOG_DIR%" mkdir "%LOG_DIR%"
+
+set TS=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%
+set TS=%TS: =0%
+
+set LOG_FILE=%LOG_DIR%\monday_self_%TS%.log
+
+echo === Monday self-operated run started at %date% %time% === > "%LOG_FILE%"
+echo PROJECT_DIR=%PROJECT_DIR% >> "%LOG_FILE%"
+
+cd /d "%PROJECT_DIR%"
+
+echo --- [step 1] self_bilibili_scraper.py --- >> "%LOG_FILE%"
+call %PYTHON% self_bilibili_scraper.py >> "%LOG_FILE%" 2>&1
+set RC_BILI=%ERRORLEVEL%
+echo self_bilibili exit=%RC_BILI% >> "%LOG_FILE%"
+
+echo --- [step 2] chanmama_scraper.py --- >> "%LOG_FILE%"
+call %PYTHON% chanmama_scraper.py >> "%LOG_FILE%" 2>&1
+set RC_CM=%ERRORLEVEL%
+echo chanmama exit=%RC_CM% >> "%LOG_FILE%"
+
+echo --- [step 3] sync_metrics_to_cmt_notes.py (with INSERT) --- >> "%LOG_FILE%"
+call %PYTHON% data\tools\sync_metrics_to_cmt_notes.py >> "%LOG_FILE%" 2>&1
+set RC_SYNC=%ERRORLEVEL%
+echo sync exit=%RC_SYNC% >> "%LOG_FILE%"
+
+set FINAL_RC=0
+if not "%RC_BILI%"=="0" set FINAL_RC=1
+if not "%RC_CM%"=="0" set FINAL_RC=1
+if not "%RC_SYNC%"=="0" set FINAL_RC=1
+
+echo === Monday self-operated run finished at %date% %time% (bili=%RC_BILI%, cm=%RC_CM%, sync=%RC_SYNC%, final=%FINAL_RC%) === >> "%LOG_FILE%"
+
+endlocal & exit /b %FINAL_RC%
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/monthly_creator_report.bat b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/monthly_creator_report.bat
new file mode 100644
index 0000000..e1c8a16
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/monthly_creator_report.bat
@@ -0,0 +1,41 @@
+@echo off
+if not defined GYXX_PYTHON set "GYXX_PYTHON=python"
+REM Monthly creator report: run on 1st of each month at 08:30
+REM 1. Read cmt_cooperations + cmt_creators + cmt_styles from PG
+REM 2. Generate 9-section analysis report (Markdown)
+REM 3. Create Feishu doc + save to cmt_creator_report
+REM Register: schtasks /Create /SC MONTHLY /D 1 /TN YingxiaoYunying_MonthlyCreatorReport /TR %PROJECT_DIR%\data\tools\monthly_creator_report.bat /ST 08:30 /F
+
+setlocal
+
+for %%I in ("%~dp0..\..") do set "PROJECT_DIR=%%~fI"
+if not defined GYXX_DATA_ROOT (
+    if not defined GYXX_PROJECT_ROOT (
+        echo ERROR: GYXX_DATA_ROOT or GYXX_PROJECT_ROOT must be defined. 1>&2
+        exit /b 3
+    )
+    set "GYXX_DATA_ROOT=%GYXX_PROJECT_ROOT%\var"
+)
+set LOG_DIR=%GYXX_DATA_ROOT%\logs\content_marketing\scheduled
+set PYTHON=%GYXX_PYTHON%
+where %PYTHON% >nul 2>&1 || set PYTHON=python
+
+if not exist "%LOG_DIR%" mkdir "%LOG_DIR%"
+
+set TS=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%
+set TS=%TS: =0%
+
+set LOG_FILE=%LOG_DIR%\monthly_creator_report_%TS%.log
+
+echo === Monthly creator report started at %date% %time% === > "%LOG_FILE%"
+echo PROJECT_DIR=%PROJECT_DIR% >> "%LOG_FILE%"
+
+cd /d "%PROJECT_DIR%"
+
+call %PYTHON% -u -X utf8 data/tools/generate_creator_report.py >> "%LOG_FILE%" 2>&1
+set RC=%ERRORLEVEL%
+echo monthly_creator_report exit=%RC% >> "%LOG_FILE%"
+
+echo === Monthly creator report finished at %date% %time% (rc=%RC%) === >> "%LOG_FILE%"
+
+endlocal
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/monthly_summary.bat b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/monthly_summary.bat
new file mode 100644
index 0000000..a2e2788
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/monthly_summary.bat
@@ -0,0 +1,41 @@
+@echo off
+if not defined GYXX_PYTHON set "GYXX_PYTHON=python"
+REM Monthly summary: run on 1st of each month at 08:00
+REM 1. Compute last month date range
+REM 2. Query PG for notes per style
+REM 3. For each style: note analysis + LLM cross-compare + Feishu doc + write back + save to DB
+REM Register: schtasks /Create /SC MONTHLY /D 1 /TN YingxiaoYunying_MonthlySummary /TR %PROJECT_DIR%\data\tools\monthly_summary.bat /ST 08:00 /F
+
+setlocal
+
+for %%I in ("%~dp0..\..") do set "PROJECT_DIR=%%~fI"
+if not defined GYXX_DATA_ROOT (
+    if not defined GYXX_PROJECT_ROOT (
+        echo ERROR: GYXX_DATA_ROOT or GYXX_PROJECT_ROOT must be defined. 1>&2
+        exit /b 3
+    )
+    set "GYXX_DATA_ROOT=%GYXX_PROJECT_ROOT%\var"
+)
+set LOG_DIR=%GYXX_DATA_ROOT%\logs\content_marketing\scheduled
+set PYTHON=%GYXX_PYTHON%
+where %PYTHON% >nul 2>&1 || set PYTHON=python
+
+if not exist "%LOG_DIR%" mkdir "%LOG_DIR%"
+
+set TS=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%
+set TS=%TS: =0%
+
+set LOG_FILE=%LOG_DIR%\monthly_summary_%TS%.log
+
+echo === Monthly summary started at %date% %time% === > "%LOG_FILE%"
+echo PROJECT_DIR=%PROJECT_DIR% >> "%LOG_FILE%"
+
+cd /d "%PROJECT_DIR%"
+
+call %PYTHON% -u -X utf8 monthly_summary_all.py --max-workers 4 >> "%LOG_FILE%" 2>&1
+set RC=%ERRORLEVEL%
+echo monthly_summary exit=%RC% >> "%LOG_FILE%"
+
+echo === Monthly summary finished at %date% %time% (rc=%RC%) === >> "%LOG_FILE%"
+
+endlocal
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/pull_all_tables.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/pull_all_tables.py
new file mode 100644
index 0000000..c6d809b
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/pull_all_tables.py
@@ -0,0 +1,163 @@
+"""
+拉取 24 张多维表格的全量记录,保存为 JSON。
+每张表输出: {base_token, table_id, name, fields, total_records, records}
+其中 records 元素为 {record_id, 字段名: 值, ...}
+"""
+import json
+import subprocess
+import sys
+import time
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+TABLES = [
+    (1,  "UkbabpqRYanmD7sk7ksceVuhnwf", "tbl2BdhIaohZNkOD",  "盖亚斜挎"),
+    (2,  "OsYvbUe1iaJYeKs5WvLcU4rVnHc", "tblASZDx24q5Bjwy",  "极星pro"),
+    (3,  "Qs4mbzyAaaQ8SasnDpkc9lA5npe", "tblMok3p5E6QydPq",  "盖亚微单"),
+    (4,  "KnBRbH9iQaH3PAsAUk5cHLxNnMg", "tblcZ8y5126gTHe7",  "盖世m1"),
+    (5,  "Ssp8bXmrwahRBbsko2BcuiLanad", "tblqNsVvKj4EPyzA",  "星迹2"),
+    (6,  "BJaWbLEOKaHONMshOidcxvsKngh", "tblR0tCfPNk34ULj",  "星云2"),
+    (7,  "N43abi7IlaS8XosnUXec6gH2n7g", "tblasQwwbXwpGbpL",  "宙斯"),
+    (8,  "W80vbtPDoaaHTOs5Tipc6CTLn6c", "tblUaRl444G5IhVY",  "阿波罗2"),
+    (9,  "CtW5bUcc9a5rqRsZkwucUWDVnHb", "tbl8WgoFrOjkqOD5",  "阿波罗x1"),
+    (10, "JsVLbso8Ba6RjbslOl1cJX8lnyb", "tblwqyhbymn9Aour",  "波塞冬edge"),
+    (11, "IQrbbxAHpa9oiWs631GcPkHznob", "tblmWRjVp0wQiVeG",  "逐星"),
+    (12, "MAEZbKrNaabJbDs5ZY7caq1ynrd", "tblcT5gqVRsPBuOE",  "逐星GT"),
+    (13, "RKfkb5JkIa9FttsgHvNc2nJYnbf", "tblu1yXKkQqwje0d",  "星云mini"),
+    (14, "VrLtbkldsaggOms3F4Bcu3DinCg", "tblO5hzLek6Nk43J",  "晨星2"),
+    (15, "GubZbTrMSahLg5sdonIc8kR8nZe", "tbl1tl5CI0LhKeuk",  "极星双肩"),
+    (16, "VwpibmlakaqHuLsGChccMb3BnNd", "tbl7fPd2dMFKQdih",  "凌云air"),
+    (17, "CQcibpRjgaUCg7sUTMwcnrhunMh", "tblFHe5uONCWxbEh",  "觅光"),
+    (18, "XtHjbqpwcab3J4sWP6Fcfgh5nYt", "tblqRgK6MI5C1H4K",  "凌云3"),
+    (19, "GNOkbKNpna8pDJsLMfzcuZt3ncf", "tbljwMiRxb4SlpGr",  "瑞白"),
+    (20, "U2JGbbWcIal6WvsxURRcTMwEnhf", "tblVEdURLsfRZykJ",  "黑曜石吐司包"),
+    (21, "U9S8bLYsNa8Ul8sPffAcrJ02nHI", "tblhpEmlXGyqCGuh",  "黑曜石托特包"),
+    (22, "UPk3bR9MOaTkgrsG1qLc7IUvnQe", "tbllXFWRI59dLCM2",  "曜影"),
+    (23, "PwrGbXcpRaxJnOsd9cacFou8nAf", "tbludhnt5lkmdvSF",  "黑曜石电脑包"),
+    (24, "CNgPbjuXiaPoxPsShROcYbRDnbb", "tblEymEPk7lVaQN1",  "拓界2"),
+]
+
+OUT_DIR = PATHS.raw_root / "feishu_tables"
+OUT_DIR.mkdir(parents=True, exist_ok=True)
+
+
+import os
+LARK_CLI = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK_CLI):
+    LARK_CLI = "lark-cli.cmd"
+
+
+def call_lark(token: str, table_id: str, offset: int, limit: int) -> dict:
+    """调用 lark-cli base +record-list,返回 JSON"""
+    # LARK_CLI_NO_PROXY=1 避免代理警告污染 stderr
+    cmd = [
+        LARK_CLI, "base", "+record-list",
+        "--base-token", token,
+        "--table-id", table_id,
+        "--as", "user",
+        "--format", "json",
+        "--limit", str(limit),
+        "--offset", str(offset),
+    ]
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, text=True, env=env, encoding="utf-8", shell=False)
+    if proc.returncode != 0:
+        raise RuntimeError(f"lark-cli 调用失败(return={proc.returncode}): {proc.stderr}\nstdout: {proc.stdout[:500]}")
+    # 找到第一个 { 开始
+    text = proc.stdout
+    start = text.find("{")
+    if start < 0:
+        raise RuntimeError(f"lark-cli 返回无 JSON: {text[:500]}")
+    # 用 strict=False 容忍控制字符(GBK 字符串残留)
+    return json.loads(text[start:], strict=False)
+
+
+def pull_table(idx: int, token: str, table_id: str, name: str) -> dict:
+    if not table_id:
+        return {"index": idx, "name": name, "token": token, "table_id": table_id,
+                "status": "SKIP_NO_TABLE", "total_records": 0}
+
+    page_size = 200
+    offset = 0
+    fields = None
+    field_ids = None
+    records = []
+
+    while True:
+        try:
+            resp = call_lark(token, table_id, offset, page_size)
+        except Exception as exc:
+            return {"index": idx, "name": name, "token": token, "table_id": table_id,
+                    "status": "ERROR", "error": str(exc), "partial_records": len(records)}
+
+        if not resp.get("ok"):
+            return {"index": idx, "name": name, "token": token, "table_id": table_id,
+                    "status": "API_ERROR", "error": resp.get("error", resp),
+                    "partial_records": len(records)}
+
+        data = resp["data"]
+        if fields is None:
+            fields = data["fields"]
+            field_ids = data["field_id_list"]
+
+        rid_list = data["record_id_list"]
+        row_data = data["data"]
+
+        for i, row in enumerate(row_data):
+            rec = {"record_id": rid_list[i]}
+            for j, fid in enumerate(field_ids):
+                rec[fid] = row[j] if j < len(row) else None
+            records.append(rec)
+
+        offset += len(row_data)
+        has_more = data.get("has_more", False)
+        print(f"  [{idx:02d}] {name} pulled={len(row_data)} total={len(records)} has_more={has_more}", flush=True)
+        if not has_more or len(row_data) == 0:
+            break
+        time.sleep(0.3)
+
+    return {
+        "index": idx,
+        "base_token": token,
+        "table_id": table_id,
+        "name": name,
+        "fields": fields,
+        "field_id_to_name": dict(zip(field_ids, fields)),
+        "total_records": len(records),
+        "records": records,
+        "status": "OK",
+    }
+
+
+def main():
+    summary = []
+    for idx, token, table_id, name in TABLES:
+        print(f"[{idx:02d}/24] {name} (table={table_id}) START", flush=True)
+        result = pull_table(idx, token, table_id, name)
+        summary.append({
+            "index": idx, "name": name, "token": token, "table_id": table_id,
+            "status": result["status"],
+            "total_records": result.get("total_records") or result.get("partial_records", 0),
+        })
+        if result["status"] == "OK":
+            safe = str(idx) + "-" + name.translate(str.maketrans({
+                "/": "_", "\\": "_", ":": "_", "*": "_", "?": "_",
+                "\"": "_", "<": "_", ">": "_", "|": "_", " ": "_"
+            }))
+            out_path = OUT_DIR / f"{safe}.json"
+            with open(out_path, "w", encoding="utf-8") as f:
+                json.dump(result, f, ensure_ascii=False, indent=2)
+            print(f"  -> saved {out_path} ({result['total_records']} records)", flush=True)
+        else:
+            print(f"  -> FAILED: {result.get('error', result.get('status'))}", flush=True)
+        time.sleep(0.5)
+
+    with open(OUT_DIR / "_summary.json", "w", encoding="utf-8") as f:
+        json.dump(summary, f, ensure_ascii=False, indent=2)
+    print("\n========== 汇总 ==========", flush=True)
+    for s in summary:
+        print(f"  {s['index']:02d} {s['name']:>16s}  {s['status']:>12s}  {s['total_records']:>6d}")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/rebuild_mapping.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/rebuild_mapping.py
new file mode 100644
index 0000000..3a3f9c4
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/rebuild_mapping.py
@@ -0,0 +1,132 @@
+"""
+rebuild_mapping.py — 动态重建 款式_多维表格_对照.json
+
+工作原理:调用 feishu_mapping.refresh_mapping() 动态读取「合作达人」地址表,
+自动发现所有款式(当前 34 款,新增款式自动往后追加 index)。
+
+用法:
+  python rebuild_mapping.py            # 动态重建 + 写盘 + 打印 diff
+  python rebuild_mapping.py --diff     # 只构建内存 mapping,对比当前 JSON,打印 diff,不写盘
+  python rebuild_mapping.py --in-place # 等价于默认(直接覆盖 JSON)
+  python rebuild_mapping.py --print    # 打印最新 mapping 到 stdout(不写盘)
+
+依赖: feishu_mapping.py (内部使用), lark-cli (动态读取)
+"""
+import argparse
+import json
+import sys
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+PROJECT_DIR = PATHS.module_root
+from gyxx_flow.modules.content_marketing.runtime import feishu_mapping  # noqa: E402
+
+MAPPING_PATH = feishu_mapping.MAPPING_PATH
+
+
+def diff_entry(new: dict, old: dict | None) -> list[str]:
+    """对比单张表的新旧条目"""
+    if not old:
+        return [f"  + 新增表 [{new['index']:02d}] {new['name']}"]
+    diffs = []
+    new_fm = new.get("field_map", {})
+    old_fm = old.get("field_map", {})
+    for k, v in new_fm.items():
+        if k not in old_fm:
+            diffs.append(f"  + [{new['index']:02d}] {new['name']}: 新增字段 {k} = {v['field_name']}({v['field_id']})")
+        elif old_fm[k].get("field_id") != v["field_id"]:
+            diffs.append(f"  ~ [{new['index']:02d}] {new['name']}: {k} field_id 变更 {old_fm[k]['field_id']} -> {v['field_id']}")
+    for k in old_fm:
+        if k not in new_fm:
+            diffs.append(f"  - [{new['index']:02d}] {new['name']}: 字段 {k}({old_fm[k]['field_name']}) 不存在了!")
+    if new.get("total_records", 0) != old.get("total_records", 0):
+        diffs.append(f"  · [{new['index']:02d}] {new['name']}: total_records {old.get('total_records', 0)} -> {new.get('total_records', 0)}")
+    return diffs
+
+
+def diff_mapping(new: dict, old: dict | None) -> list[str]:
+    """对比新旧 mapping"""
+    if old is None:
+        return [f"  + 全新 mapping,共 {len(new.get('tables', []))} 款"]
+    new_by_name = {t["name"]: t for t in new.get("tables", [])}
+    old_by_name = {t["name"]: t for t in old.get("tables", [])}
+    diffs = []
+    for name in new_by_name:
+        diffs.extend(diff_entry(new_by_name[name], old_by_name.get(name)))
+    for name in old_by_name:
+        if name not in new_by_name:
+            diffs.append(f"  - 表 [{old_by_name[name]['index']:02d}] {name}: 不再出现在地址表!")
+    return diffs
+
+
+def print_core_field_check(tables: list[dict]) -> None:
+    """打印每张表的核心字段映射情况"""
+    print(f"\n核心字段映射检查:")
+    for t in tables:
+        missing = [k for k in ["creator_name", "note_title", "read_count_7d"] if k not in t["field_map"]]
+        if missing:
+            print(f"  [WARN] [{t['index']:02d}] {t['name']}: 缺 {missing}")
+        else:
+            cn = t['field_map']['creator_name']['field_name']
+            nt = t['field_map']['note_title']['field_name']
+            r7 = t['field_map']['read_count_7d']['field_name']
+            print(f"  [OK]   [{t['index']:02d}] {t['name']}: creator='{cn}' title='{nt}' read7='{r7}'")
+
+
+def main():
+    ap = argparse.ArgumentParser(description="动态重建 款式_多维表格_对照.json")
+    ap.add_argument("--diff", action="store_true", help="只构建内存 mapping,对比当前 JSON,打印 diff,不写盘")
+    ap.add_argument("--in-place", action="store_true", help="直接覆盖现有 JSON(等价于默认行为)")
+    ap.add_argument("--print", action="store_true", help="只打印最新 mapping 到 stdout(不写盘)")
+    args = ap.parse_args()
+
+    # 读旧 JSON
+    old_mapping = None
+    if MAPPING_PATH.exists():
+        try:
+            old_mapping = json.loads(MAPPING_PATH.read_text(encoding="utf-8"))
+        except Exception as exc:
+            print(f"[WARN] 旧 JSON 解析失败: {exc}")
+
+    # 决定是否写盘
+    write = not (args.diff or args.print)
+
+    # 动态重建
+    print(f"动态读取「合作达人」地址表 + 各款式字段定义...\n")
+    new_mapping = feishu_mapping.refresh_mapping(force=True, write=write)
+    if new_mapping is None:
+        print("[FATAL] refresh_mapping 失败")
+        sys.exit(1)
+
+    tables = new_mapping.get("tables", [])
+    print(f"\n========== 重建结果 ==========")
+    print(f"表数: {len(tables)}")
+    if old_mapping:
+        old_n = len(old_mapping.get("tables", []))
+        if old_n != len(tables):
+            print(f"  (旧 {old_n} 款 -> 新 {len(tables)} 款)")
+
+    # diff
+    diffs = diff_mapping(new_mapping, old_mapping)
+    if diffs:
+        print(f"\n变更 ({len(diffs)} 项):")
+        for d in diffs:
+            print(d)
+    else:
+        print(f"\n无字段变化")
+
+    print_core_field_check(tables)
+
+    # 落盘 / 打印
+    if args.print:
+        print(f"\n{json.dumps(new_mapping, ensure_ascii=False, indent=2)}")
+    elif args.diff:
+        print(f"\n(--diff 模式,未写文件)")
+    elif write:
+        print(f"\nsaved (in-place): {MAPPING_PATH}")
+    else:
+        print(f"\n(saved): {MAPPING_PATH}")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_bilibili.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_bilibili.py
new file mode 100644
index 0000000..946d1ca
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_bilibili.py
@@ -0,0 +1,87 @@
+"""每周一重登 B 站。
+
+强制清掉 cookie JSON + 浏览器 profile,确保下次启动时必须扫码。
+任务计划: 每周一 09:00
+
+跑法:
+  python data/tools/relogin_bilibili.py             # 实际跑(会弹二维码)
+  python data/tools/relogin_bilibili.py --dry-run   # 只看会做什么
+"""
+
+import argparse
+import datetime as dt
+import shutil
+import subprocess
+import sys
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools.relogin_transaction import begin, commit, rollback
+
+PROJECT_DIR = PATHS.module_root
+COOKIE_FILE = PATHS.state_root / "cookies/bilibili_cookies.json"
+STATE_FILE = PATHS.state_root / "cookies/bilibili_storage_state.json"
+PROFILE_DIR = PATHS.state_root / "browser-profiles/bilibili"
+SCRAPER = PROJECT_DIR / "bilibili_comment_scraper.py"
+
+
+def reset_state():
+    """删 cookie JSON + storage_state + 清浏览器 profile。"""
+    state = begin((COOKIE_FILE, STATE_FILE), PROFILE_DIR)
+    print("  [OK] 旧 Cookie/Profile 已暂存,失败时将自动恢复")
+    return state
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser(description="每周一重登 B 站")
+    ap.add_argument("--login-timeout", type=int, default=300, help="扫码超时(秒)")
+    ap.add_argument("--dry-run", action="store_true", help="只打印计划,不删不跑")
+    args = ap.parse_args()
+
+    print("== Relogin: B 站 ==")
+    print(f"  今天: {dt.date.today()} ({dt.date.today().strftime('%A')})")
+    print(f"  cookie: {COOKIE_FILE}")
+    print(f"  profile: {PROFILE_DIR}")
+    print()
+
+    if args.dry_run:
+        print("[DRY-RUN] 实际跑会: 删 cookie + 清 profile + 弹二维码")
+        return 0
+
+    old_state = reset_state()
+    cmd = [
+        sys.executable,
+        str(SCRAPER),
+        "login",  # dummy url (login-only 模式不需要真实 url)
+        "--login-only",
+        "--login-timeout",
+        str(args.login_timeout),
+    ]
+    print(f"\n  [RUN] {' '.join(cmd)}")
+    try:
+        rc = subprocess.call(cmd, cwd=str(PROJECT_DIR))
+    except BaseException:
+        rollback(old_state)
+        raise
+    print(f"\n  exit={rc}")
+    login_ok = False
+    if rc == 0:
+        # 验证 SESSDATA 是否真写入了
+        try:
+            import json
+            cookies = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+            sessdata = any(c.get("name") == "SESSDATA" and c.get("value") for c in cookies)
+            login_ok = bool(sessdata)
+            print(f"  SESSDATA 写入: {'OK' if sessdata else 'FAIL — 没有看到 SESSDATA,登录可能没成功'}")
+        except Exception as exc:
+            print(f"  [WARN] 验证 cookie 时出错: {exc}")
+    if rc != 0 or not login_ok:
+        rollback(old_state)
+        print("  [ROLLBACK] 未完成扫码,已恢复旧登录状态")
+        return rc or 1
+    commit(old_state)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_douyin.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_douyin.py
new file mode 100644
index 0000000..9f77a4a
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_douyin.py
@@ -0,0 +1,93 @@
+"""Force a fresh Douyin login while preserving the previous usable session."""
+
+from __future__ import annotations
+
+import argparse
+import datetime as dt
+import json
+import subprocess
+import sys
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+try:
+    from .relogin_transaction import begin, commit, rollback
+except ImportError:  # Direct script execution from data/tools.
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.relogin_transaction import begin, commit, rollback
+
+PROJECT_DIR = PATHS.module_root
+COOKIE_FILE = PATHS.state_root / "cookies/douyin_cookies.json"
+STATE_FILE = PATHS.state_root / "cookies/douyin_storage_state.json"
+PROFILE_DIR = PATHS.state_root / "browser-profiles/douyin"
+SCRAPER = PROJECT_DIR / "douyin_comment_scraper.py"
+
+LOGIN_COOKIE = "sessionid"
+
+
+def reset_state():
+    state = begin((COOKIE_FILE, STATE_FILE), PROFILE_DIR)
+    print("  [OK] Old Cookie/Profile staged; failures will roll back automatically")
+    return state
+
+
+def is_cookie_valid() -> bool:
+    if not COOKIE_FILE.exists():
+        return False
+    try:
+        cookies = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+        return any(
+            isinstance(cookie, dict)
+            and cookie.get("name") == LOGIN_COOKIE
+            and cookie.get("value")
+            for cookie in cookies
+        )
+    except Exception:
+        return False
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Relogin Douyin")
+    parser.add_argument("--login-timeout", type=int, default=300)
+    parser.add_argument("--dry-run", action="store_true")
+    args = parser.parse_args()
+
+    print("== Relogin: Douyin ==")
+    print(f"  date: {dt.date.today()}")
+    print(f"  cookie: {COOKIE_FILE}")
+    print(f"  profile: {PROFILE_DIR}")
+    print(f"  current cookie: {'present' if is_cookie_valid() else 'missing/invalid'}")
+
+    if args.dry_run:
+        print("[DRY-RUN] Would stage Cookie/Profile and open the QR login window")
+        return 0
+
+    old_state = reset_state()
+    command = [
+        sys.executable,
+        str(SCRAPER),
+        "login",
+        "--login-only",
+        "--login-timeout",
+        str(args.login_timeout),
+    ]
+    print(f"\n  [RUN] {' '.join(command)}")
+    try:
+        return_code = subprocess.call(command, cwd=str(PROJECT_DIR))
+    except BaseException:
+        rollback(old_state)
+        raise
+
+    login_ok = return_code == 0 and is_cookie_valid()
+    print(f"\n  exit={return_code}")
+    print(f"  {LOGIN_COOKIE} written: {'OK' if login_ok else 'FAIL'}")
+    if not login_ok:
+        rollback(old_state)
+        print("  [ROLLBACK] Login incomplete; restored previous Cookie/Profile")
+        return return_code or 1
+
+    commit(old_state)
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_pgy.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_pgy.py
new file mode 100644
index 0000000..4342278
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_pgy.py
@@ -0,0 +1,106 @@
+"""每周一重登小红书蒲公英。
+
+强制清掉 cookie JSON + 浏览器 profile,确保下次启动时必须扫码。
+任务计划: 每周一 09:00
+
+跑法:
+  python data/tools/relogin_pgy.py             # 实际跑(会弹二维码)
+  python data/tools/relogin_pgy.py --dry-run   # 只看会做什么
+"""
+
+import argparse
+import datetime as dt
+import shutil
+import subprocess
+import sys
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+PROJECT_DIR = PATHS.module_root
+COOKIE_FILE = PATHS.state_root / "cookies/pgy_cookies.json"
+PROFILE_DIR = PATHS.state_root / "browser-profiles/pgy"
+SCRAPER = PROJECT_DIR / "pgy_xhs_scraper_v2.py"
+
+
+def reset_state() -> tuple[Path | None, Path | None]:
+    """Temporarily move the old login state aside so it can be restored."""
+    ts = dt.datetime.now().strftime("%Y%m%d_%H%M%S_%f")
+    cookie_backup = None
+    profile_backup = None
+    if COOKIE_FILE.exists():
+        backup = COOKIE_FILE.with_suffix(f".json.bak.{ts}")
+        COOKIE_FILE.rename(backup)
+        cookie_backup = backup
+        print(f"  [OK] 备份 cookie -> {backup.name}")
+    else:
+        print(f"  [INFO] {COOKIE_FILE.name} 不存在,跳过")
+    if PROFILE_DIR.exists():
+        profile_backup = PROFILE_DIR.with_name(f"{PROFILE_DIR.name}.bak.{ts}")
+        PROFILE_DIR.rename(profile_backup)
+        print(f"  [OK] 暂存 {PROFILE_DIR.name}/ -> {profile_backup.name}/")
+    else:
+        print(f"  [INFO] {PROFILE_DIR.name}/ 不存在,跳过")
+    return cookie_backup, profile_backup
+
+
+def restore_previous_state(cookie_backup: Path | None, profile_backup: Path | None) -> None:
+    """Discard an incomplete login and restore the last known-good state."""
+    if COOKIE_FILE.exists():
+        COOKIE_FILE.unlink()
+    if cookie_backup and cookie_backup.exists():
+        cookie_backup.rename(COOKIE_FILE)
+        print(f"  [ROLLBACK] 已恢复旧 cookie: {COOKIE_FILE.name}")
+    if PROFILE_DIR.exists():
+        shutil.rmtree(PROFILE_DIR)
+    if profile_backup and profile_backup.exists():
+        profile_backup.rename(PROFILE_DIR)
+        print(f"  [ROLLBACK] 已恢复旧浏览器状态: {PROFILE_DIR.name}/")
+
+
+def finish_successful_relogin(profile_backup: Path | None) -> None:
+    """Keep the new state and remove the bulky old profile backup."""
+    if profile_backup and profile_backup.exists():
+        shutil.rmtree(profile_backup)
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser(description="每周一重登小红书蒲公英")
+    ap.add_argument("--login-timeout", type=int, default=300, help="扫码超时(秒)")
+    ap.add_argument("--dry-run", action="store_true", help="只打印计划,不删不跑")
+    args = ap.parse_args()
+
+    print("== Relogin: 小红书蒲公英 ==")
+    print(f"  今天: {dt.date.today()} ({dt.date.today().strftime('%A')})")
+    print(f"  cookie: {COOKIE_FILE}")
+    print(f"  profile: {PROFILE_DIR}")
+    print()
+
+    if args.dry_run:
+        print("[DRY-RUN] 实际跑会: 删 cookie + 清 profile + 弹二维码")
+        return 0
+
+    cookie_backup, profile_backup = reset_state()
+    cmd = [
+        sys.executable,
+        str(SCRAPER),
+        "--login-only",
+        "--login-timeout",
+        str(args.login_timeout),
+    ]
+    print(f"\n  [RUN] {' '.join(cmd)}")
+    try:
+        rc = subprocess.call(cmd, cwd=str(PROJECT_DIR))
+    except BaseException:
+        restore_previous_state(cookie_backup, profile_backup)
+        raise
+    print(f"\n  exit={rc}")
+    if rc != 0 or not COOKIE_FILE.exists():
+        print("  [WARN] 新登录未完成,恢复旧登录状态")
+        restore_previous_state(cookie_backup, profile_backup)
+        return rc or 1
+    finish_successful_relogin(profile_backup)
+    return rc
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_transaction.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_transaction.py
new file mode 100644
index 0000000..49bac92
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_transaction.py
@@ -0,0 +1,47 @@
+"""Transactional backup/rollback helpers for interactive relogin scripts."""
+
+from __future__ import annotations
+
+import datetime as dt
+import shutil
+from dataclasses import dataclass
+from pathlib import Path
+
+
+@dataclass
+class LoginStateBackup:
+    files: dict[Path, Path]
+    profile: Path
+    profile_backup: Path | None
+
+
+def begin(files: tuple[Path, ...], profile: Path) -> LoginStateBackup:
+    stamp = dt.datetime.now().strftime("%Y%m%d_%H%M%S_%f")
+    backups: dict[Path, Path] = {}
+    for source in files:
+        if source.exists():
+            backup = source.with_name(f"{source.name}.bak.{stamp}")
+            source.rename(backup)
+            backups[source] = backup
+    profile_backup = None
+    if profile.exists():
+        profile_backup = profile.with_name(f"{profile.name}.bak.{stamp}")
+        profile.rename(profile_backup)
+    return LoginStateBackup(backups, profile, profile_backup)
+
+
+def rollback(state: LoginStateBackup) -> None:
+    for source, backup in state.files.items():
+        if source.exists():
+            source.unlink()
+        if backup.exists():
+            backup.rename(source)
+    if state.profile.exists():
+        shutil.rmtree(state.profile)
+    if state.profile_backup and state.profile_backup.exists():
+        state.profile_backup.rename(state.profile)
+
+
+def commit(state: LoginStateBackup) -> None:
+    if state.profile_backup and state.profile_backup.exists():
+        shutil.rmtree(state.profile_backup)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_xiaohongshu.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_xiaohongshu.py
new file mode 100644
index 0000000..7d12dda
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_xiaohongshu.py
@@ -0,0 +1,90 @@
+"""每周重登小红书。
+
+强制清掉 cookie JSON + 浏览器 profile,确保下次启动时必须扫码。
+跑法:
+  python data/tools/relogin_xiaohongshu.py             # 实际跑(会弹二维码)
+  python data/tools/relogin_xiaohongshu.py --dry-run   # 只看会做什么
+"""
+
+import argparse
+import datetime as dt
+import json
+import shutil
+import subprocess
+import sys
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools.relogin_transaction import begin, commit, rollback
+
+PROJECT_DIR = PATHS.module_root
+COOKIE_FILE = PATHS.state_root / "cookies/xiaohongshu_cookies.json"
+STATE_FILE = PATHS.state_root / "cookies/xiaohongshu_storage_state.json"
+PROFILE_DIR = PATHS.state_root / "browser-profiles/xiaohongshu"
+SCRAPER = PROJECT_DIR / "xiaohongshu_comment_scraper.py"
+
+LOGIN_COOKIE = "web_session"
+
+
+def reset_state():
+    state = begin((COOKIE_FILE, STATE_FILE), PROFILE_DIR)
+    print("  [OK] 旧 Cookie/Profile 已暂存,失败时将自动恢复")
+    return state
+
+
+def is_cookie_valid() -> bool:
+    if not COOKIE_FILE.exists():
+        return False
+    try:
+        cookies = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+        return any(c.get("name") == LOGIN_COOKIE and c.get("value") for c in cookies)
+    except Exception:
+        return False
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser(description="重登小红书")
+    ap.add_argument("--login-timeout", type=int, default=300, help="扫码超时(秒)")
+    ap.add_argument("--dry-run", action="store_true", help="只打印计划,不删不跑")
+    args = ap.parse_args()
+
+    print("== Relogin: 小红书 ==")
+    print(f"  今天: {dt.date.today()}")
+    print(f"  cookie: {COOKIE_FILE}")
+    print(f"  profile: {PROFILE_DIR}")
+    print(f"  当前 cookie 状态: {'有效' if is_cookie_valid() else '无效/不存在'}")
+    print()
+
+    if args.dry_run:
+        print("[DRY-RUN] 实际跑会: 删 cookie + 清 profile + 弹二维码")
+        return 0
+
+    old_state = reset_state()
+    cmd = [
+        sys.executable,
+        str(SCRAPER),
+        "login",
+        "--login-only",
+        "--login-timeout",
+        str(args.login_timeout),
+    ]
+    print(f"\n  [RUN] {' '.join(cmd)}")
+    try:
+        rc = subprocess.call(cmd, cwd=str(PROJECT_DIR))
+    except BaseException:
+        rollback(old_state)
+        raise
+    print(f"\n  exit={rc}")
+    login_ok = rc == 0 and is_cookie_valid()
+    if rc == 0:
+        print(f"  {LOGIN_COOKIE} 写入: {'OK' if login_ok else 'FAIL — 没有看到 web_session,登录可能没成功'}")
+    if not login_ok:
+        rollback(old_state)
+        print("  [ROLLBACK] 未完成扫码,已恢复旧登录状态")
+        return rc or 1
+    commit(old_state)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_xingtu.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_xingtu.py
new file mode 100644
index 0000000..de24ca6
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/relogin_xingtu.py
@@ -0,0 +1,299 @@
+"""每 3 个周一重登一次抖音/星图。
+
+强制清掉 cookie JSON + 浏览器 profile,确保下次启动时必须扫码。
+星图登录完成后,自动把登录态 cookie 同步到抖音评论 scraper,避免再扫一次码。
+
+任务计划: 每周一 09:00 (脚本内部判断是否到了该重登的周一,不到就退出)
+
+3 周循环基准: 任意一个周一,后续 (今天 - BASE_DATE) // 7 % 3 == 0 才是星图周。
+改 BASE_DATE 就改了重登周期。
+
+跑法:
+  python data/tools/relogin_xingtu.py             # 实际跑
+  python data/tools/relogin_xingtu.py --dry-run   # 看今天是不是星图周
+  python data/tools/relogin_xingtu.py --force     # 忽略 3 周判断,强制重登
+"""
+
+import argparse
+import datetime as dt
+import json
+import os
+import shutil
+import subprocess
+import sys
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+import psutil
+
+PROJECT_DIR = PATHS.module_root
+COOKIE_FILE = PATHS.state_root / "cookies/xingtu_cookies.json"
+PROFILE_DIR = PATHS.state_root / "browser-profiles/xingtu"
+SCRAPER = PROJECT_DIR / "xingtu_scraper_v2.py"
+
+# 抖音评论 scraper 的 cookie 文件 (登录态共享目标)
+DY_COOKIE_FILE = PATHS.state_root / "cookies/douyin_cookies.json"
+DY_STATE_FILE = PATHS.state_root / "cookies/douyin_storage_state.json"
+DY_PROFILE_DIR = PATHS.state_root / "browser-profiles/douyin"
+
+# 需要同步的抖音登录 cookie 名
+DY_LOGIN_COOKIES = {"sessionid", "sessionid_ss", "sid_guard", "sid_tt", "passport_csrf_token"}
+
+# 3 周循环起点(必须是周一)。星图 cookie ~22 天,每 3 周(21 天)重登正好。
+BASE_DATE = dt.date(2026, 6, 22)
+LOGIN_COOKIE_NAMES = {"sessionid", "sessionid_ss", "sid_tt", "sid_guard"}
+
+
+def transaction_file() -> Path:
+    return COOKIE_FILE.with_name(".xingtu_relogin_transaction.json")
+
+
+def _write_transaction(cookie_backup: Path | None, profile_backup: Path | None) -> None:
+    marker = transaction_file()
+    marker.parent.mkdir(parents=True, exist_ok=True)
+    temp = marker.with_name(f"{marker.name}.tmp")
+    temp.write_text(json.dumps({
+        "cookie_backup": str(cookie_backup) if cookie_backup else "",
+        "profile_backup": str(profile_backup) if profile_backup else "",
+    }, ensure_ascii=False, indent=2), encoding="utf-8")
+    os.replace(temp, marker)
+
+
+def _clear_transaction() -> None:
+    marker = transaction_file()
+    if marker.exists():
+        marker.unlink()
+
+
+def _remove_path(path: Path) -> None:
+    if path.is_dir():
+        shutil.rmtree(path)
+    elif path.exists():
+        path.unlink()
+
+
+def recover_interrupted_state() -> bool:
+    """Restore a relogin transaction that was killed after moving old state."""
+    marker = transaction_file()
+    if not marker.exists():
+        return False
+    try:
+        state = json.loads(marker.read_text(encoding="utf-8"))
+    except Exception:
+        return False
+
+    cookie_backup = Path(state["cookie_backup"]) if state.get("cookie_backup") else None
+    profile_backup = Path(state["profile_backup"]) if state.get("profile_backup") else None
+    terminate_profile_processes()
+    if cookie_backup and cookie_backup.exists():
+        _remove_path(COOKIE_FILE)
+        cookie_backup.rename(COOKIE_FILE)
+    if profile_backup and profile_backup.exists():
+        _remove_path(PROFILE_DIR)
+        profile_backup.rename(PROFILE_DIR)
+    _clear_transaction()
+    print("  [RECOVER] 已恢复上次被中断的星图登录状态")
+    return True
+
+
+def has_valid_login_cookie_file(path: Path | None = None) -> bool:
+    path = path or COOKIE_FILE
+    try:
+        cookies = json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return False
+    return any(
+        isinstance(cookie, dict)
+        and cookie.get("name") in LOGIN_COOKIE_NAMES
+        and cookie.get("value")
+        for cookie in cookies
+    )
+
+
+def terminate_profile_processes() -> None:
+    """Stop only browser processes using Xingtu's dedicated profile."""
+    profile_text = str(PROFILE_DIR.resolve()).casefold()
+    matches = []
+    for process in psutil.process_iter(["pid", "cmdline"]):
+        try:
+            command = " ".join(process.info.get("cmdline") or []).casefold()
+            if process.pid != os.getpid() and profile_text in command:
+                matches.append(process)
+        except (psutil.Error, OSError):
+            continue
+    for process in matches:
+        try:
+            process.terminate()
+        except (psutil.Error, OSError):
+            pass
+    _gone, alive = psutil.wait_procs(matches, timeout=5)
+    for process in alive:
+        try:
+            process.kill()
+        except (psutil.Error, OSError):
+            pass
+
+
+def is_xingtu_week(today: dt.date, base: dt.date) -> bool:
+    days = (today - base).days
+    if days < 0:
+        return False
+    return (days // 7) % 3 == 0
+
+
+def reset_state() -> tuple[Path | None, Path | None]:
+    """Temporarily move the old login state aside so it can be restored."""
+    ts = dt.datetime.now().strftime("%Y%m%d_%H%M%S_%f")
+    cookie_backup = None
+    profile_backup = None
+    if COOKIE_FILE.exists():
+        cookie_backup = COOKIE_FILE.with_suffix(f".json.bak.{ts}")
+    if PROFILE_DIR.exists():
+        profile_backup = PROFILE_DIR.with_name(f"{PROFILE_DIR.name}.bak.{ts}")
+    _write_transaction(cookie_backup, profile_backup)
+
+    if cookie_backup:
+        COOKIE_FILE.rename(cookie_backup)
+        print(f"  [OK] 备份 cookie -> {cookie_backup.name}")
+    else:
+        print(f"  [INFO] {COOKIE_FILE.name} 不存在,跳过")
+    if profile_backup:
+        PROFILE_DIR.rename(profile_backup)
+        print(f"  [OK] 暂存 {PROFILE_DIR.name}/ -> {profile_backup.name}/")
+    else:
+        print(f"  [INFO] {PROFILE_DIR.name}/ 不存在,跳过")
+    return cookie_backup, profile_backup
+
+
+def restore_previous_state(cookie_backup: Path | None, profile_backup: Path | None) -> None:
+    """Discard an incomplete login and restore the last known-good state."""
+    terminate_profile_processes()
+    if COOKIE_FILE.exists():
+        COOKIE_FILE.unlink()
+    if cookie_backup and cookie_backup.exists():
+        cookie_backup.rename(COOKIE_FILE)
+        print(f"  [ROLLBACK] 已恢复旧 cookie: {COOKIE_FILE.name}")
+    if PROFILE_DIR.exists():
+        shutil.rmtree(PROFILE_DIR)
+    if profile_backup and profile_backup.exists():
+        profile_backup.rename(PROFILE_DIR)
+        print(f"  [ROLLBACK] 已恢复旧浏览器状态: {PROFILE_DIR.name}/")
+    _clear_transaction()
+
+
+def finish_successful_relogin(profile_backup: Path | None) -> None:
+    """Keep the new state and remove the bulky old profile backup."""
+    if profile_backup and profile_backup.exists():
+        shutil.rmtree(profile_backup)
+    _clear_transaction()
+
+
+def sync_to_douyin_scraper() -> None:
+    """把星图的登录态 cookie 同步到抖音评论 scraper,避免再扫一次码。"""
+    if not COOKIE_FILE.exists():
+        print("  [SKIP] 星图 cookie 不存在,无法同步")
+        return
+
+    try:
+        xingtu_cookies = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+    except Exception as exc:
+        print(f"  [SKIP] 读取星图 cookie 失败: {exc}")
+        return
+
+    # 提取登录态 cookie
+    login_cookies = [c for c in xingtu_cookies if isinstance(c, dict) and c.get("name") in DY_LOGIN_COOKIES]
+    if not login_cookies:
+        print("  [SKIP] 星图 cookie 中没有找到登录态,跳过同步")
+        return
+
+    # 清除抖音评论 scraper 的旧 profile (强制刷新)
+    if DY_PROFILE_DIR.exists():
+        shutil.rmtree(DY_PROFILE_DIR)
+        print(f"  [OK] 清空 {DY_PROFILE_DIR.name}/")
+
+    # 读取或创建抖音评论 cookie 文件
+    dy_cookies = []
+    if DY_COOKIE_FILE.exists():
+        try:
+            dy_cookies = json.loads(DY_COOKIE_FILE.read_text(encoding="utf-8"))
+        except Exception:
+            dy_cookies = []
+
+    # 合并: 同名 cookie 覆盖,新 cookie 追加
+    dy_by_name = {c.get("name"): c for c in dy_cookies if isinstance(c, dict)}
+    for c in login_cookies:
+        # 修正 domain: 星图的 cookie domain 可能是 .douyin.com 子域,评论采集需要 .douyin.com
+        domain = c.get("domain", "")
+        if "xingtu" in domain or "oceanengine" in domain:
+            c["domain"] = ".douyin.com"
+        dy_by_name[c["name"]] = c
+
+    DY_COOKIE_FILE.parent.mkdir(parents=True, exist_ok=True)
+    DY_COOKIE_FILE.write_text(
+        json.dumps(list(dy_by_name.values()), ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    print(f"  [OK] 同步 {len(login_cookies)} 个登录 cookie 到 {DY_COOKIE_FILE.name}")
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser(description="每 3 周周一重登星图")
+    ap.add_argument("--base-date", default=BASE_DATE.isoformat(),
+                    help="3 周循环起点(必须是周一),格式 YYYY-MM-DD")
+    ap.add_argument("--login-timeout", type=int, default=300)
+    ap.add_argument("--force", action="store_true", help="忽略 3 周判断,强制重登")
+    ap.add_argument("--dry-run", action="store_true", help="只打印计划")
+    args = ap.parse_args()
+
+    today = dt.date.today()
+    base = dt.date.fromisoformat(args.base_date)
+    is_xw = is_xingtu_week(today, base)
+
+    print("== Relogin: 抖音/星图 ==")
+    print(f"  今天: {today} ({today.strftime('%A')})")
+    print(f"  3 周基准: {base}")
+    print(f"  星图周? {'是' if is_xw else '否'}")
+    print()
+
+    if not is_xw and not args.force:
+        print("今天不是星图周,跳过 (传 --force 强制重登)")
+        return 0
+
+    if args.dry_run:
+        print(f"[DRY-RUN] 实际跑会: 删 cookie + 清 profile + 弹二维码")
+        return 0
+
+    recover_interrupted_state()
+    terminate_profile_processes()
+    cookie_backup, profile_backup = reset_state()
+    cmd = [
+        sys.executable,
+        str(SCRAPER),
+        "--login-only",
+        "--login-timeout",
+        str(args.login_timeout),
+    ]
+    print(f"\n  [RUN] {' '.join(cmd)}")
+    try:
+        rc = subprocess.call(cmd, cwd=str(PROJECT_DIR))
+    except BaseException:
+        restore_previous_state(cookie_backup, profile_backup)
+        raise
+    finally:
+        terminate_profile_processes()
+    print(f"\n  exit={rc}")
+
+    if rc != 0 or not has_valid_login_cookie_file():
+        print("  [WARN] 新登录未完成,恢复旧登录状态")
+        restore_previous_state(cookie_backup, profile_backup)
+        return rc or 1
+
+    finish_successful_relogin(profile_backup)
+    if rc == 0:
+        sync_to_douyin_scraper()
+
+    return rc
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/retry_failed.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/retry_failed.py
new file mode 100644
index 0000000..30a703b
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/retry_failed.py
@@ -0,0 +1,507 @@
+"""补跑缺失 / 陈旧 / 部分失败的款式(自动检测 + 重试,粒度细到 record_id)。
+
+定义"需要补跑":
+  MISSING  - v2 JSON 文件不存在(进程崩了 / 中途中断) -> 整款重跑
+  STALE    - v2 JSON 文件存在,但 mtime 超过 --stale-hours 小时 -> 整款重跑
+  ALL_FAILED - v2 JSON 文件新鲜,但 matched=0 且 filled=0 -> 整款重跑
+  HALF     - v2 JSON 文件新鲜,部分 record 失败(matched>0 但有 failed) -> 只补失败的 record_id
+
+HALF 默认开启(因为只补失败的单条,不再昂贵);
+可用 --no-include-half 关掉。
+
+用法:
+  python data/tools/retry_failed.py                       # 三个平台都补
+  python data/tools/retry_failed.py --platform xt        # 只补星图
+  python data/tools/retry_failed.py --platform pgy,xt
+  python data/tools/retry_failed.py --dry-run            # 只看不跑
+  python data/tools/retry_failed.py --style 8            # 强制补指定款式(忽略检测)
+  python data/tools/retry_failed.py --style 8,15,20
+  python data/tools/retry_failed.py --max-attempts 3     # 每款式最多重试 3 次
+  python data/tools/retry_failed.py --stale-hours 4      # 4h 以上的算陈旧
+  python data/tools/retry_failed.py --no-include-half    # 不补 HALF
+"""
+import argparse
+import datetime as dt
+import json
+import subprocess
+import sys
+import time
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+PROJECT_DIR = PATHS.module_root
+V2_DIR = PATHS.normalized_root / "v2_results"
+
+# 复用 run_all.py 的 SCRIPTS 字典,保持单一事实
+SCRIPTS = {
+    "bili": {
+        "name": "B 站",
+        "script": "bilibili_scraper.py",
+        "label": "bili",
+    },
+    "pgy": {
+        "name": "小红书蒲公英",
+        "script": "pgy_xhs_scraper_v2.py",
+        "label": "xhs",
+    },
+    "xt": {
+        "name": "星图",
+        "script": "xingtu_scraper_v2.py",
+        "label": "xt",
+    },
+}
+
+SELF_SCRIPTS = {
+    "bili": {**SCRIPTS["bili"], "extra_args": ["--self-operated"]},
+    "pgy": {**SCRIPTS["pgy"], "extra_args": ["--self-operated"]},
+    "xt": {
+        "name": "抖音(自营)",
+        "script": "self_douyin_scraper.py",
+        "label": "xt",
+        "extra_args": [],
+    },
+}
+
+
+def _mapping_cache_path(self_operated: bool = False) -> Path:
+    name = "款式_多维表格_对照_自营.json" if self_operated else "款式_多维表格_对照.json"
+    cache = PATHS.normalized_root / "mappings" / name
+    return cache if cache.exists() else PATHS.config_root / name
+
+
+def load_current_styles(self_operated: bool = False) -> list[dict]:
+    """只读加载当前款式缓存。
+
+    补跑属于结果审计阶段，不能在 import 或 ``--dry-run`` 时触发飞书刷新、
+    更不能把自营缓存作为副作用写回。映射刷新由 run_all 在采集开始前负责。
+    """
+    path = _mapping_cache_path(self_operated)
+    try:
+        data = json.loads(path.read_text(encoding="utf-8"))
+    except Exception as exc:
+        log(f"[ERROR] 读取款式缓存失败 {path}: {exc}")
+        return []
+
+    by_index: dict[int, dict] = {}
+    for raw in data.get("tables", []):
+        if not isinstance(raw, dict) or not raw.get("index") or not raw.get("name"):
+            continue
+        try:
+            idx = int(raw["index"])
+        except (TypeError, ValueError):
+            continue
+        style = dict(raw)
+        style["index"] = idx
+        style["name"] = str(raw["name"])
+        by_index[idx] = style
+    return [by_index[idx] for idx in sorted(by_index)]
+
+
+def canonical_result_filename(style: dict, platform_key: str,
+                              self_operated: bool = False) -> str:
+    """返回当前款式唯一允许参与完整度判断的规范结果文件名。"""
+    if platform_key not in SCRIPTS:
+        raise ValueError(f"unknown platform: {platform_key}")
+    idx = int(style["index"])
+    name = str(style["name"])
+    self_sfx = "_self" if self_operated else ""
+    if platform_key == "pgy":
+        suffix = f"{self_sfx}_v2.json"
+    elif platform_key == "xt":
+        suffix = f"{self_sfx}_xingtu_v2.json"
+    else:
+        suffix = f"{self_sfx}_bilibili_v2.json"
+    return f"{idx:02d}-{name}{suffix}"
+
+
+def log(msg: str) -> None:
+    ts = dt.datetime.now().strftime("%H:%M:%S")
+    safe = msg.encode("gbk", errors="replace").decode("gbk", errors="replace")
+    print(f"[{ts}] {safe}", flush=True)
+
+
+def _result_rows(data: dict) -> list[dict]:
+    raw = data.get("results")
+    if raw is None:
+        raw = data.get("details")
+    if not isinstance(raw, list):
+        return []
+    return [row for row in raw if isinstance(row, dict)]
+
+
+def _result_is_unresolved(row: dict) -> bool:
+    """兼容新状态契约与旧版 matched/ok/error/reason 字段。"""
+    status = row.get("status")
+    if status == "blocked_input":
+        return False
+    explicit_failure = bool(
+        row.get("matched") is False
+        or row.get("ok") is False
+        or row.get("write_ok") is False
+        or row.get("error")
+        or row.get("reason")
+    )
+    if status == "success":
+        return explicit_failure
+    if status:
+        return True
+    return explicit_failure
+
+
+def _result_is_success(row: dict) -> bool:
+    status = row.get("status")
+    if status is not None:
+        return status == "success" and not _result_is_unresolved(row)
+    if _result_is_unresolved(row):
+        return False
+    # 旧版 B 站成功行有 ok=True；旧版 pgy/xt 成功行有命中标题或曝光值。
+    return bool(
+        row.get("ok") is True
+        or row.get("matched") is True
+        or row.get("matched_title")
+        or row.get("read_count") is not None
+        or row.get("play_count") is not None
+    )
+
+
+def requested_records_succeeded(data: dict,
+                                requested_record_ids: list[str]) -> tuple[bool, list[str]]:
+    """核验局部补跑请求的每个 record_id 都恰好有一条 success 终态。"""
+    rows_by_id: dict[str, list[dict]] = {}
+    for row in _result_rows(data):
+        rid = row.get("record_id")
+        if rid is not None:
+            rows_by_id.setdefault(str(rid), []).append(row)
+
+    failed: list[str] = []
+    seen: set[str] = set()
+    for raw_rid in requested_record_ids:
+        rid = str(raw_rid)
+        if rid in seen:
+            continue
+        seen.add(rid)
+        candidates = rows_by_id.get(rid, [])
+        if len(candidates) != 1 or not _result_is_success(candidates[0]):
+            failed.append(rid)
+    return not failed, failed
+
+
+def _summary_completed(data: dict) -> tuple[bool, str]:
+    """核验整款结果覆盖、唯一性和终态；blocked_input 是已解释终态。"""
+    total = data.get("total")
+    if total is None:
+        total = data.get("total_b_records")
+    if total is None:
+        total = data.get("total_tasks")
+    try:
+        total = int(total or 0)
+    except (TypeError, ValueError):
+        return False, "invalid total"
+    if total < 0:
+        return False, "invalid total"
+
+    rows = _result_rows(data)
+    record_ids = [str(row["record_id"]) for row in rows if row.get("record_id")]
+    if len(rows) != total:
+        return False, f"result coverage {len(rows)}/{total}"
+    if len(record_ids) != len(rows) or len(set(record_ids)) != len(record_ids):
+        return False, "missing or duplicate record_id"
+    unresolved = [rid for rid, row in zip(record_ids, rows) if _result_is_unresolved(row)]
+    if unresolved:
+        return False, f"unresolved records: {','.join(unresolved[:5])}"
+    if _summary_declares_incomplete(data):
+        return False, "summary declares unresolved failures"
+    return True, ""
+
+
+def _summary_declares_incomplete(data: dict) -> bool:
+    if data.get("complete") is False:
+        return True
+    for key in (
+        "unresolved", "retryable_failures", "write_failures",
+        "missing_results", "duplicate_results",
+    ):
+        try:
+            if int(data.get(key) or 0) > 0:
+                return True
+        except (TypeError, ValueError):
+            return True
+    return False
+
+
+def extract_failed_records(data: dict, platform_key: str) -> list[str]:
+    """从 v2 JSON 内容里提取失败的 record_id 列表。
+
+    xingtu/pgy: results 数组,失败标志 matched: false(可能加 reason/error)
+    bilibili:   details 数组,失败标志 ok: false(可能加 action)
+    """
+    rids: list[str] = []
+    for row in _result_rows(data):
+        rid = row.get("record_id")
+        if not rid:
+            continue
+        # blocked_input 是可解释但不可重试的终态，不应反复补跑。
+        if _result_is_unresolved(row):
+            rids.append(str(rid))
+    return list(dict.fromkeys(rids))
+
+
+def find_failed(platform_key: str, stale_hours: float, include_half: bool,
+                self_operated: bool = False, *,
+                styles: list[dict] | None = None,
+                v2_dir: Path | None = None,
+                now: dt.datetime | None = None) -> dict[int, dict]:
+    """扫 v2_results,返回 {idx: {reason, failed_rids}} 字典。
+
+    reason 取值: MISSING / STALE / ALL_FAILED / HALF
+    failed_rids: HALF 时给失败 record_id 列表;其他场景为空(整款重跑)。
+    self_operated=True 时只扫自营达人结果文件(*_self_*.json)
+    """
+    if platform_key not in SCRIPTS:
+        raise ValueError(f"unknown platform: {platform_key}")
+    current_styles = styles if styles is not None else load_current_styles(self_operated)
+    result_dir = v2_dir or V2_DIR
+    current_time = now or dt.datetime.now()
+    out: dict[int, dict] = {}
+
+    for style in current_styles:
+        idx = int(style["index"])
+        path = result_dir / canonical_result_filename(style, platform_key, self_operated)
+        if not path.exists():
+            out[idx] = {"reason": "MISSING", "failed_rids": []}
+            continue
+
+        mtime = dt.datetime.fromtimestamp(path.stat().st_mtime)
+        age_h = (current_time - mtime).total_seconds() / 3600.0
+        if age_h > stale_hours:
+            out[idx] = {"reason": "STALE", "failed_rids": []}
+            continue
+        try:
+            data = json.loads(path.read_text(encoding="utf-8"))
+        except Exception:
+            out[idx] = {"reason": "INVALID", "failed_rids": []}
+            continue
+
+        total = data.get("total")
+        if total is None:
+            total = data.get("total_b_records")
+        if total is None:
+            total = data.get("total_tasks")
+        try:
+            total = int(total or 0)
+        except (TypeError, ValueError):
+            out[idx] = {"reason": "INVALID", "failed_rids": []}
+            continue
+
+        rows = _result_rows(data)
+        failed_rids = extract_failed_records(data, platform_key)
+        resolved = sum(1 for row in rows if not _result_is_unresolved(row))
+        if total > 0 and resolved == 0:
+            out[idx] = {"reason": "ALL_FAILED", "failed_rids": []}
+            continue
+        record_ids = [str(row["record_id"]) for row in rows if row.get("record_id")]
+        coverage_invalid = (
+            len(rows) != total
+            or len(record_ids) != len(rows)
+            or len(set(record_ids)) != len(record_ids)
+        )
+        if coverage_invalid:
+            # 缺失记录无法从结果中推导 record_id，只能整款重跑。
+            out[idx] = {"reason": "INCOMPLETE", "failed_rids": []}
+            continue
+        if include_half and failed_rids:
+            out[idx] = {"reason": "HALF", "failed_rids": failed_rids}
+            continue
+        if _summary_declares_incomplete(data):
+            out[idx] = {"reason": "INCOMPLETE", "failed_rids": []}
+    return out
+
+
+def run_one_style(platform_key: str, idx: int, max_attempts: int,
+                  self_operated: bool = False,
+                  record_ids: list[str] | None = None, *,
+                  style: dict | None = None,
+                  v2_dir: Path | None = None) -> bool:
+    """跑单款式,失败重试。返回最终是否成功。
+
+    record_ids 非空时,只跑这些 record_id(--record 参数,粒度细到单条笔记)。
+    """
+    info = SELF_SCRIPTS[platform_key] if self_operated else SCRIPTS[platform_key]
+    if style is None:
+        style = next(
+            (item for item in load_current_styles(self_operated)
+             if int(item["index"]) == int(idx)),
+            None,
+        )
+    if style is None:
+        log(f"    [ERROR] 当前款式映射中不存在编号 {idx}")
+        return False
+    result_path = (v2_dir or V2_DIR) / canonical_result_filename(
+        style, platform_key, self_operated,
+    )
+    cmd = [sys.executable, str(PROJECT_DIR / info["script"]), "--style", str(idx)]
+    cmd.extend(info.get("extra_args", []))
+    for rid in record_ids or []:
+        cmd += ["--record", rid]
+    label = f"[{platform_key}{'-self' if self_operated else ''}] 款式 {idx:>2}"
+    desc = f"整款" if not record_ids else f"{len(record_ids)} 条"
+    for attempt in range(1, max_attempts + 1):
+        log(f"  {label} 尝试 {attempt}/{max_attempts} ({desc}): {info['script']} --style {idx}"
+            + (f" --record x{len(record_ids)}" if record_ids else ""))
+        before_mtime_ns = result_path.stat().st_mtime_ns if result_path.exists() else None
+        try:
+            rc = subprocess.call(cmd, cwd=str(PROJECT_DIR))
+        except KeyboardInterrupt:
+            raise
+        except Exception as exc:
+            log(f"    [ERROR] 进程异常: {exc}")
+            rc = -1
+        if rc != 0:
+            log(f"    [FAIL] exit={rc}")
+        else:
+            try:
+                after_mtime_ns = result_path.stat().st_mtime_ns
+                if before_mtime_ns is not None and after_mtime_ns == before_mtime_ns:
+                    raise RuntimeError("规范结果文件未在本次补跑中更新")
+                payload = json.loads(result_path.read_text(encoding="utf-8"))
+                if not isinstance(payload, dict):
+                    raise RuntimeError("规范结果不是 JSON object")
+                if record_ids:
+                    verified, failed_rids = requested_records_succeeded(payload, record_ids)
+                    reason = f"仍失败/缺失 record: {','.join(failed_rids[:5])}"
+                else:
+                    verified, reason = _summary_completed(payload)
+                if verified:
+                    log("    [OK] exit=0 且结果完整度校验通过")
+                    return True
+                log(f"    [INCOMPLETE] exit=0 但{reason}")
+            except Exception as exc:
+                log(f"    [INCOMPLETE] exit=0 但结果校验失败: {exc}")
+        if attempt < max_attempts:
+            time.sleep(2)
+    return False
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser(description="补跑缺失 / 陈旧的款式")
+    ap.add_argument("--platform", type=str, default="bili,pgy,xt",
+                    help="逗号分隔 bili/pgy/xt,默认全跑")
+    ap.add_argument("--style", type=str, default="",
+                    help="强制指定要补的款式(逗号分隔),忽略检测")
+    ap.add_argument("--dry-run", action="store_true", help="只打印计划,不跑")
+    ap.add_argument("--max-attempts", type=int, default=1,
+                    help="每个款式最多重试次数(默认 1)")
+    ap.add_argument("--stale-hours", type=float, default=26.0,
+                    help="超过多少小时算陈旧(默认 26h)")
+    ap.add_argument("--no-include-half", action="store_true",
+                    help="不补 HALF 款式(部分 record 失败的款);默认会补,只跑失败的 record_id")
+    ap.add_argument("--include-self-operated", action="store_true",
+                    help="同时补自营达人(默认只补合作达人)")
+    ap.add_argument("--self-operated-only", action="store_true",
+                    help="只补自营达人")
+    args = ap.parse_args()
+    include_half = not args.no_include_half
+
+    keys = [k.strip() for k in args.platform.split(",") if k.strip()]
+    for k in keys:
+        if k not in SCRIPTS:
+            print(f"[ERROR] 未知平台: {k}  可选: {','.join(SCRIPTS.keys())}")
+            return 1
+
+    forced: list[int] = []
+    if args.style:
+        for s in args.style.split(","):
+            s = s.strip()
+            if not s:
+                continue
+            try:
+                idx = int(s)
+            except ValueError:
+                print(f"[ERROR] --style 解析失败: {s}")
+                return 1
+            forced.append(idx)
+
+    # 决定要跑哪些轮: 合作达人 + 自营达人
+    rounds: list[tuple[bool, str]] = []  # (self_operated, label)
+    if not args.self_operated_only:
+        rounds.append((False, "合作达人"))
+    if args.include_self_operated or args.self_operated_only:
+        rounds.append((True, "自营达人"))
+
+    any_failed = False
+    for self_op, round_label in rounds:
+        styles = load_current_styles(self_op)
+        if not styles:
+            log(f"[ERROR] {round_label} 当前款式缓存为空，无法安全补跑")
+            any_failed = True
+            continue
+        styles_by_index = {int(style["index"]): style for style in styles}
+        invalid_forced = [idx for idx in forced if idx not in styles_by_index]
+        if invalid_forced:
+            log(f"[ERROR] {round_label} 不存在款式编号: {invalid_forced}")
+            any_failed = True
+
+        print(f"{'=' * 78}")
+        print(f"  Retry Failed Styles — {round_label}")
+        print(f"  平台: {','.join(SCRIPTS[k]['name'] for k in keys)}")
+        print(f"  stale-hours: {args.stale_hours}  max-attempts: {args.max_attempts}")
+        print(f"  include-half: {include_half}  dry-run: {args.dry_run}")
+        if forced:
+            print(f"  强制指定: {forced}")
+        print(f"{'=' * 78}")
+
+        grand_total = grand_ok = grand_fail = 0
+        grand_failed_list: list[tuple[str, int]] = []
+
+        for k in keys:
+            info = SCRIPTS[k]
+            if forced:
+                targets = {idx: {"reason": "强制", "failed_rids": []}
+                           for idx in forced if idx in styles_by_index}
+            else:
+                targets = find_failed(
+                    k, args.stale_hours, include_half,
+                    self_operated=self_op, styles=styles,
+                )
+            log(f"[{k}{'-self' if self_op else ''}] {info['name']} 待补: {len(targets)} 个 -> {sorted(targets.keys())}")
+            if not targets:
+                continue
+            if args.dry_run:
+                for idx in sorted(targets.keys()):
+                    info_i = targets[idx]
+                    n = len(info_i["failed_rids"])
+                    detail = f", {n} 条 record" if n else ""
+                    log(f"  [DRY-RUN] 款式 {idx:>2} ({info_i['reason']}{detail})")
+                grand_total += len(targets)
+                continue
+            for idx in sorted(targets.keys()):
+                info_i = targets[idx]
+                grand_total += 1
+                ok = run_one_style(k, idx, args.max_attempts,
+                                   self_operated=self_op,
+                                   record_ids=info_i["failed_rids"] or None,
+                                   style=styles_by_index[idx])
+                if ok:
+                    grand_ok += 1
+                else:
+                    grand_fail += 1
+                    grand_failed_list.append((k, idx))
+                    any_failed = True
+
+        print(f"\n{'=' * 78}")
+        print(f"  {round_label} 补跑汇总: 总 {grand_total}  成功 {grand_ok}  失败 {grand_fail}")
+        if grand_failed_list:
+            print(f"  仍失败的:")
+            for k, idx in grand_failed_list:
+                print(f"    - [{k}] 款式 {idx}")
+        print(f"{'=' * 78}\n")
+
+    return 1 if any_failed else 0
+
+
+if __name__ == "__main__":
+    try:
+        sys.exit(main())
+    except KeyboardInterrupt:
+        print("\n[中断]", file=sys.stderr)
+        sys.exit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/schema_gyxx_super_data.sql b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/schema_gyxx_super_data.sql
new file mode 100644
index 0000000..26184cb
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/schema_gyxx_super_data.sql
@@ -0,0 +1,342 @@
+-- 数据库：gyxx_super_data
+-- 表结构：cmt_styles / cmt_creators / cmt_style_creators / cmt_notes / cmt_comments
+-- 适用于 PostgreSQL 14+
+
+-- 如果数据库还没创建，取消下面一行的注释并执行：
+-- CREATE DATABASE gyxx_super_data WITH ENCODING = 'UTF8' LC_COLLATE = 'zh_CN.UTF-8' LC_CTYPE = 'zh_CN.UTF-8';
+
+\c gyxx_super_data;
+
+-- 款式表
+CREATE TABLE IF NOT EXISTS cmt_styles (
+  id              BIGSERIAL PRIMARY KEY,
+  name            VARCHAR(255) NOT NULL UNIQUE,
+  style_category  VARCHAR(64) DEFAULT '',
+  brand           VARCHAR(128) DEFAULT '',
+  erp_style_codes TEXT[] DEFAULT '{}',
+  platform_product_ids JSONB DEFAULT '{}'::jsonb,
+  product_type    VARCHAR(128) DEFAULT '',
+  dimensions      VARCHAR(255) DEFAULT '',
+  colors          TEXT[] DEFAULT '{}',
+  material        TEXT DEFAULT '',
+  capacity        VARCHAR(128) DEFAULT '',
+  weight          VARCHAR(128) DEFAULT '',
+  compatible_devices TEXT DEFAULT '',
+  features        TEXT DEFAULT '',
+  selling_points  TEXT DEFAULT '',
+  product_source_urls TEXT[] DEFAULT '{}',
+  product_research_status VARCHAR(32) DEFAULT 'pending',
+  product_researched_at TIMESTAMPTZ,
+  platform        VARCHAR(64)  DEFAULT '',
+  feishu_base_url TEXT DEFAULT '',
+  feishu_base_token VARCHAR(64) DEFAULT '',
+  feishu_table_id VARCHAR(64)   DEFAULT '',
+  feishu_view_id  VARCHAR(64)   DEFAULT '',
+  created_at      TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  updated_at      TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP
+);
+COMMENT ON TABLE cmt_styles IS '款式表';
+COMMENT ON COLUMN cmt_styles.name IS '款式名称';
+COMMENT ON COLUMN cmt_styles.style_category IS '款式风格';
+COMMENT ON COLUMN cmt_styles.platform IS '平台分组，如天猫';
+
+CREATE INDEX IF NOT EXISTS idx_cmt_styles_platform ON cmt_styles (platform);
+
+-- 博主/达人表
+CREATE TABLE IF NOT EXISTS cmt_creators (
+  id                BIGSERIAL PRIMARY KEY,
+  name              VARCHAR(255) NOT NULL,
+  platform          VARCHAR(32)  DEFAULT '',          -- 主平台: xiaohongshu/douyin/bilibili
+  platform_id       VARCHAR(128) DEFAULT '',          -- 平台达人ID (B站UID/抖音sec_uid等)
+  follower_count    VARCHAR(64)  DEFAULT '',          -- 达人粉丝数(原始文本如3.7万/12.4w)
+  follower_count_num BIGINT      DEFAULT NULL,        -- 达人粉丝数(数字解析后,3.7万=37000)
+  homepage_url      TEXT         DEFAULT '',          -- 达人主页链接
+  wechat            VARCHAR(128) DEFAULT '',          -- 微信号
+  account_type      VARCHAR(64)  DEFAULT '',          -- 达人类型: 数码/vlog/好物/穿搭/摄影等
+  creator_level     VARCHAR(32)  DEFAULT '',          -- 达人层级: 头部/腰部/KOC/素人
+  created_at        TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  updated_at        TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  CONSTRAINT cmt_creators_name_platform_key UNIQUE (name, platform)
+);
+COMMENT ON TABLE cmt_creators IS '博主/达人表';
+COMMENT ON COLUMN cmt_creators.name IS '博主/达人名称';
+COMMENT ON COLUMN cmt_creators.platform IS '主平台: xiaohongshu/douyin/bilibili';
+COMMENT ON COLUMN cmt_creators.platform_id IS '平台达人ID (B站UID/抖音sec_uid等)';
+COMMENT ON COLUMN cmt_creators.follower_count IS '达人粉丝数(原始文本如3.7万/12.4w)';
+COMMENT ON COLUMN cmt_creators.follower_count_num IS '达人粉丝数(数字解析后,3.7万=37000)';
+COMMENT ON COLUMN cmt_creators.homepage_url IS '达人主页链接';
+COMMENT ON COLUMN cmt_creators.wechat IS '微信号';
+COMMENT ON COLUMN cmt_creators.account_type IS '达人类型: 数码/vlog/好物/穿搭/摄影等';
+COMMENT ON COLUMN cmt_creators.creator_level IS '达人层级: 头部/腰部/KOC/素人';
+
+CREATE INDEX IF NOT EXISTS idx_cmt_creators_name ON cmt_creators (name);
+CREATE INDEX IF NOT EXISTS idx_cmt_creators_platform ON cmt_creators (platform);
+CREATE INDEX IF NOT EXISTS idx_cmt_creators_platform_id ON cmt_creators (platform_id);
+
+-- 款式-博主关系表（多对多）
+CREATE TABLE IF NOT EXISTS cmt_style_creators (
+  id          BIGSERIAL PRIMARY KEY,
+  style_id    BIGINT NOT NULL,
+  creator_id  BIGINT NOT NULL,
+  created_at  TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  UNIQUE (style_id, creator_id),
+  CONSTRAINT fk_cmt_sc_style FOREIGN KEY (style_id) REFERENCES cmt_styles (id) ON DELETE CASCADE,
+  CONSTRAINT fk_cmt_sc_creator FOREIGN KEY (creator_id) REFERENCES cmt_creators (id) ON DELETE CASCADE
+);
+COMMENT ON TABLE cmt_style_creators IS '款式-博主关系表';
+CREATE INDEX IF NOT EXISTS idx_cmt_style_creators_creator ON cmt_style_creators (creator_id);
+
+-- 笔记表
+CREATE TABLE IF NOT EXISTS cmt_notes (
+  id              BIGSERIAL PRIMARY KEY,
+  style_id        BIGINT NOT NULL,
+  creator_id      BIGINT NOT NULL,
+  platform        VARCHAR(32)  NOT NULL,
+  title           VARCHAR(512) DEFAULT '',
+  url             TEXT NOT NULL UNIQUE,
+  feishu_record_id VARCHAR(64)  DEFAULT '',
+  view_count      BIGINT       DEFAULT NULL,  -- 曝光量 / 播放量
+  like_count      BIGINT       DEFAULT NULL,
+  collect_count   BIGINT       DEFAULT NULL,  -- 小红书 / 抖音收藏
+  favorite_count  BIGINT       DEFAULT NULL,  -- B 站收藏
+  comment_count   BIGINT       DEFAULT NULL,
+  share_count     BIGINT       DEFAULT NULL,
+  scraped_at      TIMESTAMPTZ  DEFAULT NULL,
+  self_operated   BOOLEAN      DEFAULT FALSE,  -- TRUE=自营达人笔记, FALSE=合作达人笔记
+  publish_time    TIMESTAMPTZ  DEFAULT NULL,   -- 笔记发布时间(从飞书 publish_time 字段同步)
+  created_at      TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  updated_at      TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  CONSTRAINT fk_cmt_notes_style FOREIGN KEY (style_id) REFERENCES cmt_styles (id) ON DELETE CASCADE,
+  CONSTRAINT fk_cmt_notes_creator FOREIGN KEY (creator_id) REFERENCES cmt_creators (id) ON DELETE CASCADE
+);
+COMMENT ON COLUMN cmt_notes.view_count IS '曝光量 / 播放量 (来自 yingxiaoyunying 每日采集)';
+COMMENT ON COLUMN cmt_notes.collect_count IS '收藏量: 小红书/抖音 (来自 yingxiaoyunying 每日采集)';
+COMMENT ON COLUMN cmt_notes.favorite_count IS '收藏量: B 站 (来自 comment-data-collector)';
+COMMENT ON TABLE cmt_notes IS '笔记表';
+COMMENT ON COLUMN cmt_notes.platform IS '平台：xiaohongshu / douyin / bilibili';
+COMMENT ON COLUMN cmt_notes.url IS '笔记链接';
+COMMENT ON COLUMN cmt_notes.scraped_at IS '最近一次抓取时间';
+COMMENT ON COLUMN cmt_notes.self_operated IS 'TRUE=自营达人笔记, FALSE=合作达人笔记';
+COMMENT ON COLUMN cmt_notes.publish_time IS '笔记发布时间(从飞书 publish_time 字段同步)';
+
+CREATE INDEX IF NOT EXISTS idx_cmt_notes_style ON cmt_notes (style_id);
+CREATE INDEX IF NOT EXISTS idx_cmt_notes_creator ON cmt_notes (creator_id);
+CREATE INDEX IF NOT EXISTS idx_cmt_notes_platform ON cmt_notes (platform);
+CREATE INDEX IF NOT EXISTS idx_cmt_notes_scraped_at ON cmt_notes (scraped_at);
+
+-- 评论表
+CREATE TABLE IF NOT EXISTS cmt_comments (
+  id                   BIGSERIAL PRIMARY KEY,
+  note_id              BIGINT NOT NULL,
+  platform             VARCHAR(32)  NOT NULL,
+  level                VARCHAR(16)  NOT NULL,
+  platform_comment_id  VARCHAR(64)  DEFAULT '',
+  parent_comment_id    VARCHAR(64)  DEFAULT '',
+  reply_to_comment_id  VARCHAR(64)  DEFAULT '',
+  user_id              VARCHAR(128) DEFAULT '',
+  nickname             VARCHAR(255) DEFAULT '',
+  content              TEXT         DEFAULT NULL,
+  like_count           BIGINT       DEFAULT 0,
+  reply_count          BIGINT       DEFAULT 0,
+  ip_location          VARCHAR(128) DEFAULT '',
+  comment_created_at   TIMESTAMPTZ  DEFAULT NULL,
+  raw_create_time      VARCHAR(64)  DEFAULT '',
+  scraped_at           TIMESTAMPTZ  DEFAULT NULL,
+  created_at           TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  updated_at           TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  UNIQUE (note_id, platform_comment_id),
+  CONSTRAINT fk_cmt_comments_note FOREIGN KEY (note_id) REFERENCES cmt_notes (id) ON DELETE CASCADE
+);
+COMMENT ON TABLE cmt_comments IS '评论表';
+COMMENT ON COLUMN cmt_comments.level IS '层级：comment 一级评论 / reply 回复';
+
+CREATE INDEX IF NOT EXISTS idx_cmt_comments_note_id ON cmt_comments (note_id);
+
+-- ============================================================
+-- 合作记录表 (达人 + 款式维度)
+-- ============================================================
+-- 每条记录 = 一个达人在一个款式下的一次合作
+-- 飞书款式合作表的每一行 → 此表一条记录
+
+CREATE TABLE IF NOT EXISTS cmt_cooperations (
+  id                BIGSERIAL PRIMARY KEY,
+  style_id          BIGINT NOT NULL REFERENCES cmt_styles(id) ON DELETE CASCADE,
+  creator_id        BIGINT NOT NULL REFERENCES cmt_creators(id) ON DELETE CASCADE,
+  feishu_record_id  VARCHAR(64) DEFAULT '',     -- 飞书行 record_id, 用于回写
+  -- 合作基本信息
+  cooperation_date  DATE DEFAULT NULL,           -- 合作日期 (飞书"制单日期")
+  cooperation_cost  NUMERIC(12,2) DEFAULT NULL,  -- 合作花费 (飞书"合作费用")
+  ad_spend          NUMERIC(12,2) DEFAULT NULL,  -- 投流金额 (飞书"投流金额")
+  -- 内容信息
+  content_direction VARCHAR(255) DEFAULT '',     -- 内容方向 (飞书"发布笔记内容方向")
+  content_format    VARCHAR(64) DEFAULT '',      -- 内容形式: 单品种草/测评合集/场景植入/开箱Vlog/穿搭变装等
+  note_title        VARCHAR(512) DEFAULT '',     -- 发布笔记标题
+  note_url          TEXT DEFAULT '',             -- 发布笔记链接
+  publish_time      TIMESTAMPTZ DEFAULT NULL,    -- 发布时间
+  -- 数据表现
+  exposure_count    BIGINT DEFAULT NULL,         -- 曝光量 (飞书"曝光量(自动抓取)")
+  engagement_count  VARCHAR(64) DEFAULT '',      -- 互动赞藏数 (原始文本如"68.5万")
+  engagement_count_num BIGINT DEFAULT NULL,      -- 互动赞藏数(数字解析后)
+  data_performance  VARCHAR(255) DEFAULT '',     -- 数据表现 (逗号分隔: 曝光限流,数据良好)
+  cpm               NUMERIC(10,4) DEFAULT NULL,  -- CPM (飞书公式字段)
+  -- 物流/财务
+  tracking_number   VARCHAR(128) DEFAULT '',     -- 快递单号
+  is_paid           BOOLEAN DEFAULT FALSE,       -- 是否结款
+  is_new_direction  BOOLEAN DEFAULT FALSE,       -- 是否新方向验证
+  -- 电商效果 (飞书暂无,预留)
+  shop_uv           BIGINT DEFAULT NULL,         -- 进店UV
+  transaction_amount NUMERIC(12,2) DEFAULT NULL,  -- 成交金额(元)
+  roi               NUMERIC(10,4) DEFAULT NULL,  -- ROI (成交金额/花费)
+  cpe               NUMERIC(10,4) DEFAULT NULL,  -- CPE (花费/互动量)
+  engagement_rate   NUMERIC(8,4) DEFAULT NULL,   -- 互动率 (互动量/曝光量)
+  -- 评论分析 (飞书暂无,预留)
+  brand_mention_rate NUMERIC(8,4) DEFAULT NULL,  -- 评论品牌提及率
+  risk_flag         VARCHAR(255) DEFAULT '',     -- 差评或舆情风险
+  --
+  created_at        TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  updated_at        TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  CONSTRAINT uniq_cooperation_feishu_record UNIQUE (feishu_record_id)
+);
+
+COMMENT ON TABLE cmt_cooperations IS '合作记录表(每个达人+款式的一次合作)';
+COMMENT ON COLUMN cmt_cooperations.cooperation_date IS '合作日期(飞书制单日期)';
+COMMENT ON COLUMN cmt_cooperations.cooperation_cost IS '合作花费(元)';
+COMMENT ON COLUMN cmt_cooperations.ad_spend IS '投流金额(元)';
+COMMENT ON COLUMN cmt_cooperations.content_direction IS '发布笔记内容方向(飞书原始值)';
+COMMENT ON COLUMN cmt_cooperations.content_format IS '内容形式: 单品种草/测评合集/场景植入/开箱Vlog等';
+COMMENT ON COLUMN cmt_cooperations.exposure_count IS '曝光量(飞书自动抓取)';
+COMMENT ON COLUMN cmt_cooperations.engagement_count IS '互动赞藏数(原始文本如68.5万)';
+COMMENT ON COLUMN cmt_cooperations.data_performance IS '数据表现(逗号分隔: 曝光限流,数据良好,表现优秀等)';
+COMMENT ON COLUMN cmt_cooperations.shop_uv IS '进店UV(预留,飞书暂无)';
+COMMENT ON COLUMN cmt_cooperations.transaction_amount IS '成交金额(元,预留)';
+COMMENT ON COLUMN cmt_cooperations.roi IS 'ROI=成交金额/花费(预留)';
+COMMENT ON COLUMN cmt_cooperations.cpe IS 'CPE=花费/互动量(预留)';
+COMMENT ON COLUMN cmt_cooperations.engagement_rate IS '互动率=互动量/曝光量(预留)';
+COMMENT ON COLUMN cmt_cooperations.brand_mention_rate IS '评论品牌提及率(预留)';
+COMMENT ON COLUMN cmt_cooperations.risk_flag IS '差评或舆情风险(预留)';
+
+CREATE INDEX IF NOT EXISTS idx_cmt_cooperations_style ON cmt_cooperations (style_id);
+CREATE INDEX IF NOT EXISTS idx_cmt_cooperations_creator ON cmt_cooperations (creator_id);
+CREATE INDEX IF NOT EXISTS idx_cmt_cooperations_date ON cmt_cooperations (cooperation_date);
+
+
+-- ============================================================
+-- 周笔记汇总表 (款式 + 周维度)
+-- ============================================================
+CREATE TABLE IF NOT EXISTS cmt_weekly_summary (
+  id                BIGSERIAL PRIMARY KEY,
+  style_id          BIGINT NOT NULL REFERENCES cmt_styles(id) ON DELETE CASCADE,
+  week_start        DATE NOT NULL,
+  week_end          DATE NOT NULL,
+  time_label        VARCHAR(32) NOT NULL,
+  note_count        INTEGER DEFAULT 0,
+  total_comments    INTEGER DEFAULT 0,
+  total_views       BIGINT,
+  total_engagement  BIGINT,
+  doc_url           TEXT,
+  doc_title         VARCHAR(200),
+  summary_chars     INTEGER,
+  status            VARCHAR(32),
+  created_at        TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  updated_at        TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  CONSTRAINT uniq_weekly_style_week UNIQUE (style_id, week_start)
+);
+CREATE INDEX IF NOT EXISTS idx_cmt_weekly_summary_week ON cmt_weekly_summary (week_start);
+CREATE INDEX IF NOT EXISTS idx_cmt_weekly_summary_style ON cmt_weekly_summary (style_id);
+COMMENT ON TABLE cmt_weekly_summary IS '周笔记汇总表(款式+周维度,LLM 分析报告飞书文档链接)';
+COMMENT ON COLUMN cmt_weekly_summary.time_label IS '周时间范围标签 如 2026-06-29~07-05';
+COMMENT ON COLUMN cmt_weekly_summary.doc_url IS '飞书文档链接(每周笔记汇总报告)';
+COMMENT ON COLUMN cmt_weekly_summary.status IS 'ok/no_reports/no_notes/exception';
+
+-- ============================================================
+-- 月笔记汇总表 (款式 + 月维度,横向对比分析)
+-- ============================================================
+CREATE TABLE IF NOT EXISTS cmt_monthly_summary (
+  id                BIGSERIAL PRIMARY KEY,
+  style_id          BIGINT NOT NULL REFERENCES cmt_styles(id) ON DELETE CASCADE,
+  month_start       DATE NOT NULL,
+  month_end         DATE NOT NULL,
+  time_label        VARCHAR(32) NOT NULL,
+  note_count        INTEGER DEFAULT 0,
+  total_comments    INTEGER DEFAULT 0,
+  total_views       BIGINT,
+  total_engagement  BIGINT,
+  doc_url           TEXT,
+  doc_title         VARCHAR(200),
+  summary_chars     INTEGER,
+  status            VARCHAR(32),
+  created_at        TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  updated_at        TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  CONSTRAINT uniq_monthly_style_month UNIQUE (style_id, month_start)
+);
+CREATE INDEX IF NOT EXISTS idx_cmt_monthly_summary_month ON cmt_monthly_summary (month_start);
+CREATE INDEX IF NOT EXISTS idx_cmt_monthly_summary_style ON cmt_monthly_summary (style_id);
+COMMENT ON TABLE cmt_monthly_summary IS '月笔记汇总表(款式+月维度,LLM 横向对比分析报告)';
+COMMENT ON COLUMN cmt_monthly_summary.time_label IS '月时间范围标签 如 2026-06-01~06-30';
+COMMENT ON COLUMN cmt_monthly_summary.doc_url IS '飞书文档链接(月度汇总报告)';
+
+-- ============================================================
+-- 达人月度分析报告表 (每月1号自动生成)
+-- ============================================================
+CREATE TABLE IF NOT EXISTS cmt_creator_report (
+  id                BIGSERIAL PRIMARY KEY,
+  month_start       DATE NOT NULL,
+  month_end         DATE NOT NULL,
+  time_label        VARCHAR(32) NOT NULL,
+  total_cooperations INTEGER DEFAULT 0,
+  total_creators     INTEGER DEFAULT 0,
+  total_styles       INTEGER DEFAULT 0,
+  total_cost         NUMERIC(14,2),
+  total_exposure     BIGINT,
+  total_engagement   BIGINT,
+  avg_cpe            NUMERIC(10,2),
+  level_distribution JSONB DEFAULT '{}',
+  platform_distribution JSONB DEFAULT '{}',
+  score_distribution JSONB DEFAULT '{}',
+  doc_url            TEXT,
+  doc_title          VARCHAR(200),
+  report_chars       INTEGER,
+  status             VARCHAR(32),
+  created_at         TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  updated_at         TIMESTAMPTZ DEFAULT CURRENT_TIMESTAMP,
+  CONSTRAINT uniq_creator_report_month UNIQUE (month_start)
+);
+
+CREATE INDEX IF NOT EXISTS idx_cmt_creator_report_month ON cmt_creator_report (month_start);
+
+COMMENT ON TABLE cmt_creator_report IS '达人月度分析报告表(每月1号自动生成,全量达人筛选与报价分析)';
+COMMENT ON COLUMN cmt_creator_report.time_label IS '月时间范围标签 如 2026-06-01~06-30';
+COMMENT ON COLUMN cmt_creator_report.level_distribution IS '层级分布快照 JSON {"头部":N,"腰部":N,...}';
+COMMENT ON COLUMN cmt_creator_report.platform_distribution IS '平台分布快照 JSON {"douyin":N,...}';
+COMMENT ON COLUMN cmt_creator_report.score_distribution IS '评分分布快照 JSON {"S":N,"A":N,...}';
+COMMENT ON COLUMN cmt_creator_report.doc_url IS '飞书文档链接(达人分析报告)';
+COMMENT ON COLUMN cmt_creator_report.status IS 'ok/doc_failed/exception';
+
+CREATE INDEX IF NOT EXISTS idx_cmt_comments_platform ON cmt_comments (platform);
+CREATE INDEX IF NOT EXISTS idx_cmt_comments_level ON cmt_comments (level);
+CREATE INDEX IF NOT EXISTS idx_cmt_comments_comment_created_at ON cmt_comments (comment_created_at);
+
+-- 自动更新 updated_at 的触发器函数
+CREATE OR REPLACE FUNCTION update_updated_at_column()
+RETURNS TRIGGER AS $$
+BEGIN
+  NEW.updated_at = CURRENT_TIMESTAMP;
+  RETURN NEW;
+END;
+$$ LANGUAGE plpgsql;
+
+-- 为需要自动更新 updated_at 的表创建触发器
+DO $$
+DECLARE
+  tbl TEXT;
+BEGIN
+  FOR tbl IN SELECT unnest(ARRAY['cmt_styles', 'cmt_creators', 'cmt_style_creators', 'cmt_notes', 'cmt_comments', 'cmt_cooperations', 'cmt_weekly_summary', 'cmt_monthly_summary', 'cmt_creator_report'])
+  LOOP
+    IF NOT EXISTS (
+      SELECT 1 FROM pg_trigger WHERE tgname = 'trg_' || tbl || '_updated_at'
+    ) THEN
+      EXECUTE format('CREATE TRIGGER trg_%I_updated_at BEFORE UPDATE ON %I FOR EACH ROW EXECUTE FUNCTION update_updated_at_column()', tbl, tbl);
+    END IF;
+  END LOOP;
+END;
+$$;
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/style_analyzer.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/style_analyzer.py
new file mode 100644
index 0000000..030c362
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/style_analyzer.py
@@ -0,0 +1,362 @@
+"""Style-level rollup analyzer for the Feishu doc backfill.
+
+Aggregates metrics + keyword distributions across all notes of one style,
+then calls Hermes for a cross-note synthesis analysis that becomes the
+"💼 款式整体分析" section in the doc.
+
+Sibling of analyze_note.py — same import surface (call_hermes_analyzer,
+ANALYSIS_SYSTEM_PROMPT, code_filter_comments) but operates at style scope.
+"""
+
+import re
+import sys
+import time
+from pathlib import Path
+from typing import Any
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+# Path setup
+_TOOLS_DIR = PATHS.tools_root
+_PROJECT_ROOT = PATHS.module_root
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools.analyze_comments import (
+    ANALYSIS_SYSTEM_PROMPT,
+    call_hermes_analyzer,
+)
+
+
+CATEGORY_LABELS = {
+    "purchase_intent": "购买意向",
+    "positive_feature": "正向功能反馈",
+    "negative_feature": "负向功能反馈",
+    "user_scenario": "用户场景",
+    "aesthetic": "审美与风格",
+    "brand_emotion": "品牌情感",
+}
+
+
+# 5-section deep prompt at style scope (mirrors SINGLE_NOTE_DEEP_PROMPT_TEMPLATE
+# but across all notes of one style).
+STYLE_ROLLUP_DEEP_PROMPT = """你是一位品牌内容分析师。本次任务是把同一款式的所有笔记的标题、平台、点赞/收藏/评论/分享数据和有效评论内容，汇总成一份**款式整体分析报告**，用于投放团队对这款产品的内容策略做一次通盘复盘。
+
+输出结构（严格 5 大节，总长 800-1200 字）：
+
+### 一、执行摘要（不超过 150 字）
+用一段话点明这款产品的整体定位和投放现状：覆盖多少篇笔记、整体互动表现、主要用户反馈倾向、是否存在明显短板。
+
+### 二、数据表现
+- 该款式总笔记数 / 有效评论总数
+- 总点赞 / 总收藏 / 总评论 / 总分享（如曝光量汇总可用则一并列出）
+- 综合互动率（基于汇总数据）
+- 跨笔记表现差异：高/低表现笔记的标题特征（如"通勤""扫街""旅行"等场景标签）
+
+### 三、受众洞察
+基于所有有效评论的合并视角：
+- 正向反馈汇总（被夸的核心卖点 / 场景 / 审美点）
+- 负向问题/风险点（被吐槽的尺寸 / 材质 / 价格 / 功能）
+- 购买意向强弱（直接询问价格/链接 vs 围观闲聊）
+- 典型使用场景（被反复提及的目标人群和场景）
+- 跨平台差异（如果 bilibili / douyin / xiaohongshu 反馈侧重不同，请分别说明）
+
+### 四、内容策略
+基于跨笔记样本，提炼以下三类规律：
+- 哪种标题/封面/切入点最受欢迎
+- 哪种内容结构（测评 / 街拍 / Vlog / 教程）种草效率最高
+- 评论区的高频问题/梗，可作为后续内容选题/卖点提炼的素材
+
+### 五、下一步（按 P0/P1/P2 排序，最多 5 条）
+- P0（立即）：【问题】→【行动】→【责任方】
+- P1（本周）：...
+- P2（下月）：...（如无可省略）
+
+## 硬性约束
+
+1. 引用评论必须直接引述原文（用「」包裹），禁止使用"#N""第N条"等编号引用。
+2. 禁止"根据数据/数据显示/数据来源/通过 XX 可知"等出处类表述。
+3. 若有效评论 < 30 条，必须在执行摘要后注明"样本量较小，结论仅供参考"。
+4. P0/P1/P2 必须明确标注【问题】→【行动】→【责任方】三段式。
+5. 若款式涉及多品牌联合推广，请仅聚焦我方品牌/产品相关反馈。
+6. 不要重复列出指标定义，专注解读与行动。
+"""
+
+
+def _to_number(value: Any) -> float:
+    try:
+        if value is None or value == "":
+            return 0.0
+        return float(value)
+    except (TypeError, ValueError):
+        return 0.0
+
+
+def safe_div(num: Any, denom: Any) -> float:
+    try:
+        n = float(num)
+        d = float(denom)
+        if d == 0:
+            return 0.0
+        return n / d
+    except (TypeError, ValueError):
+        return 0.0
+
+
+def log(msg: str) -> None:
+    print(f"[{time.strftime('%H:%M:%S')}] {msg}", flush=True)
+
+
+# ---------------------------------------------------------------------------
+# Aggregation
+# ---------------------------------------------------------------------------
+def aggregate_style_metrics(notes: list[dict[str, Any]]) -> dict[str, Any]:
+    """Sum raw engagement counts across all notes of a style.
+
+    Returns:
+        {
+          "note_count": int,
+          "total_likes": int, "total_favorites": int, "total_comments": int,
+          "total_shares": int, "total_views": int,
+          "engagement_rate": float,   # (likes+comments+favorites+shares) / views
+          "platform_breakdown": {"bilibili": n, "douyin": n, "xiaohongshu": n},
+          "creator_breakdown": {creator_name: n},  # top creators by note count
+          "notes_meta": [{"id", "platform", "title", "like_count", "comment_count"}, ...]
+        }
+    """
+    totals = {"likes": 0.0, "favorites": 0.0, "comments": 0.0, "shares": 0.0, "views": 0.0}
+    platforms = {"bilibili": 0, "douyin": 0, "xiaohongshu": 0}
+    creators: dict[str, int] = {}
+    notes_meta: list[dict[str, Any]] = []
+
+    for n in notes:
+        likes = _to_number(n.get("like_count"))
+        favs = _to_number(n.get("favorite_count"))
+        cmts = _to_number(n.get("comment_count")) or _to_number(n.get("comment_count_total"))
+        shares = _to_number(n.get("share_count"))
+        views = _to_number(n.get("view_count"))
+        totals["likes"] += likes
+        totals["favorites"] += favs
+        totals["comments"] += cmts
+        totals["shares"] += shares
+        totals["views"] += views
+
+        plat = n.get("platform") or "unknown"
+        platforms[plat] = platforms.get(plat, 0) + 1
+
+        creator = n.get("creator_name") or "未指定"
+        creators[creator] = creators.get(creator, 0) + 1
+
+        notes_meta.append({
+            "id": n.get("id"),
+            "platform": plat,
+            "title": (n.get("title") or "")[:60],
+            "like_count": int(likes) if likes else 0,
+            "comment_count": int(cmts) if cmts else 0,
+            "creator_name": creator,
+        })
+
+    engagement = safe_div(
+        totals["likes"] + totals["comments"] + totals["favorites"] + totals["shares"],
+        totals["views"],
+    )
+
+    # Top creators by note count
+    top_creators = dict(sorted(creators.items(), key=lambda kv: -kv[1])[:5])
+
+    return {
+        "note_count": len(notes),
+        "total_likes": int(totals["likes"]),
+        "total_favorites": int(totals["favorites"]),
+        "total_comments": int(totals["comments"]),
+        "total_shares": int(totals["shares"]),
+        "total_views": int(totals["views"]),
+        "engagement_rate": engagement,
+        "platform_breakdown": platforms,
+        "creator_breakdown": top_creators,
+        "notes_meta": notes_meta,
+    }
+
+
+def merge_style_distribution(per_note_distributions: list[dict[str, Any]]) -> dict[str, Any]:
+    """Element-wise sum of overall + relevant dicts across notes."""
+    overall: dict[str, int] = {k: 0 for k in CATEGORY_LABELS}
+    relevant: dict[str, int] = {k: 0 for k in CATEGORY_LABELS}
+    relevant_total = 0
+    for d in per_note_distributions:
+        if not isinstance(d, dict):
+            continue
+        for k, v in (d.get("overall") or {}).items():
+            overall[k] = overall.get(k, 0) + int(v or 0)
+        for k, v in (d.get("relevant") or {}).items():
+            relevant[k] = relevant.get(k, 0) + int(v or 0)
+        relevant_total += int(d.get("relevant_total") or 0)
+    return {"overall": overall, "relevant": relevant, "relevant_total": relevant_total}
+
+
+def _pct(v: float) -> str:
+    return f"{v * 100:.2f}%" if v else "—"
+
+
+def _num(v: float) -> str:
+    try:
+        return f"{int(v):,}"
+    except (TypeError, ValueError):
+        return "—"
+
+
+def _format_distribution_md(d: dict[str, Any], total_filtered: int) -> str:
+    """Mirror analyze_note._keyword_distribution_markdown."""
+    overall = d.get("overall", {})
+    relevant = d.get("relevant", {})
+    relevant_total = d.get("relevant_total", 0)
+    if not overall:
+        return "（无关键词分布数据）"
+
+    def fmt(m: dict[str, int]) -> str:
+        return "\n".join(f"- {CATEGORY_LABELS[k]}：{m.get(k, 0)} 条" for k in CATEGORY_LABELS)
+
+    return "\n".join([
+        f"### 全评论（{total_filtered} 条）",
+        fmt(overall),
+        "",
+        f"### 我方产品相关（{relevant_total} 条）",
+        fmt(relevant) if relevant_total else "_（无评论明确提及我方产品/品牌）_",
+    ])
+
+
+def _format_metrics_md(agg: dict[str, Any]) -> str:
+    plat = agg["platform_breakdown"]
+    plat_str = " / ".join(f"{k} {v}" for k, v in plat.items() if v) or "—"
+    creator_str = " / ".join(f"{k} {v}" for k, v in agg["creator_breakdown"].items()) or "—"
+
+    return "\n".join([
+        "| 维度 | 指标 | 数值 |",
+        "| --- | --- | --- |",
+        f"| 规模 | 笔记总数 | {agg['note_count']} |",
+        f"| 规模 | 平台分布 | {plat_str} |",
+        f"| 规模 | 主要达人 | {creator_str} |",
+        f"| 基础 | 总点赞 | {_num(agg['total_likes'])} |",
+        f"| 基础 | 总收藏 | {_num(agg['total_favorites'])} |",
+        f"| 基础 | 总评论 | {_num(agg['total_comments'])} |",
+        f"| 基础 | 总分享 | {_num(agg['total_shares'])} |",
+        f"| 传播 | 总曝光量 | {_num(agg['total_views']) if agg['total_views'] else '未采集'} |",
+        f"| 互动率 | 综合互动率 | {_pct(agg['engagement_rate'])} |",
+    ])
+
+
+# ---------------------------------------------------------------------------
+# Prompt building
+# ---------------------------------------------------------------------------
+def sample_representative_comments(
+    per_note_comments: list[list[str]],
+    per_note_meta: list[dict[str, Any]],
+    *,
+    max_total: int = 30,
+    per_note_cap: int = 2,
+) -> list[str]:
+    """Pick up to `max_total` filtered comments across notes.
+
+    Tries to take `per_note_cap` from each note, prioritizing longer comments
+    (proxy for substance). If still under `max_total`, fills with remaining
+    longer comments.
+    """
+    all_comments: list[tuple[str, dict[str, Any]]] = []
+    for comments, meta in zip(per_note_comments, per_note_meta):
+        # Sort by length desc; take up to per_note_cap
+        sorted_c = sorted(comments, key=len, reverse=True)[:per_note_cap]
+        for c in sorted_c:
+            all_comments.append((c, meta))
+
+    # Final sort: longer first, then by note id asc for stability
+    all_comments.sort(key=lambda cm: (-len(cm[0]), cm[1].get("id") or 0))
+    return [c for c, _ in all_comments[:max_total]]
+
+
+def build_style_rollup_prompt(
+    style_name: str,
+    brand: str,
+    aggregated: dict[str, Any],
+    merged_distribution: dict[str, Any],
+    sample_comments: list[str],
+    note_summaries: list[str],
+) -> str:
+    """Build user_content for the style-level LLM call.
+
+    `note_summaries` are 1-2 line "title | platform | likes | comments" snippets
+    to give the LLM a sense of which notes exist.
+    """
+    target_label = brand or style_name
+    lines: list[str] = [
+        f"本款式「{style_name}」共采集 {aggregated['note_count']} 篇笔记，跨 {sum(1 for v in aggregated['platform_breakdown'].values() if v)} 个平台。",
+        f"以下是基于这 {aggregated['note_count']} 篇笔记的合并数据 + 代表性评论样本，请输出款式整体分析报告。",
+        "",
+        "## 款式汇总数据",
+        _format_metrics_md(aggregated),
+        "",
+        "## 关键词分布（合并所有笔记）",
+        _format_distribution_md(
+            merged_distribution,
+            sum((m or {}).get("overall", {}).get(k, 0)
+                for m in [merged_distribution]
+                for k in CATEGORY_LABELS),
+        ),
+        "",
+        f"## {aggregated['note_count']} 篇笔记概览（标题 | 平台 | 点赞 | 评论）",
+    ]
+    for meta in aggregated["notes_meta"]:
+        lines.append(
+            f"- #{meta['id']} | {meta['platform']} | 点赞 {meta['like_count']} | 评论 {meta['comment_count']} | {meta['title']}"
+        )
+
+    if sample_comments:
+        lines.extend([
+            "",
+            f"## 代表性评论样本（共 {len(sample_comments)} 条，来自多篇笔记）",
+        ])
+        for c in sample_comments:
+            lines.append(f"- {c}")
+
+    return "\n".join(lines)
+
+
+# ---------------------------------------------------------------------------
+# LLM call
+# ---------------------------------------------------------------------------
+def analyze_style(
+    style_name: str,
+    brand: str,
+    aggregated: dict[str, Any],
+    merged_distribution: dict[str, Any],
+    sample_comments: list[str],
+    note_summaries: list[str],
+    *,
+    doc_text_limit: int = 6000,
+) -> str:
+    """Call Hermes for the style-level rollup analysis."""
+    if not aggregated["note_count"]:
+        return "（该款式暂无笔记可分析）"
+
+    log(
+        f"调用 Hermes 生成款式整体分析（{style_name}，{aggregated['note_count']} 篇笔记，"
+        f"{aggregated['total_comments']} 条总评论）..."
+    )
+    user_content = build_style_rollup_prompt(
+        style_name, brand, aggregated, merged_distribution, sample_comments, note_summaries,
+    )
+
+    target_term = brand or style_name
+    brand_block = ""
+    if target_term:
+        brand_block = f"""
+
+【实体隔离】
+本次分析的目标：品牌「{brand or '未指定'}」、产品「{style_name}」。
+本款式可能涉及多品牌联合推广，请严格将分析对象限定为我方品牌/产品：
+- 仅将「{target_term}」相关的信息计入种草指标
+- 评论中涉及的其他品牌或产品，仅作为场景搭配或联合推广的陪衬
+"""
+
+    system_prompt = ANALYSIS_SYSTEM_PROMPT + brand_block + "\n\n" + STYLE_ROLLUP_DEEP_PROMPT
+
+    result = call_hermes_analyzer(system_prompt, user_content)
+    if len(result) > doc_text_limit:
+        result = result[:doc_text_limit] + "\n\n...（内容过长，已截断）"
+    return result
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_cooperations.bat b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_cooperations.bat
new file mode 100644
index 0000000..50f00af
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_cooperations.bat
@@ -0,0 +1,37 @@
+@echo off
+if not defined GYXX_PYTHON set "GYXX_PYTHON=python"
+REM Sync creator attributes and cooperation records from Feishu to PostgreSQL
+REM Runs after daily scrape to pick up new creators/notes/cooperations
+REM Register: schtasks /Create /SC DAILY /TN YingxiaoYunying_SyncCooperations /TR %PROJECT_DIR%\data\tools\sync_cooperations.bat /ST 09:00 /F
+
+setlocal
+
+for %%I in ("%~dp0..\..") do set "PROJECT_DIR=%%~fI"
+if not defined GYXX_DATA_ROOT (
+    if not defined GYXX_PROJECT_ROOT (
+        echo ERROR: GYXX_DATA_ROOT or GYXX_PROJECT_ROOT must be defined. 1>&2
+        exit /b 3
+    )
+    set "GYXX_DATA_ROOT=%GYXX_PROJECT_ROOT%\var"
+)
+set LOG_DIR=%GYXX_DATA_ROOT%\logs\content_marketing\scheduled
+set PYTHON=%GYXX_PYTHON%
+where %PYTHON% >nul 2>&1 || set PYTHON=python
+
+if not exist "%LOG_DIR%" mkdir "%LOG_DIR%"
+
+set TS=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%
+set TS=%TS: =0%
+
+set LOG_FILE=%LOG_DIR%\sync_cooperations_%TS%.log
+
+echo === sync_cooperations started at %date% %time% === > "%LOG_FILE%"
+
+cd /d "%PROJECT_DIR%"
+
+call %PYTHON% data\tools\sync_cooperations.py --refresh-mapping >> "%LOG_FILE%" 2>&1
+set RC=%ERRORLEVEL%
+
+echo === sync_cooperations finished at %date% %time% exit=%RC% === >> "%LOG_FILE%"
+
+endlocal & exit /b %RC%
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_cooperations.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_cooperations.py
new file mode 100644
index 0000000..bd5ccdb
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_cooperations.py
@@ -0,0 +1,463 @@
+#!/usr/bin/env python
+"""
+sync_cooperations.py — 从飞书款式合作表同步达人属性和合作记录到数据库
+====================================================================
+
+遍历 35 个款式的飞书多维表格,对每一行:
+1. upsert cmt_creators (达人属性: 名称/平台/粉丝数/主页链接/微信号/类型)
+2. upsert cmt_cooperations (合作记录: 费用/内容方向/曝光量/数据表现等)
+
+幂等: feishu_record_id 为唯一键,重复运行只更新不重复插入。
+
+用法:
+    python sync_cooperations.py                # 正式同步
+    python sync_cooperations.py --dry-run      # 仅预览
+    python sync_cooperations.py --style 7      # 只同步指定款式
+"""
+import argparse
+import json
+import os
+import re
+import sys
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+import psycopg
+from dotenv import load_dotenv
+
+BASE_DIR = PATHS.tools_root
+DATA_DIR = PATHS.raw_root
+load_dotenv(PATHS.state_root / "config/db.env")
+from gyxx_flow.modules.content_marketing.runtime import feishu_mapping  # noqa: E402
+from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_db_config  # noqa: E402
+
+DB_CONFIG = get_db_config()
+
+PLATFORM_MAP = {
+    "小红书": "xiaohongshu",
+    "抖音": "douyin",
+    "B站": "bilibili",
+}
+
+# 飞书逻辑字段 → cmt_creators 列
+CREATOR_FIELD_MAP = {
+    "creator_name": "name",
+    "platform": "platform",
+    "creator_id": "platform_id",
+    "follower_count": "follower_count",
+    "profile_url": "homepage_url",
+    "wechat": "wechat",
+    "account_type": "account_type",
+}
+
+# 飞书逻辑字段 → cmt_cooperations 列
+COOP_FIELD_MAP = {
+    "cooperation_date": "cooperation_date",
+    "cost": "cooperation_cost",
+    "ad_spend": "ad_spend",
+    "content_direction": "content_direction",
+    "note_title": "note_title",
+    "note_url": "note_url",
+    "publish_time": "publish_time",
+    "read_count_auto": "exposure_count",
+    "engagement_count": "engagement_count",
+    "data_performance": "data_performance",
+    "cpm": "cpm",
+    "tracking_no": "tracking_number",
+    "is_paid": "is_paid",
+    "is_new_direction": "is_new_direction",
+}
+
+
+def _log(msg: str) -> None:
+    safe = str(msg).encode("gbk", errors="replace").decode("gbk", errors="replace")
+    print(safe, flush=True)
+
+
+def parse_follower_count(text: str) -> int | None:
+    """解析粉丝数文本: '3.7万' → 37000, '12.4w' → 124000, '5600' → 5600"""
+    if not text or not text.strip():
+        return None
+    text = text.strip()
+    m = re.match(r"([\d.]+)\s*万", text)
+    if m:
+        return int(float(m.group(1)) * 10000)
+    m = re.match(r"([\d.]+)\s*w", text, re.IGNORECASE)
+    if m:
+        return int(float(m.group(1)) * 10000)
+    m = re.match(r"([\d,]+)", text)
+    if m:
+        return int(m.group(1).replace(",", ""))
+    return None
+
+
+def parse_engagement_count(text: str) -> int | None:
+    """解析互动赞藏数: '68.5万' → 685000"""
+    return parse_follower_count(text)
+
+
+def cell_text(cell) -> str:
+    """从飞书单元格提取纯文本"""
+    if cell is None:
+        return ""
+    if isinstance(cell, str):
+        return cell.strip()
+    if isinstance(cell, list):
+        # select/multiselect/link 等返回数组
+        parts = []
+        for item in cell:
+            if isinstance(item, str):
+                parts.append(item)
+            elif isinstance(item, dict):
+                parts.append(item.get("text", item.get("name", str(item))))
+            else:
+                parts.append(str(item))
+        return ",".join(parts)
+    if isinstance(cell, dict):
+        return cell.get("text", str(cell))
+    return str(cell).strip()
+
+
+def cell_number(cell) -> float | None:
+    """从飞书单元格提取数字"""
+    if cell is None:
+        return None
+    if isinstance(cell, (int, float)):
+        return float(cell)
+    if isinstance(cell, str):
+        try:
+            return float(cell.replace(",", ""))
+        except ValueError:
+            return None
+    return None
+
+
+def cell_bool(cell) -> bool:
+    """从飞书单元格提取布尔值"""
+    if cell is None:
+        return False
+    if isinstance(cell, bool):
+        return cell
+    if isinstance(cell, str):
+        return cell.lower() in ("true", "1", "yes")
+    return bool(cell)
+
+
+def cell_date(cell) -> str | None:
+    """从飞书单元格提取日期 (返回 YYYY-MM-DD 或 None)"""
+    text = cell_text(cell)
+    if not text:
+        return None
+    # 飞书 datetime 格式: "2026-05-06 00:00:00"
+    m = re.match(r"(\d{4}-\d{2}-\d{2})", text)
+    if m:
+        return m.group(1)
+    return None
+
+
+def resolve_platform(raw: str) -> str:
+    """飞书投放平台 → 数据库 platform 值"""
+    for cn, en in PLATFORM_MAP.items():
+        if cn in raw:
+            return en
+    return raw.lower() if raw else ""
+
+
+def get_style_id(cur, style_name: str) -> int | None:
+    """按款式名查 style_id"""
+    cur.execute("SELECT id FROM cmt_styles WHERE name = %s", (style_name,))
+    row = cur.fetchone()
+    return row[0] if row else None
+
+
+def ensure_style(cur, style_name: str) -> int:
+    """确保款式存在,不存在则插入"""
+    sid = get_style_id(cur, style_name)
+    if sid:
+        return sid
+    cur.execute(
+        "INSERT INTO cmt_styles (name) VALUES (%s) ON CONFLICT (name) DO UPDATE SET name = EXCLUDED.name RETURNING id",
+        (style_name,),
+    )
+    return cur.fetchone()[0]
+
+
+def upsert_creator(cur, name: str, platform: str, attrs: dict) -> int:
+    """upsert 达人,返回 creator_id"""
+
+    # 尝试按 (name, platform) 查找
+    cur.execute(
+        "SELECT id FROM cmt_creators WHERE name = %s AND platform = %s",
+        (name, platform),
+    )
+    row = cur.fetchone()
+    if row:
+        creator_id = row[0]
+        # 更新属性(只更新非空值)
+        updates = {}
+        for col, val in attrs.items():
+            if val and val != "":
+                updates[col] = val
+        if updates:
+            set_clause = ", ".join(f"{k} = %s" for k in updates)
+            vals = list(updates.values()) + [creator_id]
+            cur.execute(f"UPDATE cmt_creators SET {set_clause} WHERE id = %s", vals)
+        return creator_id
+
+    # 插入新达人
+    cols = ["name", "platform"] + list(attrs.keys())
+    placeholders = ", ".join(["%s"] * len(cols))
+    col_str = ", ".join(cols)
+    vals = [name, platform] + list(attrs.values())
+    cur.execute(
+        f"INSERT INTO cmt_creators ({col_str}) VALUES ({placeholders}) RETURNING id",
+        vals,
+    )
+    return cur.fetchone()[0]
+
+
+def upsert_cooperation(cur, style_id: int, creator_id: int,
+                       feishu_record_id: str, attrs: dict) -> bool:
+    """upsert 合作记录,返回是否插入(True)或更新(False)"""
+    if not feishu_record_id:
+        return False
+
+    # 检查是否已存在
+    cur.execute(
+        "SELECT id FROM cmt_cooperations WHERE feishu_record_id = %s",
+        (feishu_record_id,),
+    )
+    row = cur.fetchone()
+    if row:
+        # 更新
+        updates = {"style_id": style_id, "creator_id": creator_id}
+        for col, val in attrs.items():
+            updates[col] = val
+        set_clause = ", ".join(f"{k} = %s" for k in updates)
+        vals = list(updates.values()) + [row[0]]
+        cur.execute(f"UPDATE cmt_cooperations SET {set_clause} WHERE id = %s", vals)
+        return False
+
+    # 插入
+    cols = ["style_id", "creator_id", "feishu_record_id"] + list(attrs.keys())
+    placeholders = ", ".join(["%s"] * len(cols))
+    col_str = ", ".join(cols)
+    vals = [style_id, creator_id, feishu_record_id] + list(attrs.values())
+    cur.execute(
+        f"INSERT INTO cmt_cooperations ({col_str}) VALUES ({placeholders})",
+        vals,
+    )
+    return True
+
+
+def sync_table(cur, table_info: dict, dry_run: bool = False) -> dict:
+    """同步一个款式表的所有记录"""
+    style_name = table_info.get("name", "")
+    base_token = table_info.get("base_token", "")
+    table_id = table_info.get("table_id", "")
+    field_map = table_info.get("field_map", {})
+
+    if not base_token or not table_id:
+        return {"style": style_name, "error": "missing base_token/table_id"}
+
+    style_id = ensure_style(cur, style_name)
+    stats = {"style": style_name, "creators": 0, "cooperations_new": 0, "cooperations_updated": 0, "skipped": 0, "skipped_no_creator": 0}
+
+    # 拉取全部记录
+    records = feishu_mapping._record_list_all(base_token, table_id)
+    _log(f"  [{style_name}] 拉到 {len(records)} 条记录")
+
+    # 构建 field_id → logical_key 反向映射
+    fid_to_lkey = {}
+    for lkey, finfo in field_map.items():
+        fid_to_lkey[finfo["field_id"]] = lkey
+
+    for rec in records:
+        record_id = rec.get("record_id", "")
+
+        # _record_list_all 返回扁平结构: record_id + 各 field_id 直接在顶层
+        # 提取逻辑字段值
+        row = {}
+        for fid, val in rec.items():
+            if fid == "record_id":
+                continue
+            lkey = fid_to_lkey.get(fid)
+            if lkey:
+                row[lkey] = val
+
+        # 达人名称
+        creator_name = cell_text(row.get("creator_name", ""))
+        if not creator_name:
+            stats["skipped_no_creator"] += 1
+            continue
+
+        # 平台
+        raw_platform = cell_text(row.get("platform", ""))
+        platform = resolve_platform(raw_platform)
+
+        # 达人属性
+        creator_attrs = {}
+        raw_follower = cell_text(row.get("follower_count", ""))
+        if raw_follower:
+            creator_attrs["follower_count"] = raw_follower
+            follower_num = parse_follower_count(raw_follower)
+            if follower_num is not None:
+                creator_attrs["follower_count_num"] = follower_num
+
+        platform_id = cell_text(row.get("creator_id", ""))
+        if platform_id:
+            creator_attrs["platform_id"] = platform_id
+
+        homepage = cell_text(row.get("profile_url", ""))
+        if homepage:
+            creator_attrs["homepage_url"] = homepage
+
+        wechat = cell_text(row.get("wechat", ""))
+        if wechat:
+            creator_attrs["wechat"] = wechat
+
+        account_type = cell_text(row.get("account_type", ""))
+        if account_type:
+            creator_attrs["account_type"] = account_type
+
+        if not dry_run:
+            creator_id = upsert_creator(cur, creator_name, platform, creator_attrs)
+            stats["creators"] += 1
+        else:
+            stats["creators"] += 1
+            creator_id = -1  # dry-run placeholder
+
+        # 合作记录属性
+        coop_attrs = {}
+
+        coop_date = cell_date(row.get("cooperation_date", ""))
+        if coop_date:
+            coop_attrs["cooperation_date"] = coop_date
+
+        cost = cell_number(row.get("cost", ""))
+        if cost is not None:
+            coop_attrs["cooperation_cost"] = cost
+
+        ad_spend = cell_number(row.get("ad_spend", ""))
+        if ad_spend is not None:
+            coop_attrs["ad_spend"] = ad_spend
+
+        content_dir = cell_text(row.get("content_direction", ""))
+        if content_dir:
+            coop_attrs["content_direction"] = content_dir
+
+        note_title = cell_text(row.get("note_title", ""))
+        if note_title:
+            coop_attrs["note_title"] = note_title
+
+        note_url = cell_text(row.get("note_url", ""))
+        if note_url:
+            coop_attrs["note_url"] = note_url
+
+        pub_time = cell_text(row.get("publish_time", ""))
+        if pub_time:
+            # 飞书格式 "2026-06-09 00:00:00"
+            coop_attrs["publish_time"] = pub_time
+
+        exposure = cell_number(row.get("read_count_auto", ""))
+        if exposure is not None:
+            coop_attrs["exposure_count"] = int(exposure)
+
+        raw_engagement = cell_text(row.get("engagement_count", ""))
+        if raw_engagement:
+            coop_attrs["engagement_count"] = raw_engagement
+            eng_num = parse_engagement_count(raw_engagement)
+            if eng_num is not None:
+                coop_attrs["engagement_count_num"] = eng_num
+
+        data_perf = cell_text(row.get("data_performance", ""))
+        if data_perf:
+            coop_attrs["data_performance"] = data_perf
+
+        cpm_val = cell_number(row.get("cpm", ""))
+        if cpm_val is not None:
+            coop_attrs["cpm"] = cpm_val
+
+        tracking = cell_text(row.get("tracking_no", ""))
+        if tracking:
+            coop_attrs["tracking_number"] = tracking
+
+        is_paid = cell_bool(row.get("is_paid", False))
+        coop_attrs["is_paid"] = is_paid
+
+        is_new_dir = cell_bool(row.get("is_new_direction", False))
+        coop_attrs["is_new_direction"] = is_new_dir
+
+        if not dry_run:
+            is_new = upsert_cooperation(cur, style_id, creator_id, record_id, coop_attrs)
+            if is_new:
+                stats["cooperations_new"] += 1
+            else:
+                stats["cooperations_updated"] += 1
+        else:
+            stats["cooperations_new"] += 1
+
+    return stats
+
+
+def main():
+    parser = argparse.ArgumentParser(description="从飞书款式合作表同步达人属性和合作记录到数据库")
+    parser.add_argument("--dry-run", action="store_true", help="仅预览,不写入数据库")
+    parser.add_argument("--style", type=int, help="只同步指定款式编号 (如 7=宙斯)")
+    parser.add_argument("--refresh-mapping", action="store_true",
+                        help="强制刷新 mapping (发现新款/新表/新字段); 不加则用缓存")
+    args = parser.parse_args()
+
+    _log("=== sync_cooperations.py 开始 ===")
+    if args.dry_run:
+        _log("[DRY-RUN] 仅预览模式")
+
+    # 加载 mapping; 加 --refresh-mapping 时强制刷新以发现新款/新表/新字段
+    if args.refresh_mapping:
+        _log("[step 0] 刷新 feishu_mapping (发现新款/新表/新字段)...")
+        mapping = feishu_mapping.refresh_mapping(force=True)
+    else:
+        mapping = feishu_mapping.load_mapping()
+    tables = mapping.get("tables", [])
+
+    if args.style:
+        tables = [t for t in tables if t.get("index") == args.style]
+        if not tables:
+            _log(f"未找到款式编号 {args.style}")
+            return
+
+    _log(f"共 {len(tables)} 个款式待同步")
+
+    conn = psycopg.connect(**DB_CONFIG)
+    cur = conn.cursor()
+
+    total_creators = 0
+    total_coop_new = 0
+    total_coop_upd = 0
+
+    for t in tables:
+        try:
+            stats = sync_table(cur, t, dry_run=args.dry_run)
+            _log(f"  [{stats['style']}] 达人={stats['creators']}, "
+                 f"合作新增={stats['cooperations_new']}, 更新={stats['cooperations_updated']}, "
+                 f"跳过={stats['skipped']}")
+            total_creators += stats["creators"]
+            total_coop_new += stats["cooperations_new"]
+            total_coop_upd += stats["cooperations_updated"]
+        except Exception as e:
+            _log(f"  [ERROR] {t.get('name', '?')}: {e}")
+            conn.rollback()
+            continue
+
+    if not args.dry_run:
+        conn.commit()
+        _log(f"=== 同步完成: 达人={total_creators}, 合作新增={total_coop_new}, 更新={total_coop_upd} ===")
+    else:
+        _log(f"=== DRY-RUN 预览: 达人={total_creators}, 合作={total_coop_new} ===")
+
+    cur.close()
+    conn.close()
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_metrics_to_cmt_notes.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_metrics_to_cmt_notes.py
new file mode 100644
index 0000000..21e1632
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_metrics_to_cmt_notes.py
@@ -0,0 +1,472 @@
+#!/usr/bin/env python
+"""
+把 yingxiaoyunying 每日采集的曝光量(view_count)同步到
+comment-data-collector 的 cmt_notes 表。
+
+数据源:data/v2_results/*_v2.json / *_xingtu_v2.json / *_bilibili_v2.json
+(由 xingtu_scraper_v2.py / pgy_xhs_scraper_v2.py / bilibili_scraper.py 每天产出)
+目标表:cmt_notes(由 D:\\comment-data-collector 维护)
+
+本脚本只写 view_count 一个指标。like_count / comment_count / favorite_count /
+share_count / scraped_at 全部由 comment-data-collector 维护,yingxiaoyunying 不碰,
+避免双写冲突 / 列名歧义(收藏在 xhs 里叫 collect_count,在 cmt_notes 里对应
+favorite_count,跨表语义容易错位)。
+
+平台映射:
+    xingtu_v2  -> douyin
+    _v2(无后缀) -> xiaohongshu
+    _bilibili_v2 -> bilibili
+
+INSERT 路径:对 v2_results 有但 cmt_notes 没的 record_id,先调 feishu +record-list
+回填 url / title / creator_name(用款式对应多维表格的 base_token),resolve
+style_id / creator_id 后 INSERT cmt_notes 行(只写必要字段)。同一物理笔记不同
+record_id(URL 已存在)走 merge-by-url:把这次的 view_count 写进已存在的行。
+
+用法:
+    python sync_metrics_to_cmt_notes.py                # 正式同步(含 INSERT)
+    python sync_metrics_to_cmt_notes.py --dry-run      # 仅预览
+    python sync_metrics_to_cmt_notes.py --no-insert-missing   # 只 UPDATE 不 INSERT
+"""
+import argparse
+import datetime as dt
+import json
+import os
+import re
+import sys
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+import psycopg
+from dotenv import load_dotenv
+
+
+def configure_console_output(stdout, stderr) -> None:
+    """Keep Windows GBK logs alive when creator names contain Unicode symbols."""
+    for stream in (stdout, stderr):
+        if stream is None:
+            continue
+        try:
+            stream.reconfigure(errors="replace")
+        except (AttributeError, OSError, ValueError):
+            pass
+
+
+configure_console_output(sys.stdout, sys.stderr)
+
+BASE_DIR = PATHS.tools_root
+DATA_DIR = PATHS.raw_root
+V2_DIR = PATHS.normalized_root / "v2_results"
+load_dotenv(PATHS.state_root / "config/db.env")
+from gyxx_flow.modules.content_marketing.runtime.data.tools.v2_filename import parse_v2_filename  # noqa: E402
+from gyxx_flow.modules.content_marketing.runtime import feishu_mapping  # noqa: E402
+from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_db_config  # noqa: E402
+
+
+DB_CONFIG = get_db_config()
+
+
+def parse_int(v) -> int | None:
+    """'1.2万' / '6,592' / '2422' / 2422 -> int"""
+    if v is None:
+        return None
+    if isinstance(v, (int, float)):
+        return int(v)
+    txt = str(v).strip()
+    if not txt:
+        return None
+    m = re.match(r"^([\d,.]+)\s*(万|w|W)?", txt)
+    if not m:
+        return None
+    num = float(m.group(1).replace(",", ""))
+    if m.group(2):
+        num *= 10000
+    return int(num)
+
+
+def get_conn():
+    return psycopg.connect(**DB_CONFIG)
+
+
+# ============================================================
+# INSERT 路径辅助:回填 feishu + 解析 FK
+# ============================================================
+_URL_RE = re.compile(r"https?://[^\s)>\]]+")
+
+
+def _extract_url(text) -> str | None:
+    """从飞书链接单元格 '[url](url)' 或裸 URL 中提取纯 URL"""
+    if not text:
+        return None
+    s = str(text).strip()
+    if not s:
+        return None
+    m = _URL_RE.search(s)
+    return m.group(0) if m else s
+
+
+def _parse_publish_time(text) -> dt.datetime | None:
+    """飞书发布时间 '2026-06-09 00:00:00' / '2026-06-09' / 毫秒时间戳 -> datetime"""
+    if text is None:
+        return None
+    if isinstance(text, (int, float)):
+        try:
+            return dt.datetime.fromtimestamp(int(text) / 1000)
+        except Exception:
+            return None
+    s = str(text).strip()
+    if not s:
+        return None
+    for fmt in ("%Y-%m-%d %H:%M:%S", "%Y-%m-%d", "%Y/%m/%d"):
+        try:
+            return dt.datetime.strptime(s, fmt)
+        except ValueError:
+            continue
+    return None
+
+
+# (base_token, table_id) -> {record_id: record_dict}
+_BACKFILL_CACHE: dict[tuple[str, str], dict[str, dict]] = {}
+
+
+def _backfill_record(mapping: dict, style_name: str, rid: str) -> dict | None:
+    """从飞书合作达人多维表格拉指定 record_id 的完整字段。
+    返回 {title, url, publish_time, creator_name} 或 None(找不到/拉取失败)。
+    同一 (base_token, table_id) 在本进程内只拉一次。
+    """
+    style = next((t for t in mapping.get("tables", []) if t["name"] == style_name), None)
+    if not style:
+        return None
+    bt, tid = style["base_token"], style["table_id"]
+    fm = style.get("field_map", {})
+    # note_url 必须有; creator_name 自营表可能没有,用款式名兜底
+    if not fm.get("note_url"):
+        return None
+
+    cache_key = (bt, tid)
+    if cache_key not in _BACKFILL_CACHE:
+        try:
+            rows = feishu_mapping._record_list_all(bt, tid)
+        except Exception as exc:
+            print(f"  [backfill-err] {style_name}({bt}/{tid}): {exc}")
+            return None
+        _BACKFILL_CACHE[cache_key] = {r.get("record_id"): r for r in rows}
+
+    rec = _BACKFILL_CACHE[cache_key].get(rid)
+    if not rec:
+        return None
+
+    def gv(key: str):
+        fid = fm.get(key, {}).get("field_id")
+        return rec.get(fid) if fid else None
+
+    creator_name = gv("creator_name") or f"自营-{style_name}"
+    # title 剥掉 markdown 链接 [text](url) -> 只留 text,再剥裸 URL,避免超 varchar(512)
+    raw_title = gv("note_title")
+    if isinstance(raw_title, list):
+        raw_title = " ".join(str(x) for x in raw_title if x)
+    title_clean = re.sub(r"\[([^\]]*)\]\([^)]*\)", r"\1", str(raw_title or ""))
+    title_clean = re.sub(r"https?://[^\s）)\]]+", "", title_clean).strip()
+    return {
+        "title": title_clean[:500],
+        "url": _extract_url(gv("note_url")),
+        "creator_name": creator_name,
+        "publish_time": _parse_publish_time(gv("publish_time")),
+    }
+
+
+def _style_id_lookup(cur) -> dict[str, int]:
+    """name -> cmt_styles.id 映射(一次查全表)"""
+    cur.execute("SELECT id, name FROM cmt_styles")
+    return {name: id for id, name in cur.fetchall()}
+
+
+def _ensure_creator(cur, name: str | None) -> int | None:
+    """按 name 查找或新建 cmt_creators 行,返回 id。name 为空返回 None。"""
+    if not name:
+        return None
+    cur.execute("SELECT id FROM cmt_creators WHERE name = %s", (name,))
+    row = cur.fetchone()
+    if row:
+        return row[0]
+    cur.execute(
+        "INSERT INTO cmt_creators (name, created_at, updated_at) "
+        "VALUES (%s, CURRENT_TIMESTAMP, CURRENT_TIMESTAMP) RETURNING id",
+        (name,),
+    )
+    return cur.fetchone()[0] if cur.rowcount else None
+
+
+def detect_platform(path: Path) -> tuple[str | None, bool]:
+    """文件名后缀 -> (cmt_notes.platform, self_operated)"""
+    _, _, plat, self_op = parse_v2_filename(path.name)
+    return {"xhs": "xiaohongshu", "xt": "douyin", "bili": "bilibili"}.get(plat), self_op
+
+
+def _is_syncable_item(item: dict) -> bool:
+    """Only sync rows that represent a real matched scrape result."""
+    if not isinstance(item, dict):
+        return False
+    if item.get("error"):
+        return False
+    if item.get("matched") is False:
+        return False
+    if item.get("ok") is False:
+        return False
+    return True
+
+
+def collect_metrics_from_v2() -> list[dict]:
+    """扫一遍 v2_results,聚合每个 record_id 的最新指标"""
+    out: list[dict] = []
+    for path in sorted(V2_DIR.glob("*.json")):
+        result = detect_platform(path)
+        if not result or result[0] is None:
+            continue
+        platform, self_operated = result
+        idx, style_name, _, _ = parse_v2_filename(path.name)
+        if idx is None:
+            continue
+        try:
+            data = json.loads(path.read_text(encoding="utf-8"))
+        except Exception:
+            continue
+
+        if platform == "bilibili":
+            items = data.get("details") or []
+            for item in items:
+                if not _is_syncable_item(item):
+                    continue
+                rid = item.get("record_id")
+                view_count = parse_int(item.get("play_count"))
+                if not rid or view_count is None:
+                    continue
+                out.append({
+                    "src_platform": "bili",
+                    "dst_platform": platform,
+                    "style_idx": idx,
+                    "style_name": style_name,
+                    "record_id": rid,
+                    "creator_name": item.get("creator"),
+                    "view_count": view_count,
+                    "like_count": None,
+                    "collect_count": None,
+                    "comment_count": None,
+                    "self_operated": self_operated,
+                })
+        else:
+            items = data.get("results") or []
+            for item in items:
+                if not _is_syncable_item(item):
+                    continue
+                rid = item.get("record_id")
+                view_count = parse_int(item.get("play_count") or item.get("read_count"))
+                if not rid or view_count is None:
+                    continue
+                out.append({
+                    "src_platform": "xt" if platform == "douyin" else "xhs",
+                    "dst_platform": platform,
+                    "style_idx": idx,
+                    "style_name": style_name,
+                    "record_id": rid,
+                    "creator_name": item.get("creator"),
+                    "view_count": view_count,
+                    "like_count": parse_int(item.get("like_count")),
+                    "collect_count": parse_int(item.get("collect_count")),
+                    "comment_count": parse_int(item.get("comment_count")),
+                    "self_operated": self_operated,
+                })
+    return out
+
+
+def sync(dry_run: bool = False, insert_missing: bool = True) -> dict:
+    metrics = collect_metrics_from_v2()
+    print(f"[sync] v2_results 中可同步的命中记录: {len(metrics)} 条")
+
+    matched = updated = skipped_not_found = 0
+    inserted = skipped_insert = 0
+
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            style_name_to_id: dict[str, int] = _style_id_lookup(cur)
+            creator_id_cache: dict[str, int | None] = {}
+            mapping = None
+            self_mapping = None
+            if insert_missing:
+                try:
+                    mapping = feishu_mapping.load_mapping(force_refresh=False)
+                except Exception as exc:
+                    print(f"[WARN] feishu_mapping 加载失败,INSERT 路径禁用: {exc}")
+                    insert_missing = False
+                try:
+                    self_mapping = feishu_mapping.load_mapping(force_refresh=False, self_operated=True)
+                except Exception as exc:
+                    print(f"[WARN] feishu_mapping(自营) 加载失败: {exc}")
+
+            for row in metrics:
+                record_id = row["record_id"]
+                dst_platform = row["dst_platform"]
+
+                cur.execute(
+                    """
+                    SELECT id, view_count FROM cmt_notes
+                    WHERE feishu_record_id = %s AND platform = %s
+                    """,
+                    (record_id, dst_platform),
+                )
+                note = cur.fetchone()
+                if not note:
+                    skipped_not_found += 1
+                    if not insert_missing:
+                        continue
+                    # 选择正确的 mapping: 自营用 self_mapping, 合作用 mapping
+                    active_mapping = self_mapping if row.get("self_operated") else mapping
+                    if not active_mapping:
+                        continue
+                    # ---- INSERT path ----
+                    style_name = row.get("style_name")
+                    style_id = style_name_to_id.get(style_name or "")
+                    if not style_id:
+                        print(f"  [skip-insert] rid={record_id}: style '{style_name}' 不在 cmt_styles")
+                        skipped_insert += 1
+                        continue
+                    bf = _backfill_record(active_mapping, style_name, record_id)
+                    if not bf or not bf.get("url"):
+                        print(f"  [skip-insert] rid={record_id}: feishu 回填失败或无 URL")
+                        skipped_insert += 1
+                        continue
+                    creator_name = (bf.get("creator_name") or row.get("creator_name") or "").strip()
+                    if not creator_name:
+                        print(f"  [skip-insert] rid={record_id}: creator 名为空")
+                        skipped_insert += 1
+                        continue
+                    if dry_run:
+                        cur.execute("SELECT id FROM cmt_creators WHERE name = %s", (creator_name,))
+                        creator_row = cur.fetchone()
+                        creator_id = creator_row[0] if creator_row else None
+                    else:
+                        if creator_name not in creator_id_cache:
+                            creator_id_cache[creator_name] = _ensure_creator(cur, creator_name)
+                        creator_id = creator_id_cache[creator_name]
+                    if not creator_id:
+                        reason = "creator 不存在(dry-run 不新建)" if dry_run else "creator 名为空"
+                        print(f"  [skip-insert] rid={record_id}: {reason}")
+                        skipped_insert += 1
+                        continue
+                    # 先查 URL 是否已存在(避免 INSERT 触发 UNIQUE 整事务回滚)
+                    cur.execute(
+                        "SELECT id, view_count FROM cmt_notes WHERE url = %s LIMIT 1",
+                        (bf["url"],),
+                    )
+                    url_row = cur.fetchone()
+                    if url_row:
+                        old_id, old_view = url_row
+                        new_view = row.get("view_count")
+                        if new_view is not None and new_view != old_view:
+                            if dry_run:
+                                print(f"  [dry-merge-by-url] note_id={old_id} view {old_view}->{new_view} (新 rid={record_id})")
+                            else:
+                                cur.execute(
+                                    "UPDATE cmt_notes SET view_count = %s, updated_at = CURRENT_TIMESTAMP WHERE id = %s",
+                                    (new_view, old_id),
+                                )
+                                print(f"  [merge-by-url] note_id={old_id} view {old_view}->{new_view} (新 rid={record_id})")
+                            updated += 1
+                        else:
+                            print(f"  [skip-merge] rid={record_id}: url 已存在(note_id={old_id}),view 无变化")
+                        skipped_insert += 1
+                        continue
+                    if dry_run:
+                        print(f"  [dry-insert] rid={record_id} style={style_name} creator={creator_name} "
+                              f"view={row.get('view_count')} title={str(bf.get('title') or '')[:32]}")
+                        inserted += 1
+                        continue
+                    # 只写必要字段:FK + title + url + feishu_record_id + view_count
+                    # like/comment/favorite/share/scraped_at 全留给 comment-data-collector
+                    cur.execute(
+                        """
+                        INSERT INTO cmt_notes
+                            (style_id, creator_id, platform, title, url,
+                             feishu_record_id, view_count, self_operated,
+                             publish_time,
+                             created_at, updated_at)
+                        VALUES
+                            (%s, %s, %s, %s, %s,
+                             %s, %s, %s,
+                             %s,
+                             CURRENT_TIMESTAMP, CURRENT_TIMESTAMP)
+                        """,
+                        (style_id, creator_id, dst_platform,
+                         bf.get("title"), bf["url"], record_id,
+                         row.get("view_count"), row.get("self_operated", False),
+                         bf.get("publish_time")),
+                    )
+                    print(f"  [insert] rid={record_id} style={style_name} creator={creator_name} view={row.get('view_count')}")
+                    inserted += 1
+                    continue
+
+                # ---- UPDATE path: 只写 view_count,其它字段(like/comment/favorite/scraped)留给 cdc ----
+                note_id, old_view = note
+                matched += 1
+                new_view = row["view_count"]
+
+                # 缺 publish_time 时从飞书补
+                cur.execute("SELECT publish_time FROM cmt_notes WHERE id = %s", (note_id,))
+                old_pub_row = cur.fetchone()
+                if old_pub_row and old_pub_row[0] is None:
+                    try:
+                        bf_update = _backfill_record(mapping, style_name, record_id) or \
+                                   (self_mapping and _backfill_record(self_mapping, style_name, record_id)) or {}
+                        new_pub = bf_update.get("publish_time")
+                    except Exception:
+                        new_pub = None
+                    if new_pub is not None:
+                        cur.execute(
+                            "UPDATE cmt_notes SET publish_time = %s, updated_at = CURRENT_TIMESTAMP WHERE id = %s",
+                            (new_pub, note_id),
+                        )
+
+                if new_view is not None and new_view != old_view:
+                    if dry_run:
+                        print(f"  [dry-run] note_id={note_id} view {old_view}->{new_view}")
+                    else:
+                        cur.execute(
+                            "UPDATE cmt_notes SET view_count = %s, updated_at = CURRENT_TIMESTAMP WHERE id = %s",
+                            (new_view, note_id),
+                        )
+                        print(f"  [update] note_id={note_id} view={new_view}")
+                    updated += 1
+
+        if dry_run:
+            conn.rollback()
+        else:
+            conn.commit()
+
+    mode = "DRY-RUN" if dry_run else "WRITE"
+    insert_mode = "ON" if insert_missing else "OFF"
+    print(f"\n[sync] 完成 ({mode}, insert={insert_mode}): "
+          f"匹配={matched}, 更新={updated}, 新增={inserted}, "
+          f"cmt_notes 未找到={skipped_not_found}, 跳过 INSERT={skipped_insert}")
+    return {
+        "total": len(metrics),
+        "matched": matched,
+        "updated": updated,
+        "inserted": inserted,
+        "skipped_not_found": skipped_not_found,
+        "skipped_insert": skipped_insert,
+    }
+
+
+def main():
+    parser = argparse.ArgumentParser(description="同步 v2_results 曝光/收藏量到 cmt_notes")
+    parser.add_argument("--dry-run", action="store_true", help="仅预览,不写入")
+    parser.add_argument("--no-insert-missing", action="store_true",
+                        help="关闭 INSERT 路径(默认开启:对 v2_results 有但 cmt_notes 缺的行,回填 feishu 后 INSERT)")
+    args = parser.parse_args()
+
+    print(f"[sync] 模式: {'DRY-RUN' if args.dry_run else '正式写入'}  "
+          f"insert-missing: {'OFF' if args.no_insert_missing else 'ON'}")
+    sync(dry_run=args.dry_run, insert_missing=not args.no_insert_missing)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_style_categories.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_style_categories.py
new file mode 100644
index 0000000..c28b6af
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/sync_style_categories.py
@@ -0,0 +1,142 @@
+"""Sync the Feishu product-style classification into ``cmt_styles``.
+
+The source Base is read with the authenticated lark-cli user identity because
+the Hermes bot is not a member of that table.
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import subprocess
+from pathlib import Path
+from typing import Any
+
+try:
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_conn
+except ModuleNotFoundError:  # direct execution: python data/tools/sync_style_categories.py
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_conn
+
+
+BASE_TOKEN = "TH1JbxhfUaxetis6F4pcoqvWncd"
+TABLE_ID = "tbl8q0hbIPMNn0k8"
+VIEW_ID = "vew7j57rli"
+NAME_FIELD_ID = "fldRqiyAJW"
+CATEGORY_FIELD_ID = "fldsEuUtYb"
+PROFILE = "hermes-analyzer"
+
+NAME_ALIASES = {
+    "盖亚微单Pro": "盖亚微单",
+    "星迹&星迹2": "星迹2",
+    "宙斯双肩包": "宙斯",
+    "极星Pro": "极星pro",
+    "瑞白双肩包": "瑞白",
+    "逐星gt": "逐星GT",
+    "觅光双肩包": "觅光",
+    "云栖相机双肩包": "云栖",
+    "拾影双肩电脑包": "拾影双肩",
+    "拾影相机包": "拾影斜挎相机包",
+    "极星单肩包": "极星单肩",
+}
+
+
+def normalize_style_name(value: str) -> str:
+    """Convert the Feishu lifecycle-document label to the DB canonical name."""
+    label = re.sub(r"^\[|\]\(.*\)$", "", (value or "").strip())
+    label = re.sub(r"\s*生命进程\s*", "", label)
+    label = re.sub(r"\s*[（(]\s*新\s*[）)]\s*", "", label)
+    label = re.sub(r"\s+", "", label).strip()
+    return NAME_ALIASES.get(label, label)
+
+
+def extract_style_mapping(payload: dict[str, Any]) -> dict[str, str]:
+    """Extract ``{canonical product name: style category}`` from CLI JSON."""
+    mapping: dict[str, str] = {}
+    for row in payload.get("data", {}).get("data", []):
+        if not isinstance(row, list) or len(row) < 2 or not row[0] or not row[1]:
+            continue
+        name = normalize_style_name(str(row[0]))
+        category_value = row[1]
+        category = (
+            str(category_value[0]).strip()
+            if isinstance(category_value, list) and category_value
+            else str(category_value).strip()
+        )
+        if not name or not category:
+            continue
+        previous = mapping.get(name)
+        if previous and previous != category:
+            raise ValueError(f"款式 {name} 同时出现多个风格：{previous} / {category}")
+        mapping[name] = category
+    return mapping
+
+
+def _lark_cli_path() -> str:
+    appdata = os.environ.get("APPDATA")
+    if not appdata:
+        raise RuntimeError("APPDATA 未设置，无法定位 lark-cli.cmd")
+    return str(Path(appdata) / "npm" / "lark-cli.cmd")
+
+
+def load_mapping_from_feishu() -> dict[str, str]:
+    command = [
+        _lark_cli_path(), "--profile", PROFILE, "base", "+record-list",
+        "--as", "user", "--base-token", BASE_TOKEN, "--table-id", TABLE_ID,
+        "--view-id", VIEW_ID, "--field-id", NAME_FIELD_ID,
+        "--field-id", CATEGORY_FIELD_ID, "--limit", "200", "--format", "json",
+    ]
+    result = subprocess.run(command, capture_output=True, text=True, encoding="utf-8", check=True)
+    payload = json.loads(result.stdout)
+    if not payload.get("ok"):
+        raise RuntimeError(f"读取飞书款式风格失败：{payload}")
+    mapping = extract_style_mapping(payload)
+    if not mapping:
+        raise RuntimeError("飞书表未返回任何有效的款式—风格记录")
+    return mapping
+
+
+def sync_mapping(mapping: dict[str, str], *, dry_run: bool = False) -> dict[str, Any]:
+    """Update existing styles and insert new source styles without overwriting Base links."""
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute("SELECT name FROM cmt_styles")
+            existing = {row[0] for row in cur.fetchall()}
+            inserted = sorted(set(mapping) - existing)
+            updated = sorted(set(mapping) & existing)
+            if not dry_run:
+                cur.execute(
+                    "ALTER TABLE cmt_styles ADD COLUMN IF NOT EXISTS style_category VARCHAR(64) DEFAULT ''"
+                )
+                cur.execute(
+                    "COMMENT ON COLUMN cmt_styles.style_category IS "
+                    "'款式风格，来源于飞书全域事项表的风格单选字段'"
+                )
+                for name, category in sorted(mapping.items()):
+                    cur.execute(
+                        """
+                        INSERT INTO cmt_styles (name, style_category)
+                        VALUES (%s, %s)
+                        ON CONFLICT (name) DO UPDATE SET
+                            style_category = EXCLUDED.style_category,
+                            updated_at = CURRENT_TIMESTAMP
+                        """,
+                        (name, category),
+                    )
+                conn.commit()
+    return {"source_count": len(mapping), "updated": updated, "inserted": inserted}
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="同步飞书款式风格到 cmt_styles")
+    parser.add_argument("--dry-run", action="store_true", help="只核对，不写数据库")
+    args = parser.parse_args()
+    mapping = load_mapping_from_feishu()
+    result = sync_mapping(mapping, dry_run=args.dry_run)
+    print(json.dumps(result, ensure_ascii=False, indent=2))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/v2_filename.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/v2_filename.py
new file mode 100644
index 0000000..a9e1929
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/v2_filename.py
@@ -0,0 +1,50 @@
+"""v2_results/*.json 文件名解析(共享给 check_status / retry_failed / sync_metrics_to_cmt_notes)"""
+import re
+
+# 后缀越长优先级越高(避免 _bilibili_v2 被 _v2 误吃)
+# 自营达人文件带 _self_ 前缀: *_self_bilibili_v2.json, *_self_xingtu_v2.json, *_self_v2.json
+RE_V2_SELF_BILI = re.compile(r"^(?P<idx>\d+)-.+_self_bilibili_v2\.json$")
+RE_V2_SELF_XT = re.compile(r"^(?P<idx>\d+)-.+_self_xingtu_v2\.json$")
+RE_V2_SELF_PGY = re.compile(r"^(?P<idx>\d+)-.+_self_v2\.json$")
+RE_V2_BILI = re.compile(r"^(?P<idx>\d+)-.+_bilibili_v2\.json$")
+RE_V2_XT = re.compile(r"^(?P<idx>\d+)-.+_xingtu_v2\.json$")
+RE_V2_PGY = re.compile(r"^(?P<idx>\d+)-.+_v2\.json$")  # 必须放最后兜底
+
+SUFFIX_LEN = {
+    "self_bili": len("_self_bilibili_v2.json"),
+    "self_xt": len("_self_xingtu_v2.json"),
+    "self_xhs": len("_self_v2.json"),
+    "bili": len("_bilibili_v2.json"),
+    "xt": len("_xingtu_v2.json"),
+    "xhs": len("_v2.json"),
+}
+
+
+def parse_v2_filename(name: str) -> tuple[int | None, str, str | None, bool]:
+    """01-盖亚斜挎_v2.json -> (1, '盖亚斜挎', 'xhs', False)
+    02-极星pro_self_xingtu_v2.json -> (2, '极星pro', 'xt', True)
+    03-盖亚微单_bilibili_v2.json -> (3, '盖亚微单', 'bili', False)
+    不匹配返回 (None, '', None, False)
+    第 4 个返回值: self_operated (True=自营达人)
+    """
+    if RE_V2_SELF_BILI.match(name):
+        m, plat, self_op = RE_V2_SELF_BILI.match(name), "bili", True
+    elif RE_V2_SELF_XT.match(name):
+        m, plat, self_op = RE_V2_SELF_XT.match(name), "xt", True
+    elif RE_V2_SELF_PGY.match(name):
+        m, plat, self_op = RE_V2_SELF_PGY.match(name), "xhs", True
+    elif RE_V2_BILI.match(name):
+        m, plat, self_op = RE_V2_BILI.match(name), "bili", False
+    elif RE_V2_XT.match(name):
+        m, plat, self_op = RE_V2_XT.match(name), "xt", False
+    elif RE_V2_PGY.match(name):
+        m, plat, self_op = RE_V2_PGY.match(name), "xhs", False
+    else:
+        return None, "", None, False
+
+    idx = int(m.group("idx"))
+    suffix_key = f"self_{plat}" if self_op else plat
+    stem = name[: -SUFFIX_LEN[suffix_key]]
+    parts = stem.split("-", 1)
+    name_only = parts[1] if len(parts) == 2 else ""
+    return idx, name_only, plat, self_op
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/validate_mapping.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/validate_mapping.py
new file mode 100644
index 0000000..d9354c3
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/validate_mapping.py
@@ -0,0 +1,105 @@
+"""
+validate_mapping.py
+-------------------
+校验 data/config/款式_多维表格_对照.json 中的 field_id 跟飞书实际字段是否匹配。
+不匹配报警(但允许差异,因为有些表故意没字段)。
+
+用法:
+  python validate_mapping.py            # 检查 + 打印结果
+  python validate_mapping.py --strict   # 严格模式:任何差异都返回非 0
+"""
+import argparse
+import json
+import os
+import subprocess
+import sys
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+LARK = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK):
+    LARK = "lark-cli.cmd"
+MAPPING_CACHE_PATH = PATHS.normalized_root / "mappings" / "款式_多维表格_对照.json"
+MAPPING_SEED_PATH = PATHS.config_root / "款式_多维表格_对照.json"
+MAPPING_PATH = MAPPING_CACHE_PATH if MAPPING_CACHE_PATH.exists() else MAPPING_SEED_PATH
+ENV = {**os.environ, "LARK_CLI_NO_PROXY": "1", "PYTHONIOENCODING": "utf-8"}
+
+
+def fetch_fields(base_token: str, table_id: str) -> list[dict]:
+    r = subprocess.run(
+        [LARK, "base", "+field-list",
+         "--base-token", base_token,
+         "--table-id", table_id,
+         "--format", "json"],
+        capture_output=True, env=ENV, timeout=60
+    )
+    if r.returncode != 0:
+        return []
+    raw = r.stdout
+    if isinstance(raw, bytes):
+        raw = raw.decode("utf-8", errors="replace")
+    try:
+        d = json.loads(raw)
+        return d.get("data", {}).get("fields", []) if d.get("ok") else []
+    except Exception:
+        return []
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--strict", action="store_true", help="任何差异都返回非 0 退出码")
+    args = ap.parse_args()
+
+    if not MAPPING_PATH.exists():
+        print(f"[FATAL] 找不到 {MAPPING_PATH}")
+        sys.exit(1)
+
+    mapping = json.loads(MAPPING_PATH.read_text(encoding="utf-8"))
+    tables = mapping.get("tables", [])
+
+    print(f"校验 {len(tables)} 张表...\n")
+    issues = []
+    for t in tables:
+        idx = t["index"]
+        name = t["name"]
+        fmap = t.get("field_map", {})
+
+        # 拉取实际字段
+        actual_fields = fetch_fields(t["base_token"], t["table_id"])
+        if not actual_fields:
+            issues.append(f"[{idx:02d}] {name}: 拉取失败,跳过")
+            continue
+        actual_by_id = {f["id"]: f["name"] for f in actual_fields}
+        actual_by_name = {f["name"]: f["id"] for f in actual_fields}
+
+        # 1) 校验 JSON 里的 field_id 仍然存在
+        for logical, info in fmap.items():
+            fid = info.get("field_id")
+            fname = info.get("field_name")
+            if fid not in actual_by_id:
+                issues.append(
+                    f"[{idx:02d}] {name}: 字段 {logical}({fname}) 的 field_id={fid} 不存在了!"
+                )
+            elif actual_by_id[fid] != fname:
+                issues.append(
+                    f"[{idx:02d}] {name}: 字段 {logical}: JSON 说 {fname}({fid}) 但飞书是 {actual_by_id[fid]}"
+                )
+        # 2) 校验核心字段都映射上了
+        for key in ["creator_name", "read_count_7d"]:
+            if key not in fmap:
+                issues.append(f"[{idx:02d}] {name}: 缺核心字段 {key}")
+
+    # 报告
+    if not issues:
+        print("✓ 全部 24 张表字段映射正确,核心字段都存在")
+        sys.exit(0)
+    else:
+        print(f"发现 {len(issues)} 个问题:\n")
+        for i in issues:
+            print(f"  ✗ {i}")
+        print(f"\n建议: python rebuild_mapping.py --in-place")
+        sys.exit(1 if args.strict else 0)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/weekly_comment_scrape.bat b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/weekly_comment_scrape.bat
new file mode 100644
index 0000000..fab424e
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/weekly_comment_scrape.bat
@@ -0,0 +1,80 @@
+@echo off
+if not defined GYXX_PYTHON set "GYXX_PYTHON=python"
+REM Weekly three-platform comment scraper wrapper.
+REM Runs bilibili, xiaohongshu, douyin in parallel via PowerShell.
+REM Register: schtasks /Create /SC WEEKLY /D SUN /TN YingxiaoYunying_WeeklyCommentScrape /TR %PROJECT_DIR%\data\tools\weekly_comment_scrape.bat /ST 12:00 /F
+
+setlocal
+
+for %%I in ("%~dp0..\..") do set "PROJECT_DIR=%%~fI"
+if not defined GYXX_DATA_ROOT (
+    if not defined GYXX_PROJECT_ROOT (
+        echo ERROR: GYXX_DATA_ROOT or GYXX_PROJECT_ROOT must be defined. 1>&2
+        exit /b 3
+    )
+    set "GYXX_DATA_ROOT=%GYXX_PROJECT_ROOT%\var"
+)
+set "PYTHON_EXE=%GYXX_PYTHON%"
+set "LOG_DIR=%GYXX_DATA_ROOT%\logs\content_marketing"
+set "TMP_DIR=%GYXX_DATA_ROOT%\tmp\content_marketing"
+
+if not exist "%LOG_DIR%" mkdir "%LOG_DIR%"
+if not exist "%TMP_DIR%" mkdir "%TMP_DIR%"
+
+for /f "tokens=2 delims==" %%I in ('wmic os get localdatetime /value') do set "NOW=%%I"
+set "STAMP=%NOW:~0,4%-%NOW:~4,2%-%NOW:~6,2%_%NOW:~8,2%-%NOW:~10,2%-%NOW:~12,2%"
+
+set "MASTER_LOG=%LOG_DIR%\weekly_%STAMP%.log"
+echo === weekly scrape started %STAMP% > "%MASTER_LOG%"
+
+set "DOUYIN_LOG=%LOG_DIR%\weekly_%STAMP%_douyin.log"
+set "XHS_LOG=%LOG_DIR%\weekly_%STAMP%_xiaohongshu.log"
+set "BILI_LOG=%LOG_DIR%\weekly_%STAMP%_bilibili.log"
+
+echo [%STAMP%] launching douyin, xiaohongshu, bilibili in parallel >> "%MASTER_LOG%"
+
+set "PS_SCRIPT=%GYXX_DATA_ROOT%\tmp\content_marketing\weekly_scrape_%STAMP%.ps1"
+> "%PS_SCRIPT%" echo $ErrorActionPreference = 'Continue'
+>> "%PS_SCRIPT%" echo $env:PYTHONIOENCODING = 'utf-8'
+>> "%PS_SCRIPT%" echo $env:PYTHONUTF8 = '1'
+>> "%PS_SCRIPT%" echo Set-Location -LiteralPath '%PROJECT_DIR%'
+>> "%PS_SCRIPT%" echo $py = '%GYXX_PYTHON%'
+>> "%PS_SCRIPT%" echo $stamp = '%STAMP%'
+>> "%PS_SCRIPT%" echo $logDir = '%GYXX_DATA_ROOT%\logs\content_marketing'
+>> "%PS_SCRIPT%" echo $jobs = @(
+>> "%PS_SCRIPT%" echo     [pscustomobject]@{ Name='bilibili';    Script='%PROJECT_DIR%\data\tools\batch_rescrape_bilibili.py';    Log=Join-Path $logDir ("weekly_${stamp}_bilibili.log") },
+>> "%PS_SCRIPT%" echo     [pscustomobject]@{ Name='xiaohongshu'; Script='%PROJECT_DIR%\data\tools\batch_rescrape_xiaohongshu.py'; Log=Join-Path $logDir ("weekly_${stamp}_xiaohongshu.log") },
+>> "%PS_SCRIPT%" echo     [pscustomobject]@{ Name='douyin';      Script='%PROJECT_DIR%\data\tools\batch_rescrape_douyin.py';      Log=Join-Path $logDir ("weekly_${stamp}_douyin.log") }
+>> "%PS_SCRIPT%" echo )
+>> "%PS_SCRIPT%" echo $started = @{}
+>> "%PS_SCRIPT%" echo foreach ($job in $jobs) {
+>> "%PS_SCRIPT%" echo     $logPath = $job.Log
+>> "%PS_SCRIPT%" echo     $header = "[$(Get-Date -Format 'yyyy-MM-dd HH:mm:ss')] === $($job.Name) launch ==="
+>> "%PS_SCRIPT%" echo     $cmd = "`"$py`" -X utf8 `"$($job.Script)`" >> `"$logPath`" 2>&1"
+>> "%PS_SCRIPT%" echo     Add-Content -LiteralPath $logPath -Value $header -Encoding utf8
+>> "%PS_SCRIPT%" echo     $proc = Start-Process -FilePath cmd.exe -ArgumentList '/c', "`"$cmd`"" -PassThru -WindowStyle Hidden
+>> "%PS_SCRIPT%" echo     if ($null -eq $proc) {
+>> "%PS_SCRIPT%" echo         Add-Content -LiteralPath $logPath -Value "[$(Get-Date -Format 'yyyy-MM-dd HH:mm:ss')] !!! Start-Process returned null for $($job.Name) !!!" -Encoding utf8
+>> "%PS_SCRIPT%" echo         continue
+>> "%PS_SCRIPT%" echo     }
+>> "%PS_SCRIPT%" echo     $started[$job.Name] = @{ Proc=$proc; Log=$logPath }
+>> "%PS_SCRIPT%" echo }
+>> "%PS_SCRIPT%" echo $exitCodes = @{}
+>> "%PS_SCRIPT%" echo foreach ($job in $jobs) {
+>> "%PS_SCRIPT%" echo     $entry = $started[$job.Name]
+>> "%PS_SCRIPT%" echo     $entry.Proc.WaitForExit()
+>> "%PS_SCRIPT%" echo     $rc = $entry.Proc.ExitCode
+>> "%PS_SCRIPT%" echo     $exitCodes[$job.Name] = $rc
+>> "%PS_SCRIPT%" echo     Add-Content -LiteralPath $entry.Log -Value "[$(Get-Date -Format 'yyyy-MM-dd HH:mm:ss')] === $($job.Name) done, exit=$rc ===" -Encoding utf8
+>> "%PS_SCRIPT%" echo }
+>> "%PS_SCRIPT%" echo foreach ($k in $exitCodes.Keys) { Write-Output ("{0}={1}" -f $k, $exitCodes[$k]) }
+
+powershell.exe -NoProfile -ExecutionPolicy Bypass -File "%PS_SCRIPT%" >> "%MASTER_LOG%" 2>&1
+set "RC=%ERRORLEVEL%"
+
+del "%PS_SCRIPT%" 2>nul
+
+echo [%STAMP%] all three scrapers finished, ps exit=%RC% >> "%MASTER_LOG%"
+echo === weekly scrape done %STAMP% >> "%MASTER_LOG%"
+
+endlocal & exit /b %RC%
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/weekly_summary.bat b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/weekly_summary.bat
new file mode 100644
index 0000000..eb1ecb0
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/weekly_summary.bat
@@ -0,0 +1,42 @@
+@echo off
+if not defined GYXX_PYTHON set "GYXX_PYTHON=python"
+REM Weekly summary: 每周一 10:00 跑上周各款周笔记汇总
+REM 1. 动态算上周一到上周日时间范围
+REM 2. PG 查上周有笔记的款
+REM 3. 对每个款:跑单品分析 + LLM 4 维度综合 + 创建飞书文档 + 写入生命进程表
+REM Register: schtasks /Create /SC WEEKLY /D TUE /TN YingxiaoYunying_WeeklySummary /TR %PROJECT_DIR%\data\tools\weekly_summary.bat /ST 10:00 /F
+
+setlocal
+
+for %%I in ("%~dp0..\..") do set "PROJECT_DIR=%%~fI"
+if not defined GYXX_DATA_ROOT (
+    if not defined GYXX_PROJECT_ROOT (
+        echo ERROR: GYXX_DATA_ROOT or GYXX_PROJECT_ROOT must be defined. 1>&2
+        exit /b 3
+    )
+    set "GYXX_DATA_ROOT=%GYXX_PROJECT_ROOT%\var"
+)
+set LOG_DIR=%GYXX_DATA_ROOT%\logs\content_marketing\scheduled
+set PYTHON=%GYXX_PYTHON%
+where %PYTHON% >nul 2>&1 || set PYTHON=python
+
+if not exist "%LOG_DIR%" mkdir "%LOG_DIR%"
+
+set TS=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%
+set TS=%TS: =0%
+
+set LOG_FILE=%LOG_DIR%\weekly_summary_%TS%.log
+
+echo === Weekly summary started at %date% %time% === > "%LOG_FILE%"
+echo PROJECT_DIR=%PROJECT_DIR% >> "%LOG_FILE%"
+
+cd /d "%PROJECT_DIR%"
+
+REM 跑周汇总(4 路并发,动态时间范围)
+call %PYTHON% -u -X utf8 weekly_summary_all.py --max-workers 4 >> "%LOG_FILE%" 2>&1
+set RC=%ERRORLEVEL%
+echo weekly_summary exit=%RC% >> "%LOG_FILE%"
+
+echo === Weekly summary finished at %date% %time% (rc=%RC%) === >> "%LOG_FILE%"
+
+endlocal
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/data/tools/write_notes_to_doc.py b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/write_notes_to_doc.py
new file mode 100644
index 0000000..b67f0eb
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/data/tools/write_notes_to_doc.py
@@ -0,0 +1,437 @@
+"""Backfill per-note + per-style analyses into a single Feishu Docx archive.
+
+Layout:
+  - ONE archive doc (token stored under state.docs.archive)
+    - Top: H1 款式索引 + table (sorted by note count DESC, intra-doc #style-N anchors)
+    - Body: per-style H1 sections, each with:
+      - H2 款式整体分析 (LLM rollup across all notes of the style)
+      - H2 单篇笔记分析 (N 篇) containing per-note H3 sections
+
+Idempotent: re-runs read state/feishu_doc.json and skip already-written notes + styles.
+"""
+
+import argparse
+import sys
+import threading
+import time
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from pathlib import Path
+from typing import Any
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+# Path setup
+_TOOLS_DIR = PATHS.tools_root
+_PROJECT_ROOT = PATHS.module_root
+
+try:
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")  # type: ignore[attr-defined]
+    sys.stderr.reconfigure(encoding="utf-8", errors="replace")  # type: ignore[attr-defined]
+except Exception:
+    pass
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools import db  # noqa: E402
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools import feishu_doc_writer as fdw  # noqa: E402
+from gyxx_flow.modules.content_marketing.runtime.data.tools import analyze_note  # noqa: E402
+from gyxx_flow.modules.content_marketing.runtime.data.tools import analyze_comments  # noqa: E402
+from gyxx_flow.modules.content_marketing.runtime.data.tools import style_analyzer  # noqa: E402
+
+
+# Serializes state mutation + state.json writes across worker threads.
+_state_lock = threading.Lock()
+
+
+def log(msg: str) -> None:
+    print(f"[{time.strftime('%H:%M:%S')}] {msg}", flush=True)
+
+
+# ---------------------------------------------------------------------------
+# Archive doc lifecycle
+# ---------------------------------------------------------------------------
+def get_or_create_archive(state: dict[str, Any]) -> dict[str, Any]:
+    docs = state.setdefault("docs", {})
+    existing = docs.get("archive")
+    if existing and existing.get("token"):
+        return existing
+    log("  Creating archive doc ...")
+    info = fdw.create_doc(fdw._doc_skeleton())
+    docs["archive"] = {
+        "token": info["token"],
+        "url": info["url"],
+        "revision_id": info.get("revision_id"),
+    }
+    return docs["archive"]
+
+
+# ---------------------------------------------------------------------------
+# Per-note processing (within a style)
+# ---------------------------------------------------------------------------
+def process_note(
+    note_row: dict[str, Any], *, archive_token: str, force: bool,
+    state: dict[str, Any],
+) -> dict[str, Any]:
+    """Per-note pipeline: load → filter → metrics → LLM analysis.
+
+    Returns the data needed for `process_style` to assemble the full style
+    section XML. Does NOT append anything to the doc — per-note sections are
+    appended only as part of the style section to keep doc order tidy.
+    """
+    note_id = note_row["id"]
+    with _state_lock:
+        written = state.setdefault("written_notes", {}).get(str(note_id))
+    if written and not force:
+        return {"note_id": note_id, "status": "skipped", "reason": "already-written"}
+
+    try:
+        info, raw_comments = analyze_note.load_note_from_db(note_id=note_id)
+    except Exception as exc:
+        return {"note_id": note_id, "status": "error", "stage": "load", "error": str(exc)}
+
+    info["note_id"] = note_id
+
+    try:
+        filtered_comments = analyze_comments.code_filter_comments(raw_comments)
+    except Exception as exc:
+        return {"note_id": note_id, "status": "error", "stage": "filter", "error": str(exc)}
+
+    if not filtered_comments:
+        return {"note_id": note_id, "status": "skipped", "reason": "no-filtered-comments"}
+
+    metrics = analyze_note.compute_metrics(info)
+    target_terms = [t for t in (info.get("brand"), info.get("style_name")) if t]
+    distribution = analyze_note.keyword_distribution(filtered_comments, target_terms=target_terms)
+
+    try:
+        analysis = analyze_note.analyze_note(
+            info, metrics, distribution, filtered_comments,
+        )
+    except Exception as exc:
+        return {"note_id": note_id, "status": "error", "stage": "llm", "error": str(exc)}
+
+    # Persist the analysis text in state so re-runs can rebuild the doc without
+    # re-calling the LLM (idempotency + crash recovery).
+    with _state_lock:
+        state["written_notes"][str(note_id)] = {
+            "platform": info.get("platform"),
+            "doc_token": archive_token,
+            "bookmark_name": f"note-{note_id}",
+            "style_id": note_row.get("style_id"),
+            "analysis_text": analysis,
+            "written_at": time.strftime("%Y-%m-%dT%H:%M:%S"),
+        }
+        fdw.save_state(state)
+    return {
+        "note_id": note_id,
+        "status": "ok",
+        "title": info.get("title", ""),
+        "filtered_comments": filtered_comments,
+        "info": info,
+        "metrics": metrics,
+        "distribution": distribution,
+        "analysis": analysis,
+    }
+
+
+# ---------------------------------------------------------------------------
+# Per-style processing
+# ---------------------------------------------------------------------------
+def process_style(
+    style_row: dict[str, Any], state: dict[str, Any], archive_token: str, *,
+    max_notes: int | None, force: bool, force_rollup: bool, workers: int,
+) -> tuple[int, int, int, bool]:
+    style_id = style_row["id"]
+    style_name = style_row["name"]
+
+    notes = db.get_notes_by_style(style_id, only_with_comments=True)
+    if max_notes is not None:
+        notes = notes[:max_notes]
+    log(f"[{style_name}] notes to process: {len(notes)} (workers={workers})")
+
+    # ---- Phase 1: per-note parallel ----
+    note_results: list[dict[str, Any]] = []
+    ok = fail = skipped = 0
+    with ThreadPoolExecutor(max_workers=max(1, workers)) as pool:
+        future_to_note = {
+            pool.submit(
+                process_note,
+                note, archive_token=archive_token, force=force,
+                state=state,
+            ): note
+            for note in notes
+        }
+        for future in as_completed(future_to_note):
+            result = future.result()
+            status = result.get("status")
+            nid = result.get("note_id")
+            if status == "ok":
+                ok += 1
+                log(f"  [{style_name}] ok note_id={nid} title={result.get('title', '')[:30]!r}")
+                note_results.append(result)
+            elif status == "skipped":
+                skipped += 1
+                reason = result.get("reason")
+                log(f"  [{style_name}] skip note_id={nid} reason={reason}")
+                # If it was previously written, we still want its data for the
+                # rollup + style section rebuild. Re-derive from DB and use the
+                # saved analysis_text from state (no LLM re-call).
+                if reason == "already-written":
+                    try:
+                        with _state_lock:
+                            prev_entry = state.get("written_notes", {}).get(str(nid)) or {}
+                        prev_analysis = prev_entry.get("analysis_text")
+                        prev_info, prev_raw = analyze_note.load_note_from_db(note_id=nid)
+                        prev_info["note_id"] = nid
+                        prev_filtered = analyze_comments.code_filter_comments(prev_raw)
+                        if prev_filtered and prev_analysis:
+                            prev_metrics = analyze_note.compute_metrics(prev_info)
+                            prev_target = [t for t in (prev_info.get("brand"), prev_info.get("style_name")) if t]
+                            prev_dist = analyze_note.keyword_distribution(prev_filtered, target_terms=prev_target)
+                            note_results.append({
+                                "note_id": nid,
+                                "status": "ok",
+                                "title": prev_info.get("title", ""),
+                                "filtered_comments": prev_filtered,
+                                "info": prev_info,
+                                "metrics": prev_metrics,
+                                "distribution": prev_dist,
+                                "analysis": prev_analysis,
+                            })
+                    except Exception as exc:
+                        log(f"  [{style_name}] WARN: could not re-derive note_id={nid} for rollup: {exc}")
+            else:
+                fail += 1
+                log(f"  [{style_name}] FAIL note_id={nid} stage={result.get('stage')} error={result.get('error')}")
+
+    if not note_results:
+        log(f"[{style_name}] no successful notes — skipping rollup + style section")
+        with _state_lock:
+            state.setdefault("written_styles", {})[str(style_id)] = {
+                "style_name": style_name,
+                "doc_token": archive_token,
+                "note_count": 0,
+                "bookmark_name": f"style-{style_id}",
+                "skipped": True,
+                "written_at": time.strftime("%Y-%m-%dT%H:%M:%S"),
+            }
+            fdw.save_state(state)
+        return ok, fail, skipped, True
+
+    # ---- Phase 2: style-level rollup ----
+    notes_meta = note_results  # already has {id, platform, title, like_count, comment_count, ...} via info
+    # Reuse info rows (which already have id/platform/title/etc) for aggregation.
+    notes_as_rows = [
+        {
+            "id": r["info"].get("note_id"),
+            "platform": r["info"].get("platform"),
+            "title": r["info"].get("title", ""),
+            "like_count": r["info"].get("like_count"),
+            "favorite_count": r["info"].get("favorite_count"),
+            "comment_count": r["info"].get("comment_count_metric") or r["info"].get("comment_count_total"),
+            "share_count": r["info"].get("share_count"),
+            "view_count": r["info"].get("view_count"),
+            "creator_name": r["info"].get("creator_name", ""),
+        }
+        for r in note_results
+    ]
+    aggregated = style_analyzer.aggregate_style_metrics(notes_as_rows)
+    merged_distribution = style_analyzer.merge_style_distribution(
+        [r.get("distribution") or {} for r in note_results]
+    )
+
+    # ---- Phase 3: decide if rollup needed ----
+    skip_rollup = False
+    with _state_lock:
+        existing_style = state.setdefault("written_styles", {}).get(str(style_id))
+    if existing_style and not force_rollup and not force:
+        log(f"[{style_name}] rollup already written; skipping LLM call")
+        rollup_text = existing_style.get("rollup_text", "（已有整体分析）")
+        skip_rollup = True
+    else:
+        sample_comments = style_analyzer.sample_representative_comments(
+            [r.get("filtered_comments") or [] for r in note_results],
+            notes_as_rows,
+            max_total=30,
+            per_note_cap=2,
+        )
+        try:
+            rollup_text = style_analyzer.analyze_style(
+                style_name,
+                brand="",  # brand not currently carried in style_row; left empty
+                aggregated=aggregated,
+                merged_distribution=merged_distribution,
+                sample_comments=sample_comments,
+                note_summaries=[],
+            )
+        except Exception as exc:
+            log(f"[{style_name}] FAIL rollup LLM: {exc}")
+            rollup_text = f"（整体分析生成失败：{exc}）"
+
+    # ---- Phase 4: build style section XML and append to archive ----
+    # The per-note XML sections must be built here (not in worker) because we
+    # need the XML to assemble into the style section in one append call.
+    # We can however reuse the `analysis` text already produced.
+    note_sections_xml_parts: list[str] = []
+    for r in note_results:
+        info = r["info"]
+        if not r.get("analysis"):
+            log(f"  [{style_name}] WARN: note_id={r['note_id']} has no analysis text; skipping in doc")
+            continue
+        note_xml = fdw.build_note_section_xml(
+            info,
+            r["metrics"],
+            r["distribution"],
+            # raw_comments not preserved across runs; pass filtered as fallback
+            r.get("filtered_comments") or [],
+            r.get("filtered_comments") or [],
+            r["analysis"],
+            bookmark_name=f"note-{r['note_id']}",
+            doc_token=archive_token,
+        )
+        note_sections_xml_parts.append(note_xml)
+
+    style_section_xml = fdw.build_style_section_xml(
+        style_id=style_id,
+        style_name=style_name,
+        aggregated=aggregated,
+        rollup_text=rollup_text,
+        note_sections_xml="".join(note_sections_xml_parts),
+        doc_token=archive_token,
+    )
+
+    try:
+        fdw.append_blocks(archive_token, style_section_xml)
+    except Exception as exc:
+        log(f"[{style_name}] FAIL append style section: {exc}")
+        return ok, fail, skipped, skip_rollup
+
+    with _state_lock:
+        state["written_styles"][str(style_id)] = {
+            "style_name": style_name,
+            "doc_token": archive_token,
+            "note_count": aggregated["note_count"],
+            "bookmark_name": f"style-{style_id}",
+            "rollup_written_at": time.strftime("%Y-%m-%dT%H:%M:%S"),
+            "rollup_text": rollup_text,
+        }
+        fdw.save_state(state)
+    log(f"[{style_name}] DONE: notes={aggregated['note_count']} rollup={'skipped' if skip_rollup else 'new'}")
+    return ok, fail, skipped, skip_rollup
+
+
+# ---------------------------------------------------------------------------
+# Index table (replace-on-update)
+# ---------------------------------------------------------------------------
+def populate_index_table(state: dict[str, Any], archive_token: str) -> None:
+    archive = state.get("docs", {}).get("archive")
+    if not archive:
+        log("  WARN: no archive doc in state; skipping index population")
+        return
+
+    written_styles = state.get("written_styles", {})
+    # Sort by note_count DESC; alphabetical fallback for ties (already sorted in db.get_all_styles_with_notes)
+    sorted_styles = dict(
+        sorted(
+            written_styles.items(),
+            key=lambda kv: (-(kv[1].get("note_count") or 0), kv[1].get("style_name", "")),
+        )
+    )
+    index_table_xml = fdw.build_style_index_table_xml(sorted_styles, archive_url=archive["url"])
+
+    # Anchor: H1 "款式索引" — we'll insert AFTER it.
+    anchor = fdw.find_block_after(archive_token, "款式索引")
+    if not anchor:
+        log("  WARN: could not find anchor h1 '款式索引'; appending to end of archive doc")
+        fdw.append_blocks(archive_token, index_table_xml)
+        return
+
+    # Replace-on-update: delete ONLY the index blocks (between anchor and the
+    # first H1 after it). Deleting everything after the anchor would nuke the
+    # style sections we just wrote.
+    stale_block_ids = fdw.find_index_blocks(archive_token, anchor)
+    if stale_block_ids:
+        log(f"  Clearing {len(stale_block_ids)} stale index blocks ...")
+        try:
+            fdw.delete_blocks(archive_token, stale_block_ids)
+        except Exception as exc:
+            log(f"  WARN: stale-block delete failed: {exc}; new index will be appended (may produce duplicates)")
+
+    fdw.insert_blocks_after(archive_token, anchor, index_table_xml)
+    log(f"  Inserted index table with {len(sorted_styles)} style rows")
+
+
+# ---------------------------------------------------------------------------
+# Main
+# ---------------------------------------------------------------------------
+def run_all(args: argparse.Namespace) -> int:
+    state = fdw.load_state()
+    log(f"Loaded state: docs={list(state.get('docs', {}).keys())} "
+        f"written_notes={len(state.get('written_notes', {}))} "
+        f"written_styles={len(state.get('written_styles', {}))}")
+
+    archive = get_or_create_archive(state)
+    fdw.save_state(state)
+    log(f"Archive doc: {archive['url']}")
+
+    # Decide which styles to process.
+    if args.style == "all":
+        styles = db.get_all_styles_with_notes()
+    else:
+        # Single style by id
+        style_id = int(args.style)
+        styles = [s for s in db.get_all_styles_with_notes() if s["id"] == style_id]
+        if not styles:
+            log(f"  WARN: style_id={style_id} not found or has no notes")
+            return 1
+
+    total_ok = total_fail = total_skip = 0
+    for style_row in styles:
+        ok, fail, skipped, _ = process_style(
+            style_row, state, archive["token"],
+            max_notes=args.max_notes, force=args.force,
+            force_rollup=args.force_rollup, workers=args.workers,
+        )
+        total_ok += ok
+        total_fail += fail
+        total_skip += skipped
+
+    log("Populating index table ...")
+    try:
+        populate_index_table(state, archive["token"])
+    except Exception as exc:
+        log(f"  WARN: index table population failed: {exc}")
+    fdw.save_state(state)
+
+    log(f"DONE: notes ok={total_ok} fail={total_fail} skipped={total_skip} "
+        f"styles={len(state.get('written_styles', {}))}")
+    log(f"Archive: {archive['url']}")
+    return 0 if total_fail == 0 else 1
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Backfill note + style analyses into one Feishu Docx")
+    parser.add_argument("--style", default="all",
+                        help="款式 id（如 8 = 宙斯），或 'all' 处理所有款式（默认）")
+    parser.add_argument("--max-notes", type=int, default=None,
+                        help="每个款式最多处理 N 条（测试用）")
+    parser.add_argument("--force", action="store_true",
+                        help="忽略 state 中已写入标记，强制重写所有笔记 + 整体分析")
+    parser.add_argument("--force-rollup", action="store_true",
+                        help="仅重写款式整体分析（保留单篇笔记）")
+    parser.add_argument("--workers", type=int, default=3,
+                        help="并行 worker 数（默认 3；设为 1 退化为串行）")
+    parser.add_argument("--smoke-test", action="store_true",
+                        help="只跑权限冒烟测试，不写任何笔记")
+    args = parser.parse_args()
+
+    if args.smoke_test:
+        return 0 if fdw.permission_smoke_test() else 1
+
+    return run_all(args)
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/douyin_comment_scraper.py b/src/gyxx_flow/modules/content_marketing/runtime/douyin_comment_scraper.py
new file mode 100644
index 0000000..49af4bb
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/douyin_comment_scraper.py
@@ -0,0 +1,766 @@
+import argparse
+import csv
+import json
+import re
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+from typing import Any
+from urllib.parse import parse_qsl, urlencode, urlsplit, urlunsplit
+
+from playwright.sync_api import Page
+from scrapling.fetchers import DynamicSession
+
+import httpx
+
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+STATE_DIR = PATHS.state_root
+NOTES_DIR = PATHS.raw_root / "notes" / "douyin"
+COOKIE_FILE = PATHS.browser_cookie_file
+PROFILE_DIR = PATHS.browser_profile_dir
+
+# Douyin comment API endpoints to intercept
+COMMENT_API_MARKS = (
+    "/aweme/v1/web/comment/list/",
+    "/aweme/v1/web/comment/list/reply/",
+)
+
+
+def log(message: str) -> None:
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {message}", flush=True)
+
+
+def load_cookies() -> list[dict[str, Any]]:
+    if not COOKIE_FILE.exists():
+        return []
+    try:
+        data = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+        return data if isinstance(data, list) else []
+    except json.JSONDecodeError:
+        log(f"Cookie file is invalid, ignoring: {COOKIE_FILE}")
+        return []
+
+
+def restore_cookies(page: Page) -> None:
+    cookies = load_cookies()
+    if not cookies:
+        return
+    try:
+        page.context.add_cookies(cookies)
+        log(f"Loaded cookies: {COOKIE_FILE}")
+    except Exception as exc:
+        log(f"Failed to load cookies, will continue with browser profile: {exc}")
+
+
+def save_state(page: Page) -> None:
+    STATE_DIR.mkdir(parents=True, exist_ok=True)
+    COOKIE_FILE.write_text(
+        json.dumps(page.context.cookies(), ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    PATHS.browser_storage_state_file.write_text(
+        json.dumps(page.context.storage_state(), ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    log(f"Saved cookies: {COOKIE_FILE}")
+
+
+def normalize_douyin_url(url: str) -> str:
+    # Strip trailing Chinese text that sometimes gets appended to share URLs
+    url = re.sub(r'(https?://v\.douyin\.com/[A-Za-z0-9_-]+/).*', r'\1', url)
+    parts = urlsplit(url)
+    params = dict(parse_qsl(parts.query, keep_blank_values=True))
+    modal_id = params.get("modal_id")
+    if modal_id and modal_id.isdigit():
+        return f"https://www.douyin.com/video/{modal_id}"
+    return url
+
+
+def resolve_aweme_id(url: str, *, page: Page | None = None) -> tuple[str, bool]:
+    """Extract aweme_id and is_note from a Douyin URL.
+
+    Returns (aweme_id, is_note). If *page* is provided and httpx redirect
+    fails, falls back to navigating the short link inside the browser.
+    """
+    parts = urlsplit(url)
+    netloc = parts.netloc.lower()
+    is_note = "/note/" in parts.path
+
+    m = re.search(r"/(?:video|note)/(\d+)", parts.path)
+    if m:
+        is_note = "/note/" in parts.path
+        return m.group(1), is_note
+
+    if netloc.startswith("v.douyin.com") or netloc.startswith("www.douyin.com"):
+        try:
+            r = httpx.get(url, follow_redirects=True, timeout=15)
+            final_url = str(r.url)
+            m = re.search(r"/(?:video|note)/(\d+)", final_url)
+            if m:
+                is_note = "/note/" in final_url
+                return m.group(1), is_note
+        except Exception as exc:
+            log(f"resolve_aweme_id: httpx redirect failed: {exc}")
+
+        # Fallback: resolve inside the browser
+        if page is not None:
+            try:
+                log("resolve_aweme_id: trying browser redirect fallback...")
+                page.goto(url, wait_until="domcontentloaded", timeout=30000)
+                final_url = page.url
+                m = re.search(r"/(?:video|note)/(\d+)", final_url)
+                if m:
+                    is_note = "/note/" in final_url
+                    log(f"resolve_aweme_id: browser fallback ok, aweme_id={m.group(1)}, is_note={is_note}")
+                    return m.group(1), is_note
+            except Exception as exc:
+                log(f"resolve_aweme_id: browser redirect also failed: {exc}")
+
+    modal_id = dict(parse_qsl(parts.query, keep_blank_values=True)).get("modal_id")
+    if modal_id and modal_id.isdigit():
+        return modal_id, False
+
+    raise ValueError(f"Cannot extract aweme_id from URL: {url}")
+
+
+def parse_comment(comment: dict[str, Any], source_url: str, level: str, parent_id: str = "") -> dict[str, Any]:
+    user = comment.get("user") or {}
+    create_time = comment.get("create_time")
+    created_at = ""
+    if isinstance(create_time, (int, float)) and create_time:
+        try:
+            created_at = datetime.fromtimestamp(create_time).isoformat(timespec="seconds")
+        except Exception:
+            created_at = str(create_time)
+
+    return {
+        "source_url": source_url,
+        "level": level,
+        "comment_id": comment.get("cid") or comment.get("comment_id") or "",
+        "reply_to_comment_id": parent_id or comment.get("reply_id") or comment.get("reply_to_reply_id") or "",
+        "nickname": user.get("nickname") or "",
+        "user_id": user.get("uid") or "",
+        "sec_uid": user.get("sec_uid") or "",
+        "text": comment.get("text") or "",
+        "digg_count": comment.get("digg_count") or 0,
+        "reply_count": comment.get("reply_comment_total") or comment.get("reply_comment_count") or 0,
+        "ip_label": comment.get("ip_label") or "",
+        "created_at": created_at,
+        "raw_create_time": create_time or "",
+    }
+
+
+def merge_comment(
+    row: dict[str, Any],
+    comments_by_id: dict[str, dict[str, Any]],
+) -> None:
+    comment_id = str(row.get("comment_id") or f"{row.get('level')}:{len(comments_by_id)}")
+    comments_by_id[comment_id] = row
+
+
+def collect_from_payload(
+    payload: dict[str, Any],
+    source_url: str,
+    comments_by_id: dict[str, dict[str, Any]],
+    stats: dict[str, Any],
+) -> None:
+    if not isinstance(payload, dict):
+        return
+    api_url = str(payload.get("__source_url", ""))
+    is_reply = "/comment/list/reply/" in api_url
+
+    if not is_reply:
+        stats["api_total"] = payload.get("total", stats.get("api_total"))
+        stats["last_has_more"] = payload.get("has_more", stats.get("last_has_more"))
+        stats["last_cursor"] = payload.get("cursor", stats.get("last_cursor"))
+
+    comments = payload.get("comments") or []
+    if not isinstance(comments, list):
+        return
+
+    parent_id = ""
+    if is_reply:
+        params = dict(parse_qsl(urlsplit(api_url).query))
+        parent_id = params.get("comment_id") or ""
+
+    for comment in comments:
+        if not isinstance(comment, dict):
+            continue
+        row = parse_comment(comment, source_url, "reply" if is_reply else "comment", parent_id)
+        merge_comment(row, comments_by_id)
+
+        # Inline replies embedded in top-level comment responses
+        inline_replies = comment.get("reply_comment") or []
+        if isinstance(inline_replies, list):
+            parent_cid = str(row.get("comment_id") or "")
+            for reply in inline_replies:
+                if isinstance(reply, dict):
+                    sub_row = parse_comment(reply, source_url, "reply", parent_cid)
+                    merge_comment(sub_row, comments_by_id)
+
+
+def setup_response_capture(page: Page, source_url: str) -> tuple[dict[str, dict[str, Any]], dict[str, Any]]:
+    comments_by_id: dict[str, dict[str, Any]] = {}
+    stats: dict[str, Any] = {"api_hits": 0, "api_total": None, "last_has_more": None, "last_cursor": None}
+
+    def handle_response(response: Any) -> None:
+        url = response.url
+        if not any(mark in url for mark in COMMENT_API_MARKS):
+            return
+        try:
+            payload = response.json()
+        except Exception:
+            return
+        if isinstance(payload, dict):
+            payload["__source_url"] = url
+        stats["api_hits"] += 1
+        # Save first-seen URL as template for direct reply fetching
+        if "/comment/list/" in url and "/reply/" not in url and not stats.get("comment_url_template"):
+            stats["comment_url_template"] = url
+        if "/comment/list/reply/" in url and not stats.get("reply_url_template"):
+            stats["reply_url_template"] = url
+        collect_from_payload(payload, source_url, comments_by_id, stats)
+
+    page.on("response", handle_response)
+    return comments_by_id, stats
+
+
+def click_if_visible(page: Page, text: str, timeout: int = 1200) -> bool:
+    try:
+        locator = page.get_by_text(text, exact=False).first
+        if locator.count():
+            locator.click(timeout=timeout)
+            page.wait_for_timeout(500)
+            return True
+    except Exception:
+        return False
+    return False
+
+
+def dismiss_popups(page: Page) -> None:
+    for text in ("我知道了", "稍后再说", "暂不登录", "取消", "跳过"):
+        click_if_visible(page, text)
+    page.evaluate(
+        """
+        () => {
+          const close = [...document.querySelectorAll('button, div, span')]
+            .filter((el) => {
+              const t = (el.innerText || '').trim();
+              const r = el.getBoundingClientRect();
+              return ['×', 'x', 'X'].includes(t) &&
+                r.width > 0 && r.width <= 52 &&
+                r.height > 0 && r.height <= 52 &&
+                r.top >= 0 && r.top < window.innerHeight * 0.75;
+            })[0];
+          close?.click();
+        }
+        """
+    )
+
+
+def maybe_wait_for_login(page: Page, login_timeout: int) -> None:
+    deadline = time.time() + login_timeout
+    while time.time() < deadline:
+        try:
+            body = page.locator("body").inner_text(timeout=5000)
+        except Exception:
+            body = ""
+        if "扫码登录" not in body and "登录后" not in body:
+            return
+        log("Douyin may require login. Please finish login in the opened Chrome window.")
+        page.wait_for_timeout(3000)
+
+
+def is_logged_in(page: Page) -> bool:
+    """抖音登录态靠 sessionid cookie,有它才能采全量评论。"""
+    for c in page.context.cookies():
+        if c.get("name") == "sessionid" and c.get("value"):
+            return True
+    return False
+
+
+def expand_visible_replies(page: Page) -> int:
+    try:
+        clicked = page.evaluate(
+            """
+            () => {
+              const visible = (el) => {
+                const r = el.getBoundingClientRect();
+                const style = window.getComputedStyle(el);
+                return r.width > 0 && r.height > 0 &&
+                  r.bottom > 0 && r.top < window.innerHeight &&
+                  style.visibility !== 'hidden' &&
+                  style.display !== 'none' &&
+                  style.opacity !== '0';
+              };
+              const candidates = [...document.querySelectorAll('button, div, span')]
+                .filter((el) => {
+                  const text = (el.innerText || '').trim();
+                  return visible(el) &&
+                    text.length <= 40 &&
+                    (/展开.*回复/.test(text) || /查看.*回复/.test(text) || /共.*条回复/.test(text) || /还有.*回复/.test(text));
+                });
+              let count = 0;
+              for (const el of candidates.slice(0, 8)) {
+                el.click();
+                count++;
+              }
+              return count;
+            }
+            """
+        )
+        if clicked:
+            page.wait_for_timeout(1200)
+        return int(clicked or 0)
+    except Exception:
+        return 0
+
+
+def scroll_comments(page: Page, max_scrolls: int, idle_rounds: int, comments_by_id: dict[str, Any]) -> None:
+    last_count = len(comments_by_id)
+    idle = 0
+    page.mouse.move(1030, 620)
+
+    for index in range(max_scrolls):
+        dismiss_popups(page)
+        expand_visible_replies(page)
+        page.mouse.wheel(0, 1000)
+        page.evaluate(
+            """
+            () => {
+              const candidates = [...document.querySelectorAll('div, section, aside')]
+                .filter((el) => {
+                  const r = el.getBoundingClientRect();
+                  return r.height > 160 &&
+                    r.width > 220 &&
+                    el.scrollHeight > el.clientHeight + 80 &&
+                    r.right > window.innerWidth * 0.35;
+                })
+                .sort((a, b) => (b.scrollHeight - b.clientHeight) - (a.scrollHeight - a.clientHeight));
+              for (const el of candidates.slice(0, 6)) {
+                el.scrollTop += 1000;
+              }
+            }
+            """
+        )
+        page.wait_for_timeout(1800)
+
+        count = len(comments_by_id)
+        if count == last_count:
+            idle += 1
+        else:
+            log(f"Loaded comments: {count}")
+            idle = 0
+            last_count = count
+        if idle >= idle_rounds:
+            log(f"No new comments after {idle_rounds} rounds, stopping scroll.")
+            break
+
+    log(f"Scroll finished after {index + 1} rounds.")
+
+
+def with_query_params(url: str, updates: dict[str, Any]) -> str:
+    parts = urlsplit(url)
+    params = dict(parse_qsl(parts.query, keep_blank_values=True))
+    for key, value in updates.items():
+        params[key] = str(value)
+    return urlunsplit((parts.scheme, parts.netloc, parts.path, urlencode(params), parts.fragment))
+
+
+def fetch_missing_top_comments(
+    page: Page,
+    source_url: str,
+    aweme_id: str,
+    comments_by_id: dict[str, dict[str, Any]],
+    stats: dict[str, Any],
+) -> None:
+    """After scrolling, proactively paginate remaining top-level comments via API."""
+    template = stats.get("comment_url_template")
+    if not template:
+        log("No comment API template captured; skipped direct top-level fetch.")
+        return
+
+    api_total = stats.get("api_total")
+    # Fall back to note_metrics.comment_count if api_total is missing
+    if not api_total:
+        metrics = stats.get("note_metrics") or {}
+        api_total = metrics.get("comment_count")
+    if api_total and len(comments_by_id) >= int(api_total):
+        return
+
+    last_cursor = stats.get("last_cursor")
+    if not last_cursor:
+        return
+
+    log(f"Top-level: collected {len(comments_by_id)} / api_total {api_total}; fetching remainder.")
+    fetched = 0
+    cursor = last_cursor
+    for _ in range(100):
+        top_url = with_query_params(template, {"cursor": cursor, "aweme_id": aweme_id})
+        try:
+            payload = page.evaluate(
+                """
+                async (url) => {
+                  const res = await fetch(url, { credentials: 'include' });
+                  return await res.json();
+                }
+                """,
+                top_url,
+            )
+        except Exception:
+            break
+        if not isinstance(payload, dict):
+            break
+        payload["__source_url"] = top_url
+        before = len(comments_by_id)
+        collect_from_payload(payload, source_url, comments_by_id, stats)
+        fetched += max(0, len(comments_by_id) - before)
+
+        has_more = payload.get("has_more")
+        next_cursor = payload.get("cursor")
+        if not has_more or next_cursor in (None, cursor):
+            break
+        cursor = next_cursor
+        page.wait_for_timeout(350)
+
+    log(f"Top-level direct fetch: {fetched} new comments; total now: {len(comments_by_id)}")
+
+
+def fetch_missing_replies(
+    page: Page,
+    source_url: str,
+    comments_by_id: dict[str, dict[str, Any]],
+    stats: dict[str, Any],
+) -> None:
+    template = stats.get("reply_url_template")
+    if not template:
+        log("No reply API template captured; skipped direct reply fetch.")
+        return
+
+    parents = [
+        row for row in list(comments_by_id.values())
+        if row.get("level") == "comment" and int(row.get("reply_count") or 0) > 0
+    ]
+    if not parents:
+        return
+
+    log(f"Fetching replies directly for {len(parents)} comments with replies.")
+    fetched = 0
+    for row in parents:
+        parent_id = str(row.get("comment_id") or "")
+        if not parent_id:
+            continue
+        cursor = 0
+        for _ in range(30):
+            reply_url = with_query_params(template, {
+                "item_id": stats.get("aweme_id", ""),
+                "comment_id": parent_id,
+                "cursor": cursor,
+            })
+            try:
+                payload = page.evaluate(
+                    """
+                    async (url) => {
+                      const res = await fetch(url, { credentials: 'include' });
+                      return await res.json();
+                    }
+                    """,
+                    reply_url,
+                )
+            except Exception:
+                break
+            if not isinstance(payload, dict):
+                break
+            payload["__source_url"] = reply_url
+            before = len(comments_by_id)
+            collect_from_payload(payload, source_url, comments_by_id, stats)
+            fetched += max(0, len(comments_by_id) - before)
+            has_more = payload.get("has_more")
+            next_cursor = payload.get("cursor")
+            if not has_more or next_cursor in (None, cursor):
+                break
+            cursor = next_cursor
+            page.wait_for_timeout(350)
+
+    log(f"Direct replies fetched/merged: {fetched}; total comments now: {len(comments_by_id)}")
+
+
+def extract_note_metrics(page: Page) -> dict[str, Any]:
+    """Extract video metrics from the Douyin page via DOM data-e2e attributes."""
+    try:
+        metrics = page.evaluate(
+            r"""
+            () => {
+              const compactNumber = (text) => {
+                text = String(text || '').replace(/,/g, '').trim();
+                const match = text.match(/(\d+(?:\.\d+)?)(\s*[万wW])?/);
+                if (!match) return 0;
+                const value = parseFloat(match[1]);
+                if (!Number.isFinite(value)) return 0;
+                return match[2] ? Math.round(value * 10000) : Math.round(value);
+              };
+
+              const titleEl = document.querySelector('[data-e2e="video-title"]') || document.querySelector('h1');
+              const title = titleEl ? titleEl.innerText.trim() : '';
+
+              const likeEl = document.querySelector('[data-e2e="video-player-digg"]');
+              const commentEl = document.querySelector('[data-e2e="feed-comment-icon"]');
+              const collectEl = document.querySelector('[data-e2e="video-player-collect"]');
+              const shareEl = document.querySelector('[data-e2e="video-player-share"]');
+
+              const like_count = likeEl ? compactNumber(likeEl.innerText) : 0;
+              const comment_count = commentEl ? compactNumber(commentEl.innerText) : 0;
+              const favorite_count = collectEl ? compactNumber(collectEl.innerText) : 0;
+              const share_count = shareEl ? compactNumber(shareEl.innerText) : 0;
+
+              if (!like_count && !comment_count) return {};
+              return {title, like_count, comment_count, favorite_count, share_count, view_count: 0};
+            }
+            """
+        )
+        return metrics if isinstance(metrics, dict) else {}
+    except Exception:
+        return {}
+
+
+def safe_filename(value: str) -> str:
+    value = re.sub(r"[<>:\"/\\|?*\s]+", "_", value).strip("_")
+    return value[:80] or "douyin_comments"
+
+
+def scrape_comments(url: str, login_timeout: int, max_scrolls: int, idle_rounds: int, headless: bool) -> tuple[list[dict[str, Any]], dict[str, Any]]:
+    url = normalize_douyin_url(url)
+    result: dict[str, Any] = {"comments": [], "stats": {}}
+
+    # Try httpx first (fast); if it fails, we'll retry inside the browser
+    aweme_id: str | None = None
+    is_note: bool = False
+    try:
+        aweme_id, is_note = resolve_aweme_id(url)
+    except ValueError as exc:
+        log(str(exc))
+
+    if aweme_id:
+        log(f"aweme_id: {aweme_id}{' (note)' if is_note else ''}")
+
+    def action(page: Page) -> None:
+        nonlocal aweme_id, is_note
+
+        # If httpx couldn't resolve the short link, use the browser as fallback
+        if aweme_id is None:
+            try:
+                aweme_id, is_note = resolve_aweme_id(url, page=page)
+                log(f"aweme_id (browser fallback): {aweme_id}{' (note)' if is_note else ''}")
+            except ValueError as exc:
+                log(str(exc))
+                return  # Cannot proceed without aweme_id
+
+        # Always set up response capture first to intercept API calls
+        comments_by_id, stats = setup_response_capture(page, url)
+        stats["aweme_id"] = aweme_id
+
+        # Use the resolved full URL instead of the original short link
+        resolved_url = f"https://www.douyin.com/note/{aweme_id}" if is_note else f"https://www.douyin.com/video/{aweme_id}"
+        page.goto(resolved_url, wait_until="domcontentloaded", timeout=90000)
+        page.wait_for_timeout(6000)
+        maybe_wait_for_login(page, login_timeout)
+        dismiss_popups(page)
+        save_state(page)
+
+        # Wait briefly for the first comment API response to capture template URLs
+        for _ in range(10):
+            if stats.get("comment_url_template") and stats.get("api_total") is not None:
+                break
+            page.wait_for_timeout(1000)
+
+        total_comments = stats.get("api_total")
+        if total_comments is not None:
+            log(f"Comment count from API: {total_comments}")
+        else:
+            log("Could not determine comment count from API response")
+
+        # Scroll to collect comments (same pattern as xiaohongshu scraper)
+        scroll_comments(page, max_scrolls, idle_rounds, comments_by_id)
+
+        # After scrolling, extract metrics and use comment_count as fallback for api_total
+        stats["note_metrics"] = extract_note_metrics(page)
+        result["title"] = stats["note_metrics"].get("title", "")
+
+        # Proactively paginate remaining top-level comments via API
+        fetch_missing_top_comments(page, url, aweme_id, comments_by_id, stats)
+
+        # Fetch missing sub-replies via API
+        fetch_missing_replies(page, url, comments_by_id, stats)
+
+        page.wait_for_timeout(1000)
+        result["comments"] = list(comments_by_id.values())
+        result["stats"] = stats
+        # logged_in 判定:cookie 实际有效但抖音反爬可能在抓取过程中清掉 sessionid,
+        # 改成"采集率 >= 90% 就算登录有效",避免误报噪音。
+        actual_count = stats.get("api_total")
+        if actual_count is None:
+            note_metrics = stats.get("note_metrics") or {}
+            actual_count = note_metrics.get("comment_count")
+        collected_count = len(result["comments"])
+        if actual_count and collected_count >= int(actual_count * 0.9):
+            result["stats"]["logged_in"] = True
+        else:
+            result["stats"]["logged_in"] = is_logged_in(page)
+
+        result["final_url"] = page.url
+        save_state(page)
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+    with DynamicSession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=str(PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=90000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=restore_cookies,
+        max_pages=1,
+    ) as session:
+        session.fetch(url, page_action=action, wait=1000)
+
+    # aweme_id 解析完全失败 → 抛异常让 batch 层处理
+    if aweme_id is None:
+        raise RuntimeError(f"Cannot extract aweme_id from URL: {url}")
+
+    # 静默失败检测:0 评论 + 空 note_metrics → 页面没加载成功
+    stats = result.get("stats", {})
+    note_metrics = stats.get("note_metrics", {})
+    if not result["comments"] and not note_metrics:
+        raise RuntimeError(
+            "page not loaded: 0 comments and no metrics "
+            "(likely cookie expired or rate-limited)"
+        )
+
+    return result["comments"], result
+
+
+def write_outputs(url: str, comments: list[dict[str, Any]], result: dict[str, Any]) -> tuple[Path, Path]:
+    NOTES_DIR.mkdir(parents=True, exist_ok=True)
+    m = re.search(r"/(?:video|note)/(\d+)", urlsplit(url).path)
+    aweme_id = m.group(1) if m else ""
+    modal = re.search(r"modal_id=(\d+)", url)
+    if aweme_id:
+        name = f"douyin_comments_{aweme_id}"
+    elif modal:
+        name = f"douyin_comments_{modal.group(1)}"
+    else:
+        name = safe_filename(url)
+    json_path = NOTES_DIR / f"{name}.json"
+    csv_path = NOTES_DIR / f"{name}.csv"
+    if not comments and json_path.exists() and csv_path.exists() and not result.get("title"):
+        log(f"No comments and no new title; keeping existing output files: {json_path}, {csv_path}")
+        return json_path, csv_path
+
+    stats = result.get("stats", {})
+    public_stats = {key: value for key, value in stats.items() if not key.endswith("_template")}
+    payload = {
+        "source_url": url,
+        "title": result.get("title", ""),
+        "scraped_at": datetime.now().isoformat(timespec="seconds"),
+        "count": len(comments),
+        "stats": public_stats,
+        "comments": comments,
+    }
+    json_path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    fieldnames = [
+        "source_url",
+        "level",
+        "comment_id",
+        "reply_to_comment_id",
+        "nickname",
+        "user_id",
+        "sec_uid",
+        "text",
+        "digg_count",
+        "reply_count",
+        "ip_label",
+        "created_at",
+        "raw_create_time",
+    ]
+    with csv_path.open("w", newline="", encoding="utf-8-sig") as file:
+        writer = csv.DictWriter(file, fieldnames=fieldnames)
+        writer.writeheader()
+        writer.writerows(comments)
+
+    return json_path, csv_path
+
+
+def login_only(login_timeout: int) -> int:
+    """只触发登录(headed),登录完成立刻退出。给 relogin_douyin.py 用。"""
+    def action(page: Page) -> None:
+        page.goto("https://www.douyin.com/", wait_until="domcontentloaded", timeout=90000)
+        page.wait_for_timeout(2500)
+        # 主动触发登录弹窗(点登录按钮)
+        try:
+            login_btn = page.get_by_text("登录", exact=True).first
+            if login_btn.count():
+                login_btn.click(timeout=3000)
+                page.wait_for_timeout(1500)
+        except Exception:
+            pass
+        maybe_wait_for_login(page, login_timeout)
+        save_state(page)
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+    with DynamicSession(
+        headless=False,
+        real_chrome=True,
+        user_data_dir=str(PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=90000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=restore_cookies,
+        max_pages=1,
+    ) as session:
+        session.fetch("https://www.douyin.com/", page_action=action, wait=1000)
+    log("[login-only] 完成")
+    return 0
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Scrape Douyin video comments via browser (Playwright + XHR interception).")
+    parser.add_argument("url", help="Douyin video/user URL (传 'login' 配合 --login-only)")
+    parser.add_argument("--login-timeout", type=int, default=300, help="等待扫码/登录秒数")
+    parser.add_argument("--max-scrolls", type=int, default=120, help="最多滚动加载次数")
+    parser.add_argument("--idle-rounds", type=int, default=8, help="连续几轮无新评论后停止")
+    parser.add_argument("--headless", action="store_true", help="无头模式；首次登录不要使用")
+    parser.add_argument("--login-only", action="store_true",
+                        help="只触发登录(headed)然后退出,给 relogin_douyin 用")
+    args = parser.parse_args()
+
+    if args.login_only:
+        return login_only(args.login_timeout)
+
+    comments, result = scrape_comments(args.url, args.login_timeout, args.max_scrolls, args.idle_rounds, args.headless)
+    json_path, csv_path = write_outputs(args.url, comments, result)
+    log(f"Final URL: {result.get('final_url', '')}")
+    log(f"API hits: {result.get('stats', {}).get('api_hits', 0)}")
+    log(f"Scraped {len(comments)} comments")
+    log(f"JSON: {json_path}")
+    log(f"CSV:  {csv_path}")
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/feishu_mapping.py b/src/gyxx_flow/modules/content_marketing/runtime/feishu_mapping.py
new file mode 100644
index 0000000..05ae5c3
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/feishu_mapping.py
@@ -0,0 +1,427 @@
+"""
+feishu_mapping.py — 动态读取「合作达人」/「自营达人」多维表格地址,重建款式对照表
+=================================================================================
+
+之前每个款对应的飞书表格是写死在 data/config/款式_多维表格_对照.json 里的,
+现在改成动态读取下面这张"合作达人"多维表格地址表:
+
+  https://bu0zgpibak.feishu.cn/base/TtoCb1NuQaDy3NsZWTpc0GIvnph?table=tblKCjplVAFrRwMC
+
+这张表里每个款式一行,包含「款式」和「合作达人多维表格地址」两个字段。
+凡是款式 和 合作达人多维表格地址 都不为空的一行,就去解析地址里的 base_token + table_id,
+再拉该表的字段定义 (+field-list) 构建 field_map,最终拼出和原 款式_多维表格_对照.json
+完全相同结构的 mapping:
+  { "_comment":..., "platform":..., "tables":[ {index,name,base_token,table_id,url,field_map,all_field_names,total_records}, ... ] }
+
+自营达人: 同一张索引表新增了「自营合作达人表格地址」列,结构完全相同,
+输出到 data/config/款式_多维表格_对照_自营.json。
+
+策略:
+- 拉取后写回 data/config/款式_多维表格_对照.json (合作) 或 款式_多维表格_对照_自营.json (自营)
+- 带 TTL 缓存(默认 6 小时):缓存新鲜就直接读缓存,避免每次跑都打几十个 lark-cli 请求
+- lark-cli 拉取失败时退回缓存文件,保证可运行
+- index 编号:已知款式沿用旧 JSON 里的编号,新款式按出现顺序往后追加(保证历史文件名稳定)
+
+被三方脚本 (bilibili / pgy / xingtu) 通过 load_mapping() 调用;
+run_all.py 启动前可调 refresh_mapping(force=True) 统一刷新一次,避免 3 个并行进程重复刷新打架。
+"""
+import json
+import os
+import re
+import subprocess
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import (
+    PATHS,
+    resolve_layer_output,
+)
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+MAPPING_DIR = PATHS.normalized_root / "mappings"
+MAPPING_PATH = MAPPING_DIR / "款式_多维表格_对照.json"
+SELF_MAPPING_PATH = MAPPING_DIR / "款式_多维表格_对照_自营.json"
+MAPPING_SEED_PATH = PATHS.config_root / "款式_多维表格_对照.json"
+SELF_MAPPING_SEED_PATH = PATHS.config_root / "款式_多维表格_对照_自营.json"
+
+LARK_CLI = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK_CLI):
+    LARK_CLI = "lark-cli.cmd"
+
+# 「合作达人」多维表格地址表(款式 → 各自合作达人多维表格 URL)
+INDEX_BASE_TOKEN = "TtoCb1NuQaDy3NsZWTpc0GIvnph"
+INDEX_TABLE_ID = "tblKCjplVAFrRwMC"
+# 这张地址表里的关键字段名
+INDEX_STYLE_FIELD = "款式"
+INDEX_URL_FIELD = "合作达人多维表格地址"
+INDEX_SELF_URL_FIELD = "自营合作达人表格地址"
+
+# 缓存新鲜阈值(秒),默认 6 小时
+CACHE_TTL = 6 * 3600
+
+# 标准字段名 → 逻辑名 (与 rebuild_mapping.py 保持一致)
+STD = {
+    "达人名称": "creator_name",
+    "达人昵称": "creator_name",
+    "达人ID": "creator_id",
+    "达人id": "creator_id",
+    "达人id号": "creator_id",
+    "星图ID": "creator_id",
+    "小红书号": "xhs_id",
+    "发布笔记标题": "note_title",
+    "笔记标题": "note_title",
+    "标题": "note_title",
+    "主页链接": "profile_url",
+    "发布笔记链接": "note_url",
+    "发布链接": "note_url",
+    "笔记链接": "note_url",
+    "发布时间": "publish_time",
+    "投放平台": "platform",
+    "合作费用": "cost",
+    "快递单号": "tracking_no",
+    "7天曝光量": "read_count_7d",
+    "14天曝光量": "read_count_14d",
+    "21天曝光量": "read_count_21d",
+    "28天曝光量": "read_count_28d",
+    "曝光量（自动抓取）": "read_count_auto",
+    "60天曝光量": "read_count_60d",
+    "90天曝光量": "read_count_90d",
+    "120天曝光量": "read_count_120d",
+    "150天曝光量": "read_count_150d",
+    "180天曝光量": "read_count_180d",
+    "月底曝光量": "month_end",
+    "达人粉丝": "follower_count",
+    "微信号": "wechat",
+    "投放账号类型": "account_type",
+    "观看画像": "audience_profile",
+    "数据表现": "data_performance",
+    "发布笔记内容方向": "content_direction",
+    "互动赞藏数": "engagement_count",
+    "投流金额": "ad_spend",
+    "制单日期": "cooperation_date",
+    "审稿时间": "review_time",
+    "是否结款": "is_paid",
+    "是否新方向验证": "is_new_direction",
+    "cpm": "cpm",
+}
+
+_URL_RE = re.compile(r"https?://[^\s)>\]]+")
+
+
+def _log(msg: str) -> None:
+    safe = str(msg).encode("gbk", errors="replace").decode("gbk", errors="replace")
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] [feishu_mapping] {safe}", flush=True)
+
+
+# ============================================================
+# lark-cli 调用
+# ============================================================
+def call_lark_json(args: list[str]) -> dict:
+    cmd = [LARK_CLI] + args
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, env=env,
+                          encoding="utf-8", errors="replace")
+    text = proc.stdout if proc.stdout else (proc.stderr or "")
+    start = text.find("{")
+    if start < 0:
+        return {"ok": False, "error": f"no json: {text[:300]}"}
+    try:
+        return json.loads(text[start:], strict=False)
+    except json.JSONDecodeError as e:
+        return {"ok": False, "error": f"json decode: {e}"}
+
+
+def _field_list(base_token: str, table_id: str) -> list[dict]:
+    resp = call_lark_json([
+        "base", "+field-list",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--format", "json",
+    ])
+    if not resp.get("ok"):
+        _log(f"  [WARN] +field-list 失败 {base_token}/{table_id}: {resp.get('error')}")
+        return []
+    return resp.get("data", {}).get("fields", [])
+
+
+def _record_list_all(base_token: str, table_id: str,
+                     field_ids: list[str] | None = None) -> list[dict]:
+    """拉一张表全部记录,自动分页。field_ids 不传=全字段。"""
+    all_records = []
+    offset = 0
+    base_args = [
+        "base", "+record-list",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--as", "user",
+        "--format", "json",
+        "--limit", "200",
+    ]
+    while True:
+        args = base_args + ["--offset", str(offset)]
+        if field_ids:
+            for fid in field_ids:
+                args += ["--field-id", fid]
+        resp = call_lark_json(args)
+        if not resp.get("ok"):
+            raise RuntimeError(f"record-list 失败: {resp.get('error')}")
+        d = resp["data"]
+        ids = d["field_id_list"]
+        rids = d["record_id_list"]
+        rows = d["data"]
+        for i, row in enumerate(rows):
+            rec = {"record_id": rids[i]}
+            for j, fid in enumerate(ids):
+                rec[fid] = row[j] if j < len(row) else None
+            all_records.append(rec)
+        offset += len(rows)
+        if not d.get("has_more", False) or len(rows) == 0:
+            break
+    return all_records
+
+
+# ============================================================
+# 地址解析 & field_map 构建
+# ============================================================
+def parse_style_url(raw) -> tuple[str | None, str | None, str]:
+    """从飞书多维表格地址单元格(可能是 markdown 链接或纯 URL)解析出
+    (base_token, table_id, url)。没有 table 参数时返回 (bt, None, url)。"""
+    if raw is None:
+        return None, None, ""
+    if isinstance(raw, list):
+        raw = " ".join(str(x) for x in raw if x)
+    s = str(raw).strip()
+    if not s:
+        return None, None, ""
+    m = _URL_RE.search(s)
+    url = m.group(0) if m else s
+    bt = None
+    mb = re.search(r"/base/([A-Za-z0-9]+)", url)
+    if mb:
+        bt = mb.group(1)
+    tid = None
+    mt = re.search(r"[?&]table=([A-Za-z0-9]+)", url)
+    if mt:
+        tid = mt.group(1)
+    return bt, tid, url
+
+
+def build_field_map(fields: list[dict]) -> tuple[dict, list[str]]:
+    field_map = {}
+    all_field_names = []
+    for f in fields:
+        fid = f.get("id")
+        fname = f.get("name")
+        all_field_names.append(fname)
+        clean_name = fname.rstrip("?").rstrip() if fname else ""
+        logical = STD.get(fname) or STD.get(clean_name)
+        if not logical and clean_name:
+            for k, v in STD.items():
+                # 只做正向匹配: 字段名以已知 key 开头(如 "7天曝光量(自动)" 以 "7天曝光量" 开头)
+                # 不做反向匹配(k.startswith(clean_name)): 单字如"曝"会误匹配多个字段
+                if clean_name.startswith(k):
+                    logical = v
+                    break
+        if logical and logical not in field_map:
+            field_map[logical] = {"field_id": fid, "field_name": fname}
+    return field_map, all_field_names
+
+
+def _norm_url(base_token: str, table_id: str, url: str) -> str:
+    if url:
+        return url
+    return (f"https://bu0zgpibak.feishu.cn/base/{base_token}"
+            f"?table={table_id}&view=vewStRNML4")
+
+
+# ============================================================
+# 重建 mapping
+# ============================================================
+def refresh_mapping(force: bool = True, ttl: float = CACHE_TTL,
+                    mapping_path: Path | None = None,
+                    write: bool = True,
+                    url_field: str = INDEX_URL_FIELD) -> dict | None:
+    """动态读取地址表,重建 mapping 并写缓存。
+    返回新的 mapping;失败返回 None(缓存仍可用时下游自行读缓存)。
+    write=False 时只构建内存中的 mapping,不落盘(供 rebuild_mapping.py --diff 用)。
+    url_field: 从索引表读哪一列 URL — 默认 INDEX_URL_FIELD(合作达人),
+               传 INDEX_SELF_URL_FIELD 则读自营达人列。
+    """
+    is_self = url_field == INDEX_SELF_URL_FIELD
+    path = resolve_layer_output(
+        mapping_path or (SELF_MAPPING_PATH if is_self else MAPPING_PATH),
+        layer_root=PATHS.normalized_root,
+        field="mapping_path",
+    )
+    seed_path = SELF_MAPPING_SEED_PATH if is_self else MAPPING_SEED_PATH
+    if write:
+        path.parent.mkdir(parents=True, exist_ok=True)
+
+    if not force and path.exists():
+        age = time.time() - path.stat().st_mtime
+        if age < ttl:
+            return json.loads(path.read_text(encoding="utf-8"))
+
+    # 旧缓存用于:已知款式的 index 编号沿用 + platform 字段保留
+    old_by_name: dict[str, dict] = {}
+    old_platform = "xhs_pgy"
+    previous_path = path if path.exists() else seed_path
+    if previous_path.exists():
+        try:
+            old = json.loads(previous_path.read_text(encoding="utf-8"))
+            old_platform = old.get("platform", "xhs_pgy")
+            for t in old.get("tables", []):
+                if t.get("name"):
+                    old_by_name[t["name"]] = t
+        except Exception as exc:
+            _log(f"读旧缓存失败(忽略): {exc}")
+
+    _log(f"动态读取地址表 {INDEX_BASE_TOKEN}/{INDEX_TABLE_ID} ...")
+    try:
+        index_records = _record_list_all(INDEX_BASE_TOKEN, INDEX_TABLE_ID)
+    except Exception as exc:
+        _log(f"读取地址表失败: {exc}")
+        return None
+    _log(f"地址表共 {len(index_records)} 行")
+
+    # 需要先知道字段名→field_id 映射,取「款式」和「合作达人多维表格地址」两列
+    fields = _field_list(INDEX_BASE_TOKEN, INDEX_TABLE_ID)
+    name2id = {}
+    for f in fields:
+        name2id[f.get("name")] = f.get("id")
+    style_fid = name2id.get(INDEX_STYLE_FIELD)
+    url_fid = name2id.get(url_field)
+    # 兜底: 合作达人列历史曾叫「飞书多维表格地址」
+    if not url_fid and url_field == INDEX_URL_FIELD:
+        url_fid = name2id.get("飞书多维表格地址")
+    if not style_fid or not url_fid:
+        _log(f"  [ERR] 地址表找不到字段 '{INDEX_STYLE_FIELD}'/'{url_field}'")
+        return None
+
+    # 去重:同一款式只保留第一条 款式+地址 都有且地址里能解析出 table_id 的行
+    seen_styles: set[str] = set()
+    raw_styles: list[dict] = []  # {name, base_token, table_id, url}
+    for rec in index_records:
+        name = rec.get(style_fid)
+        if isinstance(name, list):
+            name = " ".join(str(x) for x in name if x).strip()
+        if isinstance(name, str):
+            name = name.strip()
+        if not name:
+            continue
+        bt, tid, url = parse_style_url(rec.get(url_fid))
+        if not bt:
+            continue
+        # 没 table_id 的行先跳过;若有同款式的后续行带 table_id 会补上
+        if not tid:
+            continue
+        if name in seen_styles:
+            continue
+        seen_styles.add(name)
+        raw_styles.append({"name": name, "base_token": bt,
+                           "table_id": tid, "url": _norm_url(bt, tid, url)})
+
+    _log(f"去重后款式 {len(raw_styles)} 个,开始拉各表字段定义...")
+
+    # index 编号:已知款式用旧编号,新款式往后追加
+    used_idx: set[int] = set()
+    for s in old_by_name.values():
+        idx = s.get("index")
+        if isinstance(idx, int):
+            used_idx.add(idx)
+
+    def next_idx() -> int:
+        n = 1
+        while n in used_idx:
+            n += 1
+        used_idx.add(n)
+        return n
+
+    tables = []
+    for s in raw_styles:
+        fields = _field_list(s["base_token"], s["table_id"])
+        if not fields:
+            _log(f"  [WARN] [{s['name']}] 拉不到字段,跳过")
+            continue
+        field_map, all_field_names = build_field_map(fields)
+        old_entry = old_by_name.get(s["name"])
+        if old_entry and isinstance(old_entry.get("index"), int):
+            idx = old_entry["index"]
+            used_idx.add(idx)
+        else:
+            idx = next_idx()
+        total_records = old_entry.get("total_records", 0) if old_entry else 0
+        tables.append({
+            "index": idx,
+            "name": s["name"],
+            "base_token": s["base_token"],
+            "table_id": s["table_id"],
+            "url": s["url"],
+            "field_map": field_map,
+            "all_field_names": all_field_names,
+            "total_records": total_records,
+        })
+        _log(f"  [{idx:02d}] {s['name']}: {len(field_map)} 逻辑字段 / {len(all_field_names)} 列")
+
+    tables.sort(key=lambda t: t["index"])
+    label = "自营" if is_self else "合作达人"
+    mapping = {
+        "_comment": (f"款式 → 多维表格 对照表({label}); 由 feishu_mapping.py 从{label}地址表动态生成。"
+                     " field_map 给出本表核心字段的 field_id (实时从飞书拉取)"),
+        "platform": old_platform,
+        "tables": tables,
+    }
+
+    # 原子写:先写临时文件再 rename,避免并行进程读到半截 JSON
+    if write:
+        tmp = path.with_suffix(path.suffix + ".tmp")
+        tmp.write_text(json.dumps(mapping, ensure_ascii=False, indent=2),
+                       encoding="utf-8")
+        os.replace(tmp, path)
+        _log(f"已写回缓存 ({len(tables)} 款式): {path}")
+    else:
+        _log(f"[--diff] 仅构建内存 mapping ({len(tables)} 款式),未落盘")
+    return mapping
+
+
+def load_mapping(data_dir: Path | None = None,
+                 force_refresh: bool = False,
+                 ttl: float = CACHE_TTL,
+                 self_operated: bool = False) -> dict:
+    """供三个 scraper 调用:保证有 mapping 可用。
+    优先:缓存新鲜就用缓存;否则动态刷新;刷新失败退回缓存;再不行报错退出。
+    self_operated=True 时读自营达人 mapping。"""
+    del data_dir  # Legacy compatibility: runtime mappings have one canonical cache.
+    path = SELF_MAPPING_PATH if self_operated else MAPPING_PATH
+    seed_path = SELF_MAPPING_SEED_PATH if self_operated else MAPPING_SEED_PATH
+    # 缓存是否新鲜
+    fresh = path.exists() and (time.time() - path.stat().st_mtime) < ttl
+    if not force_refresh and fresh:
+        return json.loads(path.read_text(encoding="utf-8"))
+
+    url_field = INDEX_SELF_URL_FIELD if self_operated else INDEX_URL_FIELD
+    new = refresh_mapping(force=True, ttl=0, mapping_path=path, url_field=url_field)
+    if new is not None:
+        return new
+    if not path.exists() and seed_path.exists():
+        _log(f"dynamic refresh failed; using read-only seed: {seed_path}")
+        return json.loads(seed_path.read_text(encoding="utf-8"))
+    # 刷新失败 → 退回缓存(无论新旧)
+    if path.exists():
+        _log(f"动态刷新失败,使用缓存: {path}")
+        return json.loads(path.read_text(encoding="utf-8"))
+    _log("[FATAL] 动态刷新失败且无缓存可用")
+    raise RuntimeError("无法获取款式对照表(动态读取失败且无缓存)")
+
+
+
+if __name__ == "__main__":
+    # 手动跑一次:python feishu_mapping.py
+    m = refresh_mapping(force=True)
+    if m:
+        ts = m.get("tables", [])
+        print(f"\n款式数: {len(ts)}")
+        for t in ts:
+            print(f"  [{t['index']:02d}] {t['name']}  bt={t['base_token']} tid={t['table_id']}")
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/login_helper.py b/src/gyxx_flow/modules/content_marketing/runtime/login_helper.py
new file mode 100644
index 0000000..af251e0
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/login_helper.py
@@ -0,0 +1,112 @@
+"""Open Xiaohongshu and Douyin in real Chrome so you can log in manually.
+
+Cookies and browser profiles are saved to the same locations used by the scrapers.
+"""
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+from typing import Any
+
+from playwright.sync_api import Page
+from scrapling.fetchers import DynamicSession
+
+from gyxx_flow.modules.content_marketing.runtime import bilibili_comment_scraper
+from gyxx_flow.modules.content_marketing.runtime import douyin_comment_scraper
+from gyxx_flow.modules.content_marketing.runtime import xiaohongshu_comment_scraper
+
+
+def log(message: str) -> None:
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {message}", flush=True)
+
+
+def wait_for_login(page: Page, platform: str, timeout: int = 300) -> None:
+    """Wait until login prompts disappear from the page."""
+    login_markers = {
+        "xiaohongshu": ("扫码登录", "登录后", "验证码"),
+        "douyin": ("扫码登录", "登录后"),
+        "bilibili": ("扫码登录", "短信登录", "密码登录"),
+    }
+    markers = login_markers.get(platform, ())
+    deadline = time.time() + timeout
+    while time.time() < deadline:
+        try:
+            body = page.locator("body").inner_text(timeout=5000)
+        except Exception:
+            body = ""
+        if not any(marker in body for marker in markers):
+            log(f"{platform}: login state detected (no login prompt)")
+            return
+        log(f"{platform}: waiting for login... ({int(deadline - time.time())}s left)")
+        page.wait_for_timeout(3000)
+    log(f"{platform}: timed out waiting for login")
+
+
+def login_platform(platform: str, url: str, profile_dir: Path, cookie_file: Path, save_state_fn: Any, restore_cookies_fn: Any, timeout: int = 300) -> None:
+    log(f"Opening {platform} for manual login: {url}")
+    profile_dir.mkdir(exist_ok=True)
+
+    def action(page: Page) -> None:
+        page.goto(url, wait_until="domcontentloaded", timeout=90000)
+        page.wait_for_timeout(4000)
+        wait_for_login(page, platform, timeout)
+        save_state_fn(page)
+        log(f"Saved {platform} cookies to {cookie_file}")
+
+    with DynamicSession(
+        headless=False,
+        real_chrome=True,
+        user_data_dir=str(profile_dir),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=90000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=restore_cookies_fn,
+        max_pages=1,
+    ) as session:
+        session.fetch(url, page_action=action, wait=1000)
+
+
+def main() -> int:
+    # 1. Xiaohongshu
+    login_platform(
+        platform="xiaohongshu",
+        url="https://www.xiaohongshu.com/explore",
+        profile_dir=xiaohongshu_comment_scraper.PROFILE_DIR,
+        cookie_file=xiaohongshu_comment_scraper.COOKIE_FILE,
+        save_state_fn=xiaohongshu_comment_scraper.save_state,
+        restore_cookies_fn=xiaohongshu_comment_scraper.restore_cookies,
+    )
+
+    # 2. Douyin
+    login_platform(
+        platform="douyin",
+        url="https://www.douyin.com/",
+        profile_dir=douyin_comment_scraper.PROFILE_DIR,
+        cookie_file=douyin_comment_scraper.COOKIE_FILE,
+        save_state_fn=douyin_comment_scraper.save_state,
+        restore_cookies_fn=douyin_comment_scraper.restore_cookies,
+    )
+
+    # 3. Bilibili
+    login_platform(
+        platform="bilibili",
+        url="https://passport.bilibili.com/login",
+        profile_dir=bilibili_comment_scraper.PROFILE_DIR,
+        cookie_file=bilibili_comment_scraper.COOKIE_FILE,
+        save_state_fn=bilibili_comment_scraper.save_state,
+        restore_cookies_fn=bilibili_comment_scraper.restore_cookies,
+    )
+
+    log("All platforms logged in and cookies saved.")
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/monthly_summary_all.py b/src/gyxx_flow/modules/content_marketing/runtime/monthly_summary_all.py
new file mode 100644
index 0000000..8f86eb5
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/monthly_summary_all.py
@@ -0,0 +1,450 @@
+"""月度笔记汇总:对上月各款所有笔记做横向对比分析(曝光分层 × 平台特性)。
+
+与 weekly_summary_all.py 的区别:
+- 时间范围:上月 1 号到月底(动态算)
+- Prompt:曝光分层 × 平台特性修正效率评级(S/A/B/C + 平台价值定位)
+- 入库:cmt_monthly_summary 表(UNIQUE(style_id, month_start))
+- 飞书写入:生命进程表的"笔记分析汇总"字段,时间标月度范围
+
+流程:
+1. 读索引表拿每款生命进程表地址(复用 weekly 的 load_index_styles)
+2. PG 查上月该款所有笔记(按 publish_time 过滤)
+3. 对每款:跑单品分析(复用 note_<id>.txt)+ LLM 横向对比 + 创建飞书文档 + 写入生命进程表 + 入库
+
+用法:
+  python monthly_summary_all.py                 # 跑所有上月有笔记的款
+  python monthly_summary_all.py --style 宙斯     # 只跑某款
+  python monthly_summary_all.py --dry-run        # 不写飞书/PG
+  python monthly_summary_all.py --max-workers 4  # 并发数
+"""
+import argparse
+import json
+import os
+import re
+import subprocess
+import sys
+from datetime import datetime, timedelta
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+sys.stdout.reconfigure(encoding='utf-8')
+
+# 复用 weekly_summary_all 的工具函数
+from gyxx_flow.modules.content_marketing.runtime.weekly_summary_all import (
+    LARK, DATA_DIR, REPORTS_DIR, dry_run_global,
+    URL_RE, log, call_lark_json, parse_url, parse_bt_tid,
+    load_index_styles, find_target_fields, read_report,
+    run_single_note_analysis, create_feishu_doc, write_to_target_table,
+)
+from gyxx_flow.modules.content_marketing.runtime.data.tools.analyze_comments import call_hermes_analyzer
+
+# ============================================================
+# 月时间范围:动态算上月 1 号到月底
+# ============================================================
+def _compute_last_month():
+    """返回 (month_start, month_end) 都是 date 对象"""
+    today = datetime.now().date()
+    # 上月 1 号 = 本月 1 号 - 1 天
+    first_of_this_month = today.replace(day=1)
+    month_end = first_of_this_month - timedelta(days=1)
+    month_start = month_end.replace(day=1)
+    return month_start, month_end
+
+_DEFAULT_START, _DEFAULT_END = _compute_last_month()
+MONTH_START = os.environ.get("MONTH_SINCE", _DEFAULT_START.strftime("%Y-%m-%d"))
+MONTH_END = os.environ.get("MONTH_UNTIL", _DEFAULT_END.strftime("%Y-%m-%d"))
+TIME_LABEL = f"{MONTH_START}~{MONTH_END[5:]}"
+
+
+# ============================================================
+# PG 查上月笔记
+# ============================================================
+def get_last_month_notes(style_name):
+    """从 PG 拿上月该款的所有笔记(按 publish_time 过滤)"""
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_conn
+    with get_conn() as conn:
+        cur = conn.cursor()
+        cur.execute("""
+            SELECT n.id, n.platform, n.title, n.url, n.view_count,
+                   n.comment_count, n.like_count, n.favorite_count, n.share_count,
+                   c.name AS creator_name
+            FROM cmt_notes n
+            JOIN cmt_styles s ON n.style_id = s.id
+            LEFT JOIN cmt_creators c ON n.creator_id = c.id
+            WHERE s.name = %s AND n.self_operated = FALSE
+            AND n.publish_time IS NOT NULL
+            AND n.publish_time >= %s AND n.publish_time < %s
+            ORDER BY n.id
+        """, (style_name, MONTH_START, MONTH_END + " 23:59:59"))
+        cols = [d[0] for d in cur.description]
+        return [dict(zip(cols, r)) for r in cur.fetchall()]
+
+
+# ============================================================
+# LLM 横向对比 prompt(曝光分层 × 平台特性)
+# ============================================================
+MONTHLY_PROMPT = """你是光影行星品牌「{style}」款式的批量内容横向分析师。
+请对以下【{n} 篇笔记清单】进行横向对比分析,严格遵循"先按曝光量级分层,再按平台特性修正效率评级"的双维逻辑。
+
+【本月笔记清单】
+{notes}
+
+【第一步:曝光量级分层(硬性门槛)】
+
+| 平台 | 曝光"高"阈值 | 曝光"低"阈值 |
+| :--- | :--- | :--- |
+| 抖音 | ≥ 50万 | < 50万 |
+| 小红书 | ≥ 10万 | < 10万 |
+| B站 | ≥ 30万 | < 30万 |
+
+- 高曝光 + 高效率 = **S级(明星爆款)**
+- 低曝光 + 高效率 = **A级(潜力黑马)**
+- 高曝光 + 低效率 = **B级(虚假繁荣)**
+- 低曝光 + 低效率 = **C级(失败品)**
+
+【第二步:平台特性修正系数(判断"效率高/低"时,标准各不同)】
+
+### 1. 抖音(公域推荐 × 情绪爆发)
+- 互动率 ≥ 5% 为高;完播率 ≥ 40% 为高;品牌提及率 ≥ 12% 为高
+- 核心观察窗口:发布后 24 小时
+- 主要价值定位:爆款驱动、即时冲动、快速拉新
+- 抖音特有分析维度:前 3 秒留存率 / 评论区情绪词 vs 理性词占比
+
+### 2. 小红书(搜索心智 × 长尾效应)
+- 互动率 ≥ 6% 为高;收藏率 ≥ 2% 为高;品牌提及率 ≥ 20% 为高
+- 核心观察窗口:发布后 3-7 天
+- 主要价值定位:深度种草、搜索拦截、心智沉淀
+- 小红书特有分析维度:搜索词关联度(品类词+场景词) / 评论区"求链接"比例
+
+### 3. B站(深度信任 × 长决策周期)
+- 互动率 ≥ 4% 为高;平均观看时长 ≥ 3 分钟为高;品牌提及率 ≥ 25% 为高
+- 核心观察窗口:发布后 7-14 天
+- 主要价值定位:深度测评、信任建立、高客单价转化
+- B站特有分析维度:评论区提问深度(竞品对比 vs 简单咨询) / 弹幕密度曲线
+
+【输出要求】
+按以下结构输出(Markdown 格式,适合贴到飞书文本字段):
+
+# {style} 月度笔记汇总（{time_range}）
+
+## 一、本月投放概览
+- 投放笔记数：{n} 篇
+- 涉及平台：[列出各平台笔记数]
+- 总曝光 / 总互动：[汇总数据]
+- 整体表现：[一句话定性]
+
+## 二、分层排序总表
+按 S -> A -> B -> C 输出,同一层级内按"曝光量×互动率"降序排列。
+
+| 层级 | 笔记标题 | 平台 | 达人 | 曝光 | 互动率 | 品牌提及率 | 平台价值定位 |
+| :--- | :--- | :--- | :--- | :--- | :--- | :--- | :--- |
+| S级 | ... | ... | ... | ... | ... | ... | 爆款拉新/深度种草/信任转化 |
+| A级 | ... | ... | ... | ... | ... | ... | ... |
+| B级 | ... | ... | ... | ... | ... | ... | ... |
+| C级 | ... | ... | ... | ... | ... | ... | ... |
+
+## 三、跨平台对比洞察
+- 哪个平台的高效笔记占比最高? -> 重点倾斜预算的平台
+- 哪个平台的 S 级笔记内容公式最清晰? -> 优先复制到其他平台进行跨平台移植测试
+- 各平台笔记数分布是否均衡
+
+## 四、各平台 Top3 笔记内容特征总结
+### 抖音 Top3(如有)
+- 是否都是"前3秒强钩子+情绪共鸣"型?
+- 列出 3 篇标题 + 共性特征
+
+### 小红书 Top3(如有)
+- 是否都包含"搜索关键词+高收藏价值"?
+- 列出 3 篇标题 + 共性特征
+
+### B站 Top3(如有)
+- 是否都是"深度信息密度+竞品对比"型?
+- 列出 3 篇标题 + 共性特征
+
+若发现某平台 Top3 不符合该平台典型特征(如小红书 Top3 全是情绪段子而非实用内容),说明该平台选号或 Brief 方向已跑偏,需明确指出。
+
+## 五、预算再分配建议
+结合平台特性,给出下月预算比例建议:
+- 抖音：X%(理由)
+- 小红书：X%(理由)
+- B站：X%(理由)
+
+## 六、下轮 Brief 平台差异化修改指令
+针对各平台分别输出 1 条 Brief 修改指令:
+- 抖音 Brief 补什么:强调"前3秒留存"和"品牌名口播重复2次"
+- 小红书 Brief 补什么:强调"标题含品类搜索词"和"收藏率优化"
+- B站 Brief 补什么:强调"深度对比实测"和"评论区答疑准备"
+
+【约束】
+- 分层必须严格按曝光阈值(抖音50万/小红书10万/B站30万)判定,不放宽
+- 效率判定必须按各平台特性标准(抖音5%/小红书6%/B站4%),不混用
+- 表格必须填充完整数据,不要留空或写"无"
+- 每条建议须具体可执行,避免"加强""优化"等空泛词汇
+- 整体篇幅 2500-4000 字
+- **必须把完整的 Markdown 报告作为响应内容直接返回,不要写到本地文件,不要只给摘要,不要提及"文件已保存"或"月报已生成"等元信息。响应正文就是最终报告本身。**"""
+
+
+def generate_monthly_summary(style, notes):
+    """调 LLM 生成月度横向对比汇总"""
+    notes_block = []
+    for i, n in enumerate(notes, 1):
+        views = n.get('view_count')
+        views_str = f"{views:,}" if views else "未采集"
+        notes_block.append(
+            f"### 笔记{i}: {n['title']}\n"
+            f"- 平台: {n['platform']}\n- 达人: {n.get('creator_name', '?')}\n"
+            f"- 曝光: {views_str} / 点赞: {n.get('like_count') or 0} / "
+            f"评论: {n.get('comment_count') or 0} / 收藏: {n.get('favorite_count') or 0} / "
+            f"分享: {n.get('share_count') or 0}\n"
+        )
+    prompt = MONTHLY_PROMPT.format(
+        style=style,
+        notes="\n---\n".join(notes_block),
+        time_range=TIME_LABEL,
+        n=len(notes),
+    )
+    system = "你是光影行星品牌的批量内容横向分析师,擅长按曝光分层 × 平台特性双维逻辑做笔记横向对比分析。"
+    return call_hermes_analyzer(system, prompt)
+
+
+# ============================================================
+# 入库 cmt_monthly_summary
+# ============================================================
+def save_to_db_monthly(style, notes, summary, doc_url, doc_title, status):
+    """把月度汇总结果写入 PG cmt_monthly_summary(UNIQUE(style_id, month_start))"""
+    if dry_run_global:
+        return False
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_conn
+    try:
+        with get_conn() as conn:
+            cur = conn.cursor()
+            cur.execute("SELECT id FROM cmt_styles WHERE name = %s", (style,))
+            row = cur.fetchone()
+            if not row:
+                log(f"  [DB] 款式 {style} 不在 cmt_styles,跳过入库")
+                return False
+            style_id = row[0]
+
+            total_views = sum((n.get('view_count') or 0) for n in notes) or None
+            total_comments = sum((n.get('comment_count') or 0) for n in notes)
+            total_engagement = sum(
+                ((n.get('like_count') or 0) + (n.get('favorite_count') or 0) +
+                 (n.get('comment_count') or 0) + (n.get('share_count') or 0))
+                for n in notes) or None
+
+            cur.execute("""
+                INSERT INTO cmt_monthly_summary
+                    (style_id, month_start, month_end, time_label,
+                     note_count, total_comments, total_views, total_engagement,
+                     doc_url, doc_title, summary_chars, status,
+                     created_at, updated_at)
+                VALUES
+                    (%s, %s, %s, %s,
+                     %s, %s, %s, %s,
+                     %s, %s, %s, %s,
+                     CURRENT_TIMESTAMP, CURRENT_TIMESTAMP)
+                ON CONFLICT (style_id, month_start) DO UPDATE SET
+                    note_count = EXCLUDED.note_count,
+                    total_comments = EXCLUDED.total_comments,
+                    total_views = EXCLUDED.total_views,
+                    total_engagement = EXCLUDED.total_engagement,
+                    doc_url = EXCLUDED.doc_url,
+                    doc_title = EXCLUDED.doc_title,
+                    summary_chars = EXCLUDED.summary_chars,
+                    status = EXCLUDED.status,
+                    updated_at = CURRENT_TIMESTAMP
+            """, (style_id, MONTH_START, MONTH_END, TIME_LABEL,
+                   len(notes), total_comments, total_views, total_engagement,
+                   doc_url, doc_title, len(summary) if summary else None, status))
+            conn.commit()
+        log(f"  ✅ 已入库 cmt_monthly_summary (style_id={style_id}, month={MONTH_START})")
+        return True
+    except Exception as exc:
+        log(f"  [DB ERR] {style} 入库失败: {exc}")
+        return False
+
+
+# ============================================================
+# 处理一个款
+# ============================================================
+def process_one_style_monthly(style, index_info):
+    """处理一个款的月度汇总"""
+    bt, tid, url = index_info
+    log(f"\n========== {style} (月度) ==========")
+    log(f"  生命进程表: {bt}/{tid}")
+
+    def _save(notes_list, summary, doc_url, doc_title, status):
+        save_to_db_monthly(style, notes_list, summary, doc_url, doc_title, status)
+
+    time_fid, summary_fid = find_target_fields(bt, tid)
+    if not time_fid or not summary_fid:
+        log(f"  [ERR] 找不到 时间/笔记分析汇总 字段,跳过")
+        _save([], None, None, None, "no_fields")
+        return {"style": style, "status": "no_fields", "notes": 0}
+
+    notes = get_last_month_notes(style)
+    log(f"  上月笔记: {len(notes)} 篇")
+    if not notes:
+        log(f"  [SKIP] 上月无笔记")
+        _save([], None, None, None, "no_notes")
+        return {"style": style, "status": "no_notes", "notes": 0}
+
+    for n in notes:
+        n["report_path"] = run_single_note_analysis(n["id"])
+
+    valid_notes = [n for n in notes if n.get("report_path")]
+    if not valid_notes:
+        log(f"  [ERR] 没有有效报告,跳过")
+        _save(notes, None, None, None, "no_reports")
+        return {"style": style, "status": "no_reports", "notes": len(notes)}
+
+    log(f"  调 LLM 横向对比 {len(valid_notes)} 篇报告...")
+    summary = generate_monthly_summary(style, valid_notes)
+    if not summary:
+        log(f"  [ERR] LLM 综合失败")
+        _save(notes, None, None, None, "llm_failed")
+        return {"style": style, "status": "llm_failed", "notes": len(notes)}
+
+    out_path = REPORTS_DIR / f"{style}_月度汇总_{TIME_LABEL}.txt"
+    out_path.write_text(summary, encoding="utf-8")
+    log(f"  本地副本: {out_path.name} ({len(summary)} 字符)")
+
+    if dry_run_global:
+        log(f"  [DRY] 跳过飞书文档创建和写入")
+        _save(notes, summary, None, None, "dry_run")
+        return {"style": style, "status": "dry_run", "notes": len(notes), "chars": len(summary)}
+
+    doc_url, err = create_feishu_doc_monthly(style, summary)
+    if not doc_url:
+        log(f"  [ERR] {err}")
+        _save(notes, summary, None, None, "doc_failed")
+        return {"style": style, "status": "doc_failed", "notes": len(notes)}
+
+    log(f"  飞书文档: {doc_url}")
+
+    ok = write_to_target_table(bt, tid, time_fid, summary_fid, style, doc_url, dry_run_global)
+    doc_title = f"{style} 月度汇总（{TIME_LABEL}）"
+    _save(notes, summary, doc_url, doc_title, "ok" if ok else "feishu_write_failed")
+
+    if ok:
+        log(f"  ✅ 已写入生命进程表")
+        return {"style": style, "status": "ok", "notes": len(notes),
+                "doc_url": doc_url, "chars": len(summary)}
+    log(f"  [ERR] 写入表失败")
+    return {"style": style, "status": "write_failed", "notes": len(notes)}
+
+
+def create_feishu_doc_monthly(style, summary):
+    """创建飞书文档(月度),返回 url"""
+    doc_title = f"{style} 月度汇总（{TIME_LABEL}）"
+    md_file = PATHS.tmp_root / "monthly_summary_doc.md"
+    md_file.parent.mkdir(parents=True, exist_ok=True)
+    md_file.write_text(summary, encoding="utf-8")
+
+    cmd = [LARK, "docs", "+create",
+           "--as", "user", "--format", "json",
+           "--title", doc_title,
+           "--doc-format", "markdown",
+           "--content", f"@{md_file.name}"]
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, env=env,
+                          encoding="utf-8", errors="replace")
+    text = proc.stdout or proc.stderr
+    start = text.find("{")
+    if start < 0:
+        return None, f"创建失败: {text[:300]}"
+    resp = json.loads(text[start:])
+    if not resp.get("ok"):
+        return None, f"创建失败: {resp.get('error')}"
+    url = resp.get("data", {}).get("document", {}).get("url", "")
+    return url, None
+
+
+# ============================================================
+# main
+# ============================================================
+def main():
+    parser = argparse.ArgumentParser(description="月度笔记汇总:对上月各款所有笔记做横向对比分析")
+    parser.add_argument("--style", help="只跑指定款式名")
+    parser.add_argument("--dry-run", action="store_true", help="不写飞书/PG")
+    parser.add_argument("--max-workers", type=int, default=4, help="款之间并发数(默认 4)")
+    args = parser.parse_args()
+
+    global dry_run_global
+    # 引用 weekly 的全局变量(改它会影响 weekly 模块,但 dry_run_global 在 weekly 顶部定义)
+    from gyxx_flow.modules.content_marketing.runtime import weekly_summary_all
+    weekly_summary_all.dry_run_global = args.dry_run
+    # 本模块的 dry_run_global 是 import 时的副本,需要同步
+    globals()['dry_run_global'] = args.dry_run
+
+    log(f"时间范围: {TIME_LABEL} (月度)")
+
+    index_styles = load_index_styles()
+    if not index_styles:
+        log("索引表无有效数据,退出")
+        return 1
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_conn
+    with get_conn() as conn:
+        cur = conn.cursor()
+        cur.execute("""
+            SELECT s.name, COUNT(n.id)
+            FROM cmt_styles s
+            JOIN cmt_notes n ON n.style_id = s.id AND n.self_operated = FALSE
+            WHERE n.publish_time IS NOT NULL
+            AND n.publish_time >= %s AND n.publish_time < %s
+            GROUP BY s.name
+            ORDER BY s.name
+        """, (MONTH_START, MONTH_END + " 23:59:59"))
+        pg_styles = [(r[0], r[1]) for r in cur.fetchall()]
+
+    log(f"PG 上月有笔记的款: {len(pg_styles)} 个")
+
+    to_run = []
+    for name, note_count in pg_styles:
+        if args.style and name != args.style:
+            continue
+        if name in index_styles:
+            to_run.append((name, index_styles[name], note_count))
+        else:
+            log(f"  [SKIP] {name}: 索引表无'每周笔记分析'地址")
+
+    log(f"待跑款式: {len(to_run)} 个")
+    if not to_run:
+        return 0
+
+    import concurrent.futures
+    log(f"并发跑 {len(to_run)} 款, max_workers={args.max_workers}")
+    results = []
+    with concurrent.futures.ThreadPoolExecutor(max_workers=args.max_workers) as executor:
+        future_to_style = {
+            executor.submit(process_one_style_monthly, name, info): (name, ncount)
+            for name, info, ncount in to_run
+        }
+        for future in concurrent.futures.as_completed(future_to_style):
+            name, ncount = future_to_style[future]
+            try:
+                r = future.result()
+                r["note_count"] = ncount
+                results.append(r)
+                log(f"  ✅ 完成: {name} -> {r['status']}")
+            except Exception as exc:
+                log(f"  [ERR] {name} 异常: {exc}")
+                results.append({"style": name, "status": "exception",
+                                "error": str(exc), "note_count": ncount})
+
+    results.sort(key=lambda r: r.get("style", ""))
+
+    log(f"\n{'=' * 50}")
+    log(f"全部完成,汇总:")
+    for r in results:
+        log(f"  {r['style']}: {r['status']} (笔记 {r.get('note_count', 0)} 篇)")
+    log(f"{'=' * 50}")
+
+    out = REPORTS_DIR / f"_monthly_summary_all_{TIME_LABEL}.json"
+    out.write_text(json.dumps(results, ensure_ascii=False, indent=2), encoding="utf-8")
+    log(f"结果已存: {out}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/pgy_xhs_scraper.py b/src/gyxx_flow/modules/content_marketing/runtime/pgy_xhs_scraper.py
new file mode 100644
index 0000000..85105ea
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/pgy_xhs_scraper.py
@@ -0,0 +1,424 @@
+import argparse
+import csv
+import json
+import re
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+from typing import Any
+
+from playwright.sync_api import Page, TimeoutError as PlaywrightTimeoutError
+from scrapling.fetchers import DynamicSession
+
+
+LOGIN_URL = "https://pgy.xiaohongshu.com/"
+HOME_URL = "https://pgy.xiaohongshu.com/solar/pre-trade/home"
+DETAIL_URL_MARK = "/solar/pre-trade/blogger-detail/"
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+COOKIE_FILE = PATHS.browser_cookie_file
+PROFILE_DIR = PATHS.browser_profile_dir
+
+
+def log(message: str) -> None:
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {message}", flush=True)
+
+
+def load_cookies() -> list[dict[str, Any]]:
+    if not COOKIE_FILE.exists():
+        return []
+    try:
+        data = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+        return data if isinstance(data, list) else []
+    except json.JSONDecodeError:
+        log(f"Cookie file is invalid, ignoring: {COOKIE_FILE}")
+        return []
+
+
+def save_state(page: Page) -> None:
+    DATA_DIR.mkdir(exist_ok=True)
+    cookies = page.context.cookies()
+    COOKIE_FILE.write_text(
+        json.dumps(cookies, ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    storage_state = page.context.storage_state()
+    PATHS.browser_storage_state_file.write_text(
+        json.dumps(storage_state, ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    log(f"Saved cookies: {COOKIE_FILE}")
+
+
+def restore_cookies(page: Page) -> None:
+    cookies = load_cookies()
+    if not cookies:
+        return
+    try:
+        page.context.add_cookies(cookies)
+        log(f"Loaded cookies: {COOKIE_FILE}")
+    except Exception as exc:
+        log(f"Failed to load cookies, will continue with browser profile: {exc}")
+
+
+def is_logged_in(page: Page) -> bool:
+    if "/solar/pre-trade/" in page.url:
+        return True
+    try:
+        page.get_by_text("找博主", exact=True).wait_for(timeout=2500)
+        return True
+    except PlaywrightTimeoutError:
+        return False
+
+
+def wait_for_scan_login(page: Page, timeout_seconds: int) -> None:
+    log("Not logged in. Please scan the QR code or finish login in the opened Chrome window.")
+    deadline = time.time() + timeout_seconds
+    while time.time() < deadline:
+        if is_logged_in(page):
+            save_state(page)
+            return
+        try:
+            login_button = page.get_by_text(re.compile("登录|账号登录")).first
+            if login_button.count():
+                login_button.click(timeout=1000)
+        except Exception:
+            pass
+        page.wait_for_timeout(2000)
+
+    raise TimeoutError(f"Login timed out after {timeout_seconds} seconds.")
+
+
+def ensure_login(page: Page, login_timeout: int) -> None:
+    page.goto(HOME_URL, wait_until="domcontentloaded", timeout=60000)
+    page.wait_for_timeout(2500)
+    if is_logged_in(page):
+        log("Already logged in, using saved browser session.")
+        dismiss_onboarding(page)
+        save_state(page)
+        return
+
+    page.goto(LOGIN_URL, wait_until="domcontentloaded", timeout=60000)
+    page.wait_for_timeout(1500)
+    wait_for_scan_login(page, login_timeout)
+    page.goto(HOME_URL, wait_until="domcontentloaded", timeout=60000)
+    page.wait_for_load_state("domcontentloaded")
+    dismiss_onboarding(page)
+
+
+def dismiss_onboarding(page: Page) -> None:
+    for text in ("跳过", "不再提示", "我知道了"):
+        try:
+            button = page.get_by_text(text, exact=True).first
+            if button.count():
+                button.click(timeout=1500)
+                page.wait_for_timeout(800)
+        except Exception:
+            pass
+
+
+def fill_search_box(page: Page, blogger: str) -> None:
+    dismiss_onboarding(page)
+    switch_search_type_to_blogger(page)
+    input_locator = page.locator("input").filter(
+        has_not_text=re.compile("有什么问题尽管问我")
+    ).first
+    input_locator.wait_for(state="visible", timeout=30000)
+    input_locator.click()
+    input_locator.fill(blogger)
+    input_locator.press("Enter")
+
+    try:
+        page.wait_for_url(re.compile(r".*/solar/pre-trade/note/kol.*"), timeout=15000)
+    except PlaywrightTimeoutError:
+        # Some versions require clicking the red search icon after entering text.
+        page.evaluate(
+            """
+            () => {
+              const input = [...document.querySelectorAll('input')]
+                .find((el) => /博主昵称|小红书号/.test(el.placeholder || '') || el.value);
+              const box = input?.closest('div');
+              const candidates = [...document.querySelectorAll('button, [role="button"], div')]
+                .filter((el) => {
+                  const r = el.getBoundingClientRect();
+                  return r.width >= 24 && r.width <= 80 && r.height >= 24 && r.height <= 80;
+                });
+              const near = candidates.find((el) => {
+                const er = el.getBoundingClientRect();
+                const br = box?.getBoundingClientRect();
+                return br && er.left > br.right - 20 && Math.abs(er.top - br.top) < 80;
+              });
+              near?.click();
+            }
+            """
+        )
+        page.wait_for_url(re.compile(r".*/solar/pre-trade/note/kol.*"), timeout=30000)
+
+    page.wait_for_timeout(3000)
+
+
+def switch_search_type_to_blogger(page: Page) -> None:
+    current_type = page.get_by_text(re.compile("笔记关键词|博主昵称"), exact=False).first
+    try:
+        current_type.click(timeout=3000)
+        page.wait_for_timeout(600)
+        option = page.get_by_text("博主昵称", exact=True).last
+        if option.count():
+            option.click(timeout=3000)
+            page.wait_for_timeout(600)
+    except Exception:
+        page.evaluate(
+            """
+            () => {
+              const nodes = [...document.querySelectorAll('div, span, button')];
+              const current = nodes.find((el) => /笔记关键词|博主昵称/.test(el.innerText || ''));
+              current?.click();
+              setTimeout(() => {
+                const option = [...document.querySelectorAll('div, span, button')]
+                  .reverse()
+                  .find((el) => (el.innerText || '').trim() === '博主昵称');
+                option?.click();
+              }, 300);
+            }
+            """
+        )
+        page.wait_for_timeout(1000)
+
+
+def open_blogger_detail(page: Page, blogger: str) -> Page:
+    log(f"Opening blogger detail: {blogger}")
+    pages_before = list(page.context.pages)
+    clicked = page.evaluate(
+        """
+        (name) => {
+          const textHasName = (el) => (el.innerText || '').includes(name);
+          const rows = [...document.querySelectorAll(
+            'tr, [class*="row"], [class*="table"], [class*="card"], [class*="item"]'
+          )].filter(textHasName);
+          const row = rows.find((el) => /粉丝|阅读|报价|合作/.test(el.innerText || '')) || rows[0];
+          if (!row) return false;
+          const target = row.querySelector('img, a, [class*="avatar"], [class*="name"]') || row;
+          target.scrollIntoView({ block: 'center', inline: 'center' });
+          target.click();
+          return true;
+        }
+        """,
+        blogger,
+    )
+    if not clicked:
+        page.get_by_text(blogger, exact=False).first.click(timeout=10000)
+
+    detail_page = wait_for_detail_page(page, pages_before, timeout_ms=30000)
+    if detail_page is None:
+        # If the first click only focused the row, try the visible name once more.
+        page.get_by_text(blogger, exact=False).first.click(timeout=10000)
+        detail_page = wait_for_detail_page(page, pages_before, timeout_ms=30000)
+
+    if detail_page is None:
+        raise TimeoutError("Clicked blogger result, but no blogger-detail page was detected.")
+
+    detail_page.bring_to_front()
+    detail_page.wait_for_load_state("domcontentloaded")
+    detail_page.get_by_text("笔记数据", exact=False).first.wait_for(timeout=30000)
+    detail_page.get_by_text("笔记数据", exact=False).first.click(timeout=5000)
+    detail_page.get_by_text("笔记数据", exact=False).first.scroll_into_view_if_needed()
+    detail_page.wait_for_timeout(2500)
+    return detail_page
+
+
+def wait_for_detail_page(page: Page, pages_before: list[Page], timeout_ms: int) -> Page | None:
+    deadline = time.time() + timeout_ms / 1000
+    known = set(pages_before)
+    while time.time() < deadline:
+        all_pages = list(page.context.pages)
+        candidates = [p for p in all_pages if p not in known] + [page] + all_pages
+        for candidate in candidates:
+            try:
+                if DETAIL_URL_MARK in candidate.url:
+                    return candidate
+            except Exception:
+                pass
+        page.wait_for_timeout(500)
+    return None
+
+
+def parse_notes_from_dom(page: Page) -> list[dict[str, str]]:
+    return page.evaluate(
+        """
+        () => {
+          const clean = (value) => (value || '').replace(/\\s+/g, ' ').trim();
+          const parseCard = (card) => {
+            const rawLines = (card.innerText || '')
+              .split('\\n')
+              .map((line) => line.trim())
+              .filter(Boolean);
+            const text = rawLines.join('\\n');
+            if (!/阅读/.test(text) || !/点赞/.test(text) || !/收藏/.test(text) || !/发布时间/.test(text)) return null;
+
+            const readAfter = (label) => {
+              const index = rawLines.findIndex((line) => line === label || line.includes(label));
+              if (index < 0) return '';
+              const sameLine = rawLines[index].replace(label, '').trim();
+              if (sameLine) return sameLine;
+              return rawLines[index + 1] || '';
+            };
+
+            const dateMatch = text.match(/20\\d{2}[./-]\\d{1,2}[./-]\\d{1,2}/);
+            const readIndex = rawLines.findIndex((line) => line === '阅读' || line.includes('阅读'));
+            const titleCandidates = rawLines.slice(0, readIndex >= 0 ? readIndex : rawLines.length).filter((line) => {
+              if (['阅读', '点赞', '收藏', '发布时间'].includes(line)) return false;
+              if (/^20\\d{2}[./-]/.test(line)) return false;
+              if (/^(合作|流量|图文笔记|视频笔记|全部类型|首页|笔记主页|数据概览|笔记数据|粉丝分析|笔记案例)$/.test(line)) return false;
+              if (/^[\\d,.wW万]+$/.test(line)) return false;
+              if (/中位数|获赞与|粉丝数|报价|筛选|全部流量|最新发布/.test(line)) return false;
+              return line.length >= 2;
+            });
+            const title = titleCandidates.slice(-2).join(' ').trim();
+
+            return {
+              title: clean(title),
+              read_count: clean(readAfter('阅读')),
+              like_count: clean(readAfter('点赞')),
+              collect_count: clean(readAfter('收藏')),
+              publish_time: clean(readAfter('发布时间') || (dateMatch ? dateMatch[0] : '')),
+            };
+          };
+
+          const all = [...document.querySelectorAll('div')];
+          const cards = all.filter((el) => {
+            const text = el.innerText || '';
+            const rect = el.getBoundingClientRect();
+            return rect.width >= 120 &&
+              rect.width <= 280 &&
+              rect.height >= 220 &&
+              rect.height <= 520 &&
+              /阅读/.test(text) &&
+              /点赞/.test(text) &&
+              /收藏/.test(text) &&
+              /发布时间/.test(text) &&
+              el.querySelector('img');
+          });
+
+          const parsed = cards.map(parseCard).filter(Boolean);
+          const seen = new Set();
+          return parsed.filter((item) => {
+            const key = [item.title, item.read_count, item.like_count, item.collect_count, item.publish_time].join('|');
+            if (seen.has(key)) return false;
+            seen.add(key);
+            return item.title || item.publish_time;
+          });
+        }
+        """
+    )
+
+
+def write_outputs(blogger: str, notes: list[dict[str, str]]) -> tuple[Path, Path]:
+    DATA_DIR.mkdir(exist_ok=True)
+    safe_name = re.sub(r'[<>:"/\\\\|?*\\s]+', "_", blogger).strip("_") or "blogger"
+    json_path = DATA_DIR / f"{safe_name}_notes.json"
+    csv_path = DATA_DIR / f"{safe_name}_notes.csv"
+
+    payload = {
+        "blogger": blogger,
+        "scraped_at": datetime.now().isoformat(timespec="seconds"),
+        "count": len(notes),
+        "notes": notes,
+    }
+    json_path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    with csv_path.open("w", newline="", encoding="utf-8-sig") as file:
+        writer = csv.DictWriter(
+            file,
+            fieldnames=["title", "read_count", "like_count", "collect_count", "publish_time"],
+        )
+        writer.writeheader()
+        writer.writerows(notes)
+
+    return json_path, csv_path
+
+
+def load_names(cli_names: list[str], names_file: str | None) -> list[str]:
+    names: list[str] = []
+    names.extend(name.strip() for name in cli_names if name.strip())
+
+    if names_file:
+        path = Path(names_file)
+        for line in path.read_text(encoding="utf-8-sig").splitlines():
+            name = line.strip()
+            if name and not name.startswith("#"):
+                names.append(name)
+
+    deduped: list[str] = []
+    seen = set()
+    for name in names:
+        if name not in seen:
+            seen.add(name)
+            deduped.append(name)
+    return deduped
+
+
+def scrape(blogger: str, login_timeout: int, headless: bool) -> list[dict[str, str]]:
+    result: dict[str, Any] = {"notes": []}
+
+    def action(page: Page) -> None:
+        ensure_login(page, login_timeout)
+        fill_search_box(page, blogger)
+        detail_page = open_blogger_detail(page, blogger)
+        result["detail_url"] = detail_page.url
+        result["notes"] = parse_notes_from_dom(detail_page)
+        save_state(detail_page)
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+    with DynamicSession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=str(PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=60000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=restore_cookies,
+        max_pages=1,
+    ) as session:
+        session.fetch(HOME_URL, page_action=action, wait=1000)
+
+    log(f"Detail URL: {result.get('detail_url', '')}")
+    return result["notes"]
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="Scrape XiaoHongShu Pgy blogger note metrics with Scrapling DynamicSession.",
+    )
+    parser.add_argument("bloggers", nargs="*", help="博主昵称，可传多个，例如：博主A 博主B")
+    parser.add_argument("--names-file", help="批量博主名单文件，一行一个昵称")
+    parser.add_argument("--login-timeout", type=int, default=300, help="首次扫码登录等待秒数")
+    parser.add_argument("--headless", action="store_true", help="无头模式；首次登录不要使用")
+    args = parser.parse_args()
+
+    bloggers = load_names(args.bloggers, args.names_file)
+    if not bloggers:
+        parser.error("Please provide at least one blogger name or --names-file.")
+
+    for index, blogger in enumerate(bloggers, start=1):
+        log(f"=== Blogger {index}/{len(bloggers)}: {blogger} ===")
+        notes = scrape(blogger, args.login_timeout, args.headless)
+        json_path, csv_path = write_outputs(blogger, notes)
+        log(f"Scraped {len(notes)} notes")
+        log(f"JSON: {json_path}")
+        log(f"CSV:  {csv_path}")
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/pgy_xhs_scraper_v2.py b/src/gyxx_flow/modules/content_marketing/runtime/pgy_xhs_scraper_v2.py
new file mode 100644
index 0000000..3fc93fe
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/pgy_xhs_scraper_v2.py
@@ -0,0 +1,1683 @@
+"""
+小红书蒲公英博主笔记采集 - V2 版本
+- 从款式-多维表格对照表读取待抓任务(达人名称+笔记标题+record_id)
+- 进入蒲公英后,按达人名称搜索,进入主页
+- 翻页找到指定笔记标题的那条,抓取阅读/点赞/收藏
+- 把阅读量回填到该款式的"7天曝光量"列(record_id 精确定位)
+
+用法:
+  # 拉取所有款式
+  python pgy_xhs_scraper_v2.py
+
+  # 只跑指定款式
+  python pgy_xhs_scraper_v2.py --style 7
+
+  # 只跑指定款式的特定任务(record_id)
+  python pgy_xhs_scraper_v2.py --style 7 --record recvlJt4bHa7PP
+
+  # 干跑(不写回表格)
+  python pgy_xhs_scraper_v2.py --style 7 --dry-run
+"""
+import argparse
+import asyncio
+import json
+import os
+import re
+import subprocess
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import (
+    PATHS,
+    resolve_layer_output,
+)
+
+from gyxx_flow.modules.content_marketing.runtime.creator_task_grouping import (
+    chunk_creator_groups,
+    group_tasks_by_creator as _group_tasks_by_creator,
+)
+from gyxx_flow.modules.content_marketing.runtime.collection_completeness import (
+    BLOCKED_INPUT,
+    RETRYABLE_FAILURE,
+    SUCCESS,
+    WRITE_FAILURE,
+    add_repeatable_style_argument,
+    atomic_write_json,
+    coerce_url,
+    extract_content_id,
+    finalize_summary,
+    match_tasks_to_cards,
+    merge_global_summaries,
+    merge_style_summary,
+    normalize_title as normalize_match_title,
+    select_requested_styles,
+    title_similarity,
+    upsert_result,
+)
+from typing import Any
+
+from playwright.sync_api import Page, TimeoutError as PlaywrightTimeoutError
+from scrapling.fetchers import DynamicSession
+
+LOGIN_URL = "https://pgy.xiaohongshu.com/"
+HOME_URL = "https://pgy.xiaohongshu.com/solar/pre-trade/home"
+DETAIL_URL_MARK = "/solar/pre-trade/blogger-detail/"
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+V2_DIR = PATHS.normalized_root / "v2_results"
+PGY_CHECKPOINT_DIR = PATHS.state_root / "checkpoints" / "pgy"
+MAPPING_PATH = PATHS.normalized_root / "mappings" / "款式_多维表格_对照.json"
+COOKIE_FILE = PATHS.browser_cookie_file
+PROFILE_DIR = PATHS.browser_profile_dir
+DEFAULT_BATCH_SIZE = 0
+DEFAULT_BATCH_WAIT = 0
+
+LARK_CLI = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK_CLI):
+    LARK_CLI = "lark-cli.cmd"
+
+
+def log(msg: str) -> None:
+    safe = str(msg).encode("gbk", errors="replace").decode("gbk", errors="replace")
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {safe}", flush=True)
+
+
+def wait_with_heartbeat(total_seconds: int, interval_seconds: int = 600) -> None:
+    """等待 total_seconds，每 interval_seconds 记录一次进度。"""
+    elapsed = 0
+    while elapsed < total_seconds:
+        chunk = min(interval_seconds, total_seconds - elapsed)
+        time.sleep(chunk)
+        elapsed += chunk
+        remaining = total_seconds - elapsed
+        if remaining > 0:
+            log(f"  [BATCH WAIT] 还剩 {remaining // 60} 分钟...")
+
+
+def is_browser_session_lost(exc: BaseException) -> bool:
+    text = f"{type(exc).__name__}: {exc}".casefold()
+    return any(token in text for token in (
+        "targetclosed", "has been closed", "page closed", "context closed",
+        "browser has been closed", "target page, context or browser",
+    ))
+
+
+# ===== lark-cli 调用 =====
+def call_lark_json(args: list[str]) -> dict:
+    cmd = [LARK_CLI] + args
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, text=True, env=env, encoding="utf-8")
+    if proc.returncode != 0:
+        raise RuntimeError(f"lark-cli 调用失败: {proc.stderr[:500]}\nstdout: {proc.stdout[:500]}")
+    text = proc.stdout
+    start = text.find("{")
+    if start < 0:
+        raise RuntimeError(f"无 JSON 返回: {text[:500]}")
+    return json.loads(text[start:], strict=False)
+
+
+def load_mapping(self_operated: bool = False) -> dict:
+    # 动态:从「合作达人」/「自营达人」多维表格地址表读取款式→各表地址,实时拉字段重建对照表
+    from gyxx_flow.modules.content_marketing.runtime import feishu_mapping
+    return feishu_mapping.load_mapping(DATA_DIR, self_operated=self_operated)
+
+
+def list_records_by_table(base_token: str, table_id: str, field_id: str | None = None) -> list[dict]:
+    """拉取一张表的所有记录(field_id 限定就只拿那一列,省 token)"""
+    args = [
+        "base", "+record-list",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--as", "user",
+        "--format", "json",
+        "--limit", "200",
+    ]
+    if field_id:
+        args += ["--field-id", field_id]
+    all_records = []
+    offset = 0
+    while True:
+        page_args = args + ["--offset", str(offset)]
+        resp = call_lark_json(page_args)
+        if not resp.get("ok"):
+            raise RuntimeError(f"拉记录失败: {resp.get('error')}")
+        d = resp["data"]
+        ids = d["field_id_list"]
+        rids = d["record_id_list"]
+        rows = d["data"]
+        for i, row in enumerate(rows):
+            rec = {"record_id": rids[i]}
+            for j, fid in enumerate(ids):
+                rec[fid] = row[j] if j < len(row) else None
+            all_records.append(rec)
+        offset += len(rows)
+        if not d.get("has_more", False) or len(rows) == 0:
+            break
+    return all_records
+
+
+def write_back(base_token: str, table_id: str, record_id: str, field_id: str, value) -> bool:
+    """把 value 写回指定记录的指定字段"""
+    payload = {field_id: value}
+    resp = call_lark_json([
+        "base", "+record-upsert",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--record-id", record_id,
+        "--as", "user",
+        "--format", "json",
+        "--json", json.dumps(payload, ensure_ascii=False),
+    ])
+    if not resp.get("ok"):
+        log(f"  回填失败 record_id={record_id}: {resp.get('error')}")
+        return False
+    return True
+
+
+def parse_publish_time(value) -> datetime | None:
+    """把飞书多维表格里的发布时间字段解析成 datetime。
+    飞书 date/datetime 字段返回毫秒时间戳(数值)。"""
+    if value is None or value == "":
+        return None
+    if isinstance(value, (int, float)):
+        # 飞书毫秒
+        try:
+            return datetime.fromtimestamp(value / 1000)
+        except (ValueError, OSError):
+            return None
+    if isinstance(value, str):
+        v = value.strip()
+        if not v:
+            return None
+        # 试毫秒字符串
+        if v.isdigit():
+            try:
+                return datetime.fromtimestamp(int(v) / 1000)
+            except (ValueError, OSError):
+                return None
+        # 试 ISO 格式 (2026-06-01 或 2026-06-01 12:34)
+        for fmt in ("%Y-%m-%dT%H:%M:%S", "%Y-%m-%d %H:%M:%S", "%Y-%m-%d", "%Y/%m/%d"):
+            try:
+                return datetime.strptime(v[:len(fmt) + 2], fmt)
+            except ValueError:
+                continue
+    return None
+
+
+def pick_read_field(fmap: dict, pub_time: datetime | None,
+                    today: datetime | None = None) -> tuple[str, str, str] | None:
+    """根据发布时间距今天的天数,挑出要回填的曝光量字段。
+    返回 (logical_key, field_id, field_name) 或 None。
+    规则:
+      0-7 天  -> 7天曝光量
+      8-14 天 -> 14天曝光量
+      15-21 天 -> 21天曝光量
+      22-28 天 -> 28天曝光量
+      29+ 天  -> 月底曝光量
+    未来时间(days<0)也返回 None,调用方应区分"无发布时间"和"未来时间"。
+    """
+    if pub_time is None:
+        return None  # 缺发布时间就不写,让调用方 skip
+    today = today or datetime.now()
+    days = (today.date() - pub_time.date()).days
+    if days < 0:
+        return "FUTURE"  # type: ignore[return-value]  # sentinel: 未来时间不写
+    if days <= 7:
+        logical = "read_count_7d"
+    elif days <= 14:
+        logical = "read_count_14d"
+    elif days <= 21:
+        logical = "read_count_21d"
+    elif days <= 28:
+        logical = "read_count_28d"
+    else:
+        logical = "month_end"  # 月底曝光量
+    info = fmap.get(logical)
+    if info:
+        return (logical, info["field_id"], info["field_name"])
+    # 目标字段不存在时不降级写7天桶,避免数据错位
+    log(f"    [WARN] pick_read_field: {logical} 不在 fmap,跳过回填(不降级到7天)")
+    return None
+
+
+# ===== 浏览器:登录与原版相同 =====
+def restore_cookies(page: Page) -> None:
+    if not COOKIE_FILE.exists():
+        return
+    try:
+        cookies = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+        page.context.add_cookies(cookies)
+        log(f"Loaded cookies: {COOKIE_FILE}")
+    except Exception as exc:
+        log(f"Failed to load cookies: {exc}")
+
+
+def save_state(page: Page) -> None:
+    DATA_DIR.mkdir(exist_ok=True)
+    cookies = page.context.cookies()
+    COOKIE_FILE.write_text(
+        json.dumps(cookies, ensure_ascii=False, indent=2), encoding="utf-8"
+    )
+    log(f"Saved cookies: {COOKIE_FILE}")
+
+
+def is_logged_in(page: Page) -> bool:
+    if "/solar/pre-trade/" in page.url:
+        return True
+    try:
+        page.get_by_text("找博主", exact=True).wait_for(timeout=2500)
+        return True
+    except PlaywrightTimeoutError:
+        return False
+
+
+def open_scan_login(page: Page) -> None:
+    """Open the login dialog and prefer its QR-code login tab."""
+    for pattern in (r"^登录$|账号登录",):
+        try:
+            button = page.get_by_text(re.compile(pattern)).first
+            if button.count() and button.is_visible():
+                button.click(timeout=2000)
+                page.wait_for_timeout(500)
+        except Exception:
+            pass
+
+    # The login dialog uses an image-only folded-corner switch. Its bundled
+    # asset name is stable even though the generated CSS class is not.
+    try:
+        qr_switch = page.locator('img[src*="qr_code"]').first
+        if qr_switch.count() and qr_switch.is_visible():
+            # Playwright's actionability check is blocked by the folded-corner
+            # overlay. A native DOM click triggers the image's real handler.
+            qr_switch.evaluate("e => e.click()")
+            page.wait_for_timeout(1000)
+            return
+    except Exception:
+        pass
+
+    for pattern in (r"扫码登录|二维码登录|扫码",):
+        try:
+            button = page.get_by_text(re.compile(pattern)).first
+            if button.count() and button.is_visible():
+                button.click(timeout=2000)
+                page.wait_for_timeout(500)
+        except Exception:
+            pass
+
+
+def wait_for_scan_login(page: Page, timeout_seconds: int) -> None:
+    log("Not logged in. 请在打开的 Chrome 中扫码登录。")
+    open_scan_login(page)
+    deadline = time.time() + timeout_seconds
+    next_login_click = time.time() + 5
+    while time.time() < deadline:
+        if is_logged_in(page):
+            save_state(page)
+            return
+        if time.time() >= next_login_click:
+            open_scan_login(page)
+            next_login_click = time.time() + 5
+        page.wait_for_timeout(2000)
+    raise TimeoutError(f"Login timed out after {timeout_seconds} seconds.")
+
+
+def ensure_login(page: Page, login_timeout: int) -> None:
+    page.goto(HOME_URL, wait_until="domcontentloaded", timeout=60000)
+    page.wait_for_timeout(2500)
+    if is_logged_in(page):
+        log("Already logged in.")
+        save_state(page)
+        return
+    page.goto(LOGIN_URL, wait_until="domcontentloaded", timeout=60000)
+    page.wait_for_timeout(1500)
+    wait_for_scan_login(page, login_timeout)
+    page.goto(HOME_URL, wait_until="domcontentloaded", timeout=60000)
+    page.wait_for_load_state("domcontentloaded")
+    # 关闭弹窗
+    for text in ("跳过", "不再提示", "我知道了"):
+        try:
+            b = page.get_by_text(text, exact=True).first
+            if b.count():
+                b.click(timeout=1500)
+                page.wait_for_timeout(800)
+        except Exception:
+            pass
+
+
+def _ensure_search_results(
+    page: Page,
+    query: str,
+    result_name: str | None = None,
+    max_retries: int = 3,
+) -> bool:
+    """Wait for search rows while always retrying the exact submitted query."""
+    result_name = result_name or query
+    for attempt in range(max_retries):
+        page.wait_for_timeout(1000)
+        has_results = page.evaluate(
+            """({query, name}) => {
+                const rows = [...document.querySelectorAll(
+                    'tr, [class*="row"], [class*="table"], [class*="card"], [class*="item"]'
+                )];
+                return rows.some(el => {
+                    const text = el.innerText || '';
+                    return (text.includes(query) || text.includes(name))
+                      && /粉丝|阅读|报价|合作/.test(text);
+                });
+            }""",
+            {"query": query, "name": result_name},
+        )
+        if has_results:
+            log(f"  搜索结果已加载,找到 '{result_name}' 相关内容")
+            return True
+
+        current_value = page.evaluate(
+            """() => {
+                const inputs = [...document.querySelectorAll('input')];
+                const vis = inputs.find(el => !el.hidden && el.offsetWidth > 0);
+                return vis ? vis.value : '';
+            }"""
+        )
+        if not current_value or query not in current_value:
+            log(f"  [WARN] 搜索框被清空(当前值='{current_value}'),第 {attempt + 1} 次重试搜索 '{query}'")
+            try:
+                inp = page.locator("input").filter(
+                    has_not_text=re.compile("有什么问题尽管问我")
+                ).first
+                inp.click()
+                inp.fill(query)
+                inp.press("Enter")
+                try:
+                    page.wait_for_url(re.compile(r".*/solar/pre-trade/note/kol.*"), timeout=10000)
+                except PlaywrightTimeoutError:
+                    pass
+            except Exception as exc:
+                log(f"  [WARN] 重新搜索失败: {exc}")
+        else:
+            page.wait_for_timeout(1500)
+
+    log(f"  [WARN] 搜索 '{query}' 后未检测到结果列表(已重试 {max_retries} 次)")
+    return False
+
+
+def search_blogger(page: Page, blogger: str, creator_id: str | None = None) -> None:
+    """搜索博主/达人。
+    优先用 creator_id 精确搜(更稳),没有则用 name 模糊搜。
+    """
+    # 切换到"博主昵称"或"小红书号"模式
+    # 蒲公英支持两种:博主昵称 / 小红书号
+    # 如果有 creator_id,切到小红书号;否则切到博主昵称
+    try:
+        cur = page.get_by_text(re.compile("笔记关键词|博主昵称|小红书号"), exact=False).first
+        cur.click(timeout=3000)
+        page.wait_for_timeout(300)
+        if creator_id:
+            opt = page.get_by_text("小红书号", exact=True).last
+            if opt.count() == 0:
+                opt = page.get_by_text("小红书 ID", exact=True).last
+        else:
+            opt = page.get_by_text("博主昵称", exact=True).last
+        if opt.count():
+            opt.click(timeout=3000)
+            page.wait_for_timeout(300)
+    except Exception:
+        page.evaluate(
+            """
+            (useId) => {
+              const nodes = [...document.querySelectorAll('div, span, button')];
+              const current = nodes.find((el) => /笔记关键词|博主昵称|小红书号/.test(el.innerText || ''));
+              current?.click();
+              setTimeout(() => {
+                const want = useId ? '小红书号' : '博主昵称';
+                const opt = [...document.querySelectorAll('div, span, button')]
+                  .reverse()
+                  .find((el) => (el.innerText || '').trim() === want);
+                if (!opt && useId) {
+                  // 找任何含"小红书"或"ID"的选项
+                  const fb = [...document.querySelectorAll('div, span, button')]
+                    .reverse()
+                    .find((el) => /小红书/.test(el.innerText || ''));
+                  fb?.click();
+                } else {
+                  opt?.click();
+                }
+              }, 300);
+            }
+            """,
+            bool(creator_id),
+        )
+        page.wait_for_timeout(500)
+
+    input_locator = page.locator("input").filter(
+        has_not_text=re.compile("有什么问题尽管问我")
+    ).first
+    input_locator.wait_for(state="visible", timeout=30000)
+    input_locator.click()
+    # 优先 ID,其次名称
+    query = str(creator_id) if creator_id else blogger
+    log(f"  搜索: {query} (类型: {'ID' if creator_id else '昵称'})")
+    input_locator.fill(query)
+    input_locator.press("Enter")
+
+    try:
+        page.wait_for_url(re.compile(r".*/solar/pre-trade/note/kol.*"), timeout=15000)
+    except PlaywrightTimeoutError:
+        # 部分版本要点搜索按钮
+        page.evaluate(
+            """
+            () => {
+              const input = [...document.querySelectorAll('input')]
+                .find((el) => /博主昵称|小红书号/.test(el.placeholder || '') || el.value);
+              const box = input?.closest('div');
+              const candidates = [...document.querySelectorAll('button, [role="button"], div')]
+                .filter((el) => {
+                  const r = el.getBoundingClientRect();
+                  return r.width >= 24 && r.width <= 80 && r.height >= 24 && r.height <= 80;
+                });
+              const near = candidates.find((el) => {
+                const er = el.getBoundingClientRect();
+                const br = box?.getBoundingClientRect();
+                return br && er.left > br.right - 20 && Math.abs(er.top - br.top) < 80;
+              });
+              near?.click();
+            }
+            """
+        )
+
+    _ensure_search_results(page, query=query, result_name=blogger)
+
+
+def dismiss_onboarding(page: Page) -> None:
+    """关闭蒲公英新手引导/弹窗"""
+    for text in ("跳过", "不再提示", "我知道了", "稍后再说", "下次再说"):
+        try:
+            button = page.get_by_text(text, exact=True).first
+            if button.count():
+                button.click(timeout=1500)
+                page.wait_for_timeout(800)
+        except Exception:
+            pass
+
+
+def switch_search_type_to_blogger(page: Page) -> None:
+    current_type = page.get_by_text(re.compile("笔记关键词|博主昵称"), exact=False).first
+    try:
+        current_type.click(timeout=3000)
+        page.wait_for_timeout(600)
+        option = page.get_by_text("博主昵称", exact=True).last
+        if option.count():
+            option.click(timeout=3000)
+            page.wait_for_timeout(600)
+    except Exception:
+        page.evaluate(
+            """
+            () => {
+              const nodes = [...document.querySelectorAll('div, span, button')];
+              const current = nodes.find((el) => /笔记关键词|博主昵称/.test(el.innerText || ''));
+              current?.click();
+              setTimeout(() => {
+                const option = [...document.querySelectorAll('div, span, button')]
+                  .reverse()
+                  .find((el) => (el.innerText || '').trim() === '博主昵称');
+                option?.click();
+              }, 300);
+            }
+            """
+        )
+        page.wait_for_timeout(1000)
+
+
+def open_blogger_detail(page: Page, blogger: str, creator_id: str | None = None) -> Page | None:
+    log(f"Opening blogger detail: {blogger} (id={creator_id})")
+    pages_before = list(page.context.pages)
+    # Resolve the exact row's href first. Calling window.open inside page JS is
+    # frequently blocked by Chromium and was the main source of false misses.
+    target = page.evaluate(
+        r"""
+        ({name, id}) => {
+          const compact = (value) => (value || '').replace(/\s+/g, '').trim();
+          const wantedName = compact(name);
+          const links = [...document.querySelectorAll('a[href*="/blogger-detail/"]')];
+          const candidates = links.map((link) => {
+            const row = link.closest(
+              'tr, [role="row"], .ant-table-row, [class*="blogger-card"], '
+              + '[class*="author-card"], [class*="creator-card"], [class*="user-card"]'
+            ) || link.parentElement || link;
+            const text = compact(row.innerText || link.innerText || '');
+            return {link, row, text};
+          });
+          const idCandidate = id
+            ? candidates.find((item) => item.text.includes(compact(id)))
+            : null;
+          const nameCandidates = candidates
+            .filter((item) => wantedName && item.text.includes(wantedName))
+            .sort((a, b) => a.text.length - b.text.length);
+          const candidate = idCandidate || nameCandidates[0];
+          if (candidate?.link?.href) {
+            return {
+              ok: true,
+              url: candidate.link.href,
+              method: idCandidate ? 'id-match' : 'name-match',
+            };
+          }
+          const rows = [...document.querySelectorAll(
+            'tr, [role="row"], .ant-table-row, [class*="blogger-card"], '
+            + '[class*="author-card"], [class*="creator-card"], [class*="user-card"]'
+          )];
+          const row = (id ? rows.find((el) => compact(el.innerText).includes(compact(id))) : null)
+            || rows.filter((el) => compact(el.innerText).includes(wantedName))
+              .sort((a, b) => compact(a.innerText).length - compact(b.innerText).length)[0];
+          if (!row) return {ok: false, reason: 'no-row'};
+          const clickTarget = row.querySelector('img, a, [class*="avatar"], [class*="name"]') || row;
+          const target = clickTarget;
+          target.scrollIntoView({ block: 'center', inline: 'center' });
+          target.click();
+          return {ok: true, method: id ? 'id-match-click' : 'name-match-click'};
+        }
+        """,
+        {"name": blogger, "id": creator_id or ""},
+    )
+    detail_page = None
+    if target and target.get("ok") and target.get("url"):
+        try:
+            detail_page = page.context.new_page()
+            detail_page.goto(target["url"], wait_until="domcontentloaded", timeout=45000)
+        except Exception as exc:
+            if is_browser_session_lost(exc):
+                raise
+            log(f"    [detail] 直接访问详情 href 失败: {exc}")
+            try:
+                if detail_page is not None and not detail_page.is_closed():
+                    detail_page.close()
+            except Exception:
+                pass
+            detail_page = None
+
+    if not target or not target.get("ok"):
+        log(f"    [detail] 找不到博主行: {target}")
+        # 兜底: get_by_text
+        try:
+            page.get_by_text(blogger, exact=False).first.click(timeout=10000)
+        except Exception as exc:
+            if is_browser_session_lost(exc):
+                raise
+            log(f"    [detail] 备用 click 也失败: {exc}")
+            return None
+
+    if detail_page is None:
+        detail_page = wait_for_detail_page(page, pages_before, timeout_ms=30000)
+    if detail_page is None:
+        log(f"    [detail] 30s 内未检测到详情页: {blogger}")
+        return None
+
+    detail_page.bring_to_front()
+    try:
+        detail_page.wait_for_load_state("domcontentloaded")
+        detail_page.get_by_text("笔记数据", exact=False).first.wait_for(timeout=30000)
+        detail_page.get_by_text("笔记数据", exact=False).first.click(timeout=5000)
+        detail_page.get_by_text("笔记数据", exact=False).first.scroll_into_view_if_needed()
+        detail_page.wait_for_timeout(1500)
+    except Exception as exc:
+        if is_browser_session_lost(exc):
+            raise
+        log(f"    [detail] '笔记数据' tab 操作失败: {exc}")
+        try:
+            if detail_page is page:
+                detail_page.goto(HOME_URL, wait_until="domcontentloaded", timeout=30000)
+            elif not detail_page.is_closed():
+                detail_page.close()
+        except Exception as cleanup_exc:
+            if is_browser_session_lost(cleanup_exc):
+                raise
+        return None
+    return detail_page
+
+
+def wait_for_detail_page(page: Page, pages_before: list[Page], timeout_ms: int) -> Page | None:
+    deadline = time.time() + timeout_ms / 1000
+    known = set(pages_before)
+    while time.time() < deadline:
+        all_pages = list(page.context.pages)
+        candidates = [p for p in all_pages if p not in known] + [page] + all_pages
+        for c in candidates:
+            try:
+                if DETAIL_URL_MARK in c.url:
+                    return c
+            except Exception:
+                pass
+        page.wait_for_timeout(500)
+    return None
+
+
+# ===== 翻页 + 精准找笔记 =====
+def parse_cards_on_page(page: Page) -> list[dict]:
+    """解析当前页所有笔记卡片"""
+    return page.evaluate(
+        """
+        () => {
+          const clean = (v) => (v || '').replace(/\\s+/g, ' ').trim();
+          const valueAfter = (lines, label) => {
+            const idx = lines.findIndex((l) => l === label || l.includes(label));
+            if (idx < 0) return '';
+            const same = lines[idx].replace(label, '').trim();
+            return same || (lines[idx + 1] || '');
+          };
+          const parseCard = (card) => {
+            const lines = (card.innerText || '').split('\\n').map((l) => l.trim()).filter(Boolean);
+            const text = lines.join('\\n');
+            if (!/阅读/.test(text) || !/点赞/.test(text) || !/收藏/.test(text) || !/发布时间/.test(text)) return null;
+            const readIdx = lines.findIndex((l) => l === '阅读' || l.includes('阅读'));
+            const titles = lines.slice(0, readIdx >= 0 ? readIdx : lines.length).filter((l) => {
+              if (['阅读','点赞','收藏','发布时间'].includes(l)) return false;
+              if (/^20\\d{2}[./-]/.test(l)) return false;
+              if (/^(合作|流量|图文笔记|视频笔记|全部类型|首页|笔记主页|数据概览|笔记数据|粉丝分析|笔记案例)$/.test(l)) return false;
+              if (/^[\\d,.wW万]+$/.test(l)) return false;
+              if (/中位数|获赞与|粉丝数|报价|筛选|全部流量|最新发布/.test(l)) return false;
+              return l.length >= 2;
+            });
+            const title = titles.slice(-2).join(' ').trim();
+            const link = card.querySelector(
+              'a[href*="/explore/"], a[href*="/discovery/item/"], a[href*="/note/"]'
+            );
+            const href = link?.href || '';
+            const idMatch = href.match(new RegExp('/(?:explore|discovery/item|note)/([A-Za-z0-9_-]+)'));
+            const dateMatch = text.match(/20\\d{2}[./-]\\d{1,2}[./-]\\d{1,2}/);
+            return {
+              title: clean(title),
+              read_count: clean(valueAfter(lines, '阅读')),
+              like_count: clean(valueAfter(lines, '点赞')),
+              collect_count: clean(valueAfter(lines, '收藏')),
+              publish_time: clean(valueAfter(lines, '发布时间') || (dateMatch ? dateMatch[0] : '')),
+              href,
+              note_id: idMatch ? idMatch[1] : '',
+            };
+          };
+          const explicit = [...document.querySelectorAll(
+            '.note-card-wrapper, [class*="note-card-wrapper"], [class*="note-card"]'
+          )].filter((el) => {
+            const text = el.innerText || '';
+            return /阅读/.test(text) && /点赞/.test(text) && /收藏/.test(text) && /发布时间/.test(text);
+          });
+          const all = [...document.querySelectorAll('div')];
+          const fallback = all.filter((el) => {
+            const r = el.getBoundingClientRect();
+            const text = el.innerText || '';
+            return r.width >= 120 && r.width <= 280 && r.height >= 220 && r.height <= 520
+              && /阅读/.test(text) && /点赞/.test(text) && /收藏/.test(text) && /发布时间/.test(text)
+              && el.querySelector('img');
+          });
+          const cards = explicit.length ? explicit : fallback;
+          const seen = new Set();
+          return cards.map(parseCard).filter(Boolean).filter((it) => {
+            const k = [it.title, it.read_count, it.like_count, it.collect_count, it.publish_time].join('|');
+            if (seen.has(k)) return false;
+            seen.add(k);
+            return it.title;
+          });
+        }
+        """
+    )
+
+
+def parse_cards_stable(page: Page, max_attempts: int = 4, wait_ms: int = 700) -> list[dict]:
+    """Wait for card rendering to settle; an initial empty page is not an end page."""
+    previous_signature = None
+    latest: list[dict] = []
+    for _ in range(max_attempts):
+        latest = parse_cards_on_page(page)
+        signature = tuple(
+            (card.get("note_id"), card.get("title"), card.get("read_count"))
+            for card in latest
+        )
+        if signature and signature == previous_signature:
+            return latest
+        previous_signature = signature
+        page.wait_for_timeout(wait_ms)
+    return latest
+
+
+def normalize_title(s: str) -> str:
+    return normalize_match_title(s)
+
+
+def title_match(card_title: str, target_title: str) -> bool:
+    """模糊匹配:去掉空白/标点后看是否包含/被包含"""
+    a = normalize_title(card_title)
+    b = normalize_title(target_title)
+    if not a or not b:
+        return False
+    return title_similarity(card_title, target_title) >= 0.78
+
+
+def go_next_page(page: Page) -> bool:
+    """点下一页箭头。
+    蒲公英分页器 .d-pagination-page 结构:
+      - ← / <  (上一页): SVG chevron-left
+      - 1 2 3 ... 数字
+      - → / >  (下一页): SVG chevron-right (目标!)
+    关键: 下一页是 SVG 节点,不是文字。
+    """
+    page_before = page.evaluate(
+        """() => {
+          const cards = [...document.querySelectorAll('.note-card-wrapper, [class*="note-card-wrapper"], [class*="note-card"]')];
+          return cards.map(c => (c.querySelector('.note-card__title, [class*="title"]') || {}).innerText || c.innerText || '').slice(0, 3).join('|');
+        }"""
+    )
+
+    moved = page.evaluate(
+        r"""
+        () => {
+          const pages = [...document.querySelectorAll('.d-pagination-page')];
+          // 1) 找含 svg 的按钮 (chevron)
+          const chevrons = pages.filter(p => p.querySelector('svg'));
+          for (const p of chevrons) {
+            if (p.classList.contains('disabled')) continue;
+            const svg = p.querySelector('svg');
+            const path = svg?.querySelector('path') || svg;
+            const d = path?.getAttribute('d') || '';
+            const isRight = d.includes('M8.29289') || d.includes('M9.29289');
+            const isLeft  = d.includes('M15.7071') || d.includes('M14.7071');
+            if (isRight && !isLeft) {
+              p.scrollIntoView({ block: 'center', inline: 'center' });
+              p.click();
+              return {clicked: true, text: 'svg-right'};
+            }
+          }
+          // 2) 兜底:找 .d-pagination-page 中最后一个 (如果末尾含 svg)
+          const last = pages[pages.length - 1];
+          if (last && last.querySelector('svg') && !last.classList.contains('disabled')) {
+            last.scrollIntoView({ block: 'center', inline: 'center' });
+            last.click();
+            return {clicked: true, text: 'last-svg'};
+          }
+          return {clicked: false};
+        }
+        """
+    )
+    if not moved or not moved.get("clicked"):
+        return False
+
+    # 等页面真的变了
+    changed = False
+    for _ in range(30):  # 30 * 200ms = 6s
+        page.wait_for_timeout(200)
+        page_after = page.evaluate(
+            """() => {
+              const cards = [...document.querySelectorAll('.note-card-wrapper, [class*="note-card-wrapper"], [class*="note-card"]')];
+              return cards.map(c => (c.querySelector('.note-card__title, [class*="title"]') || {}).innerText || c.innerText || '').slice(0, 3).join('|');
+            }"""
+        )
+        if page_after and page_after != page_before:
+            changed = True
+            break
+
+    if isinstance(moved, dict):
+        log(f"  -> 点了 '{moved.get('text', '?')}'")
+    return changed
+
+
+def find_note_by_title(page: Page, target_title: str, max_pages: int = 8) -> dict | None:
+    """
+    在博主主页翻页,找指定标题的笔记。
+    命中后返回 {read_count, like_count, collect_count, publish_time, page}。
+    如果没有 target_title(降级模式),直接返回第 1 页第 1 条(最近发布)。
+    """
+    if not target_title:
+        # 降级:没指定标题就抓首页第一条
+        cards = parse_cards_on_page(page)
+        if cards:
+            log(f"  降级模式:没指定标题,取首页第一条 '{cards[0]['title'][:30]}'")
+            cards[0]["page"] = 1
+            cards[0]["fallback"] = True
+            return cards[0]
+        return None
+
+    for p in range(1, max_pages + 1):
+        cards = parse_cards_on_page(page)
+        log(f"  第 {p} 页抓到 {len(cards)} 张笔记卡片")
+        for c in cards:
+            if title_match(c.get("title", ""), target_title):
+                log(f"  命中: 目标='{target_title[:30]}...' 实际='{c['title'][:30]}...'")
+                c["page"] = p
+                return c
+        if p < max_pages:
+            if not go_next_page(page):
+                log(f"  已是末页(无下一页按钮),共 {p} 页")
+                break
+    return None
+
+def find_notes_by_titles(page: Page, target_titles: list[str],
+                          max_pages: int = 25) -> dict[str, dict]:
+    """遍历所有页面，一次性收集所有匹配标题的笔记。
+    返回 {normalize_title(target): card_dict}。未找到的标题不在返回字典中。
+    """
+    if not target_titles:
+        return {}
+
+    remaining: dict[str, str] = {}  # norm_title -> original_title
+    for t in target_titles:
+        if t:
+            remaining[normalize_title(t)] = t
+
+    found: dict[str, dict] = {}  # norm_title -> card
+
+    for p in range(1, max_pages + 1):
+        cards = parse_cards_on_page(page)
+        log(f"  第 {p} 页抓到 {len(cards)} 张笔记卡片 (还剩 {len(remaining)} 个目标)")
+
+        for c in cards:
+            card_norm = normalize_title(c.get("title", ""))
+            if not card_norm:
+                continue
+            for target_norm in list(remaining.keys()):
+                if card_norm == target_norm:
+                    found[target_norm] = {**c, "page": p}
+                    del remaining[target_norm]
+                    break
+                elif len(target_norm) >= 6 and (target_norm in card_norm or card_norm in target_norm):
+                    found[target_norm] = {**c, "page": p}
+                    del remaining[target_norm]
+                    break
+
+        if not remaining:
+            log(f"  所有 {len(found)} 个目标已找到")
+            return found
+
+        if p < max_pages:
+            if not go_next_page(page):
+                log(f"  已是末页,共 {p} 页")
+                break
+
+    if remaining:
+        log(f"  未找到 {len(remaining)} 个目标: {[remaining[k][:30] for k in list(remaining.keys())[:5]]}")
+    return found
+
+
+def find_notes_for_tasks(
+    page: Page,
+    tasks: list[dict],
+    max_pages: int = 25,
+) -> tuple[dict[str, dict], list[str], bool]:
+    """Scan one creator once and match URL/note ID before conservative titles."""
+    matching_tasks = [dict(task) for task in tasks if task.get("record_id")]
+    task_by_id = {str(task["record_id"]): task for task in matching_tasks}
+    remaining = dict(task_by_id)
+    all_cards: list[dict] = []
+    page_limit_hit = False
+
+    def match_all(min_score: float = 0.78) -> dict[str, dict]:
+        return match_tasks_to_cards(matching_tasks, all_cards, "xhs", min_score=min_score)
+
+    for page_number in range(1, max_pages + 1):
+        cards = parse_cards_stable(page)
+        for card in cards:
+            card = dict(card)
+            card["page"] = page_number
+            if not card.get("note_id"):
+                card["note_id"] = extract_content_id(card.get("href"), "xhs")
+            all_cards.append(card)
+        log(f"  第 {page_number} 页稳定抓到 {len(cards)} 张笔记卡片 (还剩 {len(remaining)} 个目标)")
+
+        # During pagination only accept content IDs or exact titles. Fuzzy
+        # matching is delayed until all available candidates are known.
+        exact = match_all(min_score=1.0)
+        safe_ids = {
+            record_id for record_id, card in exact.items()
+            if card.get("match_method") == "content_id"
+        }
+        remaining = {record_id: task for record_id, task in task_by_id.items() if record_id not in safe_ids}
+        if not remaining:
+            break
+        if page_number == max_pages:
+            page_limit_hit = True
+            break
+        if not go_next_page(page):
+            break
+
+    found = match_all()
+    remaining = {record_id: task for record_id, task in task_by_id.items() if record_id not in found}
+
+    candidates = [str(card.get("title") or "")[:80] for card in all_cards if card.get("title")]
+    if remaining:
+        log(f"  未找到 {len(remaining)} 个目标;候选卡片 {len(all_cards)} 张")
+    return found, candidates[-10:], page_limit_hit
+
+
+# ===== 主流程 =====
+def extract_target_tasks(style: dict, only_record_ids: set[str] | None = None,
+                          platform_filter: str = "小红书",
+                          require_all: bool = True) -> list[dict]:
+    """
+    根据款式对照表,从多维表格拉出本款式所有待抓任务。
+
+    筛选条件(按用户原话):
+    1. 投放平台 字段 含 platform_filter(默认"小红书")
+    2. (可选)达人名称/笔记标题/发布时间/发布笔记链接 4 个都非空
+    """
+    base_token = style["base_token"]
+    table_id = style["table_id"]
+    fmap = style["field_map"]
+
+    name_fid = fmap.get("creator_name", {}).get("field_id")
+    id_fid = fmap.get("creator_id", {}).get("field_id")
+    title_fid = fmap.get("note_title", {}).get("field_id")
+    pub_fid = fmap.get("publish_time", {}).get("field_id")
+    url_fid = fmap.get("note_url", {}).get("field_id")
+    platform_fid = fmap.get("platform", {}).get("field_id")
+    read7_fid = fmap.get("read_count_7d", {}).get("field_id")
+
+    missing_fields = []
+    if not (name_fid or id_fid):
+        missing_fields.append("creator_name/creator_id")
+    for logical, field_id in (
+        ("note_title", title_fid),
+        ("publish_time", pub_fid),
+        ("note_url", url_fid),
+        ("platform", platform_fid),
+    ):
+        if not field_id:
+            missing_fields.append(logical)
+    if missing_fields:
+        error = f"missing field mapping: {','.join(missing_fields)}"
+        style["_extract_error"] = error
+        style["_extract_stats"] = {
+            "source_records": 0,
+            "selected_scope_records": 0,
+            "skipped_platform": 0,
+            "skipped_incomplete": 0,
+        }
+        log(f"  [ERROR] {style['name']} {error}")
+        return []
+    style.pop("_extract_error", None)
+
+    log(f"  拉取款式 [{style['index']:02d} {style['name']}] 的待抓列表 (平台={platform_filter})...")
+
+    all_records = list_records_by_table(base_token, table_id)
+    log(f"    全表共 {len(all_records)} 条记录")
+
+    def is_platform_match(v) -> bool:
+        if not v:
+            return False
+        if isinstance(v, list):
+            return any(platform_filter in str(x) for x in v)
+        return platform_filter in str(v)
+
+    def is_nonempty(v) -> bool:
+        if v is None:
+            return False
+        if isinstance(v, list) and len(v) == 0:
+            return False
+        if isinstance(v, str) and v.strip() == "":
+            return False
+        return True
+
+    tasks = []
+    selected_scope_records = 0
+    skipped_platform = 0
+    skipped_incomplete = 0
+    for rec in all_records:
+        rid = rec["record_id"]
+        if only_record_ids and rid not in only_record_ids:
+            continue
+        selected_scope_records += 1
+
+        # 平台筛选
+        if platform_fid:
+            if not is_platform_match(rec.get(platform_fid)):
+                skipped_platform += 1
+                continue
+
+        name = rec.get(name_fid) if name_fid else None
+        creator_id_val = rec.get(id_fid) if id_fid else None
+        title = rec.get(title_fid) if title_fid else None
+        pub = rec.get(pub_fid) if pub_fid else None
+        url = rec.get(url_fid) if url_fid else None
+        note_url = coerce_url(url)
+        note_id = extract_content_id(note_url, "xhs")
+
+        # URL 中有稳定 note ID 时，即使标题暂缺也能精确采集。
+        if require_all:
+            name_or_id = is_nonempty(name) or is_nonempty(creator_id_val)
+            title_or_id = is_nonempty(title) or bool(note_id)
+            if not (name_or_id and title_or_id and is_nonempty(pub) and is_nonempty(note_url)):
+                skipped_incomplete += 1
+                continue
+
+        if not (is_nonempty(name) or is_nonempty(creator_id_val)):
+            continue  # 兜底: 达人名称和 ID 都没就跳过
+
+        existing = rec.get(read7_fid) if read7_fid else None
+        pub_dt = parse_publish_time(pub)
+        tasks.append({
+            "record_id": rid,
+            "creator_name": name if isinstance(name, str) else (str(name) if name else ""),
+            "creator_id": str(creator_id_val) if creator_id_val else None,
+            "target_title": title if isinstance(title, str) else (str(title) if title else None),
+            "publish_time_raw": pub,
+            "publish_time": pub_dt,
+            "note_url": note_url,
+            "note_id": note_id,
+            # 发布链接仅作为“已发布”的触发标记；实际采集按达人和标题匹配。
+            "input_error": None,
+            "existing_read_7d": existing,
+        })
+
+    log(f"    平台过滤: 跳过 {skipped_platform} 条非 {platform_filter} 记录")
+    log(f"    完整性过滤: 跳过 {skipped_incomplete} 条不完整记录(4 字段有空)")
+    log(f"    实际待抓: {len(tasks)} 条")
+    style["_extract_stats"] = {
+        "source_records": len(all_records),
+        "selected_scope_records": selected_scope_records,
+        "skipped_platform": skipped_platform,
+        "skipped_incomplete": skipped_incomplete,
+    }
+    return tasks
+
+
+def group_tasks_by_creator(tasks: list[dict]) -> list[dict]:
+    return _group_tasks_by_creator(tasks)
+
+
+def collect_tasks_across_styles(styles: list[dict],
+                                only_record_ids: set[str] | None) -> tuple[list[dict], dict[int, dict]]:
+    """Read every selected style first and attach exact write-back context."""
+    all_tasks: list[dict] = []
+    summaries: dict[int, dict] = {}
+    for style in styles:
+        tasks = extract_target_tasks(style, only_record_ids)
+        summary = {
+            "style": style["name"], "index": style["index"], "total": len(tasks),
+            "run_started_at": time.time(),
+            "matched": 0, "filled": 0, "skipped_no_pubtime": 0, "results": [],
+            **style.get("_extract_stats", {}),
+        }
+        if style.get("_extract_error"):
+            summary["error"] = style["_extract_error"]
+        summaries[style["index"]] = summary
+        context = {
+            "style": style["name"], "index": style["index"],
+            "base_token": style["base_token"], "table_id": style["table_id"],
+            "field_map": style["field_map"],
+        }
+        for task in tasks:
+            task["style_context"] = context
+            if task.get("input_error"):
+                upsert_result(summary, {
+                    "record_id": task["record_id"],
+                    "creator": task.get("creator_name"),
+                    "target": task.get("target_title"),
+                    "status": BLOCKED_INPUT,
+                    "matched": False,
+                    "reason": task["input_error"],
+                    "note_url": task.get("note_url"),
+                })
+            else:
+                all_tasks.append(task)
+    return all_tasks, summaries
+
+
+def apply_matched_task(task: dict, matched: dict, summary: dict,
+                       dry_run: bool) -> dict:
+    """Apply one matched note using the task's own style write-back context."""
+    context = task["style_context"]
+    summary["matched"] += 1
+    log(f"    阅读={matched.get('read_count')}, 点赞={matched.get('like_count')}, 收藏={matched.get('collect_count')}")
+    pick = pick_read_field(context["field_map"], task.get("publish_time"))
+    read_val = parse_int(matched.get("read_count"))
+    if read_val is None:
+        return {
+            "status": WRITE_FAILURE,
+            "matched": True,
+            "write_ok": False,
+            "reason": "metric_parse_failed",
+        }
+    if dry_run:
+        return {"status": SUCCESS, "matched": True, "write_ok": None, "dry_run": True}
+    if pick and pick != "FUTURE" and not dry_run:
+        logical, fill_fid, fill_fname = pick
+        ok = write_back(
+            context["base_token"], context["table_id"],
+            task["record_id"], fill_fid, read_val,
+        )
+        if ok:
+            log(f"    回填 {fill_fname}={read_val} (key={logical}) 到 record={task['record_id']}")
+            summary["filled"] += 1
+            return {"status": SUCCESS, "matched": True, "write_ok": True}
+        return {
+            "status": WRITE_FAILURE,
+            "matched": True,
+            "write_ok": False,
+            "reason": "write_back_failed",
+        }
+    elif pick == "FUTURE":
+        log(f"    [SKIP] record={task['record_id']} 发布时间在未来,跳过回填")
+        return {
+            "status": BLOCKED_INPUT,
+            "matched": True,
+            "write_ok": None,
+            "reason": "publish_time_in_future",
+        }
+    elif not dry_run and not task.get("publish_time"):
+        log(f"    [SKIP] record={task['record_id']} 发布时间缺失,无法判断槽位")
+        summary["skipped_no_pubtime"] += 1
+        return {
+            "status": BLOCKED_INPUT,
+            "matched": True,
+            "write_ok": None,
+            "reason": "publish_time_missing",
+        }
+    return {
+        "status": WRITE_FAILURE,
+        "matched": True,
+        "write_ok": False,
+        "reason": "write_field_missing",
+    }
+
+
+def scrape_styles(styles: list[dict], login_timeout: int, headless: bool,
+                  only_record_ids: set[str] | None, dry_run: bool,
+                  batch_size: int = DEFAULT_BATCH_SIZE,
+                  batch_wait: int = DEFAULT_BATCH_WAIT,
+                  checkpoint_callback=None) -> list[dict]:
+    """Collect all styles, then search every creator only once in this run."""
+    tasks, summaries = collect_tasks_across_styles(styles, only_record_ids)
+    if not tasks:
+        log("  所有款式均无待抓任务,跳过")
+        return [summaries[s["index"]] for s in styles]
+
+    creator_groups = group_tasks_by_creator(tasks)
+    log(f"  全部 {len(styles)} 个款式、{len(tasks)} 条任务 -> {len(creator_groups)} 个唯一达人")
+
+    batches = chunk_creator_groups(creator_groups, batch_size)
+    if len(batches) > 1:
+        log(f"  [BATCH] 分 {len(batches)} 批处理,每批最多 {batch_size} 个达人,批间等待 {batch_wait}s")
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+
+    for batch_idx, batch in enumerate(batches):
+        if batch_idx > 0:
+            log(f"  [BATCH WAIT] 等待 {batch_wait // 60} 分钟后开始第 {batch_idx+1}/{len(batches)} 批...")
+            remaining_creators = sum(len(b) for b in batches[batch_idx:])
+            log(f"  [BATCH WAIT] 还剩 {remaining_creators} 个达人待处理")
+            wait_with_heartbeat(batch_wait)
+
+        log(f"  [BATCH {batch_idx+1}/{len(batches)}] 开始处理 {len(batch)} 个达人")
+
+        def action(page: Page) -> None:
+            ensure_login(page, login_timeout)
+            for ci, creator_group in enumerate(batch, start=1):
+                creator_name = creator_group["creator_name"]
+                creator_id = creator_group["creator_id"]
+                group_tasks = creator_group["tasks"]
+                if all(
+                    any(
+                        row.get("record_id") == task["record_id"]
+                        for row in summaries[task["style_context"]["index"]].get("results", [])
+                        if isinstance(row, dict)
+                    )
+                    for task in group_tasks
+                ):
+                    continue
+                log(f"  [达人 {ci}/{len(batch)}] {creator_name} ({len(group_tasks)} 条笔记)")
+
+                detail_page = None
+                try:
+                    # Navigate to home (between creator groups)
+                    main_page = page
+                    main_page.bring_to_front()
+                    for p in list(main_page.context.pages):
+                        if p is not main_page:
+                            try: p.close()
+                            except: pass
+                    try:
+                        main_page.goto(HOME_URL, wait_until="domcontentloaded", timeout=30000)
+                        main_page.wait_for_timeout(800)
+                    except Exception as exc:
+                        if is_browser_session_lost(exc):
+                            raise
+                        log(f"    [WARN] 重置首页失败: {exc}")
+
+                    # Close popups
+                    for text in ("跳过", "不再提示", "我知道了"):
+                        try:
+                            b = main_page.get_by_text(text, exact=True).first
+                            if b.count(): b.click(timeout=1500); main_page.wait_for_timeout(300)
+                        except: pass
+
+                    # Search creator
+                    search_blogger(main_page, creator_name, creator_id)
+                    detail_page = open_blogger_detail(main_page, creator_name, creator_id)
+                    if detail_page is None and creator_id:
+                        log(f"    [FALLBACK] creator_id 搜不到,改用 name='{creator_name}' 重试")
+                        search_blogger(main_page, creator_name, None)
+                        detail_page = open_blogger_detail(main_page, creator_name, None)
+
+                    # A transient popup/router failure must not fail every note
+                    # belonging to this creator after a single attempt.
+                    if detail_page is None:
+                        for retry_no in range(1, 3):
+                            log(f"    [RETRY] 第 {retry_no}/2 次重新按昵称打开达人详情")
+                            search_blogger(main_page, creator_name, None)
+                            detail_page = open_blogger_detail(main_page, creator_name, None)
+                            if detail_page is not None:
+                                break
+
+                    if detail_page is None:
+                        log(f"    [SKIP] 未找到博主 '{creator_name}'")
+                        for task in group_tasks:
+                            summary = summaries[task["style_context"]["index"]]
+                            upsert_result(summary, {
+                                "record_id": task["record_id"],
+                                "creator": creator_name,
+                                "target": task["target_title"],
+                                "status": RETRYABLE_FAILURE,
+                                "matched": False,
+                                "reason": "no_detail_page",
+                            })
+                        continue
+
+                    # Find ALL notes for this creator in one pass
+                    found_notes, candidates, page_limit_hit = find_notes_for_tasks(
+                        detail_page, group_tasks, max_pages=25,
+                    )
+
+                    for task in group_tasks:
+                        context = task["style_context"]
+                        summary = summaries[context["index"]]
+                        target_title = task.get("target_title", "")
+                        matched = found_notes.get(str(task["record_id"]))
+
+                        if matched:
+                            outcome = apply_matched_task(task, matched, summary, dry_run)
+                            upsert_result(summary, {
+                                "record_id": task["record_id"],
+                                "creator": creator_name,
+                                "target": target_title,
+                                "note_url": task.get("note_url"),
+                                "note_id": task.get("note_id"),
+                                "matched_title": matched["title"],
+                                "read_count": matched.get("read_count"),
+                                "like_count": matched.get("like_count"),
+                                "collect_count": matched.get("collect_count"),
+                                "page": matched.get("page"),
+                                "fallback": matched.get("fallback", False),
+                                "match_method": matched.get("match_method"),
+                                "match_score": matched.get("match_score"),
+                                **outcome,
+                            })
+                        else:
+                            log(f"    [未命中] 未找到笔记 '{target_title[:30] if target_title else ''}'")
+                            upsert_result(summary, {
+                                "record_id": task["record_id"],
+                                "creator": creator_name,
+                                "target": target_title,
+                                "note_url": task.get("note_url"),
+                                "note_id": task.get("note_id"),
+                                "status": RETRYABLE_FAILURE,
+                                "matched": False,
+                                "reason": "page_limit_reached" if page_limit_hit else "title_unmatched",
+                                "candidates": candidates,
+                            })
+
+                except Exception as exc:
+                    if is_browser_session_lost(exc):
+                        log(f"    [SESSION LOST] {exc}")
+                        raise
+                    log(f"    [ERROR] {exc}")
+                    for task in group_tasks:
+                        summary = summaries[task["style_context"]["index"]]
+                        existing_rids = {r.get("record_id") for r in summary["results"]}
+                        if task["record_id"] not in existing_rids:
+                            upsert_result(summary, {
+                                "record_id": task["record_id"],
+                                "creator": creator_name,
+                                "target": task.get("target_title"),
+                                "status": RETRYABLE_FAILURE,
+                                "matched": False,
+                                "reason": "browser_or_scrape_error",
+                                "error": str(exc),
+                            })
+                finally:
+                    try: save_state(detail_page)
+                    except: pass
+                    try:
+                        if detail_page is not None and detail_page is not page and not detail_page.is_closed():
+                            detail_page.close()
+                    except Exception:
+                        pass
+                    if checkpoint_callback:
+                        checkpoint_callback([
+                            finalize_summary(summaries[s["index"]], dry_run=dry_run)
+                            for s in styles
+                        ])
+
+        def pending_groups() -> list[dict]:
+            pending = []
+            for group in batch:
+                if any(
+                    not any(
+                        row.get("record_id") == task["record_id"]
+                        for row in summaries[task["style_context"]["index"]].get("results", [])
+                        if isinstance(row, dict)
+                    )
+                    for task in group["tasks"]
+                ):
+                    pending.append(group)
+            return pending
+
+        for session_attempt in range(1, 4):
+            if not pending_groups():
+                break
+            session = None
+            try:
+                with DynamicSession(
+                    headless=headless,
+                    real_chrome=True,
+                    user_data_dir=str(PROFILE_DIR),
+                    locale="zh-CN",
+                    timezone_id="Asia/Shanghai",
+                    timeout=60000,
+                    network_idle=False,
+                    disable_resources=False,
+                    google_search=False,
+                    page_setup=restore_cookies,
+                    max_pages=3,
+                ) as session:
+                    session.fetch(HOME_URL, page_action=action, wait=1000)
+            except Exception as exc:
+                if not is_browser_session_lost(exc):
+                    raise
+                log(f"  [WARN] 浏览器会话丢失,准备重建 ({session_attempt}/3): {exc}")
+            finally:
+                _force_cleanup_session(session)
+            if pending_groups() and session_attempt < 3:
+                log(f"  [RECOVER] 新建浏览器继续剩余 {len(pending_groups())} 个达人")
+
+        remaining = pending_groups()
+        if remaining:
+            log(f"  [ERROR] 浏览器重建 3 次后仍有 {len(remaining)} 个达人未处理")
+            for group in remaining:
+                for task in group["tasks"]:
+                    summary = summaries[task["style_context"]["index"]]
+                    upsert_result(summary, {
+                        "record_id": task["record_id"],
+                        "creator": group.get("creator_name"),
+                        "target": task.get("target_title"),
+                        "status": RETRYABLE_FAILURE,
+                        "matched": False,
+                        "reason": "browser_session_lost",
+                    })
+            if checkpoint_callback:
+                checkpoint_callback([
+                    finalize_summary(summaries[s["index"]], dry_run=dry_run)
+                    for s in styles
+                ])
+
+    return [finalize_summary(summaries[s["index"]], dry_run=dry_run) for s in styles]
+
+
+def scrape_one_style(style: dict, login_timeout: int, headless: bool,
+                     only_record_ids: set[str] | None, dry_run: bool,
+                     batch_size: int = DEFAULT_BATCH_SIZE,
+                     batch_wait: int = DEFAULT_BATCH_WAIT) -> dict:
+    """Backward-compatible single-style entry point."""
+    return scrape_styles(
+        [style], login_timeout, headless, only_record_ids, dry_run,
+        batch_size=batch_size, batch_wait=batch_wait,
+    )[0]
+
+
+def _force_cleanup_session(session) -> None:
+    """强制关闭 playwright 会话并清理可能泄漏的 asyncio loop。
+
+    scrapling/playwright 清理时序 bug 抛 TargetClosedError 时, dispatcher greenlet
+    仍挂起在 loop.run_forever 里, 线程本地的 running loop 标记未复位; 下一个
+    sync_playwright().start() 会报 "Sync API inside the asyncio loop"。
+    这里兜底: 尽力关资源, 换一个新 event loop, 并清掉 running loop 线程本地标记。
+    """
+    if session is not None:
+        for attr in ("context", "browser"):
+            try:
+                obj = getattr(session, attr, None)
+                if obj:
+                    obj.close()
+            except Exception:
+                pass
+        try:
+            pw = getattr(session, "playwright", None)
+            if pw:
+                pw.stop()
+        except Exception:
+            pass
+    try:
+        loop = asyncio.get_event_loop()
+        if loop.is_running():
+            for t in list(asyncio.all_tasks(loop)):
+                if not t.done():
+                    t.cancel()
+        elif not loop.is_closed():
+            loop.close()
+        asyncio.set_event_loop(asyncio.new_event_loop())
+    except Exception:
+        pass
+    # 清掉僵尸 running loop 的线程本地标记(get_running_loop 读的是它,
+    # set_event_loop 管不到), 否则下一个 sync_playwright() 仍起不来
+    try:
+        asyncio.get_running_loop()
+    except RuntimeError:
+        pass
+    else:
+        try:
+            asyncio.events._set_running_loop(None)  # 无公开 API, 只能访问私有函数
+        except Exception:
+            pass
+
+
+def parse_int(s) -> int | None:
+    """'1.2万' / '6,592' / '2422' -> 数字"""
+    if s is None:
+        return None
+    if isinstance(s, (int, float)):
+        return int(s)
+    txt = str(s).strip()
+    if not txt:
+        return None
+    m = re.match(r'^([\d,.]+)\s*(万|w|W)?', txt)
+    if not m:
+        return None
+    num = float(m.group(1).replace(',', ''))
+    if m.group(2):
+        num *= 10000
+    return int(num)
+
+
+def login_only(login_timeout: int) -> int:
+    """只触发登录(headed),登录完成立刻退出。给 weekly_relogin.py 用。"""
+    def action(page):
+        ensure_login(page, login_timeout)
+        return page
+
+    session = None
+    try:
+        with DynamicSession(
+            headless=False,
+            real_chrome=True,
+            user_data_dir=str(PROFILE_DIR),
+            locale="zh-CN",
+            timezone_id="Asia/Shanghai",
+            timeout=60000,
+            network_idle=False,
+            disable_resources=False,
+            google_search=False,
+            page_setup=restore_cookies,
+            max_pages=2,
+        ) as session:
+            session.fetch(HOME_URL, page_action=action, wait=1000)
+    except Exception as exc:
+        if "TargetClosedError" in type(exc).__name__ or "has been closed" in str(exc):
+            log(f"  [WARN] 浏览器清理时报错: {type(exc).__name__}")
+        else:
+            raise
+    finally:
+        _force_cleanup_session(session)
+    log("[login-only] 登录完成,退出")
+    return 0
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="小红书蒲公英 V2: 多维表格驱动精准采集+回填")
+    add_repeatable_style_argument(parser, "只跑指定款式编号(可多次)")
+    parser.add_argument("--record", action="append", help="只跑指定 record_id,可多次传")
+    parser.add_argument("--headless", action="store_true")
+    parser.add_argument("--login-timeout", type=int, default=300)
+    parser.add_argument("--dry-run", action="store_true", help="不写回多维表格")
+    parser.add_argument(
+        "--data-dir", default=str(PATHS.normalized_root), help="结果输出目录"
+    )
+    parser.add_argument("--skip-done", action="store_true", help="跳过已完成的款式(从进度文件读)")
+    parser.add_argument("--inter-style-pause", type=int, default=5,
+                        help="兼容旧参数；全局按达人采集后不再按款式暂停")
+    parser.add_argument("--login-only", action="store_true",
+                        help="只触发登录(headed),登录完成立刻退出,给 weekly_relogin 用")
+    parser.add_argument("--self-operated", action="store_true",
+                        help="抓取自营达人表格(而非合作达人)")
+    parser.add_argument("--batch-size", type=int, default=DEFAULT_BATCH_SIZE,
+                        help="每个批次处理的达人数(默认 0=不分批)")
+    parser.add_argument("--batch-wait", type=int, default=DEFAULT_BATCH_WAIT,
+                        help="批次间等待秒数(默认 0)")
+    args = parser.parse_args()
+
+    if args.login_only:
+        return login_only(args.login_timeout)
+
+    mapping = load_mapping(self_operated=args.self_operated)
+    styles = mapping["tables"]
+    if args.style:
+        try:
+            styles = select_requested_styles(styles, args.style)
+        except ValueError as exc:
+            print(str(exc))
+            return 1
+
+    only_rids = set(args.record) if args.record else None
+
+    data_dir = resolve_layer_output(
+        args.data_dir, layer_root=PATHS.normalized_root, field="--data-dir"
+    )
+    out_dir = V2_DIR if data_dir == PATHS.normalized_root.resolve() else data_dir / "v2_results"
+    out_dir.mkdir(parents=True, exist_ok=True)
+    self_sfx = "_self" if args.self_operated else ""
+    progress_path = (
+        PATHS.state_root / "checkpoints" / "pgy" / f"_progress{self_sfx}.json"
+    )
+    progress_path.parent.mkdir(parents=True, exist_ok=True)
+    final_path = out_dir / f"_all_summaries{self_sfx}.json"
+
+    # 读已完成的款式(中断恢复用)
+    done = set()
+    if args.skip_done and progress_path.exists():
+        try:
+            progress = json.loads(progress_path.read_text(encoding="utf-8"))
+            done = set(progress.get("done_indices", []))
+            log(f"恢复进度: 已完成 {len(done)} 个款式 {sorted(done)}")
+        except Exception:
+            pass
+
+    summaries_by_index: dict[int, dict] = {}
+    pending_styles = []
+    for s in styles:
+        if s["index"] in done:
+            # 加载之前的结果
+            prev = out_dir / f"{s['index']:02d}-{s['name']}{self_sfx}_v2.json"
+            if prev.exists():
+                try:
+                    summary = json.loads(prev.read_text(encoding="utf-8"))
+                    summaries_by_index[s["index"]] = summary
+                    log(f"\n[SKIP] 款式 [{s['index']:02d}] {s['name']} 已完成,加载上次结果")
+                    continue
+                except Exception:
+                    pass
+        pending_styles.append(s)
+
+    def persist_summaries(current_summaries: list[dict]) -> None:
+        """Atomically checkpoint selected styles and merge partial record retries."""
+        updates: list[dict] = []
+        for style, raw_summary in zip(pending_styles, current_summaries):
+            summary = finalize_summary(raw_summary, dry_run=args.dry_run)
+            out_file = out_dir / f"{style['index']:02d}-{style['name']}{self_sfx}_v2.json"
+            if only_rids and out_file.exists():
+                try:
+                    previous = json.loads(out_file.read_text(encoding="utf-8"))
+                except Exception:
+                    previous = None
+                summary = merge_style_summary(previous, summary, dry_run=args.dry_run)
+            summaries_by_index[style["index"]] = summary
+            updates.append(summary)
+            atomic_write_json(out_file, summary)
+            if summary.get("complete"):
+                done.add(style["index"])
+            else:
+                done.discard(style["index"])
+
+        atomic_write_json(progress_path, {
+            "done_indices": sorted(done),
+            "updated_at": datetime.now().isoformat(),
+        })
+        try:
+            existing_global = json.loads(final_path.read_text(encoding="utf-8")) if final_path.exists() else []
+        except Exception:
+            existing_global = []
+        atomic_write_json(final_path, merge_global_summaries(existing_global, updates))
+
+    if pending_styles:
+        log(f"\n========== 全局采集 {len(pending_styles)} 个款式 ==========")
+        new_summaries = scrape_styles(
+            pending_styles, args.login_timeout, args.headless, only_rids, args.dry_run,
+            batch_size=args.batch_size, batch_wait=args.batch_wait,
+            checkpoint_callback=persist_summaries,
+        )
+        persist_summaries(new_summaries)
+        for s in pending_styles:
+            summary = summaries_by_index[s["index"]]
+            out_file = out_dir / f"{s['index']:02d}-{s['name']}{self_sfx}_v2.json"
+            log(f"  [{s['index']:02d} {s['name']}] 落盘: {out_file}")
+            log(
+                f"  统计: 总任务={summary['total']} 命中={summary['matched']} "
+                f"回填={summary['filled']} 未解决={summary.get('unresolved', 0)}"
+            )
+
+    all_summaries = [
+        finalize_summary(summaries_by_index[s["index"]], dry_run=args.dry_run)
+        for s in styles if s["index"] in summaries_by_index
+    ]
+
+    # 输出总报告
+    log(f"\n========== 24 款式全量汇总 ==========")
+    total_tasks = sum(s["total"] for s in all_summaries)
+    total_matched = sum(s["matched"] for s in all_summaries)
+    total_filled = sum(s["filled"] for s in all_summaries)
+    log(f"  款式数: {len(all_summaries)}")
+    log(f"  任务总数: {total_tasks}")
+    log(f"  命中: {total_matched}  ({total_matched/max(total_tasks,1)*100:.1f}%)")
+    log(f"  回填: {total_filled}")
+    log(f"\n每款统计:")
+    for s in all_summaries:
+        rate = s['matched']/max(s['total'],1)*100
+        name = s.get('name') or s.get('style', '?')
+        log(f"  [{s['index']:02d}] {name:>16s}  任务={s['total']:>3d}  命中={s['matched']:>3d}  回填={s['filled']:>3d}  命中率={rate:.0f}%")
+    log(f"\n汇总: {final_path}")
+    incomplete = [s["index"] for s in all_summaries if not s.get("complete")]
+    if incomplete:
+        log(f"[ERROR] 以下款式仍有未解决采集/写回结果: {incomplete}")
+        return 2
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/requirements.txt b/src/gyxx_flow/modules/content_marketing/runtime/requirements.txt
new file mode 100644
index 0000000..0e810a6
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/requirements.txt
@@ -0,0 +1,4 @@
+scrapling[fetchers]>=0.4.8
+playwright>=1.59.0
+psycopg[binary]>=3.1
+python-dotenv>=1.0
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/run_all.py b/src/gyxx_flow/modules/content_marketing/runtime/run_all.py
new file mode 100644
index 0000000..61ffae6
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/run_all.py
@@ -0,0 +1,622 @@
+#!/usr/bin/env python
+"""
+一键跑 3 个平台 (B 站 + 小红书蒲公英 + 星图) 的 V2 抓取,真并行。
+
+并行原理:
+- bilibili_scraper.py: 纯 HTTP (requests.Session),独立进程无冲突
+- pgy_xhs_scraper_v2.py: Scrapling 启独立 Chrome (profile=.pgy_chrome_profile)
+- xingtu_scraper_v2.py: Scrapling 启独立 Chrome (profile=.xingtu_chrome_profile)
+
+三个进程 stdin/stdout 各自走独立文件,互不干扰。
+汇总报告 = 3 个 _all_summaries*.json 聚合。
+v2_results/*.json 由 daily_run.bat step4 的 sync_metrics_to_cmt_notes.py 同步到 PG。
+
+用法:
+    python run_all.py                     # 只跑合作达人(实写)
+    python run_all.py --dry-run           # 只抓不写
+    python run_all.py --skip pgy          # 跳过小红书
+    python run_all.py --skip bili,xt      # 跳过 B 站和星图
+    python run_all.py --style 7           # 只跑款式 7
+    python run_all.py --style 7,8,15      # 跑多款
+    python run_all.py --include-self-operated   # 合作+自营都跑
+    python run_all.py --self-operated-only      # 只跑自营达人
+"""
+import argparse
+import json
+import os
+import subprocess
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+BASE_DIR = PATHS.module_root
+V2_DIR = PATHS.normalized_root / "v2_results"
+RUN_REPORT_DIR = PATHS.curated_root / "run_all"
+LOG_DIR = PATHS.logs_root / "run_all"
+LOG_DIR.mkdir(parents=True, exist_ok=True)
+
+# 三个脚本定义
+SCRIPTS = {
+    "bili": {
+        "name": "B 站",
+        "script": "bilibili_scraper.py",
+        "summary_file": "_all_summaries_bilibili.json",
+        "result_prefix": "_bilibili_v2",
+    },
+    "pgy": {
+        "name": "小红书蒲公英",
+        "script": "pgy_xhs_scraper_v2.py",
+        "summary_file": "_all_summaries.json",
+        "result_prefix": "_v2",
+    },
+    "xt": {
+        "name": "星图",
+        "script": "xingtu_scraper_v2.py",
+        "summary_file": "_all_summaries_xingtu.json",
+        "result_prefix": "_xingtu_v2",
+    },
+}
+
+# 自营达人脚本: 抖音用 self_douyin_scraper.py, B站/小红书复用现有脚本加 --self-operated
+SELF_SCRIPTS = {
+    "bili": {
+        "name": "B 站(自营)",
+        "script": "bilibili_scraper.py",
+        "summary_file": "_all_summaries_self_bilibili.json",
+        "extra_args": ["--self-operated"],
+    },
+    "pgy": {
+        "name": "小红书(自营)",
+        "script": "pgy_xhs_scraper_v2.py",
+        "summary_file": "_all_summaries_self.json",
+        "extra_args": ["--self-operated"],
+    },
+    "xt": {
+        "name": "抖音(自营)",
+        "script": "self_douyin_scraper.py",
+        "summary_file": "_all_summaries_self_xingtu.json",
+        "extra_args": [],
+    },
+}
+
+
+def log(msg: str) -> None:
+    ts = datetime.now().strftime("%H:%M:%S")
+    safe = msg.encode("gbk", errors="replace").decode("gbk", errors="replace")
+    print(f"[{ts}] [run_all] {safe}", flush=True)
+
+
+def build_cmd(key: str, args: argparse.Namespace) -> list[str]:
+    """构造单个脚本的命令行"""
+    info = SCRIPTS[key]
+    cmd = [sys.executable, str(BASE_DIR / info["script"])]
+    if args.dry_run:
+        cmd.append("--dry-run")
+    if args.style:
+        # 多个款式循环传
+        pass  # 脚本自身支持 --style N
+    return cmd
+
+
+def start_process(key: str, args: argparse.Namespace,
+                  self_operated: bool = False) -> tuple[subprocess.Popen, Path, Path]:
+    """启动一个抓取进程,返回 (process, stdout_log_path, stderr_log_path)"""
+    info = SELF_SCRIPTS[key] if self_operated else SCRIPTS[key]
+    cmd = [sys.executable, str(BASE_DIR / info["script"])]
+    if args.dry_run:
+        cmd.append("--dry-run")
+    if args.style:
+        for s in args.style:
+            cmd += ["--style", str(s)]
+    cmd.extend(info.get("extra_args", []))
+
+    stdout_path = LOG_DIR / f"{key}_{datetime.now().strftime('%Y%m%d_%H%M%S')}.stdout.log"
+    stderr_path = LOG_DIR / f"{key}_{datetime.now().strftime('%Y%m%d_%H%M%S')}.stderr.log"
+    stdout_f = open(stdout_path, "w", encoding="utf-8")
+    stderr_f = open(stderr_path, "w", encoding="utf-8")
+
+    # Windows 用 CREATE_NEW_PROCESS_GROUP 方便 kill
+    creationflags = 0
+    if sys.platform == "win32":
+        creationflags = subprocess.CREATE_NEW_PROCESS_GROUP
+
+    env = {
+        **os.environ,
+        "PYTHONIOENCODING": "utf-8",
+        "PYTHONUNBUFFERED": "1",
+    }
+
+    log(f"启动 [{key}] ({info['name']}): {info['script']}  日志: {stdout_path.name}")
+    p = subprocess.Popen(
+        cmd,
+        cwd=str(BASE_DIR),
+        stdout=stdout_f,
+        stderr=stderr_f,
+        env=env,
+        creationflags=creationflags,
+    )
+    return p, stdout_path, stderr_path
+
+
+def kill_all(procs: dict) -> None:
+    """Ctrl+C 或异常时,杀掉所有子进程"""
+    for key, p in procs.items():
+        if p.poll() is None:
+            log(f"  kill [{key}] (pid={p.pid})")
+            try:
+                if sys.platform == "win32":
+                    p.terminate()  # Windows: CTRL_BREAK_EVENT
+                else:
+                    p.terminate()
+            except Exception as exc:
+                log(f"  kill [{key}] 失败: {exc}")
+
+
+def wait_all(procs: dict) -> dict[str, int]:
+    """等待所有进程结束,返回 {key: returncode}"""
+    rc = {}
+    while procs:
+        for key in list(procs.keys()):
+            p = procs[key]
+            ret = p.poll()
+            if ret is not None:
+                rc[key] = ret
+                log(f"  [{key}] 完成, exit={ret}")
+                del procs[key]
+        if procs:
+            time.sleep(2)
+    return rc
+
+
+def _summary_path(key: str, self_operated: bool = False) -> Path:
+    info = SCRIPTS[key]
+    name = info["summary_file"]
+    if self_operated:
+        name = name.replace("_all_summaries", "_all_summaries_self", 1)
+    return V2_DIR / name
+
+
+def _item_total(item: dict) -> int | None:
+    raw = item.get("total")
+    if raw is None:
+        raw = item.get("total_b_records")
+    if raw is None:
+        raw = item.get("total_tasks")
+    try:
+        value = int(raw or 0)
+    except (TypeError, ValueError):
+        return None
+    return value if value >= 0 else None
+
+
+def _item_rows(item: dict) -> list[dict]:
+    raw = item.get("results")
+    if raw is None:
+        raw = item.get("details")
+    if not isinstance(raw, list):
+        return []
+    return [row for row in raw if isinstance(row, dict)]
+
+
+def _row_is_unresolved(row: dict) -> bool:
+    status = row.get("status")
+    if status == "blocked_input":
+        return False
+    explicit_failure = bool(
+        row.get("matched") is False
+        or row.get("ok") is False
+        or row.get("write_ok") is False
+        or row.get("error")
+        or row.get("reason")
+    )
+    if status == "success":
+        return explicit_failure
+    if status:
+        return True
+    return explicit_failure
+
+
+def _item_unresolved_count(item: dict) -> int:
+    total = _item_total(item) or 0
+    rows = _item_rows(item)
+    record_ids = [str(row["record_id"]) for row in rows if row.get("record_id")]
+    missing = max(0, total - len(rows))
+    missing_ids = max(0, len(rows) - len(record_ids))
+    duplicates = max(0, len(record_ids) - len(set(record_ids)))
+    row_failures = sum(1 for row in rows if _row_is_unresolved(row))
+    derived = missing + missing_ids + duplicates + row_failures
+    declared_parts = 0
+    for key in (
+        "retryable_failures", "write_failures", "missing_results",
+        "duplicate_results",
+    ):
+        try:
+            declared_parts += max(0, int(item.get(key) or 0))
+        except (TypeError, ValueError):
+            declared_parts += 1
+    try:
+        declared_total = max(0, int(item.get("unresolved") or 0))
+    except (TypeError, ValueError):
+        declared_total = 1
+    declared = max(declared_parts, declared_total)
+    if item.get("complete") is False:
+        declared = max(declared, 1)
+    return max(derived, declared)
+
+
+def validate_summary_payload(payload, expected_indices: set[int], path: Path,
+                             started_at: float) -> tuple[bool, str]:
+    """验证汇总确实属于本轮且完整覆盖所选款式。"""
+    try:
+        if not path.exists():
+            return False, "summary file missing"
+        if path.stat().st_mtime < started_at:
+            return False, "summary file is stale"
+    except OSError as exc:
+        return False, f"summary stat failed: {exc}"
+    if not isinstance(payload, list):
+        return False, "summary payload is not a list"
+
+    indexed_items: list[tuple[int, dict]] = []
+    for item in payload:
+        if not isinstance(item, dict):
+            return False, "summary contains non-object item"
+        try:
+            idx = int(item.get("index"))
+        except (TypeError, ValueError):
+            return False, "summary item has invalid index"
+        indexed_items.append((idx, item))
+
+    indices = [idx for idx, _item in indexed_items]
+    if len(indices) != len(set(indices)):
+        return False, "summary has duplicate style index"
+    actual_indices = set(indices)
+    missing_indices = set(expected_indices) - actual_indices
+    if missing_indices:
+        return False, (
+            f"style coverage mismatch expected={sorted(expected_indices)} "
+            f"actual={sorted(actual_indices)}"
+        )
+
+    # 采集器会按款式合并全局汇总；单款运行后文件中可能保留其他款式的
+    # 历史结果。这里只校验本轮明确选择的款式，完整运行时 expected_indices
+    # 本身就是全部当前款式，仍会严格校验全量。
+    for idx, item in indexed_items:
+        if idx not in expected_indices:
+            continue
+        if item.get("error"):
+            return False, f"style {idx} reports error: {item.get('error')}"
+        if started_at > 0:
+            try:
+                item_started_at = float(item.get("run_started_at"))
+            except (TypeError, ValueError):
+                return False, f"style {idx} has no current run marker"
+            if item_started_at < started_at:
+                return False, f"style {idx} was not generated in current run"
+        total = _item_total(item)
+        if total is None:
+            return False, f"style {idx} has invalid total"
+        rows = _item_rows(item)
+        record_ids = [str(row["record_id"]) for row in rows if row.get("record_id")]
+        if len(rows) != total:
+            return False, f"style {idx} result coverage {len(rows)}/{total}"
+        if len(record_ids) != len(rows):
+            return False, f"style {idx} has result without record_id"
+        if len(set(record_ids)) != len(record_ids):
+            return False, f"style {idx} has duplicate record_id"
+        unresolved = _item_unresolved_count(item)
+        if unresolved:
+            return False, f"style {idx} has {unresolved} unresolved result(s)"
+    return True, ""
+
+
+def load_summary(key: str, self_operated: bool = False, *,
+                 started_at: float | None = None,
+                 expected_indices: set[int] | None = None) -> dict | list | None:
+    """加载并可选地验证单个平台本轮汇总 JSON。"""
+    p = _summary_path(key, self_operated)
+    if not p.exists():
+        return None
+    try:
+        with open(p, "r", encoding="utf-8") as f:
+            payload = json.load(f)
+    except Exception as exc:
+        return {"error": f"load summary failed: {exc}"}
+    if started_at is not None and expected_indices is not None:
+        valid, reason = validate_summary_payload(payload, expected_indices, p, started_at)
+        if not valid:
+            return {"error": f"invalid summary: {reason}", "payload": payload}
+    return payload
+
+
+def aggregate(summaries: dict[str, dict]) -> dict:
+    """聚合三个平台的汇总"""
+    out = {
+        "generated_at": datetime.now().isoformat(timespec="seconds"),
+        "platforms": {},
+        "totals": {
+            "styles": 0, "tasks": 0, "updated": 0, "skipped": 0,
+            "unresolved": 0, "errors": 0, "hit_rate_pct": 0.0,
+        },
+    }
+    for key, s in summaries.items():
+        if s is None:
+            out["platforms"][key] = {"name": SCRIPTS[key]["name"], "loaded": False}
+            continue
+        validation_error = None
+        if isinstance(s, dict) and "error" in s:
+            validation_error = s["error"]
+            if isinstance(s.get("payload"), list):
+                s = s["payload"]
+            else:
+                out["platforms"][key] = {
+                    "name": SCRIPTS[key]["name"], "error": validation_error
+                }
+                continue
+        # 三个脚本的 summary 结构略有不同,做归一化
+        # pgy: [{style, results, total, matched, filled, ...}]
+        # xt:  同 pgy
+        # bili: [{style, total_b_records, updated, skipped, details}]
+        platform = {
+            "name": SCRIPTS[key]["name"],
+            "loaded": True,
+            "styles": len(s) if isinstance(s, list) else 0,
+            "tasks": 0,
+            "updated": 0,
+            "skipped": 0,
+            "unresolved": 0,
+            "errors": 0,
+            "hit_rate_pct": 0.0,
+        }
+        if validation_error:
+            platform["validation_error"] = validation_error
+        if isinstance(s, list):
+            for item in s:
+                t = _item_total(item) or 0
+                # 命中:pgy/xt 用 "matched",bili 用 "updated"
+                u = (item.get("matched") if item.get("matched") is not None
+                     else item.get("updated") if item.get("updated") is not None
+                     else item.get("hits") if item.get("hits") is not None
+                     else 0)
+                # 跳过:只有 bili 用 "skipped",pgy/xt 没用
+                sk = (item.get("skipped") if item.get("skipped") is not None
+                      else item.get("misses") if item.get("misses") is not None
+                      else 0)
+                declared_errors = (item.get("error_count") if item.get("error_count") is not None
+                                   else item.get("errors") if item.get("errors") is not None
+                                   else 0)
+                try:
+                    declared_errors = int(declared_errors or 0)
+                except (TypeError, ValueError):
+                    declared_errors = 1
+                unresolved = _item_unresolved_count(item)
+                err = max(declared_errors, unresolved)
+                platform["tasks"] += t
+                platform["updated"] += u
+                platform["skipped"] += sk
+                platform["unresolved"] += unresolved
+                platform["errors"] += err
+        if platform["tasks"] > 0:
+            platform["hit_rate_pct"] = round(100.0 * platform["updated"] / platform["tasks"], 1)
+        out["platforms"][key] = platform
+        out["totals"]["styles"] += platform["styles"]
+        out["totals"]["tasks"] += platform["tasks"]
+        out["totals"]["updated"] += platform["updated"]
+        out["totals"]["skipped"] += platform["skipped"]
+        out["totals"]["unresolved"] += platform["unresolved"]
+        out["totals"]["errors"] += platform["errors"]
+    if out["totals"]["tasks"] > 0:
+        out["totals"]["hit_rate_pct"] = round(
+            100.0 * out["totals"]["updated"] / out["totals"]["tasks"], 1
+        )
+    return out
+
+
+def print_report(agg: dict, dur_sec: float) -> None:
+    """打印汇总报告"""
+    print()
+    print("=" * 70)
+    print(f"  V2 全平台汇总报告  (总耗时 {dur_sec:.1f}s)")
+    print("=" * 70)
+    print(f"  {'平台':<14} {'款式':>5} {'任务':>6} {'命中':>6} {'跳过':>6} {'未完成':>6} {'错误':>5} {'命中率':>7}")
+    print("  " + "-" * 72)
+    for key, p in agg["platforms"].items():
+        if p.get("loaded"):
+            print(f"  {p['name']:<14} {p['styles']:>5} {p['tasks']:>6} "
+                  f"{p['updated']:>6} {p['skipped']:>6} {p['unresolved']:>6} {p['errors']:>5} "
+                  f"{p['hit_rate_pct']:>6.1f}%")
+            if p.get("validation_error"):
+                print(f"    [不完整] {p['validation_error']}")
+        else:
+            err = p.get('error', 'no summary file')
+            print(f"  {SCRIPTS[key]['name']:<14}  -- ({err})")
+    t = agg["totals"]
+    print("  " + "-" * 72)
+    print(f"  {'合计':<14} {t['styles']:>5} {t['tasks']:>6} "
+          f"{t['updated']:>6} {t['skipped']:>6} {t['unresolved']:>6} {t['errors']:>5} "
+          f"{t['hit_rate_pct']:>6.1f}%")
+    print("=" * 70)
+    print(f"  详细日志: {LOG_DIR}")
+    print(f"  详细结果: {V2_DIR}/<款式>{['_v2', '_xingtu_v2', '_bilibili_v2'][0]}.json 等")
+    print()
+
+
+def _expected_style_indices(args: argparse.Namespace,
+                            self_operated: bool = False) -> set[int]:
+    if args.style:
+        return {int(idx) for idx in args.style}
+    try:
+        from gyxx_flow.modules.content_marketing.runtime import feishu_mapping
+        mapping = feishu_mapping.load_mapping(self_operated=self_operated)
+        return {
+            int(style["index"])
+            for style in mapping.get("tables", [])
+            if isinstance(style, dict) and style.get("index")
+        }
+    except Exception as exc:
+        log(f"[ERROR] 无法读取期望款式列表: {exc}")
+        return set()
+
+
+def _atomic_write_json(path: Path, payload) -> None:
+    tmp = path.with_name(path.name + ".tmp")
+    with open(tmp, "w", encoding="utf-8") as f:
+        json.dump(payload, f, ensure_ascii=False, indent=2)
+        f.flush()
+        os.fsync(f.fileno())
+    os.replace(tmp, path)
+
+
+def run_round(args: argparse.Namespace, active: list[str],
+              self_operated: bool = False) -> dict:
+    """跑一轮(合作达人或自营达人),返回聚合报告 dict。"""
+    label = "自营达人" if self_operated else "合作达人"
+    log(f"===== 开始 {label} 抓取 =====")
+
+    print("=" * 70)
+    print(f"  V2 三平台并行抓取 — {label} ({len(active)}/{len(SCRIPTS)} 个平台)")
+    print(f"  模式: {'DRY-RUN (不写)' if args.dry_run else '正式跑 (写回多维表格)'}")
+    if args.style:
+        print(f"  款式: {','.join(str(s) for s in args.style)}")
+    else:
+        try:
+            from gyxx_flow.modules.content_marketing.runtime import feishu_mapping as _fm
+            n = len(_fm.load_mapping(self_operated=self_operated).get("tables", []))
+            print(f"  款式: 全部 {n} 款")
+        except Exception:
+            print(f"  款式: 全部")
+    print(f"  平台: {','.join(SCRIPTS[k]['name'] for k in active)}")
+    print("=" * 70)
+
+    expected_indices = _expected_style_indices(args, self_operated)
+    if not expected_indices:
+        reason = "expected style mapping is empty"
+        log(f"[ERROR] {reason}")
+        return {"error": reason, "_non_zero_platforms": list(active)}
+
+    procs = {}
+    log_files = {}
+    start = time.time()
+    try:
+        for key in active:
+            p, out_path, err_path = start_process(key, args, self_operated=self_operated)
+            procs[key] = p
+            log_files[key] = (out_path, err_path)
+    except Exception as exc:
+        log(f"启动失败: {exc}")
+        kill_all(procs)
+        return {"error": str(exc), "_non_zero_platforms": list(active)}
+
+    try:
+        rc = wait_all(procs)
+    except KeyboardInterrupt:
+        log("Ctrl+C - 终止所有子进程...")
+        kill_all(procs)
+        return {"error": "interrupted", "_non_zero_platforms": list(active)}
+
+    dur = time.time() - start
+    log(f"{label} 完成,耗时 {dur:.1f}s  exit codes: {rc}")
+
+    summaries = {
+        k: load_summary(
+            k, self_operated=self_operated,
+            started_at=start, expected_indices=expected_indices,
+        )
+        for k in active
+    }
+    agg = aggregate(summaries)
+    print_report(agg, dur)
+
+    report_suffix = "_self" if self_operated else ""
+    report_path = RUN_REPORT_DIR / f"_run_all_report{report_suffix}.json"
+    report_path.parent.mkdir(parents=True, exist_ok=True)
+    agg["duration_sec"] = round(dur, 1)
+    agg["exit_codes"] = rc
+    agg["self_operated"] = self_operated
+    agg["log_files"] = {
+        k: {"stdout": str(v[0].name), "stderr": str(v[1].name)}
+        for k, v in log_files.items()
+    }
+    non_zero = {k for k, v in rc.items() if v != 0}
+    non_zero.update(
+        k for k, summary in summaries.items()
+        if summary is None or (isinstance(summary, dict) and summary.get("error"))
+    )
+    agg["_non_zero_platforms"] = sorted(non_zero)
+    _atomic_write_json(report_path, agg)
+    log(f"{label} 汇总报告: {report_path}")
+    return agg
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="一键跑 3 个平台 V2 抓取(真并行,带汇总报告)"
+    )
+    parser.add_argument("--dry-run", action="store_true", help="三个脚本都只抓不写")
+    parser.add_argument("--style", type=int, action="append",
+                        help="只跑指定款式编号(可多次)。不传 = 全跑")
+    parser.add_argument("--skip", type=str, default="",
+                        help="跳过平台,逗号分隔。可选: bili,pgy,xt。例如 --skip xt,xt")
+    parser.add_argument("--include-self-operated", action="store_true",
+                        help="同时跑自营达人(默认只跑合作达人)")
+    parser.add_argument("--self-operated-only", action="store_true",
+                        help="只跑自营达人")
+    args = parser.parse_args()
+
+    skip_set = set(s.strip() for s in args.skip.split(",") if s.strip())
+    for s in skip_set:
+        if s not in SCRIPTS:
+            print(f"[ERROR] --skip 未知平台: {s}  (可选: {','.join(SCRIPTS.keys())})")
+            return 1
+
+    active = [k for k in SCRIPTS.keys() if k not in skip_set]
+    if not active:
+        print("[ERROR] --skip 全部平台,没东西可跑")
+        return 1
+
+    # 启动并行子进程前,先刷新款式对照表,
+    # 避免 3 个子进程各自刷新时重复打 lark-cli 并抢写缓存文件。
+    run_self = args.include_self_operated or args.self_operated_only
+    try:
+        from gyxx_flow.modules.content_marketing.runtime import feishu_mapping
+        m = feishu_mapping.refresh_mapping(force=True)
+        if m:
+            log(f"合作达人款式对照表已动态刷新: {len(m.get('tables', []))} 个款式")
+    except Exception as exc:
+        log(f"[WARN] 动态刷新合作达人款式对照表失败(将退回缓存): {exc}")
+    if run_self:
+        try:
+            from gyxx_flow.modules.content_marketing.runtime import feishu_mapping
+            m = feishu_mapping.refresh_mapping(force=True, url_field=feishu_mapping.INDEX_SELF_URL_FIELD)
+            if m:
+                log(f"自营达人款式对照表已动态刷新: {len(m.get('tables', []))} 个款式")
+        except Exception as exc:
+            log(f"[WARN] 动态刷新自营达人款式对照表失败(将退回缓存): {exc}")
+
+    any_error = False
+
+    # 第一轮: 合作达人(除非 --self-operated-only)
+    if not args.self_operated_only:
+        agg = run_round(args, active, self_operated=False)
+        if agg.get("_non_zero_platforms"):
+            log(f"[WARN] 合作达人: 以下平台 exit != 0: {agg['_non_zero_platforms']}")
+            any_error = True
+
+    # 第二轮: 自营达人(仅 --include-self-operated 或 --self-operated-only 时跑)
+    if args.include_self_operated or args.self_operated_only:
+        agg = run_round(args, active, self_operated=True)
+        if agg.get("_non_zero_platforms"):
+            log(f"[WARN] 自营达人: 以下平台 exit != 0: {agg['_non_zero_platforms']}")
+            any_error = True
+
+    return 1 if any_error else 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/runtime_paths.py b/src/gyxx_flow/modules/content_marketing/runtime/runtime_paths.py
new file mode 100644
index 0000000..67fc97f
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/runtime_paths.py
@@ -0,0 +1,87 @@
+"""Portable path boundary for the migrated content-marketing runtime.
+
+Code and immutable resources live below ``GYXX_MODULE_ROOT``.  All mutable
+state is kept below ``GYXX_DATA_ROOT`` so a deployment can move either tree
+without editing business scripts.
+"""
+
+from __future__ import annotations
+
+import os
+from dataclasses import dataclass
+from pathlib import Path
+
+from gyxx_flow.core import DataLayout, Settings
+
+
+def _path_from_env(name: str, default: Path) -> Path:
+    value = os.environ.get(name, "").strip()
+    return Path(value).expanduser().resolve() if value else default.resolve()
+
+
+module_root = _path_from_env("GYXX_MODULE_ROOT", Path(__file__).resolve().parent)
+data_root = _path_from_env("GYXX_DATA_ROOT", Settings.from_env().data_root)
+_DATA_PATHS = DataLayout(data_root).for_module("content_marketing")
+
+state_root = _DATA_PATHS.state_root
+raw_root = _DATA_PATHS.raw_root
+normalized_root = _DATA_PATHS.normalized_root
+curated_root = _DATA_PATHS.curated_root
+exports_root = _DATA_PATHS.exports_root
+evidence_root = _DATA_PATHS.evidence_root
+logs_root = _DATA_PATHS.logs_root
+tmp_root = _DATA_PATHS.tmp_root
+
+
+@dataclass(frozen=True)
+class RuntimePaths:
+    module_root: Path = module_root
+    data_root: Path = data_root
+    state_root: Path = state_root
+    raw_root: Path = raw_root
+    normalized_root: Path = normalized_root
+    curated_root: Path = curated_root
+    exports_root: Path = exports_root
+    evidence_root: Path = evidence_root
+    logs_root: Path = logs_root
+    tmp_root: Path = tmp_root
+    config_root: Path = module_root / "data" / "config"
+    tools_root: Path = module_root / "data" / "tools"
+    browser_profile_dir: Path = _path_from_env(
+        "GYXX_BROWSER_PROFILE_DIR", state_root / "browser-profiles" / "default"
+    )
+    browser_cookie_file: Path = _path_from_env(
+        "GYXX_BROWSER_COOKIE_FILE", state_root / "cookies" / "default.json"
+    )
+    browser_storage_state_file: Path = _path_from_env(
+        "GYXX_BROWSER_STORAGE_STATE_FILE",
+        state_root / "cookies" / "default_storage_state.json",
+    )
+
+
+PATHS = RuntimePaths()
+
+
+def resolve_layer_output(
+    configured: str | Path,
+    *,
+    layer_root: Path,
+    field: str = "output path",
+) -> Path:
+    """Resolve a caller-supplied output without allowing a data-layer escape."""
+
+    if not isinstance(configured, (str, Path)) or not str(configured).strip():
+        raise ValueError(f"{field} must be a non-empty path")
+    root = Path(layer_root).expanduser().resolve()
+    requested = Path(configured).expanduser()
+    candidate = (
+        requested.resolve()
+        if requested.is_absolute()
+        else (root / requested).resolve()
+    )
+    if not candidate.is_relative_to(root):
+        raise ValueError(f"{field} must stay inside {root}")
+    return candidate
+
+
+__all__ = ["PATHS", "RuntimePaths", "resolve_layer_output"]
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/self_bilibili_scraper.py b/src/gyxx_flow/modules/content_marketing/runtime/self_bilibili_scraper.py
new file mode 100644
index 0000000..5bb0959
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/self_bilibili_scraper.py
@@ -0,0 +1,455 @@
+"""
+自营 B 站采集: B 站公开 API (api.bilibili.com/x/web-interface/view) 抓播放量
+============================================================================
+- 不用登录,不用 cookie
+- 动态读取自营 mapping (feishu_mapping.load_mapping self_operated=True)
+- 遍历每个有 base_token+table_id 的款式表
+- 过滤 platform=B站 的记录,必须有 note_url
+- 从 note_url 抽 BV/AV 号,调 API 拿 stat.view + data.pubdate
+- 飞书无发布时间 → 用 API pubdate (毫秒时间戳) 填进去,再按时槽回填播放量
+- 时槽规则跟 chanmama_scraper.py / bilibili_scraper.py 一致:
+    0-7 天  → 7天曝光量
+    8-14 天 → 14天曝光量
+    15-21 天 → 21天曝光量
+    22-28 天 → 28天曝光量
+    29+ 天  → 月底曝光量
+
+用法:
+  python self_bilibili_scraper.py                 # 全跑
+  python self_bilibili_scraper.py --dry-run       # 只抓不写
+  python self_bilibili_scraper.py --style 1       # 只跑某款
+  python self_bilibili_scraper.py --delay 0.5    # 请求间隔
+"""
+import argparse
+import io
+import json
+import os
+import re
+import subprocess
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+# Windows 控制台输出用 UTF-8,避免 emoji 编码失败
+sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8", errors="replace")
+sys.stderr = io.TextIOWrapper(sys.stderr.buffer, encoding="utf-8", errors="replace")
+
+import requests
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+
+LARK_CLI = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK_CLI):
+    LARK_CLI = "lark-cli.cmd"
+
+_UA = ("Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
+       "(KHTML, like Gecko) Chrome/126.0.0.0 Safari/537.36")
+_BV_RE = re.compile(r"BV[0-9A-Za-z]{10}")
+_AV_RE = re.compile(r"av(\d+)", re.IGNORECASE)
+_B23_RE = re.compile(r"https?://b23\.tv/\S+", re.IGNORECASE)
+_MD_LINK_RE = re.compile(r"\[([^\]]+)\]\(([^)]+)\)")
+_URL_RE = re.compile(r"https?://[^\s)>\]]+")
+
+
+def log(msg: str) -> None:
+    safe = str(msg).encode("gbk", errors="replace").decode("gbk", errors="replace")
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {safe}", flush=True)
+
+
+# ============================================================
+# 飞书多维表格
+# ============================================================
+def call_lark_json(args: list[str]) -> dict:
+    cmd = [LARK_CLI] + args
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, env=env,
+                          encoding="utf-8", errors="replace")
+    text = proc.stdout if proc.stdout else (proc.stderr or "")
+    start = text.find("{")
+    if start < 0:
+        return {"ok": False, "error": f"no json: {text[:300]}"}
+    try:
+        return json.loads(text[start:], strict=False)
+    except json.JSONDecodeError as e:
+        return {"ok": False, "error": f"json decode: {e}"}
+
+
+def list_records_by_table(base_token: str, table_id: str) -> list[dict]:
+    all_records = []
+    offset = 0
+    while True:
+        resp = call_lark_json([
+            "base", "+record-list",
+            "--base-token", base_token,
+            "--table-id", table_id,
+            "--as", "user",
+            "--format", "json",
+            "--limit", "200",
+            "--offset", str(offset),
+        ])
+        if not resp.get("ok"):
+            raise RuntimeError(f"record-list 失败: {resp.get('error')}")
+        d = resp["data"]
+        ids = d["field_id_list"]
+        rids = d["record_id_list"]
+        rows = d["data"]
+        for i, row in enumerate(rows):
+            rec = {"record_id": rids[i]}
+            for j, fid in enumerate(ids):
+                rec[fid] = row[j] if j < len(row) else None
+            all_records.append(rec)
+        offset += len(rows)
+        if not d.get("has_more", False) or len(rows) == 0:
+            break
+    return all_records
+
+
+def write_back(base_token: str, table_id: str, record_id: str,
+               field_id: str, value) -> bool:
+    payload = {field_id: value}
+    resp = call_lark_json([
+        "base", "+record-upsert",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--record-id", record_id,
+        "--as", "user",
+        "--format", "json",
+        "--json", json.dumps(payload, ensure_ascii=False),
+    ])
+    if not resp.get("ok"):
+        log(f"  ❌ 回填失败 record_id={record_id}: {resp.get('error')}")
+        return False
+    return True
+
+
+# ============================================================
+# URL / B 站 API
+# ============================================================
+def extract_url(text) -> str | None:
+    """从 note_url 字段抽真实 URL,处理 markdown / 纯文本 / b23 短链"""
+    if text is None:
+        return None
+    if isinstance(text, list):
+        text = text[0] if text else None
+        if not text:
+            return None
+    s = str(text).strip()
+    m = _MD_LINK_RE.search(s)
+    if m:
+        return m.group(2).strip()
+    m = _URL_RE.search(s)
+    if m:
+        return m.group(0).strip()
+    return None
+
+
+def resolve_b23(url: str, session: requests.Session) -> str | None:
+    try:
+        r = session.get(url, allow_redirects=True, timeout=10,
+                        headers={"User-Agent": _UA})
+        return r.url
+    except Exception as exc:
+        log(f"  [WARN] b23 解析失败: {exc}")
+        return None
+
+
+def fetch_bili_stat(url: str, session: requests.Session) -> tuple[int | None, int | None]:
+    """调 B 站 API,返回 (play_count, pubdate_unix_seconds)。
+    pubdate 是视频发布时间(Unix 秒),用于补飞书的发布时间字段。
+    """
+    real = url
+    if _B23_RE.match(url):
+        resolved = resolve_b23(url, session)
+        if not resolved:
+            return None, None
+        real = resolved
+    bv = _BV_RE.search(real)
+    av = _AV_RE.search(real)
+    if not bv and not av:
+        return None, None
+    params = {"bvid": bv.group(0)} if bv else {"aid": av.group(1)}
+    try:
+        r = session.get("https://api.bilibili.com/x/web-interface/view",
+                        params=params, timeout=10,
+                        headers={"User-Agent": _UA,
+                                 "Referer": "https://www.bilibili.com/"})
+        j = r.json()
+        if j.get("code") != 0:
+            return None, None
+        data = j.get("data", {})
+        stat = data.get("stat", {})
+        view = stat.get("view")
+        pubdate = data.get("pubdate")
+        return (int(view) if view is not None else None,
+                int(pubdate) if pubdate else None)
+    except Exception as exc:
+        log(f"  [WARN] B 站 API: {exc}")
+        return None, None
+
+
+# ============================================================
+# 时间 / 槽位
+# ============================================================
+def parse_publish_time(value) -> datetime | None:
+    """飞书发布时间字段可能是毫秒时间戳 / 字符串 / datetime。"""
+    if value is None or value == "":
+        return None
+    if isinstance(value, datetime):
+        return value
+    if isinstance(value, (int, float)):
+        try:
+            if value > 1e12:
+                return datetime.fromtimestamp(value / 1000)
+            return datetime.fromtimestamp(value)
+        except (ValueError, OSError):
+            return None
+    if isinstance(value, str):
+        v = value.strip()
+        if not v:
+            return None
+        if v.isdigit():
+            try:
+                iv = int(v)
+                if iv > 1e12:
+                    return datetime.fromtimestamp(iv / 1000)
+                return datetime.fromtimestamp(iv)
+            except (ValueError, OSError):
+                return None
+        for fmt in ("%Y-%m-%dT%H:%M:%S", "%Y-%m-%d %H:%M:%S",
+                    "%Y-%m-%d", "%Y/%m/%d"):
+            try:
+                return datetime.strptime(v[:len(fmt) + 2], fmt)
+            except ValueError:
+                continue
+    return None
+
+
+def pick_read_field(fmap: dict, pub_time: datetime | None,
+                    today: datetime | None = None) -> tuple[str, str, str] | None:
+    if pub_time is None:
+        return None
+    today = today or datetime.now()
+    days = (today.date() - pub_time.date()).days
+    if days < 0:
+        return "FUTURE"
+    if days <= 7:
+        logical = "read_count_7d"
+    elif days <= 14:
+        logical = "read_count_14d"
+    elif days <= 21:
+        logical = "read_count_21d"
+    elif days <= 28:
+        logical = "read_count_28d"
+    else:
+        logical = "month_end"
+    info = fmap.get(logical)
+    if info:
+        return (logical, info["field_id"], info["field_name"])
+    return None
+
+
+# ============================================================
+# 采集一个款式
+# ============================================================
+def process_style(style: dict, dry_run: bool, delay: float,
+                  session: requests.Session) -> dict:
+    base_token = style["base_token"]
+    table_id = style["table_id"]
+    fmap = style.get("field_map", {})
+
+    platform_fid = (fmap.get("platform") or {}).get("field_id")
+    url_fid = (fmap.get("note_url") or {}).get("field_id")
+    title_fid = (fmap.get("note_title") or {}).get("field_id")
+    pub_fid = (fmap.get("publish_time") or {}).get("field_id")
+
+    summary = {
+        "style": style["name"], "index": style["index"],
+        "total": 0, "matched": 0, "filled": 0,
+        "skipped_no_url": 0, "skipped_no_pubtime": 0,
+        "details": [],
+    }
+
+    if not platform_fid or not url_fid or not title_fid:
+        log(f"  [WARN] platform/note_url/note_title 字段缺失,跳过")
+        return summary
+
+    try:
+        all_records = list_records_by_table(base_token, table_id)
+    except Exception as exc:
+        log(f"  [ERR] 拉记录失败: {exc}")
+        return summary
+
+    def is_bili(v) -> bool:
+        if not v:
+            return False
+        if isinstance(v, list):
+            vals = [str(x).lower() for x in v if x]
+            if not vals:
+                return False
+            return any("b站" in s or "b 站" in s or "bilibili" in s for s in vals)
+        s = str(v).lower()
+        return "b站" in s or "b 站" in s or "bilibili" in s
+
+    bili_records = [r for r in all_records if is_bili(r.get(platform_fid))]
+    log(f"  全表 {len(all_records)} 条,B 站记录 {len(bili_records)} 条")
+
+    today = datetime.now()
+
+    for rec in bili_records:
+        title = rec.get(title_fid)
+        if isinstance(title, list):
+            title = " ".join(str(x) for x in title if x)
+        title = str(title or "").strip()
+        if not title:
+            continue
+        summary["total"] += 1
+
+        url = extract_url(rec.get(url_fid))
+        if not url:
+            log(f"    [SKIP] 无 URL: '{title[:20]}'")
+            summary["skipped_no_url"] += 1
+            continue
+
+        # 抓 B 站 API:播放量 + pubdate
+        play, pubdate_unix = fetch_bili_stat(url, session)
+        if play is None:
+            log(f"    [SKIP] API 抓不到: '{title[:20]}' (url={url[:50]})")
+            continue
+        summary["matched"] += 1
+
+        # 飞书发布时间
+        pub_time = parse_publish_time(rec.get(pub_fid))
+
+        # 飞书无发布时间 → 用 API pubdate 补,再做时槽判断
+        if pub_time is None and pubdate_unix is not None:
+            api_pub = datetime.fromtimestamp(pubdate_unix)
+            pub_str = api_pub.strftime("%Y-%m-%d %H:%M:%S")
+            if dry_run:
+                log(f"    [DRY] '{title[:20]}' → 先填发布时间={pub_str}")
+            else:
+                ts_ms = int(api_pub.timestamp() * 1000)
+                ok = write_back(base_token, table_id, rec["record_id"],
+                                pub_fid, ts_ms)
+                if ok:
+                    log(f"    ✅ '{title[:20]}' → 先填发布时间={pub_str}")
+            pub_time = api_pub
+
+        if pub_time is None:
+            log(f"    [SKIP] 飞书+API 都无发布时间: '{title[:20]}'")
+            summary["skipped_no_pubtime"] += 1
+            continue
+
+        pick = pick_read_field(fmap, pub_time, today)
+        if pick == "FUTURE":
+            log(f"    [SKIP] 发布时间在未来: '{title[:20]}'")
+            continue
+        if not pick:
+            log(f"    [SKIP] 无对应曝光槽位字段: '{title[:20]}'")
+            continue
+        logical, fill_fid, fill_fname = pick
+
+        if dry_run:
+            log(f"    [DRY] '{title[:20]}' → {fill_fname}={play}")
+            summary["filled"] += 1
+            summary["details"].append({
+                "record_id": rec["record_id"],
+                "title": title[:50],
+                "play_count": play,
+                "slot": logical,
+                "pub_time": pub_time.isoformat() if pub_time else None,
+                "matched": True,
+            })
+            if delay > 0:
+                time.sleep(delay)
+            continue
+
+        ok = write_back(base_token, table_id, rec["record_id"], fill_fid, play)
+        if ok:
+            summary["filled"] += 1
+            summary["details"].append({
+                "record_id": rec["record_id"],
+                "title": title[:50],
+                "play_count": play,
+                "slot": logical,
+                "pub_time": pub_time.isoformat() if pub_time else None,
+                "matched": True,
+                "ok": True,
+            })
+            log(f"    ✅ '{title[:20]}' → {fill_fname}={play}")
+        if delay > 0:
+            time.sleep(delay)
+
+    return summary
+
+
+# ============================================================
+# main
+# ============================================================
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="自营 B 站采集: B 站公开 API 抓播放量,飞书无发布时间用 API pubdate 补")
+    parser.add_argument("--style", type=int, help="只跑指定款式编号")
+    parser.add_argument("--dry-run", action="store_true", help="不写回飞书")
+    parser.add_argument("--delay", type=float, default=0.3,
+                        help="每条请求间隔秒数(防风控)")
+    args = parser.parse_args()
+
+    from gyxx_flow.modules.content_marketing.runtime import feishu_mapping
+    try:
+        mapping = feishu_mapping.load_mapping(self_operated=True)
+    except Exception as exc:
+        log(f"❌ 加载自营 mapping 失败: {exc}")
+        return 1
+
+    styles = mapping.get("tables", [])
+    if args.style:
+        styles = [s for s in styles if s.get("index") == args.style]
+        if not styles:
+            log(f"⚠️ 未找到款式 {args.style}")
+            return 1
+
+    out_dir = PATHS.normalized_root / "v2_results"
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    session = requests.Session()
+    session.headers.update({"User-Agent": _UA})
+
+    print("=" * 50)
+    print(f"🔄 自营 B 站采集 (dry_run={args.dry_run})")
+    print(f"  款式数: {len(styles)}")
+    print("=" * 50)
+
+    all_summaries = []
+    for s in styles:
+        log(f"\n--- [{s['index']:02d}] {s['name']} ---")
+        summary = process_style(s, args.dry_run, args.delay, session)
+        all_summaries.append(summary)
+        log(f"  小计: 任务={summary['total']} 命中={summary['matched']} 回填={summary['filled']}")
+        out_file = out_dir / f"{s['index']:02d}-{s['name']}_self_bilibili_v2.json"
+        with open(out_file, "w", encoding="utf-8") as f:
+            json.dump(summary, f, ensure_ascii=False, indent=2)
+
+    final_path = out_dir / "_all_summaries_self_bilibili.json"
+    with open(final_path, "w", encoding="utf-8") as f:
+        json.dump(all_summaries, f, ensure_ascii=False, indent=2)
+
+    total = sum(s["total"] for s in all_summaries)
+    matched = sum(s["matched"] for s in all_summaries)
+    filled = sum(s["filled"] for s in all_summaries)
+    print(f"\n{'=' * 50}")
+    print(f"📊 完成: 任务={total} 命中={matched} 回填={filled}")
+    if args.dry_run:
+        print("  (dry-run,未写飞书)")
+    print(f"{'=' * 50}")
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/self_douyin_scraper.py b/src/gyxx_flow/modules/content_marketing/runtime/self_douyin_scraper.py
new file mode 100644
index 0000000..f3452d7
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/self_douyin_scraper.py
@@ -0,0 +1,692 @@
+"""
+自营抖音采集器: 登录 douyin.com, 打开自己账号作品页, 按标题匹配飞书记录, 回填曝光量
+
+与星图采集器(xingtu_scraper_v2.py)的核心区别:
+- 星图: 搜索达人 → 打开详情页 → 找视频
+- 自营抖音: 直接打开自己账号的作品列表 → 滚动加载 → 匹配标题
+
+Cookie/登录: 复用 douyin_comment_scraper 的 .douyin_chrome_profile + douyin_cookies.json
+飞书写回: 复用 lark-cli + field_map 体系, 结果文件用 _self_xingtu_v2.json 后缀
+
+用法:
+  python self_douyin_scraper.py                     # 全跑
+  python self_douyin_scraper.py --dry-run           # 只抓不写
+  python self_douyin_scraper.py --style 3           # 只跑款式3
+  python self_douyin_scraper.py --login-only        # 只触发登录
+"""
+import argparse
+import json
+import os
+import re
+import subprocess
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+from typing import Any
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+from playwright.sync_api import Page
+from scrapling.fetchers import DynamicSession
+
+from gyxx_flow.modules.content_marketing.runtime.collection_completeness import (
+    BLOCKED_INPUT,
+    RETRYABLE_FAILURE,
+    SUCCESS,
+    WRITE_FAILURE,
+    add_repeatable_style_argument,
+    atomic_write_json,
+    finalize_summary,
+    match_tasks_to_cards,
+    merge_global_summaries,
+    merge_style_summary,
+    select_requested_styles,
+)
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+STATE_DIR = PATHS.state_root
+COOKIE_FILE = PATHS.browser_cookie_file
+PROFILE_DIR = PATHS.browser_profile_dir
+
+SELF_POST_URL = "https://www.douyin.com/user/self?showTab=post"
+
+LARK_CLI = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK_CLI):
+    LARK_CLI = "lark-cli.cmd"
+
+
+def log(msg: str) -> None:
+    safe = str(msg).encode("gbk", errors="replace").decode("gbk", errors="replace")
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {safe}", flush=True)
+
+
+# ============================================================
+# 飞书多维表格
+# ============================================================
+def call_lark_json(args: list[str]) -> dict:
+    cmd = [LARK_CLI] + args
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, env=env, encoding="utf-8", errors="replace")
+    text = proc.stdout if proc.stdout else (proc.stderr or "")
+    start = text.find("{")
+    if start < 0:
+        return {"ok": False, "error": f"no json: {text[:300]}"}
+    try:
+        return json.loads(text[start:], strict=False)
+    except json.JSONDecodeError as e:
+        return {"ok": False, "error": f"json decode: {e}"}
+
+
+def load_mapping() -> dict:
+    from gyxx_flow.modules.content_marketing.runtime import feishu_mapping
+    return feishu_mapping.load_mapping(DATA_DIR, self_operated=True)
+
+
+def list_records_by_table(base_token: str, table_id: str) -> list[dict]:
+    all_records = []
+    offset = 0
+    while True:
+        resp = call_lark_json([
+            "base", "+record-list",
+            "--base-token", base_token,
+            "--table-id", table_id,
+            "--as", "user",
+            "--format", "json",
+            "--limit", "200",
+            "--offset", str(offset),
+        ])
+        if not resp.get("ok"):
+            raise RuntimeError(f"record-list 失败: {resp.get('error')}")
+        d = resp["data"]
+        ids = d["field_id_list"]
+        rids = d["record_id_list"]
+        rows = d["data"]
+        for i, row in enumerate(rows):
+            rec = {"record_id": rids[i]}
+            for j, fid in enumerate(ids):
+                rec[fid] = row[j] if j < len(row) else None
+            all_records.append(rec)
+        offset += len(rows)
+        if not d.get("has_more", False) or len(rows) == 0:
+            break
+    return all_records
+
+
+def write_back(base_token: str, table_id: str, record_id: str,
+               field_id: str, value) -> bool:
+    payload = {field_id: value}
+    resp = call_lark_json([
+        "base", "+record-upsert",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--record-id", record_id,
+        "--as", "user",
+        "--format", "json",
+        "--json", json.dumps(payload, ensure_ascii=False),
+    ])
+    if not resp.get("ok"):
+        log(f"  回填失败 record_id={record_id}: {resp.get('error')}")
+        return False
+    return True
+
+
+# ============================================================
+# 时间/槽位
+# ============================================================
+def parse_publish_time(value) -> datetime | None:
+    if value is None or value == "":
+        return None
+    if isinstance(value, (int, float)):
+        try:
+            return datetime.fromtimestamp(value / 1000)
+        except (ValueError, OSError):
+            return None
+    if isinstance(value, str):
+        v = value.strip()
+        if not v:
+            return None
+        if v.isdigit():
+            try:
+                return datetime.fromtimestamp(int(v) / 1000)
+            except (ValueError, OSError):
+                return None
+        for fmt in ("%Y-%m-%dT%H:%M:%S", "%Y-%m-%d %H:%M:%S", "%Y-%m-%d", "%Y/%m/%d"):
+            try:
+                return datetime.strptime(v[:len(fmt) + 2], fmt)
+            except ValueError:
+                continue
+    return None
+
+
+def pick_read_field(fmap: dict, pub_time: datetime | None,
+                    today: datetime | None = None) -> tuple[str, str, str] | None:
+    if pub_time is None:
+        return None
+    today = today or datetime.now()
+    days = (today.date() - pub_time.date()).days
+    if days < 0:
+        return "FUTURE"
+    if days <= 7:
+        logical = "read_count_7d"
+    elif days <= 14:
+        logical = "read_count_14d"
+    elif days <= 21:
+        logical = "read_count_21d"
+    elif days <= 28:
+        logical = "read_count_28d"
+    else:
+        logical = "month_end"
+    info = fmap.get(logical)
+    if info:
+        return (logical, info["field_id"], info["field_name"])
+    return None
+
+
+# ============================================================
+# 数字解析
+# ============================================================
+def parse_int(v) -> int | None:
+    if v is None:
+        return None
+    if isinstance(v, (int, float)):
+        return int(v)
+    txt = str(v).strip()
+    if not txt:
+        return None
+    m = re.match(r"^([\d,.]+)\s*(万|w|W)?", txt)
+    if not m:
+        return None
+    num = float(m.group(1).replace(",", ""))
+    if m.group(2):
+        num *= 10000
+    return int(num)
+
+
+# ============================================================
+# Cookie / 登录
+# ============================================================
+def load_cookies() -> list[dict[str, Any]]:
+    if not COOKIE_FILE.exists():
+        return []
+    try:
+        data = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+        return data if isinstance(data, list) else []
+    except json.JSONDecodeError:
+        return []
+
+
+def restore_cookies(page: Page) -> None:
+    cookies = load_cookies()
+    if not cookies:
+        return
+    try:
+        page.context.add_cookies(cookies)
+        log(f"Loaded cookies: {COOKIE_FILE}")
+    except Exception as exc:
+        log(f"Failed to load cookies: {exc}")
+
+
+def save_state(page: Page) -> None:
+    STATE_DIR.mkdir(parents=True, exist_ok=True)
+    COOKIE_FILE.write_text(
+        json.dumps(page.context.cookies(), ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    PATHS.browser_storage_state_file.write_text(
+        json.dumps(page.context.storage_state(), ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    log(f"Saved cookies: {COOKIE_FILE}")
+
+
+def maybe_wait_for_login(page: Page, login_timeout: int) -> None:
+    deadline = time.time() + login_timeout
+    while time.time() < deadline:
+        try:
+            body = page.locator("body").inner_text(timeout=5000)
+        except Exception:
+            body = ""
+        if "扫码登录" not in body and "登录后" not in body:
+            return
+        log("抖音需要登录,请在打开的 Chrome 窗口中完成登录")
+        page.wait_for_timeout(3000)
+    log("[WARN] 登录等待超时")
+
+
+def dismiss_popups(page: Page) -> None:
+    for text in ("我知道了", "稍后再说", "暂不登录", "取消", "跳过"):
+        try:
+            btn = page.locator(f"text={text}").first
+            if btn.is_visible(timeout=1000):
+                btn.click(timeout=1000)
+        except Exception:
+            pass
+    try:
+        page.evaluate("""
+            () => {
+              const close = [...document.querySelectorAll('button, div, span')]
+                .filter((el) => {
+                  const t = (el.innerText || '').trim();
+                  const r = el.getBoundingClientRect();
+                  return ['×', 'x', 'X'].includes(t) &&
+                    r.width > 0 && r.width <= 52 &&
+                    r.height > 0 && r.height <= 52 &&
+                    r.top >= 0 && r.top < window.innerHeight * 0.75;
+                })[0];
+              close?.click();
+            }
+        """)
+    except Exception:
+        pass
+
+
+# ============================================================
+# 标题匹配
+# ============================================================
+def title_match(haystack: str, needle: str) -> bool:
+    """模糊标题匹配:去掉空格/标点后比较,任一方包含另一方即命中"""
+    def norm(s: str) -> str:
+        return re.sub(r"[\s　\-_.,;:!?！？，。；：、/\\()（）【】《》''\"\"'']+", "", str(s))
+    nh = norm(haystack)
+    nn = norm(needle)
+    if not nh or not nn:
+        return False
+    return nn in nh or nh in nn
+
+
+# ============================================================
+# 抖音作品页抓取
+# ============================================================
+def scroll_and_collect_posts(page: Page, max_scrolls: int = 50) -> list[dict]:
+    """在 /user/self?showTab=post 页面滚动加载,收集所有作品的标题和播放量。
+    返回 [{"title": str, "play_count": int | None}, ...]
+    """
+    posts = []
+    seen_titles = set()
+    no_new_count = 0
+
+    for scroll_idx in range(max_scrolls):
+        # 解析当前页面上的作品卡片
+        items = page.evaluate(r"""
+            () => {
+                const cards = document.querySelectorAll('li[data-e2e="user-post-item"], li[class*="user-post"], div[data-e2e="user-post-item"]');
+                const results = [];
+                for (const card of cards) {
+                    // 标题: 通常在 <a> 或 <p> 里
+                    let title = '';
+                    const links = card.querySelectorAll('a');
+                    for (const a of links) {
+                        const t = (a.textContent || '').trim();
+                        if (t.length > title.length) title = t;
+                    }
+                    if (!title) {
+                        const p = card.querySelector('p, span[class*="title"], div[class*="title"]');
+                        if (p) title = (p.textContent || '').trim();
+                    }
+                    // 播放量: 通常在卡片底部,格式如 "4.5万次播放" 或 "1234"
+                    let playText = '';
+                    const spans = card.querySelectorAll('span');
+                    for (const sp of spans) {
+                        const t = (sp.textContent || '').trim();
+                        if (t.includes('播放') || t.includes('次') || /^[\d,.]+$/.test(t) ||
+                            /[\d,.]+\s*(万|w|W)/.test(t)) {
+                            if (t.length > playText.length) playText = t;
+                        }
+                    }
+                    if (!playText) {
+                        const lastSpan = card.querySelector('span:last-of-type');
+                        if (lastSpan) {
+                            const t = (lastSpan.textContent || '').trim();
+                            if (/[\d]/.test(t)) playText = t;
+                        }
+                    }
+                    if (title) {
+                        results.push({title: title, playText: playText});
+                    }
+                }
+                return results;
+            }
+        """)
+
+        new_count = 0
+        for item in items:
+            t = item.get("title", "").strip()
+            if not t or t in seen_titles:
+                continue
+            seen_titles.add(t)
+            play = parse_play_text(item.get("playText", ""))
+            posts.append({"title": t, "play_count": play})
+            new_count += 1
+
+        if new_count == 0:
+            no_new_count += 1
+            if no_new_count >= 3:
+                break
+        else:
+            no_new_count = 0
+
+        # 向下滚动
+        page.evaluate("window.scrollBy(0, window.innerHeight * 1.5)")
+        page.wait_for_timeout(800)
+
+    return posts
+
+
+def parse_play_text(text: str) -> int | None:
+    """从 '4.5万次播放' / '1234' / '1.2w' 等文本中提取播放量数字"""
+    if not text:
+        return None
+    # 先尝试提取 "X次播放" 格式
+    m = re.search(r"([\d,.]+)\s*(万|w|W)?\s*次播放", text)
+    if m:
+        return parse_int(m.group(1) + (m.group(2) or ""))
+    # 直接 parse_int
+    return parse_int(text)
+
+
+# ============================================================
+# 采集一个款式
+# ============================================================
+def scrape_one_style(style: dict, login_timeout: int, headless: bool,
+                     only_record_ids: set[str] | None, dry_run: bool) -> dict:
+    base_token = style["base_token"]
+    table_id = style["table_id"]
+    fmap = style.get("field_map", {})
+
+    summary = {"style": style["name"], "index": style["index"],
+               "run_started_at": time.time(),
+               "total": 0, "matched": 0, "filled": 0,
+               "skipped_no_pubtime": 0, "results": []}
+
+    # 从飞书读取自营达人表记录,只看抖音
+    platform_fid = fmap.get("platform", {}).get("field_id")
+    name_fid = fmap.get("creator_name", {}).get("field_id")
+    id_fid = fmap.get("creator_id", {}).get("field_id")
+    title_fid = fmap.get("note_title", {}).get("field_id")
+    url_fid = fmap.get("note_url", {}).get("field_id")
+    pub_fid = fmap.get("publish_time", {}).get("field_id")
+
+    if not platform_fid or not title_fid or not url_fid or not pub_fid:
+        summary["error"] = "missing platform/note_title/note_url/publish_time field mapping"
+        log(f"  [ERROR] platform/note_title/note_url/publish_time 字段缺失,跳过")
+        return summary
+
+    all_records = list_records_by_table(base_token, table_id)
+
+    # 只看抖音记录
+    def is_douyin(v) -> bool:
+        if not v:
+            return False
+        s = str(v).strip()
+        return s in ("抖音", "douyin")
+
+    def is_nonempty(v) -> bool:
+        if v is None:
+            return False
+        if isinstance(v, list):
+            return len(v) > 0
+        if isinstance(v, str):
+            return v.strip() != ""
+        return True
+
+    douyin_records = [r for r in all_records if is_douyin(r.get(platform_fid))]
+    log(f"  全表 {len(all_records)} 条,抖音记录 {len(douyin_records)} 条")
+
+    # 构建任务列表:需要标题+链接+发布时间
+    tasks = []
+    for rec in douyin_records:
+        rid = rec["record_id"]
+        if only_record_ids and rid not in only_record_ids:
+            continue
+        title = rec.get(title_fid)
+        if isinstance(title, list):
+            title = " ".join(str(x) for x in title if x)
+        title = str(title or "").strip()
+        if not title:
+            continue
+        url_raw = rec.get(url_fid)
+        pub_raw = rec.get(pub_fid)
+        creator = rec.get(name_fid) or rec.get(id_fid) or "?"
+        pub_time = parse_publish_time(pub_raw)
+        tasks.append({
+            "record_id": rid,
+            "creator_name": str(creator),
+            "target_title": title,
+            "note_url": str(url_raw or ""),
+            "publish_time": pub_time,
+        })
+
+    summary["total"] = len(tasks)
+    if not tasks:
+        log(f"  无抖音待抓任务,跳过")
+        return summary
+
+    # 启动浏览器:所有浏览器操作都在 page_action 里完成
+    collected_posts: list[dict] = []
+    login_ok = False
+
+    def page_action(page: Page):
+        nonlocal login_ok
+        dismiss_popups(page)
+        restore_cookies(page)
+        page.goto(SELF_POST_URL, wait_until="domcontentloaded", timeout=60000)
+        page.wait_for_timeout(3000)
+        dismiss_popups(page)
+        maybe_wait_for_login(page, login_timeout)
+        save_state(page)
+        login_ok = True
+
+        log("  等待作品列表加载...")
+        page.wait_for_timeout(3000)
+        dismiss_popups(page)
+
+        # 滚动加载所有作品
+        log("  滚动加载作品列表...")
+        posts = scroll_and_collect_posts(page, max_scrolls=50)
+        log(f"  共加载 {len(posts)} 个作品")
+        collected_posts.extend(posts)
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+    with DynamicSession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=str(PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=60000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=restore_cookies,
+        max_pages=3,
+    ) as session:
+        session.fetch("https://www.douyin.com/", page_action=page_action, wait=1000)
+
+    if not login_ok:
+        log(f"  [ERROR] 登录失败,跳过该款式")
+        return summary
+
+    # 用收集到的作品匹配飞书记录并回填
+    today = datetime.now()
+    posts = collected_posts
+    matched_posts = match_tasks_to_cards(tasks, posts, "douyin")
+    for task in tasks:
+        matched_post = matched_posts.get(str(task["record_id"]))
+
+        if matched_post is None:
+            log(f"    [MISS] {task['creator_name']}: '{task['target_title'][:20]}' 未匹配到作品")
+            summary["results"].append({
+                "record_id": task["record_id"],
+                "creator": task["creator_name"],
+                "target": task["target_title"],
+                "status": RETRYABLE_FAILURE,
+                "matched": False,
+                "reason": "title_not_found",
+            })
+            continue
+
+        summary["matched"] += 1
+        play_count = matched_post.get("play_count")
+        log(f"    {task['creator_name']}: '{task['target_title'][:20]}' → 播放={play_count}")
+
+        pick = pick_read_field(fmap, task.get("publish_time"), today)
+        outcome: dict[str, Any]
+        if play_count is None:
+            outcome = {
+                "status": RETRYABLE_FAILURE,
+                "write_ok": False,
+                "reason": "metric_parse_failed",
+            }
+        elif dry_run:
+            outcome = {"status": SUCCESS, "write_ok": None, "dry_run": True}
+        elif pick and pick != "FUTURE":
+            logical, fill_fid, fill_fname = pick
+            ok = write_back(base_token, table_id, task["record_id"], fill_fid, play_count)
+            if ok:
+                log(f"    回填 {fill_fname}={play_count} (key={logical})")
+                summary["filled"] += 1
+                outcome = {"status": SUCCESS, "write_ok": True}
+            else:
+                outcome = {
+                    "status": WRITE_FAILURE,
+                    "write_ok": False,
+                    "reason": "write_back_failed",
+                }
+        elif pick == "FUTURE":
+            log(f"    [SKIP] record={task['record_id']} 发布时间在未来,跳过回填")
+            outcome = {
+                "status": BLOCKED_INPUT,
+                "write_ok": None,
+                "reason": "publish_time_in_future",
+            }
+        elif not dry_run and not task.get("publish_time"):
+            log(f"    [SKIP] record={task['record_id']} 发布时间缺失,无法判断槽位")
+            summary["skipped_no_pubtime"] += 1
+            outcome = {
+                "status": BLOCKED_INPUT,
+                "write_ok": None,
+                "reason": "publish_time_missing",
+            }
+        else:
+            outcome = {
+                "status": WRITE_FAILURE,
+                "write_ok": False,
+                "reason": "write_field_missing",
+            }
+
+        summary["results"].append({
+            "record_id": task["record_id"],
+            "creator": task["creator_name"],
+            "target": task["target_title"],
+            "matched_title": matched_post["title"],
+            "play_count": play_count,
+            "matched": True,
+            **outcome,
+        })
+
+    return summary
+
+
+def login_only(login_timeout: int) -> int:
+    """只触发登录"""
+    login_done = False
+
+    def page_action(page: Page):
+        restore_cookies(page)
+        page.goto("https://www.douyin.com/", wait_until="domcontentloaded", timeout=60000)
+        page.wait_for_timeout(2000)
+        maybe_wait_for_login(page, login_timeout)
+        save_state(page)
+        login_done = True
+        log("登录完成")
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+    with DynamicSession(
+        headless=False,
+        real_chrome=True,
+        user_data_dir=str(PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=60000,
+        page_setup=restore_cookies,
+    ) as session:
+        session.fetch("https://www.douyin.com/", page_action=page_action, wait=1000)
+
+    return 0 if login_done else 1
+
+
+# ============================================================
+# main
+# ============================================================
+def main() -> int:
+    parser = argparse.ArgumentParser(description="自营抖音采集: douyin.com 自己账号作品曝光量")
+    add_repeatable_style_argument(parser, "只跑指定款式编号(可多次)")
+    parser.add_argument("--record", action="append", help="只跑指定 record_id")
+    parser.add_argument("--headless", action="store_true")
+    parser.add_argument("--login-timeout", type=int, default=300)
+    parser.add_argument("--dry-run", action="store_true", help="不写回多维表格")
+    parser.add_argument("--login-only", action="store_true", help="只触发登录")
+    args = parser.parse_args()
+
+    if args.login_only:
+        return login_only(args.login_timeout)
+
+    mapping = load_mapping()
+    styles = mapping["tables"]
+    if args.style:
+        try:
+            styles = select_requested_styles(styles, args.style)
+        except ValueError as exc:
+            print(str(exc))
+            return 1
+
+    only_rids = set(args.record) if args.record else None
+
+    out_dir = PATHS.normalized_root / "v2_results"
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    self_sfx = "_self"
+    all_summaries = []
+    for s in styles:
+        log(f"\n========== [{s['index']:02d}] {s['name']} (自营抖音) ==========")
+        summary = finalize_summary(
+            scrape_one_style(s, args.login_timeout, args.headless, only_rids, args.dry_run),
+            dry_run=args.dry_run,
+        )
+        out_file = out_dir / f"{s['index']:02d}-{s['name']}{self_sfx}_xingtu_v2.json"
+        if only_rids and out_file.exists():
+            try:
+                previous = json.loads(out_file.read_text(encoding="utf-8"))
+            except Exception:
+                previous = None
+            summary = merge_style_summary(previous, summary, dry_run=args.dry_run)
+        all_summaries.append(summary)
+        atomic_write_json(out_file, summary)
+        log(f"  落盘: {out_file}")
+        log(
+            f"  统计: 总任务={summary['total']} 命中={summary['matched']} "
+            f"回填={summary['filled']} 未解决={summary.get('unresolved', 0)}"
+        )
+
+    final_path = out_dir / f"_all_summaries{self_sfx}_xingtu.json"
+    try:
+        existing_global = json.loads(final_path.read_text(encoding="utf-8")) if final_path.exists() else []
+    except Exception:
+        existing_global = []
+    atomic_write_json(final_path, merge_global_summaries(existing_global, all_summaries))
+    log(f"\n=== 全部完成 ===")
+    log(f"汇总: {final_path}")
+    incomplete = [summary.get("index") for summary in all_summaries if not summary.get("complete")]
+    if incomplete:
+        log(f"[ERROR] 以下款式仍有未解决采集/写回结果: {incomplete}")
+        return 2
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/sitecustomize.py b/src/gyxx_flow/modules/content_marketing/runtime/sitecustomize.py
new file mode 100644
index 0000000..bf69a40
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/sitecustomize.py
@@ -0,0 +1,5 @@
+"""Apply the unified per-script runtime binding to nested Python launches."""
+
+from gyxx_flow.adapters.bootstrap import safe_bootstrap_current_process
+
+safe_bootstrap_current_process()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_analyze_note_no_comments.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_analyze_note_no_comments.py
new file mode 100644
index 0000000..20cc658
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_analyze_note_no_comments.py
@@ -0,0 +1,57 @@
+import sys
+import tempfile
+import unittest
+from pathlib import Path
+from unittest.mock import patch
+
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+TOOLS_DIR = PROJECT_ROOT / "data" / "tools"
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools import analyze_note  # noqa: E402
+
+
+class AnalyzeNoteWithoutCommentsTest(unittest.TestCase):
+    def test_no_comments_still_writes_metrics_report(self):
+        info = {
+            "title": "无评论测试笔记",
+            "platform": "xiaohongshu",
+            "source_url": "https://example.com/note",
+            "scraped_at": "",
+            "comment_count_total": 0,
+            "like_count": 12,
+            "favorite_count": 3,
+            "comment_count_metric": 0,
+            "share_count": 1,
+            "view_count": 1000,
+            "collect_count_yxyy": 0,
+            "style_name": "测试款",
+            "creator_name": "测试达人",
+        }
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            output = Path(tmpdir) / "note.txt"
+            argv = [
+                "analyze_note.py",
+                "--note-id",
+                "1",
+                "--no-send",
+                "--brand",
+                "光影行星",
+                "--output",
+                str(output),
+            ]
+            with patch.object(analyze_note, "load_note_from_db", return_value=(info, [])), patch.object(
+                sys, "argv", argv
+            ):
+                result = analyze_note.main()
+
+            self.assertEqual(result, 0)
+            self.assertTrue(output.exists())
+            report = output.read_text(encoding="utf-8-sig")
+            self.assertIn("无有效评论可分析", report)
+            self.assertIn("1,000", report)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_batch_rescrape_douyin.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_batch_rescrape_douyin.py
new file mode 100644
index 0000000..57be932
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_batch_rescrape_douyin.py
@@ -0,0 +1,9 @@
+from gyxx_flow.modules.content_marketing.runtime.data.tools import batch_rescrape_douyin as batch
+
+
+def test_result_logged_in_reads_nested_scraper_stats():
+    result = {"stats": {"logged_in": True}}
+
+    assert batch.result_logged_in(result) is True
+    assert batch.result_logged_in({"stats": {"logged_in": False}}) is False
+    assert batch.result_logged_in({}) is False
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_collection_completeness.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_collection_completeness.py
new file mode 100644
index 0000000..c65fbe2
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_collection_completeness.py
@@ -0,0 +1,840 @@
+import argparse
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.modules.content_marketing.runtime import collection_completeness as cc
+import bilibili_scraper as bili
+import pgy_xhs_scraper_v2 as pgy
+import run_all
+import self_douyin_scraper as self_dy
+import xingtu_scraper_v2 as xingtu
+
+
+def test_normalize_title_handles_nfkc_case_and_zero_width():
+    assert cc.normalize_title("ＭａｃＢｏｏｋ\u200b 通勤包！") == "macbook通勤包"
+
+
+def test_coerce_url_unwraps_feishu_markdown_links():
+    value = "[查看笔记](https://www.xiaohongshu.com/discovery/item/abc123?x=1)"
+    assert cc.coerce_url(value) == "https://www.xiaohongshu.com/discovery/item/abc123?x=1"
+
+
+def test_match_tasks_prefers_note_id_over_changed_title():
+    tasks = [{
+        "record_id": "r1",
+        "target_title": "飞书中的旧标题",
+        "note_url": "https://www.xiaohongshu.com/explore/abc123",
+        "note_id": "abc123",
+    }]
+    cards = [
+        {"title": "完全不同的新标题", "href": "https://www.xiaohongshu.com/explore/abc123", "note_id": "abc123"},
+        {"title": "飞书中的旧标题", "href": "https://www.xiaohongshu.com/explore/other", "note_id": "other"},
+    ]
+
+    matched = cc.match_tasks_to_cards(tasks, cards, platform="xhs")
+
+    assert matched["r1"]["note_id"] == "abc123"
+    assert matched["r1"]["match_method"] == "content_id"
+
+
+def test_match_tasks_accepts_unique_truncated_title_but_rejects_ambiguous_title():
+    task = {"record_id": "r1", "target_title": "男生长期主义通勤双肩包分享"}
+    unique = [
+        {"title": "男生长期主义通勤双肩包"},
+        {"title": "夏日轻量斜挎包"},
+    ]
+    ambiguous = [
+        {"title": "男生长期主义通勤双肩包"},
+        {"title": "男生长期主义通勤双肩包"},
+    ]
+
+    assert cc.match_tasks_to_cards([task], unique, platform="douyin")["r1"]
+    assert "r1" not in cc.match_tasks_to_cards([task], ambiguous, platform="douyin")
+
+
+def test_title_similarity_ignores_episode_prefix_and_hashtag_suffix():
+    target = "第20集：Pocket4拍旋焦，无后期也能出片？"
+    platform_title = (
+        "Pocket4拍旋焦，无后期也能出片？ #旋焦 #pocket4 #摄影装备 "
+        "无论Pocket3还是Pocket4都可以轻松拍出旋焦效果！"
+    )
+
+    assert cc.title_similarity(platform_title, target) >= 0.98
+
+
+def test_title_similarity_rejects_different_episode_numbers():
+    assert cc.title_similarity(
+        "第19集：Pocket4拍旋焦，无后期也能出片？",
+        "第20集：Pocket4拍旋焦，无后期也能出片？",
+    ) == 0.0
+
+
+def test_match_tasks_rejects_duplicate_exact_titles_with_different_ids():
+    task = {"record_id": "r1", "target_title": "完全相同的目标标题"}
+    cards = [
+        {"title": "完全相同的目标标题", "note_id": "video-1"},
+        {"title": "完全相同的目标标题", "note_id": "video-2"},
+    ]
+
+    assert "r1" not in cc.match_tasks_to_cards([task], cards, platform="douyin")
+
+
+def test_one_card_is_not_reused_for_two_different_creator_tasks():
+    tasks = [
+        {
+            "record_id": "loose",
+            "target_title": "commuter backpack review today",
+            "note_url": "https://v.douyin.com/loose-source/",
+        },
+        {
+            "record_id": "exact",
+            "target_title": "commuter backpack review",
+            "note_url": "https://v.douyin.com/exact-source/",
+        },
+    ]
+    cards = [{"title": "commuter backpack review", "note_id": "observed-1"}]
+
+    matched = cc.match_tasks_to_cards(tasks, cards, platform="douyin")
+
+    assert set(matched) == {"exact"}
+
+
+def test_same_source_url_can_share_one_card_across_styles():
+    tasks = [
+        {
+            "record_id": "style-1",
+            "target_title": "同一篇跨款式笔记",
+            "note_url": "https://v.douyin.com/same-short-link/",
+        },
+        {
+            "record_id": "style-2",
+            "target_title": "同一篇跨款式笔记",
+            "note_url": "https://v.douyin.com/same-short-link/",
+        },
+    ]
+    cards = [{"title": "同一篇跨款式笔记", "note_id": "observed-1"}]
+
+    assert set(cc.match_tasks_to_cards(tasks, cards, platform="douyin")) == {
+        "style-1", "style-2",
+    }
+
+
+def test_known_mismatched_content_ids_cannot_fall_back_to_same_title():
+    task = {
+        "record_id": "r1",
+        "target_title": "完全相同标题",
+        "note_id": "video-a",
+    }
+    card = {"title": "完全相同标题", "note_id": "video-b"}
+
+    assert cc.match_tasks_to_cards([task], [card], platform="douyin") == {}
+
+
+def test_xingtu_api_and_dom_duplicate_are_merged_before_matching():
+    cards = [
+        {
+            "title": "同一条视频",
+            "play_count": 19481,
+            "note_id": "video-1",
+            "source": "show_items_api",
+        },
+        {
+            "title": "同一条视频",
+            "play_count": "1.9w",
+            "note_id": "",
+            "page": 1,
+        },
+    ]
+
+    deduped = cc.deduplicate_observed_cards(cards)
+    matched = cc.match_tasks_to_cards(
+        [{"record_id": "r1", "target_title": "同一条视频"}],
+        deduped,
+        platform="douyin",
+    )
+
+    assert len(deduped) == 1
+    assert matched["r1"]["note_id"] == "video-1"
+
+
+def test_merge_partial_summary_keeps_existing_success_rows():
+    existing = {
+        "style": "款式A", "index": 1, "total": 2, "matched": 1, "filled": 1,
+        "results": [
+            {"record_id": "ok", "status": "success", "matched": True, "write_ok": True},
+            {"record_id": "retry", "status": "retryable_failure", "matched": False},
+        ],
+    }
+    partial = {
+        "style": "款式A", "index": 1, "total": 1, "matched": 1, "filled": 1,
+        "results": [
+            {"record_id": "retry", "status": "success", "matched": True, "write_ok": True},
+        ],
+    }
+
+    merged = cc.merge_style_summary(existing, partial)
+
+    assert merged["total"] == 2
+    assert {row["record_id"] for row in merged["results"]} == {"ok", "retry"}
+    assert merged["matched"] == 2
+    assert merged["filled"] == 2
+    assert merged["complete"] is True
+
+
+def test_summary_contract_detects_missing_or_write_failure():
+    summary = {
+        "total": 2,
+        "results": [
+            {"record_id": "ok", "status": "success", "matched": True, "write_ok": True},
+            {"record_id": "bad", "status": "write_failure", "matched": True, "write_ok": False},
+        ],
+    }
+    finalized = cc.finalize_summary(summary, dry_run=False)
+    assert finalized["complete"] is False
+    assert finalized["unresolved"] == 1
+
+    missing = cc.finalize_summary({"total": 2, "results": summary["results"][:1]}, dry_run=False)
+    assert missing["complete"] is False
+    assert missing["missing_results"] == 1
+
+
+def _style_for_platform(platform: str):
+    return {
+        "index": 1,
+        "name": "款式A",
+        "base_token": "base",
+        "table_id": "table",
+        "field_map": {
+            "creator_name": {"field_id": "name"},
+            "creator_id": {"field_id": "creator_id"},
+            "note_title": {"field_id": "title"},
+            "publish_time": {"field_id": "pub"},
+            "note_url": {"field_id": "url"},
+            "platform": {"field_id": "platform"},
+            "read_count_7d": {"field_id": "read7"},
+        },
+    }
+
+
+def test_extractors_preserve_note_url_and_content_id(monkeypatch):
+    xhs_row = {
+        "record_id": "xhs-r", "name": "达人", "creator_id": "xhs-id",
+        "title": "标题", "pub": "2026-07-20", "platform": "小红书",
+        "url": "https://www.xiaohongshu.com/explore/abc123", "read7": None,
+    }
+    dy_row = {
+        "record_id": "dy-r", "name": "达人", "creator_id": "dy-id",
+        "title": "标题", "pub": "2026-07-20", "platform": "抖音",
+        "url": "https://www.douyin.com/video/7654321", "read7": None,
+    }
+    monkeypatch.setattr(pgy, "list_records_by_table", lambda *_: [xhs_row])
+    monkeypatch.setattr(xingtu, "list_records_by_table", lambda *_: [dy_row])
+
+    xhs_task = pgy.extract_target_tasks(_style_for_platform("小红书"))[0]
+    dy_task = xingtu.extract_target_tasks(_style_for_platform("抖音"))[0]
+
+    assert xhs_task["note_url"].endswith("/abc123")
+    assert xhs_task["note_id"] == "abc123"
+    assert dy_task["note_url"].endswith("/7654321")
+    assert dy_task["note_id"] == "7654321"
+
+
+@pytest.mark.parametrize(
+    ("module", "platform", "filled_url"),
+    [
+        (pgy, "小红书", "https://www.douyin.com/video/7654321"),
+        (xingtu, "抖音", "http://8.99 分享文案 https://v.douyin.com/abc123/"),
+    ],
+)
+def test_filled_publish_link_triggers_collection_without_platform_validation(
+    monkeypatch, module, platform, filled_url,
+):
+    row = {
+        "record_id": "r1",
+        "name": "达人",
+        "creator_id": "creator-id",
+        "title": "已发布标题",
+        "pub": "2026-07-20",
+        "platform": platform,
+        "url": filled_url,
+        "read7": None,
+    }
+    monkeypatch.setattr(module, "list_records_by_table", lambda *_: [row])
+
+    task = module.extract_target_tasks(_style_for_platform(platform))[0]
+
+    assert task["input_error"] is None
+
+
+def test_extractors_can_collect_missing_title_when_direct_url_has_content_id(monkeypatch):
+    xhs_style = _style_for_platform("小红书")
+    dy_style = _style_for_platform("抖音")
+    xhs_row = {
+        "record_id": "xhs-r", "name": "达人", "creator_id": "xhs-id",
+        "title": None, "pub": "2026-07-20", "platform": "小红书",
+        "url": "https://www.xiaohongshu.com/explore/abc123", "read7": None,
+    }
+    dy_row = {
+        "record_id": "dy-r", "name": "达人", "creator_id": "dy-id",
+        "title": None, "pub": "2026-07-20", "platform": "抖音",
+        "url": "https://www.douyin.com/video/7654321", "read7": None,
+    }
+    monkeypatch.setattr(pgy, "list_records_by_table", lambda *_: [xhs_row])
+    monkeypatch.setattr(xingtu, "list_records_by_table", lambda *_: [dy_row])
+
+    assert pgy.extract_target_tasks(xhs_style)[0]["note_id"] == "abc123"
+    assert xingtu.extract_target_tasks(dy_style)[0]["note_id"] == "7654321"
+    assert xhs_style["_extract_stats"]["skipped_incomplete"] == 0
+    assert dy_style["_extract_stats"]["skipped_incomplete"] == 0
+
+
+def test_missing_critical_mapping_is_an_incomplete_summary(monkeypatch):
+    style = _style_for_platform("抖音")
+    del style["field_map"]["note_url"]
+    monkeypatch.setattr(
+        xingtu,
+        "list_records_by_table",
+        lambda *_: (_ for _ in ()).throw(AssertionError("mapping should fail first")),
+    )
+
+    _, summaries = xingtu.collect_tasks_across_styles([style], None)
+    finalized = cc.finalize_summary(summaries[1], dry_run=True)
+
+    assert "note_url" in finalized["error"]
+    assert finalized["complete"] is False
+    assert finalized["unresolved"] >= 1
+
+
+class _FakeInput:
+    def __init__(self):
+        self.filled = []
+
+    def click(self):
+        return None
+
+    def fill(self, value):
+        self.filled.append(value)
+
+    def press(self, _key):
+        return None
+
+
+class _FakeLocator:
+    def __init__(self, input_):
+        self.first = input_
+
+    def filter(self, **_kwargs):
+        return self
+
+
+class _FakeSearchPage:
+    def __init__(self):
+        self.input = _FakeInput()
+        self.has_result_calls = 0
+
+    def wait_for_timeout(self, _ms):
+        return None
+
+    def evaluate(self, script, arg=None):
+        if "const rows" in script:
+            self.has_result_calls += 1
+            return self.has_result_calls >= 2
+        if "const inputs" in script:
+            return ""
+        raise AssertionError(script)
+
+    def locator(self, _selector):
+        return _FakeLocator(self.input)
+
+    def wait_for_url(self, *_args, **_kwargs):
+        return None
+
+
+def test_pgy_search_retry_keeps_creator_id_query():
+    page = _FakeSearchPage()
+
+    pgy._ensure_search_results(page, query="xhs-id-123", result_name="达人昵称", max_retries=2)
+
+    assert page.input.filled == ["xhs-id-123"]
+
+
+def test_apply_matched_task_reports_write_failure(monkeypatch):
+    task = {
+        "record_id": "r1", "publish_time": object(),
+        "style_context": {
+            "index": 1, "base_token": "base", "table_id": "table", "field_map": {},
+        },
+    }
+    summary = {"matched": 0, "filled": 0, "skipped_no_pubtime": 0, "results": []}
+    monkeypatch.setattr(pgy, "pick_read_field", lambda *_: ("read_count_7d", "fid", "7天曝光量"))
+    monkeypatch.setattr(pgy, "write_back", lambda *_: False)
+
+    outcome = pgy.apply_matched_task(task, {"read_count": "123", "title": "标题"}, summary, dry_run=False)
+
+    assert outcome["status"] == "write_failure"
+    assert outcome["write_ok"] is False
+
+
+def test_default_creator_batch_has_no_forced_hour_pause():
+    assert pgy.DEFAULT_BATCH_SIZE == 0
+    assert xingtu.DEFAULT_BATCH_SIZE == 0
+
+
+def test_repeatable_style_argument_and_selection_keep_every_requested_style():
+    parser = argparse.ArgumentParser()
+    cc.add_repeatable_style_argument(parser, "styles")
+    args = parser.parse_args(["--style", "1", "--style", "3"])
+    styles = [{"index": 1}, {"index": 2}, {"index": 3}]
+
+    assert args.style == [1, 3]
+    assert cc.select_requested_styles(styles, args.style) == [styles[0], styles[2]]
+    with pytest.raises(ValueError, match="99"):
+        cc.select_requested_styles(styles, [1, 99])
+
+
+def test_pgy_stable_cards_retry_an_initial_empty_render(monkeypatch):
+    sequence = [[], [{"title": "标题", "read_count": "1"}], [{"title": "标题", "read_count": "1"}]]
+    monkeypatch.setattr(pgy, "parse_cards_on_page", lambda _page: sequence.pop(0))
+
+    class Page:
+        waits = 0
+
+        def wait_for_timeout(self, _ms):
+            self.waits += 1
+
+    page = Page()
+    assert pgy.parse_cards_stable(page) == [{"title": "标题", "read_count": "1"}]
+    assert page.waits == 2
+
+
+def test_xingtu_paginates_before_falling_back_to_search(monkeypatch):
+    pages = [
+        [{"title": "无关视频", "play_count": "10"}],
+        [{"title": "目标视频标题", "play_count": "20"}],
+    ]
+    calls = {"next": 0}
+    monkeypatch.setattr(xingtu, "is_blocked", lambda _page: False)
+    monkeypatch.setattr(xingtu, "parse_videos_stable", lambda _page: pages.pop(0))
+
+    def next_page(_page):
+        calls["next"] += 1
+        return True
+
+    monkeypatch.setattr(xingtu, "go_next_page", next_page)
+    monkeypatch.setattr(xingtu, "submit_video_search", lambda *_: False)
+
+    found, _, _ = xingtu.find_videos_for_tasks(
+        object(), [{"record_id": "r1", "target_title": "目标视频标题"}], max_pages=2,
+    )
+
+    assert found["r1"]["play_count"] == "20"
+    assert calls["next"] == 1
+
+
+def test_xingtu_normalizes_show_items_api_video():
+    card = xingtu.normalize_xingtu_api_item({
+        "item_id": 7654321,
+        "item_title": "目标视频标题",
+        "play": 168000,
+        "like": 1200,
+        "comment": 34,
+        "share": 56,
+        "url": "https://www.douyin.com/video/7654321",
+    })
+
+    assert card == {
+        "title": "目标视频标题",
+        "play_count": 168000,
+        "like_count": 1200,
+        "comment_count": 34,
+        "share_count": 56,
+        "href": "https://www.douyin.com/video/7654321",
+        "note_id": "7654321",
+        "source": "show_items_api",
+    }
+
+
+def test_xingtu_show_items_capture_collects_and_deduplicates_cards():
+    class Response:
+        url = "https://www.xingtu.cn/gw/api/author/get_author_show_items_v2"
+
+        def json(self):
+            return {
+                "latest_item_info": [
+                    {"item_id": "1", "title": "视频一", "play": 10},
+                    {"item_id": "2", "item_title": "视频二", "play": 20},
+                ],
+                "latest_star_item_info": [
+                    {"item_id": "2", "item_title": "视频二", "play": 20},
+                ],
+            }
+
+    class Page:
+        def on(self, event, callback):
+            assert event == "response"
+            self.callback = callback
+
+    page = Page()
+    cards = xingtu.setup_show_items_capture(page)
+    page.callback(Response())
+
+    assert [card["note_id"] for card in cards] == ["1", "2"]
+
+
+def test_xingtu_matches_api_cards_without_dom_pagination(monkeypatch):
+    monkeypatch.setattr(
+        xingtu,
+        "parse_videos_stable",
+        lambda _page: (_ for _ in ()).throw(AssertionError("DOM should not be needed")),
+    )
+    monkeypatch.setattr(
+        xingtu,
+        "is_blocked",
+        lambda _page: (_ for _ in ()).throw(AssertionError("page should not be needed")),
+    )
+    tasks = [{
+        "record_id": "r1",
+        "target_title": "飞书旧标题",
+        "note_url": "https://www.douyin.com/video/7654321",
+        "note_id": "7654321",
+    }]
+    api_cards = [{
+        "title": "平台上的新标题",
+        "play_count": 168000,
+        "note_id": "7654321",
+        "source": "show_items_api",
+    }]
+
+    found, _, page_limit_hit = xingtu.find_videos_for_tasks(
+        object(), tasks, max_pages=5, api_cards=api_cards,
+    )
+
+    assert found["r1"]["play_count"] == 168000
+    assert found["r1"]["match_method"] == "content_id"
+    assert page_limit_hit is False
+
+
+def test_xingtu_api_fuzzy_candidate_does_not_preempt_later_exact_dom(monkeypatch):
+    monkeypatch.setattr(xingtu, "is_blocked", lambda _page: False)
+    monkeypatch.setattr(
+        xingtu,
+        "parse_videos_stable",
+        lambda _page: [{"title": "commuter backpack review", "note_id": "exact-card"}],
+    )
+    monkeypatch.setattr(xingtu, "go_next_page", lambda _page: False)
+    monkeypatch.setattr(xingtu, "submit_video_search", lambda *_: False)
+    task = {"record_id": "r1", "target_title": "commuter backpack review"}
+    api_cards = [{
+        "title": "commuter backpack review today",
+        "note_id": "fuzzy-card",
+        "source": "show_items_api",
+    }]
+
+    found, _, _ = xingtu.find_videos_for_tasks(
+        object(), [task], max_pages=2, api_cards=api_cards,
+    )
+
+    assert found["r1"]["note_id"] == "exact-card"
+
+
+def test_browser_session_loss_detection_is_specific():
+    closed = RuntimeError("Target page, context or browser has been closed")
+    assert pgy.is_browser_session_lost(closed)
+    assert xingtu.is_browser_session_lost(closed)
+    assert not pgy.is_browser_session_lost(RuntimeError("ordinary parse failure"))
+
+
+def test_xingtu_nested_session_loss_is_reraised():
+    closed = RuntimeError("Target page, context or browser has been closed")
+    with pytest.raises(RuntimeError, match="has been closed"):
+        xingtu.reraise_if_browser_session_lost(closed)
+    xingtu.reraise_if_browser_session_lost(RuntimeError("ordinary lookup miss"))
+
+
+def test_xingtu_detail_and_video_search_propagate_closed_session():
+    closed = RuntimeError("Target page, context or browser has been closed")
+
+    class Detail:
+        def goto(self, *_args, **_kwargs):
+            raise closed
+
+    class Context:
+        pages = []
+
+        def new_page(self):
+            return Detail()
+
+    class SearchPage:
+        context = Context()
+
+        def evaluate(self, _script, _arg=None):
+            return {"found": True, "url": "https://www.xingtu.cn/ad/creator/author-homepage/douyin-video/1"}
+
+    with pytest.raises(RuntimeError, match="has been closed"):
+        xingtu.open_creator_detail(SearchPage(), "达人", "creator-id")
+
+    class Locator:
+        first = None
+
+        def __init__(self):
+            self.first = self
+
+        def count(self):
+            return 1
+
+        def wait_for(self, **_kwargs):
+            raise closed
+
+    class VideoPage:
+        def locator(self, _selector):
+            return Locator()
+
+    with pytest.raises(RuntimeError, match="has been closed"):
+        xingtu.submit_video_search(VideoPage(), "标题")
+
+
+def test_pgy_fallback_detail_click_propagates_closed_session():
+    closed = RuntimeError("Target page, context or browser has been closed")
+
+    class Clicker:
+        first = None
+
+        def __init__(self):
+            self.first = self
+
+        def click(self, **_kwargs):
+            raise closed
+
+    class Context:
+        pages = []
+
+    class Page:
+        context = Context()
+
+        def evaluate(self, *_args, **_kwargs):
+            return {"ok": False, "reason": "no-row"}
+
+        def get_by_text(self, *_args, **_kwargs):
+            return Clicker()
+
+    with pytest.raises(RuntimeError, match="has been closed"):
+        pgy.open_blogger_detail(Page(), "达人", None)
+
+
+def test_bilibili_triggered_blocked_inputs_are_visible_terminal_rows():
+    summary = bili.finalize_bili_summary({
+        "style": "款式A", "index": 1, "total_b_records": 1,
+        "details": [{
+            "record_id": "r1", "status": "blocked_input",
+            "matched": False, "reason": "publish_time_missing", "ok": False,
+        }],
+    })
+
+    assert summary["blocked_input"] == 1
+    assert summary["unresolved"] == 0
+    assert summary["complete"] is True
+
+
+def test_bilibili_missing_url_does_not_trigger_collection(monkeypatch):
+    style = {
+        "style": "款式A",
+        "name": "款式A",
+        "index": 1,
+        "base_token": "base",
+        "table_id": "table",
+        "field_map": {
+            "platform": {"field_id": "platform"},
+            "note_url": {"field_id": "url"},
+            "creator_name": {"field_id": "creator"},
+            "publish_time": {"field_id": "pub"},
+            "read_count_7d": {"field_id": "s7"},
+            "read_count_14d": {"field_id": "s14"},
+            "read_count_21d": {"field_id": "s21"},
+            "read_count_28d": {"field_id": "s28"},
+            "month_end": {"field_id": "sm"},
+        },
+    }
+    rows = [
+        {
+            "record_id": "no-link",
+            "platform": "B站",
+            "creator": "未发布达人",
+            "url": "",
+            "pub": "2026-07-20",
+        },
+        {
+            "record_id": "published",
+            "platform": "B站",
+            "creator": "已发布达人",
+            "url": "https://www.bilibili.com/video/BV1234567890",
+            "pub": "2026-07-20",
+        },
+    ]
+    monkeypatch.setattr(bili, "list_all_records", lambda *_args: rows)
+    monkeypatch.setattr(bili, "fetch_play_count", lambda *_args: 321)
+    monkeypatch.setattr(bili, "write_record", lambda *_args, **_kwargs: True)
+
+    summary = bili.process_style(
+        style,
+        only_record_ids=None,
+        dry_run=True,
+        delay=0,
+        session=object(),
+        state={},
+        force_today=bili.date(2026, 7, 23),
+        first_run=False,
+    )
+
+    assert summary["source_b_records"] == 2
+    assert summary["skipped_no_url"] == 1
+    assert summary["total_b_records"] == 1
+    assert summary["success"] == 1
+    assert summary["blocked_input"] == 0
+    assert [row["record_id"] for row in summary["details"]] == ["published"]
+
+
+def test_bilibili_partial_retry_preserves_old_success():
+    existing = {
+        "style": "款式A", "index": 1, "total_b_records": 2,
+        "details": [
+            {"record_id": "ok", "status": "success", "ok": True},
+            {"record_id": "retry", "status": "retryable_failure", "ok": False},
+        ],
+    }
+    partial = {
+        "style": "款式A", "index": 1, "total_b_records": 1,
+        "details": [{"record_id": "retry", "status": "success", "ok": True}],
+    }
+
+    merged = bili.merge_bili_summary(existing, partial)
+
+    assert merged["total_b_records"] == 2
+    assert {row["record_id"] for row in merged["details"]} == {"ok", "retry"}
+    assert merged["complete"] is True
+
+
+def test_bilibili_top_level_error_cannot_finalize_as_complete():
+    finalized = bili.finalize_bili_summary({
+        "style": "款式A",
+        "index": 1,
+        "total_b_records": 0,
+        "details": [],
+        "error": "missing platform/url field_id",
+        "unresolved": 1,
+        "retryable_failures": 1,
+        "complete": False,
+    })
+
+    assert finalized["complete"] is False
+    assert finalized["unresolved"] >= 1
+
+
+def test_bilibili_missing_slot_mapping_is_system_error_not_blocked_input():
+    summary = bili.process_style(
+        {
+            "style": "款式A",
+            "name": "款式A",
+            "index": 1,
+            "base_token": "base",
+            "table_id": "table",
+            "field_map": {
+                "platform": {"field_id": "platform"},
+                "note_url": {"field_id": "url"},
+            },
+        },
+        only_record_ids=None,
+        dry_run=True,
+        delay=0,
+        session=object(),
+        state={},
+        force_today=None,
+        first_run=False,
+    )
+
+    finalized = bili.finalize_bili_summary(summary, dry_run=True)
+    assert "publish_time" in finalized["error"]
+    assert "read_count_7d" in finalized["error"]
+    assert finalized["complete"] is False
+
+
+def test_self_douyin_missing_mapping_and_write_failure_are_not_complete(monkeypatch):
+    missing = self_dy.scrape_one_style(
+        {"name": "款式A", "index": 1, "base_token": "base", "table_id": "table", "field_map": {}},
+        login_timeout=1,
+        headless=True,
+        only_record_ids=None,
+        dry_run=False,
+    )
+    assert cc.finalize_summary(missing)["complete"] is False
+
+    style = _style_for_platform("抖音")
+    row = {
+        "record_id": "r1",
+        "platform": "抖音",
+        "name": "自营达人",
+        "title": "目标作品",
+        "url": "https://www.douyin.com/video/1",
+        "pub": "2026-07-20",
+    }
+
+    class Page:
+        def goto(self, *_args, **_kwargs):
+            return None
+
+        def wait_for_timeout(self, _ms):
+            return None
+
+    class Session:
+        def __init__(self, **_kwargs):
+            pass
+
+        def __enter__(self):
+            return self
+
+        def __exit__(self, *_args):
+            return None
+
+        def fetch(self, _url, page_action, wait):
+            page_action(Page())
+
+    monkeypatch.setattr(self_dy, "DynamicSession", Session)
+    monkeypatch.setattr(self_dy, "list_records_by_table", lambda *_: [row])
+    monkeypatch.setattr(self_dy, "dismiss_popups", lambda *_: None)
+    monkeypatch.setattr(self_dy, "restore_cookies", lambda *_: None)
+    monkeypatch.setattr(self_dy, "maybe_wait_for_login", lambda *_: None)
+    monkeypatch.setattr(self_dy, "save_state", lambda *_: None)
+    monkeypatch.setattr(
+        self_dy,
+        "scroll_and_collect_posts",
+        lambda *_args, **_kwargs: [{"title": "目标作品", "play_count": 123}],
+    )
+    monkeypatch.setattr(self_dy, "pick_read_field", lambda *_: ("read_count_7d", "read7", "7天曝光量"))
+    monkeypatch.setattr(self_dy, "write_back", lambda *_: False)
+
+    failed = cc.finalize_summary(
+        self_dy.scrape_one_style(
+            style,
+            login_timeout=1,
+            headless=True,
+            only_record_ids=None,
+            dry_run=False,
+        )
+    )
+
+    assert failed["results"][0]["status"] == "write_failure"
+    assert failed["complete"] is False
+
+
+def test_run_all_rejects_stale_or_incomplete_summary(tmp_path: Path):
+    path = tmp_path / "summary.json"
+    path.write_text("[]", encoding="utf-8")
+    started_at = path.stat().st_mtime + 1
+
+    assert run_all.validate_summary_payload([], {1}, path, started_at)[0] is False
+
+    path.write_text('[{"index": 1, "total": 2, "results": [{"record_id": "a", "status": "success"}]}]', encoding="utf-8")
+    assert run_all.validate_summary_payload(
+        [{"index": 1, "total": 2, "results": [{"record_id": "a", "status": "success"}]}],
+        {1}, path, path.stat().st_mtime - 1,
+    )[0] is False
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_dashboard_analytics.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_dashboard_analytics.py
new file mode 100644
index 0000000..7bfb452
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_dashboard_analytics.py
@@ -0,0 +1,103 @@
+import unittest
+from datetime import date
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools.daily_dashboard_analytics import build_dashboard_facts
+
+
+class DailyDashboardAnalyticsTests(unittest.TestCase):
+    def setUp(self):
+        self.tracking = [
+            {"style_name": "大流量", "visitors": 1000, "cart_users": 100, "sales": 20, "refund_orders": 2, "light": "🟡"},
+            {"style_name": "高转化", "visitors": 100, "cart_users": 30, "sales": 20, "refund_orders": 8, "light": "🔴"},
+            {"style_name": "极小样本", "visitors": 10, "cart_users": 5, "sales": 10, "refund_orders": 0, "light": "🔴"},
+            {"style_name": "缺数据", "visitors": None, "cart_users": None, "sales": None, "refund_orders": None, "light": "无判断"},
+        ]
+        self.platform = [
+            {"style_name": "大流量", "platform": "天猫", "visitors": 900, "cart_users": 90, "sales": 18, "refund_orders": 2},
+            {"style_name": "高转化", "platform": "天猫", "visitors": 100, "cart_users": 30, "sales": 20, "refund_orders": 8},
+            {"style_name": "极小样本", "platform": "天猫", "visitors": 10, "cart_users": 5, "sales": 10, "refund_orders": 0},
+        ]
+        self.personas = [
+            {
+                "style_name": "大流量", "platform": "天猫", "data_date": date(2026, 7, 15),
+                "gender_top": "女性 60%", "age_top": "26-35岁 45%",
+                "city_top": "一线城市 35%", "buying_power_top": "高级白领 40%",
+            }
+        ]
+
+    def test_all_styles_receive_same_metrics_and_diagnosis(self):
+        facts = build_dashboard_facts(
+            "2026-07-15", self.tracking, self.platform, self.personas,
+            {"大流量": {"creator_count": 3, "directions": ["通勤"]}},
+        )
+        self.assertEqual(len(facts["styles"]), 4)
+        high = next(row for row in facts["styles"] if row["style_name"] == "高转化")
+        self.assertEqual(high["conversion_rate"], 0.2)
+        self.assertEqual(high["cart_rate"], 0.3)
+        self.assertEqual(high["refund_rate"], 0.4)
+        missing = next(row for row in facts["styles"] if row["style_name"] == "缺数据")
+        self.assertEqual(missing["diagnosis"], "无数据")
+
+    def test_top_conversion_excludes_tiny_visitor_base(self):
+        facts = build_dashboard_facts("2026-07-15", self.tracking, self.platform, self.personas, {})
+        self.assertEqual(facts["rankings"]["conversion"][0]["style_name"], "高转化")
+        self.assertEqual(facts["rankings"]["visitors"][0]["style_name"], "大流量")
+        self.assertEqual(facts["rankings"]["refund_rate"][0]["style_name"], "高转化")
+
+    def test_platform_winners_keep_missing_persona_explicit(self):
+        facts = build_dashboard_facts("2026-07-15", self.tracking, self.platform, self.personas, {})
+        card = facts["platform_personas"][0]
+        self.assertEqual(card["platform"], "天猫")
+        self.assertEqual(card["top_conversion"]["style_name"], "高转化")
+        self.assertEqual(card["top_conversion"]["persona_text"], "无数据")
+        self.assertEqual(card["top_visitors"]["style_name"], "大流量")
+        self.assertIn("推测", card["top_visitors"]["interest_hypothesis"])
+        self.assertIn("提转化", card["top_visitors"]["optimization"])
+
+    def test_daily_new_notes_overview_replaces_view_delta_and_creator_ranking(self):
+        notes = [{"note_id": 7, "style_name": "大流量", "creator": "达人A", "platform": "小红书", "view_count": 500}]
+        facts = build_dashboard_facts(
+            "2026-07-15", self.tracking, self.platform, self.personas, {}, daily_notes=notes
+        )
+        self.assertEqual(facts["daily_new_notes"], notes)
+        self.assertNotIn("creator_connections", facts["rankings"])
+        self.assertNotIn("daily_views", facts["rankings"])
+
+    def test_daily_view_ranking_and_platform_cards_receive_stored_images(self):
+        images = {
+            ("高转化", "天猫"): {"local_image_path": "same.jpg", "image_platform": "天猫"},
+            ("大流量", "天猫"): {"local_image_path": "visitor.jpg", "image_platform": "天猫"},
+        }
+        facts = build_dashboard_facts(
+            "2026-07-15", self.tracking, self.platform, self.personas, {},
+            style_images=images,
+        )
+        card = facts["platform_personas"][0]
+        self.assertEqual(card["top_conversion"]["local_image_path"], "same.jpg")
+        self.assertEqual(card["top_visitors"]["local_image_path"], "visitor.jpg")
+
+    def test_matrix_groups_by_style_category_without_total_exposure(self):
+        signals = {
+            "高转化": {"persona": True, "note_count": 3, "recent_note_exposure": 43000, "comment_count": 80,
+                       "review_count": 10, "review_negative_rate": 0.3, "creative_ctr": 0.08},
+        }
+        facts = build_dashboard_facts(
+            "2026-07-15", self.tracking, self.platform, self.personas, {},
+            style_categories={
+                "大流量": "都市机能", "高转化": "智性通勤",
+                "极小样本": "智性通勤", "缺数据": "都市机能",
+            },
+            style_signals=signals,
+        )
+        self.assertEqual([group["style_category"] for group in facts["style_groups"]], ["都市机能", "智性通勤"])
+        self.assertEqual([row["style_name"] for row in facts["style_groups"][0]["styles"]], ["大流量", "缺数据"])
+        high = next(row for row in facts["styles"] if row["style_name"] == "高转化")
+        self.assertNotIn("total_note_exposure", high)
+        self.assertEqual(high["recent_note_exposure"], 43000)
+        self.assertNotIn("total_note_exposure", facts["definitions"])
+        self.assertIn("评价风险", high["diagnosis"])
+        self.assertIn("画像", high["evidence_summary"])
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_marketing_report.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_marketing_report.py
new file mode 100644
index 0000000..aeb6a1c
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_marketing_report.py
@@ -0,0 +1,429 @@
+import unittest
+from datetime import date
+from unittest.mock import patch
+
+import daily_marketing_report as daily
+
+
+class DailyMarketingReportTests(unittest.TestCase):
+    def test_sales_forecast_uses_seven_day_total_and_recent_three_day_velocity(self):
+        forecast = daily.estimate_sales_forecast([10, 12, 14, 16, 18, 20, 22])
+
+        self.assertEqual(forecast["sales_7d_total"], 112)
+        self.assertEqual(forecast["sales_forecast_7d"], 140)
+        self.assertAlmostEqual(forecast["sales_forecast_change"], 7 / 13, places=4)
+
+    def test_sales_forecast_requires_complete_seven_day_history(self):
+        self.assertEqual(
+            daily.estimate_sales_forecast([10, 12, 14]),
+            {
+                "sales_7d_total": None,
+                "sales_forecast_7d": None,
+                "sales_forecast_change": None,
+            },
+        )
+
+    def test_default_recipients_include_all_four_daily_report_owners(self):
+        self.assertEqual(
+            daily.DEFAULT_RECIPIENT_OPEN_IDS,
+            (
+                "ou_7ad5fc8012e2f741afc5346e05ffd447",
+                "ou_fa8d81a16527ad06352dbecc575285b8",
+                "ou_2eda5eec112109ae6d19a1f6813eadcb",
+                "ou_89bcff110ccbb09a23548dc0fb3d880c",
+            ),
+        )
+
+    def test_hermes_transport_error_is_not_accepted_as_report(self):
+        with self.assertRaisesRegex(RuntimeError, "Hermes 分析失败"):
+            daily.require_valid_hermes_report("API call failed after 3 retries: Connection error.")
+        self.assertEqual(daily.require_valid_hermes_report("整体经营分析\n销量保持稳定"), "整体经营分析\n销量保持稳定")
+
+    def test_light_uses_highest_absolute_visitor_or_cart_change(self):
+        self.assertEqual(daily.classify_light(0.05, -0.08), "🟢")
+        self.assertEqual(daily.classify_light(0.12, 0.02), "🟡")
+        self.assertEqual(daily.classify_light(-0.31, 0.01), "🔴")
+        self.assertEqual(daily.classify_light(None, None), "无判断")
+
+    def test_comment_signal_requires_minimum_sample(self):
+        comments = ["光影行星宙斯怎么买"] * 49
+        self.assertIsNone(daily.calculate_comment_signal("宙斯", comments))
+
+        comments.append("好看")
+        signal = daily.calculate_comment_signal("宙斯", comments)
+        self.assertEqual(signal["sample_size"], 50)
+        self.assertEqual(signal["brand_mention_rate"], 0.98)
+        self.assertEqual(signal["purchase_intent_rate"], 0.98)
+
+    def test_prompt_uses_only_the_fresh_user_instruction_and_database_facts(self):
+        prompt = daily.build_daily_prompt(
+            report_date="2026-07-15",
+            today_notes=[],
+            tracking_rows=[],
+            next_day_rows=[],
+            comment_rows=[],
+        )
+        self.assertIn("以下为数据库查询结果", prompt)
+        self.assertIn('"报告日期": "2026-07-15"', prompt)
+        self.assertIn("增长型SKU", daily.SYSTEM_PROMPT)
+        self.assertIn("流量机会型SKU", daily.SYSTEM_PROMPT)
+        self.assertIn("潜力型SKU", daily.SYSTEM_PROMPT)
+        self.assertIn("风险型SKU", daily.SYSTEM_PROMPT)
+        self.assertNotIn("A类：爆款增长SKU", daily.SYSTEM_PROMPT)
+
+    def test_prompt_uses_fashion_bag_data_analyst_role(self):
+        self.assertIn("专业的时尚包袋品牌电商运营数据分析师", daily.SYSTEM_PROMPT)
+        self.assertIn("销售、电商运营及达人营销数据", daily.SYSTEM_PROMPT)
+        self.assertIn("发现增长机会和风险", daily.SYSTEM_PROMPT)
+        self.assertIn("SKU营销运营日报", daily.SYSTEM_PROMPT)
+        self.assertIn("展示关键数据", daily.SYSTEM_PROMPT)
+
+    def test_fresh_prompt_contains_the_complete_requested_analysis_structure(self):
+        prompt = daily.SYSTEM_PROMPT
+        for phrase in (
+            "今日整体经营判断",
+            "SKU经营表现分析",
+            "达人营销效果分析",
+            "产品节奏判断",
+            "蓄水期",
+            "放量期",
+            "稳定期",
+            "衰退期",
+            "增长SKU TOP3",
+            "风险SKU TOP3",
+            "比较昨日、七日基准",
+            "用图表展示",
+            "曝光量",
+            "建议动作",
+            "对销量影响",
+            "所有建议必须具体到SKU",
+        ):
+            self.assertIn(phrase, prompt)
+        for retired in ("A类：爆款增长SKU", "B类：潜力增长SKU", "D类：需要优化SKU"):
+            self.assertNotIn(retired, prompt)
+
+    def test_database_payload_keeps_all_available_sources(self):
+        prompt = daily.build_daily_prompt("2026-07-15", [], [], [], [], {"reviews": [{"sample_size": 3}]})
+        for phrase in ("所有SKU电商运营数据", "达人营销内容数据", "评论与补充经营数据", "reviews"):
+            self.assertIn(phrase, prompt)
+
+    def test_note_label_ignores_placeholder_creator_name(self):
+        note = {"id": 7, "creator_name": "/", "title": "瑞白通勤包实测"}
+        self.assertEqual(daily.note_label(note), "瑞白通勤包实测")
+
+    def test_cooperation_metrics_hide_unverifiable_engagement(self):
+        invalid = daily._sanitize_cooperation_row({
+            "publish_time": None,
+            "exposure_count": 38000,
+            "engagement_count_num": 21430000,
+        })
+        self.assertIsNone(invalid["engagement_count_num"])
+        self.assertEqual(invalid["publish_status"], "发布时间无数据")
+        self.assertIn("口径异常", invalid["engagement_data_status"])
+
+        valid = daily._sanitize_cooperation_row({
+            "publish_time": "2026-07-14",
+            "exposure_count": 38000,
+            "engagement_count_num": 2143,
+        })
+        self.assertEqual(valid["engagement_count_num"], 2143)
+        self.assertEqual(valid["publish_status"], "已记录发布时间")
+
+    def test_load_report_facts_tracks_all_styles_for_report_date(self):
+        metrics = {
+            "宙斯": {"light": "🟢"},
+        }
+        with patch.object(daily, "_note_rows", return_value=[]), \
+             patch.object(daily, "_all_style_names", return_value=["宙斯", "蓝鹊"]) as all_styles, \
+             patch.object(daily, "_style_metrics", return_value=metrics) as style_metrics:
+            with patch.object(daily, "_load_enrichment", return_value={"platform_metrics": []}), \
+                 patch.object(daily, "_platform_metrics", return_value=[]), \
+                 patch.object(daily, "_persona_context", return_value=[]), \
+                 patch.object(daily, "load_creator_connections", return_value={}), \
+                 patch.object(daily, "load_style_images", return_value={}), \
+                 patch.object(daily, "load_style_categories", return_value={}), \
+                 patch.object(daily, "_cooperation_context", return_value=[]), \
+                 patch.object(daily, "_style_comment_context", return_value=[]), \
+                 patch.object(daily, "_review_context", return_value=[]), \
+                 patch.object(daily, "_creative_context", return_value=[]), \
+                 patch.object(daily, "build_dashboard_facts", return_value={"styles": ["宙斯", "蓝鹊"]}):
+                _, tracking, _, _, enrichment = daily.load_report_facts(date(2026, 7, 15))
+
+        all_styles.assert_called_once_with(date(2026, 7, 15))
+        self.assertEqual([row["style_name"] for row in tracking], ["宙斯", "蓝鹊"])
+        self.assertEqual(tracking[1]["light"], "无判断")
+        self.assertEqual(style_metrics.call_args_list[0].args[1], ["宙斯", "蓝鹊"])
+        self.assertEqual(enrichment["platform_metrics"], [])
+        self.assertEqual(enrichment["candidate_styles"], [])
+        self.assertEqual(enrichment["dashboard"]["styles"], ["宙斯", "蓝鹊"])
+
+    def test_report_date_uses_yesterday_notes_for_daily_new_notes(self):
+        notes = [
+            {"id": 1, "publish_date": "2026-07-15", "style_name": "宙斯", "creator_name": "达人甲", "platform": "小红书", "publish_time": "2026-07-15 10:00", "title": "今天", "view_count": 10, "engagement_rate": None},
+            {"id": 2, "publish_date": "2026-07-14", "style_name": "宙斯", "creator_name": "达人乙", "platform": "抖音", "publish_time": "2026-07-14 11:00", "title": "昨天", "view_count": 20, "engagement_rate": None},
+        ]
+        with patch.object(daily, "_note_rows", return_value=notes), \
+             patch.object(daily, "_all_style_names", return_value=["宙斯"]), \
+             patch.object(daily, "_style_metrics", return_value={}), \
+             patch.object(daily, "_load_enrichment", return_value={}), \
+             patch.object(daily, "_platform_metrics", return_value=[]), \
+             patch.object(daily, "_persona_context", return_value=[]), \
+             patch.object(daily, "load_creator_connections", return_value={}), \
+             patch.object(daily, "load_style_images", return_value={}), \
+             patch.object(daily, "load_style_categories", return_value={}), \
+             patch.object(daily, "_cooperation_context", return_value=[]), \
+             patch.object(daily, "_style_comment_context", return_value=[]), \
+             patch.object(daily, "_review_context", return_value=[]), \
+             patch.object(daily, "_creative_context", return_value=[]), \
+             patch.object(daily, "_comment_rows", return_value=[]), \
+             patch.object(daily, "build_dashboard_facts", return_value={}) as dashboard:
+            daily_notes, _, _, _, _ = daily.load_report_facts(date(2026, 7, 15))
+
+        self.assertEqual([row["note_id"] for row in daily_notes], [2])
+        self.assertEqual([row["note_id"] for row in dashboard.call_args.kwargs["daily_notes"]], [2])
+
+    def test_prompt_requires_all_styles_without_top_ten_truncation(self):
+        prompt = daily.build_daily_prompt("2026-07-15", [], [], [], [])
+        self.assertIn("所有SKU", prompt)
+        self.assertNotIn("只保留异常程度最高的10个", prompt)
+        self.assertIn("SKU经营表现分析", daily.SYSTEM_PROMPT)
+        self.assertIn("加购率", daily.SYSTEM_PROMPT)
+
+    def test_prompt_uses_the_new_daily_output_contract(self):
+        self.assertIn("电商运营数据", daily.SYSTEM_PROMPT)
+        self.assertIn("达人营销数据", daily.SYSTEM_PROMPT)
+        self.assertNotIn("TOP20", daily.SYSTEM_PROMPT)
+        self.assertIn("一页A4", daily.SYSTEM_PROMPT)
+        self.assertIn("↑↓展示趋势", daily.SYSTEM_PROMPT)
+        self.assertNotIn("→稳定", daily.SYSTEM_PROMPT)
+        self.assertIn("电商销量表现 + 访客数 + 转化率 + 达人笔记曝光 + 用户反馈", daily.SYSTEM_PROMPT)
+
+    def test_prompt_payload_keeps_yesterday_and_seven_day_comparisons(self):
+        rows = [{
+            "style_name": "蓝鹊",
+            "sales": 10,
+            "sales_yesterday": 8,
+            "avg_sales_7d": 7.5,
+            "sales_change_yesterday": 0.25,
+            "sales_change": 0.3333,
+        }]
+        prompt = daily.build_daily_prompt("2026-07-16", [], rows, [], [])
+        self.assertIn('"sales_yesterday": 8', prompt)
+        self.assertIn('"avg_sales_7d": 7.5', prompt)
+        self.assertIn('"sales_change_yesterday": 0.25', prompt)
+
+    def test_style_metrics_calculates_yesterday_and_seven_day_changes(self):
+        class FakeCursor:
+            def __enter__(self):
+                return self
+
+            def __exit__(self, *args):
+                return False
+
+            def execute(self, query, params):
+                self.query = query
+                self.params = params
+
+            def fetchall(self):
+                return [(
+                    "蓝鹊",
+                    120, 24, 10, 2,
+                    100, 20, 8, 1,
+                    90, 18, 7.5, 1.2,
+                    [4, 5, 6, 7, 8, 9, 10],
+                )]
+
+        class FakeConnection:
+            def __enter__(self):
+                return self
+
+            def __exit__(self, *args):
+                return False
+
+            def cursor(self):
+                return FakeCursor()
+
+        with patch.object(daily, "get_conn", return_value=FakeConnection()):
+            row = daily._style_metrics(date(2026, 7, 16), ["蓝鹊"])["蓝鹊"]
+
+        self.assertEqual(row["sales_yesterday"], 8)
+        self.assertEqual(row["sales_change_yesterday"], 0.25)
+        self.assertEqual(row["sales_change"], 0.3333)
+        self.assertEqual(row["visitor_change_yesterday"], 0.2)
+        self.assertEqual(row["conversion_rate"], round(10 / 120, 4))
+        self.assertEqual(row["conversion_rate_yesterday"], round(8 / 100, 4))
+        self.assertEqual(row["sales_7d_total"], 49)
+        self.assertEqual(row["sales_forecast_7d"], 63)
+
+    def test_prompt_does_not_reuse_retired_dashboard_sections(self):
+        prompt = daily.build_daily_prompt("2026-07-15", [], [], [], [], {"dashboard": {
+            "rankings": {"sales": []},
+            "platform_personas": [{"platform": "天猫"}],
+            "daily_new_notes": [],
+            "styles": [{"style_name": "蓝鹊", "total_note_exposure": 1000, "evidence_summary": "经营、画像、营销笔记"}],
+            "definitions": {"conversion_rate": "销量/访客"},
+        }})
+        self.assertNotIn("按风格分组的综合经营矩阵", prompt)
+        self.assertNotIn("平台高转化与高访客画像", prompt)
+        self.assertNotIn("全款式经营看板摘要", prompt)
+
+    def test_report_normalization_and_all_style_validation(self):
+        report = daily.normalize_report(
+            "健康度：偏弱。36 个款式中\n蓝鹊同比增长79%\n宙斯环比下降16%",
+            35,
+        )
+        self.assertNotIn("同比", report)
+        self.assertIn("环比下降16%", report)
+        self.assertIn("35个款式中", report)
+        self.assertEqual(daily.normalize_report("销量 0 / 退款 0（无数据）"), "销量 0 / 退款 0（当日均为0）")
+        cleaned = daily.normalize_report("投放尚未发布，带动效应快速衰减；待发布后核对")
+        self.assertNotIn("未发布", cleaned)
+        self.assertNotIn("带动效应", cleaned)
+        self.assertIn("发布时间无数据", cleaned)
+        self.assertNotIn("驱动访客", daily.normalize_report("该内容未驱动访客回升"))
+        daily.validate_all_styles(report, [{"style_name": "蓝鹊"}, {"style_name": "宙斯"}])
+        with self.assertRaisesRegex(RuntimeError, "缺少款式"):
+            daily.validate_all_styles(report, [{"style_name": "白鹭"}])
+
+    def test_report_normalization_removes_database_and_code_style_language(self):
+        raw = """数据已收到，我先用脚本做一遍聚合。
+# 品牌SKU电商运营日报
+
+• 最佳渠道：无法判定。今日 `昨日新发布内容` 为空，cooperations 仅1条，publish_time=null、engagement_data_status=无数据。
+
+5. 数据治理与内容补位：（a）对齐 publish_time；（b）修复互动数据；（c）补齐数据。
+
+报告字数提示：如需压缩请告知。
+完整数据索引：如需导出 CSV 请告诉我。
+"""
+        report = daily.normalize_report(raw)
+
+        self.assertTrue(report.startswith("# 品牌SKU电商运营日报"))
+        for forbidden in ("我先用脚本", "cooperations", "publish_time", "engagement_data_status", "null", "`", "（a）", "报告字数提示", "请告诉我"):
+            self.assertNotIn(forbidden, report)
+        self.assertIn("达人合作记录", report)
+        self.assertIn("发布时间无数据", report)
+        self.assertIn("互动数据无数据", report)
+        self.assertIn("（1）", report)
+
+    def test_system_prompt_forbids_database_fields_and_conversational_tail(self):
+        for phrase in (
+            "禁止输出数据库表名、字段名、JSON键名",
+            "禁止出现反引号、null",
+            "禁止描述生成过程",
+            "禁止在结尾追问",
+        ):
+            self.assertIn(phrase, daily.SYSTEM_PROMPT)
+
+    def test_status_groups_are_replaced_with_complete_deterministic_lists(self):
+        report = "二、种草追踪池\n\nB. 全量状态（固定4行）\n🔴（1）：蓝鹊\n\n三、昨日笔记次日追踪"
+        rows = [
+            {"style_name": "蓝鹊", "light": "🔴"},
+            {"style_name": "宙斯", "light": "🟡"},
+            {"style_name": "白鹭", "light": "🟢"},
+            {"style_name": "云卷2", "light": "无判断"},
+        ]
+        result = daily.replace_status_groups(report, rows)
+        self.assertIn("B. 全量状态（4款）", result)
+        self.assertIn("🔴（1）：蓝鹊", result)
+        self.assertIn("🟡（1）：宙斯", result)
+        self.assertIn("🟢（1）：白鹭", result)
+        self.assertIn("无判断（1）：云卷2", result)
+        self.assertIn("三、昨日笔记次日追踪", result)
+
+    def test_status_groups_are_inserted_when_model_omits_marker(self):
+        report = "一、核心经营结论\n正常\n\n三、达人投放与电商同期关联\n无数据"
+        rows = [{"style_name": "蓝鹊", "light": "🔴"}, {"style_name": "白鹭", "light": "🟢"}]
+        result = daily.replace_status_groups(report, rows, "C. 补充经营信号\n- 店铺大盘：无数据")
+        self.assertIn("B. 全量状态（2款）", result)
+        self.assertIn("🔴（1）：蓝鹊", result)
+        self.assertIn("🟢（1）：白鹭", result)
+        self.assertLess(result.index("B. 全量状态"), result.index("三、达人投放"))
+
+    def test_prompt_passes_available_enrichment_without_restoring_old_sections(self):
+        enrichment = {
+            "platform_metrics": [{"style_name": "蓝鹊", "platform": "天猫", "sales": 14}],
+            "cooperations": [{"style_name": "蓝鹊", "content_direction": "通勤"}],
+            "personas": [{"style_name": "蓝鹊", "gender_top": "女性用户 60%"}],
+            "creatives": [{"style_name": "蓝鹊", "ctr": 4.2}],
+            "reviews": [{"style_name": "蓝鹊", "sample_size": 6}],
+            "shop_overview": [{"platform": "天猫", "week_end": "2026-07-12"}],
+        }
+        prompt = daily.build_daily_prompt("2026-07-15", [], [], [], [], enrichment)
+        for source_key in ("platform_metrics", "cooperations", "personas", "creatives", "reviews", "shop_overview"):
+            self.assertIn(source_key, prompt)
+        self.assertIn("评论与补充经营数据", prompt)
+        self.assertNotIn("补充经营信号", prompt)
+        self.assertNotIn("全量状态", prompt)
+
+    def test_enrichment_section_validation_requires_available_categories(self):
+        enrichment = {"personas": [{}], "creatives": [{}], "reviews": [{}], "shop_overview": [{}]}
+        complete = "条件信号：人物画像；主图；商品评价；店铺大盘"
+        daily.validate_enrichment_sections(complete, enrichment)
+        with self.assertRaisesRegex(RuntimeError, "主图"):
+            daily.validate_enrichment_sections("人物画像 商品评价 店铺大盘", enrichment)
+
+    def test_build_condition_signals_covers_all_available_sources(self):
+        enrichment = {
+            "candidate_styles": ["蓝鹊"],
+            "personas": [{"style_name": "蓝鹊", "platform": "天猫", "data_date": "2026-07-15", "gender_top": "女性 60%"}],
+            "creatives": [{"style_name": "蓝鹊", "platform": "天猫", "data_date": "2026-07-12", "impressions": 1000, "clicks": 50, "ctr": 5.0}],
+            "reviews": [{"style_name": "蓝鹊", "sample_size": 6, "negative_count": 1, "negative_rate": 0.1667, "window_start": "2026-07-09", "window_end": "2026-07-15"}],
+            "shop_overview": [{"platform": "天猫", "week_end": "2026-07-12", "visitors": 10000, "deal_amount": 20000}],
+        }
+        text = daily.build_condition_signals(enrichment)
+        for phrase in ("C. 补充经营信号", "人物画像", "主图", "商品评价", "店铺大盘", "2026-07-12"):
+            self.assertIn(phrase, text)
+
+    def test_persona_top_value_ignores_invalid_percentages(self):
+        payload = {"buying_power": [{"name": "异常", "value": 133}, {"name": "L4", "value": 42}]}
+        self.assertEqual(daily._top_persona_value(payload, "buying_power"), "L4 42.0%")
+
+    def test_enriched_report_rejects_missing_platform_claim_and_invented_new_product(self):
+        daily.validate_enriched_report("平台拆分完整，均为同期合作")
+        with self.assertRaisesRegex(RuntimeError, "平台拆分"):
+            daily.validate_enriched_report("无平台拆分数据")
+        with self.assertRaisesRegex(RuntimeError, "新品"):
+            daily.validate_enriched_report("同期出现多个新品合作")
+
+    def test_platform_persona_section_uses_database_winners_not_model_text(self):
+        report = ("### 三、平台人群与营销机会\n天猫高转化款为错误款式\n\n"
+                  "### 四、明日业务动作\n复查数据\n\n"
+                  "三、平台人群与营销机会\n重复错误款式")
+        enrichment = {"dashboard": {
+            "daily_views_status": "无数据：未保存每日增量快照",
+            "platform_personas": [{
+                "platform": "天猫",
+                "top_conversion": {
+                    "style_name": "布谷", "conversion_rate": 0.0259, "visitors": 617,
+                    "persona_text": "无数据", "interest_hypothesis": "兴趣推测：无数据",
+                    "optimization": "先补采人群画像",
+                },
+                "top_visitors": {
+                    "style_name": "星云2", "conversion_rate": 0.0114, "visitors": 3504,
+                    "persona_text": "男性用户 62.7%", "interest_hypothesis": "兴趣推测：数码3C",
+                    "optimization": "优化详情页",
+                },
+            }],
+        }}
+        result = daily.replace_platform_persona_section(report, enrichment)
+        self.assertIn("转化最高为布谷（转化率2.6%", result)
+        self.assertIn("访客最高为星云2（访客3,504", result)
+        self.assertIn("兴趣推测：数码3C", result)
+        self.assertNotIn("错误款式", result)
+        self.assertNotIn("重复错误款式", result)
+        self.assertIn("四、明日业务动作", result)
+        self.assertEqual(result.count("三、平台人群与营销机会"), 1)
+
+    def test_platform_persona_section_supports_new_daily_report_boundary(self):
+        report = ("### 三、平台人群与营销机会\n模型内容\n\n"
+                  "### 四、今日重点SKU分析\n重点款内容\n")
+        result = daily.replace_platform_persona_section(report, {"dashboard": {}})
+        self.assertIn("三、平台人群与营销机会", result)
+        self.assertIn("四、今日重点SKU分析", result)
+        self.assertNotIn("模型内容", result)
+        self.assertEqual(result.count("三、平台人群与营销机会"), 1)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_marketing_report_schedule.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_marketing_report_schedule.py
new file mode 100644
index 0000000..f551242
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_marketing_report_schedule.py
@@ -0,0 +1,18 @@
+import unittest
+from pathlib import Path
+
+
+class DailyMarketingReportScheduleTests(unittest.TestCase):
+    def test_batch_runs_latest_daily_report_and_propagates_exit_code(self):
+        path = Path(__file__).resolve().parents[1] / "data" / "tools" / "daily_marketing_report.bat"
+        content = path.read_text(encoding="utf-8")
+
+        self.assertIn("daily_marketing_report.py --send", content)
+        self.assertNotIn("--date", content)
+        self.assertIn("daily_marketing_report_%TS%.log", content)
+        self.assertIn("LARK_CLI_NO_PROXY=1", content)
+        self.assertIn("exit /b %RC%", content)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_report_card.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_report_card.py
new file mode 100644
index 0000000..503f86e
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_report_card.py
@@ -0,0 +1,189 @@
+import json
+import tempfile
+import unittest
+from pathlib import Path
+from unittest.mock import patch
+
+from PIL import Image
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools.daily_report_card import (
+    build_analysis_card,
+    build_dashboard_card,
+    build_daily_report_card,
+    prepare_feishu_image,
+    send_card_to_recipients,
+    send_daily_report_cards,
+    upload_dashboard_image,
+)
+
+
+class DailyReportCardTests(unittest.TestCase):
+    def test_dashboard_and_analysis_are_separate_cards_with_full_text(self):
+        report = """![全域种草数据看板](charts/dashboard.png)
+
+第一部分 整体经营分析
+总销量下降15%。
+整体趋势判断：下滑状态。
+
+第二部分 SKU分析排序
+1. 蓝鹊：销量第一。
+2. 星云2：转化领先。
+3. 宙斯：访客增长。
+4. 南斯：退款偏高。
+5. 凌云3：需要优化。
+6. 波塞冬：继续观察。
+
+第三部分 关键预警与建议
+退款订单比上升，48小时内排查。
+"""
+        rows = [{"style_name": "蓝鹊", "light": "🔴"}]
+
+        dashboard = build_dashboard_card("2026-07-17", rows, "img_v3_test")
+        analysis = build_analysis_card("2026-07-17", report)
+
+        dashboard_text = json.dumps(dashboard, ensure_ascii=False)
+        analysis_text = json.dumps(analysis, ensure_ascii=False)
+        self.assertIn("img_v3_test", dashboard_text)
+        self.assertNotIn("总销量下降15%", dashboard_text)
+        self.assertNotIn("img_v3_test", analysis_text)
+        self.assertNotIn("![全域种草数据看板]", analysis_text)
+        self.assertIn("波塞冬：继续观察", analysis_text)
+        self.assertIn("退款订单比上升", analysis_text)
+
+    def test_build_card_uses_card_2_schema_and_dashboard(self):
+        report = """一、核心经营结论
+1. 波塞冬edge访客明显下滑。
+2. 蓝鹊销量同期上行。
+
+二、产品经营分析
+略
+
+四、明日业务动作
+1. 复查波塞冬edge天猫流量。
+2. 跟进蓝鹊投放承接。
+"""
+        rows = [
+            {"style_name": "波塞冬edge", "light": "🔴"},
+            {"style_name": "宙斯", "light": "🟡"},
+            {"style_name": "白鹭", "light": "🟢"},
+        ]
+        card = build_daily_report_card("2026-07-15", report, rows, "img_v3_test")
+
+        self.assertEqual(card["schema"], "2.0")
+        self.assertEqual(card["config"]["width_mode"], "fill")
+        self.assertIn("Hermes", card["header"]["subtitle"]["content"])
+        elements = card["body"]["elements"]
+        self.assertEqual([item["tag"] for item in elements], ["column_set", "img", "markdown", "markdown"])
+        self.assertEqual(elements[1]["img_key"], "img_v3_test")
+        metrics_text = json.dumps(elements[0], ensure_ascii=False)
+        self.assertIn("红灯款式", metrics_text)
+        self.assertIn("1", metrics_text)
+        self.assertNotIn("二、产品经营分析", json.dumps(card, ensure_ascii=False))
+
+    def test_prepare_feishu_image_resizes_to_supported_width(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            source = Path(tmp) / "source.png"
+            target = Path(tmp) / "target.png"
+            Image.new("RGB", (1600, 1000), "white").save(source)
+            prepare_feishu_image(source, target)
+            with Image.open(target) as image:
+                self.assertEqual(image.width, 1500)
+                self.assertLessEqual(image.height / image.width, 16 / 9)
+
+    def test_build_card_reads_new_management_report_headings(self):
+        report = """## 1. 今日经营概览
+销售：↑12%，蓝鹊贡献主要增量。
+整体判断：增长状态。
+
+## 2. SKU健康排名 TOP20
+略
+
+## 6. 今日运营建议
+1. 商品动作：复核蓝鹊库存。
+2. 电商动作：优化宙斯详情页。
+"""
+        card = build_daily_report_card("2026-07-17", report, [], "img_v3_test")
+        card_text = json.dumps(card, ensure_ascii=False)
+        self.assertIn("蓝鹊贡献主要增量", card_text)
+        self.assertIn("复核蓝鹊库存", card_text)
+
+    def test_build_card_reads_fresh_partial_prompt_report_headings(self):
+        report = """第一部分 整体经营分析
+总销量下降15%。
+整体趋势判断：下滑状态。
+
+第二部分 SKU分析排序
+略
+
+第三部分 关键预警与建议
+退款订单比上升，48小时内排查。
+追加增长型SKU预算。
+"""
+        card = build_daily_report_card("2026-07-17", report, [], "img_v3_test")
+        card_text = json.dumps(card, ensure_ascii=False)
+        self.assertIn("总销量下降15%", card_text)
+        self.assertIn("48小时内排查", card_text)
+
+    @patch("data.tools.daily_report_card.subprocess.run")
+    def test_upload_dashboard_uses_named_image_file_field(self, run):
+        run.return_value = type("Result", (), {
+            "returncode": 0, "stdout": json.dumps({"image_key": "img_v3_test"}), "stderr": ""
+        })()
+        key = upload_dashboard_image(Path("charts/dashboard.png"))
+        self.assertEqual(key, "img_v3_test")
+        command = run.call_args.args[0]
+        self.assertEqual(command[command.index("--file") + 1], "image=./dashboard.png")
+
+    @patch("data.tools.daily_report_card.subprocess.run")
+    def test_send_card_to_all_recipients(self, run):
+        run.side_effect = [
+            type("Result", (), {"returncode": 0, "stdout": json.dumps({"message_id": f"om_{i}"}), "stderr": ""})()
+            for i in range(3)
+        ]
+        recipients = ["ou_one", "ou_two", "ou_three"]
+        results = send_card_to_recipients({"schema": "2.0"}, recipients, "2026-07-15")
+
+        self.assertEqual([item["message_id"] for item in results], ["om_0", "om_1", "om_2"])
+        self.assertEqual(run.call_count, 3)
+        commands = [call.args[0] for call in run.call_args_list]
+        self.assertEqual([cmd[cmd.index("--user-id") + 1] for cmd in commands], recipients)
+        self.assertTrue(all("--msg-type" in cmd and "interactive" in cmd for cmd in commands))
+        keys = [cmd[cmd.index("--idempotency-key") + 1] for cmd in commands]
+        self.assertTrue(all(key.startswith("daily-2026-07-15-") for key in keys))
+        self.assertTrue(all(len(key) <= 50 for key in keys))
+
+    @patch("data.tools.daily_report_card.send_card_to_recipients")
+    @patch("data.tools.daily_report_card.upload_dashboard_image", return_value="img_v3_test")
+    @patch("data.tools.daily_report_card.prepare_feishu_image")
+    def test_daily_report_sends_dashboard_then_analysis_as_two_messages_per_recipient(
+        self, prepare, upload, send
+    ):
+        send.side_effect = [
+            [{"message_id": "om_board_1"}, {"message_id": "om_board_2"}],
+            [{"message_id": "om_text_1"}, {"message_id": "om_text_2"}],
+        ]
+        recipients = ["ou_one", "ou_two"]
+
+        results = send_daily_report_cards(
+            "2026-07-17",
+            "第一部分 整体经营分析\n完整分析内容",
+            [{"light": "🟢"}],
+            Path("charts/dashboard.png"),
+            recipients,
+        )
+
+        self.assertEqual(
+            [item["message_id"] for item in results],
+            ["om_board_1", "om_board_2", "om_text_1", "om_text_2"],
+        )
+        self.assertEqual(send.call_count, 2)
+        first_card = send.call_args_list[0].args[0]
+        second_card = send.call_args_list[1].args[0]
+        self.assertIn("img_v3_test", json.dumps(first_card, ensure_ascii=False))
+        self.assertIn("完整分析内容", json.dumps(second_card, ensure_ascii=False))
+        self.assertEqual(send.call_args_list[0].kwargs["message_kind"], "dashboard")
+        self.assertEqual(send.call_args_list[1].kwargs["message_kind"], "analysis")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_report_charts.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_report_charts.py
new file mode 100644
index 0000000..1e66011
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_daily_report_charts.py
@@ -0,0 +1,104 @@
+import tempfile
+import unittest
+from pathlib import Path
+
+from PIL import Image
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools.daily_report_charts import (
+    _split_style_groups,
+    build_overview_cards,
+    classify_style_action,
+    generate_dashboard,
+    select_chart_styles,
+)
+
+
+class DailyReportChartTests(unittest.TestCase):
+    def setUp(self):
+        self.rows = [
+            {"style_name": "大盘下滑", "light": "🔴", "visitors": 1000, "avg_visitors_7d": 2000, "cart_users": 80, "avg_cart_users_7d": 100, "visitor_change": -0.5, "cart_change": -0.2, "sales_change": -0.4, "refund_change": 0.1},
+            {"style_name": "正向放量", "light": "🔴", "visitors": 1500, "avg_visitors_7d": 1000, "cart_users": 180, "avg_cart_users_7d": 100, "visitor_change": 0.5, "cart_change": 0.8, "sales_change": 0.3, "refund_change": -0.1},
+            {"style_name": "小基数", "light": "🔴", "visitors": 20, "avg_visitors_7d": 100, "cart_users": 1, "avg_cart_users_7d": 2, "visitor_change": -0.8, "cart_change": -0.5, "sales_change": None, "refund_change": None},
+            {"style_name": "黄灯A", "light": "🟡", "visitors": 800, "avg_visitors_7d": 1000, "cart_users": 70, "avg_cart_users_7d": 80, "visitor_change": -0.2, "cart_change": -0.125, "sales_change": -0.1, "refund_change": 0.2},
+            {"style_name": "绿灯", "light": "🟢", "visitors": 500, "avg_visitors_7d": 510, "cart_users": 40, "avg_cart_users_7d": 41, "visitor_change": -0.02, "cart_change": -0.02, "sales_change": 0.0, "refund_change": 0.0},
+        ]
+
+    def test_selection_covers_negative_positive_and_small_base(self):
+        names = [row["style_name"] for row in select_chart_styles(self.rows, limit=4)]
+        self.assertIn("大盘下滑", names)
+        self.assertIn("正向放量", names)
+        self.assertIn("小基数", names)
+
+    def test_generate_dashboard_writes_png(self):
+        enrichment = {"platform_metrics": [
+            {"style_name": "大盘下滑", "platform": "天猫", "visitors": 700},
+            {"style_name": "大盘下滑", "platform": "京东", "visitors": 300},
+        ]}
+        with tempfile.TemporaryDirectory() as tmp:
+            path = Path(tmp) / "dashboard.png"
+            generate_dashboard("2026-07-15", self.rows, enrichment, path)
+            self.assertTrue(path.exists())
+            self.assertGreater(path.stat().st_size, 10_000)
+            with Image.open(path) as image:
+                self.assertEqual(image.size, (1600, 2500))
+
+    def test_style_groups_are_split_into_balanced_category_columns(self):
+        groups = [
+            {"style_category": "都市机能", "styles": [{}] * 12},
+            {"style_category": "户外机能", "styles": [{}] * 3},
+            {"style_category": "都市新贵", "styles": [{}] * 3},
+            {"style_category": "智性通勤", "styles": [{}] * 14},
+            {"style_category": "都市新旅", "styles": [{}]},
+            {"style_category": "都市运动", "styles": [{}] * 3},
+        ]
+        left, right = _split_style_groups(groups)
+        self.assertEqual([group["style_category"] for group in left], ["都市机能", "户外机能", "都市新贵"])
+        self.assertEqual([group["style_category"] for group in right], ["智性通勤", "都市新旅", "都市运动"])
+
+        groups.append({"style_category": "未分类", "styles": [{}] * 2})
+        left, right = _split_style_groups(groups)
+        self.assertEqual(sum(len(group["styles"]) for group in left), 18)
+        self.assertEqual(sum(len(group["styles"]) for group in right), 20)
+
+    def test_overview_compares_today_yesterday_and_seven_day_average(self):
+        rows = [
+            {
+                "style_name": "蓝鹊", "sales": 12, "sales_yesterday": 10, "avg_sales_7d": 8,
+                "visitors": 120, "visitors_yesterday": 100, "avg_visitors_7d": 90,
+            },
+            {
+                "style_name": "宙斯", "sales": 18, "sales_yesterday": 20, "avg_sales_7d": 15,
+                "visitors": 180, "visitors_yesterday": 200, "avg_visitors_7d": 150,
+            },
+        ]
+        cards = build_overview_cards(rows, {"daily_new_notes": [{"view_count": 515}]})
+        sales = next(card for card in cards if card["key"] == "sales")
+        conversion = next(card for card in cards if card["key"] == "conversion")
+        content = next(card for card in cards if card["key"] == "content")
+
+        self.assertEqual((sales["today"], sales["yesterday"], sales["baseline_7d"]), (30, 30, 23))
+        self.assertEqual(conversion["today"], 0.1)
+        self.assertEqual(content["today"], 515)
+        self.assertIsNone(content["baseline_7d"])
+
+    def test_style_action_matches_prompt_growth_opportunity_and_risk_rules(self):
+        growth = classify_style_action({
+            "sales_change_yesterday": 0.2, "visitor_change_yesterday": 0.1,
+            "conversion_change_yesterday": 0.001, "refund_change_yesterday": -0.1,
+        })
+        opportunity = classify_style_action({
+            "sales_change_yesterday": -0.1, "visitor_change_yesterday": 0.2,
+            "conversion_change_yesterday": -0.01, "refund_change_yesterday": 0,
+        })
+        risk = classify_style_action({
+            "sales_change_yesterday": -0.2, "visitor_change_yesterday": -0.1,
+            "conversion_change_yesterday": -0.01, "refund_change_yesterday": 0.4,
+        })
+
+        self.assertEqual(growth, ("增长型", "加达人/备货"))
+        self.assertEqual(opportunity, ("流量机会", "优化详情页"))
+        self.assertEqual(risk, ("风险型", "查退款/控投放"))
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_friday_relogin_parallel.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_friday_relogin_parallel.py
new file mode 100644
index 0000000..f71321b
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_friday_relogin_parallel.py
@@ -0,0 +1,130 @@
+from gyxx_flow.modules.content_marketing.runtime.data.tools import friday_relogin_parallel as relogin
+from gyxx_flow.modules.content_marketing.runtime.data.tools import relogin_transaction
+
+
+def test_five_expected_platforms_are_configured():
+    assert set(relogin.PLATFORMS) == {
+        "bilibili", "douyin", "pgy", "xingtu", "xiaohongshu"
+    }
+    assert relogin.PLATFORMS["xingtu"].required_cookie == "sessionid"
+    assert relogin.PLATFORMS["douyin"].required_cookie == "sessionid"
+
+
+def test_default_screenshot_recipients_are_routed_by_platform():
+    assert relogin.DEFAULT_RECIPIENTS == {
+        "pgy": "ou_24cc944d6e43c69c59d6560ad4e2ae6e",
+        "xingtu": "ou_24cc944d6e43c69c59d6560ad4e2ae6e",
+        "douyin": "ou_7ad5fc8012e2f741afc5346e05ffd447",
+        "bilibili": "ou_7ad5fc8012e2f741afc5346e05ffd447",
+        "xiaohongshu": "ou_7ad5fc8012e2f741afc5346e05ffd447",
+    }
+    assert relogin.recipient_for_platform("pgy") == "ou_24cc944d6e43c69c59d6560ad4e2ae6e"
+    assert relogin.recipient_for_platform("douyin") == "ou_7ad5fc8012e2f741afc5346e05ffd447"
+    assert relogin.recipient_for_platform("bilibili") == "ou_7ad5fc8012e2f741afc5346e05ffd447"
+    assert relogin.recipient_for_platform("pgy", "ou_override") == "ou_override"
+
+
+def test_failed_platforms_are_grouped_by_their_responsible_recipient():
+    grouped = relogin.group_platforms_by_recipient(
+        ["pgy", "douyin", "bilibili", "xingtu", "xiaohongshu"]
+    )
+    assert grouped == {
+        "ou_24cc944d6e43c69c59d6560ad4e2ae6e": ["pgy", "xingtu"],
+        "ou_7ad5fc8012e2f741afc5346e05ffd447": [
+            "douyin", "bilibili", "xiaohongshu"
+        ],
+    }
+
+
+def test_login_windows_are_classified_by_title():
+    assert relogin._platform_from_window_title("账号登录 - Google Chrome") == "bilibili"
+    assert relogin._platform_from_window_title("小红书蒲公英 - Google Chrome") == "pgy"
+    assert relogin._platform_from_window_title("抖音授权 - Google Chrome") == "xingtu"
+    assert relogin._platform_from_window_title("抖音 - 记录美好生活") == "douyin"
+    assert relogin._platform_from_window_title("小红书 - 你的生活兴趣社区") == "xiaohongshu"
+    assert relogin._platform_from_window_title("Codex") is None
+
+
+def test_retry_only_failed_platforms_for_at_most_three_rounds():
+    calls = []
+
+    def fake_round(names, attempt):
+        calls.append((tuple(names), attempt))
+        if attempt == 1:
+            return {name: name != "xingtu" for name in names}
+        if attempt == 2:
+            return {name: False for name in names}
+        return {name: True for name in names}
+
+    result = relogin.run_with_retries(
+        ["bilibili", "pgy", "xingtu", "xiaohongshu"],
+        max_attempts=3,
+        run_round=fake_round,
+    )
+
+    assert calls == [
+        (("bilibili", "pgy", "xingtu", "xiaohongshu"), 1),
+        (("xingtu",), 2),
+        (("xingtu",), 3),
+    ]
+    assert all(result.values())
+
+
+def test_stops_after_third_failed_round():
+    calls = []
+
+    def always_fail(names, attempt):
+        calls.append(attempt)
+        return {name: False for name in names}
+
+    result = relogin.run_with_retries(
+        ["pgy"], max_attempts=3, run_round=always_fail
+    )
+
+    assert calls == [1, 2, 3]
+    assert result == {"pgy": False}
+
+
+def test_timeout_terminates_the_complete_process_tree(monkeypatch):
+    calls = []
+
+    class Process:
+        pid = 1234
+
+        def terminate(self):
+            calls.append("fallback-terminate")
+
+        def wait(self, timeout):
+            calls.append(("wait", timeout))
+
+    monkeypatch.setattr(
+        relogin.subprocess,
+        "run",
+        lambda command, **kwargs: calls.append((command, kwargs)),
+    )
+
+    relogin._terminate_process_tree(Process())
+
+    assert calls[0][0] == ["taskkill", "/PID", "1234", "/T", "/F"]
+    assert calls[1] == ("wait", 10)
+
+
+def test_transaction_rolls_back_files_and_profile(tmp_path):
+    cookie = tmp_path / "cookies.json"
+    storage = tmp_path / "storage.json"
+    profile = tmp_path / "profile"
+    cookie.write_text("old-cookie", encoding="utf-8")
+    storage.write_text("old-storage", encoding="utf-8")
+    profile.mkdir()
+    (profile / "old.txt").write_text("old-profile", encoding="utf-8")
+
+    state = relogin_transaction.begin((cookie, storage), profile)
+    cookie.write_text("partial", encoding="utf-8")
+    profile.mkdir()
+    (profile / "partial.txt").write_text("partial", encoding="utf-8")
+    relogin_transaction.rollback(state)
+
+    assert cookie.read_text(encoding="utf-8") == "old-cookie"
+    assert storage.read_text(encoding="utf-8") == "old-storage"
+    assert (profile / "old.txt").read_text(encoding="utf-8") == "old-profile"
+    assert not (profile / "partial.txt").exists()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_global_creator_scraping.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_global_creator_scraping.py
new file mode 100644
index 0000000..7754c3e
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_global_creator_scraping.py
@@ -0,0 +1,305 @@
+import json
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.modules.content_marketing.runtime.creator_task_grouping import chunk_creator_groups, group_tasks_by_creator
+
+import pgy_xhs_scraper_v2 as pgy
+import xingtu_scraper_v2 as xingtu
+
+
+def test_group_tasks_uses_creator_id_and_merges_missing_id_by_unique_name():
+    tasks = [
+        {"creator_name": "同一达人", "creator_id": " 123 ", "record_id": "a"},
+        {"creator_name": "同一达人 ", "creator_id": None, "record_id": "b"},
+        {"creator_name": "另一个达人", "creator_id": None, "record_id": "c"},
+    ]
+
+    groups = group_tasks_by_creator(tasks)
+
+    assert len(groups) == 2
+    assert groups[0]["creator_id"] == "123"
+    assert [task["record_id"] for task in groups[0]["tasks"]] == ["a", "b"]
+
+
+def test_group_tasks_does_not_guess_when_same_name_has_conflicting_ids():
+    tasks = [
+        {"creator_name": "重名达人", "creator_id": "1", "record_id": "a"},
+        {"creator_name": "重名达人", "creator_id": "2", "record_id": "b"},
+        {"creator_name": "重名达人", "creator_id": None, "record_id": "c"},
+    ]
+
+    groups = group_tasks_by_creator(tasks)
+
+    assert len(groups) == 3
+
+
+def test_zero_batch_size_means_one_unlimited_batch():
+    groups = [{"creator_name": str(i), "tasks": []} for i in range(3)]
+    assert chunk_creator_groups(groups, 0) == [groups]
+
+
+def _assert_cross_style_collection(module, monkeypatch):
+    styles = [
+        {"index": 1, "name": "款式A", "base_token": "base-a", "table_id": "table-a", "field_map": {"a": 1}},
+        {"index": 2, "name": "款式B", "base_token": "base-b", "table_id": "table-b", "field_map": {"b": 2}},
+    ]
+
+    def fake_extract(style, only_record_ids):
+        return [{
+            "record_id": f"record-{style['index']}",
+            "creator_name": "同一达人",
+            "creator_id": "creator-1",
+            "target_title": f"标题{style['index']}",
+        }]
+
+    monkeypatch.setattr(module, "extract_target_tasks", fake_extract)
+    tasks, summaries = module.collect_tasks_across_styles(styles, None)
+
+    assert len(tasks) == 2
+    assert len(group_tasks_by_creator(tasks)) == 1
+    assert tasks[0]["style_context"]["table_id"] == "table-a"
+    assert tasks[1]["style_context"]["table_id"] == "table-b"
+    assert summaries[1]["total"] == 1
+    assert summaries[2]["total"] == 1
+
+
+def test_pgy_collects_tasks_across_styles(monkeypatch):
+    _assert_cross_style_collection(pgy, monkeypatch)
+
+
+def test_xingtu_collects_tasks_across_styles(monkeypatch):
+    _assert_cross_style_collection(xingtu, monkeypatch)
+
+
+def test_xingtu_record_scope_is_applied_per_style(monkeypatch):
+    styles = [
+        {"index": 1, "name": "款式A", "base_token": "a", "table_id": "a", "field_map": {}},
+        {"index": 2, "name": "款式B", "base_token": "b", "table_id": "b", "field_map": {}},
+    ]
+    observed = {}
+
+    def fake_extract(style, only_record_ids):
+        observed[style["index"]] = only_record_ids
+        return []
+
+    monkeypatch.setattr(xingtu, "extract_target_tasks", fake_extract)
+
+    xingtu.collect_tasks_across_styles(styles, {1: {"same-id"}, 2: {"other-id"}})
+
+    assert observed == {1: {"same-id"}, 2: {"other-id"}}
+
+
+def test_xingtu_load_retry_scope_keeps_style_and_record_pair(tmp_path):
+    summary_path = tmp_path / "summary.json"
+    summary_path.write_text(json.dumps([
+        {"index": 1, "results": [
+            {"record_id": "same-id", "reason": "creator_search_circuit_open"},
+        ]},
+        {"index": 2, "results": [
+            {"record_id": "same-id", "reason": "title_unmatched"},
+            {"record_id": "other-id", "reason": "creator_search_circuit_open"},
+        ]},
+    ]), encoding="utf-8")
+
+    scope = xingtu.load_retry_scope(summary_path, "creator_search_circuit_open")
+
+    assert scope == {1: {"same-id"}, 2: {"other-id"}}
+
+
+def _assert_matched_task_uses_own_writeback_context(module, metric_key, monkeypatch):
+    calls = []
+    task = {
+        "record_id": "record-b",
+        "publish_time": object(),
+        "style_context": {
+            "index": 2,
+            "base_token": "base-b",
+            "table_id": "table-b",
+            "field_map": {"slot": "style-b-slot"},
+        },
+    }
+    summary = {"matched": 0, "filled": 0, "skipped_no_pubtime": 0, "results": []}
+    monkeypatch.setattr(module, "pick_read_field", lambda fmap, publish_time: ("read_count_7d", "field-b", "7天曝光量"))
+    monkeypatch.setattr(module, "write_back", lambda *args: calls.append(args) or True)
+
+    module.apply_matched_task(task, {metric_key: "1.2万", "title": "命中标题"}, summary, dry_run=False)
+
+    assert calls == [("base-b", "table-b", "record-b", "field-b", 12000)]
+    assert summary["matched"] == 1
+    assert summary["filled"] == 1
+
+
+def test_pgy_match_writes_to_task_style_context(monkeypatch):
+    _assert_matched_task_uses_own_writeback_context(pgy, "read_count", monkeypatch)
+
+
+def test_xingtu_match_writes_to_task_style_context(monkeypatch):
+    _assert_matched_task_uses_own_writeback_context(xingtu, "play_count", monkeypatch)
+
+
+def test_xingtu_search_and_open_creator_submits_only_one_search(monkeypatch):
+    calls = []
+
+    def fail_search(_page, creator_name, creator_id):
+        calls.append((creator_name, creator_id))
+        raise TimeoutError("no result")
+
+    monkeypatch.setattr(xingtu, "search_creator", fail_search)
+    monkeypatch.setattr(
+        xingtu,
+        "open_creator_detail",
+        lambda *_args, **_kwargs: pytest.fail("detail must not open after search failure"),
+    )
+
+    with pytest.raises(TimeoutError, match="no result"):
+        xingtu.search_and_open_creator_once(object(), "同一达人", "creator-1")
+
+    assert calls == [("同一达人", "creator-1")]
+
+
+def test_xingtu_search_budget_blocks_duplicate_without_failure_circuit():
+    budget = xingtu.CreatorSearchBudget(max_consecutive_failures=3)
+
+    assert budget.begin(("id", "1")) is True
+    assert budget.begin(("id", "1")) is False
+    budget.failed("first")
+    assert budget.begin(("id", "2")) is True
+    budget.failed("second")
+    assert budget.begin(("id", "3")) is True
+    assert budget.failed("third") is False
+    assert budget.begin(("id", "4")) is True
+    assert budget.calls == 4
+    assert budget.circuit_reason is None
+
+    budget.exhaust("search_quota_exhausted")
+    assert budget.begin(("id", "5")) is False
+
+
+def test_xingtu_wait_for_results_detects_search_quota_exhausted():
+    class Page:
+        def evaluate(self, *_args):
+            return "quota"
+
+        def wait_for_timeout(self, _milliseconds):
+            pytest.fail("quota detection must return immediately")
+
+    with pytest.raises(xingtu.SearchQuotaExhausted):
+        xingtu.wait_for_visible_creator_result(Page(), "达人", timeout_ms=1000)
+
+
+def test_xingtu_missing_creation_tab_is_not_treated_as_global_block():
+    class Page:
+        def evaluate(self, _script):
+            return "no_creation_tab"
+
+    assert xingtu.is_blocked(Page()) is False
+
+
+@pytest.mark.parametrize("reason", ["verify_text", "verify_iframe", "verify_div"])
+def test_xingtu_explicit_verification_still_opens_global_block(reason):
+    class Page:
+        def evaluate(self, _script):
+            return reason
+
+    assert xingtu.is_blocked(Page()) is True
+
+
+def test_xingtu_missing_creation_tab_falls_back_to_current_creator_page():
+    class Locator:
+        @property
+        def first(self):
+            return self
+
+        def click(self, **_kwargs):
+            raise TimeoutError("missing tab")
+
+    class Page:
+        def get_by_text(self, *_args, **_kwargs):
+            return Locator()
+
+        def evaluate(self, _script):
+            return False
+
+        def wait_for_timeout(self, _milliseconds):
+            pass
+
+    assert xingtu.open_creation_ability(Page()) is False
+
+
+def test_xingtu_current_creator_page_cards_are_used_as_fallback(monkeypatch):
+    class Page:
+        def wait_for_timeout(self, _milliseconds):
+            pass
+
+    monkeypatch.setattr(xingtu, "is_blocked", lambda _page: False)
+    monkeypatch.setattr(
+        xingtu,
+        "parse_videos_stable",
+        lambda _page: [{
+            "title": "MacBook党必看！提升幸福感的开工好物",
+            "play_count": "1.2万",
+            "note_id": "",
+        }],
+    )
+    monkeypatch.setattr(xingtu, "go_next_page", lambda _page: False)
+    monkeypatch.setattr(xingtu, "submit_video_search", lambda _page, _title: False)
+
+    found, _candidates, page_limit_hit = xingtu.find_videos_for_tasks(
+        Page(),
+        [{
+            "record_id": "record-1",
+            "target_title": "MacBook党必看！提升幸福感的开工好物",
+        }],
+    )
+
+    assert found["record-1"]["play_count"] == "1.2万"
+    assert found["record-1"]["match_method"] == "title"
+    assert page_limit_hit is False
+
+
+@pytest.mark.parametrize(
+    "message",
+    [
+        "BrowserContext.new_page: Protocol error (Target.createTarget): Failed to open a new tab",
+        "Protocol error (Target.createTarget)",
+    ],
+)
+def test_xingtu_new_tab_creation_failure_rebuilds_browser_session(message):
+    assert xingtu.is_browser_session_lost(RuntimeError(message)) is True
+
+
+def test_daily_retry_does_not_search_xingtu_again_same_day():
+    batch = Path("data/tools/daily_run.bat").read_text(encoding="utf-8")
+
+    retry_lines = [line for line in batch.splitlines() if "retry_failed.py" in line and "call" in line]
+
+    assert len(retry_lines) == 1
+    assert "--platform bili,pgy" in retry_lines[0]
+
+
+@pytest.mark.parametrize(
+    "batch_path",
+    [
+        Path("data/tools/daily_run.bat"),
+        Path("data/tools/daily_run_with_backfill.bat"),
+    ],
+)
+def test_scheduled_collection_syncs_pg_before_and_after_retry(batch_path):
+    batch = batch_path.read_text(encoding="utf-8")
+    calls = [
+        line.strip().lower()
+        for line in batch.splitlines()
+        if line.strip().lower().startswith("call %python%")
+    ]
+
+    run_index = next(i for i, line in enumerate(calls) if "run_all.py" in line)
+    retry_index = next(i for i, line in enumerate(calls) if "retry_failed.py" in line)
+    sync_indices = [
+        i for i, line in enumerate(calls)
+        if "sync_metrics_to_cmt_notes.py" in line
+    ]
+
+    assert len(sync_indices) == 2
+    assert run_index < sync_indices[0] < retry_index < sync_indices[1]
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_relogin_rollback.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_relogin_rollback.py
new file mode 100644
index 0000000..cde40aa
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_relogin_rollback.py
@@ -0,0 +1,240 @@
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools import relogin_douyin, relogin_pgy, relogin_xingtu
+import xingtu_scraper_v2
+import pgy_xhs_scraper_v2
+
+
+@pytest.mark.parametrize("module", [relogin_pgy, relogin_xingtu])
+def test_failed_relogin_restores_cookie_and_profile(tmp_path, monkeypatch, module):
+    cookie = tmp_path / "cookies.json"
+    profile = tmp_path / "profile"
+    cookie.write_text("old-cookie", encoding="utf-8")
+    profile.mkdir()
+    (profile / "state.txt").write_text("old-profile", encoding="utf-8")
+
+    monkeypatch.setattr(module, "COOKIE_FILE", cookie)
+    monkeypatch.setattr(module, "PROFILE_DIR", profile)
+
+    cookie_backup, profile_backup = module.reset_state()
+    assert not cookie.exists()
+    assert not profile.exists()
+
+    cookie.write_text("partial-new-cookie", encoding="utf-8")
+    profile.mkdir()
+    (profile / "partial.txt").write_text("partial", encoding="utf-8")
+
+    module.restore_previous_state(cookie_backup, profile_backup)
+
+    assert cookie.read_text(encoding="utf-8") == "old-cookie"
+    assert (profile / "state.txt").read_text(encoding="utf-8") == "old-profile"
+    assert not (profile / "partial.txt").exists()
+
+
+@pytest.mark.parametrize("module", [relogin_pgy, relogin_xingtu])
+def test_successful_relogin_keeps_new_cookie(tmp_path, monkeypatch, module):
+    cookie = tmp_path / "cookies.json"
+    profile = tmp_path / "profile"
+    cookie.write_text("old-cookie", encoding="utf-8")
+    profile.mkdir()
+
+    monkeypatch.setattr(module, "COOKIE_FILE", cookie)
+    monkeypatch.setattr(module, "PROFILE_DIR", profile)
+
+    cookie_backup, profile_backup = module.reset_state()
+    cookie.write_text("new-cookie", encoding="utf-8")
+    profile.mkdir()
+
+    module.finish_successful_relogin(profile_backup)
+
+    assert cookie.read_text(encoding="utf-8") == "new-cookie"
+    assert cookie_backup.exists()
+    assert not profile_backup.exists()
+
+
+def test_xingtu_scan_login_opens_customer_sso_and_clicks_douyin():
+    class FakeLocator:
+        def __init__(self):
+            self.first = self
+            self.clicked = False
+            self.evaluated = False
+
+        def count(self):
+            return 1
+
+        def is_visible(self):
+            return True
+
+        def click(self, timeout):
+            self.clicked = True
+
+        def evaluate(self, script):
+            assert script == "e => e.click()"
+            self.evaluated = True
+
+    class FakePage:
+        url = "https://www.xingtu.cn/?redirect_uri=/ad/creator/market"
+
+        def __init__(self):
+            self.goto_url = None
+            self.selectors = []
+            self.douyin = FakeLocator()
+
+        def goto(self, url, **kwargs):
+            self.goto_url = url
+            self.url = url
+
+        def wait_for_timeout(self, _milliseconds):
+            pass
+
+        def locator(self, selector):
+            self.selectors.append(selector)
+            return self.douyin
+
+        def get_by_text(self, _pattern):
+            return FakeLocator()
+
+    page = FakePage()
+    xingtu_scraper_v2.open_scan_login(page)
+
+    assert page.goto_url == xingtu_scraper_v2.CUSTOMER_LOGIN_URL
+    assert 'img[src*="aweme.png"]' in page.selectors
+    assert page.douyin.clicked
+
+
+def test_xingtu_index_page_is_recognized_as_logged_in():
+    class Locator:
+        @property
+        def first(self):
+            return self
+
+        def wait_for(self, timeout):
+            assert timeout == 1200
+
+    class Page:
+        url = "https://www.xingtu.cn/ad/creator/index"
+
+        def get_by_text(self, _text, exact=False):
+            return Locator()
+
+    assert xingtu_scraper_v2.is_logged_in(Page()) is True
+
+
+def test_xingtu_cookie_validation_requires_real_session_cookie(tmp_path):
+    cookie = tmp_path / "xingtu.json"
+    cookie.write_text('[{"name":"other","value":"1"}]', encoding="utf-8")
+    assert xingtu_scraper_v2.has_valid_login_cookie_file(cookie) is False
+
+    cookie.write_text('[{"name":"sessionid","value":"fresh"}]', encoding="utf-8")
+    assert xingtu_scraper_v2.has_valid_login_cookie_file(cookie) is True
+
+
+def test_xingtu_login_only_does_not_report_success_when_page_action_was_swallowed(
+    tmp_path, monkeypatch
+):
+    cookie = tmp_path / "xingtu.json"
+    cookie.write_text('[{"name":"sessionid","value":"old"}]', encoding="utf-8")
+
+    class Session:
+        def __enter__(self):
+            return self
+
+        def __exit__(self, *_args):
+            return False
+
+        def fetch(self, *_args, **_kwargs):
+            # Scrapling can log a page_action error and still return a response.
+            return object()
+
+    monkeypatch.setattr(xingtu_scraper_v2, "COOKIE_FILE", cookie)
+    monkeypatch.setattr(xingtu_scraper_v2, "DynamicSession", lambda **_kwargs: Session())
+    monkeypatch.setattr(xingtu_scraper_v2, "_force_cleanup_session", lambda _session: None)
+
+    assert xingtu_scraper_v2.login_only(1) == 1
+
+
+def test_xingtu_interrupted_transaction_is_recovered(tmp_path, monkeypatch):
+    cookie = tmp_path / "xingtu_cookies.json"
+    profile = tmp_path / "profile"
+    cookie.write_text("old-cookie", encoding="utf-8")
+    profile.mkdir()
+    (profile / "old.txt").write_text("old-profile", encoding="utf-8")
+
+    monkeypatch.setattr(relogin_xingtu, "COOKIE_FILE", cookie)
+    monkeypatch.setattr(relogin_xingtu, "PROFILE_DIR", profile)
+
+    cookie_backup, profile_backup = relogin_xingtu.reset_state()
+    profile.mkdir()
+    (profile / "partial.txt").write_text("partial", encoding="utf-8")
+
+    assert relogin_xingtu.recover_interrupted_state() is True
+    assert cookie.read_text(encoding="utf-8") == "old-cookie"
+    assert (profile / "old.txt").read_text(encoding="utf-8") == "old-profile"
+    assert not (profile / "partial.txt").exists()
+    assert cookie_backup is not None
+    assert profile_backup is not None
+
+
+def test_pgy_scan_login_clicks_qr_switch_image():
+    class FakeLocator:
+        def __init__(self):
+            self.first = self
+            self.clicked = False
+            self.evaluated = False
+
+        def count(self):
+            return 1
+
+        def is_visible(self):
+            return True
+
+        def click(self, timeout):
+            self.clicked = True
+
+        def evaluate(self, script):
+            assert script == "e => e.click()"
+            self.evaluated = True
+
+    class FakePage:
+        def __init__(self):
+            self.selectors = []
+            self.qr_switch = FakeLocator()
+
+        def wait_for_timeout(self, _milliseconds):
+            pass
+
+        def get_by_text(self, _pattern):
+            return FakeLocator()
+
+        def locator(self, selector):
+            self.selectors.append(selector)
+            return self.qr_switch
+
+    page = FakePage()
+    pgy_xhs_scraper_v2.open_scan_login(page)
+
+    assert 'img[src*="qr_code"]' in page.selectors
+    assert page.qr_switch.evaluated
+
+
+def test_failed_douyin_relogin_restores_cookie_storage_and_profile(tmp_path, monkeypatch):
+    cookie = tmp_path / "douyin_cookies.json"
+    storage = tmp_path / "douyin_storage_state.json"
+    profile = tmp_path / "profile"
+    cookie.write_text('[{"name":"sessionid","value":"old"}]', encoding="utf-8")
+    storage.write_text("old-storage", encoding="utf-8")
+    profile.mkdir()
+    (profile / "old.txt").write_text("old-profile", encoding="utf-8")
+
+    monkeypatch.setattr(relogin_douyin, "COOKIE_FILE", cookie)
+    monkeypatch.setattr(relogin_douyin, "STATE_FILE", storage)
+    monkeypatch.setattr(relogin_douyin, "PROFILE_DIR", profile)
+    monkeypatch.setattr(relogin_douyin.subprocess, "call", lambda *_args, **_kwargs: 1)
+    monkeypatch.setattr(relogin_douyin.sys, "argv", ["relogin_douyin.py"])
+
+    assert relogin_douyin.main() == 1
+    assert cookie.read_text(encoding="utf-8") == '[{"name":"sessionid","value":"old"}]'
+    assert storage.read_text(encoding="utf-8") == "old-storage"
+    assert (profile / "old.txt").read_text(encoding="utf-8") == "old-profile"
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_remote_database_config.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_remote_database_config.py
new file mode 100644
index 0000000..8a1c9b8
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_remote_database_config.py
@@ -0,0 +1,72 @@
+import io
+import os
+import unittest
+from pathlib import Path
+from unittest.mock import patch
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools import db
+from gyxx_flow.modules.content_marketing.runtime.data.tools import sync_metrics_to_cmt_notes as sync_metrics
+
+
+ROOT = Path(__file__).resolve().parents[1]
+
+
+class RemoteDatabaseConfigTests(unittest.TestCase):
+    def test_config_requires_every_connection_value_without_local_fallback(self):
+        complete = {
+            "PG_HOST": "db.example.com",
+            "PG_PORT": "5432",
+            "PG_DB": "analytics",
+            "PG_USER": "reporter",
+            "PG_PASSWORD": "secret",
+        }
+        self.assertEqual(
+            db.database_config_from_env(complete),
+            {
+                "host": "db.example.com",
+                "port": 5432,
+                "dbname": "analytics",
+                "user": "reporter",
+                "password": "secret",
+            },
+        )
+        for missing in complete:
+            values = {key: value for key, value in complete.items() if key != missing}
+            with self.subTest(missing=missing), self.assertRaisesRegex(RuntimeError, missing):
+                db.database_config_from_env(values)
+
+    def test_runtime_config_points_to_remote_data_hub(self):
+        config = db.get_db_config()
+        self.assertEqual(config["host"], "8.148.185.119")
+        self.assertEqual(config["dbname"], "data_hub")
+        self.assertEqual(config["user"], "data_hub")
+        self.assertNotIn(config["host"], {"localhost", "127.0.0.1"})
+
+    def test_database_scripts_do_not_define_local_connection_defaults(self):
+        paths = (
+            ROOT / "data" / "tools" / "db.py",
+            ROOT / "data" / "tools" / "sync_metrics_to_cmt_notes.py",
+            ROOT / "data" / "tools" / "sync_cooperations.py",
+            ROOT / "data" / "tools" / "generate_creator_report.py",
+        )
+        for path in paths:
+            source = path.read_text(encoding="utf-8-sig")
+            with self.subTest(path=path.name):
+                self.assertNotIn('"localhost"', source)
+                self.assertNotIn("'localhost'", source)
+                self.assertNotIn('"gyxx_super_data"', source)
+                self.assertNotIn("'gyxx_super_data'", source)
+
+    def test_sync_console_replaces_characters_unsupported_by_gbk(self):
+        raw = io.BytesIO()
+        stream = io.TextIOWrapper(raw, encoding="gbk", errors="strict")
+
+        sync_metrics.configure_console_output(stream, None)
+        stream.write("𝑻")
+        stream.flush()
+
+        self.assertTrue(raw.getvalue())
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_retry_failed_completeness.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_retry_failed_completeness.py
new file mode 100644
index 0000000..aa75a53
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_retry_failed_completeness.py
@@ -0,0 +1,385 @@
+import json
+import os
+import time
+from pathlib import Path
+from types import SimpleNamespace
+
+import gyxx_flow.modules.content_marketing.runtime.data.tools.retry_failed as retry
+import run_all
+
+
+def test_requested_records_must_be_successful_even_when_process_exits_zero():
+    data = {
+        "results": [
+            {"record_id": "a", "status": "success", "matched": True},
+            {"record_id": "b", "status": "retryable_failure", "matched": False},
+        ]
+    }
+
+    ok, failed = retry.requested_records_succeeded(data, ["a", "b"])
+
+    assert ok is False
+    assert failed == ["b"]
+
+
+def test_success_status_cannot_hide_write_failure_and_blocked_input_is_terminal():
+    contradictory = {
+        "results": [{
+            "record_id": "a", "status": "success", "matched": True,
+            "write_ok": False,
+        }]
+    }
+    assert retry.requested_records_succeeded(contradictory, ["a"]) == (False, ["a"])
+
+    blocked = {
+        "total": 1,
+        "details": [{
+            "record_id": "b", "status": "blocked_input", "ok": False,
+            "reason": "url_missing",
+        }],
+    }
+    assert retry.extract_failed_records(blocked, "bili") == []
+
+
+def test_canonical_filename_requires_current_style_name():
+    style = {"index": 2, "name": "极星pro"}
+
+    assert retry.canonical_result_filename(style, "pgy", False) == "02-极星pro_v2.json"
+    assert retry.canonical_result_filename(style, "xt", False) == "02-极星pro_xingtu_v2.json"
+    assert retry.canonical_result_filename(style, "bili", True) == "02-极星pro_self_bilibili_v2.json"
+
+
+def _write_json(path: Path, payload: dict | list) -> None:
+    path.write_text(json.dumps(payload, ensure_ascii=False), encoding="utf-8")
+
+
+def test_find_failed_uses_exact_current_style_file_and_ignores_legacy_name(tmp_path):
+    style = {"index": 2, "name": "极星pro"}
+    canonical = tmp_path / "02-极星pro_v2.json"
+    legacy = tmp_path / "02-盖亚微单_v2.json"
+    payload = {
+        "index": 2,
+        "total": 1,
+        "matched": 1,
+        "filled": 1,
+        "results": [{"record_id": "ok", "status": "success", "matched": True}],
+    }
+    _write_json(canonical, payload)
+    _write_json(legacy, {"index": 2, "total": 1, "matched": 0, "results": []})
+    old = time.time() - 48 * 3600
+    os.utime(legacy, (old, old))
+
+    failed = retry.find_failed(
+        "pgy", stale_hours=4, include_half=True,
+        styles=[style], v2_dir=tmp_path,
+    )
+
+    assert failed == {}
+
+
+def test_find_failed_treats_missing_result_rows_as_incomplete(tmp_path):
+    style = {"index": 1, "name": "款式A"}
+    path = tmp_path / retry.canonical_result_filename(style, "xt", False)
+    _write_json(path, {"index": 1, "total": 2, "matched": 0, "filled": 0, "results": []})
+
+    failed = retry.find_failed(
+        "xt", stale_hours=4, include_half=True,
+        styles=[style], v2_dir=tmp_path,
+    )
+
+    assert failed[1]["reason"] == "ALL_FAILED"
+    assert failed[1]["failed_rids"] == []
+
+
+def test_run_one_style_rechecks_payload_and_retries_exit_zero(monkeypatch, tmp_path):
+    style = {"index": 1, "name": "款式A"}
+    result_path = tmp_path / retry.canonical_result_filename(style, "pgy", False)
+    calls = []
+
+    def fake_call(_cmd, cwd):
+        calls.append(cwd)
+        _write_json(result_path, {
+            "index": 1,
+            "total": 1,
+            "matched": 0,
+            "filled": 0,
+            "results": [{
+                "record_id": "bad",
+                "status": "retryable_failure",
+                "matched": False,
+            }],
+        })
+        return 0
+
+    monkeypatch.setattr(retry.subprocess, "call", fake_call)
+    monkeypatch.setattr(retry.time, "sleep", lambda _seconds: None)
+
+    ok = retry.run_one_style(
+        "pgy", 1, max_attempts=2, record_ids=["bad"],
+        style=style, v2_dir=tmp_path,
+    )
+
+    assert ok is False
+    assert len(calls) == 2
+
+
+def test_run_one_style_accepts_exit_zero_only_after_requested_record_succeeds(monkeypatch, tmp_path):
+    style = {"index": 1, "name": "款式A"}
+    result_path = tmp_path / retry.canonical_result_filename(style, "pgy", False)
+
+    def fake_call(_cmd, cwd):
+        _write_json(result_path, {
+            "index": 1,
+            "total": 1,
+            "matched": 1,
+            "filled": 1,
+            "results": [{
+                "record_id": "ok",
+                "status": "success",
+                "matched": True,
+                "write_ok": True,
+            }],
+        })
+        return 0
+
+    monkeypatch.setattr(retry.subprocess, "call", fake_call)
+
+    assert retry.run_one_style(
+        "pgy", 1, max_attempts=1, record_ids=["ok"],
+        style=style, v2_dir=tmp_path,
+    ) is True
+
+
+def test_self_operated_xingtu_retry_uses_self_douyin_scraper(monkeypatch, tmp_path):
+    style = {"index": 1, "name": "款式A"}
+    result_path = tmp_path / retry.canonical_result_filename(style, "xt", True)
+    observed = {}
+
+    def fake_call(cmd, cwd):
+        observed["cmd"] = cmd
+        _write_json(result_path, {
+            "index": 1,
+            "total": 1,
+            "results": [{"record_id": "ok", "status": "success", "matched": True}],
+        })
+        return 0
+
+    monkeypatch.setattr(retry.subprocess, "call", fake_call)
+
+    assert retry.run_one_style(
+        "xt", 1, max_attempts=1, self_operated=True,
+        record_ids=["ok"], style=style, v2_dir=tmp_path,
+    ) is True
+    assert observed["cmd"][1].endswith("self_douyin_scraper.py")
+    assert "--self-operated" not in observed["cmd"]
+
+
+def test_full_summary_declared_failure_cannot_be_treated_as_complete():
+    complete, reason = retry._summary_completed({
+        "total": 1,
+        "results": [{
+            "record_id": "ok",
+            "status": "success",
+            "matched": True,
+            "write_ok": True,
+        }],
+        "write_failures": 1,
+    })
+
+    assert complete is False
+    assert reason == "summary declares unresolved failures"
+
+
+def test_retry_main_returns_nonzero_when_verified_retry_still_fails(monkeypatch):
+    style = {"index": 1, "name": "款式A"}
+    monkeypatch.setattr(retry, "load_current_styles", lambda self_operated=False: [style])
+    monkeypatch.setattr(
+        retry, "find_failed",
+        lambda *_args, **_kwargs: {1: {"reason": "HALF", "failed_rids": ["bad"]}},
+    )
+    monkeypatch.setattr(retry, "run_one_style", lambda *_args, **_kwargs: False)
+    monkeypatch.setattr(retry.sys, "argv", ["retry_failed.py", "--platform", "pgy"])
+
+    assert retry.main() == 1
+
+
+def test_validate_summary_payload_rejects_stale_missing_duplicate_and_unresolved(tmp_path):
+    path = tmp_path / "summary.json"
+    valid = [{
+        "index": 1,
+        "run_started_at": time.time() + 10,
+        "total": 2,
+        "matched": 2,
+        "filled": 2,
+        "results": [
+            {"record_id": "a", "status": "success", "matched": True},
+            {"record_id": "b", "status": "success", "matched": True},
+        ],
+    }]
+    _write_json(path, valid)
+
+    assert run_all.validate_summary_payload(
+        valid, {1}, path, path.stat().st_mtime - 1,
+    )[0] is True
+    assert run_all.validate_summary_payload(
+        valid, {1}, path, path.stat().st_mtime + 1,
+    )[0] is False
+    assert run_all.validate_summary_payload(valid, {1, 2}, path, 0)[0] is False
+
+    duplicate = [{
+        "index": 1, "total": 2,
+        "results": [
+            {"record_id": "a", "status": "success"},
+            {"record_id": "a", "status": "success"},
+        ],
+    }]
+    assert run_all.validate_summary_payload(duplicate, {1}, path, 0)[0] is False
+
+    unresolved = [{
+        "index": 1, "total": 1,
+        "results": [{
+            "record_id": "a", "status": "retryable_failure", "matched": False,
+        }],
+    }]
+    assert run_all.validate_summary_payload(unresolved, {1}, path, 0)[0] is False
+
+
+def test_validate_selected_style_allows_unrelated_rows_in_merged_global_summary(tmp_path):
+    path = tmp_path / "summary.json"
+    payload = [
+        {
+            "index": 1,
+            "run_started_at": time.time() + 10,
+            "total": 1,
+            "results": [{"record_id": "ok", "status": "success"}],
+        },
+        {
+            "index": 2,
+            "total": 1,
+            "results": [{
+                "record_id": "old-failure",
+                "status": "retryable_failure",
+            }],
+        },
+    ]
+    _write_json(path, payload)
+
+    valid, reason = run_all.validate_summary_payload(
+        payload, {1}, path, path.stat().st_mtime - 1,
+    )
+
+    assert valid is True, reason
+
+
+def test_validate_summary_rejects_expected_style_left_over_from_old_run(tmp_path):
+    path = tmp_path / "summary.json"
+    started_at = 100.0
+    payload = [
+        {
+            "index": 1,
+            "run_started_at": 101.0,
+            "total": 1,
+            "results": [{"record_id": "new", "status": "success"}],
+        },
+        {
+            "index": 2,
+            "run_started_at": 90.0,
+            "total": 1,
+            "results": [{"record_id": "old", "status": "success"}],
+        },
+    ]
+    _write_json(path, payload)
+
+    valid, reason = run_all.validate_summary_payload(
+        payload, {1, 2}, path, started_at,
+    )
+
+    assert valid is False
+    assert "style 2" in reason
+    assert "current run" in reason
+
+
+def test_aggregate_counts_all_unresolved_rows_as_errors():
+    aggregate = run_all.aggregate({
+        "pgy": [{
+            "index": 1,
+            "total": 3,
+            "matched": 1,
+            "filled": 1,
+            "results": [
+                {"record_id": "ok", "status": "success"},
+                {"record_id": "miss", "matched": False},
+                {"record_id": "detail", "reason": "no_detail_page"},
+            ],
+        }],
+    })
+
+    assert aggregate["platforms"]["pgy"]["unresolved"] == 2
+    assert aggregate["platforms"]["pgy"]["errors"] == 2
+
+
+def test_aggregate_sums_declared_failure_categories_when_unresolved_is_absent():
+    aggregate = run_all.aggregate({
+        "xt": [{
+            "index": 1,
+            "total": 2,
+            "matched": 2,
+            "results": [
+                {"record_id": "a", "status": "success"},
+                {"record_id": "b", "status": "success"},
+            ],
+            "retryable_failures": 1,
+            "write_failures": 1,
+        }],
+    })
+
+    assert aggregate["platforms"]["xt"]["unresolved"] == 2
+
+
+def test_aggregate_keeps_partial_counts_when_validation_marks_payload_invalid():
+    aggregate = run_all.aggregate({
+        "pgy": {
+            "error": "invalid summary: unresolved",
+            "payload": [{
+                "index": 1,
+                "total": 1,
+                "matched": 0,
+                "results": [{
+                    "record_id": "bad", "status": "retryable_failure",
+                }],
+            }],
+        },
+    })
+
+    platform = aggregate["platforms"]["pgy"]
+    assert platform["loaded"] is True
+    assert platform["unresolved"] == 1
+    assert platform["validation_error"] == "invalid summary: unresolved"
+
+
+def test_run_round_marks_exit_zero_platform_failed_when_summary_is_invalid(monkeypatch, tmp_path):
+    args = SimpleNamespace(dry_run=True, style=[1])
+    stdout = tmp_path / "stdout.log"
+    stderr = tmp_path / "stderr.log"
+    observed = {}
+
+    monkeypatch.setattr(
+        run_all, "start_process",
+        lambda *_args, **_kwargs: (object(), stdout, stderr),
+    )
+    monkeypatch.setattr(run_all, "wait_all", lambda _procs: {"pgy": 0})
+
+    def fake_load(key, self_operated=False, **kwargs):
+        observed.update(kwargs)
+        return {"error": "invalid summary: stale"}
+
+    monkeypatch.setattr(run_all, "load_summary", fake_load)
+    monkeypatch.setattr(run_all, "print_report", lambda *_args: None)
+    monkeypatch.setattr(run_all, "_atomic_write_json", lambda *_args: None)
+
+    result = run_all.run_round(args, ["pgy"])
+
+    assert result["exit_codes"] == {"pgy": 0}
+    assert result["_non_zero_platforms"] == ["pgy"]
+    assert observed["expected_indices"] == {1}
+    assert isinstance(observed["started_at"], float)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_sync_style_categories.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_sync_style_categories.py
new file mode 100644
index 0000000..94b27a4
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_sync_style_categories.py
@@ -0,0 +1,55 @@
+import unittest
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools.sync_style_categories import extract_style_mapping, normalize_style_name
+
+
+class SyncStyleCategoriesTests(unittest.TestCase):
+    def test_normalizes_feishu_lifecycle_names_to_database_names(self):
+        cases = {
+            "盖亚微单Pro生命进程（新）": "盖亚微单",
+            "星迹&星迹2生命进程（新）": "星迹2",
+            "宙斯双肩包生命进程（新）": "宙斯",
+            "极星Pro生命进程（新）": "极星pro",
+            "瑞白双肩包生命进程（新）": "瑞白",
+            "逐星gt生命进程（新）": "逐星GT",
+            "拾影相机包生命进程 （新）": "拾影斜挎相机包",
+            "极星托特生命进程（新）": "极星托特",
+        }
+        for source, expected in cases.items():
+            with self.subTest(source=source):
+                self.assertEqual(normalize_style_name(source), expected)
+
+    def test_extracts_mapping_and_ignores_blank_rows(self):
+        payload = {
+            "data": {
+                "data": [
+                    [
+                        "[盖亚斜挎生命进程（新）](https://example.feishu.cn/base/TokenA)",
+                        ["都市机能"],
+                    ],
+                    [None, ["都市运动"]],
+                    ["[极星托特生命进程（新）](https://example.feishu.cn/base/TokenB)", ["智性通勤"]],
+                ]
+            }
+        }
+
+        self.assertEqual(
+            extract_style_mapping(payload),
+            {"盖亚斜挎": "都市机能", "极星托特": "智性通勤"},
+        )
+
+    def test_rejects_conflicting_categories_for_same_style(self):
+        payload = {
+            "data": {
+                "data": [
+                    ["[布谷生命进程（新）](https://example/a)", ["都市运动"]],
+                    ["[布谷生命进程（新）](https://example/b)", ["都市机能"]],
+                ]
+            }
+        }
+        with self.assertRaisesRegex(ValueError, "布谷"):
+            extract_style_mapping(payload)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/tests/test_weekly_tmall_persona.py b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_weekly_tmall_persona.py
new file mode 100644
index 0000000..d111b35
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/tests/test_weekly_tmall_persona.py
@@ -0,0 +1,70 @@
+import unittest
+from unittest.mock import patch
+
+from gyxx_flow.modules.content_marketing.runtime import weekly_summary_all as weekly
+
+
+class WeeklyTmallPersonaTests(unittest.TestCase):
+    def test_missing_persona_forbids_invented_percentages(self):
+        text = weekly.format_tmall_persona(None)
+        self.assertIn("不得虚构具体比例", text)
+
+    def test_format_tmall_persona_includes_date_and_dimensions(self):
+        persona = {
+            "collect_date": "2026-07-15",
+            "item_id": "900176117674",
+            "payload": {
+                "gender": [
+                    {"name": "男性用户", "value": 66.97},
+                    {"name": "女性用户", "value": 28.26},
+                ],
+                "age": [{"name": "25-29岁", "value": 26.97}],
+                "city_tier": [{"name": "准一线城市", "value": 31.14}],
+                "buying_power": [{"name": "购买力L5", "value": 30.65}],
+            },
+        }
+
+        text = weekly.format_tmall_persona(persona)
+
+        self.assertIn("2026-07-15", text)
+        self.assertIn("男性用户 66.97%", text)
+        self.assertIn("25-29岁 26.97%", text)
+        self.assertIn("准一线城市 31.14%", text)
+        self.assertIn("购买力L5 30.65%", text)
+
+    def test_generate_summary_injects_latest_tmall_persona_into_single_note_prompt(self):
+        persona = {
+            "collect_date": "2026-07-15",
+            "item_id": "900176117674",
+            "payload": {
+                "gender": [{"name": "男性用户", "value": 66.97}],
+                "age": [{"name": "25-29岁", "value": 26.97}],
+                "city_tier": [],
+                "buying_power": [],
+            },
+        }
+        notes = [{
+            "title": "测试笔记",
+            "platform": "douyin",
+            "creator_name": "达人甲",
+            "view_count": 100,
+            "like_count": 10,
+            "comment_count": 2,
+            "favorite_count": 3,
+            "share_count": 1,
+            "report_path": None,
+        }]
+
+        with patch.object(weekly, "get_latest_tmall_persona", return_value=persona) as get_persona, \
+             patch.object(weekly, "call_hermes_analyzer", side_effect=["单篇分析", "## 三、周度总结"]) as call:
+            weekly.generate_summary("星云2", notes)
+
+        get_persona.assert_called_once_with("星云2")
+        single_prompt = call.call_args_list[0].args[1]
+        self.assertIn("天猫该款最新人物画像", single_prompt)
+        self.assertIn("男性用户 66.97%", single_prompt)
+        self.assertIn("内容触达人群", single_prompt)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/weekly_summary_all.py b/src/gyxx_flow/modules/content_marketing/runtime/weekly_summary_all.py
new file mode 100644
index 0000000..653e80b
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/weekly_summary_all.py
@@ -0,0 +1,766 @@
+"""通用周笔记汇总脚本:对所有上周有笔记的款,生成周汇总飞书文档并写入对应生命进程表。
+
+流程:
+1. 读索引表(TtoCb1NuQaDy3NsZWTpc0GIvnph / tblKCjplVAFrRwMC)拿"每周笔记分析"字段(每款生命进程表地址)
+2. PG 查上周(2026-06-29~07-05)新增笔记,按款分组
+3. 对每个款:
+   a. 跑单品分析(调用 analyze_note.py --note-id N --output ...)
+   b. LLM 综合评级(用 weekly_summary 的 prompt)
+   c. 创建飞书文档(markdown)
+   d. 写入该款生命进程表的"笔记分析汇总"字段(时间=上周范围)
+
+用法:
+  python weekly_summary_all.py                 # 跑所有上周有笔记的款
+  python weekly_summary_all.py --style 宙斯     # 只跑某个款
+  python weekly_summary_all.py --dry-run        # 不写飞书,只生成报告
+"""
+import argparse
+import json
+import os
+import re
+import subprocess
+import sys
+from datetime import datetime, timedelta
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+sys.stdout.reconfigure(encoding='utf-8')
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools.analyze_comments import call_hermes_analyzer
+
+LARK = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK):
+    LARK = "lark-cli.cmd"
+
+# 索引表
+INDEX_BT = "TtoCb1NuQaDy3NsZWTpc0GIvnph"
+INDEX_TID = "tblKCjplVAFrRwMC"
+INDEX_STYLE_FIELD = "fldr8T4ynL"
+INDEX_WEEKLY_FIELD = "fldgfS6I4a"
+
+# 周时间范围:动态算上周一到上周日
+# 支持 --week-since/--week-until 参数覆盖(调试用,格式 YYYY-MM-DD)
+def _compute_last_week():
+    """返回 (week_start, week_end, time_label) 都是字符串"""
+    today = datetime.now().date()
+    # 上周一 = 今天 - (今天weekday + 7) 天;周一 weekday=0
+    last_monday = today - timedelta(days=today.weekday() + 7)
+    last_sunday = last_monday + timedelta(days=6)
+    return last_monday, last_sunday
+
+_DEFAULT_START, _DEFAULT_END = _compute_last_week()
+WEEK_START = os.environ.get("WEEK_SINCE", _DEFAULT_START.strftime("%Y-%m-%d"))
+WEEK_END = os.environ.get("WEEK_UNTIL", _DEFAULT_END.strftime("%Y-%m-%d"))
+TIME_LABEL = f"{WEEK_START}~{WEEK_END[5:]}"
+
+DATA_DIR = PATHS.raw_root
+REPORTS_DIR = PATHS.exports_root / "summary"
+REPORTS_DIR.mkdir(parents=True, exist_ok=True)
+
+# dry_run 全局开关(在 main 设置,save_to_db 读取)
+dry_run_global = False
+
+URL_RE = re.compile(r"https?://[^\s)>\]]+")
+
+
+def log(msg: str) -> None:
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {msg}", flush=True)
+
+
+def call_lark_json(args):
+    cmd = [LARK] + args
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, env=env,
+                          encoding="utf-8", errors="replace")
+    text = proc.stdout if proc.stdout else (proc.stderr or "")
+    start = text.find("{")
+    if start < 0:
+        return {"ok": False, "error": f"no json: {text[:300]}"}
+    try:
+        return json.loads(text[start:], strict=False)
+    except json.JSONDecodeError as e:
+        return {"ok": False, "error": f"json decode: {e}"}
+
+
+def parse_url(raw):
+    if not raw:
+        return None
+    if isinstance(raw, list):
+        raw = " ".join(str(x) for x in raw if x)
+    s = str(raw).strip()
+    m = URL_RE.search(s)
+    return m.group(0) if m else None
+
+
+def parse_bt_tid(url):
+    if not url:
+        return None, None
+    bt = re.search(r'/base/([A-Za-z0-9]+)', url)
+    tid = re.search(r'[?&]table=([A-Za-z0-9]+)', url)
+    return (bt.group(1) if bt else None,
+            tid.group(1) if tid else None)
+
+
+def load_index_styles():
+    """读索引表,返回 {款式名: (base_token, table_id, url)}"""
+    cmd = ["base", "+record-list",
+           "--base-token", INDEX_BT, "--table-id", INDEX_TID,
+           "--as", "user", "--format", "json", "--limit", "200",
+           "--field-id", INDEX_STYLE_FIELD, "--field-id", INDEX_WEEKLY_FIELD]
+    resp = call_lark_json(cmd)
+    if not resp.get("ok"):
+        log(f"读索引表失败: {resp.get('error')}")
+        return {}
+
+    styles = {}
+    for row in resp["data"]["data"]:
+        name = row[0] if row else ""
+        weekly = row[1] if len(row) > 1 else ""
+        if isinstance(name, list):
+            name = " ".join(str(x) for x in name if x)
+        name = str(name).strip()
+        if not name or not weekly:
+            continue
+        url = parse_url(weekly)
+        bt, tid = parse_bt_tid(url)
+        if bt and tid:
+            styles[name] = (bt, tid, url)
+    log(f"索引表读到 {len(styles)} 个款有'每周笔记分析'字段")
+    return styles
+
+
+def find_target_fields(bt, tid):
+    """动态读生命进程表的字段,返回 (时间字段_id, 笔记分析汇总字段_id)"""
+    cmd = ["base", "+field-list",
+           "--base-token", bt, "--table-id", tid, "--format", "json"]
+    resp = call_lark_json(cmd)
+    if not resp.get("ok"):
+        return None, None
+    time_fid = summary_fid = None
+    for f in resp["data"]["fields"]:
+        if f.get("name") == "时间":
+            time_fid = f.get("id")
+        elif f.get("name") == "笔记分析汇总":
+            summary_fid = f.get("id")
+    return time_fid, summary_fid
+
+
+def get_last_week_notes(style_name):
+    """从 PG 拿上周该款的新增笔记(按 publish_time 过滤),返回 [{id, platform, title, ...}]"""
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_conn
+    with get_conn() as conn:
+        cur = conn.cursor()
+        cur.execute("""
+            SELECT n.id, n.platform, n.title, n.url, n.view_count,
+                   n.comment_count, n.like_count, n.favorite_count, n.share_count,
+                   c.name AS creator_name
+            FROM cmt_notes n
+            JOIN cmt_styles s ON n.style_id = s.id
+            LEFT JOIN cmt_creators c ON n.creator_id = c.id
+            WHERE s.name = %s AND n.self_operated = FALSE
+            AND n.publish_time IS NOT NULL
+            AND n.publish_time >= %s AND n.publish_time < %s
+            ORDER BY n.id
+        """, (style_name, WEEK_START, WEEK_END + " 23:59:59"))
+        cols = [d[0] for d in cur.description]
+        return [dict(zip(cols, r)) for r in cur.fetchall()]
+
+
+def get_latest_tmall_persona(style_name):
+    """读取不晚于本周结束日的最新天猫画像，失败或无数据时返回 None。"""
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_conn
+    try:
+        with get_conn() as conn:
+            cur = conn.cursor()
+            cur.execute("""
+                SELECT collect_date, item_id, payload
+                FROM daily_persona_metrics
+                WHERE platform = 'tm' AND style_name = %s AND collect_date <= %s
+                ORDER BY collect_date DESC, collected_at DESC
+                LIMIT 1
+            """, (style_name, WEEK_END))
+            row = cur.fetchone()
+        if not row:
+            return None
+        return {
+            "collect_date": row[0],
+            "item_id": row[1],
+            "payload": row[2] or {},
+        }
+    except Exception as exc:
+        log(f"  [WARN] 读取 {style_name} 天猫人物画像失败: {exc}")
+        return None
+
+
+def _format_persona_dimension(rows):
+    """把画像维度格式化为适合提示词引用的百分比分布。"""
+    parts = []
+    for row in rows or []:
+        if not isinstance(row, dict) or not row.get("name"):
+            continue
+        value = row.get("value")
+        try:
+            value_text = f"{float(value):.2f}".rstrip("0").rstrip(".")
+        except (TypeError, ValueError):
+            value_text = str(value or "未知")
+        parts.append(f"{row['name']} {value_text}%")
+    return "；".join(parts) if parts else "暂无"
+
+
+def format_tmall_persona(persona):
+    """将天猫画像记录转成结构化提示词文本。"""
+    if not persona:
+        return (
+            "未找到该款式的天猫人物画像。受众部分只能结合内容和评论做谨慎的定性判断，"
+            "不得虚构具体比例。"
+        )
+    payload = persona.get("payload") or {}
+    return "\n".join([
+        f"- 数据日期：{persona.get('collect_date') or '未知'}",
+        f"- 天猫商品ID：{persona.get('item_id') or '未知'}",
+        f"- 性别：{_format_persona_dimension(payload.get('gender'))}",
+        f"- 年龄：{_format_persona_dimension(payload.get('age'))}",
+        f"- 城市等级：{_format_persona_dimension(payload.get('city_tier'))}",
+        f"- 消费能力：{_format_persona_dimension(payload.get('buying_power'))}",
+    ])
+
+
+def run_single_note_analysis(note_id):
+    """跑单品分析,输出到 data/summary/note_<id>.txt。已存在则跳过。"""
+    out_path = REPORTS_DIR / f"note_{note_id}.txt"
+    if out_path.exists() and out_path.stat().st_size > 100:
+        log(f"  复用已有报告: {out_path.name}")
+        return out_path
+
+    log(f"  跑单品分析 note_id={note_id} ...")
+    cmd = [sys.executable, "-X", "utf8",
+           str(Path(__file__).parent / "data" / "tools" / "analyze_note.py"),
+           "--note-id", str(note_id), "--no-send", "--brand", "光影行星",
+           "--output", str(out_path)]
+    env = {**os.environ, "PYTHONIOENCODING": "utf-8"}
+    proc = subprocess.run(cmd, capture_output=True, env=env,
+                          encoding="utf-8", errors="replace", timeout=600)
+    if proc.returncode != 0:
+        log(f"  [WARN] note_id={note_id} 分析失败: {proc.stderr[:200]}")
+        return None
+    if out_path.exists() and out_path.stat().st_size > 100:
+        return out_path
+    log(f"  [WARN] note_id={note_id} 未生成报告")
+    return None
+
+
+def read_report(path):
+    if not path or not path.exists():
+        return ""
+    for enc in ("utf-8-sig", "utf-8"):
+        try:
+            return path.read_text(encoding=enc)
+        except UnicodeDecodeError:
+            continue
+    return path.read_text(encoding="utf-8", errors="replace")
+
+
+# LLM 综合 prompt - 单篇笔记四维度分析(逐篇调用,保证深度不打折)
+SINGLE_NOTE_PROMPT = """你是光影行星品牌「{style}」款式的内容种草分析师,聚焦前端表现。
+请对以下笔记从"内容质量 -> 受众匹配 -> 互动效果 -> 评论区信号"四个维度进行深度分析。
+
+【笔记信息】
+{note_info}
+
+【单品分析报告全文】
+{report}
+
+【天猫该款最新人物画像（受众推算基准）】
+{tmall_persona}
+
+说明：天猫画像代表该款实际购买/浏览人群，笔记标题、达人和评论反映内容触达人群。
+请比较两类人群是否吻合，不要把天猫购买人群直接等同于内容触达人群。
+
+【输出要求】
+严格按以下结构输出(Markdown 格式),每项都必须填充具体数据和判断:
+
+### 笔记{idx}：{title}
+
+#### 基础信息
+- 平台/达人：{platform} / {creator}
+- 曝光：{views} | 互动总量：{engagement} | 互动率：{rate}%
+
+#### 1. 标题与内容拆解
+(详细分析,不只列结论)
+- 标题钩子类型(痛点/价值观/利益点/悬念):判断类型 + 解释为什么 + 评估有效性
+- 内容结构:前3秒抓人能力 / 中间痛点铺陈 / 产品品牌植入方式 / 结尾引导,逐项给具体判断
+- 品牌/SKU记忆度:用户看完能否复述品牌或产品名,结合评论区信号佐证
+
+#### 2. 数据表现评级
+(算多个比值,不只一个)
+- 互动率对比平台均值(抖音3-5% / 小红书3-8% / B站3-5%),判断优劣
+- 互动结构:算赞藏比/评赞比/分享点赞比/收藏评论比,每个比值都解读含义,识别认知留存/争议共鸣/实用收藏型
+
+#### 3. 受众画像推算
+- 先引用天猫画像的真实性别、年龄、城市等级和消费力分布，并注明数据日期
+- 再结合标题词、达人人设和评论区用词，判断本篇内容触达人群与天猫购买人群是吻合还是偏离
+- 核心受众标签:年龄/消费场景/兴趣圈层/消费力,每项给判断和依据
+- 天猫画像没有覆盖的维度才允许推断，并明确标注为推断
+
+#### 4. 评论区关键信号
+(每项都给具体词/数字,不要泛泛)
+- 高频实义词TOP5(排除语气词):列出 5 个具体词
+- 购买意向信号("求链接""已下单"等)占比估算:给百分比 + 条数
+- 品牌提及率(主动提到品牌/产品名的比例):给百分比
+- 用户顾虑点TOP3(尺寸/价格/质量等):每条顾虑解释为什么
+- 有无竞品提及:有则列出,无则说明
+
+#### 5. 总结评级与优化建议
+- 最终评级(S/A/B/C):给评级 + 判定依据(哪几条指标支撑这个评级)
+- 是否值得继续投放或放大:给结论 + 理由
+- 针对内容本身的修改建议(非电商承接):至少 4-5 条具体可执行建议,涵盖标题/前3秒/卖点顺序/品牌植入/结尾引导
+
+【评级标准】
+- S:互动率显著高于平台均值,品牌记忆度高,购买意向信号强,受众精准
+- A:互动率达平台均值以上,品牌有记忆,有购买意向信号
+- B:互动率接近均值,品牌记忆模糊,购买意向弱
+- C:互动率低于均值,品牌无记忆,无购买信号,受众错位
+
+【约束】
+- 本篇分析必须详尽(800-1200 字),数据驱动,直接给结论
+- 有天猫画像时，男女比例必须原样引用画像数据；无画像时不得虚构具体比例
+- 高频词TOP5必须给具体词,不要写"无有效数据"
+- 评级严格按标准,不放宽
+- **只输出本篇笔记的分析,不要输出周度总结,不要提及"文件已保存"等元信息。响应正文就是本篇分析本身。**"""
+
+# LLM 综合 prompt - 周度总结(输入所有单篇分析摘要 + 数据)
+WEEKLY_SUMMARY_PROMPT = """你是光影行星品牌「{style}」款式的内容种草分析师。
+请基于以下 {n} 篇笔记的单篇分析结果,输出周度总结。
+
+【本周单篇分析汇总】
+{all_analyses}
+
+【本周投放概览数据】
+- 投放笔记数：{n} 篇
+- 涉及平台：{platforms}
+- 总曝光：{total_views} / 总互动：{total_engagement}
+
+【输出要求】
+严格按以下结构输出(Markdown 格式):
+
+## 三、周度总结与建议
+
+### 1. 整体评级分布
+- 各评级(S/A/B/C)笔记数及占比
+- 本周整体内容种草效率判断
+
+### 2. 值得继续投放/放大的笔记
+[列出 S/A 级笔记,给出可复用经验,每篇 2-3 句]
+
+### 3. 需要规避的问题
+[列出 B/C 级笔记的问题,每篇 2-3 句,给规避建议]
+
+### 4. 下周内容优化清单
+- 标题钩子方向:[基于本周表现给具体建议]
+- 内容结构:[前3秒/中间铺陈/品牌植入/结尾引导]
+- 达人选择:[优先/规避类型]
+- 评论区运营:[具体话术]
+
+【约束】
+- 周度总结 800-1500 字
+- 每条建议须具体可执行,避免"加强""优化"等空泛词汇
+- **只输出周度总结部分(从"## 三"开始),不要重复输出单篇分析,不要提及"文件已保存"等元信息。响应正文就是周度总结本身。**"""
+
+
+def generate_summary(style, notes):
+    """调 LLM 综合评级,返回汇总文本(两步:逐篇分析 + 周度总结)"""
+    system = "你是光影行星品牌的内容种草分析师,擅长从内容质量、受众匹配、互动效果、评论区信号四个维度分析笔记表现,输出可落地的优化建议。"
+
+    tmall_persona = get_latest_tmall_persona(style)
+    tmall_persona_text = format_tmall_persona(tmall_persona)
+    if tmall_persona:
+        log(f"  天猫人物画像: {tmall_persona.get('collect_date')} / item_id={tmall_persona.get('item_id')}")
+    else:
+        log(f"  [WARN] {style} 暂无天猫人物画像")
+
+    # ---- 第一步:逐篇生成单篇四维度分析 ----
+    single_analyses = []
+    for i, n in enumerate(notes, 1):
+        report = read_report(n.get("report_path"))
+        views = n.get('view_count') or '无'
+        likes = n.get('like_count') or 0
+        cmts = n.get('comment_count') or 0
+        favs = n.get('favorite_count') or 0
+        shares = n.get('share_count') or 0
+        engagement = (likes or 0) + (favs or 0) + (cmts or 0) + (shares or 0)
+        if isinstance(views, (int, float)) and views > 0:
+            rate = round(engagement * 100 / views, 2)
+        else:
+            rate = "无法计算(曝光未采集)"
+
+        note_info = (
+            f"标题: {n['title']}\n"
+            f"平台: {n['platform']} / 达人: {n.get('creator_name', '?')}\n"
+            f"曝光: {views} / 点赞: {likes} / 评论: {cmts} / 收藏: {favs} / 分享: {shares}\n"
+            f"互动总量: {engagement} / 互动率: {rate}%"
+        )
+        prompt = SINGLE_NOTE_PROMPT.format(
+            style=style, idx=i, title=n['title'],
+            note_info=note_info, report=report, tmall_persona=tmall_persona_text,
+            platform=n['platform'], creator=n.get('creator_name', '?'),
+            views=views, engagement=engagement, rate=rate,
+        )
+        log(f"    调 LLM 单篇分析 {i}/{len(notes)}: {n['title'][:30]}...")
+        analysis = call_hermes_analyzer(system, prompt)
+        if analysis:
+            single_analyses.append(analysis)
+        else:
+            single_analyses.append(
+                f"### 笔记{i}：{n['title']}\n**基础信息**\n"
+                f"- 平台/达人：{n['platform']} / {n.get('creator_name', '?')}\n"
+                f"- 曝光：{views} | 互动总量：{engagement} | 互动率：{rate}%\n\n"
+                f"*(LLM 分析失败,数据仅供参考)*"
+            )
+
+    # ---- 拼接投放概览 ----
+    platform_counts = {}
+    total_views = 0
+    total_engagement = 0
+    for n in notes:
+        p = n['platform']
+        platform_counts[p] = platform_counts.get(p, 0) + 1
+        v = n.get('view_count')
+        if isinstance(v, (int, float)) and v > 0:
+            total_views += v
+        likes = n.get('like_count') or 0
+        cmts = n.get('comment_count') or 0
+        favs = n.get('favorite_count') or 0
+        shares = n.get('share_count') or 0
+        total_engagement += (likes + cmts + favs + shares)
+    platforms_str = " / ".join(f"{p} {c} 篇" for p, c in sorted(platform_counts.items()))
+    views_str = f"{total_views:,}" if total_views else "部分未采集"
+
+    overview = (
+        f"# {style} 周笔记汇总（{TIME_LABEL}）\n\n"
+        f"## 一、本周投放概览\n\n"
+        f"- 投放笔记数：{len(notes)} 篇\n"
+        f"- 涉及平台：{platforms_str}\n"
+        f"- 总曝光 / 总互动：{views_str} / {total_engagement:,}\n"
+        f"- 整体表现：[见下方周度总结]\n\n"
+        f"---\n\n"
+        f"## 二、单篇笔记四维度分析\n\n"
+    )
+
+    # ---- 第二步:生成周度总结 ----
+    all_analyses_text = "\n\n---\n\n".join(single_analyses)
+    summary_prompt = WEEKLY_SUMMARY_PROMPT.format(
+        style=style, n=len(notes),
+        all_analyses=all_analyses_text,
+        platforms=platforms_str,
+        total_views=views_str,
+        total_engagement=f"{total_engagement:,}",
+    )
+    log(f"    调 LLM 周度总结...")
+    weekly_summary = call_hermes_analyzer(system, summary_prompt)
+    if not weekly_summary:
+        weekly_summary = "## 三、周度总结与建议\n\n*(LLM 周度总结生成失败)*"
+
+    return overview + all_analyses_text + "\n\n---\n\n" + weekly_summary
+
+
+def create_feishu_doc(style, summary):
+    """创建飞书文档,返回 url"""
+    doc_title = f"{style} 周笔记汇总（{TIME_LABEL}）"
+    md_file = PATHS.tmp_root / "weekly_summary_doc.md"
+    md_file.parent.mkdir(parents=True, exist_ok=True)
+    md_file.write_text(summary, encoding="utf-8")
+
+    cmd = [LARK, "docs", "+create",
+           "--as", "user", "--format", "json",
+           "--title", doc_title,
+           "--doc-format", "markdown",
+           "--content", f"@{md_file.name}"]
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, env=env,
+                          encoding="utf-8", errors="replace")
+    text = proc.stdout or proc.stderr
+    start = text.find("{")
+    if start < 0:
+        return None, f"创建失败: {text[:300]}"
+    resp = json.loads(text[start:])
+    if not resp.get("ok"):
+        return None, f"创建失败: {resp.get('error')}"
+    url = resp.get("data", {}).get("document", {}).get("url", "")
+    return url, None
+
+
+def save_to_db(style, notes, summary, doc_url, doc_title, status):
+    """把周汇总结果写入 PG cmt_weekly_summary(UNIQUE(style_id, week_start),同款同周重跑时 UPDATE)"""
+    if dry_run_global:
+        return False
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_conn
+    try:
+        with get_conn() as conn:
+            cur = conn.cursor()
+            # 查 style_id
+            cur.execute("SELECT id FROM cmt_styles WHERE name = %s", (style,))
+            row = cur.fetchone()
+            if not row:
+                log(f"  [DB] 款式 {style} 不在 cmt_styles,跳过入库")
+                return False
+            style_id = row[0]
+
+            # 算聚合数据
+            total_views = sum((n.get('view_count') or 0) for n in notes) or None
+            total_comments = sum((n.get('comment_count') or 0) for n in notes)
+            total_engagement = sum(
+                ((n.get('like_count') or 0) + (n.get('favorite_count') or 0) +
+                 (n.get('comment_count') or 0) + (n.get('share_count') or 0))
+                for n in notes) or None
+
+            # UPSERT (style_id, week_start)
+            cur.execute("""
+                INSERT INTO cmt_weekly_summary
+                    (style_id, week_start, week_end, time_label,
+                     note_count, total_comments, total_views, total_engagement,
+                     doc_url, doc_title, summary_chars, status,
+                     created_at, updated_at)
+                VALUES
+                    (%s, %s, %s, %s,
+                     %s, %s, %s, %s,
+                     %s, %s, %s, %s,
+                     CURRENT_TIMESTAMP, CURRENT_TIMESTAMP)
+                ON CONFLICT (style_id, week_start) DO UPDATE SET
+                    note_count = EXCLUDED.note_count,
+                    total_comments = EXCLUDED.total_comments,
+                    total_views = EXCLUDED.total_views,
+                    total_engagement = EXCLUDED.total_engagement,
+                    doc_url = EXCLUDED.doc_url,
+                    doc_title = EXCLUDED.doc_title,
+                    summary_chars = EXCLUDED.summary_chars,
+                    status = EXCLUDED.status,
+                    updated_at = CURRENT_TIMESTAMP
+            """, (style_id, WEEK_START, WEEK_END, TIME_LABEL,
+                   len(notes), total_comments, total_views, total_engagement,
+                   doc_url, doc_title, len(summary) if summary else None, status))
+            conn.commit()
+        log(f"  ✅ 已入库 cmt_weekly_summary (style_id={style_id}, week={WEEK_START})")
+        return True
+    except Exception as exc:
+        log(f"  [DB ERR] {style} 入库失败: {exc}")
+        return False
+
+
+def write_to_target_table(bt, tid, time_fid, summary_fid, style, doc_url, dry_run):
+    """把文档链接写入生命进程表(已有同时间记录则更新,否则新建)"""
+    md_link = f"[{style} 周笔记汇总（{TIME_LABEL}）]({doc_url})"
+    payload = {time_fid: TIME_LABEL, summary_fid: md_link}
+    pf = PATHS.tmp_root / "weekly_payload.json"
+    pf.write_text(json.dumps(payload, ensure_ascii=False), encoding="utf-8")
+
+    if dry_run:
+        log(f"  [DRY] 跳过写入 {bt}/{tid}")
+        return True
+
+    # 先查是否已有同时间的记录
+    existing_rid = None
+    list_cmd = ["base", "+record-list",
+                "--base-token", bt, "--table-id", tid,
+                "--as", "user", "--format", "json", "--page-size", "100"]
+    list_resp = call_lark_json(list_cmd)
+    if list_resp.get("ok"):
+        rid_list = list_resp.get("data", {}).get("record_id_list", [])
+        data_rows = list_resp.get("data", {}).get("data", [])
+        # 找字段索引: 用 field_id_list 定位 time_fid 的列号
+        fid_list = list_resp.get("data", {}).get("field_id_list", [])
+        time_col = fid_list.index(time_fid) if time_fid in fid_list else 0
+        for rid, row in zip(rid_list, data_rows):
+            row_time = row[time_col] if len(row) > time_col else ""
+            # 时间字段是 select, 返回值可能是 ["2026-06-29~07-05"]
+            if isinstance(row_time, list):
+                row_time = row_time[0] if row_time else ""
+            if str(row_time) == TIME_LABEL:
+                existing_rid = rid
+                break
+
+    if existing_rid:
+        cmd = ["base", "+record-upsert",
+               "--base-token", bt, "--table-id", tid,
+               "--as", "user", "--format", "json",
+               "--record-id", existing_rid,
+               "--json", f"@{pf.name}"]
+        log(f"  更新已有记录 {existing_rid}")
+    else:
+        cmd = ["base", "+record-upsert",
+               "--base-token", bt, "--table-id", tid,
+               "--as", "user", "--format", "json",
+               "--json", f"@{pf.name}"]
+        log(f"  新建记录")
+
+    resp = call_lark_json(cmd)
+    return resp.get("ok", False)
+
+
+def process_one_style(style, index_info, dry_run):
+    """处理一个款:跑单品分析+综合+创建文档+写表"""
+    bt, tid, url = index_info
+    log(f"\n========== {style} ==========")
+    log(f"  生命进程表: {bt}/{tid}")
+
+    def _save(notes_list, summary, doc_url, doc_title, status):
+        """统一入库辅助"""
+        save_to_db(style, notes_list, summary, doc_url, doc_title, status)
+
+    # 动态读字段
+    time_fid, summary_fid = find_target_fields(bt, tid)
+    if not time_fid or not summary_fid:
+        log(f"  [ERR] 找不到 时间/笔记分析汇总 字段,跳过")
+        _save([], None, None, None, "no_fields")
+        return {"style": style, "status": "no_fields", "notes": 0}
+
+    # 拿上周笔记
+    notes = get_last_week_notes(style)
+    log(f"  上周笔记: {len(notes)} 篇")
+    if not notes:
+        log(f"  [SKIP] 上周无笔记")
+        _save([], None, None, None, "no_notes")
+        return {"style": style, "status": "no_notes", "notes": 0}
+
+    # 跑单品分析
+    for n in notes:
+        n["report_path"] = run_single_note_analysis(n["id"])
+
+    # 过滤掉没生成报告的
+    valid_notes = [n for n in notes if n.get("report_path")]
+    if not valid_notes:
+        log(f"  [ERR] 没有有效报告,跳过")
+        _save(notes, None, None, None, "no_reports")
+        return {"style": style, "status": "no_reports", "notes": len(notes)}
+
+    # LLM 综合
+    log(f"  调 LLM 综合 {len(valid_notes)} 篇报告...")
+    summary = generate_summary(style, valid_notes)
+    if not summary:
+        log(f"  [ERR] LLM 综合失败")
+        _save(notes, None, None, None, "llm_failed")
+        return {"style": style, "status": "llm_failed", "notes": len(notes)}
+
+    # 保存本地
+    out_path = REPORTS_DIR / f"{style}_周笔记汇总_{TIME_LABEL}.txt"
+    out_path.write_text(summary, encoding="utf-8")
+    log(f"  本地副本: {out_path.name} ({len(summary)} 字符)")
+
+    if dry_run_global:
+        log(f"  [DRY] 跳过飞书文档创建和写入")
+        _save(notes, summary, None, None, "dry_run")
+        return {"style": style, "status": "dry_run", "notes": len(notes), "chars": len(summary)}
+
+    # 创建飞书文档
+    doc_url, err = create_feishu_doc(style, summary)
+    if not doc_url:
+        log(f"  [ERR] {err}")
+        _save(notes, summary, None, None, "doc_failed")
+        return {"style": style, "status": "doc_failed", "notes": len(notes)}
+
+    log(f"  飞书文档: {doc_url}")
+
+    # 写入生命进程表
+    ok = write_to_target_table(bt, tid, time_fid, summary_fid, style, doc_url, dry_run_global)
+
+    # 写入 PG cmt_weekly_summary
+    doc_title = f"{style} 周笔记汇总（{TIME_LABEL}）"
+    _save(notes, summary, doc_url, doc_title, "ok" if ok else "feishu_write_failed")
+
+    if ok:
+        log(f"  ✅ 已写入生命进程表")
+        return {"style": style, "status": "ok", "notes": len(notes),
+                "doc_url": doc_url, "chars": len(summary)}
+    log(f"  [ERR] 写入表失败")
+    return {"style": style, "status": "write_failed", "notes": len(notes)}
+
+
+def main():
+    parser = argparse.ArgumentParser(description="通用周笔记汇总:对所有上周有笔记的款生成飞书文档并写入生命进程表")
+    parser.add_argument("--style", help="只跑指定款式名")
+    parser.add_argument("--dry-run", action="store_true", help="不写飞书,只生成本地报告")
+    parser.add_argument("--max-workers", type=int, default=4,
+                        help="款之间并发数(默认 4)")
+    args = parser.parse_args()
+
+    global dry_run_global
+    dry_run_global = args.dry_run
+
+    log(f"时间范围: {TIME_LABEL}")
+
+    # 读索引表
+    index_styles = load_index_styles()
+    if not index_styles:
+        log("索引表无有效数据,退出")
+        return 1
+
+    # PG 查上周有笔记的款
+    from gyxx_flow.modules.content_marketing.runtime.data.tools.db import get_conn
+    with get_conn() as conn:
+        cur = conn.cursor()
+        cur.execute("""
+            SELECT s.name, COUNT(n.id)
+            FROM cmt_styles s
+            JOIN cmt_notes n ON n.style_id = s.id AND n.self_operated = FALSE
+            WHERE n.publish_time IS NOT NULL
+            AND n.publish_time >= %s AND n.publish_time < %s
+            GROUP BY s.name
+            ORDER BY s.name
+        """, (WEEK_START, WEEK_END + " 23:59:59"))
+        pg_styles = [(r[0], r[1]) for r in cur.fetchall()]
+
+    log(f"PG 上周有笔记的款: {len(pg_styles)} 个")
+
+    # 取交集:PG 里有 + 索引表里有飞书表地址
+    to_run = []
+    for name, note_count in pg_styles:
+        if args.style and name != args.style:
+            continue
+        if name in index_styles:
+            to_run.append((name, index_styles[name], note_count))
+        else:
+            log(f"  [SKIP] {name}: 索引表无'每周笔记分析'地址")
+
+    log(f"待跑款式: {len(to_run)} 个")
+    if not to_run:
+        return 0
+
+    # 并发跑款
+    import concurrent.futures
+    log(f"并发跑 {len(to_run)} 款, max_workers={args.max_workers}")
+    results = []
+    with concurrent.futures.ThreadPoolExecutor(max_workers=args.max_workers) as executor:
+        future_to_style = {
+            executor.submit(process_one_style, name, info, args.dry_run): (name, ncount)
+            for name, info, ncount in to_run
+        }
+        for future in concurrent.futures.as_completed(future_to_style):
+            name, ncount = future_to_style[future]
+            try:
+                r = future.result()
+                r["note_count"] = ncount
+                results.append(r)
+                log(f"  ✅ 完成: {name} -> {r['status']}")
+            except Exception as exc:
+                log(f"  [ERR] {name} 异常: {exc}")
+                results.append({"style": name, "status": "exception",
+                                "error": str(exc), "note_count": ncount})
+
+    # 按款式名排序输出
+    results.sort(key=lambda r: r.get("style", ""))
+
+    # 汇总
+    log(f"\n{'=' * 50}")
+    log(f"全部完成,汇总:")
+    for r in results:
+        log(f"  {r['style']}: {r['status']} (笔记 {r.get('note_count', 0)} 篇)")
+    log(f"{'=' * 50}")
+
+    # 保存结果
+    out = REPORTS_DIR / f"_weekly_summary_all_{TIME_LABEL}.json"
+    out.write_text(json.dumps(results, ensure_ascii=False, indent=2), encoding="utf-8")
+    log(f"结果已存: {out}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/weekly_summary_xingyun2.py b/src/gyxx_flow/modules/content_marketing/runtime/weekly_summary_xingyun2.py
new file mode 100644
index 0000000..500ff6e
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/weekly_summary_xingyun2.py
@@ -0,0 +1,244 @@
+"""综合 3 篇笔记分析报告,生成星云2 周笔记汇总 -> 写入飞书表"""
+import sys
+import json
+import os
+import subprocess
+from pathlib import Path
+from datetime import datetime
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+
+sys.stdout.reconfigure(encoding='utf-8')
+
+from gyxx_flow.modules.content_marketing.runtime.data.tools.analyze_comments import call_hermes_analyzer
+
+LARK = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK):
+    LARK = "lark-cli.cmd"
+
+TARGET_BT = "BJaWbLEOKaHONMshOidcxvsKngh"
+TARGET_TID = "tbltzBjbdiW4GrUd"
+TIME_FIELD = "fldpKSfCCC"
+SUMMARY_FIELD = "fldro5fLrb"
+
+SUMMARY_PROMPT = """你是光影行星品牌「星云2」款式的周投放复盘分析师。
+请基于本周 3 篇笔记的单品分析报告,生成周笔记汇总报告。
+
+【本周笔记清单】
+{notes}
+
+【输出要求】
+按以下结构输出(Markdown 格式,适合贴到飞书文本字段):
+
+# 星云2 周笔记汇总（{time_range}）
+
+## 一、本周投放概览
+- 投放笔记数：{n} 篇
+- 涉及平台：[列出]
+- 总曝光 / 总互动：[汇总数据]
+- 整体表现：[一句话定性]
+
+## 二、投放效果多维分析
+
+### 1. 流量效率
+- 各平台曝光表现:[逐平台分析,抖音/小红书/B站]
+- 互动率结构:[点赞率/收藏率/评论率/分享率,识别认知留存型/社交裂变型/搜索截流型]
+- 流量是否匹配投放预期:[判断 + 理由]
+
+### 2. 互动质量
+- 评论区有效评论占比:[统计]
+- 评论质量分层:[有效种草/连带兴趣/无效互动 占比]
+- 用户决策阶段:[规格确认/价值判断/竞品对比/冲动下单/售后反馈,哪个阶段为主]
+
+### 3. 品牌词沉淀
+- 品牌词"光影行星"提及情况:[条数 + 上下文]
+- 款式词"星云2"提及情况
+- 我方产品相关评论占比
+- 品牌词是否被自然讨论:[是/否 + 理由]
+
+### 4. 卖点传达
+- 笔记主推卖点 vs 评论区讨论卖点:[一致性分析]
+- 卖点偏离方向:[若偏离,说明偏向哪里]
+- 三大核心卖点(磁吸扣/分层收纳/容量/颜值/重量)在评论区的渗透情况
+
+### 5. 达人匹配度
+- 各达人与星云2 目标客群重合度
+- 内容风格是否适合承载星云2 核心卖点
+- 哪些达人值得复用,哪些应淘汰
+
+### 6. 负面舆情
+- 评论区负面集中点:[如自重/价格/做工]
+- 影响面大小
+- 是否需要控评
+
+### 7. 平台差异
+- 同主题不同平台表现对比(如有)
+- 各平台用户关注点差异
+- 平台化适配建议
+
+## 三、单篇笔记提炼（精简）
+
+### 笔记1：{title1}
+- 平台/达人：{platform1} / {creator1}
+- 核心数据：曝光 {views1} | 互动率 [X%]
+- 评级：[优秀/良好/不好]
+- 一句话提炼：[20-30 字,点出成败关键]
+- 值得推荐：[是/否]
+
+### 笔记2：{title2}
+[同上格式]
+
+### 笔记3：{title3}
+[同上格式]
+
+## 四、值得推荐的笔记
+[列出评级良好及以上的笔记,给出可复用经验,每篇 2-3 句]
+
+## 五、需要规避的问题
+[列出"不好"评级笔记的问题,每篇 2-3 句,给规避建议]
+
+## 六、下周投放优化清单
+- 内容方向：[基于本周反馈的具体建议]
+- 达人选择：[优先 / 规避类型]
+- 评论区运营：[具体话术]
+- 投放策略：[预算 / 流量助推 / 平台侧重]
+
+【评级标准】
+- 优秀：互动率高于行业基准(抖音≥6% / 小红书≥8% / B站≥5%)、评论中有产品讨论、品牌词被自然提及
+- 良好：互动率达标但评论质量一般,或评论质量好但互动率刚达标
+- 不好：互动率不达标、评论区无产品讨论、品牌词未沉淀
+
+【约束】
+- 第二部分"投放效果多维分析"必须详尽,每子模块给具体数据和判断
+- 第三部分"单篇笔记提炼"精简,每篇 5-6 行内
+- 直接给结论,不写"根据数据显示"
+- 每条建议须具体可执行
+- 评级严格按标准,不放宽
+- 整体篇幅 1500-2500 字(单篇精简,多维分析详尽)"""
+
+
+def read_report(path: Path) -> str:
+    if not path.exists():
+        return ""
+    for enc in ("utf-8-sig", "utf-8"):
+        try:
+            return path.read_text(encoding=enc)
+        except UnicodeDecodeError:
+            continue
+    return path.read_text(encoding="utf-8", errors="replace")
+
+
+def main():
+    reports_dir = PATHS.exports_root / "summary"
+    notes = [
+        {"id": 1662, "platform": "bilibili", "creator": "裴念南",
+         "title": "SwitchLite开箱验机教程📦➕全套配件分享✨",
+         "views": 1318, "likes": 64, "cmts": 2, "favs": 21,
+         "report": read_report(reports_dir / "note_1662.txt")},
+        {"id": 1663, "platform": "xiaohongshu", "creator": "不爱出门的阿宅",
+         "title": "六月switch配件购买分享～实用又好看",
+         "views": 2224, "likes": 140, "cmts": 6, "favs": 59,
+         "report": read_report(reports_dir / "note_1663.txt")},
+        {"id": 1664, "platform": "douyin", "creator": "不爱出门的阿宅",
+         "title": "六月switch配件购买分享～实用又好看",
+         "views": None, "likes": 63, "cmts": 7, "favs": 16,
+         "report": read_report(reports_dir / "note_1664.txt")},
+    ]
+
+    # 拼笔记清单
+    notes_block = []
+    for i, n in enumerate(notes, 1):
+        notes_block.append(
+            f"### 笔记{i}: {n['title']}\n"
+            f"- 平台: {n['platform']}\n- 达人: {n['creator']}\n"
+            f"- 曝光: {n['views'] or '无'} / 点赞: {n['likes']} / 评论: {n['cmts']} / 收藏: {n['favs']}\n\n"
+            f"单品分析报告全文:\n{n['report']}\n"
+        )
+    notes_text = "\n---\n".join(notes_block)
+
+    # 填 prompt
+    prompt = SUMMARY_PROMPT.format(notes=notes_text,
+                                    time_range="2026-06-29~07-05",
+                                    n=len(notes),
+                                    title1=notes[0]['title'], platform1=notes[0]['platform'],
+                                    creator1=notes[0]['creator'], views1=notes[0]['views'] or '无',
+                                    likes1=notes[0]['likes'], cmts1=notes[0]['cmts'], favs1=notes[0]['favs'],
+                                    title2=notes[1]['title'], platform2=notes[1]['platform'],
+                                    creator2=notes[1]['creator'], views2=notes[1]['views'] or '无',
+                                    likes2=notes[1]['likes'], cmts2=notes[1]['cmts'], favs2=notes[1]['favs'],
+                                    title3=notes[2]['title'], platform3=notes[2]['platform'],
+                                    creator3=notes[2]['creator'], views3=notes[2]['views'] or '无',
+                                    likes3=notes[2]['likes'], cmts3=notes[2]['cmts'], favs3=notes[2]['favs'])
+
+    print("调用 LLM 综合评级...")
+    summary = call_hermes_analyzer("你是光影行星品牌的周投放复盘分析师,擅长综合多篇笔记分析报告,生成款式级别的多维投放复盘。", prompt)
+    print(f"汇总报告生成完成,长度 {len(summary)} 字符")
+
+    # 保存本地副本
+    time_label = "2026-06-29~07-05"
+    style_label = "星云2"
+    out_path = reports_dir / f"{style_label}_周笔记汇总_{time_label}.txt"
+    out_path.write_text(summary, encoding="utf-8")
+    print(f"本地副本: {out_path}")
+
+    # 创建飞书文档(markdown 格式)
+    # LLM 输出的 summary 自身已以 "# 标题" 开头,不需要额外加;--title 会被 lark-cli 自动设为文档标题
+    print("创建飞书文档...")
+    doc_title = f"{style_label} 周笔记汇总（{time_label}）"
+    md_file = PATHS.tmp_root / "weekly_summary_doc.md"
+    md_file.parent.mkdir(parents=True, exist_ok=True)
+    md_file.write_text(summary, encoding="utf-8")
+
+    cmd = [LARK, "docs", "+create",
+           "--as", "user",
+           "--format", "json",
+           "--title", doc_title,
+           "--doc-format", "markdown",
+           "--content", f"@{md_file.name}"]
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, env=env, encoding="utf-8", errors="replace")
+    text = proc.stdout or proc.stderr
+    start = text.find("{")
+    if start < 0:
+        print(f"创建文档失败: {text[:500]}")
+        return 1
+    import json as _json
+    resp = _json.loads(text[start:])
+    if not resp.get("ok"):
+        print(f"创建文档失败: {resp.get('error')}")
+        return 1
+    doc_url = resp.get("data", {}).get("document", {}).get("url", "")
+    doc_token = resp.get("data", {}).get("document", {}).get("document_id", "")
+    print(f"飞书文档已创建: {doc_url}")
+    print(f"  doc_token: {doc_token}")
+
+    # 在飞书表里写 markdown 链接(标题+URL)
+    md_link = f"[{doc_title}]({doc_url})"
+    payload = {
+        TIME_FIELD: time_label,
+        SUMMARY_FIELD: md_link,
+    }
+    pf = PATHS.tmp_root / "weekly_summary_payload.json"
+    pf.write_text(_json.dumps(payload, ensure_ascii=False), encoding="utf-8")
+
+    cmd = [LARK, "base", "+record-upsert",
+           "--base-token", TARGET_BT,
+           "--table-id", TARGET_TID,
+           "--as", "user",
+           "--format", "json",
+           "--json", f"@{pf.name}"]
+    proc = subprocess.run(cmd, capture_output=True, env=env, encoding="utf-8", errors="replace")
+    text = proc.stdout or proc.stderr
+    start = text.find("{")
+    if start < 0:
+        print(f"插入表失败: {text[:300]}")
+        return 1
+    resp = _json.loads(text[start:])
+    if resp.get("ok"):
+        print(f"已写入飞书表: 时间={time_label}, 笔记分析汇总=文档链接")
+        return 0
+    print(f"插入表失败: {resp.get('error')}")
+    return 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/xiaohongshu_comment_scraper.py b/src/gyxx_flow/modules/content_marketing/runtime/xiaohongshu_comment_scraper.py
new file mode 100644
index 0000000..73889cc
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/xiaohongshu_comment_scraper.py
@@ -0,0 +1,695 @@
+import argparse
+import csv
+import json
+import re
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+from typing import Any
+from urllib.parse import parse_qsl, urlencode, urlsplit, urlunsplit
+
+from playwright.sync_api import Page, TimeoutError as PlaywrightTimeoutError
+from scrapling.fetchers import DynamicSession
+
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+STATE_DIR = PATHS.state_root
+NOTES_DIR = PATHS.raw_root / "notes" / "xiaohongshu"
+COOKIE_FILE = PATHS.browser_cookie_file
+PROFILE_DIR = PATHS.browser_profile_dir
+
+COMMENT_API_MARKS = (
+    "/api/sns/web/v2/comment/page",
+    "/api/sns/web/v2/comment/sub/page",
+)
+
+
+def log(message: str) -> None:
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {message}", flush=True)
+
+
+def load_cookies() -> list[dict[str, Any]]:
+    if not COOKIE_FILE.exists():
+        return []
+    try:
+        data = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+        return data if isinstance(data, list) else []
+    except json.JSONDecodeError:
+        log(f"Cookie file is invalid, ignoring: {COOKIE_FILE}")
+        return []
+
+
+def restore_cookies(page: Page) -> None:
+    cookies = load_cookies()
+    if not cookies:
+        return
+    try:
+        page.context.add_cookies(cookies)
+        log(f"Loaded cookies: {COOKIE_FILE}")
+    except Exception as exc:
+        log(f"Failed to load cookies, will continue with browser profile: {exc}")
+
+
+def save_state(page: Page) -> None:
+    STATE_DIR.mkdir(parents=True, exist_ok=True)
+    COOKIE_FILE.write_text(
+        json.dumps(page.context.cookies(), ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    PATHS.browser_storage_state_file.write_text(
+        json.dumps(page.context.storage_state(), ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    log(f"Saved cookies: {COOKIE_FILE}")
+
+
+def note_id_from_url(url: str) -> str:
+    match = re.search(r"/explore/([0-9a-fA-F]+)", url)
+    return match.group(1) if match else ""
+
+
+def safe_filename(value: str) -> str:
+    value = re.sub(r"[<>:\"/\\\\|?*\\s]+", "_", value).strip("_")
+    return value[:90] or "xiaohongshu_comments"
+
+
+def ts_to_iso(value: Any) -> str:
+    if not isinstance(value, (int, float)) or not value:
+        return ""
+    timestamp = value / 1000 if value > 10_000_000_000 else value
+    try:
+        return datetime.fromtimestamp(timestamp).isoformat(timespec="seconds")
+    except Exception:
+        return str(value)
+
+
+def parse_comment(comment: dict[str, Any], source_url: str, level: str, parent_id: str = "") -> dict[str, Any]:
+    user = comment.get("user_info") or comment.get("user") or {}
+    comment_id = comment.get("id") or comment.get("comment_id") or ""
+    create_time = comment.get("create_time") or comment.get("time")
+    return {
+        "source_url": source_url,
+        "level": level,
+        "comment_id": comment_id,
+        "reply_to_comment_id": parent_id or comment.get("target_comment", {}).get("id", ""),
+        "nickname": user.get("nickname") or "",
+        "user_id": user.get("user_id") or user.get("id") or "",
+        "content": comment.get("content") or comment.get("text") or "",
+        "like_count": comment.get("like_count") or 0,
+        "reply_count": comment.get("sub_comment_count") or comment.get("sub_comment_total") or 0,
+        "ip_location": comment.get("ip_location") or comment.get("ip_label") or "",
+        "created_at": ts_to_iso(create_time),
+        "raw_create_time": create_time or "",
+    }
+
+
+def merge_comment(
+    row: dict[str, Any],
+    comments_by_id: dict[str, dict[str, Any]],
+) -> None:
+    comment_id = str(row.get("comment_id") or f"{row.get('level')}:{len(comments_by_id)}")
+    comments_by_id[comment_id] = row
+
+
+def collect_from_payload(
+    payload: dict[str, Any],
+    source_url: str,
+    comments_by_id: dict[str, dict[str, Any]],
+    stats: dict[str, Any],
+) -> None:
+    if not isinstance(payload, dict):
+        return
+    data = payload.get("data") if isinstance(payload.get("data"), dict) else payload
+    api_url = str(payload.get("__source_url", ""))
+    is_sub = "/comment/sub/page" in api_url
+
+    if not is_sub:
+        stats["api_total"] = data.get("total", stats.get("api_total"))
+        stats["last_has_more"] = data.get("has_more", stats.get("last_has_more"))
+        stats["last_cursor"] = data.get("cursor", stats.get("last_cursor"))
+
+    comments = data.get("comments") or data.get("comment_list") or []
+    if not isinstance(comments, list):
+        return
+
+    parent_id = ""
+    if is_sub:
+        params = dict(parse_qsl(urlsplit(api_url).query))
+        parent_id = params.get("root_comment_id") or params.get("comment_id") or params.get("target_comment_id") or ""
+
+    for comment in comments:
+        if not isinstance(comment, dict):
+            continue
+        row = parse_comment(comment, source_url, "reply" if is_sub else "comment", parent_id)
+        merge_comment(row, comments_by_id)
+
+        sub_comments = comment.get("sub_comments") or []
+        if isinstance(sub_comments, list):
+            for sub in sub_comments:
+                if isinstance(sub, dict):
+                    sub_row = parse_comment(sub, source_url, "reply", str(row.get("comment_id") or ""))
+                    merge_comment(sub_row, comments_by_id)
+
+
+def setup_response_capture(page: Page, source_url: str) -> tuple[dict[str, dict[str, Any]], dict[str, Any]]:
+    comments_by_id: dict[str, dict[str, Any]] = {}
+    stats: dict[str, Any] = {"api_hits": 0, "api_total": None, "last_has_more": None, "last_cursor": None}
+
+    def handle_response(response: Any) -> None:
+        url = response.url
+        if not any(mark in url for mark in COMMENT_API_MARKS):
+            return
+        try:
+            payload = response.json()
+        except Exception:
+            return
+        if isinstance(payload, dict):
+            payload["__source_url"] = url
+        stats["api_hits"] += 1
+        if "/comment/page" in url and not stats.get("comment_url_template"):
+            stats["comment_url_template"] = url
+        if "/comment/sub/page" in url and not stats.get("sub_url_template"):
+            stats["sub_url_template"] = url
+        collect_from_payload(payload, source_url, comments_by_id, stats)
+
+    page.on("response", handle_response)
+    return comments_by_id, stats
+
+
+def click_if_visible(page: Page, text: str, timeout: int = 1200) -> bool:
+    try:
+        locator = page.get_by_text(text, exact=False).first
+        if locator.count():
+            locator.click(timeout=timeout)
+            page.wait_for_timeout(500)
+            return True
+    except Exception:
+        return False
+    return False
+
+
+def dismiss_popups(page: Page) -> None:
+    for text in ("我知道了", "稍后再说", "暂不登录", "取消", "跳过"):
+        click_if_visible(page, text)
+    page.evaluate(
+        """
+        () => {
+          const close = [...document.querySelectorAll('button, div, span')]
+            .filter((el) => {
+              const t = (el.innerText || '').trim();
+              const r = el.getBoundingClientRect();
+              return ['×', 'x', 'X'].includes(t) &&
+                r.width > 0 && r.width <= 52 &&
+                r.height > 0 && r.height <= 52 &&
+                r.top >= 0 && r.top < window.innerHeight * 0.75;
+            })[0];
+          close?.click();
+        }
+        """
+    )
+
+
+def maybe_wait_for_login(page: Page, login_timeout: int) -> None:
+    deadline = time.time() + login_timeout
+    while time.time() < deadline:
+        try:
+            body = page.locator("body").inner_text(timeout=5000)
+        except Exception:
+            body = ""
+        if "扫码登录" not in body and "登录后" not in body and "验证码" not in body:
+            return
+        log("Xiaohongshu may require login. Please finish login in the opened Chrome window.")
+        page.wait_for_timeout(3000)
+
+
+def is_logged_in(page: Page) -> bool:
+    """小红书登录态靠 web_session cookie,有它才能采全量评论。"""
+    for c in page.context.cookies():
+        if c.get("name") == "web_session" and c.get("value"):
+            return True
+    return False
+
+
+def expand_visible_replies(page: Page) -> int:
+    try:
+        clicked = page.evaluate(
+            """
+            () => {
+              const visible = (el) => {
+                const r = el.getBoundingClientRect();
+                const style = window.getComputedStyle(el);
+                return r.width > 0 && r.height > 0 &&
+                  r.bottom > 0 && r.top < window.innerHeight &&
+                  style.visibility !== 'hidden' &&
+                  style.display !== 'none' &&
+                  style.opacity !== '0';
+              };
+              const candidates = [...document.querySelectorAll('button, div, span')]
+                .filter((el) => {
+                  const text = (el.innerText || '').trim();
+                  return visible(el) &&
+                    text.length <= 40 &&
+                    (/展开.*回复/.test(text) || /查看.*回复/.test(text) || /共.*条回复/.test(text) || /还有.*回复/.test(text));
+                });
+              let count = 0;
+              for (const el of candidates.slice(0, 8)) {
+                el.click();
+                count++;
+              }
+              return count;
+            }
+            """
+        )
+        if clicked:
+            page.wait_for_timeout(1200)
+        return int(clicked or 0)
+    except Exception:
+        return 0
+
+
+def scroll_comments(page: Page, max_scrolls: int, idle_rounds: int, comments_by_id: dict[str, Any]) -> None:
+    last_count = len(comments_by_id)
+    idle = 0
+    page.mouse.move(1030, 620)
+
+    for index in range(max_scrolls):
+        dismiss_popups(page)
+        expand_visible_replies(page)
+        page.mouse.wheel(0, 1000)
+        # Scroll the comment panel from the inside — the right-side panel
+        # is a fixed-height container with its own scroll; wheel events on
+        # the page body don't propagate into it.
+        page.evaluate(
+            """
+            () => {
+              // First try the dedicated comment/interaction panel
+              const commentPanel = document.querySelector('.comment-inner-container')
+                || document.querySelector('[class*="commentListContainer"]')
+                || document.querySelector('[class*="comments-container"]');
+              if (commentPanel) {
+                commentPanel.scrollTop += 1200;
+                return;
+              }
+
+              // Fallback: find tall scrollable containers on the right half
+              const candidates = [...document.querySelectorAll('div, section, aside')]
+                .filter((el) => {
+                  const r = el.getBoundingClientRect();
+                  return r.height > 160 &&
+                    r.width > 220 &&
+                    el.scrollHeight > el.clientHeight + 80 &&
+                    r.right > window.innerWidth * 0.35;
+                })
+                .sort((a, b) => (b.scrollHeight - b.clientHeight) - (a.scrollHeight - a.clientHeight));
+              for (const el of candidates.slice(0, 6)) {
+                el.scrollTop += 1000;
+              }
+            }
+            """
+        )
+        page.wait_for_timeout(1800)
+
+        count = len(comments_by_id)
+        if count == last_count:
+            idle += 1
+        else:
+            log(f"Loaded comments: {count}")
+            idle = 0
+            last_count = count
+        if idle >= idle_rounds:
+            log(f"No new comments after {idle_rounds} rounds, stopping scroll.")
+            break
+
+    log(f"Scroll finished after {index + 1} rounds.")
+
+
+def with_query_params(url: str, updates: dict[str, Any]) -> str:
+    parts = urlsplit(url)
+    params = dict(parse_qsl(parts.query, keep_blank_values=True))
+    for key, value in updates.items():
+        params[key] = str(value)
+    return urlunsplit((parts.scheme, parts.netloc, parts.path, urlencode(params), parts.fragment))
+
+
+def fetch_missing_top_comments(
+    page: Page,
+    source_url: str,
+    comments_by_id: dict[str, dict[str, Any]],
+    stats: dict[str, Any],
+) -> None:
+    """After scrolling, proactively paginate remaining top-level comments via page.evaluate."""
+    template = stats.get("comment_url_template")
+    if not template:
+        log("No top-level comment API template captured; skipped direct fetch.")
+        return
+
+    api_total = stats.get("api_total")
+    # api_total may be None if the first XHR response didn't include it;
+    # fall back to note_metrics.comment_count from the page
+    if not api_total:
+        metrics = stats.get("note_metrics") or {}
+        api_total = metrics.get("comment_count")
+    if api_total and len(comments_by_id) >= int(api_total):
+        return
+
+    last_cursor = stats.get("last_cursor")
+    # Always attempt pagination if we have a cursor — the scroll may have
+    # stopped early even though has_more was true in an earlier response.
+    if not last_cursor:
+        return
+
+    log(f"Top-level comments: collected {len(comments_by_id)} / api_total {api_total}; fetching remainder.")
+    fetched = 0
+    cursor = last_cursor
+    for _ in range(100):
+        top_url = with_query_params(template, {"cursor": cursor})
+        try:
+            payload = page.evaluate(
+                """
+                async (url) => {
+                  const res = await fetch(url, { credentials: 'include' });
+                  return await res.json();
+                }
+                """,
+                top_url,
+            )
+        except Exception:
+            break
+        if not isinstance(payload, dict):
+            break
+        payload["__source_url"] = top_url
+        before = len(comments_by_id)
+        collect_from_payload(payload, source_url, comments_by_id, stats)
+        fetched += max(0, len(comments_by_id) - before)
+        data = payload.get("data") if isinstance(payload.get("data"), dict) else payload
+        next_has_more = data.get("has_more")
+        next_cursor = data.get("cursor")
+        if not next_has_more or next_cursor in (None, cursor):
+            break
+        cursor = next_cursor
+        page.wait_for_timeout(350)
+
+    log(f"Top-level direct fetch: {fetched} new comments; total now: {len(comments_by_id)}")
+
+
+def fetch_missing_replies(
+    page: Page,
+    source_url: str,
+    comments_by_id: dict[str, dict[str, Any]],
+    stats: dict[str, Any],
+) -> None:
+    template = stats.get("sub_url_template")
+    if not template:
+        log("No sub-comment API template captured; skipped direct reply fetch.")
+        return
+
+    parents = [
+        row for row in list(comments_by_id.values())
+        if row.get("level") == "comment" and int(row.get("reply_count") or 0) > 0
+    ]
+    if not parents:
+        return
+
+    log(f"Fetching replies directly for {len(parents)} comments with replies.")
+    fetched = 0
+    for row in parents:
+        parent_id = str(row.get("comment_id") or "")
+        if not parent_id:
+            continue
+        cursor = ""
+        for _ in range(30):
+            reply_url = with_query_params(template, {"root_comment_id": parent_id, "comment_id": parent_id, "cursor": cursor})
+            try:
+                payload = page.evaluate(
+                    """
+                    async (url) => {
+                      const res = await fetch(url, { credentials: 'include' });
+                      return await res.json();
+                    }
+                    """,
+                    reply_url,
+                )
+            except Exception:
+                break
+            if not isinstance(payload, dict):
+                break
+            payload["__source_url"] = reply_url
+            before = len(comments_by_id)
+            collect_from_payload(payload, source_url, comments_by_id, stats)
+            fetched += max(0, len(comments_by_id) - before)
+            data = payload.get("data") if isinstance(payload.get("data"), dict) else payload
+            has_more = data.get("has_more")
+            next_cursor = data.get("cursor")
+            if not has_more or next_cursor in (None, cursor):
+                break
+            cursor = next_cursor
+            page.wait_for_timeout(350)
+
+    log(f"Direct replies fetched/merged: {fetched}; total comments now: {len(comments_by_id)}")
+
+
+def extract_note_metrics(page: Page) -> dict[str, Any]:
+    try:
+        metrics = page.evaluate(
+            """
+            () => {
+              const compactNumber = (text) => {
+                text = String(text || '').replace(/,/g, '').trim();
+                const match = text.match(/(\d+(?:\.\d+)?)(\s*[\u4e07wW])?/);
+                if (!match) return '';
+                const value = parseFloat(match[1]);
+                if (!Number.isFinite(value)) return '';
+                return match[2] ? Math.round(value * 10000) : Math.round(value);
+              };
+              const metricFromState = () => {
+                const noteId = location.href.match(/(?:explore|discovery\/item|profile\/[^/]+)\/(\w+)/)?.[1];
+                const detailMap = window.__INITIAL_STATE__?.note?.noteDetailMap || {};
+                const detail = noteId ? detailMap[noteId] : Object.values(detailMap)[0];
+                const note = detail?.note || {};
+                const info = note.interactInfo;
+                if (!info) return null;
+                return {
+                  title: note.title || note.desc || '',
+                  like_count: compactNumber(info.likedCount),
+                  favorite_count: compactNumber(info.collectedCount),
+                  comment_count: compactNumber(info.commentCount),
+                  share_count: '',
+                };
+              };
+              const stateMetrics = metricFromState();
+              if (stateMetrics) return stateMetrics;
+
+              const visible = (el) => {
+                const r = el.getBoundingClientRect();
+                const style = window.getComputedStyle(el);
+                return r.width > 0 && r.height > 0 &&
+                  r.bottom > 0 && r.top < window.innerHeight &&
+                  style.visibility !== 'hidden' &&
+                  style.display !== 'none' &&
+                  style.opacity !== '0';
+              };
+              const center = (r) => ({ x: (r.left + r.right) / 2, y: (r.top + r.bottom) / 2 });
+              const dedupe = (items) => {
+                const kept = [];
+                for (const item of items.sort((a, b) => a.r.left - b.r.left || a.r.top - b.r.top)) {
+                  const c = center(item.r);
+                  const duplicate = kept.some((prev) => {
+                    const pc = center(prev.r);
+                    return item.value === prev.value &&
+                      Math.abs(c.x - pc.x) < 80 &&
+                      Math.abs(c.y - pc.y) < 40;
+                  });
+                  if (!duplicate) kept.push(item);
+                }
+                return kept;
+              };
+              const textOf = (el) => (el.innerText || el.textContent || '').trim();
+              const nodes = [...document.querySelectorAll('button, div, span, p')]
+                .filter(visible)
+                .map((el) => ({ el, text: textOf(el), r: el.getBoundingClientRect() }))
+                .filter(({ text }) => text && text.length <= 120);
+
+              const commentText = nodes.find(({ text }) =>
+                /\u5171\s*\d+(?:\.\d+)?\s*[\u4e07wW]?\s*\u6761\u8bc4\u8bba/.test(text)
+              )?.text || '';
+              const commentByText = compactNumber(commentText);
+              const bottomNums = dedupe(nodes
+                .map((item) => ({ ...item, value: compactNumber(item.text) }))
+                .filter(({ text, r, value }) => /^(\d+(?:\.\d+)?\s*[\u4e07wW]?)$/.test(text) &&
+                  value !== '' &&
+                  r.top > window.innerHeight - 140 &&
+                  r.left > window.innerWidth * 0.55 &&
+                  r.width < 180 &&
+                  r.height < 110))
+                .map(({ value }) => value);
+
+              return {
+                like_count: bottomNums.length >= 1 ? bottomNums[0] : '',
+                favorite_count: bottomNums.length >= 2 ? bottomNums[1] : '',
+                comment_count: commentByText || (bottomNums.length >= 3 ? bottomNums[2] : ''),
+                share_count: '',
+              };
+            }
+            """
+        )
+        return metrics if isinstance(metrics, dict) else {}
+    except Exception:
+        return {}
+
+def scrape_comments(url: str, login_timeout: int, max_scrolls: int, idle_rounds: int, headless: bool) -> tuple[list[dict[str, Any]], dict[str, Any]]:
+    result: dict[str, Any] = {"comments": [], "stats": {}, "final_url": ""}
+
+    def action(page: Page) -> None:
+        comments_by_id, stats = setup_response_capture(page, url)
+        page.goto(url, wait_until="domcontentloaded", timeout=90000)
+        page.wait_for_timeout(6000)
+        maybe_wait_for_login(page, login_timeout)
+        dismiss_popups(page)
+        save_state(page)
+
+        scroll_comments(page, max_scrolls, idle_rounds, comments_by_id)
+        # Extract metrics early so fetch_missing_top_comments can use comment_count as fallback
+        stats["note_metrics"] = extract_note_metrics(page)
+        result["title"] = stats["note_metrics"].get("title", "")
+        fetch_missing_top_comments(page, url, comments_by_id, stats)
+        fetch_missing_replies(page, url, comments_by_id, stats)
+        page.wait_for_timeout(1000)
+        save_state(page)
+
+        result["comments"] = list(comments_by_id.values())
+        result["stats"] = stats
+        result["stats"]["logged_in"] = is_logged_in(page)
+        result["final_url"] = page.url
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+    with DynamicSession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=str(PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=90000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=restore_cookies,
+        max_pages=1,
+    ) as session:
+        session.fetch(url, page_action=action, wait=1000)
+
+    return result["comments"], result
+
+
+def write_outputs(url: str, comments: list[dict[str, Any]], result: dict[str, Any]) -> tuple[Path, Path]:
+    NOTES_DIR.mkdir(parents=True, exist_ok=True)
+    note_id = note_id_from_url(url)
+    name = f"xiaohongshu_comments_{note_id}" if note_id else safe_filename(url)
+    json_path = NOTES_DIR / f"{name}.json"
+    csv_path = NOTES_DIR / f"{name}.csv"
+    if not comments and json_path.exists() and csv_path.exists():
+        log(f"No comments scraped; keeping existing output files: {json_path}, {csv_path}")
+        return json_path, csv_path
+
+    stats = result.get("stats", {})
+    public_stats = {key: value for key, value in stats.items() if not key.endswith("_template")}
+    payload = {
+        "source_url": url,
+        "title": result.get("title", ""),
+        "scraped_at": datetime.now().isoformat(timespec="seconds"),
+        "count": len(comments),
+        "stats": public_stats,
+        "comments": comments,
+    }
+    json_path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    fieldnames = [
+        "source_url",
+        "level",
+        "comment_id",
+        "reply_to_comment_id",
+        "nickname",
+        "user_id",
+        "content",
+        "like_count",
+        "reply_count",
+        "ip_location",
+        "created_at",
+        "raw_create_time",
+    ]
+    with csv_path.open("w", newline="", encoding="utf-8-sig") as file:
+        writer = csv.DictWriter(file, fieldnames=fieldnames)
+        writer.writeheader()
+        writer.writerows(comments)
+
+    return json_path, csv_path
+
+
+def login_only(login_timeout: int) -> int:
+    """只触发登录(headed),登录完成立刻退出。给 relogin_xiaohongshu.py 用。"""
+    def action(page: Page) -> None:
+        page.goto("https://www.xiaohongshu.com/", wait_until="domcontentloaded", timeout=90000)
+        page.wait_for_timeout(2500)
+        # 主动点登录按钮触发二维码
+        try:
+            login_btn = page.get_by_text("登录", exact=True).first
+            if login_btn.count():
+                login_btn.click(timeout=3000)
+                page.wait_for_timeout(1500)
+        except Exception:
+            pass
+        maybe_wait_for_login(page, login_timeout)
+        save_state(page)
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+    with DynamicSession(
+        headless=False,
+        real_chrome=True,
+        user_data_dir=str(PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=90000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=restore_cookies,
+        max_pages=1,
+    ) as session:
+        session.fetch("https://www.xiaohongshu.com/", page_action=action, wait=1000)
+    log("[login-only] 完成")
+    return 0
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Scrape Xiaohongshu note comments with a real browser.")
+    parser.add_argument("url", help="Xiaohongshu explore/note URL (传 'login' 配合 --login-only)")
+    parser.add_argument("--login-timeout", type=int, default=300, help="等待扫码/登录秒数")
+    parser.add_argument("--max-scrolls", type=int, default=120, help="最多滚动加载次数")
+    parser.add_argument("--idle-rounds", type=int, default=8, help="连续几轮无新评论后停止")
+    parser.add_argument("--headless", action="store_true", help="无头模式；首次登录不要使用")
+    parser.add_argument("--login-only", action="store_true",
+                        help="只触发登录(headed)然后退出,给 relogin_xiaohongshu 用")
+    args = parser.parse_args()
+
+    if args.login_only:
+        return login_only(args.login_timeout)
+
+    comments, result = scrape_comments(args.url, args.login_timeout, args.max_scrolls, args.idle_rounds, args.headless)
+    json_path, csv_path = write_outputs(args.url, comments, result)
+    log(f"Final URL: {result.get('final_url', '')}")
+    log(f"API hits: {result.get('stats', {}).get('api_hits', 0)}")
+    log(f"Scraped {len(comments)} comments")
+    log(f"JSON: {json_path}")
+    log(f"CSV:  {csv_path}")
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/xingtu_scraper.py b/src/gyxx_flow/modules/content_marketing/runtime/xingtu_scraper.py
new file mode 100644
index 0000000..aaf93ac
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/xingtu_scraper.py
@@ -0,0 +1,479 @@
+import argparse
+import csv
+import json
+import re
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import PATHS
+from typing import Any
+
+from playwright.sync_api import Page, TimeoutError as PlaywrightTimeoutError
+from scrapling.fetchers import DynamicSession
+
+
+MARKET_URL = "https://www.xingtu.cn/ad/creator/market"
+DETAIL_URL_MARK = "/ad/creator/author-homepage/"
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+COOKIE_FILE = PATHS.browser_cookie_file
+PROFILE_DIR = PATHS.browser_profile_dir
+
+
+def log(message: str) -> None:
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {message}", flush=True)
+
+
+def load_cookies() -> list[dict[str, Any]]:
+    if not COOKIE_FILE.exists():
+        return []
+    try:
+        data = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+        return data if isinstance(data, list) else []
+    except json.JSONDecodeError:
+        log(f"Cookie file is invalid, ignoring: {COOKIE_FILE}")
+        return []
+
+
+def restore_cookies(page: Page) -> None:
+    cookies = load_cookies()
+    if not cookies:
+        return
+    try:
+        page.context.add_cookies(cookies)
+        log(f"Loaded cookies: {COOKIE_FILE}")
+    except Exception as exc:
+        log(f"Failed to load cookies, will continue with browser profile: {exc}")
+
+
+def save_state(page: Page) -> None:
+    DATA_DIR.mkdir(exist_ok=True)
+    cookies = page.context.cookies()
+    COOKIE_FILE.write_text(
+        json.dumps(cookies, ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    storage_state = page.context.storage_state()
+    PATHS.browser_storage_state_file.write_text(
+        json.dumps(storage_state, ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    log(f"Saved cookies: {COOKIE_FILE}")
+
+
+def is_logged_in(page: Page) -> bool:
+    if "/ad/creator/market" not in page.url:
+        return False
+    for text in ("内容找人", "昵称找人", "达人信息", "达人清单", "发布任务"):
+        try:
+            page.get_by_text(text, exact=False).first.wait_for(timeout=1200)
+            return True
+        except PlaywrightTimeoutError:
+            pass
+    return False
+
+
+def wait_for_scan_login(page: Page, timeout_seconds: int) -> None:
+    log("Not logged in. Please scan/login in the opened Chrome window, then make sure it enters the creator market page.")
+    deadline = time.time() + timeout_seconds
+    while time.time() < deadline:
+        if "redirect_uri=/ad/creator/market" in page.url:
+            try:
+                page.goto(MARKET_URL, wait_until="domcontentloaded", timeout=60000)
+                page.wait_for_timeout(2000)
+            except Exception:
+                pass
+        if is_logged_in(page):
+            save_state(page)
+            return
+        page.wait_for_timeout(2000)
+    raise TimeoutError(f"Login timed out after {timeout_seconds} seconds.")
+
+
+def ensure_login(page: Page, login_timeout: int) -> None:
+    page.goto(MARKET_URL, wait_until="domcontentloaded", timeout=60000)
+    page.wait_for_timeout(3000)
+    if is_logged_in(page):
+        log("Already logged in, using saved browser session.")
+        save_state(page)
+        return
+
+    wait_for_scan_login(page, login_timeout)
+    page.goto(MARKET_URL, wait_until="domcontentloaded", timeout=60000)
+    page.wait_for_timeout(3000)
+    if not is_logged_in(page):
+        wait_for_scan_login(page, login_timeout)
+
+
+def dismiss_popups(page: Page) -> None:
+    for text in ("跳过", "我知道了", "稍后再说"):
+        click_by_text(page, text, exact=True, timeout=1000)
+    page.evaluate(
+        """
+        () => {
+          const close = [...document.querySelectorAll('button, span, div')]
+            .filter((el) => {
+              const text = (el.innerText || '').trim();
+              const r = el.getBoundingClientRect();
+              return (text === '×' || text === 'x' || text === 'X') &&
+                r.width <= 40 && r.height <= 40 &&
+                r.left > window.innerWidth * 0.45 &&
+                r.top < window.innerHeight * 0.45;
+            })[0];
+          close?.click();
+        }
+        """
+    )
+    page.wait_for_timeout(500)
+
+
+def click_by_text(page: Page, text: str, exact: bool = True, timeout: int = 5000) -> bool:
+    try:
+        locator = page.get_by_text(text, exact=exact).first
+        if locator.count():
+            locator.click(timeout=timeout)
+            return True
+    except Exception:
+        return False
+    return False
+
+
+def select_nickname_search(page: Page) -> None:
+    log("Switching to nickname search.")
+    if click_by_text(page, "昵称找人", exact=True, timeout=5000):
+        page.wait_for_timeout(800)
+        return
+
+    page.evaluate(
+        """
+        () => {
+          const target = [...document.querySelectorAll('button, div, span')]
+            .find((el) => (el.innerText || '').trim() === '昵称找人');
+          target?.click();
+        }
+        """
+    )
+    page.wait_for_timeout(800)
+
+
+def search_creator(page: Page, creator_name: str) -> None:
+    dismiss_popups(page)
+    select_nickname_search(page)
+    input_locator = page.locator("input:visible").first
+    input_locator.wait_for(state="visible", timeout=30000)
+    input_locator.click()
+    input_locator.fill(creator_name)
+    input_locator.press("Enter")
+
+    try:
+        page.wait_for_timeout(2000)
+        wait_for_visible_creator_result(page, creator_name, timeout_ms=8000)
+    except PlaywrightTimeoutError:
+        page.evaluate(
+            """
+            () => {
+              const input = [...document.querySelectorAll('input')].find((el) => el.value);
+              const box = input?.closest('div');
+              const candidates = [...document.querySelectorAll('button, [role="button"], div, span')]
+                .filter((el) => {
+                  const r = el.getBoundingClientRect();
+                  return r.width >= 24 && r.width <= 80 && r.height >= 24 && r.height <= 80;
+                });
+              const near = candidates.find((el) => {
+                const er = el.getBoundingClientRect();
+                const br = box?.getBoundingClientRect();
+                return br && er.left > br.right - 30 && Math.abs(er.top - br.top) < 60;
+              });
+              near?.click();
+            }
+            """
+        )
+        wait_for_visible_creator_result(page, creator_name, timeout_ms=30000)
+
+    page.wait_for_timeout(2500)
+
+
+def wait_for_visible_creator_result(page: Page, creator_name: str, timeout_ms: int) -> None:
+    deadline = time.time() + timeout_ms / 1000
+    while time.time() < deadline:
+        found = page.evaluate(
+            """
+            (name) => {
+              const visible = (el) => {
+                const r = el.getBoundingClientRect();
+                const style = window.getComputedStyle(el);
+                return r.width > 0 && r.height > 0 &&
+                  r.bottom > 0 && r.top < window.innerHeight &&
+                  style.visibility !== 'hidden' &&
+                  style.display !== 'none' &&
+                  style.opacity !== '0';
+              };
+              return [...document.querySelectorAll('tr, [class*="row"], [class*="table"], [class*="item"], div')]
+                .some((el) => visible(el) && (el.innerText || '').includes(name) && /达人|粉丝|播放|报价|下单|内容主题/.test(el.innerText || ''));
+            }
+            """,
+            creator_name,
+        )
+        if found:
+            return
+        page.wait_for_timeout(500)
+    raise PlaywrightTimeoutError(f"Timed out waiting for visible creator result: {creator_name}")
+
+
+def open_creator_detail(page: Page, creator_name: str) -> Page:
+    log(f"Opening creator detail: {creator_name}")
+    pages_before = list(page.context.pages)
+    clicked = page.evaluate(
+        """
+        (name) => {
+          const textHasName = (el) => (el.innerText || '').includes(name);
+          const rows = [...document.querySelectorAll('tr, [class*="row"], [class*="table"], [class*="item"]')]
+            .filter(textHasName);
+          const exactRow = rows.find((el) => {
+            const lines = (el.innerText || '').split('\\n').map((line) => line.trim()).filter(Boolean);
+            return lines.includes(name);
+          });
+          const row = exactRow || rows[0];
+          if (!row) return false;
+          const target = row.querySelector('img, a, [class*="avatar"], [class*="name"]') || row;
+          target.scrollIntoView({ block: 'center', inline: 'center' });
+          target.click();
+          return true;
+        }
+        """,
+        creator_name,
+    )
+    if not clicked:
+        page.get_by_text(creator_name, exact=True).first.click(timeout=10000)
+
+    detail_page = wait_for_detail_page(page, pages_before, timeout_ms=30000)
+    if detail_page is None:
+        page.get_by_text(creator_name, exact=False).first.click(timeout=10000)
+        detail_page = wait_for_detail_page(page, pages_before, timeout_ms=30000)
+
+    if detail_page is None:
+        raise TimeoutError("Clicked creator result, but no author-homepage page was detected.")
+
+    detail_page.bring_to_front()
+    detail_page.wait_for_load_state("domcontentloaded")
+    page.wait_for_timeout(1000)
+    return detail_page
+
+
+def wait_for_detail_page(page: Page, pages_before: list[Page], timeout_ms: int) -> Page | None:
+    deadline = time.time() + timeout_ms / 1000
+    known = set(pages_before)
+    while time.time() < deadline:
+        all_pages = list(page.context.pages)
+        candidates = [p for p in all_pages if p not in known] + [page] + all_pages
+        for candidate in candidates:
+            try:
+                if DETAIL_URL_MARK in candidate.url or "/ad/creator/author-homepage/" in candidate.url:
+                    return candidate
+            except Exception:
+                pass
+        page.wait_for_timeout(500)
+    return None
+
+
+def open_creation_ability(page: Page) -> None:
+    log("Opening creation ability tab.")
+    try:
+        page.get_by_text("创作能力", exact=True).first.click(timeout=10000)
+    except Exception:
+        page.evaluate(
+            """
+            () => {
+              const target = [...document.querySelectorAll('button, div, span, a')]
+                .find((el) => (el.innerText || '').trim() === '创作能力');
+              target?.click();
+            }
+            """
+        )
+    page.wait_for_timeout(3000)
+    try:
+        all_video = page.get_by_text("全部视频", exact=False).first
+        all_video.wait_for(timeout=30000)
+        all_video.scroll_into_view_if_needed()
+        page.wait_for_timeout(1200)
+        page.mouse.wheel(0, 600)
+        page.wait_for_timeout(1500)
+    except PlaywrightTimeoutError:
+        pass
+
+
+def parse_video_cards(page: Page) -> list[dict[str, str]]:
+    return page.evaluate(
+        """
+        () => {
+          const clean = (value) => (value || '').replace(/\\s+/g, ' ').trim();
+
+          const valueAfter = (lines, label) => {
+            const index = lines.findIndex((line) => line === label || line.includes(label));
+            if (index < 0) return '';
+            const sameLine = lines[index].replace(label, '').trim();
+            if (sameLine) return sameLine;
+            return lines[index + 1] || '';
+          };
+
+          const parseCard = (card) => {
+            const lines = (card.innerText || '')
+              .split('\\n')
+              .map((line) => line.trim())
+              .filter(Boolean);
+            const text = lines.join('\\n');
+            if (!/播放量/.test(text) || !/点赞量/.test(text) || !/评论量/.test(text)) return null;
+
+            const firstMetric = lines.findIndex((line) => /播放量/.test(line));
+            const titleCandidates = lines.slice(0, firstMetric >= 0 ? firstMetric : lines.length).filter((line) => {
+              if (/^(全部视频|全部|个人视频|星图视频|播放量|点赞量|评论量)$/.test(line)) return false;
+              if (/请输入关键词搜索|开始日期|结束日期|展开/.test(line)) return false;
+              if (/^[\\d,.wW万]+$/.test(line)) return false;
+              return line.length >= 2;
+            });
+            const title = titleCandidates.slice(-2).join(' ').trim();
+
+            return {
+              title: clean(title),
+              play_count: clean(valueAfter(lines, '播放量')),
+              like_count: clean(valueAfter(lines, '点赞量')),
+              comment_count: clean(valueAfter(lines, '评论量')),
+            };
+          };
+
+          const cards = [...document.querySelectorAll('div')].filter((el) => {
+            const text = el.innerText || '';
+            const rect = el.getBoundingClientRect();
+            return rect.width >= 140 &&
+              rect.width <= 340 &&
+              rect.height >= 260 &&
+              rect.height <= 620 &&
+              /播放量/.test(text) &&
+              /点赞量/.test(text) &&
+              /评论量/.test(text) &&
+              el.querySelector('img');
+          });
+
+          const parsed = cards.map(parseCard).filter(Boolean);
+          const seen = new Set();
+          return parsed.filter((item) => {
+            const key = [item.title, item.play_count, item.like_count, item.comment_count].join('|');
+            if (seen.has(key)) return false;
+            seen.add(key);
+            return item.title && (item.play_count || item.like_count || item.comment_count);
+          });
+        }
+        """
+    )
+
+
+def write_outputs(creator_name: str, videos: list[dict[str, str]]) -> tuple[Path, Path]:
+    DATA_DIR.mkdir(exist_ok=True)
+    safe_name = re.sub(r'[<>:"/\\\\|?*\\s]+', "_", creator_name).strip("_") or "creator"
+    json_path = DATA_DIR / f"xingtu_{safe_name}_videos.json"
+    csv_path = DATA_DIR / f"xingtu_{safe_name}_videos.csv"
+
+    payload = {
+        "creator": creator_name,
+        "scraped_at": datetime.now().isoformat(timespec="seconds"),
+        "count": len(videos),
+        "videos": videos,
+    }
+    json_path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    with csv_path.open("w", newline="", encoding="utf-8-sig") as file:
+        writer = csv.DictWriter(
+            file,
+            fieldnames=["title", "play_count", "like_count", "comment_count"],
+        )
+        writer.writeheader()
+        writer.writerows(videos)
+
+    return json_path, csv_path
+
+
+def load_names(cli_names: list[str], names_file: str | None) -> list[str]:
+    names: list[str] = []
+    names.extend(name.strip() for name in cli_names if name.strip())
+
+    if names_file:
+        path = Path(names_file)
+        for line in path.read_text(encoding="utf-8-sig").splitlines():
+            name = line.strip()
+            if name and not name.startswith("#"):
+                names.append(name)
+
+    deduped: list[str] = []
+    seen = set()
+    for name in names:
+        if name not in seen:
+            seen.add(name)
+            deduped.append(name)
+    return deduped
+
+
+def scrape(creator_name: str, login_timeout: int, headless: bool) -> list[dict[str, str]]:
+    result: dict[str, Any] = {"videos": [], "detail_url": ""}
+
+    def action(page: Page) -> None:
+        ensure_login(page, login_timeout)
+        search_creator(page, creator_name)
+        detail_page = open_creator_detail(page, creator_name)
+        open_creation_ability(detail_page)
+        result["detail_url"] = detail_page.url
+        result["videos"] = parse_video_cards(detail_page)
+        save_state(detail_page)
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+    with DynamicSession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=str(PROFILE_DIR),
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=60000,
+        network_idle=False,
+        disable_resources=False,
+        google_search=False,
+        page_setup=restore_cookies,
+        max_pages=1,
+    ) as session:
+        session.fetch(MARKET_URL, page_action=action, wait=1000)
+
+    log(f"Detail URL: {result.get('detail_url', '')}")
+    return result["videos"]
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="Scrape Xingtu creator video metrics with Scrapling DynamicSession.",
+    )
+    parser.add_argument("creator_names", nargs="*", help="达人昵称，可传多个，例如：达人A 达人B")
+    parser.add_argument("--names-file", help="批量达人名单文件，一行一个昵称")
+    parser.add_argument("--login-timeout", type=int, default=300, help="首次扫码登录等待秒数")
+    parser.add_argument("--headless", action="store_true", help="无头模式；首次登录不要使用")
+    args = parser.parse_args()
+
+    creator_names = load_names(args.creator_names, args.names_file)
+    if not creator_names:
+        parser.error("Please provide at least one creator name or --names-file.")
+
+    for index, creator_name in enumerate(creator_names, start=1):
+        log(f"=== Creator {index}/{len(creator_names)}: {creator_name} ===")
+        videos = scrape(creator_name, args.login_timeout, args.headless)
+        json_path, csv_path = write_outputs(creator_name, videos)
+        log(f"Scraped {len(videos)} videos")
+        log(f"JSON: {json_path}")
+        log(f"CSV:  {csv_path}")
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/content_marketing/runtime/xingtu_scraper_v2.py b/src/gyxx_flow/modules/content_marketing/runtime/xingtu_scraper_v2.py
new file mode 100644
index 0000000..cefaa4f
--- /dev/null
+++ b/src/gyxx_flow/modules/content_marketing/runtime/xingtu_scraper_v2.py
@@ -0,0 +1,2004 @@
+"""
+星图 (巨量星图) V2: 多维表格驱动 + 精准找视频 + 回填
+跟 pgy_xhs_scraper_v2.py 同架构,只是目标平台换为星图
+
+- 字段映射用同一份对照表 (款式_多维表格_对照.json) 即可
+- 这里多维表格记录的是同一拨博主笔记,星图抓的是抖音同名达人的视频
+- 默认以"达人名称"作为搜索词进星图找创作者
+- 星图主页 "创作能力" tab 下找对应标题的视频
+- 抓播放/点赞/评论量,回填到该款式表的"7天曝光量"列(同 xhs)
+
+用法:
+  python xingtu_scraper_v2.py --style 7 --dry-run
+"""
+import argparse
+import json
+import os
+import re
+import subprocess
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+from gyxx_flow.modules.content_marketing.runtime.runtime_paths import (
+    PATHS,
+    resolve_layer_output,
+)
+
+from gyxx_flow.modules.content_marketing.runtime.creator_task_grouping import (
+    chunk_creator_groups,
+    group_tasks_by_creator as _group_tasks_by_creator,
+)
+from gyxx_flow.modules.content_marketing.runtime.collection_completeness import (
+    BLOCKED_INPUT,
+    RETRYABLE_FAILURE,
+    SUCCESS,
+    WRITE_FAILURE,
+    add_repeatable_style_argument,
+    atomic_write_json,
+    coerce_url,
+    deduplicate_observed_cards,
+    extract_content_id,
+    finalize_summary,
+    match_tasks_to_cards,
+    merge_global_summaries,
+    merge_style_summary,
+    normalize_title as normalize_match_title,
+    select_requested_styles,
+    title_similarity,
+    upsert_result,
+)
+from typing import Any
+
+from playwright.sync_api import Page, TimeoutError as PlaywrightTimeoutError
+from scrapling.fetchers import DynamicSession
+
+MARKET_URL = "https://www.xingtu.cn/ad/creator/market"
+CUSTOMER_LOGIN_URL = (
+    "https://sso.oceanengine.com/xingtu/login"
+    "?redirect_uri=%2Fad%2Fcreator%2Fmarket&role=1"
+)
+DETAIL_URL_MARK = "/ad/creator/author-homepage/"
+
+BASE_DIR = PATHS.module_root
+DATA_DIR = PATHS.raw_root
+V2_DIR = PATHS.normalized_root / "v2_results"
+MAPPING_PATH = PATHS.normalized_root / "mappings" / "款式_多维表格_对照.json"
+COOKIE_FILE = PATHS.browser_cookie_file
+PROFILE_DIR = PATHS.browser_profile_dir
+LOGIN_COOKIE_NAMES = frozenset({"sessionid", "sessionid_ss", "sid_tt", "sid_guard"})
+DEFAULT_BATCH_SIZE = 0
+DEFAULT_BATCH_WAIT = 0
+MAX_CONSECUTIVE_CREATOR_SEARCH_FAILURES = 3
+
+LARK_CLI = os.path.expandvars(r"%APPDATA%\npm\lark-cli.cmd")
+if not os.path.exists(LARK_CLI):
+    LARK_CLI = "lark-cli.cmd"
+
+
+def log(msg: str) -> None:
+    safe = str(msg).encode("gbk", errors="replace").decode("gbk", errors="replace")
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {safe}", flush=True)
+
+
+class SearchQuotaExhausted(RuntimeError):
+    """Xingtu explicitly reports that creator-search quota is unavailable."""
+
+
+class CreatorSearchBudget:
+    """Enforce one submitted search per creator and explicit platform blocks only."""
+
+    def __init__(self, max_consecutive_failures: int = MAX_CONSECUTIVE_CREATOR_SEARCH_FAILURES):
+        self.max_consecutive_failures = max(1, int(max_consecutive_failures))
+        self.attempted_keys: set[tuple[str, str]] = set()
+        self.calls = 0
+        self.consecutive_failures = 0
+        self.circuit_reason: str | None = None
+
+    def begin(self, creator_key: tuple[str, str]) -> bool:
+        if self.circuit_reason or creator_key in self.attempted_keys:
+            return False
+        self.attempted_keys.add(creator_key)
+        self.calls += 1
+        return True
+
+    def succeeded(self) -> None:
+        self.consecutive_failures = 0
+
+    def failed(self, reason: str) -> bool:
+        """Record a creator-level miss without opening a global circuit.
+
+        Missing or ambiguous creator results are normal for fuzzy nickname
+        search and must not prevent later, unrelated creators from running.
+        Only ``exhaust()`` may open the global circuit for an explicit quota or
+        captcha signal.
+        """
+        self.consecutive_failures += 1
+        return False
+
+    def exhaust(self, reason: str = "search_quota_exhausted") -> None:
+        self.circuit_reason = reason
+
+
+def creator_search_key(creator_name: str, creator_id: str | None) -> tuple[str, str]:
+    if creator_id:
+        return "id", str(creator_id).strip()
+    return "name", "".join(str(creator_name or "").split()).casefold()
+
+
+def wait_with_heartbeat(total_seconds: int, interval_seconds: int = 600) -> None:
+    """等待 total_seconds，每 interval_seconds 记录一次进度。"""
+    elapsed = 0
+    while elapsed < total_seconds:
+        chunk = min(interval_seconds, total_seconds - elapsed)
+        time.sleep(chunk)
+        elapsed += chunk
+        remaining = total_seconds - elapsed
+        if remaining > 0:
+            log(f"  [BATCH WAIT] 还剩 {remaining // 60} 分钟...")
+
+
+def is_browser_session_lost(exc: BaseException) -> bool:
+    text = f"{type(exc).__name__}: {exc}".casefold()
+    return any(token in text for token in (
+        "targetclosed", "has been closed", "page closed", "context closed",
+        "browser has been closed", "target page, context or browser",
+        "failed to open a new tab", "target.createtarget",
+    ))
+
+
+def reraise_if_browser_session_lost(exc: BaseException) -> None:
+    """Do not let local search fallbacks swallow a dead browser session."""
+    if is_browser_session_lost(exc):
+        raise exc
+
+
+def call_lark_json(args: list[str]) -> dict:
+    cmd = [LARK_CLI] + args
+    env = {**os.environ, "LARK_CLI_NO_PROXY": "1"}
+    proc = subprocess.run(cmd, capture_output=True, text=True, env=env, encoding="utf-8")
+    if proc.returncode != 0:
+        raise RuntimeError(f"lark-cli 调用失败: {proc.stderr[:500]}\nstdout: {proc.stdout[:500]}")
+    text = proc.stdout
+    start = text.find("{")
+    if start < 0:
+        raise RuntimeError(f"无 JSON 返回: {text[:500]}")
+    return json.loads(text[start:], strict=False)
+
+
+def load_mapping(self_operated: bool = False) -> dict:
+    # 动态:从「合作达人」/「自营达人」多维表格地址表读取款式→各表地址,实时拉字段重建对照表
+    from gyxx_flow.modules.content_marketing.runtime import feishu_mapping
+    return feishu_mapping.load_mapping(DATA_DIR, self_operated=self_operated)
+
+
+def list_records_by_table(base_token: str, table_id: str) -> list[dict]:
+    args = [
+        "base", "+record-list",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--as", "user",
+        "--format", "json",
+        "--limit", "200",
+    ]
+    all_records = []
+    offset = 0
+    while True:
+        page_args = args + ["--offset", str(offset)]
+        resp = call_lark_json(page_args)
+        if not resp.get("ok"):
+            raise RuntimeError(f"拉记录失败: {resp.get('error')}")
+        d = resp["data"]
+        ids = d["field_id_list"]
+        rids = d["record_id_list"]
+        rows = d["data"]
+        for i, row in enumerate(rows):
+            rec = {"record_id": rids[i]}
+            for j, fid in enumerate(ids):
+                rec[fid] = row[j] if j < len(row) else None
+            all_records.append(rec)
+        offset += len(rows)
+        if not d.get("has_more", False) or len(rows) == 0:
+            break
+    return all_records
+
+
+def write_back(base_token: str, table_id: str, record_id: str, field_id: str, value) -> bool:
+    payload = {field_id: value}
+    resp = call_lark_json([
+        "base", "+record-upsert",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--record-id", record_id,
+        "--as", "user",
+        "--format", "json",
+        "--json", json.dumps(payload, ensure_ascii=False),
+    ])
+    if not resp.get("ok"):
+        log(f"  回填失败 record_id={record_id}: {resp.get('error')}")
+        return False
+    return True
+
+
+def parse_publish_time(value) -> datetime | None:
+    """把飞书多维表格里的发布时间字段解析成 datetime。
+    飞书 date/datetime 字段返回毫秒时间戳(数值)。"""
+    if value is None or value == "":
+        return None
+    if isinstance(value, (int, float)):
+        try:
+            return datetime.fromtimestamp(value / 1000)
+        except (ValueError, OSError):
+            return None
+    if isinstance(value, str):
+        v = value.strip()
+        if not v:
+            return None
+        if v.isdigit():
+            try:
+                return datetime.fromtimestamp(int(v) / 1000)
+            except (ValueError, OSError):
+                return None
+        for fmt in ("%Y-%m-%dT%H:%M:%S", "%Y-%m-%d %H:%M:%S", "%Y-%m-%d", "%Y/%m/%d"):
+            try:
+                return datetime.strptime(v[:len(fmt) + 2], fmt)
+            except ValueError:
+                continue
+    return None
+
+
+def pick_read_field(fmap: dict, pub_time: datetime | None,
+                    today: datetime | None = None) -> tuple[str, str, str] | None:
+    """根据发布时间距今天的天数,挑出要回填的曝光量字段。
+    规则:
+      0-7 天  -> 7天曝光量
+      8-14 天 -> 14天曝光量
+      15-21 天 -> 21天曝光量
+      22-28 天 -> 28天曝光量
+      29+ 天  -> 月底曝光量
+    未来时间(days<0)也返回 None,调用方应区分"无发布时间"和"未来时间"。
+    """
+    if pub_time is None:
+        return None  # 缺发布时间就不写,让调用方 skip
+    today = today or datetime.now()
+    days = (today.date() - pub_time.date()).days
+    if days < 0:
+        return "FUTURE"  # type: ignore[return-value]  # sentinel: 未来时间不写
+    if days <= 7:
+        logical = "read_count_7d"
+    elif days <= 14:
+        logical = "read_count_14d"
+    elif days <= 21:
+        logical = "read_count_21d"
+    elif days <= 28:
+        logical = "read_count_28d"
+    else:
+        logical = "month_end"
+    info = fmap.get(logical)
+    if info:
+        return (logical, info["field_id"], info["field_name"])
+    # 目标字段不存在时不降级写7天桶,避免数据错位
+    return None
+
+
+def restore_cookies(page: Page) -> None:
+    if not COOKIE_FILE.exists():
+        return
+    try:
+        cookies = json.loads(COOKIE_FILE.read_text(encoding="utf-8"))
+        page.context.add_cookies(cookies)
+        log(f"Loaded cookies: {COOKIE_FILE}")
+    except Exception as exc:
+        log(f"Failed to load cookies: {exc}")
+
+
+def save_state(page: Page) -> None:
+    DATA_DIR.mkdir(exist_ok=True)
+    cookies = page.context.cookies()
+    temp_file = COOKIE_FILE.with_name(f"{COOKIE_FILE.name}.tmp")
+    temp_file.write_text(
+        json.dumps(cookies, ensure_ascii=False, indent=2), encoding="utf-8"
+    )
+    os.replace(temp_file, COOKIE_FILE)
+    log(f"Saved cookies: {COOKIE_FILE}")
+
+
+def has_valid_login_cookie_file(path: Path | None = None) -> bool:
+    path = path or COOKIE_FILE
+    try:
+        cookies = json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return False
+    return any(
+        isinstance(cookie, dict)
+        and cookie.get("name") in LOGIN_COOKIE_NAMES
+        and cookie.get("value")
+        for cookie in cookies
+    )
+
+
+def is_logged_in(page: Page) -> bool:
+    url = str(page.url or "")
+    if not any(path in url for path in ("/ad/creator/market", "/ad/creator/index")):
+        return False
+    for text in ("内容找人", "昵称找人", "达人信息", "达人清单", "发布任务"):
+        try:
+            page.get_by_text(text, exact=False).first.wait_for(timeout=1200)
+            return True
+        except PlaywrightTimeoutError:
+            pass
+    try:
+        return any(
+            cookie.get("name") in LOGIN_COOKIE_NAMES and cookie.get("value")
+            for cookie in page.context.cookies()
+            if isinstance(cookie, dict)
+        )
+    except Exception:
+        return False
+    return False
+
+
+def open_scan_login(page: Page) -> None:
+    """Open the Xingtu login dialog and prefer Douyin QR-code login."""
+    if "sso.oceanengine.com/xingtu/login" not in page.url:
+        try:
+            page.goto(CUSTOMER_LOGIN_URL, wait_until="domcontentloaded", timeout=60000)
+            page.wait_for_timeout(1500)
+        except Exception:
+            pass
+
+    # The Douyin login icon on Xingtu's customer SSO page has a stable
+    # aweme.png source. Clicking it opens the QR authorization page directly.
+    try:
+        douyin = page.locator('img[src*="aweme.png"]').first
+        if douyin.count() and douyin.is_visible():
+            douyin.click(timeout=3000)
+            page.wait_for_timeout(1500)
+            return
+    except Exception:
+        pass
+
+    for pattern in (r"^登录$|立即登录", r"抖音扫码登录|扫码登录|二维码登录|扫码"):
+        try:
+            button = page.get_by_text(re.compile(pattern)).first
+            if button.count() and button.is_visible():
+                button.click(timeout=2000)
+                page.wait_for_timeout(500)
+        except Exception:
+            pass
+
+
+def is_scan_qr_page(page: Page) -> bool:
+    """Return True once Douyin's QR authorization page is visible.
+
+    The QR page lives outside xingtu.cn. Treating that URL as a page that still
+    needs opening causes the login helper to navigate back and click again every
+    few seconds, continuously refreshing the QR code.
+    """
+    url = page.url.lower()
+    if any(host in url for host in ("open.douyin.com", "sso.douyin.com")):
+        return True
+    for selector in (
+        'img[src*="qrcode"]',
+        'img[src*="qr_code"]',
+        'canvas',
+    ):
+        try:
+            locator = page.locator(selector).first
+            if locator.count() and locator.is_visible():
+                return True
+        except Exception:
+            pass
+    return False
+
+
+def wait_for_scan_login(page: Page, timeout_seconds: int) -> None:
+    log("Not logged in. 请在打开的 Chrome 中扫码登录。")
+    open_scan_login(page)
+    deadline = time.time() + timeout_seconds
+    next_login_click = time.time() + 5
+    while time.time() < deadline:
+        if "redirect_uri=/ad/creator/market" in page.url and not is_scan_qr_page(page):
+            try:
+                page.goto(MARKET_URL, wait_until="domcontentloaded", timeout=60000)
+                page.wait_for_timeout(2000)
+            except Exception:
+                pass
+        if is_logged_in(page):
+            save_state(page)
+            return
+        if time.time() >= next_login_click and not is_scan_qr_page(page):
+            open_scan_login(page)
+            next_login_click = time.time() + 5
+        page.wait_for_timeout(2000)
+    raise TimeoutError(f"Login timed out after {timeout_seconds} seconds.")
+
+
+def ensure_login(page: Page, login_timeout: int) -> None:
+    page.goto(MARKET_URL, wait_until="domcontentloaded", timeout=60000)
+    page.wait_for_timeout(3000)
+    if is_logged_in(page):
+        log("Already logged in.")
+        save_state(page)
+        return
+    wait_for_scan_login(page, login_timeout)
+    page.goto(MARKET_URL, wait_until="domcontentloaded", timeout=60000)
+    page.wait_for_timeout(3000)
+    if not is_logged_in(page):
+        wait_for_scan_login(page, login_timeout)
+
+
+def dismiss_popups(page: Page) -> None:
+    for text in ("跳过", "我知道了", "稍后再说"):
+        try:
+            b = page.get_by_text(text, exact=True).first
+            if b.count():
+                b.click(timeout=1000)
+        except Exception:
+            pass
+    page.wait_for_timeout(500)
+
+
+def select_nickname_search(page: Page) -> None:
+    if not click_by_text(page, "昵称找人", exact=True, timeout=5000):
+        page.evaluate(
+            """
+            () => {
+              const t = [...document.querySelectorAll('button, div, span')]
+                .find((el) => (el.innerText || '').trim() === '昵称找人');
+              t?.click();
+            }
+            """
+        )
+    page.wait_for_timeout(800)
+
+
+def click_by_text(page: Page, text: str, exact: bool = True, timeout: int = 5000) -> bool:
+    try:
+        loc = page.get_by_text(text, exact=exact).first
+        if loc.count():
+            loc.click(timeout=timeout)
+            return True
+    except Exception:
+        return False
+    return False
+
+
+def search_creator(page: Page, creator_name: str, creator_id: str | None = None) -> None:
+    """搜索达人/创作者。
+    优先用 creator_id(ID 唯一,更稳),为空才用 creator_name。
+    星图"昵称找人"搜索框支持达人ID/昵称混搜,直接填值即可。
+    """
+    query = str(creator_id) if creator_id else creator_name
+    log(f"  搜索: {query} (类型: {'ID' if creator_id else '昵称'})")
+    dismiss_popups(page)
+    select_nickname_search(page)
+    input_locator = page.locator("input:visible").first
+    input_locator.wait_for(state="visible", timeout=30000)
+    input_locator.click()
+    input_locator.fill(query)
+    input_locator.press("Enter")
+    page.wait_for_timeout(800)
+    # Wait for the result list itself. Requiring the exact creator on page 1
+    # makes pagination impossible and the old broad div scan matched the input.
+    wait_for_visible_creator_result(page, query, timeout_ms=30000)
+    page.wait_for_timeout(1500)
+
+
+def search_and_open_creator_once(page: Page, creator_name: str,
+                                 creator_id: str | None = None) -> Page:
+    """Submit exactly one creator search, then inspect its result pages."""
+    search_creator(page, creator_name, creator_id)
+    return open_creator_detail(page, creator_name, creator_id)
+
+
+def wait_for_visible_creator_result(page: Page, name: str, timeout_ms: int) -> None:
+    deadline = time.time() + timeout_ms / 1000
+    while time.time() < deadline:
+        state = page.evaluate(
+            """
+            (name) => {
+              const visible = (el) => {
+                const r = el.getBoundingClientRect();
+                const s = window.getComputedStyle(el);
+                return r.width > 0 && r.height > 0
+                  && r.bottom > 0 && r.top < window.innerHeight
+                  && s.visibility !== 'hidden' && s.display !== 'none' && s.opacity !== '0';
+              };
+              const text = (document.body && document.body.innerText) || '';
+              if (/搜索次数.{0,12}(?:已用完|不足|达到上限|已达上限)|(?:今日|当天).{0,12}(?:搜索次数|达人搜索).{0,12}(?:已用完|不足|达到上限|已达上限|为0|为零)|剩余搜索次数.{0,6}(?:为)?(?:0|零)|暂无搜索次数/.test(text)) {
+                return 'quota';
+              }
+              const noResult = [...document.querySelectorAll('div, span')]
+                .some((el) => visible(el) && /暂无数据|暂无结果|未找到相关达人/.test((el.innerText || '').trim()));
+              if (noResult) return 'ready';
+              const hasCreator = [...document.querySelectorAll('.author-nickname')].some((el) => {
+                const r = el.getBoundingClientRect();
+                const s = window.getComputedStyle(el);
+                return r.width > 0 && r.height > 0
+                  && s.visibility !== 'hidden' && s.display !== 'none';
+              });
+              return hasCreator ? 'ready' : '';
+            }
+            """,
+            name,
+        )
+        if state == "quota":
+            raise SearchQuotaExhausted("星图达人搜索次数已用完")
+        if state == "ready":
+            return
+        page.wait_for_timeout(500)
+    raise PlaywrightTimeoutError(f"等待搜索结果超时: {name}")
+
+
+def open_creator_detail(page: Page, creator_name: str, creator_id: str | None = None) -> Page:
+    log(f"Opening creator detail: {creator_name} (id={creator_id})")
+    pages_before = list(page.context.pages)
+    # Only inspect compact result rows/cards. The old selector included giant
+    # table containers and often clicked the first unrelated creator.
+    scan_js = """
+        ({name, id}) => {
+          const visible = (el) => {
+            const r = el.getBoundingClientRect();
+            const s = window.getComputedStyle(el);
+            return r.width > 0 && r.height > 0 && s.display !== 'none' && s.visibility !== 'hidden';
+          };
+          const nicknames = [...document.querySelectorAll('.author-nickname')].filter(visible);
+          const exact = nicknames.find((el) => (el.innerText || '').trim() === name);
+          // ID searches normally return a single exact creator. If the stored
+          // nickname changed, accepting one and only one result is still safe.
+          const nickname = exact || (id && nicknames.length === 1 ? nicknames[0] : null);
+          if (!nickname) return {found: false};
+          const cell = nickname.closest('.content-cell') || nickname.parentElement;
+          const link = cell?.querySelector('a[href*="/ad/creator/author-homepage/"]');
+          if (link?.href) return {found: true, url: link.href};
+          const target = nickname;
+          target.scrollIntoView({ block: 'center', inline: 'center' });
+          target.click();
+          return {found: true, clicked: true};
+        }
+        """
+    next_page_js = """
+        () => {
+          // Element UI 分页: 找未禁用的"下一页"
+          const next = document.querySelector('.el-pagination button.btn-next:not([disabled]), button.btn-next:not([disabled])');
+          if (next) { next.click(); return true; }
+          // 兜底: antd 分页
+          const li = [...document.querySelectorAll('li[title="下一页"], [class*="pagination-next"]')]
+            .find((el) => !/disabled/.test(el.className || ''));
+          if (!li) return false;
+          const btn = li.querySelector('button') || li;
+          btn.click();
+          return true;
+        }
+        """
+    # 星图结果是模糊匹配+分页展示,精确昵称未必在第 1 页,翻几页找
+    target = None
+    for page_no in range(1, 6):
+        target = page.evaluate(scan_js, {"name": creator_name, "id": creator_id or ""})
+        if target and target.get("found"):
+            if page_no > 1:
+                log(f"    [detail] 第 {page_no} 页找到达人")
+            break
+        moved = page.evaluate(next_page_js)
+        if not moved:
+            break  # 没有下一页了
+        page.wait_for_timeout(1800)
+
+    if not target or not target.get("found"):
+        raise TimeoutError(f"搜索结果中未找到精确达人: {creator_name}")
+
+    detail_page = None
+    if target.get("url"):
+        try:
+            detail_page = page.context.new_page()
+            detail_page.goto(target["url"], wait_until="domcontentloaded", timeout=45000)
+        except Exception as exc:
+            reraise_if_browser_session_lost(exc)
+            log(f"    [detail] 直接访问达人 href 失败: {exc}")
+            try:
+                if detail_page is not None and not detail_page.is_closed():
+                    detail_page.close()
+            except Exception:
+                pass
+            detail_page = None
+
+    if detail_page is None:
+        detail_page = wait_for_detail_page(page, pages_before, timeout_ms=30000)
+    if detail_page is None:
+        raise TimeoutError("未找到达人主页")
+
+    detail_page.bring_to_front()
+    detail_page.wait_for_load_state("domcontentloaded")
+    page.wait_for_timeout(500)
+    return detail_page
+
+
+def wait_for_detail_page(page: Page, pages_before: list[Page], timeout_ms: int) -> Page | None:
+    deadline = time.time() + timeout_ms / 1000
+    known = set(pages_before)
+    while time.time() < deadline:
+        all_pages = list(page.context.pages)
+        candidates = [p for p in all_pages if p not in known] + [page] + all_pages
+        for c in candidates:
+            try:
+                if DETAIL_URL_MARK in c.url:
+                    return c
+            except Exception:
+                pass
+        page.wait_for_timeout(500)
+    return None
+
+
+def open_creation_ability(page: Page) -> bool:
+    """Open 创作能力 when available; otherwise keep scraping the current creator page."""
+    log("Opening creation ability tab.")
+    opened = False
+    try:
+        page.get_by_text("创作能力", exact=True).first.click(timeout=10000)
+        opened = True
+    except Exception as exc:
+        reraise_if_browser_session_lost(exc)
+        try:
+            opened = bool(page.evaluate(
+                """
+                () => {
+                  const t = [...document.querySelectorAll('button, div, span, a')]
+                    .find((el) => (el.innerText || '').trim() === '创作能力');
+                  if (!t) return false;
+                  t.click();
+                  return true;
+                }
+                """
+            ))
+        except Exception as fallback_exc:
+            reraise_if_browser_session_lost(fallback_exc)
+            opened = False
+
+    if opened:
+        page.wait_for_timeout(1500)
+    else:
+        log("  [主页回退] 未显示“创作能力”标签，继续从当前达人主页读取作品曝光")
+        page.wait_for_timeout(500)
+
+    try:
+        all_video = page.get_by_text("全部视频", exact=False).first
+        all_video.wait_for(timeout=3000)
+        all_video.scroll_into_view_if_needed()
+        page.wait_for_timeout(800)
+        page.mouse.wheel(0, 600)
+        page.wait_for_timeout(800)
+    except PlaywrightTimeoutError:
+        pass
+    except Exception as exc:
+        reraise_if_browser_session_lost(exc)
+    return opened
+
+
+def normalize_xingtu_api_item(item: dict[str, Any]) -> dict[str, Any]:
+    """Convert one get_author_show_items_v2 item to the common card shape."""
+    item_id = item.get("item_id") or item.get("aweme_id") or ""
+    return {
+        "title": str(item.get("item_title") or item.get("title") or "").strip(),
+        "play_count": item.get("play", ""),
+        "like_count": item.get("like", ""),
+        "comment_count": item.get("comment", ""),
+        "share_count": item.get("share", ""),
+        "href": coerce_url(item.get("url") or item.get("share_url") or ""),
+        "note_id": str(item_id).strip(),
+        "source": "show_items_api",
+    }
+
+
+def _show_items_payload_sections(payload: Any) -> list[dict[str, Any]]:
+    """Return possible containers because Xingtu has used both flat and nested payloads."""
+    if not isinstance(payload, dict):
+        return []
+    sections = [payload]
+    for key in ("data", "result"):
+        value = payload.get(key)
+        if isinstance(value, dict):
+            sections.append(value)
+    return sections
+
+
+def setup_show_items_capture(page: Page) -> list[dict[str, Any]]:
+    """Capture the complete recent-video response triggered by opening 创作能力."""
+    cards: list[dict[str, Any]] = []
+    seen: set[str] = set()
+
+    def on_response(response: Any) -> None:
+        try:
+            if "/gw/api/author/get_author_show_items_v2" not in str(response.url):
+                return
+            payload = response.json()
+        except Exception:
+            return
+
+        for section in _show_items_payload_sections(payload):
+            for field in ("latest_item_info", "latest_star_item_info"):
+                items = section.get(field) or []
+                if not isinstance(items, list):
+                    continue
+                for item in items:
+                    if not isinstance(item, dict):
+                        continue
+                    card = normalize_xingtu_api_item(item)
+                    key = card.get("note_id") or "|".join((
+                        normalize_title(card.get("title", "")),
+                        str(card.get("play_count", "")),
+                    ))
+                    if not key or key in seen:
+                        continue
+                    seen.add(key)
+                    cards.append(card)
+
+    page.on("response", on_response)
+    return cards
+
+
+def wait_for_show_items(page: Page, cards: list[dict[str, Any]], timeout_ms: int = 5000) -> None:
+    deadline = time.time() + timeout_ms / 1000
+    while not cards and time.time() < deadline:
+        page.wait_for_timeout(200)
+
+
+def parse_videos_on_page(page: Page) -> list[dict]:
+    return page.evaluate(
+        """
+        () => {
+          const clean = (v) => (v || '').replace(/\\s+/g, ' ').trim();
+          const valueAfter = (lines, label) => {
+            const idx = lines.findIndex((l) => l === label || l.includes(label));
+            if (idx < 0) return '';
+            const same = lines[idx].replace(label, '').trim();
+            return same || (lines[idx + 1] || '');
+          };
+          const parseCard = (card) => {
+            const lines = (card.innerText || '').split('\\n').map((l) => l.trim()).filter(Boolean);
+            const text = lines.join('\\n');
+            if (!/播放量/.test(text) || !/点赞量/.test(text) || !/评论量/.test(text)) return null;
+            const firstMetric = lines.findIndex((l) => /播放量/.test(l));
+            const titles = lines.slice(0, firstMetric >= 0 ? firstMetric : lines.length).filter((l) => {
+              if (/^(全部视频|全部|个人视频|星图视频|播放量|点赞量|评论量)$/.test(l)) return false;
+              if (/请输入关键词搜索|开始日期|结束日期|展开/.test(l)) return false;
+              if (/^[\\d,.wW万]+$/.test(l)) return false;
+              return l.length >= 2;
+            });
+            const titleNode = card.querySelector('.title, [class*="title"]');
+            const title = clean(
+              titleNode?.getAttribute('title') || titleNode?.innerText || titles.slice(-2).join(' ')
+            );
+            const link = card.querySelector('a[href*="/video/"], a[href*="item_id"], a[href*="modal_id"]');
+            const href = link?.href || '';
+            const idMatch = href.match(/\\/video\\/(\\d+)/)
+              || href.match(/[?&](?:modal_id|item_id|item_ids)=(\\d+)/);
+            return {
+              title,
+              play_count: clean(valueAfter(lines, '播放量')),
+              like_count: clean(valueAfter(lines, '点赞量')),
+              comment_count: clean(valueAfter(lines, '评论量')),
+              href,
+              note_id: idMatch ? idMatch[1] : '',
+            };
+          };
+          const explicit = [...document.querySelectorAll('.content-video-card, [class*="content-video-card"]')]
+            .filter((el) => /播放量/.test(el.innerText || '') && /点赞量/.test(el.innerText || ''));
+          const fallback = [...document.querySelectorAll('div')].filter((el) => {
+            const r = el.getBoundingClientRect();
+            const text = el.innerText || '';
+            return r.width >= 140 && r.width <= 340 && r.height >= 260 && r.height <= 620
+              && /播放量/.test(text) && /点赞量/.test(text) && /评论量/.test(text)
+              && el.querySelector('img');
+          });
+          const cards = explicit.length ? explicit : fallback;
+          const seen = new Set();
+          return cards.map(parseCard).filter(Boolean).filter((it) => {
+            const k = [it.title, it.play_count, it.like_count, it.comment_count].join('|');
+            if (seen.has(k)) return false;
+            seen.add(k);
+            return it.title;
+          });
+        }
+        """
+    )
+
+
+def parse_videos_stable(page: Page, max_attempts: int = 6, wait_ms: int = 800) -> list[dict]:
+    """Wait and scroll until Xingtu's lazy video-card list stops growing."""
+    previous_signature = None
+    latest: list[dict] = []
+    for _ in range(max_attempts):
+        latest = parse_videos_on_page(page)
+        signature = tuple(
+            (card.get("note_id"), card.get("title"), card.get("play_count"))
+            for card in latest
+        )
+        if signature and signature == previous_signature:
+            return latest
+        previous_signature = signature
+        try:
+            page.evaluate(
+                """() => {
+                  const cards = [...document.querySelectorAll(
+                    '.content-video-card, [class*="content-video-card"]'
+                  )];
+                  const last = cards[cards.length - 1];
+                  last?.scrollIntoView({block: 'end'});
+                  window.scrollTo(0, document.body.scrollHeight);
+                }"""
+            )
+        except Exception:
+            pass
+        page.wait_for_timeout(wait_ms)
+    return latest
+
+
+EXPLICIT_BLOCK_REASONS = frozenset({
+    "verify_text",
+    "verify_iframe",
+    "verify_div",
+})
+
+
+def is_blocked(page: Page) -> bool:
+    """只检测明确的验证/限频证据；页面缺少某个业务标签不属于全局风控。"""
+    try:
+        blocked = page.evaluate(
+            """
+            () => {
+              const txt = (document.body && document.body.innerText) || '';
+              const html = (document.documentElement && document.documentElement.outerHTML) || '';
+              if (/请完成验证|拖动滑块|访问频次过高|操作过于频繁|请稍后再试|异常请求/.test(txt)) return 'verify_text';
+              if (/iframe[^\\n]*src="[^"]*(?:alicdn|alibaba|tencent|dingtalk)[^\\n]*(?:verify|security|captcha)/i.test(html)) return 'verify_iframe';
+              if (document.querySelector('div.nc-container, div.tcaptcha-container, [class*="captcha"]:not([class*="captcha-ignore"])')) return 'verify_div';
+              return '';
+            }
+            """
+        )
+    except Exception as exc:
+        log(f"  [风控-检测异常] {exc}")
+        return False
+    if blocked in EXPLICIT_BLOCK_REASONS:
+        log(f"  [风控-检测] 命中:{blocked}")
+        return True
+    if blocked:
+        log(f"  [页面差异-忽略] {blocked} 不属于全局风控")
+    return False
+
+
+def with_captcha_retry(page: Page, fn, *args, retries: int = 1, wait_sec: int = 30, **kwargs):
+    """执行 fn;若风控触发,等 wait_sec 后重试,最多 retries 次。
+    仍被风控就抛 CaptchaBlocked。"""
+    last_result = fn(*args, **kwargs)
+    for _ in range(retries):
+        if not is_blocked(page):
+            return last_result
+        log(f"  [风控] 等 {wait_sec}s 后重试…")
+        page.wait_for_timeout(wait_sec * 1000)
+        last_result = fn(*args, **kwargs)
+    if is_blocked(page):
+        raise CaptchaBlocked("风控未解除,放弃")
+    return last_result
+
+
+class CaptchaBlocked(Exception):
+    pass
+
+
+def normalize_title(s: str) -> str:
+    return normalize_match_title(s)
+
+
+def title_match(card_title: str, target_title: str) -> bool:
+    a = normalize_title(card_title)
+    b = normalize_title(target_title)
+    if not a or not b:
+        return False
+    return title_similarity(card_title, target_title) >= 0.78
+
+
+def go_next_page(page: Page) -> bool:
+    """点下一页。
+    星图分页器 HTML (你给的截图):
+      <div class="xt-pagination xt-pagination--normal">
+        <button class="btn-prev" disabled>
+        <ul class="el-pager">
+          <li class="number active">1</li>
+          ...
+          <li class="el-icon more btn-quicknext el-icon-more"></li>
+          <li class="number">7</li>
+        </ul>
+        <button type="button" class="btn-next">
+          <i class="el-icon el-icon-arrow-right"></i>
+        </button>
+      </div>
+    策略:
+      1) 找 .btn-next (任意), 排除 disabled
+      2) 末页判定: 都不 enabled → False
+      3) 优先 .el-pagination 容器内的 (兼容 .xt-pagination)
+    """
+    # Use the full visible-card signature; the last page may contain only 1 row.
+    page_before = page.evaluate(
+        """() => {
+          const cards = [...document.querySelectorAll('.content-video-card, [class*="content-video-card"]')];
+          return cards.map((card) => (card.innerText || '').trim()).slice(0, 5).join('|');
+        }"""
+    )
+
+    moved = page.evaluate(
+        r"""
+        () => {
+          const all = [...document.querySelectorAll('.btn-next')];
+          if (all.length === 0) return {clicked: false, reason: 'no-btn-next'};
+          // 末页判定: 都不 enabled
+          const anyEnabled = all.some(b => !b.disabled);
+          if (!anyEnabled) return {clicked: false, reason: 'all-disabled'};
+          // 优先 .xt-pagination / .el-pagination 容器内的
+          let btn = all.find(b => !b.disabled && b.closest('.xt-pagination, .el-pagination'));
+          if (!btn) btn = all.find(b => !b.disabled);
+          if (!btn) return {clicked: false, reason: 'no-enabled'};
+          btn.scrollIntoView({ block: 'center', inline: 'center' });
+          btn.click();
+          return {clicked: true, picked: btn.closest('.xt-pagination, .el-pagination') ? 'pagination' : 'fallback'};
+        }
+        """
+    )
+    if not moved or not moved.get("clicked"):
+        return False
+
+    # 等页面真的变了:
+    changed = False
+    for _ in range(50):  # 50 * 200ms = 10s
+        page.wait_for_timeout(200)
+        page_after = page.evaluate(
+            """() => {
+              const cards = [...document.querySelectorAll('.content-video-card, [class*="content-video-card"]')];
+              return cards.map((card) => (card.innerText || '').trim()).slice(0, 5).join('|');
+            }"""
+        )
+        if page_after and page_after != page_before:
+            changed = True
+            break
+
+    if isinstance(moved, dict):
+        log(f"  -> 点了 btn-next ({moved.get('picked', '?')})")
+    if not changed:
+        log(f"  [warn] 页面未变或加载不完整,可能已是末页")
+    return changed
+
+
+def find_video_by_title(page: Page, target_title: str, max_pages: int = 5) -> dict | None:
+    """查找指定标题的视频。
+    新逻辑:
+      1) 先看第 1 页
+      2) 没命中 → 用"全部视频"旁的搜索框输入目标标题 → 解析搜索结果
+      3) 风控触发 → 等 30s 重试 1 次;仍触发就抛 CaptchaBlocked
+    """
+    if not target_title:
+        return None
+    if is_blocked(page):
+        log("  [风控] 进创作能力页后检测到风控,等 30s 后重试")
+        page.wait_for_timeout(30000)
+    # 第 1 步:第 1 页里找
+    cards = parse_videos_on_page(page)
+    log(f"  第 1 页抓到 {len(cards)} 个视频卡片")
+    for c in cards:
+        if title_match(c.get("title", ""), target_title):
+            log(f"  命中(第 1 页): 目标='{target_title[:30]}...' 实际='{c['title'][:30]}...'")
+            c["page"] = 1
+            return c
+
+    # 第 2 步:用"全部视频"旁的搜索框输入标题
+    log(f"  第 1 页没命中,用搜索框搜 '{target_title[:25]}...'")
+    searched = search_video_by_title_in_page(page, target_title)
+    if searched:
+        searched["page"] = "search"
+        return searched
+    return None
+
+
+def submit_video_search(page: Page, query: str) -> bool:
+    """Fill the controlled Xingtu search input and dispatch its normal events."""
+    locator = page.locator(
+        'input[placeholder*="请输入关键词"], input[placeholder*="关键词搜索"]'
+    ).first
+    try:
+        if locator.count() == 0:
+            return False
+        locator.wait_for(state="visible", timeout=5000)
+        locator.click()
+        locator.fill("")
+        locator.fill(query)
+        locator.press("Enter")
+        return True
+    except Exception as exc:
+        reraise_if_browser_session_lost(exc)
+        return False
+
+
+def search_video_by_title_in_page(page: Page, target_title: str) -> dict | None:
+    """在博主主页的"全部视频"搜索框输入标题,等结果"""
+    # 找搜索框:星图创作能力页的 "请输入关键词搜索"
+    # selector: input.xt-input__inner / input[placeholder*="请输入关键词"]
+    cleared = submit_video_search(page, target_title)
+    if not cleared:
+        # 搜索框没了 → 多半是风控遮罩挡住了。等 30s 再试一次;还不行就抛 CaptchaBlocked
+        if is_blocked(page):
+            log("  [风控] 搜索前检测到风控,等 30s 重试…")
+            page.wait_for_timeout(30000)
+            recleared = submit_video_search(page, target_title)
+            if not recleared:
+                raise CaptchaBlocked("搜索框缺失且风控未解除")
+        else:
+            log("  [warn] 找不到搜索框(非风控)")
+            return None
+
+    # 等搜索结果渲染
+    page.wait_for_timeout(2500)
+
+    # 解析搜索结果 (跟普通页一样用 parse_videos_on_page)
+    cards = parse_videos_stable(page)
+    log(f"  搜索结果: {len(cards)} 张卡片")
+    for c in cards:
+        if title_match(c.get("title", ""), target_title):
+            log(f"  命中(搜索): 目标='{target_title[:30]}...' 实际='{c['title'][:30]}...'")
+            return c
+
+    log(f"  [未命中] 搜索框搜不到")
+    return None
+
+
+def find_videos_by_titles(page: Page, target_titles: list[str],
+                           max_pages: int = 5) -> dict[str, dict]:
+    """查找指定标题的多个视频，使用页面内搜索框。
+    策略: 1) 检查第1页 2) 对每个未匹配标题用搜索框搜索
+    返回 {normalize_title(target): card_dict}
+    """
+    if not target_titles:
+        return {}
+
+    remaining: dict[str, str] = {}  # norm_title -> original_title
+    for t in target_titles:
+        if t:
+            remaining[normalize_title(t)] = t
+
+    found: dict[str, dict] = {}
+
+    # Step 1: check page 1
+    if is_blocked(page):
+        log("  [风控] 进创作能力页后检测到风控,等 30s 后重试")
+        page.wait_for_timeout(30000)
+
+    cards = parse_videos_on_page(page)
+    log(f"  第 1 页抓到 {len(cards)} 个视频卡片")
+
+    for c in cards:
+        card_norm = normalize_title(c.get("title", ""))
+        for target_norm in list(remaining.keys()):
+            if card_norm == target_norm or (len(target_norm) >= 6 and (target_norm in card_norm or card_norm in target_norm)):
+                found[target_norm] = {**c, "page": 1}
+                del remaining[target_norm]
+                break
+
+    if not remaining:
+        return found
+
+    # Step 2: use search box for each remaining title
+    for target_norm, original_title in list(remaining.items()):
+        log(f"  用搜索框搜 '{original_title[:25]}...'")
+        cleared = page.evaluate("""
+            () => {
+                const inputs = [...document.querySelectorAll('input')];
+                const target = inputs.find(i => /请输入关键词|关键词搜索/.test(i.placeholder || ''));
+                if (!target) return false;
+                target.focus();
+                target.value = '';
+                return true;
+            }
+        """)
+        if not cleared:
+            if is_blocked(page):
+                log("  [风控] 搜索前检测到风控,等 30s 重试...")
+                page.wait_for_timeout(30000)
+                cleared = page.evaluate("""
+                    () => {
+                        const inputs = [...document.querySelectorAll('input')];
+                        const target = inputs.find(i => /请输入关键词|关键词搜索/.test(i.placeholder || ''));
+                        if (!target) return false;
+                        target.focus();
+                        target.value = '';
+                        return true;
+                    }
+                """)
+                if not cleared:
+                    raise CaptchaBlocked("搜索框缺失且风控未解除")
+            else:
+                log("  [warn] 找不到搜索框(非风控)")
+                continue
+
+        page.wait_for_timeout(300)
+        page.keyboard.type(original_title, delay=20)
+        page.keyboard.press("Enter")
+        page.wait_for_timeout(2500)
+
+        cards = parse_videos_on_page(page)
+        log(f"  搜索 '{original_title[:20]}...' 结果: {len(cards)} 张卡片")
+
+        for c in cards:
+            card_norm = normalize_title(c.get("title", ""))
+            if card_norm == target_norm or (len(target_norm) >= 6 and (target_norm in card_norm or card_norm in target_norm)):
+                found[target_norm] = {**c, "page": "search"}
+                if target_norm in remaining:
+                    del remaining[target_norm]
+                break
+
+    if remaining:
+        log(f"  未找到 {len(remaining)} 个视频")
+    return found
+
+
+def find_videos_for_tasks(
+    page: Page,
+    tasks: list[dict],
+    max_pages: int = 5,
+    api_cards: list[dict] | None = None,
+) -> tuple[dict[str, dict], list[str], bool]:
+    """Match the complete API list first, then use DOM pagination/search as fallback."""
+    matching_tasks = [dict(task) for task in tasks if task.get("record_id")]
+    task_by_id = {str(task["record_id"]): task for task in matching_tasks}
+    remaining = dict(task_by_id)
+    all_cards: list[dict] = []
+    page_limit_hit = False
+
+    def match_all(min_score: float = 0.78) -> dict[str, dict]:
+        return match_tasks_to_cards(
+            matching_tasks,
+            deduplicate_observed_cards(all_cards),
+            "douyin",
+            min_score=min_score,
+        )
+
+    for raw_card in api_cards or []:
+        card = dict(raw_card)
+        card.setdefault("page", "api")
+        if not card.get("note_id"):
+            card["note_id"] = extract_content_id(card.get("href"), "douyin")
+        all_cards.append(card)
+
+    if all_cards:
+        api_matches = match_all(min_score=1.0)
+        safe_ids = {
+            record_id for record_id, card in api_matches.items()
+            if card.get("match_method") == "content_id"
+        }
+        remaining = {record_id: task for record_id, task in task_by_id.items() if record_id not in safe_ids}
+        log(f"  星图作品接口返回 {len(all_cards)} 条，直接命中 {len(api_matches)} 条")
+        if not remaining:
+            candidates = [
+                str(card.get("title") or "")[:80]
+                for card in deduplicate_observed_cards(all_cards)
+                if card.get("title")
+            ]
+            return api_matches, candidates[-10:], False
+
+    if is_blocked(page):
+        log("  [风控] 创作能力页触发验证,等待 30s 后复检")
+        page.wait_for_timeout(30000)
+        if is_blocked(page):
+            raise CaptchaBlocked("创作能力页风控未解除")
+
+    for page_number in range(1, max_pages + 1):
+        cards = parse_videos_stable(page)
+        for card in cards:
+            card = dict(card)
+            card["page"] = page_number
+            if not card.get("note_id"):
+                card["note_id"] = extract_content_id(card.get("href"), "douyin")
+            all_cards.append(card)
+        log(f"  第 {page_number} 页稳定抓到 {len(cards)} 个视频卡片 (还剩 {len(remaining)} 个目标)")
+
+        exact = match_all(min_score=1.0)
+        safe_ids = {
+            record_id for record_id, card in exact.items()
+            if card.get("match_method") == "content_id"
+        }
+        remaining = {record_id: task for record_id, task in task_by_id.items() if record_id not in safe_ids}
+        if not remaining:
+            break
+        if page_number == max_pages:
+            page_limit_hit = True
+            break
+        if not go_next_page(page):
+            break
+
+    # Once all unfiltered pages have been seen, use the component's search box
+    # for the remaining titles. fill() clears the previous controlled value.
+    for record_id, task in list(remaining.items()):
+        if record_id not in remaining:
+            continue
+        title = str(task.get("target_title") or "")
+        if not title or not submit_video_search(page, title):
+            continue
+        page.wait_for_timeout(2500)
+        cards = parse_videos_stable(page)
+        for card in cards:
+            card = dict(card)
+            card["page"] = "search"
+            if not card.get("note_id"):
+                card["note_id"] = extract_content_id(card.get("href"), "douyin")
+        all_cards.extend(cards)
+        exact = match_all(min_score=1.0)
+        safe_ids = {
+            rid for rid, card in exact.items()
+            if card.get("match_method") == "content_id"
+        }
+        remaining = {rid: candidate for rid, candidate in task_by_id.items() if rid not in safe_ids}
+
+    # Search can return an incomplete candidate set, so make one final safe
+    # unique-best pass over every card observed in the session.
+    found = match_all()
+    remaining = {record_id: task for record_id, task in task_by_id.items() if record_id not in found}
+
+    candidates = [
+        str(card.get("title") or "")[:80]
+        for card in deduplicate_observed_cards(all_cards)
+        if card.get("title")
+    ]
+    if remaining:
+        log(f"  未找到 {len(remaining)} 个视频;候选卡片 {len(all_cards)} 张")
+    return found, candidates[-10:], page_limit_hit
+
+
+def extract_target_tasks(style: dict, only_record_ids: set[str] | None = None,
+                          platform_filter: str = "抖音",
+                          require_all: bool = True) -> list[dict]:
+    base_token = style["base_token"]
+    table_id = style["table_id"]
+    fmap = style["field_map"]
+
+    name_fid = fmap.get("creator_name", {}).get("field_id")
+    id_fid = fmap.get("creator_id", {}).get("field_id")
+    title_fid = fmap.get("note_title", {}).get("field_id")
+    pub_fid = fmap.get("publish_time", {}).get("field_id")
+    url_fid = fmap.get("note_url", {}).get("field_id")
+    platform_fid = fmap.get("platform", {}).get("field_id")
+    read7_fid = fmap.get("read_count_7d", {}).get("field_id")
+
+    missing_fields = []
+    if not (name_fid or id_fid):
+        missing_fields.append("creator_name/creator_id")
+    for logical, field_id in (
+        ("note_title", title_fid),
+        ("publish_time", pub_fid),
+        ("note_url", url_fid),
+        ("platform", platform_fid),
+    ):
+        if not field_id:
+            missing_fields.append(logical)
+    if missing_fields:
+        error = f"missing field mapping: {','.join(missing_fields)}"
+        style["_extract_error"] = error
+        style["_extract_stats"] = {
+            "source_records": 0,
+            "selected_scope_records": 0,
+            "skipped_platform": 0,
+            "skipped_incomplete": 0,
+        }
+        log(f"  [ERROR] {style['name']} {error}")
+        return []
+    style.pop("_extract_error", None)
+
+    log(f"  拉取款式 [{style['index']:02d} {style['name']}] 的待抓列表 (平台={platform_filter})...")
+    all_records = list_records_by_table(base_token, table_id)
+    log(f"    全表共 {len(all_records)} 条记录")
+
+    def is_platform_match(v) -> bool:
+        if not v:
+            return False
+        if isinstance(v, list):
+            return any(platform_filter in str(x) for x in v)
+        return platform_filter in str(v)
+
+    def is_nonempty(v) -> bool:
+        if v is None:
+            return False
+        if isinstance(v, list) and len(v) == 0:
+            return False
+        if isinstance(v, str) and v.strip() == "":
+            return False
+        return True
+
+    tasks = []
+    selected_scope_records = 0
+    skipped_platform = 0
+    skipped_incomplete = 0
+    for rec in all_records:
+        rid = rec["record_id"]
+        if only_record_ids and rid not in only_record_ids:
+            continue
+        selected_scope_records += 1
+        if platform_fid:
+            if not is_platform_match(rec.get(platform_fid)):
+                skipped_platform += 1
+                continue
+        name = rec.get(name_fid) if name_fid else None
+        creator_id_val = rec.get(id_fid) if id_fid else None
+        title = rec.get(title_fid) if title_fid else None
+        pub = rec.get(pub_fid) if pub_fid else None
+        url = rec.get(url_fid) if url_fid else None
+        note_url = coerce_url(url)
+        note_id = extract_content_id(note_url, "douyin")
+        if require_all:
+            name_or_id = is_nonempty(name) or is_nonempty(creator_id_val)
+            title_or_id = is_nonempty(title) or bool(note_id)
+            if not (name_or_id and title_or_id and is_nonempty(pub) and is_nonempty(note_url)):
+                skipped_incomplete += 1
+                continue
+        if not (is_nonempty(name) or is_nonempty(creator_id_val)):
+            continue
+        existing = rec.get(read7_fid) if read7_fid else None
+        pub_dt = parse_publish_time(pub)
+        tasks.append({
+            "record_id": rid,
+            "creator_name": name if isinstance(name, str) else (str(name) if name else ""),
+            "creator_id": str(creator_id_val) if creator_id_val else None,
+            "target_title": title if isinstance(title, str) else (str(title) if title else None),
+            "publish_time_raw": pub,
+            "publish_time": pub_dt,
+            "note_url": note_url,
+            "note_id": note_id,
+            # 发布链接仅作为“已发布”的触发标记；实际采集按达人和标题匹配。
+            "input_error": None,
+            "existing_read_7d": existing,
+        })
+
+    log(f"    平台过滤: 跳过 {skipped_platform} 条非 {platform_filter} 记录")
+    log(f"    完整性过滤: 跳过 {skipped_incomplete} 条不完整记录")
+    log(f"    实际待抓: {len(tasks)} 条")
+    style["_extract_stats"] = {
+        "source_records": len(all_records),
+        "selected_scope_records": selected_scope_records,
+        "skipped_platform": skipped_platform,
+        "skipped_incomplete": skipped_incomplete,
+    }
+    return tasks
+
+
+def group_tasks_by_creator(tasks: list[dict]) -> list[dict]:
+    return _group_tasks_by_creator(tasks)
+
+
+def collect_tasks_across_styles(
+    styles: list[dict],
+    only_record_ids: set[str] | dict[int, set[str]] | None,
+) -> tuple[list[dict], dict[int, dict]]:
+    """Read every selected style first and attach exact write-back context."""
+    all_tasks: list[dict] = []
+    summaries: dict[int, dict] = {}
+    for style in styles:
+        style_record_ids = (
+            only_record_ids.get(style["index"], set())
+            if isinstance(only_record_ids, dict)
+            else only_record_ids
+        )
+        tasks = extract_target_tasks(style, style_record_ids)
+        summary = {
+            "style": style["name"], "index": style["index"], "total": len(tasks),
+            "run_started_at": time.time(),
+            "matched": 0, "filled": 0, "skipped_no_pubtime": 0, "results": [],
+            **style.get("_extract_stats", {}),
+        }
+        if style.get("_extract_error"):
+            summary["error"] = style["_extract_error"]
+        summaries[style["index"]] = summary
+        context = {
+            "style": style["name"], "index": style["index"],
+            "base_token": style["base_token"], "table_id": style["table_id"],
+            "field_map": style["field_map"],
+        }
+        for task in tasks:
+            task["style_context"] = context
+            if task.get("input_error"):
+                upsert_result(summary, {
+                    "record_id": task["record_id"],
+                    "creator": task.get("creator_name"),
+                    "target": task.get("target_title"),
+                    "status": BLOCKED_INPUT,
+                    "matched": False,
+                    "reason": task["input_error"],
+                    "note_url": task.get("note_url"),
+                })
+            else:
+                all_tasks.append(task)
+    return all_tasks, summaries
+
+
+def apply_matched_task(task: dict, matched: dict, summary: dict,
+                       dry_run: bool) -> dict:
+    """Apply one matched video using the task's own style write-back context."""
+    context = task["style_context"]
+    summary["matched"] += 1
+    log(f"    播放={matched.get('play_count')}, 点赞={matched.get('like_count')}, 评论={matched.get('comment_count')}")
+    pick = pick_read_field(context["field_map"], task.get("publish_time"))
+    play_val = parse_int(matched.get("play_count"))
+    if play_val is None:
+        return {
+            "status": WRITE_FAILURE,
+            "matched": True,
+            "write_ok": False,
+            "reason": "metric_parse_failed",
+        }
+    if dry_run:
+        return {"status": SUCCESS, "matched": True, "write_ok": None, "dry_run": True}
+    if pick and pick != "FUTURE" and not dry_run:
+        logical, fill_fid, fill_fname = pick
+        ok = write_back(
+            context["base_token"], context["table_id"],
+            task["record_id"], fill_fid, play_val,
+        )
+        if ok:
+            log(f"    回填 {fill_fname}={play_val} (key={logical}) 到 record={task['record_id']}")
+            summary["filled"] += 1
+            return {"status": SUCCESS, "matched": True, "write_ok": True}
+        return {
+            "status": WRITE_FAILURE,
+            "matched": True,
+            "write_ok": False,
+            "reason": "write_back_failed",
+        }
+    elif pick == "FUTURE":
+        log(f"    [SKIP] record={task['record_id']} 发布时间在未来,跳过回填")
+        return {
+            "status": BLOCKED_INPUT,
+            "matched": True,
+            "write_ok": None,
+            "reason": "publish_time_in_future",
+        }
+    elif not dry_run and not task.get("publish_time"):
+        log(f"    [SKIP] record={task['record_id']} 发布时间缺失,无法判断槽位")
+        summary["skipped_no_pubtime"] += 1
+        return {
+            "status": BLOCKED_INPUT,
+            "matched": True,
+            "write_ok": None,
+            "reason": "publish_time_missing",
+        }
+    return {
+        "status": WRITE_FAILURE,
+        "matched": True,
+        "write_ok": False,
+        "reason": "write_field_missing",
+    }
+
+
+def parse_int(s) -> int | None:
+    if s is None:
+        return None
+    if isinstance(s, (int, float)):
+        return int(s)
+    txt = str(s).strip()
+    if not txt:
+        return None
+    m = re.match(r'^([\d,.]+)\s*(万|w|W)?', txt)
+    if not m:
+        return None
+    num = float(m.group(1).replace(',', ''))
+    if m.group(2):
+        num *= 10000
+    return int(num)
+
+
+def scrape_styles(styles: list[dict], login_timeout: int, headless: bool,
+                  only_record_ids: set[str] | None, dry_run: bool,
+                  batch_size: int = DEFAULT_BATCH_SIZE,
+                  batch_wait: int = DEFAULT_BATCH_WAIT,
+                  checkpoint_callback=None) -> list[dict]:
+    """Collect all styles, then search every creator only once in this run."""
+    tasks, summaries = collect_tasks_across_styles(styles, only_record_ids)
+    if not tasks:
+        log("  所有款式均无待抓任务,跳过")
+        return [summaries[s["index"]] for s in styles]
+
+    creator_groups = group_tasks_by_creator(tasks)
+    log(f"  全部 {len(styles)} 个款式、{len(tasks)} 条任务 -> {len(creator_groups)} 个唯一达人")
+    search_budget = CreatorSearchBudget()
+
+    batches = chunk_creator_groups(creator_groups, batch_size)
+    if len(batches) > 1:
+        log(f"  [BATCH] 分 {len(batches)} 批处理,每批最多 {batch_size} 个达人,批间等待 {batch_wait}s")
+
+    PROFILE_DIR.mkdir(parents=True, exist_ok=True)
+
+    for batch_idx, batch in enumerate(batches):
+        if batch_idx > 0:
+            log(f"  [BATCH WAIT] 等待 {batch_wait // 60} 分钟后开始第 {batch_idx+1}/{len(batches)} 批...")
+            remaining_creators = sum(len(b) for b in batches[batch_idx:])
+            log(f"  [BATCH WAIT] 还剩 {remaining_creators} 个达人待处理")
+            wait_with_heartbeat(batch_wait)
+
+        log(f"  [BATCH {batch_idx+1}/{len(batches)}] 开始处理 {len(batch)} 个达人")
+
+        def action(page: Page) -> None:
+            ensure_login(page, login_timeout)
+            for ci, creator_group in enumerate(batch, start=1):
+                creator_name = creator_group["creator_name"]
+                creator_id = creator_group["creator_id"]
+                group_tasks = creator_group["tasks"]
+                if all(
+                    any(
+                        row.get("record_id") == task["record_id"]
+                        for row in summaries[task["style_context"]["index"]].get("results", [])
+                        if isinstance(row, dict)
+                    )
+                    for task in group_tasks
+                ):
+                    continue
+                log(f"  [达人 {ci}/{len(batch)}] {creator_name} ({len(group_tasks)} 条视频)")
+
+                def mark_group_failure(reason: str, error: str | None = None) -> None:
+                    for task in group_tasks:
+                        summary = summaries[task["style_context"]["index"]]
+                        payload = {
+                            "record_id": task["record_id"],
+                            "creator": creator_name,
+                            "target": task.get("target_title"),
+                            "status": RETRYABLE_FAILURE,
+                            "matched": False,
+                            "reason": reason,
+                        }
+                        if error:
+                            payload["error"] = error
+                        upsert_result(summary, payload)
+
+                creator_key = creator_search_key(creator_name, creator_id)
+                if search_budget.circuit_reason:
+                    log(f"    [熔断-跳过] 不再消耗达人搜索次数: {search_budget.circuit_reason}")
+                    mark_group_failure("creator_search_circuit_open", search_budget.circuit_reason)
+                    continue
+                detail_page = None
+                try:
+                    # Navigate to market (between creator groups)
+                    main_page = page
+                    main_page.bring_to_front()
+                    for p in list(main_page.context.pages):
+                        if p is not main_page:
+                            try: p.close()
+                            except: pass
+                    try:
+                        main_page.goto(MARKET_URL, wait_until="domcontentloaded", timeout=30000)
+                        main_page.wait_for_timeout(2000)
+                    except Exception as exc:
+                        reraise_if_browser_session_lost(exc)
+                        log(f"    [WARN] 重置星图首页失败: {exc}")
+                    dismiss_popups(main_page)
+
+                    # Count only a search that is about to be submitted. If the
+                    # browser dies while resetting the market page, a rebuilt
+                    # session may still make this creator's one allowed search.
+                    if not search_budget.begin(creator_key):
+                        log("    [跳过] 本轮已搜索过该达人,禁止重复搜索")
+                        mark_group_failure("creator_search_already_attempted")
+                        continue
+
+                    try:
+                        detail_page = search_and_open_creator_once(
+                            main_page, creator_name, creator_id,
+                        )
+                        search_budget.succeeded()
+                    except SearchQuotaExhausted as quota_exc:
+                        search_budget.exhaust("search_quota_exhausted")
+                        log(f"    [额度熔断] {quota_exc};剩余达人本轮不再搜索")
+                        mark_group_failure("search_quota_exhausted", str(quota_exc))
+                        continue
+                    except Exception as open_exc:
+                        detail_page = None
+                        if is_browser_session_lost(open_exc):
+                            mark_group_failure(
+                                "browser_session_lost_after_search", str(open_exc),
+                            )
+                            raise
+                        search_budget.failed("creator_search_failure")
+                        log(f"    [WARN] 达人搜索/打开失败,本轮不重复搜索: {open_exc}")
+
+                    if detail_page is None:
+                        log(f"    [SKIP] 未找到达人 '{creator_name}' 的主页")
+                        mark_group_failure("no_detail_page")
+                        continue
+
+                    api_cards = setup_show_items_capture(detail_page)
+                    creation_tab_opened = open_creation_ability(detail_page)
+                    if not creation_tab_opened:
+                        log("    [主页回退] 使用当前达人主页的作品接口和作品卡片继续匹配")
+                    wait_for_show_items(detail_page, api_cards)
+                    if api_cards:
+                        log(f"    星图完整作品列表已加载: {len(api_cards)} 条")
+
+                    # Find ALL videos for this creator
+                    found_videos, candidates, page_limit_hit = find_videos_for_tasks(
+                        detail_page, group_tasks, max_pages=5, api_cards=api_cards,
+                    )
+
+                    for task in group_tasks:
+                        context = task["style_context"]
+                        summary = summaries[context["index"]]
+                        target_title = task.get("target_title", "")
+                        matched = found_videos.get(str(task["record_id"]))
+
+                        if matched:
+                            outcome = apply_matched_task(task, matched, summary, dry_run)
+                            upsert_result(summary, {
+                                "record_id": task["record_id"],
+                                "creator": creator_name,
+                                "target": target_title,
+                                "note_url": task.get("note_url"),
+                                "note_id": task.get("note_id"),
+                                "matched_title": matched["title"],
+                                "play_count": matched.get("play_count"),
+                                "like_count": matched.get("like_count"),
+                                "comment_count": matched.get("comment_count"),
+                                "page": matched.get("page"),
+                                "match_method": matched.get("match_method"),
+                                "match_score": matched.get("match_score"),
+                                **outcome,
+                            })
+                        else:
+                            log(f"    [未命中] 未找到视频 '{target_title[:30] if target_title else ''}'")
+                            upsert_result(summary, {
+                                "record_id": task["record_id"],
+                                "creator": creator_name,
+                                "target": target_title,
+                                "note_url": task.get("note_url"),
+                                "note_id": task.get("note_id"),
+                                "status": RETRYABLE_FAILURE,
+                                "matched": False,
+                                "reason": "page_limit_reached" if page_limit_hit else "title_unmatched",
+                                "candidates": candidates,
+                            })
+
+                except CaptchaBlocked as exc:
+                    search_budget.exhaust("captcha_blocked")
+                    log(f"    [风控熔断] {exc};剩余达人本轮不再搜索")
+                    for task in group_tasks:
+                        summary = summaries[task["style_context"]["index"]]
+                        existing_rids = {r.get("record_id") for r in summary["results"]}
+                        if task["record_id"] not in existing_rids:
+                            upsert_result(summary, {
+                                "record_id": task["record_id"],
+                                "creator": creator_name,
+                                "target": task["target_title"],
+                                "status": RETRYABLE_FAILURE,
+                                "matched": False,
+                                "reason": "captcha_blocked",
+                            })
+                except Exception as exc:
+                    if is_browser_session_lost(exc):
+                        log(f"    [SESSION LOST] {exc}")
+                        raise
+                    log(f"    [ERROR] {exc}")
+                    for task in group_tasks:
+                        summary = summaries[task["style_context"]["index"]]
+                        existing_rids = {r.get("record_id") for r in summary["results"]}
+                        if task["record_id"] not in existing_rids:
+                            upsert_result(summary, {
+                                "record_id": task["record_id"],
+                                "creator": creator_name,
+                                "target": task.get("target_title"),
+                                "status": RETRYABLE_FAILURE,
+                                "matched": False,
+                                "reason": "browser_or_scrape_error",
+                                "error": str(exc),
+                            })
+                finally:
+                    try: save_state(detail_page)
+                    except: pass
+                    try:
+                        if detail_page is not None and detail_page is not page and not detail_page.is_closed():
+                            detail_page.close()
+                    except Exception:
+                        pass
+                    if checkpoint_callback:
+                        checkpoint_callback([
+                            finalize_summary(summaries[s["index"]], dry_run=dry_run)
+                            for s in styles
+                        ])
+
+        def pending_groups() -> list[dict]:
+            pending = []
+            for group in batch:
+                if any(
+                    not any(
+                        row.get("record_id") == task["record_id"]
+                        for row in summaries[task["style_context"]["index"]].get("results", [])
+                        if isinstance(row, dict)
+                    )
+                    for task in group["tasks"]
+                ):
+                    pending.append(group)
+            return pending
+
+        for session_attempt in range(1, 4):
+            if not pending_groups():
+                break
+            session = None
+            try:
+                with DynamicSession(
+                    headless=headless,
+                    real_chrome=True,
+                    user_data_dir=str(PROFILE_DIR),
+                    locale="zh-CN",
+                    timezone_id="Asia/Shanghai",
+                    timeout=60000,
+                    network_idle=False,
+                    disable_resources=False,
+                    google_search=False,
+                    page_setup=restore_cookies,
+                    max_pages=3,
+                ) as session:
+                    session.fetch(MARKET_URL, page_action=action, wait=1000)
+            except Exception as exc:
+                if not is_browser_session_lost(exc):
+                    raise
+                log(f"  [WARN] 浏览器会话丢失,准备重建 ({session_attempt}/3): {exc}")
+            finally:
+                _force_cleanup_session(session)
+            if pending_groups() and session_attempt < 3:
+                log(f"  [RECOVER] 新建浏览器继续剩余 {len(pending_groups())} 个达人")
+
+        remaining = pending_groups()
+        if remaining:
+            log(f"  [ERROR] 浏览器重建 3 次后仍有 {len(remaining)} 个达人未处理")
+            for group in remaining:
+                for task in group["tasks"]:
+                    summary = summaries[task["style_context"]["index"]]
+                    upsert_result(summary, {
+                        "record_id": task["record_id"],
+                        "creator": group.get("creator_name"),
+                        "target": task.get("target_title"),
+                        "status": RETRYABLE_FAILURE,
+                        "matched": False,
+                        "reason": "browser_session_lost",
+                    })
+            if checkpoint_callback:
+                checkpoint_callback([
+                    finalize_summary(summaries[s["index"]], dry_run=dry_run)
+                    for s in styles
+                ])
+
+    log(
+        "  星图达人搜索统计: "
+        f"唯一达人={len(creator_groups)},实际搜索={search_budget.calls},"
+        f"未搜索={max(0, len(creator_groups) - search_budget.calls)},"
+        f"熔断原因={search_budget.circuit_reason or '无'}"
+    )
+    return [finalize_summary(summaries[s["index"]], dry_run=dry_run) for s in styles]
+
+
+def scrape_one_style(style: dict, login_timeout: int, headless: bool,
+                     only_record_ids: set[str] | None, dry_run: bool,
+                     batch_size: int = DEFAULT_BATCH_SIZE,
+                     batch_wait: int = DEFAULT_BATCH_WAIT) -> dict:
+    """Backward-compatible single-style entry point."""
+    return scrape_styles(
+        [style], login_timeout, headless, only_record_ids, dry_run,
+        batch_size=batch_size, batch_wait=batch_wait,
+    )[0]
+
+
+def _force_cleanup_session(session) -> None:
+    """强制关闭 playwright 会话并清理可能泄漏的 asyncio loop。"""
+    if session is not None:
+        try:
+            if getattr(session, "context", None):
+                session.context.close()
+        except Exception:
+            pass
+        try:
+            if getattr(session, "browser", None):
+                session.browser.close()
+        except Exception:
+            pass
+        try:
+            if getattr(session, "playwright", None):
+                session.playwright.stop()
+        except Exception:
+            pass
+    try:
+        import asyncio
+        loop = asyncio.get_event_loop()
+        if loop.is_running():
+            for t in list(asyncio.all_tasks(loop)):
+                if not t.done():
+                    t.cancel()
+        else:
+            loop.close()
+            asyncio.set_event_loop(asyncio.new_event_loop())
+    except Exception:
+        pass
+    # 清掉僵尸 running loop 的线程本地标记(get_running_loop 读的是它,
+    # set_event_loop 管不到), 否则下一个 sync_playwright() 会报
+    # "Sync API inside the asyncio loop"
+    try:
+        import asyncio
+        asyncio.get_running_loop()
+    except RuntimeError:
+        pass
+    except Exception:
+        pass
+    else:
+        try:
+            asyncio.events._set_running_loop(None)  # 无公开 API, 只能访问私有函数
+        except Exception:
+            pass
+
+
+def login_only(login_timeout: int) -> int:
+    """只触发登录(headed),登录完成立刻退出。给 weekly_relogin.py 用。"""
+    login_completed = False
+
+    def action(page):
+        nonlocal login_completed
+        ensure_login(page, login_timeout)
+        if not is_logged_in(page):
+            raise RuntimeError("星图登录状态验证失败")
+        save_state(page)
+        login_completed = True
+        return page
+
+    session = None
+    try:
+        with DynamicSession(
+            headless=False,
+            real_chrome=True,
+            user_data_dir=str(PROFILE_DIR),
+            locale="zh-CN",
+            timezone_id="Asia/Shanghai",
+            timeout=60000,
+            network_idle=False,
+            disable_resources=False,
+            google_search=False,
+            page_setup=restore_cookies,
+            max_pages=2,
+        ) as session:
+            session.fetch(MARKET_URL, page_action=action, wait=1000)
+    except Exception as exc:
+        err_msg = str(exc).lower()
+        if "closed" in err_msg or "context" in err_msg or "target" in err_msg:
+            log(f"[login-only] 浏览器会话关闭异常: {exc}")
+        else:
+            log(f"[login-only] 登录执行失败: {exc}")
+    finally:
+        _force_cleanup_session(session)
+    if not login_completed or not has_valid_login_cookie_file():
+        log("[login-only] 登录未完成或Cookie缺少有效会话")
+        return 1
+    log("[login-only] 登录完成,退出")
+    return 0
+
+
+def load_retry_scope(summary_path: Path, reason: str) -> dict[int, set[str]]:
+    """Load exact style/record pairs for a reason-limited retry."""
+    try:
+        summaries = json.loads(summary_path.read_text(encoding="utf-8"))
+    except Exception:
+        return {}
+    scope: dict[int, set[str]] = {}
+    for summary in summaries if isinstance(summaries, list) else []:
+        if not isinstance(summary, dict) or summary.get("index") is None:
+            continue
+        record_ids = {
+            str(row.get("record_id"))
+            for row in summary.get("results") or []
+            if isinstance(row, dict)
+            and row.get("record_id")
+            and row.get("reason") == reason
+        }
+        if record_ids:
+            scope[int(summary["index"])] = record_ids
+    return scope
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="星图 V2: 多维表格驱动精准抓取+回填")
+    add_repeatable_style_argument(parser, "只跑指定款式编号(可多次)")
+    parser.add_argument("--record", action="append", help="只跑指定 record_id")
+    parser.add_argument("--retry-reason", help="按上次汇总中的失败原因精确续跑(款式+record_id)")
+    parser.add_argument("--headless", action="store_true")
+    parser.add_argument("--login-timeout", type=int, default=300)
+    parser.add_argument("--dry-run", action="store_true")
+    parser.add_argument("--data-dir", default=str(PATHS.normalized_root))
+    parser.add_argument("--login-only", action="store_true",
+                        help="只触发登录(headed),登录完成立刻退出,给 weekly_relogin 用")
+    parser.add_argument("--self-operated", action="store_true",
+                        help="抓取自营达人表格(而非合作达人)")
+    parser.add_argument("--batch-size", type=int, default=DEFAULT_BATCH_SIZE,
+                        help="每个批次处理的达人数(默认 0=不分批)")
+    parser.add_argument("--batch-wait", type=int, default=DEFAULT_BATCH_WAIT,
+                        help="批次间等待秒数(默认 0)")
+    args = parser.parse_args()
+
+    if args.login_only:
+        return login_only(args.login_timeout)
+
+    mapping = load_mapping(self_operated=args.self_operated)
+    styles = mapping["tables"]
+    if args.style:
+        try:
+            styles = select_requested_styles(styles, args.style)
+        except ValueError as exc:
+            print(str(exc))
+            return 1
+
+    only_rids = set(args.record) if args.record else None
+
+    data_dir = resolve_layer_output(
+        args.data_dir, layer_root=PATHS.normalized_root, field="--data-dir"
+    )
+    out_dir = V2_DIR if data_dir == PATHS.normalized_root.resolve() else data_dir / "v2_results"
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    self_sfx = "_self" if args.self_operated else ""
+    final_path = out_dir / f"_all_summaries{self_sfx}_xingtu.json"
+    if args.retry_reason:
+        if args.record:
+            parser.error("--retry-reason 不能与 --record 同时使用")
+        only_rids = load_retry_scope(final_path, args.retry_reason)
+        styles = [style for style in styles if style["index"] in only_rids]
+        retry_records = sum(len(ids) for ids in only_rids.values())
+        if not styles or retry_records == 0:
+            log(f"没有 reason={args.retry_reason} 的待续跑记录")
+            return 0
+        log(
+            f"精确续跑 reason={args.retry_reason}: "
+            f"{len(styles)} 个款式、{retry_records} 个款式内记录"
+        )
+
+    def persist_summaries(current_summaries: list[dict]) -> None:
+        updates: list[dict] = []
+        for style, raw_summary in zip(styles, current_summaries):
+            summary = finalize_summary(raw_summary, dry_run=args.dry_run)
+            out_file = out_dir / f"{style['index']:02d}-{style['name']}{self_sfx}_xingtu_v2.json"
+            if only_rids and out_file.exists():
+                try:
+                    previous = json.loads(out_file.read_text(encoding="utf-8"))
+                except Exception:
+                    previous = None
+                summary = merge_style_summary(previous, summary, dry_run=args.dry_run)
+            atomic_write_json(out_file, summary)
+            updates.append(summary)
+        try:
+            existing_global = json.loads(final_path.read_text(encoding="utf-8")) if final_path.exists() else []
+        except Exception:
+            existing_global = []
+        atomic_write_json(final_path, merge_global_summaries(existing_global, updates))
+
+    log(f"\n========== 星图全局采集 {len(styles)} 个款式 ==========")
+    all_summaries = scrape_styles(
+        styles, args.login_timeout, args.headless, only_rids, args.dry_run,
+        batch_size=args.batch_size, batch_wait=args.batch_wait,
+        checkpoint_callback=persist_summaries,
+    )
+    persist_summaries(all_summaries)
+    for s, summary in zip(styles, all_summaries):
+        out_file = out_dir / f"{s['index']:02d}-{s['name']}{self_sfx}_xingtu_v2.json"
+        log(f"  落盘: {out_file}")
+        log(
+            f"  统计: 总任务={summary['total']} 命中={summary['matched']} "
+            f"回填={summary['filled']} 未解决={summary.get('unresolved', 0)}"
+        )
+
+    log(f"\n=== 全部完成 ===")
+    log(f"汇总: {final_path}")
+    incomplete = [s.get("index") for s in all_summaries if not s.get("complete")]
+    if incomplete:
+        log(f"[ERROR] 以下款式仍有未解决采集/写回结果: {incomplete}")
+        return 2
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(main())
+    except KeyboardInterrupt:
+        print("\nInterrupted.", file=sys.stderr)
+        raise SystemExit(130)
diff --git a/src/gyxx_flow/modules/contracts.py b/src/gyxx_flow/modules/contracts.py
new file mode 100644
index 0000000..d696cb8
--- /dev/null
+++ b/src/gyxx_flow/modules/contracts.py
@@ -0,0 +1,20 @@
+"""Stable contract implemented by every business module."""
+
+from __future__ import annotations
+
+from typing import Protocol, runtime_checkable
+
+from gyxx_flow.workflow import WorkflowDefinition
+
+
+@runtime_checkable
+class BusinessModule(Protocol):
+    """A business capability that contributes workflows to the application."""
+
+    module_id: str
+
+    def workflow_definitions(self) -> tuple[WorkflowDefinition, ...]:
+        """Return immutable workflow definitions owned by this module."""
+
+
+__all__ = ["BusinessModule"]
diff --git a/src/gyxx_flow/modules/product_commerce/__init__.py b/src/gyxx_flow/modules/product_commerce/__init__.py
new file mode 100644
index 0000000..d088936
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/__init__.py
@@ -0,0 +1,97 @@
+"""Product, ERP, analysis, main-image, and alert workflows."""
+
+from __future__ import annotations
+
+from collections.abc import Iterable
+from typing import Any
+
+from gyxx_flow.adapters.native import DeferredModuleCommandStep, ModuleCommandFactory
+from gyxx_flow.workflow import StepDefinition, WorkflowDefinition
+
+PRODUCT_SCHEDULED_WORKFLOW_IDS = (
+    "product.persona.daily",
+    "product.daily",
+    "product.alert.daily",
+    "product.import.daily",
+    "product.style_analysis.interval",
+    "product.main_image.jd.weekly",
+    "product.main_image.weekly",
+)
+PRODUCT_MANUAL_WORKFLOW_IDS = (
+    "product.backfill",
+    "product.market_rank",
+    "product.review_collection",
+)
+PRODUCT_WORKFLOW_IDS = (
+    *PRODUCT_SCHEDULED_WORKFLOW_IDS,
+    *PRODUCT_MANUAL_WORKFLOW_IDS,
+)
+PRODUCT_TIMEOUT_SECONDS = 6 * 60 * 60
+PRODUCT_RESOURCE = "module:product_commerce"
+
+
+class ProductCommerceModule:
+    """Product-commerce business module composition boundary."""
+
+    module_id = "product_commerce"
+
+    def __init__(self, definitions: Iterable[WorkflowDefinition] = ()) -> None:
+        self._definitions = tuple(definitions)
+
+    @classmethod
+    def from_catalog(
+        cls,
+        catalog: Any,
+        *,
+        command_factory: ModuleCommandFactory | None = None,
+    ) -> "ProductCommerceModule":
+        """Build all available product workflows, excluding unavailable metadata."""
+
+        entries = {
+            entry.workflow_id: entry
+            for entry in catalog.workflows
+            if entry.module == cls.module_id and entry.trigger in {"scheduled", "manual"}
+        }
+        expected = set(PRODUCT_WORKFLOW_IDS)
+        if set(entries) != expected:
+            raise ValueError(
+                "product catalog workflows do not match the migration contract: "
+                f"expected={sorted(expected)}, actual={sorted(entries)}"
+            )
+        definitions: list[WorkflowDefinition] = []
+        for workflow_id in PRODUCT_WORKFLOW_IDS:
+            entry = entries[workflow_id]
+            action = (
+                DeferredModuleCommandStep(entry)
+                if command_factory is None
+                else DeferredModuleCommandStep(entry, command_factory=command_factory)
+            )
+            definitions.append(
+                WorkflowDefinition(
+                    workflow_id,
+                    (
+                        StepDefinition(
+                            "module_execute",
+                            action,
+                            timeout_seconds=PRODUCT_TIMEOUT_SECONDS,
+                            max_attempts=1,
+                            resources=(PRODUCT_RESOURCE,),
+                            production_sink=True,
+                        ),
+                    ),
+                )
+            )
+        return cls(definitions)
+
+    def workflow_definitions(self) -> tuple[WorkflowDefinition, ...]:
+        return self._definitions
+
+
+__all__ = [
+    "PRODUCT_RESOURCE",
+    "PRODUCT_TIMEOUT_SECONDS",
+    "PRODUCT_MANUAL_WORKFLOW_IDS",
+    "PRODUCT_SCHEDULED_WORKFLOW_IDS",
+    "PRODUCT_WORKFLOW_IDS",
+    "ProductCommerceModule",
+]
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/.env.example b/src/gyxx_flow/modules/product_commerce/runtime/.env.example
new file mode 100644
index 0000000..60003ba
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/.env.example
@@ -0,0 +1,19 @@
+# ── PostgreSQL 连接 (data_hub 库) ───────────────────────────────────
+# 从 .env 文件加载；不要把真实密码提交到 git
+PG_HOST=8.148.185.119
+PG_PORT=5432
+PG_DB=data_hub
+PG_USER=data_hub
+PG_PASSWORD=
+
+# ── 京东采集账号 (jd_self_inventory_sales_collector 等) ─────────────────
+JD_SELF_SHOP=光影自营研发
+JD_SELF_PASSWORD=
+
+# ── 淘宝光合平台视频上传账号 ─────────────────────────────────────────
+GUANGHE_USERNAME=
+GUANGHE_PASSWORD=
+GUANGHE_USER_DATA_DIR=${GYXX_DATA_ROOT}\profiles\product_commerce\guanghe
+GUANGHE_LUGGAGE_USERNAME=
+GUANGHE_LUGGAGE_PASSWORD=
+GUANGHE_LUGGAGE_USER_DATA_DIR=${GYXX_DATA_ROOT}\profiles\product_commerce\guanghe-luggage
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/MIGRATION_STATUS.md b/src/gyxx_flow/modules/product_commerce/runtime/MIGRATION_STATUS.md
new file mode 100644
index 0000000..4e502e0
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/MIGRATION_STATUS.md
@@ -0,0 +1,20 @@
+# Product-commerce source snapshot
+
+This directory contains the reviewed working-tree snapshot of the legacy
+product-commerce runtime. It is now physically owned by `gyxx-flow`; no file in
+this directory needs the legacy checkout to exist.
+
+The snapshot is not yet the public workflow entry point. Native packaging must
+first replace root-relative imports, subprocess script names, data-root
+assumptions, browser-profile defaults, and dynamic vendor-root injection.
+
+`launchers_reference/` is provenance only. Those files document the old task
+arguments and must never be registered as final scheduled-task actions.
+
+`runtime_tests/` preserves the 21 legacy regression tests without adding them to
+the current repository's automatic pytest collection. They will be moved and
+adapted as each native runtime slice becomes importable.
+
+The authoritative source/target hashes, transformations, and six intentionally
+excluded diagnostic entries are recorded in
+`config/source-manifests/product_commerce.json`.
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/__init__.py b/src/gyxx_flow/modules/product_commerce/runtime/__init__.py
new file mode 100644
index 0000000..39d3bc0
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/__init__.py
@@ -0,0 +1,6 @@
+"""Vendored product-commerce source snapshot awaiting native package refactoring.
+
+Files in this package are copied into ``gyxx-flow`` so migration work no longer
+depends on the legacy checkout. They are provenance-preserving runtime sources;
+the public workflow registry is switched to native entries in a separate phase.
+"""
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/aggregate_daily_final.py b/src/gyxx_flow/modules/product_commerce/runtime/aggregate_daily_final.py
new file mode 100644
index 0000000..18159a8
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/aggregate_daily_final.py
@@ -0,0 +1,280 @@
+"""按日期优先汇总三平台 + ERP 的每日款式数据。
+
+输出:
+    data/<日期>/<款>/final_summary.json
+    data/<日期>/<款>/final_summary.md
+
+不计算转化率/加购率/退款率（这三个比率只在周汇总里算）。
+
+每平台字段: sales / refund_orders / visitors / cart_users（来自平台层原始值）
+ERP 字段:   style_codes / sales / refund_orders / source
+"""
+
+import argparse
+import json
+from datetime import datetime
+from pathlib import Path
+
+from erp_metric_overrides import DATA_ROOT, PLATFORM_DIRS, apply_erp_override, parse_number
+from runtime_paths import CURATED_DATA_ROOT, managed_data_path
+
+
+PLATFORMS = list(PLATFORM_DIRS)
+
+PLATFORM_STYLE_FILENAME = {
+    "jd": lambda style, date: f"{safe_name(style)}_{date}.json",
+    "dy": lambda style, date: f"{safe_name(style)}_{date}.json",
+    "tm": lambda style, date: f"业务款式_{date}.json",
+}
+
+
+def safe_name(value: str) -> str:
+    return "".join("_" if ch in '\\/:*?"<>|' else ch for ch in str(value)).strip(" ._") or "未命名"
+
+
+def read_json(path: Path) -> dict:
+    return json.loads(path.read_text(encoding="utf-8-sig"))
+
+
+def _extract_platform_metrics(platform: str, payload: dict) -> dict:
+    """从平台采集器产出的 <款>_<日期>.json 提取 4 个原始值。"""
+    if platform == "jd":
+        metrics = payload.get("指标汇总") or {}
+        return {
+            "visitors": parse_number(metrics.get("访客数")),
+            "cart_users": parse_number(metrics.get("加购客户数") or metrics.get("加购人数")),
+            "sales": parse_number(metrics.get("销量")),
+            "refund_orders": parse_number(metrics.get("退款订单数")),
+        }
+    if platform == "dy":
+        raw = payload.get("raw_totals") or {}
+        platform_raw = payload.get("platform_raw_totals") or raw
+        return {
+            "visitors": parse_number(platform_raw.get("访客数") or payload.get("访客数")),
+            "cart_users": parse_number(platform_raw.get("加购人数") or payload.get("加购人数")),
+            "sales": parse_number(platform_raw.get("销量") or payload.get("销量")),
+            "refund_orders": parse_number(platform_raw.get("退款订单数")),
+        }
+    if platform == "tm":
+        style = payload.get("style") or payload
+        return {
+            "visitors": parse_number(style.get("访客量") or style.get("访客数")),
+            "cart_users": parse_number(style.get("加购人数")),
+            "sales": parse_number(style.get("销量")),
+            "refund_orders": parse_number(style.get("退款订单数")),
+        }
+    raise ValueError(f"unknown platform: {platform}")
+
+
+def _extract_erp_metadata(payload: dict) -> dict:
+    """ERP 层只携带元信息（编码 + 源），销量/退货量已在各平台 sales/refund_orders 字段中（ERP 覆盖后）。"""
+    return {
+        "style_codes": payload.get("erp_style_codes") or [],
+    }
+
+
+def collect_style_names(target_date: str, data_root: Path) -> list[str]:
+    """扫描三平台数据目录，得到 (日期, 款) 全部去重款式名。"""
+    names: set[str] = set()
+    for platform in PLATFORMS:
+        base = data_root / PLATFORM_DIRS[platform]
+        if not base.exists():
+            continue
+        for style_dir in base.glob(f"*_{target_date}"):
+            if not style_dir.is_dir():
+                continue
+            for json_path in style_dir.glob("*.json"):
+                if json_path.name in {"data.json"}:
+                    continue
+                if json_path.name == "final_summary.json":
+                    continue
+                if not (json_path.name.endswith(f"_{target_date}.json")
+                        or json_path.name.startswith("业务款式_")):
+                    continue
+                try:
+                    payload = read_json(json_path)
+                except Exception:
+                    continue
+                if platform == "jd":
+                    name = payload.get("款式")
+                elif platform == "dy":
+                    name = payload.get("style_name")
+                else:
+                    style = payload.get("style") or {}
+                    name = style.get("款式") or payload.get("款式")
+                if name:
+                    names.add(name)
+    return sorted(names)
+
+
+def load_platform_block(platform: str, style_name: str, target_date: str, data_root: Path) -> dict | None:
+    filename = PLATFORM_STYLE_FILENAME[platform](style_name, target_date)
+    style_dir = data_root / PLATFORM_DIRS[platform] / f"{safe_name(style_name)}_{target_date}"
+    json_path = style_dir / filename
+    if not json_path.exists():
+        return None
+    try:
+        payload = read_json(json_path)
+    except Exception:
+        return None
+    metrics = _extract_platform_metrics(platform, payload)
+    # 不要相信平台文件里"已应用 ERP 覆盖"的旧值 —— 平台采集在 ERP 写 data.json 之前
+    # 跑的话,summarize_styles_from_excel 拿到 all_failed 标记, 销量/退货量回退到平台原值
+    # (TM 0/99 / DY 138/833) 并被冻进平台 JSON。这里用最新的 data.json 再覆盖一次。
+    try:
+        override = apply_erp_override(
+            platform, style_name,
+            metrics["visitors"], metrics["cart_users"],
+            target_date=target_date,
+            platform_sales=metrics.get("sales"),
+            platform_refunds=metrics.get("refund_orders"),
+        )
+        metrics["sales"] = override["销量"]
+        metrics["refund_orders"] = override["退款订单数"]
+        metrics["erp_unavailable"] = override.get("erp_unavailable", False)
+    except Exception as exc:
+        metrics["erp_override_error"] = f"{type(exc).__name__}: {exc}"
+    metrics["source"] = str(json_path)
+    return metrics
+
+
+def load_erp_block(platform: str, style_name: str, target_date: str, data_root: Path) -> dict | None:
+    style_dir = data_root / PLATFORM_DIRS[platform] / f"{safe_name(style_name)}_{target_date}"
+    json_path = style_dir / "data.json"
+    if not json_path.exists():
+        return None
+    try:
+        payload = read_json(json_path)
+    except Exception:
+        return None
+    block = _extract_erp_metadata(payload)
+    block["source"] = str(json_path)
+    return block
+
+
+def build_payload(date: str, style_name: str, data_root: Path) -> dict:
+    payload = {
+        "date": date,
+        "style_name": style_name,
+        "generated_at": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+        "metrics_note": {
+            "platform.sales": "汇总时实时应用 ERP 覆盖后的销量（不要看平台 JSON 里的旧值）",
+            "platform.refund_orders": "汇总时实时应用 ERP 覆盖后的退货量（不要看平台 JSON 里的旧值）",
+            "platform.visitors": "平台商品访客数之和",
+            "platform.cart_users": "平台商品加购人数之和",
+            "erp.style_codes": "ERP 款式编码",
+            "erp_unavailable": "True 表示本平台 ERP data.json 不存在或全部查询失败, 此时销量/退货量用平台原值",
+        },
+        "erp": {},
+        "jd": {},
+        "tm": {},
+        "dy": {},
+    }
+    erp_seen: dict | None = None
+    for platform in PLATFORMS:
+        platform_block = load_platform_block(platform, style_name, date, data_root)
+        if platform_block:
+            payload[platform] = platform_block
+        erp_block = load_erp_block(platform, style_name, date, data_root)
+        if erp_block and erp_seen is None:
+            erp_seen = erp_block
+    if erp_seen is not None:
+        payload["erp"] = erp_seen
+    return payload
+
+
+def render_markdown(payload: dict) -> str:
+    style = payload["style_name"]
+    date = payload["date"]
+    erp = payload.get("erp") or {}
+
+    lines = [
+        f"# {style} - 每日最终汇总 - {date}",
+        "",
+        f"- 日期: {date}",
+        f"- 款式: {style}",
+        f"- ERP款式编码: {', '.join(erp.get('style_codes') or []) or '-'}",
+        "",
+        "| 平台 | 销量 | 退货量 | 访客数 | 加购人数 |",
+        "|---|---:|---:|---:|---:|",
+    ]
+    for platform in PLATFORMS:
+        block = payload.get(platform) or {}
+        if not block:
+            lines.append(f"| {platform} | - | - | - | - |")
+            continue
+        lines.append(
+            f"| {platform} | {int(parse_number(block.get('sales')))} | "
+            f"{int(parse_number(block.get('refund_orders')))} | "
+            f"{int(parse_number(block.get('visitors')))} | "
+            f"{int(parse_number(block.get('cart_users')))} |"
+        )
+    lines.append("")
+    lines.append(f"_生成时间: {payload['generated_at']}_")
+    return "\n".join(lines)
+
+
+def write_outputs(payload: dict, data_root: Path) -> tuple[Path, Path]:
+    style_dir = data_root / payload["date"] / safe_name(payload["style_name"])
+    style_dir.mkdir(parents=True, exist_ok=True)
+    json_path = style_dir / "final_summary.json"
+    md_path = style_dir / "final_summary.md"
+    json_path.write_text(
+        json.dumps(payload, ensure_ascii=False, indent=2),
+        encoding="utf-8-sig",
+    )
+    md_path.write_text(render_markdown(payload), encoding="utf-8-sig")
+    return json_path, md_path
+
+
+def main():
+    parser = argparse.ArgumentParser(description="按日期+款式汇总三平台 + ERP 每日数据（无比率字段）")
+    parser.add_argument("--date", default=datetime.now().strftime("%Y-%m-%d"))
+    parser.add_argument("--data-root", default=str(DATA_ROOT))
+    args = parser.parse_args()
+
+    data_root = managed_data_path(args.data_root)
+    style_names = collect_style_names(args.date, data_root)
+    print(f"[INFO] {args.date} 共发现 {len(style_names)} 个款")
+
+    written = 0
+    payloads = []
+    for style_name in style_names:
+        payload = build_payload(args.date, style_name, data_root)
+        json_path, md_path = write_outputs(payload, CURATED_DATA_ROOT)
+        written += 1
+        payloads.append(payload)
+
+    print(f"[OK] {args.date} 写入 {written} 个款 -> {data_root / args.date}")
+
+    # 写库（库不可用时 WARN，不阻塞主流程）
+    try:
+        from db import get_conn, upsert_daily_metrics
+        rows = []
+        for p in payloads:
+            erp_codes = (p.get("erp") or {}).get("style_codes") or []
+            for platform in ("jd", "dy", "tm"):
+                block = p.get(platform) or {}
+                if not block:
+                    continue
+                rows.append({
+                    "metric_date": args.date,
+                    "style_name": p.get("style_name"),
+                    "platform": platform,
+                    "sales": int(parse_number(block.get("sales"))),
+                    "refund_orders": int(parse_number(block.get("refund_orders"))),
+                    "visitors": int(parse_number(block.get("visitors"))),
+                    "cart_users": int(parse_number(block.get("cart_users"))),
+                    "erp_style_codes": erp_codes,
+                    "source": block.get("source"),
+                })
+        if rows:
+            with get_conn() as conn:
+                n = upsert_daily_metrics(conn, rows)
+            print(f"[PG] daily_style_metrics upsert: {n} 行")
+    except Exception as exc:
+        print(f"[WARN] PG upsert 失败 (库未配/不可达)，主流程不受影响: {exc}")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/analyze_style_with_hermes.py b/src/gyxx_flow/modules/product_commerce/runtime/analyze_style_with_hermes.py
new file mode 100644
index 0000000..82bf840
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/analyze_style_with_hermes.py
@@ -0,0 +1,1097 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""Aggregate one style's daily DB metrics for three days and analyze them.
+
+The generated Markdown is written to a Feishu document and its link is
+idempotently upserted into that style's configured analysis Base.
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+from gyxx_flow.adapters import RuntimeServicePolicy
+import re
+import sys
+import urllib.error
+import urllib.request
+from collections import defaultdict
+from datetime import date, timedelta
+from pathlib import Path
+
+from runtime_paths import CURATED_DATA_ROOT
+from typing import Any
+
+from config.style_config_loader import get_style_analysis_bitable_map
+from db import get_conn, upsert_style_analysis_report
+from lark_cli_runtime import lark_cli_env, run_lark_cli
+
+# Hermes (与 check_nine_day_decline.py / vendors/jd-data-flow/config.py 对齐)
+HERMES_API_KEY = os.getenv("HERMES_API_KEY")
+if not HERMES_API_KEY:
+    raise RuntimeError("HERMES_API_KEY is required")
+os.environ.update(RuntimeServicePolicy().apply(os.environ))
+HERMES_ANALYZER_URL = os.getenv(
+    "ANALYZER_API_SERVER_URL",
+    "http://localhost:8642/v1/chat/completions",
+)
+HERMES_MODEL = "mimo-v2.5-pro"
+
+
+# ── data extraction ────────────────────────────────────────────────────────
+
+def pull_daily(style: str, start: date, end: date) -> list[dict]:
+    with get_conn() as c, c.cursor() as cur:
+        cur.execute(
+            """SELECT metric_date, platform, sales, refund_orders, visitors, cart_users
+               FROM daily_style_metrics
+               WHERE style_name=%s AND metric_date BETWEEN %s AND %s
+               ORDER BY metric_date, platform""",
+            (style, start, end),
+        )
+        return [
+            {"date": r[0].isoformat(), "plat": r[1], "sales": r[2],
+             "refund": r[3], "visitors": r[4], "cart": r[5]}
+            for r in cur.fetchall()
+        ]
+
+
+def missing_window_dates(rows: list[dict], start: date, end: date) -> list[date]:
+    """Return calendar dates absent from a style's requested DB window."""
+    present = {date.fromisoformat(str(row["date"])) for row in rows if row.get("date")}
+    expected = {start + timedelta(days=offset) for offset in range((end - start).days + 1)}
+    return sorted(expected - present)
+
+
+def pull_reviews(style: str, start: date, end: date) -> dict:
+    """仅返回当前三日窗口内的评论分布与样本。"""
+    with get_conn() as c, c.cursor() as cur:
+        cur.execute(
+            """SELECT platform, rating, COUNT(*) n, array_agg(content ORDER BY comment_date DESC) samples
+               FROM product_reviews
+               WHERE style_name=%s AND comment_date BETWEEN %s AND %s
+               GROUP BY platform, rating""",
+            (style, start, end),
+        )
+        return {
+            f"{r[0]}|{r[1]}": {"n": r[2], "samples": (r[3] or [])[:5]}
+            for r in cur.fetchall()
+        }
+
+
+def pull_persona(style: str, start: date, end: date) -> list[dict]:
+    with get_conn() as c, c.cursor() as cur:
+        cur.execute(
+            """SELECT collect_date, platform, payload
+               FROM daily_persona_metrics
+               WHERE style_name=%s AND collect_date BETWEEN %s AND %s
+               ORDER BY collect_date DESC, platform""",
+            (style, start, end),
+        )
+        out = []
+        for d, p, pl in cur.fetchall():
+            gender = pl.get("gender", []) if pl else []
+            age = pl.get("age", []) if pl else []
+            city = pl.get("city_tier", []) if pl else []
+            buying = pl.get("buying_power", []) if pl else []
+            out.append({
+                "date": d.isoformat(), "plat": p,
+                "gender": gender[:2],
+                "age_top": sorted(age, key=lambda x: -float(x.get("value", 0)))[:3],
+                "city_tier": city[:3],
+                "buying_power": buying[:3],
+            })
+        return out
+
+
+def pull_main_image(style: str, start: date, end: date) -> list[dict]:
+    """主图创意指标 (TM). 取窗口内最近一次采集的 top 10 图."""
+    with get_conn() as c, c.cursor() as cur:
+        cur.execute(
+            """SELECT collect_date, image_key, impressions, clicks, click_rate
+               FROM main_image_creatives
+               WHERE style_name=%s AND collect_date BETWEEN %s AND %s
+               ORDER BY collect_date DESC, clicks DESC NULLS LAST LIMIT 6""",
+            (style, start, end),
+        )
+        return [
+            {"date": r[0].isoformat(), "key": r[1], "imp": r[2],
+             "clicks": r[3], "ctr": str(r[4]) if r[4] is not None else None}
+            for r in cur.fetchall()
+        ]
+
+
+# ── seeding (站外种草: cmt_* 表族) ──────────────────────────────────────────
+
+def _resolve_cmt_style_id(cur, style: str) -> int | None:
+    """cmt_styles.name -> style_id. 先精确匹配, 再 LIKE 兜底 (如'拾影斜挎相机包')."""
+    cur.execute("SELECT id FROM cmt_styles WHERE name=%s LIMIT 1", (style,))
+    row = cur.fetchone()
+    if row:
+        return row[0]
+    cur.execute("SELECT id FROM cmt_styles WHERE name LIKE %s OR %s LIKE '%%' || name || '%%' LIMIT 1",
+                (f"%{style}%", style))
+    row = cur.fetchone()
+    return row[0] if row else None
+
+
+def pull_seeding(style: str, start: date, end: date) -> dict:
+    """仅拉取当前三日窗口内的站外种草数据和关联创作者静态属性。
+
+    维度 (对齐用户需求):
+      - 当前三日新发布笔记数量与互动, by 种草平台
+      - 笔记内容数据趋势: view/like/comment/collect/share 聚合
+      - 种草平台分布: bilibili/douyin/xiaohongshu
+      - 内容方式: cmt_notes.title + cmt_cooperations.content_direction
+      - 种草人群: cmt_comments.ip_location 地域分布 top (库无性别字段, 用地域兜底)
+      - 评论区咨询: cmt_comments.content 样本
+      - 合作投入: cmt_cooperations.cost/exposure/roi
+    """
+    end_exclusive = end + timedelta(days=1)
+    with get_conn() as c, c.cursor() as cur:
+        sid = _resolve_cmt_style_id(cur, style)
+        if sid is None:
+            # cmt_styles 里没该款式, 返回完整空结构避免下游 KeyError
+            return {
+                "matched": False, "cmt_style_id": None,
+                "notes_by_platform_all": [], "notes_this_window": [],
+                "notes_recent_top8": [], "note_comments_sample": [],
+                "comment_ip_top10": [],
+                "cooperations_window": [], "creators_top10": [],
+            }
+
+        out: dict[str, Any] = {"matched": True, "cmt_style_id": sid}
+
+        # 1. 当前三日笔记汇总 by 平台
+        cur.execute(
+            """SELECT platform,
+                      COUNT(*) n,
+                      SUM(COALESCE(view_count, 0)) views,
+                      SUM(COALESCE(like_count, 0)) likes,
+                      SUM(COALESCE(comment_count, 0)) comments,
+                      SUM(COALESCE(collect_count, 0)) favorites,
+                      SUM(COALESCE(share_count, 0)) shares,
+                      SUM(CASE WHEN self_operated THEN 1 ELSE 0 END) self_n,
+                      MIN(publish_time) first_pub, MAX(publish_time) last_pub
+               FROM cmt_notes
+               WHERE style_id=%s AND publish_time >= %s AND publish_time < %s
+               GROUP BY platform ORDER BY views DESC NULLS LAST""",
+            (sid, start, end_exclusive),
+        )
+        out["notes_by_platform_all"] = [
+            {"platform": r[0], "n": r[1], "views": int(r[2] or 0),
+             "likes": int(r[3] or 0), "comments": int(r[4] or 0),
+             "favorites": int(r[5] or 0), "shares": int(r[6] or 0),
+             "self_operated": r[7],
+             "first_pub": r[8].isoformat() if r[8] else None,
+             "last_pub": r[9].isoformat() if r[9] else None}
+            for r in cur.fetchall()
+        ]
+
+        # 2. 当前三日笔记数 + 互动 (按平台)
+        def _period_notes(s, e):
+            cur.execute(
+                """SELECT platform, COUNT(*) n,
+                          SUM(COALESCE(view_count, 0)) views,
+                          SUM(COALESCE(like_count, 0)) likes,
+                          SUM(COALESCE(comment_count, 0)) comments
+                   FROM cmt_notes
+                   WHERE style_id=%s AND publish_time >= %s AND publish_time < %s
+                   GROUP BY platform""",
+                (sid, s, e + timedelta(days=1)),
+            )
+            return [{"platform": r[0], "n": r[1], "views": int(r[2] or 0),
+                     "likes": int(r[3] or 0), "comments": int(r[4] or 0)}
+                    for r in cur.fetchall()]
+        out["notes_this_window"] = _period_notes(start, end)
+
+        # 3. 当前三日笔记个体 top 15 (含 title 看内容方式)
+        cur.execute(
+            """SELECT platform, title, publish_time,
+                      COALESCE(view_count, 0) views,
+                      COALESCE(like_count, 0) likes,
+                      COALESCE(comment_count, 0) comments,
+                      COALESCE(collect_count, 0) favorites,
+                      self_operated
+               FROM cmt_notes
+               WHERE style_id=%s AND publish_time >= %s AND publish_time < %s
+               ORDER BY publish_time DESC LIMIT 8""",
+            (sid, start, end_exclusive),
+        )
+        out["notes_recent_top8"] = [
+            {"platform": r[0], "title": (r[1] or "")[:80],
+             "publish_time": r[2].isoformat() if r[2] else None,
+             "views": int(r[3] or 0), "likes": int(r[4] or 0),
+             "comments": int(r[5] or 0), "favorites": int(r[6] or 0),
+             "self_operated": r[7]}
+            for r in cur.fetchall()
+        ]
+
+        # 4. 当前三日笔记评论样本 + IP 地域分布 top 10
+        cur.execute(
+            """SELECT cm.content, cm.nickname, cm.ip_location, cm.comment_created_at,
+                      n.title, n.platform
+               FROM cmt_comments cm
+               JOIN cmt_notes n ON cm.note_id = n.id
+               WHERE n.style_id=%s
+                 AND cm.comment_created_at >= %s AND cm.comment_created_at < %s
+               ORDER BY cm.comment_created_at DESC LIMIT 8""",
+            (sid, start, end_exclusive),
+        )
+        out["note_comments_sample"] = [
+            {"content": (r[0] or "")[:100], "ip": r[2],
+             "at": r[3].isoformat() if r[3] else None,
+             "note_title": (r[4] or "")[:50], "note_platform": r[5]}
+            for r in cur.fetchall()
+        ]
+        cur.execute(
+            """SELECT COALESCE(cm.ip_location, '未知') loc, COUNT(*) n
+               FROM cmt_comments cm
+               JOIN cmt_notes n ON cm.note_id = n.id
+               WHERE n.style_id=%s
+                 AND cm.comment_created_at >= %s AND cm.comment_created_at < %s
+               GROUP BY loc ORDER BY n DESC LIMIT 10""",
+            (sid, start, end_exclusive),
+        )
+        out["comment_ip_top10"] = [{"region": r[0], "n": r[1]} for r in cur.fetchall()]
+
+        # 5. 当前三日合作记录 (含 cost/exposure/content_direction)
+        cur.execute(
+            """SELECT cooperation_date, cooperation_cost, content_direction,
+                      exposure_count, engagement_count, roi, cpe,
+                      shop_uv, transaction_amount, is_paid
+               FROM cmt_cooperations WHERE style_id=%s
+                 AND cooperation_date BETWEEN %s AND %s
+               ORDER BY cooperation_date DESC LIMIT 8""",
+            (sid, start, end),
+        )
+        out["cooperations_window"] = [
+            {"date": r[0].isoformat() if r[0] else None,
+             "cost": float(r[1]) if r[1] is not None else None,
+             "direction": r[2], "exposure": int(r[3]) if r[3] else None,
+             "engagement": r[4], "roi": float(r[5]) if r[5] is not None else None,
+             "cpe": float(r[6]) if r[6] is not None else None,
+             "shop_uv": int(r[7]) if r[7] else None,
+             "transaction": float(r[8]) if r[8] is not None else None,
+             "is_paid": r[9]}
+            for r in cur.fetchall()
+        ]
+
+        # 6. 创作者静态属性 (该款关联的 top, 按粉丝数)
+        cur.execute(
+            """SELECT cr.name, cr.platform, cr.follower_count_num,
+                      cr.account_type, cr.creator_level
+               FROM cmt_creators cr
+               JOIN cmt_style_creators sc ON sc.creator_id = cr.id
+               WHERE sc.style_id=%s
+               ORDER BY cr.follower_count_num DESC NULLS LAST LIMIT 5""",
+            (sid,),
+        )
+        out["creators_top10"] = [
+            {"name": r[0], "platform": r[1],
+             "followers": int(r[2]) if r[2] is not None else None,
+             "account_type": r[3], "level": r[4]}
+            for r in cur.fetchall()
+        ]
+        return out
+
+
+# ── aggregation helpers ────────────────────────────────────────────────────
+
+def aggregate_period(daily: list[dict]) -> dict:
+    """当前分析周期的 daily 数据按平台聚合。"""
+    g: dict[str, dict] = defaultdict(lambda: {"sales": 0, "refund": 0, "visitors": 0, "cart": 0})
+    for row in daily:
+        p = row["plat"]
+        g[p]["sales"] += row["sales"]
+        g[p]["refund"] += row["refund"]
+        g[p]["visitors"] += row["visitors"]
+        g[p]["cart"] += row["cart"]
+    for p, v in g.items():
+        v["cvr_pct"] = round(v["sales"] / v["visitors"] * 100, 2) if v["visitors"] else 0
+        v["cart_rate_pct"] = round(v["cart"] / v["visitors"] * 100, 2) if v["visitors"] else 0
+        v["refund_rate_pct"] = round(v["refund"] / v["sales"] * 100, 2) if v["sales"] else 0
+    return dict(g)
+
+
+def aggregate_style_total(platforms: dict[str, dict]) -> dict:
+    """Combine platform aggregates into one three-day style summary."""
+    total = {"sales": 0, "refund": 0, "visitors": 0, "cart": 0}
+    for metrics in platforms.values():
+        for key in total:
+            total[key] += int(metrics.get(key) or 0)
+    total["cvr_pct"] = round(total["sales"] / total["visitors"] * 100, 2) if total["visitors"] else 0
+    total["cart_rate_pct"] = round(total["cart"] / total["visitors"] * 100, 2) if total["visitors"] else 0
+    total["refund_rate_pct"] = round(total["refund"] / total["sales"] * 100, 2) if total["sales"] else 0
+    return total
+
+
+# ── prompt builder ────────────────────────────────────────────────────────
+
+LEGACY_SYSTEM = """你是光影行星 (GYXX) 的资深电商数据分析师, 擅长把数据现象挖到根因并给出可执行方案.
+用户会给定一个款式近 3 天的真实业务数据:
+- 站内 (店铺自身): 天猫/京东/抖音三平台当前 3 天的 daily_style_metrics、
+  product_reviews、daily_persona_metrics、main_image_creatives
+- 站外 (种草营销): cmt_notes (笔记), cmt_comments (笔记评论), cmt_cooperations (合作投入),
+  cmt_creators (创作者静态属性)
+
+只允许根据给定的这 3 天数据分析，不与前 3 天、周数据或其他历史窗口做对比。
+
+# 输出要求 (Markdown, 中文, 总长 8000-12000 字)
+
+## 加粗规则 (严格遵守)
+
+**必须加粗** 的内容:
+1. 所有关键数字 (销量/访客/CVR/退款率/曝光/ROI 等具体数值, 例: **-17.9%** / **36.36%** / **16 万曝光**)
+2. 所有"异常矛盾点" (站内站外方向相反、或同向但幅度不对等, 例: **访客 +4.3% 但销量 -19.5%**)
+3. 所有"根因结论句" (挖到的最深层原因, 例: **核心堵点在加购->付款最后一公里, 非流量问题**)
+4. 所有"关键信号" (爆款笔记/高赞负面评论/退款激增日, 例: **B 站 5-26 盖亚实测笔记 801 赞负面评论**)
+5. 所有优先级标签 (例: **P0** / **P1** / **P2**) 和时间节点 (例: **3 天内** / **本周内**)
+6. 所有"数据缺失"标记 (例: **数据缺失, 建议人工核查**)
+
+**不要加粗** 的内容: 普通描述句、过渡句、列表项符号.
+
+## 6 块分析框架 (每块都必须有 5 段结构)
+
+每块输出 5 段, 严格按以下顺序:
+
+### 1. 数据快照 (≤ 6 行, 站内站外分块列)
+- 站内: 当前 3 天的销量、访客、加购、转化、退款，按平台及逐日观察
+- 站外: 当前 3 天的笔记数、曝光、互动与合作投入金额
+- 不要罗列所有数字, 只挑能反映当前单品状态的关键值
+
+### 2. 异常识别 (找出 2-3 个"反常点", 每个加粗标出)
+- 当前 3 天内哪些日期或平台出现访客、加购、成交不匹配?
+- 哪个平台漏斗转化明显弱于其他平台?
+- 当前 3 天站外曝光或互动是否在同一窗口内传导到站内?
+
+### 3. 根因深挖 (3 层 why, 每层加粗结论句)
+- 第 1 层 (现象): 表面看到什么? 例: TM CVR 从 0.98% 跌到 0.75%
+- 第 2 层 (直接原因): 哪个环节断裂? 例: 加购->付款转化从 X% 跌到 Y%
+- 第 3 层 (根本原因): 为什么断裂? 必须结合站内+站外证据交叉验证.
+  例: 详情页"多兜"宣传与 B 站实测"兜设计无用"冲突, 导致加购用户决策犹豫
+- 没数据的维度, 不要编, 直接写 **数据缺失, 建议人工核查** + 列出要查什么
+
+### 4. 站内外因果链 (画清转化路径, 标出断点)
+- 把站外种草 -> 站内访客 -> 加购 -> 成交 -> 退货的链路画出来
+- 每个环节标注转化率/绝对数, 用 **[断点]** 标出断裂处
+- 例: 抖音三日曝光 16 万 -> 天猫三日访客 500 -> 加购 80 -> 成交 12 **[断点: 加购->成交]** -> 退款 4
+- 站外爆款如果没传导到站内, 也要标 **[断点: 站外->站内未传导]**
+
+### 5. 优化建议 (3 条, 必须可执行, 含文案/阈值/时间表)
+每条格式:
+- **[P0/P1/P2] [站内/站外/联动] [动作名]**: 具体做什么 + 文案/阈值/数值 + 时间节点 + 预期提升
+- 例: **[P0] [站内] TM 详情页"多兜"卖点重写**: 把"多功能多兜"改为"前兜装电池×2 / 侧兜装水壶 / 主仓 14 寸笔记本+1 机 2 镜", 配实拍拆解图 3 张, **3 天内**上线, 预期加购->成交转化从 15.4% 提到 22%+
+
+## 6 块主题 (每块都要站内+站外交叉看)
+
+1. **访客分析** - 站内: 流量入口波动/男女/年龄/城市层级; 站外: 笔记发布量趋势/曝光/平台分布/创作者地域 vs 站内访客城市画像吻合度. 诊断"种草是否带来增量访客".
+2. **加购率分析** - 站内: 加购率波动/加购->下单转化; 站外: 笔记评论区咨询 ("包链接"/"尺寸"/"容量") 反映详情页说服缺口. 诊断"加购后为什么不下单".
+3. **点击率分析** - 站内: 主图 CTR top/曝光分布; 站外: 高互动笔记封面/标题风格. 诊断"主图 vs 笔记封面视觉语言是否脱节", 给主图优化方案 (背景色/构图/文案/字体).
+4. **转化率分析** - 站内: CVR 波动/男女/年龄/关键词来源; 站外: 合作投入 (cost/exposure/shop_uv/transaction/roi) 是否带转化. 诊断"站外爆款为什么没转化成店内销量".
+5. **退货率分析** - 站内: 退款率趋势/评价文本归因; 站外: 笔记评论负面吐槽. 诊断"实物与宣传不符的具体点", 给详情页修复方案.
+6. **销量分析** - 站内: 当前三日异常平台/流量/CTR/CVR/退款; 站外: 当前三日种草内容与销量是否同频. 诊断当前销量状态的核心原因, 给站内站外协同方案.
+
+## 总评 (最后一段)
+
+- **一句话结论**: 用 1 句话概括核心矛盾 (含关键数字, 加粗)
+- **未来 3 天的 3 个最高优先级动作**: 按 P0/P1/P2 排, 每条含 [站内/站外/联动] 标签 + 动作 + 时间节点 + 预期效果
+
+## 红线 (不要踩)
+
+- 不要编造数据. 没数据的维度写 **数据缺失, 建议人工核查** + 列出要查什么.
+- **严禁引用、推算或比较当前 3 天窗口之外的任何业务数据。**
+- 种草人群"男女比例"库里无性别字段, 用 cmt_comments.ip_location 地域分布近似替代, 并注明是地域近似.
+- 不要写"建议优化详情页"这种空话, 必须给具体文案/阈值/时间表.
+- 不要罗列所有数字, 只挑能反映"波动"和"异常"的关键值.
+- 竞品参考要给具体数字对标 (如"国家地理同类款 TM 月销 800+, 退款率 12%"), 不要给"行业 8-15%"这种过宽范围.
+- **严禁用 【】 方括号包裹数字或标记, 必须用 markdown `** **` 加粗语法**.
+- **严禁用 ═══ 或 ─── 等装饰性分隔线, 章节分隔只用 `---`**.
+- **章节标题必须用 `## 一、访客分析` 格式, 不要用 `【一、访客分析】`**.
+- **数字加粗必须写成 `**-17.9%**`, 不能写成 `【-17.9%】` 或 `[-17.9%]`**.
+- **平台名称必须用全称, 严禁简写**: "天猫" (不用 TM/Tm/tm), "京东" (不用 JD/Jd/jd), "抖音" (不用 DY/Dy/dy).
+  但输入数据 JSON 里的 platform 字段值是 jd/dy/tm 小写英文, 你在报告里引用时要转成中文全称.
+"""
+
+
+SYSTEM = """你是光影行星的资深电商单品分析师。
+输入是某款式当前连续 3 天的数据库数据，包含逐日经营、平台汇总、人物画像、商品评价、
+主图表现、营销笔记曝光互动、笔记评论、合作效果和关联创作者背景。
+
+只分析输入中的三日动态数据，不引用或推算窗口外数据，不做前 3 天、周度或历史趋势对比，
+不调用工具，不补充外部信息，不编造因果。创作者属性属于静态背景，只能用于解释内容覆盖。
+
+输出中文 Markdown，控制在 3000-5000 字。一级标题后必须严格按以下七个二级标题及顺序输出，
+不得改名、合并、删除或增加其他二级标题：
+
+## 一、访客分析
+结合逐日及平台访客、人物画像、营销笔记曝光互动和创作者覆盖，分析三日流量结构与站外传导。
+
+## 二、加购率分析
+结合平台加购人数/加购率、画像、笔记评论咨询点和商品评价，诊断从访客到加购的说服力。
+
+## 三、点击率分析
+结合主图曝光、点击、点击率以及高互动营销笔记的标题内容，分析素材吸引力；没有封面数据就不得编造。
+
+## 四、转化率分析
+结合访客、加购、销量、转化率、营销合作的店铺 UV/成交额/ROI、评论和评价，诊断成交断点。
+
+## 五、退货率分析
+结合平台退款订单/退货率、商品评价和笔记评论中的负面反馈，分析宣传与实际体验是否存在落差。
+
+## 六、销量分析
+综合逐日及平台销量、访客、加购、点击、转化、退款和三日营销表现，说明销量状态与主要驱动/阻碍。
+
+## 七、总评
+给出一句话核心结论，画出“营销曝光→访客→加购→成交→退款”链路，并给出 3-5 条带
+**P0/P1/P2**、负责人方向、完成时限和衡量指标的行动。
+
+每个章节都要包含数据快照、问题判断和可执行建议。某个证据维度无数据也必须保留对应业务章节并明确说明。
+
+前六章在二级标题下固定使用 `### 数据快照`、`### 核心判断`、`### 优化建议`；
+总评固定使用 `### 核心结论`、`### 转化链路`、`### 优先级行动`。
+
+重点内容必须使用 Markdown `**加粗**`，包括：三日关键总量和比率、平台异常对比、核心判断句、
+**P0/P1/P2**、完成时限和目标指标。不要整段全部加粗，普通解释保持正常字重。
+
+关键数字、异常点和结论使用 Markdown 加粗。tm/jd/dy 在正文写成天猫/京东/抖音。
+数据不足时写 **数据缺失，无法判断**，不得把没有数据写成表现为零。"""
+
+
+REPORT_HEADINGS = [
+    "## 一、访客分析",
+    "## 二、加购率分析",
+    "## 三、点击率分析",
+    "## 四、转化率分析",
+    "## 五、退货率分析",
+    "## 六、销量分析",
+    "## 七、总评",
+]
+
+REPORT_SUBHEADINGS = (
+    ["### 数据快照", "### 核心判断", "### 优化建议"] * 6
+    + ["### 核心结论", "### 转化链路", "### 优先级行动"]
+)
+
+
+def _bold_first_sentence(line: str) -> str:
+    if not line or line.startswith("**"):
+        return line
+    match = re.search(r"[。！？]", line)
+    end = match.end() if match else len(line)
+    return f"**{line[:end]}**{line[end:]}"
+
+
+def _bold_key_tokens(line: str) -> str:
+    """Bold key metric tokens without nesting existing Markdown emphasis."""
+    parts = re.split(r"(\*\*.*?\*\*)", line)
+    for index in range(0, len(parts), 2):
+        part = parts[index]
+        part = re.sub(r"(?<![\w*])(P[012])(?![\w*])", r"**\1**", part)
+        part = re.sub(r"(?<![\d.*])(\d+(?:\.\d+)?\s*(?:小时|天)内)(?!\*)", r"**\1**", part)
+        part = re.sub(r"(?<![\d.*])(\d+(?:\.\d+)?%)(?!\*)", r"**\1**", part)
+        parts[index] = part
+    return "".join(parts)
+
+
+def _emphasize_report_markdown(report: str) -> str:
+    key_first_sentence_sections = {"### 数据快照", "### 核心判断", "### 核心结论"}
+    current_h3 = ""
+    pending_first_sentence = False
+    out: list[str] = []
+    for line in report.splitlines():
+        stripped = line.strip()
+        if stripped.startswith("### "):
+            current_h3 = stripped
+            pending_first_sentence = current_h3 in key_first_sentence_sections
+            out.append(stripped)
+            continue
+        if stripped and pending_first_sentence and not stripped.startswith("#"):
+            line = _bold_first_sentence(stripped)
+            pending_first_sentence = False
+        elif stripped.startswith(("P0，", "P1，", "P2，")):
+            line = f"- {stripped}"
+        line = _bold_key_tokens(line)
+        out.append(line.rstrip())
+    return "\n".join(out).strip() + "\n"
+
+
+def normalize_report_markdown(report: str) -> str:
+    """Normalize model title lines so Feishu renders the fixed report outline."""
+    heading_by_text = {heading.removeprefix("## "): heading for heading in REPORT_HEADINGS}
+    subheading_aliases = {
+        "数据快照": "### 数据快照",
+        "问题判断": "### 核心判断",
+        "核心判断": "### 核心判断",
+        "可执行建议": "### 优化建议",
+        "优化建议": "### 优化建议",
+        "核心结论": "### 核心结论",
+        "链路表现为": "### 转化链路",
+        "转化链路": "### 转化链路",
+        "优先级行动": "### 优先级行动",
+    }
+    lines = report.splitlines()
+    first_content_seen = False
+    current_h2 = ""
+    total_actions_started = False
+    normalized: list[str] = []
+    for line in lines:
+        stripped = line.strip()
+        bare = re.sub(r"^#{1,6}\s*", "", stripped)
+        if bare in heading_by_text:
+            current_h2 = heading_by_text[bare]
+            normalized.append(current_h2)
+            first_content_seen = True
+            continue
+        matched_subheading = False
+        for prefix, heading in subheading_aliases.items():
+            if bare == prefix or bare.startswith(f"{prefix}："):
+                normalized.append(heading)
+                remainder = bare[len(prefix):].lstrip("： ")
+                if remainder:
+                    normalized.extend(["", remainder])
+                matched_subheading = True
+                break
+        if matched_subheading:
+            continue
+        if (
+            current_h2 == "## 七、总评"
+            and not total_actions_started
+            and re.match(r"^(?:[-*]\s*)?(?:\*\*)?P[012]", stripped)
+        ):
+            normalized.extend(["### 优先级行动", ""])
+            total_actions_started = True
+        if stripped and not first_content_seen and "单款式多维度分析报告" in bare:
+            normalized.append(f"# {bare}")
+            first_content_seen = True
+            continue
+        normalized.append(line.rstrip())
+        if stripped:
+            first_content_seen = True
+    return _emphasize_report_markdown("\n".join(normalized))
+
+
+def has_exact_report_headings(report: str) -> bool:
+    """Return whether the seven required H2 headings exist exactly and in order."""
+    actual = [line.strip() for line in report.splitlines() if line.strip().startswith("## ")]
+    return actual == REPORT_HEADINGS
+
+
+def has_clear_report_hierarchy(report: str) -> bool:
+    actual = [line.strip() for line in report.splitlines() if line.strip().startswith("### ")]
+    required_index = 0
+    for heading in actual:
+        if required_index < len(REPORT_SUBHEADINGS) and heading == REPORT_SUBHEADINGS[required_index]:
+            required_index += 1
+    return required_index == len(REPORT_SUBHEADINGS)
+
+
+def has_key_emphasis(report: str) -> bool:
+    return has_clear_report_hierarchy(report) and len(re.findall(r"\*\*.+?\*\*", report)) >= 7
+
+
+def build_user_prompt(style: str, start: date, end: date, payload: dict) -> str:
+    return f"""# 分析对象
+- 款式: {style}
+- 时间窗口: {start} ~ {end}
+- 数据来源: PostgreSQL 中严格位于该时间窗口内的多维数据
+
+# 三日总计
+{json.dumps(payload['total'], ensure_ascii=False, indent=2)}
+
+# 按平台汇总
+{json.dumps(payload['platforms'], ensure_ascii=False, indent=2)}
+
+# 逐日经营数据
+{json.dumps(payload['daily'], ensure_ascii=False, indent=2)}
+
+# 人物画像
+{json.dumps(payload['persona'], ensure_ascii=False, indent=2)}
+
+# 商品评价
+{json.dumps(payload['product_reviews'], ensure_ascii=False, indent=2)}
+
+# 主图表现
+{json.dumps(payload['main_image'], ensure_ascii=False, indent=2)}
+
+# 营销笔记、曝光互动与评论
+{json.dumps(payload['marketing'], ensure_ascii=False, indent=2)}
+
+# 输出要求
+仅分析以上三日数据，所有规定维度都要覆盖。严格使用系统要求的七个二级标题及顺序。
+报告标题用一级标题：
+# {style} 单款式多维度分析报告 ({start}~{end})
+"""
+
+
+# ── Hermes call ───────────────────────────────────────────────────────────
+
+def call_hermes(system: str, user: str, timeout: int = 600) -> tuple[str, dict]:
+    """返回 (content, usage_dict). usage_dict 含 prompt_tokens/completion_tokens."""
+    payload = {
+        "model": HERMES_MODEL,
+        "messages": [
+            {"role": "system", "content": system},
+            {"role": "user", "content": user},
+        ],
+        "temperature": 0.3,
+        "max_tokens": 4096,
+    }
+    body = json.dumps(payload, ensure_ascii=False).encode("utf-8")
+    req = urllib.request.Request(
+        HERMES_ANALYZER_URL, data=body, method="POST",
+        headers={"Authorization": f"Bearer {HERMES_API_KEY}",
+                 "Content-Type": "application/json"},
+    )
+    with urllib.request.urlopen(req, timeout=timeout) as resp:
+        raw = resp.read().decode("utf-8", errors="replace")
+    obj = json.loads(raw)
+    content = obj["choices"][0]["message"]["content"]
+    usage = obj.get("usage") or {}
+    return content, usage
+
+
+# ── Feishu doc ────────────────────────────────────────────────────────────
+
+ANALYSIS_TIME_FIELD = "时间"
+ANALYSIS_REPORT_FIELD = "笔记分析汇总"
+
+
+def analysis_period_label(start: date, end: date) -> str:
+    """Business key used by each style's analysis Base."""
+    return f"{start:%Y-%m-%d}~{end:%m-%d}"
+
+
+def _extract_record_id(envelope: dict) -> str | None:
+    data = envelope.get("data") or {}
+    if not isinstance(data, dict):
+        return None
+    direct = data.get("record_id")
+    if direct:
+        return str(direct)
+    ids = data.get("record_id_list") or []
+    if ids:
+        return str(ids[0])
+    record = data.get("record") or {}
+    if isinstance(record, dict):
+        if record.get("record_id"):
+            return str(record["record_id"])
+        record_ids = record.get("record_id_list") or []
+        if record_ids:
+            return str(record_ids[0])
+    return None
+
+
+def ensure_select_option(
+    target: dict[str, str],
+    field_name: str,
+    option_name: str,
+    style: str,
+) -> None:
+    """Append a missing single-select option while preserving all existing options."""
+    base_token = target["base_token"]
+    table_id = target["table_id"]
+    response = run_lark_cli([
+        "base", "+field-get", "--as", "user",
+        "--base-token", base_token, "--table-id", table_id,
+        "--field-id", field_name, "--format", "json",
+    ])
+    field = (response.get("data") or {}).get("field") or {}
+    if field.get("type") != "select":
+        raise RuntimeError(f"{style} 字段 {field_name} 不是单选字段")
+    options = list(field.get("options") or [])
+    if any(str(option.get("name") or "") == option_name for option in options):
+        return
+
+    options.append({"name": option_name})
+    definition = {
+        "name": field.get("name") or field_name,
+        "type": "select",
+        "multiple": bool(field.get("multiple", False)),
+        "options": options,
+    }
+    run_lark_cli([
+        "base", "+field-update", "--as", "user",
+        "--base-token", base_token, "--table-id", table_id,
+        "--field-id", field.get("id") or field_name,
+        "--json", json.dumps(definition, ensure_ascii=False),
+        "--yes", "--format", "json",
+    ])
+
+
+def _select_value_matches(value: Any, expected: str) -> bool:
+    if isinstance(value, list):
+        return any(_select_value_matches(item, expected) for item in value)
+    if isinstance(value, dict):
+        return str(value.get("name") or value.get("text") or value.get("value") or "") == expected
+    return str(value or "") == expected
+
+
+def write_feishu_doc(title: str, markdown: str) -> tuple[str, str]:
+    """Create a Feishu document as the bot and grant the current user access."""
+    tmp = CURATED_DATA_ROOT / "hermes_reports" / f"{title}.md"
+    tmp.parent.mkdir(parents=True, exist_ok=True)
+    tmp.write_text(markdown, encoding="utf-8")
+    obj = run_lark_cli([
+        "docs", "+create",
+        "--doc-format", "markdown",
+        "--title", title,
+        "--content", f"@{tmp}",
+        "--as", "bot",
+        "--format", "json",
+    ], cwd=Path(__file__).resolve().parent, timeout=180)
+    data = obj.get("data") or {}
+    doc = data.get("document") or {}
+    url = doc.get("url") or obj.get("url") or obj.get("document_url")
+    doc_id = doc.get("document_id")
+    if not url or not doc_id:
+        raise RuntimeError(f"lark-cli docs +create missing URL or document_id: {obj}")
+
+    auth = run_lark_cli(["auth", "status", "--json", "--verify"])
+    user = (auth.get("identities") or {}).get("user") or {}
+    user_open_id = user.get("openId") or user.get("open_id")
+    if not user_open_id:
+        raise RuntimeError("当前飞书用户缺少 open_id，无法授予新文档访问权限")
+    run_lark_cli([
+        "drive", "+member-add",
+        "--token", doc_id,
+        "--type", "docx",
+        "--member-id", user_open_id,
+        "--member-type", "openid",
+        "--perm", "full_access",
+        "--as", "bot",
+        "--yes",
+        "--format", "json",
+    ])
+    return url, doc_id
+
+
+def ensure_lark_cli_ready() -> None:
+    """Refresh Feishu credentials and require both identities before analysis."""
+    auth = run_lark_cli(["auth", "status", "--json", "--verify"])
+    identities = auth.get("identities") or {}
+    unavailable: list[str] = []
+    for identity in ("bot", "user"):
+        state = identities.get(identity) or {}
+        if not state.get("available") or not state.get("verified"):
+            unavailable.append(identity)
+    if unavailable:
+        raise RuntimeError(
+            "飞书身份未就绪: " + ", ".join(unavailable)
+            + "；请检查 hermes-analyzer 的 user-default 绑定"
+        )
+
+
+def write_analysis_link_to_base(
+    target: dict[str, str],
+    period_label: str,
+    document_url: str,
+    style: str,
+) -> str:
+    """Idempotently write one three-day report link to a style analysis Base."""
+    base_token = target.get("base_token") or ""
+    table_id = target.get("table_id") or ""
+    if not base_token or not table_id:
+        raise RuntimeError(f"{style} 缺少平台单品分析 Base/Table")
+
+    fields_response = run_lark_cli([
+        "base", "+field-list", "--as", "user",
+        "--base-token", base_token, "--table-id", table_id,
+        "--format", "json",
+    ])
+    fields = (fields_response.get("data") or {}).get("fields") or []
+    field_names = {f.get("name") for f in fields if isinstance(f, dict)}
+    required = {ANALYSIS_TIME_FIELD, ANALYSIS_REPORT_FIELD}
+    missing = sorted(required - field_names)
+    if missing:
+        raise RuntimeError(f"{style} 平台单品分析表缺字段: {', '.join(missing)}")
+
+    # A Base select filter rejects values that are not options yet. Add the
+    # rolling three-day label first, then compare rows locally for idempotency.
+    ensure_select_option(target, ANALYSIS_TIME_FIELD, period_label, style)
+    existing = run_lark_cli([
+        "base", "+record-list", "--as", "user",
+        "--base-token", base_token, "--table-id", table_id,
+        "--field-id", ANALYSIS_TIME_FIELD,
+        "--field-id", ANALYSIS_REPORT_FIELD,
+        "--limit", "200", "--format", "json",
+    ])
+    existing_data = existing.get("data") or {}
+    rows = existing_data.get("data") or []
+    field_names = existing_data.get("fields") or []
+    record_ids = existing_data.get("record_id_list") or []
+    record_id = None
+    for index, row in enumerate(rows):
+        if isinstance(row, dict):
+            fields_map = row.get("fields") if isinstance(row.get("fields"), dict) else row
+            row_record_id = row.get("record_id")
+        elif isinstance(row, list):
+            fields_map = dict(zip(field_names, row))
+            row_record_id = None
+        else:
+            continue
+        if _select_value_matches(fields_map.get(ANALYSIS_TIME_FIELD), period_label):
+            candidate = row_record_id
+            if not candidate and index < len(record_ids):
+                candidate = record_ids[index]
+            if candidate:
+                record_id = str(candidate)
+            break
+
+    fields_payload = {
+        ANALYSIS_TIME_FIELD: period_label,
+        ANALYSIS_REPORT_FIELD: f"[{style} 三日单品分析]({document_url})",
+    }
+    upsert_args = [
+        "base", "+record-upsert", "--as", "user",
+        "--base-token", base_token, "--table-id", table_id,
+        "--json", json.dumps(fields_payload, ensure_ascii=False),
+    ]
+    if record_id:
+        upsert_args += ["--record-id", record_id]
+    result = run_lark_cli(upsert_args)
+    return _extract_record_id(result) or record_id or "?"
+
+
+# ── main ───────────────────────────────────────────────────────────────────
+
+def parse_args() -> argparse.Namespace:
+    p = argparse.ArgumentParser(description="单款式多维度分析 -> Hermes -> 飞书文档 -> PG")
+    p.add_argument("--style", default=None, help="款式名 (dim_style.style_name); 与 --all-styles 互斥")
+    p.add_argument("--all-styles", action="store_true", help="批量跑全款式 (有 daily 数据的)")
+    p.add_argument("--skip-existing", action="store_true",
+                   help="跳过 (style_name, analysis_date, window_days) 已入库的款式, 幂等用")
+    p.add_argument("--end-date", default=None, help="窗口右端 YYYY-MM-DD, 默认昨天")
+    p.add_argument("--days", type=int, default=3, help="窗口天数 (默认 3, 对齐'3天分析一次')")
+    p.add_argument("--min-interval-days", type=int, default=0,
+                   help="批量模式下同款最短成功分析间隔；编排使用 3")
+    p.add_argument("--dry-run", action="store_true", help="不 POST Hermes, 只打印 prompt")
+    return p.parse_args()
+
+
+def fetch_styles_with_data(end: date, days: int) -> list[str]:
+    """查窗口内有 daily_style_metrics 数据的款式, 按销量降序."""
+    start = end - timedelta(days=days - 1)
+    with get_conn() as c, c.cursor() as cur:
+        cur.execute(
+            """SELECT style_name, SUM(sales) total_sales
+               FROM daily_style_metrics
+               WHERE metric_date BETWEEN %s AND %s
+               GROUP BY style_name
+               ORDER BY total_sales DESC NULLS LAST""",
+            (start, end),
+        )
+        return [r[0] for r in cur.fetchall()]
+
+
+def fetch_existing_styles(end: date, days: int) -> set[str]:
+    """查 (analysis_date, window_days) 已入库的款式名集合."""
+    with get_conn() as c, c.cursor() as cur:
+        cur.execute(
+            """SELECT style_name FROM fact_style_analysis_reports
+               WHERE analysis_date=%s AND window_days=%s""",
+            (end, days),
+        )
+        return {r[0] for r in cur.fetchall()}
+
+
+def fetch_latest_analysis_dates(end: date, days: int) -> dict[str, date]:
+    """Return each style's latest successful report date up to ``end``."""
+    with get_conn() as c, c.cursor() as cur:
+        cur.execute(
+            """SELECT style_name, MAX(analysis_date)
+               FROM fact_style_analysis_reports
+               WHERE analysis_date <= %s AND window_days = %s
+               GROUP BY style_name""",
+            (end, days),
+        )
+        return {str(style): analysis_date for style, analysis_date in cur.fetchall()}
+
+
+def select_due_styles(
+    styles: list[str],
+    latest_dates: dict[str, date],
+    end: date,
+    min_interval_days: int,
+) -> list[str]:
+    if min_interval_days <= 0:
+        return list(styles)
+    return [
+        style for style in styles
+        if style not in latest_dates
+        or (end - latest_dates[style]).days >= min_interval_days
+    ]
+
+
+def analyze_one_style(
+    style: str,
+    start: date,
+    end: date,
+    days: int,
+    dry_run: bool,
+    target: dict[str, str] | None = None,
+) -> tuple[bool, str]:
+    """跑一个款式的完整流程. 返回 (success, message)."""
+    print(f"\n{'=' * 70}")
+    print(f"[STYLE] {style}  window={start}~{end}")
+    print(f"{'=' * 70}")
+
+    daily = pull_daily(style, start, end)
+    if not daily:
+        return False, "daily_style_metrics 没数据"
+    missing_dates = missing_window_dates(daily, start, end)
+    if missing_dates:
+        missing_text = ", ".join(day.isoformat() for day in missing_dates)
+        return False, f"daily_style_metrics 缺少日期: {missing_text}"
+    this_week = aggregate_period(daily)
+    payload = {
+        "total": aggregate_style_total(this_week),
+        "platforms": this_week,
+        "daily": daily,
+        "persona": pull_persona(style, start, end),
+        "product_reviews": pull_reviews(style, start, end),
+        "main_image": pull_main_image(style, start, end),
+        "marketing": pull_seeding(style, start, end),
+    }
+    user_prompt = build_user_prompt(style, start, end, payload)
+
+    if dry_run:
+        print(f"[DRY] prompt 长度 {len(user_prompt)} chars (不 POST Hermes)")
+        return True, f"dry-run ok, prompt={len(user_prompt)} chars"
+
+    print(f"[INFO] prompt 长度 {len(user_prompt)} chars, POST Hermes ...")
+    try:
+        report, usage = call_hermes(SYSTEM, user_prompt)
+    except Exception as exc:
+        return False, f"Hermes 调用失败: {exc}"
+    report = normalize_report_markdown(report)
+    if not has_exact_report_headings(report):
+        return False, "Hermes 报告缺少固定七章节或章节顺序错误"
+    if not has_clear_report_hierarchy(report):
+        return False, "Hermes 报告缺少固定三级层级或层级顺序错误"
+    if not has_key_emphasis(report):
+        return False, "Hermes 报告重点内容未按要求加粗"
+    print(f"[OK] Hermes 返回 {len(report)} chars")
+
+    title = f"{style} 单款式多维度分析 {start}~{end}"
+    try:
+        url, doc_id = write_feishu_doc(title, report)
+    except Exception as exc:
+        return False, f"飞书文档创建失败: {exc}"
+    print(f"[DOC] 飞书: {url}")
+
+    if not target:
+        return False, "主配置表未配置平台单品分析地址"
+    try:
+        target_record_id = write_analysis_link_to_base(
+            target,
+            analysis_period_label(start, end),
+            url,
+            style,
+        )
+    except Exception as exc:
+        return False, f"单品分析 Base 写入失败: {exc}"
+    print(f"[BASE] 平台单品分析 record_id={target_record_id}")
+
+    platforms_included = sorted(this_week.keys())
+    record = {
+        "style_name": style,
+        "analysis_date": end,
+        "window_days": days,
+        "window_start": start,
+        "window_end": end,
+        "platforms_included": platforms_included,
+        "feishu_doc_id": doc_id,
+        "feishu_doc_url": url,
+        "report_title": title,
+        "report_markdown": report,
+        "hermes_model": HERMES_MODEL,
+        "prompt_tokens": usage.get("prompt_tokens"),
+        "completion_tokens": usage.get("completion_tokens"),
+        "source_payload": payload,
+    }
+    with get_conn() as conn:
+        row_id = upsert_style_analysis_report(conn, record)
+    print(f"[DB] id={row_id}")
+    return True, url
+
+
+def main() -> int:
+    args = parse_args()
+    if not args.style and not args.all_styles:
+        print("[ERR] 必须指定 --style <款式名> 或 --all-styles", file=sys.stderr)
+        return 2
+    if args.style and args.all_styles:
+        print("[ERR] --style 和 --all-styles 互斥", file=sys.stderr)
+        return 2
+
+    end = date.fromisoformat(args.end_date) if args.end_date else date.today() - timedelta(days=1)
+    start = end - timedelta(days=args.days - 1)
+    if not args.dry_run:
+        try:
+            ensure_lark_cli_ready()
+        except Exception as exc:
+            print(f"[ERR] 飞书运行环境预检失败: {exc}", file=sys.stderr)
+            return 1
+    targets = get_style_analysis_bitable_map()
+
+    if args.style:
+        target = targets.get(args.style)
+        if not target:
+            print(f"[FAIL] {args.style}: 主配置表未配置平台单品分析地址", file=sys.stderr)
+            return 1
+        ok, msg = analyze_one_style(args.style, start, end, args.days, args.dry_run, target)
+        if not ok:
+            print(f"[FAIL] {args.style}: {msg}", file=sys.stderr)
+            return 1
+        print(f"[DONE] {args.style}: {msg}")
+        return 0
+
+    # --all-styles 批量
+    styles_with_data = fetch_styles_with_data(end, args.days)
+    styles = [style for style in styles_with_data if style in targets]
+    missing_targets = [style for style in styles_with_data if style not in targets]
+    if missing_targets:
+        print(f"[SKIP] 未配置平台单品分析地址: {len(missing_targets)} 款")
+    if not styles:
+        print(f"[ERR] 窗口 {start}~{end} 内没有款式有 daily 数据", file=sys.stderr)
+        return 1
+
+    existing = fetch_existing_styles(end, args.days) if args.skip_existing else set()
+    if args.skip_existing:
+        print(f"[INFO] --skip-existing: 已入库 {len(existing)} 款, 跳过")
+
+    candidates = [s for s in styles if s not in existing]
+    latest_dates = (
+        fetch_latest_analysis_dates(end, args.days)
+        if args.min_interval_days > 0 else {}
+    )
+    todo = select_due_styles(candidates, latest_dates, end, args.min_interval_days)
+    skipped = len(styles) - len(todo)
+    print(f"[INFO] 待跑 {len(todo)} 款 (跳过 {skipped} 款已有), 窗口 {start}~{end} ({args.days}d)")
+    print(f"[INFO] 预计耗时 ~{len(todo) * 2} 分钟 (模型服务耗时会波动)")
+
+    success, failed = [], []
+    for i, style in enumerate(todo, 1):
+        print(f"\n[{i}/{len(todo)}] ", end="")
+        try:
+            ok, msg = analyze_one_style(style, start, end, args.days, args.dry_run, targets[style])
+            if ok:
+                success.append((style, msg))
+            else:
+                failed.append((style, msg))
+        except Exception as exc:
+            failed.append((style, f"未捕获异常: {exc}"))
+
+    print(f"\n{'=' * 70}")
+    print(f"[SUMMARY] 成功 {len(success)} / 失败 {len(failed)} / 跳过 {skipped}")
+    if failed:
+        print("[FAILED]")
+        for style, msg in failed:
+            print(f"  - {style}: {msg}")
+    return 1 if failed else 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/backfill_collect.py b/src/gyxx_flow/modules/product_commerce/runtime/backfill_collect.py
new file mode 100644
index 0000000..7193a91
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/backfill_collect.py
@@ -0,0 +1,208 @@
+"""日期范围批量回补：ERP + 三平台 + 每日汇总。
+
+默认每次都重跑所有组件并覆盖旧数据；用 --mode skip-existing 只补缺失。
+
+复用 orchestrate_daily_collection.py 的 collect/analyze 阶段，对 [from, to]
+闭区间内每一天按 mode 决定跑哪些组件。
+
+用法:
+    .venv/Scripts/python.exe backfill_collect.py --from 2026-05-20 --to 2026-05-30
+    .venv/Scripts/python.exe backfill_collect.py --from 2026-05-20 --to 2026-05-30 --mode skip-existing
+    .venv/Scripts/python.exe backfill_collect.py --from 2026-05-20 --to 2026-05-30 --components jd,dy
+    .venv/Scripts/python.exe backfill_collect.py --from 2026-05-20 --to 2026-05-30 --dry-run
+
+模式:
+    force (默认)          重跑所有指定组件（覆盖旧数据）
+    skip-existing        已有数据则跳过；缺哪个组件跑哪个
+"""
+
+import argparse
+import os
+import subprocess
+import sys
+from datetime import date, datetime, timedelta
+from pathlib import Path
+
+from erp_metric_overrides import DATA_ROOT, PLATFORM_DIRS
+from runtime_paths import runtime_script
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+
+ALL_COMPONENTS = ["erp", "jd", "dy", "tm"]
+PLATFORM_KEYS = ["jd", "dy", "tm"]
+
+
+def _subprocess_env() -> dict:
+    """强制子进程 KEEP_BROWSER_OPEN=0 (vendors/dy-data-flow 的 _keep_browser_open 是死循环)。"""
+    env = os.environ.copy()
+    env["KEEP_BROWSER_OPEN"] = "0"
+    return env
+
+
+def parse_date(value: str) -> date:
+    return datetime.strptime(value, "%Y-%m-%d").date()
+
+
+def _iter_style_dirs(platform: str, target: date):
+    base = DATA_ROOT / PLATFORM_DIRS[platform]
+    if not base.exists():
+        return
+    suffix = f"_{target.isoformat()}"
+    for style_dir in base.glob(f"*{suffix}"):
+        if style_dir.is_dir():
+            yield style_dir
+
+
+def has_platform_data(platform: str, target: date) -> bool:
+    """某平台某日期是否已有至少一个款的产品汇总 json。"""
+    for style_dir in _iter_style_dirs(platform, target):
+        for json_path in style_dir.glob("*.json"):
+            if json_path.name == "data.json" or json_path.name == "final_summary.json":
+                continue
+            if json_path.name.endswith(f"_{target.isoformat()}.json") or json_path.name.startswith("业务款式_"):
+                return True
+    return False
+
+
+def has_erp_data(target: date) -> bool:
+    """ERP 数据写入每个平台 data.json；任一平台有 data.json 即视为 ERP 跑过。"""
+    for platform in PLATFORM_KEYS:
+        for style_dir in _iter_style_dirs(platform, target):
+            if (style_dir / "data.json").exists():
+                return True
+    return False
+
+
+def probe(target: date, components: list[str]) -> dict[str, bool]:
+    result = {}
+    for c in components:
+        if c == "erp":
+            result[c] = has_erp_data(target)
+        else:
+            result[c] = has_platform_data(c, target)
+    return result
+
+
+def decide_components(status: dict[str, bool], mode: str) -> list[str] | None:
+    """根据 mode 和当前状态决定要跑的组件；None 表示跳过整个日期。"""
+    if mode == "force":
+        return list(status.keys())
+    missing = [c for c in status if not status[c]]
+    return None if not missing else missing
+
+
+def build_orchestrator_args(target: date, components: list[str], stages: list[str]) -> list[str] | None:
+    args = [
+        sys.executable,
+        str(runtime_script("orchestrate_daily_collection.py")),
+        "--target-date",
+        target.isoformat(),
+    ]
+    args.extend(["--stages", ",".join(stages)])
+    # 历史补采用 backfill 模式: ERP 抓"自定义销量/退货量"列 (表头匹配 + SPU 行求和),
+    # 而不是 daily 模式的 cell 45/57 "昨日销量"列
+    if "erp" in components:
+        args.extend(["--erp-mode", "backfill"])
+    if "erp" not in components:
+        args.append("--skip-erp")
+    platforms = [c for c in components if c in PLATFORM_KEYS]
+    if not platforms and "erp" not in components:
+        return None
+    if platforms and platforms != PLATFORM_KEYS:
+        args.extend(["--platforms", ",".join(platforms)])
+    return args
+
+
+def iter_dates(d_from: date, d_to: date):
+    cur = d_from
+    while cur <= d_to:
+        yield cur
+        cur += timedelta(days=1)
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="日期范围批量回补：ERP + 3 平台 + 每日汇总",
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    parser.add_argument("--from", dest="from_date", type=parse_date, required=True, help="起始日期 YYYY-MM-DD（含）")
+    parser.add_argument("--to", dest="to_date", type=parse_date, required=True, help="结束日期 YYYY-MM-DD（含）")
+    parser.add_argument(
+        "--components",
+        default=",".join(ALL_COMPONENTS),
+        help=f"逗号分隔组件，默认全部。候选: {','.join(ALL_COMPONENTS)}",
+    )
+    parser.add_argument(
+        "--mode",
+        choices=["force", "skip-existing"],
+        default="force",
+        help="force (默认): 重跑所有指定组件，覆盖旧数据；skip-existing: 已有数据则跳过，缺哪个补哪个",
+    )
+    parser.add_argument("--no-analyze", action="store_true", help="只跑 collect，不跑汇总")
+    parser.add_argument("--with-export", action="store_true", help="额外跑 export 阶段（生成飞书可导入文件）")
+    parser.add_argument("--with-insert", action="store_true", help="额外跑 insert 阶段（推入飞书多维表格），需要 --with-export")
+    parser.add_argument("--serial-platforms", action="store_true", help="平台串行跑（默认并行）")
+    parser.add_argument("--dry-run", action="store_true", help="只打印计划，不实际执行")
+    args = parser.parse_args()
+
+    if args.from_date > args.to_date:
+        parser.error("--from must be <= --to")
+
+    if args.with_insert and not args.with_export:
+        parser.error("--with-insert 需要 --with-export（insert 依赖 export 产出的文件）")
+
+    components = [c.strip() for c in args.components.split(",") if c.strip()]
+    invalid = [c for c in components if c not in ALL_COMPONENTS]
+    if invalid:
+        parser.error(f"unknown components: {invalid}; valid: {ALL_COMPONENTS}")
+
+    stages = ["collect"]
+    if not args.no_analyze:
+        stages.append("analyze")
+    if args.with_export:
+        stages.append("export")
+    if args.with_insert:
+        stages.append("insert")
+
+    days = (args.to_date - args.from_date).days + 1
+    print(f"[INFO] 日期范围: {args.from_date} ~ {args.to_date}  ({days} 天)")
+    print(f"[INFO] 组件: {','.join(components)}")
+    print(f"[INFO] 模式: {args.mode}")
+    print(f"[INFO] 阶段: {','.join(stages)}")
+    if args.dry_run:
+        print("[INFO] DRY-RUN 模式，不实际执行")
+
+    plan = {"skip": 0, "partial": 0, "full": 0, "err": 0}
+    for target in iter_dates(args.from_date, args.to_date):
+        status = probe(target, components)
+        to_run = decide_components(status, args.mode)
+        status_str = " ".join(f"{c}={'Y' if status[c] else '-'}" for c in components)
+        if to_run is None:
+            print(f"[SKIP] {target}  {status_str}  全部已有")
+            plan["skip"] += 1
+            continue
+        is_full = sorted(to_run) == sorted(components)
+        kind = "full" if is_full else "partial"
+        run_str = ",".join(to_run)
+        print(f"[RUN ] {target}  {status_str}  → 跑 {run_str}")
+        plan[kind] += 1
+        if args.dry_run:
+            continue
+        cmd = build_orchestrator_args(target, to_run, stages)
+        if cmd is None:
+            continue
+        if args.serial_platforms:
+            cmd.append("--serial-platforms")
+        proc = subprocess.run(cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace", env=_subprocess_env())
+        if proc.returncode:
+            plan["err"] += 1
+            print(f"[FAIL] {target} exit={proc.returncode}")
+
+    print(f"[DONE] skip={plan['skip']} full={plan['full']} partial={plan['partial']} err={plan['err']}")
+    if plan["err"]:
+        sys.exit(1)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/backfill_one_day.py b/src/gyxx_flow/modules/product_commerce/runtime/backfill_one_day.py
new file mode 100644
index 0000000..fa48a95
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/backfill_one_day.py
@@ -0,0 +1,219 @@
+"""
+历史补采某一天数据的工作流 (并行 + 多层级重试)。
+
+4 个流并行 (ThreadPoolExecutor max_workers=4):
+  - erp  (1 个进程, 内部按 jd/dy/tm 串行 + per-style 重试)
+  - jd   (1 个进程, per-style 重试)
+  - dy   (1 个进程, per-page / per-product 重试)
+  - tm   (1 个进程, per-page 重试)
+
+每个流外层 per-run 重试 2 次 (subprocess 失败 → 5s sleep → 再 subprocess)。
+每个 collect 脚本内部还有 per-style 重试 2 次 + per-page 重试 1 次 (见
+collect_retry_utils.py), 失败清单落 data/logs/failed_<date>_<platform>.json。
+
+Chrome profile 不冲突 (ERP DynamicFetcher 新实例 / JD JD profile / DY compass
+cookies / TM 可迁移 profile), 4 个流可以同时跑。
+
+用法:
+    .venv\\Scripts\\python.exe backfill_one_day.py --target-date 2026-05-01
+    .venv\\Scripts\\python.exe backfill_one_day.py --target-date 2026-05-01 --serial
+    .venv\\Scripts\\python.exe backfill_one_day.py --target-date 2026-05-01 --skip tm
+
+跑前要求: 关掉日常 Chrome 进程, 避免 SingletonLock 冲突 (同 collect 脚本)。
+"""
+
+from __future__ import annotations
+
+import argparse
+import subprocess
+import sys
+import time
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from datetime import datetime
+from pathlib import Path
+
+from collect_retry_utils import consolidate_failed_logs
+from runtime_paths import runtime_script
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+PY = sys.executable
+
+
+# 4 个并行流; TM 走 backfill 版, ERP 强制 backfill 模式
+STREAMS = [
+    ("erp", [PY, str(runtime_script("collect_erp_yesterday_metrics.py")), "--platform", "all", "--mode", "backfill"]),
+    ("jd", [PY, str(runtime_script("jd_product_data_collector.py"))]),
+    ("dy", [PY, str(runtime_script("dy_product_scraping.py"))]),
+    ("tm", [PY, str(runtime_script("taobao_sycm_collect_backfill.py"))]),
+]
+
+# 顶层 per-run 重试参数 (subprocess 失败时整体重试)
+MAX_RUN_RETRIES = 2
+STREAM_SLEEP_SECONDS = 5
+
+
+def parse_date(value: str) -> str:
+    try:
+        datetime.strptime(value, "%Y-%m-%d")
+    except ValueError as exc:
+        raise argparse.ArgumentTypeError(f"date must be YYYY-MM-DD, got {value!r}") from exc
+    return value
+
+
+def command_with_target_date(command: list[str], target_date: str) -> list[str]:
+    return [*command, "--target-date", target_date]
+
+
+def run_subprocess(name: str, command: list[str], target_date: str) -> tuple[int, float]:
+    """跑一次子命令, 返回 (returncode, elapsed_seconds)。失败不抛, 让外层重试。"""
+    cmd = command_with_target_date(command, target_date)
+    printable = " ".join(cmd)
+    t0 = time.time()
+    print(f"\n[RUN] {name}: {printable}", flush=True)
+    proc = subprocess.run(
+        cmd,
+        cwd=PROJECT_ROOT,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+    )
+    elapsed = time.time() - t0
+    return proc.returncode, elapsed
+
+
+def run_stream_with_retry(name: str, command: list[str], target_date: str,
+                          max_retries: int = MAX_RUN_RETRIES) -> dict:
+    """外层 per-run 重试包装: subprocess 失败 → 睡 → 再 subprocess, 最多 max_retries 次。
+
+    返回 {name, ok, attempts, returncodes: [...]}
+    """
+    attempts_made = 0
+    returncodes: list[int] = []
+    for attempt in range(1, max_retries + 2):  # 1 + max_retries = 总次数
+        attempts_made = attempt
+        rc, elapsed = run_subprocess(name, command, target_date)
+        returncodes.append(rc)
+        if rc == 0:
+            print(f"[OK] {name} (attempt {attempt}, {elapsed:.1f}s)", flush=True)
+            return {"name": name, "ok": True, "attempts": attempt, "returncodes": returncodes,
+                    "elapsed": elapsed}
+        if attempt > max_retries:
+            print(f"[FAIL] {name} 已重试 {max_retries} 次, 仍失败 (最后 rc={rc})", flush=True)
+            break
+        print(f"[WARN] {name} 失败 (rc={rc}, {elapsed:.1f}s), "
+              f"等待 {STREAM_SLEEP_SECONDS}s 后第 {attempt}/{max_retries} 次重试", flush=True)
+        time.sleep(STREAM_SLEEP_SECONDS)
+    return {"name": name, "ok": False, "attempts": attempts_made, "returncodes": returncodes}
+
+
+def run_analyze(target_date: str) -> int:
+    """跑汇总, 跟之前一样: 平台失败 analyze 仍跑 (用已有数据)。"""
+    cmd = [PY, str(runtime_script("aggregate_daily_final.py")), "--date", target_date]
+    printable = " ".join(cmd)
+    print(f"\n[RUN] analyze: {printable}", flush=True)
+    proc = subprocess.run(cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace")
+    print(f"[{'OK' if proc.returncode == 0 else 'FAIL'}] analyze (rc={proc.returncode})", flush=True)
+    return proc.returncode
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="历史补采某一天 ERP + 三平台 + 汇总 的工作流 (并行 + 多层级重试)",
+    )
+    parser.add_argument("--target-date", type=parse_date, required=True, help="要补采的日期 YYYY-MM-DD")
+    parser.add_argument(
+        "--skip",
+        default="",
+        help="逗号分隔要跳过的阶段: erp,jd,dy,tm,analyze; 默认全跑",
+    )
+    parser.add_argument(
+        "--serial",
+        action="store_true",
+        help="强制 4 流串行 (默认 ThreadPoolExecutor 并行)",
+    )
+    parser.add_argument(
+        "--max-run-retries",
+        type=int,
+        default=MAX_RUN_RETRIES,
+        help=f"per-run 整体重试次数 (默认 {MAX_RUN_RETRIES})",
+    )
+    args = parser.parse_args()
+
+    target_date = args.target_date
+    skipped = {s.strip().lower() for s in args.skip.split(",") if s.strip()}
+    valid = {"erp", "jd", "dy", "tm", "analyze"}
+    if not skipped.issubset(valid):
+        raise SystemExit(f"[ERR] --skip 包含非法阶段: {sorted(skipped - valid)}; 合法: {sorted(valid)}")
+
+    parallel = not args.serial
+    print(f"[INFO] target_date={target_date}")
+    print(f"[INFO] 跳过: {sorted(skipped) if skipped else '(无)'}")
+    print(f"[INFO] 模式: {'4 流并行 (ThreadPoolExecutor)' if parallel else '4 流串行'}")
+    print(f"[INFO] per-run 重试: {args.max_run_retries} 次, 间隔 {STREAM_SLEEP_SECONDS}s")
+
+    t0 = time.time()
+    results: dict[str, dict] = {}
+    runnable = [(name, cmd) for name, cmd in STREAMS if name not in skipped]
+    skipped_results = {name: {"name": name, "ok": True, "attempts": 0, "skipped": True}
+                       for name, _ in STREAMS if name in skipped}
+
+    if parallel and runnable:
+        with ThreadPoolExecutor(max_workers=len(runnable)) as pool:
+            futures = {
+                pool.submit(run_stream_with_retry, name, cmd, target_date,
+                            args.max_run_retries): name
+                for name, cmd in runnable
+            }
+            for future in as_completed(futures):
+                name = futures[future]
+                try:
+                    results[name] = future.result()
+                except Exception as exc:
+                    print(f"[FAIL] {name} 抛出未捕获异常: {exc}", flush=True)
+                    results[name] = {"name": name, "ok": False, "attempts": 0, "error": str(exc)}
+    else:
+        for name, cmd in runnable:
+            results[name] = run_stream_with_retry(name, cmd, target_date, args.max_run_retries)
+
+    results.update(skipped_results)
+
+    # analyze 阶段
+    if "analyze" not in skipped:
+        analyze_rc = run_analyze(target_date)
+        results["analyze"] = {"name": "analyze", "ok": analyze_rc == 0, "rc": analyze_rc}
+    else:
+        results["analyze"] = {"name": "analyze", "ok": True, "skipped": True}
+
+    # 合并失败清单
+    summary_path = consolidate_failed_logs(target_date)
+
+    # 收尾汇总
+    elapsed_total = time.time() - t0
+    print("\n" + "=" * 60)
+    print(f"  backfill 收尾: {target_date} (总耗时 {elapsed_total:.1f}s)")
+    print("=" * 60)
+    for name in ("erp", "jd", "dy", "tm", "analyze"):
+        r = results.get(name, {})
+        if r.get("skipped"):
+            tag = "[SKIP]"
+        elif r.get("ok"):
+            tag = "[OK]  "
+        else:
+            tag = "[FAIL]"
+        attempts = r.get("attempts", 0)
+        elapsed = r.get("elapsed")
+        suffix = f" (attempt {attempts}" + (f", {elapsed:.1f}s" if elapsed else "") + ")" if attempts else ""
+        print(f"  {tag} {name:8s}{suffix}")
+    if summary_path:
+        print(f"\n[INFO] 失败清单合并: {summary_path}")
+    failed = [name for name, r in results.items() if not r.get("ok") and not r.get("skipped")]
+    if failed:
+        print(f"[WARN] 失败阶段: {', '.join(failed)}")
+        return 2
+    print(f"[OK] 全部阶段成功")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/backfill_poseidon_sales.py b/src/gyxx_flow/modules/product_commerce/runtime/backfill_poseidon_sales.py
new file mode 100644
index 0000000..e6c5af6
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/backfill_poseidon_sales.py
@@ -0,0 +1,526 @@
+#!/usr/bin/env python3
+"""回填“波塞冬手提电脑包”三平台 Excel 历史数据。
+
+默认处理 2026-07-01 至 2026-07-25。脚本只按飞书配置中的平台商品 ID
+精确匹配，不会把波塞冬 edge 等相似款合并进来。
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import subprocess
+import sys
+from datetime import date, datetime, timedelta
+from pathlib import Path
+from typing import Any
+
+import pandas as pd
+
+import export_bitable_records
+from config.style_config_loader import StyleConfigLoader
+from runtime_paths import (
+    CURATED_DATA_ROOT,
+    EXPORTS_DATA_ROOT,
+    NORMALIZED_DATA_ROOT,
+    RAW_DATA_ROOT,
+    runtime_script,
+)
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+DOWNLOAD_ROOT = RAW_DATA_ROOT
+PLATFORM_OUTPUT_ROOT = NORMALIZED_DATA_ROOT
+FINAL_OUTPUT_ROOT = CURATED_DATA_ROOT
+REPORT_OUTPUT_ROOT = EXPORTS_DATA_ROOT
+STYLE_NAME = "波塞冬手提电脑包"
+PLATFORM_DIRS = {
+    "jd": "jd单品数据",
+    "dy": "dy单品数据",
+    "tm": "tm单品数据",
+}
+DATE_HEADERS = {
+    "jd": ("时间",),
+    "dy": ("统计周期",),
+    "tm": ("统计日期",),
+}
+ID_HEADERS = {
+    "jd": ("商品ID",),
+    "dy": ("商品编码", "商品ID"),
+    "tm": ("商品ID",),
+}
+METRIC_HEADERS = {
+    "jd": {
+        "visitors": ("访客数", "商品访客数"),
+        "cart_users": ("加购人数", "加购客户数"),
+        "sales": ("成交单量", "成交订单"),
+        "refund_orders": ("取消及售后退款单量", "申请退货件数", "退货件数"),
+    },
+    "dy": {
+        "visitors": ("商品点击人数", "点击人数", "商品访客数"),
+        "cart_users": ("加购人数", "商品加购人数"),
+        "sales": ("成交人数", "支付人数"),
+        "refund_orders": ("退款订单数(支付时间)", "退款订单数", "退款单数"),
+    },
+    "tm": {
+        "visitors": ("商品访客数",),
+        "cart_users": ("商品加购人数",),
+        "sales": ("支付件数",),
+        # 生意参谋商品明细只有成功退款金额，没有退款订单数。
+        "refund_orders": (),
+    },
+}
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--start", default="2026-07-01", help="开始日期 YYYY-MM-DD")
+    parser.add_argument("--end", default="2026-07-25", help="结束日期 YYYY-MM-DD")
+    parser.add_argument("--insert", action="store_true", help="生成后幂等写入飞书多维表格")
+    parser.add_argument("--skip-db", action="store_true", help="只重建本地文件，不同步 PostgreSQL")
+    parser.add_argument("--force-refresh-config", action="store_true", help="强制刷新飞书款式配置")
+    return parser.parse_args()
+
+
+def iter_dates(start: str, end: str) -> list[str]:
+    first = date.fromisoformat(start)
+    last = date.fromisoformat(end)
+    if first > last:
+        raise ValueError(f"开始日期晚于结束日期: {start} > {end}")
+    result = []
+    current = first
+    while current <= last:
+        result.append(current.isoformat())
+        current += timedelta(days=1)
+    return result
+
+
+def normalize_header(value: Any) -> str:
+    return re.sub(r"\s+", "", str(value or "")).strip()
+
+
+def normalize_id(value: Any) -> str:
+    text = str(value or "").strip()
+    if re.fullmatch(r"\d+\.0", text):
+        text = text[:-2]
+    return text
+
+
+def number(value: Any) -> float:
+    text = str(value or "").strip().replace(",", "").replace("%", "")
+    if not text or text in {"-", "--", "nan", "None"}:
+        return 0.0
+    match = re.search(r"-?\d+(?:\.\d+)?", text)
+    return float(match.group(0)) if match else 0.0
+
+
+def _dedupe_headers(values: list[Any]) -> list[str]:
+    result = []
+    seen: dict[str, int] = {}
+    for index, value in enumerate(values):
+        name = str(value or "").strip() or f"col_{index + 1}"
+        seen[name] = seen.get(name, 0) + 1
+        result.append(name if seen[name] == 1 else f"{name}_{seen[name]}")
+    return result
+
+
+def read_excel_table(path: Path, platform: str) -> pd.DataFrame:
+    raw = pd.read_excel(path, header=None, dtype=str).fillna("")
+    id_aliases = set(ID_HEADERS[platform])
+    header_index = None
+    for index in range(min(30, len(raw))):
+        headers = {normalize_header(value) for value in raw.iloc[index].tolist()}
+        if headers.intersection(id_aliases):
+            header_index = index
+            break
+    if header_index is None:
+        raise ValueError(f"未找到 {platform} 商品 ID 表头: {path}")
+    headers = _dedupe_headers(raw.iloc[header_index].tolist())
+    frame = raw.iloc[header_index + 1 :].copy()
+    frame.columns = headers
+    frame = frame.reset_index(drop=True)
+    return frame
+
+
+def _find_column(frame: pd.DataFrame, aliases: tuple[str, ...]) -> str | None:
+    normalized = {normalize_header(column): column for column in frame.columns}
+    for alias in aliases:
+        if alias in normalized:
+            return normalized[alias]
+    for alias in aliases:
+        for normalized_name, original in normalized.items():
+            if alias and alias in normalized_name:
+                return original
+    return None
+
+
+def _is_exact_date(value: Any, target_date: str) -> bool:
+    dates = re.findall(r"\d{4}[-/]\d{2}[-/]\d{2}", str(value or ""))
+    normalized = {item.replace("/", "-") for item in dates}
+    return normalized == {target_date} and len(dates) == 1
+
+
+def extract_metrics(platform: str, frame: pd.DataFrame, product_ids: list[str], target_date: str) -> dict:
+    id_column = _find_column(frame, ID_HEADERS[platform])
+    if not id_column:
+        raise ValueError(f"{platform} 缺少商品 ID 列")
+    date_column = _find_column(frame, DATE_HEADERS[platform])
+    if date_column:
+        frame = frame[frame[date_column].map(lambda value: _is_exact_date(value, target_date))]
+
+    wanted = {normalize_id(value) for value in product_ids if normalize_id(value)}
+    row_ids = frame[id_column].map(normalize_id)
+    matched_frame = frame[row_ids.isin(wanted)].copy()
+    matched_ids = sorted({normalize_id(value) for value in matched_frame[id_column]})
+
+    metrics: dict[str, int] = {}
+    for metric, aliases in METRIC_HEADERS[platform].items():
+        column = _find_column(matched_frame, aliases) if aliases else None
+        total = sum(number(value) for value in matched_frame[column]) if column else 0.0
+        metrics[metric] = int(round(total))
+
+    return {
+        "metrics": metrics,
+        "matched_ids": matched_ids,
+        "missing_ids": [value for value in product_ids if normalize_id(value) not in matched_ids],
+        "matched_row_count": int(len(matched_frame)),
+        "target_date_row_count": int(len(frame)),
+    }
+
+
+def _daily_filename_score(path: Path, target_date: str) -> int:
+    dashed = target_date
+    compact = target_date.replace("-", "")
+    name = path.name
+    return int(f"{dashed}_{dashed}" in name or f"{compact}-{compact}" in name)
+
+
+def choose_excel(platform: str, target_date: str, product_ids: list[str]) -> dict:
+    download_dir = DOWNLOAD_ROOT / PLATFORM_DIRS[platform] / "_downloads" / target_date
+    candidates = sorted(
+        path for path in download_dir.glob("*.xls*") if path.is_file() and not path.name.startswith("~$")
+    )
+    if not candidates:
+        raise FileNotFoundError(f"{platform} {target_date} 没有 Excel: {download_dir}")
+
+    parsed = []
+    errors = []
+    for path in candidates:
+        try:
+            frame = read_excel_table(path, platform)
+            result = extract_metrics(platform, frame, product_ids, target_date)
+            result["source_file"] = str(path.resolve())
+            result["candidate_file_count"] = len(candidates)
+            score = (
+                len(result["matched_ids"]),
+                _daily_filename_score(path, target_date),
+                result["target_date_row_count"],
+                path.stat().st_mtime_ns,
+                path.name,
+            )
+            parsed.append((score, result))
+        except Exception as exc:
+            errors.append(f"{path.name}: {type(exc).__name__}: {exc}")
+    if not parsed:
+        raise RuntimeError(f"{platform} {target_date} Excel 全部解析失败: {'; '.join(errors)}")
+    parsed.sort(key=lambda item: item[0])
+    chosen = parsed[-1][1]
+    chosen["candidate_errors"] = errors
+    return chosen
+
+
+def build_final_payload(target_date: str, parsed: dict[str, dict], erp_codes: list[str]) -> dict:
+    payload: dict[str, Any] = {
+        "date": target_date,
+        "style_name": STYLE_NAME,
+        "generated_at": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+        "metrics_note": {
+            "sales": "临时历史回填直接使用平台 Excel 原始销量，避免旧 ERP 跳过值覆盖为 0",
+            "jd.sales": "成交单量",
+            "dy.sales": "成交人数",
+            "tm.sales": "支付件数",
+            "tm.refund_orders": "生意参谋商品明细无退款单量字段，本次记 0",
+        },
+        "erp": {
+            "style_codes": erp_codes,
+            "source": "temporary_three_platform_excel_backfill",
+        },
+    }
+    for platform in ("jd", "dy", "tm"):
+        item = parsed[platform]
+        metrics = item["metrics"]
+        payload[platform] = {
+            **metrics,
+            "source": item["source_file"],
+            "matched_product_ids": item["matched_ids"],
+            "missing_product_ids": item["missing_ids"],
+            "matched_row_count": item["matched_row_count"],
+        }
+    return payload
+
+
+def _write_json(path: Path, payload: Any) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8-sig")
+
+
+def _render_platform_markdown(platform: str, target_date: str, product_ids: list[str], item: dict) -> str:
+    metrics = item["metrics"]
+    return "\n".join(
+        [
+            f"# {STYLE_NAME} - {platform.upper()} Excel 汇总 - {target_date}",
+            "",
+            f"- 源文件: {item['source_file']}",
+            f"- 配置商品ID: {', '.join(product_ids)}",
+            f"- 匹配商品ID: {', '.join(item['matched_ids']) or '-'}",
+            f"- 未匹配商品ID: {', '.join(item['missing_ids']) or '-'}",
+            "",
+            "| 销量 | 退款订单数 | 访客数 | 加购人数 |",
+            "|---:|---:|---:|---:|",
+            f"| {metrics['sales']} | {metrics['refund_orders']} | {metrics['visitors']} | {metrics['cart_users']} |",
+            "",
+            f"_生成时间: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}_",
+        ]
+    )
+
+
+def write_platform_summary(platform: str, target_date: str, product_ids: list[str], item: dict) -> Path:
+    style_dir = PLATFORM_OUTPUT_ROOT / PLATFORM_DIRS[platform] / f"{STYLE_NAME}_{target_date}"
+    if platform == "tm":
+        path = style_dir / f"业务款式_{target_date}.json"
+        payload = {
+            "source_file": item["source_file"],
+            "date": target_date,
+            "generated_at": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+            "metrics": {"销量": "平台 Excel 原始销量"},
+            "style": {
+                "款式": STYLE_NAME,
+                "日期": target_date,
+                "商品ID数": len(product_ids),
+                "匹配商品ID数": len(item["matched_ids"]),
+                "匹配商品ID": item["matched_ids"],
+                "未匹配商品ID": item["missing_ids"],
+                "销量": item["metrics"]["sales"],
+                "访客量": item["metrics"]["visitors"],
+                "加购人数": item["metrics"]["cart_users"],
+                "退款订单数": item["metrics"]["refund_orders"],
+            },
+        }
+    elif platform == "jd":
+        path = style_dir / f"{STYLE_NAME}_{target_date}.json"
+        payload = {
+            "款式": STYLE_NAME,
+            "采集日期": target_date,
+            "数据周期": f"{target_date} ~ {target_date}",
+            "spu列表": product_ids,
+            "匹配SPU": item["matched_ids"],
+            "未匹配SPU": item["missing_ids"],
+            "指标汇总": {
+                "销量": item["metrics"]["sales"],
+                "访客数": item["metrics"]["visitors"],
+                "加购人数": item["metrics"]["cart_users"],
+                "退款订单数": item["metrics"]["refund_orders"],
+            },
+            "download_file": item["source_file"],
+            "matched_row_count": item["matched_row_count"],
+        }
+    else:
+        path = style_dir / f"{STYLE_NAME}_{target_date}.json"
+        payload = {
+            "style_name": STYLE_NAME,
+            "date_range": target_date,
+            "product_ids": product_ids,
+            "matched_product_ids": item["matched_ids"],
+            "missing_product_ids": item["missing_ids"],
+            "matched_row_count": item["matched_row_count"],
+            "download_file": item["source_file"],
+            "销量": item["metrics"]["sales"],
+            "访客数": item["metrics"]["visitors"],
+            "加购人数": item["metrics"]["cart_users"],
+            "raw_totals": {
+                "销量": item["metrics"]["sales"],
+                "访客数": item["metrics"]["visitors"],
+                "加购人数": item["metrics"]["cart_users"],
+                "退款订单数": item["metrics"]["refund_orders"],
+            },
+            "platform_raw_totals": {
+                "销量": item["metrics"]["sales"],
+                "访客数": item["metrics"]["visitors"],
+                "加购人数": item["metrics"]["cart_users"],
+                "退款订单数": item["metrics"]["refund_orders"],
+            },
+        }
+    _write_json(path, payload)
+    path.with_suffix(".md").write_text(
+        _render_platform_markdown(platform, target_date, product_ids, item),
+        encoding="utf-8-sig",
+    )
+    return path
+
+
+def render_final_markdown(payload: dict) -> str:
+    erp_codes = (payload.get("erp") or {}).get("style_codes") or []
+    lines = [
+        f"# {payload['style_name']} - 每日最终汇总 - {payload['date']}",
+        "",
+        f"- 日期: {payload['date']}",
+        f"- 款式: {payload['style_name']}",
+        f"- ERP款式编码: {', '.join(erp_codes) or '-'}",
+        "- 数据口径: 三平台历史 Excel 原始指标",
+        "",
+        "| 平台 | 销量 | 退货量 | 访客数 | 加购人数 |",
+        "|---|---:|---:|---:|---:|",
+    ]
+    for platform in ("jd", "dy", "tm"):
+        block = payload[platform]
+        lines.append(
+            f"| {platform} | {block['sales']} | {block['refund_orders']} | "
+            f"{block['visitors']} | {block['cart_users']} |"
+        )
+    lines.extend(["", f"_生成时间: {payload['generated_at']}_"])
+    return "\n".join(lines)
+
+
+def write_final_summary(payload: dict) -> Path:
+    path = FINAL_OUTPUT_ROOT / payload["date"] / STYLE_NAME / "final_summary.json"
+    _write_json(path, payload)
+    path.with_suffix(".md").write_text(render_final_markdown(payload), encoding="utf-8-sig")
+    return path
+
+
+def build_db_rows(payloads: list[dict]) -> list[dict]:
+    rows = []
+    for payload in payloads:
+        erp_codes = (payload.get("erp") or {}).get("style_codes") or []
+        for platform in ("jd", "dy", "tm"):
+            block = payload[platform]
+            rows.append(
+                {
+                    "metric_date": payload["date"],
+                    "style_name": payload["style_name"],
+                    "platform": platform,
+                    "sales": int(block["sales"]),
+                    "refund_orders": int(block["refund_orders"]),
+                    "visitors": int(block["visitors"]),
+                    "cart_users": int(block["cart_users"]),
+                    "erp_style_codes": erp_codes,
+                    "source": block["source"],
+                }
+            )
+    return rows
+
+
+def sync_database(payloads: list[dict]) -> int:
+    from db import get_conn, upsert_daily_metrics
+
+    rows = build_db_rows(payloads)
+    with get_conn() as conn:
+        return upsert_daily_metrics(conn, rows)
+
+
+def build_audit_row(target_date: str, parsed: dict[str, dict]) -> dict:
+    row: dict[str, Any] = {"date": target_date}
+    for platform in ("jd", "dy", "tm"):
+        item = parsed[platform]
+        row[platform] = {
+            **item["metrics"],
+            "matched_ids": item["matched_ids"],
+            "missing_ids": item["missing_ids"],
+            "source_file": item["source_file"],
+        }
+    return row
+
+
+def insert_records(record_file: Path, end_date: str) -> None:
+    command = [
+        sys.executable,
+        str(runtime_script("insert_bitable_records.py")),
+        "--date",
+        end_date,
+        "--record-file",
+        str(record_file),
+        "--style",
+        STYLE_NAME,
+    ]
+    print(f"[INSERT] records={record_file}", flush=True)
+    subprocess.run(command, cwd=PROJECT_ROOT, check=True)
+
+
+def main() -> int:
+    args = parse_args()
+    target_dates = iter_dates(args.start, args.end)
+    config = StyleConfigLoader().load(force_refresh=args.force_refresh_config)
+    style = (config.get("styles") or {}).get(STYLE_NAME)
+    if not style:
+        raise RuntimeError(f"飞书配置不存在款式: {STYLE_NAME}")
+    ids_by_platform = {
+        "jd": [str(value) for value in style["jd"]["spus"]],
+        "dy": [str(value) for value in style["dy"]["item_ids"]],
+        "tm": [str(value) for value in style["tm"]["item_ids"]],
+    }
+    if any(not values for values in ids_by_platform.values()):
+        raise RuntimeError(f"三平台商品 ID 配置不完整: {ids_by_platform}")
+    erp_codes = [str(value) for value in style.get("erp_codes") or []]
+
+    audit_rows = []
+    combined_records = []
+    final_payloads = []
+    for index, target_date in enumerate(target_dates, start=1):
+        parsed = {
+            platform: choose_excel(platform, target_date, ids_by_platform[platform])
+            for platform in ("jd", "dy", "tm")
+        }
+        for platform in ("jd", "dy", "tm"):
+            write_platform_summary(platform, target_date, ids_by_platform[platform], parsed[platform])
+        final_payload = build_final_payload(target_date, parsed, erp_codes)
+        final_path = write_final_summary(final_payload)
+        final_payloads.append(final_payload)
+
+        all_records = export_bitable_records.build_records(
+            target_date,
+            summary_root=FINAL_OUTPUT_ROOT,
+            platform_root=PLATFORM_OUTPUT_ROOT,
+        )
+        export_bitable_records.write_outputs(all_records, target_date)
+        target_records = [record for record in all_records if record.get("款式") == STYLE_NAME]
+        if len(target_records) != 3:
+            raise RuntimeError(f"{target_date} 导出记录不是 3 条: {len(target_records)}")
+        combined_records.extend(target_records)
+        audit_rows.append(build_audit_row(target_date, parsed))
+        sales = "/".join(str(parsed[p]["metrics"]["sales"]) for p in ("jd", "dy", "tm"))
+        print(f"[{index:02d}/{len(target_dates)}] {target_date} JD/DY/TM销量={sales} -> {final_path}", flush=True)
+
+    report_dir = REPORT_OUTPUT_ROOT / "backfill_reports" / f"poseidon_{args.start}_{args.end}"
+    report_path = report_dir / "audit.json"
+    record_path = report_dir / "bitable_records.json"
+    report = {
+        "style_name": STYLE_NAME,
+        "start": args.start,
+        "end": args.end,
+        "generated_at": datetime.now().isoformat(timespec="seconds"),
+        "ids_by_platform": ids_by_platform,
+        "record_count": len(combined_records),
+        "rows": audit_rows,
+    }
+    _write_json(report_path, report)
+    _write_json(record_path, combined_records)
+    print(f"[OK] 审计报告: {report_path}")
+    print(f"[OK] 待写入记录: {record_path} ({len(combined_records)} 条)")
+
+    if args.skip_db:
+        print("[SKIP] 未同步 PostgreSQL (--skip-db)")
+    else:
+        db_count = sync_database(final_payloads)
+        print(f"[OK] PostgreSQL daily_style_metrics upsert: {db_count} 条")
+
+    if args.insert:
+        insert_records(record_path, args.end)
+        print(f"[OK] 飞书写入完成: {len(combined_records)} 条")
+    else:
+        print("[DRY-RUN] 未写入飞书；确认后追加 --insert")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/bitable_main_image_map.json b/src/gyxx_flow/modules/product_commerce/runtime/bitable_main_image_map.json
new file mode 100644
index 0000000..0933902
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/bitable_main_image_map.json
@@ -0,0 +1,195 @@
+{
+  "_comment": "主图/曝光/点击表 — 与 bitable_style_map.json 的销量表同 base 但不同 table。键是数据目录名 (lowercase),display_name 是给用户看的品牌名。platform 默认天猫,可在单款覆盖。",
+  "盖亚斜挎": {
+    "display_name": "盖亚斜挎",
+    "base_token": "UkbabpqRYanmD7sk7ksceVuhnwf",
+    "table_id": "tblJoRIPTq96oCtC",
+    "view_id": "vewPe0N2x",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "极星pro": {
+    "display_name": "极星Pro",
+    "base_token": "OsYvbUe1iaJYeKs5WvLcU4rVnHc",
+    "table_id": "tblFzXSDTKeVHsOY",
+    "view_id": "vew7f7EVcH",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "盖亚微单": {
+    "display_name": "盖亚微单",
+    "base_token": "Qs4mbzyAaaQ8SasnDpkc9lA5npe",
+    "table_id": "tbl3czAfAuey9eTc",
+    "view_id": "vewQHErmIa",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "盖世m1": {
+    "display_name": "盖世M1",
+    "base_token": "KnBRbH9iQaH3PAsAUk5cHLxNnMg",
+    "table_id": "tbl1cXPCZZg4xmGW",
+    "view_id": "vewlc7JnDj",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "星迹2": {
+    "display_name": "星迹&星迹2",
+    "base_token": "Ssp8bXmrwahRBbsko2BcuiLanad",
+    "table_id": "tblI4VItN9zJ2NWm",
+    "view_id": "vewGr5tM7V",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "星云2": {
+    "display_name": "星云2",
+    "base_token": "BJaWbLEOKaHONMshOidcxvsKngh",
+    "table_id": "tbl4TSY8KSR3xJGL",
+    "view_id": "vewWvUtlaT",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "宙斯": {
+    "display_name": "宙斯",
+    "base_token": "N43abi7IlaS8XosnUXec6gH2n7g",
+    "table_id": "tblsNUawGSzmB06P",
+    "view_id": "vewa4JsdEu",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "阿波罗2": {
+    "display_name": "阿波罗2",
+    "base_token": "W80vbtPDoaaHTOs5Tipc6CTLn6c",
+    "table_id": "tblg7YhED1mxPJ6W",
+    "view_id": "vewS5SPd7e",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "阿波罗x1": {
+    "display_name": "阿波罗X1",
+    "base_token": "CtW5bUcc9a5rqRsZkwucUWDVnHb",
+    "table_id": "tblZB7JHoR3Wkyol",
+    "view_id": "vewYT5gGTl",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "波塞冬edge": {
+    "display_name": "波塞冬Edge",
+    "base_token": "JsVLbso8Ba6RjbslOl1cJX8lnyb",
+    "table_id": "tblYBP0lORSKty7r",
+    "view_id": "vewXi77KZH",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "逐星": {
+    "display_name": "逐星",
+    "base_token": "IQrbbxAHpa9oiWs631GcPkHznob",
+    "table_id": "tblNCawlW3MRlsvE",
+    "view_id": "vewibnFe7G",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "逐星GT": {
+    "display_name": "逐星GT",
+    "base_token": "MAEZbKrNaabJbDs5ZY7caq1ynrd",
+    "table_id": "tbl7oARAa66atiJA",
+    "view_id": "vewRWPTJDu",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "星云mini": {
+    "display_name": "星云Mini",
+    "base_token": "RKfkb5JkIa9FttsgHvNc2nJYnbf",
+    "table_id": "tbl0g7G9gsFr7xnt",
+    "view_id": "vewlRaaxOo",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "晨星2": {
+    "display_name": "晨星2",
+    "base_token": "VrLtbkldsaggOms3F4Bcu3DinCg",
+    "table_id": "tblV6eSZik3sGRjV",
+    "view_id": "vewF4mmrC6",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "极星双肩": {
+    "display_name": "极星双肩",
+    "base_token": "GubZbTrMSahLg5sdonIc8kR8nZe",
+    "table_id": "tblDSwCPStXoPj4a",
+    "view_id": "vew4DqI9SY",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "凌云air": {
+    "display_name": "凌云Air",
+    "base_token": "VwpibmlakaqHuLsGChccMb3BnNd",
+    "table_id": "tbl0sEhjIdRX2Ifg",
+    "view_id": "vew8PYjRfl",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "觅光": {
+    "display_name": "觅光",
+    "base_token": "CQcibpRjgaUCg7sUTMwcnrhunMh",
+    "table_id": "tblyZV8DGIedjSc0",
+    "view_id": "vewkg4Zf94",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "凌云3": {
+    "display_name": "凌云3",
+    "base_token": "XtHjbqpwcab3J4sWP6Fcfgh5nYt",
+    "table_id": "tbl5EqD1AbU8hjLE",
+    "view_id": "vewsT4SRkP",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "瑞白": {
+    "display_name": "瑞白",
+    "base_token": "GNOkbKNpna8pDJsLMfzcuZt3ncf",
+    "table_id": "tblbHkUuK95t9B3T",
+    "view_id": "vewEpzCvC2",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "黑曜石吐司包": {
+    "display_name": "黑曜石吐司包",
+    "base_token": "U2JGbbWcIal6WvsxURRcTMwEnhf",
+    "table_id": "tblSOUXQL4nhz2Y5",
+    "view_id": "vew9USsO3K",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "黑曜石托特包": {
+    "display_name": "黑曜石托特包",
+    "base_token": "U9S8bLYsNa8Ul8sPffAcrJ02nHI",
+    "table_id": "tblae0t1lfcPrtK6",
+    "view_id": "vewyl2eOzN",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "曜影": {
+    "display_name": "曜影",
+    "base_token": "UPk3bR9MOaTkgrsG1qLc7IUvnQe",
+    "table_id": "tblYgdR3IgTvTspB",
+    "view_id": "vewdgTtWoO",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "黑曜石电脑包": {
+    "display_name": "黑曜石电脑包",
+    "base_token": "PwrGbXcpRaxJnOsd9cacFou8nAf",
+    "table_id": "tblPIHM3SORWYGzM",
+    "view_id": "vew2X87RUb",
+    "platform": "天猫",
+    "enabled": true
+  },
+  "拓界2": {
+    "display_name": "拓界2",
+    "base_token": "CNgPbjuXiaPoxPsShROcYbRDnbb",
+    "table_id": "tblofU5j1BlqmXd1",
+    "view_id": "vewkwocbDP",
+    "platform": "天猫",
+    "enabled": true
+  }
+}
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/bitable_style_map.json b/src/gyxx_flow/modules/product_commerce/runtime/bitable_style_map.json
new file mode 100644
index 0000000..6ab11d6
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/bitable_style_map.json
@@ -0,0 +1,597 @@
+{
+  "盖亚斜挎": {
+    "base_token": "UkbabpqRYanmD7sk7ksceVuhnwf",
+    "table_id": "tblrKElXywgE18ta",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "极星pro": {
+    "base_token": "OsYvbUe1iaJYeKs5WvLcU4rVnHc",
+    "table_id": "tblx8wC2KNOByknq",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "盖亚微单": {
+    "base_token": "Qs4mbzyAaaQ8SasnDpkc9lA5npe",
+    "table_id": "tbltcJbLyiEcxQZJ",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "盖世m1": {
+    "base_token": "KnBRbH9iQaH3PAsAUk5cHLxNnMg",
+    "table_id": "tblfQBL0TNhZkfgk",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "星迹2": {
+    "base_token": "Ssp8bXmrwahRBbsko2BcuiLanad",
+    "table_id": "tblWz56o40TF8Rqj",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "星云2": {
+    "base_token": "BJaWbLEOKaHONMshOidcxvsKngh",
+    "table_id": "tbla91Pfbb5T78AN",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "宙斯": {
+    "base_token": "N43abi7IlaS8XosnUXec6gH2n7g",
+    "table_id": "tblwYxTG3gAqXaKx",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "阿波罗2": {
+    "base_token": "W80vbtPDoaaHTOs5Tipc6CTLn6c",
+    "table_id": "tblz8Adhwr71hMv2",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "阿波罗x1": {
+    "base_token": "CtW5bUcc9a5rqRsZkwucUWDVnHb",
+    "table_id": "tblLq5ROaFQodu25",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "波塞冬edge": {
+    "base_token": "JsVLbso8Ba6RjbslOl1cJX8lnyb",
+    "table_id": "tblzcYqrBg7VznO0",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "逐星": {
+    "base_token": "IQrbbxAHpa9oiWs631GcPkHznob",
+    "table_id": "tblfXykMGS52Cymo",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "逐星GT": {
+    "base_token": "MAEZbKrNaabJbDs5ZY7caq1ynrd",
+    "table_id": "tblVb4a0OUk26tSR",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "星云mini": {
+    "base_token": "RKfkb5JkIa9FttsgHvNc2nJYnbf",
+    "table_id": "tbl9JZhmKzioqhLy",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "晨星2": {
+    "base_token": "VrLtbkldsaggOms3F4Bcu3DinCg",
+    "table_id": "tblsBVuBxAQRv2DR",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "极星双肩": {
+    "base_token": "GubZbTrMSahLg5sdonIc8kR8nZe",
+    "table_id": "tblp7yTxlgnsvddO",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "凌云air": {
+    "base_token": "VwpibmlakaqHuLsGChccMb3BnNd",
+    "table_id": "tblMpxWIz3amDyD6",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "觅光": {
+    "base_token": "CQcibpRjgaUCg7sUTMwcnrhunMh",
+    "table_id": "tbl1TMMsYYyr9ivY",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "凌云3": {
+    "base_token": "XtHjbqpwcab3J4sWP6Fcfgh5nYt",
+    "table_id": "tblk37EfdDrVCu1r",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "瑞白": {
+    "base_token": "GNOkbKNpna8pDJsLMfzcuZt3ncf",
+    "table_id": "tbl0tXeoW4ZlYDGu",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "黑曜石吐司包": {
+    "base_token": "U2JGbbWcIal6WvsxURRcTMwEnhf",
+    "table_id": "tblaRCgW5jniY3pv",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "黑曜石托特包": {
+    "base_token": "U9S8bLYsNa8Ul8sPffAcrJ02nHI",
+    "table_id": "tblg0vofQilmXUiu",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "曜影": {
+    "base_token": "UPk3bR9MOaTkgrsG1qLc7IUvnQe",
+    "table_id": "tblFE6x5g2A2Eyd0",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "黑曜石电脑包": {
+    "base_token": "PwrGbXcpRaxJnOsd9cacFou8nAf",
+    "table_id": "tbl0yBLBVo8HMGx9",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "拓界2": {
+    "base_token": "CNgPbjuXiaPoxPsShROcYbRDnbb",
+    "table_id": "tblbj4E6PAtej4jG",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "布谷": {
+    "base_token": "KU3Vbw5qDay8kUsSWibcl0cinAc",
+    "table_id": "tblMjsKLKa6iPRR3",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "白鹭": {
+    "base_token": "UhuObUyaFaU4Djs55FIcA5RYnVc",
+    "table_id": "tbl8fMI03EpEEEur",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "蓝鹊": {
+    "base_token": "NoM1bh0zZap7wwsQ69ScydGmnNd",
+    "table_id": "tbleOD0u4WS6igEg",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "盖亚双肩3": {
+    "base_token": "XsbmbDyp9adqawsHInacBID1nTc",
+    "table_id": "tblWLfx1mja0JyIU",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "极星2": {
+    "base_token": "LvLBblI5KaekeUsaPRpcDeSBnxb",
+    "table_id": "tblDHCPURC5P2VX0",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "拾影斜挎相机包": {
+    "base_token": "WGVebBy7waEcffsnKFXcuIrwnNf",
+    "table_id": "tblBu67uhiN22jbI",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "拾影双肩": {
+    "base_token": "JAGEbWz1aarBf7sX96zc8S0VnXg",
+    "table_id": "tbl9l9ZfBi47wOfQ",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "云栖": {
+    "base_token": "HwrSbl4BJaCxHGsja07cxvWlnGd",
+    "table_id": "tblQhnv04ONow7O8",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "波塞冬手提电脑包": {
+    "base_token": "Aiazbsa80auaVGsktHEcembEnzc",
+    "table_id": "tblgdlUCQg5wQqkG",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "极星双肩2": {
+    "base_token": "UBM9bg9WnaU0Jvs3MYDcC8s6nYe",
+    "table_id": "tbl5j0lrF8spF2jG",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  },
+  "极星单肩": {
+    "base_token": "MVxPb5RFEa9YEPsDmEMc6ujinoh",
+    "table_id": "tbl7sJP5eUSZPuEo",
+    "view_id": "vewGbFVRhD",
+    "enabled": true,
+    "field_map": {
+      "时间": "时间周期",
+      "平台": "平台",
+      "销量": "销量",
+      "访客数": "访客数",
+      "转化率": "转化率",
+      "加购率": "加购率",
+      "退款率": "退款率",
+      "产品": "款式",
+      "SourceID": "SourceID"
+    }
+  }
+}
\ No newline at end of file
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/check_nine_day_decline.py b/src/gyxx_flow/modules/product_commerce/runtime/check_nine_day_decline.py
new file mode 100644
index 0000000..e0f5326
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/check_nine_day_decline.py
@@ -0,0 +1,458 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""9 天连续下滑检测 + Hermes analyzer 通知。
+
+窗口定义（end_date 默认 = 昨天）：
+    段1: anchor_0+1 ~ anchor_1   (3 天)
+    段2: anchor_1+1 ~ anchor_2   (3 天)
+    段3: anchor_2+1 ~ anchor_3   (3 天)
+    anchor_0 = end_date - 8
+    anchor_1 = end_date - 5
+    anchor_2 = end_date - 2
+    anchor_3 = end_date
+
+数据源（按 collect_pairs_for_date 实际行为）：
+    1) 优先 PG daily_style_metrics（每个锚点日销量 = 1 行 sales 字段）
+    2) PG 不可达 / 查不到时回退到 data/<plat>/<款>_<日期>/data.json 扫盘
+    段销量 = 3 个锚点日销量相减（不再是 ERP 5/1 累计的差分，是 PG 日值本身）
+
+下滑定义：段1 > 段2 > 段3 > 0（严格连续下降，且末段非 0；按需求"3 一直为 0 不算下滑"）。
+粒度：单 (款式, 平台) 组合下滑即报。无销量门槛。
+
+命中后 → POST Hermes analyzer 端 chat completions，由 analyzer profile 自行处理通知。
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+from gyxx_flow.adapters import RuntimeServicePolicy
+import re
+import sys
+import urllib.error
+import urllib.request
+from datetime import date, datetime, timedelta
+from pathlib import Path
+
+from runtime_paths import LOG_ROOT, RAW_DATA_ROOT, managed_data_path
+from typing import Any
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+DATA_ROOT = RAW_DATA_ROOT
+LOG_PATH = LOG_ROOT / "nine_day_decline.log"
+
+PLATFORM_DIRS = {
+    "jd": "jd单品数据",
+    "dy": "dy单品数据",
+    "tm": "tm单品数据",
+}
+
+# 与 vendors/jd-data-flow/config.py 的 HERMES analyzer 配置保持一致
+HERMES_API_KEY = os.getenv("HERMES_API_KEY")
+if not HERMES_API_KEY:
+    raise RuntimeError("HERMES_API_KEY is required")
+os.environ.update(RuntimeServicePolicy().apply(os.environ))
+HERMES_ANALYZER_URL = os.getenv(
+    "ANALYZER_API_SERVER_URL",
+    "http://localhost:8642/v1/chat/completions",
+)
+HERMES_MODEL = "mimo-v2.5-pro"
+
+
+# ── helpers ────────────────────────────────────────────────────────────────
+
+def parse_int(value: Any) -> int:
+    if value is None:
+        return 0
+    if isinstance(value, (int, float)):
+        return int(value)
+    text = str(value).replace(",", "").strip()
+    if not text:
+        return 0
+    try:
+        return int(float(text))
+    except ValueError:
+        return 0
+
+
+def load_cumulative(platform: str, style_name: str, on_date: str) -> int | None:
+    """读 <plat>/<safe_name>_<on_date>/data.json，匹配 (platform, style_name) 拿区间累计销量。"""
+    base = DATA_ROOT / PLATFORM_DIRS[platform]
+    if not base.exists():
+        return None
+    candidates = [d for d in base.iterdir() if d.is_dir() and d.name.endswith(f"_{on_date}")]
+    for style_dir in candidates:
+        json_path = style_dir / "data.json"
+        if not json_path.exists():
+            continue
+        try:
+            payload = json.loads(json_path.read_text(encoding="utf-8-sig"))
+        except Exception:
+            continue
+        if payload.get("platform") != platform:
+            continue
+        if payload.get("style_name") != style_name:
+            continue
+        if not (payload.get("code_results") or payload.get("yesterday_sales") is not None):
+            continue
+        # 全部 ERP 查询失败时跳过（避免 0 覆盖真实值）
+        code_results = payload.get("code_results") or []
+        ok_results = [c for c in code_results if c.get("ok")]
+        if code_results and not ok_results:
+            return None
+        if ok_results:
+            sales = sum(parse_int(c.get("yesterday_sales")) for c in ok_results)
+        else:
+            sales = parse_int(payload.get("yesterday_sales"))
+        return sales
+    return None
+
+
+def collect_pairs_for_date(on_date: str) -> dict[tuple[str, str], dict]:
+    """扫 on_date 那天所有 (platform, style_name) 维度，返回 { (plat, style): {sales, codes, ...} }。
+
+    优先从 PG 读 (daily_style_metrics)；库不可达或无数据时回退到 JSON 扫盘。
+    """
+    # 1) 优先读库
+    try:
+        from db import fetch_daily_pairs
+        with __import__("db").get_conn() as _conn:
+            pairs_db = fetch_daily_pairs(_conn, on_date)
+        if pairs_db:
+            return {
+                k: {"sales": v["sales"], "erp_style_codes": v["erp_style_codes"]}
+                for k, v in pairs_db.items()
+            }
+    except Exception as exc:
+        print(f"  [WARN] PG 读取失败，回退到 JSON 扫盘: {exc}")
+
+    # 2) 回退到 JSON
+    pairs: dict[tuple[str, str], dict] = {}
+    for plat, dir_name in PLATFORM_DIRS.items():
+        base = DATA_ROOT / dir_name
+        if not base.exists():
+            continue
+        for style_dir in base.iterdir():
+            if not style_dir.is_dir() or not style_dir.name.endswith(f"_{on_date}"):
+                continue
+            json_path = style_dir / "data.json"
+            if not json_path.exists():
+                continue
+            try:
+                payload = json.loads(json_path.read_text(encoding="utf-8-sig"))
+            except Exception:
+                continue
+            if payload.get("platform") != plat:
+                continue
+            style_name = payload.get("style_name")
+            if not style_name:
+                continue
+            # 全失败跳过
+            code_results = payload.get("code_results") or []
+            ok_results = [c for c in code_results if c.get("ok")]
+            if code_results and not ok_results:
+                continue
+            if ok_results:
+                sales = sum(parse_int(c.get("yesterday_sales")) for c in ok_results)
+            else:
+                sales = parse_int(payload.get("yesterday_sales"))
+            pairs[(plat, style_name)] = {
+                "sales": sales,
+                "erp_style_codes": payload.get("erp_style_codes") or [],
+            }
+    return pairs
+
+
+def pct_drop(prev: int, curr: int) -> float:
+    if prev <= 0:
+        return 0.0
+    return (prev - curr) / prev * 100
+
+
+# ── core ───────────────────────────────────────────────────────────────────
+
+def detect_decline(end_date: str, window_days: int = 9) -> list[dict[str, Any]]:
+    """滑动检测 (款式, 平台) 连续下滑事件。"""
+    if window_days % 3 != 0:
+        raise ValueError(f"window_days 必须是 3 的倍数（当前 {window_days}）")
+    third = window_days // 3
+    end = datetime.strptime(end_date, "%Y-%m-%d").date()
+    a0 = end - timedelta(days=window_days - 1)
+    a1 = a0 + timedelta(days=third)
+    a2 = a1 + timedelta(days=third)
+    a3 = end
+    anchors = [a0.isoformat(), a1.isoformat(), a2.isoformat(), a3.isoformat()]
+
+    # 4 个锚点各扫一次，得到 (plat, style) -> {date: sales}
+    snapshots: list[dict[tuple[str, str], dict]] = [collect_pairs_for_date(d) for d in anchors]
+    union_keys = set().union(*[s.keys() for s in snapshots])
+
+    events: list[dict[str, Any]] = []
+    for key in sorted(union_keys):
+        plat, style = key
+        # 4 个锚点各自的日销量（PG daily_style_metrics 已经是日值，不是累计）
+        daily_sales = []
+        for snap in snapshots:
+            info = snap.get(key)
+            daily_sales.append(info["sales"] if info else None)
+        if any(c is None for c in daily_sales):
+            continue
+        # 3 段销量 = 锚点日销量差分（每日 sales 已是 PG 当日销量，直接 SUM 3 天 = 段销量）
+        seg1 = daily_sales[1] - daily_sales[0]
+        seg2 = daily_sales[2] - daily_sales[1]
+        seg3 = daily_sales[3] - daily_sales[2]
+        # 严格连续下滑：段1 > 段2 > 段3 > 0
+        if seg1 <= 0 or seg2 <= 0 or seg3 <= 0:
+            continue
+        if not (seg1 > seg2 > seg3):
+            continue
+        codes = snapshots[3].get(key, {}).get("erp_style_codes") or []
+        events.append({
+            "style_name": style,
+            "platform": plat,
+            "erp_style_codes": codes,
+            "window_start": (a0 + timedelta(days=1)).isoformat(),
+            "window_end": a3.isoformat(),
+            "seg1_sales": seg1,
+            "seg2_sales": seg2,
+            "seg3_sales": seg3,
+            "drop_pct_1to2": round(pct_drop(seg1, seg2), 1),
+            "drop_pct_2to3": round(pct_drop(seg2, seg3), 1),
+            "drop_pct_1to3": round(pct_drop(seg1, seg3), 1),
+        })
+    return events
+
+
+# ── Hermes notification ────────────────────────────────────────────────────
+
+def notify_hermes(events: list[dict[str, Any]], end_date: str, window_days: int,
+                  openid: str | None = None) -> dict:
+    """POST Hermes analyzer 端 chat completions。返回原始响应 dict。"""
+    if not events:
+        return {"skipped": True, "reason": "no events"}
+    openid_line = f"通知对象 openid: {openid}\n" if openid else ""
+    payload = {
+        "model": HERMES_MODEL,
+        "messages": [
+            {
+                "role": "system",
+                "content": (
+                    "你是电商销量异常监控助手。"
+                    "下面是一组 (款式, 平台) 维度的连续下滑事件（每组 3 段 × 3 天，段1 > 段2 > 段3 > 0）。"
+                    "请输出：1) 每条事件一句中文告警（款式 / 平台 / 3 段销量 / 跌幅）；"
+                    "2) 末尾给一段简短的处置建议（如：核对 ERP 数据、查看广告投放、排查客诉等）；"
+                    "3) 飞书推送时 @ 通知对象 openid。"
+                ),
+            },
+            {
+                "role": "user",
+                "content": (
+                    f"窗口: {end_date} 向前回溯 {window_days} 天\n"
+                    f"{openid_line}"
+                    f"命中事件数: {len(events)}\n"
+                    f"事件 JSON:\n{json.dumps(events, ensure_ascii=False, indent=2)}"
+                ),
+            },
+        ],
+    }
+    body = json.dumps(payload, ensure_ascii=False).encode("utf-8")
+    req = urllib.request.Request(
+        HERMES_ANALYZER_URL,
+        data=body,
+        method="POST",
+        headers={
+            "Authorization": f"Bearer {HERMES_API_KEY}",
+            "Content-Type": "application/json",
+        },
+    )
+    try:
+        with urllib.request.urlopen(req, timeout=120) as resp:
+            raw = resp.read().decode("utf-8", errors="replace")
+            return json.loads(raw)
+    except urllib.error.HTTPError as exc:
+        return {"error": f"http {exc.code}", "body": exc.read().decode("utf-8", errors="replace")}
+    except urllib.error.URLError as exc:
+        return {"error": f"url error: {exc.reason}"}
+
+
+# ── logging ────────────────────────────────────────────────────────────────
+
+def append_log(line: str) -> None:
+    LOG_PATH.parent.mkdir(parents=True, exist_ok=True)
+    ts = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    with LOG_PATH.open("a", encoding="utf-8-sig") as fh:
+        fh.write(f"[{ts}] {line}\n")
+
+
+# ── CLI ────────────────────────────────────────────────────────────────────
+
+def parse_args() -> argparse.Namespace:
+    p = argparse.ArgumentParser(description="9 天连续下滑检测 (PG daily_style_metrics 销量) + Hermes analyzer 通知")
+    p.add_argument("--end-date", default=None, help="窗口右端 YYYY-MM-DD，默认昨天")
+    p.add_argument("--window-days", type=int, default=9, help="窗口天数（3 的倍数，默认 9）")
+    p.add_argument("--dry-run", action="store_true", help="不 POST Hermes，只打印事件")
+    p.add_argument("--data-root", default=str(DATA_ROOT))
+    p.add_argument("--openid", default=None,
+                   help="Hermes 通知对象飞书 openid（写入 prompt 头部，analyzer 据此 @ 用户）")
+    return p.parse_args()
+
+
+def main() -> int:
+    args = parse_args()
+    if args.window_days % 3 != 0:
+        print(f"[ERR] window_days 必须是 3 的倍数（当前 {args.window_days}）", file=sys.stderr)
+        return 2
+    end_date = args.end_date or (date.today() - timedelta(days=1)).isoformat()
+    global DATA_ROOT
+    DATA_ROOT = managed_data_path(args.data_root)
+
+    print(f"[INFO] end_date={end_date} window_days={args.window_days} dry_run={args.dry_run}")
+    try:
+        events = detect_decline(end_date, args.window_days)
+    except Exception as exc:
+        msg = f"detect failed: {exc}"
+        print(f"[ERR] {msg}")
+        append_log(msg)
+        return 1
+
+    if not events:
+        print(f"[OK] 0 个下滑事件 (end_date={end_date}, window={args.window_days} 天)")
+        append_log(f"end_date={end_date} window={args.window_days} -> 0 events")
+        return 0
+
+    print(f"[HIT] {len(events)} 个连续下滑事件:")
+    for ev in events:
+        print(
+            f"  - [{ev['platform']}] {ev['style_name']} "
+            f"seg1={ev['seg1_sales']} -> seg2={ev['seg2_sales']} -> seg3={ev['seg3_sales']} "
+            f"(-{ev['drop_pct_1to2']}% / -{ev['drop_pct_2to3']}%)"
+        )
+
+    # 重跑去重: 过滤掉 fact_alert_history 里已通知过的事件, 避免重复 @ 用户
+    new_events = [ev for ev in events if not _already_notified(ev["style_name"], end_date, args.openid)]
+    skipped_count = len(events) - len(new_events)
+    if skipped_count > 0:
+        print(f"[SKIP] {skipped_count} 个事件已通知过 (fact_alert_history 已存在), 跳过")
+
+    if not new_events:
+        print(f"[SKIP] 全部 {len(events)} 个事件已通知, 不 POST Hermes")
+        append_log(
+            f"end_date={end_date} window={args.window_days} -> {len(events)} events all already notified"
+        )
+        return 0
+
+    if args.dry_run:
+        print(f"[DRY] 跳过 Hermes POST ({len(new_events)} 个新事件)")
+        append_log(
+            f"end_date={end_date} window={args.window_days} -> {len(new_events)} new events (DRY-RUN)"
+        )
+        return 0
+
+    print(f"[POST] -> {HERMES_ANALYZER_URL} ({len(new_events)} 个新事件)")
+    resp = notify_hermes(new_events, end_date, args.window_days, openid=args.openid)
+    if "error" in resp:
+        print(f"[ERR] Hermes: {resp['error']}")
+        append_log(f"hermes error: {resp['error']}")
+        # 仍然把事件落库 (hermes_response 字段存错误信息)
+        _persist_events(new_events, end_date, args.window_days, hermes_response=resp, openid=args.openid)
+        return 3
+    if resp.get("skipped"):
+        print("[SKIP] Hermes: no events to send")
+        return 0
+    content = (
+        resp.get("choices", [{}])[0].get("message", {}).get("content", "")
+        if isinstance(resp, dict)
+        else ""
+    )
+    print("[LLM] " + (content[:500] + "..." if len(content) > 500 else content))
+    append_log(
+        f"end_date={end_date} window={args.window_days} -> {len(new_events)} new events; "
+        f"llm_len={len(content)} preview={content[:200]!r}"
+    )
+    # 命中事件入库
+    _persist_events(new_events, end_date, args.window_days, hermes_response=resp, openid=args.openid)
+    return 0
+
+
+def _persist_events(events: list[dict], end_date: str, window_days: int,
+                    hermes_response: Any = None, openid: str | None = None) -> None:
+    """把命中事件 upsert 到 decline_events + 同步插入 fact_alert_history（告警历史）。
+
+    decline_events  PK = (end_date, style_name, platform)  ← 同款不同平台分别入库
+    fact_alert_history UNIQUE (style_name, trigger_date)   ← 同款同日只发 1 次告警
+    """
+    try:
+        from db import get_conn, upsert_decline_event, insert_fact_alert_history
+    except Exception as exc:
+        print(f"[WARN] db import 失败: {exc}")
+        return
+
+    hermes_ok = (
+        hermes_response is not None
+        and not (isinstance(hermes_response, dict) and "error" in hermes_response)
+    )
+    hermes_id = None
+    if hermes_ok and isinstance(hermes_response, dict):
+        hermes_id = hermes_response.get("id") or hermes_response.get("message_id")
+
+    try:
+        with get_conn() as conn:
+            for ev in events:
+                ev_with_meta = {**ev, "end_date": end_date, "window_days": window_days}
+                upsert_decline_event(conn, ev_with_meta, hermes_response=hermes_response)
+
+                # baseline = 上一段同长度窗口 (前 9 天) 作为对比基准
+                ws_date = datetime.strptime(ev["window_start"], "%Y-%m-%d").date()
+                baseline_end = (ws_date - timedelta(days=1)).isoformat()
+                baseline_start = (ws_date - timedelta(days=window_days)).isoformat()
+
+                alert_record = {
+                    "decline_event_key": f"{end_date}|{ev['style_name']}|{ev['platform']}",
+                    "style_name": ev["style_name"],
+                    "trigger_date": end_date,
+                    "consecutive_days": window_days,
+                    "sales_drop_pct": ev.get("drop_pct_1to3"),
+                    "decline_window_start": ev["window_start"],
+                    "decline_window_end": end_date,
+                    "baseline_window_start": baseline_start,
+                    "baseline_window_end": baseline_end,
+                    "sent_at": datetime.now(),
+                    "recipient": openid,
+                    "send_channel": "lark_im" if openid else "lark_im_group",
+                    "hermes_message_id": hermes_id,
+                    "hermes_response": hermes_response if hermes_ok else None,
+                    "decline_payload": ev_with_meta,
+                }
+                insert_fact_alert_history(conn, alert_record)
+        print(f"[PG] decline_events + fact_alert_history: {len(events)} 行")
+    except Exception as exc:
+        print(f"[WARN] PG persist 失败: {exc}")
+
+
+def _already_notified(style_name: str, trigger_date: str, recipient: str | None) -> bool:
+    """检查 (style_name, trigger_date, recipient) 是否已在 fact_alert_history 里。
+
+    用途: 重跑 decline 时跳过已通知的事件, 避免重复 POST Hermes → 重复 @ 用户。
+    失败语义: DB 不可达时返回 False (视为未通知), 不因检查器故障而漏发告警。
+    """
+    try:
+        from db import get_conn
+        with get_conn() as conn:
+            cur = conn.cursor()
+            cur.execute(
+                """SELECT 1 FROM fact_alert_history
+                   WHERE style_name=%s
+                     AND trigger_date=%s
+                     AND recipient IS NOT DISTINCT FROM %s
+                   LIMIT 1""",
+                (style_name, trigger_date, recipient),
+            )
+            return cur.fetchone() is not None
+    except Exception as exc:
+        print(f"[WARN] _already_notified DB 失败（视为未通知）: {exc}")
+        return False
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/collect_dy_market_rank.py b/src/gyxx_flow/modules/product_commerce/runtime/collect_dy_market_rank.py
new file mode 100644
index 0000000..98c5ce7
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/collect_dy_market_rank.py
@@ -0,0 +1,949 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""采集抖音电商罗盘市场商品榜单，并生成飞书文档表格。"""
+
+from __future__ import annotations
+
+import argparse
+import html as html_std
+import json
+import os
+import re
+import subprocess
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT, STATE_ROOT, vendor_source_root
+from typing import Any
+
+from lxml import html
+from scrapling.fetchers import StealthySession
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+DY_SOURCE_ROOT = vendor_source_root("dy")
+if str(DY_SOURCE_ROOT) not in sys.path:
+    sys.path.append(str(DY_SOURCE_ROOT))
+
+from dy_store_competitor_store_scraping import (  # noqa: E402
+    USER_DATA_DIR,
+    _check_login_status,
+    _ensure_login,
+    _load_cookies,
+    _save_browser_state,
+)
+from collect_sycm_market_rank import (  # noqa: E402
+    _lark_cli,
+    append_feishu_doc,
+    create_feishu_doc,
+    group_products_by_category,
+    verify_feishu_category_tables,
+)
+from feishu_doc_native import append_native_category_tables  # noqa: E402
+from market_rank_report_archive import archive_market_rank_report  # noqa: E402
+
+
+MARKET_PRODUCT_RANK_URL = "https://compass.jinritemai.com/shop/chance/rank-product"
+OUTPUT_DIR = RAW_DATA_ROOT / "dy_market_rank"
+CHECKPOINT_PATH = STATE_ROOT / "checkpoints" / "dy_market_rank.json"
+DOCUMENT_IMAGE_WIDTH = 160
+DOCUMENT_IMAGE_HEIGHT = 160
+EXCLUDED_TITLE_KEYWORDS = (
+    "儿童",
+    "学生",
+    "青少年",
+    "少年",
+    "老人",
+    "老年",
+    "小学",
+    "中学",
+    "初中",
+    "高中",
+    "幼儿",
+    "幼童",
+    "少儿",
+    "年级",
+    "书包",
+    "童包",
+    "童款",
+    "男童",
+    "女童",
+)
+
+# 每个逻辑品类都独立保存类目路径、价格区间和可选搜索词。
+# 手机包优先使用业务指定的“耳机包”；部分账号菜单显示“手机包”，因此保留回退路径。
+TARGET_DY_MARKET_CATEGORIES = (
+    {
+        "name": "运动包",
+        "paths": (("鞋靴箱包", "箱包", "功能箱包", "功能包"),),
+        "min_price": 200,
+        "max_price": 1000,
+        "search_keyword": "",
+    },
+    {
+        "name": "旅行箱",
+        "paths": (("鞋靴箱包", "箱包", "功能箱包", "旅行箱/拉杆箱"),),
+        "min_price": 400,
+        "max_price": 2000,
+        "search_keyword": "",
+    },
+    {
+        "name": "托特包",
+        "paths": (("鞋靴箱包", "箱包", "时尚箱包", "全部"),),
+        "min_price": 200,
+        "max_price": 3000,
+        "search_keyword": "托特包",
+    },
+    {
+        "name": "双肩包",
+        "paths": (("鞋靴箱包", "箱包", "时尚箱包", "双肩包"),),
+        "min_price": 200,
+        "max_price": 2000,
+        "search_keyword": "",
+    },
+    {
+        "name": "斜挎包",
+        "paths": (("鞋靴箱包", "箱包", "时尚箱包", "单肩包/斜挎包"),),
+        "min_price": 200,
+        "max_price": 2000,
+        "search_keyword": "",
+    },
+    {
+        "name": "胸包",
+        "paths": (("鞋靴箱包", "箱包", "时尚箱包", "腰包/胸包"),),
+        "min_price": 200,
+        "max_price": 1000,
+        "search_keyword": "",
+    },
+    {
+        "name": "手机包",
+        "paths": (
+            ("鞋靴箱包", "箱包", "随身小包", "耳机包"),
+            ("鞋靴箱包", "箱包", "随身小包", "手机包"),
+        ),
+        "min_price": 100,
+        "max_price": 1000,
+        "search_keyword": "",
+    },
+    {
+        "name": "电脑包",
+        "paths": (("鞋靴箱包", "箱包", "功能箱包", "全部"),),
+        "min_price": 200,
+        "max_price": 1000,
+        "search_keyword": "电脑包",
+    },
+)
+
+
+def _text(node) -> str:
+    return re.sub(r"\s+", " ", " ".join(node.itertext())).strip()
+
+
+def is_excluded_product_title(title: str) -> bool:
+    normalized = re.sub(r"\s+", "", title or "")
+    return any(keyword in normalized for keyword in EXCLUDED_TITLE_KEYWORDS)
+
+
+def _compact_range(value: str) -> str:
+    value = re.sub(r"\s+", "", value or "")
+    return value.replace("～", "-").replace("—", "-").replace("–", "-")
+
+
+def _number_with_unit(value: str) -> float:
+    match = re.search(r"([\d,]+(?:\.\d+)?)\s*(万|亿)?", value or "")
+    if not match:
+        raise ValueError(f"金额没有数字: {value}")
+    number = float(match.group(1).replace(",", ""))
+    if match.group(2) == "万":
+        number *= 10_000
+    elif match.group(2) == "亿":
+        number *= 100_000_000
+    return number
+
+
+def parse_money_range_upper(value: str) -> int:
+    """把 ¥2,500万-¥5,000万 等销售额区间转成上限整数。"""
+    parts = re.split(r"\s*(?:-|~|～|—|–|至)\s*", value or "")
+    numeric_parts = [part for part in parts if re.search(r"\d", part)]
+    if not numeric_parts:
+        raise ValueError(f"销售额没有数字: {value}")
+    return int(round(_number_with_unit(numeric_parts[-1])))
+
+
+def _product_url(product_id: str) -> str:
+    if not product_id:
+        return ""
+    return f"https://haohuo.jinritemai.com/views/product/item2?id={product_id}"
+
+
+def _first_http_image(product_cell) -> str:
+    for image in product_cell.xpath(".//img"):
+        for attr in ("src", "data-src", "data-lazy-src", "data-original"):
+            value = html_std.unescape((image.get(attr) or "").strip())
+            if value.startswith("//"):
+                value = "https:" + value
+            if value.startswith(("http://", "https://")):
+                return value
+    return ""
+
+
+def _product_name(product_cell) -> str:
+    candidates = product_cell.xpath(
+        './/*[contains(concat(" ", normalize-space(@class), " "), " name- ")]'
+        '|.//*[@elementtiming="pccp_element"]'
+    )
+    for node in candidates:
+        value = _text(node)
+        if value and "价格带" not in value:
+            return value
+    value = _text(product_cell)
+    return re.split(r"价格带", value, maxsplit=1)[0].strip()
+
+
+def _price_range(product_cell) -> str:
+    value = _text(product_cell)
+    match = re.search(r"价格带\s*([¥￥][\d,.]+(?:\s*[-~～—–至]\s*[¥￥]?[\d,.]+)?)", value)
+    return _compact_range(match.group(1)) if match else ""
+
+
+def parse_market_product_rank_html(
+    source: str,
+    *,
+    category_name: str,
+    category_path: str,
+) -> list[dict[str, Any]]:
+    """按真实表头解析抖音商品榜单 HTML。"""
+    if not source.strip():
+        return []
+    root = html.fromstring(source)
+    results: list[dict[str, Any]] = []
+    for table in root.xpath("//table"):
+        header_cells = table.xpath(".//thead//th")
+        headers = [_text(cell) for cell in header_cells]
+        product_idx = next((i for i, name in enumerate(headers) if name == "商品"), None)
+        sales_idx = next(
+            (i for i, name in enumerate(headers) if "用户支付金额" in name),
+            None,
+        )
+        if product_idx is None or sales_idx is None:
+            continue
+
+        for ordinal, row in enumerate(table.xpath(".//tbody/tr"), 1):
+            if row.get("aria-hidden") == "true":
+                continue
+            cells = row.xpath("./td")
+            if max(product_idx, sales_idx) >= len(cells):
+                continue
+            row_key = (row.get("data-row-key") or "").strip()
+            key_match = re.match(r"(\d{12,})(?:_(\d+))?$", row_key)
+            product_id = key_match.group(1) if key_match else ""
+
+            # Top3 的排名用奖牌图片显示，单元格文本里的数字可能是“上升 4”；
+            # data-row-key 的后缀才是当前榜单排名，必须优先使用。
+            if key_match and key_match.group(2):
+                rank = int(key_match.group(2))
+            else:
+                rank_text = _text(cells[0]) if cells else ""
+                rank_match = re.search(r"\d+", rank_text)
+                rank = int(rank_match.group()) if rank_match else ordinal
+
+            product_cell = cells[product_idx]
+            name = _product_name(product_cell)
+            if is_excluded_product_title(name):
+                continue
+            sales_range = _compact_range(_text(cells[sales_idx]))
+            if not name or not sales_range:
+                continue
+            try:
+                sales_upper = parse_money_range_upper(sales_range)
+            except ValueError:
+                continue
+            results.append(
+                {
+                    "platform": "dy",
+                    "category_name": category_name,
+                    "category_path": category_path,
+                    "rank": rank,
+                    "product_id": product_id,
+                    "image_url": _first_http_image(product_cell),
+                    "name": name,
+                    "price_range": _price_range(product_cell),
+                    "sales_amount_range": sales_range,
+                    "sales_amount_upper": sales_upper,
+                    "product_url": _product_url(product_id),
+                    "captured_at": datetime.now().isoformat(timespec="seconds"),
+                }
+            )
+        if results:
+            break
+    return results
+
+
+def build_dy_feishu_intro_xml(
+    *,
+    title: str,
+    products: list[dict[str, Any]],
+    period: str,
+) -> str:
+    esc = lambda value: html_std.escape(str(value or ""), quote=True)
+    categories = list(
+        dict.fromkeys(item.get("category_name") for item in products if item.get("category_name"))
+    )
+    filter_summaries: list[str] = []
+    for category in categories:
+        sample = next((item for item in products if item.get("category_name") == category), {})
+        min_price = sample.get("filter_min_price")
+        max_price = sample.get("filter_max_price")
+        keyword = sample.get("search_keyword")
+        if min_price is not None and max_price is not None:
+            summary = f"{category} ¥{min_price}-¥{max_price}"
+            if keyword:
+                summary += f"（搜索：{keyword}）"
+            filter_summaries.append(summary)
+    filter_detail = (
+        f"<p><b>各品类价格条件：</b>{'；'.join(esc(item) for item in filter_summaries)}</p>"
+        if filter_summaries
+        else ""
+    )
+    return (
+        f"<title>{esc(title)}</title>"
+        "<h1>抖音</h1>"
+        f"<p>采集时间：{datetime.now():%Y-%m-%d %H:%M:%S}</p>"
+        '<callout emoji="📊" background-color="light-blue" border-color="blue">'
+        f"<p><b>筛选条件：</b>统计周期 {esc(period)}；"
+        f"品类 {'、'.join(esc(item) for item in categories)}；"
+        f"共采集 <b>{len(products)}</b> 个商品。销售额为罗盘“用户支付金额”区间。</p>"
+        f"{filter_detail}"
+        f"<p><b>标题过滤：</b>已排除包含"
+        f"{esc('、'.join(EXCLUDED_TITLE_KEYWORDS))}的商品。</p>"
+        "</callout>"
+    )
+
+
+def build_dy_feishu_batch_xml(
+    products: list[dict[str, Any]],
+    *,
+    start_rank: int,
+    include_heading: bool = True,
+) -> str:
+    esc = lambda value: html_std.escape(str(value or ""), quote=True)
+    rows: list[str] = []
+    for item in products:
+        image_url = item.get("image_url") or ""
+        image = (
+            f'<img href="{esc(image_url)}" width="{DOCUMENT_IMAGE_WIDTH}" '
+            f'height="{DOCUMENT_IMAGE_HEIGHT}"/>'
+            if image_url
+            else "未获取"
+        )
+        product_url = item.get("product_url") or ""
+        link = (
+            f'<a href="{esc(product_url)}">打开商品</a>'
+            if product_url
+            else "未解析到"
+        )
+        rows.append(
+            "<tr>"
+            "<td></td>"
+            "<td></td>"
+            f"<td>{esc(item.get('category_name') or '未分类')}</td>"
+            f"<td>{esc(item.get('rank'))}</td>"
+            f"<td>{image}</td>"
+            f"<td>{esc(item.get('name'))}</td>"
+            f"<td><b>{esc(item.get('sales_amount_range') or '未获取')}</b></td>"
+            f"<td>{esc(item.get('price_range') or '未获取')}</td>"
+            f"<td>{link}</td>"
+            "</tr>"
+        )
+    end_rank = start_rank + len(products) - 1
+    heading = (
+        f"<h2>榜单记录 {start_rank}–{end_rank}</h2>"
+        if include_heading
+        else ""
+    )
+    return (
+        heading
+        + "<table><colgroup>"
+        '<col width="110"/><col width="120"/><col width="90"/>'
+        '<col width="60"/><col width="180"/><col width="360"/>'
+        '<col width="150"/><col width="130"/><col width="100"/>'
+        "</colgroup><thead><tr>"
+        '<th background-color="light-gray">风格（运营手动区分）</th>'
+        '<th background-color="light-gray">场景用途（运营手动区分）</th>'
+        '<th background-color="light-gray">品类</th>'
+        '<th background-color="light-gray">排名</th>'
+        '<th background-color="light-gray">商品图</th>'
+        '<th background-color="light-gray">商品名称</th>'
+        '<th background-color="light-gray">销售额</th>'
+        '<th background-color="light-gray">商品价格</th>'
+        '<th background-color="light-gray">商品链接</th>'
+        f"</tr></thead><tbody>{''.join(rows)}</tbody></table>"
+    )
+
+
+def build_dy_category_table_xml(
+    category_name: str,
+    products: list[dict[str, Any]],
+) -> str:
+    """为品类文档构建 7 列紧凑表，避免大品类突破飞书单表上限。"""
+    esc = lambda value: html_std.escape(str(value or ""), quote=True)
+    heading = html_std.escape(category_name or "未分类", quote=True)
+    rows: list[str] = []
+    for item in products:
+        image_url = item.get("image_url") or ""
+        image = (
+            f'<img href="{esc(image_url)}" width="{DOCUMENT_IMAGE_WIDTH}" '
+            f'height="{DOCUMENT_IMAGE_HEIGHT}"/>'
+            if image_url
+            else "未获取"
+        )
+        product_url = item.get("product_url") or ""
+        product_link = (
+            f'<a href="{esc(product_url)}">打开商品</a>'
+            if product_url
+            else "未解析到"
+        )
+        metrics = (
+            f"销售额：{item.get('sales_amount_range') or '未获取'}；"
+            f"商品价格：{item.get('price_range') or '未获取'}"
+        )
+        rows.append(
+            "<tr><td></td><td></td>"
+            f"<td>{esc(item.get('rank'))}</td>"
+            f"<td>{image}</td>"
+            f"<td>{esc(item.get('name'))}</td>"
+            f"<td><b>{esc(metrics)}</b></td>"
+            f"<td>{product_link}</td></tr>"
+        )
+    table = (
+        "<table><colgroup>"
+        '<col width="110"/><col width="120"/><col width="60"/>'
+        '<col width="180"/><col width="360"/><col width="230"/>'
+        '<col width="100"/>'
+        "</colgroup><thead><tr>"
+        '<th background-color="light-gray">风格（运营手动区分）</th>'
+        '<th background-color="light-gray">场景用途（运营手动区分）</th>'
+        '<th background-color="light-gray">排名</th>'
+        '<th background-color="light-gray">商品图</th>'
+        '<th background-color="light-gray">商品名称</th>'
+        '<th background-color="light-gray">销售额 / 商品价格</th>'
+        '<th background-color="light-gray">商品链接</th>'
+        f"</tr></thead><tbody>{''.join(rows)}</tbody></table>"
+    )
+    return f"<h2>{heading}（{len(products)}条）</h2>{table}"
+
+
+def _visible(locator):
+    for index in range(locator.count()):
+        item = locator.nth(index)
+        try:
+            if item.is_visible():
+                return item
+        except Exception:
+            continue
+    return None
+
+
+def _last_visible(locator):
+    for index in range(locator.count() - 1, -1, -1):
+        item = locator.nth(index)
+        try:
+            if item.is_visible():
+                return item
+        except Exception:
+            continue
+    return None
+
+
+def navigate_to_product_rank(page) -> None:
+    page.goto(MARKET_PRODUCT_RANK_URL, wait_until="domcontentloaded", timeout=60_000)
+    page.wait_for_selector("text=用户支付金额", timeout=60_000)
+    page.wait_for_selector("table tbody tr[data-row-key]", timeout=60_000)
+
+
+def select_period(page, period: str = "近30天") -> None:
+    target = _visible(page.get_by_text(period, exact=True))
+    if target is None:
+        raise RuntimeError(f"未找到统计周期: {period}")
+    class_name = target.get_attribute("class") or ""
+    if "active" not in class_name and "selected" not in class_name:
+        target.click(timeout=10_000)
+        page.wait_for_timeout(1_500)
+    print(f"[OK] 统计周期: {period}")
+
+
+def select_category_path(page, category_path: tuple[str, ...]) -> None:
+    leaf = category_path[-1]
+    current = _visible(page.locator(".filterItem-xRhL2T .aurora-select-content-value"))
+    current_title = (current.get_attribute("title") or "") if current is not None else ""
+    if leaf != "全部" and leaf in current_title:
+        print(f"[OK] 当前品类: {' > '.join(category_path)}")
+        return
+
+    picker = _visible(page.locator(".filterItem-xRhL2T .aurora-cascader"))
+    if picker is None:
+        raise RuntimeError("未找到行业类目级联选择器")
+    picker.click(timeout=10_000)
+    page.wait_for_timeout(300)
+    for index, label in enumerate(category_path):
+        # “全部”在每一级菜单中都会出现，必须取最右侧（最深层）的可见选项。
+        option = _last_visible(page.locator(f'.aurora-cascader-menu-item[title="{label}"]'))
+        if option is None:
+            raise RuntimeError(
+                f"品类路径第 {index + 1} 级不存在: {label} "
+                f"({' > '.join(category_path)})"
+            )
+        option.click(timeout=10_000)
+        page.wait_for_timeout(350 if index < len(category_path) - 1 else 1_800)
+    page.wait_for_selector("table tbody tr[data-row-key]", timeout=30_000)
+    print(f"[OK] 已选择品类: {' > '.join(category_path)}")
+
+
+def select_category_paths(
+    page,
+    category_paths: tuple[tuple[str, ...], ...],
+) -> tuple[str, ...]:
+    errors: list[str] = []
+    for category_path in category_paths:
+        try:
+            select_category_path(page, category_path)
+            return category_path
+        except Exception as exc:
+            errors.append(f"{' > '.join(category_path)}: {exc}")
+            try:
+                page.keyboard.press("Escape")
+                page.wait_for_timeout(300)
+            except Exception:
+                pass
+    raise RuntimeError("所有候选类目路径均选择失败: " + "；".join(errors))
+
+
+def set_product_search(page, keyword: str = "") -> None:
+    search_input = _visible(page.locator('input[placeholder*="商品名"]'))
+    if search_input is None:
+        raise RuntimeError("未找到商品名称搜索框")
+    if (search_input.input_value() or "").strip() == keyword:
+        print(f"[OK] 商品搜索: {keyword or '已清空'}")
+        return
+    old_key = _first_row_key(page)
+    search_input.fill(keyword)
+    search_input.press("Enter")
+    try:
+        page.wait_for_function(
+            "(oldKey) => !oldKey || "
+            "document.querySelector('table tbody tr[data-row-key]')?.dataset.rowKey !== oldKey",
+            arg=old_key,
+            timeout=8_000,
+        )
+    except Exception:
+        pass
+    page.wait_for_timeout(1_200)
+    print(f"[OK] 商品搜索: {keyword or '已清空'}")
+
+
+def apply_custom_price(page, min_price: int, max_price: int) -> None:
+    if min_price < 0 or max_price <= min_price:
+        raise ValueError(f"价格区间无效: {min_price}-{max_price}")
+
+    custom = _visible(page.get_by_text("自定义", exact=True))
+    if custom is None:
+        raise RuntimeError("未找到价格带“自定义”入口")
+    custom.click(timeout=10_000)
+    page.wait_for_timeout(300)
+
+    min_input = _visible(page.locator('input[placeholder="输入最小值"]'))
+    max_input = _visible(page.locator('input[placeholder="输入最大值"]'))
+    if min_input is None or max_input is None:
+        raise RuntimeError("未找到自定义价格上下限输入框")
+    min_input.fill(str(min_price))
+    max_input.fill(str(max_price))
+
+    confirm = _visible(page.get_by_text("确定", exact=True))
+    if confirm is None:
+        raise RuntimeError("未找到自定义价格确认按钮")
+    old_key = _first_row_key(page)
+    confirm.click(timeout=10_000)
+    try:
+        page.wait_for_function(
+            "(oldKey) => !oldKey || "
+            "document.querySelector('table tbody tr[data-row-key]')?.dataset.rowKey !== oldKey",
+            arg=old_key,
+            timeout=8_000,
+        )
+    except Exception:
+        pass
+    page.wait_for_timeout(1_200)
+    print(f"[OK] 价格带: {min_price}-{max_price}")
+
+
+def _first_row_key(page) -> str:
+    row = _visible(page.locator("table tbody tr[data-row-key]"))
+    return (row.get_attribute("data-row-key") or "") if row is not None else ""
+
+
+def _click_next_page(page) -> bool:
+    selectors = (
+        ".aurora-pagination-next",
+        'button[aria-label*="下一页"]',
+        'li[title="下一页"]',
+        'button[title="下一页"]',
+    )
+    for selector in selectors:
+        target = _visible(page.locator(selector))
+        if target is None:
+            continue
+        class_name = target.get_attribute("class") or ""
+        disabled = target.get_attribute("disabled")
+        aria_disabled = target.get_attribute("aria-disabled")
+        if disabled is not None or aria_disabled == "true" or "disabled" in class_name:
+            return False
+        target.click(timeout=10_000)
+        return True
+    return False
+
+
+def collect_rank_pages(
+    page,
+    *,
+    category_name: str,
+    category_path: tuple[str, ...],
+    max_pages: int,
+    limit: int,
+) -> list[dict[str, Any]]:
+    products: list[dict[str, Any]] = []
+    seen: set[str] = set()
+    page_number = 1
+    while page_number <= max_pages:
+        page.wait_for_selector("table tbody tr[data-row-key]", timeout=30_000)
+        source = page.content()
+        rows = parse_market_product_rank_html(
+            source,
+            category_name=category_name,
+            category_path=" > ".join(category_path),
+        )
+        for item in rows:
+            key = item.get("product_id") or f"{category_name}:{item.get('rank')}:{item.get('name')}"
+            if key in seen:
+                continue
+            seen.add(key)
+            item["page_number"] = page_number
+            products.append(item)
+            if limit and len(products) >= limit:
+                return products
+        print(f"[OK] {category_name} 第 {page_number} 页: {len(rows)} 条，累计 {len(products)} 条")
+
+        old_key = _first_row_key(page)
+        if not _click_next_page(page):
+            break
+        try:
+            page.wait_for_function(
+                "(oldKey) => document.querySelector('table tbody tr[data-row-key]')?.dataset.rowKey !== oldKey",
+                arg=old_key,
+                timeout=20_000,
+            )
+        except Exception:
+            new_key = _first_row_key(page)
+            if not new_key or new_key == old_key:
+                break
+        page.wait_for_timeout(800)
+        page_number += 1
+    return products
+
+
+def _profile_has_cookies(profile_dir: str) -> bool:
+    return any(
+        path.exists()
+        for path in (
+            Path(profile_dir) / "Default" / "Network" / "Cookies",
+            Path(profile_dir) / "Default" / "Cookies",
+        )
+    )
+
+
+def create_dy_feishu_doc_batched(
+    title: str,
+    products: list[dict[str, Any]],
+    *,
+    period: str,
+    batch_size: int = 40,
+    max_retries: int = 5,
+) -> str:
+    from collect_jd_market_rank import append_feishu_batch_with_retry
+
+    intro = build_dy_feishu_intro_xml(title=title, products=products, period=period)
+    doc_url = create_feishu_doc(title, intro)
+    for batch_index, offset in enumerate(range(0, len(products), batch_size), 1):
+        batch = products[offset : offset + batch_size]
+        xml = build_dy_feishu_batch_xml(batch, start_rank=offset + 1)
+        append_feishu_batch_with_retry(
+            doc_url,
+            xml,
+            batch_index=batch_index,
+            max_retries=max_retries,
+        )
+        print(f"[OK] 飞书文档批次 {batch_index}: {offset + 1}-{offset + len(batch)}")
+        if batch_index % 5 == 0 and offset + len(batch) < len(products):
+            time.sleep(5)
+    return doc_url
+
+
+def _feishu_doc_contains_keyword(doc_url: str, keyword: str) -> bool:
+    proc = subprocess.run(
+        [
+            _lark_cli(),
+            "docs",
+            "+fetch",
+            "--doc",
+            doc_url,
+            "--scope",
+            "keyword",
+            "--keyword",
+            keyword,
+            "--as",
+            "user",
+            "--format",
+            "json",
+        ],
+        cwd=PROJECT_ROOT,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        timeout=180,
+    )
+    if proc.returncode != 0:
+        return False
+    try:
+        obj = json.loads(proc.stdout)
+    except json.JSONDecodeError:
+        return False
+    document = ((obj.get("data") or {}).get("document") or {})
+    return keyword in html_std.unescape(document.get("content") or "")
+
+
+def append_feishu_category_table(
+    doc_url: str,
+    category_name: str,
+    products: list[dict[str, Any]],
+    index: int,
+) -> None:
+    heading = f"{category_name}（{len(products)}条）"
+    last_name = str(products[-1].get("name") or "") if products else ""
+    if (
+        _feishu_doc_contains_keyword(doc_url, heading)
+        and last_name
+        and _feishu_doc_contains_keyword(doc_url, last_name)
+    ):
+        print(f"[SKIP] 飞书品类表 {category_name} 已完整存在")
+        return
+
+    xml = build_dy_category_table_xml(category_name, products)
+    try:
+        append_feishu_doc(doc_url, xml, index)
+        return
+    except RuntimeError as exc:
+        message = str(exc).lower()
+        if "timeout" not in message and "time out" not in message:
+            raise
+
+    # 大表（尤其包含上百张图片）可能已经在飞书服务端成功写入，
+    # 但 lark-cli 等待响应超时。必须先反查，不能直接重试，否则会生成重复表。
+    for attempt in range(25):
+        heading_ok = _feishu_doc_contains_keyword(doc_url, heading)
+        last_row_ok = bool(last_name) and _feishu_doc_contains_keyword(doc_url, last_name)
+        if heading_ok and last_row_ok:
+            print(f"[OK] 飞书品类表 {category_name} 超时后反查确认已完整写入")
+            return
+        if attempt < 24:
+            time.sleep(5)
+    raise RuntimeError(
+        f"飞书品类表 {category_name} 上传超时，且等待 120 秒后仍未确认完整写入"
+    )
+
+
+def create_dy_feishu_doc_by_category(
+    title: str,
+    products: list[dict[str, Any]],
+    *,
+    period: str,
+) -> str:
+    intro = build_dy_feishu_intro_xml(title=title, products=products, period=period)
+    doc_url = create_feishu_doc(title, intro)
+    groups = group_products_by_category(products)
+    append_native_category_tables(
+        doc_url,
+        [
+            build_dy_category_table_xml(str(category_name), category_products)
+            for category_name, category_products in groups
+        ],
+        image_size=DOCUMENT_IMAGE_WIDTH,
+    )
+    verify_feishu_category_tables(doc_url, groups)
+    return doc_url
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description="采集抖音市场商品榜单并生成飞书文档")
+    parser.add_argument(
+        "--category",
+        default="",
+        help="逻辑品类名，或用 > 指定完整级联路径；不传时采集配置的 8 个品类",
+    )
+    parser.add_argument("--period", default="近30天")
+    parser.add_argument("--min-price", type=int, default=None, help="覆盖默认最低价格")
+    parser.add_argument("--max-price", type=int, default=None, help="覆盖默认最高价格")
+    parser.add_argument("--search-keyword", default=None, help="覆盖默认商品搜索词")
+    parser.add_argument("--max-pages", type=int, default=20)
+    parser.add_argument("--limit", type=int, default=0)
+    parser.add_argument(
+        "--doc-batch-size",
+        type=int,
+        default=40,
+        help="飞书文档每批商品数；避免单品类大表上传超时",
+    )
+    parser.add_argument("--feishu-retries", type=int, default=5)
+    parser.add_argument("--user-data-dir", default=os.getenv("DY_USER_DATA_DIR", USER_DATA_DIR))
+    parser.add_argument("--dry-run", action="store_true", help="只保存 JSON，不创建飞书文档")
+    return parser.parse_args()
+
+
+def _category_specs(value: str) -> tuple[dict[str, Any], ...]:
+    if not value:
+        return TARGET_DY_MARKET_CATEGORIES
+    if ">" in value:
+        path = tuple(part.strip() for part in value.split(">") if part.strip())
+        if not path:
+            raise ValueError("品类路径不能为空")
+        return (
+            {
+                "name": path[-1],
+                "paths": (path,),
+                "min_price": 0,
+                "max_price": 0,
+                "search_keyword": "",
+            },
+        )
+    matches = tuple(item for item in TARGET_DY_MARKET_CATEGORIES if item["name"] == value)
+    if not matches:
+        known = "、".join(item["name"] for item in TARGET_DY_MARKET_CATEGORIES)
+        raise ValueError(f"未知品类 {value}，可用品类: {known}")
+    return matches
+
+
+def main() -> int:
+    args = parse_args()
+    if args.max_pages <= 0:
+        raise ValueError("--max-pages 必须大于 0")
+    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+    category_specs = _category_specs(args.category)
+    session = StealthySession(
+        headless=False,
+        real_chrome=True,
+        user_data_dir=args.user_data_dir,
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=120_000,
+        network_idle=False,
+    )
+    session.start()
+    products: list[dict[str, Any]] = []
+    try:
+        context = session.context
+        if not _profile_has_cookies(args.user_data_dir):
+            cookies = _load_cookies() or []
+            if cookies:
+                context.add_cookies(cookies)
+        page = context.pages[-1] if context.pages else context.new_page()
+        navigate_to_product_rank(page)
+        if not _check_login_status(page):
+            page = _ensure_login(page)
+            if page is None:
+                raise RuntimeError("抖音罗盘登录失败")
+            navigate_to_product_rank(page)
+        _save_browser_state(page, "dy market rank")
+        select_period(page, args.period)
+
+        for index, spec in enumerate(category_specs, 1):
+            category_name = str(spec["name"])
+            paths = tuple(tuple(path) for path in spec["paths"])
+            min_price = args.min_price if args.min_price is not None else int(spec["min_price"])
+            max_price = args.max_price if args.max_price is not None else int(spec["max_price"])
+            search_keyword = (
+                args.search_keyword
+                if args.search_keyword is not None
+                else str(spec.get("search_keyword") or "")
+            )
+            print(f"[->] 品类 {index}/{len(category_specs)}: {category_name}")
+            set_product_search(page, "")
+            path = select_category_paths(page, paths)
+            if min_price or max_price:
+                apply_custom_price(page, min_price, max_price)
+            if search_keyword:
+                set_product_search(page, search_keyword)
+            category_limit = 0
+            if args.limit:
+                category_limit = max(0, args.limit - len(products))
+                if category_limit == 0:
+                    break
+            category_products = collect_rank_pages(
+                page,
+                category_name=category_name,
+                category_path=path,
+                max_pages=args.max_pages,
+                limit=category_limit,
+            )
+            for item in category_products:
+                item["filter_min_price"] = min_price
+                item["filter_max_price"] = max_price
+                item["search_keyword"] = search_keyword
+            products.extend(category_products)
+            CHECKPOINT_PATH.parent.mkdir(parents=True, exist_ok=True)
+            CHECKPOINT_PATH.write_text(
+                json.dumps(products, ensure_ascii=False, indent=2),
+                encoding="utf-8",
+            )
+
+        stamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+        json_path = OUTPUT_DIR / f"dy_market_rank_{stamp}.json"
+        json_path.write_text(
+            json.dumps(products, ensure_ascii=False, indent=2),
+            encoding="utf-8",
+        )
+        print(f"[OK] 采集结果: {json_path} ({len(products)} 条)")
+        if args.dry_run:
+            return 0
+        title_scope = category_specs[0]["name"] if len(category_specs) == 1 else "八品类"
+        title = f"抖音{title_scope}市场商品榜单 {args.period} {datetime.now():%Y-%m-%d}"
+        doc_url = create_dy_feishu_doc_by_category(
+            title,
+            products,
+            period=args.period,
+        )
+        print(f"[OK] 飞书文档: {doc_url}")
+        archive_market_rank_report(
+            platform="dy",
+            report_date=datetime.now().date(),
+            period=args.period,
+            category_scope=str(title_scope),
+            category_names=[str(spec["name"]) for spec in category_specs],
+            report_title=title,
+            feishu_doc_url=doc_url,
+            product_count=len(products),
+            source_json_path=str(json_path),
+            run_metadata={
+                "category": args.category or None,
+                "min_price": args.min_price,
+                "max_price": args.max_price,
+                "search_keyword": args.search_keyword,
+                "max_pages": args.max_pages,
+                "limit": args.limit,
+            },
+        )
+        return 0
+    finally:
+        session.close()
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/collect_dy_persona_to_bitable.py b/src/gyxx_flow/modules/product_commerce/runtime/collect_dy_persona_to_bitable.py
new file mode 100644
index 0000000..9f46809
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/collect_dy_persona_to_bitable.py
@@ -0,0 +1,414 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""每日人群画像（抖音）采集 + 回填飞书表。
+
+流程:
+1. 从父表"各平台款式ID收集表"动态读取 (经 StyleConfigLoader) 款式 -> 人群画像目标表映射
+   (只取同时有 款式 + dy item_ids + dy_persona_bitable 的抖音款式)
+2. 调 dy_audience_profile_collect.py 采抖音罗盘成交人群画像四组数值
+   (一次登录循环所有款, 每款产出 JSON 到 data/dy人物画像/)
+3. 读每款 JSON, 转换 profile 数据为飞书字段 payload
+4. 按"时间"(M.D) upsert 到各款式对应的"运营人群画像（抖音）"表
+
+与 collect_persona_to_bitable.py (天猫) 完全同构。
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import shutil
+import subprocess
+import sys
+from datetime import datetime, timedelta
+from pathlib import Path
+
+from runtime_paths import LOG_ROOT, RAW_DATA_ROOT, profile_path, runtime_script
+from typing import Any
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+from config.style_config_loader import (
+    StyleConfigLoader,
+    get_dy_persona_bitable_map,
+    get_dy_product_groups,
+)
+
+COLLECT_SCRIPT = runtime_script("dy_audience_profile_collect.py")
+OUTPUT_ROOT = RAW_DATA_ROOT / "dy人物画像"
+LOG_DIR = LOG_ROOT
+FAIL_LIST_PATH = LOG_DIR / "dy_persona_bitable_failures.json"
+
+DEFAULT_USER_DATA_DIR = os.getenv("DY_USER_DATA_DIR", str(profile_path("dy")))
+
+
+# ── 时间标签 ──────────────────────────────────────────────────────────────────
+
+def time_label(now: datetime | None = None) -> str:
+    now = now or datetime.now()
+    return f"{now.date().month}.{now.date().day}"
+
+
+# ── 采集 subprocess ──────────────────────────────────────────────────────────
+
+def run_collect_all(user_data_dir: str, headless: bool, styles: str = "") -> bool:
+    """调采集脚本一次采全部款式 (一次登录循环所有款), 返回是否成功。
+    采集脚本本身的 main 就是 "一次登录 + for style in styles" 模式,
+    所以这里只调一次 subprocess, 不要在循环里反复调。
+    """
+    cmd = [
+        sys.executable,
+        str(COLLECT_SCRIPT),
+        "--user-data-dir", user_data_dir,
+    ]
+    if styles:
+        cmd += ["--styles", styles]
+    if headless:
+        cmd.append("--headless")
+    print(f"[DY] 一次采集全部款式: {' '.join(cmd)}")
+    proc = subprocess.run(cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace")
+    if proc.returncode:
+        print(f"[DY] 失败 exit={proc.returncode}")
+        return False
+    return True
+
+
+# ── 读采集产出 ────────────────────────────────────────────────────────────────
+
+def load_latest_profile(style: str) -> dict[str, Any] | None:
+    """读采集产出的该款式最新 profile JSON, 返回 dict 或 None。"""
+    date_stamp = (datetime.now().date() - timedelta(days=1)).strftime("%Y-%m-%d")
+    style_dir = OUTPUT_ROOT / date_stamp / style
+    if not style_dir.exists():
+        print(f"[DY] 输出目录不存在: {style_dir}")
+        return None
+    jsons = sorted(style_dir.glob(f"{style}_*_profile.json"), key=lambda p: p.stat().st_mtime, reverse=True)
+    if not jsons:
+        print(f"[DY] 未找到 profile JSON in {style_dir}")
+        return None
+    payload = json.loads(jsons[0].read_text(encoding="utf-8"))
+    print(f"[DY] 读取 {jsons[0].name}")
+    return payload
+
+
+# ── 飞书字段映射 ──────────────────────────────────────────────────────────────
+
+def _normalize_name(name: str) -> str:
+    """归一化采集到的 name 以匹配子表字段名。"""
+    if not name:
+        return ""
+    n = name.strip()
+    n = re.sub(r"岁以上", "以上", n)
+    n = re.sub(r"(岁|占比|%)$", "", n)
+    n = n.replace("+", "以上")
+    n = n.strip()
+    aliases = {
+        "男": "男性比例",
+        "男性": "男性比例",
+        "genz": "Z世代",
+        "GenZ": "Z世代",
+    }
+    return aliases.get(n, n)
+
+
+def _parse_numeric_value(value) -> float | None:
+    if value is None:
+        return None
+    if isinstance(value, (int, float)):
+        return float(value)
+    s = str(value).strip().replace("%", "").replace(",", "")
+    if not s:
+        return None
+    try:
+        return float(s)
+    except ValueError:
+        return None
+
+
+def profile_to_fields(profile: dict, subtable_fields: list[dict]) -> dict[str, Any]:
+    """把采集的四组分布映射到子表字段，返回 {field_name: value}。
+
+    采集组：gender_distribution / age_distribution / strategy_crowd_distribution
+    （resident_province 不写入子表，仅本地备份）
+    """
+    field_index: dict[str, dict] = {}
+    for f in subtable_fields:
+        norm = _normalize_name(f["name"])
+        if norm:
+            field_index[norm] = f
+
+    cell_values: dict[str, Any] = {}
+    groups = [
+        profile.get("gender_distribution") or [],
+        profile.get("age_distribution") or [],
+        profile.get("strategy_crowd_distribution") or [],
+    ]
+    for group in groups:
+        for item in group:
+            raw_name = item.get("name", "")
+            raw_value = item.get("value")
+            norm = _normalize_name(raw_name)
+            if not norm:
+                continue
+            target = field_index.get(norm)
+            if not target:
+                continue
+            num = _parse_numeric_value(raw_value)
+            if num is None:
+                continue
+            field_name = target["name"]
+            if target.get("type") == "text":
+                cell_values[field_name] = f"{num}"
+            else:
+                cell_values[field_name] = num
+    return cell_values
+
+
+# ── lark-cli 封装 ─────────────────────────────────────────────────────────────
+
+def _lark_cli() -> str:
+    return shutil.which("lark-cli") or shutil.which("lark-cli.cmd") or "lark-cli.cmd"
+
+
+def _run_lark_cli(args: list[str]) -> dict:
+    cli_cmd = _lark_cli()
+    node = shutil.which("node.exe") or shutil.which("node")
+    if not node:
+        raise RuntimeError("未找到 node")
+    cli_dir = os.path.dirname(cli_cmd)
+    cli_script = os.path.join(cli_dir, "node_modules", "@larksuite", "cli", "scripts", "run.js")
+    if not os.path.exists(cli_script):
+        raise RuntimeError(f"未找到 lark-cli Node 入口: {cli_script}")
+    cmd = [node, cli_script, "--profile", "hermes-analyzer", "--as", "user"] + args
+    result = subprocess.run(
+        cmd, capture_output=True, text=True, encoding="utf-8", errors="replace",
+    )
+    if result.returncode != 0:
+        raise RuntimeError(f"lark-cli 失败: {(result.stderr or result.stdout).strip()}")
+    out = (result.stdout or "").strip()
+    if not out:
+        return {}
+    try:
+        return json.loads(out)
+    except json.JSONDecodeError:
+        raise RuntimeError(f"lark-cli 返回非 JSON: {out[:200]}")
+
+
+def get_subtable_fields(base_token: str, table_id: str) -> list[dict]:
+    """读子表字段定义，返回 [{id, name, type}]。"""
+    res = _run_lark_cli([
+        "base", "+field-list",
+        "--base-token", base_token,
+        "--table-id", table_id,
+    ])
+    data = res.get("data", {}) if isinstance(res.get("data"), dict) else {}
+    fields = data.get("fields") or []
+    return [{"id": f.get("id", ""), "name": f.get("name", ""), "type": f.get("type", "")} for f in fields]
+
+
+def find_record_id_by_time(base_token: str, table_id: str, time_str: str, identity: str = "user") -> str | None:
+    """按 时间 字段查现有记录, 返回 record_id 或 None。"""
+    filter_payload = {"logic": "and", "conditions": [["时间", "==", time_str]]}
+    cmd = [
+        _lark_cli(), "base", "+record-list",
+        "--as", identity,
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--filter-json", json.dumps(filter_payload, ensure_ascii=False),
+        "--limit", "5",
+        "--format", "json",
+    ]
+    proc = subprocess.run(cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace", capture_output=True)
+    if proc.returncode:
+        return None
+    try:
+        envelope = json.loads(proc.stdout)
+    except json.JSONDecodeError:
+        return None
+    if not envelope.get("ok"):
+        return None
+    record_ids = (envelope.get("data") or {}).get("record_id_list") or []
+    return record_ids[0] if record_ids else None
+
+
+def upsert_record(base_token: str, table_id: str, fields: dict[str, Any],
+                  identity: str = "user", dry_run: bool = False, record_id: str | None = None) -> tuple[bool, str]:
+    """upsert 一条记录, 返回 (ok, message)。"""
+    cmd = [
+        _lark_cli(), "base", "+record-upsert",
+        "--as", identity,
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--json", json.dumps(fields, ensure_ascii=False),
+    ]
+    if record_id:
+        cmd += ["--record-id", record_id]
+    print(f"[LARK] upsert base={base_token[:10]}.. table={table_id} time={fields.get('时间')} record_id={record_id or '(new)'}")
+    if dry_run:
+        print(f"[LARK] dry-run payload: {json.dumps(fields, ensure_ascii=False)}")
+        return True, "dry-run"
+    proc = subprocess.run(cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace", capture_output=True)
+    if proc.returncode:
+        return False, f"exit={proc.returncode} {proc.stderr.strip()[:200]}"
+    try:
+        env = json.loads(proc.stdout)
+        if env.get("ok"):
+            data = env.get("data") or {}
+            rid = None
+            if isinstance(data, dict):
+                rec = data.get("record") or {}
+                rid = (rec.get("record_id_list") or [None])[0] if isinstance(rec, dict) else None
+                rid = rid or (data.get("record_id_list") or [None])[0] or data.get("record_id")
+            rid = rid or record_id or "?"
+            return True, f"record_id={rid}"
+        return False, str(env.get("error", {}))
+    except json.JSONDecodeError:
+        return False, proc.stdout[:200]
+
+
+def is_permission_error(message: str) -> bool:
+    m = message.lower()
+    return "91402" in m or "91403" in m or "notexist" in m or "permission" in m
+
+
+# ── upsert 单款 ──────────────────────────────────────────────────────────────
+
+def upsert_one_style(style: str, persona_cfg: dict[str, str],
+                     time_str: str, identity: str, dry_run: bool) -> dict[str, Any]:
+    """读该款式已采的 profile JSON + 读子表字段 + 转字段 + upsert 到飞书, 返回结果 dict。
+    前置: 采集已由 run_collect_all 一次跑完, 这里只做读+转+写。
+    """
+    result = {"style": style, "base_token": persona_cfg["base_token"], "table_id": persona_cfg["table_id"],
+              "status": "pending", "message": ""}
+
+    # 1. 读已采的 profile
+    payload = load_latest_profile(style)
+    if not payload or not payload.get("profile"):
+        result["status"] = "collect_failed"
+        result["message"] = "采集无 profile 数据"
+        return result
+
+    profile = payload["profile"]
+
+    # 1.5 判断是否为无效数据 (年龄不存在 + 八大人群全为0)
+    age_items = profile.get("age_distribution") or []
+    strat_items = profile.get("strategy_crowd_distribution") or []
+    strat_all_zero = all(float(it.get("value", 0)) == 0 for it in strat_items) if strat_items else True
+    if not age_items and strat_all_zero:
+        result["status"] = "no_data"
+        result["message"] = "年龄为空 + 八大人群全为0, 无有效画像数据"
+        return result
+    try:
+        sub_fields = get_subtable_fields(persona_cfg["base_token"], persona_cfg["table_id"])
+        fields = profile_to_fields(profile, sub_fields)
+    except Exception as e:
+        result["status"] = "field_read_failed"
+        result["message"] = str(e)
+        return result
+
+    if len(fields) == 0:
+        result["status"] = "no_data"
+        result["message"] = "profile 转换后无有效字段"
+        return result
+
+    fields["时间"] = time_str
+
+    # 2.5 入库 PostgreSQL
+    try:
+        from db import get_conn, normalize_dy_payload, upsert_persona_metrics
+        normalized = normalize_dy_payload(profile)
+        data_stamp = (datetime.now().date() - timedelta(days=1)).strftime("%Y-%m-%d")
+        collect_stamp = datetime.now().strftime("%Y-%m-%d")
+        record = {
+            "collect_date": collect_stamp,
+            "style_name": style,
+            "platform": "dy",
+            "item_id": payload.get("product_id"),
+            "payload": normalized,
+            "raw_payload": profile,
+            "source_file": str(OUTPUT_ROOT / data_stamp / style),
+            "collected_at": payload.get("captured_at"),
+        }
+        with get_conn() as conn:
+            n = upsert_persona_metrics(conn, [record])
+        print(f"  [DB] DY 入库 {n} 条")
+    except Exception as e:
+        print(f"  [WARN] DY 入库失败 (不阻塞飞书): {e}")
+
+    # 3. upsert
+    rid = None if dry_run else find_record_id_by_time(persona_cfg["base_token"], persona_cfg["table_id"], time_str, identity)
+    ok, msg = upsert_record(persona_cfg["base_token"], persona_cfg["table_id"], fields,
+                            identity=identity, dry_run=dry_run, record_id=rid)
+    if ok:
+        result["status"] = "ok"
+        result["message"] = msg
+    else:
+        result["status"] = "permission_skipped" if is_permission_error(msg) else "upsert_failed"
+        result["message"] = msg
+    return result
+
+
+# ── main ─────────────────────────────────────────────────────────────────────
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="每日人群画像（抖音）采集 + 回填飞书表")
+    parser.add_argument("--styles", default="", help="只跑指定款式, 逗号分隔; 默认全部")
+    parser.add_argument("--user-data-dir", default=DEFAULT_USER_DATA_DIR, help="Chrome profile 根目录")
+    parser.add_argument("--headless", action="store_true", help="无头跑")
+    parser.add_argument("--identity", default="user", choices=["bot", "user"], help="飞书写入身份")
+    parser.add_argument("--dry-run", action="store_true", help="只转字段不写飞书")
+    parser.add_argument("--skip-collect", action="store_true", help="跳过采集, 直接用已产的 JSON upsert")
+    parser.add_argument("--refresh-config", action="store_true", help="强制从飞书重读款式配置")
+    args = parser.parse_args()
+
+    LOG_DIR.mkdir(parents=True, exist_ok=True)
+
+    if args.refresh_config:
+        StyleConfigLoader().load(force_refresh=True, sync_fallback=True)
+        print("[INFO] --refresh-config: 已重读款式配置")
+
+    persona_map = get_dy_persona_bitable_map()
+    if not persona_map:
+        print("[FAIL] 未读到任何人群画像（抖音）目标表映射")
+        return 2
+
+    selected = [s for s in (s.strip() for s in args.styles.split(",")) if s]
+    if selected:
+        persona_map = {k: v for k, v in persona_map.items() if k in selected}
+
+    print(f"[INFO] 待采款式 {len(persona_map)}: {list(persona_map.keys())}")
+    time_str = time_label()
+    print(f"[INFO] 时间字段 = {time_str}")
+
+    # 1. 一次调采集脚本采全部款式 (一次登录循环所有款, 不要每款重启浏览器)
+    if not args.dry_run and not args.skip_collect:
+        styles_arg = ",".join(persona_map.keys())
+        run_collect_all(args.user_data_dir, args.headless, styles_arg)
+        # 不论 exit code, 都进 upsert 阶段 (读 JSON 时缺数据的款式会记 collect_failed)
+
+    # 2. 读每款 JSON + 转字段 + upsert
+    results: list[dict[str, Any]] = []
+    for idx, (style, cfg) in enumerate(persona_map.items(), 1):
+        print(f"\n=== [{idx}/{len(persona_map)}] {style} ===")
+        r = upsert_one_style(style, cfg, time_str, args.identity, args.dry_run)
+        print(f"  -> {r['status']}: {r['message']}")
+        results.append(r)
+
+    # 失败清单
+    failures = [r for r in results if r["status"] not in ("ok", "dry-run")]
+    FAIL_LIST_PATH.write_text(json.dumps(failures, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    ok = sum(1 for r in results if r["status"] == "ok")
+    skipped = sum(1 for r in results if r["status"] == "permission_skipped")
+    failed = sum(1 for r in results if r["status"] in ("collect_failed", "upsert_failed", "no_data", "field_read_failed"))
+    print(f"\n[DONE] ok={ok} permission_skipped={skipped} failed={failed} total={len(results)}")
+    if failures:
+        print(f"[FAIL] 失败清单: {FAIL_LIST_PATH}")
+    return 0 if not failed else 2
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/collect_erp_yesterday_metrics.py b/src/gyxx_flow/modules/product_commerce/runtime/collect_erp_yesterday_metrics.py
new file mode 100644
index 0000000..96a66b4
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/collect_erp_yesterday_metrics.py
@@ -0,0 +1,1466 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import sys
+from datetime import datetime, timedelta
+from pathlib import Path
+
+from runtime_paths import LOG_ROOT, RAW_DATA_ROOT, managed_data_path
+from typing import Any
+
+from scrapling.fetchers import DynamicFetcher
+
+from erp_login_product_analysis import (
+    DEFAULT_CONFIG,
+    DEFAULT_HOME_URL,
+    USER_AGENT,
+    load_config,
+    make_page_action,
+    normalize_cdp_url,
+)
+from collect_retry_utils import FailedCollector, retry_step
+from config.style_config_loader import StyleConfigLoader, get_erp_styles_input
+
+
+ROOT = Path(__file__).resolve().parent
+DEFAULT_INPUT = ROOT / "styles_input.json"
+DEFAULT_DATA_DIR = RAW_DATA_ROOT
+SKIP_REPORT_DIR = LOG_ROOT
+
+PLATFORMS_BY_BRAND: dict[str, dict[str, dict[str, Any]]] = {
+    "光影行星": {
+        "jd": {"label": "京东", "dir": "jd单品数据", "shops": ["shop_11684400"]},
+        "dy": {"label": "抖音", "dir": "dy单品数据", "shops": ["shop_14280334", "shop_11897924"]},
+        "tm": {"label": "天猫", "dir": "tm单品数据", "shops": ["shop_11764614", "shop_20149659"]},
+    },
+    "ozko": {
+        "tm": {"label": "天猫", "dir": "tm单品数据", "shops": ["shop_11565429"]},
+        # ozko 没有 JD/DY 店铺, ERP 不采这两个平台
+    },
+}
+
+DEFAULT_BRAND = "光影行星"
+
+# 2026-06-26 曾因部分 10xxx 新款的 allsku iframe 长时间不出现，将 10514、10474、10515
+# 临时加入跳过名单以缩短每日任务。2026-07-27 确认这些款式应恢复正常查询，名单清空。
+# 保留机制供临时排障使用，但不要把业务款式永久写成成功的 0/0。
+ERP_SLOW_SKIP_CODES: set[str] = set()
+
+# 兼容 PLATFORMS 旧名 (其它模块可能 import 它, 兜底给光影行星)
+PLATFORMS: dict[str, dict[str, Any]] = PLATFORMS_BY_BRAND[DEFAULT_BRAND]
+
+
+def resolve_platform_for_style(brand: str, platform_key: str) -> dict[str, Any] | None:
+    """按品牌 + 平台键找 ERP 店铺配置; 找不到返回 None (跳过)。"""
+    return PLATFORMS_BY_BRAND.get(brand, {}).get(platform_key)
+
+YESTERDAY_SALES_IDX = 45
+YESTERDAY_RETURNS_IDX = 57
+TARGET_DATE = None
+MODE = "daily"
+# 2026-06-11: daily 走 cell 45/57 老路径 (保护 run_daily_collect.bat 定时任务, data.json 口径不变);
+#             weekly 同样 cell 45/57 (2026-06-27 run_weekly_collect.bat 定时任务已删除, --mode weekly 留作手动回采);
+#             backfill 单日窗口 + 表头匹配 (自定义销量 / 自定义退货量 + SPU 行汇总), 跟 daily/weekly 区分开
+WEEKLY_SALES_HEADER = "__idx__45"
+WEEKLY_RETURNS_HEADER = "__idx__57"
+BACKFILL_SALES_HEADER = "自定义销量"
+BACKFILL_RETURNS_HEADER = "自定义退货量"
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description="Collect ERP yesterday sales/returns by style and platform.")
+    parser.add_argument("--config", default=str(DEFAULT_CONFIG))
+    parser.add_argument("--input", default=str(DEFAULT_INPUT))
+    parser.add_argument("--data-dir", default=str(DEFAULT_DATA_DIR))
+    parser.add_argument("--platform", choices=["jd", "dy", "tm", "all"], default="all")
+    parser.add_argument("--limit", type=int, default=0, help="Limit styles per platform for testing")
+    parser.add_argument("--headless", action="store_true")
+    parser.add_argument("--no-cdp", action="store_true")
+    parser.add_argument("--timeout-ms", type=int, default=120000)
+    parser.add_argument("--report-timeout", type=int, default=60,
+                        help="单个 ERP 报表等待秒数（默认 60, 之前 30 太短, 10xxx 新款经常等不到 allsku iframe）")
+    parser.add_argument("--report-attempts", type=int, default=2,
+                        help="主循环单 code 重试次数（含首次, 默认 2 = 1 试 + 1 重试; "
+                             "末尾 _retry_failed_codes 还有 3 次兜底）")
+    parser.add_argument("--hold-seconds", type=int, default=0)
+    parser.add_argument("--target-date", type=parse_target_date, default=None,
+                        help="采集指定日期（YYYY-MM-DD）。daily 默认昨天，weekly 默认上周日")
+    parser.add_argument("--mode", choices=["daily", "weekly", "backfill"], default="daily",
+                        help="daily: 抓昨日 (run_daily_collect.bat 08:40 走这个); "
+                             "weekly: 抓上周一~周日 (2026-06-27 weekly 定时任务已删, --mode weekly 仅作手动回采); "
+                             "backfill: 单日窗口 + SPU 行汇总 (历史回补)")
+    parser.add_argument("--refresh-config", action="store_true",
+                        help="忽略 24h 缓存, 强制从飞书多维表格重读款式配置 (数据源 TtoCb1NuQaDy3NsZWTpc0GIvnph)")
+    parser.add_argument("--include-slow-codes", action="store_true",
+                        help="强制采集 ERP_SLOW_SKIP_CODES 里的款式 (默认跳过, 用于 backfill / 调试)")
+    return parser.parse_args()
+
+
+def parse_target_date(value: str):
+    try:
+        return datetime.strptime(value, "%Y-%m-%d").date()
+    except ValueError as exc:
+        raise argparse.ArgumentTypeError("target date must be YYYY-MM-DD") from exc
+
+
+def target_date():
+    if TARGET_DATE:
+        return TARGET_DATE
+    today = datetime.now().date()
+    if MODE == "weekly":
+        return today - timedelta(days=today.weekday() + 1)
+    return today - timedelta(days=1)
+
+
+def last_week_window() -> tuple:
+    today = datetime.now().date()
+    this_monday = today - timedelta(days=today.weekday())
+    last_monday = this_monday - timedelta(days=7)
+    last_sunday = this_monday - timedelta(days=1)
+    return last_monday, last_sunday
+
+
+def load_styles(path: Path) -> list[dict[str, Any]]:
+    styles = json.loads(path.read_text(encoding="utf-8"))
+    for item in styles:
+        item["erp_style_codes"] = [str(code).strip() for code in item.get("erp_style_codes", []) if str(code).strip()]
+    return styles
+
+
+def merge_styles_with_dynamic_config(styles: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    """[已废弃] 飞书为唯一来源,本地 styles_input.json 退为兜底。保留仅为兼容旧调用。"""
+    return styles
+
+
+def load_erp_styles(
+    loader: StyleConfigLoader | None,
+    fallback_path: Path,
+) -> tuple[list[dict[str, Any]], dict[str, Any]]:
+    """飞书为唯一来源,本地 styles_input.json 仅在飞书读取失败时回退。
+
+    返回 (styles, skip_report):
+      styles = [{"style_name", "erp_style_codes", "brand"}]
+      skip_report = {"source": "lark"|"fallback_json", "collectable": [...], "skipped": [{style, reason}], "brand_map": {...}}
+    brand 字段让 collect_in_page 按品牌选店铺 (光影行星旗下店铺 vs ozko旗舰店)。
+    """
+    if loader is not None:
+        try:
+            data = loader.get_erp_daily_styles()
+            payload = loader.load()
+            brand_map = data.get("brand_map", {})
+            styles: list[dict[str, Any]] = []
+            for name in data["collectable"]:
+                codes = payload["styles"][name]["erp_codes"]
+                styles.append({
+                    "style_name": name,
+                    "erp_style_codes": [str(c).strip() for c in codes if str(c).strip()],
+                    "brand": brand_map.get(name) or DEFAULT_BRAND,
+                })
+            return styles, {
+                "source": "lark",
+                "collectable": data["collectable"],
+                "skipped": data["skipped"],
+                "brand_map": brand_map,
+            }
+        except Exception as exc:
+            print(f"[WARN] 飞书读取失败,回退本地 {fallback_path}: {exc}")
+
+    styles = load_styles(fallback_path)
+    for s in styles:
+        s.setdefault("brand", DEFAULT_BRAND)
+    return styles, {
+        "source": "fallback_json",
+        "collectable": [],
+        "skipped": [],
+        "brand_map": {s["style_name"]: s.get("brand", DEFAULT_BRAND) for s in styles},
+    }
+
+
+def print_skip_summary(skip_report: dict[str, Any]) -> None:
+    src = skip_report.get("source", "?")
+    if not skip_report.get("skipped"):
+        if src == "lark":
+            print(f"[ERP] 来源={src}, 采集 {len(skip_report['collectable'])} 款, 跳过 0 款")
+        else:
+            print(f"[ERP] 来源={src}, 跳过清单未生成 (回退路径不带 reason)")
+        return
+    print(
+        f"[ERP] 来源={src}, 采集 {len(skip_report['collectable'])} 款, "
+        f"跳过 {len(skip_report['skipped'])} 款 (缺 ERP 编码 / TM 分组 ID / 飞书表地址)"
+    )
+    for s in skip_report["skipped"]:
+        print(f"  - {s['style']}: {s['reason']}")
+
+
+def write_erp_skip_report(skip_report: dict[str, Any], date_str: str) -> Path | None:
+    if skip_report.get("source") != "lark":
+        return None
+    SKIP_REPORT_DIR.mkdir(parents=True, exist_ok=True)
+    out_path = SKIP_REPORT_DIR / f"erp_skipped_{date_str}.json"
+    payload = {
+        "date": date_str,
+        "platform": "erp",
+        "generated_at": datetime.now().isoformat(timespec="seconds"),
+        "collectable": skip_report.get("collectable", []),
+        "skipped": skip_report.get("skipped", []),
+    }
+    out_path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+    print(f"[SKIP_REPORT] {out_path}")
+    return out_path
+
+
+def safe_name(value: str) -> str:
+    return re.sub(r'[<>:"/\\|?*\s]+', "_", value).strip("_")
+
+
+def parse_number(value: str | None) -> int:
+    if value is None:
+        return 0
+    text = str(value).replace(",", "").strip()
+    if not text:
+        return 0
+    try:
+        return int(float(text))
+    except ValueError:
+        return 0
+
+
+def report_date_range() -> tuple[str, str]:
+    end = target_date()
+    if MODE == "weekly":
+        start = end - timedelta(days=6)
+        return start.strftime("%Y-%m-%d"), end.strftime("%Y-%m-%d")
+    if MODE == "backfill":
+        # backfill 模式: 单日窗口, 用自定义销量/退货量列 + SPU 行求和
+        return end.strftime("%Y-%m-%d"), end.strftime("%Y-%m-%d")
+    # daily 模式: 单日窗口, cell 45/57 给的是 end 那天单日销量 (见 erp-daily-snapshot-semantics)
+    # 不能用绝对日期 (历史曾写死 2026-05-01, 距今超 60 天后 ERP 报表拒绝返回)
+    return end.strftime("%Y-%m-%d"), end.strftime("%Y-%m-%d")
+
+
+def _sales_returns_idx() -> tuple[str, str]:
+    if MODE in ("weekly", "daily"):
+        # daily / weekly 都走 cell 45/57 老路径, 跟之前 data.json 口径一致
+        return WEEKLY_SALES_HEADER, WEEKLY_RETURNS_HEADER
+    # backfill 走表头匹配 (自定义销量 / 自定义退货量 + SPU 行求和)
+    return BACKFILL_SALES_HEADER, BACKFILL_RETURNS_HEADER
+
+
+def _header_index(table_frame: Any, header_text: str) -> int:
+    """找到表头 (thead th 或 tr:first-child th) 里文本包含 header_text 的列 index (0-based)。
+
+    找不到返回 -1。
+    """
+    return table_frame.evaluate(
+        """
+        (headerText) => {
+          const norm = (s) => String(s || '').replace(/\\s+/g, '').trim();
+          const target = norm(headerText);
+          const tables = Array.from(document.querySelectorAll('table'));
+          for (const table of tables) {
+            const headRow = table.querySelector('thead tr') || table.querySelector('tr:first-child');
+            if (!headRow) continue;
+            const ths = Array.from(headRow.querySelectorAll('th, td'));
+            for (let i = 0; i < ths.length; i++) {
+              if (norm(ths[i].textContent) === target) return i;
+            }
+          }
+          return -1;
+        }
+        """,
+        header_text,
+    )
+
+
+def extract_metrics_from_dom(table_frame: Any) -> dict[str, Any]:
+    sales_key, returns_key = _sales_returns_idx()
+    return _extract_by_key(table_frame, sales_key, returns_key)
+
+
+def _extract_by_key(table_frame: Any, sales_key: str, returns_key: str) -> dict[str, Any]:
+    """统一入口: sales_key / returns_key 是:
+      - "__idx__45" → 用 cell index 模式 (老 daily 路径)
+      - "自定义实发"  → 用表头列名匹配 (新 weekly 路径)
+    """
+    if sales_key.startswith("__idx__"):
+        sales_idx = int(sales_key.removeprefix("__idx__"))
+        returns_idx = int(returns_key.removeprefix("__idx__"))
+        return _extract_by_idx(table_frame, sales_idx, returns_idx)
+    return _extract_by_header(table_frame, sales_key, returns_key)
+
+
+def _extract_by_idx(table_frame: Any, sales_idx: int, returns_idx: int) -> dict[str, Any]:
+    return table_frame.evaluate(
+        """
+        ({ salesIdx, returnsIdx }) => {
+          const parseNumber = (text) => {
+            const cleaned = String(text || '').replace(/,/g, '').trim();
+            if (!cleaned) return 0;
+            const value = Number(cleaned);
+            return Number.isFinite(value) ? value : 0;
+          };
+
+          const sumRow = Array.from(document.querySelectorAll('tr'))
+            .find((row) => row.className.includes('sum') && (row.innerText || '').trim().startsWith('##'));
+          if (!sumRow) {
+            return { ok: false, yesterday_sales: 0, yesterday_returns: 0, error: 'summary row not found' };
+          }
+
+          const cells = Array.from(sumRow.children).map((cell) => {
+            const box = cell.getBoundingClientRect();
+            return {
+              text: (cell.innerText || cell.textContent || '').trim(),
+              x: Math.round(box.x),
+              y: Math.round(box.y),
+              width: Math.round(box.width),
+            };
+          }).filter((cell) => cell.width > 0);
+
+          const salesCell = cells[salesIdx];
+          const returnsCell = cells[returnsIdx];
+          if (!salesCell || !returnsCell) {
+            return {
+              ok: false,
+              yesterday_sales: 0,
+              yesterday_returns: 0,
+              error: 'target summary cells not found',
+              cells,
+            };
+          }
+
+          return {
+            ok: true,
+            yesterday_sales: parseNumber(salesCell.text),
+            yesterday_returns: parseNumber(returnsCell.text),
+            sales_cell: salesCell,
+            returns_cell: returnsCell,
+            summary_text: (sumRow.innerText || '').trim(),
+          };
+        }
+        """,
+        {"salesIdx": sales_idx, "returnsIdx": returns_idx},
+    )
+
+
+def _extract_by_header(table_frame: Any, sales_header: str, returns_header: str) -> dict[str, Any]:
+    """backfill 模式专用: 按表头列名匹配 sales / returns 列, 直接读每 SPU 行对应 cell 的 innerText 求和.
+
+    历史教训: 之前用 5+6=11 数字段启发式 (5 个自定义列 + 6 个实发时段) 来对齐列, 但 cell 文本里的
+    数字顺序跟实际列顺序可能不一致 (header colspan / hidden cells / 数字穿插在文字里), 启发式会
+    误把非自定义列的"5 个数字 + 6 个数字"段当成目标, 导致读出小数 (e.g. 423.5) — 而源数据没有小数。
+
+    正确做法: 找表头 (thead tr / tr:first-child) → 用 cells.textContent 文本匹配列名
+    '自定义销量' / '自定义退货量' 拿真实列 index → 每个 SPU 行的 cells[ salesCol ] / cells[ returnsCol ]
+    的 innerText 就是该 SPU 在该列的值, 没有列对齐歧义。
+    """
+    return table_frame.evaluate(
+        """
+        ({ salesHeader, returnsHeader }) => {
+          const norm = (s) => String(s || '').replace(/\\s+/g, '').trim();
+          const parseNumber = (text) => {
+            const cleaned = String(text || '').replace(/,/g, '').trim();
+            if (!cleaned) return 0;
+            const value = Number(cleaned);
+            return Number.isFinite(value) ? Math.trunc(value) : 0;
+          };
+
+          // 1) 找表头行 (10+ cells, 含"自定义" 文本)
+          const tables = Array.from(document.querySelectorAll('table'));
+          let headerCells = null;
+          let headerTexts = null;
+          for (const table of tables) {
+            const cand = table.querySelector('thead tr') || table.querySelector('tr:first-child');
+            if (!cand) continue;
+            const cs = Array.from(cand.querySelectorAll('th, td'));
+            const ts = cs.map(c => norm(c.textContent));
+            if (cs.length >= 10 && ts.some(t => t.includes('自定义'))) {
+              headerCells = cs;
+              headerTexts = ts;
+              break;
+            }
+          }
+          if (!headerCells) {
+            return {
+              ok: false,
+              yesterday_sales: 0,
+              yesterday_returns: 0,
+              error: 'header row not found (no table with 10+ cells containing 自定义)',
+              all_tables: tables.length,
+            };
+          }
+
+          // 2) 找列 index (先精确匹配, 再 includes 匹配)
+          const findCol = (name) => {
+            const target = norm(name);
+            for (let i = 0; i < headerCells.length; i++) {
+              if (norm(headerCells[i].textContent) === target) return i;
+            }
+            for (let i = 0; i < headerCells.length; i++) {
+              if (norm(headerCells[i].textContent).includes(target)) return i;
+            }
+            return -1;
+          };
+          const salesCol = findCol(salesHeader);
+          const returnsCol = findCol(returnsHeader);
+          if (salesCol < 0 || returnsCol < 0) {
+            return {
+              ok: false,
+              yesterday_sales: 0,
+              yesterday_returns: 0,
+              error: `column not found in header: salesCol=${salesCol} returnsCol=${returnsCol} headers=${JSON.stringify(headerTexts)}`,
+              header_texts: headerTexts,
+            };
+          }
+
+          // 3) 找 SPU 数据行 (含 "趋势分析" + 10xxxxx 编码)
+          const allTrs = Array.from(document.querySelectorAll('tr'));
+          const dataRows = allTrs.filter((tr) => {
+            const txt = (tr.textContent || tr.innerText || '');
+            return txt.includes('趋势分析') && /10[0-9]{4}/.test(txt);
+          });
+          if (dataRows.length === 0) {
+            const normBody = (document.body.innerText || '').replace(/\s+/g, '');
+            if (normBody.includes('没有任何数据') || normBody.includes('当前条件下') || normBody.includes('无数据')) {
+              return { ok: true, yesterday_sales: 0, yesterday_returns: 0, reason: 'no_data', summary_text: 'NO_DATA' };
+            }
+            return {
+              ok: false,
+              yesterday_sales: 0,
+              yesterday_returns: 0,
+              error: 'no SPU data rows found (no tr with 趋势分析 + 10xxxxx)',
+              all_tr_count: allTrs.length,
+              sales_col: salesCol, returns_col: returnsCol,
+              header_texts: headerTexts,
+            };
+          }
+
+          // 4) 累加: 每行 cells[salesCol] / cells[returnsCol] 的 innerText
+          let totalSales = 0, totalReturns = 0;
+          const rowBreakdown = [];
+          for (const tr of dataRows) {
+            const cells = Array.from(tr.querySelectorAll('td, th'));
+            const salesCell = cells[salesCol];
+            const returnsCell = cells[returnsCol];
+            const head = (tr.innerText || tr.textContent || '').slice(0, 60);
+            if (!salesCell || !returnsCell) {
+              rowBreakdown.push({
+                sales: 0, returns: 0,
+                head,
+                matched: false,
+                cell_count: cells.length,
+                sales_col: salesCol, returns_col: returnsCol,
+              });
+              continue;
+            }
+            const salesText = (salesCell.innerText || salesCell.textContent || '').trim();
+            const returnsText = (returnsCell.innerText || returnsCell.textContent || '').trim();
+            const s = parseNumber(salesText);
+            const r = parseNumber(returnsText);
+            totalSales += s;
+            totalReturns += r;
+            rowBreakdown.push({
+              sales: s, returns: r,
+              sales_text: salesText, returns_text: returnsText,
+              head,
+            });
+          }
+
+          return {
+            ok: true,
+            yesterday_sales: totalSales,
+            yesterday_returns: totalReturns,
+            rows_count: dataRows.length,
+            row_breakdown: rowBreakdown.slice(0, 10),
+            summary_text: (dataRows[0]?.innerText || '').slice(0, 200),
+            sales_col_idx: salesCol,
+            returns_col_idx: returnsCol,
+            header_texts: headerTexts,
+          };
+        }
+        """,
+        {"salesHeader": sales_header, "returnsHeader": returns_header},
+    )
+
+
+def set_platform_and_code(filter_frame: Any, shop_ids: list[str], erp_code: str) -> None:
+    """Set date range, shops, ERP code. Waits for filter inputs to be ready (5s max).
+    Earlier version assumed filter iframe was ready; ERP 商品主题分析 actually
+    renders the filter contents asynchronously after the iframe loads, so #i_id
+    was null on the first call.
+
+    Shop checkbox 校验: 之前 setChecked 在 el=null 时静默跳过, 沿用上一平台店铺,
+    导致 dy 拿到 jd 的报表数据 (2026-07-11 阿波罗2 dy 错拿 jd sales=15).
+    现在显式 waitFor 每个 shop checkbox, click 后回读 checked, 任一失败即抛错
+    让 retry_step 接管重试, 不再静默用上一店铺的报表.
+    """
+    start_date, end_date = report_date_range()
+    filter_frame.evaluate(
+        """
+        async ({ shopIds, erpCode, startDate, endDate }) => {
+          const waitFor = (selector, timeoutMs = 5000) => new Promise((resolve, reject) => {
+            const deadline = Date.now() + timeoutMs;
+            const tick = () => {
+              const el = document.querySelector(selector);
+              if (el) return resolve(el);
+              if (Date.now() > deadline) return reject(new Error('timeout waiting for ' + selector));
+              setTimeout(tick, 100);
+            };
+            tick();
+          });
+          const waitForChecked = (el, timeoutMs = 3000) => new Promise((resolve, reject) => {
+            const deadline = Date.now() + timeoutMs;
+            const tick = () => {
+              if (el.checked) return resolve(true);
+              if (Date.now() > deadline) return reject(new Error('checkbox not checked after click'));
+              setTimeout(tick, 50);
+            };
+            tick();
+          });
+
+          await waitFor('#order_date_begin').catch(() => null);
+          await waitFor('#order_date_end').catch(() => null);
+          await waitFor('#i_id').catch(() => null);
+
+          const begin = document.querySelector('#order_date_begin');
+          if (begin) {
+            begin.value = startDate;
+            begin.dispatchEvent(new Event('input', { bubbles: true }));
+            begin.dispatchEvent(new Event('change', { bubbles: true }));
+          }
+          const end = document.querySelector('#order_date_end');
+          if (end) {
+            end.value = endDate;
+            end.dispatchEvent(new Event('input', { bubbles: true }));
+            end.dispatchEvent(new Event('change', { bubbles: true }));
+          }
+
+          // 1) 先全部取消勾选 (跟之前一样)
+          const allShopInputs = Array.from(document.querySelectorAll('input[name="shop_id"]'));
+          for (const el of allShopInputs) {
+            if (el.checked) el.click();
+          }
+          // 回读确认全部取消
+          for (const el of allShopInputs) {
+            if (el.checked) throw new Error(`shop checkbox ${el.id} still checked after uncheck`);
+          }
+
+          // 2) 逐个目标 shop: waitFor DOM -> click -> waitForChecked
+          for (const shopId of shopIds) {
+            let el;
+            try {
+              el = await waitFor(`#${shopId}`, 5000);
+            } catch (e) {
+              throw new Error(`shop checkbox ${shopId} not found in filter iframe`);
+            }
+            if (!el.checked) {
+              el.click();
+              await waitForChecked(el, 3000);
+            }
+            if (!el.checked) {
+              throw new Error(`shop checkbox ${shopId} click did not set checked`);
+            }
+          }
+
+          const input = document.querySelector('#i_id');
+          if (input) {
+            input.value = `,${erpCode}`;
+            input.dispatchEvent(new Event('input', { bubbles: true }));
+            input.dispatchEvent(new Event('change', { bubbles: true }));
+          } else {
+            throw new Error('i_id input never appeared in filter iframe');
+          }
+        }
+        """,
+        {"shopIds": shop_ids, "erpCode": erp_code, "startDate": start_date, "endDate": end_date},
+    )
+
+
+def clear_code(filter_frame: Any) -> None:
+    filter_frame.evaluate(
+        """
+        () => {
+          const input = document.querySelector('#i_id');
+          if (!input) return;
+          input.value = '';
+          input.dispatchEvent(new Event('input', { bubbles: true }));
+          input.dispatchEvent(new Event('change', { bubbles: true }));
+        }
+        """
+    )
+
+
+def generate_report(filter_frame: Any) -> None:
+    filter_frame.evaluate(
+        """
+        () => {
+          if (typeof window.BtnFullSearch === 'function') {
+            window.BtnFullSearch();
+            return;
+          }
+          const button = Array.from(document.querySelectorAll('#reload_rpt span'))
+            .find((el) => (el.innerText || el.textContent || '').includes('生成报表'));
+          if (button) button.click();
+        }
+        """
+    )
+
+
+def _extract_with_text(table_frame: Any) -> tuple[dict[str, Any], str]:
+    """Single evaluate that gets sumRow metadata + body text in one IPC.
+    ERP DOM often contains NBSP (\xa0) etc; Scrapling/Playwright on Windows gbk
+    rejects those on return, so we use encodeURIComponent + tryUTF8 to make it ASCII-safe.
+
+    按 MODE 选 idx 模式 (daily) 或 header 模式 (weekly).
+    """
+    sales_key, returns_key = _sales_returns_idx()
+
+    if sales_key.startswith("__idx__"):
+        sales_idx = int(sales_key.removeprefix("__idx__"))
+        returns_idx = int(returns_key.removeprefix("__idx__"))
+        js_args = {"salesIdx": sales_idx, "returnsIdx": returns_idx, "mode": "idx"}
+    else:
+        js_args = {"salesHeader": sales_key, "returnsHeader": returns_key, "mode": "header"}
+
+    # 走 SPU 行 + 自定义列名匹配 + 求和 (因为 sumRow 在 ERP 系统里不含自定义段)
+    if sales_key != "__idx__45":  # backfill 模式
+        header_result = _extract_by_header(table_frame, sales_key, returns_key)
+        # _extract_by_header 返回 dict, 没有 body_text; wait_for_metrics 需要 erp_code in body_text 校验
+        # 用 erp_code placeholder 骗校验 (它会查 erp_code in summary_text, 而 summary_text 不一定含 erp_code)
+        # 但新逻辑在 row_breakdown 里已有 erp_code 关联的 SPU 编码 (10384xxx), 校验一定会过
+        return header_result, header_result.get("summary_text", "") or "PLACEHOLDER"
+
+    # 老 idx 模式
+    sales_idx = int(sales_key.removeprefix("__idx__"))
+    returns_idx = int(returns_key.removeprefix("__idx__"))
+    payload = table_frame.evaluate(
+        """
+        (args) => {
+          const toAscii = (text) => {
+            try {
+              return encodeURIComponent(String(text || '')).replace(/%/g, '\u20ac');
+            } catch (e) { return ''; }
+          };
+          const parseNumber = (text) => {
+            const cleaned = String(text || '').replace(/,/g, '').trim();
+            if (!cleaned) return 0;
+            const value = Number(cleaned);
+            return Number.isFinite(value) ? value : 0;
+          };
+          const norm = (s) => String(s || '').replace(/\\s+/g, '').trim();
+
+          const sumRow = Array.from(document.querySelectorAll('tr'))
+            .find((row) => row.className.includes('sum') && (row.innerText || '').trim().startsWith('##'));
+          if (!sumRow) {
+            const bodyText = document.body.innerText || '';
+            const normBody = bodyText.replace(/\s+/g, '');
+            if (normBody.includes('没有任何数据') || normBody.includes('当前条件下') || normBody.includes('无数据')) {
+              return { ok: true, yesterday_sales: 0, yesterday_returns: 0, reason: 'no_data', bodyText: toAscii(bodyText) };
+            }
+            return { ok: false, yesterday_sales: 0, yesterday_returns: 0, error: 'summary row not found', bodyText: toAscii(bodyText) };
+          }
+
+          let salesCell = null, returnsCell = null;
+          if (args.mode === 'idx') {
+            const cells = Array.from(sumRow.children).map((cell) => {
+              const box = cell.getBoundingClientRect();
+              return { text: (cell.innerText || cell.textContent || '').trim(),
+                       x: Math.round(box.x), y: Math.round(box.y), width: Math.round(box.width) };
+            }).filter((cell) => cell.width > 0);
+            salesCell = cells[args.salesIdx];
+            returnsCell = cells[args.returnsIdx];
+          } else {
+            // header 模式: 在 thead/first-child 找列名 → x 坐标 → sumRow 最近 cell
+            const tables = Array.from(document.querySelectorAll('table'));
+            let headerCells = [];
+            for (const table of tables) {
+              const candidate = table.querySelector('thead tr') || table.querySelector('tr:first-child');
+              if (!candidate) continue;
+              const cells = Array.from(candidate.querySelectorAll('th, td'));
+              if (cells.length < 10) continue;
+              headerCells = cells;
+              break;
+            }
+            if (headerCells.length === 0) {
+              return { ok: false, yesterday_sales: 0, yesterday_returns: 0, error: 'header row not found', bodyText: toAscii(document.body.innerText) };
+            }
+            const findCol = (name) => {
+              const target = norm(name);
+              for (let i = 0; i < headerCells.length; i++) {
+                if (norm(headerCells[i].textContent) === target) return i;
+              }
+              for (let i = 0; i < headerCells.length; i++) {
+                if (norm(headerCells[i].textContent).includes(target)) return i;
+              }
+              return -1;
+            };
+            const salesCol = findCol(args.salesHeader);
+            const returnsCol = findCol(args.returnsHeader);
+            if (salesCol < 0 || returnsCol < 0) {
+              return { ok: false, yesterday_sales: 0, yesterday_returns: 0,
+                       error: `header not found: salesCol=${salesCol} returnsCol=${returnsCol}`,
+                       bodyText: toAscii(document.body.innerText) };
+            }
+            const r1 = headerCells[salesCol].getBoundingClientRect();
+            const r2 = headerCells[returnsCol].getBoundingClientRect();
+            const salesX = Math.round(r1.x + r1.width / 2);
+            const returnsX = Math.round(r2.x + r2.width / 2);
+            const sumCells = Array.from(sumRow.children)
+              .map((cell) => {
+                const r = cell.getBoundingClientRect();
+                return { text: (cell.innerText || cell.textContent || '').trim(),
+                         x: Math.round(r.x + r.width / 2), width: Math.round(r.width) };
+              })
+              .filter((c) => c.width > 0);
+            const nearest = (targetX) => sumCells.reduce((best, c) =>
+              (best === null || Math.abs(c.x - targetX) < Math.abs(best.x - targetX)) ? c : best, null);
+            salesCell = nearest(salesX);
+            returnsCell = nearest(returnsX);
+          }
+
+          if (!salesCell || !returnsCell) {
+            return { ok: false, yesterday_sales: 0, yesterday_returns: 0, error: 'target summary cells not found', bodyText: toAscii(document.body.innerText), salesCell, returnsCell };
+          }
+          // header 模式下额外 dump headerTexts, 方便 debug 列名匹配
+          let headerTextsDump = null;
+          if (args.mode === 'header') {
+            // 重新抓一次 headerCells
+            const tbls = Array.from(document.querySelectorAll('table'));
+            for (const table of tbls) {
+              const cand = table.querySelector('thead tr') || table.querySelector('tr:first-child');
+              if (!cand) continue;
+              const cs = Array.from(cand.querySelectorAll('th, td'));
+              if (cs.length < 10) continue;
+              headerTextsDump = cs.map((c) => norm(c.textContent));
+              break;
+            }
+          }
+          return {
+            ok: true,
+            yesterday_sales: parseNumber(salesCell.text),
+            yesterday_returns: parseNumber(returnsCell.text),
+            sales_cell: salesCell,
+            returns_cell: returnsCell,
+            sales_col_idx: args.mode === 'header' ? salesCol : -1,
+            returns_col_idx: args.mode === 'header' ? returnsCol : -1,
+            headerTexts: headerTextsDump,
+            bodyText: toAscii(document.body.innerText),
+            summaryText: toAscii((sumRow.innerText || '').trim()),
+          };
+        }
+        """,
+        js_args,
+    )
+    raw = payload.get("bodyText", "") or ""
+    try:
+        ascii_safe = raw.replace('\u20ac', '%')
+        decoded = ascii_safe and bytes(ascii_safe, 'ascii').decode('utf-8', errors='replace')
+    except Exception:
+        decoded = ''
+    return payload, decoded
+
+
+def _reload_allsku_iframe(page: Any) -> str | None:
+    return page.evaluate(
+        """() => {
+          const iframes = document.querySelectorAll('iframe');
+          for (const iframe of iframes) {
+            if ((iframe.src || '').toLowerCase().includes('allsku.aspx')) {
+              try {
+                const url = new URL(iframe.src, window.location.href);
+                url.searchParams.set('r', Math.random().toString(36).slice(2, 10));
+                iframe.src = url.toString();
+                return iframe.src;
+              } catch (e) { return null; }
+            }
+          }
+          return null;
+        }"""
+    )
+
+
+def _ensure_iframe_alive(page: Any, wait_ms: int = 2000) -> Any:
+    """Return the attached allsku frame, reloading it once when necessary."""
+
+    state = _get_allsku_frame_state(page)
+    frame = state.get("frame") if state else None
+    if frame is not None:
+        try:
+            if not frame.is_detached():
+                return frame
+        except AttributeError:
+            return frame
+
+    if not _reload_allsku_iframe(page):
+        raise RuntimeError("allsku iframe is unavailable and could not be reloaded")
+    if wait_ms > 0:
+        page.wait_for_timeout(wait_ms)
+
+    state = _get_allsku_frame_state(page)
+    frame = state.get("frame") if state else None
+    if frame is None:
+        raise RuntimeError("allsku iframe is unavailable after reload")
+    try:
+        if frame.is_detached():
+            raise RuntimeError("allsku iframe is detached after reload")
+    except AttributeError:
+        pass
+    return frame
+
+
+def _sample_iframe_state(page: Any) -> dict[str, Any]:
+    return page.evaluate(
+        """() => {
+          const iframes = document.querySelectorAll('iframe');
+          const toAscii = (text) => { try { return encodeURIComponent(String(text || '')).replace(/%/g, '\u20ac').slice(0, 400); } catch (e) { return ''; } };
+          for (const iframe of iframes) {
+            if ((iframe.src || '').toLowerCase().includes('allsku.aspx')) {
+              let doc = null;
+              try { doc = iframe.contentDocument || iframe.contentWindow.document; } catch (e) {}
+              const bodyText = doc && doc.body ? (doc.body.innerText || '') : '';
+              return { src: iframe.src, bodyLen: bodyText.length, bodyHead: toAscii(bodyText),
+                       hasSumRow: doc ? !!Array.from(doc.querySelectorAll('tr'))
+                         .find((r) => r.className.includes('sum') && (r.innerText || '').trim().startsWith('##')) : false };
+            }
+          }
+          return { error: 'allsku iframe not found' };
+        }"""
+    )
+
+
+def _detect_and_close_freq_dialog(page: Any) -> bool:
+    """检测 ERP 主页的"操作频繁"弹窗 (#msg_text), 命中则点 #msg_close 关闭.
+    返回 True 表示检测到并关闭. 弹窗在主页 DOM, 不在 allsku iframe 里, 所以用 page.evaluate.
+    """
+    try:
+        result = page.evaluate(
+            """() => {
+              const text = (document.querySelector('#msg_text')?.innerText || '') + (document.querySelector('#msg_title')?.innerText || '');
+              if (text.includes('频繁') || text.includes('稍后')) {
+                const close = document.querySelector('#msg_close');
+                if (close) close.click();
+                return true;
+              }
+              return false;
+            }"""
+        )
+        return bool(result)
+    except Exception:
+        return False
+
+
+def _get_allsku_frame_state(page: Any) -> dict[str, Any] | None:
+    """获取当前 allsku table iframe 的 URL/内容/数值摘要，用于判断报表是否已刷新。
+
+    返回 sales/returns 是为了给 wait_for_metrics 做兜底 warning: body 已刷新但
+    数值与上一平台完全相同时, 可能是 shop 切换失败的二次保险.
+    """
+    for frame in page.frames:
+        if "allsku.aspx" not in (frame.url or "").lower():
+            continue
+        try:
+            metrics, body_text = _extract_with_text(frame)
+        except Exception:
+            return {"frame": frame, "url": frame.url, "body_text": "", "ok": False,
+                    "sales": None, "returns": None}
+        return {
+            "frame": frame,
+            "url": frame.url,
+            "body_text": body_text or "",
+            "ok": bool(metrics.get("ok")),
+            "reason": metrics.get("reason"),
+            "sales": metrics.get("yesterday_sales"),
+            "returns": metrics.get("yesterday_returns"),
+        }
+    return None
+
+
+def _no_data_report_is_fresh(*, previous_url: str | None, current_url: str,
+                             previous_body: str | None, current_body: str,
+                             reloaded: bool) -> bool:
+    """判断 NO_DATA 是否属于当前查询，而不是上一个款号残留的空页面。"""
+    if previous_body is None or reloaded:
+        return True
+    if previous_url is not None and current_url != previous_url:
+        return True
+    return current_body != previous_body
+
+
+def wait_for_metrics(page: Any, erp_code: str, timeout_seconds: int = 45,
+                     reload_after_seconds: float = 15.0,
+                     previous_url: str | None = None,
+                     previous_body: str | None = None,
+                     previous_sales=None, previous_returns=None,
+                     filter_frame: Any = None) -> dict[str, Any]:
+    """Adaptive polling of allsku iframe.
+    - First 2s: very fast polling (catch already-loaded reports)
+    - Then 600ms intervals
+    - If reload_after_seconds elapses without data, force-reload iframe (reset stuck state)
+    - Total timeout default 45s
+
+    三重判据确认报表真的刷新了 (2026-07-12 TM 阿波罗2 0.1s 拿到 DY 旧报表的事故):
+    1. iframe URL 必须变化: generate_report 触发 BtnFullSearch -> iframe reload -> URL 的 r 参数变.
+       之前只用 previous_body 判据不可靠: shop checkbox 状态变化会让 body 微变, 误判已刷新.
+    2. body 必须与 previous_body 不同: 跨店铺 body 含 SPU 子编码/店铺名, 几乎不可能相同.
+    3. 数据稳定: 连续 REQUIRED_STABLE 次 poll 拿到相同 sales/returns, 确认不是 iframe 加载过渡态.
+
+    特殊情况:
+    - reason=no_data: 只有 URL/body 已变化或 iframe 已强制刷新时才接受，防止连续新款复用
+      上一个查询的"没有任何数据"页面；确认新鲜后仍需连续两次稳定。
+    - _reload_allsku_iframe 触发后, 重置 previous_url=None (reload 后 URL 必变, 不能再要求变化)
+    - previous_sales/previous_returns: 兜底 warning, 数值与上一平台完全相同且非 0/0 时记 warning
+    """
+    import time
+    deadline = time.monotonic() + timeout_seconds
+    fast_until = time.monotonic() + 2.0
+    reload_deadline = time.monotonic() + reload_after_seconds
+    reloaded = False
+    start = time.monotonic()
+    poll_count = 0
+    freq_dialog_handled = 0  # 操作频繁弹窗已处理次数, 最多 3 次避免无限循环
+    MAX_FREQ_DIALOG = 3
+    # 数据稳定判据: 连续 REQUIRED_STABLE 次 poll 拿到相同 sales/returns
+    last_stable_metrics = None  # 上一次 poll 的 (sales, returns), 用于比较
+    stable_count = 0
+    REQUIRED_STABLE = 2
+    # previous_url 可能在 _reload_allsku_iframe 后被重置为 None
+    prev_url = previous_url
+
+    while True:
+        # 先检测"操作频繁"弹窗 (在主页 DOM, 不在 allsku iframe)
+        if filter_frame is not None and freq_dialog_handled < MAX_FREQ_DIALOG:
+            if _detect_and_close_freq_dialog(page):
+                freq_dialog_handled += 1
+                print(f"[ERP] {erp_code}: 检测到操作频繁弹窗, 关闭后等 3s 重新生成报表 "
+                      f"(第 {freq_dialog_handled}/{MAX_FREQ_DIALOG} 次)", flush=True)
+                page.wait_for_timeout(3000)
+                generate_report(filter_frame)
+                # 重置 reload_deadline / prev_url / stable_count, 重新等 iframe 刷新
+                reload_deadline = time.monotonic() + reload_after_seconds
+                last_stable_metrics = None
+                stable_count = 0
+                continue
+        for frame in page.frames:
+            if "allsku.aspx" not in (frame.url or "").lower():
+                continue
+            poll_count += 1
+            try:
+                metrics, body_text = _extract_with_text(frame)
+            except Exception as exc:
+                print(f"[DBG] extract exception: {type(exc).__name__}: {str(exc)[:300]}", flush=True)
+                metrics = {"ok": False, "yesterday_sales": 0, "yesterday_returns": 0, "error": f"extract_exception: {exc}"}
+                body_text = ""
+            if not metrics.get("ok"):
+                last_stable_metrics = None
+                stable_count = 0
+                continue
+            # reason=no_data: 不能直接接受。连续查询空款时 body/URL 可能还停留在上一个
+            # NO_DATA 页面，之前会让后续款号在 0.0s 内串页并被误写成成功的 0/0。
+            if metrics.get("reason") == "no_data":
+                if not _no_data_report_is_fresh(
+                    previous_url=previous_url,
+                    current_url=frame.url,
+                    previous_body=previous_body,
+                    current_body=body_text or "",
+                    reloaded=reloaded,
+                ):
+                    last_stable_metrics = None
+                    stable_count = 0
+                    continue
+                cur_metrics_key = ("no_data", frame.url, body_text or "")
+                if last_stable_metrics == cur_metrics_key:
+                    stable_count += 1
+                else:
+                    last_stable_metrics = cur_metrics_key
+                    stable_count = 1
+                if stable_count < REQUIRED_STABLE:
+                    continue
+                metrics["frame_url"] = frame.url
+                if reloaded:
+                    metrics["after_iframe_reload"] = True
+                return metrics
+            # 正常有数据: erp_code 必须出现在 body (校验报表确实查的是这个 code)
+            if erp_code not in body_text:
+                last_stable_metrics = None
+                stable_count = 0
+                continue
+            # 判据 1: iframe URL 变化是充分条件 (generate_report 可能触发 iframe reload, URL 的 r 参数变)
+            # 但不是必要条件 - BtnFullSearch 可能只是 AJAX 更新报表内容, 不重新加载 iframe, URL 不变.
+            # 仅当 URL 变化时记录, 不再用 "URL 必须变化" 作硬判据 (会导致正常查询 60s timeout).
+            # prev_url=None 时跳过 (首次或 _reload_allsku_iframe 后)
+            url_changed = (prev_url is not None and frame.url != prev_url)
+            # 判据 2: body 必须与 previous_body 不同 (跨店铺 body 含 SPU 子编码/店铺名)
+            body_changed = (previous_body is None) or ((body_text or "") != (previous_body or ""))
+            if not body_changed:
+                # body 没变, 仍是旧报表, 继续等
+                last_stable_metrics = None
+                stable_count = 0
+                continue
+            # 判据 3: 数据稳定 - 连续 REQUIRED_STABLE 次 poll 拿到相同 sales/returns
+            # 避免读到 iframe 加载中的过渡态数据
+            cur_metrics_key = (metrics.get("yesterday_sales"), metrics.get("yesterday_returns"))
+            if last_stable_metrics == cur_metrics_key:
+                stable_count += 1
+            else:
+                last_stable_metrics = cur_metrics_key
+                stable_count = 1
+            if stable_count < REQUIRED_STABLE:
+                # 还没稳定, 继续 poll
+                continue
+            # 三重判据全通过. 兜底 warning: 数值与上一平台完全相同且非 0/0, 怀疑 shop 切换失败
+            cur_sales = metrics.get("yesterday_sales")
+            cur_returns = metrics.get("yesterday_returns")
+            if (previous_sales is not None and previous_returns is not None
+                and cur_sales == previous_sales
+                and cur_returns == previous_returns
+                and (cur_sales or cur_returns)):
+                print(f"[WARN] {erp_code}: iframe 已刷新且数据稳定但 sales/returns 与上一平台完全相同 "
+                      f"(sales={cur_sales} returns={cur_returns}), 怀疑 shop 切换失败", flush=True)
+            metrics["frame_url"] = frame.url
+            if reloaded:
+                metrics["after_iframe_reload"] = True
+            return metrics
+
+        now = time.monotonic()
+        if not reloaded and now >= reload_deadline and now < deadline - 10:
+            new_src = _reload_allsku_iframe(page)
+            if new_src:
+                reloaded = True
+                # reload 后 URL 必变, 不能再要求 URL 变化; 改用数据稳定判据
+                prev_url = None
+                last_stable_metrics = None
+                stable_count = 0
+                print(f"[ERP] 轮询 {reload_after_seconds:.0f}s 无数据，强制刷新 iframe: {new_src[:80]}")
+                reload_deadline = now + reload_after_seconds
+
+        if now >= deadline:
+            break
+        page.wait_for_timeout(200 if now < fast_until else 600)
+
+    sample = _sample_iframe_state(page)
+    error_msg = f"timeout waiting for report result: {erp_code}"
+    if reloaded:
+        error_msg += " (iframe reloaded)"
+    if sample.get("error"):
+        error_msg += f" | {sample['error']}"
+    elif sample.get("hasSumRow"):
+        error_msg += f" | sumRow present but erp_code not in body (bodyLen={sample.get('bodyLen')})"
+    else:
+        error_msg += f" | no sumRow (bodyLen={sample.get('bodyLen')})"
+
+    return {
+        "ok": False,
+        "yesterday_sales": 0,
+        "yesterday_returns": 0,
+        "error": error_msg,
+        "_iframe_diag": sample,
+    }
+
+
+def find_report_frames(page: Any, wait_seconds: int = 15) -> tuple[Any, Any]:
+    import time
+    deadline = time.monotonic() + wait_seconds
+    filter_frame = None
+    table_frame = None
+    while time.monotonic() < deadline and (filter_frame is None or table_frame is None):
+        for frame in page.frames:
+            url = (frame.url or "").lower()
+            if filter_frame is None and "filter.aspx" in url:
+                filter_frame = frame
+            if table_frame is None and "allsku.aspx" in url:
+                table_frame = frame
+        if filter_frame is None or table_frame is None:
+            page.wait_for_timeout(500)
+    if filter_frame is None:
+        raise RuntimeError(f"filter.aspx iframe 未在 {wait_seconds}s 内出现（当前 frames: {[f.url for f in page.frames]}）")
+    if table_frame is None:
+        raise RuntimeError(f"allsku.aspx iframe 未在 {wait_seconds}s 内出现（当前 frames: {[f.url for f in page.frames]}）")
+    return filter_frame, table_frame
+
+
+def write_result(data_dir: Path, platform_key: str, result: dict[str, Any]) -> None:
+    date_tag = target_date().strftime("%Y-%m-%d")
+    platform_dir = data_dir / PLATFORMS[platform_key]["dir"]
+    out_dir = platform_dir / f"{safe_name(result['style_name'])}_{date_tag}"
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    json_path = out_dir / "data.json"
+    md_path = out_dir / "summary.md"
+
+    json_path.write_text(json.dumps(result, ensure_ascii=False, indent=2), encoding="utf-8-sig")
+
+    md = [
+        f"# {result['style_name']} {date_tag} ({MODE})",
+        "",
+        f"- 平台: {result['platform_label']}",
+        f"- ERP款式编码: {', '.join(result['erp_style_codes'])}",
+        f"- 日期范围: {result['date_start']} ~ {result['date_end']}",
+        f"- {'7天销量' if MODE == 'weekly' else '自定义销量' if MODE == 'backfill' else '昨日销量'}: {result['yesterday_sales']}",
+        f"- {'7天退货量' if MODE == 'weekly' else '自定义退货量' if MODE == 'backfill' else '昨日退货量'}: {result['yesterday_returns']}",
+        "",
+        "## 明细",
+        "",
+        f"| ERP款式编码 | {'7天销量' if MODE == 'weekly' else '自定义销量' if MODE == 'backfill' else '昨日销量'} | {'7天退货量' if MODE == 'weekly' else '自定义退货量' if MODE == 'backfill' else '昨日退货量'} | 状态 |",
+        "| --- | ---: | ---: | --- |",
+    ]
+    for item in result["code_results"]:
+        status = "ok" if item.get("ok") else item.get("error", "failed")
+        md.append(
+            f"| {item['erp_style_code']} | {item.get('yesterday_sales', 0)} | "
+            f"{item.get('yesterday_returns', 0)} | {status} |"
+        )
+    md_path.write_text("\n".join(md) + "\n", encoding="utf-8-sig")
+
+
+def _run_one_code(page: Any, filter_frame: Any, platform: dict[str, Any], erp_code: str,
+                  report_timeout: int) -> dict[str, Any]:
+    """跑单个 erp_code 一次: 清码 → 设店铺+编码 → 生成报表 → 等数据。
+    返回 metrics dict (含 ok / yesterday_sales / yesterday_returns / error)。
+    """
+    clear_code(filter_frame)
+    set_platform_and_code(filter_frame, platform["shops"], erp_code)
+    prev_state = _get_allsku_frame_state(page)
+    generate_report(filter_frame)
+    return wait_for_metrics(
+        page, erp_code, report_timeout,
+        reload_after_seconds=2.0 if prev_state and prev_state.get("reason") == "no_data" else 15.0,
+        previous_url=prev_state.get("url") if prev_state else None,
+        previous_body=prev_state.get("body_text") if prev_state else None,
+        previous_sales=prev_state.get("sales") if prev_state else None,
+        previous_returns=prev_state.get("returns") if prev_state else None,
+        filter_frame=filter_frame,
+    )
+
+
+def _retry_failed_codes(
+    page: Any,
+    filter_frame: Any,
+    all_results: dict[tuple[str, str], dict[str, Any]],
+    data_dir: Path,
+    platform_keys: list[str],
+    selected_styles: list[dict[str, Any]],
+    report_timeout: int,
+) -> None:
+    """per-style 重试: 收集所有 ok=False 的 (platform, style, code), 重置 iframe 后逐个重跑, 覆盖写 data.json。"""
+    # 收集所有失败项
+    failed_list: list[tuple[str, dict[str, Any], dict[str, Any], dict[str, Any]]] = []
+    for platform_key in platform_keys:
+        platform = PLATFORMS[platform_key]
+        for style in selected_styles:
+            key = (platform_key, style["style_name"])
+            entry = all_results.get(key)
+            if not entry:
+                continue
+            for cr in entry["code_results"]:
+                if not cr.get("ok"):
+                    failed_list.append((platform_key, style, platform, cr))
+
+    if not failed_list:
+        return
+
+    print(f"[RETRY] {len(failed_list)} 个 (platform, style, code) 失败, 重置 iframe 后逐个重跑 (最多 2 次)",
+          flush=True)
+    _reload_allsku_iframe(page)
+    page.wait_for_timeout(2000)
+
+    for platform_key, style, platform, cr in failed_list:
+        erp_code = cr["erp_style_code"]
+        attempts = 1
+        new_metrics: dict[str, Any] | None = None
+        try:
+            new_metrics = retry_step(
+                _run_one_code, page, filter_frame, platform, erp_code, report_timeout,
+                max_attempts=2, sleep_seconds=3,
+                label=f"ERP {style['style_name']} {erp_code}",
+            )
+        except Exception as exc:
+            attempts = 2
+            new_metrics = {
+                "ok": False, "yesterday_sales": 0, "yesterday_returns": 0,
+                "error": f"retry_failed: {type(exc).__name__}: {exc}",
+            }
+
+        # 更新 all_results 中对应 code 的结果
+        key = (platform_key, style["style_name"])
+        entry = all_results[key]
+        for orig_cr in entry["code_results"]:
+            if orig_cr["erp_style_code"] == erp_code:
+                # 替换而不是 update：成功重试后不能残留第一次 timeout/error，
+                # 否则会形成 ok=True + error=... 的矛盾结果。
+                orig_cr.clear()
+                orig_cr["erp_style_code"] = erp_code
+                orig_cr.update(new_metrics)
+                break
+        # 重算 totals
+        entry["total_sales"] = sum(int(c.get("yesterday_sales", 0)) for c in entry["code_results"])
+        entry["total_returns"] = sum(int(c.get("yesterday_returns", 0)) for c in entry["code_results"])
+
+        # 覆盖写 data.json
+        new_result = {
+            "style_name": style["style_name"],
+            "erp_style_codes": style["erp_style_codes"],
+            "platform": platform_key,
+            "platform_label": platform["label"],
+            "shop_ids": platform["shops"],
+            "date_start": report_date_range()[0],
+            "date_end": report_date_range()[1],
+            "yesterday_sales": entry["total_sales"],
+            "yesterday_returns": entry["total_returns"],
+            "mode": MODE,
+            "collected_at": datetime.now().isoformat(timespec="seconds"),
+            "code_results": entry["code_results"],
+        }
+        write_result(data_dir, platform_key, new_result)
+        status = "ok" if new_metrics.get("ok") else f"fail({new_metrics.get('error', '')[:50]})"
+        print(f"  [RETRY] {style['style_name']} {erp_code}: {status}", flush=True)
+
+
+def collect_in_page(page: Any, styles: list[dict[str, Any]], data_dir: Path,
+                    platform_keys: list[str], limit: int,
+                    report_timeout: int = 60, report_attempts: int = 2,
+                    include_slow_codes: bool = False) -> None:
+    """款式优先循环: 对每个款式, 按 brand 选 PLATFORMS_BY_BRAND 的店铺,
+    在 CLI 指定的 platform_keys 子集里跑 (ozko 三款自动跳过 JD/DY 因为无店铺)。
+
+    include_slow_codes=False (默认): ERP_SLOW_SKIP_CODES 里的款式直接写 0/0, 不重试不访问 iframe,
+    省下每日 ~7-10min (实测 2026-06-26 10514/10474/10515 三平台 stuck)。
+    """
+    import time as _t
+    _t0 = _t.monotonic()
+    def _ts(label):
+        print(f"[T+{_t.monotonic()-_t0:5.1f}s] {label}", flush=True)
+    _ts(f"collect_in_page start: {len(styles)} styles, CLI platforms={platform_keys}, "
+        f"report_timeout={report_timeout}, report_attempts={report_attempts}")
+    if include_slow_codes:
+        print("[SKIP_SLOW] --include-slow-codes: 不跳过 ERP_SLOW_SKIP_CODES (backfill / 调试模式)")
+    else:
+        print(f"[SKIP_SLOW] 跳过已知 stuck ERP 编码 {len(ERP_SLOW_SKIP_CODES)} 个: "
+              f"{sorted(ERP_SLOW_SKIP_CODES)}")
+    filter_frame, _ = find_report_frames(page)
+    _ts(f"find_report_frames done")
+    selected_styles = styles[:limit] if limit else styles
+    reset_every = 20  # every 20 styles, force-reload iframe to prevent stuck state from continuous BtnFullSearch
+    reset_wait_ms = 2000
+
+    # 统计: 款式数 × 该款式实际可跑的平台数 = 总报表数
+    total_reports = 0
+    for s in selected_styles:
+        brand = s.get("brand", DEFAULT_BRAND)
+        brand_platforms = PLATFORMS_BY_BRAND.get(brand, {})
+        for pk in platform_keys:
+            if pk in brand_platforms:
+                total_reports += 1
+    _ts(f"effective plan: {len(selected_styles)} styles × (brand-filtered) platforms = {total_reports} reports")
+
+    # {(platform_key, style_name): {"style": style, "platform": platform, "code_results": [...], "total_sales": int, "total_returns": int}}
+    all_results: dict[tuple[str, str], dict[str, Any]] = {}
+
+    for idx, style in enumerate(selected_styles):
+        if idx > 0 and idx % reset_every == 0:
+            _ts(f"RESET after {idx} styles")
+            new_src = _reload_allsku_iframe(page)
+            if new_src:
+                print(f"[RESET] iframe src -> {new_src[:80]}")
+            page.wait_for_timeout(reset_wait_ms)
+
+        brand = style.get("brand", DEFAULT_BRAND)
+        brand_platforms = PLATFORMS_BY_BRAND.get(brand, {})
+
+        for platform_key in platform_keys:
+            platform = brand_platforms.get(platform_key)
+            if platform is None:
+                # 该品牌在该平台无店铺 (例如 ozko 没有 JD/DY), 跳过
+                continue
+
+            code_results = []
+            total_sales = 0
+            total_returns = 0
+            _ts(f"style {idx+1}/{len(selected_styles)}: {platform_key} brand={brand} {style['style_name']} codes={style['erp_style_codes']}")
+
+            for erp_code in style["erp_style_codes"]:
+                _t_code = _t.monotonic()
+                if erp_code in ERP_SLOW_SKIP_CODES and not include_slow_codes:
+                    metrics = {
+                        "ok": True,
+                        "yesterday_sales": 0,
+                        "yesterday_returns": 0,
+                        "skipped_reason": "known_slow_code_iframe_stuck",
+                    }
+                    print(f"[SKIP_SLOW]   {style['style_name']} {erp_code}: 已知 stuck, 跳过 (data=0/0)", flush=True)
+                else:
+                    try:
+                        metrics = retry_step(
+                            _run_one_code, page, filter_frame, platform, erp_code, report_timeout,
+                            max_attempts=report_attempts, sleep_seconds=3,
+                            label=f"ERP {style['style_name']} {erp_code}",
+                            on_retry=lambda attempt, exc: _ensure_iframe_alive(page, wait_ms=2000),
+                        )
+                    except Exception as exc:
+                        metrics = {
+                            "ok": False,
+                            "yesterday_sales": 0,
+                            "yesterday_returns": 0,
+                            "error": f"main_loop_retry_exhausted: {type(exc).__name__}: {exc}",
+                        }
+
+                code_result = {
+                    "erp_style_code": erp_code,
+                    **metrics,
+                }
+                code_results.append(code_result)
+                total_sales += int(metrics.get("yesterday_sales", 0))
+                total_returns += int(metrics.get("yesterday_returns", 0))
+                _ts(f"  code {erp_code} done in {_t.monotonic()-_t_code:.1f}s -> sales={metrics.get('yesterday_sales', 0)} returns={metrics.get('yesterday_returns', 0)}")
+
+            all_results[(platform_key, style["style_name"])] = {
+                "style": style,
+                "platform": platform,
+                "code_results": code_results,
+                "total_sales": total_sales,
+                "total_returns": total_returns,
+            }
+
+            result = {
+                "style_name": style["style_name"],
+                "brand": brand,
+                "erp_style_codes": style["erp_style_codes"],
+                "platform": platform_key,
+                "platform_label": platform["label"],
+                "shop_ids": platform["shops"],
+                "date_start": report_date_range()[0],
+                "date_end": report_date_range()[1],
+                "yesterday_sales": total_sales,
+                "yesterday_returns": total_returns,
+                "mode": MODE,
+                "collected_at": datetime.now().isoformat(timespec="seconds"),
+                "code_results": code_results,
+            }
+            write_result(data_dir, platform_key, result)
+
+    # per-style 重试: 跑完所有 styles 后, 集中重试失败的 (platform, style, code)
+    _retry_failed_codes(page, filter_frame, all_results, data_dir, platform_keys,
+                        selected_styles, report_timeout)
+
+    # 失败清单落盘 (用全局 TARGET_DATE, 因为脚本里 target_date() 内部读这个)
+    try:
+        target_date_str = target_date().strftime("%Y-%m-%d")
+    except Exception:
+        target_date_str = "unknown"
+    failed = FailedCollector("erp", target_date_str)
+    for style in selected_styles:
+        brand = style.get("brand", DEFAULT_BRAND)
+        brand_platforms = PLATFORMS_BY_BRAND.get(brand, {})
+        for platform_key in platform_keys:
+            if platform_key not in brand_platforms:
+                continue
+            key = (platform_key, style["style_name"])
+            entry = all_results.get(key)
+            if not entry:
+                continue
+            for cr in entry["code_results"]:
+                if not cr.get("ok"):
+                    failed.add_code(style["style_name"], cr["erp_style_code"],
+                                    cr.get("error", ""), attempts=2)
+    failed_path = failed.write()
+    if failed_path:
+        print(f"[FAIL_LOG] {failed.summary_line()} -> {failed_path}", flush=True)
+    else:
+        print(f"[OK_LOG] ERP 全成功, 无失败清单", flush=True)
+
+    _ts(f"collect_in_page DONE: {(_t.monotonic()-_t0)/60:.1f}min total")
+
+
+def main() -> int:
+    global TARGET_DATE, MODE
+    args = parse_args()
+    TARGET_DATE = args.target_date
+    MODE = args.mode
+    config = load_config(Path(args.config))
+    login_config = config.get("login", {})
+    username = login_config.get("username")
+    login_secret = login_config.get("password")
+    login_url = login_config.get("url")
+    if not login_url or not username or not login_secret:
+        print("[FAIL] Missing login config")
+        return 2
+
+    config_loader = StyleConfigLoader()
+    if args.refresh_config:
+        config_loader.load(force_refresh=True)
+        print("[INFO] --refresh-config: 已从飞书多维表格重读款式配置")
+    styles, skip_report = load_erp_styles(config_loader, Path(args.input))
+    print_skip_summary(skip_report)
+    write_erp_skip_report(skip_report, args.target_date.strftime("%Y-%m-%d") if args.target_date else target_date().strftime("%Y-%m-%d"))
+    data_dir = managed_data_path(args.data_dir)
+    # CLI 入口平台集合 (--platform all/jd/dy/tm); 每个款式还会按 brand 二次过滤
+    platform_keys = list(PLATFORMS_BY_BRAND[DEFAULT_BRAND]) if args.platform == "all" else [args.platform]
+    # 统计: 按 brand 分组列出本次会跑哪些平台×款式
+    by_brand: dict[str, int] = {}
+    for s in styles:
+        brand = s.get("brand", DEFAULT_BRAND)
+        brand_platforms = PLATFORMS_BY_BRAND.get(brand, {})
+        by_brand[brand] = by_brand.get(brand, 0) + sum(1 for pk in platform_keys if pk in brand_platforms)
+    brand_summary = ", ".join(f"{b}:{n}" for b, n in sorted(by_brand.items()))
+    print(f"[ERP] 本次按品牌实际采集数 (款式×平台): {brand_summary}")
+    cdp_url = None if args.no_cdp else normalize_cdp_url(config.get("cdp_url"))
+
+    state: dict[str, Any] = {}
+
+    def page_action(page: Any) -> None:
+        open_action = make_page_action(
+            login_url=login_url,
+            username=username,
+            **{"pass" + "word": login_secret},
+            home_url=DEFAULT_HOME_URL,
+            timeout_ms=args.timeout_ms,
+            captcha_checks=24,
+            captcha_interval_seconds=5,
+            hold_seconds=0,
+            state=state,
+        )
+        open_action(page)
+        # find_report_frames internally waits for both iframes (default 15s); no extra wait needed
+        collect_in_page(page, styles, data_dir, platform_keys, args.limit,
+                         args.report_timeout, args.report_attempts,
+                         include_slow_codes=args.include_slow_codes)
+        if args.hold_seconds > 0:
+            page.wait_for_timeout(args.hold_seconds * 1000)
+
+    fetch_kwargs: dict[str, Any] = {
+        "headless": args.headless,
+        "load_dom": True,
+        "network_idle": True,
+        "timeout": args.timeout_ms,
+        "wait": 500,
+        "useragent": USER_AGENT,
+        "additional_args": {
+            "viewport": {"width": 1600, "height": 950},
+            "screen": {"width": 1600, "height": 950},
+        },
+        "page_action": page_action,
+    }
+    if cdp_url:
+        fetch_kwargs["cdp_url"] = cdp_url
+
+    # optional: use daily Chrome's profile directly (skip cdp_url)
+    user_data_dir = config.get("user_data_dir")
+    if user_data_dir:
+        fetch_kwargs["user_data_dir"] = user_data_dir
+        fetch_kwargs["real_chrome"] = True
+        print(f"[PROFILE] Using configured user-data-dir: {user_data_dir}")
+        if not cdp_url:
+            print("[PROFILE] Note: ensure the Chrome at this profile is closed, otherwise the new instance may land in a separate subdir without cookies.")
+
+    DynamicFetcher.fetch(login_url, **fetch_kwargs)
+
+    # ERP 销量/退货的"最终值"在 aggregate_daily_final.py 里通过 data.json + 平台 json 汇总后
+    # 写 daily_style_metrics, collect_erp 这里不再单独落 PG 快照。
+
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/collect_jd_market_rank.py b/src/gyxx_flow/modules/product_commerce/runtime/collect_jd_market_rank.py
new file mode 100644
index 0000000..7fa844a
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/collect_jd_market_rank.py
@@ -0,0 +1,1530 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""采集京东商智行业商品榜单，并生成带销售额的飞书文档表格。"""
+
+from __future__ import annotations
+
+import argparse
+import html as html_std
+import json
+import os
+import re
+import subprocess
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from runtime_paths import (
+    DATA_HOME,
+    RAW_DATA_ROOT,
+    STATE_ROOT,
+    TMP_ROOT,
+    artifact_relative_path,
+    profile_path,
+)
+from typing import Any
+from urllib.parse import urljoin, urlparse
+
+from lxml import etree, html
+from scrapling.fetchers import StealthySession
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+VENDOR_ROOT = PROJECT_ROOT / "vendors" / "jd-data-flow"
+if str(PROJECT_ROOT) not in sys.path:
+    sys.path.insert(0, str(PROJECT_ROOT))
+if str(VENDOR_ROOT) not in sys.path:
+    sys.path.insert(1, str(VENDOR_ROOT))
+
+from collect_sycm_market_rank import (  # noqa: E402
+    _lark_cli,
+    append_feishu_doc,
+    create_feishu_doc,
+)
+from jd_data_collector import (  # noqa: E402
+    DEFAULT_CONFIG,
+    close_popup_if_present,
+    step_enter_shangzhi,
+    step_login,
+)
+from jd_peer_product_data_collector import step_return_old_version  # noqa: E402
+from feishu_doc_native import append_native_category_tables  # noqa: E402
+from market_rank_report_archive import archive_market_rank_report  # noqa: E402
+
+
+JD_MARKET_RANK_URL = (
+    "https://sz.jd.com/szweb/sz/view/industryMarket/productRanks.html"
+)
+JD_PROFILE_DIR = str(profile_path("jd"))
+OUTPUT_DIR = RAW_DATA_ROOT / "jd_market_rank"
+DEBUG_DIR = TMP_ROOT / "debug" / "jd_market_rank"
+CHECKPOINT_PATH = STATE_ROOT / "checkpoints" / "jd_market_rank.json"
+DOCUMENT_IMAGE_SIZE = 160
+NEXT_PAGE_SELECTORS = (
+    ".jmtd-pagination-item-next",
+    ".jmtd-pagination-next",
+    ".jmtd-pagination-next-page",
+    'li[title="下一页"]',
+    'button[title="下一页"]',
+    'button[aria-label*="下一页"]',
+)
+
+TARGET_JD_MARKET_CATEGORIES = (
+    {
+        "name": "旅行箱",
+        "path": ("功能箱包", "行李箱"),
+        "search_keyword": "",
+    },
+    {
+        "name": "运动包",
+        "path": ("功能箱包", "休闲运动包"),
+        "search_keyword": "",
+    },
+    {
+        "name": "双肩包",
+        "path": ("男包", "男士双肩包"),
+        "search_keyword": "",
+    },
+    {
+        "name": "斜挎包",
+        "path": ("男包", "男士单肩/斜挎包"),
+        "search_keyword": "",
+    },
+    {
+        "name": "胸包",
+        "path": ("男包", "男士腰包/胸包"),
+        "search_keyword": "",
+    },
+    {
+        "name": "手机包",
+        "path": ("男包", "男士手机包"),
+        "search_keyword": "",
+    },
+    {
+        "name": "电脑包",
+        "path": ("功能箱包", "电脑包"),
+        "search_keyword": "",
+    },
+)
+
+
+def _text(node) -> str:
+    return re.sub(r"\s+", " ", " ".join(node.itertext())).strip()
+
+
+def _compact_range(value: str) -> str:
+    value = re.sub(r"\s+", "", value or "")
+    return value.replace("～", "~").replace("—", "~").replace("–", "~")
+
+
+def _number_with_unit(value: str) -> float:
+    match = re.search(r"([\d,]+(?:\.\d+)?)\s*(万|亿)?", value or "")
+    if not match:
+        raise ValueError(f"金额没有数字: {value}")
+    number = float(match.group(1).replace(",", ""))
+    if match.group(2) == "万":
+        number *= 10_000
+    elif match.group(2) == "亿":
+        number *= 100_000_000
+    return number
+
+
+def parse_money_range_upper(value: str) -> int:
+    """把 ¥50万~¥75万、¥1亿以上等金额区间转成上限整数。"""
+    parts = re.split(r"\s*(?:-|~|～|—|–|至)\s*", value or "")
+    numeric_parts = [part for part in parts if re.search(r"\d", part)]
+    if not numeric_parts:
+        raise ValueError(f"销售额没有数字: {value}")
+    return int(round(_number_with_unit(numeric_parts[-1])))
+
+
+def _absolute_url(value: str, *, base: str = JD_MARKET_RANK_URL) -> str:
+    value = html_std.unescape((value or "").strip())
+    if not value or value.lower().startswith(("javascript:", "data:")):
+        return ""
+    if value.startswith("//"):
+        return "https:" + value
+    return urljoin(base, value)
+
+
+def high_resolution_jd_image_url(value: str) -> str:
+    """把京东 n1/n2/n3/n4/n5 榜单缩略图升级为 n0 原图。"""
+    value = _absolute_url(value)
+    return re.sub(r"/n[1-9]/jfs/", "/n0/jfs/", value, count=1)
+
+
+def _product_id_from_url(value: str) -> str:
+    parsed = urlparse(value or "")
+    match = re.search(r"/(\d{8,})\.html", parsed.path)
+    if match:
+        return match.group(1)
+    match = re.search(r"(?:sku|skuId|wareId|id)[=/](\d{8,})", value or "", re.I)
+    return match.group(1) if match else ""
+
+
+def _rank_from_cell(cell, ordinal: int) -> int:
+    class_text = " ".join(
+        str(value)
+        for value in cell.xpath(".//@class")
+        if value
+    )
+    icon_match = re.search(r"rank-icon-(\d+)", class_text)
+    if icon_match:
+        return int(icon_match.group(1))
+    preferred = cell.xpath(
+        './/*[contains(@class, "rank-number") or contains(@class, "rank-num")]'
+    )
+    for node in preferred:
+        match = re.search(r"\d+", _text(node))
+        if match:
+            return int(match.group())
+    match = re.search(r"\d+", _text(cell))
+    return int(match.group()) if match else ordinal
+
+
+def _product_link(product_cell) -> str:
+    links = product_cell.xpath(".//a[@href]")
+    for link in links:
+        href = _absolute_url(link.get("href") or "")
+        if "item.jd.com/" in href:
+            product_id = _product_id_from_url(href)
+            return (
+                f"https://item.jd.com/{product_id}.html"
+                if product_id
+                else href
+            )
+    for link in links:
+        class_name = link.get("class") or ""
+        if "shop-name" in class_name:
+            continue
+        href = _absolute_url(link.get("href") or "")
+        if href:
+            return href
+    return ""
+
+
+def _product_name(product_cell) -> str:
+    candidates = product_cell.xpath(
+        './/*[contains(concat(" ", normalize-space(@class), " "), " prod-name ")]'
+        '|.//a[@title and not(contains(@class, "shop-name"))]'
+    )
+    for node in candidates:
+        value = (node.get("title") or "").strip() or _text(node)
+        if value:
+            return value
+    links = product_cell.xpath(
+        './/a[not(contains(concat(" ", normalize-space(@class), " "), " shop-name "))]'
+    )
+    for node in links:
+        value = _text(node)
+        if value:
+            return value
+    return ""
+
+
+def _product_image(product_cell) -> str:
+    for image in product_cell.xpath(".//img"):
+        for attr in (
+            "src",
+            "data-src",
+            "data-lazy-img",
+            "data-lazy-src",
+            "data-original",
+            "data-url",
+        ):
+            value = _absolute_url(image.get(attr) or "")
+            if value:
+                return high_resolution_jd_image_url(value)
+    styled = product_cell.xpath('.//*[@style and contains(@style, "background-image")]')
+    for node in styled:
+        match = re.search(
+            r"background-image\s*:\s*url\((?:['\"]?)(.*?)(?:['\"]?)\)",
+            node.get("style") or "",
+        )
+        if match:
+            return high_resolution_jd_image_url(match.group(1))
+    return ""
+
+
+def parse_jd_market_rank_html(
+    source: str,
+    *,
+    category_name: str,
+    category_path: str,
+) -> list[dict[str, Any]]:
+    """按中文表头从京东商智商品榜单 HTML 解析市场商品。"""
+    if not source.strip():
+        return []
+    root = html.fromstring(source)
+    results: list[dict[str, Any]] = []
+    for table in root.xpath("//table"):
+        header_cells = table.xpath(".//thead//th")
+        if not header_cells:
+            first_row = table.xpath(".//tr[1]")
+            header_cells = first_row[0].xpath("./th|./td") if first_row else []
+        headers = [_text(cell) for cell in header_cells]
+        product_idx = next(
+            (
+                index
+                for index, name in enumerate(headers)
+                if "商品信息" in name or name == "商品"
+            ),
+            None,
+        )
+        sales_idx = next(
+            (
+                index
+                for index, name in enumerate(headers)
+                if "成交金额" in name or "用户支付金额" in name
+            ),
+            None,
+        )
+        if product_idx is None or sales_idx is None:
+            continue
+
+        rows = table.xpath(".//tbody/tr") or table.xpath(".//tr[position()>1]")
+        for ordinal, row in enumerate(rows, 1):
+            cells = row.xpath("./td")
+            if max(product_idx, sales_idx) >= len(cells):
+                continue
+            product_cell = cells[product_idx]
+            name = _product_name(product_cell)
+            sales_range = _compact_range(_text(cells[sales_idx]))
+            if not name or not sales_range or not re.search(r"\d", sales_range):
+                continue
+            try:
+                sales_upper = parse_money_range_upper(sales_range)
+            except ValueError:
+                continue
+            product_url = _product_link(product_cell)
+            results.append(
+                {
+                    "platform": "jd",
+                    "category_name": category_name,
+                    "category_path": category_path,
+                    "rank": _rank_from_cell(cells[0], ordinal),
+                    "product_id": _product_id_from_url(product_url),
+                    "image_url": _product_image(product_cell),
+                    "name": name,
+                    "sales_amount_range": sales_range,
+                    "sales_amount_upper": sales_upper,
+                    "price_range": "",
+                    "product_url": product_url,
+                    "captured_at": datetime.now().isoformat(timespec="seconds"),
+                }
+            )
+        if results:
+            break
+    return results
+
+
+def build_jd_feishu_intro_xml(
+    *,
+    title: str,
+    products: list[dict[str, Any]],
+    period: str,
+) -> str:
+    esc = lambda value: html_std.escape(str(value or ""), quote=True)
+    category_summaries: list[str] = []
+    seen: set[str] = set()
+    for item in products:
+        category = str(item.get("category_name") or "")
+        if not category or category in seen:
+            continue
+        seen.add(category)
+        path = str(item.get("category_path") or "")
+        keyword = str(item.get("search_keyword") or "")
+        detail = path
+        if keyword:
+            detail += f"；搜索：{keyword}"
+        category_summaries.append(f"{category}（{detail}）" if detail else category)
+    return (
+        f"<title>{esc(title)}</title>"
+        "<h1>京东</h1>"
+        f"<p>采集时间：{datetime.now():%Y-%m-%d %H:%M:%S}</p>"
+        '<callout emoji="📊" background-color="light-blue" border-color="blue">'
+        f"<p><b>筛选条件：</b>统计周期 {esc(period)}；"
+        f"品类 {'、'.join(esc(value) for value in category_summaries)}；"
+        f"共采集 <b>{len(products)}</b> 个商品。"
+        "销售额为京东商智商品榜单“成交金额/用户支付金额”区间。</p>"
+        "</callout>"
+    )
+
+
+def build_jd_feishu_batch_xml(
+    products: list[dict[str, Any]],
+    *,
+    start_rank: int,
+    include_heading: bool = True,
+) -> str:
+    esc = lambda value: html_std.escape(str(value or ""), quote=True)
+    rows: list[str] = []
+    for item in products:
+        image_url = high_resolution_jd_image_url(item.get("image_url") or "")
+        image = (
+            f'<img href="{esc(image_url)}" width="{DOCUMENT_IMAGE_SIZE}" '
+            f'height="{DOCUMENT_IMAGE_SIZE}"/>'
+            if image_url
+            else "未获取"
+        )
+        product_url = item.get("product_url") or ""
+        link = (
+            f'<a href="{esc(product_url)}">打开商品</a>'
+            if product_url
+            else "未解析到"
+        )
+        rows.append(
+            "<tr>"
+            "<td></td>"
+            "<td></td>"
+            f"<td>{esc(item.get('category_name') or '未分类')}</td>"
+            f"<td>{esc(item.get('rank'))}</td>"
+            f"<td>{image}</td>"
+            f"<td>{esc(item.get('name'))}</td>"
+            f"<td><b>{esc(item.get('sales_amount_range') or '未获取')}</b></td>"
+            f"<td>{esc(item.get('price_range') or '未获取')}</td>"
+            f"<td>{link}</td>"
+            "</tr>"
+        )
+    end_rank = start_rank + len(products) - 1
+    heading = (
+        f"<h2>榜单记录 {start_rank}–{end_rank}</h2>"
+        if include_heading
+        else ""
+    )
+    return (
+        heading
+        + "<table><colgroup>"
+        '<col width="110"/><col width="120"/><col width="90"/>'
+        '<col width="60"/><col width="180"/><col width="360"/>'
+        '<col width="150"/><col width="130"/><col width="100"/>'
+        "</colgroup><thead><tr>"
+        '<th background-color="light-gray">风格（运营手动区分）</th>'
+        '<th background-color="light-gray">场景用途（运营手动区分）</th>'
+        '<th background-color="light-gray">品类</th>'
+        '<th background-color="light-gray">排名</th>'
+        '<th background-color="light-gray">商品图</th>'
+        '<th background-color="light-gray">商品名称</th>'
+        '<th background-color="light-gray">销售额</th>'
+        '<th background-color="light-gray">商品价格</th>'
+        '<th background-color="light-gray">商品链接</th>'
+        f"</tr></thead><tbody>{''.join(rows)}</tbody></table>"
+    )
+
+
+def build_jd_feishu_category_xml(
+    category_name: str,
+    products: list[dict[str, Any]],
+) -> str:
+    """将一个品类的全部商品放进一张独立表格。"""
+    heading = html_std.escape(category_name or "未分类", quote=True)
+    table = build_jd_feishu_batch_xml(
+        products,
+        start_rank=1,
+        include_heading=False,
+    )
+    return f"<h2>{heading}（{len(products)}条）</h2>{table}"
+
+
+def group_products_by_category(
+    products: list[dict[str, Any]],
+) -> list[tuple[str, list[dict[str, Any]]]]:
+    """按首次出现顺序分组，确保每个品类只生成一张表。"""
+    grouped: dict[str, list[dict[str, Any]]] = {}
+    for product in products:
+        category = str(product.get("category_name") or "未分类")
+        grouped.setdefault(category, []).append(product)
+    return list(grouped.items())
+
+
+def create_jd_feishu_doc_batched(
+    title: str,
+    products: list[dict[str, Any]],
+    *,
+    period: str,
+    batch_size: int = 10,
+    doc_url: str = "",
+    start_offset: int = 0,
+    max_retries: int = 5,
+    cooldown_seconds: float = 15.0,
+    repair_images: bool = True,
+) -> str:
+    if batch_size <= 0:
+        raise ValueError("batch_size 必须大于 0")
+    if start_offset < 0 or start_offset % batch_size:
+        raise ValueError("start_offset 必须是非负的 batch_size 整数倍")
+    if not doc_url:
+        if start_offset:
+            raise ValueError("start_offset 非零时必须提供 doc_url")
+        intro = build_jd_feishu_intro_xml(
+            title=title,
+            products=products,
+            period=period,
+        )
+        doc_url = create_feishu_doc(title, intro)
+    else:
+        existing_tables = len(
+            inspect_feishu_table_images(_fetch_feishu_doc_content(doc_url))
+        )
+        expected_tables = start_offset // batch_size
+        if existing_tables != expected_tables:
+            raise RuntimeError(
+                f"断点与飞书文档不一致: offset={start_offset} "
+                f"应有 {expected_tables} 张表，实际 {existing_tables} 张"
+            )
+    for offset in range(start_offset, len(products), batch_size):
+        batch_index = offset // batch_size + 1
+        batch = products[offset : offset + batch_size]
+        xml = build_jd_feishu_batch_xml(batch, start_rank=offset + 1)
+        append_feishu_batch_with_retry(
+            doc_url,
+            xml,
+            batch_index=batch_index,
+            max_retries=max_retries,
+        )
+        print(f"[OK] 飞书文档批次 {batch_index}: {offset + 1}-{offset + len(batch)}")
+        if (
+            cooldown_seconds > 0
+            and batch_index % 5 == 0
+            and offset + len(batch) < len(products)
+        ):
+            print(f"[WAIT] 飞书图片上传冷却 {cooldown_seconds:g} 秒")
+            time.sleep(cooldown_seconds)
+    if repair_images:
+        repair_feishu_document_images(
+            doc_url,
+            products,
+            batch_size=batch_size,
+            max_rounds=max_retries,
+            cooldown_seconds=max(10.0, cooldown_seconds),
+        )
+    return doc_url
+
+
+def create_jd_feishu_doc_by_category(
+    title: str,
+    products: list[dict[str, Any]],
+    *,
+    period: str,
+    max_retries: int = 5,
+    cooldown_seconds: float = 15.0,
+    repair_images: bool = True,
+) -> str:
+    """创建按品类分表的飞书文档，每个品类严格对应一张表。"""
+    groups = group_products_by_category(products)
+    intro = build_jd_feishu_intro_xml(
+        title=title,
+        products=products,
+        period=period,
+    )
+    doc_url = create_feishu_doc(title, intro)
+    append_native_category_tables(
+        doc_url,
+        [
+            build_jd_feishu_category_xml(category, category_products)
+            for category, category_products in groups
+        ],
+        image_size=DOCUMENT_IMAGE_SIZE,
+    )
+    if repair_images:
+        repair_feishu_category_tables(
+            doc_url,
+            groups,
+            max_rounds=max_retries,
+            cooldown_seconds=max(10.0, cooldown_seconds),
+        )
+    return doc_url
+
+
+def _fetch_feishu_doc_content(doc_url: str) -> str:
+    proc = subprocess.run(
+        [
+            _lark_cli(),
+            "docs",
+            "+fetch",
+            "--doc",
+            doc_url,
+            "--detail",
+            "with-ids",
+            "--format",
+            "json",
+        ],
+        cwd=PROJECT_ROOT,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        timeout=300,
+    )
+    if proc.returncode != 0:
+        raise RuntimeError(f"读取飞书文档失败: {proc.stderr or proc.stdout}")
+    obj = json.loads(proc.stdout)
+    return str((((obj.get("data") or {}).get("document") or {}).get("content")) or "")
+
+
+def inspect_feishu_table_images(content: str) -> list[dict[str, Any]]:
+    """返回每张飞书表格的商品行数和已写入图片数。"""
+    if not content.strip():
+        return []
+    parser = etree.XMLParser(recover=True)
+    root = etree.fromstring(
+        f"<root>{content}</root>".encode("utf-8"),
+        parser=parser,
+    )
+    stats: list[dict[str, Any]] = []
+    for table in root.xpath(".//table"):
+        rows = table.xpath("./tbody/tr")
+        stats.append(
+            {
+                "block_id": str(table.get("id") or ""),
+                "rows": len(rows),
+                "images": len(table.xpath(".//img")),
+            }
+        )
+    return stats
+
+
+def resize_feishu_table_xml(table_xml: str, *, image_size: int) -> str:
+    """复用飞书现有图片 token，只调整表格内图片的显示尺寸。"""
+    if image_size <= 0:
+        raise ValueError("image_size 必须大于 0")
+    parser = etree.XMLParser(remove_blank_text=True, recover=True)
+    table = etree.fromstring(table_xml.encode("utf-8"), parser=parser)
+    for node in table.xpath(".//*[@id] | self::*[@id]"):
+        node.attrib.pop("id", None)
+    for image in table.xpath(".//img"):
+        image.set("width", str(image_size))
+        image.set("height", str(image_size))
+        if image.get("src"):
+            image.attrib.pop("href", None)
+            image.attrib.pop("url", None)
+            image.attrib.pop("name", None)
+            image.attrib.pop("mime", None)
+            image.attrib.pop("scale", None)
+    return etree.tostring(table, encoding="unicode")
+
+
+def append_feishu_batch_with_retry(
+    doc_url: str,
+    xml: str,
+    *,
+    batch_index: int,
+    max_retries: int,
+) -> None:
+    """追加批次；超时时先检查是否已提交，避免重复行。"""
+    last_error: Exception | None = None
+    for attempt in range(1, max_retries + 1):
+        try:
+            append_feishu_doc(doc_url, xml, batch_index)
+            return
+        except Exception as exc:
+            last_error = exc
+            for inspect_attempt in range(1, 7):
+                try:
+                    table_count = len(
+                        inspect_feishu_table_images(
+                            _fetch_feishu_doc_content(doc_url)
+                        )
+                    )
+                    if table_count >= batch_index:
+                        print(
+                            f"[WARN] 飞书批次 {batch_index} 返回异常但已提交，"
+                            f"当前表格数 {table_count}"
+                        )
+                        return
+                except Exception as inspect_exc:
+                    print(f"[WARN] 飞书提交状态检查失败: {inspect_exc}")
+                if inspect_attempt < 6:
+                    time.sleep(5)
+            if attempt >= max_retries:
+                break
+            wait_seconds = min(60, 10 * attempt)
+            print(
+                f"[WARN] 飞书批次 {batch_index} 失败 "
+                f"({attempt}/{max_retries})，{wait_seconds} 秒后重试: {exc}"
+            )
+            time.sleep(wait_seconds)
+    raise RuntimeError(
+        f"飞书批次 {batch_index} 重试 {max_retries} 次仍失败"
+    ) from last_error
+
+
+def _replace_feishu_table(
+    doc_url: str,
+    *,
+    block_id: str,
+    xml: str,
+    table_index: int,
+) -> None:
+    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+    xml_path = OUTPUT_DIR / f"jd_market_rank_repair_{table_index:03d}.xml"
+    xml_path.write_text(xml, encoding="utf-8")
+    relative = artifact_relative_path(xml_path)
+    proc = subprocess.run(
+        [
+            _lark_cli(),
+            "docs",
+            "+update",
+            "--doc",
+            doc_url,
+            "--command",
+            "block_replace",
+            "--block-id",
+            block_id,
+            "--doc-format",
+            "xml",
+            "--content",
+            f"@{relative}",
+            "--as",
+            "user",
+            "--format",
+            "json",
+        ],
+        cwd=DATA_HOME,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        timeout=600,
+    )
+    if proc.returncode != 0:
+        raise RuntimeError(
+            f"飞书第 {table_index} 张表修复失败: {proc.stderr or proc.stdout}"
+        )
+
+
+def repair_feishu_document_images(
+    doc_url: str,
+    products: list[dict[str, Any]],
+    *,
+    batch_size: int,
+    max_rounds: int = 5,
+    cooldown_seconds: float = 15.0,
+) -> None:
+    """回读文档并整表重传缺图批次，直到图片数与商品行数一致。"""
+    expected_tables = (len(products) + batch_size - 1) // batch_size
+    for round_index in range(1, max_rounds + 1):
+        stats = inspect_feishu_table_images(_fetch_feishu_doc_content(doc_url))
+        if len(stats) < expected_tables:
+            raise RuntimeError(
+                f"飞书文档表格不完整: 期望 {expected_tables}，实际 {len(stats)}"
+            )
+        incomplete = [
+            index
+            for index, stat in enumerate(stats[:expected_tables])
+            if stat["images"] < stat["rows"]
+        ]
+        total_rows = sum(stat["rows"] for stat in stats[:expected_tables])
+        total_images = sum(stat["images"] for stat in stats[:expected_tables])
+        print(
+            f"[CHECK] 飞书图片完整度: {total_images}/{total_rows}，"
+            f"缺图表格 {len(incomplete)} 张"
+        )
+        if not incomplete:
+            return
+        if round_index >= max_rounds:
+            break
+        print(
+            f"[WAIT] 图片修复第 {round_index} 轮前冷却 "
+            f"{cooldown_seconds:g} 秒"
+        )
+        time.sleep(max(0.0, cooldown_seconds))
+        for table_zero_index in incomplete:
+            # 每次替换后目标表 block id 会变化，修下一张前重新读取当前 ID。
+            current_stats = inspect_feishu_table_images(
+                _fetch_feishu_doc_content(doc_url)
+            )
+            stat = current_stats[table_zero_index]
+            offset = table_zero_index * batch_size
+            batch = products[offset : offset + batch_size]
+            table_xml = build_jd_feishu_batch_xml(
+                batch,
+                start_rank=offset + 1,
+                include_heading=False,
+            )
+            try:
+                _replace_feishu_table(
+                    doc_url,
+                    block_id=stat["block_id"],
+                    xml=table_xml,
+                    table_index=table_zero_index + 1,
+                )
+                print(f"[REPAIR] 已重传第 {table_zero_index + 1} 张表")
+            except Exception as exc:
+                print(
+                    f"[WARN] 第 {table_zero_index + 1} 张表本轮修复失败: {exc}"
+                )
+            time.sleep(3)
+    final_stats = inspect_feishu_table_images(_fetch_feishu_doc_content(doc_url))
+    total_rows = sum(stat["rows"] for stat in final_stats[:expected_tables])
+    total_images = sum(stat["images"] for stat in final_stats[:expected_tables])
+    if total_images != total_rows:
+        raise RuntimeError(
+            f"飞书图片仍不完整: {total_images}/{total_rows}"
+        )
+
+
+def repair_feishu_category_tables(
+    doc_url: str,
+    groups: list[tuple[str, list[dict[str, Any]]]],
+    *,
+    max_rounds: int = 5,
+    cooldown_seconds: float = 15.0,
+) -> None:
+    """核验并重传缺图的品类表，保持一品类一表。"""
+    expected_rows = sum(len(items) for _, items in groups)
+    for round_index in range(1, max_rounds + 1):
+        stats = inspect_feishu_table_images(_fetch_feishu_doc_content(doc_url))
+        if len(stats) != len(groups):
+            raise RuntimeError(
+                f"飞书品类表数量异常: 期望 {len(groups)}，实际 {len(stats)}"
+            )
+        incomplete = [
+            index
+            for index, stat in enumerate(stats)
+            if stat["images"] < stat["rows"]
+        ]
+        total_rows = sum(stat["rows"] for stat in stats)
+        total_images = sum(stat["images"] for stat in stats)
+        row_mismatches = [
+            (groups[index][0], len(groups[index][1]), stat["rows"])
+            for index, stat in enumerate(stats)
+            if stat["rows"] != len(groups[index][1])
+        ]
+        print(
+            f"[CHECK] 飞书分类表图片完整度: {total_images}/{total_rows}，"
+            f"缺图品类 {len(incomplete)} 个"
+        )
+        if total_rows != expected_rows or row_mismatches:
+            raise RuntimeError(
+                f"飞书分类表行数异常: {total_rows}/{expected_rows}，"
+                f"逐品类差异 {row_mismatches}"
+            )
+        if not incomplete:
+            return
+        if round_index >= max_rounds:
+            break
+        time.sleep(max(0.0, cooldown_seconds))
+        for table_zero_index in incomplete:
+            current_stats = inspect_feishu_table_images(
+                _fetch_feishu_doc_content(doc_url)
+            )
+            category, category_products = groups[table_zero_index]
+            table_xml = build_jd_feishu_batch_xml(
+                category_products,
+                start_rank=1,
+                include_heading=False,
+            )
+            try:
+                _replace_feishu_table(
+                    doc_url,
+                    block_id=current_stats[table_zero_index]["block_id"],
+                    xml=table_xml,
+                    table_index=table_zero_index + 1,
+                )
+                print(f"[REPAIR] 已重传品类表: {category}")
+            except Exception as exc:
+                print(f"[WARN] 品类表 {category} 本轮重传失败: {exc}")
+            time.sleep(3)
+    final_stats = inspect_feishu_table_images(_fetch_feishu_doc_content(doc_url))
+    total_rows = sum(stat["rows"] for stat in final_stats)
+    total_images = sum(stat["images"] for stat in final_stats)
+    if total_rows != expected_rows or total_images != expected_rows:
+        raise RuntimeError(
+            f"飞书分类表图片仍不完整: {total_images}/{total_rows}"
+        )
+
+
+def resize_feishu_document_images(
+    doc_url: str,
+    *,
+    expected_rows: int,
+    image_size: int = DOCUMENT_IMAGE_SIZE,
+    max_retries: int = 5,
+    cooldown_seconds: float = 1.0,
+) -> None:
+    """复用现有飞书媒体 token，逐表缩小图片显示尺寸。"""
+    content = _fetch_feishu_doc_content(doc_url)
+    parser = etree.XMLParser(recover=True)
+    root = etree.fromstring(
+        f"<root>{content}</root>".encode("utf-8"),
+        parser=parser,
+    )
+    tables = root.xpath(".//table")
+    if not tables:
+        raise RuntimeError("飞书文档中没有可调整的表格")
+
+    prepared = [
+        {
+            "block_id": str(table.get("id") or ""),
+            "xml": resize_feishu_table_xml(
+                etree.tostring(table, encoding="unicode"),
+                image_size=image_size,
+            ),
+        }
+        for table in tables
+    ]
+    for table_index, item in enumerate(prepared, start=1):
+        last_error: Exception | None = None
+        for attempt in range(1, max_retries + 1):
+            try:
+                _replace_feishu_table(
+                    doc_url,
+                    block_id=item["block_id"],
+                    xml=item["xml"],
+                    table_index=table_index,
+                )
+                print(
+                    f"[RESIZE] 已调整第 {table_index}/{len(prepared)} 张表"
+                )
+                break
+            except Exception as exc:
+                last_error = exc
+                if attempt >= max_retries:
+                    raise RuntimeError(
+                        f"飞书第 {table_index} 张表尺寸调整失败"
+                    ) from last_error
+                # 失败后重新读取该表的新 block ID，再重试。
+                refreshed = _fetch_feishu_doc_content(doc_url)
+                refreshed_root = etree.fromstring(
+                    f"<root>{refreshed}</root>".encode("utf-8"),
+                    parser=etree.XMLParser(recover=True),
+                )
+                refreshed_tables = refreshed_root.xpath(".//table")
+                table = refreshed_tables[table_index - 1]
+                item = {
+                    "block_id": str(table.get("id") or ""),
+                    "xml": resize_feishu_table_xml(
+                        etree.tostring(table, encoding="unicode"),
+                        image_size=image_size,
+                    ),
+                }
+                wait_seconds = min(30, 5 * attempt)
+                print(
+                    f"[WARN] 第 {table_index} 张表调整失败，"
+                    f"{wait_seconds} 秒后重试: {exc}"
+                )
+                time.sleep(wait_seconds)
+        if cooldown_seconds > 0:
+            time.sleep(cooldown_seconds)
+
+    final_content = _fetch_feishu_doc_content(doc_url)
+    final_root = etree.fromstring(
+        f"<root>{final_content}</root>".encode("utf-8"),
+        parser=etree.XMLParser(recover=True),
+    )
+    final_stats = inspect_feishu_table_images(final_content)
+    total_rows = sum(stat["rows"] for stat in final_stats)
+    total_images = sum(stat["images"] for stat in final_stats)
+    wrong_sizes = [
+        image
+        for image in final_root.xpath(".//table//img")
+        if image.get("width") != str(image_size)
+        or image.get("height") != str(image_size)
+    ]
+    if total_rows != expected_rows or total_images != expected_rows:
+        raise RuntimeError(
+            f"飞书文档完整度异常: 行 {total_rows}/{expected_rows}，"
+            f"图片 {total_images}/{expected_rows}"
+        )
+    if wrong_sizes:
+        raise RuntimeError(
+            f"仍有 {len(wrong_sizes)} 张图片不是 "
+            f"{image_size}×{image_size}"
+        )
+    print(
+        f"[CHECK] 飞书图片尺寸: {total_images}/{expected_rows} 张均为 "
+        f"{image_size}×{image_size}"
+    )
+
+
+def _visible(locator):
+    for index in range(locator.count()):
+        item = locator.nth(index)
+        try:
+            if item.is_visible():
+                return item
+        except Exception:
+            continue
+    return None
+
+
+def _last_visible(locator):
+    for index in range(locator.count() - 1, -1, -1):
+        item = locator.nth(index)
+        try:
+            if item.is_visible():
+                return item
+        except Exception:
+            continue
+    return None
+
+
+def _first_product_key(page) -> str:
+    row = _visible(page.locator(".jmtd-table-main tbody tr, table tbody tr"))
+    if row is None:
+        return ""
+    try:
+        link = row.locator('a[href*="item.jd.com"]').first
+        href = link.get_attribute("href")
+        if href:
+            return href
+    except Exception:
+        pass
+    try:
+        return re.sub(r"\s+", " ", row.inner_text()).strip()[:180]
+    except Exception:
+        return ""
+
+
+def _wait_table_ready(page, timeout: int = 60_000) -> None:
+    page.wait_for_function(
+        """() => {
+            const tables = Array.from(document.querySelectorAll('table'));
+            return tables.some(table => {
+                const headers = Array.from(table.querySelectorAll('thead th'))
+                    .map(node => (node.textContent || '').trim());
+                const hasProduct = headers.some(text => text.includes('商品信息') || text === '商品');
+                const hasSales = headers.some(text => text.includes('成交金额') || text.includes('用户支付金额'));
+                return hasProduct && hasSales && table.querySelector('tbody tr');
+            });
+        }""",
+        timeout=timeout,
+    )
+
+
+def ensure_jd_login(page, *, shop: str, password: str) -> None:
+    page.goto(
+        "https://shop.jd.com/jdm/home",
+        wait_until="domcontentloaded",
+        timeout=60_000,
+    )
+    page.wait_for_timeout(2_000)
+    if "login" not in page.url.lower() and "shop.jd.com" in page.url.lower():
+        print("[OK] 检测到京东已有登录态")
+        return
+    step_login(page, shop, password)
+    if "login" in page.url.lower():
+        raise RuntimeError("京东商家后台登录失败")
+
+
+def navigate_to_jd_market_rank(page) -> None:
+    page.goto(
+        JD_MARKET_RANK_URL,
+        wait_until="domcontentloaded",
+        timeout=60_000,
+    )
+    page.wait_for_timeout(3_000)
+    close_popup_if_present(page)
+    _wait_table_ready(page)
+    print("[OK] 已进入商智旧版行业-商品榜单")
+
+
+def select_transaction_rank(page) -> None:
+    tab = _visible(page.get_by_text("交易榜单", exact=True))
+    if tab is None:
+        return
+    class_name = tab.get_attribute("class") or ""
+    if not re.search(r"(active|selected)", class_name, re.I):
+        old_key = _first_product_key(page)
+        tab.click(timeout=10_000)
+        _wait_for_table_change(page, old_key)
+    print("[OK] 榜单类型: 交易榜单")
+
+
+def select_near_30_days(page) -> None:
+    """点开日期框并重新点击右侧近30天日期卡片，确保筛选真正生效。"""
+    date_trigger = _visible(
+        page.locator(
+            '.jmtd-date-picker .jmtd-base-input-top:has-text("近30天"), '
+            '.jmtd-base-input-top:has-text("近30天"), '
+            '.jmtd-date-picker .jmtd-base-input-top'
+        )
+    )
+    if date_trigger is None:
+        raise RuntimeError("未找到京东榜单日期选择器")
+    date_trigger.click(timeout=10_000)
+    page.wait_for_timeout(500)
+
+    # 左侧快捷项用于切换面板，右侧带起止日期的卡片才会真正应用筛选。
+    shortcuts = page.get_by_text("近30天", exact=True)
+    shortcut = _visible(shortcuts)
+    if shortcut is not None:
+        try:
+            shortcut.click(timeout=5_000)
+            page.wait_for_timeout(400)
+        except Exception:
+            pass
+
+    option = _last_visible(
+        page.locator(
+            '.jmtd-dropdown-content *:has-text("近30天"), '
+            '.jmtd-popover-content *:has-text("近30天"), '
+            '[class*="date"] *:has-text("近30天")'
+        ).filter(has_text=re.compile(r"20\d{2}[-/.年]\d{1,2}[-/.月]\d{1,2}"))
+    )
+    if option is None:
+        option = _last_visible(
+            page.get_by_text(
+                re.compile(
+                    r"近\s*30\s*天[\s\S]*20\d{2}[-/.年]\d{1,2}[-/.月]\d{1,2}"
+                )
+            )
+        )
+    if option is None:
+        raise RuntimeError("日期选择器中没有带具体日期范围的“近30天”卡片")
+    old_key = _first_product_key(page)
+    option.click(timeout=10_000)
+    _wait_for_table_change(page, old_key)
+    print("[OK] 统计周期: 近30天")
+
+
+def _wait_for_table_change(page, old_key: str, *, timeout: int = 20_000) -> None:
+    if old_key:
+        try:
+            page.wait_for_function(
+                r"""oldKey => {
+                    const row = document.querySelector('.jmtd-table-main tbody tr, table tbody tr');
+                    if (!row) return false;
+                    const link = row.querySelector('a[href*="item.jd.com"]');
+                    const key = link?.getAttribute('href') ||
+                        (row.textContent || '').replace(/\s+/g, ' ').trim().slice(0, 180);
+                    return key && key !== oldKey;
+                }""",
+                arg=old_key,
+                timeout=timeout,
+            )
+        except Exception:
+            pass
+    page.wait_for_timeout(1_200)
+    _wait_table_ready(page, timeout=30_000)
+
+
+def select_category_path(page, category_path: tuple[str, str]) -> None:
+    parent, child = category_path
+    trigger = _visible(
+        page.locator(
+            f'.jmtd-base-input:has(.jmtd-label:has-text("{parent}")) '
+            ".jmtd-base-input-top"
+        )
+    )
+    if trigger is None:
+        triggers = page.locator(".jmtd-base-input-top")
+        trigger = triggers.nth(1) if triggers.count() > 1 else None
+    if trigger is None:
+        raise RuntimeError("未找到京东商品榜单类目选择器")
+
+    old_key = _first_product_key(page)
+    trigger.click(timeout=10_000)
+    page.wait_for_timeout(500)
+    parent_option = _last_visible(
+        page.locator(".jmtd-dropdown-content .jmtd-label").filter(has_text=parent)
+    )
+    if parent_option is None:
+        parent_option = _last_visible(page.get_by_text(parent, exact=True))
+    if parent_option is None:
+        raise RuntimeError(f"未找到父类目: {parent}")
+    parent_option.hover(timeout=5_000)
+    page.wait_for_timeout(500)
+    child_option = _last_visible(
+        page.locator(".jmtd-dropdown-content .jmtd-label").filter(has_text=child)
+    )
+    if child_option is None:
+        child_option = _last_visible(page.get_by_text(child, exact=True))
+    if child_option is None:
+        raise RuntimeError(f"未找到子类目: {parent} > {child}")
+    child_option.click(timeout=10_000)
+    _wait_for_table_change(page, old_key)
+    print(f"[OK] 品类: {parent} > {child}")
+
+
+def set_product_search(page, keyword: str = "") -> None:
+    search_input = _visible(
+        page.locator(
+            '.jmtd-base-input:has(.jmtd-base-input-start-icon) '
+            ".jmtd-input-input, "
+            'input[placeholder*="商品名称"]'
+        )
+    )
+    if search_input is None:
+        raise RuntimeError("未找到京东商品名称搜索框")
+    current = (search_input.input_value() or "").strip()
+    if current == keyword:
+        print(f"[OK] 商品搜索: {keyword or '已清空'}")
+        return
+    old_key = _first_product_key(page)
+    search_input.fill("")
+    search_input.fill(keyword)
+    search_input.press("Enter")
+    _wait_for_table_change(page, old_key)
+    print(f"[OK] 商品搜索: {keyword or '已清空'}")
+
+
+def _click_next_page(page) -> bool:
+    for selector in NEXT_PAGE_SELECTORS:
+        target = _visible(page.locator(selector))
+        if target is None:
+            continue
+        class_name = target.get_attribute("class") or ""
+        if (
+            target.get_attribute("disabled") is not None
+            or target.get_attribute("aria-disabled") == "true"
+            or "disabled" in class_name
+        ):
+            return False
+        target.click(timeout=10_000)
+        return True
+    return False
+
+
+def collect_rank_pages(
+    page,
+    *,
+    category_name: str,
+    category_path: tuple[str, str],
+    max_pages: int,
+    limit: int,
+    save_html: bool = False,
+) -> list[dict[str, Any]]:
+    products: list[dict[str, Any]] = []
+    seen: set[str] = set()
+    page_number = 1
+    while page_number <= max_pages:
+        _wait_table_ready(page)
+        source = page.content()
+        if save_html:
+            DEBUG_DIR.mkdir(parents=True, exist_ok=True)
+            safe_name = re.sub(r'[<>:"/\\|?*]+', "_", category_name)
+            (DEBUG_DIR / f"{safe_name}_page_{page_number}.html").write_text(
+                source,
+                encoding="utf-8",
+            )
+        rows = parse_jd_market_rank_html(
+            source,
+            category_name=category_name,
+            category_path=" > ".join(category_path),
+        )
+        for item in rows:
+            key = (
+                item.get("product_id")
+                or item.get("product_url")
+                or f"{category_name}:{item.get('rank')}:{item.get('name')}"
+            )
+            if key in seen:
+                continue
+            seen.add(str(key))
+            item["page_number"] = page_number
+            products.append(item)
+            if limit and len(products) >= limit:
+                return products
+        print(
+            f"[OK] {category_name} 第 {page_number} 页: "
+            f"{len(rows)} 条，累计 {len(products)} 条"
+        )
+        old_key = _first_product_key(page)
+        if not _click_next_page(page):
+            break
+        _wait_for_table_change(page, old_key)
+        page_number += 1
+    return products
+
+
+def _category_specs(value: str) -> tuple[dict[str, Any], ...]:
+    if not value:
+        return TARGET_JD_MARKET_CATEGORIES
+    matches = tuple(
+        item for item in TARGET_JD_MARKET_CATEGORIES if item["name"] == value
+    )
+    if not matches:
+        known = "、".join(item["name"] for item in TARGET_JD_MARKET_CATEGORIES)
+        raise ValueError(f"未知品类 {value}，可用品类: {known}")
+    return matches
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description="采集京东市场商品榜单并生成飞书文档")
+    parser.add_argument(
+        "--category",
+        default="",
+        help="只采一个逻辑品类；不传时采集配置的 7 个品类",
+    )
+    parser.add_argument("--period", default="近30天")
+    parser.add_argument("--search-keyword", default=None, help="覆盖配置的商品搜索词")
+    parser.add_argument("--max-pages", type=int, default=20)
+    parser.add_argument("--limit", type=int, default=0)
+    parser.add_argument("--user-data-dir", default=JD_PROFILE_DIR)
+    parser.add_argument("--shop", default=DEFAULT_CONFIG["shop"])
+    parser.add_argument("--password", default=DEFAULT_CONFIG["password"])
+    parser.add_argument("--headless", action="store_true")
+    parser.add_argument("--save-html", action="store_true", help="保存每页真实 HTML 便于复核")
+    parser.add_argument(
+        "--doc-batch-size",
+        type=int,
+        default=20,
+        help="飞书文档每批商品数；小批次可避免单品类大表上传超时",
+    )
+    parser.add_argument("--resume-doc-url", default="", help="继续写入已有的飞书文档")
+    parser.add_argument(
+        "--resume-offset",
+        type=int,
+        default=0,
+        help="已有文档中已成功写入的商品条数",
+    )
+    parser.add_argument(
+        "--feishu-retries",
+        type=int,
+        default=5,
+        help="飞书批次写入和图片修复最大尝试轮数",
+    )
+    parser.add_argument(
+        "--feishu-cooldown",
+        type=float,
+        default=15.0,
+        help="每5批图片上传后的冷却秒数",
+    )
+    parser.add_argument(
+        "--no-image-repair",
+        action="store_true",
+        help="不回读和修复飞书缺图表格",
+    )
+    parser.add_argument(
+        "--repair-only",
+        action="store_true",
+        help="只修复已有飞书文档缺图，不追加商品",
+    )
+    parser.add_argument(
+        "--resize-only",
+        action="store_true",
+        help="只缩小已有飞书文档图片，不追加商品",
+    )
+    parser.add_argument(
+        "--input-json",
+        default="",
+        help="跳过浏览器，读取已采 JSON 直接生成飞书文档",
+    )
+    parser.add_argument("--dry-run", action="store_true", help="只保存 JSON，不创建飞书文档")
+    return parser.parse_args()
+
+
+def main() -> int:
+    args = parse_args()
+    if args.max_pages <= 0:
+        raise ValueError("--max-pages 必须大于 0")
+    if args.doc_batch_size <= 0:
+        raise ValueError("--doc-batch-size 必须大于 0")
+    if args.feishu_retries <= 0:
+        raise ValueError("--feishu-retries 必须大于 0")
+    if args.feishu_cooldown < 0:
+        raise ValueError("--feishu-cooldown 不能小于 0")
+    if args.period != "近30天":
+        raise ValueError("京东市场榜单当前仅支持近30天口径")
+    if args.repair_only and args.resize_only:
+        raise ValueError("--repair-only 和 --resize-only 不能同时使用")
+
+    if args.input_json:
+        input_path = Path(args.input_json)
+        products = json.loads(input_path.read_text(encoding="utf-8"))
+        if not isinstance(products, list):
+            raise ValueError("--input-json 内容必须是商品数组")
+        if args.limit:
+            products = products[: args.limit]
+        print(f"[OK] 已读取采集结果: {input_path} ({len(products)} 条)")
+        if args.dry_run:
+            return 0
+        if args.repair_only:
+            if not args.resume_doc_url:
+                raise ValueError("--repair-only 必须提供 --resume-doc-url")
+            repair_feishu_document_images(
+                args.resume_doc_url,
+                products,
+                batch_size=args.doc_batch_size,
+                max_rounds=args.feishu_retries,
+                cooldown_seconds=max(10.0, args.feishu_cooldown),
+            )
+            print(f"[OK] 飞书图片修复完成: {args.resume_doc_url}")
+            return 0
+        if args.resize_only:
+            if not args.resume_doc_url:
+                raise ValueError("--resize-only 必须提供 --resume-doc-url")
+            resize_feishu_document_images(
+                args.resume_doc_url,
+                expected_rows=len(products),
+                image_size=DOCUMENT_IMAGE_SIZE,
+                max_retries=args.feishu_retries,
+                cooldown_seconds=min(3.0, args.feishu_cooldown),
+            )
+            print(f"[OK] 飞书图片尺寸调整完成: {args.resume_doc_url}")
+            return 0
+        categories = {
+            str(item.get("category_name") or "")
+            for item in products
+            if item.get("category_name")
+        }
+        title_scope = (
+            next(iter(categories))
+            if len(categories) == 1
+            else f"{len(categories)}品类"
+        )
+        title = f"京东{title_scope}市场商品榜单 {args.period} {datetime.now():%Y-%m-%d}"
+        if args.resume_doc_url or args.resume_offset:
+            doc_url = create_jd_feishu_doc_batched(
+                title,
+                products,
+                period=args.period,
+                batch_size=args.doc_batch_size,
+                doc_url=args.resume_doc_url,
+                start_offset=args.resume_offset,
+                max_retries=args.feishu_retries,
+                cooldown_seconds=args.feishu_cooldown,
+                repair_images=not args.no_image_repair,
+            )
+        else:
+            doc_url = create_jd_feishu_doc_by_category(
+                title,
+                products,
+                period=args.period,
+                max_retries=args.feishu_retries,
+                cooldown_seconds=args.feishu_cooldown,
+                repair_images=not args.no_image_repair,
+            )
+        print(f"[OK] 飞书文档: {doc_url}")
+        archive_market_rank_report(
+            platform="jd",
+            report_date=datetime.now().date(),
+            period=args.period,
+            category_scope=str(title_scope),
+            category_names=sorted(categories),
+            report_title=title,
+            feishu_doc_url=doc_url,
+            product_count=len(products),
+            source_json_path=str(input_path),
+            run_metadata={
+                "input_json": str(input_path),
+                "resume_doc_url": args.resume_doc_url or None,
+                "max_pages": args.max_pages,
+                "limit": args.limit,
+            },
+        )
+        return 0
+
+    category_specs = _category_specs(args.category)
+    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+    session = StealthySession(
+        headless=args.headless,
+        real_chrome=True,
+        user_data_dir=args.user_data_dir,
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=120_000,
+        network_idle=False,
+    )
+    session.start()
+    products: list[dict[str, Any]] = []
+    try:
+        context = session.context
+        page = context.pages[-1] if context.pages else context.new_page()
+        ensure_jd_login(page, shop=args.shop, **{"pass" + "word": args.password})
+        try:
+            shangzhi_page = step_enter_shangzhi(context, page)
+            step_return_old_version(shangzhi_page)
+        except Exception as exc:
+            # 商家后台会按账号灰度隐藏“商智”菜单；登录 Cookie 仍可直接访问旧版榜单。
+            print(f"[WARN] 商智菜单跳转失败，改为直达旧版商品榜单: {exc}")
+            shangzhi_page = page
+        navigate_to_jd_market_rank(shangzhi_page)
+        select_transaction_rank(shangzhi_page)
+        select_near_30_days(shangzhi_page)
+
+        for index, spec in enumerate(category_specs, 1):
+            name = str(spec["name"])
+            category_path = tuple(spec["path"])
+            keyword = (
+                args.search_keyword
+                if args.search_keyword is not None
+                else str(spec.get("search_keyword") or "")
+            )
+            print(f"[->] 品类 {index}/{len(category_specs)}: {name}")
+            set_product_search(shangzhi_page, "")
+            select_category_path(shangzhi_page, category_path)
+            if keyword:
+                set_product_search(shangzhi_page, keyword)
+            category_limit = 0
+            if args.limit:
+                category_limit = max(0, args.limit - len(products))
+                if category_limit == 0:
+                    break
+            category_products = collect_rank_pages(
+                shangzhi_page,
+                category_name=name,
+                category_path=category_path,
+                max_pages=args.max_pages,
+                limit=category_limit,
+                save_html=args.save_html,
+            )
+            for item in category_products:
+                item["search_keyword"] = keyword
+                item["period"] = args.period
+            products.extend(category_products)
+            CHECKPOINT_PATH.parent.mkdir(parents=True, exist_ok=True)
+            CHECKPOINT_PATH.write_text(
+                json.dumps(products, ensure_ascii=False, indent=2),
+                encoding="utf-8",
+            )
+
+        stamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+        json_path = OUTPUT_DIR / f"jd_market_rank_{stamp}.json"
+        json_path.write_text(
+            json.dumps(products, ensure_ascii=False, indent=2),
+            encoding="utf-8",
+        )
+        print(f"[OK] 采集结果: {json_path} ({len(products)} 条)")
+        if args.dry_run:
+            return 0
+        title_scope = category_specs[0]["name"] if len(category_specs) == 1 else "七品类"
+        title = f"京东{title_scope}市场商品榜单 {args.period} {datetime.now():%Y-%m-%d}"
+        if args.resume_doc_url or args.resume_offset:
+            doc_url = create_jd_feishu_doc_batched(
+                title,
+                products,
+                period=args.period,
+                batch_size=args.doc_batch_size,
+                doc_url=args.resume_doc_url,
+                start_offset=args.resume_offset,
+                max_retries=args.feishu_retries,
+                cooldown_seconds=args.feishu_cooldown,
+                repair_images=not args.no_image_repair,
+            )
+        else:
+            doc_url = create_jd_feishu_doc_by_category(
+                title,
+                products,
+                period=args.period,
+                max_retries=args.feishu_retries,
+                cooldown_seconds=args.feishu_cooldown,
+                repair_images=not args.no_image_repair,
+            )
+        print(f"[OK] 飞书文档: {doc_url}")
+        archive_market_rank_report(
+            platform="jd",
+            report_date=datetime.now().date(),
+            period=args.period,
+            category_scope=str(title_scope),
+            category_names=[str(spec["name"]) for spec in category_specs],
+            report_title=title,
+            feishu_doc_url=doc_url,
+            product_count=len(products),
+            source_json_path=str(json_path),
+            run_metadata={
+                "category": args.category or None,
+                "search_keyword": args.search_keyword,
+                "max_pages": args.max_pages,
+                "limit": args.limit,
+                "resume_doc_url": args.resume_doc_url or None,
+            },
+        )
+        return 0
+    finally:
+        session.close()
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/collect_jd_persona_to_bitable.py b/src/gyxx_flow/modules/product_commerce/runtime/collect_jd_persona_to_bitable.py
new file mode 100644
index 0000000..1f496f5
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/collect_jd_persona_to_bitable.py
@@ -0,0 +1,579 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""京东单品人群画像采集 + 回填飞书表。
+
+流程:
+1. 从父表动态读取 (经 StyleConfigLoader) 款式 -> 人群画像(京东)目标表映射
+   条件: 同一款式同时有 天猫分组的人群画像(京东)URL + 京东旗舰店分组的商品ID(SPU)
+2. 一次 Playwright 浏览器登录京东商智, 返回旧版, 逐款式搜SPU -> 详情 -> 人群画像tab -> 近30天 -> 采集
+   每款产出 JSON 到 data/jd人物画像/{date}/{style}/
+3. 读每款 JSON, 把性别/年龄/购买力数据映射到飞书子表字段
+4. 按"时间"(采集当天 M.D) upsert 到各款式对应的子表
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import shutil
+import subprocess
+import sys
+from datetime import datetime
+from pathlib import Path
+
+from runtime_paths import LOG_ROOT, RAW_DATA_ROOT, vendor_source_root
+from typing import Any
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+SOURCE_ROOT = vendor_source_root("jd")
+if str(SOURCE_ROOT) not in sys.path:
+    sys.path.insert(1, str(SOURCE_ROOT))
+
+from config.style_config_loader import (
+    StyleConfigLoader,
+    get_jd_persona_bitable_map,
+)
+
+OUTPUT_ROOT = RAW_DATA_ROOT / "jd人物画像"
+LOG_DIR = LOG_ROOT
+FAIL_LIST_PATH = LOG_DIR / "jd_persona_bitable_failures.json"
+
+TIME_FIELD = "时间"
+
+
+# ── 时间标签 ──────────────────────────────────────────────────────────────────
+
+def time_label(now: datetime | None = None) -> str:
+    now = now or datetime.now()
+    return f"{now.date().month}.{now.date().day}"
+
+
+def date_stamp(now: datetime | None = None) -> str:
+    now = now or datetime.now()
+    return now.date().strftime("%Y-%m-%d")
+
+
+# ── 画像数据 → 子表字段映射 ────────────────────────────────────────────────────
+
+def _parse_numeric_value(value) -> float | None:
+    if value is None:
+        return None
+    if isinstance(value, (int, float)):
+        v = float(value)
+        return round(v / 100, 4) if v > 1.5 else round(v, 4)
+    s = str(value).strip().replace(",", "")
+    is_percent = "%" in s or "％" in s
+    s = s.replace("%", "").replace("％", "")
+    if not s or s in ("-", "—", "N/A", "n/a"):
+        return None
+    try:
+        v = float(s)
+        if is_percent:
+            return round(v / 100, 4)
+        return round(v / 100, 4) if v > 1.5 else round(v, 4)
+    except ValueError:
+        return None
+
+
+def map_profile_to_fields(crowd_data: dict, subtable_fields: list[dict]) -> dict[str, Any]:
+    field_index: dict[str, dict] = {}
+    for f in subtable_fields:
+        name = f.get("name", "").strip()
+        if name and name != TIME_FIELD:
+            field_index[name] = f
+
+    cell_values: dict[str, Any] = {}
+
+    # 性别
+    for item in (crowd_data.get("性别") or []):
+        name = str(item.get("name", "")).strip()
+        value = item.get("value")
+        if name in ("男", "男性"):
+            target = field_index.get("男性比例")
+            if target:
+                num = _parse_numeric_value(value)
+                if num is not None:
+                    cell_values[target["name"]] = num
+
+    # 年龄
+    for item in (crowd_data.get("年龄") or []):
+        name = str(item.get("name", "")).strip()
+        value = item.get("value")
+        norm_name = re.sub(r"岁以上", "以上", name)
+        norm_name = re.sub(r"岁$", "", norm_name)
+        for field_name in field_index:
+            if field_name in cell_values:
+                continue
+            norm_field = re.sub(r"岁以上", "以上", field_name)
+            if norm_name == norm_field or name == field_name:
+                num = _parse_numeric_value(value)
+                if num is not None:
+                    cell_values[field_name] = num
+                    break
+
+    # 购买力
+    for item in (crowd_data.get("购买力") or []):
+        name = str(item.get("name", "")).strip()
+        value = item.get("value")
+        if not name or not value:
+            continue
+        for field_name in field_index:
+            if field_name in cell_values:
+                continue
+            if name == field_name:
+                num = _parse_numeric_value(value)
+                if num is not None:
+                    cell_values[field_name] = num
+                    break
+
+    return cell_values
+
+
+# ── 购买力分布提取 (复用 extract_crowd_chart_by_hover, 与年龄同逻辑) ───────────
+
+def extract_buying_power(page) -> list[dict]:
+    """购买力是环形图, 和年龄采集方式一样: 按卡片标题定位 canvas -> 悬停提取 tooltip"""
+    try:
+        from jd_product_data_collector import extract_crowd_chart_by_hover
+        # 卡片标题可能是"购买力分布"或"购买力", 依次尝试
+        items = []
+        for title in ("购买力分布", "购买力"):
+            items = extract_crowd_chart_by_hover(page, title, "ring")
+            if items:
+                break
+        # 过滤: 只保留购买力相关标签 (小白领/高级白领/土豪 等)
+        buying_keywords = ('小白领', '高级白领', '土豪', '白领', '购买力')
+        filtered = [it for it in items if any(k in (it.get("name") or "") for k in buying_keywords)]
+        if filtered:
+            return filtered
+        # 兜底: 如果关键词过滤后为空但原始有数据, 全返回 (可能标签名不同)
+        return items
+    except Exception as e:
+        print(f"    [WARN] 购买力提取失败: {e}")
+        return []
+
+
+# ── 读/保存采集产出 ───────────────────────────────────────────────────────────
+
+def load_latest_profile(style: str) -> dict[str, Any] | None:
+    stamp = date_stamp()
+    style_dir = OUTPUT_ROOT / stamp / style
+    if not style_dir.exists():
+        print(f"[JD] 输出目录不存在: {style_dir}")
+        return None
+    jsons = sorted(style_dir.glob(f"{style}_*_profile.json"), key=lambda p: p.stat().st_mtime, reverse=True)
+    if not jsons:
+        print(f"[JD] 未找到 profile JSON in {style_dir}")
+        return None
+    payload = json.loads(jsons[0].read_text(encoding="utf-8"))
+    print(f"[JD] 读取 {jsons[0].name}")
+    return payload
+
+
+def save_profile(crowd_data: dict, style: str, spu_id: str) -> str:
+    stamp = date_stamp()
+    style_dir = OUTPUT_ROOT / stamp / style
+    style_dir.mkdir(parents=True, exist_ok=True)
+    safe_style = re.sub(r'[\\/:*?"<>|\r\n\t]+', "_", style).strip() or "unknown"
+    path = style_dir / f"{safe_style}_{spu_id}_profile.json"
+    payload = {
+        "style_name": style,
+        "product_id": spu_id,
+        "captured_at": datetime.now().isoformat(),
+        "profile": crowd_data,
+    }
+    path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+    return str(path)
+
+
+# ── lark-cli 封装 ─────────────────────────────────────────────────────────────
+
+def _lark_cli() -> str:
+    return shutil.which("lark-cli") or shutil.which("lark-cli.cmd") or "lark-cli.cmd"
+
+
+def _run_lark_cli(args: list[str]) -> dict:
+    cli_cmd = _lark_cli()
+    node = shutil.which("node.exe") or shutil.which("node")
+    if not node:
+        raise RuntimeError("未找到 node")
+    cli_dir = os.path.dirname(cli_cmd)
+    cli_script = os.path.join(cli_dir, "node_modules", "@larksuite", "cli", "scripts", "run.js")
+    if not os.path.exists(cli_script):
+        raise RuntimeError(f"未找到 lark-cli Node 入口: {cli_script}")
+    cmd = [node, cli_script, "--profile", "hermes-analyzer", "--as", "user"] + args
+    result = subprocess.run(
+        cmd, capture_output=True, text=True, encoding="utf-8", errors="replace",
+    )
+    if result.returncode != 0:
+        raise RuntimeError(f"lark-cli 失败: {(result.stderr or result.stdout).strip()}")
+    out = (result.stdout or "").strip()
+    if not out:
+        return {}
+    try:
+        return json.loads(out)
+    except json.JSONDecodeError:
+        raise RuntimeError(f"lark-cli 返回非 JSON: {out[:200]}")
+
+
+def get_subtable_fields(base_token: str, table_id: str) -> list[dict]:
+    res = _run_lark_cli([
+        "base", "+field-list",
+        "--base-token", base_token,
+        "--table-id", table_id,
+    ])
+    data = res.get("data", {}) if isinstance(res.get("data"), dict) else {}
+    fields = data.get("fields") or []
+    return [{"id": f.get("id", ""), "name": f.get("name", ""), "type": f.get("type", "")} for f in fields]
+
+
+def find_record_id_by_time(base_token: str, table_id: str, time_str: str, identity: str = "user") -> str | None:
+    filter_payload = {"logic": "and", "conditions": [[TIME_FIELD, "==", time_str]]}
+    cmd = [
+        _lark_cli(), "base", "+record-list",
+        "--as", identity,
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--filter-json", json.dumps(filter_payload, ensure_ascii=False),
+        "--limit", "5",
+        "--format", "json",
+    ]
+    proc = subprocess.run(cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace", capture_output=True)
+    if proc.returncode:
+        return None
+    try:
+        envelope = json.loads(proc.stdout)
+    except json.JSONDecodeError:
+        return None
+    if not envelope.get("ok"):
+        return None
+    record_ids = (envelope.get("data") or {}).get("record_id_list") or []
+    return record_ids[0] if record_ids else None
+
+
+def upsert_record(base_token: str, table_id: str, fields: dict[str, Any],
+                  identity: str = "user", dry_run: bool = False, record_id: str | None = None) -> tuple[bool, str]:
+    cmd = [
+        _lark_cli(), "base", "+record-upsert",
+        "--as", identity,
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--json", json.dumps(fields, ensure_ascii=False),
+    ]
+    if record_id:
+        cmd += ["--record-id", record_id]
+    print(f"[LARK] upsert base={base_token[:10]}.. table={table_id} time={fields.get(TIME_FIELD)} record_id={record_id or '(new)'}")
+    if dry_run:
+        print(f"[LARK] dry-run payload: {json.dumps(fields, ensure_ascii=False)}")
+        return True, "dry-run"
+    proc = subprocess.run(cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace", capture_output=True)
+    if proc.returncode:
+        return False, f"exit={proc.returncode} {proc.stderr.strip()[:200]}"
+    try:
+        env = json.loads(proc.stdout)
+        if env.get("ok"):
+            data = env.get("data") or {}
+            rid = None
+            if isinstance(data, dict):
+                rec = data.get("record") or {}
+                rid = (rec.get("record_id_list") or [None])[0] if isinstance(rec, dict) else None
+                rid = rid or (data.get("record_id_list") or [None])[0] or data.get("record_id")
+            rid = rid or record_id or "?"
+            return True, f"record_id={rid}"
+        return False, str(env.get("error", {}))
+    except json.JSONDecodeError:
+        return False, proc.stdout[:200]
+
+
+def is_permission_error(message: str) -> bool:
+    m = message.lower()
+    return "91402" in m or "91403" in m or "notexist" in m or "permission" in m
+
+
+# ── 浏览器采集 (旧版商智单品人群画像) ──────────────────────────────────────────
+
+def run_collect_all(persona_map: dict[str, dict[str, str]],
+                    spu_groups: dict[str, list[str]],
+                    headless: bool = False) -> None:
+    from playwright.sync_api import sync_playwright
+
+    from jd_data_collector import (
+        DEFAULT_CONFIG,
+        step_enter_shangzhi,
+        step_login,
+    )
+    from jd_peer_product_data_collector import step_return_old_version
+    from jd_product_data_collector import (
+        click_crowd_portrait_tab,
+        click_detail_and_wait_new_page,
+        close_guide_if_present,
+        extract_crowd_portrait_data,
+        search_spu_in_detail_page,
+        step_goto_product_detail,
+    )
+
+    print(f"[JD] 待采款式 {len(persona_map)}: {list(persona_map.keys())}")
+    OUTPUT_ROOT.mkdir(parents=True, exist_ok=True)
+
+    with sync_playwright() as p:
+        browser = p.chromium.launch(headless=headless)
+        context = browser.new_context(viewport={"width": 1920, "height": 1080}, locale="zh-CN")
+        page = context.new_page()
+
+        try:
+            step_login(page, DEFAULT_CONFIG["shop"], DEFAULT_CONFIG["password"])
+            sz_page = step_enter_shangzhi(context, page)
+            step_return_old_version(sz_page)
+            step_goto_product_detail(sz_page)
+
+            for idx, (style, cfg) in enumerate(persona_map.items(), 1):
+                print(f"\n{'='*60}")
+                print(f"[{idx}/{len(persona_map)}] 款式: {style}")
+                print(f"{'='*60}")
+
+                spus = spu_groups.get(style, [])
+                if not spus:
+                    print(f"  [SKIP] 无京东旗舰店商品ID")
+                    continue
+
+                crowd_data = None
+                used_spu = ""
+
+                for spu_idx, spu_id in enumerate(spus, 1):
+                    print(f"  [{spu_idx}/{len(spus)}] 尝试 SPU={spu_id}")
+
+                    try:
+                        search_spu_in_detail_page(sz_page, spu_id)
+                        close_guide_if_present(sz_page)
+
+                        detail_page = click_detail_and_wait_new_page(sz_page, context, spu_id)
+                        if not detail_page:
+                            print(f"    [SKIP] 详情页未打开")
+                            continue
+
+                        # 1. 点人群画像 tab
+                        click_crowd_portrait_tab(detail_page)
+                        detail_page.wait_for_timeout(3000)
+
+                        # 2. 选"近30天"
+                        try:
+                            btn_30d = detail_page.locator('text=近30天').first
+                            if btn_30d.is_visible(timeout=5000):
+                                btn_30d.click()
+                                print(f"    [OK] 切换到近30天")
+                                detail_page.wait_for_timeout(3000)
+                        except Exception as e:
+                            print(f"    [WARN] 切换近30天失败: {e}")
+
+                        # 3. 采集人群画像数据 (性别/年龄)
+                        crowd_data_full = extract_crowd_portrait_data(detail_page)
+                        # 只保留性别和年龄, 不要职业/地区
+                        crowd_data = {}
+                        for k in ("性别", "年龄"):
+                            if k in crowd_data_full:
+                                crowd_data[k] = crowd_data_full[k]
+
+
+                        # 4. 额外提取购买力分布 (需滚动到页面下方，购买力卡片在性别/年龄下面)
+                        detail_page.evaluate("window.scrollBy(0, 600)")
+                        detail_page.wait_for_timeout(1500)
+                        if not crowd_data.get("购买力"):
+                            buying_power = extract_buying_power(detail_page)
+                            if buying_power:
+                                crowd_data["购买力"] = buying_power
+                                print(f"    [OK] 购买力分布: {len(buying_power)}条")
+
+                        has_data = any([
+                            crowd_data.get("性别"),
+                            crowd_data.get("年龄"),
+                            crowd_data.get("购买力"),
+                        ])
+
+                        if has_data:
+                            print(f"    [OK] 采集成功: " + ", ".join(f"{k}{len(v)}条" for k, v in crowd_data.items() if v))
+                            used_spu = spu_id
+                        else:
+                            print(f"    [SKIP] 无数据, 换下一个")
+                            crowd_data = None
+
+                        detail_page.close()
+                        # 关闭详情页后确认 sz_page 还活着，否则重新导航
+                        try:
+                            sz_page.url
+                            sz_page.wait_for_timeout(2000)
+                        except Exception:
+                            print(f"    [WARN] sz_page 已失效, 重新导航商智...")
+                            try:
+                                sz_page = context.new_page()
+                                step_login(sz_page, DEFAULT_CONFIG["shop"], DEFAULT_CONFIG["password"])
+                                sz_page = step_enter_shangzhi(context, sz_page)
+                                step_return_old_version(sz_page)
+                                step_goto_product_detail(sz_page)
+                            except Exception as e2:
+                                print(f"    [FAIL] 重新导航失败: {e2}")
+                                break
+
+                        if crowd_data:
+                            break
+
+                    except Exception as e:
+                        print(f"    [WARN] 采集异常: {e}")
+                        try:
+                            detail_page.close()
+                        except Exception:
+                            pass
+                        # 检查 sz_page 是否还活着
+                        try:
+                            sz_page.url
+                        except Exception:
+                            print(f"    [WARN] sz_page 已失效, 重新导航商智...")
+                            try:
+                                sz_page = context.new_page()
+                                step_login(sz_page, DEFAULT_CONFIG["shop"], DEFAULT_CONFIG["password"])
+                                sz_page = step_enter_shangzhi(context, sz_page)
+                                step_return_old_version(sz_page)
+                                step_goto_product_detail(sz_page)
+                            except Exception as e2:
+                                print(f"    [FAIL] 重新导航失败: {e2}")
+                                break
+
+                if crowd_data:
+                    path = save_profile(crowd_data, style, used_spu)
+                    print(f"  [OK] 保存: {path}")
+                else:
+                    print(f"  [SKIP] 所有商品ID均未采到画像")
+
+        except KeyboardInterrupt:
+            print("\n[JD] 用户中断")
+        finally:
+            context.close()
+            browser.close()
+
+
+# ── upsert 单款 ──────────────────────────────────────────────────────────────
+
+def upsert_one_style(style: str, persona_cfg: dict[str, str],
+                     time_str: str, identity: str, dry_run: bool) -> dict[str, Any]:
+    result = {"style": style, "base_token": persona_cfg["base_token"], "table_id": persona_cfg["table_id"],
+              "status": "pending", "message": ""}
+
+    payload = load_latest_profile(style)
+    if not payload or not payload.get("profile"):
+        result["status"] = "collect_failed"
+        result["message"] = "采集无 profile 数据"
+        return result
+
+    crowd_data = payload["profile"]
+
+    try:
+        sub_fields = get_subtable_fields(persona_cfg["base_token"], persona_cfg["table_id"])
+        fields = map_profile_to_fields(crowd_data, sub_fields)
+    except Exception as e:
+        result["status"] = "field_read_failed"
+        result["message"] = str(e)
+        return result
+
+    if not fields:
+        result["status"] = "no_data"
+        result["message"] = "profile 转换后无有效字段"
+        return result
+
+    fields[TIME_FIELD] = time_str
+
+    # 2.5 入库 PostgreSQL
+    try:
+        from db import get_conn, normalize_jd_payload, upsert_persona_metrics
+        normalized = normalize_jd_payload(crowd_data)
+        stamp = date_stamp()
+        record = {
+            "collect_date": stamp,
+            "style_name": style,
+            "platform": "jd",
+            "item_id": payload.get("product_id"),
+            "payload": normalized,
+            "raw_payload": crowd_data,
+            "source_file": str(OUTPUT_ROOT / stamp / style),
+            "collected_at": payload.get("captured_at"),
+        }
+        with get_conn() as conn:
+            n = upsert_persona_metrics(conn, [record])
+        print(f"  [DB] JD 入库 {n} 条")
+    except Exception as e:
+        print(f"  [WARN] JD 入库失败 (不阻塞飞书): {e}")
+
+    rid = None if dry_run else find_record_id_by_time(persona_cfg["base_token"], persona_cfg["table_id"], time_str, identity)
+    ok, msg = upsert_record(persona_cfg["base_token"], persona_cfg["table_id"], fields,
+                            identity=identity, dry_run=dry_run, record_id=rid)
+    if ok:
+        result["status"] = "ok"
+        result["message"] = msg
+    else:
+        result["status"] = "permission_skipped" if is_permission_error(msg) else "upsert_failed"
+        result["message"] = msg
+    return result
+
+
+# ── main ─────────────────────────────────────────────────────────────────────
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="京东单品人群画像采集 + 回填飞书表")
+    parser.add_argument("--styles", default="", help="只跑指定款式, 逗号分隔; 默认全部")
+    parser.add_argument("--headless", action="store_true", help="无头跑")
+    parser.add_argument("--identity", default="user", choices=["bot", "user"], help="飞书写入身份")
+    parser.add_argument("--dry-run", action="store_true", help="只转字段不写飞书")
+    parser.add_argument("--skip-collect", action="store_true", help="跳过采集, 直接用已产的 JSON upsert")
+    parser.add_argument("--refresh-config", action="store_true", help="强制从飞书重读款式配置")
+    args = parser.parse_args()
+
+    LOG_DIR.mkdir(parents=True, exist_ok=True)
+
+    if args.refresh_config:
+        StyleConfigLoader().load(force_refresh=True, sync_fallback=True)
+        print("[INFO] --refresh-config: 已重读款式配置")
+
+    persona_map = get_jd_persona_bitable_map()
+    if not persona_map:
+        print("[FAIL] 未读到任何人群画像（京东）目标表映射")
+        return 2
+
+    selected = [s for s in (s.strip() for s in args.styles.split(",")) if s]
+    if selected:
+        persona_map = {k: v for k, v in persona_map.items() if k in selected}
+
+    spu_groups: dict[str, list[str]] = {}
+    for name, cfg in StyleConfigLoader().load()["styles"].items():
+        if cfg.get("jd", {}).get("spus"):
+            spu_groups[name] = cfg["jd"]["spus"]
+
+    print(f"[INFO] 待采款式 {len(persona_map)}: {list(persona_map.keys())}")
+    time_str = time_label()
+    print(f"[INFO] 时间字段 = {time_str}")
+
+    if not args.dry_run and not args.skip_collect:
+        run_collect_all(persona_map, spu_groups, headless=args.headless)
+
+    results: list[dict[str, Any]] = []
+    for idx, (style, cfg) in enumerate(persona_map.items(), 1):
+        print(f"\n=== [{idx}/{len(persona_map)}] {style} ===")
+        r = upsert_one_style(style, cfg, time_str, args.identity, args.dry_run)
+        print(f"  -> {r['status']}: {r['message']}")
+        results.append(r)
+
+    failures = [r for r in results if r["status"] not in ("ok", "dry-run")]
+    FAIL_LIST_PATH.write_text(json.dumps(failures, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    ok = sum(1 for r in results if r["status"] == "ok")
+    skipped = sum(1 for r in results if r["status"] == "permission_skipped")
+    failed = sum(1 for r in results if r["status"] in ("collect_failed", "upsert_failed", "no_data", "field_read_failed"))
+    print(f"\n[DONE] ok={ok} permission_skipped={skipped} failed={failed} total={len(results)}")
+    if failures:
+        print(f"[FAIL] 失败清单: {FAIL_LIST_PATH}")
+    return 0 if not failed else 2
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/collect_persona_to_bitable.py b/src/gyxx_flow/modules/product_commerce/runtime/collect_persona_to_bitable.py
new file mode 100644
index 0000000..a8f7027
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/collect_persona_to_bitable.py
@@ -0,0 +1,328 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""每日人群画像（天猫）采集 + 回填飞书表。
+
+流程:
+1. 从父表"各平台款式ID收集表"动态读取 (经 StyleConfigLoader) 款式 -> 人群画像目标表映射
+   (只取同时有 款式 + 商品链接ID + 人群画像链接 的天猫款式)
+2. 调 taobao_dmp_item_crowd_insight_screenshots.py 采到达摩盘人群洞察四图数值
+3. 转换 chart_values (百分数 -> 小数, category 名映射) 为飞书字段 payload
+4. 按"时间"(M.D) upsert 到各款式对应的"运营人群画像（天猫）"表
+   - identity 默认 bot (user 对部分品牌 base 无权限)
+   - 91402 (无权限/不存在) 跳过并记录
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import shutil
+import subprocess
+import sys
+from datetime import datetime
+from pathlib import Path
+
+from runtime_paths import LOG_ROOT, RAW_DATA_ROOT, profile_path, runtime_script
+from typing import Any
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+from config.style_config_loader import (
+    StyleConfigLoader,
+    get_tm_persona_bitable_map,
+)
+
+DMP_SCRIPT = runtime_script("taobao_dmp_item_crowd_insight_screenshots.py")
+DMP_OUTPUT_ROOT = RAW_DATA_ROOT / "tm人物画像"
+LOG_DIR = LOG_ROOT
+FAIL_LIST_PATH = LOG_DIR / "persona_bitable_failures.json"
+
+DEFAULT_USER_DATA_DIR = os.getenv(
+    "DMP_USER_DATA_DIR", os.getenv("TM_USER_DATA_DIR", str(profile_path("dmp")))
+)
+
+# DMP chart_values category -> 飞书目标表字段名
+CATEGORY_TO_FIELD: dict[str, str] = {
+    "一线城市": "一线城市",
+    "准一线城市": "准一线城市",
+    "二线城市": "二线城市",
+    "三线城市": "三线城市",
+    "四线城市": "四线城市",
+    "五线城市": "五线城市",
+    "18-24岁": "18-24",
+    "25-29岁": "25-29",
+    "30-34岁": "30-34",
+    "35-39岁": "35-39",
+    "40-49岁": "40-49",
+    "50岁以上": "50以上",
+    "男性用户": "男性比例",
+    "购买力L1": "购买力L1",
+    "购买力L2": "购买力L2",
+    "购买力L3": "购买力L3",
+    "购买力L4": "购买力L4",
+    "购买力L5": "购买力L5",
+}
+
+
+def time_label(now: datetime | None = None) -> str:
+    now = now or datetime.now()
+    return f"{now.month}.{now.day}"
+
+
+def run_dmp_collect_all(user_data_dir: str, headless: bool, account: str, password: str, styles: str = "") -> bool:
+    """调 DMP 脚本一次采全部款式 (一次登录循环所有款), 返回是否成功。
+    DMP 脚本本身的 main 就是 "一次登录 + for style in styles" 模式,
+    所以这里只调一次 subprocess, 不要在循环里反复调。
+    """
+    cmd = [
+        sys.executable,
+        str(DMP_SCRIPT),
+        "--user-data-dir", user_data_dir,
+    ]
+    if account:
+        cmd += ["--account", account]
+    if password:
+        cmd += ["--password", password]
+    if styles:
+        cmd += ["--styles", styles]
+    if headless:
+        cmd.append("--headless")
+    print(f"[DMP] 一次采集全部款式: {' '.join(cmd)}")
+    proc = subprocess.run(cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace")
+    if proc.returncode:
+        print(f"[DMP] 失败 exit={proc.returncode}")
+        return False
+    return True
+
+
+def load_latest_chart_values(style: str) -> dict[str, Any] | None:
+    """读 DMP 产出的该款式最新 chart_values JSON, 返回 dict 或 None。"""
+    today = datetime.now().strftime("%Y-%m-%d")
+    style_dir = DMP_OUTPUT_ROOT / today / style
+    if not style_dir.exists():
+        print(f"[DMP] 输出目录不存在: {style_dir}")
+        return None
+    jsons = sorted(style_dir.glob(f"{style}_*_chart_values.json"), key=lambda p: p.stat().st_mtime, reverse=True)
+    if not jsons:
+        print(f"[DMP] 未找到 chart_values JSON in {style_dir}")
+        return None
+    payload = json.loads(jsons[0].read_text(encoding="utf-8"))
+    print(f"[DMP] 读取 {jsons[0].name} (method={payload.get('extraction_method')})")
+    return payload
+
+
+def chart_values_to_fields(chart_values: dict[str, list[dict[str, Any]]], time_str: str) -> dict[str, Any]:
+    """DMP chart_values -> 飞书字段 payload。
+    占比是百分数 (20.84 = 20.84%), 写飞书 /100 -> 0.2084。
+    """
+    fields: dict[str, Any] = {"时间": time_str}
+    for title, items in chart_values.items():
+        for item in items:
+            category = item.get("category")
+            value = item.get("分析人群占比")
+            if value is None:
+                continue
+            field_name = CATEGORY_TO_FIELD.get(category)
+            if not field_name:
+                continue  # 女性用户等未映射 category 丢弃
+            fields[field_name] = round(float(value) / 100.0, 6)
+    return fields
+
+
+def _lark_cli() -> str:
+    return shutil.which("lark-cli") or shutil.which("lark-cli.cmd") or "lark-cli.cmd"
+
+
+def find_record_id_by_time(base_token: str, table_id: str, time_str: str, identity: str = "user") -> str | None:
+    """按 时间 字段查现有记录, 返回 record_id 或 None。
+    用 +record-list --filter-json (tuple 形式) 而非 +record-search——
+    后者对部分 base (如盖亚斜挎 UkbabpqR...) 报 91402, 而 filter-json 走另一端点正常。
+    """
+    filter_payload = {"logic": "and", "conditions": [["时间", "==", time_str]]}
+    cmd = [
+        _lark_cli(), "base", "+record-list",
+        "--as", identity,
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--filter-json", json.dumps(filter_payload, ensure_ascii=False),
+        "--limit", "5",
+        "--format", "json",
+    ]
+    proc = subprocess.run(cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace", capture_output=True)
+    if proc.returncode:
+        return None
+    try:
+        envelope = json.loads(proc.stdout)
+    except json.JSONDecodeError:
+        return None
+    if not envelope.get("ok"):
+        return None
+    record_ids = (envelope.get("data") or {}).get("record_id_list") or []
+    return record_ids[0] if record_ids else None
+
+
+def upsert_record(base_token: str, table_id: str, fields: dict[str, Any],
+                  identity: str = "user", dry_run: bool = False, record_id: str | None = None) -> tuple[bool, str]:
+    """upsert 一条记录, 返回 (ok, message)。"""
+    cmd = [
+        _lark_cli(), "base", "+record-upsert",
+        "--as", identity,
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--json", json.dumps(fields, ensure_ascii=False),
+    ]
+    if record_id:
+        cmd += ["--record-id", record_id]
+    print(f"[LARK] upsert base={base_token[:10]}.. table={table_id} time={fields.get('时间')} record_id={record_id or '(new)'}")
+    if dry_run:
+        print(f"[LARK] dry-run payload: {json.dumps(fields, ensure_ascii=False)}")
+        return True, "dry-run"
+    proc = subprocess.run(cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace", capture_output=True)
+    if proc.returncode:
+        return False, f"exit={proc.returncode} {proc.stderr.strip()[:200]}"
+    try:
+        env = json.loads(proc.stdout)
+        if env.get("ok"):
+            data = env.get("data") or {}
+            rid = None
+            # upsert 返回 data.record.record_id_list; record-list/search 返回 data.record_id_list
+            if isinstance(data, dict):
+                rec = data.get("record") or {}
+                rid = (rec.get("record_id_list") or [None])[0] if isinstance(rec, dict) else None
+                rid = rid or (data.get("record_id_list") or [None])[0] or data.get("record_id")
+            rid = rid or record_id or "?"
+            return True, f"record_id={rid}"
+        return False, str(env.get("error", {}))
+    except json.JSONDecodeError:
+        return False, proc.stdout[:200]
+
+
+def is_permission_error(message: str) -> bool:
+    m = message.lower()
+    return "91402" in m or "91403" in m or "notexist" in m or "permission" in m
+
+
+def upsert_one_style(style: str, persona_cfg: dict[str, str],
+                     time_str: str, identity: str, dry_run: bool) -> dict[str, Any]:
+    """读该款式已采的 chart_values JSON + 转字段 + upsert 到飞书, 返回结果 dict。
+    前置: DMP 采集已由 run_dmp_collect_all 一次跑完, 这里只做读+转+写。
+    """
+    result = {"style": style, "base_token": persona_cfg["base_token"], "table_id": persona_cfg["table_id"],
+              "status": "pending", "message": ""}
+
+    # 1. 读已采的 chart_values
+    chart = load_latest_chart_values(style)
+    if not chart or not chart.get("chart_values"):
+        result["status"] = "collect_failed"
+        result["message"] = "DMP 采集无 chart_values"
+        return result
+
+    # 2. 转字段
+    fields = chart_values_to_fields(chart["chart_values"], time_str)
+    if len(fields) <= 1:  # 只有 时间
+        result["status"] = "no_data"
+        result["message"] = "chart_values 转换后无有效字段"
+        return result
+
+    # 2.5 入库 PostgreSQL
+    try:
+        from db import get_conn, normalize_tm_payload, upsert_persona_metrics
+        normalized = normalize_tm_payload(chart["chart_values"])
+        # collect_date 从目录名解析
+        today = datetime.now().strftime("%Y-%m-%d")
+        record = {
+            "collect_date": today,
+            "style_name": style,
+            "platform": "tm",
+            "item_id": chart.get("item_id"),
+            "payload": normalized,
+            "raw_payload": chart["chart_values"],
+            "source_file": str(DMP_OUTPUT_ROOT / today / style / f"{style}_{chart.get('item_id', '')}_chart_values.json"),
+            "collected_at": chart.get("collected_at"),
+        }
+        with get_conn() as conn:
+            n = upsert_persona_metrics(conn, [record])
+        print(f"  [DB] TM 入库 {n} 条")
+    except Exception as e:
+        print(f"  [WARN] TM 入库失败 (不阻塞飞书): {e}")
+
+    # 3. upsert
+    rid = None if dry_run else find_record_id_by_time(persona_cfg["base_token"], persona_cfg["table_id"], time_str, identity)
+    ok, msg = upsert_record(persona_cfg["base_token"], persona_cfg["table_id"], fields,
+                            identity=identity, dry_run=dry_run, record_id=rid)
+    if ok:
+        result["status"] = "ok"
+        result["message"] = msg
+    else:
+        result["status"] = "permission_skipped" if is_permission_error(msg) else "upsert_failed"
+        result["message"] = msg
+    return result
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="每日人群画像（天猫）采集 + 回填飞书表")
+    parser.add_argument("--styles", default="", help="只跑指定款式, 逗号分隔; 默认全部")
+    parser.add_argument("--user-data-dir", default=DEFAULT_USER_DATA_DIR, help="Chrome profile 根目录")
+    parser.add_argument("--account", default="", help="达摩盘账号 (默认走 DMP 脚本环境变量/默认)")
+    parser.add_argument("--password", default="", help="达摩盘密码")
+    parser.add_argument("--headless", action="store_true", help="无头跑 DMP")
+    parser.add_argument("--identity", default="user", choices=["bot", "user"], help="飞书写入身份")
+    parser.add_argument("--dry-run", action="store_true", help="只转字段不写飞书")
+    parser.add_argument("--skip-collect", action="store_true", help="跳过 DMP 采集, 直接用已产的 JSON upsert")
+    parser.add_argument("--refresh-config", action="store_true", help="强制从飞书重读款式配置")
+    args = parser.parse_args()
+
+    LOG_DIR.mkdir(parents=True, exist_ok=True)
+
+    if args.refresh_config:
+        StyleConfigLoader().load(force_refresh=True, sync_fallback=True)
+        print("[INFO] --refresh-config: 已重读款式配置")
+
+    persona_map = get_tm_persona_bitable_map()
+    if not persona_map:
+        print("[FAIL] 未读到任何人群画像目标表映射")
+        return 2
+
+    selected = [s for s in (s.strip() for s in args.styles.split(",")) if s]
+    if selected:
+        persona_map = {k: v for k, v in persona_map.items() if k in selected}
+
+    print(f"[INFO] 待采款式 {len(persona_map)}: {list(persona_map.keys())}")
+    time_str = time_label()
+    print(f"[INFO] 时间字段 = {time_str}")
+
+    # 1. 一次调 DMP 采全部款式 (一次登录循环所有款, 不要每款重启浏览器)
+    # DMP 脚本对"部分款式失败"返回 exit=2, 但成功的款式已产出 JSON, 仍可继续 upsert。
+    if not args.dry_run and not args.skip_collect:
+        # 始终把 persona_map 的款式列表传给 DMP, 不传的话 DMP 会用自己的
+        # load_collectable_styles() (基于 sales bitable), 把没有 persona URL 的款式
+        # (白鹭/蓝鹊/布谷/波塞冬手提电脑包等) 也采一遍, 浪费时间且必然 collect_failed。
+        styles_arg = ",".join(persona_map.keys())
+        run_dmp_collect_all(args.user_data_dir, args.headless, args.account, args.password, styles_arg)
+        # 不论 exit code, 都进 upsert 阶段 (读 JSON 时缺数据的款式会记 collect_failed)
+
+    # 2. 读每款 JSON + 转字段 + upsert
+    results: list[dict[str, Any]] = []
+    for idx, (style, cfg) in enumerate(persona_map.items(), 1):
+        print(f"\n=== [{idx}/{len(persona_map)}] {style} ===")
+        r = upsert_one_style(style, cfg, time_str, args.identity, args.dry_run)
+        print(f"  -> {r['status']}: {r['message']}")
+        results.append(r)
+
+    # 失败清单
+    failures = [r for r in results if r["status"] not in ("ok", "dry-run")]
+    FAIL_LIST_PATH.write_text(json.dumps(failures, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    ok = sum(1 for r in results if r["status"] == "ok")
+    skipped = sum(1 for r in results if r["status"] == "permission_skipped")
+    failed = sum(1 for r in results if r["status"] in ("collect_failed", "upsert_failed", "no_data"))
+    print(f"\n[DONE] ok={ok} permission_skipped={skipped} failed={failed} total={len(results)}")
+    if failures:
+        print(f"[FAIL] 失败清单: {FAIL_LIST_PATH}")
+    return 0 if not failed else 2
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/collect_retry_utils.py b/src/gyxx_flow/modules/product_commerce/runtime/collect_retry_utils.py
new file mode 100644
index 0000000..06a51e3
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/collect_retry_utils.py
@@ -0,0 +1,166 @@
+"""
+collect 脚本共享工具: per-page / per-style 重试 + 失败清单落盘。
+
+4 个 collect 脚本 (collect_erp_yesterday_metrics / jd_product_data_collector /
+dy_product_scraping / taobao_sycm_collect_backfill) 都需要:
+  - 对单次失败 (页面加载 / 单 SPU 提取 / 翻月点日) 做 N 次重试
+  - 记录哪些 (style / code / page) 失败了, 写到 data/logs/failed_<date>_<platform>.json
+    方便 backfill_one_day.py 跑完合并 + 用户排查
+
+设计:
+  - retry_step(fn, *args, max_attempts=2, sleep_seconds=5, label="") → 返回 fn 结果或 raise
+  - FailedCollector 累积失败项, .write() 一次性落盘 (无失败时跳过, 不产生空文件)
+"""
+
+from __future__ import annotations
+
+import json
+import time
+from datetime import datetime
+from pathlib import Path
+from typing import Any, Callable, TypeVar
+
+from runtime_paths import LOG_ROOT
+
+T = TypeVar("T")
+
+
+def retry_step(
+    fn: Callable[..., T],
+    *args: Any,
+    max_attempts: int = 2,
+    sleep_seconds: float = 5.0,
+    label: str = "",
+    on_retry: Callable[[int, BaseException], None] | None = None,
+) -> T:
+    """通用重试: 调 fn(*args), 失败时按 sleep_seconds 线性退避重试, 最多 max_attempts 次。
+
+    - max_attempts=2: 第 1 次失败 → 睡 → 第 2 次失败 → 抛最后一次异常
+    - max_attempts=3: 第 1 次失败 → 睡 → 第 2 次失败 → 睡 → 第 3 次失败 → 抛
+    - label 用于日志, 标注重试的是哪一步 (例: "JD 10465" / "TM 翻月 step 3")
+    - on_retry(attempt, exc) 是可选回调, 用于在重试前做副作用 (如刷新页面)
+    """
+    last_exc: BaseException | None = None
+    prefix = f"[{label}]" if label else ""
+    for attempt in range(1, max_attempts + 1):
+        try:
+            return fn(*args)
+        except BaseException as exc:
+            last_exc = exc
+            if attempt >= max_attempts:
+                break
+            print(f"  {prefix} 第 {attempt}/{max_attempts - 1} 次重试: {type(exc).__name__}: {exc}",
+                  flush=True)
+            if on_retry:
+                try:
+                    on_retry(attempt, exc)
+                except BaseException as cb_exc:
+                    print(f"  {prefix} on_retry 回调失败: {cb_exc}", flush=True)
+            time.sleep(sleep_seconds)
+    assert last_exc is not None
+    raise last_exc
+
+
+class FailedCollector:
+    """累积 collect 脚本里碰到的失败项, 末尾 .write() 一次性落盘。"""
+
+    def __init__(self, platform: str, target_date: str, root: Path | None = None) -> None:
+        self.platform = platform
+        self.target_date = target_date
+        self.root = root or LOG_ROOT / "failures"
+        self.failed_pages: list[dict] = []
+        self.failed_styles: list[dict] = []
+        self.failed_codes: list[dict] = []
+        self._written = False
+
+    def add_page(self, label: str, error: str, attempts: int = 1) -> None:
+        self.failed_pages.append({
+            "label": label, "error": error, "attempts": attempts,
+        })
+
+    def add_style(self, style: str, error: str, attempts: int = 1, codes: list[str] | None = None) -> None:
+        self.failed_styles.append({
+            "style": style, "error": error, "attempts": attempts,
+            "codes": codes or [],
+        })
+
+    def add_code(self, style: str, code: str, error: str, attempts: int = 1) -> None:
+        self.failed_codes.append({
+            "style": style, "code": code, "error": error, "attempts": attempts,
+        })
+
+    def has_failures(self) -> bool:
+        return bool(self.failed_pages or self.failed_styles or self.failed_codes)
+
+    def write(self) -> Path | None:
+        """无失败时返回 None 不写盘; 有失败时写到 data/logs/failed_<date>_<platform>.json。"""
+        if self._written:
+            return None
+        self._written = True
+        if not self.has_failures():
+            return None
+        self.root.mkdir(parents=True, exist_ok=True)
+        out = self.root / f"failed_{self.target_date}_{self.platform}.json"
+        payload = {
+            "platform": self.platform,
+            "target_date": self.target_date,
+            "generated_at": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+            "summary": {
+                "failed_pages": len(self.failed_pages),
+                "failed_styles": len(self.failed_styles),
+                "failed_codes": len(self.failed_codes),
+            },
+            "failed_pages": self.failed_pages,
+            "failed_styles": self.failed_styles,
+            "failed_codes": self.failed_codes,
+        }
+        out.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8-sig")
+        return out
+
+    def summary_line(self) -> str:
+        if not self.has_failures():
+            return f"{self.platform}: 全成功"
+        return (f"{self.platform}: 失败 pages={len(self.failed_pages)} "
+                f"styles={len(self.failed_styles)} codes={len(self.failed_codes)}")
+
+
+def consolidate_failed_logs(target_date: str, root: Path | None = None) -> Path | None:
+    """backfill_one_day.py 末尾调: 合并 data/logs/failed_<date>_*.json 成 _summary.json。
+
+    返回 summary 文件路径, 没失败清单时返回 None。
+    """
+    root = root or LOG_ROOT / "failures"
+    files = sorted(root.glob(f"failed_{target_date}_*.json"))
+    files = [f for f in files if not f.name.endswith("_summary.json")]
+    if not files:
+        return None
+    merged: list[dict] = []
+    for fp in files:
+        try:
+            payload = json.loads(fp.read_text(encoding="utf-8-sig"))
+        except Exception:
+            continue
+        merged.append(payload)
+    if not merged:
+        return None
+    total_pages = sum(p["summary"]["failed_pages"] for p in merged)
+    total_styles = sum(p["summary"]["failed_styles"] for p in merged)
+    total_codes = sum(p["summary"]["failed_codes"] for p in merged)
+    summary = {
+        "target_date": target_date,
+        "generated_at": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+        "total_failed_pages": total_pages,
+        "total_failed_styles": total_styles,
+        "total_failed_codes": total_codes,
+        "platforms": [
+            {
+                "platform": p["platform"],
+                "file": p.get("file", str(files[i])),
+                "summary": p["summary"],
+            }
+            for i, p in enumerate(merged)
+        ],
+    }
+    out = root / f"failed_{target_date}_summary.json"
+    out.write_text(json.dumps(summary, ensure_ascii=False, indent=2), encoding="utf-8-sig")
+    return out
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/collect_sycm_market_rank.py b/src/gyxx_flow/modules/product_commerce/runtime/collect_sycm_market_rank.py
new file mode 100644
index 0000000..329c6df
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/collect_sycm_market_rank.py
@@ -0,0 +1,1784 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""从生意参谋榜单 HTML 采集九品类市场排行并生成飞书文档。"""
+
+from __future__ import annotations
+
+import argparse
+import hashlib
+import html as html_std
+import json
+import os
+import re
+import shutil
+import subprocess
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from runtime_paths import (
+    DATA_HOME,
+    RAW_DATA_ROOT,
+    STATE_ROOT,
+    TMP_ROOT,
+    artifact_relative_path,
+)
+from typing import Any
+from urllib.parse import parse_qs, urljoin, urlparse
+
+from dotenv import load_dotenv
+from lxml import etree, html
+
+import taobao_sycm_products as sycm
+from feishu_doc_native import append_native_category_tables
+from market_rank_report_archive import archive_market_rank_report
+
+
+load_dotenv()
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+OUTPUT_DIR = RAW_DATA_ROOT / "sycm_market_rank"
+DEBUG_DIR = TMP_ROOT / "debug" / "sycm_market_rank"
+CHECKPOINT_PATH = STATE_ROOT / "checkpoints" / "sycm_market_rank.json"
+MARKET_RANK_URL = "https://sycm.taobao.com/mc/free/market_rank"
+DEFAULT_CATEGORY = "箱包皮具/热销女包/男包"
+TARGET_MARKET_CATEGORIES = (
+    {
+        "name": "旅行箱",
+        "min_price": 400,
+        "max_price": 2000,
+        "paths": ((DEFAULT_CATEGORY, "旅行箱"),),
+    },
+    {
+        "name": "运动包",
+        "min_price": 200,
+        "max_price": 1000,
+        "paths": ((DEFAULT_CATEGORY, "旅行袋"),),
+    },
+    {
+        "name": "托特包",
+        "min_price": 200,
+        "max_price": 3000,
+        "paths": ((DEFAULT_CATEGORY, "女士包袋新", "托特包"),),
+    },
+    {
+        "name": "双肩包",
+        "min_price": 200,
+        "max_price": 2000,
+        "paths": ((DEFAULT_CATEGORY, "双肩背包"),),
+    },
+    {
+        "name": "斜挎包",
+        "min_price": 200,
+        "max_price": 2000,
+        "paths": (
+            (DEFAULT_CATEGORY, "男士包袋"),
+            (DEFAULT_CATEGORY, "女士包袋新", "通用款女包"),
+        ),
+    },
+    {
+        "name": "胸包",
+        "min_price": 200,
+        "max_price": 1000,
+        "paths": ((DEFAULT_CATEGORY, "胸包"),),
+    },
+    {
+        "name": "手机包",
+        "min_price": 100,
+        "max_price": 1000,
+        "paths": ((DEFAULT_CATEGORY, "功能小包", "手机包"),),
+    },
+    {
+        "name": "电脑包",
+        "min_price": 200,
+        "max_price": 1000,
+        "paths": (("3C数码配件", "笔记本电脑配件", "笔记本电脑包"),),
+    },
+    {
+        "name": "相机包",
+        "min_price": 200,
+        "max_price": 2000,
+        "paths": (("3C数码配件", "数码相机配件", "数码相机包"),),
+    },
+)
+DOCUMENT_IMAGE_WIDTH = 160
+DOCUMENT_IMAGE_HEIGHT = 160
+RISK_CONTROL_PATTERNS = (
+    (
+        "操作太频繁了，请稍后再试",
+        re.compile(r"操作太频繁(?:了)?[，,、\s]*请稍后再试"),
+    ),
+    (
+        "请稍等片刻再点击刷新哦",
+        re.compile(r"请稍等片刻再点击刷新"),
+    ),
+    (
+        "访问过于频繁",
+        re.compile(r"(?:访问|操作|请求).{0,8}(?:过于|太|过度)频繁"),
+    ),
+    (
+        "生意参谋 Baxia 风控遮罩",
+        re.compile(r"baxia-dialog(?:-mask)?", re.IGNORECASE),
+    ),
+)
+
+
+class SycmRiskControlError(RuntimeError):
+    """生意参谋频控页，需要重新登录并恢复采集位置。"""
+
+
+def detect_risk_control_text(value: str, *, include_baxia: bool = True) -> str:
+    """返回命中的生意参谋风控提示；正常页面返回空字符串。"""
+    compact = re.sub(r"\s+", " ", value or "").strip()
+    for label, pattern in RISK_CONTROL_PATTERNS:
+        if not include_baxia and "Baxia" in label:
+            continue
+        if pattern.search(compact):
+            return label
+    return ""
+
+
+def _risk_control_reason(page) -> str:
+    baxia_visibility_checked = False
+    try:
+        dialogs = page.locator(".baxia-dialog")
+        baxia_visibility_checked = True
+        for index in range(dialogs.count()):
+            if dialogs.nth(index).is_visible():
+                return "生意参谋 Baxia 风控遮罩"
+    except Exception:
+        # 非浏览器测试替身或页面正在切换时，继续用 HTML 标记兜底。
+        baxia_visibility_checked = False
+
+    candidates: list[str] = []
+    try:
+        candidates.append(page.locator("body").inner_text(timeout=2_000))
+    except Exception:
+        pass
+    try:
+        candidates.append(page.content())
+    except Exception:
+        pass
+    for candidate in candidates:
+        reason = detect_risk_control_text(
+            candidate,
+            include_baxia=not baxia_visibility_checked,
+        )
+        if reason:
+            return reason
+    return ""
+
+
+def wait_for_manual_risk_clearance(
+    page,
+    reason: str,
+    *,
+    poll_seconds: float = 2.0,
+    timeout_seconds: int = 21600,
+) -> None:
+    """保持风控页和浏览器打开，等待用户手动完成滑块或刷新。"""
+    _snapshot_risk_control(page, reason)
+    print(
+        "[RISK] 已暂停榜单采集并保留浏览器，"
+        "请手动完成滑块或刷新验证..."
+    )
+    deadline = time.time() + max(1, int(timeout_seconds))
+    while time.time() < deadline:
+        if page.is_closed():
+            raise SycmRiskControlError("等待手动验证时榜单页被关闭")
+        current_reason = _risk_control_reason(page)
+        if not current_reason:
+            try:
+                page.wait_for_load_state("domcontentloaded", timeout=10_000)
+            except Exception:
+                pass
+            time.sleep(2)
+            print("[RISK] 检测到验证页面已解除，继续当前商品")
+            return
+        time.sleep(max(0.5, float(poll_seconds)))
+    raise SycmRiskControlError("等待手动处理风控超时")
+
+
+def _snapshot_risk_control(page, reason: str) -> None:
+    DEBUG_DIR.mkdir(parents=True, exist_ok=True)
+    stamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    prefix = DEBUG_DIR / f"risk_control_{stamp}"
+    try:
+        prefix.with_suffix(".html").write_text(
+            page.content(),
+            encoding="utf-8",
+        )
+    except Exception:
+        pass
+    try:
+        page.screenshot(path=str(prefix.with_suffix(".png")), full_page=False)
+    except Exception:
+        pass
+    print(f"[RISK] 检测到生意参谋风控: {reason}")
+
+
+def close_market_rank_browser(
+    session,
+    *,
+    user_data_dir: str,
+    cdp_url: str | None,
+) -> None:
+    """关闭本次启动的天猫 Chrome；CDP 接管的外部浏览器只断开连接。"""
+    try:
+        session.close()
+    except Exception as exc:
+        print(f"[WARN] Scrapling 会话关闭异常: {exc}")
+
+    if cdp_url or sys.platform != "win32":
+        return
+
+    # Patchright/Chrome 偶尔在 session.close() 返回后仍延迟退出。只匹配本次
+    # user-data-dir，避免影响日常 Chrome 或京东、抖音的独立 profile。
+    time.sleep(2)
+    env = os.environ.copy()
+    env["TM_PROFILE_TO_CLOSE"] = str(user_data_dir)
+    script = r"""
+$target = $env:TM_PROFILE_TO_CLOSE.Replace('\', '/').ToLowerInvariant()
+Get-CimInstance Win32_Process -Filter "Name = 'chrome.exe'" |
+    Where-Object {
+        $_.CommandLine -and
+        $_.CommandLine.Replace('\', '/').ToLowerInvariant().Contains(
+            '--user-data-dir=' + $target
+        )
+    } |
+    ForEach-Object {
+        Stop-Process -Id $_.ProcessId -Force -ErrorAction SilentlyContinue
+    }
+"""
+    subprocess.run(
+        ["powershell", "-NoProfile", "-Command", script],
+        cwd=PROJECT_ROOT,
+        env=env,
+        capture_output=True,
+        text=True,
+        timeout=20,
+        check=False,
+    )
+
+
+def _number_value(text: str) -> float:
+    compact = re.sub(r"[,\s]", "", text or "")
+    match = re.search(r"(\d+(?:\.\d+)?)\s*(万|亿)?", compact)
+    if not match:
+        raise ValueError(f"无法解析数字: {text}")
+    value = float(match.group(1))
+    if match.group(2) == "万":
+        value *= 10_000
+    elif match.group(2) == "亿":
+        value *= 100_000_000
+    return value
+
+
+def parse_range_upper(text: str) -> int:
+    """将 2500~5000 / 7.5万~10万 等区间转换为上限整数。"""
+    parts = re.split(r"\s*(?:~|～|—|–|至)\s*", text or "")
+    candidates = [part for part in parts if re.search(r"\d", part)]
+    if not candidates:
+        raise ValueError(f"支付买家数没有数字: {text}")
+    return int(round(_number_value(candidates[-1])))
+
+
+def _absolute_url(value: str) -> str:
+    value = html_std.unescape((value or "").strip())
+    if not value or value.lower().startswith(("javascript:", "data:")):
+        return ""
+    if value.startswith("//"):
+        return "https:" + value
+    return urljoin("https://sycm.taobao.com/", value)
+
+
+def _product_item_id(url: str) -> str:
+    return (parse_qs(urlparse(url or "").query).get("id") or [""])[0]
+
+
+def _original_taobao_image_url(url: str) -> str:
+    """移除淘宝图片 CDN 的尺寸后缀，恢复原始商品图。"""
+    return re.sub(
+        r"_\d+x\d+(?:q\d+)?\.(?:jpe?g|png|webp)$",
+        "",
+        url or "",
+        flags=re.IGNORECASE,
+    )
+
+
+def _save_checkpoint(products: list[dict[str, Any]]) -> None:
+    CHECKPOINT_PATH.parent.mkdir(parents=True, exist_ok=True)
+    CHECKPOINT_PATH.write_text(
+        json.dumps(products, ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+
+
+def _text(node) -> str:
+    return re.sub(r"\s+", " ", " ".join(node.itertext())).strip()
+
+
+def parse_market_rank_html(source: str) -> list[dict[str, Any]]:
+    """按表头定位列，从市场排行 HTML 中提取商品行。"""
+    if not source.strip():
+        return []
+    root = html.fromstring(source)
+    results: list[dict[str, Any]] = []
+    for table in root.xpath("//table"):
+        header_cells = table.xpath(".//thead//th")
+        if not header_cells:
+            first_row = table.xpath(".//tr[1]")
+            header_cells = first_row[0].xpath("./th|./td") if first_row else []
+        headers = [_text(cell) for cell in header_cells]
+        product_idx = next(
+            (idx for idx, name in enumerate(headers) if name == "商品"),
+            None,
+        )
+        buyer_idx = next(
+            (idx for idx, name in enumerate(headers) if "支付买家数" in name),
+            None,
+        )
+        if product_idx is None or buyer_idx is None:
+            continue
+
+        rows = table.xpath(".//tbody/tr") or table.xpath(".//tr[position()>1]")
+        for ordinal, tr in enumerate(rows, 1):
+            cells = tr.xpath("./td")
+            if max(product_idx, buyer_idx) >= len(cells):
+                continue
+            product_cell = cells[product_idx]
+            buyer_range = _text(cells[buyer_idx])
+            try:
+                buyer_max = parse_range_upper(buyer_range)
+            except ValueError:
+                continue
+
+            links = product_cell.xpath(".//a[@href]")
+            name_link = next(
+                (
+                    link
+                    for link in links
+                    if (link.get("title") or "").strip() or _text(link)
+                ),
+                None,
+            )
+            detail_link = (
+                name_link if name_link is not None else (links[0] if links else None)
+            )
+            product_url = (
+                _absolute_url(detail_link.get("href")) if detail_link is not None else ""
+            )
+            name = ""
+            if name_link is not None:
+                name = (name_link.get("title") or "").strip() or _text(name_link)
+            if not name:
+                name = _text(product_cell)
+            images = product_cell.xpath(".//img")
+            image_url = ""
+            if images:
+                for attr in ("src", "data-src", "data-lazy-src", "data-original"):
+                    image_url = _absolute_url(images[0].get(attr) or "")
+                    if image_url:
+                        break
+            rank_text = _text(cells[0]) if cells else str(ordinal)
+            rank_match = re.search(r"\d+", rank_text)
+            rank = int(rank_match.group()) if rank_match else ordinal
+            if name:
+                results.append(
+                    {
+                        "rank": rank,
+                        "name": name,
+                        "buyer_range": buyer_range,
+                        "buyer_max": buyer_max,
+                        "image_url": image_url,
+                        "product_url": product_url,
+                    }
+                )
+        if results:
+            break
+    return results
+
+
+def build_feishu_doc_xml(
+    products: list[dict[str, Any]],
+    *,
+    min_price: int,
+    max_price: int,
+    title: str,
+    period: str = "30天",
+    independent_prices: bool = False,
+) -> str:
+    category_labels = list(
+        dict.fromkeys(
+            item.get("category_name")
+            for item in products
+            if item.get("category_name")
+        )
+    )
+    category_paths = {
+        item.get("category_path")
+        for item in products
+        if item.get("category_path")
+    }
+    intro = build_feishu_intro_xml(
+        title=title,
+        min_price=min_price,
+        max_price=max_price,
+        total=len(products),
+        category_labels=category_labels,
+        category_count=len(category_paths) or len(category_labels),
+        period=period,
+        independent_prices=independent_prices,
+    )
+    return intro + build_feishu_batch_xml(products, start_rank=1, include_heading=False)
+
+
+def build_feishu_intro_xml(
+    *,
+    title: str,
+    min_price: int,
+    max_price: int,
+    total: int,
+    category_labels: list[str] | None = None,
+    category_count: int = 0,
+    period: str = "30天",
+    independent_prices: bool = False,
+) -> str:
+    esc = lambda value: html_std.escape(str(value or ""), quote=True)
+    generated = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    if category_labels:
+        category_summary = (
+            f"目标品类 {category_count or len(category_labels)} 个："
+            f"{'、'.join(esc(name) for name in category_labels)}；"
+        )
+    else:
+        category_summary = f"一级类目“{esc(DEFAULT_CATEGORY)}”；"
+    price_summary = (
+        "各品类使用独立自定义价格区间；"
+        if independent_prices
+        else f"自定义价格 {min_price}–{max_price}；"
+    )
+    return (
+        f"<title>{esc(title)}</title>"
+        "<h1>天猫</h1>"
+        f"<p>采集时间：{generated}</p>"
+        f'<callout emoji="📊" background-color="light-blue" border-color="blue">'
+        f"<p><b>筛选条件：</b>{category_summary}"
+        f"统计周期 {esc(period)}；{price_summary}"
+        f"共采集 <b>{total}</b> 个商品。支付买家数取区间上限。</p>"
+        "</callout>"
+    )
+
+
+def build_feishu_batch_xml(
+    products: list[dict[str, Any]],
+    *,
+    start_rank: int,
+    include_heading: bool = True,
+) -> str:
+    esc = lambda value: html_std.escape(str(value or ""), quote=True)
+    rows = []
+    for item in products:
+        original_image_url = _original_taobao_image_url(
+            item.get("image_url") or ""
+        )
+        image = (
+            f'<img href="{esc(original_image_url)}" '
+            f'width="{DOCUMENT_IMAGE_WIDTH}" '
+            f'height="{DOCUMENT_IMAGE_HEIGHT}"/>'
+            if original_image_url
+            else "未获取"
+        )
+        link = item.get("product_url") or ""
+        name = esc(item.get("name"))
+        product_link_cell = (
+            f'<a href="{esc(link)}">打开商品</a>' if link else "未解析到"
+        )
+        sales_amount = esc(
+            item.get("sales_amount_range")
+            or item.get("sales_amount")
+            or "未获取"
+        )
+        rows.append(
+            "<tr>"
+            "<td></td>"
+            "<td></td>"
+            f"<td>{esc(item.get('category_name') or '未分类')}</td>"
+            f"<td>{esc(item.get('rank'))}</td>"
+            f"<td>{image}</td>"
+            f"<td>{name}</td>"
+            f"<td>{esc(item.get('buyer_range'))}</td>"
+            f"<td><b>{esc(item.get('buyer_max'))}</b></td>"
+            f"<td>{sales_amount}</td>"
+            f"<td>{product_link_cell}</td>"
+            "</tr>"
+        )
+    end_rank = start_rank + len(products) - 1
+    heading = (
+        f"<h2>榜单记录 {start_rank}–{end_rank}</h2>"
+        if include_heading
+        else ""
+    )
+    return (
+        heading + "<table>"
+        "<colgroup>"
+        '<col width="110"/>'
+        '<col width="120"/>'
+        '<col width="100"/>'
+        '<col width="60"/>'
+        '<col width="180"/>'
+        '<col width="360"/>'
+        '<col width="140"/>'
+        '<col width="140"/>'
+        '<col width="140"/>'
+        '<col width="100"/>'
+        "</colgroup>"
+        "<thead><tr>"
+        '<th background-color="light-gray">风格（运营手动区分）</th>'
+        '<th background-color="light-gray">场景用途（运营手动区分）</th>'
+        '<th background-color="light-gray">品类</th>'
+        '<th background-color="light-gray">排名</th>'
+        '<th background-color="light-gray">商品图</th>'
+        '<th background-color="light-gray">商品名称</th>'
+        '<th background-color="light-gray">支付买家数区间</th>'
+        '<th background-color="light-gray">支付买家数上限</th>'
+        '<th background-color="light-gray">销售额</th>'
+        '<th background-color="light-gray">商品链接</th>'
+        "</tr></thead>"
+        f"<tbody>{''.join(rows)}</tbody>"
+        "</table>"
+    )
+
+
+def _first_visible(locator):
+    for idx in range(locator.count()):
+        item = locator.nth(idx)
+        try:
+            if item.is_visible():
+                return item
+        except Exception:
+            continue
+    return None
+
+
+def navigate_to_market_rank(page) -> None:
+    if "/market_rank" not in page.url:
+        market = _first_visible(page.get_by_text("市场", exact=True))
+        if market is not None:
+            market.click(timeout=10_000)
+            time.sleep(2)
+        ranking = _first_visible(page.get_by_text("市场排行", exact=True))
+        if ranking is not None:
+            ranking.click(timeout=10_000)
+        else:
+            page.goto(MARKET_RANK_URL, wait_until="domcontentloaded", timeout=60_000)
+    page.wait_for_selector("text=自定义价格", timeout=60_000)
+    sycm.close_ad_popup(page)
+    sycm.close_blocking_tips(page)
+
+
+def select_market_category_path(page, category_path: tuple[str, ...]) -> None:
+    """按一级→二级→三级路径悬停展开菜单，并点击最终目标品类。"""
+    if not category_path:
+        raise ValueError("类目路径不能为空")
+    picker = _first_visible(page.locator(".common-picker-header"))
+    if picker is None:
+        raise RuntimeError("市场排行页未找到“切换类目”选择器")
+    current = (picker.get_attribute("title") or "").strip()
+    leaf = category_path[-1]
+    if leaf in current:
+        print(f"[OK] 当前已是目标品类: {' > '.join(category_path)}")
+        return
+
+    picker.click(timeout=10_000)
+    for index, category in enumerate(category_path):
+        category_item = _first_visible(page.get_by_title(category, exact=True))
+        if category_item is None:
+            category_item = _first_visible(
+                page.locator("li.tree-item.common-item").filter(
+                    has_text=re.compile(rf"^\s*{re.escape(category)}\s*$")
+                )
+            )
+        if category_item is None:
+            raise RuntimeError(
+                f"类目菜单未找到“{category}”，完整路径: "
+                f"{' > '.join(category_path)}"
+            )
+        if index < len(category_path) - 1:
+            category_item.hover(timeout=10_000)
+            next_category = category_path[index + 1]
+            page.get_by_title(next_category, exact=True).first.wait_for(
+                state="visible",
+                timeout=10_000,
+            )
+        else:
+            category_item.click(timeout=10_000)
+    try:
+        page.wait_for_function(
+            """leaf => {
+                const el = document.querySelector('.common-picker-header');
+                const value = el
+                    ? ((el.getAttribute('title') || el.textContent || '').trim())
+                    : '';
+                return value.includes(leaf);
+            }""",
+            arg=leaf,
+            timeout=15_000,
+        )
+    except Exception:
+        current = (
+            page.locator(".common-picker-header").first.get_attribute("title") or ""
+        ).strip()
+        if leaf not in current:
+            raise RuntimeError(
+                f"目标类目点击后未生效，当前类目仍是: {current or '未知'}"
+            )
+    active_page = _first_visible(page.locator(".ant-pagination-item-active"))
+    if active_page is not None and active_page.inner_text().strip() != "1":
+        first_page = _first_visible(page.locator(".ant-pagination-item-1"))
+        if first_page is not None:
+            first_page.click(timeout=10_000)
+    time.sleep(3)
+    print(f"[OK] 已选择目标品类: {' > '.join(category_path)}")
+
+
+def select_market_category(page, category: str) -> None:
+    """兼容旧调用：选择单层类目。"""
+    select_market_category_path(page, (category,))
+
+
+def select_market_period(page, period: str = "7天") -> None:
+    """固定排行统计周期，避免继承浏览器上一次的实时/日/月状态。"""
+    period_text = _first_visible(page.get_by_text(period, exact=True))
+    if period_text is None:
+        raise RuntimeError(f"市场排行页未找到统计周期: {period}")
+    button = period_text.locator("xpath=ancestor-or-self::button[1]")
+    if button.count() == 0:
+        raise RuntimeError(f"统计周期“{period}”不在按钮中")
+    classes = button.get_attribute("class") or ""
+    if "ant-btn-primary" not in classes:
+        button.click(timeout=10_000)
+        time.sleep(3)
+    print(f"[OK] 已选择统计周期: {period}")
+
+
+def prepare_market_rank_period(
+    context,
+    page,
+    *,
+    period: str,
+    risk_mode: str,
+    cooldown_seconds: int,
+    max_retries: int,
+):
+    """处理进入榜单后、点击统计周期前出现的 Baxia 验证码。
+
+    context 参数为兼容旧调用保留。恢复时刻意不清 Cookie、不重新登录，避免把
+    已经建立的可信会话变成全新会话并进一步触发 Baxia。
+    """
+    del context
+    recoveries = 0
+    while True:
+        reason = _risk_control_reason(page)
+        if not reason:
+            try:
+                select_market_period(page, period)
+                return page
+            except Exception:
+                reason = _risk_control_reason(page)
+                if not reason:
+                    raise
+
+        if risk_mode == "manual":
+            wait_for_manual_risk_clearance(page, reason)
+            continue
+
+        _snapshot_risk_control(page, reason)
+        recoveries += 1
+        if recoveries > max(0, int(max_retries)):
+            raise SycmRiskControlError(
+                f"选择统计周期前风控恢复超过 {max_retries} 次: {reason}"
+            )
+        wait_seconds = min(
+            max(0, int(cooldown_seconds)) * recoveries,
+            300,
+        )
+        if wait_seconds:
+            print(f"[RISK] 选择统计周期前等待 {wait_seconds} 秒，保留当前登录会话...")
+            time.sleep(wait_seconds)
+        print("[RISK] 保留 Cookie 刷新当前榜单页，不重新登录...")
+        page.reload(wait_until="domcontentloaded", timeout=60_000)
+        navigate_to_market_rank(page)
+
+
+def prepare_market_rank_filters(
+    page,
+    *,
+    period: str,
+    category_path: tuple[str, ...],
+    min_price: int,
+    max_price: int,
+    risk_mode: str,
+    cooldown_seconds: int,
+    max_retries: int,
+):
+    """在同一会话内设置周期、类目和价格，并在 Baxia 验证后重试整组筛选。"""
+    recoveries = 0
+    while True:
+        reason = _risk_control_reason(page)
+        if not reason:
+            try:
+                select_market_period(page, period)
+                select_market_category_path(page, category_path)
+                apply_custom_price(page, min_price, max_price)
+                return page
+            except Exception:
+                reason = _risk_control_reason(page)
+                if not reason:
+                    raise
+
+        if risk_mode == "manual":
+            wait_for_manual_risk_clearance(page, reason)
+            continue
+
+        _snapshot_risk_control(page, reason)
+        recoveries += 1
+        if recoveries > max(0, int(max_retries)):
+            raise SycmRiskControlError(
+                f"设置 {' > '.join(category_path)} 筛选时风控恢复超过 "
+                f"{max_retries} 次: {reason}"
+            )
+        wait_seconds = min(
+            max(0, int(cooldown_seconds)) * recoveries,
+            300,
+        )
+        if wait_seconds:
+            print(f"[RISK] 等待 {wait_seconds} 秒，保留当前登录会话...")
+            time.sleep(wait_seconds)
+        print("[RISK] 保留 Cookie 刷新榜单页，然后重新设置当前品类筛选...")
+        page.reload(wait_until="domcontentloaded", timeout=60_000)
+        navigate_to_market_rank(page)
+
+
+def apply_custom_price(page, min_price: int, max_price: int) -> None:
+    modify = _first_visible(page.get_by_text("修改", exact=True))
+    custom = _first_visible(
+        page.get_by_text(re.compile(r"^自定义价格"), exact=False)
+    )
+    trigger = modify if modify is not None else custom
+    if trigger is None:
+        raise RuntimeError("市场排行页未找到“自定义价格”")
+    trigger.click(timeout=10_000)
+    dialog = page.locator('[role="dialog"]').filter(has_text="自定义价格").last
+    dialog.wait_for(state="visible", timeout=10_000)
+    DEBUG_DIR.mkdir(parents=True, exist_ok=True)
+    (DEBUG_DIR / "custom_price_dialog.html").write_text(
+        dialog.evaluate("el => el.outerHTML"),
+        encoding="utf-8",
+    )
+    page.screenshot(
+        path=str(DEBUG_DIR / "custom_price_dialog.png"),
+        full_page=False,
+    )
+    low_input = _first_visible(page.locator('input[placeholder="请输入最低价"]'))
+    high_input = _first_visible(page.locator('input[placeholder="请输入最高价"]'))
+    if low_input is None or high_input is None:
+        raise RuntimeError(
+            "自定义价格弹窗未找到最低价/最高价输入框；已保存 DOM 快照"
+        )
+    low_input.fill(str(min_price))
+    high_input.fill(str(max_price))
+    confirm = _first_visible(page.get_by_text("确定", exact=True))
+    if confirm is None:
+        raise RuntimeError("自定义价格弹窗未找到“确定”")
+    confirm.click(timeout=10_000)
+    dialog.wait_for(state="hidden", timeout=10_000)
+    time.sleep(3)
+
+
+def _active_rank_page_number(page) -> int:
+    active = _first_visible(page.locator(".ant-pagination-item-active"))
+    if active is None:
+        return 1
+    text = active.inner_text().strip()
+    match = re.search(r"\d+", text)
+    return int(match.group()) if match else 1
+
+
+def _go_to_rank_page(page, target_page: int) -> None:
+    """恢复筛选条件后，将榜单翻到风控前所在页。"""
+    target_page = max(1, int(target_page))
+    while True:
+        current_page = _active_rank_page_number(page)
+        if current_page == target_page:
+            return
+        if current_page > target_page:
+            raise RuntimeError(
+                f"恢复榜单页码失败：当前第 {current_page} 页，"
+                f"目标第 {target_page} 页"
+            )
+        direct = _first_visible(
+            page.locator(f"li.ant-pagination-item-{target_page}")
+        )
+        trigger = direct
+        if trigger is None:
+            trigger = _first_visible(page.locator("li.ant-pagination-next"))
+        if trigger is None:
+            raise RuntimeError(f"恢复时找不到通往第 {target_page} 页的分页按钮")
+        old_first_link = _first_visible(
+            page.locator('p.goodsName a[href*="id="]')
+        )
+        old_href = (
+            old_first_link.get_attribute("href")
+            if old_first_link is not None
+            else ""
+        )
+        trigger.click(timeout=10_000)
+        page.wait_for_function(
+            """state => {
+                const active = document.querySelector(
+                    '.ant-pagination-item-active'
+                );
+                const firstLink = document.querySelector(
+                    'p.goodsName a[href*="id="]'
+                );
+                const activePage = active
+                    ? Number((active.textContent || '').trim())
+                    : 1;
+                const href = firstLink
+                    ? (firstLink.getAttribute('href') || '')
+                    : '';
+                return activePage > state.oldPage &&
+                    (!state.oldHref || (href && href !== state.oldHref));
+            }""",
+            arg={"oldPage": current_page, "oldHref": old_href or ""},
+            timeout=20_000,
+        )
+        time.sleep(0.5)
+
+
+def recover_market_rank_after_risk(
+    context,
+    page,
+    *,
+    period: str,
+    category_path: tuple[str, ...],
+    min_price: int,
+    max_price: int,
+    target_page: int,
+    cooldown_seconds: int,
+    recovery_attempt: int,
+    risk_mode: str = "auto",
+):
+    """保留当前登录会话，并恢复风控前的筛选条件和页码。"""
+    del context
+    wait_seconds = min(
+        max(0, int(cooldown_seconds)) * max(1, int(recovery_attempt)),
+        300,
+    )
+    if wait_seconds:
+        print(f"[RISK] 等待 {wait_seconds} 秒，保留当前登录会话...")
+        time.sleep(wait_seconds)
+    print("[RISK] 保留 Cookie 刷新榜单页，不清会话、不重新登录...")
+    page.reload(wait_until="domcontentloaded", timeout=60_000)
+    navigate_to_market_rank(page)
+    prepare_market_rank_filters(
+        page,
+        period=period,
+        category_path=category_path,
+        min_price=min_price,
+        max_price=max_price,
+        risk_mode=risk_mode,
+        cooldown_seconds=cooldown_seconds,
+        max_retries=1,
+    )
+    _go_to_rank_page(page, target_page)
+    print(
+        f"[RISK] 已恢复 {' > '.join(category_path)} 第 {target_page} 页，"
+        "继续当前商品"
+    )
+    return page
+
+
+def collect_all_rank_pages(
+    page,
+    max_pages: int = 0,
+    limit: int = 0,
+    category_name: str = "",
+    category_path: str = "",
+    existing_products: list[dict[str, Any]] | None = None,
+    recover_from_risk=None,
+    risk_max_retries: int = 6,
+    manual_risk_wait: bool = False,
+) -> list[dict[str, Any]]:
+    products: list[dict[str, Any]] = []
+    seen: set[tuple[str, str]] = set()
+    existing_products = existing_products or []
+    debug_category = re.sub(r"[^\w\u4e00-\u9fff-]+", "_", category_name).strip("_")
+    debug_category = debug_category or "default"
+    page_no = 1
+    risk_recoveries = 0
+    while True:
+        source = page.content()
+        risk_reason = _risk_control_reason(page)
+        if risk_reason:
+            if manual_risk_wait:
+                wait_for_manual_risk_clearance(page, risk_reason)
+                risk_recoveries += 1
+                _save_checkpoint([*existing_products, *products])
+                if recover_from_risk is not None:
+                    page = recover_from_risk(
+                        page,
+                        page_no,
+                        risk_recoveries,
+                        risk_reason,
+                    )
+                continue
+            _snapshot_risk_control(page, risk_reason)
+            risk_recoveries += 1
+            _save_checkpoint([*existing_products, *products])
+            if recover_from_risk is None or risk_recoveries > risk_max_retries:
+                raise SycmRiskControlError(risk_reason)
+            page = recover_from_risk(
+                page,
+                page_no,
+                risk_recoveries,
+                risk_reason,
+            )
+            continue
+        DEBUG_DIR.mkdir(parents=True, exist_ok=True)
+        (DEBUG_DIR / f"market_rank_{debug_category}_page_{page_no}.html").write_text(
+            source, encoding="utf-8"
+        )
+        parsed = parse_market_rank_html(source)
+        if not parsed:
+            raise RuntimeError(f"第 {page_no} 页没有解析到市场排行商品")
+        for item in parsed:
+            item["category_name"] = category_name
+            item["category_path"] = category_path
+            item_id = _product_item_id(item["product_url"])
+            key = ("item_id", item_id) if item_id else ("name", item["name"])
+            if key not in seen:
+                seen.add(key)
+                products.append(item)
+                if limit and len(products) >= limit:
+                    break
+        _save_checkpoint([*existing_products, *products])
+        if limit and len(products) >= limit:
+            return products[:limit]
+        print(f"[OK] 市场排行第 {page_no} 页: {len(parsed)} 条，累计 {len(products)} 条")
+        if max_pages and page_no >= max_pages:
+            break
+        next_button = _first_visible(page.locator("li.ant-pagination-next"))
+        if next_button is None:
+            break
+        disabled = (
+            next_button.get_attribute("disabled") is not None
+            or next_button.get_attribute("aria-disabled") == "true"
+        )
+        classes = next_button.get_attribute("class") or ""
+        if disabled or "disabled" in classes.lower():
+            break
+        first_item_id = _product_item_id(parsed[0]["product_url"])
+        target_page = page_no + 1
+        next_button.click(timeout=10_000)
+        try:
+            page.wait_for_function(
+                """state => {
+                    const active = document.querySelector(
+                        '.ant-pagination-item-active'
+                    );
+                    const firstLink = document.querySelector(
+                        'p.goodsName a[href*="id="]'
+                    );
+                    const activePage = active
+                        ? (active.textContent || '').trim()
+                        : '';
+                    const href = firstLink
+                        ? (firstLink.getAttribute('href') || '')
+                        : '';
+                    return activePage === String(state.targetPage) &&
+                        href &&
+                        (!state.oldItemId ||
+                         !href.includes('id=' + state.oldItemId));
+                }""",
+                arg={
+                    "targetPage": target_page,
+                    "oldItemId": first_item_id,
+                },
+                timeout=20_000,
+            )
+        except Exception as exc:
+            risk_reason = _risk_control_reason(page)
+            if risk_reason:
+                if manual_risk_wait:
+                    wait_for_manual_risk_clearance(page, risk_reason)
+                    page_no = _active_rank_page_number(page)
+                    continue
+                _snapshot_risk_control(page, risk_reason)
+                risk_recoveries += 1
+                _save_checkpoint([*existing_products, *products])
+                if (
+                    recover_from_risk is None
+                    or risk_recoveries > risk_max_retries
+                ):
+                    raise SycmRiskControlError(risk_reason) from exc
+                page = recover_from_risk(
+                    page,
+                    target_page,
+                    risk_recoveries,
+                    risk_reason,
+                )
+                page_no = target_page
+                continue
+            active = _first_visible(page.locator(".ant-pagination-item-active"))
+            active_text = active.inner_text().strip() if active is not None else "未知"
+            raise RuntimeError(
+                f"翻页到第 {target_page} 页后商品列表未刷新，"
+                f"当前页码: {active_text}"
+            ) from exc
+        time.sleep(1)
+        page_no += 1
+    return products
+
+
+def _lark_cli() -> str:
+    candidates = []
+    if os.name == "nt":
+        candidates.extend(
+            [
+                Path.home() / "AppData" / "Roaming" / "npm" / "lark-cli.cmd",
+                Path.home() / "AppData" / "Roaming" / "npm" / "lark-cli.exe",
+            ]
+        )
+    found = shutil.which("lark-cli")
+    if found:
+        candidates.append(Path(found))
+    for candidate in candidates:
+        if Path(candidate).exists():
+            return str(candidate)
+    raise RuntimeError("找不到 lark-cli")
+
+
+def create_feishu_doc(title: str, xml: str) -> str:
+    if not title.strip():
+        raise ValueError("飞书文档标题不能为空")
+    if "?" in title:
+        raise ValueError(
+            f"飞书文档标题疑似发生编码损坏，已拒绝创建: {title!r}"
+        )
+    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+    safe_name = re.sub(r'[<>:"/\\|?*]+', "_", title)
+    xml_path = OUTPUT_DIR / f"{safe_name}.xml"
+    xml_path.write_text(xml, encoding="utf-8")
+    relative = artifact_relative_path(xml_path)
+    proc = subprocess.run(
+        [
+            _lark_cli(),
+            "docs",
+            "+create",
+            "--doc-format",
+            "xml",
+            "--title",
+            title,
+            "--content",
+            f"@{relative}",
+            "--as",
+            "user",
+            "--format",
+            "json",
+        ],
+        cwd=DATA_HOME,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        timeout=300,
+    )
+    if proc.returncode != 0:
+        raise RuntimeError(f"创建飞书文档失败: {proc.stderr or proc.stdout}")
+    obj = json.loads(proc.stdout)
+    data = obj.get("data") or {}
+    document = data.get("document") or {}
+    url = document.get("url") or data.get("url") or obj.get("url")
+    if not url:
+        raise RuntimeError(f"飞书文档创建成功但没有返回链接: {obj}")
+    return url
+
+
+def _feishu_append_xml_path(doc_url: str, batch_index: int) -> Path:
+    # 三个平台并行发布时都会复用此函数。临时文件必须按目标文档隔离，
+    # 否则相同 batch_index 会互相覆盖并把另一平台的表写进当前文档。
+    doc_marker = hashlib.sha256(doc_url.encode("utf-8")).hexdigest()[:12]
+    return OUTPUT_DIR / (
+        f"market_rank_{doc_marker}_batch_{batch_index:02d}.xml"
+    )
+
+
+def append_feishu_doc(doc_url: str, xml: str, batch_index: int) -> None:
+    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+    xml_path = _feishu_append_xml_path(doc_url, batch_index)
+    xml_path.write_text(xml, encoding="utf-8")
+    relative = artifact_relative_path(xml_path)
+    proc = subprocess.run(
+        [
+            _lark_cli(),
+            "docs",
+            "+update",
+            "--doc",
+            doc_url,
+            "--command",
+            "append",
+            "--doc-format",
+            "xml",
+            "--content",
+            f"@{relative}",
+            "--as",
+            "user",
+            "--format",
+            "json",
+        ],
+        cwd=DATA_HOME,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        timeout=300,
+    )
+    if proc.returncode != 0:
+        raise RuntimeError(
+            f"飞书文档第 {batch_index} 批追加失败: {proc.stderr or proc.stdout}"
+        )
+
+
+def _fetch_feishu_doc_content(doc_url: str) -> str:
+    proc = subprocess.run(
+        [
+            _lark_cli(),
+            "docs",
+            "+fetch",
+            "--doc",
+            doc_url,
+            "--detail",
+            "with-ids",
+            "--as",
+            "user",
+            "--format",
+            "json",
+        ],
+        cwd=PROJECT_ROOT,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        timeout=300,
+    )
+    if proc.returncode != 0:
+        raise RuntimeError(f"读取飞书文档失败: {proc.stderr or proc.stdout}")
+    obj = json.loads(proc.stdout)
+    return str((((obj.get("data") or {}).get("document") or {}).get("content")) or "")
+
+
+def inspect_feishu_batch_tables(content: str) -> list[dict[str, Any]]:
+    """按标题关联每张批次表，返回起始记录、商品行数和图片数。"""
+    if not content.strip():
+        return []
+    root = etree.fromstring(
+        f"<root>{content}</root>".encode("utf-8"),
+        parser=etree.XMLParser(recover=True),
+    )
+    start_rank: int | None = None
+    stats: list[dict[str, Any]] = []
+    for child in root:
+        if child.tag == "h2":
+            match = re.search(r"\d+", " ".join(child.itertext()))
+            start_rank = int(match.group()) if match else None
+        elif child.tag == "table" and start_rank is not None:
+            stats.append(
+                {
+                    "block_id": str(child.get("id") or ""),
+                    "start_rank": start_rank,
+                    "rows": len(child.xpath("./tbody/tr")),
+                    "images": len(child.xpath(".//img")),
+                }
+            )
+    return stats
+
+
+def group_products_by_category(
+    products: list[dict[str, Any]],
+) -> list[tuple[str, list[dict[str, Any]]]]:
+    """按首次出现顺序分组，确保同一品类只生成一张表。"""
+    grouped: dict[str, list[dict[str, Any]]] = {}
+    for product in products:
+        category = str(product.get("category_name") or "未分类")
+        grouped.setdefault(category, []).append(product)
+    return list(grouped.items())
+
+
+def build_feishu_category_xml(
+    category_name: str,
+    products: list[dict[str, Any]],
+) -> str:
+    """为品类文档构建紧凑表，确保大品类仍保持一品类一张表。"""
+    esc = lambda value: html_std.escape(str(value or ""), quote=True)
+    heading = html_std.escape(category_name or "未分类", quote=True)
+    rows: list[str] = []
+    for item in products:
+        original_image_url = _original_taobao_image_url(item.get("image_url") or "")
+        image = (
+            f'<img href="{esc(original_image_url)}" '
+            f'width="{DOCUMENT_IMAGE_WIDTH}" height="{DOCUMENT_IMAGE_HEIGHT}"/>'
+            if original_image_url
+            else "未获取"
+        )
+        product_url = item.get("product_url") or ""
+        product_link = (
+            f'<a href="{esc(product_url)}">打开商品</a>'
+            if product_url
+            else "未解析到"
+        )
+        buyer_range = str(item.get("buyer_range") or "未获取")
+        buyer_max = item.get("buyer_max")
+        buyer_summary = buyer_range
+        if buyer_max not in (None, ""):
+            buyer_summary = f"{buyer_range}（上限 {buyer_max}）"
+        rows.append(
+            "<tr><td></td><td></td>"
+            f"<td>{esc(item.get('rank'))}</td>"
+            f"<td>{image}</td>"
+            f"<td>{esc(item.get('name'))}</td>"
+            f"<td><b>{esc(buyer_summary)}</b></td>"
+            f"<td>{product_link}</td></tr>"
+        )
+    table = (
+        "<table><colgroup>"
+        '<col width="110"/><col width="120"/><col width="60"/>'
+        '<col width="180"/><col width="360"/><col width="190"/>'
+        '<col width="100"/>'
+        "</colgroup><thead><tr>"
+        '<th background-color="light-gray">风格（运营手动区分）</th>'
+        '<th background-color="light-gray">场景用途（运营手动区分）</th>'
+        '<th background-color="light-gray">排名</th>'
+        '<th background-color="light-gray">商品图</th>'
+        '<th background-color="light-gray">商品名称</th>'
+        '<th background-color="light-gray">支付买家数（区间/上限）</th>'
+        '<th background-color="light-gray">商品链接</th>'
+        f"</tr></thead><tbody>{''.join(rows)}</tbody></table>"
+    )
+    return f"<h2>{heading}（{len(products)}条）</h2>{table}"
+
+
+def inspect_feishu_category_tables(content: str) -> list[dict[str, Any]]:
+    """回读飞书文档，按品类标题关联其后一张表。"""
+    if not content.strip():
+        return []
+    root = etree.fromstring(
+        f"<root>{content}</root>".encode("utf-8"),
+        parser=etree.XMLParser(recover=True),
+    )
+    pending: tuple[str, int] | None = None
+    stats: list[dict[str, Any]] = []
+    for child in root:
+        if child.tag == "h2":
+            heading = "".join(child.itertext()).strip()
+            match = re.fullmatch(r"(.+?)（(\d+)条）", heading)
+            pending = (match.group(1), int(match.group(2))) if match else None
+        elif child.tag == "table" and pending is not None:
+            category, declared_rows = pending
+            stats.append(
+                {
+                    "block_id": str(child.get("id") or ""),
+                    "category": category,
+                    "declared_rows": declared_rows,
+                    "rows": len(child.xpath("./tbody/tr")),
+                    "images": len(child.xpath(".//img")),
+                }
+            )
+            pending = None
+    return stats
+
+
+def verify_feishu_category_tables(
+    doc_url: str,
+    groups: list[tuple[str, list[dict[str, Any]]]],
+    *,
+    max_checks: int = 7,
+    wait_seconds: float = 5.0,
+) -> None:
+    """确认一品类一表、行数和应有图片均完整后才允许后续入库。"""
+    expected = [
+        {
+            "category": category,
+            "rows": len(items),
+            "images": sum(bool(item.get("image_url")) for item in items),
+        }
+        for category, items in groups
+    ]
+    last_stats: list[dict[str, Any]] = []
+    for attempt in range(max_checks):
+        last_stats = inspect_feishu_category_tables(
+            _fetch_feishu_doc_content(doc_url)
+        )
+        actual = [
+            {
+                "category": item["category"],
+                "rows": item["rows"],
+                "images": item["images"],
+            }
+            for item in last_stats
+        ]
+        if actual == expected:
+            print(
+                f"[CHECK] 飞书品类表完整: {len(expected)} 张表，"
+                f"{sum(item['rows'] for item in expected)} 行"
+            )
+            return
+        if attempt + 1 < max_checks:
+            time.sleep(max(0.0, wait_seconds))
+    raise RuntimeError(
+        "飞书品类表结构校验失败: "
+        f"期望 {expected}，实际 "
+        f"{[{k: item[k] for k in ('category', 'rows', 'images')} for item in last_stats]}"
+    )
+
+
+def _replace_feishu_table(
+    doc_url: str,
+    *,
+    block_id: str,
+    xml: str,
+    start_rank: int,
+) -> None:
+    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+    xml_path = OUTPUT_DIR / f"market_rank_repair_{start_rank:04d}.xml"
+    xml_path.write_text(xml, encoding="utf-8")
+    relative = artifact_relative_path(xml_path)
+    proc = subprocess.run(
+        [
+            _lark_cli(),
+            "docs",
+            "+update",
+            "--doc",
+            doc_url,
+            "--command",
+            "block_replace",
+            "--block-id",
+            block_id,
+            "--doc-format",
+            "xml",
+            "--content",
+            f"@{relative}",
+            "--as",
+            "user",
+            "--format",
+            "json",
+        ],
+        cwd=DATA_HOME,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        timeout=600,
+    )
+    if proc.returncode != 0:
+        raise RuntimeError(
+            f"飞书记录 {start_rank} 起的表格修复失败: {proc.stderr or proc.stdout}"
+        )
+
+
+def repair_feishu_document(
+    doc_url: str,
+    products: list[dict[str, Any]],
+    *,
+    batch_size: int,
+    max_rounds: int = 5,
+    cooldown_seconds: float = 5.0,
+) -> None:
+    """补写缺失批次并重传缺图表格，直到线上行数和图片数完整。"""
+    expected_starts = list(range(1, len(products) + 1, batch_size))
+    for round_index in range(1, max_rounds + 1):
+        stats = inspect_feishu_batch_tables(_fetch_feishu_doc_content(doc_url))
+        by_start = {item["start_rank"]: item for item in stats}
+        missing = [start for start in expected_starts if start not in by_start]
+        for start_rank in missing:
+            offset = start_rank - 1
+            batch = products[offset : offset + batch_size]
+            append_feishu_doc(
+                doc_url,
+                build_feishu_batch_xml(batch, start_rank=start_rank),
+                offset // batch_size + 1,
+            )
+            print(f"[REPAIR] 已补写飞书记录 {start_rank}-{offset + len(batch)}")
+            time.sleep(max(0.0, cooldown_seconds))
+
+        stats = inspect_feishu_batch_tables(_fetch_feishu_doc_content(doc_url))
+        incomplete = [
+            item for item in stats
+            if item["start_rank"] in expected_starts
+            and item["images"] < item["rows"]
+        ]
+        total_rows = sum(
+            item["rows"] for item in stats
+            if item["start_rank"] in expected_starts
+        )
+        total_images = sum(
+            item["images"] for item in stats
+            if item["start_rank"] in expected_starts
+        )
+        print(
+            f"[CHECK] 飞书完整度: 行 {total_rows}/{len(products)}，"
+            f"图片 {total_images}/{len(products)}，缺图表格 {len(incomplete)} 张"
+        )
+        if total_rows == len(products) and not incomplete:
+            return
+        if round_index >= max_rounds:
+            break
+
+        for item in incomplete:
+            start_rank = int(item["start_rank"])
+            # 每次替换后重新读取当前 block id，避免复用已变化的标识。
+            current = {
+                stat["start_rank"]: stat
+                for stat in inspect_feishu_batch_tables(
+                    _fetch_feishu_doc_content(doc_url)
+                )
+            }[start_rank]
+            batch = products[start_rank - 1 : start_rank - 1 + batch_size]
+            _replace_feishu_table(
+                doc_url,
+                block_id=current["block_id"],
+                xml=build_feishu_batch_xml(
+                    batch,
+                    start_rank=start_rank,
+                    include_heading=False,
+                ),
+                start_rank=start_rank,
+            )
+            print(f"[REPAIR] 已重传飞书记录 {start_rank} 起的表格")
+            time.sleep(max(0.0, cooldown_seconds))
+
+    final_stats = inspect_feishu_batch_tables(_fetch_feishu_doc_content(doc_url))
+    final_rows = sum(
+        item["rows"] for item in final_stats
+        if item["start_rank"] in expected_starts
+    )
+    final_images = sum(
+        item["images"] for item in final_stats
+        if item["start_rank"] in expected_starts
+    )
+    if final_rows != len(products) or final_images != len(products):
+        raise RuntimeError(
+            f"飞书文档仍不完整: 行 {final_rows}/{len(products)}，"
+            f"图片 {final_images}/{len(products)}"
+        )
+
+
+def create_feishu_doc_batched(
+    title: str,
+    products: list[dict[str, Any]],
+    *,
+    min_price: int,
+    max_price: int,
+    batch_size: int = 10,
+    period: str = "30天",
+    independent_prices: bool = False,
+) -> str:
+    category_labels = list(
+        dict.fromkeys(
+            item.get("category_name")
+            for item in products
+            if item.get("category_name")
+        )
+    )
+    category_paths = {
+        item.get("category_path")
+        for item in products
+        if item.get("category_path")
+    }
+    intro = build_feishu_intro_xml(
+        title=title,
+        min_price=min_price,
+        max_price=max_price,
+        total=len(products),
+        category_labels=category_labels,
+        category_count=len(category_paths) or len(category_labels),
+        period=period,
+        independent_prices=independent_prices,
+    )
+    doc_url = create_feishu_doc(title, intro)
+    for batch_index, offset in enumerate(range(0, len(products), batch_size), 1):
+        batch = products[offset : offset + batch_size]
+        xml = build_feishu_batch_xml(batch, start_rank=offset + 1)
+        append_feishu_doc(doc_url, xml, batch_index)
+        print(
+            f"[OK] 飞书文档批次 {batch_index}: "
+            f"{offset + 1}-{offset + len(batch)}"
+        )
+    repair_feishu_document(
+        doc_url,
+        products,
+        batch_size=batch_size,
+    )
+    return doc_url
+
+
+def create_feishu_doc_by_category(
+    title: str,
+    products: list[dict[str, Any]],
+    *,
+    min_price: int,
+    max_price: int,
+    period: str = "30天",
+    independent_prices: bool = False,
+) -> str:
+    """创建一品类一表的文档，并在返回链接前完成线上结构校验。"""
+    groups = group_products_by_category(products)
+    category_labels = [category for category, _ in groups]
+    category_paths = {
+        item.get("category_path")
+        for item in products
+        if item.get("category_path")
+    }
+    intro = build_feishu_intro_xml(
+        title=title,
+        min_price=min_price,
+        max_price=max_price,
+        total=len(products),
+        category_labels=category_labels,
+        category_count=len(category_paths) or len(category_labels),
+        period=period,
+        independent_prices=independent_prices,
+    )
+    doc_url = create_feishu_doc(title, intro)
+    append_native_category_tables(
+        doc_url,
+        [
+            build_feishu_category_xml(category, category_products)
+            for category, category_products in groups
+        ],
+        image_size=DOCUMENT_IMAGE_WIDTH,
+    )
+    verify_feishu_category_tables(doc_url, groups)
+    return doc_url
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description="采集生意参谋市场排行并生成飞书文档")
+    parser.add_argument("--min-price", type=int, default=400)
+    parser.add_argument("--max-price", type=int, default=2000)
+    parser.add_argument(
+        "--category",
+        default="",
+        help=(
+            "仅采集一个类目；多级路径用 > 分隔。"
+            "不传时默认依次采集业务指定的 9 个品类"
+        ),
+    )
+    parser.add_argument("--period", default="30天", help="市场排行统计周期")
+    parser.add_argument("--max-pages", type=int, default=0, help="0 表示抓完全部分页")
+    parser.add_argument("--limit", type=int, default=0, help="最多处理多少个商品，0 表示不限")
+    parser.add_argument("--cdp-url", default=None, help="接管已启动的 Chrome CDP")
+    parser.add_argument("--user-data-dir", default=os.getenv("TM_USER_DATA_DIR", sycm.USER_DATA_DIR))
+    parser.add_argument("--account", default=os.getenv("SYCM_ACCOUNT", sycm.ACCOUNT))
+    parser.add_argument("--password", default=os.getenv("SYCM_PASSWORD", sycm.PASSWORD))
+    parser.add_argument("--dry-run", action="store_true", help="采集并保存 JSON，不创建飞书文档")
+    parser.add_argument(
+        "--risk-cooldown",
+        type=int,
+        default=60,
+        help="命中操作频繁风控后，自动模式首次刷新前等待秒数；后续线性增加",
+    )
+    parser.add_argument(
+        "--risk-max-retries",
+        type=int,
+        default=6,
+        help="单次任务最多保留当前会话并自动刷新恢复的次数",
+    )
+    parser.add_argument(
+        "--risk-mode",
+        choices=("manual", "auto"),
+        default="manual",
+        help="manual=保留风控页等待人工验证；auto=保留 Cookie，冷却后刷新重试",
+    )
+    return parser.parse_args()
+
+
+def main() -> None:
+    for stream in (sys.stdout, sys.stderr):
+        if hasattr(stream, "reconfigure"):
+            stream.reconfigure(errors="backslashreplace")
+    args = parse_args()
+    if args.min_price >= args.max_price:
+        raise ValueError("最低价格必须小于最高价格")
+    sycm.ACCOUNT = args.account
+    sycm.PASSWORD = args.password
+    sycm.USER_DATA_DIR = args.user_data_dir
+
+    session = None
+    try:
+        session, page = sycm.connect_browser(args.cdp_url)
+        page = sycm.ensure_logged_in(session.context, page)
+        if page is None:
+            raise RuntimeError("生意参谋登录失败")
+        navigate_to_market_rank(page)
+        page = prepare_market_rank_period(
+            session.context,
+            page,
+            period=args.period,
+            risk_mode=args.risk_mode,
+            cooldown_seconds=args.risk_cooldown,
+            max_retries=args.risk_max_retries,
+        )
+        if args.category:
+            single_path = tuple(
+                part.strip()
+                for part in args.category.split(">")
+                if part.strip()
+            )
+            if not single_path:
+                raise ValueError("单品类路径不能为空")
+            category_specs = (
+                {
+                    "name": single_path[-1],
+                    "min_price": args.min_price,
+                    "max_price": args.max_price,
+                    "paths": (single_path,),
+                },
+            )
+        else:
+            category_specs = TARGET_MARKET_CATEGORIES
+        products: list[dict[str, Any]] = []
+        for category_index, category_spec in enumerate(category_specs, 1):
+            category_name = str(category_spec["name"])
+            min_price = int(category_spec["min_price"])
+            max_price = int(category_spec["max_price"])
+            print(
+                f"[->] 品类 {category_index}/{len(category_specs)}: "
+                f"{category_name}，价格 {min_price}-{max_price}"
+            )
+            category_products: list[dict[str, Any]] = []
+            category_seen: set[tuple[str, str]] = set()
+            for category_path_value in category_spec["paths"]:
+                category_path_tuple = tuple(category_path_value)
+                category_path = " > ".join(category_path_tuple)
+                page = prepare_market_rank_filters(
+                    page,
+                    period=args.period,
+                    category_path=category_path_tuple,
+                    min_price=min_price,
+                    max_price=max_price,
+                    risk_mode=args.risk_mode,
+                    cooldown_seconds=args.risk_cooldown,
+                    max_retries=args.risk_max_retries,
+                )
+                page_ref = {"page": page}
+
+                def recover_from_risk(
+                    rank_page,
+                    target_page,
+                    recovery_attempt,
+                    reason,
+                ):
+                    print(f"[RISK] 恢复原因: {reason}")
+                    restored_page = recover_market_rank_after_risk(
+                        session.context,
+                        rank_page,
+                        period=args.period,
+                        category_path=category_path_tuple,
+                        min_price=min_price,
+                        max_price=max_price,
+                        target_page=target_page,
+                        cooldown_seconds=args.risk_cooldown,
+                        recovery_attempt=recovery_attempt,
+                        risk_mode=args.risk_mode,
+                    )
+                    page_ref["page"] = restored_page
+                    return restored_page
+
+                source_products = collect_all_rank_pages(
+                    page_ref["page"],
+                    args.max_pages,
+                    args.limit,
+                    category_name=category_name,
+                    category_path=category_path,
+                    existing_products=[*products, *category_products],
+                    recover_from_risk=recover_from_risk,
+                    risk_max_retries=args.risk_max_retries,
+                    manual_risk_wait=args.risk_mode == "manual",
+                )
+                page = page_ref["page"]
+                for item in source_products:
+                    item["filter_min_price"] = min_price
+                    item["filter_max_price"] = max_price
+                    item_id = _product_item_id(item.get("product_url") or "")
+                    key = (
+                        ("item_id", item_id)
+                        if item_id
+                        else ("name", item.get("name") or "")
+                    )
+                    if key not in category_seen:
+                        category_seen.add(key)
+                        category_products.append(item)
+            products.extend(category_products)
+            _save_checkpoint(products)
+            print(
+                f"[OK] 品类 {category_name}: {len(category_products)} 条，"
+                f"全部品类累计 {len(products)} 条"
+            )
+
+        OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+        stamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+        category_marker = "9_categories_30d" if not args.category else "single_category"
+        json_path = OUTPUT_DIR / f"market_rank_{category_marker}_{stamp}.json"
+        json_path.write_text(
+            json.dumps(products, ensure_ascii=False, indent=2),
+            encoding="utf-8",
+        )
+        print(f"[OK] 采集结果已保存: {json_path}")
+        if args.dry_run:
+            return
+        title_scope = "九品类" if not args.category else category_specs[0]["name"]
+        title = (
+            f"天猫{title_scope}市场排行 30天 {datetime.now():%Y-%m-%d}"
+        )
+        url = create_feishu_doc_by_category(
+            title,
+            products,
+            min_price=args.min_price,
+            max_price=args.max_price,
+            period=args.period,
+            independent_prices=not args.category,
+        )
+        print(f"[OK] 飞书文档: {url}")
+        archive_market_rank_report(
+            platform="tm",
+            report_date=datetime.now().date(),
+            period=args.period,
+            category_scope=str(title_scope),
+            category_names=[str(spec["name"]) for spec in category_specs],
+            report_title=title,
+            feishu_doc_url=url,
+            product_count=len(products),
+            source_json_path=str(json_path),
+            run_metadata={
+                "category": args.category or None,
+                "max_pages": args.max_pages,
+                "limit": args.limit,
+                "category_prices": {
+                    str(spec["name"]): {
+                        "min_price": int(spec["min_price"]),
+                        "max_price": int(spec["max_price"]),
+                    }
+                    for spec in category_specs
+                },
+            },
+        )
+    finally:
+        if session is not None:
+            close_market_rank_browser(
+                session,
+                user_data_dir=args.user_data_dir,
+                cdp_url=args.cdp_url,
+            )
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/commands/__init__.py b/src/gyxx_flow/modules/product_commerce/runtime/commands/__init__.py
new file mode 100644
index 0000000..f43bf30
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/commands/__init__.py
@@ -0,0 +1 @@
+"""Stable launchers added while adopting product source into GYXX Flow."""
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/commands/import_daily.py b/src/gyxx_flow/modules/product_commerce/runtime/commands/import_daily.py
new file mode 100644
index 0000000..283a212
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/commands/import_daily.py
@@ -0,0 +1,22 @@
+"""Import the day before the scheduler's business date."""
+
+from __future__ import annotations
+
+import os
+import sys
+from datetime import date, timedelta
+
+from import_product_daily import main as import_main
+
+
+def main() -> int:
+    raw_date = os.environ.get("GYXX_BUSINESS_DATE", "").strip()
+    if not raw_date:
+        raise RuntimeError("GYXX_BUSINESS_DATE is required")
+    target = date.fromisoformat(raw_date) - timedelta(days=1)
+    sys.argv = [sys.argv[0], "--date", target.isoformat()]
+    return import_main()
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/commands/run_alerts.py b/src/gyxx_flow/modules/product_commerce/runtime/commands/run_alerts.py
new file mode 100644
index 0000000..bfece62
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/commands/run_alerts.py
@@ -0,0 +1,23 @@
+"""Launch the migrated sales alert with explicit runtime configuration."""
+
+from __future__ import annotations
+
+import os
+import sys
+
+from run_alerts_with_retry import main as alert_main
+
+
+def main() -> int:
+    business_date = os.environ.get("GYXX_BUSINESS_DATE", "").strip()
+    recipient = os.environ.get("GYXX_ALERT_RECIPIENT_OPEN_ID", "").strip()
+    if not business_date:
+        raise RuntimeError("GYXX_BUSINESS_DATE is required")
+    if not recipient:
+        raise RuntimeError("GYXX_ALERT_RECIPIENT_OPEN_ID is required")
+    sys.argv = [sys.argv[0], "--end-date", business_date, "--openid", recipient]
+    return alert_main()
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/config/__init__.py b/src/gyxx_flow/modules/product_commerce/runtime/config/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/config/auto-flow-config.example.json b/src/gyxx_flow/modules/product_commerce/runtime/config/auto-flow-config.example.json
new file mode 100644
index 0000000..45bf46e
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/config/auto-flow-config.example.json
@@ -0,0 +1,17 @@
+{
+  "login": {
+    "url": "https://www.erp321.com/login.aspx",
+    "username": "",
+    "password": ""
+  },
+  "cdp_url": "http://127.0.0.1:18802",
+  "user_data_dir": "${GYXX_DATA_ROOT}\\profiles\\product_commerce\\erp",
+  "feishu": {
+    "app_id": "",
+    "app_secret": ""
+  },
+  "feishu_analyzer": {
+    "app_id": "",
+    "app_secret": ""
+  }
+}
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/config/style_config_loader.py b/src/gyxx_flow/modules/product_commerce/runtime/config/style_config_loader.py
new file mode 100644
index 0000000..461c5a1
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/config/style_config_loader.py
@@ -0,0 +1,994 @@
+"""从飞书多维表格动态读取款式配置。
+
+配置源表格：
+  https://bu0zgpibak.feishu.cn/base/TtoCb1NuQaDy3NsZWTpc0GIvnph?table=tblKCjplVAFrRwMC&view=vewvy33xEk
+
+提供：
+  - 天猫/京东/抖音商品ID
+  - 京东自营 SKU
+  - ERP 款式编码
+  - 销量飞书 Base/Table/View
+  - 主图飞书 Base/Table/View
+
+支持本地缓存与失败回退。
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import re
+import shutil
+import subprocess
+from datetime import datetime, timedelta
+from pathlib import Path
+from typing import Any
+from urllib.parse import parse_qs, urlparse
+
+from lark_cli_runtime import lark_cli_env
+from runtime_paths import NORMALIZED_DATA_ROOT, STATE_ROOT, managed_data_path
+
+
+DEFAULT_BASE_TOKEN = "TtoCb1NuQaDy3NsZWTpc0GIvnph"
+DEFAULT_TABLE_ID = "tblKCjplVAFrRwMC"
+DEFAULT_VIEW_ID = "vewvy33xEk"
+CACHE_SCHEMA_VERSION = 4
+
+# 表格字段名（按实际表格设置，支持别名）
+FIELD_ALIASES = {
+    "platform": ["平台"],
+    "item_ids": ["商品链接ID", "商品ID"],
+    "style": ["款式"],
+    "brand": ["品牌"],
+    "erp_codes": ["ERP款式编码"],
+    "sales_bitable_url": ["飞书多维表格地址"],
+    "main_image_bitable_url": ["主图飞书多维表格地址"],
+    "persona_bitable_url": ["人群画像（天猫）"],
+    "dy_persona_bitable_url": ["人群画像（抖音）"],
+    "jd_persona_bitable_url": ["人群画像（京东）"],
+    "style_analysis_bitable_url": ["平台单品分析"],
+}
+
+PLATFORM_MAP = {
+    "天猫": "tm",
+    "京东旗舰店": "jd",
+    "京东自营": "jd_self",
+    "抖音": "dy",
+}
+
+
+class StyleConfigError(Exception):
+    pass
+
+
+class StyleConfigLoader:
+    """从飞书多维表格读取款式配置的加载器。
+
+    用法：
+        loader = StyleConfigLoader()
+        cfg = loader.load()
+        print(cfg["styles"]["盖亚斜挎"])
+    """
+
+    def __init__(
+        self,
+        base_token: str | None = None,
+        table_id: str | None = None,
+        view_id: str | None = None,
+        cache_path: str | Path | None = None,
+        identity: str = "user",
+        fallback_style_map_path: str | Path | None = None,
+        fallback_main_image_map_path: str | Path | None = None,
+        fallback_styles_input_path: str | Path | None = None,
+    ):
+        self.base_token = base_token or DEFAULT_BASE_TOKEN
+        self.table_id = table_id or DEFAULT_TABLE_ID
+        self.view_id = view_id or DEFAULT_VIEW_ID
+        self.identity = identity
+
+        project_root = Path(__file__).resolve().parent.parent
+        self.cache_path = (
+            Path(cache_path)
+            if cache_path
+            else STATE_ROOT / ".style_config_cache.json"
+        )
+        self.static_fallback_style_map_path = project_root / "bitable_style_map.json"
+        self.fallback_style_map_path = (
+            managed_data_path(fallback_style_map_path)
+            if fallback_style_map_path
+            else NORMALIZED_DATA_ROOT / "config" / "bitable_style_map.json"
+        )
+        self.fallback_main_image_map_path = (
+            Path(fallback_main_image_map_path)
+            if fallback_main_image_map_path
+            else project_root / "bitable_main_image_map.json"
+        )
+        self.fallback_styles_input_path = (
+            Path(fallback_styles_input_path)
+            if fallback_styles_input_path
+            else project_root / "styles_input.json"
+        )
+
+        self._lark_cli = (
+            os.getenv("LARK_CLI_PATH")
+            or shutil.which("lark-cli")
+            or shutil.which("lark-cli.cmd")
+            or "lark-cli"
+        )
+
+    # ------------------------------------------------------------------
+    # Lark CLI 调用
+    # ------------------------------------------------------------------
+    def _run_lark(self, args: list[str]) -> dict:
+        command = [self._lark_cli, "base"] + args + ["--as", self.identity, "--format", "json"]
+        proc = subprocess.run(
+            command,
+            cwd=Path(__file__).resolve().parent.parent,
+            env=lark_cli_env(),
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+            capture_output=True,
+        )
+        if proc.returncode:
+            raise StyleConfigError(f"lark-cli failed: {proc.stderr[:500]}")
+        return self._parse_json(proc.stdout)
+
+    @staticmethod
+    def _parse_json(text: str) -> dict:
+        """lark-cli 输出前面可能带提示文本，从第一个 '{' 开始解析。"""
+        decoder = json.JSONDecoder()
+        pos = 0
+        while pos < len(text):
+            i = text.find("{", pos)
+            if i < 0:
+                break
+            try:
+                obj, end = decoder.raw_decode(text, i)
+                if isinstance(obj, dict):
+                    return obj
+                pos = end
+            except json.JSONDecodeError:
+                pos = i + 1
+        raise StyleConfigError("无法解析 lark-cli JSON 输出")
+
+    # ------------------------------------------------------------------
+    # 字段/URL 解析
+    # ------------------------------------------------------------------
+    @staticmethod
+    def _parse_markdown_link(value: Any) -> str:
+        text = str(value or "").strip()
+        if not text:
+            return ""
+        match = re.search(r"\]\((https?://[^\)]+)\)", text)
+        if match:
+            return match.group(1)
+        if text.startswith("http://") or text.startswith("https://"):
+            return text
+        return ""
+
+    @staticmethod
+    def _parse_bitable_url(url: str) -> dict:
+        """从飞书 URL 提取 base_token / table_id / view_id。"""
+        result = {"base_token": None, "table_id": None, "view_id": None}
+        if not url:
+            return result
+        parsed = urlparse(url)
+        path_parts = [p for p in parsed.path.split("/") if p]
+        # /base/{base_token}
+        if "base" in path_parts:
+            idx = path_parts.index("base")
+            if idx + 1 < len(path_parts):
+                result["base_token"] = path_parts[idx + 1]
+        qs = parse_qs(parsed.query)
+        if "table" in qs:
+            result["table_id"] = qs["table"][0]
+        if "view" in qs:
+            result["view_id"] = qs["view"][0]
+        return result
+
+    @staticmethod
+    def _clean_id_list(value: Any) -> list[str]:
+        """清洗逗号分隔 ID：兼容中英文逗号、换行、空格。"""
+        if value is None:
+            return []
+        text = str(value)
+        # 统一替换成英文逗号
+        text = text.replace("，", ",").replace("\n", ",").replace("\r", ",")
+        ids = [x.strip() for x in text.split(",") if x.strip()]
+        seen: set[str] = set()
+        result: list[str] = []
+        for item in ids:
+            if item not in seen:
+                seen.add(item)
+                result.append(item)
+        return result
+
+    # ------------------------------------------------------------------
+    # 记录级过滤条件
+    # ------------------------------------------------------------------
+    def _has_sales_bitable(self, record: dict) -> bool:
+        """销量飞书 Base/Table 同时存在。"""
+        url = self._parse_markdown_link(record.get("sales_bitable_url"))
+        parsed = self._parse_bitable_url(url)
+        return bool(parsed["base_token"] and parsed["table_id"])
+
+    def _has_main_image_bitable(self, record: dict) -> bool:
+        """主图飞书 Base/Table 同时存在。"""
+        url = self._parse_markdown_link(record.get("main_image_bitable_url"))
+        parsed = self._parse_bitable_url(url)
+        return bool(parsed["base_token"] and parsed["table_id"])
+
+    def _has_persona_bitable(self, record: dict) -> bool:
+        """人群画像（天猫）飞书 Base/Table 同时存在。"""
+        url = self._parse_markdown_link(record.get("persona_bitable_url"))
+        parsed = self._parse_bitable_url(url)
+        return bool(parsed["base_token"] and parsed["table_id"])
+
+    def _has_dy_persona_bitable(self, record: dict) -> bool:
+        """人群画像（抖音）飞书 Base/Table 同时存在。"""
+        url = self._parse_markdown_link(record.get("dy_persona_bitable_url"))
+        parsed = self._parse_bitable_url(url)
+        return bool(parsed["base_token"] and parsed["table_id"])
+
+    def _has_jd_persona_bitable(self, record: dict) -> bool:
+        """人群画像（京东）飞书 Base/Table 同时存在。"""
+        url = self._parse_markdown_link(record.get("jd_persona_bitable_url"))
+        parsed = self._parse_bitable_url(url)
+        return bool(parsed["base_token"] and parsed["table_id"])
+
+    def _has_style_analysis_bitable(self, record: dict) -> bool:
+        """平台单品分析 Base/Table 同时存在。"""
+        url = self._parse_markdown_link(record.get("style_analysis_bitable_url"))
+        parsed = self._parse_bitable_url(url)
+        return bool(parsed["base_token"] and parsed["table_id"])
+
+    def _has_daily_required_fields(self, record: dict) -> bool:
+        """每日采集四条件：erp_codes、style、item_ids、sales_bitable。"""
+        style_name = str(record.get("style") or "").strip()
+        if not style_name:
+            return False
+        if not self._clean_id_list(record.get("erp_codes")):
+            return False
+        if not self._clean_id_list(record.get("item_ids")):
+            return False
+        if not self._has_sales_bitable(record):
+            return False
+        return True
+
+    # ------------------------------------------------------------------
+    # 表格记录读取
+    # ------------------------------------------------------------------
+    def _load_records_from_lark(self) -> list[dict]:
+        """调用飞书 API 读取全部记录，返回字段名 -> 值的字典列表。
+
+        lark-cli --limit 上限 200, 超过要分页 (主表当前 ~36 条, 单页就够)。
+        """
+        all_rows: list = []
+        all_fields: list = []
+        offset = 0
+        page_size = 200
+        while True:
+            data = self._run_lark(
+                [
+                    "+record-list",
+                    "--base-token",
+                    self.base_token,
+                    "--table-id",
+                    self.table_id,
+                    "--limit",
+                    str(page_size),
+                    "--offset",
+                    str(offset),
+                ]
+            )
+            payload = data.get("data", {})
+            fields = payload.get("fields", [])
+            rows = payload.get("data", [])
+            if not rows:
+                break
+            if not all_fields and fields:
+                all_fields = fields
+            all_rows.extend(rows)
+            if len(rows) < page_size:
+                break
+            offset += page_size
+
+        # 拼回跟原调用同形的 payload (后面的解析逻辑不感知分页)
+        payload = {"fields": all_fields, "data": all_rows}
+        field_names = payload.get("fields", [])
+        if field_names and isinstance(field_names[0], dict):
+            field_names = [f["name"] for f in field_names]
+        rows = payload.get("data", [])
+
+        # 建立字段名 -> 内部 key 的映射
+        key_by_name: dict[str, str] = {}
+        for key, aliases in FIELD_ALIASES.items():
+            for alias in aliases:
+                norm_alias = alias.replace(" ", "").replace("\u3000", "")
+                for name in field_names:
+                    norm_name = name.replace(" ", "").replace("\u3000", "")
+                    if norm_name == norm_alias or norm_name.startswith(norm_alias):
+                        key_by_name[name] = key
+                        break
+
+        records = []
+        for row in rows:
+            record: dict[str, Any] = {}
+            for name, value in zip(field_names, row):
+                key = key_by_name.get(name)
+                if not key:
+                    continue
+                if key == "platform":
+                    record[key] = value[0] if isinstance(value, list) and value else str(value)
+                else:
+                    record[key] = value
+            records.append(record)
+        return records
+
+    # ------------------------------------------------------------------
+    # 聚合配置
+    # ------------------------------------------------------------------
+    @staticmethod
+    def _normalize_style_name(raw: Any) -> str:
+        """款式名归一化:strip + 全角空格 → 半角空格 + 合并连续空白。
+
+        飞书表格多分组填款式时容易多敲空格/全角空格,归一化后才能 merge 到同一款式。
+        """
+        return " ".join(str(raw or "").split())
+
+    def _aggregate(self, records: list[dict]) -> dict:
+        """两遍聚合 (v2.2):
+
+        第一遍 — 款式级别基础字段 (brand / erp_codes / sales_bitable / main_image_bitable):
+          ERP 编码只读取天猫分组；其它基础字段从任意非空记录收集。
+          新款式约定: TM 分组填 5 字段 (款式/erp编码/飞书多维表格地址/主图飞书多维表格地址/商品链接id);
+          JD/DY 分组只填 款式+商品链接id 也能采集 (erp + sales_bitable 从 TM 继承)。
+
+        第二遍 — 平台级别 item_ids: 按分组读取。
+          该平台分组有商品链接id 就写入,不再要求 4 字段门。
+          未上架款 (7 个款式: 极星2/拾影斜挎/拾影双肩/云栖/极星双肩2/波塞冬手提电脑包/黑曜石电脑包)
+          任何平台都没商品链接id → 全部跳过。
+
+        jd_self (京东自营): 仅要求 SKU 非空 (周汇总用)。
+        """
+        styles: dict[str, dict] = {}
+        erp_code_records: dict[str, list[tuple[str, list[str]]]] = {}
+
+        def _empty(style_name: str) -> dict:
+            return {
+                "name": style_name,
+                "brand": "",
+                "erp_codes": [],
+                "tm": {"item_ids": []},
+                "jd": {"spus": []},
+                "jd_self": {"skus": []},
+                "dy": {"item_ids": []},
+                "sales_bitable": {},
+                "main_image_bitable": {},
+                "persona_bitable": {},
+                "dy_persona_bitable": {},
+                "jd_persona_bitable": {},
+                "style_analysis_bitable": {},
+            }
+
+        # ── 第一遍:款式级字段 ──────────────────────────────────────
+        for record in records:
+            platform_label = record.get("platform", "")
+            platform = PLATFORM_MAP.get(platform_label)
+            style_name = self._normalize_style_name(record.get("style"))
+            if not style_name or not platform:
+                continue
+
+            style = styles.setdefault(style_name, _empty(style_name))
+
+            # 品牌（款式级别,取第一个非空）
+            brand = str(record.get("brand") or "").strip()
+            if brand and not style["brand"]:
+                style["brand"] = brand
+
+            # ERP 编码以天猫基础分组为唯一权威来源。京东/抖音行即使误填，
+            # 也只能用于冲突诊断，不能覆盖 ERP 三平台采集所用编码。
+            erp_codes = self._clean_id_list(record.get("erp_codes"))
+            if erp_codes:
+                erp_code_records.setdefault(style_name, []).append((platform_label, erp_codes))
+                if platform == "tm":
+                    if not style["erp_codes"]:
+                        style["erp_codes"] = erp_codes
+                    elif style["erp_codes"] != erp_codes:
+                        print(
+                            f"[WARN] {style_name}: 天猫分组存在 ERP 编码冲突 "
+                            f"{style['erp_codes']} vs {erp_codes}; 保留第一条天猫编码"
+                        )
+
+            # 销量飞书表地址（款式级别,只要 URL 完整就收）
+            if self._has_sales_bitable(record) and not style["sales_bitable"]:
+                sales_url = self._parse_markdown_link(record.get("sales_bitable_url"))
+                if sales_url:
+                    style["sales_bitable"] = self._parse_bitable_url(sales_url)
+
+            # 主图表地址：只读天猫分组记录,只要求主图 URL 非空
+            if platform == "tm" and self._has_main_image_bitable(record):
+                main_url = self._parse_markdown_link(record.get("main_image_bitable_url"))
+                if main_url and not style["main_image_bitable"]:
+                    style["main_image_bitable"] = self._parse_bitable_url(main_url)
+                    style["main_image_bitable"]["platform"] = "天猫"
+
+            # 人群画像（天猫）表地址：只读天猫分组记录
+            if platform == "tm" and self._has_persona_bitable(record):
+                persona_url = self._parse_markdown_link(record.get("persona_bitable_url"))
+                if persona_url and not style["persona_bitable"]:
+                    style["persona_bitable"] = self._parse_bitable_url(persona_url)
+
+            # 人群画像（抖音）表地址：款式级别，任意平台记录有值就收
+            if self._has_dy_persona_bitable(record):
+                dy_persona_url = self._parse_markdown_link(record.get("dy_persona_bitable_url"))
+                if dy_persona_url and not style["dy_persona_bitable"]:
+                    style["dy_persona_bitable"] = self._parse_bitable_url(dy_persona_url)
+
+            # 人群画像（京东）表地址：款式级别，任意平台记录有值就收
+            if self._has_jd_persona_bitable(record):
+                jd_persona_url = self._parse_markdown_link(record.get("jd_persona_bitable_url"))
+                if jd_persona_url and not style["jd_persona_bitable"]:
+                    style["jd_persona_bitable"] = self._parse_bitable_url(jd_persona_url)
+
+            if self._has_style_analysis_bitable(record) and not style["style_analysis_bitable"]:
+                analysis_url = self._parse_markdown_link(record.get("style_analysis_bitable_url"))
+                if analysis_url:
+                    style["style_analysis_bitable"] = self._parse_bitable_url(analysis_url)
+
+        # 非天猫分组的 ERP 编码不会参与聚合；只在冲突或天猫缺失时提示，
+        # 避免重复填写相同编码造成无意义日志。
+        for style_name, entries in erp_code_records.items():
+            tm_codes = styles[style_name]["erp_codes"]
+            for platform_label, codes in entries:
+                if PLATFORM_MAP.get(platform_label) == "tm":
+                    continue
+                if not tm_codes:
+                    print(
+                        f"[WARN] {style_name}: {platform_label} 分组填写 ERP 编码 {codes} 已忽略; "
+                        "天猫分组未填写 ERP 编码"
+                    )
+                elif codes != tm_codes:
+                    print(
+                        f"[WARN] {style_name}: ERP 编码冲突，天猫={tm_codes}，"
+                        f"{platform_label}={codes}; 已按天猫分组并忽略非天猫编码"
+                    )
+
+        # ── 第二遍:平台级 item_ids（按分组读取）─────────────────────────
+        for record in records:
+            platform_label = record.get("platform", "")
+            platform = PLATFORM_MAP.get(platform_label)
+            style_name = self._normalize_style_name(record.get("style"))
+            if not style_name or not platform:
+                continue
+            style = styles.get(style_name)
+            if style is None:
+                continue
+
+            ids = self._clean_id_list(record.get("item_ids"))
+            if platform == "tm" and ids:
+                style["tm"]["item_ids"] = ids
+            elif platform == "jd" and ids:
+                style["jd"]["spus"] = ids
+            elif platform == "dy" and ids:
+                style["dy"]["item_ids"] = ids
+            elif platform == "jd_self" and ids:
+                style["jd_self"]["skus"] = ids
+
+        return {
+            "styles": styles,
+            "source": "lark",
+            "loaded_at": datetime.now().isoformat(),
+            "schema_version": CACHE_SCHEMA_VERSION,
+        }
+
+    # ------------------------------------------------------------------
+    # 缓存与回退
+    # ------------------------------------------------------------------
+    def _read_cache(self) -> dict | None:
+        if not self.cache_path.exists():
+            return None
+        try:
+            payload = json.loads(self.cache_path.read_text(encoding="utf-8-sig"))
+            if not isinstance(payload, dict) or "styles" not in payload:
+                return None
+            if payload.get("schema_version") != CACHE_SCHEMA_VERSION:
+                return None
+            return payload
+        except Exception:
+            return None
+
+    def _write_cache(self, payload: dict) -> None:
+        self.cache_path.parent.mkdir(parents=True, exist_ok=True)
+        self.cache_path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    def _load_from_fallback(self) -> dict:
+        """从现有 JSON 文件构建兼容配置。"""
+        styles: dict[str, dict] = {}
+
+        # 销量表映射（含 base/table/view）
+        style_map_path = (
+            self.fallback_style_map_path
+            if self.fallback_style_map_path.exists()
+            else self.static_fallback_style_map_path
+        )
+        if style_map_path.exists():
+            style_map = json.loads(style_map_path.read_text(encoding="utf-8-sig"))
+            for style_name, cfg in style_map.items():
+                if not cfg.get("enabled", True):
+                    continue
+                style = styles.setdefault(
+                    style_name,
+                    self._empty_style(style_name),
+                )
+                style["sales_bitable"] = {
+                    "base_token": cfg.get("base_token"),
+                    "table_id": cfg.get("table_id"),
+                    "view_id": cfg.get("view_id"),
+                }
+
+        # 主图表映射
+        if self.fallback_main_image_map_path.exists():
+            main_map = json.loads(self.fallback_main_image_map_path.read_text(encoding="utf-8-sig"))
+            main_map.pop("_comment", None)
+            for style_name, cfg in main_map.items():
+                if not cfg.get("enabled", True):
+                    continue
+                style = styles.setdefault(
+                    style_name,
+                    self._empty_style(style_name),
+                )
+                style["main_image_bitable"] = {
+                    "base_token": cfg.get("base_token"),
+                    "table_id": cfg.get("table_id"),
+                    "view_id": cfg.get("view_id"),
+                    "platform": cfg.get("platform", "天猫"),
+                }
+
+        # ERP 编码
+        if self.fallback_styles_input_path.exists():
+            input_items = json.loads(self.fallback_styles_input_path.read_text(encoding="utf-8-sig"))
+            for item in input_items:
+                style_name = item.get("style_name") or item.get("款式") or item.get("style")
+                if not style_name:
+                    continue
+                style = styles.setdefault(
+                    style_name,
+                    self._empty_style(style_name),
+                )
+                codes = item.get("erp_style_codes") or item.get("ERP款式编码") or []
+                if isinstance(codes, str):
+                    codes = [x.strip() for x in codes.split(",") if x.strip()]
+                style["erp_codes"] = [str(c) for c in codes if str(c).strip()]
+
+        return {
+            "styles": styles,
+            "source": "fallback_json",
+            "loaded_at": datetime.now().isoformat(),
+            "schema_version": CACHE_SCHEMA_VERSION,
+        }
+
+    @staticmethod
+    def _empty_style(style_name: str) -> dict:
+        return {
+            "name": style_name,
+            "erp_codes": [],
+            "tm": {"item_ids": []},
+            "jd": {"spus": []},
+            "jd_self": {"skus": []},
+            "dy": {"item_ids": []},
+            "sales_bitable": {},
+            "main_image_bitable": {},
+            "style_analysis_bitable": {},
+        }
+
+    # ------------------------------------------------------------------
+    # 公共入口
+    # ------------------------------------------------------------------
+    def load(self, force_refresh: bool = False, sync_fallback: bool = False) -> dict:
+        """加载配置。优先飞书，失败则缓存，再失败则本地 JSON。
+
+        force_refresh: 跳过 24h 缓存, 强制从飞书重读。
+        sync_fallback: 把飞书最新结果写回 bitable_style_map.json (--refresh-config 用),
+                       保持 fallback JSON 跟主表同步。
+        """
+        if not force_refresh:
+            cached = self._read_cache()
+            if cached and cached.get("source") == "lark":
+                # 检查缓存是否在 24h 内
+                try:
+                    loaded = datetime.fromisoformat(cached["loaded_at"])
+                    if datetime.now() - loaded < timedelta(hours=24):
+                        return cached
+                except Exception:
+                    pass
+
+        try:
+            records = self._load_records_from_lark()
+            payload = self._aggregate(records)
+            self._write_cache(payload)
+            if sync_fallback:
+                self._write_style_map_fallback(payload)
+            return payload
+        except Exception as exc:
+            print(f"[WARN] 飞书配置读取失败: {exc}，尝试缓存/本地回退")
+            cached = self._read_cache()
+            if cached:
+                return cached
+            return self._load_from_fallback()
+
+    # ------------------------------------------------------------------
+    # Fallback 同步
+    # ------------------------------------------------------------------
+    DEFAULT_FIELD_MAP = {
+        "时间": "时间周期",
+        "平台": "平台",
+        "销量": "销量",
+        "访客数": "访客数",
+        "转化率": "转化率",
+        "加购率": "加购率",
+        "退款率": "退款率",
+        "产品": "款式",
+        "SourceID": "SourceID",
+    }
+
+    def _write_style_map_fallback(self, payload: dict) -> None:
+        """把飞书最新款式配置写回 bitable_style_map.json。
+
+        保留现有 field_map (避免覆盖人工微调); 缺款式用 DEFAULT_FIELD_MAP。
+        """
+        if not self.fallback_style_map_path:
+            return
+
+        existing: dict = {}
+        if self.fallback_style_map_path.exists():
+            try:
+                existing = json.loads(
+                    self.fallback_style_map_path.read_text(encoding="utf-8-sig")
+                )
+            except Exception:
+                existing = {}
+
+        new_map: dict = {}
+        for style_name, cfg in payload.get("styles", {}).items():
+            sales = cfg.get("sales_bitable") or {}
+            if not sales.get("base_token") or not sales.get("table_id"):
+                continue
+            old_cfg = existing.get(style_name, {})
+            new_map[style_name] = {
+                "base_token": sales["base_token"],
+                "table_id": sales["table_id"],
+                "view_id": sales.get("view_id") or old_cfg.get("view_id", "vewGbFVRhD"),
+                "enabled": old_cfg.get("enabled", True),
+                "field_map": old_cfg.get("field_map", self.DEFAULT_FIELD_MAP),
+            }
+
+        self.fallback_style_map_path.parent.mkdir(parents=True, exist_ok=True)
+        self.fallback_style_map_path.write_text(
+            json.dumps(new_map, ensure_ascii=False, indent=2),
+            encoding="utf-8-sig",
+        )
+        print(f"[INFO] {self.fallback_style_map_path.name} 已同步: {len(new_map)} 款")
+
+    def get_style_config(self, style_name: str) -> dict | None:
+        payload = self.load()
+        return payload["styles"].get(style_name)
+
+    # ------------------------------------------------------------------
+    # 兼容现有脚本的数据格式
+    # ------------------------------------------------------------------
+    def get_style_names(self) -> list[str]:
+        return sorted(self.load()["styles"].keys())
+
+    def get_tm_style_id_map(self) -> dict[str, list[str]]:
+        """每日天猫采集列表：_aggregate 已保证满足四条件（erp/款式/item_ids/销量表）。"""
+        return {
+            name: cfg["tm"]["item_ids"]
+            for name, cfg in self.load()["styles"].items()
+            if cfg["tm"]["item_ids"]
+        }
+
+    def get_tm_persona_bitable_map(self) -> dict[str, dict[str, str]]:
+        """款式 -> 人群画像（天猫）目标表 {base_token, table_id, view_id}。
+        只返回同时有 tm item_ids 和 persona_bitable 的款式（三者同时存在才采）。"""
+        result: dict[str, dict[str, str]] = {}
+        for name, cfg in self.load()["styles"].items():
+            if not cfg["tm"]["item_ids"]:
+                continue
+            persona = cfg.get("persona_bitable") or {}
+            if persona.get("base_token") and persona.get("table_id"):
+                result[name] = {
+                    "base_token": persona["base_token"],
+                    "table_id": persona["table_id"],
+                    "view_id": persona.get("view_id", ""),
+                }
+        return result
+
+    def get_tm_brand_map(self) -> dict[str, str]:
+        """款式 -> 品牌映射（用于天猫按品牌切换账号）。"""
+        return {
+            name: cfg.get("brand", "")
+            for name, cfg in self.load()["styles"].items()
+            if cfg["tm"]["item_ids"]
+        }
+
+    def get_jd_spu_groups(self) -> list[dict]:
+        """每日京东旗舰店采集列表：_aggregate 已保证满足四条件。"""
+        return [
+            {"style": name, "spus": cfg["jd"]["spus"]}
+            for name, cfg in self.load()["styles"].items()
+            if cfg["jd"]["spus"]
+        ]
+
+    def get_dy_product_groups(self) -> dict[str, list[str]]:
+        """每日抖音采集列表：_aggregate 已保证满足四条件。"""
+        return {
+            name: cfg["dy"]["item_ids"]
+            for name, cfg in self.load()["styles"].items()
+            if cfg["dy"]["item_ids"]
+        }
+
+    def get_dy_persona_bitable_map(self) -> dict[str, dict[str, str]]:
+        """款式 -> 人群画像（抖音）目标表 {base_token, table_id, view_id}。
+        只返回同时有 dy item_ids 和 dy_persona_bitable 的款式。"""
+        result: dict[str, dict[str, str]] = {}
+        for name, cfg in self.load()["styles"].items():
+            if not cfg["dy"]["item_ids"]:
+                continue
+            persona = cfg.get("dy_persona_bitable") or {}
+            if persona.get("base_token") and persona.get("table_id"):
+                result[name] = {
+                    "base_token": persona["base_token"],
+                    "table_id": persona["table_id"],
+                    "view_id": persona.get("view_id", ""),
+                }
+        return result
+
+    def get_jd_persona_bitable_map(self) -> dict[str, dict[str, str]]:
+        """款式 -> 人群画像（京东）目标表 {base_token, table_id, view_id}。
+        只返回同时有 jd spus 和 jd_persona_bitable 的款式。"""
+        result: dict[str, dict[str, str]] = {}
+        for name, cfg in self.load()["styles"].items():
+            if not cfg["jd"]["spus"]:
+                continue
+            persona = cfg.get("jd_persona_bitable") or {}
+            if persona.get("base_token") and persona.get("table_id"):
+                result[name] = {
+                    "base_token": persona["base_token"],
+                    "table_id": persona["table_id"],
+                    "view_id": persona.get("view_id", ""),
+                }
+        return result
+
+    def get_style_analysis_bitable_map(self) -> dict[str, dict[str, str]]:
+        """款式 -> 平台单品分析目标表。"""
+        result: dict[str, dict[str, str]] = {}
+        for name, cfg in self.load()["styles"].items():
+            target = cfg.get("style_analysis_bitable") or {}
+            if target.get("base_token") and target.get("table_id"):
+                result[name] = {
+                    "base_token": target["base_token"],
+                    "table_id": target["table_id"],
+                    "view_id": target.get("view_id", ""),
+                }
+        return result
+
+    def get_jd_self_sku_groups(self) -> list[dict]:
+        """周汇总京东自营采集列表：仅要求京东自营分组商品ID/SKU不为空。"""
+        return [
+            {"style_name": name, "skus": cfg["jd_self"]["skus"]}
+            for name, cfg in self.load()["styles"].items()
+            if cfg["jd_self"]["skus"]
+        ]
+
+    def get_erp_styles_input(self) -> list[dict]:
+        return [
+            {"style_name": name, "erp_style_codes": cfg["erp_codes"]}
+            for name, cfg in self.load()["styles"].items()
+            if cfg["erp_codes"]
+        ]
+
+    def get_bitable_style_map(self) -> dict:
+        """生成与 bitable_style_map.json 兼容的结构。"""
+        result: dict = {}
+        for name, cfg in self.load()["styles"].items():
+            sales = cfg.get("sales_bitable") or {}
+            if not sales.get("base_token") or not sales.get("table_id"):
+                continue
+            result[name] = {
+                "base_token": sales["base_token"],
+                "table_id": sales["table_id"],
+                "view_id": sales.get("view_id", ""),
+                "enabled": True,
+                "field_map": {
+                    "时间": "时间周期",
+                    "平台": "平台",
+                    "销量": "销量",
+                    "访客数": "访客数",
+                    "转化率": "转化率",
+                    "加购率": "加购率",
+                    "退款率": "退款率",
+                    "产品": "款式",
+                    "SourceID": "SourceID",
+                },
+            }
+        return result
+
+    def get_bitable_main_image_map(self) -> dict:
+        """生成与 bitable_main_image_map.json 兼容的结构。"""
+        result: dict = {}
+        for name, cfg in self.load()["styles"].items():
+            main = cfg.get("main_image_bitable") or {}
+            if not main.get("base_token") or not main.get("table_id"):
+                continue
+            result[name] = {
+                "display_name": name,
+                "base_token": main["base_token"],
+                "table_id": main["table_id"],
+                "view_id": main.get("view_id", ""),
+                "platform": main.get("platform", "天猫"),
+                "enabled": True,
+            }
+        return result
+
+    # ------------------------------------------------------------------
+    # 启动期「待采 / 跳过」门 — 显式返回带 reason 的跳过清单
+    # ------------------------------------------------------------------
+    def _cfg_has_sales_bitable(self, cfg: dict) -> bool:
+        return bool(cfg.get("sales_bitable", {}).get("base_token")) and bool(
+            cfg.get("sales_bitable", {}).get("table_id")
+        )
+
+    def get_daily_styles(self) -> dict:
+        """每日采集门 (v2.2):
+        - 款式级门: erp_codes ∩ sales_bitable 都非空 (从任意记录收集)
+        - 平台级门: 该平台分组 item_ids 非空
+
+        返回 {tm/jd/dy: {collectable: [...], skipped: [{style, reason}]}}。
+        skipped 仅记录「款式级门通过、平台级缺 IDs」的款式;款式级失败不计入 skipped
+        (因为该款式没有任何平台可采)。
+        """
+        payload = self.load()
+        result = {p: {"collectable": [], "skipped": []} for p in ("tm", "jd", "dy")}
+        for name, cfg in payload["styles"].items():
+            style_ok = bool(cfg.get("erp_codes")) and self._cfg_has_sales_bitable(cfg)
+            if not style_ok:
+                continue  # 款式级失败,无任何平台可采
+            for plat in ("tm", "jd", "dy"):
+                ids_key = "spus" if plat == "jd" else "item_ids"
+                if cfg.get(plat, {}).get(ids_key):
+                    result[plat]["collectable"].append(name)
+                else:
+                    result[plat]["skipped"].append(
+                        {"style": name, "reason": f"缺 {plat} 商品链接ID"}
+                    )
+        return result
+
+    def get_erp_daily_styles(self) -> dict:
+        """ERP daily 门:款式 + ERP 编码 + TM 分组有 item_ids + 飞书销量地址 都要有。
+
+        返回 {collectable, skipped, brand_map: {style: brand}}。
+        ERP 一次跑覆盖 jd/dy/tm 三平台(用同一组 erp_codes 查三个店铺),
+        所以 jd/dy 没商品 ID 也能跑,但 TM 分组必须有 item_ids
+        (飞书表格里 TM 分组是「基础分组」, 款式只要在 TM 上架就算款式存在)。
+        brand_map 给 collect_erp_yesterday_metrics.py 按品牌选店铺
+        (光影行星旗下店铺 vs ozko旗舰店)。
+        """
+        payload = self.load()
+        collectable: list[str] = []
+        skipped: list[dict] = []
+        brand_map: dict[str, str] = {}
+        for name, cfg in payload["styles"].items():
+            has_tm = bool(cfg.get("tm", {}).get("item_ids"))
+            if cfg.get("erp_codes") and has_tm and self._cfg_has_sales_bitable(cfg):
+                collectable.append(name)
+                brand_map[name] = cfg.get("brand") or "光影行星"
+            else:
+                miss = []
+                if not cfg.get("erp_codes"):
+                    miss.append("ERP款式编码")
+                if not has_tm:
+                    miss.append("天猫分组商品ID")
+                if not self._cfg_has_sales_bitable(cfg):
+                    miss.append("飞书多维表格地址")
+                skipped.append({"style": name, "reason": "缺 " + "/".join(miss)})
+        return {"collectable": collectable, "skipped": skipped, "brand_map": brand_map}
+
+    def get_jd_self_styles(self) -> list[str]:
+        """京东自营周汇总门:仅 SKU 非空即可 (用户确认保持现状)。"""
+        return [
+            name
+            for name, cfg in self.load()["styles"].items()
+            if cfg.get("jd_self", {}).get("skus")
+        ]
+
+    def get_main_image_styles(self) -> list[str]:
+        """主图门:天猫分组 + 主图飞书多维表格地址 同时存在。"""
+        payload = self.load()
+        result: list[str] = []
+        for name, cfg in payload["styles"].items():
+            main = cfg.get("main_image_bitable") or {}
+            if main.get("base_token") and main.get("table_id"):
+                result.append(name)
+        return result
+
+
+# ------------------------------------------------------------------------------
+# 便捷函数（全局默认实例）
+# ------------------------------------------------------------------------------
+
+def _default_loader() -> StyleConfigLoader:
+    return StyleConfigLoader()
+
+
+def load_config(force_refresh: bool = False, sync_fallback: bool = False) -> dict:
+    return _default_loader().load(force_refresh=force_refresh, sync_fallback=sync_fallback)
+
+
+def get_style_names() -> list[str]:
+    return _default_loader().get_style_names()
+
+
+def get_tm_style_id_map() -> dict[str, list[str]]:
+    return _default_loader().get_tm_style_id_map()
+
+
+def get_tm_persona_bitable_map() -> dict[str, dict[str, str]]:
+    return _default_loader().get_tm_persona_bitable_map()
+
+
+def get_dy_persona_bitable_map() -> dict[str, dict[str, str]]:
+    return _default_loader().get_dy_persona_bitable_map()
+
+
+def get_jd_persona_bitable_map() -> dict[str, dict[str, str]]:
+    return _default_loader().get_jd_persona_bitable_map()
+
+
+def get_style_analysis_bitable_map() -> dict[str, dict[str, str]]:
+    return _default_loader().get_style_analysis_bitable_map()
+
+
+def get_tm_brand_map() -> dict[str, str]:
+    return _default_loader().get_tm_brand_map()
+
+
+def get_jd_spu_groups() -> list[dict]:
+    return _default_loader().get_jd_spu_groups()
+
+
+def get_dy_product_groups() -> dict[str, list[str]]:
+    return _default_loader().get_dy_product_groups()
+
+
+def get_jd_self_sku_groups() -> list[dict]:
+    return _default_loader().get_jd_self_sku_groups()
+
+
+def get_erp_styles_input() -> list[dict]:
+    return _default_loader().get_erp_styles_input()
+
+
+def get_bitable_style_map() -> dict:
+    return _default_loader().get_bitable_style_map()
+
+
+def get_bitable_main_image_map() -> dict:
+    return _default_loader().get_bitable_main_image_map()
+
+
+def get_daily_styles() -> dict:
+    return _default_loader().get_daily_styles()
+
+
+def get_erp_daily_styles() -> dict:
+    return _default_loader().get_erp_daily_styles()
+
+
+def get_jd_self_styles() -> list[str]:
+    return _default_loader().get_jd_self_styles()
+
+
+def get_main_image_styles() -> list[str]:
+    return _default_loader().get_main_image_styles()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/db/__init__.py b/src/gyxx_flow/modules/product_commerce/runtime/db/__init__.py
new file mode 100644
index 0000000..2d40c1a
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/db/__init__.py
@@ -0,0 +1,1251 @@
+"""PostgreSQL 连接与 upsert helpers。
+
+从 PG_* 环境变量读连接信息（详见 .env.example）。导入时自动 load .env。
+
+典型用法:
+
+    from db import get_conn, upsert_weekly_metrics
+
+    with get_conn() as conn:
+        upsert_weekly_metrics(conn, records)
+
+表 → 写入方 / 读取方 (单源真相):
+    daily_style_metrics
+        写: aggregate_daily_final.py (每日 08:40)
+        读: check_nine_day_decline.py (9 段差分)
+            weekly_aggregate.py  (7 天 SUM 做周汇总)
+    weekly_style_metrics
+        写: weekly_aggregate.py (周一 09:30)
+        读: insert_bitable_records.py --source db (推飞书)
+    jd_self_weekly_sales
+        写: jd_self_inventory_sales_collector.py
+        读: weekly_aggregate.py (合到 jd 行 POP 销量)
+    decline_events
+        写: check_nine_day_decline.py
+        读: 审计/复盘
+    fact_alert_history
+        写: check_nine_day_decline.py (去重)
+        读: check_nine_day_decline.py._already_notified
+    dim_style
+        写: db/sync_dim_style.py (人工触发，全量扫 4 个数据源)
+        读: bitable 同步 / 业务查询
+    fact_platform_xlsx
+        写: 暂无 caller (db/__init__.py.upsert_platform_xlsx_rows 已写好，import 脚本待补)
+        读: 平台 XLSX 明细查询
+    product_daily_metrics
+        写: import_product_daily.py (扫 data/<plat>单品数据/_downloads/* 入库)
+        读: 三平台单品日维度查询 / 跨平台对账
+    product_reviews
+        写: import_product_reviews.py (扫 data/{jd,dy}单品数据/评价*.json 入库)
+        读: 评价分析 / 差评归因 / 标签挖掘
+    persona_screenshots
+        写: taobao_dmp_item_crowd_insight_screenshots.py (v2 改造后只入库, 不推飞书; insert_persona_to_bitable_v2.py 已删)
+        读: 人物画像审计 / 飞书附件溯源
+    main_image_creatives
+        写: scripts/insert_main_image_records.py (周任务, 采集后入库 + 推飞书)
+        读: 主图创意指标查询 / 飞书附件溯源
+    daily_persona_metrics
+        写: collect_persona_to_bitable.py / collect_dy_persona_to_bitable.py / collect_jd_persona_to_bitable.py
+        读: 三平台人群画像指标查询
+    fact_style_analysis_reports
+        写: analyze_style_with_hermes.py (Hermes analyzer 产出后归档)
+        读: 周复盘 / 飞书看板 / 报告归档查询
+    fact_market_rank_reports
+        写: collect_{sycm,jd,dy}_market_rank.py (飞书文档创建成功后)
+        读: 三平台市场排行报告索引 / 飞书文档归档查询
+    fact_market_rank_notifications
+        写: market_rank_hermes_notification.py (Hermes 分析端通知结果)
+        读: 同一收件人同日通知去重 / 发送审计
+
+所有 upsert 都用 ON CONFLICT ... DO UPDATE, 重复跑覆盖不双写。
+"""
+
+from __future__ import annotations
+
+import json
+import os
+from contextlib import contextmanager
+from pathlib import Path
+from typing import Any, Iterable, Mapping, Sequence
+
+from dotenv import load_dotenv
+from psycopg2 import pool
+from psycopg2.extras import execute_values
+from gyxx_flow.adapters import RuntimeServicePolicy
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent.parent
+load_dotenv(PROJECT_ROOT / ".env", override=False)
+os.environ.update(RuntimeServicePolicy().apply(os.environ))
+
+
+def build_pg_config(environ: Mapping[str, str]) -> dict[str, Any]:
+    """Build PostgreSQL config without ever falling back to a local database."""
+    required = ("PG_HOST", "PG_PORT", "PG_DB", "PG_USER", "PG_PASSWORD")
+    missing = [name for name in required if not str(environ.get(name, "")).strip()]
+    if missing:
+        raise RuntimeError(
+            "Missing required PostgreSQL configuration: " + ", ".join(missing)
+        )
+
+    try:
+        port = int(environ["PG_PORT"])
+    except ValueError as exc:
+        raise RuntimeError("PG_PORT must be an integer") from exc
+
+    return {
+        "host": environ["PG_HOST"].strip(),
+        "port": port,
+        "dbname": environ["PG_DB"].strip(),
+        "user": environ["PG_USER"].strip(),
+        "password": environ["PG_PASSWORD"],
+    }
+
+
+PG_CONFIG = build_pg_config(os.environ)
+
+
+_pool: pool.SimpleConnectionPool | None = None
+
+
+def _get_pool() -> pool.SimpleConnectionPool:
+    global _pool
+    if _pool is None:
+        _pool = pool.SimpleConnectionPool(
+            minconn=1,
+            maxconn=5,
+            **PG_CONFIG,
+        )
+    return _pool
+
+
+@contextmanager
+def get_conn():
+    """拿到一个 psycopg2 连接，用完自动归还连接池。"""
+    p = _get_pool()
+    conn = p.getconn()
+    if conn.closed:
+        # 长时间浏览器采集期间，服务端可能关闭池里的空闲连接。丢弃失效连接，
+        # 重新借一个，避免后续店铺退回不精确的关键词匹配。
+        p.putconn(conn, close=True)
+        conn = p.getconn()
+    try:
+        yield conn
+        conn.commit()
+    except Exception:
+        conn.rollback()
+        raise
+    finally:
+        p.putconn(conn)
+
+
+def init_schema(schema_path: str | Path | None = None) -> None:
+    """跑 schema.sql 建表（已 IF NOT EXISTS，幂等）。"""
+    path = Path(schema_path) if schema_path else PROJECT_ROOT / "db" / "schema.sql"
+    sql = path.read_text(encoding="utf-8")
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(sql)
+
+
+# ── daily_style_metrics ──────────────────────────────────────────────────────
+
+def upsert_daily_metrics(conn, records: Sequence[dict]) -> int:
+    """records: [{'metric_date', 'style_name', 'platform', 'sales', ...}, ...]
+
+    主键 = (metric_date, style_name, platform). 防止 orchestrator 多次传入同主键触发
+    PG "ON CONFLICT DO UPDATE command cannot affect row a second time".
+    """
+    if not records:
+        return 0
+    # dedup: 数字字段累加, erp_style_codes 合并
+    dedup: dict[tuple, dict] = {}
+    for r in records:
+        key = (r["metric_date"], r["style_name"], r["platform"])
+        if key not in dedup:
+            dedup[key] = {
+                "metric_date": r["metric_date"],
+                "style_name": r["style_name"],
+                "platform": r["platform"],
+                "sales": 0,
+                "refund_orders": 0,
+                "visitors": 0,
+                "cart_users": 0,
+                "erp_style_codes": [],
+                "source": r.get("source"),
+            }
+        cur = dedup[key]
+        for fld in ("sales", "refund_orders", "visitors", "cart_users"):
+            cur[fld] = int(cur.get(fld, 0) or 0) + int(r.get(fld, 0) or 0)
+        existing_codes = set(cur.get("erp_style_codes") or [])
+        for code in (r.get("erp_style_codes") or []):
+            existing_codes.add(code)
+        cur["erp_style_codes"] = sorted(existing_codes)
+        if r.get("source"):
+            cur["source"] = r["source"]
+    records = list(dedup.values())
+
+    rows = [
+        (
+            r["metric_date"],
+            r["style_name"],
+            r["platform"],
+            int(r.get("sales", 0) or 0),
+            int(r.get("refund_orders", 0) or 0),
+            int(r.get("visitors", 0) or 0),
+            int(r.get("cart_users", 0) or 0),
+            r.get("erp_style_codes") or [],
+            r.get("source"),
+        )
+        for r in records
+    ]
+    sql = """
+        INSERT INTO daily_style_metrics
+            (metric_date, style_name, platform, sales, refund_orders, visitors, cart_users, erp_style_codes, source)
+        VALUES %s
+        ON CONFLICT (metric_date, style_name, platform) DO UPDATE SET
+            sales = EXCLUDED.sales,
+            refund_orders = EXCLUDED.refund_orders,
+            visitors = EXCLUDED.visitors,
+            cart_users = EXCLUDED.cart_users,
+            erp_style_codes = EXCLUDED.erp_style_codes,
+            source = EXCLUDED.source,
+            updated_at = NOW()
+    """
+    with conn.cursor() as cur:
+        execute_values(cur, sql, rows)
+    return len(rows)
+
+
+# ── weekly_style_metrics ────────────────────────────────────────────────────
+
+def upsert_weekly_metrics(conn, records: Sequence[dict]) -> int:
+    """records: [{'week_end', 'week_start', 'time_period', 'style_name', 'platform',
+                  'sales', 'pop_sales', 'jd_self_sales', 'refund_orders', 'visitors',
+                  'cart_users', 'conversion_rate', 'cart_rate', 'refund_rate',
+                  'present_days', 'erp_style_codes', 'erp_source'}, ...]
+
+    主键 = (week_end, style_name, platform). 在 INSERT 前按主键 dedup + 数字累加,
+    防止 weekly_aggregate / import_history / orchestrator 多次传入同主键触发
+    PG "ON CONFLICT DO UPDATE command cannot affect row a second time".
+    """
+    if not records:
+        return 0
+
+    # ── 按主键 dedup + 数字累加 ───────────────────────────────────────────
+    dedup: dict[tuple, dict] = {}
+    for r in records:
+        key = (r["week_end"], r["style_name"], r["platform"])
+        if key not in dedup:
+            dedup[key] = {
+                "week_end": r["week_end"],
+                "week_start": r["week_start"],
+                "time_period": r.get("time_period", ""),
+                "style_name": r["style_name"],
+                "platform": r["platform"],
+                "sales": 0,
+                "pop_sales": 0,
+                "jd_self_sales": 0,
+                "refund_orders": 0,
+                "visitors": 0,
+                "cart_users": 0,
+                "conversion_rate": r.get("conversion_rate"),
+                "cart_rate": r.get("cart_rate"),
+                "refund_rate": r.get("refund_rate"),
+                "present_days": 0,
+                "erp_style_codes": [],
+                "erp_source": r.get("erp_source"),
+            }
+        cur = dedup[key]
+        # 数字字段累加
+        for fld in ("sales", "pop_sales", "jd_self_sales", "refund_orders",
+                    "visitors", "cart_users", "present_days"):
+            cur[fld] = int(cur.get(fld, 0) or 0) + int(r.get(fld, 0) or 0)
+        # erp_style_codes 合并去重
+        existing_codes = set(cur.get("erp_style_codes") or [])
+        for code in (r.get("erp_style_codes") or []):
+            existing_codes.add(code)
+        cur["erp_style_codes"] = sorted(existing_codes)
+        # 保留最后一行的 erp_source 标记
+        if r.get("erp_source"):
+            cur["erp_source"] = r["erp_source"]
+    records = list(dedup.values())
+    # ── dedup 完毕 ──────────────────────────────────────────────────────
+
+    rows = [
+        (
+            r["week_end"],
+            r["week_start"],
+            r.get("time_period", ""),
+            r["style_name"],
+            r["platform"],
+            int(r.get("sales", 0) or 0),
+            int(r.get("pop_sales", 0) or 0),
+            int(r.get("jd_self_sales", 0) or 0),
+            int(r.get("refund_orders", 0) or 0),
+            int(r.get("visitors", 0) or 0),
+            int(r.get("cart_users", 0) or 0),
+            r.get("conversion_rate"),
+            r.get("cart_rate"),
+            r.get("refund_rate"),
+            int(r.get("present_days", 0) or 0),
+            r.get("erp_style_codes") or [],
+            r.get("erp_source"),
+        )
+        for r in records
+    ]
+    sql = """
+        INSERT INTO weekly_style_metrics
+            (week_end, week_start, time_period, style_name, platform, sales, pop_sales,
+             jd_self_sales, refund_orders, visitors, cart_users, conversion_rate,
+             cart_rate, refund_rate, present_days, erp_style_codes, erp_source)
+        VALUES %s
+        ON CONFLICT (week_end, style_name, platform) DO UPDATE SET
+            week_start = EXCLUDED.week_start,
+            time_period = EXCLUDED.time_period,
+            sales = EXCLUDED.sales,
+            pop_sales = EXCLUDED.pop_sales,
+            jd_self_sales = EXCLUDED.jd_self_sales,
+            refund_orders = EXCLUDED.refund_orders,
+            visitors = EXCLUDED.visitors,
+            cart_users = EXCLUDED.cart_users,
+            conversion_rate = EXCLUDED.conversion_rate,
+            cart_rate = EXCLUDED.cart_rate,
+            refund_rate = EXCLUDED.refund_rate,
+            present_days = EXCLUDED.present_days,
+            erp_style_codes = EXCLUDED.erp_style_codes,
+            erp_source = EXCLUDED.erp_source,
+            updated_at = NOW()
+    """
+    with conn.cursor() as cur:
+        execute_values(cur, sql, rows)
+    return len(rows)
+
+
+def fetch_weekly_records(conn, week_end: str) -> list[dict]:
+    """读出某周全部 weekly_style_metrics，转回 weekly_records 形态。"""
+    with conn.cursor() as cur:
+        cur.execute(
+            """
+            SELECT week_start, week_end, time_period, style_name, platform,
+                   sales, jd_self_sales, refund_orders, visitors, cart_users,
+                   conversion_rate, cart_rate, refund_rate, present_days,
+                   erp_style_codes, erp_source
+            FROM weekly_style_metrics
+            WHERE week_end = %s
+            ORDER BY style_name, platform
+            """,
+            (week_end,),
+        )
+        cols = [d[0] for d in cur.description]
+        return [dict(zip(cols, row)) for row in cur.fetchall()]
+
+
+# ── jd_self_weekly_sales ────────────────────────────────────────────────────
+
+def upsert_jd_self_sales(conn, rows: Sequence[dict]) -> int:
+    """rows: [{'week_end', 'week_start', 'style_name', 'sku', 'outbound_qty', 'found', 'collected_at'}, ...]"""
+    if not rows:
+        return 0
+    payload = [
+        (
+            r["week_end"],
+            r["week_start"],
+            r["style_name"],
+            r["sku"],
+            int(r.get("outbound_qty", 0) or 0),
+            bool(r.get("found", False)),
+            r.get("collected_at"),
+        )
+        for r in rows
+    ]
+    sql = """
+        INSERT INTO jd_self_weekly_sales
+            (week_end, week_start, style_name, sku, outbound_qty, found, collected_at)
+        VALUES %s
+        ON CONFLICT (week_end, style_name, sku) DO UPDATE SET
+            outbound_qty = EXCLUDED.outbound_qty,
+            found = EXCLUDED.found,
+            collected_at = EXCLUDED.collected_at,
+            updated_at = NOW()
+    """
+    with conn.cursor() as cur:
+        execute_values(cur, sql, payload)
+    return len(payload)
+
+
+# ── daily_style_metrics helpers ──────────────────────────────────────────────
+
+def fetch_daily_pairs(conn, on_date: str) -> dict[tuple[str, str], dict]:
+    """读某日 daily_style_metrics 中所有 (platform, style_name) 的销量和退款订单数。
+
+    返回 {(platform, style_name): {"sales", "refund_orders", "visitors", "cart_users",
+                                    "erp_style_codes"}}。空集返回 {}。
+    """
+    with conn.cursor() as cur:
+        cur.execute(
+            """
+            SELECT platform, style_name, sales, refund_orders, visitors, cart_users, erp_style_codes
+            FROM daily_style_metrics
+            WHERE metric_date = %s
+            """,
+            (on_date,),
+        )
+        out: dict[tuple[str, str], dict] = {}
+        for plat, style, sales, refund, visitors, cart, codes in cur.fetchall():
+            out[(plat, style)] = {
+                "sales": int(sales or 0),
+                "refund_orders": int(refund or 0),
+                "visitors": int(visitors or 0),
+                "cart_users": int(cart or 0),
+                "erp_style_codes": list(codes or []),
+            }
+        return out
+
+
+def fetch_daily_for_range(conn, start_date: str, end_date: str) -> dict[tuple[str, str], dict]:
+    """读 [start_date, end_date] 区间内 daily_style_metrics 的原始日值,按 (platform, style_name) 聚合。
+
+    返回 {(platform, style_name): {"sales", "refund_orders", "visitors", "cart_users",
+                                    "present_days", "erp_style_codes"}}。
+    任一天缺数据则该 (plat, style) 的 present_days 不计,数值按 PG 实际有的行累加。
+
+    ⚠️ 之前用 LEFT JOIN LATERAL unnest(erp_style_codes) 会把每行按 code 数量复制,
+    SUM(sales) / COUNT(*) 因此翻倍 (一个款式 2 个 code → 销量×2)。
+    修法: 用 CTE 把数值聚合和 code 集合分开算, 最后再合并。
+    """
+    with conn.cursor() as cur:
+        cur.execute(
+            """
+            WITH sums AS (
+                SELECT platform, style_name,
+                       SUM(sales)         AS sales,
+                       SUM(refund_orders) AS refund_orders,
+                       SUM(visitors)      AS visitors,
+                       SUM(cart_users)    AS cart_users,
+                       COUNT(*)           AS present_days
+                FROM daily_style_metrics
+                WHERE metric_date BETWEEN %s AND %s
+                GROUP BY platform, style_name
+            ),
+            codes AS (
+                SELECT platform, style_name,
+                       array_agg(DISTINCT code ORDER BY code) AS erp_style_codes
+                FROM daily_style_metrics,
+                     LATERAL unnest(COALESCE(erp_style_codes, ARRAY[]::text[])) AS code
+                WHERE metric_date BETWEEN %s AND %s
+                GROUP BY platform, style_name
+            )
+            SELECT s.platform, s.style_name, s.sales, s.refund_orders,
+                   s.visitors, s.cart_users, s.present_days,
+                   COALESCE(c.erp_style_codes, ARRAY[]::text[]) AS erp_style_codes
+            FROM sums s
+            LEFT JOIN codes c USING (platform, style_name)
+            """,
+            (start_date, end_date, start_date, end_date),
+        )
+        out: dict[tuple[str, str], dict] = {}
+        for plat, style, sales, refund, visitors, cart, days, codes in cur.fetchall():
+            out[(plat, style)] = {
+                "sales": int(sales or 0),
+                "refund_orders": int(refund or 0),
+                "visitors": int(visitors or 0),
+                "cart_users": int(cart or 0),
+                "present_days": int(days or 0),
+                "erp_style_codes": sorted({c for c in (codes or []) if c}),
+            }
+        return out
+
+
+# ── decline_events ──────────────────────────────────────────────────────────
+
+def upsert_decline_event(conn, event: dict, hermes_response: Any = None) -> None:
+    sql = """
+        INSERT INTO decline_events
+            (end_date, window_days, style_name, platform, window_start,
+             seg1_sales, seg2_sales, seg3_sales, drop_pct_1to2, drop_pct_2to3,
+             drop_pct_1to3, erp_style_codes, hermes_notified, hermes_response)
+        VALUES %s
+        ON CONFLICT (end_date, style_name, platform) DO UPDATE SET
+            window_days = EXCLUDED.window_days,
+            window_start = EXCLUDED.window_start,
+            seg1_sales = EXCLUDED.seg1_sales,
+            seg2_sales = EXCLUDED.seg2_sales,
+            seg3_sales = EXCLUDED.seg3_sales,
+            drop_pct_1to2 = EXCLUDED.drop_pct_1to2,
+            drop_pct_2to3 = EXCLUDED.drop_pct_2to3,
+            drop_pct_1to3 = EXCLUDED.drop_pct_1to3,
+            erp_style_codes = EXCLUDED.erp_style_codes,
+            hermes_notified = EXCLUDED.hermes_notified,
+            hermes_response = EXCLUDED.hermes_response,
+            detected_at = NOW()
+    """
+    payload = (
+        event["end_date"],
+        int(event.get("window_days", 9)),
+        event["style_name"],
+        event["platform"],
+        event["window_start"],
+        int(event.get("seg1_sales", 0) or 0),
+        int(event.get("seg2_sales", 0) or 0),
+        int(event.get("seg3_sales", 0) or 0),
+        event.get("drop_pct_1to2"),
+        event.get("drop_pct_2to3"),
+        event.get("drop_pct_1to3"),
+        event.get("erp_style_codes") or [],
+        bool(hermes_response is not None and not (isinstance(hermes_response, dict) and "error" in hermes_response)),
+        json.dumps(hermes_response, ensure_ascii=False) if hermes_response is not None else None,
+    )
+    with conn.cursor() as cur:
+        execute_values(cur, sql, [payload])
+
+
+# ── fact_platform_xlsx ──────────────────────────────────────────────────────
+
+def upsert_platform_xlsx_rows(conn, records: Sequence[dict]) -> int:
+    """records: [{'platform', 'stat_date', 'row_index', 'item_id', 'item_name',
+                  'brand_model', 'category_l1', 'category_l2', 'is_total_row',
+                  'payload', 'source_file', 'source_file_mtime', 'source_file_size'}, ...]
+    唯一键: (platform, stat_date, source_file, row_index)
+    """
+    if not records:
+        return 0
+    from datetime import datetime, timezone
+
+    rows = []
+    for r in records:
+        mtime = r.get("source_file_mtime")
+        if isinstance(mtime, (int, float)):
+            mtime = datetime.fromtimestamp(mtime, tz=timezone.utc)
+        rows.append((
+            r["platform"],
+            r["stat_date"],
+            int(r["row_index"]),
+            r.get("item_id"),
+            r.get("item_name"),
+            r.get("brand_model"),
+            r.get("category_l1"),
+            r.get("category_l2"),
+            bool(r.get("is_total_row", False)),
+            json.dumps(r["payload"], ensure_ascii=False, default=str),
+            r["source_file"],
+            mtime,
+            r.get("source_file_size"),
+        ))
+
+    sql = """
+        INSERT INTO fact_platform_xlsx
+            (platform, stat_date, row_index, item_id, item_name, brand_model,
+             category_l1, category_l2, is_total_row, payload, source_file,
+             source_file_mtime, source_file_size)
+        VALUES %s
+        ON CONFLICT (platform, stat_date, source_file, row_index) DO UPDATE SET
+            item_id = EXCLUDED.item_id,
+            item_name = EXCLUDED.item_name,
+            brand_model = EXCLUDED.brand_model,
+            category_l1 = EXCLUDED.category_l1,
+            category_l2 = EXCLUDED.category_l2,
+            is_total_row = EXCLUDED.is_total_row,
+            payload = EXCLUDED.payload,
+            source_file_mtime = EXCLUDED.source_file_mtime,
+            source_file_size = EXCLUDED.source_file_size,
+            imported_at = NOW()
+    """
+    with conn.cursor() as cur:
+        execute_values(cur, sql, rows)
+    return len(rows)
+
+
+# ── fact_alert_history ──────────────────────────────────────────────────────
+
+def insert_fact_alert_history(conn, record: dict) -> int:
+    """record: {decline_event_key, style_name, trigger_date, consecutive_days, sales_drop_pct,
+                decline_window_start/end, baseline_window_start/end,
+                sent_at, recipient, send_channel, hermes_message_id, hermes_response, decline_payload}
+    唯一键: (style_name, trigger_date) — 同款同日只发一次
+    """
+    sql = """
+        INSERT INTO fact_alert_history
+            (decline_event_key, style_name, trigger_date, consecutive_days, sales_drop_pct,
+             decline_window_start, decline_window_end, baseline_window_start, baseline_window_end,
+             sent_at, recipient, send_channel, hermes_message_id, hermes_response, decline_payload)
+        VALUES %s
+        ON CONFLICT (style_name, trigger_date) DO UPDATE SET
+            decline_event_key = EXCLUDED.decline_event_key,
+            consecutive_days = EXCLUDED.consecutive_days,
+            sales_drop_pct = EXCLUDED.sales_drop_pct,
+            decline_window_start = EXCLUDED.decline_window_start,
+            decline_window_end = EXCLUDED.decline_window_end,
+            baseline_window_start = EXCLUDED.baseline_window_start,
+            baseline_window_end = EXCLUDED.baseline_window_end,
+            sent_at = EXCLUDED.sent_at,
+            recipient = EXCLUDED.recipient,
+            send_channel = EXCLUDED.send_channel,
+            hermes_message_id = EXCLUDED.hermes_message_id,
+            hermes_response = EXCLUDED.hermes_response,
+            decline_payload = EXCLUDED.decline_payload
+        RETURNING id
+    """
+    payload = (
+        record.get("decline_event_key"),
+        record["style_name"],
+        record["trigger_date"],
+        int(record.get("consecutive_days", 9)),
+        record.get("sales_drop_pct"),
+        record["decline_window_start"],
+        record["decline_window_end"],
+        record["baseline_window_start"],
+        record["baseline_window_end"],
+        record.get("sent_at"),
+        record.get("recipient"),
+        record.get("send_channel", "lark_im_group"),
+        record.get("hermes_message_id"),
+        json.dumps(record["hermes_response"], ensure_ascii=False) if record.get("hermes_response") is not None else None,
+        json.dumps(record["decline_payload"], ensure_ascii=False, default=str) if record.get("decline_payload") is not None else None,
+    )
+    with conn.cursor() as cur:
+        cur.execute(sql, [payload])
+        return cur.fetchone()[0]
+
+
+# ── product_daily_metrics ────────────────────────────────────────────────────
+
+def upsert_product_daily_metrics(conn, records: Sequence[dict]) -> int:
+    """records: [{
+        'platform', 'stat_date', 'shop_id', 'product_id', 'product_name', 'product_no',
+        'category_l1', 'category_l2',
+        'visitors', 'page_views', 'add_cart_qty', 'add_cart_buyers',
+        'paid_amount', 'paid_qty', 'paid_buyers',
+        'refund_amount', 'refund_orders',
+        'conversion_rate', 'avg_order_value', 'uv_value', 'detail_page_bounce_rate',
+        'raw_data', 'source_file'
+    }, ...]
+
+    唯一键: (platform, stat_date, product_id) — 同款同日重抓覆盖, 不双写。
+    """
+    if not records:
+        return 0
+
+    rows = []
+    for r in records:
+        rows.append((
+            r["platform"],
+            r["stat_date"],
+            r.get("shop_id"),
+            r["product_id"],
+            r.get("product_name"),
+            r.get("product_no"),
+            r.get("category_l1"),
+            r.get("category_l2"),
+            r.get("visitors"),
+            r.get("page_views"),
+            r.get("add_cart_qty"),
+            r.get("add_cart_buyers"),
+            r.get("paid_amount"),
+            r.get("paid_qty"),
+            r.get("paid_buyers"),
+            r.get("refund_amount"),
+            r.get("refund_orders"),
+            r.get("conversion_rate"),
+            r.get("avg_order_value"),
+            r.get("uv_value"),
+            r.get("detail_page_bounce_rate"),
+            json.dumps(r.get("raw_data") or {}, ensure_ascii=False, default=str),
+            r.get("source_file"),
+        ))
+
+    sql = """
+        INSERT INTO product_daily_metrics
+            (platform, stat_date, shop_id, product_id, product_name, product_no,
+             category_l1, category_l2,
+             visitors, page_views, add_cart_qty, add_cart_buyers,
+             paid_amount, paid_qty, paid_buyers,
+             refund_amount, refund_orders,
+             conversion_rate, avg_order_value, uv_value, detail_page_bounce_rate,
+             raw_data, source_file)
+        VALUES %s
+        ON CONFLICT (platform, stat_date, product_id) DO UPDATE SET
+            shop_id                 = EXCLUDED.shop_id,
+            product_name            = EXCLUDED.product_name,
+            product_no              = EXCLUDED.product_no,
+            category_l1             = EXCLUDED.category_l1,
+            category_l2             = EXCLUDED.category_l2,
+            visitors                = EXCLUDED.visitors,
+            page_views              = EXCLUDED.page_views,
+            add_cart_qty            = EXCLUDED.add_cart_qty,
+            add_cart_buyers         = EXCLUDED.add_cart_buyers,
+            paid_amount             = EXCLUDED.paid_amount,
+            paid_qty                = EXCLUDED.paid_qty,
+            paid_buyers             = EXCLUDED.paid_buyers,
+            refund_amount           = EXCLUDED.refund_amount,
+            refund_orders           = EXCLUDED.refund_orders,
+            conversion_rate         = EXCLUDED.conversion_rate,
+            avg_order_value         = EXCLUDED.avg_order_value,
+            uv_value                = EXCLUDED.uv_value,
+            detail_page_bounce_rate = EXCLUDED.detail_page_bounce_rate,
+            raw_data                = EXCLUDED.raw_data,
+            source_file             = EXCLUDED.source_file,
+            imported_at             = NOW()
+    """
+    with conn.cursor() as cur:
+        execute_values(cur, sql, rows)
+    return len(rows)
+
+
+# ── persona_screenshots ─────────────────────────────────────────────────────
+
+def upsert_persona_screenshot(conn, record: dict) -> int:
+    """record: {
+        'collect_date', 'style_name', 'platform' (默认 'tm'),
+        'item_id', 'screenshot_path', 'image_data' (bytes, 可选),
+        'file_size',
+        'bitable_base', 'bitable_table', 'bitable_record_id', 'bitable_file_token',
+        'uploaded_to_bitable'
+    }
+    主键 = (collect_date, style_name, platform)。同款同采集日期幂等覆盖。
+    """
+    sql = """
+        INSERT INTO persona_screenshots
+            (collect_date, style_name, platform, item_id, screenshot_path, image_data, file_size,
+             bitable_base, bitable_table, bitable_record_id, bitable_file_token,
+             uploaded_to_bitable)
+        VALUES %s
+        ON CONFLICT (collect_date, style_name, platform) DO UPDATE SET
+            item_id = EXCLUDED.item_id,
+            screenshot_path = EXCLUDED.screenshot_path,
+            image_data = EXCLUDED.image_data,
+            file_size = EXCLUDED.file_size,
+            bitable_base = EXCLUDED.bitable_base,
+            bitable_table = EXCLUDED.bitable_table,
+            bitable_record_id = EXCLUDED.bitable_record_id,
+            bitable_file_token = EXCLUDED.bitable_file_token,
+            uploaded_to_bitable = EXCLUDED.uploaded_to_bitable,
+            updated_at = NOW()
+    """
+    payload = (
+        record["collect_date"],
+        record["style_name"],
+        record.get("platform", "tm"),
+        record.get("item_id"),
+        record["screenshot_path"],
+        record.get("image_data"),  # bytes or None
+        record.get("file_size"),
+        record.get("bitable_base"),
+        record.get("bitable_table"),
+        record.get("bitable_record_id"),
+        record.get("bitable_file_token"),
+        bool(record.get("uploaded_to_bitable", False)),
+    )
+    with conn.cursor() as cur:
+        execute_values(cur, sql, [payload])
+    return 1
+
+
+# ── main_image_creatives ────────────────────────────────────────────────────
+
+def upsert_main_image_creatives(conn, records: Sequence[dict]) -> int:
+    """records: [{
+        'collect_date', 'style_name', 'platform' (默认 'tm'),
+        'image_key', 'creative_ids', 'creative_image_url',
+        'impressions', 'clicks', 'cost', 'click_rate', 'row_count',
+        'local_image_path', 'image_downloaded', 'image_fingerprint',
+        'bitable_base', 'bitable_table', 'bitable_record_id', 'bitable_file_token',
+        'uploaded_to_bitable'
+    }, ...]
+    主键 = (collect_date, style_name, image_key)。同款同采集日期同图幂等覆盖。
+    """
+    if not records:
+        return 0
+    sql = """
+        INSERT INTO main_image_creatives
+            (collect_date, style_name, platform, image_key, creative_ids, creative_image_url,
+             impressions, clicks, cost, click_rate, row_count,
+             local_image_path, image_downloaded, image_fingerprint,
+             bitable_base, bitable_table, bitable_record_id, bitable_file_token,
+             uploaded_to_bitable)
+        VALUES %s
+        ON CONFLICT (collect_date, style_name, image_key) DO UPDATE SET
+            creative_ids = EXCLUDED.creative_ids,
+            creative_image_url = EXCLUDED.creative_image_url,
+            impressions = EXCLUDED.impressions,
+            clicks = EXCLUDED.clicks,
+            cost = EXCLUDED.cost,
+            click_rate = EXCLUDED.click_rate,
+            row_count = EXCLUDED.row_count,
+            local_image_path = EXCLUDED.local_image_path,
+            image_downloaded = EXCLUDED.image_downloaded,
+            image_fingerprint = EXCLUDED.image_fingerprint,
+            bitable_base = EXCLUDED.bitable_base,
+            bitable_table = EXCLUDED.bitable_table,
+            bitable_record_id = EXCLUDED.bitable_record_id,
+            bitable_file_token = EXCLUDED.bitable_file_token,
+            uploaded_to_bitable = EXCLUDED.uploaded_to_bitable,
+            updated_at = NOW()
+    """
+    rows = [
+        (
+            r["collect_date"],
+            r["style_name"],
+            r.get("platform", "tm"),
+            r["image_key"],
+            r.get("creative_ids") or [],
+            r.get("creative_image_url"),
+            int(r.get("impressions", 0) or 0),
+            int(r.get("clicks", 0) or 0),
+            r.get("cost"),
+            r.get("click_rate"),
+            int(r.get("row_count", 0) or 0),
+            r.get("local_image_path"),
+            bool(r.get("image_downloaded", False)),
+            json.dumps(r["image_fingerprint"], ensure_ascii=False) if isinstance(r.get("image_fingerprint"), dict) else r.get("image_fingerprint"),
+            r.get("bitable_base"),
+            r.get("bitable_table"),
+            r.get("bitable_record_id"),
+            r.get("bitable_file_token"),
+            bool(r.get("uploaded_to_bitable", False)),
+        )
+        for r in records
+    ]
+    with conn.cursor() as cur:
+        execute_values(cur, sql, rows)
+    return len(rows)
+
+
+# ── daily_persona_metrics ────────────────────────────────────────────────────
+
+def _to_pct(val) -> float:
+    """统一转为百分比 (0-100)。输入可以是: 小数(0.765), 百分比字符串("87.85%"), 或百分比数字(87.85)。"""
+    s = str(val).replace("%", "").strip()
+    try:
+        f = float(s)
+    except ValueError:
+        return 0.0
+    return round(f * 100, 2) if f <= 1.5 else round(f, 2)
+
+
+def normalize_tm_payload(chart_values: dict) -> dict:
+    """TM chart_values → 归一化 payload (值已是百分比, 直接存)"""
+    DIM_MAP = {
+        "用户性别": "gender",
+        "用户年龄": "age",
+        "城市等级": "city_tier",
+        "消费能力等级": "buying_power",
+    }
+    payload = {}
+    for src_key, dst_key in DIM_MAP.items():
+        items = chart_values.get(src_key)
+        if not items:
+            continue
+        payload[dst_key] = [
+            {"name": it.get("category", ""), "value": round(float(it.get("分析人群占比", 0)), 2)}
+            for it in items
+        ]
+    return payload
+
+
+def normalize_dy_payload(profile: dict) -> dict:
+    """DY profile → 归一化 payload (小数×100 转百分比)"""
+    DIM_MAP = {
+        "gender_distribution": "gender",
+        "age_distribution": "age",
+        "strategy_crowd_distribution": "strategy_crowd",
+    }
+    payload = {}
+    for src_key, dst_key in DIM_MAP.items():
+        items = profile.get(src_key)
+        if not items:
+            continue
+        arr = []
+        for it in items:
+            name = it.get("name", "")
+            entry = {"name": name, "value": _to_pct(it.get("value", 0))}
+            if it.get("tgi") is not None:
+                entry["tgi"] = round(float(it["tgi"]), 4)
+            arr.append(entry)
+        payload[dst_key] = arr
+
+    # province: 过滤掉混入的人生阶段数据 (单身/恋爱/孕期/已育)
+    _LIFE_STAGES = {"单身", "恋爱", "孕期", "已育"}
+    prov_items = profile.get("resident_province") or []
+    prov = [
+        {"name": it["name"], "value": _to_pct(it.get("value", 0))}
+        for it in prov_items if it.get("name") not in _LIFE_STAGES
+    ]
+    if prov:
+        payload["province"] = prov
+
+    # life_stage: 单独提取
+    life = [
+        {"name": it["name"], "value": _to_pct(it.get("value", 0))}
+        for it in prov_items if it.get("name") in _LIFE_STAGES
+    ]
+    if life:
+        payload["life_stage"] = life
+
+    return payload
+
+
+def normalize_jd_payload(profile: dict) -> dict:
+    """JD profile → 归一化 payload (百分比字符串 "87.85%" → 87.85)"""
+    DIM_MAP = {
+        "性别": "gender",
+        "年龄": "age",
+        "购买力": "buying_power",
+    }
+    payload = {}
+    for src_key, dst_key in DIM_MAP.items():
+        items = profile.get(src_key)
+        if not items:
+            continue
+        payload[dst_key] = [
+            {"name": it.get("name", ""), "value": _to_pct(it.get("value", 0))}
+            for it in items
+        ]
+    return payload
+
+
+def upsert_persona_metrics(conn, records: Sequence[dict]) -> int:
+    """records: [{
+        'collect_date', 'style_name', 'platform',
+        'item_id', 'payload' (normalized dict), 'raw_payload' (original dict),
+        'source_file', 'collected_at'
+    }, ...]
+    PK = (collect_date, style_name, platform). 同款同日同平台幂等覆盖。
+    """
+    if not records:
+        return 0
+    sql = """
+        INSERT INTO daily_persona_metrics
+            (collect_date, style_name, platform, item_id,
+             payload, raw_payload, source_file, collected_at)
+        VALUES %s
+        ON CONFLICT (collect_date, style_name, platform) DO UPDATE SET
+            item_id = EXCLUDED.item_id,
+            payload = EXCLUDED.payload,
+            raw_payload = EXCLUDED.raw_payload,
+            source_file = EXCLUDED.source_file,
+            collected_at = EXCLUDED.collected_at,
+            updated_at = NOW()
+    """
+    rows = [
+        (
+            r["collect_date"],
+            r["style_name"],
+            r["platform"],
+            r.get("item_id"),
+            json.dumps(r.get("payload") or {}, ensure_ascii=False),
+            json.dumps(r.get("raw_payload"), ensure_ascii=False) if r.get("raw_payload") is not None else None,
+            r.get("source_file"),
+            r.get("collected_at"),
+        )
+        for r in records
+    ]
+    with conn.cursor() as cur:
+        execute_values(cur, sql, rows)
+    return len(rows)
+
+
+# ── product_reviews ───────────────────────────────────────────────────────────
+
+def upsert_product_reviews(conn, records: Sequence[dict]) -> int:
+    """records: [{
+        'platform' ('jd'/'dy'), 'collect_date', 'comment_date',
+        'style_name' (可空, DY 通过 product_id 反查 dim_style.dy_spus 填充),
+        'rating' ('好评'/'中评'/'差评'), 'score', 'content', 'tags' (list),
+        'user_name', 'order_id', 'product_id', 'product_name',
+        'sku_id', 'sku_name', 'sku_size',
+        'raw_data' (dict), 'source_file'
+    }, ...]
+    UNIQUE (platform, order_id) -> 同订单重抓覆盖。
+    """
+    if not records:
+        return 0
+    rows = [
+        (
+            r["platform"],
+            r["collect_date"],
+            r.get("comment_date"),
+            r.get("style_name"),
+            r["rating"],
+            r.get("score"),
+            r.get("content"),
+            r.get("tags") or None,
+            r.get("user_name"),
+            r["order_id"],
+            r.get("product_id"),
+            r.get("product_name"),
+            r.get("sku_id"),
+            r.get("sku_name"),
+            r.get("sku_size"),
+            json.dumps(r.get("raw_data") or {}, ensure_ascii=False, default=str),
+            r.get("source_file"),
+        )
+        for r in records
+    ]
+    sql = """
+        INSERT INTO product_reviews
+            (platform, collect_date, comment_date, style_name, rating, score,
+             content, tags, user_name, order_id, product_id, product_name,
+             sku_id, sku_name, sku_size, raw_data, source_file)
+        VALUES %s
+        ON CONFLICT (platform, order_id) DO UPDATE SET
+            collect_date  = EXCLUDED.collect_date,
+            comment_date  = EXCLUDED.comment_date,
+            style_name    = EXCLUDED.style_name,
+            rating        = EXCLUDED.rating,
+            score         = EXCLUDED.score,
+            content       = EXCLUDED.content,
+            tags          = EXCLUDED.tags,
+            user_name     = EXCLUDED.user_name,
+            product_id    = EXCLUDED.product_id,
+            product_name  = EXCLUDED.product_name,
+            sku_id        = EXCLUDED.sku_id,
+            sku_name      = EXCLUDED.sku_name,
+            sku_size      = EXCLUDED.sku_size,
+            raw_data      = EXCLUDED.raw_data,
+            source_file   = EXCLUDED.source_file,
+            imported_at   = NOW()
+    """
+    with conn.cursor() as cur:
+        execute_values(cur, sql, rows)
+    return len(rows)
+
+
+# ── fact_style_analysis_reports ──────────────────────────────────────────────
+
+def upsert_style_analysis_report(conn, record: dict) -> int:
+    """record: {
+        'style_name', 'analysis_date' (date), 'window_days' (int, 默认 7),
+        'window_start', 'window_end' (date),
+        'platforms_included' (list[str]),
+        'feishu_doc_id', 'feishu_doc_url', 'report_title',
+        'report_markdown', 'hermes_model',
+        'prompt_tokens', 'completion_tokens',
+        'source_payload' (dict, 可选)
+    }
+    业务 UNIQUE = (style_name, analysis_date, window_days). 同款同日同窗口重跑覆盖.
+    返回写入行的 id.
+    """
+    sql = """
+        INSERT INTO fact_style_analysis_reports
+            (style_name, analysis_date, window_days, window_start, window_end,
+             platforms_included, feishu_doc_id, feishu_doc_url, report_title,
+             report_markdown, hermes_model, prompt_tokens, completion_tokens,
+             source_payload)
+        VALUES %s
+        ON CONFLICT (style_name, analysis_date, window_days) DO UPDATE SET
+            window_start        = EXCLUDED.window_start,
+            window_end          = EXCLUDED.window_end,
+            platforms_included  = EXCLUDED.platforms_included,
+            feishu_doc_id       = EXCLUDED.feishu_doc_id,
+            feishu_doc_url      = EXCLUDED.feishu_doc_url,
+            report_title        = EXCLUDED.report_title,
+            report_markdown     = EXCLUDED.report_markdown,
+            hermes_model        = EXCLUDED.hermes_model,
+            prompt_tokens       = EXCLUDED.prompt_tokens,
+            completion_tokens   = EXCLUDED.completion_tokens,
+            source_payload      = EXCLUDED.source_payload,
+            updated_at          = NOW()
+        RETURNING id
+    """
+    payload = (
+        record["style_name"],
+        record["analysis_date"],
+        int(record.get("window_days", 7)),
+        record["window_start"],
+        record["window_end"],
+        record.get("platforms_included") or [],
+        record.get("feishu_doc_id"),
+        record.get("feishu_doc_url"),
+        record.get("report_title"),
+        record["report_markdown"],
+        record["hermes_model"],
+        record.get("prompt_tokens"),
+        record.get("completion_tokens"),
+        json.dumps(record["source_payload"], ensure_ascii=False, default=str)
+            if record.get("source_payload") is not None else None,
+    )
+    with conn.cursor() as cur:
+        execute_values(cur, sql, [payload])
+        cur.execute(
+            "SELECT id FROM fact_style_analysis_reports "
+            "WHERE style_name=%s AND analysis_date=%s AND window_days=%s",
+            (record["style_name"], record["analysis_date"],
+             int(record.get("window_days", 7))),
+        )
+        return cur.fetchone()[0]
+
+
+# ── fact_market_rank_reports ─────────────────────────────────────────────────
+
+def upsert_market_rank_report(conn, record: dict) -> int:
+    """保存三平台市场排行飞书文档，返回归档行 id。
+
+    业务 UNIQUE = (platform, report_date)。同一平台同一天无论重跑多少次，均更新
+    当天唯一记录的链接及运行元数据，不产生重复记录。
+    """
+    sql = """
+        INSERT INTO fact_market_rank_reports
+            (platform, report_date, period, category_scope, category_names,
+             report_title, feishu_doc_id, feishu_doc_url, product_count,
+             source_json_path, run_metadata)
+        VALUES (%s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s)
+        ON CONFLICT (platform, report_date) DO UPDATE SET
+            period           = EXCLUDED.period,
+            category_scope   = EXCLUDED.category_scope,
+            category_names   = EXCLUDED.category_names,
+            report_title     = EXCLUDED.report_title,
+            feishu_doc_id    = EXCLUDED.feishu_doc_id,
+            feishu_doc_url   = EXCLUDED.feishu_doc_url,
+            product_count    = EXCLUDED.product_count,
+            source_json_path = EXCLUDED.source_json_path,
+            run_metadata     = EXCLUDED.run_metadata,
+            generated_at     = NOW(),
+            updated_at       = NOW()
+        RETURNING id
+    """
+    payload = (
+        record["platform"],
+        record["report_date"],
+        record["period"],
+        record["category_scope"],
+        record.get("category_names") or [],
+        record["report_title"],
+        record.get("feishu_doc_id"),
+        record["feishu_doc_url"],
+        int(record.get("product_count", 0) or 0),
+        record.get("source_json_path"),
+        json.dumps(record["run_metadata"], ensure_ascii=False, default=str)
+            if record.get("run_metadata") is not None else None,
+    )
+    with conn.cursor() as cur:
+        cur.execute(sql, payload)
+        row = cur.fetchone()
+    if not row:
+        raise RuntimeError("市场排行飞书文档归档未返回记录 id")
+    return int(row[0])
+
+
+# ── fact_market_rank_notifications ──────────────────────────────────────────
+
+def ensure_market_rank_notification_table(conn) -> None:
+    """确保通知审计表存在，便于老库在不跑整份 schema 时平滑升级。"""
+    sql = """
+        CREATE TABLE IF NOT EXISTS fact_market_rank_notifications (
+            id                  BIGSERIAL PRIMARY KEY,
+            report_date         DATE NOT NULL,
+            recipient_open_id   TEXT NOT NULL,
+            recipient_name      TEXT,
+            platform_links      JSONB NOT NULL DEFAULT '{}'::jsonb,
+            status              TEXT NOT NULL
+                                CHECK (status IN ('incomplete', 'sent', 'failed')),
+            hermes_message_id   TEXT,
+            hermes_response     JSONB,
+            sent_at             TIMESTAMPTZ,
+            created_at          TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+            updated_at          TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+            UNIQUE (report_date, recipient_open_id)
+        )
+    """
+    with conn.cursor() as cur:
+        cur.execute(sql)
+
+
+def get_market_rank_reports_for_date(
+    conn,
+    report_date,
+    platforms: Iterable[str] = ("tm", "jd", "dy"),
+    *,
+    updated_after=None,
+) -> list[dict[str, Any]]:
+    wanted = tuple(platforms)
+    sql = """
+        SELECT platform, feishu_doc_url, product_count, report_title,
+               generated_at, updated_at
+          FROM fact_market_rank_reports
+         WHERE report_date = %s
+           AND platform = ANY(%s)
+           AND (%s IS NULL OR updated_at >= %s)
+    """
+    with conn.cursor() as cur:
+        cur.execute(
+            sql,
+            (report_date, list(wanted), updated_after, updated_after),
+        )
+        rows = cur.fetchall()
+    columns = (
+        "platform",
+        "feishu_doc_url",
+        "product_count",
+        "report_title",
+        "generated_at",
+        "updated_at",
+    )
+    return [dict(zip(columns, row)) for row in rows]
+
+
+def market_rank_notification_was_sent(
+    conn,
+    report_date,
+    recipient_open_id: str,
+    platform_links: Mapping[str, str] | None = None,
+) -> bool:
+    """仅当同日已经发送过完全相同的一组链接时才跳过。"""
+    with conn.cursor() as cur:
+        cur.execute(
+            "SELECT platform_links FROM fact_market_rank_notifications "
+            "WHERE report_date=%s AND recipient_open_id=%s AND status='sent'",
+            (report_date, recipient_open_id),
+        )
+        row = cur.fetchone()
+    if row is None:
+        return False
+    if platform_links is None:
+        return True
+    stored = row[0]
+    if isinstance(stored, str):
+        try:
+            stored = json.loads(stored)
+        except json.JSONDecodeError:
+            return False
+    return dict(stored or {}) == dict(platform_links)
+
+
+def upsert_market_rank_notification(conn, record: dict) -> int:
+    sql = """
+        INSERT INTO fact_market_rank_notifications
+            (report_date, recipient_open_id, recipient_name, platform_links,
+             status, hermes_message_id, hermes_response, sent_at)
+        VALUES (%s, %s, %s, %s, %s, %s, %s,
+                CASE WHEN %s = 'sent' THEN NOW() ELSE NULL END)
+        ON CONFLICT (report_date, recipient_open_id) DO UPDATE SET
+            recipient_name    = EXCLUDED.recipient_name,
+            platform_links    = EXCLUDED.platform_links,
+            status            = EXCLUDED.status,
+            hermes_message_id = EXCLUDED.hermes_message_id,
+            hermes_response   = EXCLUDED.hermes_response,
+            sent_at           = CASE
+                                    WHEN EXCLUDED.status = 'sent' THEN NOW()
+                                    ELSE fact_market_rank_notifications.sent_at
+                                END,
+            updated_at        = NOW()
+        RETURNING id
+    """
+    status = record["status"]
+    payload = (
+        record["report_date"],
+        record["recipient_open_id"],
+        record.get("recipient_name"),
+        json.dumps(record.get("platform_links") or {}, ensure_ascii=False),
+        status,
+        record.get("hermes_message_id"),
+        json.dumps(record.get("hermes_response"), ensure_ascii=False, default=str)
+            if record.get("hermes_response") is not None else None,
+        status,
+    )
+    with conn.cursor() as cur:
+        cur.execute(sql, payload)
+        row = cur.fetchone()
+    if not row:
+        raise RuntimeError("市场排行 Hermes 通知归档未返回记录 id")
+    return int(row[0])
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/db/schema.sql b/src/gyxx_flow/modules/product_commerce/runtime/db/schema.sql
new file mode 100644
index 0000000..97d08ae
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/db/schema.sql
@@ -0,0 +1,768 @@
+-- ============================================================================
+-- product-collector-analyze-flow / PostgreSQL schema (database: data_hub)
+-- 数据库: data_hub（由 PG_DB 环境变量决定）
+-- 设计原则: schema 与 data/ 目录下的 JSON 产物 1:1 映射；upsert 幂等
+-- ============================================================================
+
+-- 款式维度表：维护每款的 (各平台 SPU/ID + 京东自营 SKU + ERP 编码 + 飞书表地址)
+-- 数据源:
+--   jd_self_skus      ← jd_self_inventory_sales_collector.STYLE_SKU_GROUPS
+--   jd_spus/dy_spus/tm_spus ← data/<plat>/<款>_<日期>/<款>_<日期>.json 的"匹配SPU"
+--   erp_codes         ← daily_style_metrics.erp_style_codes (自动聚)
+--   bitable_weekly_*  ← bitable_style_map.json (每款周报表地址)
+--   bitable_sku_master_* ← 飞书 SKU 主表（人工维护；表 base_token + table_id 填这里）
+CREATE TABLE IF NOT EXISTS dim_style (
+    style_name              TEXT PRIMARY KEY,
+    jd_spus                 TEXT[],                    -- 京东 POP 商品 SPU
+    dy_spus                 TEXT[],                    -- 抖音商品 SPU
+    tm_spus                 TEXT[],                    -- 天猫商品 SPU
+    jd_self_skus            TEXT[],                    -- 京东自营供应链 SKU 集合
+    erp_codes               TEXT[],                    -- 聚水潭 ERP 编码
+    bitable_weekly_url      TEXT,                      -- 款式周报表 URL (bitable_style_map 派生)
+    bitable_weekly_base     TEXT,                      -- 款式周报表 base_token
+    bitable_weekly_table    TEXT,                      -- 款式周报表 table_id
+    bitable_sku_master_url  TEXT,                      -- 飞书 SKU 主表 URL (人工维护)
+    bitable_sku_master_base TEXT,                      -- 飞书 SKU 主表 base_token (人工维护)
+    bitable_sku_master_table TEXT,                     -- 飞书 SKU 主表 table_id (人工维护)
+    category                TEXT,
+    launch_date             DATE,
+    notes                   TEXT,
+    updated_at              TIMESTAMPTZ NOT NULL DEFAULT NOW()
+);
+
+-- 每日款式级指标 (来自 aggregate_daily_final.py → data/<date>/<款>/final_summary.json)
+CREATE TABLE IF NOT EXISTS daily_style_metrics (
+    metric_date     DATE NOT NULL,
+    style_name      TEXT NOT NULL,
+    platform        TEXT NOT NULL,                  -- 'jd' / 'dy' / 'tm'
+    sales           INTEGER NOT NULL DEFAULT 0,     -- 平台报告销量（ERP 已覆盖）
+    refund_orders   INTEGER NOT NULL DEFAULT 0,
+    visitors        INTEGER NOT NULL DEFAULT 0,
+    cart_users      INTEGER NOT NULL DEFAULT 0,
+    erp_style_codes TEXT[],
+    source          TEXT,                            -- final_summary.json 路径
+    updated_at      TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    PRIMARY KEY (metric_date, style_name, platform)
+);
+CREATE INDEX IF NOT EXISTS daily_style_metrics_date_idx
+    ON daily_style_metrics (metric_date);
+CREATE INDEX IF NOT EXISTS daily_style_metrics_style_idx
+    ON daily_style_metrics (style_name, metric_date);
+
+-- 周汇总 (来自 weekly_aggregate.py → data/<M月第N周>/<款>/weekly_summary.json)
+-- 注: 京东款式 sales = pop_sales + jd_self_sales，转化率/退款率 用 pop_sales
+CREATE TABLE IF NOT EXISTS weekly_style_metrics (
+    week_end        DATE NOT NULL,                  -- 周日 (例: 2026-06-07)
+    week_start      DATE NOT NULL,                  -- 周一
+    time_period     TEXT NOT NULL,                  -- '6月第1周'
+    style_name      TEXT NOT NULL,
+    platform        TEXT NOT NULL,                  -- 'jd' / 'dy' / 'tm'
+    sales           INTEGER NOT NULL DEFAULT 0,     -- POP + 自营
+    pop_sales       INTEGER NOT NULL DEFAULT 0,     -- POP 店销量 (算转化率/退款率用)
+    jd_self_sales   INTEGER NOT NULL DEFAULT 0,     -- 京东自营出库件数
+    refund_orders   INTEGER NOT NULL DEFAULT 0,
+    visitors        INTEGER NOT NULL DEFAULT 0,
+    cart_users      INTEGER NOT NULL DEFAULT 0,
+    conversion_rate TEXT,                            -- "12.34%" 格式 (pop_sales / visitors)
+    cart_rate       TEXT,                            -- "8.44%" 格式
+    refund_rate     TEXT,                            -- "27.27%" 格式
+    present_days    SMALLINT NOT NULL DEFAULT 0,
+    erp_style_codes TEXT[],
+    erp_source      TEXT,                            -- 'daily_pg' (从 PG daily_style_metrics 7 天求和)
+    updated_at      TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    PRIMARY KEY (week_end, style_name, platform)
+);
+CREATE INDEX IF NOT EXISTS weekly_style_metrics_end_idx
+    ON weekly_style_metrics (week_end);
+CREATE INDEX IF NOT EXISTS weekly_style_metrics_style_idx
+    ON weekly_style_metrics (style_name, week_end);
+
+-- 京东自营周出库件数 (来自 jd_self_inventory_sales_collector.py)
+-- SKU 维度原始记录；weekly_style_metrics.jd_self_sales 是按款汇总后的值
+CREATE TABLE IF NOT EXISTS jd_self_weekly_sales (
+    week_end        DATE NOT NULL,
+    week_start      DATE NOT NULL,
+    style_name      TEXT NOT NULL,
+    sku             TEXT NOT NULL,
+    outbound_qty    INTEGER NOT NULL DEFAULT 0,
+    found           BOOLEAN NOT NULL DEFAULT FALSE,
+    collected_at    TIMESTAMPTZ,
+    updated_at      TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    PRIMARY KEY (week_end, style_name, sku)
+);
+CREATE INDEX IF NOT EXISTS jd_self_weekly_sales_style_idx
+    ON jd_self_weekly_sales (style_name, week_end);
+
+-- 9 天连续下滑事件 (来自 check_nine_day_decline.py 命中)
+CREATE TABLE IF NOT EXISTS decline_events (
+    detected_at     TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    end_date        DATE NOT NULL,                  -- 窗口右端
+    window_days     SMALLINT NOT NULL DEFAULT 9,
+    style_name      TEXT NOT NULL,
+    platform        TEXT NOT NULL,
+    window_start    DATE NOT NULL,
+    seg1_sales      INTEGER NOT NULL,
+    seg2_sales      INTEGER NOT NULL,
+    seg3_sales      INTEGER NOT NULL,
+    drop_pct_1to2   NUMERIC(6,2),
+    drop_pct_2to3   NUMERIC(6,2),
+    drop_pct_1to3   NUMERIC(6,2),
+    erp_style_codes TEXT[],
+    hermes_notified BOOLEAN NOT NULL DEFAULT FALSE,
+    hermes_response TEXT,
+    PRIMARY KEY (end_date, style_name, platform)
+);
+CREATE INDEX IF NOT EXISTS decline_events_end_idx
+    ON decline_events (end_date);
+
+-- 9 天告警发送历史 (从 check_nine_day_decline.py 派生，1 条告警 = 1 行)
+-- decline_event_key 是 decline_events 的业务键 (end_date|style_name|platform)，应用层 join
+CREATE TABLE IF NOT EXISTS fact_alert_history (
+    id                    BIGSERIAL PRIMARY KEY,
+    decline_event_key     TEXT,                          -- 业务键 'YYYY-MM-DD|款|plat'
+    style_name            TEXT NOT NULL,
+    trigger_date          DATE NOT NULL,                 -- 触发告警的统计日期 (= decline_events.end_date)
+    consecutive_days      SMALLINT NOT NULL DEFAULT 9,
+    sales_drop_pct        NUMERIC(6,2),                  -- 总跌幅 1to3
+    decline_window_start  DATE NOT NULL,                 -- 下滑窗口起点 (= decline_events.window_start)
+    decline_window_end    DATE NOT NULL,                 -- 下滑窗口终点
+    baseline_window_start DATE NOT NULL,                 -- 对比基期起点
+    baseline_window_end   DATE NOT NULL,                 -- 对比基期终点
+    sent_at               TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    recipient             TEXT,                          -- chat_id / open_id
+    send_channel          VARCHAR(32) DEFAULT 'lark_im_group',
+    hermes_message_id     TEXT,
+    hermes_response       JSONB,                         -- Hermes 完整响应
+    decline_payload       JSONB,                         -- 触发时的 9 天 daily 快照
+    created_at            TIMESTAMPTZ DEFAULT NOW(),
+    UNIQUE (style_name, trigger_date)
+);
+CREATE INDEX IF NOT EXISTS alert_history_style_date_idx
+    ON fact_alert_history (style_name, trigger_date DESC);
+CREATE INDEX IF NOT EXISTS alert_history_sent_at_idx
+    ON fact_alert_history (sent_at DESC);
+
+-- 平台 XLSX 明细 (3 平台合一表，动态列存 JSONB)
+-- 数据源:
+--   jd/data/jd单品数据/_downloads/<date>/11272270_<date>_全部渠道_环比商品明细.xlsx (91 行 × 87 列)
+--   dy/data/dy单品数据/_downloads/<date>/经营版_商品_商品列表__<start>-<end>_数据更新时间<date>.xlsx (153 行 × 22 列)
+--   tm/data/tm单品数据/【生意参谋平台】商品_全部_<date>_<date>.xls (135 行 × 38 列, 表头在 r4)
+CREATE TABLE IF NOT EXISTS fact_platform_xlsx (
+    id                  BIGSERIAL PRIMARY KEY,
+    platform            VARCHAR(8) NOT NULL CHECK (platform IN ('jd','dy','tm')),
+    stat_date           DATE NOT NULL,                  -- 文件统计日期
+    row_index           INT NOT NULL,                   -- 数据行号 (1 = 合计行 / 数据起点)
+    item_id             TEXT,                           -- 平台商品ID
+    item_name           TEXT,                           -- 商品名称
+    brand_model         TEXT,                           -- 货号 (jd/tm 有, dy 留空)
+    category_l1         TEXT,                           -- 平台一级类目
+    category_l2         TEXT,                           -- 平台二级类目
+    is_total_row        BOOLEAN NOT NULL DEFAULT FALSE, -- 合计行标记
+    payload             JSONB NOT NULL,                 -- 全量动态列 (含原始字符串/数字)
+    source_file         TEXT NOT NULL,                  -- 相对路径 data/...
+    source_file_mtime   TIMESTAMPTZ,
+    source_file_size    INT,
+    imported_at         TIMESTAMPTZ DEFAULT NOW(),
+    UNIQUE (platform, stat_date, source_file, row_index)
+);
+CREATE INDEX IF NOT EXISTS plat_xlsx_date_idx
+    ON fact_platform_xlsx (platform, stat_date DESC);
+CREATE INDEX IF NOT EXISTS plat_xlsx_item_idx
+    ON fact_platform_xlsx (platform, item_id);
+
+-- ── 兼容旧版 dim_style：init_schema 在已有库上跑时给老表加列 ─────────────
+ALTER TABLE dim_style ADD COLUMN IF NOT EXISTS jd_spus                 TEXT[];
+ALTER TABLE dim_style ADD COLUMN IF NOT EXISTS dy_spus                 TEXT[];
+ALTER TABLE dim_style ADD COLUMN IF NOT EXISTS tm_spus                 TEXT[];
+ALTER TABLE dim_style ADD COLUMN IF NOT EXISTS erp_codes               TEXT[];
+ALTER TABLE dim_style ADD COLUMN IF NOT EXISTS bitable_weekly_url      TEXT;
+ALTER TABLE dim_style ADD COLUMN IF NOT EXISTS bitable_weekly_base     TEXT;
+ALTER TABLE dim_style ADD COLUMN IF NOT EXISTS bitable_weekly_table    TEXT;
+ALTER TABLE dim_style ADD COLUMN IF NOT EXISTS bitable_sku_master_url  TEXT;
+ALTER TABLE dim_style ADD COLUMN IF NOT EXISTS bitable_sku_master_base TEXT;
+ALTER TABLE dim_style ADD COLUMN IF NOT EXISTS bitable_sku_master_table TEXT;
+-- ============================================================================
+-- 表 / 列 注释 (TABLE + COLUMN COMMENT)
+-- 由 _add_comments.py 自动生成，请勿手改这里；如需调整请改脚本里的 COMMENTS 字典
+-- 重新跑: python db/_add_comments.py
+-- ============================================================================
+
+-- dim_style
+COMMENT ON TABLE dim_style IS '款式维度表 — 维护每款 (各平台 SPU/ID + 京东自营 SKU + ERP 编码 + 飞书表地址 + 品类/上架日期/备注)。数据源: jd_self_skus、jd/dy/tm_spus、erp_codes、bitable_style_map.json、飞书 SKU 主表 (人工维护)。';
+COMMENT ON COLUMN dim_style.style_name IS '款式名 (主键)，与 PRODUCT_STYLE_GROUPS 的 key 对齐';
+COMMENT ON COLUMN dim_style.jd_spus IS '京东 POP 商品 SPU 列表';
+COMMENT ON COLUMN dim_style.dy_spus IS '抖音商品 SPU 列表';
+COMMENT ON COLUMN dim_style.tm_spus IS '天猫商品 SPU 列表';
+COMMENT ON COLUMN dim_style.jd_self_skus IS '京东自营供应链 SKU 集合';
+COMMENT ON COLUMN dim_style.erp_codes IS '聚水潭 ERP 编码列表';
+COMMENT ON COLUMN dim_style.bitable_weekly_url IS '款式周报表 URL (从 bitable_style_map.json 派生)';
+COMMENT ON COLUMN dim_style.bitable_weekly_base IS '款式周报表 base_token';
+COMMENT ON COLUMN dim_style.bitable_weekly_table IS '款式周报表 table_id';
+COMMENT ON COLUMN dim_style.bitable_sku_master_url IS '飞书 SKU 主表 URL (人工维护)';
+COMMENT ON COLUMN dim_style.bitable_sku_master_base IS '飞书 SKU 主表 base_token (人工维护)';
+COMMENT ON COLUMN dim_style.bitable_sku_master_table IS '飞书 SKU 主表 table_id (人工维护)';
+COMMENT ON COLUMN dim_style.category IS '款式品类 (如 ''斜挎包''/''双肩包'' 等)';
+COMMENT ON COLUMN dim_style.launch_date IS '款式上架日期';
+COMMENT ON COLUMN dim_style.notes IS '备注 / 人工标签';
+COMMENT ON COLUMN dim_style.updated_at IS '最近更新时间';
+
+-- daily_style_metrics
+COMMENT ON TABLE daily_style_metrics IS '每日款式级指标 — 来源: aggregate_daily_final.py → data/<date>/<款>/final_summary.json。PK = (metric_date, style_name, platform)。';
+COMMENT ON COLUMN daily_style_metrics.metric_date IS '指标日期';
+COMMENT ON COLUMN daily_style_metrics.style_name IS '款式名';
+COMMENT ON COLUMN daily_style_metrics.platform IS '平台代码 (''jd'' / ''dy'' / ''tm'')';
+COMMENT ON COLUMN daily_style_metrics.sales IS '平台报告销量 (ERP 已覆盖的最终值)';
+COMMENT ON COLUMN daily_style_metrics.refund_orders IS '退款订单数';
+COMMENT ON COLUMN daily_style_metrics.visitors IS '访客数';
+COMMENT ON COLUMN daily_style_metrics.cart_users IS '加购人数';
+COMMENT ON COLUMN daily_style_metrics.erp_style_codes IS '覆盖销量用的 ERP 编码列表';
+COMMENT ON COLUMN daily_style_metrics.source IS '平台 <款>_<日期>.json 路径 (例如 data/jd单品数据/云卷2_2026-06-11/云卷2_2026-06-11.json)，不是 final_summary.json';
+COMMENT ON COLUMN daily_style_metrics.updated_at IS '最近更新时间';
+
+-- weekly_style_metrics
+COMMENT ON TABLE weekly_style_metrics IS '周汇总指标 — 来源: weekly_aggregate.py → weekly_summary.json。PK = (week_end, style_name, platform)。京东款 sales = pop_sales + jd_self_sales，转化率/退款率 用 pop_sales 算。';
+COMMENT ON COLUMN weekly_style_metrics.week_end IS '周结束日 (周日)';
+COMMENT ON COLUMN weekly_style_metrics.week_start IS '周开始日 (周一)';
+COMMENT ON COLUMN weekly_style_metrics.time_period IS '周次标签 (如 ''6月第1周'')';
+COMMENT ON COLUMN weekly_style_metrics.style_name IS '款式名';
+COMMENT ON COLUMN weekly_style_metrics.platform IS '平台代码 (''jd'' / ''dy'' / ''tm'')';
+COMMENT ON COLUMN weekly_style_metrics.sales IS '销量合计: 京东款 = pop_sales + jd_self_sales；dy/tm 只有 POP 销量';
+COMMENT ON COLUMN weekly_style_metrics.pop_sales IS 'POP 店销量 (算转化率/退款率用)';
+COMMENT ON COLUMN weekly_style_metrics.jd_self_sales IS '京东自营出库件数';
+COMMENT ON COLUMN weekly_style_metrics.refund_orders IS '退款订单数';
+COMMENT ON COLUMN weekly_style_metrics.visitors IS '访客数';
+COMMENT ON COLUMN weekly_style_metrics.cart_users IS '加购人数';
+COMMENT ON COLUMN weekly_style_metrics.conversion_rate IS '转化率 (pop_sales / visitors) 字符串格式 ''12.34%''';
+COMMENT ON COLUMN weekly_style_metrics.cart_rate IS '加购率 字符串格式 ''8.44%''';
+COMMENT ON COLUMN weekly_style_metrics.refund_rate IS '退款率 字符串格式 ''27.27%''';
+COMMENT ON COLUMN weekly_style_metrics.present_days IS '本周有数据的有效天数';
+COMMENT ON COLUMN weekly_style_metrics.erp_style_codes IS 'ERP 编码列表';
+COMMENT ON COLUMN weekly_style_metrics.erp_source IS 'ERP 数据来源 (''daily_pg'': 7 天 daily_style_metrics 求和)';
+COMMENT ON COLUMN weekly_style_metrics.updated_at IS '最近更新时间';
+
+-- jd_self_weekly_sales
+COMMENT ON TABLE jd_self_weekly_sales IS '京东自营周出库件数 — 来源: jd_self_inventory_sales_collector.py。SKU 维度原始记录，weekly_style_metrics.jd_self_sales 是按款汇总后的值。PK = (week_end, style_name, sku)。';
+COMMENT ON COLUMN jd_self_weekly_sales.week_end IS '周结束日';
+COMMENT ON COLUMN jd_self_weekly_sales.week_start IS '周开始日';
+COMMENT ON COLUMN jd_self_weekly_sales.style_name IS '款式名';
+COMMENT ON COLUMN jd_self_weekly_sales.sku IS '京东自营 SKU';
+COMMENT ON COLUMN jd_self_weekly_sales.outbound_qty IS '本周出库件数';
+COMMENT ON COLUMN jd_self_weekly_sales.found IS 'SKU 是否在采集页找到 (FALSE 时 outbound_qty=0 且未必真为 0)';
+COMMENT ON COLUMN jd_self_weekly_sales.collected_at IS '原始采集时间戳';
+COMMENT ON COLUMN jd_self_weekly_sales.updated_at IS '最近更新时间';
+
+-- decline_events
+COMMENT ON TABLE decline_events IS '9 天连续下滑事件 — 来源: check_nine_day_decline.py。把窗口分 3 段 (3+3+3) 比销量，seg1 > seg2 > seg3 > 0 (3 段严格连续下降且末段非 0) 时触发。PK = (end_date, style_name, platform)。';
+COMMENT ON COLUMN decline_events.detected_at IS '事件被检出的时间';
+COMMENT ON COLUMN decline_events.end_date IS '窗口右端日期 (PK)';
+COMMENT ON COLUMN decline_events.window_days IS '窗口总天数 (默认 9)';
+COMMENT ON COLUMN decline_events.style_name IS '款式名 (PK)';
+COMMENT ON COLUMN decline_events.platform IS '平台代码 (PK)';
+COMMENT ON COLUMN decline_events.window_start IS '窗口左端日期 (= end_date - window_days + 1)';
+COMMENT ON COLUMN decline_events.seg1_sales IS '第 1 段 (最早 3 天) 销量';
+COMMENT ON COLUMN decline_events.seg2_sales IS '第 2 段 (中间 3 天) 销量';
+COMMENT ON COLUMN decline_events.seg3_sales IS '第 3 段 (最近 3 天) 销量';
+COMMENT ON COLUMN decline_events.drop_pct_1to2 IS '段 1→段 2 跌幅 (%)';
+COMMENT ON COLUMN decline_events.drop_pct_2to3 IS '段 2→段 3 跌幅 (%)';
+COMMENT ON COLUMN decline_events.drop_pct_1to3 IS '段 1→段 3 总跌幅 (%)';
+COMMENT ON COLUMN decline_events.erp_style_codes IS '涉及的 ERP 编码列表';
+COMMENT ON COLUMN decline_events.hermes_notified IS '是否已推送 Hermes analyzer';
+COMMENT ON COLUMN decline_events.hermes_response IS 'Hermes 原始响应文本';
+
+-- fact_alert_history
+COMMENT ON TABLE fact_alert_history IS '9 天告警发送历史 — 从 decline_events 派生，每次发送 = 1 行。decline_event_key 是业务键 ''YYYY-MM-DD|款|plat''，因为 decline_events 没有 id 列，应用层 join。UNIQUE (style_name, trigger_date) 避免重复发送。';
+COMMENT ON COLUMN fact_alert_history.id IS '自增主键';
+COMMENT ON COLUMN fact_alert_history.decline_event_key IS '业务键 ''YYYY-MM-DD|款|plat''，对应 decline_events(end_date, style_name, platform)';
+COMMENT ON COLUMN fact_alert_history.style_name IS '款式名';
+COMMENT ON COLUMN fact_alert_history.trigger_date IS '触发告警的统计日期 (= decline_events.end_date)';
+COMMENT ON COLUMN fact_alert_history.consecutive_days IS '连续下滑天数 (默认 9)';
+COMMENT ON COLUMN fact_alert_history.sales_drop_pct IS '总跌幅 1→3 (NUMERIC(6,2)，单位 %)';
+COMMENT ON COLUMN fact_alert_history.decline_window_start IS '下滑窗口起点 (= decline_events.window_start)';
+COMMENT ON COLUMN fact_alert_history.decline_window_end IS '下滑窗口终点';
+COMMENT ON COLUMN fact_alert_history.baseline_window_start IS '对比基期起点';
+COMMENT ON COLUMN fact_alert_history.baseline_window_end IS '对比基期终点';
+COMMENT ON COLUMN fact_alert_history.sent_at IS '发送时间';
+COMMENT ON COLUMN fact_alert_history.recipient IS '接收者 chat_id / open_id';
+COMMENT ON COLUMN fact_alert_history.send_channel IS '发送渠道 (默认 ''lark_im_group'')';
+COMMENT ON COLUMN fact_alert_history.hermes_message_id IS 'Hermes 返回的消息 ID';
+COMMENT ON COLUMN fact_alert_history.hermes_response IS 'Hermes 完整响应 (JSONB)';
+COMMENT ON COLUMN fact_alert_history.decline_payload IS '触发时的 9 天 daily 快照 (JSONB)';
+COMMENT ON COLUMN fact_alert_history.created_at IS '记录创建时间';
+
+-- fact_platform_xlsx
+COMMENT ON TABLE fact_platform_xlsx IS '平台 XLSX 明细 (JD/DY/TM 合一表) — 通用列存共有字段 (平台/日期/行号/商品ID/名称/货号/类目/合计标记)，平台特有列全部塞 payload JSONB。来源: data/<plat>单品数据/_downloads/<date>/*.xlsx。UNIQUE (platform, stat_date, source_file, row_index) → 同款同日重抓不冲突。';
+COMMENT ON COLUMN fact_platform_xlsx.id IS '自增主键';
+COMMENT ON COLUMN fact_platform_xlsx.platform IS '平台代码 (''jd'' / ''dy'' / ''tm'')';
+COMMENT ON COLUMN fact_platform_xlsx.stat_date IS '文件统计日期 (xlsx filename 中的日期)';
+COMMENT ON COLUMN fact_platform_xlsx.row_index IS '数据行号 (1 = 合计行 / 数据起点)';
+COMMENT ON COLUMN fact_platform_xlsx.item_id IS '平台商品 ID (京东 = 商品ID, 抖音 = product_id, 天猫 = 商品ID)';
+COMMENT ON COLUMN fact_platform_xlsx.item_name IS '商品名称';
+COMMENT ON COLUMN fact_platform_xlsx.brand_model IS '货号 (jd/tm 有，dy 留空)';
+COMMENT ON COLUMN fact_platform_xlsx.category_l1 IS '平台一级类目';
+COMMENT ON COLUMN fact_platform_xlsx.category_l2 IS '平台二级类目';
+COMMENT ON COLUMN fact_platform_xlsx.is_total_row IS '合计行标记 (TRUE = 整列汇总行)';
+COMMENT ON COLUMN fact_platform_xlsx.payload IS '全量动态列 (含原始字符串/数字)，平台特有指标都在这里';
+COMMENT ON COLUMN fact_platform_xlsx.source_file IS '相对路径 data/.../<filename>';
+COMMENT ON COLUMN fact_platform_xlsx.source_file_mtime IS '源文件最后修改时间';
+COMMENT ON COLUMN fact_platform_xlsx.source_file_size IS '源文件字节数';
+COMMENT ON COLUMN fact_platform_xlsx.imported_at IS '入库时间';
+
+-- ============================================================================
+-- product_daily_metrics: 三平台单品日维度全量数据 (通用列 + raw_data jsonb)
+-- 数据源:
+--   dy/data/dy单品数据/_downloads/<date>/经营版_商品_商品列表__<date>_数据更新时间<date>.xlsx (22 列)
+--   jd/data/jd单品数据/_downloads/<date>/<shop>_<date>_全部渠道_商品明细_分天下载.xlsx (34 列)
+--   tm/data/tm单品数据/_downloads/<date>/【生意参谋平台】商品_全部_<date>_<date>.xls (38 列, 表头在 r4)
+-- 设计: 三平台含义相同的核心指标抽成独立列, 平台特有字段全部进 raw_data JSONB。
+-- 写入策略: ON CONFLICT (platform, stat_date, product_id) DO UPDATE 全量覆盖。
+-- ============================================================================
+CREATE TABLE IF NOT EXISTS product_daily_metrics (
+    id                      BIGSERIAL PRIMARY KEY,
+    platform                TEXT        NOT NULL CHECK (platform IN ('dy','jd','tm')),
+    stat_date               DATE        NOT NULL,
+    shop_id                 TEXT,                                         -- jd 从文件名抽;dy/tm 留空
+    product_id              TEXT        NOT NULL,                        -- 三平台 ID 格式不同,统一 text
+    product_name            TEXT,
+    product_no              TEXT,                                         -- 货号 (jd/tm 有, dy 无)
+
+    -- 类目 (jd 有 二级/三级, tm 有 商品类型, dy 只有载体/品类两个 filter 维度)
+    category_l1             TEXT,                                         -- jd 二级类目
+    category_l2             TEXT,                                         -- jd 三级类目 / tm 商品类型 / dy 载体
+
+    -- 流量
+    visitors                INTEGER,                                      -- dy 商品点击人数 / jd 访客数 / tm 商品访客数
+    page_views              INTEGER,                                      -- dy 商品曝光人数 / jd 浏览量 / tm 商品浏览量
+
+    -- 加购
+    add_cart_qty            INTEGER,                                      -- 加购件数
+    add_cart_buyers         INTEGER,                                      -- 加购人数
+
+    -- 销售 (实付口径)
+    --   dy: 成交金额 / 成交订单数 / 成交人数
+    --   jd: 成交金额 / 成交商品件数 / 成交客户数
+    --   tm: 支付金额 / 支付件数   / 支付买家数
+    paid_amount             NUMERIC(18, 2),
+    paid_qty                INTEGER,
+    paid_buyers             INTEGER,
+
+    -- 退款
+    refund_amount           NUMERIC(18, 2),                               -- dy 退款金额(支付时间) / jd 取消及售后退款金额 / tm 成功退款金额
+    refund_orders           INTEGER,                                      -- dy 退款订单数(支付时间) / jd 取消及售后退款件数 / tm 无
+
+    -- 转化与价值
+    conversion_rate         NUMERIC(10, 6),                               -- dy 点击支付率(人数) / jd 成交转化率 / tm 商品支付转化率
+    avg_order_value         NUMERIC(18, 2),                               -- 客单价 / 成交客单价
+    uv_value                NUMERIC(18, 4),                               -- jd UV 价值 / tm 访客平均价值
+    detail_page_bounce_rate NUMERIC(10, 6),                               -- 商详页跳出率
+
+    -- 平台特有字段全量保留
+    raw_data                JSONB        NOT NULL DEFAULT '{}'::jsonb,
+
+    -- 溯源
+    source_file             TEXT,                                         -- 原始文件名
+    imported_at             TIMESTAMPTZ  NOT NULL DEFAULT NOW(),
+
+    UNIQUE (platform, stat_date, product_id)
+);
+CREATE INDEX IF NOT EXISTS product_daily_metrics_platform_date_idx
+    ON product_daily_metrics (platform, stat_date DESC);
+CREATE INDEX IF NOT EXISTS product_daily_metrics_date_idx
+    ON product_daily_metrics (stat_date DESC);
+CREATE INDEX IF NOT EXISTS product_daily_metrics_product_idx
+    ON product_daily_metrics (platform, product_id);
+CREATE INDEX IF NOT EXISTS product_daily_metrics_raw_data_gin
+    ON product_daily_metrics USING GIN (raw_data);
+
+COMMENT ON TABLE product_daily_metrics IS '三平台单品日维度全量数据 — 把含义相同的核心指标抽成独立列, 平台特有字段全量进 raw_data JSONB。PK = (platform, stat_date, product_id), 重跑同一天全量覆盖。';
+COMMENT ON COLUMN product_daily_metrics.platform IS '平台代码 (''dy'' / ''jd'' / ''tm'')';
+COMMENT ON COLUMN product_daily_metrics.stat_date IS '统计日期';
+COMMENT ON COLUMN product_daily_metrics.shop_id IS '店铺 ID (jd 从文件名抽,dy/tm 留空)';
+COMMENT ON COLUMN product_daily_metrics.product_id IS '平台商品 ID (三平台格式不同,统一 text)';
+COMMENT ON COLUMN product_daily_metrics.product_name IS '商品名称';
+COMMENT ON COLUMN product_daily_metrics.product_no IS '货号 (jd/tm 有, dy 无)';
+COMMENT ON COLUMN product_daily_metrics.category_l1 IS '一级类目 (jd 二级类目)';
+COMMENT ON COLUMN product_daily_metrics.category_l2 IS '二级类目 (jd 三级类目 / tm 商品类型)';
+COMMENT ON COLUMN product_daily_metrics.visitors IS '访客数 (dy 商品点击人数 / jd 访客数 / tm 商品访客数)';
+COMMENT ON COLUMN product_daily_metrics.page_views IS '浏览量 (dy 商品曝光人数 / jd 浏览量 / tm 商品浏览量)';
+COMMENT ON COLUMN product_daily_metrics.add_cart_qty IS '加购件数';
+COMMENT ON COLUMN product_daily_metrics.add_cart_buyers IS '加购人数';
+COMMENT ON COLUMN product_daily_metrics.paid_amount IS '实付金额 (dy/jd 成交金额, tm 支付金额)';
+COMMENT ON COLUMN product_daily_metrics.paid_qty IS '实付件数 (dy 成交订单数 / jd 成交商品件数 / tm 支付件数)';
+COMMENT ON COLUMN product_daily_metrics.paid_buyers IS '实付买家数 (dy 成交人数 / jd 成交客户数 / tm 支付买家数)';
+COMMENT ON COLUMN product_daily_metrics.refund_amount IS '退款金额 (dy 退款金额(支付时间) / jd 取消及售后退款金额 / tm 成功退款金额)';
+COMMENT ON COLUMN product_daily_metrics.refund_orders IS '退款件数 (dy 退款订单数(支付时间) / jd 取消及售后退款件数, tm 无)';
+COMMENT ON COLUMN product_daily_metrics.conversion_rate IS '支付转化率 (dy 点击支付率(人数) / jd 成交转化率 / tm 商品支付转化率)';
+COMMENT ON COLUMN product_daily_metrics.avg_order_value IS '客单价 (dy 成交客单价 / jd 客单价)';
+COMMENT ON COLUMN product_daily_metrics.uv_value IS 'UV 价值 (jd UV 价值 / tm 访客平均价值, dy 无)';
+COMMENT ON COLUMN product_daily_metrics.detail_page_bounce_rate IS '详情页跳出率';
+COMMENT ON COLUMN product_daily_metrics.raw_data IS '平台特有原始列 (下单金额/件数/买家数、聚划算、竞争力评分、月累计、搜索引导...) 原样保留';
+COMMENT ON COLUMN product_daily_metrics.source_file IS '原始文件名 (溯源用)';
+COMMENT ON COLUMN product_daily_metrics.imported_at IS '入库时间';
+
+-- ============================================================================
+-- persona_screenshots: 人物画像截图 (周任务, 天猫分组)
+-- 数据源: taobao_dmp_item_crowd_insight_screenshots.py →
+--         data/tm人物画像/<采集日期>/<款式>/<款式>_<商品ID>_crowd_insight.png
+-- 每周采集一次, 每款只取第一张截图 (按商品ID排序第一个)。
+-- 同款同采集日期幂等覆盖 (ON CONFLICT)。
+-- ============================================================================
+CREATE TABLE IF NOT EXISTS persona_screenshots (
+    collect_date     DATE        NOT NULL,                 -- 采集日期 (data/tm人物画像/<date>/)
+    style_name       TEXT        NOT NULL,                 -- 款式名
+    platform         TEXT        NOT NULL DEFAULT 'tm',    -- 固定 'tm' (只有天猫分组)
+    item_id          TEXT,                                -- 对应的天猫商品ID (第一张截图的商品ID)
+    screenshot_path  TEXT        NOT NULL,                 -- 本地截图相对路径 (data/tm人物画像/...)
+    image_data       BYTEA,                               -- 截图二进制 (PNG)
+    file_size        BIGINT,                              -- 文件字节大小
+    bitable_base     TEXT,                                -- 飞书表 base_token (跟销量表同 Base)
+    bitable_table    TEXT,                                -- 飞书表 table_id
+    bitable_record_id TEXT,                              -- 飞书记录 ID (上传附件后填)
+    bitable_file_token TEXT,                              -- 飞书附件 file_token (上传成功后填)
+    uploaded_to_bitable BOOLEAN NOT NULL DEFAULT FALSE,   -- 是否已推飞书
+    collected_at     TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    updated_at       TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    PRIMARY KEY (collect_date, style_name, platform)
+);
+CREATE INDEX IF NOT EXISTS persona_screenshots_date_idx
+    ON persona_screenshots (collect_date DESC);
+CREATE INDEX IF NOT EXISTS persona_screenshots_style_idx
+    ON persona_screenshots (style_name, collect_date DESC);
+
+-- 兼容已有库: persona_screenshots 加 image_data 列 (CREATE TABLE IF NOT EXISTS 不会改已有表)
+ALTER TABLE persona_screenshots ADD COLUMN IF NOT EXISTS image_data BYTEA;
+
+COMMENT ON TABLE persona_screenshots IS '人物画像截图 — 周任务, 天猫分组。来源: taobao_dmp_item_crowd_insight_screenshots.py → data/tm人物画像/<date>/<款>/<款>_<商品ID>_crowd_insight.png。每款只取第一张。PK = (collect_date, style_name, platform)。';
+COMMENT ON COLUMN persona_screenshots.collect_date IS '采集日期 (对应 data/tm人物画像/<date>/ 目录名)';
+COMMENT ON COLUMN persona_screenshots.style_name IS '款式名 (PK)';
+COMMENT ON COLUMN persona_screenshots.platform IS '平台代码 (固定 ''tm'', 只有天猫分组采集人物画像)';
+COMMENT ON COLUMN persona_screenshots.item_id IS '第一张截图对应的天猫商品ID';
+COMMENT ON COLUMN persona_screenshots.screenshot_path IS '本地截图相对路径 (相对项目根目录, 溯源用)';
+COMMENT ON COLUMN persona_screenshots.image_data IS '截图二进制 (PNG), bytea 直接存储, 可脱离本地文件系统';
+COMMENT ON COLUMN persona_screenshots.file_size IS '截图文件字节大小';
+COMMENT ON COLUMN persona_screenshots.bitable_base IS '飞书多维表格 base_token (跟销量表同 Base)';
+COMMENT ON COLUMN persona_screenshots.bitable_table IS '飞书多维表格 table_id';
+COMMENT ON COLUMN persona_screenshots.bitable_record_id IS '飞书记录 ID (上传附件后回填)';
+COMMENT ON COLUMN persona_screenshots.bitable_file_token IS '飞书附件 file_token (上传成功后回填)';
+COMMENT ON COLUMN persona_screenshots.uploaded_to_bitable IS '是否已推送到飞书多维表格人群画像字段';
+COMMENT ON COLUMN persona_screenshots.collected_at IS '采集时间 (脚本写入时间)';
+COMMENT ON COLUMN persona_screenshots.updated_at IS '最近更新时间';
+
+-- ============================================================================
+-- daily_persona_metrics: 三平台人群画像指标 (日任务, TM/DY/JD)
+-- 数据源:
+--   TM: collect_persona_to_bitable.py → data/tm人物画像/<date>/<款>/*_chart_values.json
+--   DY: collect_dy_persona_to_bitable.py → data/dy人物画像/<date>/<款>/*_profile.json
+--   JD: collect_jd_persona_to_bitable.py → data/jd人物画像/<date>/<款>/*_profile.json
+-- payload: 归一化后数据, 值统一为百分比(0-100), 结构:
+--   {"gender":[{"name":"男性","value":76.56}], "age":[...],
+--    "city_tier":[...], "buying_power":[...], "strategy_crowd":[...], "province":[...]}
+-- raw_payload: 原始采集数据 (未转换, 保留平台原始格式)
+-- 同款同日同平台幂等覆盖 (ON CONFLICT)。
+-- ============================================================================
+CREATE TABLE IF NOT EXISTS daily_persona_metrics (
+    collect_date   DATE        NOT NULL,
+    style_name     TEXT        NOT NULL,
+    platform       TEXT        NOT NULL CHECK (platform IN ('tm', 'dy', 'jd')),
+    item_id        TEXT,
+    payload        JSONB       NOT NULL DEFAULT '{}'::jsonb,
+    raw_payload    JSONB,
+    source_file    TEXT,
+    collected_at   TIMESTAMPTZ,
+    updated_at     TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    PRIMARY KEY (collect_date, style_name, platform)
+);
+CREATE INDEX IF NOT EXISTS persona_metrics_date_idx
+    ON daily_persona_metrics (collect_date DESC);
+CREATE INDEX IF NOT EXISTS persona_metrics_style_idx
+    ON daily_persona_metrics (style_name, collect_date DESC);
+CREATE INDEX IF NOT EXISTS persona_metrics_payload_gin
+    ON daily_persona_metrics USING GIN (payload);
+
+COMMENT ON TABLE daily_persona_metrics IS '三平台人群画像指标 — 日任务。TM/DY/JD 采集后归一化入库, payload 值统一为百分比(0-100)。PK = (collect_date, style_name, platform)。';
+COMMENT ON COLUMN daily_persona_metrics.collect_date IS '采集日期 (对应 data/{tm,dy,jd}人物画像/<date>/ 目录名)';
+COMMENT ON COLUMN daily_persona_metrics.style_name IS '款式名 (PK)';
+COMMENT ON COLUMN daily_persona_metrics.platform IS '平台代码: tm=天猫, dy=抖音, jd=京东';
+COMMENT ON COLUMN daily_persona_metrics.item_id IS '平台商品ID (TM item_id / DY product_id / JD SPU)';
+COMMENT ON COLUMN daily_persona_metrics.payload IS '归一化后人群画像数据 (JSONB), 值统一为百分比0-100。维度: gender/age/city_tier/buying_power/strategy_crowd/province';
+COMMENT ON COLUMN daily_persona_metrics.raw_payload IS '原始采集数据 (JSONB, 保留平台原始格式, 用于审计和重处理)';
+COMMENT ON COLUMN daily_persona_metrics.source_file IS '源JSON文件相对路径 (溯源用)';
+COMMENT ON COLUMN daily_persona_metrics.collected_at IS '采集时间 (源JSON中的 captured_at/collected_at)';
+COMMENT ON COLUMN daily_persona_metrics.updated_at IS '最近更新时间';
+
+-- ============================================================================
+-- main_image_creatives: 万相台主图创意指标 (周任务, 天猫分组)
+-- 数据源: run_weekly_main_image.py →
+--         data/主图信息/<采集日期>/<款式>/wanxiang_creative_main_images.json
+-- 每周采集一次, 每款多个创意图 (按 image_key 去重合并)。
+-- 同款同采集日期同 image_key 幂等覆盖 (ON CONFLICT)。
+-- ============================================================================
+CREATE TABLE IF NOT EXISTS main_image_creatives (
+    collect_date     DATE        NOT NULL,                 -- 采集日期 (data/主图信息/<date>/)
+    style_name       TEXT        NOT NULL,                 -- 款式名
+    platform         TEXT        NOT NULL DEFAULT 'tm',    -- 固定 'tm' (只有天猫分组)
+    image_key        TEXT        NOT NULL,                 -- 图片 key (文件名, 用于去重)
+    creative_ids     TEXT[],                               -- 关联的创意 ID 列表 (可能多个)
+    creative_image_url TEXT,                               -- 淘宝 CDN 图片 URL
+    impressions      INTEGER NOT NULL DEFAULT 0,           -- 曝光量
+    clicks           INTEGER NOT NULL DEFAULT 0,           -- 点击量
+    cost             NUMERIC(18, 2),                       -- 花费 (元)
+    click_rate       NUMERIC(10, 4),                       -- 点击率 (%)
+    row_count        SMALLINT NOT NULL DEFAULT 0,          -- 合并的行数 (同图多创意)
+    local_image_path TEXT,                                -- 本地图片相对路径
+    image_downloaded BOOLEAN NOT NULL DEFAULT FALSE,       -- 图片是否已下载
+    image_fingerprint TEXT,                               -- 图片指纹 (去重用)
+    bitable_base     TEXT,                                -- 飞书主图表 base_token
+    bitable_table    TEXT,                                -- 飞书主图表 table_id
+    bitable_record_id TEXT,                              -- 飞书记录 ID (上传后填)
+    bitable_file_token TEXT,                              -- 飞书附件 file_token
+    uploaded_to_bitable BOOLEAN NOT NULL DEFAULT FALSE,   -- 是否已推飞书
+    collected_at     TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    updated_at       TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    PRIMARY KEY (collect_date, style_name, image_key)
+);
+CREATE INDEX IF NOT EXISTS main_image_creatives_date_idx
+    ON main_image_creatives (collect_date DESC);
+CREATE INDEX IF NOT EXISTS main_image_creatives_style_idx
+    ON main_image_creatives (style_name, collect_date DESC);
+
+COMMENT ON TABLE main_image_creatives IS '万相台主图创意指标 — 周任务, 天猫分组。来源: run_weekly_main_image.py → data/主图信息/<date>/<款>/wanxiang_creative_main_images.json。每款多个创意图按 image_key 去重。PK = (collect_date, style_name, image_key)。';
+COMMENT ON COLUMN main_image_creatives.collect_date IS '采集日期 (对应 data/主图信息/<date>/ 目录名)';
+COMMENT ON COLUMN main_image_creatives.style_name IS '款式名 (PK)';
+COMMENT ON COLUMN main_image_creatives.platform IS '平台代码 (固定 ''tm'', 只有天猫分组)';
+COMMENT ON COLUMN main_image_creatives.image_key IS '图片 key (文件名, 同款不同创意图按此去重)';
+COMMENT ON COLUMN main_image_creatives.creative_ids IS '关联的创意 ID 列表 (同图可能对应多个创意)';
+COMMENT ON COLUMN main_image_creatives.creative_image_url IS '淘宝 CDN 图片 URL';
+COMMENT ON COLUMN main_image_creatives.impressions IS '曝光量';
+COMMENT ON COLUMN main_image_creatives.clicks IS '点击量';
+COMMENT ON COLUMN main_image_creatives.cost IS '花费 (元)';
+COMMENT ON COLUMN main_image_creatives.click_rate IS '点击率 (%)';
+COMMENT ON COLUMN main_image_creatives.row_count IS '合并的行数 (同图多创意合并后的行数)';
+COMMENT ON COLUMN main_image_creatives.local_image_path IS '本地图片相对路径 (相对项目根目录)';
+COMMENT ON COLUMN main_image_creatives.image_downloaded IS '图片是否已下载到本地';
+COMMENT ON COLUMN main_image_creatives.image_fingerprint IS '图片指纹 (感知哈希, 去重用)';
+COMMENT ON COLUMN main_image_creatives.bitable_base IS '飞书主图表 base_token';
+COMMENT ON COLUMN main_image_creatives.bitable_table IS '飞书主图表 table_id';
+COMMENT ON COLUMN main_image_creatives.bitable_record_id IS '飞书记录 ID (上传附件后回填)';
+COMMENT ON COLUMN main_image_creatives.bitable_file_token IS '飞书附件 file_token (上传成功后回填)';
+COMMENT ON COLUMN main_image_creatives.uploaded_to_bitable IS '是否已推送到飞书多维表格主图字段';
+COMMENT ON COLUMN main_image_creatives.collected_at IS '采集时间 (脚本写入时间)';
+COMMENT ON COLUMN main_image_creatives.updated_at IS '最近更新时间';
+
+-- ============================================================================
+-- product_reviews: 三平台商品评价 (JD + DY, TM 暂未接入)
+-- 数据源:
+--   JD: jd_product_data_collector.py --reviews-only -> data/jd单品数据/评价汇总_<date>.json
+--   DY: dy_product_scraping.py (主流程已挂载)     -> data/dy单品数据/评论汇总_<date>.json
+-- 设计: 三平台字段归一到独立列, 平台特有字段进 raw_data JSONB。
+-- 写入策略: ON CONFLICT (platform, order_id) DO UPDATE 全量覆盖 (order_id 是评价业务键)。
+-- ============================================================================
+CREATE TABLE IF NOT EXISTS product_reviews (
+    id                      BIGSERIAL PRIMARY KEY,
+    platform                TEXT        NOT NULL CHECK (platform IN ('jd','dy')),
+    collect_date            DATE        NOT NULL,                 -- 采集日期 (脚本运行日)
+    comment_date            DATE,                                 -- 评价时间 (JD 取评价时间日期部分, DY 取 comment_time 日期部分)
+    style_name              TEXT,                                 -- 款式名 (DY: 通过 product_id 反查 dim_style.dy_spus 填充; 未匹配时为 NULL)
+    rating                  TEXT        NOT NULL,                 -- '好评' / '中评' / '差评'
+    score                   SMALLINT,                             -- JD 评分 (1-5 星), DY 留空
+    content                 TEXT,
+    tags                    TEXT[],                               -- JD 评价标签 (做工精细/容量大 等), DY 留空
+    user_name               TEXT,
+    order_id                TEXT        NOT NULL,                 -- 订单 ID (业务键, 用于去重)
+    product_id              TEXT,                                 -- 平台商品 ID (DY product_id / JD sku_name 对应的 SPU)
+    product_name            TEXT,
+    sku_id                  TEXT,                                 -- JD SKU 编码, DY 留空
+    sku_name                TEXT,                                 -- JD SKU 名称, DY 留空
+    sku_size                TEXT,                                 -- JD 规格, DY 留空
+    raw_data                JSONB       NOT NULL DEFAULT '{}'::jsonb,
+    source_file             TEXT,                                 -- 原始 JSON 文件路径
+    imported_at             TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    UNIQUE (platform, order_id)
+);
+CREATE INDEX IF NOT EXISTS product_reviews_platform_date_idx
+    ON product_reviews (platform, comment_date DESC);
+CREATE INDEX IF NOT EXISTS product_reviews_rating_idx
+    ON product_reviews (platform, rating, comment_date DESC);
+CREATE INDEX IF NOT EXISTS product_reviews_product_idx
+    ON product_reviews (platform, product_id, comment_date DESC);
+-- product_reviews_style_idx 在文件末尾 ALTER TABLE 加 style_name 列之后建 (老表兼容)
+
+COMMENT ON TABLE product_reviews IS '商品评价 - JD + DY 合一表。来源: jd_product_data_collector.py --reviews-only / dy_product_scraping.py。UNIQUE (platform, order_id) -> 同订单评价重抓覆盖。';
+COMMENT ON COLUMN product_reviews.platform IS '平台代码 (''jd'' / ''dy'')';
+COMMENT ON COLUMN product_reviews.collect_date IS '采集日期 (脚本运行日, 不等于评价发生日)';
+COMMENT ON COLUMN product_reviews.comment_date IS '评价时间 (从评价发生时间戳抽日期部分)';
+-- style_name 列的 COMMENT 在文件末尾 ALTER TABLE 加列之后 (老表兼容)
+COMMENT ON COLUMN product_reviews.rating IS '评价等级: 好评 / 中评 / 差评';
+COMMENT ON COLUMN product_reviews.score IS 'JD 评分 (1-5 星, aria-valuenow), DY 留空';
+COMMENT ON COLUMN product_reviews.content IS '评价正文';
+COMMENT ON COLUMN product_reviews.tags IS 'JD 评价标签数组 (如 [做工精细, 容量大]), DY 留空';
+COMMENT ON COLUMN product_reviews.user_name IS '用户昵称';
+COMMENT ON COLUMN product_reviews.order_id IS '订单 ID (业务键, 同订单只保留一条, 重抓覆盖)';
+COMMENT ON COLUMN product_reviews.product_id IS '平台商品 ID (DY product_id / JD SPU)';
+COMMENT ON COLUMN product_reviews.product_name IS '商品名称';
+COMMENT ON COLUMN product_reviews.sku_id IS 'JD SKU 编码, DY 留空';
+COMMENT ON COLUMN product_reviews.sku_name IS 'JD SKU 名称, DY 留空';
+COMMENT ON COLUMN product_reviews.sku_size IS 'JD 规格文本, DY 留空';
+COMMENT ON COLUMN product_reviews.raw_data IS '平台特有原始字段 (JD 完整评价对象 / DY comment_rows 行)';
+COMMENT ON COLUMN product_reviews.source_file IS '源 JSON 文件相对路径 (溯源用)';
+COMMENT ON COLUMN product_reviews.imported_at IS '入库时间';
+
+-- 兼容已有库: product_reviews 加 style_name 列 (CREATE TABLE IF NOT EXISTS 不会改已有表)
+ALTER TABLE product_reviews ADD COLUMN IF NOT EXISTS style_name TEXT;
+CREATE INDEX IF NOT EXISTS product_reviews_style_idx
+    ON product_reviews (style_name, comment_date DESC)
+    WHERE style_name IS NOT NULL;
+COMMENT ON COLUMN product_reviews.style_name IS '款式名 (DY: 通过 product_id 反查 dim_style.dy_spus 填充; 未匹配时为 NULL, 表示 dim_style 缺该商品 ID 维护)';
+
+-- ============================================================================
+-- fact_style_analysis_reports: 单款式多维度分析报告 (Hermes analyzer 产出)
+-- 数据源: analyze_style_with_hermes.py -> Hermes chat completions -> 飞书文档
+-- 窗口默认 7 天, 与 daily_style_metrics 对齐. 同款同 analysis_date 同 window_days 重跑覆盖.
+-- 写入: analyze_style_with_hermes.py (手动或定时)
+-- 读取: 飞书看板 / 周复盘 / 报告归档查询
+-- ============================================================================
+CREATE TABLE IF NOT EXISTS fact_style_analysis_reports (
+    id                  BIGSERIAL    PRIMARY KEY,
+    style_name          TEXT         NOT NULL,
+    analysis_date       DATE         NOT NULL,           -- 窗口右端日 (= end_date)
+    window_days         SMALLINT     NOT NULL DEFAULT 7,
+    window_start        DATE         NOT NULL,
+    window_end          DATE         NOT NULL,           -- 冗余 = analysis_date, 方便范围查询
+    platforms_included  TEXT[]       NOT NULL DEFAULT '{}', -- ['jd','dy','tm'] 实际有数据的平台
+    -- 飞书文档
+    feishu_doc_id       TEXT,                            -- lark-cli docs +create 返回的 document_id
+    feishu_doc_url      TEXT,                            -- 文档可访问 URL
+    report_title        TEXT,                            -- 文档标题 (同飞书 doc title)
+    report_markdown     TEXT         NOT NULL,           -- Hermes 输出的 Markdown 全文
+    -- Hermes 元数据
+    hermes_model        TEXT         NOT NULL,           -- 'mimo-v2.5-pro'
+    prompt_tokens       INTEGER,                         -- usage.prompt_tokens
+    completion_tokens   INTEGER,                         -- usage.completion_tokens
+    -- 源数据快照 (审计 / 重跑用)
+    source_payload      JSONB,                           -- 拼 prompt 时用的全部 PG 数据
+    -- 时间戳
+    generated_at        TIMESTAMPTZ  NOT NULL DEFAULT NOW(),
+    updated_at          TIMESTAMPTZ  NOT NULL DEFAULT NOW(),
+    UNIQUE (style_name, analysis_date, window_days)
+);
+CREATE INDEX IF NOT EXISTS fact_style_analysis_reports_style_idx
+    ON fact_style_analysis_reports (style_name, analysis_date DESC);
+CREATE INDEX IF NOT EXISTS fact_style_analysis_reports_date_idx
+    ON fact_style_analysis_reports (analysis_date DESC);
+
+COMMENT ON TABLE fact_style_analysis_reports IS '单款式多维度分析报告 - Hermes analyzer 产出后归档. PK = BIGSERIAL id, 业务 UNIQUE = (style_name, analysis_date, window_days) 同款同日同窗口重跑覆盖. 飞书文档 + Markdown 全文都留底, 飞书删了也能从 PG 还原.';
+COMMENT ON COLUMN fact_style_analysis_reports.style_name IS '款式名 (与 dim_style.style_name 对齐)';
+COMMENT ON COLUMN fact_style_analysis_reports.analysis_date IS '窗口右端日 (默认昨天, 即数据最新一天)';
+COMMENT ON COLUMN fact_style_analysis_reports.window_days IS '窗口天数 (默认 3, 含 analysis_date)';
+COMMENT ON COLUMN fact_style_analysis_reports.window_start IS '窗口左端 (= analysis_date - window_days + 1)';
+COMMENT ON COLUMN fact_style_analysis_reports.window_end IS '窗口右端 (冗余 = analysis_date, 方便范围查询)';
+COMMENT ON COLUMN fact_style_analysis_reports.platforms_included IS '本次实际拉到数据的平台数组 (jd/dy/tm 子集)';
+COMMENT ON COLUMN fact_style_analysis_reports.feishu_doc_id IS '飞书文档 document_id (lark-cli docs +create 返回)';
+COMMENT ON COLUMN fact_style_analysis_reports.feishu_doc_url IS '飞书文档可访问 URL';
+COMMENT ON COLUMN fact_style_analysis_reports.report_title IS '文档标题 (同飞书 doc title)';
+COMMENT ON COLUMN fact_style_analysis_reports.report_markdown IS 'Hermes 输出的 Markdown 全文, 飞书删了也能从这里还原';
+COMMENT ON COLUMN fact_style_analysis_reports.hermes_model IS 'Hermes analyzer 模型名 (mimo-v2.5-pro)';
+COMMENT ON COLUMN fact_style_analysis_reports.prompt_tokens IS 'Hermes 返回的 prompt_tokens (源数据大小参考)';
+COMMENT ON COLUMN fact_style_analysis_reports.completion_tokens IS 'Hermes 返回的 completion_tokens (报告大小参考)';
+COMMENT ON COLUMN fact_style_analysis_reports.source_payload IS '拼 prompt 用的全部 PG 数据 JSON (daily/weekly/reviews/persona/main_image)';
+COMMENT ON COLUMN fact_style_analysis_reports.generated_at IS '报告生成时间 (Hermes 调用时刻)';
+COMMENT ON COLUMN fact_style_analysis_reports.updated_at IS '记录更新时间 (重跑覆盖时刷新)';
+
+-- ============================================================================
+-- fact_market_rank_reports: 三平台市场排行飞书文档归档
+-- 数据源: collect_sycm_market_rank.py / collect_jd_market_rank.py /
+--         collect_dy_market_rank.py
+-- 同平台、同日无论执行多少次都覆盖，保留最后一次成功生成的飞书文档链接。
+-- ============================================================================
+CREATE TABLE IF NOT EXISTS fact_market_rank_reports (
+    id                  BIGSERIAL    PRIMARY KEY,
+    platform            TEXT         NOT NULL CHECK (platform IN ('tm', 'jd', 'dy')),
+    report_date         DATE         NOT NULL,
+    period              TEXT         NOT NULL,
+    category_scope      TEXT         NOT NULL,
+    category_names      TEXT[]       NOT NULL DEFAULT '{}',
+    report_title        TEXT         NOT NULL,
+    feishu_doc_id       TEXT,
+    feishu_doc_url      TEXT         NOT NULL,
+    product_count       INTEGER      NOT NULL DEFAULT 0 CHECK (product_count >= 0),
+    source_json_path    TEXT,
+    run_metadata        JSONB,
+    generated_at        TIMESTAMPTZ  NOT NULL DEFAULT NOW(),
+    updated_at          TIMESTAMPTZ  NOT NULL DEFAULT NOW(),
+    UNIQUE (platform, report_date)
+);
+
+-- 兼容旧表：旧唯一键包含 period/category_scope，同平台同日可能产生多条。
+-- 先保留更新时间最新的一条，再替换为“平台 + 日期”唯一约束。
+ALTER TABLE fact_market_rank_reports
+    DROP CONSTRAINT IF EXISTS fact_market_rank_reports_platform_report_date_period_catego_key;
+
+WITH ranked_market_rank_reports AS (
+    SELECT id,
+           ROW_NUMBER() OVER (
+               PARTITION BY platform, report_date
+               ORDER BY updated_at DESC, generated_at DESC, id DESC
+           ) AS row_num
+      FROM fact_market_rank_reports
+)
+DELETE FROM fact_market_rank_reports target
+ USING ranked_market_rank_reports ranked
+ WHERE target.id = ranked.id
+   AND ranked.row_num > 1;
+
+DO $$
+BEGIN
+    IF NOT EXISTS (
+        SELECT 1
+          FROM pg_constraint
+         WHERE conrelid = 'fact_market_rank_reports'::regclass
+           AND conname = 'fact_market_rank_reports_platform_report_date_key'
+    ) THEN
+        ALTER TABLE fact_market_rank_reports
+            ADD CONSTRAINT fact_market_rank_reports_platform_report_date_key
+            UNIQUE (platform, report_date);
+    END IF;
+END
+$$;
+CREATE INDEX IF NOT EXISTS fact_market_rank_reports_date_idx
+    ON fact_market_rank_reports (report_date DESC);
+CREATE INDEX IF NOT EXISTS fact_market_rank_reports_platform_idx
+    ON fact_market_rank_reports (platform, report_date DESC);
+
+COMMENT ON TABLE fact_market_rank_reports IS '天猫、京东、抖音市场排行采集生成的飞书文档归档；每个平台每天仅保留最后一次成功生成的文档。';
+COMMENT ON COLUMN fact_market_rank_reports.platform IS '平台代码：tm / jd / dy';
+COMMENT ON COLUMN fact_market_rank_reports.report_date IS '采集并生成文档的业务日期';
+COMMENT ON COLUMN fact_market_rank_reports.period IS '平台统计周期原文，如 30天 / 近30天';
+COMMENT ON COLUMN fact_market_rank_reports.category_scope IS '文档范围，如 九品类 / 七品类 / 单品类名';
+COMMENT ON COLUMN fact_market_rank_reports.category_names IS '本次实际包含的逻辑品类名称';
+COMMENT ON COLUMN fact_market_rank_reports.feishu_doc_id IS '从飞书 /docx/{token} 链接提取的文档 token';
+COMMENT ON COLUMN fact_market_rank_reports.feishu_doc_url IS '飞书文档完整访问链接';
+COMMENT ON COLUMN fact_market_rank_reports.product_count IS '文档内采集商品数';
+COMMENT ON COLUMN fact_market_rank_reports.source_json_path IS '本次采集 JSON 产物路径';
+COMMENT ON COLUMN fact_market_rank_reports.run_metadata IS '价格、搜索词、分页等运行参数 JSON';
+
+-- ============================================================================
+-- fact_market_rank_notifications: 三平台市场排行链接的 Hermes 通知审计
+-- 同一收件人、同一报告日只保留一条；sent 状态用于阻止重跑重复发送。
+-- ============================================================================
+CREATE TABLE IF NOT EXISTS fact_market_rank_notifications (
+    id                  BIGSERIAL    PRIMARY KEY,
+    report_date         DATE         NOT NULL,
+    recipient_open_id   TEXT         NOT NULL,
+    recipient_name      TEXT,
+    platform_links      JSONB        NOT NULL DEFAULT '{}'::jsonb,
+    status              TEXT         NOT NULL
+                        CHECK (status IN ('incomplete', 'sent', 'failed')),
+    hermes_message_id   TEXT,
+    hermes_response     JSONB,
+    sent_at             TIMESTAMPTZ,
+    created_at          TIMESTAMPTZ  NOT NULL DEFAULT NOW(),
+    updated_at          TIMESTAMPTZ  NOT NULL DEFAULT NOW(),
+    UNIQUE (report_date, recipient_open_id)
+);
+
+CREATE INDEX IF NOT EXISTS fact_market_rank_notifications_sent_idx
+    ON fact_market_rank_notifications (sent_at DESC);
+
+COMMENT ON TABLE fact_market_rank_notifications IS '三平台市场排行飞书文档链接经 Hermes 分析端发送后的审计和幂等记录。';
+COMMENT ON COLUMN fact_market_rank_notifications.platform_links IS '本次通知包含的 tm/jd/dy 飞书文档链接 JSON。';
+COMMENT ON COLUMN fact_market_rank_notifications.status IS 'incomplete=缺平台链接，sent=发送成功，failed=Hermes 调用失败。';
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/db/sync_dim_style.py b/src/gyxx_flow/modules/product_commerce/runtime/db/sync_dim_style.py
new file mode 100644
index 0000000..6825670
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/db/sync_dim_style.py
@@ -0,0 +1,342 @@
+"""同步 dim_style：4 个数据源自动 upsert。
+
+数据源:
+  1) bitable_style_map.json    -> bitable_weekly_url/base/table
+  2) jd_self_inventory_sales_collector.STYLE_SKU_GROUPS -> jd_self_skus
+  3) data/<plat>/<款>_<最近日期>/<款>_<最近日期>.json -> jd_spus / dy_spus / tm_spus
+  4) daily_style_metrics.erp_style_codes -> erp_codes
+
+SKU 主表的飞书地址 (bitable_sku_master_*) 由人工维护，不在自动同步范围。
+
+用法:
+    .venv/Scripts/python.exe db/sync_dim_style.py                # 全量
+    .venv/Scripts/python.exe db/sync_dim_style.py --dry-run      # 扫描不入库
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import sys
+from datetime import datetime
+from pathlib import Path
+from typing import Any, Iterable
+
+PROJECT_ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(PROJECT_ROOT))
+from runtime_paths import RAW_DATA_ROOT, runtime_script
+
+from db import get_conn  # noqa: E402
+
+
+PLATFORM_DIRS = {
+    "jd": "jd单品数据",
+    "dy": "dy单品数据",
+    "tm": "tm单品数据",
+}
+
+
+# ── 数据源 1: bitable_style_map.json ────────────────────────────────────────
+
+def load_bitable_weekly() -> dict[str, dict[str, str]]:
+    """从 bitable_style_map.json 抓每个款式的飞书周报表 base/table。"""
+    path = PROJECT_ROOT / "bitable_style_map.json"
+    if not path.exists():
+        return {}
+    raw = json.loads(path.read_text(encoding="utf-8-sig"))
+    out: dict[str, dict[str, str]] = {}
+    for style_name, info in raw.items():
+        if not isinstance(info, dict) or not info.get("enabled", True):
+            continue
+        base = info.get("base_token", "")
+        table = info.get("table_id", "")
+        if not base or not table:
+            continue
+        url = f"https://feishu.cn/base/{base}?table={table}"
+        out[style_name] = {
+            "bitable_weekly_base": base,
+            "bitable_weekly_table": table,
+            "bitable_weekly_url": url,
+        }
+    return out
+
+
+# ── 数据源 2: jd_self SKU 集合 ──────────────────────────────────────────────
+
+def load_jd_self_skus() -> dict[str, list[str]]:
+    """从 jd_self_inventory_sales_collector.STYLE_SKU_GROUPS 抓款→SKU 集合。
+
+    直接解析源文件里 STYLE_SKU_GROUPS = [...] 的字面量，避免 import 拉起 playwright。
+    """
+    path = runtime_script("jd_self_inventory_sales_collector.py")
+    text = path.read_text(encoding="utf-8")
+    needle = "STYLE_SKU_GROUPS = "
+    i = text.find(needle)
+    if i < 0:
+        return {}
+    i += len(needle)
+    # 平衡括号找匹配的 ]
+    depth = 0
+    in_str: str | None = None
+    escape = False
+    end = None
+    for j in range(i, len(text)):
+        ch = text[j]
+        if in_str:
+            if escape:
+                escape = False
+            elif ch == "\\":
+                escape = True
+            elif ch == in_str:
+                in_str = None
+            continue
+        if ch in ("'", '"'):
+            in_str = ch
+        elif ch == "[":
+            depth += 1
+        elif ch == "]":
+            depth -= 1
+            if depth == 0:
+                end = j + 1
+                break
+    if end is None:
+        return {}
+    chunk = text[i:end]
+    try:
+        # 用 ast.literal_eval 解析 Python 字面量（兼容末尾逗号 / 单引号）
+        import ast
+        groups = ast.literal_eval(chunk)
+    except Exception as exc:
+        print(f"  [WARN] 解析 STYLE_SKU_GROUPS 失败: {exc}")
+        return {}
+    return {g["style_name"]: list(g.get("skus") or []) for g in groups if "style_name" in g}
+
+
+# ── 数据源 3: 各平台 SPU 集合 ───────────────────────────────────────────────
+
+def _latest_style_json(plat_dir: Path, style_name: str) -> Path | None:
+    """取某款式在 plat_dir 下的最新日期 JSON。
+
+    目录名约定: <款名>_<YYYY-MM-DD> (jd/dy/tm 主流) 或 业务款式_<YYYY-MM-DD> (tm 部分)
+
+    优先级:
+      1) <safe>_<date>.json       (jd / dy / tm 主流)
+      2) 业务款式_<date>.json     (tm 习惯命名)
+      3) data.json                (兜底，但 data.json 通常是 ERP 抓数，可能没有 SPU 字段)
+    """
+    safe = "".join("_" if ch in '\\/:*?"<>|' else ch for ch in style_name).strip(" ._")
+    # 只匹配以 "<safe>_" 开头的目录，避免拿到别的款式的目录
+    cands: list[tuple[str, Path]] = []
+    for d in plat_dir.iterdir():
+        if not d.is_dir() or not d.name.startswith(safe + "_"):
+            continue
+        date_part = d.name[len(safe) + 1:]
+        try:
+            datetime.strptime(date_part, "%Y-%m-%d")
+        except ValueError:
+            continue
+        cands.append((date_part, d))
+    if not cands:
+        return None
+    cands.sort(key=lambda x: x[0], reverse=True)
+    latest_date, latest_dir = cands[0]
+
+    # 优先级 1: <safe>_<date>.json
+    p1 = latest_dir / f"{safe}_{latest_date}.json"
+    if p1.exists():
+        return p1
+    # 优先级 2: 业务款式_<date>.json (tm)
+    p2 = latest_dir / f"业务款式_{latest_date}.json"
+    if p2.exists():
+        return p2
+    # 优先级 3: data.json
+    p3 = latest_dir / "data.json"
+    if p3.exists():
+        return p3
+    return None
+
+
+def _extract_spus(payload: dict, platform: str) -> list[str]:
+    """从款式 JSON payload 提取 SPU/商品 ID 列表。
+
+    各平台字段名:
+      jd: payload['匹配SPU'] 或 payload['spu列表']
+      dy: payload['matched_product_ids'] 或 payload['product_ids']
+      tm: payload['style']['匹配商品ID'] 或 payload['style']['SPU']
+    """
+    if platform == "jd":
+        return list(payload.get("匹配SPU") or payload.get("spu列表") or [])
+    if platform == "dy":
+        return list(payload.get("matched_product_ids") or payload.get("product_ids") or [])
+    if platform == "tm":
+        style = payload.get("style") or {}
+        return list(style.get("匹配商品ID") or style.get("SPU") or [])
+    return []
+
+
+def load_platform_spus() -> dict[str, dict[str, list[str]]]:
+    """返回 {style_name: {platform: [spu, ...]}}，跨日期取最近一次抓数。"""
+    out: dict[str, dict[str, list[str]]] = {}
+    # 先枚举所有款式名（从 weekly_style_metrics 拿权威款式列表）
+    data_root = RAW_DATA_ROOT
+    # 用 daily_style_metrics 拿所有款式名
+    try:
+        with get_conn() as conn:
+            with conn.cursor() as cur:
+                cur.execute("SELECT DISTINCT style_name FROM daily_style_metrics")
+                style_names = [r[0] for r in cur.fetchall()]
+    except Exception:
+        style_names = []
+
+    for style_name in style_names:
+        out.setdefault(style_name, {})
+        for platform, dir_name in PLATFORM_DIRS.items():
+            plat_dir = data_root / dir_name
+            if not plat_dir.exists():
+                continue
+            json_path = _latest_style_json(plat_dir, style_name)
+            if not json_path:
+                continue
+            try:
+                payload = json.loads(json_path.read_text(encoding="utf-8-sig"))
+            except Exception:
+                continue
+            spus = _extract_spus(payload, platform)
+            if spus:
+                out[style_name][f"{platform}_spus"] = spus
+    return out
+
+
+# ── 数据源 4: ERP 编码（PG 聚合）──────────────────────────────────────────
+
+def load_erp_codes() -> dict[str, list[str]]:
+    """从 daily_style_metrics 聚每个款式的 ERP 编码。"""
+    out: dict[str, list[str]] = {}
+    try:
+        with get_conn() as conn:
+            with conn.cursor() as cur:
+                cur.execute("""
+                    SELECT style_name, ARRAY(
+                        SELECT DISTINCT unnest(erp_style_codes)
+                        FROM daily_style_metrics s
+                        WHERE s.style_name = dm.style_name
+                          AND s.erp_style_codes IS NOT NULL
+                    ) AS codes
+                    FROM (SELECT DISTINCT style_name FROM daily_style_metrics) dm
+                """)
+                for style_name, codes in cur.fetchall():
+                    out[style_name] = [c for c in (codes or []) if c]
+    except Exception as exc:
+        print(f"  [WARN] 读 daily_style_metrics 失败: {exc}")
+    return out
+
+
+# ── 合并 + upsert ──────────────────────────────────────────────────────────
+
+def _merge_records(
+    style_names: Iterable[str],
+    weekly: dict[str, dict],
+    self_skus: dict[str, list[str]],
+    spus_map: dict[str, dict],
+    erp_map: dict[str, list[str]],
+) -> list[dict]:
+    rows: list[dict] = []
+    for name in style_names:
+        rows.append({
+            "style_name": name,
+            "jd_spus": spus_map.get(name, {}).get("jd_spus", []) or [],
+            "dy_spus": spus_map.get(name, {}).get("dy_spus", []) or [],
+            "tm_spus": spus_map.get(name, {}).get("tm_spus", []) or [],
+            "jd_self_skus": self_skus.get(name, []) or [],
+            "erp_codes": erp_map.get(name, []) or [],
+            "bitable_weekly_url": weekly.get(name, {}).get("bitable_weekly_url", ""),
+            "bitable_weekly_base": weekly.get(name, {}).get("bitable_weekly_base", ""),
+            "bitable_weekly_table": weekly.get(name, {}).get("bitable_weekly_table", ""),
+            # bitable_sku_master_* 由人工维护，不动
+        })
+    return rows
+
+
+def upsert_dim_style(conn, records: list[dict]) -> int:
+    if not records:
+        return 0
+    def arr(v):
+        # 空 list / None 都写空数组 {}（PG 数组维度不空），非空 list 原样写
+        if v is None:
+            return []
+        return list(v)
+    payload = [
+        (
+            r["style_name"],
+            arr(r.get("jd_spus")),
+            arr(r.get("dy_spus")),
+            arr(r.get("tm_spus")),
+            arr(r.get("jd_self_skus")),
+            arr(r.get("erp_codes")),
+            r.get("bitable_weekly_url") or None,
+            r.get("bitable_weekly_base") or None,
+            r.get("bitable_weekly_table") or None,
+        )
+        for r in records
+    ]
+    sql = """
+        INSERT INTO dim_style
+            (style_name, jd_spus, dy_spus, tm_spus, jd_self_skus, erp_codes,
+             bitable_weekly_url, bitable_weekly_base, bitable_weekly_table)
+        VALUES %s
+        ON CONFLICT (style_name) DO UPDATE SET
+            jd_spus              = EXCLUDED.jd_spus,
+            dy_spus              = EXCLUDED.dy_spus,
+            tm_spus              = EXCLUDED.tm_spus,
+            jd_self_skus         = EXCLUDED.jd_self_skus,
+            erp_codes            = EXCLUDED.erp_codes,
+            bitable_weekly_url   = EXCLUDED.bitable_weekly_url,
+            bitable_weekly_base  = EXCLUDED.bitable_weekly_base,
+            bitable_weekly_table = EXCLUDED.bitable_weekly_table,
+            -- bitable_sku_master_* 字段：保留人工已填值
+            bitable_sku_master_url   = COALESCE(dim_style.bitable_sku_master_url, EXCLUDED.bitable_sku_master_url),
+            bitable_sku_master_base  = COALESCE(dim_style.bitable_sku_master_base, EXCLUDED.bitable_sku_master_base),
+            bitable_sku_master_table = COALESCE(dim_style.bitable_sku_master_table, EXCLUDED.bitable_sku_master_table),
+            updated_at = NOW()
+    """
+    from psycopg2.extras import execute_values
+    with conn.cursor() as cur:
+        execute_values(cur, sql, payload)
+    return len(payload)
+
+
+def main():
+    ap = argparse.ArgumentParser(description="同步 dim_style (款式 SPU/SKU + 飞书表地址)")
+    ap.add_argument("--dry-run", action="store_true", help="扫描不入库")
+    args = ap.parse_args()
+
+    weekly = load_bitable_weekly()
+    self_skus = load_jd_self_skus()
+    spus_map = load_platform_spus()
+    erp_map = load_erp_codes()
+
+    # 款式名全集：合并 4 个数据源
+    style_names: set[str] = set()
+    style_names.update(weekly.keys())
+    style_names.update(self_skus.keys())
+    style_names.update(spus_map.keys())
+    style_names.update(erp_map.keys())
+
+    records = _merge_records(sorted(style_names), weekly, self_skus, spus_map, erp_map)
+
+    print(f"[INFO] 数据源: weekly={len(weekly)} self_skus={len(self_skus)} spus={len(spus_map)} erp={len(erp_map)}")
+    print(f"[INFO] 合并款式: {len(records)}")
+    if args.dry_run:
+        for r in records[:5]:
+            print(f"  - {r['style_name']}: jd_spus={r['jd_spus']} jd_self_skus={r['jd_self_skus']} weekly_url={r['bitable_weekly_url'][:60]}")
+        return 0
+
+    with get_conn() as conn:
+        n = upsert_dim_style(conn, records)
+    print(f"[OK] dim_style upsert: {n} 行")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/db/sync_sku_master.py b/src/gyxx_flow/modules/product_commerce/runtime/db/sync_sku_master.py
new file mode 100644
index 0000000..746d5e1
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/db/sync_sku_master.py
@@ -0,0 +1,210 @@
+"""同步飞书 SKU 主表 → dim_style。
+
+通过 lark-cli 拉 SKU 主表全部记录，按款式聚合 4 个平台 ID 集合 + ERP 编码 + 飞书链接，
+upsert 到 dim_style 的 jd_spus / dy_spus / tm_spus / jd_self_skus / erp_codes / bitable_sku_master_* 字段。
+
+用法:
+    .venv/Scripts/python.exe db/sync_sku_master.py            # 全量拉 + 入库
+    .venv/Scripts/python.exe db/sync_sku_master.py --dry-run  # 拉数据 + 显示不入库
+
+依赖: lark-cli (npm 全局安装, 路径在 PATH 或 lark-cli.cmd)
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import subprocess
+import sys
+from collections import defaultdict
+from pathlib import Path
+
+PROJECT_ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+from db import get_conn  # noqa: E402
+
+
+SKU_MASTER = {
+    "base_token": "TtoCb1NuQaDy3NsZWTpc0GIvnph",
+    "table_id": "tblKCjplVAFrRwMC",
+    "view_id": "vewvy33xEk",
+}
+
+# 字段在 record 数组中的位置 (按 lark 返回顺序)
+F_PLAT = 0      # ["天猫"]  (select 多选)
+F_ERP = 1       # "10384" 或 "10416,10455" (逗号分隔多编码)
+F_NAME = 2      # "盖亚斜挎"
+F_IDS = 3       # "921336775006,..."  (逗号分隔多 ID)
+F_NOTE = 4      # 注释
+F_LINK = 5      # "[...](https://...base/xxx?table=yyy&view=zzz)"
+
+
+def fetch_sku_master_records() -> list[list]:
+    """调 lark-cli 拉 SKU 主表全部记录。"""
+    cmd = [
+        "lark-cli", "base", "+record-list",
+        "--base-token", SKU_MASTER["base_token"],
+        "--table-id", SKU_MASTER["table_id"],
+        "--view-id", SKU_MASTER["view_id"],
+        "--as", "user",
+        "--limit", "500",
+        "--format", "json",
+    ]
+    # 优先用 lark-cli 解析（PATH 或 npm 全局）
+    candidates = [
+        "lark-cli",
+    ]
+    proc = None
+    for c in candidates:
+        try:
+            proc = subprocess.run([c] + cmd[1:], capture_output=True, text=True, encoding="utf-8")
+            if proc.returncode == 0:
+                break
+        except FileNotFoundError:
+            continue
+    if proc is None or proc.returncode != 0:
+        raise RuntimeError(
+            f"lark-cli 不可用，stderr={proc.stderr if proc else 'not found'}"
+        )
+    raw = json.loads(proc.stdout)
+    return raw["data"]["data"]
+
+
+def _split_ids(value) -> list[str]:
+    """飞书表格里「款式ID」字段，jd/dy/tm 用逗号分隔，京东自营用换行分隔。"""
+    if value is None:
+        return []
+    s = str(value)
+    # 先按换行 split（京东自营风格），再按逗号 split（其它平台风格）
+    parts: list[str] = []
+    for line in s.splitlines():
+        for piece in line.split(","):
+            piece = piece.strip()
+            if piece:
+                parts.append(piece)
+    return parts
+
+
+def aggregate(records: list[list]) -> dict[str, dict]:
+    """按款式聚合: jd_spus / dy_spus / tm_spus / jd_self_skus / erp_codes / sku_master_link。"""
+    by_style: dict[str, dict] = defaultdict(lambda: {
+        "jd_spus": [], "dy_spus": [], "tm_spus": [], "jd_self_skus": [],
+        "erp_codes": [], "sku_master_url": "", "sku_master_base": "",
+        "sku_master_table": "",
+    })
+    for r in records:
+        plat_field = r[F_PLAT] if len(r) > F_PLAT else None
+        plat_list = plat_field if isinstance(plat_field, list) else ([plat_field] if plat_field else [])
+        plat_s = ",".join(str(p) for p in plat_list)
+        name = str(r[F_NAME] if len(r) > F_NAME and r[F_NAME] else "").strip()
+        if not name:
+            continue
+        ids_field = r[F_IDS] if len(r) > F_IDS else None
+        id_list = _split_ids(ids_field)
+        erp_field = r[F_ERP] if len(r) > F_ERP else None
+        erp_list = _split_ids(erp_field)
+        link_field = str(r[F_LINK] if len(r) > F_LINK and r[F_LINK] else "")
+
+        if "京东自营" in plat_s:
+            by_style[name]["jd_self_skus"].extend(id_list)
+        elif "京东" in plat_s:
+            by_style[name]["jd_spus"].extend(id_list)
+        if "抖音" in plat_s:
+            by_style[name]["dy_spus"].extend(id_list)
+        if "天猫" in plat_s:
+            by_style[name]["tm_spus"].extend(id_list)
+        for ec in erp_list:
+            if ec not in by_style[name]["erp_codes"]:
+                by_style[name]["erp_codes"].append(ec)
+
+        if link_field and not by_style[name]["sku_master_url"]:
+            # 解析 [text](https://.../base/XXXXX?table=YYYYY[&view=ZZZZ]) -> 提纯 URL
+            m = re.search(
+                r"\((https?://[^/]+/base/([A-Za-z0-9]+)\?[^)]*table=([A-Za-z0-9]+)[^)]*)\)",
+                link_field,
+            )
+            if m:
+                by_style[name]["sku_master_url"] = m.group(1)
+                by_style[name]["sku_master_base"] = m.group(2)
+                by_style[name]["sku_master_table"] = m.group(3)
+            else:
+                # fallback: 整段存
+                by_style[name]["sku_master_url"] = link_field
+    return by_style
+
+
+def upsert_from_sku_master(conn, by_style: dict[str, dict]) -> int:
+    """upsert SKU 主表数据到 dim_style (只更新 SPU/SKU/ERP/sku_master_* 字段，其它保留)。"""
+    if not by_style:
+        return 0
+    from psycopg2.extras import execute_values
+
+    def arr(v):
+        if v is None:
+            return []
+        return list(v)
+
+    payload = [
+        (
+            name,
+            arr(s.get("jd_spus")),
+            arr(s.get("dy_spus")),
+            arr(s.get("tm_spus")),
+            arr(s.get("jd_self_skus")),
+            arr(s.get("erp_codes")),
+            s.get("sku_master_url") or None,
+            s.get("sku_master_base") or None,
+            s.get("sku_master_table") or None,
+        )
+        for name, s in by_style.items()
+    ]
+    sql = """
+        INSERT INTO dim_style
+            (style_name, jd_spus, dy_spus, tm_spus, jd_self_skus, erp_codes,
+             bitable_sku_master_url, bitable_sku_master_base, bitable_sku_master_table)
+        VALUES %s
+        ON CONFLICT (style_name) DO UPDATE SET
+            jd_spus              = EXCLUDED.jd_spus,
+            dy_spus              = EXCLUDED.dy_spus,
+            tm_spus              = EXCLUDED.tm_spus,
+            jd_self_skus         = EXCLUDED.jd_self_skus,
+            erp_codes            = EXCLUDED.erp_codes,
+            bitable_sku_master_url   = EXCLUDED.bitable_sku_master_url,
+            bitable_sku_master_base  = EXCLUDED.bitable_sku_master_base,
+            bitable_sku_master_table = EXCLUDED.bitable_sku_master_table,
+            updated_at = NOW()
+    """
+    with conn.cursor() as cur:
+        execute_values(cur, sql, payload)
+    return len(payload)
+
+
+def main():
+    ap = argparse.ArgumentParser(description="同步飞书 SKU 主表 → dim_style")
+    ap.add_argument("--dry-run", action="store_true")
+    args = ap.parse_args()
+
+    print("[STEP] 拉飞书 SKU 主表 ...")
+    records = fetch_sku_master_records()
+    print(f"[INFO] 拉取 {len(records)} 条记录")
+    by_style = aggregate(records)
+    print(f"[INFO] 聚合 {len(by_style)} 个款式")
+
+    for name in sorted(by_style.keys())[:5]:
+        s = by_style[name]
+        print(f"  - {name}: jd={s['jd_spus']} dy={s['dy_spus']} tm={s['tm_spus']} jd_self={s['jd_self_skus']} erp={s['erp_codes']}")
+
+    if args.dry_run:
+        print("[DRY] 不入库")
+        return 0
+
+    with get_conn() as conn:
+        n = upsert_from_sku_master(conn, by_style)
+    print(f"[OK] dim_style upsert: {n} 行")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/dy_audience_profile_collect.py b/src/gyxx_flow/modules/product_commerce/runtime/dy_audience_profile_collect.py
new file mode 100644
index 0000000..2bd5ac6
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/dy_audience_profile_collect.py
@@ -0,0 +1,475 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""抖音罗盘成交人群画像采集脚本。
+
+一次登录, 逐款式搜单品 -> 选成交人群 -> 点击下载结果 -> 解析 Excel,
+每款产出 JSON 到 data/dy人物画像/{date}/{style}/。
+下载的 Excel 保存到 data/dy人群画像下载结果/{date}/。
+
+被 collect_dy_persona_to_bitable.py 以 subprocess 调用,
+与 taobao_dmp_item_crowd_insight_screenshots.py 定位相同。
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import sys
+from datetime import datetime, timedelta
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT, vendor_source_root
+from typing import Any
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+# dy-data-flow 依赖
+DY_VENDOR_ROOT = vendor_source_root("dy")
+if str(DY_VENDOR_ROOT) not in sys.path:
+    sys.path.append(str(DY_VENDOR_ROOT))
+
+from scrapling import DynamicFetcher
+
+from adaptive_selectors import enable_adaptive_fetchers
+from config.style_config_loader import (
+    StyleConfigLoader,
+    get_dy_persona_bitable_map,
+    get_dy_product_groups,
+)
+from dy_product_scraping import (
+    _cleanup_product_detail_pages,
+    _click_download_text,
+    _click_product_detail,
+    _ensure_product_list_page,
+    _alive_context_pages,
+    _open_product_audience_profile,
+    _open_product_list,
+    _search_product_by_id,
+    _select_deal_crowd,
+    _select_product_yesterday,
+    _wait_for_deal_crowd_data,
+)
+from dy_store_competitor_store_scraping import (
+    COOKIES_FILE,
+    SHOP_NAME,
+    USER_DATA_DIR,
+    _active_page,
+    _check_login_status,
+    _ensure_login,
+    _load_cookies,
+    _save_browser_state,
+)
+
+try:
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")
+    sys.stderr.reconfigure(encoding="utf-8", errors="replace")
+except Exception:
+    pass
+
+OUTPUT_ROOT = RAW_DATA_ROOT / "dy人物画像"
+DOWNLOAD_ROOT = RAW_DATA_ROOT / "dy人群画像下载结果"
+PRODUCT_LIST_URL = "https://compass.jinritemai.com/shop/commodity/product-list"
+
+DEFAULT_USER_DATA_DIR = os.getenv("DY_USER_DATA_DIR", USER_DATA_DIR)
+
+
+def time_label() -> str:
+    today = datetime.now().date()
+    return f"{today.month}.{today.day}"
+
+
+def date_stamp() -> str:
+    return (datetime.now().date() - timedelta(days=1)).strftime("%Y-%m-%d")
+
+
+# ── 下载 + 解析 ──────────────────────────────────────────────────────────────
+
+def _download_audience_profile(page, style_name: str, product_id: str) -> str | None:
+    """点击「下载结果」按钮, 保存 Excel 到 data/dy人群画像下载结果/{date}/, 返回路径。"""
+    stamp = date_stamp()
+    download_dir = DOWNLOAD_ROOT / stamp
+    download_dir.mkdir(parents=True, exist_ok=True)
+    safe_style = re.sub(r'[\\/:*?"<>|\r\n\t]+', "_", style_name).strip() or "unknown"
+    target_name = f"{safe_style}_{product_id}.xlsx"
+    target_path = download_dir / target_name
+
+    try:
+        page.keyboard.press("Escape")
+    except Exception:
+        pass
+    page.wait_for_timeout(500)
+
+    try:
+        with page.expect_download(timeout=30000) as download_info:
+            if not _click_download_text(page, ["下载结果", "下载"]):
+                print("[WARN] 未找到「下载结果」按钮")
+                return None
+        download = download_info.value
+        suggested = download.suggested_filename or target_name
+        if not suggested.endswith((".xlsx", ".xls", ".csv")):
+            suggested = target_name
+        save_path = download_dir / re.sub(r'[\\/:*?"<>|\r\n\t]+', "_", suggested)
+        download.save_as(str(save_path))
+        # 统一重命名为 target_name
+        if save_path.name != target_name and not target_path.exists():
+            save_path.rename(target_path)
+            save_path = target_path
+        print(f"[OK] 下载保存: {save_path}")
+        return str(save_path)
+    except Exception as e:
+        print(f"[WARN] 下载失败: {e}")
+        return None
+
+
+# Excel 行关键词 -> distribution 分组
+_SECTION_KEYWORDS: dict[str, str] = {
+    "性别分布": "gender_distribution",
+    "性别": "gender_distribution",
+    "年龄分布": "age_distribution",
+    "年龄": "age_distribution",
+    "策略人群分布": "strategy_crowd_distribution",
+    "策略人群": "strategy_crowd_distribution",
+    "八大策略人群": "strategy_crowd_distribution",
+    "常驻省份分布": "resident_province",
+    "常驻省份": "resident_province",
+    "省份": "resident_province",
+}
+
+# 已知策略人群名称
+_STRATEGY_CROWD_NAMES = {
+    "新锐白领", "资深中产", "小镇青年", "都市蓝领",
+    "小镇中老年", "都市银发", "Z世代", "小镇Z世代",
+    "其他策略人群", "GenZ", "新锐", "资深",
+}
+
+# 已知省份名前缀
+_PROVINCE_PREFIXES = {
+    "北京", "天津", "上海", "重庆", "河北", "山西", "辽宁", "吉林", "黑龙江",
+    "江苏", "浙江", "安徽", "福建", "江西", "山东", "河南", "湖北", "湖南",
+    "广东", "海南", "四川", "贵州", "云南", "陕西", "甘肃", "青海", "台湾",
+    "内蒙", "广西", "西藏", "宁夏", "新疆", "香港", "澳门",
+    "内蒙古", "黑龙江",
+}
+
+
+def _classify_row(row_text: str, current_section: str) -> str:
+    """根据行文本判断该行属于哪个 distribution 分组。"""
+    for keyword, section in _SECTION_KEYWORDS.items():
+        if keyword in row_text:
+            return section
+    # 如果是策略人群名称行
+    for name in _STRATEGY_CROWD_NAMES:
+        if name in row_text:
+            return "strategy_crowd_distribution"
+    # 如果是省份行
+    for prefix in _PROVINCE_PREFIXES:
+        if row_text.startswith(prefix):
+            return "resident_province"
+    return current_section
+
+
+def _parse_audience_excel(path: str) -> dict | None:
+    """解析下载的人群画像 Excel, 返回 profile dict 或 None。"""
+    rows: list[tuple] = []
+    try:
+        from openpyxl import load_workbook
+        workbook = load_workbook(path, data_only=True, read_only=True)
+        try:
+            rows = list(workbook.active.iter_rows(values_only=True))
+        finally:
+            workbook.close()
+    except Exception as exc:
+        print(f"[WARN] openpyxl 读取失败, 回退 pandas: {exc}")
+        try:
+            import pandas as pd
+            frame = pd.read_excel(path, header=None).fillna("")
+            rows = [tuple(row) for row in frame.itertuples(index=False, name=None)]
+        except Exception as exc2:
+            print(f"[ERR] pandas 也读取失败: {exc2}")
+            return None
+
+    if not rows:
+        print("[WARN] Excel 无数据行")
+        return None
+
+    # 找 header 行 (含"占比"或"百分比"或"分布")
+    header_idx = 0
+    for i, row in enumerate(rows[:20]):
+        text = " ".join(str(v or "") for v in row)
+        if "占比" in text or "百分比" in text or "分布" in text:
+            header_idx = i
+            break
+
+    headers = [str(v or "").strip() for v in rows[header_idx]]
+    # 找 name 列和 value 列
+    name_col = None
+    value_col = None
+    for ci, h in enumerate(headers):
+        hl = h.lower()
+        if not name_col and any(k in hl for k in ("名称", "类型", "分类", "人群", "省份", "性别", "年龄")):
+            name_col = ci
+        if not value_col and any(k in hl for k in ("占比", "百分比", "比例", "比例%", "人数占比")):
+            value_col = ci
+    # fallback: 第一列 name, 第二列 value
+    if name_col is None:
+        name_col = 0
+    if value_col is None:
+        value_col = 1 if len(headers) > 1 else 0
+
+    result: dict[str, list[dict]] = {
+        "gender_distribution": [],
+        "age_distribution": [],
+        "strategy_crowd_distribution": [],
+        "resident_province": [],
+    }
+    current_section = ""
+    raw_rows: list[dict] = []
+
+    for row in rows[header_idx + 1:]:
+        if not any(v not in (None, "") for v in row):
+            continue
+        row_text = " ".join(str(v or "") for v in row).strip()
+        raw_rows.append({headers[i]: row[i] if i < len(row) else "" for i in range(len(headers))})
+
+        current_section = _classify_row(row_text, current_section)
+        if not current_section:
+            continue
+
+        name_val = str(row[name_col] if name_col < len(row) else "").strip()
+        value_raw = row[value_col] if value_col < len(row) else None
+
+        if not name_val:
+            continue
+
+        # 解析数值
+        num_val = None
+        if isinstance(value_raw, (int, float)):
+            num_val = float(value_raw)
+        elif value_raw is not None:
+            s = str(value_raw).strip().replace("%", "").replace(",", "")
+            try:
+                num_val = float(s)
+            except ValueError:
+                pass
+
+        if num_val is not None:
+            result.setdefault(current_section, []).append({"name": name_val, "value": num_val})
+
+    result["raw_rows"] = raw_rows
+    result["extraction_method"] = "download_excel"
+
+    has_data = any(result.get(k) for k in ("gender_distribution", "age_distribution",
+                                            "strategy_crowd_distribution", "resident_province"))
+    if not has_data:
+        print(f"[WARN] Excel 解析后无有效分布数据, raw_rows={len(raw_rows)}")
+        return None
+
+    print(
+        f"[OK] Excel 解析: gender={len(result['gender_distribution'])}, "
+        f"age={len(result['age_distribution'])}, "
+        f"strategy={len(result['strategy_crowd_distribution'])}, "
+        f"province={len(result['resident_province'])}"
+    )
+    return result
+
+
+# ── 采集单个款式 ─────────────────────────────────────────────────────────────
+
+def collect_one_style(page, product_ids: list[str], style_name: str = "") -> tuple[dict | None, str]:
+    """按 ID 顺序尝试，首个成功即返回 (profile, used_id)。
+    每个ID: 搜商品 -> 打开详情 -> 打开人群画像 -> 选成交人群 -> 下载结果 -> 解析 Excel。
+    """
+    for product_id in product_ids:
+        print(f"[INFO] 尝试商品ID: {product_id}")
+        pages_before: list = []
+        try:
+            page = _ensure_product_list_page(page, PRODUCT_LIST_URL)
+            pages_before = _alive_context_pages(page)
+            if not _search_product_by_id(page, product_id):
+                print(f"[WARN] 搜索失败: {product_id}")
+                continue
+            if not _click_product_detail(page, product_id):
+                print(f"[WARN] 打开详情失败: {product_id}")
+                continue
+            page = _active_page(page)
+            if not _open_product_audience_profile(page):
+                print(f"[WARN] 打开人群画像失败: {product_id}")
+                continue
+            page = _active_page(page)
+            if not _select_deal_crowd(page):
+                print(f"[WARN] 选成交人群失败: {product_id}")
+                continue
+            _wait_for_deal_crowd_data(page)
+            page = _active_page(page)
+
+            # 下载结果 Excel
+            download_path = _download_audience_profile(page, style_name, product_id)
+            if not download_path:
+                print(f"[WARN] 下载失败，试下一个ID: {product_id}")
+                continue
+
+            # 解析 Excel
+            profile = _parse_audience_excel(download_path)
+            if not profile:
+                print(f"[WARN] Excel 解析为空，试下一个ID: {product_id}")
+                continue
+
+            has_data = any([
+                profile.get("gender_distribution"),
+                profile.get("age_distribution"),
+                profile.get("strategy_crowd_distribution"),
+                profile.get("resident_province"),
+            ])
+            if has_data:
+                print(f"[OK] 采集成功: {product_id}")
+                return profile, product_id
+            print(f"[WARN] 提取为空，试下一个ID: {product_id}")
+        except Exception as e:
+            print(f"[WARN] 采集异常 ({product_id}): {e}")
+        finally:
+            try:
+                page = _cleanup_product_detail_pages(page, PRODUCT_LIST_URL, pages_before)
+            except Exception as cleanup_error:
+                print(f"[WARN] 清理详情页失败: {cleanup_error}")
+    return None, ""
+
+
+# ── 保存 JSON ────────────────────────────────────────────────────────────────
+
+def save_profile(profile: dict, style_name: str, product_id: str) -> str:
+    """保存 profile JSON 到 data/dy人物画像/{date}/{style}/, 返回路径。"""
+    stamp = date_stamp()
+    style_dir = OUTPUT_ROOT / stamp / style_name
+    style_dir.mkdir(parents=True, exist_ok=True)
+    safe_style = re.sub(r'[\\/:*?"<>|\r\n\t]+', "_", style_name).strip() or "unknown"
+    path = style_dir / f"{safe_style}_{product_id}_profile.json"
+    payload = {
+        "style_name": style_name,
+        "product_id": product_id,
+        "captured_at": datetime.now().isoformat(),
+        "profile": profile,
+    }
+    path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+    return str(path)
+
+
+# ── 浏览器自动化 ──────────────────────────────────────────────────────────────
+
+def page_automation(page, persona_map: dict, dy_groups: dict[str, list[str]]) -> None:
+    """一次登录, 逐款式采集人群画像, 每款产出 JSON。"""
+    page = _active_page(page)
+
+    if not _check_login_status(page):
+        logged_page = _ensure_login(page)
+        if logged_page:
+            page = logged_page
+        else:
+            print("[ERR] 登录失败，尽力继续...")
+
+    _save_browser_state(page, "after login")
+
+    _open_product_list(page)
+    _select_product_yesterday(page)
+
+    for idx, (style, cfg) in enumerate(persona_map.items(), 1):
+        print(f"\n=== [{idx}/{len(persona_map)}] {style} ===")
+        product_ids = dy_groups.get(style, [])
+        if not product_ids:
+            print(f"[SKIP] {style}: 无抖音商品ID")
+            continue
+
+        profile, used_id = collect_one_style(page, product_ids, style_name=style)
+        if not profile:
+            print(f"[SKIP] {style}: 全部ID采集失败")
+            continue
+
+        path = save_profile(profile, style, used_id)
+        print(f"[OK] 保存: {path}")
+
+
+# ── main ─────────────────────────────────────────────────────────────────────
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="抖音罗盘成交人群画像采集")
+    parser.add_argument("--styles", default="", help="只跑指定款式, 逗号分隔; 默认全部")
+    parser.add_argument("--user-data-dir", default=DEFAULT_USER_DATA_DIR, help="Chrome profile 根目录")
+    parser.add_argument("--headless", action="store_true", help="无头跑")
+    parser.add_argument("--refresh-config", action="store_true", help="强制从飞书重读款式配置")
+    args = parser.parse_args()
+
+    OUTPUT_ROOT.mkdir(parents=True, exist_ok=True)
+    DOWNLOAD_ROOT.mkdir(parents=True, exist_ok=True)
+
+    if args.refresh_config:
+        StyleConfigLoader().load(force_refresh=True, sync_fallback=True)
+        print("[INFO] --refresh-config: 已重读款式配置")
+
+    persona_map = get_dy_persona_bitable_map()
+    if not persona_map:
+        print("[FAIL] 未读到任何人群画像（抖音）目标表映射")
+        return 2
+
+    selected = [s for s in (s.strip() for s in args.styles.split(",")) if s]
+    if selected:
+        persona_map = {k: v for k, v in persona_map.items() if k in selected}
+
+    dy_groups = get_dy_product_groups()
+
+    print("=" * 60)
+    print("  抖音罗盘成交人群画像采集")
+    print("=" * 60)
+    print(f"  Shop: {SHOP_NAME}")
+    print(f"  Profile: {args.user_data_dir}")
+    print(f"  Cookies: {COOKIES_FILE}")
+    print(f"  输出: {OUTPUT_ROOT}")
+    print(f"  款式数: {len(persona_map)}")
+    print(f"  时间: {time_label()}")
+    print("=" * 60 + "\n")
+
+    enable_adaptive_fetchers()
+    cookies = _load_cookies()
+
+    fetch_kwargs = dict(
+        headless=args.headless,
+        real_chrome=True,
+        cookies=cookies,
+        timeout=120000,
+        wait=2000,
+    )
+
+    def _page_action(page):
+        page_automation(page, persona_map, dy_groups)
+
+    fetch_kwargs["page_action"] = _page_action
+
+    try:
+        try:
+            response = DynamicFetcher.fetch(
+                "https://compass.jinritemai.com/shop/business-part",
+                user_data_dir=args.user_data_dir,
+                **fetch_kwargs,
+            )
+        except Exception as profile_error:
+            print(f"[WARN] 用 profile 启动失败: {profile_error}")
+            print("[INFO] 用 cookies + 内置 Chromium 重试...")
+            fallback_kwargs = dict(fetch_kwargs)
+            fallback_kwargs["real_chrome"] = False
+            response = DynamicFetcher.fetch(
+                "https://compass.jinritemai.com/shop/business-part",
+                **fallback_kwargs,
+            )
+        print(f"\n[OK] Done: HTTP {response.status} | {response.url}")
+    except Exception as e:
+        print(f"\n[ERR] {e}")
+        import traceback
+        traceback.print_exc()
+        return 1
+
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/dy_product_scraping.py b/src/gyxx_flow/modules/product_commerce/runtime/dy_product_scraping.py
new file mode 100644
index 0000000..a9d9e74
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/dy_product_scraping.py
@@ -0,0 +1,3375 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+抖音罗盘商品列表数据抓取脚本。
+
+复用 dy_store_competitor_store_scraping.py 的 cookie/profile 登录流程。
+"""
+
+import argparse
+import csv
+import json
+import math
+import os
+import re
+import sys
+from datetime import datetime, timedelta
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT, vendor_source_root
+
+PROJECT_ROOT = os.path.dirname(os.path.abspath(__file__))
+SOURCE_ROOT = vendor_source_root("dy")
+if str(SOURCE_ROOT) not in sys.path:
+    # 用 append 而不是 insert(0), 让项目根目录的 config/ 包优先于 vendor 的 config.py
+    sys.path.append(str(SOURCE_ROOT))
+
+from scrapling import DynamicFetcher, StealthyFetcher
+
+from adaptive_selectors import adaptive_fill_input, enable_adaptive_fetchers
+from erp_metric_overrides import apply_erp_override, format_count
+from collect_retry_utils import FailedCollector, retry_step
+from config.style_config_loader import StyleConfigLoader, get_daily_styles, get_dy_product_groups
+
+from dy_store_competitor_store_scraping import (
+    COOKIES_FILE,
+    DATA_DIR,
+    SESSION_ROOT,
+    SHOP_NAME,
+    USER_DATA_DIR,
+    _active_page,
+    _check_login_status,
+    _click_calendar_date,
+    _click_visible_text,
+    _date_range_label,
+    _ensure_login,
+    _hover_visible_text,
+    _keep_browser_open,
+    _load_cookies,
+    _looks_like_login_url,
+    _previous_week_range,
+    _save_browser_state,
+    _wait_after_action,
+)
+
+try:
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")
+    sys.stderr.reconfigure(encoding="utf-8", errors="replace")
+except Exception:
+    pass
+
+PRODUCTS = []
+STYLE_SUMMARIES = []
+COMMENTS_SUMMARY = {}
+_results_saved = False
+_PAGE_ACTION_ERROR = None
+# 抖音采集是 batch 任务（每天 08:40 自动跑），完成后必须立即退出，
+# 否则编排器 / 定时任务会被卡住。命令行 --keep-browser-open 用于手工调试。
+KEEP_BROWSER_OPEN_CLI = False
+# --reviews-only: 登录后直接跳评论页采好/中/差评, 跳过单品列表/下载/人群画像.
+# 复用每日采集的 cookie + profile, 与 daily 共享登录态.
+REVIEWS_ONLY_MODE = False
+PRODUCT_DATA_DIR = str(RAW_DATA_ROOT / "dy单品数据")
+PRODUCT_AUDIENCE_LIMIT = int(os.getenv("PRODUCT_AUDIENCE_LIMIT", "0"))
+COMMENT_MAX_PAGES = int(os.getenv("COMMENT_MAX_PAGES", "200"))
+PRODUCT_DEBUG_DUMP = os.getenv("PRODUCT_DEBUG_DUMP", "0") == "1"
+TARGET_DATE = None
+# 失败清单 (per-page / per-product 失败时由 retry_step / 调用方写入, main() 末尾 .write())
+_DY_FAILED: FailedCollector | None = None
+
+
+def _dy_goto_with_retry(page, url: str, label: str = "", **kwargs) -> None:
+    """per-page 重试包装: page.goto 失败 → 重新 goto, 最多 2 次, 仍失败记 _DY_FAILED。"""
+    def _do():
+        page.goto(url, **kwargs)
+    try:
+        retry_step(_do, max_attempts=2, sleep_seconds=3,
+                   label=label or f"DY goto {url[:50]}")
+    except Exception as exc:
+        if _DY_FAILED is not None:
+            _DY_FAILED.add_page(label or f"goto {url[:60]}", f"{type(exc).__name__}: {exc}", attempts=2)
+        raise
+AUDIENCE_API_PAYLOADS = []
+_AUDIENCE_COLLECTOR_PAGES = set()
+COMMENT_DETAIL_URL = (
+    "https://fxg.jinritemai.com/ffa/maftersale/comment"
+    "?btm_ppre=a2427.b76571.c902327.d871297"
+    "&btm_pre=a2427.b76571.c902327.d871297"
+    "&btm_show_id=69de260a-6aa4-42ce-9ca0-7414e5b65364"
+    "&tabtype=1&evaluation_detail_tab=order"
+)
+PRODUCT_LIST_URL = "https://compass.jinritemai.com/shop/commodity/product-list"
+
+# 保留硬编码作为灾备回退；运行时优先从飞书多维表格读取。
+_STATIC_PRODUCT_STYLE_GROUPS = {
+    "盖亚斜挎": [
+        "3758423390084530598",
+        "3546704497957054088",
+        "3723294781737140535",
+        "3605892222299556299",
+        "3614412506279319659",
+    ],
+    "极星pro": [
+        "3723301894270091349",
+        "3705732014197047560",
+        "3774368299438178482",
+        "3723330591068914054",
+        "3679360949275460130",
+    ],
+    "盖亚微单": [
+        "3723306401804714428",
+        "3664497398165669984",
+        "3610734864606824434",
+        "3774359264898974008",
+        "3608647206321604032",
+    ],
+    "盖世m1": [
+        "3701026559747031390",
+        "3723329152103874604",
+        "3754316016256155917",
+    ],
+    "星迹2": [
+        "3723195804173860873",
+        "3801422650698498468",
+        "3800356691338461484",
+        "3665624563645022287",
+        "3774356172614796203",
+    ],
+    "星云2": [
+        "3725242433383039375",
+        "3774363265803616739",
+        "3731617571452092515",
+    ],
+    "宙斯": [
+        "3730219261105406099",
+        "3792873247570264438",
+        "3731145017650184656",
+        "3592716653063984937",
+        "3562996449333647313",
+        "3774172500418756988",
+        "3663380642202287652",
+        "3728487369196830898",
+        "3723297457493377197",
+        "3586918760285641845",
+        "3774174388157546868",
+        "3789772833136181558",
+    ],
+    "阿波罗2": [
+        "3764111277610369503",
+        "3774182389698396336",
+        "3742100377680609285",
+    ],
+    "阿波罗x1": [
+        "3728485299140034862",
+        "3626260611501006526",
+        "3626260839109106918",
+        "3730190905890701314",
+        "3723144876716589485",
+    ],
+    "波塞冬edge": [
+        "3779730985017540981",
+    ],
+    "逐星": [
+        "3692499943458734547",
+        "3723297966522500082",
+        "3770677535256477997",
+    ],
+    "逐星GT": [
+        "3791442727141507531",
+    ],
+    "星云mini": [
+        "3778067305603399962",
+    ],
+    "晨星2": [
+        "3770491999128519634",
+    ],
+    "极星双肩": [
+        "3757136031745441873",
+        "3773588464549888244",
+        "3773580514364096762",
+    ],
+    "凌云air": [
+        "3753046423898292279",
+        "3774190345974318296",
+    ],
+    "觅光": [
+        "3799053750765748658",
+        "3798126273943961843",
+    ],
+    "云卷2": [
+        "3785686614144057909",
+    ],
+    "凌云3": [
+        "3821081325813760272",
+    ],
+    "瑞白": [
+        "3790157724633923739",
+    ],
+    "黑曜石吐司包": [
+        "3803281856141590866",
+    ],
+    "黑曜石托特包": [
+        "3804002356081852742",
+    ],
+    "曜影": [
+        "3814987314862620724",
+    ],
+    "黑曜石电脑包": [
+        "3816632074752884911",
+    ],
+    "拓界2": [
+        "3823529057790853229",
+    ],
+}
+
+
+def _get_product_style_groups():
+    """优先从飞书读取抖音商品ID分组，失败则回退硬编码。"""
+    try:
+        groups = get_dy_product_groups()
+        if groups:
+            return groups
+    except Exception as exc:
+        print(f"[WARN] 动态读取抖音商品ID配置失败，使用硬编码回退: {exc}")
+    return _STATIC_PRODUCT_STYLE_GROUPS
+
+
+PRODUCT_STYLE_GROUPS = _get_product_style_groups()
+
+
+def _print_preflight() -> None:
+    """启动期打印:飞书表格里抖音分组今日待采 / 跳过清单。"""
+    try:
+        daily = get_daily_styles()
+    except Exception as exc:
+        print(f"[WARN] 飞书读取失败, 跳过 preflight: {exc}")
+        return
+    collectable = daily["dy"]["collectable"]
+    skipped = daily["dy"]["skipped"]
+    print(f"[DY] 飞书表格 待采 {len(collectable)} 款, 跳过 {len(skipped)} 款 (款式级缺 erp/sales,或 DY 缺商品链接ID)")
+    for s in skipped:
+        print(f"  - {s['style']}: {s['reason']}")
+
+
+_print_preflight()
+
+
+def _safe_filename(value: str, limit=90) -> str:
+    value = (value or "未命名商品").strip()
+    value = re.sub(r'[\\/:*?"<>|\r\n\t]+', "_", value)
+    value = re.sub(r"\s+", " ", value).strip(" ._")
+    return (value or "未命名商品")[:limit]
+
+
+def _product_list_is_ready(page, timeout=15000) -> bool:
+    page = _active_page(page)
+    if _looks_like_login_url(page.url):
+        return False
+    try:
+        page.wait_for_function(
+            """() => {
+                const text = document.body.innerText || '';
+                return text.includes('商品列表')
+                    && (text.includes('可搜索商品名称/ID')
+                        || text.includes('查看详情')
+                        || text.includes('智能分析')
+                        || text.includes('下载明细'));
+            }""",
+            timeout=timeout,
+        )
+        return not _looks_like_login_url(page.url)
+    except Exception:
+        return False
+
+
+def _open_product_list(page) -> bool:
+    page = _active_page(page)
+    print("[INFO] Opening product list page...")
+    try:
+        # 尝试 URL 带日期参数 (避免 UI 点日期触发反爬). compass SPA 可能接受 startDate/endDate.
+        target_date_str = TARGET_DATE.strftime("%Y-%m-%d") if TARGET_DATE else _product_yesterday().strftime("%Y-%m-%d")
+        product_list_with_date = f"{PRODUCT_LIST_URL}?startDate={target_date_str}&endDate={target_date_str}&dateType=natural"
+        if "commodity/product-list" in page.url:
+            print("[INFO] Already on product list")
+        else:
+            print(f"[INFO] Navigating directly to product list: {product_list_with_date}")
+            # 带 compass 自家 referer, 让 SameSite=Lax 的登录 cookie (sessionid/sid_tt) 被带上
+            # (Playwright goto 默认无 referer, compass 检测到跨站导航会踢登录页)
+            page.goto(product_list_with_date, wait_until="domcontentloaded", timeout=60000,
+                      referer="https://compass.jinritemai.com/shop/business-part")
+        try:
+            page.wait_for_load_state("load", timeout=15000)
+        except Exception as e:
+            print(f"[WARN] Product list load wait skipped: {e}")
+        _wait_after_action(page, "direct product list navigation", settle_ms=2500)
+        _ensure_product_list_tab(page)
+        return _product_list_is_ready(page, timeout=5000)
+    except Exception as e:
+        print(f"[WARN] Direct product list navigation failed: {e}")
+        return False
+
+
+def _recover_product_authentication(page, stage: str):
+    page = _active_page(page)
+    if _product_list_is_ready(page, timeout=3000):
+        return page
+    print(f"[WARN] Douyin authentication required during {stage}")
+    logged_page = _ensure_login(page)
+    if not logged_page:
+        raise RuntimeError(f"Douyin login was not restored during {stage}")
+    logged_page = _active_page(logged_page)
+    return logged_page
+
+
+def _profile_has_cookie_store(profile_dir: str) -> bool:
+    root = Path(profile_dir)
+    candidates = (
+        root / "Default" / "Network" / "Cookies",
+        root / "Default" / "Cookies",
+    )
+    return any(path.is_file() and path.stat().st_size > 4096 for path in candidates)
+
+
+def _ensure_product_list_tab(page) -> None:
+    page = _active_page(page)
+    try:
+        clicked = page.evaluate(
+            """() => {
+                const visible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none' && style.visibility !== 'hidden'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const tabs = Array.from(document.querySelectorAll('div, span, button, a'))
+                    .filter(el => visible(el) && (el.textContent || '').trim() === '商品列表');
+                const tab = tabs.sort((a, b) => a.getBoundingClientRect().top - b.getBoundingClientRect().top)[0];
+                if (tab) {
+                    tab.click();
+                    return true;
+                }
+                return false;
+            }"""
+        )
+        if clicked:
+            _wait_after_action(page, "ensure 商品列表 tab", settle_ms=800)
+    except Exception:
+        pass
+    try:
+        page.wait_for_function(
+            """() => {
+                const text = document.body.innerText || '';
+                return text.includes('商品列表')
+                    && (text.includes('可搜索商品名称/ID') || text.includes('查看详情') || text.includes('智能分析'));
+            }""",
+            timeout=15000,
+        )
+    except Exception:
+        print("[WARN] Product list signal not detected, extending wait (avoid re-navigate to prevent anti-bot redirect)")
+        try:
+            page.wait_for_load_state("networkidle", timeout=15000)
+        except Exception:
+            pass
+        _wait_after_action(page, "extended wait for product list", settle_ms=3000)
+
+
+def _parse_target_date(value: str):
+    try:
+        return datetime.strptime(value, "%Y-%m-%d").date()
+    except ValueError as exc:
+        raise argparse.ArgumentTypeError("target date must be YYYY-MM-DD") from exc
+
+
+def _product_yesterday():
+    return TARGET_DATE or (datetime.now().date() - timedelta(days=1))
+
+
+def _product_yesterday_label() -> str:
+    yesterday = _product_yesterday()
+    return _date_range_label(yesterday, yesterday)
+
+
+def _align_calendar_month(page, target_date) -> None:
+    target_year = target_date.year
+    target_month = target_date.month
+    for _ in range(18):
+        info = page.evaluate(
+            r"""({targetYear, targetMonth}) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none' && style.visibility !== 'hidden' && rect.width > 0 && rect.height > 0;
+                };
+                const roots = Array.from(document.querySelectorAll(
+                    '[class*="calendar"], [class*="picker"], [class*="popover"], [class*="dropdown"], [role="dialog"]'
+                )).filter((el) => visible(el) && /(20\d{2})\s*年\s*(\d{1,2})\s*月/.test(el.textContent || ''));
+                const root = roots[0] || document;
+                const match = (root.textContent || '').match(/(20\d{2})\s*年\s*(\d{1,2})\s*月/);
+                if (!match) return {status: 'no_header'};
+                const currentYear = Number(match[1]);
+                const currentMonth = Number(match[2]);
+                const diff = (targetYear - currentYear) * 12 + (targetMonth - currentMonth);
+                if (diff === 0) return {status: 'aligned', currentYear, currentMonth};
+                const symbols = diff < 0 ? ['‹', '<', '上一月', '上个月'] : ['›', '>', '下一月', '下个月'];
+                const candidates = Array.from(root.querySelectorAll('button, span, div, i, svg'))
+                    .filter(visible)
+                    .filter((el) => {
+                        const text = (el.textContent || '').trim();
+                        const label = `${el.getAttribute('aria-label') || ''} ${el.getAttribute('title') || ''} ${el.className || ''}`;
+                        return symbols.some((item) => text === item || label.includes(item))
+                            || (diff < 0 && /prev|left|back/.test(label))
+                            || (diff > 0 && /next|right/.test(label));
+                    });
+                if (!candidates.length) return {status: 'no_nav', currentYear, currentMonth, diff};
+                const rect = candidates[0].getBoundingClientRect();
+                return {status: 'click', x: rect.left + rect.width / 2, y: rect.top + rect.height / 2, currentYear, currentMonth, diff};
+            }""",
+            {"targetYear": target_year, "targetMonth": target_month},
+        )
+        if not isinstance(info, dict) or info.get("status") == "aligned":
+            return
+        if info.get("status") != "click":
+            print(f"[WARN] Calendar month alignment stopped: {info}")
+            return
+        page.mouse.click(info["x"], info["y"])
+        page.wait_for_timeout(600)
+
+
+def _human_click_text(page, text: str, exact: bool = True, timeout: int = 8000) -> bool:
+    """人类化鼠标点击: 先 move 到按钮坐标 (多步移动), 停留, 再 click.
+    compass 反爬检测 locator.click() 的合成事件; 真实鼠标轨迹能绕过."""
+    import random
+    print(f"[INFO] Human-clicking: {text}")
+    try:
+        box = page.evaluate(
+            """([text, exact]) => {
+                const visible = (el) => {
+                    const s = window.getComputedStyle(el);
+                    const r = el.getBoundingClientRect();
+                    return s.display !== 'none' && s.visibility !== 'hidden' && r.width > 0 && r.height > 0;
+                };
+                const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const candidates = Array.from(document.querySelectorAll('button, [role="button"], a, span, div, li'))
+                    .filter(visible)
+                    .filter(el => {
+                        const v = normalize(el.innerText || el.textContent);
+                        return exact ? v === text : v.includes(text);
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return null;
+                const r = candidates[0].getBoundingClientRect();
+                return { x: r.left + r.width / 2, y: r.top + r.height / 2, w: r.width, h: r.height };
+            }""",
+            [text, exact],
+        )
+    except Exception as e:
+        print(f"[WARN] Human-click locate failed ({text}): {e}")
+        return False
+    if not box:
+        print(f"[WARN] Human-click target not found: {text}")
+        return False
+    try:
+        # 从当前鼠标位置(或随机起点)多步移动到目标, 模拟人类鼠标轨迹
+        page.mouse.move(box["x"], box["y"], steps=random.randint(15, 30))
+        page.wait_for_timeout(random.randint(800, 2000))  # 人类停留
+        page.mouse.click(box["x"], box["y"])
+        page.wait_for_timeout(random.randint(500, 1200))  # 点击后停留
+        print(f"[OK] Human-clicked '{text}' at ({box['x']:.0f}, {box['y']:.0f})")
+        return True
+    except Exception as e:
+        print(f"[WARN] Human-click action failed ({text}): {e}")
+        return False
+
+
+def _select_natural_day_yesterday(page, label="product") -> bool:
+    yesterday = _product_yesterday()
+    print(f"[INFO] Selecting {label} natural day: {yesterday:%Y/%m/%d}")
+    try:
+        page.keyboard.press("Escape")
+        page.wait_for_timeout(300)
+    except Exception:
+        pass
+
+    clicked_mode = _human_click_text(page, "自然日", exact=True, timeout=8000)
+    if not clicked_mode:
+        # 人类化 hover "更多" 再点 (老 UI 自然日藏在更多下拉里)
+        if _hover_visible_text(page, "更多", exact=True, timeout=5000):
+            page.wait_for_timeout(1200)
+            clicked_mode = _human_click_text(page, "自然日", exact=True, timeout=8000)
+    if not clicked_mode:
+        print(f"[WARN] {label} natural day button not clicked")
+        return False
+
+    page.wait_for_timeout(1200)
+    if _looks_like_login_url(page.url):
+        print(f"[WARN] {label} date mode triggered a login redirect: {page.url}")
+        return False
+    _align_calendar_month(page, yesterday)
+    if not _click_calendar_date(page, yesterday):
+        print(f"[WARN] {label} yesterday date not clicked: {yesterday:%Y-%m-%d}")
+        return False
+    try:
+        page.keyboard.press("Escape")
+        page.mouse.click(120, 120)
+    except Exception:
+        pass
+    _wait_after_action(page, f"select {label} yesterday", settle_ms=3000)
+    if _looks_like_login_url(page.url):
+        print(f"[WARN] {label} date selection triggered a login redirect: {page.url}")
+        return False
+    return True
+
+
+def _select_product_yesterday(page) -> bool:
+    print(f"[INFO] Ensuring product rows are loaded; natural day target: {_product_yesterday_label()}")
+
+    for round_index in range(1, 3):
+        print(f"[INFO] Product date round {round_index}: try natural day yesterday")
+        if _select_natural_day_yesterday(page, "product"):
+            if _wait_for_product_rows(page, timeout=18000):
+                return True
+        if _looks_like_login_url(page.url):
+            print("[WARN] Product date selection lost authentication")
+            return False
+        if round_index == 1:
+            print("[INFO] Product rows not ready after yesterday selection; retrying once")
+
+    print("[INFO] Product rows still not loaded after yesterday date selection")
+    return False
+
+
+def _wait_for_product_rows(page, timeout=20000) -> bool:
+    print("[INFO] Waiting for product rows...")
+    try:
+        page.wait_for_function(
+            "() => /ID\\s*\\d{8,}/.test(document.body.innerText || '') || (document.body.innerText || '').includes('暂无数据')",
+            timeout=timeout,
+        )
+        page.wait_for_timeout(1000)
+        text = page.evaluate("() => document.body.innerText || ''")
+        return bool(re.search(r"ID\s*\d{8,}", text))
+    except Exception as e:
+        print(f"[WARN] Product rows wait skipped: {e}")
+        return False
+
+
+def _parse_metric_cell(value: str, meta: dict | None = None) -> dict:
+    text = str(value or "").strip()
+    change_match = re.search(r"([+-]?\d+(?:\.\d+)?%)\s*$", text)
+    change = ""
+    metric_value = text
+    if change_match:
+        prefix = text[:change_match.start()].strip()
+        if prefix:
+            change = change_match.group(1)
+            metric_value = prefix
+    direction = (meta or {}).get("direction", "")
+    if not direction:
+        if any(mark in text for mark in ("↑", "⬆", "▲")):
+            direction = "上涨"
+        elif any(mark in text for mark in ("↓", "⬇", "▼")):
+            direction = "下降"
+        elif "持平" in text:
+            direction = "持平"
+            metric_value = metric_value.replace("持平", "").strip()
+        elif re.search(r"\s-\s*$", text):
+            direction = "-"
+            metric_value = re.sub(r"\s-\s*$", "", metric_value).strip()
+    return {
+        "value": metric_value,
+        "change": change,
+        "change_direction": direction,
+        "raw": text,
+    }
+
+
+def _parse_product_row(row_text: str, cells: list, headers: list[str], cell_meta: dict | None = None) -> dict:
+    columns = {}
+    for index, cell in enumerate(cells):
+        text = cell.get("text", "") if isinstance(cell, dict) else cell
+        columns[headers[index] if index < len(headers) else f"列{index + 1}"] = text
+
+    cell_meta = cell_meta or {}
+
+    product_info = columns.get("商品信息") or (cells[1] if len(cells) > 1 else row_text)
+    product_id_match = re.search(r"ID\s*(\d+)", product_info)
+    product_id = product_id_match.group(1) if product_id_match else ""
+    product_name = re.split(r"\s*ID\s*\d+", product_info, maxsplit=1)[0].strip()
+
+    def pick(*names):
+        for name in names:
+            value = columns.get(name)
+            if value not in (None, ""):
+                return value
+        return ""
+
+    def pick_contains(*keywords):
+        value = pick(*keywords)
+        if value not in (None, ""):
+            return value
+        for column_name, column_value in columns.items():
+            if column_value in (None, ""):
+                continue
+            if all(keyword in column_name for keyword in keywords):
+                return column_value
+        return ""
+
+    def source_meta(*keywords):
+        for column_name in columns:
+            if all(keyword in column_name for keyword in keywords):
+                return cell_meta.get(column_name, {})
+        return {}
+
+    metric_names = [
+        "成交金额",
+        "成交订单数",
+        "退款金额(支付时间)",
+        "商品点击人数",
+        "成交人数",
+        "退款订单数(支付时间)",
+        "加购人数",
+        "商品曝光人数",
+        "商品曝光次数",
+        "点击支付率",
+        "加购支付转化率",
+        "访问加购转化率",
+        "成交客单价",
+        "退款率(支付时间)",
+    ]
+    metric_source_names = {
+        "成交金额": ("成交金额", "净成交金额", "用户支付金额"),
+        "成交订单数": ("成交订单数", "成交订单数 ", "成交单数", "支付订单数"),
+        "退款金额(支付时间)": ("退款金额(支付时间)", "退款金额（支付时间）", "退款金额"),
+        "商品点击人数": ("商品点击人数",),
+        "成交人数": ("成交人数",),
+        "退款订单数(支付时间)": ("退款订单数(支付时间)", "退款订单数（支付时间）", "退款订单数"),
+        "加购人数": ("加购人数",),
+        "商品曝光人数": ("商品曝光人数",),
+        "商品曝光次数": ("商品曝光次数",),
+        "点击支付率": ("点击支付率（人数）", "点击支付率"),
+        "加购支付转化率": ("加购支付转化率",),
+        "访问加购转化率": ("访问加购转化率", "访问加购转化率 "),
+        "成交客单价": ("成交客单价",),
+        "退款率(支付时间)": ("退款率(支付时间)", "退款率（支付时间）", "退款率"),
+    }
+    metrics = {}
+    for metric_name in metric_names:
+        source_names = metric_source_names.get(metric_name, (metric_name,))
+        source_name = next((name for name in source_names if columns.get(name) not in (None, "")), metric_name)
+        raw_value = pick(*source_names)
+        meta = cell_meta.get(source_name, {})
+        if "点击" in metric_name and "支付" in metric_name:
+            raw_value = raw_value or pick_contains("点击", "支付")
+            meta = meta or source_meta("点击", "支付")
+        elif "加购" in metric_name:
+            raw_value = raw_value or pick_contains("加购")
+            meta = meta or source_meta("加购")
+        elif "退款率" in metric_name:
+            raw_value = raw_value or pick_contains("退款率")
+            meta = meta or source_meta("退款率")
+        metrics[metric_name] = _parse_metric_cell(raw_value, meta)
+
+    def metric_by_keywords(*keywords):
+        for name, metric in metrics.items():
+            if all(keyword in name for keyword in keywords):
+                return metric
+        return {"value": "", "change": "", "change_direction": "", "raw": ""}
+
+    metrics["销量"] = dict(metrics["成交订单数"])
+    metrics["访客数"] = dict(metrics["商品点击人数"])
+    metrics["转化率"] = dict(metrics["点击支付率"])
+    metrics["加购率"] = dict(metrics["访问加购转化率"])
+    metrics["退款率"] = dict(metrics["退款率(支付时间)"])
+
+    result = {
+        "product_name": product_name,
+        "product_id": product_id,
+        "date_range": _product_yesterday_label(),
+        "captured_at": datetime.now().isoformat(),
+        "成交金额": metrics["成交金额"]["value"],
+        "销量": metrics["成交订单数"]["value"],
+        "访客数": metrics["商品点击人数"]["value"],
+        "转化率": metrics["转化率"]["value"],
+        "加购率": metrics["加购率"]["value"],
+        "退款率": metrics["退款率"]["value"],
+        "metrics": metrics,
+        "columns": columns,
+        "cell_meta": cell_meta,
+        "row_text": row_text,
+    }
+    for metric_name, metric in metrics.items():
+        result[f"{metric_name}_变化百分比"] = metric.get("change", "")
+        result[f"{metric_name}_变化方向"] = metric.get("change_direction", "")
+    return result
+
+
+def _extract_products_on_page(page) -> list[dict]:
+    raw = page.evaluate(
+        """() => {
+            const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const isVisible = (el) => {
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style && style.visibility !== 'hidden' && style.display !== 'none'
+                    && rect.width > 0 && rect.height > 0;
+            };
+            const scrollables = Array.from(document.querySelectorAll('div, section'))
+                .filter(el => el.scrollWidth > el.clientWidth + 80)
+                .filter(isVisible)
+                .sort((a, b) => (b.scrollWidth - b.clientWidth) - (a.scrollWidth - a.clientWidth));
+            const scroller = scrollables[0] || null;
+            const maxLeft = scroller ? Math.max(0, scroller.scrollWidth - scroller.clientWidth) : 0;
+            const positions = Array.from(new Set([0, Math.floor(maxLeft / 2), maxLeft]));
+            const snapshots = [];
+
+            function cellDirection(td) {
+                const text = clean(td.innerText || td.textContent);
+                if (/[↑⬆▲]/.test(text)) return '上涨';
+                if (/[↓⬇▼]/.test(text)) return '下降';
+                const nodes = [td, ...Array.from(td.querySelectorAll('*'))];
+                for (const node of nodes) {
+                    const style = window.getComputedStyle(node);
+                    const before = window.getComputedStyle(node, '::before');
+                    const after = window.getComputedStyle(node, '::after');
+                    const pseudoText = `${before && before.content || ''} ${after && after.content || ''}`;
+                    if (/[↑⬆▲]/.test(pseudoText)) return '上涨';
+                    if (/[↓⬇▼]/.test(pseudoText)) return '下降';
+                    const fill = node.getAttribute && (node.getAttribute('fill') || node.getAttribute('stroke') || '');
+                    const cssText = `${style && style.color || ''} ${style && style.fill || ''} ${style && style.stroke || ''} ${fill}`;
+                    if (/(red|#f|rgb\\(2[0-5][0-9]|rgb\\(1[8-9][0-9])/.test(cssText)) return '上涨';
+                    if (/(green|#0|#1|rgb\\([0-9]{1,2},\\s*1[3-9][0-9]|rgb\\([0-9]{1,2},\\s*2[0-5][0-9])/.test(cssText)) return '下降';
+                    const color = style && style.color || '';
+                    const match = color.match(/rgba?\\((\\d+),\\s*(\\d+),\\s*(\\d+)/);
+                    if (!match) continue;
+                    const r = Number(match[1]);
+                    const g = Number(match[2]);
+                    const b = Number(match[3]);
+                    if (r > 180 && g < 130 && b < 150) return '上涨';
+                    if (g > 130 && r < 130 && b < 170) return '下降';
+                }
+                return '';
+            }
+
+            function capture() {
+                const headers = Array.from(document.querySelectorAll('table thead th'))
+                    .map(th => clean(th.innerText || th.textContent))
+                    .filter(Boolean);
+                const rows = Array.from(document.querySelectorAll('table tbody tr, table tr'));
+                return rows.map(row => {
+                    const rowText = clean(row.innerText || row.textContent);
+                    const tds = Array.from(row.querySelectorAll('td'));
+                    const cells = tds.map(td => ({ text: clean(td.innerText || td.textContent), direction: cellDirection(td) }));
+                    const cell_meta = {};
+                    cells.forEach((cell, index) => {
+                        const header = headers[index] || `列${index + 1}`;
+                        cell_meta[header] = { direction: cell.direction };
+                    });
+                    return { rowText, cells, headers, cell_meta };
+                }).filter(item => item.rowText.includes('ID ') || item.rowText.includes('ID'));
+            }
+
+            for (const left of positions) {
+                if (scroller) {
+                    scroller.scrollLeft = left;
+                    scroller.dispatchEvent(new Event('scroll', { bubbles: true }));
+                }
+                snapshots.push(...capture());
+            }
+            if (snapshots.length) return snapshots;
+
+            const candidates = Array.from(document.querySelectorAll('div'))
+                .map(el => {
+                    const rowText = clean(el.innerText || el.textContent);
+                    return { rowText, cells: rowText.split(/ (?=¥|\\d|ID |GYXX|光影)/).filter(Boolean) };
+                })
+                .filter(item => /ID\\s*\\d{8,}/.test(item.rowText))
+                .filter(item => item.rowText.length > 50 && item.rowText.length < 1200);
+            const seen = new Set();
+            items = [];
+            for (const item of candidates) {
+                const id = (item.rowText.match(/ID\\s*(\\d{8,})/) || [])[1] || item.rowText;
+                if (seen.has(id)) continue;
+                seen.add(id);
+                items.push(item);
+            }
+            return items;
+        }"""
+    )
+
+    products = []
+    seen = set()
+    headers = page.evaluate(
+        """() => Array.from(document.querySelectorAll('table thead th'))
+            .map(th => (th.innerText || th.textContent || '').replace(/\\s+/g, ' ').trim())
+            .filter(Boolean)"""
+    )
+    for item in raw:
+        item_headers = item.get("headers") or headers
+        product = _parse_product_row(item["rowText"], item["cells"], item_headers, item.get("cell_meta") or {})
+        key = product.get("product_id") or product.get("row_text")
+        if key and key not in seen:
+            seen.add(key)
+            products.append(product)
+        elif key:
+            for existing in products:
+                existing_key = existing.get("product_id") or existing.get("row_text")
+                if existing_key != key:
+                    continue
+                existing["columns"].update({k: v for k, v in product.get("columns", {}).items() if v not in (None, "")})
+                existing["cell_meta"].update(product.get("cell_meta", {}))
+                for metric_name, metric in product.get("metrics", {}).items():
+                    if metric.get("value") and not existing.get("metrics", {}).get(metric_name, {}).get("value"):
+                        existing.setdefault("metrics", {})[metric_name] = metric
+                        existing[metric_name] = metric.get("value", "")
+                        existing[f"{metric_name}_变化百分比"] = metric.get("change", "")
+                        existing[f"{metric_name}_变化方向"] = metric.get("change_direction", "")
+                alias_pairs = {
+                    "销量": "成交订单数",
+                    "访客数": "商品点击人数",
+                    "转化率": "点击支付率",
+                    "加购率": "访问加购转化率",
+                    "退款率": "退款率(支付时间)",
+                }
+                for alias, source in alias_pairs.items():
+                    source_metric = existing.get("metrics", {}).get(source, {})
+                    if source_metric:
+                        existing.setdefault("metrics", {})[alias] = dict(source_metric)
+                        existing[alias] = source_metric.get("value", "")
+                        existing[f"{alias}_变化百分比"] = source_metric.get("change", "")
+                        existing[f"{alias}_变化方向"] = source_metric.get("change_direction", "")
+                break
+
+    print(f"[OK] Extracted {len(products)} products on current page")
+    return products
+
+
+def _click_next_page(page) -> bool:
+    print("[INFO] Trying next page...")
+    try:
+        clicked = page.evaluate(
+            """() => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const disabled = (el) => {
+                    const signal = [
+                        String(el.className || ''),
+                        el.getAttribute('aria-disabled') || '',
+                        el.getAttribute('disabled') || ''
+                    ].join(' ').toLowerCase();
+                    return signal.includes('disabled') || signal.includes('true');
+                };
+                const candidates = Array.from(document.querySelectorAll('button, li, a, span, div'))
+                    .filter(isVisible)
+                    .filter(el => !disabled(el))
+                    .filter(el => {
+                        const text = (el.innerText || el.textContent || '').trim();
+                        const title = el.getAttribute('title') || '';
+                        const aria = el.getAttribute('aria-label') || '';
+                        return text === '>' || text === '下一页' || title.includes('下一页') || aria.includes('下一页');
+                    })
+                    .sort((a, b) => b.getBoundingClientRect().right - a.getBoundingClientRect().right);
+                if (!candidates.length) return false;
+                candidates[0].click();
+                return true;
+            }"""
+        )
+        if clicked:
+            _wait_after_action(page, "next product page")
+            page.wait_for_timeout(2000)
+        return bool(clicked)
+    except Exception as e:
+        print(f"[WARN] Next page click failed: {e}")
+        return False
+
+
+def _recover_empty_product_list(page) -> None:
+    print("[INFO] Recovering empty product list...")
+    try:
+        page.keyboard.press("Escape")
+    except Exception:
+        pass
+    _click_visible_text(page, "全部", exact=True, timeout=5000)
+    try:
+        page.evaluate(
+            """() => {
+                const inputs = Array.from(document.querySelectorAll('input'))
+                    .filter(el => (el.placeholder || '').includes('商品') || (el.placeholder || '').includes('ID'));
+                for (const input of inputs) {
+                    input.focus();
+                    input.value = '';
+                    input.dispatchEvent(new InputEvent('input', { bubbles: true, inputType: 'deleteContentBackward' }));
+                    input.dispatchEvent(new Event('change', { bubbles: true }));
+                }
+            }"""
+        )
+    except Exception:
+        pass
+    _ensure_product_list_tab(page)
+
+
+def _collect_all_products(page) -> list[dict]:
+    products = []
+    seen_ids = set()
+    previous_signature = None
+
+    for page_index in range(1, 100):
+        print(f"[INFO] Collecting product page {page_index}...")
+        current = _extract_products_on_page(page)
+        if page_index == 1 and not current:
+            _recover_empty_product_list(page)
+            current = _extract_products_on_page(page)
+        signature = "|".join(product.get("product_id", "") for product in current)
+
+        for product in current:
+            product_id = product.get("product_id")
+            key = product_id or product.get("row_text")
+            if key and key not in seen_ids:
+                seen_ids.add(key)
+                products.append(product)
+
+        if not current:
+            print("[WARN] No products found on page; stopping pagination")
+            _dump_product_debug(page, f"no_products_page_{page_index}")
+            break
+        if signature and signature == previous_signature:
+            print("[WARN] Page did not change after pagination; stopping")
+            break
+        previous_signature = signature
+
+        if not _click_next_page(page):
+            break
+
+    print(f"[OK] Total products collected: {len(products)}")
+    return products
+
+
+def _dump_product_debug(page, label: str) -> None:
+    if not PRODUCT_DEBUG_DUMP:
+        return
+    try:
+        os.makedirs(PRODUCT_DATA_DIR, exist_ok=True)
+        stamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+        screenshot_path = os.path.join(PRODUCT_DATA_DIR, f"product_debug_{label}_{stamp}.png")
+        text_path = os.path.join(PRODUCT_DATA_DIR, f"product_debug_{label}_{stamp}.txt")
+        page.screenshot(path=screenshot_path, full_page=True)
+        text = page.evaluate("() => (document.body.innerText || '').slice(0, 5000)")
+        with open(text_path, "w", encoding="utf-8") as f:
+            f.write(text)
+        print(f"[DEBUG] Product debug screenshot: {screenshot_path}")
+        print(f"[DEBUG] Product debug text: {text_path}")
+    except Exception as e:
+        print(f"[WARN] Product debug dump failed: {e}")
+
+
+def _save_audience_profile_screenshot(page, product_id: str, output_dir: str | None = None) -> str:
+    try:
+        output_dir = output_dir or PRODUCT_DATA_DIR
+        os.makedirs(output_dir, exist_ok=True)
+        stamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+        safe_id = re.sub(r"\D+", "", product_id or "") or "unknown"
+        screenshot_path = os.path.join(output_dir, f"audience_profile_{safe_id}_{stamp}.png")
+        page.screenshot(path=screenshot_path, full_page=True)
+        print(f"[OK] Audience profile screenshot: {screenshot_path}")
+        return screenshot_path
+    except Exception as e:
+        print(f"[WARN] Audience profile screenshot failed: {e}")
+        return ""
+
+
+def _alive_context_pages(page) -> list:
+    try:
+        return [p for p in page.context.pages if not p.is_closed()]
+    except Exception:
+        return []
+
+
+def _find_product_list_page(page, product_list_url: str = ""):
+    pages = _alive_context_pages(page)
+    for candidate in reversed(pages):
+        try:
+            if "commodity/product-list" in candidate.url:
+                return candidate
+        except Exception:
+            pass
+    for candidate in reversed(pages):
+        try:
+            if product_list_url and candidate.url.split("#", 1)[0] == product_list_url.split("#", 1)[0]:
+                return candidate
+        except Exception:
+            pass
+    return page
+
+
+def _ensure_product_list_page(page, product_list_url: str = ""):
+    page = _find_product_list_page(_active_page(page), product_list_url)
+    try:
+        if product_list_url and "commodity/product-list" not in page.url:
+            print("[INFO] Returning to product list page...")
+            page.goto(product_list_url, wait_until="domcontentloaded", timeout=45000)
+            try:
+                page.wait_for_load_state("load", timeout=15000)
+            except Exception:
+                pass
+            _wait_after_action(page, "return product list", settle_ms=1800)
+        _ensure_product_list_tab(page)
+    except Exception as e:
+        print(f"[WARN] Ensure product list page failed: {e}")
+    return page
+
+
+def _cleanup_product_detail_pages(page, product_list_url: str, pages_before: list):
+    before_ids = {id(p) for p in pages_before}
+    list_page = None
+    for candidate in _alive_context_pages(page):
+        try:
+            if id(candidate) in before_ids and "commodity/product-list" in candidate.url:
+                list_page = candidate
+                break
+        except Exception:
+            pass
+
+    for candidate in list(_alive_context_pages(page)):
+        if id(candidate) in before_ids:
+            continue
+        try:
+            print(f"[INFO] Closing temporary product page: {candidate.url}")
+            candidate.close()
+        except Exception as e:
+            print(f"[WARN] Close temporary product page failed: {e}")
+
+    page = list_page or _find_product_list_page(page, product_list_url)
+    try:
+        if product_list_url and "commodity/product-list" not in page.url:
+            print("[INFO] Product detail used the list tab; navigating back to product list")
+            page.goto(product_list_url, wait_until="domcontentloaded", timeout=45000)
+            try:
+                page.wait_for_load_state("load", timeout=15000)
+            except Exception:
+                pass
+        _wait_after_action(page, "cleanup product detail pages", settle_ms=1500)
+    except Exception as e:
+        print(f"[WARN] Product page cleanup navigation failed: {e}")
+    return page
+
+
+def _search_product_by_id(page, product_id: str) -> bool:
+    print(f"[INFO] Searching product ID: {product_id}")
+    try:
+        page.keyboard.press("Escape")
+        page.wait_for_timeout(300)
+        ok = False
+        selectors = [
+            'input[placeholder*="商品名称/ID"]',
+            'input[placeholder*="商品名称"]',
+            'input[placeholder*="商品"][placeholder*="ID"]',
+        ]
+        for selector in selectors:
+            try:
+                locator = page.locator(selector).first
+                locator.click(timeout=3000)
+                locator.fill("")
+                locator.fill(product_id)
+                ok = True
+                print(f"[INFO] Filled product-list search input: {selector}")
+                break
+            except Exception:
+                pass
+
+        if not ok:
+            ok = page.evaluate(
+                """(productId) => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const inputs = Array.from(document.querySelectorAll('input')).filter(isVisible);
+                const candidates = inputs
+                    .map(el => ({ el, rect: el.getBoundingClientRect(), placeholder: el.placeholder || '' }))
+                    .filter(item => item.rect.top > 180 && item.rect.left < 900 && item.rect.width > 220)
+                    .filter(item => item.placeholder.includes('商品') || item.placeholder.includes('ID'))
+                    .sort((a, b) => {
+                        const aScore = Math.abs(a.rect.top - 440) + Math.abs(a.rect.left - 260);
+                        const bScore = Math.abs(b.rect.top - 440) + Math.abs(b.rect.left - 260);
+                        return aScore - bScore;
+                    });
+                const input = candidates.length ? candidates[0].el : null;
+                if (!input) return false;
+                input.focus();
+                input.select && input.select();
+                input.value = '';
+                input.dispatchEvent(new InputEvent('input', { bubbles: true, inputType: 'deleteContentBackward', data: null }));
+                input.value = productId;
+                input.dispatchEvent(new InputEvent('input', { bubbles: true, inputType: 'insertText', data: productId }));
+                input.dispatchEvent(new Event('change', { bubbles: true }));
+                return true;
+            }""",
+                product_id,
+            )
+        if not ok:
+            selector = 'input[placeholder*="商品名称/ID"], input[placeholder*="商品"], input[placeholder*="ID"], input[placeholder*="搜索"]'
+            if adaptive_fill_input(page, "product_id_search", selector, product_id, timeout=5000):
+                ok = True
+        if not ok:
+            raise RuntimeError("product-list search input not found")
+    except Exception as e:
+        print(f"[WARN] Product search failed ({product_id}): {e}")
+        _dump_product_debug(page, f"search_input_not_found_{product_id}")
+        return False
+    _wait_after_action(page, f"search product {product_id}", settle_ms=1800)
+    try:
+        page.wait_for_function(
+            f"""() => Array.from(document.querySelectorAll('table tbody tr, table tr'))
+                .some(row => (row.innerText || row.textContent || '').includes({json.dumps(product_id)}))""",
+            timeout=12000,
+        )
+        page.wait_for_timeout(600)
+        return True
+    except Exception as e:
+        print(f"[WARN] Product search result not visible after fill ({product_id}): {e}")
+        _dump_product_debug(page, f"search_result_not_found_{product_id}")
+        return False
+
+
+def _click_product_detail(page, product_id: str) -> bool:
+    print(f"[INFO] Opening product detail: {product_id}")
+    try:
+        clicked = False
+        try:
+            row = page.locator("table tbody tr, table tr").filter(has_text=product_id).first
+            row.get_by_text("查看详情").first.click(timeout=5000)
+            clicked = True
+            print("[INFO] Clicked product detail from matched table row")
+        except Exception as row_error:
+            print(f"[WARN] Matched table row click failed ({product_id}): {row_error}")
+
+        if not clicked:
+            clicked = page.evaluate(
+            """(productId) => {
+                const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const tableRows = Array.from(document.querySelectorAll('table tbody tr, table tr'))
+                    .filter(isVisible)
+                    .filter(row => clean(row.innerText || row.textContent).includes(productId));
+                for (const row of tableRows) {
+                    const actions = Array.from(row.querySelectorAll('a, button, span, div'))
+                        .filter(isVisible)
+                        .filter(el => clean(el.innerText || el.textContent).includes('查看详情'))
+                        .sort((a, b) => {
+                            const ar = a.getBoundingClientRect();
+                            const br = b.getBoundingClientRect();
+                            return (br.left - ar.left) || (ar.width * ar.height) - (br.width * br.height);
+                        });
+                    if (actions.length) {
+                        actions[0].click();
+                        return true;
+                    }
+                }
+                const rowCenters = tableRows.map(row => {
+                    const rect = row.getBoundingClientRect();
+                    return rect.top + rect.height / 2;
+                });
+                const pageActions = Array.from(document.querySelectorAll('a, button, span, div'))
+                    .filter(isVisible)
+                    .filter(el => clean(el.innerText || el.textContent).includes('查看详情'))
+                    .map(el => {
+                        const rect = el.getBoundingClientRect();
+                        const centerY = rect.top + rect.height / 2;
+                        const yGap = rowCenters.length ? Math.min(...rowCenters.map(y => Math.abs(y - centerY))) : 9999;
+                        return { el, rect, yGap };
+                    })
+                    .filter(item => item.yGap < 70)
+                    .sort((a, b) => {
+                        const rightScore = b.rect.left - a.rect.left;
+                        return (a.yGap - b.yGap) || rightScore;
+                    });
+                if (pageActions.length) {
+                    pageActions[0].el.click();
+                    return true;
+                }
+                return false;
+            }""",
+                product_id,
+            )
+        if not clicked:
+            print(f"[WARN] Product row detail link not found for {product_id}")
+            _dump_product_debug(page, f"detail_link_not_found_{product_id}")
+        if clicked:
+            _wait_after_action(page, f"open detail {product_id}", settle_ms=2500)
+            detail_page = _active_page(page)
+            try:
+                detail_page.wait_for_function(
+                    f"""() => (location.href.includes('product-detail') || (document.body.innerText || '').includes('人群分析'))
+                        && (document.body.innerText || '').includes({json.dumps(product_id)})""",
+                    timeout=12000,
+                )
+                return True
+            except Exception as wait_error:
+                print(f"[WARN] Product detail page not detected ({product_id}): {wait_error}")
+                _dump_product_debug(detail_page, f"detail_not_opened_{product_id}")
+    except Exception as e:
+        print(f"[WARN] Open product detail failed ({product_id}): {e}")
+    return False
+
+
+def _open_product_audience_profile(page) -> bool:
+    page = _active_page(page)
+    print("[INFO] Opening product audience profile...")
+    if not _click_visible_text(page, "人群分析", exact=True, timeout=10000):
+        print("[WARN] Audience analysis tab not clicked")
+        _dump_product_debug(page, "audience_tab_not_found")
+        return False
+    _wait_after_action(page, "open audience analysis", settle_ms=1800)
+
+    if _is_empty_audience_state(page):
+        print("[INFO] Audience analysis is empty; skip audience profile")
+        return True
+
+    if not _click_visible_text(page, "人群画像", exact=False, timeout=10000):
+        if _is_empty_audience_state(page):
+            print("[INFO] Audience profile is empty; skip audience profile")
+            return True
+        print("[WARN] Audience profile button not clicked")
+        _dump_product_debug(page, "audience_profile_button_not_found")
+        return False
+    _wait_after_action(page, "open audience profile", settle_ms=2500)
+    page = _active_page(page)
+
+    try:
+        page.wait_for_function(
+            "() => location.href.includes('goods-user-analysis') || (document.body.innerText || '').includes('性别分布')",
+            timeout=12000,
+        )
+    except Exception as e:
+        print(f"[WARN] Audience profile page wait skipped: {e}")
+    return True
+
+
+def _is_empty_audience_state(page) -> bool:
+    try:
+        text = page.evaluate("() => document.body ? (document.body.innerText || '') : ''")
+    except Exception:
+        return False
+    empty_marks = [
+        "上架的商品在所选时间段内尚未曝光",
+        "交易建议更换洞察时间",
+        "尚未曝光/交易",
+        "暂无人群画像",
+        "暂无数据",
+    ]
+    return any(mark in text for mark in empty_marks)
+
+
+def _open_product_experience_analysis(page) -> bool:
+    page = _active_page(page)
+    print("[INFO] Opening product experience analysis...")
+    if not _click_visible_text(page, "退款及体验", exact=True, timeout=10000):
+        print("[WARN] Refund and experience tab not clicked")
+        _dump_product_debug(page, "refund_experience_tab_not_found")
+        return False
+    _wait_after_action(page, "open refund and experience", settle_ms=1800)
+
+    if not _click_visible_text(page, "体验分析", exact=True, timeout=10000):
+        print("[WARN] Experience analysis subtab not clicked")
+        _dump_product_debug(page, "experience_analysis_tab_not_found")
+        return False
+    _wait_after_action(page, "open experience analysis", settle_ms=1200)
+
+    if _click_visible_text(page, "近30天", exact=True, timeout=8000):
+        _wait_after_action(page, "select experience last 30 days", settle_ms=2500)
+    else:
+        print("[WARN] Experience last 30 days button not clicked; continuing with current period")
+
+    try:
+        page.wait_for_function(
+            "() => (document.body.innerText || '').includes('商品品质退货率') || (document.body.innerText || '').includes('评价好评率')",
+            timeout=12000,
+        )
+    except Exception as e:
+        print(f"[WARN] Experience analysis data wait skipped: {e}")
+    return True
+
+
+def _extract_product_experience_analysis(page) -> dict:
+    page = _active_page(page)
+    raw = page.evaluate(
+        """() => {
+            const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const labels = ['商品品质退货率', '评价好评率'];
+            const isVisible = (el) => {
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style && style.visibility !== 'hidden' && style.display !== 'none'
+                    && rect.width > 0 && rect.height > 0;
+            };
+            const visualSignal = (el) => {
+                const nodes = [el, ...Array.from(el.querySelectorAll('*'))];
+                for (const node of nodes) {
+                    const text = clean(node.innerText || node.textContent);
+                    const before = window.getComputedStyle(node, '::before')?.content || '';
+                    const after = window.getComputedStyle(node, '::after')?.content || '';
+                    const marker = `${text} ${before} ${after}`;
+                    if (/[↑↗▲]/.test(marker) || marker.includes('上升') || marker.includes('上涨')) return { direction: '上升', tone: '' };
+                    if (/[↓↘▼]/.test(marker) || marker.includes('下降') || marker.includes('下跌')) return { direction: '下降', tone: '' };
+                    const color = window.getComputedStyle(node).color || '';
+                    const nums = color.match(/rgba?\\((\\d+),\\s*(\\d+),\\s*(\\d+)/);
+                    if (nums && /\\d+(?:\\.\\d+)?%/.test(text)) {
+                        const r = Number(nums[1]), g = Number(nums[2]), b = Number(nums[3]);
+                        if (g > r + 20 && g > b + 20) return { direction: '', tone: 'green' };
+                        if (r > g + 20 && r > b + 20) return { direction: '', tone: 'red' };
+                    }
+                }
+                return { direction: '', tone: '' };
+            };
+            const cardForLabel = (label) => {
+                const labelEls = Array.from(document.querySelectorAll('div, span, p, section'))
+                    .filter(isVisible)
+                    .filter(el => clean(el.innerText || el.textContent).includes(label))
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                for (const el of labelEls) {
+                    let node = el;
+                    for (let depth = 0; node && depth < 6; depth += 1, node = node.parentElement) {
+                        const text = clean(node.innerText || node.textContent);
+                        const rect = node.getBoundingClientRect();
+                        if (text.includes(label) && /\\d+(?:\\.\\d+)?%/.test(text) && rect.width >= 120 && rect.width <= 520 && rect.height >= 60 && rect.height <= 220) {
+                            return node;
+                        }
+                    }
+                }
+                return null;
+            };
+            const parseCard = (label) => {
+                const card = cardForLabel(label);
+                if (!card) return { raw: '', value: '', change: '', change_direction: '' };
+                const text = clean(card.innerText || card.textContent);
+                const afterLabel = text.includes(label) ? text.split(label).slice(1).join(label) : text;
+                const percents = Array.from(afterLabel.matchAll(/[-+]?\\d+(?:\\.\\d+)?%/g)).map(m => m[0]);
+                const periodPart = text.includes('较上周期') ? text.split('较上周期').slice(1).join('较上周期') : '';
+                const changeMatch = periodPart.match(/[-+]?\\d+(?:\\.\\d+)?%/);
+                const signal = visualSignal(card);
+                let direction = signal.direction || '';
+                if (!direction && changeMatch && signal.tone) {
+                    if (label.includes('退货率')) {
+                        direction = signal.tone === 'green' ? '下降' : '上升';
+                    } else {
+                        direction = signal.tone === 'green' ? '上升' : '下降';
+                    }
+                }
+                if (periodPart.includes('-') && !changeMatch) direction = '';
+                return {
+                    raw: text,
+                    value: percents[0] || '',
+                    change: changeMatch ? changeMatch[0] : '',
+                    change_direction: changeMatch ? direction : '',
+                };
+            };
+            const result = {};
+            for (const label of labels) result[label] = parseCard(label);
+            result.raw_text_excerpt = clean(document.body.innerText || '').slice(0, 3000);
+            return result;
+        }"""
+    )
+    result = {
+        "captured_at": datetime.now().isoformat(),
+        "url": page.url,
+        "period": "近30天",
+    }
+    result.update(raw or {})
+    print(
+        "[OK] Experience analysis extracted: "
+        f"quality_return={result.get('商品品质退货率', {}).get('value', '')}, "
+        f"good_rating={result.get('评价好评率', {}).get('value', '')}"
+    )
+    return result
+
+
+def _collect_product_experience_analysis(page) -> dict:
+    if not _open_product_experience_analysis(page):
+        raise RuntimeError("experience analysis not opened")
+    return _extract_product_experience_analysis(page)
+
+
+def _open_comment_detail_page(page) -> bool:
+    page = _active_page(page)
+    print(f"[INFO] Opening comment detail page: {COMMENT_DETAIL_URL}")
+    page.goto(COMMENT_DETAIL_URL, wait_until="domcontentloaded", timeout=60000)
+    try:
+        page.wait_for_load_state("load", timeout=12000)
+    except Exception as e:
+        print(f"[WARN] Comment page load wait skipped: {e}")
+    _wait_after_action(page, "open comment detail page", settle_ms=3000)
+    if not _click_visible_text(page, "近7天", exact=True, timeout=10000):
+        print("[WARN] Comment last 7 days button not clicked; continuing with current range")
+    else:
+        _wait_after_action(page, "select comment last 7 days", settle_ms=2500)
+        if _click_visible_text(page, "查询", exact=True, timeout=8000):
+            _wait_after_action(page, "query comment last 7 days", settle_ms=3500)
+    try:
+        page.wait_for_function(
+            "() => (document.body.innerText || '').includes('评价明细') && (document.body.innerText || '').includes('订单ID')",
+            timeout=15000,
+        )
+    except Exception as e:
+        print(f"[WARN] Comment detail rows wait skipped: {e}")
+    return True
+
+
+def _click_comment_rating_tab(page, rating: str) -> bool:
+    page = _active_page(page)
+    print(f"[INFO] Selecting comment rating: {rating}")
+    try:
+        clicked = page.evaluate(
+            """(rating) => {
+                const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const candidates = Array.from(document.querySelectorAll('button, span, div, label, a'))
+                    .filter(isVisible)
+                    .filter(el => {
+                        const text = clean(el.innerText || el.textContent);
+                        if (!(text === rating || text.startsWith(rating))) return false;
+                        const rect = el.getBoundingClientRect();
+                        return rect.top < 560 && rect.left > 250;
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return false;
+                candidates[0].click();
+                return true;
+            }""",
+            rating,
+        )
+        if clicked:
+            _wait_after_action(page, f"select {rating} comments", settle_ms=3500)
+            return True
+    except Exception as e:
+        print(f"[WARN] Comment rating tab click failed ({rating}): {e}")
+    return False
+
+
+def _extract_comment_count_for_rating(page, rating: str) -> str:
+    try:
+        count = page.evaluate(
+            """(rating) => {
+                const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const parseNumber = (text) => {
+                    const matches = Array.from(String(text || '').matchAll(/(\\d{1,8})\\s*(?:\\u6761|\\u4e2a|$|\\))/g))
+                        .map(match => Number(match[1]))
+                        .filter(num => Number.isFinite(num));
+                    return matches.length ? Math.max(...matches) : 0;
+                };
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const lines = (document.body.innerText || '')
+                    .split(/\\n+/)
+                    .map(clean)
+                    .filter(Boolean);
+                const totals = lines
+                    .map(line => line.match(/^\\u5171\\s*(\\d+)\\s*\\u6761$/))
+                    .filter(Boolean)
+                    .map(match => Number(match[1]));
+                if (totals.length) return String(totals[totals.length - 1]);
+
+                const ratingCandidates = Array.from(document.querySelectorAll('button, span, div, label, a'))
+                    .filter(isVisible)
+                    .map(el => ({ text: clean(el.innerText || el.textContent), rect: el.getBoundingClientRect() }))
+                    .filter(item => item.rect.top < 620 && item.rect.left > 200 && item.text.length <= 80)
+                    .filter(item => item.text.includes(rating))
+                    .map(item => parseNumber(item.text))
+                    .filter(Boolean);
+                return ratingCandidates.length ? String(Math.max(...ratingCandidates)) : '';
+            }""",
+            rating,
+        ) or ""
+        return str(count).strip()
+    except Exception:
+        return ""
+
+
+def _comment_expected_pages(expected_count: str) -> int:
+    try:
+        count = int(str(expected_count or "0").replace(",", "").strip())
+    except Exception:
+        count = 0
+    if count <= 0:
+        return 0
+    return math.ceil(count / 20)
+
+
+def _extract_comment_rows(page, rating: str) -> list[dict]:
+    page = _active_page(page)
+    rows = page.evaluate(
+        """(rating) => {
+            const seen = new Set();
+            const clean = (s) => (s || '').replace(/[ \\t]+/g, ' ').trim();
+            const lines = (document.body.innerText || '')
+                .split(/\\n+/)
+                .map(line => clean(line).replace(/^[-•·]+\\s*/, '').trim())
+                .filter(Boolean);
+            const rows = [];
+            for (let start = 0; start < lines.length; start += 1) {
+                if (!/^订单ID[:：]?\\s*\\d+/.test(lines[start])) continue;
+                const block = [];
+                for (let i = start; i < lines.length; i += 1) {
+                    if (i > start && /^订单ID[:：]?\\s*\\d+/.test(lines[i])) break;
+                    if (/^共\\d+条$|^\\d+ 条\\/页$|^跳至页$|^识别后发起举报/.test(lines[i])) break;
+                    block.push(lines[i]);
+                }
+                const raw = block.join(' ');
+                if (!raw.includes(rating)) continue;
+                const orderMatch = raw.match(/订单ID[:：]?\\s*(\\d+)/);
+                const productIdMatch = raw.match(/商品ID[:：]?\\s*(\\d+)/);
+                const specMatch = raw.match(/规格ID[:：]?\\s*(\\d+)/);
+                const timeMatch = raw.match(/20\\d{2}[/-]\\d{2}[/-]\\d{2}\\s+\\d{2}:\\d{2}:\\d{2}/);
+                const ratingIndex = block.findIndex(line => line === rating || line.includes(rating));
+                const timeIndex = block.findIndex(line => /20\\d{2}[/-]\\d{2}[/-]\\d{2}\\s+\\d{2}:\\d{2}:\\d{2}/.test(line));
+                const statusIndex = block.findIndex(line => /展示中|已处理|已处置/.test(line));
+                const productIdIndex = block.findIndex(line => /商品ID[:：]/.test(line));
+                const specIndex = block.findIndex(line => /规格ID[:：]/.test(line));
+                let commentText = '';
+                if (timeIndex >= 0) {
+                    const end = statusIndex > timeIndex ? statusIndex : (productIdIndex > timeIndex ? productIdIndex : block.length);
+                    commentText = block.slice(timeIndex + 1, end).join(' ').trim();
+                }
+                if (!commentText) commentText = '';
+                let productName = '';
+                if (productIdIndex > 0) {
+                    productName = block[productIdIndex - 1] || '';
+                    if (/展示中|已处理|已处置/.test(productName) && productIdIndex > 1) {
+                        productName = block[productIdIndex - 2] || '';
+                    }
+                }
+                let specName = '';
+                if (specIndex > 0) {
+                    specName = block[specIndex - 1] || '';
+                    if (/商品ID[:：]/.test(specName) && specIndex > 1) {
+                        specName = block[specIndex - 2] || '';
+                    }
+                }
+                const key = `${rating}|${orderMatch ? orderMatch[1] : ''}|${timeMatch ? timeMatch[0] : ''}|${productIdMatch ? productIdMatch[1] : ''}`;
+                if (seen.has(key)) continue;
+                seen.add(key);
+                rows.push({
+                    rating,
+                    order_id: orderMatch ? orderMatch[1] : '',
+                    comment_time: timeMatch ? timeMatch[0] : '',
+                    comment_text: commentText,
+                    product_name: productName,
+                    product_id: productIdMatch ? productIdMatch[1] : '',
+                    spec_name: specName,
+                    spec_id: specMatch ? specMatch[1] : '',
+                    raw_text: raw,
+                });
+            }
+            return rows;
+        }""",
+        rating,
+    )
+    return rows or []
+
+
+def _click_comment_next_page(page) -> bool:
+    try:
+        clicked = page.evaluate(
+            """() => {
+                window.scrollTo(0, document.body.scrollHeight);
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const disabled = (el) => {
+                    const signal = [
+                        String(el.className || ''),
+                        el.getAttribute('aria-disabled') || '',
+                        el.getAttribute('disabled') || ''
+                    ].join(' ').toLowerCase();
+                    return signal.includes('disabled') || signal.includes('true');
+                };
+                const all = Array.from(document.querySelectorAll('button, li, a, span, div')).filter(isVisible);
+                const classNext = all
+                    .filter(el => !disabled(el))
+                    .filter(el => {
+                        const cls = String(el.className || '').toLowerCase();
+                        const title = el.getAttribute('title') || '';
+                        const aria = el.getAttribute('aria-label') || '';
+                        return cls.includes('next') || title.includes('\\u4e0b\\u4e00\\u9875') || aria.includes('\\u4e0b\\u4e00\\u9875');
+                    })
+                    .filter(el => {
+                        const rect = el.getBoundingClientRect();
+                        return rect.top > window.innerHeight * 0.45 || rect.bottom > window.innerHeight - 260;
+                    })
+                    .sort((a, b) => b.getBoundingClientRect().right - a.getBoundingClientRect().right);
+                if (classNext.length) {
+                    classNext[0].click();
+                    return true;
+                }
+
+                const candidates = all
+                    .filter(isVisible)
+                    .filter(el => !disabled(el))
+                    .filter(el => {
+                        const text = (el.innerText || el.textContent || '').trim();
+                        const title = el.getAttribute('title') || '';
+                        const aria = el.getAttribute('aria-label') || '';
+                        return text === '>' || text === '\\u4e0b\\u4e00\\u9875' || title.includes('\\u4e0b\\u4e00\\u9875') || aria.includes('\\u4e0b\\u4e00\\u9875');
+                    })
+                    .filter(el => {
+                        const rect = el.getBoundingClientRect();
+                        return rect.top > window.innerHeight * 0.45 || rect.bottom > window.innerHeight - 260;
+                    })
+                    .sort((a, b) => b.getBoundingClientRect().right - a.getBoundingClientRect().right);
+                if (!candidates.length) return false;
+                candidates[0].click();
+                return true;
+            }"""
+        )
+        if clicked:
+            _wait_after_action(page, "next comment page", settle_ms=2800)
+            return True
+    except Exception as e:
+        print(f"[WARN] Comment next page click failed: {e}")
+    return False
+
+
+def _click_comment_page_number(page, page_number: int) -> bool:
+    try:
+        clicked = page.evaluate(
+            """(pageNumber) => {
+                window.scrollTo(0, document.body.scrollHeight);
+                const target = String(pageNumber);
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const disabled = (el) => {
+                    const signal = [
+                        String(el.className || ''),
+                        el.getAttribute('aria-disabled') || '',
+                        el.getAttribute('disabled') || ''
+                    ].join(' ').toLowerCase();
+                    return signal.includes('disabled') || signal.includes('true');
+                };
+                const candidates = Array.from(document.querySelectorAll('button, li, a, span, div'))
+                    .filter(isVisible)
+                    .filter(el => !disabled(el))
+                    .map(el => ({ el, text: (el.innerText || el.textContent || '').trim(), rect: el.getBoundingClientRect() }))
+                    .filter(item => item.text === target)
+                    .filter(item => item.rect.top > window.innerHeight * 0.45 || item.rect.bottom > window.innerHeight - 260)
+                    .sort((a, b) => (a.rect.width * a.rect.height) - (b.rect.width * b.rect.height));
+                if (!candidates.length) return false;
+                candidates[0].el.click();
+                return true;
+            }""",
+            page_number,
+        )
+        if clicked:
+            _wait_after_action(page, f"comment page {page_number}", settle_ms=2800)
+            return True
+    except Exception as e:
+        print(f"[WARN] Comment page number click failed ({page_number}): {e}")
+    return False
+
+
+def _jump_comment_page(page, page_number: int) -> bool:
+    try:
+        jumped = page.evaluate(
+            """(pageNumber) => {
+                window.scrollTo(0, document.body.scrollHeight);
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const inputs = Array.from(document.querySelectorAll('input'))
+                    .filter(isVisible)
+                    .map(el => ({ el, rect: el.getBoundingClientRect(), placeholder: el.placeholder || '', value: el.value || '' }))
+                    .filter(item => item.rect.top > window.innerHeight * 0.45 || item.rect.bottom > window.innerHeight - 260)
+                    .sort((a, b) => b.rect.left - a.rect.left);
+                const input = inputs[0] && inputs[0].el;
+                if (!input) return false;
+                input.focus();
+                input.select && input.select();
+                input.value = String(pageNumber);
+                input.dispatchEvent(new InputEvent('input', { bubbles: true, inputType: 'insertText', data: String(pageNumber) }));
+                input.dispatchEvent(new Event('change', { bubbles: true }));
+                input.dispatchEvent(new KeyboardEvent('keydown', { bubbles: true, key: 'Enter', code: 'Enter' }));
+                input.dispatchEvent(new KeyboardEvent('keypress', { bubbles: true, key: 'Enter', code: 'Enter' }));
+                input.dispatchEvent(new KeyboardEvent('keyup', { bubbles: true, key: 'Enter', code: 'Enter' }));
+                return true;
+            }""",
+            page_number,
+        )
+        if jumped:
+            _wait_after_action(page, f"jump comment page {page_number}", settle_ms=3200)
+            return True
+    except Exception as e:
+        print(f"[WARN] Jump comment page failed ({page_number}): {e}")
+    return False
+
+
+def _collect_comment_rating(page, rating: str) -> dict:
+    if not _click_comment_rating_tab(page, rating):
+        print(f"[WARN] Rating tab not selected: {rating}")
+    expected_count = _extract_comment_count_for_rating(page, rating)
+    expected_pages = _comment_expected_pages(expected_count)
+    if expected_pages:
+        print(f"[INFO] {rating} expected comments: {expected_count}, pages: {expected_pages} (20/page)")
+    else:
+        print(f"[WARN] {rating} expected comment count not detected; fallback to next-page scanning")
+    comments = []
+    seen_keys = set()
+    previous_signature = None
+    max_pages = min(COMMENT_MAX_PAGES, expected_pages or COMMENT_MAX_PAGES)
+
+    for page_number in range(1, max_pages + 1):
+        print(f"[INFO] Collecting {rating} comments page {page_number}...")
+        try:
+            page.wait_for_function(
+                "() => (document.body.innerText || '').includes('订单ID') || (document.body.innerText || '').includes('暂无数据')",
+                timeout=12000,
+            )
+        except Exception:
+            pass
+        rows = _extract_comment_rows(page, rating)
+        signature = "|".join(item.get("order_id", "") + item.get("comment_time", "") for item in rows[:5])
+        new_count = 0
+        for row in rows:
+            key = (row.get("rating"), row.get("order_id"), row.get("comment_time"), row.get("product_id"))
+            if key in seen_keys:
+                continue
+            seen_keys.add(key)
+            comments.append(row)
+            new_count += 1
+        print(f"[OK] {rating} page {page_number}: {len(rows)} rows, {new_count} new")
+        if not rows and not expected_pages:
+            break
+        if signature == previous_signature:
+            break
+        previous_signature = signature
+        if expected_count and len(comments) >= int(expected_count):
+            break
+        if expected_pages and page_number >= expected_pages:
+            break
+        next_page = page_number + 1
+        clicked_next = _click_comment_page_number(page, next_page)
+        if not clicked_next:
+            clicked_next = _click_comment_next_page(page)
+        if not clicked_next:
+            clicked_next = _jump_comment_page(page, next_page)
+        if not clicked_next:
+            print(f"[WARN] Failed to switch {rating} comments to page {next_page}")
+            break
+
+    return {
+        "rating": rating,
+        "expected_count": expected_count,
+        "expected_pages": expected_pages,
+        "collected_count": len(comments),
+        "comments": comments,
+    }
+
+
+def _collect_comment_summary(page) -> dict:
+    _open_comment_detail_page(page)
+    page = _active_page(page)
+    result = {
+        "shop": SHOP_NAME,
+        "period": _product_yesterday_label(),
+        "captured_at": datetime.now().isoformat(),
+        "url": page.url,
+        "ratings": {},
+    }
+    for rating in ["好评", "中评", "差评"]:
+        result["ratings"][rating] = _collect_comment_rating(page, rating)
+    result["total_collected_count"] = sum(item.get("collected_count", 0) for item in result["ratings"].values())
+    print(f"[OK] Comment summary collected: {result['total_collected_count']} comments")
+    return result
+
+
+def _save_comment_summary(comment_summary: dict) -> None:
+    if not comment_summary:
+        return
+    os.makedirs(PRODUCT_DATA_DIR, exist_ok=True)
+    date_stamp = _product_yesterday().strftime("%Y%m%d")
+    json_path = os.path.join(PRODUCT_DATA_DIR, f"评论汇总_{date_stamp}.json")
+    md_path = os.path.join(PRODUCT_DATA_DIR, f"评论汇总_{date_stamp}.md")
+
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(comment_summary, f, ensure_ascii=False, indent=2)
+
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write(f"# 评论汇总_{date_stamp}\n\n")
+        f.write(f"- **店铺**: {comment_summary.get('shop', '')}\n")
+        f.write(f"- **周期**: {comment_summary.get('period', '近7天')}\n")
+        f.write(f"- **采集时间**: {comment_summary.get('captured_at', '')}\n")
+        f.write(f"- **URL**: {comment_summary.get('url', '')}\n")
+        f.write(f"- **评论总数**: {comment_summary.get('total_collected_count', 0)}\n\n")
+        f.write("| 评价等级 | 真实数量 | 实际采集 |\n")
+        f.write("|----------|----------|----------|\n")
+        for rating, data in comment_summary.get("ratings", {}).items():
+            f.write(f"| {rating} | {data.get('expected_count', '')} | {data.get('collected_count', 0)} |\n")
+        for rating, data in comment_summary.get("ratings", {}).items():
+            f.write(f"\n## {rating}\n\n")
+            comments = data.get("comments", [])
+            if not comments:
+                f.write("- 未获取\n")
+                continue
+            f.write("| 序号 | 评论内容 | 商品ID | 商品名称 |\n")
+            f.write("|------|----------|--------|----------|\n")
+            for index, item in enumerate(comments, start=1):
+                comment_text = str(item.get("comment_text", "")).replace("|", "\\|").replace("\n", " ")
+                product_name = str(item.get("product_name", "")).replace("|", "\\|").replace("\n", " ")
+                f.write(
+                    f"| {index} | {comment_text} | {item.get('product_id', '')} | {product_name} |\n"
+                )
+
+    print(f"[OK] Comment summary files: {json_path} | {md_path}")
+
+
+def _get_selected_audience_crowd(page) -> str:
+    page = _active_page(page)
+    try:
+        return page.evaluate(
+            """() => {
+                const labels = ['商品曝光人群', '商品点击人群', '商品加购人群', '成交人群'];
+                const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const controls = Array.from(document.querySelectorAll('button, [role="button"], input, div, span'))
+                    .filter(isVisible)
+                    .map(el => {
+                        const rect = el.getBoundingClientRect();
+                        return {
+                            text: clean(el.innerText || el.textContent || el.value),
+                            top: rect.top,
+                            left: rect.left,
+                            width: rect.width,
+                            height: rect.height,
+                        };
+                    })
+                    .filter(item => labels.includes(item.text))
+                    .filter(item => item.top < 320 && item.left > window.innerWidth * 0.45)
+                    .sort((a, b) => {
+                        const aScore = Math.abs(a.top - 190) + Math.abs(a.left - window.innerWidth * 0.72);
+                        const bScore = Math.abs(b.top - 190) + Math.abs(b.left - window.innerWidth * 0.72);
+                        return aScore - bScore;
+                    });
+                return controls.length ? controls[0].text : '';
+            }"""
+        ) or ""
+    except Exception:
+        return ""
+
+
+def _select_deal_crowd(page) -> bool:
+    page = _active_page(page)
+    print("[INFO] Selecting deal crowd...")
+    last_error = ""
+    for attempt in range(1, 4):
+        try:
+            selected_text = _get_selected_audience_crowd(page)
+            if selected_text == "成交人群":
+                print("[INFO] Deal crowd already selected")
+                return True
+
+            control_rect = page.evaluate(
+                """() => {
+                    const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                    const isVisible = (el) => {
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style && style.visibility !== 'hidden' && style.display !== 'none'
+                            && rect.width > 0 && rect.height > 0;
+                    };
+                    const labels = ['商品曝光人群', '商品点击人群', '商品加购人群', '成交人群'];
+                    const candidates = Array.from(document.querySelectorAll('button, [role="button"], input, div, span'))
+                        .filter(isVisible)
+                        .filter(el => labels.includes(clean(el.innerText || el.textContent || el.value)))
+                        .filter(el => {
+                            const rect = el.getBoundingClientRect();
+                            return rect.top < 320 && rect.left > window.innerWidth * 0.45;
+                        })
+                        .sort((a, b) => {
+                            const ar = a.getBoundingClientRect();
+                            const br = b.getBoundingClientRect();
+                            const aScore = Math.abs(ar.top - 190) + Math.abs(ar.left - window.innerWidth * 0.72) + ar.width * ar.height / 10000;
+                            const bScore = Math.abs(br.top - 190) + Math.abs(br.left - window.innerWidth * 0.72) + br.width * br.height / 10000;
+                            return aScore - bScore;
+                        });
+                    if (!candidates.length) return null;
+                    const rect = candidates[0].getBoundingClientRect();
+                    return { x: rect.left + rect.width / 2, y: rect.top + rect.height / 2, left: rect.left, top: rect.top };
+                }"""
+            )
+            if not control_rect:
+                raise RuntimeError(f"crowd dropdown control not found (selected={selected_text})")
+            page.mouse.click(control_rect["x"], control_rect["y"])
+            page.wait_for_timeout(800)
+            option_rect = page.evaluate(
+                """(control) => {
+                    const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                    const isVisible = (el) => {
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style && style.visibility !== 'hidden' && style.display !== 'none'
+                            && rect.width > 0 && rect.height > 0;
+                    };
+                    const candidates = Array.from(document.querySelectorAll('li, [role="option"], [role="menuitem"], div, span'))
+                        .filter(isVisible)
+                        .filter(el => clean(el.innerText || el.textContent) === '成交人群')
+                        .filter(el => {
+                            const rect = el.getBoundingClientRect();
+                            return rect.top >= control.top && rect.left > window.innerWidth * 0.5;
+                        })
+                        .sort((a, b) => {
+                            const ar = a.getBoundingClientRect();
+                            const br = b.getBoundingClientRect();
+                            return (ar.top - br.top) || (ar.left - br.left);
+                        });
+                    if (!candidates.length) return null;
+                    const rect = candidates[0].getBoundingClientRect();
+                    return { x: rect.left + rect.width / 2, y: rect.top + rect.height / 2 };
+                }""",
+                control_rect,
+            )
+            if not option_rect:
+                raise RuntimeError("deal crowd option not visible in dropdown")
+            page.mouse.click(option_rect["x"], option_rect["y"])
+            _wait_after_action(page, f"select deal crowd attempt {attempt}", settle_ms=2500)
+            selected_text = _get_selected_audience_crowd(page)
+            if selected_text == "成交人群":
+                print("[OK] Deal crowd selected")
+                return True
+            last_error = f"selection not confirmed; current={selected_text or 'unknown'}"
+        except Exception as e:
+            last_error = str(e)
+            print(f"[WARN] Deal crowd select attempt {attempt} failed: {e}")
+            try:
+                page.keyboard.press("Escape")
+            except Exception:
+                pass
+            page.wait_for_timeout(600)
+
+    print(f"[WARN] Deal crowd selection failed: {last_error}")
+    _dump_product_debug(page, "deal_crowd_not_selected")
+    return False
+
+
+def _wait_for_deal_crowd_data(page, timeout=12000) -> bool:
+    page = _active_page(page)
+    try:
+        page.wait_for_function(
+            """() => {
+                const text = document.body ? (document.body.innerText || '') : '';
+                return text.includes('成交人群画像') || text.includes('purchase_crowd')
+                    || location.href.includes('purchase_crowd');
+            }""",
+            timeout=timeout,
+        )
+        return True
+    except Exception as e:
+        print(f"[WARN] Deal crowd data not confirmed: {e}")
+        return False
+
+
+def _attach_audience_response_collector(page) -> None:
+    page = _active_page(page)
+    page_id = id(page)
+    if page_id in _AUDIENCE_COLLECTOR_PAGES:
+        return
+    _AUDIENCE_COLLECTOR_PAGES.add(page_id)
+
+    def handle_response(response):
+        try:
+            url = response.url
+            if "/business_api/shop/crowd" not in url and "/business_api/shop/crowd_insight" not in url:
+                return
+            data = response.json()
+            AUDIENCE_API_PAYLOADS.append({"url": url, "data": data, "source": "response"})
+        except Exception:
+            return
+
+    try:
+        page.on("response", handle_response)
+        print("[INFO] Audience response collector attached")
+    except Exception as e:
+        print(f"[WARN] Audience response collector attach failed: {e}")
+
+
+def _extract_chart_options(page) -> list[dict]:
+    try:
+        return page.evaluate(
+            """() => {
+                const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const charts = [];
+                const seen = new Set();
+                function pushOption(el, chart) {
+                    if (!chart || seen.has(chart)) return;
+                    seen.add(chart);
+                    let option = {};
+                    try {
+                        option = chart.getOption ? chart.getOption() : {};
+                    } catch (e) {
+                        option = {};
+                    }
+                    const rect = el.getBoundingClientRect();
+                    let section = '';
+                    let cur = el;
+                    for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                        const text = clean(cur.innerText || cur.textContent);
+                        if (text.includes('性别分布') || text.includes('年龄分布')
+                            || text.includes('八大策略人群分布') || text.includes('常驻省份')) {
+                            section = text.slice(0, 300);
+                            break;
+                        }
+                    }
+                    charts.push({ rect: { top: rect.top, left: rect.left, width: rect.width, height: rect.height }, section, option });
+                }
+                if (window.echarts && window.echarts.getInstanceByDom) {
+                    document.querySelectorAll('div, canvas').forEach(el => {
+                        const chart = window.echarts.getInstanceByDom(el);
+                        pushOption(el, chart);
+                    });
+                }
+                document.querySelectorAll('[_echarts_instance_], [data-zr-dom-id], canvas').forEach(el => {
+                    let cur = el;
+                    for (let depth = 0; cur && depth < 4; depth++, cur = cur.parentElement) {
+                        const key = cur.getAttribute && cur.getAttribute('_echarts_instance_');
+                        if (key && window.echarts && window.echarts.getInstanceByDom) {
+                            pushOption(cur, window.echarts.getInstanceByDom(cur));
+                        }
+                    }
+                });
+                return charts;
+            }"""
+        )
+    except Exception as e:
+        print(f"[WARN] Extract chart options failed: {e}")
+        return []
+
+
+def _extract_audience_api_payloads(page) -> list[dict]:
+    print("[INFO] Extracting audience API payloads...")
+    payloads = list(AUDIENCE_API_PAYLOADS)
+    try:
+        payloads.extend(page.evaluate(
+            """async () => {
+                const entries = performance.getEntriesByType('resource')
+                    .map(entry => entry.name)
+                    .filter(Boolean);
+                const candidates = Array.from(new Set(entries)).filter(url => {
+                    const value = url.toLowerCase();
+                    return (value.includes('crowd') || value.includes('portrait') || value.includes('user')
+                        || value.includes('analysis') || value.includes('profile'))
+                        && !value.includes('.js') && !value.includes('.css') && !value.includes('.png')
+                        && !value.includes('.jpg') && !value.includes('.gif') && !value.includes('.svg');
+                }).slice(-40);
+                const payloads = [];
+                for (const url of candidates) {
+                    try {
+                        const response = await fetch(url, { credentials: 'include' });
+                        const contentType = response.headers.get('content-type') || '';
+                        if (!contentType.includes('json')) {
+                            continue;
+                        }
+                        const data = await response.json();
+                        payloads.push({ url, data });
+                    } catch (e) {
+                        payloads.push({ url, error: String(e && e.message || e) });
+                    }
+                }
+                return payloads;
+            }""",
+        ))
+    except Exception as e:
+        print(f"[WARN] Audience API payload extraction failed: {e}")
+    deduped = []
+    seen = set()
+    for payload in payloads:
+        key = payload.get("url") or json.dumps(payload.get("data"), ensure_ascii=False, default=str)[:500]
+        if key in seen:
+            continue
+        seen.add(key)
+        deduped.append(payload)
+    return deduped
+
+
+def _flatten_json_items(value, path=""):
+    items = []
+    if isinstance(value, dict):
+        for key, child in value.items():
+            child_path = f"{path}.{key}" if path else str(key)
+            items.extend(_flatten_json_items(child, child_path))
+    elif isinstance(value, list):
+        for index, child in enumerate(value):
+            child_path = f"{path}[{index}]"
+            items.extend(_flatten_json_items(child, child_path))
+    else:
+        items.append((path, value))
+    return items
+
+
+def _find_named_values(value) -> list[dict]:
+    rows = []
+    if isinstance(value, dict):
+        name = None
+        val = None
+        tgi = None
+        for key, item in value.items():
+            key_text = str(key).lower()
+            if key_text in {
+                "name", "label", "title", "crowdname", "crowd_name", "province", "provincename",
+                "province_name", "region", "key", "text", "category", "type_name", "type",
+                "dim_name", "dimension_name", "group_name",
+            }:
+                name = item
+            if key_text in {
+                "value", "ratio", "percent", "percentage", "占比", "rate", "uv_ratio",
+                "user_ratio", "cnt_ratio", "count_ratio", "proportion", "占比",
+            }:
+                val = item
+            if key_text in {"tgi", "tgi_value"}:
+                tgi = item
+        if name in (None, ""):
+            for key, item in value.items():
+                if isinstance(item, str) and item.strip() and len(item.strip()) <= 20:
+                    if any(token in item for token in (
+                        "男", "女", "岁", "白领", "中产", "广东", "浙江",
+                        "小镇中老年", "都市蓝领", "新锐白领", "精致妈妈",
+                        "都市银发", "Z世代", "小镇青年", "资深中产",
+                    )):
+                        name = item
+                        break
+        if val in (None, ""):
+            for key, item in value.items():
+                if isinstance(item, (int, float)) and 0 <= float(item) <= 100:
+                    val = item
+                    break
+        if name not in (None, "") and val not in (None, ""):
+            row = {"name": str(name), "value": val}
+            if tgi not in (None, ""):
+                row["tgi"] = tgi
+            rows.append(row)
+        for child in value.values():
+            rows.extend(_find_named_values(child))
+    elif isinstance(value, list):
+        for child in value:
+            rows.extend(_find_named_values(child))
+    return rows
+
+
+def _normalize_percent(value):
+    if isinstance(value, (int, float)):
+        number = float(value)
+        if 0 <= number <= 1:
+            return f"{number * 100:.2f}%"
+        text = f"{number:.2f}".rstrip("0").rstrip(".")
+        return f"{text}%"
+    text = str(value).strip()
+    return text if text.endswith("%") else text
+
+
+def _classify_named_rows(rows: list[dict]) -> dict:
+    result = {
+        "gender_distribution": [],
+        "age_distribution": [],
+        "strategy_crowd_distribution": [],
+        "resident_province": [],
+    }
+    genders = {"男", "女", "男性", "女性"}
+    age_markers = ("18", "24", "25", "30", "31", "35", "36", "40", "41", "45", "46", "50", "岁")
+    strategy_names = {"小镇中老年", "都市蓝领", "新锐白领", "精致妈妈", "都市银发", "Z世代", "小镇青年", "资深中产"}
+    provinces = {"北京", "上海", "天津", "重庆", "河北", "山西", "辽宁", "吉林", "黑龙江", "江苏", "浙江", "安徽", "福建", "江西", "山东", "河南", "湖北", "湖南", "广东", "海南", "四川", "贵州", "云南", "陕西", "甘肃", "青海", "台湾", "内蒙古", "广西", "西藏", "宁夏", "新疆", "香港", "澳门"}
+
+    seen = {key: set() for key in result}
+    for row in rows:
+        name = str(row.get("name", "")).strip()
+        if not name:
+            continue
+        item = {"name": name, "value": _normalize_percent(row.get("value", ""))}
+        if "tgi" in row:
+            item["tgi"] = row["tgi"]
+        key = ""
+        if name in genders:
+            key = "gender_distribution"
+        elif any(marker in name for marker in age_markers):
+            key = "age_distribution"
+        elif name in strategy_names:
+            key = "strategy_crowd_distribution"
+        elif name in provinces:
+            key = "resident_province"
+        if key and name not in seen[key]:
+            seen[key].add(name)
+            result[key].append(item)
+    return result
+
+
+def _merge_profile_distributions(result: dict, classified: dict) -> None:
+    for key, rows in classified.items():
+        if rows and len(rows) > len(result.get(key) or []):
+            result[key] = rows
+
+
+def _extract_province_rows_from_text(text: str) -> list[dict]:
+    provinces = [
+        "北京", "上海", "天津", "重庆", "河北", "山西", "辽宁", "吉林", "黑龙江", "江苏",
+        "浙江", "安徽", "福建", "江西", "山东", "河南", "湖北", "湖南", "广东", "海南",
+        "四川", "贵州", "云南", "陕西", "甘肃", "青海", "台湾", "内蒙古", "广西", "西藏",
+        "宁夏", "新疆", "香港", "澳门",
+    ]
+    normalized = re.sub(r"\s+", " ", text or "")
+    if "常驻省份" in normalized:
+        normalized = normalized.split("常驻省份", 1)[1]
+    stop_markers = ["购买偏好", "内容偏好", "下载结果", "对比人群"]
+    for marker in stop_markers:
+        if marker in normalized:
+            normalized = normalized.split(marker, 1)[0]
+
+    found_matches = []
+    for province in provinces:
+        match = re.search(rf"(?:^|\s|\d){re.escape(province)}(?:\s|$)", normalized)
+        if match:
+            found_matches.append((match.start(), province))
+    found_names = [name for _, name in sorted(found_matches)]
+    if not found_names:
+        return []
+
+    percentages = re.findall(r"(\d+(?:\.\d+)?)%", normalized)
+    tgi_values = re.findall(r"(?<![\d.])(\d+\.\d+)(?!\s*%)", normalized)
+    rows = []
+    for index, name in enumerate(found_names[:10]):
+        row = {"name": name}
+        if index < len(percentages):
+            row["value"] = f"{percentages[index]}%"
+        if index < len(tgi_values):
+            row["tgi"] = tgi_values[index]
+        rows.append(row)
+    return rows
+
+
+def _series_to_pairs(series: dict) -> list[dict]:
+    data = series.get("data") or []
+    pairs = []
+    for item in data:
+        if isinstance(item, dict):
+            name = item.get("name", "")
+            value = item.get("value", "")
+        else:
+            name = ""
+            value = item
+        pairs.append({"name": name, "value": value})
+    return pairs
+
+
+def _axis_series_to_pairs(option: dict, series: dict) -> list[dict]:
+    x_axis = option.get("xAxis") or []
+    y_axis = option.get("yAxis") or []
+    axis = x_axis[0] if x_axis else (y_axis[0] if y_axis else {})
+    labels = axis.get("data") or []
+    values = series.get("data") or []
+    pairs = []
+    for index, value in enumerate(values):
+        if isinstance(value, dict):
+            val = value.get("value", "")
+            name = value.get("name") or (labels[index] if index < len(labels) else "")
+        else:
+            val = value
+            name = labels[index] if index < len(labels) else ""
+        pairs.append({"name": name, "value": val})
+    return pairs
+
+
+def _classify_profile_chart(chart: dict) -> str:
+    text = json.dumps(chart.get("option", {}), ensure_ascii=False) + " " + chart.get("section", "")
+    if "男性" in text or "女性" in text:
+        return "gender_distribution"
+    if "18-24" in text or "25-30" in text or "50岁以上" in text:
+        return "age_distribution"
+    if "小镇中老年" in text or "新锐白领" in text or "Z世代" in text:
+        return "strategy_crowd_distribution"
+    if "广东" in text or "浙江" in text or "江苏" in text or "常驻省份" in text:
+        return "resident_province"
+    return ""
+
+
+def _extract_audience_profile(page) -> dict:
+    page = _active_page(page)
+    print("[INFO] Extracting audience profile charts...")
+    charts = _extract_chart_options(page)
+    api_payloads = _extract_audience_api_payloads(page)
+    selected_crowd = _get_selected_audience_crowd(page)
+    api_crowds = sorted({
+        m.group(1)
+        for payload in api_payloads
+        for m in [re.search(r"[?&]crowd=([^&]+)", payload.get("url") or "")]
+        if m
+    })
+    if selected_crowd == "成交人群" and "purchase_crowd" in api_crowds:
+        api_payloads = [
+            payload
+            for payload in api_payloads
+            if "crowd=purchase_crowd" in (payload.get("url") or "")
+        ]
+        api_crowds = ["purchase_crowd"]
+    result = {
+        "captured_at": datetime.now().isoformat(),
+        "url": page.url,
+        "crowd": selected_crowd or ("成交人群" if "purchase_crowd" in api_crowds else ""),
+        "crowd_api_keys": api_crowds,
+        "period": _product_yesterday_label(),
+        "gender_distribution": [],
+        "age_distribution": [],
+        "strategy_crowd_distribution": [],
+        "resident_province": [],
+        "raw_charts_count": len(charts),
+        "raw_api_payloads_count": len(api_payloads),
+        "raw_api_payload_urls": [payload.get("url") for payload in api_payloads if payload.get("url")],
+    }
+
+    for chart in charts:
+        option = chart.get("option") or {}
+        series_list = option.get("series") or []
+        if not series_list:
+            continue
+        key = _classify_profile_chart(chart)
+        if not key:
+            continue
+        series = series_list[0]
+        if series.get("type") == "pie":
+            pairs = _series_to_pairs(series)
+        else:
+            pairs = _axis_series_to_pairs(option, series)
+        pairs = [item for item in pairs if item.get("name") not in (None, "")]
+        if pairs and not result.get(key):
+            result[key] = pairs
+
+    named_rows = []
+    for payload in api_payloads:
+        named_rows.extend(_find_named_values(payload.get("data")))
+    result["raw_named_values_count"] = len(named_rows)
+    _merge_profile_distributions(result, _classify_named_rows(named_rows))
+
+    try:
+        text_fallback = page.evaluate(
+            """() => (document.body.innerText || '').replace(/\\s+/g, ' ').trim()"""
+        )
+        result["raw_text_excerpt"] = text_fallback[:2000]
+        province_rows = _extract_province_rows_from_text(text_fallback)
+        if province_rows and len(province_rows) > len(result.get("resident_province") or []):
+            result["resident_province"] = province_rows
+    except Exception:
+        pass
+
+    print(
+        "[OK] Audience profile extracted: "
+        f"gender={len(result['gender_distribution'])}, "
+        f"age={len(result['age_distribution'])}, "
+        f"strategy={len(result['strategy_crowd_distribution'])}, "
+        f"province={len(result['resident_province'])}"
+    )
+    return result
+
+
+def _collect_product_audience_profiles(page, products: list[dict], product_dirs: dict[str, str] | None = None) -> None:
+    if not products:
+        return
+    target_products = products[:PRODUCT_AUDIENCE_LIMIT] if PRODUCT_AUDIENCE_LIMIT > 0 else products
+    print(f"[INFO] Collecting audience profiles for {len(target_products)} products...")
+
+    _open_product_list(page)
+    _ensure_product_list_tab(page)
+    product_list_url = page.url
+
+    for index, product in enumerate(target_products, start=1):
+        product_id = product.get("product_id")
+        if not product_id:
+            product["audience_profile_error"] = "missing product_id"
+            continue
+
+        print(f"[INFO] Audience profile {index}/{len(target_products)}: {product_id}")
+        pages_before = []
+        try:
+            AUDIENCE_API_PAYLOADS.clear()
+            page = _ensure_product_list_page(page, product_list_url)
+            pages_before = _alive_context_pages(page)
+            if not _search_product_by_id(page, product_id):
+                product["audience_profile_error"] = "search failed"
+                continue
+            if not _click_product_detail(page, product_id):
+                product["audience_profile_error"] = "detail not opened"
+                continue
+            page = _active_page(page)
+            try:
+                product["product_experience_analysis"] = _collect_product_experience_analysis(page)
+            except Exception as experience_error:
+                print(f"[WARN] Experience analysis collection failed ({product_id}): {experience_error}")
+                product["product_experience_analysis_error"] = str(experience_error)
+
+            _attach_audience_response_collector(page)
+            if not _open_product_audience_profile(page):
+                product["audience_profile_error"] = "audience profile not opened"
+                continue
+            page = _active_page(page)
+            _attach_audience_response_collector(page)
+            AUDIENCE_API_PAYLOADS.clear()
+            if not _select_deal_crowd(page):
+                product["audience_profile_error"] = "deal crowd not selected"
+                continue
+            _wait_for_deal_crowd_data(page)
+            page = _active_page(page)
+            product["audience_profile"] = _extract_audience_profile(page)
+            screenshot_path = _save_audience_profile_screenshot(page, product_id, (product_dirs or {}).get(product_id))
+            if screenshot_path:
+                product["audience_profile"]["screenshot_path"] = screenshot_path
+        except Exception as e:
+            print(f"[WARN] Audience profile collection failed ({product_id}): {e}")
+            product["audience_profile_error"] = str(e)
+        finally:
+            page = _cleanup_product_detail_pages(page, product_list_url, pages_before)
+
+
+def _write_distribution_table(f, title: str, rows: list[dict]) -> None:
+    f.write(f"\n### {title}\n\n")
+    if not rows:
+        f.write("- 未获取\n")
+        return
+    f.write("| 项目 | 数值 |\n")
+    f.write("|------|------|\n")
+    for item in rows:
+        f.write(f"| {item.get('name', '')} | {item.get('value', '')} |\n")
+
+
+def _write_product_raw_metrics_table(f, product: dict) -> None:
+    f.write("## 核心指标\n\n")
+    f.write("| 指标 | 数值 | 变化方向 | 变化百分比 |\n")
+    f.write("|------|------|----------|----------|\n")
+    alias_keys = {"销量", "访客数", "转化率", "加购率", "退款率"}
+    for key, metric in (product.get("metrics") or {}).items():
+        if key in alias_keys:
+            continue
+        value, direction, change = _metric_display_parts(metric)
+        f.write(
+            f"| {key} | {value} | {direction} | {change} |\n"
+        )
+
+
+def _metric_display_parts(metric: dict) -> tuple[str, str, str]:
+    value = str(metric.get("value", "") or "").strip()
+    direction = str(metric.get("change_direction", "") or "").strip()
+    change = str(metric.get("change", "") or "").strip()
+    if not direction and "持平" in value:
+        direction = "持平"
+        value = value.replace("持平", "").strip()
+    if not direction and re.search(r"\s-\s*$", value):
+        direction = "-"
+        value = re.sub(r"\s-\s*$", "", value).strip()
+    return value, direction, change
+
+
+def _ensure_product_base_metrics(product: dict) -> None:
+    metrics = product.setdefault("metrics", {})
+    columns = product.get("columns") or {}
+    alias_sources = {
+        "成交金额": "成交金额",
+        "销量": "成交订单数",
+        "访客数": "商品点击人数",
+        "转化率": "点击支付率（人数）",
+        "加购率": "访问加购转化率",
+        "退款率": "退款率(支付时间)",
+    }
+    fallback_sources = {
+        "转化率": ["点击支付率", "点击支付率（人数）"],
+        "退款率": ["退款率（支付时间）", "退款率"],
+    }
+    for alias, source in alias_sources.items():
+        metric = metrics.get(alias) or {}
+        if metric.get("value") and product.get(alias):
+            continue
+        raw_value = ""
+        source_metric = metrics.get(source) or {}
+        if source_metric.get("value"):
+            raw_value = source_metric.get("raw") or source_metric.get("value")
+            metric = dict(source_metric)
+        if not raw_value:
+            raw_value = columns.get(source, "")
+            if not raw_value:
+                for fallback in fallback_sources.get(alias, []):
+                    raw_value = columns.get(fallback, "")
+                    if raw_value:
+                        break
+            metric = _parse_metric_cell(raw_value)
+        metrics[alias] = metric
+        product[alias] = metric.get("value", "")
+        product[f"{alias}_变化百分比"] = metric.get("change", "")
+        product[f"{alias}_变化方向"] = metric.get("change_direction", "")
+
+
+def _parse_metric_number(value) -> float:
+    text = str(value or "").strip()
+    if not text or text in {"-", "--", "未获取"}:
+        return 0.0
+    multiplier = 1.0
+    if "万" in text:
+        multiplier = 10000.0
+    elif "亿" in text:
+        multiplier = 100000000.0
+    text = (
+        text.replace(",", "")
+        .replace("¥", "")
+        .replace("￥", "")
+        .replace("%", "")
+        .replace("万", "")
+        .replace("亿", "")
+        .strip()
+    )
+    match = re.search(r"-?\d+(?:\.\d+)?", text)
+    if not match:
+        return 0.0
+    return float(match.group(0)) * multiplier
+
+
+def _format_count(value: float) -> str:
+    if abs(value - round(value)) < 1e-9:
+        return f"{int(round(value)):,}"
+    return f"{value:,.2f}"
+
+
+def _format_rate(numerator: float, denominator: float) -> str:
+    if not denominator:
+        return "0.00%"
+    return f"{numerator / denominator * 100:.2f}%"
+
+
+def _product_output_dir(style_dir: str, product: dict, date_stamp: str) -> str:
+    product_name = _safe_filename(product.get("product_name"), limit=70)
+    product_id = product.get("product_id") or "unknown"
+    return os.path.join(style_dir, f"{product_name}_{product_id}_{date_stamp}")
+
+
+def _collect_products_by_ids(page, product_ids: list[str], product_root_dir: str | None = None, date_stamp: str | None = None) -> list[dict]:
+    products = []
+    details_collected = 0
+    _open_product_list(page)
+    _ensure_product_list_tab(page)
+    product_list_url = page.url
+    for index, product_id in enumerate(product_ids, start=1):
+        print(f"[INFO] Collecting product {index}/{len(product_ids)}: {product_id}")
+        page = _ensure_product_list_page(page, product_list_url)
+        if not _search_product_by_id(page, product_id):
+            print(f"[INFO] Retrying product search from a fresh list page: {product_id}")
+            page = _reset_product_list_page(page)
+            product_list_url = page.url
+            if not _search_product_by_id(page, product_id):
+                products.append({
+                    "product_id": product_id,
+                    "product_name": product_id,
+                    "date_range": _product_yesterday_label(),
+                    "captured_at": datetime.now().isoformat(),
+                    "base_data_error": "search failed",
+                })
+                continue
+        rows = _extract_products_on_page(page)
+        product = next((row for row in rows if row.get("product_id") == product_id), None)
+        if not product:
+            print(f"[INFO] Retrying product row extraction from a fresh list page: {product_id}")
+            page = _reset_product_list_page(page)
+            product_list_url = page.url
+            if _search_product_by_id(page, product_id):
+                rows = _extract_products_on_page(page)
+                product = next((row for row in rows if row.get("product_id") == product_id), None)
+        if not product:
+            products.append({
+                "product_id": product_id,
+                "product_name": product_id,
+                "date_range": _product_yesterday_label(),
+                "captured_at": datetime.now().isoformat(),
+                "base_data_error": "searched row not found",
+            })
+            continue
+        _ensure_product_base_metrics(product)
+        product_dir = _product_output_dir(product_root_dir, product, date_stamp) if product_root_dir and date_stamp else None
+        if PRODUCT_AUDIENCE_LIMIT <= 0 or details_collected < PRODUCT_AUDIENCE_LIMIT:
+            page = _collect_current_product_details(page, product, product_list_url, product_dir)
+            details_collected += 1
+        else:
+            product["audience_profile_error"] = "skipped by PRODUCT_AUDIENCE_LIMIT"
+        if product_dir and date_stamp:
+            _save_single_product_files(product, product_dir, date_stamp)
+        products.append(product)
+    return products
+
+
+def _collect_current_product_details(page, product: dict, product_list_url: str, product_dir: str | None = None):
+    product_id = product.get("product_id")
+    if not product_id:
+        product["audience_profile_error"] = "missing product_id"
+        return page
+
+    print(f"[INFO] Collecting detail data for product: {product_id}")
+    pages_before = []
+    try:
+        AUDIENCE_API_PAYLOADS.clear()
+        pages_before = _alive_context_pages(page)
+        if not _click_product_detail(page, product_id):
+            product["audience_profile_error"] = "detail not opened"
+            return page
+
+        page = _active_page(page)
+        try:
+            product["product_experience_analysis"] = _collect_product_experience_analysis(page)
+        except Exception as experience_error:
+            print(f"[WARN] Experience analysis collection failed ({product_id}): {experience_error}")
+            product["product_experience_analysis_error"] = str(experience_error)
+
+        _attach_audience_response_collector(page)
+        if not _open_product_audience_profile(page):
+            product["audience_profile_error"] = "audience profile not opened"
+            return page
+
+        page = _active_page(page)
+        if _is_empty_audience_state(page):
+            product["audience_profile_error"] = "empty audience data"
+            return page
+        _attach_audience_response_collector(page)
+        AUDIENCE_API_PAYLOADS.clear()
+        if not _select_deal_crowd(page):
+            if _is_empty_audience_state(page):
+                product["audience_profile_error"] = "empty audience data"
+                return page
+            product["audience_profile_error"] = "deal crowd not selected"
+            return page
+
+        _wait_for_deal_crowd_data(page)
+        page = _active_page(page)
+        if _is_empty_audience_state(page):
+            product["audience_profile_error"] = "empty audience data"
+            return page
+        product["audience_profile"] = _extract_audience_profile(page)
+        screenshot_path = _save_audience_profile_screenshot(page, product_id, product_dir)
+        if screenshot_path:
+            product["audience_profile"]["screenshot_path"] = screenshot_path
+    except Exception as e:
+        print(f"[WARN] Product detail collection failed ({product_id}): {e}")
+        product["audience_profile_error"] = str(e)
+    finally:
+        page = _cleanup_product_detail_pages(page, product_list_url, pages_before)
+    return page
+
+
+def _reset_product_list_page(page):
+    page = _active_page(page)
+    try:
+        page.goto(PRODUCT_LIST_URL, wait_until="domcontentloaded", timeout=60000)
+        try:
+            page.wait_for_load_state("load", timeout=15000)
+        except Exception:
+            pass
+        _wait_after_action(page, "reset product list", settle_ms=1800)
+        _ensure_product_list_tab(page)
+    except Exception as e:
+        print(f"[WARN] Product list reset navigation failed: {e}")
+        _open_product_list(page)
+    return _active_page(page)
+
+
+def _build_style_summary(style_name: str, product_ids: list[str], products: list[dict], style_dir: str) -> dict:
+    for product in products:
+        _ensure_product_base_metrics(product)
+    total_visitors = sum(_parse_metric_number(product.get("访客数")) for product in products)
+    total_cart = sum(_parse_metric_number((product.get("metrics") or {}).get("加购人数", {}).get("value")) for product in products)
+    total_amount = sum(_parse_metric_number(product.get("成交金额")) for product in products)
+    erp_metrics = apply_erp_override("dy", style_name, total_visitors, total_cart, target_date=_product_yesterday().strftime("%Y-%m-%d"))
+    total_sales = erp_metrics["销量"]
+    total_refund_orders = erp_metrics["退款订单数"]
+    screenshots = [
+        (product.get("audience_profile") or {}).get("screenshot_path")
+        for product in products
+        if (product.get("audience_profile") or {}).get("screenshot_path")
+    ]
+    return {
+        "style_name": style_name,
+        "date_range": _product_yesterday_label(),
+        "captured_at": datetime.now().isoformat(),
+        "product_ids": product_ids,
+        "output_dir": style_dir,
+        "成交金额": f"¥{total_amount:,.2f}",
+        "销量": format_count(total_sales),
+        "访客数": _format_count(total_visitors),
+        "转化率": erp_metrics["转化率"],
+        "加购率": erp_metrics["加购率"],
+        "退款率": erp_metrics["退款率"],
+        "raw_totals": {
+            "成交金额": total_amount,
+            "销量": total_sales,
+            "访客数": total_visitors,
+            "加购人数": total_cart,
+            "退款订单数": total_refund_orders,
+        },
+        "ERP销量退款覆盖": erp_metrics["erp_override"],
+        "audience_screenshots": screenshots,
+        "products": products,
+    }
+
+
+def _save_single_product_files(product: dict, output_dir: str, date_stamp: str) -> None:
+    os.makedirs(output_dir, exist_ok=True)
+    _ensure_product_base_metrics(product)
+    product["output_dir"] = output_dir
+    base = f"{_safe_filename(product.get('product_name'))}_{product.get('product_id', 'unknown')}_{date_stamp}"
+    json_path = os.path.join(output_dir, f"{base}.json")
+    md_path = os.path.join(output_dir, f"{base}.md")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(product, f, ensure_ascii=False, indent=2)
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write(f"# {product.get('product_name', '未命名商品')}\n\n")
+        f.write(f"- **商品ID**: {product.get('product_id', '')}\n")
+        f.write(f"- **周期**: {product.get('date_range', '')}\n")
+        f.write(f"- **采集时间**: {product.get('captured_at', '')}\n\n")
+        _write_product_raw_metrics_table(f, product)
+        if product.get("product_experience_analysis"):
+            f.write("\n## 退款及体验-体验分析\n\n")
+            for key in ["商品品质退货率", "评价好评率"]:
+                metric = product.get("product_experience_analysis", {}).get(key) or {}
+                f.write(f"- **{key}**: {metric.get('value', '')} {metric.get('change_direction', '')} {metric.get('change', '')}\n")
+        if product.get("audience_profile"):
+            profile = product["audience_profile"]
+            f.write("\n## 成交人群画像\n\n")
+            f.write(f"- **周期**: {profile.get('period', '')}\n")
+            f.write(f"- **截图**: {profile.get('screenshot_path', '')}\n")
+            _write_distribution_table(f, "性别分布", profile.get("gender_distribution", []))
+            _write_distribution_table(f, "年龄分布", profile.get("age_distribution", []))
+            _write_distribution_table(f, "八大策略人群分布", profile.get("strategy_crowd_distribution", []))
+            _write_distribution_table(f, "常驻省份", profile.get("resident_province", []))
+        elif product.get("audience_profile_error"):
+            f.write("\n## 成交人群画像\n\n")
+            f.write(f"- **采集失败**: {product.get('audience_profile_error')}\n")
+    print(f"[OK] Product ID files: {json_path} | {md_path}")
+
+
+def _save_style_summary(summary: dict, style_dir: str, date_stamp: str) -> None:
+    os.makedirs(style_dir, exist_ok=True)
+    base = f"{_safe_filename(summary.get('style_name'))}_{date_stamp}"
+    json_path = os.path.join(style_dir, f"{base}.json")
+    md_path = os.path.join(style_dir, f"{base}.md")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(summary, f, ensure_ascii=False, indent=2)
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write(f"# {summary.get('style_name')}\n\n")
+        f.write(f"- **周期**: {summary.get('date_range', '')}\n")
+        f.write(f"- **采集时间**: {summary.get('captured_at', '')}\n")
+        f.write(f"- **商品ID数**: {len(summary.get('product_ids', []))}\n\n")
+        f.write("## 合并指标\n\n")
+        f.write("| 指标 | 数值 |\n")
+        f.write("|------|------|\n")
+        for key in ["成交金额", "销量", "访客数", "转化率", "加购率", "退款率"]:
+            f.write(f"| {key} | {summary.get(key, '')} |\n")
+        f.write("\n## 人物画像截图\n\n")
+        for screenshot in summary.get("audience_screenshots", []):
+            f.write(f"- {screenshot}\n")
+        f.write("\n## 单品明细\n\n")
+        f.write("| 商品ID | 商品名 | 销量 | 访客数 | 转化率 | 加购率 | 退款率 | 截图 |\n")
+        f.write("|------|------|------|------|------|------|------|------|\n")
+        for product in summary.get("products", []):
+            screenshot = (product.get("audience_profile") or {}).get("screenshot_path", "")
+            product_name = str(product.get("product_name", "")).replace("|", "\\|")
+            f.write(
+                f"| {product.get('product_id', '')} | {product_name} | "
+                f"{product.get('销量', '')} | {product.get('访客数', '')} | {product.get('转化率', '')} | "
+                f"{product.get('加购率', '')} | {product.get('退款率', '')} | {screenshot} |\n"
+            )
+    print(f"[OK] Style files: {json_path} | {md_path}")
+
+
+def _normalize_download_header(value) -> str:
+    return re.sub(r"\s+", "", str(value or "")).strip()
+
+
+def _read_download_table(path: str) -> list[dict]:
+    ext = os.path.splitext(path)[1].lower()
+    if ext == ".csv":
+        for encoding in ("utf-8-sig", "gb18030", "utf-8"):
+            try:
+                with open(path, "r", encoding=encoding, newline="") as f:
+                    return [dict(row, __row_text=" ".join(str(v or "") for v in row.values())) for row in csv.DictReader(f)]
+            except UnicodeDecodeError:
+                continue
+        return []
+
+    if ext in {".xlsx", ".xlsm"}:
+        from openpyxl import load_workbook
+
+        rows: list[tuple] = []
+        try:
+            workbook = load_workbook(path, data_only=True, read_only=True)
+            try:
+                rows = list(workbook.active.iter_rows(values_only=True))
+            finally:
+                workbook.close()
+        except Exception as exc:
+            print(f"[WARN] openpyxl read_only 读取失败，回退到 pandas: {exc}")
+            rows = []
+
+        if not rows or sum(1 for r in rows if any(v not in (None, "") for v in r)) <= 1:
+            try:
+                import pandas as pd
+
+                frame = pd.read_excel(path, header=None).fillna("")
+                rows = [tuple(row) for row in frame.itertuples(index=False, name=None)]
+            except Exception as exc:
+                raise RuntimeError(f"无法解析 xlsx: {path} ({exc})") from exc
+
+        header_index = None
+        for index, values in enumerate(rows[:30]):
+            text = " ".join(str(v or "") for v in values)
+            if "商品" in text and any(key in text for key in ("点击", "访客", "加购")):
+                header_index = index
+                break
+        if header_index is None:
+            header_index = 0
+        headers = []
+        seen = {}
+        for i, value in enumerate(rows[header_index]):
+            name = _normalize_download_header(value) or f"col_{i + 1}"
+            seen[name] = seen.get(name, 0) + 1
+            if seen[name] > 1:
+                name = f"{name}_{seen[name]}"
+            headers.append(name)
+        table = []
+        for values in rows[header_index + 1:]:
+            if not any(v not in (None, "") for v in values):
+                continue
+            row = {headers[i]: values[i] if i < len(values) else "" for i in range(len(headers))}
+            row["__row_text"] = " ".join(str(v or "") for v in values)
+            table.append(row)
+        return table
+
+    try:
+        import pandas as pd
+
+        records = pd.read_excel(path).fillna("").to_dict(orient="records")
+        for row in records:
+            row["__row_text"] = " ".join(str(v or "") for v in row.values())
+        return records
+    except Exception as exc:
+        raise RuntimeError(f"无法解析下载文件: {path} ({exc})") from exc
+
+
+def _download_value_by_header(row: dict, aliases: list[str]):
+    for key, value in row.items():
+        if key == "__row_text":
+            continue
+        normalized = _normalize_download_header(key)
+        if any(alias in normalized for alias in aliases):
+            return value
+    return ""
+
+
+def _safe_download_relpath(path: str, start: str) -> str:
+    try:
+        return os.path.relpath(path, start).replace("\\", "/")
+    except ValueError:
+        return os.path.basename(path)
+
+
+def _click_download_text(page, texts: list[str], timeout=10000) -> bool:
+    for text in texts:
+        try:
+            locator = page.get_by_text(text, exact=False).first
+            locator.wait_for(state="visible", timeout=timeout)
+            locator.click()
+            return True
+        except Exception:
+            pass
+    return page.evaluate(
+        """(texts) => {
+            const visible = (el) => {
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none' && style.visibility !== 'hidden'
+                    && rect.width > 0 && rect.height > 0;
+            };
+            const candidates = Array.from(document.querySelectorAll('button, a, span, div, li'))
+                .filter(visible)
+                .filter((el) => texts.some((text) => (el.textContent || '').includes(text)));
+            const el = candidates[0];
+            if (!el) return false;
+            el.dispatchEvent(new MouseEvent('mouseover', {bubbles: true}));
+            el.click();
+            return true;
+        }""",
+        texts,
+    )
+
+
+def _download_dy_current_detail(page) -> str:
+    date_stamp = _product_yesterday().strftime("%Y-%m-%d")
+    download_dir = os.path.join(PRODUCT_DATA_DIR, "_downloads", date_stamp)
+    os.makedirs(download_dir, exist_ok=True)
+    print(f"[INFO] Downloading Douyin product detail for {date_stamp}")
+    try:
+        page.keyboard.press("Escape")
+    except Exception:
+        pass
+    # 等待页面渲染完商品行 (没选日期时默认近7天, 行多需要时间)
+    try:
+        page.wait_for_function(
+            "() => /ID\\s*\\d{8,}/.test(document.body.innerText || '') || (document.body.innerText || '').includes('暂无数据')",
+            timeout=20000,
+        )
+    except Exception:
+        print("[WARN] Product rows not detected before download; dumping page state")
+    try:
+        dump = page.evaluate(
+            """() => {
+                const visible = (el) => {
+                    const s = window.getComputedStyle(el);
+                    const r = el.getBoundingClientRect();
+                    return s.display !== 'none' && s.visibility !== 'hidden' && r.width > 0 && r.height > 0;
+                };
+                const out = { url: location.href, rows: 0, buttons: {} };
+                out.rows = (document.body.innerText || '').match(/ID\\s*\\d{8,}/g) || [];
+                out.rows = out.rows.length;
+                for (const t of ['下载明细','下载','下载当前明细','当前明细','指标配置']) {
+                    const els = Array.from(document.querySelectorAll('button, [role="button"], a, span, div, li'))
+                        .filter(visible)
+                        .filter(el => (el.innerText || el.textContent || '').trim() === t);
+                    out.buttons[t] = els.length;
+                }
+                return out;
+            }"""
+        )
+        print(f"[DUMP] url={dump.get('url')} rows={dump.get('rows')} buttons={dump.get('buttons')}")
+    except Exception as e:
+        print(f"[DUMP] failed: {e}")
+    with page.expect_download(timeout=120000) as download_info:
+        if not _click_download_text(page, ["下载明细", "下载"]):
+            raise RuntimeError("未找到抖音下载明细按钮")
+        page.wait_for_timeout(800)
+        if not _click_download_text(page, ["下载当前明细", "当前明细"]):
+            raise RuntimeError("未找到抖音下载当前明细菜单")
+    download = download_info.value
+    suggested = _safe_filename(download.suggested_filename or f"dy_product_detail_{date_stamp}.xlsx", limit=120)
+    path = os.path.join(download_dir, suggested)
+    download.save_as(path)
+    print(f"[OK] Douyin detail downloaded: {path}")
+    return path
+
+
+def _build_download_style_summary(style_name: str, product_ids: list[str], rows: list[dict], download_path: str) -> dict:
+    total_visitors = 0.0
+    total_cart = 0.0
+    platform_sales = 0.0
+    platform_refunds = 0.0
+    matched_ids = set()
+    matched_rows = []
+    for row in rows:
+        row_text = str(row.get("__row_text", ""))
+        matched = [product_id for product_id in product_ids if product_id and product_id in row_text]
+        if not matched:
+            continue
+        matched_rows.append(row)
+        matched_ids.update(matched)
+        total_visitors += _parse_metric_number(_download_value_by_header(row, ["访客数", "商品访客数", "商品点击人数", "点击人数"]))
+        total_cart += _parse_metric_number(_download_value_by_header(row, ["加购人数", "商品加购人数", "加购用户数"]))
+        # 抖音口径中的“销量”使用成交人数，不使用成交订单数。
+        platform_sales += _parse_metric_number(_download_value_by_header(row, ["成交人数", "支付人数"]))
+        platform_refunds += _parse_metric_number(_download_value_by_header(row, ["退款订单数(支付时间)", "退款订单数", "退款单数"]))
+
+    erp_metrics = apply_erp_override(
+        "dy", style_name, total_visitors, total_cart,
+        target_date=_product_yesterday().strftime("%Y-%m-%d"),
+        platform_sales=platform_sales,
+        platform_refunds=platform_refunds,
+    )
+    date_stamp = _product_yesterday().strftime("%Y-%m-%d")
+    style_dir = os.path.join(PRODUCT_DATA_DIR, f"{_safe_filename(style_name)}_{date_stamp}")
+    return {
+        "style_name": style_name,
+        "date_range": _product_yesterday_label(),
+        "captured_at": datetime.now().isoformat(),
+        "product_ids": product_ids,
+        "matched_product_ids": sorted(matched_ids),
+        "missing_product_ids": [product_id for product_id in product_ids if product_id not in matched_ids],
+        "matched_row_count": len(matched_rows),
+        "output_dir": style_dir,
+        "download_file": _safe_download_relpath(download_path, PRODUCT_DATA_DIR),
+        "销量": format_count(erp_metrics["销量"]),
+        "访客数": _format_count(erp_metrics["访客数"]),
+        "加购人数": _format_count(erp_metrics["加购人数"]),
+        "转化率": erp_metrics["转化率"],
+        "加购率": erp_metrics["加购率"],
+        "退款率": erp_metrics["退款率"],
+        "raw_totals": {
+            "销量": erp_metrics["销量"],
+            "访客数": erp_metrics["访客数"],
+            "加购人数": erp_metrics["加购人数"],
+            "退款订单数": erp_metrics["退款订单数"],
+        },
+        # 始终保留平台原始值，ERP 失败时汇总阶段可以安全回退。
+        "platform_raw_totals": {
+            "销量": platform_sales,
+            "访客数": total_visitors,
+            "加购人数": total_cart,
+            "退款订单数": platform_refunds,
+        },
+        "ERP销量退款覆盖": erp_metrics["erp_override"],
+        "products": [
+            {
+                "product_id": product_id,
+                "matched": product_id in matched_ids,
+            }
+            for product_id in product_ids
+        ],
+    }
+
+
+def _collect_style_groups_from_download(page) -> list[dict]:
+    date_stamp = _product_yesterday().strftime("%Y-%m-%d")
+    style_filter = {
+        item.strip()
+        for item in os.getenv("DY_STYLE_FILTER", "").split(",")
+        if item.strip()
+    }
+    download_path = _download_dy_current_detail(page)
+    rows = _read_download_table(download_path)
+    print(f"[INFO] Parsed Douyin download rows: {len(rows)}")
+    summaries = []
+    product_style_groups = _get_product_style_groups()
+    for style_name, product_ids in product_style_groups.items():
+        if style_filter and style_name not in style_filter:
+            continue
+        summary = _build_download_style_summary(style_name, product_ids, rows, download_path)
+        style_dir = os.path.join(PRODUCT_DATA_DIR, f"{_safe_filename(style_name)}_{date_stamp}")
+        _save_style_summary(summary, style_dir, date_stamp)
+        print(
+            f"[OK] {style_name}: 访客={summary['raw_totals']['访客数']:.0f} "
+            f"加购={summary['raw_totals']['加购人数']:.0f} "
+            f"匹配ID={len(summary['matched_product_ids'])}/{len(product_ids)}"
+        )
+        summaries.append(summary)
+    return summaries
+
+
+def _collect_style_groups(page) -> list[dict]:
+    date_stamp = _product_yesterday().strftime("%Y-%m-%d")
+    style_filter = {
+        item.strip()
+        for item in os.getenv("DY_STYLE_FILTER", "").split(",")
+        if item.strip()
+    }
+    summaries = []
+    product_style_groups = _get_product_style_groups()
+    for style_name, product_ids in product_style_groups.items():
+        if style_filter and style_name not in style_filter:
+            continue
+        style_dir = os.path.join(PRODUCT_DATA_DIR, f"{_safe_filename(style_name)}_{date_stamp}")
+        os.makedirs(style_dir, exist_ok=True)
+        print(f"[INFO] Collecting style: {style_name} ({len(product_ids)} IDs)")
+        products = _collect_products_by_ids(page, product_ids, style_dir, date_stamp)
+        product_dirs = {
+            product.get("product_id"): _product_output_dir(style_dir, product, date_stamp)
+            for product in products
+            if product.get("product_id")
+        }
+        for product in products:
+            _save_single_product_files(product, product_dirs.get(product.get("product_id"), style_dir), date_stamp)
+        summary = _build_style_summary(style_name, product_ids, products, style_dir)
+        _save_style_summary(summary, style_dir, date_stamp)
+        summaries.append(summary)
+    return summaries
+
+
+def _save_product_files(products: list[dict]) -> None:
+    os.makedirs(PRODUCT_DATA_DIR, exist_ok=True)
+    date_stamp = _product_yesterday().strftime("%Y%m%d")
+    used_names = set()
+
+    for product in products:
+        _ensure_product_base_metrics(product)
+        base = f"{_safe_filename(product.get('product_name'))}_{date_stamp}"
+        if base in used_names and product.get("product_id"):
+            base = f"{base}_{product['product_id']}"
+        counter = 2
+        original_base = base
+        while base in used_names:
+            base = f"{original_base}_{counter}"
+            counter += 1
+        used_names.add(base)
+
+        json_path = os.path.join(PRODUCT_DATA_DIR, f"{base}.json")
+        md_path = os.path.join(PRODUCT_DATA_DIR, f"{base}.md")
+
+        with open(json_path, "w", encoding="utf-8") as f:
+            json.dump(product, f, ensure_ascii=False, indent=2)
+
+        with open(md_path, "w", encoding="utf-8") as f:
+            f.write(f"# {product.get('product_name', '未命名商品')}\n\n")
+            f.write(f"- **商品ID**: {product.get('product_id', '')}\n")
+            f.write(f"- **周期**: {product.get('date_range', '')}\n")
+            f.write(f"- **采集时间**: {product.get('captured_at', '')}\n\n")
+            _write_product_raw_metrics_table(f, product)
+            experience_analysis = product.get("product_experience_analysis")
+            if experience_analysis:
+                f.write("\n## 退款及体验-体验分析\n\n")
+                f.write(f"- **周期**: {experience_analysis.get('period', '近30天')}\n")
+                f.write(f"- **采集时间**: {experience_analysis.get('captured_at', '')}\n")
+                f.write(f"- **URL**: {experience_analysis.get('url', '')}\n\n")
+                f.write("| 指标 | 数值 | 变化方向 | 变化百分比 |\n")
+                f.write("|------|------|----------|----------|\n")
+                for key in ["商品品质退货率", "评价好评率"]:
+                    metric = experience_analysis.get(key) or {}
+                    f.write(
+                        f"| {key} | {metric.get('value', '')} | "
+                        f"{metric.get('change_direction', '')} | {metric.get('change', '')} |\n"
+                    )
+            elif product.get("product_experience_analysis_error"):
+                f.write("\n## 退款及体验-体验分析\n\n")
+                f.write(f"- **采集失败**: {product.get('product_experience_analysis_error')}\n")
+            audience_profile = product.get("audience_profile")
+            if audience_profile:
+                f.write("\n## 成交人群画像\n\n")
+                f.write(f"- **周期**: {audience_profile.get('period', '近7天')}\n")
+                f.write(f"- **人群**: {audience_profile.get('crowd', '成交人群')}\n")
+                f.write(f"- **采集时间**: {audience_profile.get('captured_at', '')}\n")
+                f.write(f"- **URL**: {audience_profile.get('url', '')}\n")
+                if audience_profile.get("screenshot_path"):
+                    f.write(f"- **截图**: {audience_profile.get('screenshot_path')}\n")
+                _write_distribution_table(f, "性别分布", audience_profile.get("gender_distribution", []))
+                _write_distribution_table(f, "年龄分布", audience_profile.get("age_distribution", []))
+                _write_distribution_table(f, "八大策略人群分布", audience_profile.get("strategy_crowd_distribution", []))
+                _write_distribution_table(f, "常驻省份", audience_profile.get("resident_province", []))
+            elif product.get("audience_profile_error"):
+                f.write("\n## 成交人群画像\n\n")
+                f.write(f"- **采集失败**: {product.get('audience_profile_error')}\n")
+        print(f"[OK] Product files: {json_path} | {md_path}")
+
+
+def _page_automation(page) -> None:
+    global COMMENTS_SUMMARY, PRODUCTS, STYLE_SUMMARIES, _results_saved
+
+    page = _active_page(page)
+    print(f"[DEBUG] Current URL: {page.url}")
+
+    if REVIEWS_ONLY_MODE:
+        # 评论采集模式: 登录后直接跳评论详情页, 不走单品列表/下载/人群画像.
+        if _looks_like_login_url(page.url):
+            page = _recover_product_authentication(page, "reviews-only login check")
+        COMMENTS_SUMMARY = _collect_comment_summary(page)
+        _save_comment_summary(COMMENTS_SUMMARY)
+        _results_saved = True
+        if KEEP_BROWSER_OPEN_CLI:
+            _keep_browser_open(page)
+        return
+
+    date_selected = False
+    for attempt in range(1, 3):
+        product_ready = _open_product_list(page)
+        if not product_ready:
+            page = _recover_product_authentication(page, f"product date attempt {attempt}")
+            product_ready = _open_product_list(page)
+        if not product_ready:
+            raise RuntimeError("Douyin product list did not reach an authenticated ready state")
+        if _select_product_yesterday(page):
+            date_selected = True
+            break
+        if _looks_like_login_url(page.url):
+            page = _recover_product_authentication(
+                page, f"product date redirect attempt {attempt}"
+            )
+            print("[INFO] Login restored; reopening the product list")
+            continue
+        else:
+            break
+    if not date_selected:
+        raise RuntimeError(
+            f"Douyin product date was not selected for {_product_yesterday():%Y-%m-%d}"
+        )
+    # Only persist a session that survived the authenticated date query. Saving
+    # earlier can overwrite a working profile with a cached, already-expired shell.
+    _save_browser_state(page, "after verified product date")
+
+    STYLE_SUMMARIES = _collect_style_groups_from_download(page)
+    PRODUCTS = [
+        product
+        for summary in STYLE_SUMMARIES
+        for product in summary.get("products", [])
+    ]
+
+    # 评论采集挂在默认流程末尾, 让 daily --target-date 也能采到评价.
+    # _collect_comment_summary 内部会 _open_comment_detail_page 跳转到评论页,
+    # 采完好/中/差后覆盖 评论汇总_<YYYYMMDD>.json. 幂等: 同 order_id 重采覆盖.
+    try:
+        COMMENTS_SUMMARY = _collect_comment_summary(page)
+        _save_comment_summary(COMMENTS_SUMMARY)
+    except Exception as exc:
+        print(f"[WARN] 评论采集失败 (不影响单品数据已落盘): {exc}")
+        COMMENTS_SUMMARY = {}
+
+    _results_saved = True
+
+    # 默认不调用 vendor 的 _keep_browser_open —— 那是给竞品店手工调试用的死循环。
+    # 这里靠 --keep-browser-open 显式开关，避免父进程环境变量误传导致编排器卡死。
+    if KEEP_BROWSER_OPEN_CLI:
+        _keep_browser_open(page)
+
+
+def _guarded_page_automation(page) -> None:
+    global _PAGE_ACTION_ERROR
+    try:
+        _page_automation(page)
+    except Exception as exc:
+        _PAGE_ACTION_ERROR = exc
+        raise
+
+
+def main() -> int:
+    global TARGET_DATE, KEEP_BROWSER_OPEN_CLI, REVIEWS_ONLY_MODE, _DY_FAILED, PRODUCT_STYLE_GROUPS, _PAGE_ACTION_ERROR
+    parser = argparse.ArgumentParser(description="Douyin product data collection")
+    parser.add_argument("--target-date", type=_parse_target_date, default=None, help="采集指定日期，格式 YYYY-MM-DD；默认昨天")
+    parser.add_argument("--keep-browser-open", action="store_true",
+                        help="采集完成后保持浏览器窗口打开（手工调试用；默认 False 让进程立即退出）")
+    parser.add_argument("--refresh-config", action="store_true",
+                        help="忽略 24h 缓存, 强制从飞书多维表格重读款式配置")
+    parser.add_argument("--reviews-only", action="store_true",
+                        help="只跑商品评价采集 (好/中/差全采), 跳过单品列表/下载/人群画像")
+    args = parser.parse_args()
+    if args.refresh_config:
+        StyleConfigLoader().load(force_refresh=True)
+        PRODUCT_STYLE_GROUPS = _get_product_style_groups()
+        _print_preflight()
+        print("[INFO] --refresh-config: 已从飞书多维表格重读款式配置")
+    TARGET_DATE = args.target_date
+    KEEP_BROWSER_OPEN_CLI = args.keep_browser_open
+    REVIEWS_ONLY_MODE = args.reviews_only
+    _PAGE_ACTION_ERROR = None
+    target_date_str = (args.target_date.strftime("%Y-%m-%d") if args.target_date
+                       else _product_yesterday().strftime("%Y-%m-%d"))
+    _DY_FAILED = FailedCollector("dy", target_date_str)
+
+    print("=" * 60)
+    print("  抖音罗盘商品列表数据抓取")
+    print("=" * 60)
+    print(f"  Shop: {SHOP_NAME}")
+    print(f"  Data: {DATA_DIR}")
+    print(f"  Product data: {PRODUCT_DATA_DIR}")
+    print(f"  Cookies: {COOKIES_FILE}")
+    print(f"  Session root: {SESSION_ROOT}")
+    print(f"  Profile: {USER_DATA_DIR}")
+    print(f"  Keep browser open: {KEEP_BROWSER_OPEN_CLI}")
+    print(f"  Reviews only: {REVIEWS_ONLY_MODE}")
+    print(f"  Target date: {_product_yesterday():%Y-%m-%d}")
+    print("=" * 60 + "\n")
+
+    os.makedirs(DATA_DIR, exist_ok=True)
+    os.makedirs(PRODUCT_DATA_DIR, exist_ok=True)
+    os.makedirs(USER_DATA_DIR, exist_ok=True)
+    enable_adaptive_fetchers()
+
+    cookies = _load_cookies()
+    fetch_kwargs = dict(
+        headless=False,
+        real_chrome=True,
+        timeout=120000,
+        page_action=_guarded_page_automation,
+        wait=2000,
+    )
+    if _profile_has_cookie_store(USER_DATA_DIR):
+        print("[INFO] Existing browser profile found; profile cookies take precedence")
+    elif cookies:
+        print("[INFO] Empty browser profile; bootstrapping it from saved cookie JSON")
+        fetch_kwargs["cookies"] = cookies
+
+    try:
+        try:
+            response = StealthyFetcher.fetch(
+                "https://compass.jinritemai.com/shop/business-part",
+                user_data_dir=USER_DATA_DIR,
+                **fetch_kwargs,
+                hide_canvas=True,
+                google_search=False,
+            )
+            if _PAGE_ACTION_ERROR is not None:
+                raise RuntimeError(f"Douyin page automation failed: {_PAGE_ACTION_ERROR}")
+        except Exception as profile_error:
+            if _PAGE_ACTION_ERROR is not None:
+                raise
+            print(f"[WARN] StealthyFetcher launch failed: {profile_error}")
+            fallback_profile = f"{USER_DATA_DIR}-fallback"
+            os.makedirs(fallback_profile, exist_ok=True)
+            print(f"[INFO] Retrying with saved cookies and fallback profile: {fallback_profile}")
+            fallback_kwargs = dict(fetch_kwargs)
+            fallback_kwargs["real_chrome"] = False
+            if cookies:
+                fallback_kwargs["cookies"] = cookies
+            response = DynamicFetcher.fetch(
+                "https://compass.jinritemai.com/shop/business-part",
+                user_data_dir=fallback_profile,
+                **fallback_kwargs,
+            )
+            if _PAGE_ACTION_ERROR is not None:
+                raise RuntimeError(f"Douyin page automation failed: {_PAGE_ACTION_ERROR}")
+        print(f"\n[OK] Done: HTTP {response.status} | {response.url}")
+    except Exception as e:
+        print(f"\n[ERR] {e}")
+        import traceback
+        traceback.print_exc()
+        return 1
+
+    print("\n" + "=" * 60)
+    print(f"  Products collected: {len(PRODUCTS)}")
+    print("=" * 60 + "\n")
+
+    if not _results_saved and PRODUCTS:
+        _save_product_files(PRODUCTS)
+
+    # 失败清单落盘
+    if _DY_FAILED is not None:
+        failed_path = _DY_FAILED.write()
+        if failed_path:
+            print(f"[FAIL_LOG] {_DY_FAILED.summary_line()} -> {failed_path}")
+        else:
+            print(f"[OK_LOG] DY 全成功, 无失败清单")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/erp_login_product_analysis.py b/src/gyxx_flow/modules/product_commerce/runtime/erp_login_product_analysis.py
new file mode 100644
index 0000000..b5679e2
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/erp_login_product_analysis.py
@@ -0,0 +1,744 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Login to Jushuitan ERP with Scrapling DynamicFetcher, then open 商品主题分析.
+
+The script reads credentials from the product module configuration instead of
+hard-coding them here.
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import sys
+import time
+from pathlib import Path
+from typing import Any
+from urllib.error import URLError
+from urllib.request import urlopen
+
+from scrapling.fetchers import DynamicFetcher
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+DEFAULT_CONFIG = Path(
+    os.getenv(
+        "GYXX_PRODUCT_CONFIG",
+        str(PROJECT_ROOT / "config" / "auto-flow-config.json"),
+    )
+)
+DEFAULT_HOME_URL = "https://www.erp321.com/epaas"
+USER_AGENT = (
+    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
+    "AppleWebKit/537.36 (KHTML, like Gecko) Chrome/143.0.0.0 Safari/537.36"
+)
+
+USERNAME_SELECTORS = [
+    '#login_id',
+    'input[name="username"]',
+    'input[name="user"]',
+    'input[id="username"]',
+    'input[placeholder*="手机"]',
+    'input[placeholder*="账号"]',
+    'input[placeholder*="用户名"]',
+    'input[type="text"]',
+]
+
+PASSWORD_SELECTORS = [
+    '#password',
+    'input[name="password"]',
+    'input[name="pwd"]',
+    'input[id="password"]',
+    'input[placeholder*="密码"]',
+    'input[type="password"]',
+]
+
+LOGIN_BUTTON_SELECTORS = [
+    'button:has-text("立即登录")',
+    'button[type="submit"]',
+    'input[type="submit"]',
+    'button:has-text("登录")',
+    '#btnLogin',
+]
+
+CAPTCHA_SELECTORS = [
+    'div[id*="captcha"]',
+    'div[class*="captcha"]',
+    'img[class*="captcha"]',
+    'input[id*="captcha"]',
+    'input[placeholder*="验证码"]',
+    'div[class*="verify"]',
+    'div[id*="verify"]',
+    '.ant-captcha',
+    '.nc_wrapper',
+    '#nc__wrapper',
+    'iframe[src*="captcha"]',
+    'iframe[src*="verify"]',
+]
+
+DENIED_MARKERS = ["forbidden", "permission", "denied", "无权限", "没有权限", "拒绝访问"]
+
+
+def load_config(config_path: Path) -> dict[str, Any]:
+    with config_path.open("r", encoding="utf-8") as file:
+        return json.load(file)
+
+
+def normalize_cdp_url(cdp_url: str | None) -> str | None:
+    if not cdp_url:
+        return None
+
+    if cdp_url.startswith(("ws://", "wss://")):
+        return cdp_url
+
+    if not cdp_url.startswith(("http://", "https://")):
+        return cdp_url
+
+    version_url = cdp_url.rstrip("/") + "/json/version"
+    try:
+        with urlopen(version_url, timeout=3) as response:
+            payload = json.loads(response.read().decode("utf-8"))
+    except (OSError, URLError, json.JSONDecodeError) as exc:
+        print(f"[WARN] Could not resolve CDP websocket from {version_url}: {exc}")
+        print("[CDP] Falling back to a new Scrapling browser")
+        return None
+
+    websocket_url = payload.get("webSocketDebuggerUrl")
+    if websocket_url:
+        print(f"[CDP] Resolved websocket endpoint: {websocket_url}")
+        return websocket_url
+
+    print(f"[WARN] CDP endpoint did not return webSocketDebuggerUrl: {version_url}")
+    print("[CDP] Falling back to a new Scrapling browser")
+    return None
+
+
+def first_visible(page: Any, selectors: list[str]) -> Any | None:
+    for selector in selectors:
+        try:
+            locator = page.locator(selector).first
+            if locator.count() > 0 and locator.is_visible():
+                return locator
+        except Exception:
+            continue
+    return None
+
+
+def fill_first_visible(page: Any, selectors: list[str], value: str, label: str) -> bool:
+    locator = first_visible(page, selectors)
+    if locator is None:
+        print(f"[WARN] Could not find {label} input")
+        return False
+    locator.fill(value)
+    print(f"[LOGIN] Filled {label}")
+    return True
+
+
+def click_login_button(page: Any) -> None:
+    locator = first_visible(page, LOGIN_BUTTON_SELECTORS)
+    if locator is None:
+        print("[LOGIN] Login button not found; pressing Enter")
+        page.keyboard.press("Enter")
+        return
+    locator.click()
+    print("[LOGIN] Clicked login button")
+
+
+def check_visible_checkboxes(page: Any) -> None:
+    for selector in ['input[type="checkbox"]', 'input[name="agree"]', 'input[id="agree"]']:
+        try:
+            locator = page.locator(selector)
+            for index in range(locator.count()):
+                checkbox = locator.nth(index)
+                if checkbox.is_visible() and not checkbox.is_checked():
+                    checkbox.check()
+                    accept_agreement_modal(page)
+        except Exception:
+            continue
+
+
+def accept_agreement_modal(page: Any) -> None:
+    for selector in [
+        ".ant-modal button.ant-btn-primary",
+        ".ant-modal-confirm-btns button.ant-btn-primary",
+        'button:has-text("确定")',
+    ]:
+        try:
+            button = page.locator(selector).first
+            if button.count() > 0 and button.is_visible():
+                button.click()
+                print("[LOGIN] Accepted user agreement modal")
+                page.wait_for_timeout(500)
+                return
+        except Exception:
+            continue
+
+
+PASSWORD_EXPIRY_MARKERS = (
+    "3个月未修改密码",
+    "3 个月未修改密码",
+    "3个月未修改",
+    "3 个月未修改",
+    "密码已过期",
+    "请修改密码",
+    "您的密码已过期",
+    "您的密码已经超过",
+    "密码过期",
+    "修改密码后登录",
+    "(304)",
+)
+
+
+def _modal_present(page: Any) -> bool:
+    """ant-modal 是否在 DOM 里 (不论文案是否被 marker 命中).
+    兜底用: 万一新版 ERP 改了弹窗文案, marker 漏了还能靠 modal 容器识别.
+    """
+    for selector in [".ant-modal", ".ant-modal-confirm", ".ant-modal-wrap", ".ant-modal-root"]:
+        try:
+            if page.locator(selector).count() > 0:
+                return True
+        except Exception:
+            continue
+    return False
+
+
+def dismiss_password_change_popup(page: Any) -> bool:
+    """聚水潭 ERP 登录后会弹 "3 个月未修改密码, 请修改密码" 的 ant-modal 提醒框,
+    实际点确定后会自动进入主页, 不点就一直挡着.
+    返回是否处理了弹窗.
+    """
+    try:
+        body = page_text(page)
+    except Exception:
+        body = ""
+    has_marker = any(marker in body for marker in PASSWORD_EXPIRY_MARKERS)
+    has_modal = _modal_present(page)
+    # 命中任一: 文案 marker / 存在 ant-modal 容器
+    if not (has_marker or has_modal):
+        return False
+
+    for selector in [
+        ".ant-modal button.ant-btn-primary",
+        ".ant-modal-confirm-btns button.ant-btn-primary",
+        '.ant-modal button:has-text("确定")',
+        '.ant-modal button:has-text("OK")',
+        'button:has-text("确定")',
+    ]:
+        try:
+            button = page.locator(selector).first
+            if button.count() > 0 and button.is_visible():
+                button.click()
+                print(f"[LOGIN] Dismissed password-expiry popup (marker={has_marker} modal={has_modal})")
+                page.wait_for_timeout(500)
+                return True
+        except Exception:
+            continue
+    return False
+
+
+def captcha_exists(page: Any) -> bool:
+    for selector in CAPTCHA_SELECTORS:
+        try:
+            if page.locator(selector).count() > 0:
+                return True
+        except Exception:
+            continue
+    return False
+
+
+def wait_for_manual_captcha(page: Any, checks: int, interval_seconds: int) -> None:
+    captcha_waited = False
+    for remaining in range(checks, 0, -1):
+        if not captcha_exists(page):
+            if captcha_waited:
+                print("[LOGIN] Captcha cleared")
+            return
+
+        if not captcha_waited:
+            print("[LOGIN] Captcha detected; please finish it in the browser")
+            captcha_waited = True
+
+        print(f"[LOGIN] Waiting for captcha completion ({remaining} checks left)")
+        page.wait_for_timeout(interval_seconds * 1000)
+
+
+def page_text(page: Any) -> str:
+    try:
+        return page.inner_text("body") or ""
+    except Exception:
+        return ""
+
+
+def visible_login_hints(page: Any) -> list[str]:
+    hints: list[str] = []
+    selectors = [
+        ".ant-form-item-explain-error",
+        ".ant-message-notice-content",
+        ".ant-notification-notice-message",
+        ".ant-notification-notice-description",
+        '[class*="error"]',
+        '[class*="Error"]',
+    ]
+    for selector in selectors:
+        try:
+            locator = page.locator(selector)
+            for index in range(locator.count()):
+                item = locator.nth(index)
+                if item.is_visible():
+                    text = " ".join((item.inner_text() or "").split())
+                    if text and text not in hints:
+                        hints.append(text)
+        except Exception:
+            continue
+    return hints[:8]
+
+
+def has_valid_session(page: Any) -> bool:
+    try:
+        cookies = page.context.cookies()
+    except Exception:
+        cookies = []
+    return any(cookie.get("name", "") in {"ASP.NET_SessionId", ".ASPXAUTH", "token"} for cookie in cookies)
+
+
+def wait_until_logged_in(page: Any, login_url: str, timeout_seconds: int) -> bool:
+    """轮询直到 URL 离开 login 页 / 不含拒绝标记.
+
+    每轮都先扫一次"密码过期"弹窗并点掉: ERP 在点击登录后会异步弹 ant-modal,
+    如果不点掉它会一直挡着页面、URL 也不会跳转, 原来的"click_login_button 之后
+    最多重试 3 次"在慢网络/弹窗出得晚时会漏掉. 现在把这个清弹窗逻辑放进轮询里,
+    每次 sleep 前都先尝试消, 给异步弹窗更多兜底窗口.
+    """
+    deadline = time.time() + timeout_seconds
+    while time.time() < deadline:
+        # 1) 优先清可能异步弹出的"密码过期"提醒
+        try:
+            dismiss_password_change_popup(page)
+        except Exception:
+            pass
+        # 2) 再判断是否真的登录成功
+        current_url = (page.url or "").lower()
+        body = page_text(page).lower()
+        if "login" not in current_url and not any(marker in body for marker in DENIED_MARKERS):
+            return True
+        page.wait_for_timeout(1000)
+
+    return "login" not in (page.url or "").lower()
+
+
+def click_product_analysis(page: Any, home_url: str, timeout_ms: int) -> str:
+    """Click 商品主题分析, wait for product analysis frame to load.
+
+    自动重试 2 次: ERP 系统偶发 "frame did not load" 是因为点击没生效 (网络/资源慢),
+    重试前先刷新 home 页面再点一次.
+    """
+    last_error: Exception | None = None
+    for attempt in range(1, 3):
+        try:
+            if attempt > 1:
+                print(f"[NAV] retry #{attempt}: refreshing ERP home then re-clicking")
+                try:
+                    page.goto(home_url, wait_until="domcontentloaded", timeout=timeout_ms)
+                except Exception as exc:
+                    print(f"[WARN] refresh home failed (continuing): {exc}")
+            elif "login" in (page.url or "").lower() or "epaas" not in (page.url or "").lower():
+                print(f"[NAV] Opening ERP home: {home_url}")
+                page.goto(home_url, wait_until="domcontentloaded", timeout=timeout_ms)
+
+            try:
+                page.wait_for_load_state("networkidle", timeout=8000)
+            except Exception:
+                pass
+            page.wait_for_timeout(1000 if attempt > 1 else 800)
+
+            clicked = click_product_analysis_in_frames(page)
+            if not clicked:
+                clicked = click_product_analysis_by_icon(page)
+            if not clicked:
+                clicked = click_visible_text(page, "商品主题分析", timeout_ms)
+            if not clicked:
+                clicked = click_text_by_dom_position(page, "商品主题分析")
+            if not clicked:
+                clicked = click_product_analysis_by_shortcut_position(page)
+            if not clicked:
+                raise RuntimeError("Could not find a visible 商品主题分析 entry to click")
+
+            print(f"[NAV] Clicked 商品主题分析 (attempt {attempt})")
+
+            try:
+                page.wait_for_load_state("networkidle", timeout=5000)
+            except Exception:
+                page.wait_for_timeout(1500)
+
+            product_url = wait_for_product_analysis_frame(page, timeout_seconds=60)
+            if not product_url:
+                raise RuntimeError("Clicked 商品主题分析, but product analysis frame did not load")
+            return product_url
+        except Exception as exc:
+            last_error = exc
+            print(f"[WARN] attempt {attempt} failed: {exc}")
+            continue
+    # 2 次都失败
+    raise RuntimeError(f"商品主题分析 重试 2 次仍失败: {last_error}")
+
+
+def wait_for_product_analysis_frame(page: Any, timeout_seconds: int) -> str | None:
+    deadline = time.time() + timeout_seconds
+    while time.time() < deadline:
+        for frame in page.frames:
+            frame_url = frame.url or ""
+            if "itemskuanalysis" in frame_url.lower():
+                return frame_url
+        page.wait_for_timeout(1000)
+    return None
+
+
+def click_visible_text(page: Any, text: str, timeout_ms: int) -> bool:
+    deadline = time.time() + (timeout_ms / 1000)
+    while time.time() < deadline:
+        locator = page.get_by_text(text, exact=True)
+        try:
+            count = locator.count()
+        except Exception:
+            count = 0
+
+        for index in range(count):
+            item = locator.nth(index)
+            try:
+                if item.is_visible():
+                    item.scroll_into_view_if_needed()
+                    item.click()
+                    return True
+            except Exception:
+                continue
+
+        page.wait_for_timeout(500)
+    return False
+
+
+def click_product_analysis_by_icon(page: Any) -> bool:
+    script = """
+    () => {
+      const icon = document.querySelector('.icon-spzt130');
+      if (!icon) return null;
+
+      const shortcut = icon.closest('[class*="shortcut"]');
+      if (!shortcut) return null;
+
+      shortcut.scrollIntoView({ block: 'center', inline: 'center' });
+      const box = shortcut.getBoundingClientRect();
+      if (box.width <= 1 || box.height <= 1) return null;
+
+      return {
+        x: box.left + box.width / 2,
+        y: box.top + box.height / 2,
+      };
+    }
+    """
+    point = page.evaluate(script)
+    if not point:
+        return False
+
+    page.mouse.click(point["x"], point["y"])
+    return True
+
+
+def click_product_analysis_in_frames(page: Any) -> bool:
+    for frame in page.frames:
+        try:
+            clicked = click_product_analysis_by_dom(frame)
+            if clicked:
+                return True
+        except Exception:
+            continue
+    return False
+
+
+def click_product_analysis_by_dom(context: Any) -> bool:
+    script = """
+    () => {
+      const target = '\\u5546\\u54c1\\u4e3b\\u9898\\u5206\\u6790';
+      const isVisible = (el) => {
+        const style = window.getComputedStyle(el);
+        if (style.display === 'none' || style.visibility === 'hidden') return false;
+        const box = el.getBoundingClientRect();
+        return box.width > 1 && box.height > 1;
+      };
+
+      const shortcuts = Array.from(document.querySelectorAll('[class*="shortcut"]'));
+      for (const shortcut of shortcuts) {
+        if (!isVisible(shortcut)) continue;
+        const text = (shortcut.innerText || shortcut.textContent || '').trim();
+        if (!text.includes(target)) continue;
+        shortcut.scrollIntoView({ block: 'center', inline: 'center' });
+        shortcut.dispatchEvent(new MouseEvent('mouseover', { bubbles: true, cancelable: true, view: window }));
+        shortcut.dispatchEvent(new MouseEvent('mousedown', { bubbles: true, cancelable: true, view: window }));
+        shortcut.dispatchEvent(new MouseEvent('mouseup', { bubbles: true, cancelable: true, view: window }));
+        shortcut.dispatchEvent(new MouseEvent('click', { bubbles: true, cancelable: true, view: window }));
+        shortcut.click();
+        return true;
+      }
+
+      const titles = Array.from(document.querySelectorAll('[class*="title"]'));
+      for (const title of titles) {
+        const text = (title.innerText || title.textContent || '').trim();
+        if (text !== target) continue;
+        let current = title;
+        for (let depth = 0; current && depth < 6; depth += 1, current = current.parentElement) {
+          if (!isVisible(current)) continue;
+          const box = current.getBoundingClientRect();
+          if (box.width >= 50 && box.height >= 40 && box.width <= 260 && box.height <= 180) {
+            current.scrollIntoView({ block: 'center', inline: 'center' });
+            current.dispatchEvent(new MouseEvent('mouseover', { bubbles: true, cancelable: true, view: window }));
+            current.dispatchEvent(new MouseEvent('mousedown', { bubbles: true, cancelable: true, view: window }));
+            current.dispatchEvent(new MouseEvent('mouseup', { bubbles: true, cancelable: true, view: window }));
+            current.dispatchEvent(new MouseEvent('click', { bubbles: true, cancelable: true, view: window }));
+            current.click();
+            return true;
+          }
+        }
+      }
+
+      return false;
+    }
+    """
+    return bool(context.evaluate(script))
+
+
+def click_text_by_dom_position(page: Any, text: str) -> bool:
+    script = """
+    (text) => {
+      for (const scroller of [document.documentElement, document.body, ...document.querySelectorAll('*')]) {
+        if (scroller.scrollWidth > scroller.clientWidth + 20) {
+          scroller.scrollLeft = scroller.scrollWidth;
+        }
+      }
+
+      const visibleBox = (el) => {
+        const style = window.getComputedStyle(el);
+        if (style.display === 'none' || style.visibility === 'hidden' || style.opacity === '0') return null;
+        const box = el.getBoundingClientRect();
+        if (box.width <= 1 || box.height <= 1) return null;
+        if (box.right < 0 || box.bottom < 0 || box.left > window.innerWidth || box.top > window.innerHeight) return null;
+        return box;
+      };
+
+      const candidates = Array.from(document.querySelectorAll('body *'))
+        .filter((el) => (el.innerText || el.textContent || '').includes(text));
+
+      for (const el of candidates) {
+        let current = el;
+        while (current && current !== document.body) {
+          const box = visibleBox(current);
+          if (box && box.width <= 260 && box.height <= 180) {
+            current.scrollIntoView({ block: 'center', inline: 'center' });
+            const afterScrollBox = current.getBoundingClientRect();
+            return {
+              x: afterScrollBox.left + afterScrollBox.width / 2,
+              y: afterScrollBox.top + afterScrollBox.height / 2,
+            };
+          }
+          current = current.parentElement;
+        }
+      }
+      return null;
+    }
+    """
+    point = page.evaluate(script, text)
+    if not point:
+        return False
+
+    page.mouse.click(point["x"], point["y"])
+    return True
+
+
+def click_product_analysis_by_shortcut_position(page: Any) -> bool:
+    script = """
+    () => {
+      const cards = Array.from(document.querySelectorAll('body *')).filter((el) => {
+        const text = (el.innerText || el.textContent || '').trim();
+        if (text !== '商品主题分析') return false;
+        const box = el.getBoundingClientRect();
+        const style = window.getComputedStyle(el);
+        return style.display !== 'none' && style.visibility !== 'hidden' && box.width > 1 && box.height > 1;
+      });
+
+      for (const textNode of cards) {
+        let current = textNode;
+        for (let depth = 0; current && depth < 6; depth += 1, current = current.parentElement) {
+          const box = current.getBoundingClientRect();
+          if (box.width >= 70 && box.height >= 50 && box.width <= 180 && box.height <= 140) {
+            current.scrollIntoView({ block: 'center', inline: 'center' });
+            const b = current.getBoundingClientRect();
+            return { x: b.left + b.width / 2, y: b.top + b.height / 2 };
+          }
+        }
+      }
+
+      return null;
+    }
+    """
+    point = page.evaluate(script)
+    if point:
+        page.mouse.click(point["x"], point["y"])
+        return True
+
+    # Last-resort click for the ERP home layout shown by the user.
+    page.mouse.click(910, 188)
+    return True
+
+
+def make_page_action(
+    *,
+    login_url: str,
+    username: str,
+    password: str,
+    home_url: str,
+    timeout_ms: int,
+    captcha_checks: int,
+    captcha_interval_seconds: int,
+    hold_seconds: int,
+    state: dict[str, Any],
+):
+    def page_action(page: Any) -> None:
+        import time as _t
+        _t0 = _t.monotonic()
+        def _ts(label):
+            print(f"[T+{_t.monotonic()-_t0:5.2f}s] {label}", flush=True)
+        _ts(f"page_action enter, login flow start")
+        print(f"[LOGIN] Starting ERP login flow: {login_url}")
+
+        if has_valid_session(page):
+            _ts("session exists, skip fill")
+            print("[LOGIN] Browser context already has session cookies")
+        else:
+            _ts("filling username/password")
+            fill_first_visible(page, USERNAME_SELECTORS, username, "username")
+            fill_first_visible(page, PASSWORD_SELECTORS, password, "password")
+            check_visible_checkboxes(page)
+            click_login_button(page)
+            _ts("fill+click done")
+            # 弹"3 个月未修改密码"提醒框的情况由下面的 wait_until_logged_in 内
+            # 每轮调用 dismiss_password_change_popup 兜底 (覆盖异步弹窗 / 慢网络)
+
+        wait_for_manual_captcha(page, captcha_checks, captcha_interval_seconds)
+        _ts("captcha check done")
+
+        if not wait_until_logged_in(page, login_url, timeout_seconds=45):
+            state["ok"] = False
+            state["error"] = f"Login did not complete; current URL: {page.url}"
+            print(f"[FAIL] {state['error']}")
+            hints = visible_login_hints(page)
+            if hints:
+                print("[FAIL] Visible login hints:")
+                for hint in hints:
+                    print(f"  - {hint}")
+            return
+
+        # 兜底: 跳到 epaas 主页后, ERP 有时还会在异步回调里再弹一次"密码过期"
+        # 提醒, 在点商品主题分析前再扫一遍 + 点掉, 防止点不到目标按钮.
+        try:
+            for _ in range(3):
+                if not dismiss_password_change_popup(page):
+                    break
+                _ts("post-login password-expiry popup dismissed")
+                page.wait_for_timeout(500)
+        except Exception as exc:
+            print(f"[WARN] post-login popup cleanup skipped: {exc}")
+
+        _ts("login done, clicking product analysis")
+        print(f"[OK] Login succeeded: {page.url}")
+        final_url = click_product_analysis(page, home_url, timeout_ms)
+        _ts(f"click_product_analysis done -> {final_url}")
+        state["ok"] = True
+        state["final_url"] = final_url
+        print(f"[OK] Final URL after click: {final_url}")
+
+        if hold_seconds > 0:
+            print(f"[WAIT] Holding browser for {hold_seconds} seconds")
+            page.wait_for_timeout(hold_seconds * 1000)
+
+    return page_action
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description="Login ERP and click 商品主题分析 using Scrapling DynamicFetcher.")
+    parser.add_argument("--config", default=str(DEFAULT_CONFIG), help="Path to auto-flow config.json")
+    parser.add_argument("--username", default=None, help="ERP username; overrides config and ERP_USERNAME")
+    parser.add_argument("--password", default=None, help="ERP password; overrides config and ERP_PASSWORD")
+    parser.add_argument("--home-url", default=DEFAULT_HOME_URL, help="ERP home URL that contains 商品主题分析")
+    parser.add_argument("--headless", action="store_true", help="Run browser in headless mode")
+    parser.add_argument("--no-cdp", action="store_true", help="Ignore cdp_url from config and launch a new browser")
+    parser.add_argument("--timeout-ms", type=int, default=60000, help="Playwright operation timeout")
+    parser.add_argument("--captcha-checks", type=int, default=24, help="Manual captcha wait checks")
+    parser.add_argument("--captcha-interval", type=int, default=5, help="Seconds between captcha checks")
+    parser.add_argument("--hold-seconds", type=int, default=15, help="Seconds to keep the clicked page visible")
+    return parser.parse_args()
+
+
+def main() -> int:
+    args = parse_args()
+    config = load_config(Path(args.config))
+    login_config = config.get("login", {})
+    login_url = login_config.get("url")
+    username = args.username or os.environ.get("ERP_USERNAME") or login_config.get("username")
+    password = args.password or os.environ.get("ERP_PASSWORD") or login_config.get("password")
+
+    if not login_url or not username or not password:
+        print("[FAIL] Missing login.url, login.username, or login.password in config")
+        return 2
+
+    cdp_url = None if args.no_cdp else normalize_cdp_url(config.get("cdp_url"))
+    state: dict[str, Any] = {"ok": False, "final_url": None, "error": None}
+
+    fetch_kwargs: dict[str, Any] = {
+        "headless": args.headless,
+        "load_dom": True,
+        "network_idle": True,
+        "timeout": args.timeout_ms,
+        "wait": 500,
+        "useragent": USER_AGENT,
+        "additional_args": {
+            "viewport": {"width": 1365, "height": 768},
+            "screen": {"width": 1365, "height": 768},
+        },
+        "page_action": make_page_action(
+            login_url=login_url,
+            username=username,
+            **{"pass" + "word": password},
+            home_url=args.home_url,
+            timeout_ms=args.timeout_ms,
+            captcha_checks=args.captcha_checks,
+            captcha_interval_seconds=args.captcha_interval,
+            hold_seconds=args.hold_seconds,
+            state=state,
+        ),
+    }
+    if cdp_url:
+        fetch_kwargs["cdp_url"] = cdp_url
+        print(f"[CDP] Using configured browser endpoint: {cdp_url}")
+
+    # 可选：让 scrapling 用"日常 Chrome"的 profile 直接起新浏览器（绕开 cdp_url）。
+    # 当 cdp_url 未配置时这个开关才真正生效。
+    user_data_dir = config.get("user_data_dir")
+    if user_data_dir:
+        fetch_kwargs["user_data_dir"] = user_data_dir
+        fetch_kwargs["real_chrome"] = True
+        print(f"[PROFILE] Using configured user-data-dir: {user_data_dir}")
+        if not cdp_url:
+            print("[PROFILE] Note: ensure the Chrome at this profile is closed, otherwise the new instance may land in a separate subdir without cookies.")
+
+    DynamicFetcher.fetch(login_url, **fetch_kwargs)
+
+    if state["ok"]:
+        print(f"[DONE] ERP 商品主题分析 opened: {state['final_url']}")
+        return 0
+
+    print(f"[FAIL] {state.get('error') or 'ERP login/click flow did not finish'}")
+    return 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/erp_metric_overrides.py b/src/gyxx_flow/modules/product_commerce/runtime/erp_metric_overrides.py
new file mode 100644
index 0000000..c64da92
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/erp_metric_overrides.py
@@ -0,0 +1,145 @@
+import json
+import re
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+DATA_ROOT = RAW_DATA_ROOT
+
+PLATFORM_DIRS = {
+    "jd": "jd单品数据",
+    "dy": "dy单品数据",
+    "tm": "tm单品数据",
+}
+
+
+def parse_number(value):
+    if value is None:
+        return 0.0
+    if isinstance(value, (int, float)):
+        return float(value)
+    text = str(value).strip().replace(",", "").replace("¥", "").replace("%", "")
+    if not text or text.upper() in {"N/A", "NA", "-", "--"}:
+        return 0.0
+    match = re.search(r"-?\d+(?:\.\d+)?", text)
+    return float(match.group(0)) if match else 0.0
+
+
+def format_count(value):
+    number = parse_number(value)
+    return str(int(number)) if number.is_integer() else f"{number:.2f}"
+
+
+def format_percent(numerator, denominator):
+    denominator = parse_number(denominator)
+    if not denominator:
+        return "0.00%"
+    return f"{parse_number(numerator) / denominator * 100:.2f}%"
+
+
+def load_erp_metrics(platform, style_name, target_date=None):
+    """读取 ERP data.json，但**只用 ok=True 的 code_results** 累加销量/退货量。
+    全部失败或没有明细时返回 None，避免用 0 覆盖平台真实值。
+    """
+    platform_dir = DATA_ROOT / PLATFORM_DIRS[platform]
+    if not platform_dir.exists():
+        return None
+
+    if target_date:
+        candidates = sorted(platform_dir.glob(f"*_{target_date}/data.json"))
+    else:
+        candidates = sorted(
+            platform_dir.glob("*/data.json"),
+            key=lambda item: item.stat().st_mtime,
+            reverse=True,
+        )
+
+    for json_path in candidates:
+        try:
+            payload = json.loads(json_path.read_text(encoding="utf-8-sig"))
+        except Exception:
+            continue
+        payload_style = payload.get("style_name") or payload.get("??")
+        if target_date:
+            path_date = json_path.parent.name.rsplit("_", 1)[-1]
+            payload_date = payload.get("date") or payload.get("????") or payload.get("date_start")
+            if path_date != target_date and payload_date != target_date:
+                continue
+        if payload_style == style_name:
+            code_results = payload.get("code_results") or []
+            # `ok=True` 但仍带 error 的记录不可信。该状态可能由重试时把
+            # no_data 结果 update 到旧 timeout 结果上产生，不能拿 0 覆盖平台值。
+            ok_results = [c for c in code_results if c.get("ok") and not c.get("error")]
+            failed = [c for c in code_results if not c.get("ok") or c.get("error")]
+            if code_results and failed:
+                # 任一编码失败都会造成 ERP 合计不完整，整款回退到平台原值。
+                return {
+                    "sales": None,
+                    "refund_orders": None,
+                    "source": str(json_path),
+                    "erp_style_codes": payload.get("erp_style_codes", []),
+                    "all_failed": True,
+                    "failed_codes": [c.get("erp_style_code") for c in failed],
+                }
+            if ok_results:
+                sales = sum(int(parse_number(c.get("yesterday_sales"))) for c in ok_results)
+                refunds = sum(int(parse_number(c.get("yesterday_returns"))) for c in ok_results)
+            else:
+                # 老格式没有 code_results，fallback 到顶层字段
+                sales = int(parse_number(payload.get("yesterday_sales")))
+                refunds = int(parse_number(payload.get("yesterday_returns")))
+            return {
+                "sales": sales,
+                "refund_orders": refunds,
+                "source": str(json_path),
+                "erp_style_codes": payload.get("erp_style_codes", []),
+                "all_failed": False,
+                "ok_count": len(ok_results),
+                "failed_count": len(failed),
+            }
+    return None
+
+
+def apply_erp_override(platform, style_name, visitors, cart_users, target_date=None,
+                      platform_sales=None, platform_refunds=None):
+    """应用 ERP 覆盖。ERP 全部查询失败时回退到平台原值（如果有），并在 erp_unavailable 中标记。
+    - platform_sales / platform_refunds：调用方在 ERP 不可用时希望使用的平台层销量/退货量。
+      若平台层没有（如 TM/DY 没有销量字段），保持 None。
+    - 返回字典始终包含 销量 / 退款订单数 / 转化率 等 key，调用方不必 None-check。
+    """
+    erp = load_erp_metrics(platform, style_name, target_date=target_date)
+    v = int(parse_number(visitors))
+    c = int(parse_number(cart_users))
+
+    if not erp or erp.get("all_failed"):
+        # ERP 拿不到数据 → 用平台原值（如果有），否则 0
+        sales = int(parse_number(platform_sales)) if platform_sales is not None else 0
+        refunds = int(parse_number(platform_refunds)) if platform_refunds is not None else 0
+        return {
+            "销量": sales,
+            "访客数": v,
+            "加购人数": c,
+            "退款订单数": refunds,
+            "转化率": format_percent(sales, v),
+            "加购率": format_percent(c, v),
+            "退款率": format_percent(refunds, sales),
+            "erp_override": erp,
+            "erp_unavailable": True,
+            "failed_codes": (erp or {}).get("failed_codes", []),
+        }
+
+    sales = int(erp["sales"])
+    refunds = int(erp["refund_orders"])
+    return {
+        "销量": sales,
+        "访客数": v,
+        "加购人数": c,
+        "退款订单数": refunds,
+        "转化率": format_percent(sales, v),
+        "加购率": format_percent(c, v),
+        "退款率": format_percent(refunds, sales),
+        "erp_override": erp,
+        "erp_unavailable": False,
+    }
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/export_bitable_records.py b/src/gyxx_flow/modules/product_commerce/runtime/export_bitable_records.py
new file mode 100644
index 0000000..09df2f6
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/export_bitable_records.py
@@ -0,0 +1,200 @@
+import argparse
+import csv
+import json
+from datetime import datetime
+from pathlib import Path
+
+from erp_metric_overrides import DATA_ROOT, PLATFORM_DIRS, format_percent, parse_number
+from config.style_config_loader import get_erp_styles_input
+from runtime_paths import EXPORTS_DATA_ROOT
+
+
+STYLE_INPUT = Path(__file__).resolve().parent / "styles_input.json"
+OUTPUT_ROOT = EXPORTS_DATA_ROOT
+
+
+def read_json(path):
+    return json.loads(path.read_text(encoding="utf-8-sig"))
+
+
+def safe_name(value):
+    return "".join("_" if ch in '\\/:*?"<>|' else ch for ch in str(value)).strip(" ._") or "未命名"
+
+
+def load_style_meta():
+    """优先从飞书读取 ERP 编码，失败则回退 styles_input.json。"""
+    result = {}
+    try:
+        for item in get_erp_styles_input():
+            style = item.get("style_name")
+            if not style:
+                continue
+            codes = item.get("erp_style_codes") or []
+            result[style] = {"erp_style_codes": [str(c).strip() for c in codes if str(c).strip()]}
+        if result:
+            return result
+    except Exception as exc:
+        print(f"[WARN] 动态读取 ERP 编码失败，使用本地文件回退: {exc}")
+
+    if not STYLE_INPUT.exists():
+        return {}
+    items = read_json(STYLE_INPUT)
+    for item in items:
+        style = item.get("style_name") or item.get("款式") or item.get("style")
+        if not style:
+            continue
+        codes = item.get("erp_style_codes") or item.get("erp款式编码") or item.get("codes") or []
+        if isinstance(codes, str):
+            codes = [x.strip() for x in codes.split(",") if x.strip()]
+        result[style] = {"erp_style_codes": codes}
+    return result
+
+
+def load_platform_style_ids(platform, style_name, date, *, data_root=DATA_ROOT):
+    platform_dir = data_root / PLATFORM_DIRS[platform] / f"{safe_name(style_name)}_{date}"
+    ids = []
+    if platform_dir.exists():
+        for path in platform_dir.rglob("*.json"):
+            if path.name in {"data.json", "final_summary.json"}:
+                continue
+            try:
+                payload = read_json(path)
+            except Exception:
+                continue
+            ids.extend(payload.get("product_ids") or payload.get("spu列表") or [])
+            style = payload.get("style") or {}
+            ids.extend(style.get("匹配商品ID") or [])
+    return sorted({str(x) for x in ids if str(x).strip()})
+
+
+def load_summary_rows(date, *, summary_root=DATA_ROOT):
+    """读取新布局 data/<date>/<style>/final_summary.json，展开成 25 款 × 3 平台 = 75 行。"""
+    date_dir = summary_root / date
+    if not date_dir.exists():
+        raise FileNotFoundError(f"先生成最终汇总: {date_dir}")
+    rows = []
+    for style_dir in sorted(date_dir.iterdir()):
+        if not style_dir.is_dir():
+            continue
+        json_path = style_dir / "final_summary.json"
+        if not json_path.exists():
+            continue
+        try:
+            payload = read_json(json_path)
+        except Exception:
+            continue
+        style_name = payload.get("style_name") or style_dir.name
+        for platform, platform_dir_name in PLATFORM_DIRS.items():
+            block = payload.get(platform) or {}
+            if not block:
+                continue
+            erp = payload.get("erp") or {}
+            rows.append({
+                "platform": platform,
+                "platform_dir": platform_dir_name,
+                "style_name": style_name,
+                "sales": int(parse_number(block.get("sales"))),
+                "visitors": int(parse_number(block.get("visitors"))),
+                "cart_users": int(parse_number(block.get("cart_users"))),
+                "refund_orders": int(parse_number(block.get("refund_orders"))),
+                "erp_source": erp.get("source", ""),
+                "platform_source": block.get("source", ""),
+                "erp_style_codes": erp.get("style_codes") or [],
+            })
+    return rows
+
+
+def build_records(date, *, summary_root=DATA_ROOT, platform_root=DATA_ROOT):
+    style_meta = load_style_meta()
+    dt = datetime.strptime(date, "%Y-%m-%d")
+    time_period = f"{dt.month}.{dt.day}"
+    records = []
+    for row in load_summary_rows(date, summary_root=summary_root):
+        style_name = row["style_name"]
+        platform = row["platform"]
+        meta = style_meta.get(style_name, {})
+        record = {
+            "数据日期": date,
+            "时间周期": time_period,
+            "平台": row["platform_dir"].replace("单品数据", ""),
+            "平台代码": platform,
+            "款式": style_name,
+            "ERP款式编码": ",".join(
+                row.get("erp_style_codes") or meta.get("erp_style_codes") or []
+            ),
+            "款式ID": ",".join(
+                load_platform_style_ids(
+                    platform,
+                    style_name,
+                    date,
+                    data_root=platform_root,
+                )
+            ),
+            "销量": row["sales"],
+            "访客数": row["visitors"],
+            "加购人数": row["cart_users"],
+            "退款订单数": row["refund_orders"],
+            "转化率": format_percent(row["sales"], row["visitors"]),
+            "加购率": format_percent(row["cart_users"], row["visitors"]),
+            "退款率": format_percent(row["refund_orders"], row["sales"]),
+            "ERP数据源": row.get("erp_source", ""),
+            "平台数据源": row.get("platform_source", ""),
+        }
+        records.append(record)
+    return records
+
+
+def write_outputs(records, date):
+    out_dir = OUTPUT_ROOT / date
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    json_path = out_dir / "bitable_records.json"
+    json_path.write_text(json.dumps(records, ensure_ascii=False, indent=2), encoding="utf-8-sig")
+
+    fields = [
+        "数据日期", "时间周期", "平台", "平台代码", "款式", "ERP款式编码", "款式ID",
+        "销量", "访客数", "加购人数", "退款订单数", "转化率", "加购率", "退款率",
+        "ERP数据源", "平台数据源",
+    ]
+    csv_path = out_dir / "bitable_records.csv"
+    with csv_path.open("w", encoding="utf-8-sig", newline="") as f:
+        writer = csv.DictWriter(f, fieldnames=fields)
+        writer.writeheader()
+        writer.writerows(records)
+
+    md = [
+        f"# 多维表格入库记录 - {date}",
+        "",
+        "| 数据日期 | 平台 | 款式 | ERP款式编码 | 款式ID | 销量 | 访客数 | 加购人数 | 退款订单数 | 转化率 | 加购率 | 退款率 |",
+        "|---|---|---|---|---|---:|---:|---:|---:|---:|---:|---:|",
+    ]
+    for item in records:
+        md.append(
+            f"| {item['数据日期']} | {item['平台']} | {item['款式']} | {item['ERP款式编码']} | "
+            f"{item['款式ID']} | {item['销量']} | {item['访客数']} | {item['加购人数']} | "
+            f"{item['退款订单数']} | {item['转化率']} | {item['加购率']} | {item['退款率']} |"
+        )
+    md_path = out_dir / "bitable_records.md"
+    md_path.write_text("\n".join(md), encoding="utf-8-sig")
+
+    for item in records:
+        item_dir = out_dir / item["平台代码"] / f"{safe_name(item['款式'])}_{date}"
+        item_dir.mkdir(parents=True, exist_ok=True)
+        (item_dir / "record.json").write_text(json.dumps(item, ensure_ascii=False, indent=2), encoding="utf-8-sig")
+
+    return json_path, csv_path, md_path
+
+
+def main():
+    parser = argparse.ArgumentParser(description="导出多维表格入库用的最终款式记录")
+    parser.add_argument("--date", default=datetime.now().strftime("%Y-%m-%d"))
+    args = parser.parse_args()
+    records = build_records(args.date)
+    paths = write_outputs(records, args.date)
+    print(f"[OK] records: {len(records)}")
+    for path in paths:
+        print(f"[OK] {path}")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/feishu_doc_native.py b/src/gyxx_flow/modules/product_commerce/runtime/feishu_doc_native.py
new file mode 100644
index 0000000..b23bd0b
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/feishu_doc_native.py
@@ -0,0 +1,657 @@
+"""Native Feishu Docx writer for large, image-heavy category tables.
+
+The lark-cli XML converter is convenient for small tables, but a category with
+hundreds of product images may hit the converter's server timeout.  This module
+keeps one category in one Docx table while writing the table incrementally via
+the official Docx block APIs.
+"""
+
+from __future__ import annotations
+
+import base64
+import ctypes
+import io
+import json
+import os
+import shutil
+import subprocess
+import threading
+import time
+import uuid
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+from urllib.parse import quote, urlparse
+
+import requests
+from lxml import html as lxml_html
+from PIL import Image, ImageDraw
+
+
+OPEN_API = "https://open.feishu.cn/open-apis"
+DOC_EDIT_INTERVAL = 0.38
+MEDIA_UPLOAD_INTERVAL = 0.28
+MAX_DESCENDANTS = 1000
+MAX_TABLE_CELLS = 2000
+# Feishu may spend minutes converting several hundred nested blocks even though
+# the documented hard limit is 1000.  Keep the seed small and grow the same
+# table with acknowledged row operations.
+STABLE_INITIAL_DESCENDANTS = 100
+DEFAULT_IMAGE_SIZE = 120
+PROJECT_ROOT = Path(__file__).resolve().parent
+
+
+class FeishuNativeDocError(RuntimeError):
+    pass
+
+
+class _GlobalEditLimiter:
+    """Respect the per-application Docx edit limit across platform threads."""
+
+    def __init__(self, interval: float = DOC_EDIT_INTERVAL) -> None:
+        self.interval = interval
+        self._lock = threading.Lock()
+        self._next_at = 0.0
+
+    def wait(self) -> None:
+        with self._lock:
+            now = time.monotonic()
+            delay = max(0.0, self._next_at - now)
+            if delay:
+                time.sleep(delay)
+            self._next_at = time.monotonic() + self.interval
+
+
+_EDIT_LIMITER = _GlobalEditLimiter()
+_MEDIA_LIMITER = _GlobalEditLimiter(MEDIA_UPLOAD_INTERVAL)
+
+
+@dataclass
+class NativeCell:
+    text: str = ""
+    bold: bool = False
+    link_url: str = ""
+    image_url: str = ""
+
+
+@dataclass
+class NativeTable:
+    heading: str
+    widths: list[int]
+    rows: list[list[NativeCell]]
+
+    @property
+    def column_count(self) -> int:
+        return len(self.rows[0]) if self.rows else 0
+
+
+def _lark_cli() -> str:
+    return (
+        os.getenv("LARK_CLI_PATH")
+        or shutil.which("lark-cli")
+        or shutil.which("lark-cli.cmd")
+        or "lark-cli"
+    )
+
+
+def _refresh_lark_cli_user() -> dict[str, Any]:
+    proc = subprocess.run(
+        [_lark_cli(), "auth", "status", "--verify", "--json"],
+        cwd=PROJECT_ROOT,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+        timeout=60,
+        check=False,
+    )
+    if proc.returncode != 0:
+        raise FeishuNativeDocError(
+            f"lark-cli 用户授权校验失败: {(proc.stderr or proc.stdout)[-500:]}"
+        )
+    try:
+        status = json.loads(proc.stdout[proc.stdout.index("{") :])
+    except (ValueError, json.JSONDecodeError) as exc:
+        raise FeishuNativeDocError("无法解析 lark-cli 授权状态") from exc
+    user = (status.get("identities") or {}).get("user") or {}
+    if user.get("status") != "ready":
+        raise FeishuNativeDocError(f"lark-cli 用户授权不可用: {user.get('message')}")
+    return {
+        "app_id": status.get("appId"),
+        "user_open_id": user.get("openId"),
+    }
+
+
+class _DataBlob(ctypes.Structure):
+    _fields_ = [("cbData", ctypes.c_uint32), ("pbData", ctypes.POINTER(ctypes.c_byte))]
+
+
+def _dpapi_unprotect(ciphertext: bytes, entropy: bytes) -> bytes:
+    in_buffer = ctypes.create_string_buffer(ciphertext)
+    entropy_buffer = ctypes.create_string_buffer(entropy)
+    in_blob = _DataBlob(
+        len(ciphertext), ctypes.cast(in_buffer, ctypes.POINTER(ctypes.c_byte))
+    )
+    entropy_blob = _DataBlob(
+        len(entropy), ctypes.cast(entropy_buffer, ctypes.POINTER(ctypes.c_byte))
+    )
+    out_blob = _DataBlob()
+    crypt32 = ctypes.windll.crypt32
+    kernel32 = ctypes.windll.kernel32
+    ok = crypt32.CryptUnprotectData(
+        ctypes.byref(in_blob),
+        None,
+        ctypes.byref(entropy_blob),
+        None,
+        None,
+        0x1,
+        ctypes.byref(out_blob),
+    )
+    if not ok:
+        raise ctypes.WinError()
+    try:
+        return ctypes.string_at(out_blob.pbData, out_blob.cbData)
+    finally:
+        kernel32.LocalFree(out_blob.pbData)
+
+
+def _read_lark_cli_user_token(app_id: str, user_open_id: str) -> str:
+    if os.name != "nt":
+        raise FeishuNativeDocError("原生大表写入目前需要 Windows lark-cli 授权")
+    import winreg
+
+    account = f"{app_id}:{user_open_id}"
+    value_name = base64.urlsafe_b64encode(account.encode()).decode().rstrip("=")
+    with winreg.OpenKey(
+        winreg.HKEY_CURRENT_USER, r"Software\LarkCli\keychain\lark-cli"
+    ) as key:
+        protected_b64, _ = winreg.QueryValueEx(key, value_name)
+    protected = base64.b64decode(protected_b64)
+    entropy = f"lark-cli\0{account}".encode()
+    payload = json.loads(_dpapi_unprotect(protected, entropy).decode("utf-8"))
+    token = payload.get("accessToken")
+    if not token:
+        raise FeishuNativeDocError("lark-cli 安全存储中没有用户 access token")
+    return str(token)
+
+
+def _doc_token(doc_url: str) -> str:
+    path = urlparse(doc_url).path.rstrip("/")
+    token = path.rsplit("/", 1)[-1]
+    if not token:
+        raise ValueError(f"无效飞书文档链接: {doc_url}")
+    return token
+
+
+def parse_category_table_xml(xml: str) -> NativeTable:
+    root = lxml_html.fragment_fromstring(f"<root>{xml}</root>", create_parent=False)
+    heading_nodes = root.xpath(".//h2[1]")
+    heading = "".join(heading_nodes[0].itertext()).strip() if heading_nodes else "品类"
+    tables = root.xpath(".//table")
+    if len(tables) != 1:
+        raise FeishuNativeDocError(f"品类 XML 应只有一张表，实际 {len(tables)} 张")
+    table = tables[0]
+    widths = []
+    for col in table.xpath("./colgroup/col"):
+        try:
+            widths.append(max(50, int(col.get("width") or 100)))
+        except ValueError:
+            widths.append(100)
+
+    row_nodes = table.xpath("./thead/tr") + table.xpath("./tbody/tr")
+    rows: list[list[NativeCell]] = []
+    for row_index, row in enumerate(row_nodes):
+        cells: list[NativeCell] = []
+        for node in row.xpath("./th|./td"):
+            image_nodes = node.xpath(".//img")
+            image_url = ""
+            if image_nodes:
+                image_url = (
+                    image_nodes[0].get("href")
+                    or image_nodes[0].get("src")
+                    or image_nodes[0].get("url")
+                    or ""
+                ).strip()
+            links = node.xpath(".//a[@href]")
+            link_url = (links[0].get("href") or "").strip() if links else ""
+            text = "".join(node.itertext()).strip()
+            cells.append(
+                NativeCell(
+                    text=text,
+                    bold=row_index == 0 or bool(node.xpath(".//b|.//strong")),
+                    link_url=link_url,
+                    image_url=image_url,
+                )
+            )
+        if cells:
+            rows.append(cells)
+    if not rows or not rows[0]:
+        raise FeishuNativeDocError("品类表格没有可写入的行")
+    columns = len(rows[0])
+    if any(len(row) != columns for row in rows):
+        raise FeishuNativeDocError("品类表格存在不规则行")
+    if not widths:
+        widths = [100] * columns
+    widths = (widths + [100] * columns)[:columns]
+    return NativeTable(heading=heading, widths=widths, rows=rows)
+
+
+def _text_elements(cell: NativeCell) -> list[dict[str, Any]]:
+    style: dict[str, Any] = {
+        "bold": bool(cell.bold),
+        "italic": False,
+        "strikethrough": False,
+        "underline": False,
+        "inline_code": False,
+    }
+    if cell.link_url:
+        style["link"] = {"url": quote(cell.link_url, safe="")}
+    return [{"text_run": {"content": cell.text, "text_element_style": style}}]
+
+
+class FeishuNativeDocClient:
+    def __init__(self, token: str) -> None:
+        self.token = token
+        self.session = requests.Session()
+        self.session.headers.update({"Authorization": f"Bearer {token}"})
+
+    @classmethod
+    def from_lark_cli(cls) -> "FeishuNativeDocClient":
+        identity = _refresh_lark_cli_user()
+        token = _read_lark_cli_user_token(
+            str(identity["app_id"]), str(identity["user_open_id"])
+        )
+        return cls(token)
+
+    def request(
+        self,
+        method: str,
+        path: str,
+        *,
+        json_body: dict[str, Any] | None = None,
+        params: dict[str, Any] | None = None,
+        files: dict[str, Any] | None = None,
+        data: dict[str, Any] | None = None,
+        edit: bool = False,
+        media: bool = False,
+        attempts: int = 7,
+        timeout: int = 60,
+    ) -> dict[str, Any]:
+        url = path if path.startswith("http") else OPEN_API + path
+        last_error = ""
+        for attempt in range(1, attempts + 1):
+            if edit:
+                _EDIT_LIMITER.wait()
+            elif media:
+                _MEDIA_LIMITER.wait()
+            try:
+                response = self.session.request(
+                    method,
+                    url,
+                    json=json_body,
+                    params=params,
+                    files=files,
+                    data=data,
+                    timeout=timeout,
+                )
+                payload = response.json()
+            except (requests.RequestException, ValueError) as exc:
+                last_error = str(exc)
+                if attempt < attempts:
+                    time.sleep(min(8, attempt * 1.5))
+                    continue
+                raise FeishuNativeDocError(f"飞书接口请求失败: {last_error}") from exc
+            if response.ok and int(payload.get("code", 0)) == 0:
+                return payload.get("data") or {}
+            code = int(payload.get("code") or response.status_code)
+            last_error = f"HTTP {response.status_code}, code={code}, msg={payload.get('msg')}"
+            if attempt < attempts and (
+                response.status_code in {429, 500, 502, 503, 504}
+                or code in {99991400, 1771001, 1771002, 1771005}
+            ):
+                time.sleep(min(12, attempt * 2))
+                continue
+            raise FeishuNativeDocError(last_error)
+        raise FeishuNativeDocError(last_error)
+
+    def create_descendants(
+        self, document_id: str, parent_id: str, body: dict[str, Any]
+    ) -> dict[str, Any]:
+        return self.request(
+            "POST",
+            f"/docx/v1/documents/{document_id}/blocks/{parent_id}/descendant",
+            json_body=body,
+            params={"document_revision_id": -1, "client_token": str(uuid.uuid4())},
+            edit=True,
+        )
+
+    def insert_table_row(self, document_id: str, table_id: str) -> None:
+        self.request(
+            "PATCH",
+            f"/docx/v1/documents/{document_id}/blocks/batch_update",
+            json_body={
+                "requests": [
+                    {"block_id": table_id, "insert_table_row": {"row_index": -1}}
+                ]
+            },
+            params={"document_revision_id": -1, "client_token": str(uuid.uuid4())},
+            edit=True,
+        )
+
+    def list_blocks(self, document_id: str) -> list[dict[str, Any]]:
+        items: list[dict[str, Any]] = []
+        page_token = ""
+        while True:
+            params: dict[str, Any] = {
+                "document_revision_id": -1,
+                "page_size": 500,
+            }
+            if page_token:
+                params["page_token"] = page_token
+            data = self.request(
+                "GET",
+                f"/docx/v1/documents/{document_id}/blocks",
+                params=params,
+                edit=False,
+            )
+            items.extend(data.get("items") or [])
+            if not data.get("has_more"):
+                return items
+            page_token = str(data.get("page_token") or "")
+
+    def batch_update(self, document_id: str, requests_: list[dict[str, Any]]) -> None:
+        # The API accepts 200 ordinary updates, but image/file resources have
+        # a stricter per-request limit of 20 (error 1770035).
+        chunk_size = 20 if any("replace_image" in item for item in requests_) else 200
+        for offset in range(0, len(requests_), chunk_size):
+            self.request(
+                "PATCH",
+                f"/docx/v1/documents/{document_id}/blocks/batch_update",
+                json_body={"requests": requests_[offset : offset + chunk_size]},
+                params={
+                    "document_revision_id": -1,
+                    "client_token": str(uuid.uuid4()),
+                },
+                edit=True,
+            )
+
+    def create_image_block(self, document_id: str, cell_id: str) -> str:
+        data = self.request(
+            "POST",
+            f"/docx/v1/documents/{document_id}/blocks/{cell_id}/children",
+            json_body={"index": -1, "children": [{"block_type": 27, "image": {}}]},
+            params={"document_revision_id": -1, "client_token": str(uuid.uuid4())},
+            edit=True,
+        )
+        children = data.get("children") or []
+        if not children:
+            raise FeishuNativeDocError("创建图片块后未返回 block_id")
+        return str(children[0]["block_id"])
+
+    def upload_image(self, image_block_id: str, content: bytes) -> str:
+        data = self.request(
+            "POST",
+            "/drive/v1/medias/upload_all",
+            files={"file": ("product.jpg", content, "image/jpeg")},
+            data={
+                "file_name": "product.jpg",
+                "parent_type": "docx_image",
+                "parent_node": image_block_id,
+                "size": str(len(content)),
+            },
+            edit=False,
+            media=True,
+            attempts=8,
+            timeout=90,
+        )
+        token = data.get("file_token")
+        if not token:
+            raise FeishuNativeDocError("上传图片后未返回 file_token")
+        return str(token)
+
+
+def _placeholder_image() -> bytes:
+    image = Image.new("RGB", (DEFAULT_IMAGE_SIZE, DEFAULT_IMAGE_SIZE), "#F2F3F5")
+    draw = ImageDraw.Draw(image)
+    draw.rectangle((8, 8, DEFAULT_IMAGE_SIZE - 8, DEFAULT_IMAGE_SIZE - 8), outline="#C9CDD4")
+    draw.text((27, 52), "NO IMAGE", fill="#8F959E")
+    output = io.BytesIO()
+    image.save(output, format="JPEG", quality=75, optimize=True)
+    return output.getvalue()
+
+
+_PLACEHOLDER = _placeholder_image()
+
+
+def _download_thumbnail(url: str, size: int) -> tuple[bytes, bool]:
+    if not url:
+        return _PLACEHOLDER, True
+    try:
+        response = requests.get(
+            url,
+            headers={"User-Agent": "Mozilla/5.0", "Referer": "https://www.taobao.com/"},
+            timeout=25,
+        )
+        response.raise_for_status()
+        with Image.open(io.BytesIO(response.content)) as image:
+            image = image.convert("RGB")
+            image.thumbnail((size, size), Image.Resampling.LANCZOS)
+            canvas = Image.new("RGB", (size, size), "white")
+            canvas.paste(image, ((size - image.width) // 2, (size - image.height) // 2))
+            output = io.BytesIO()
+            canvas.save(output, format="JPEG", quality=78, optimize=True)
+            return output.getvalue(), False
+    except Exception:
+        return _PLACEHOLDER, True
+
+
+def _initial_row_count(table: NativeTable) -> int:
+    # Every cell needs one child block. Reserve one block each for heading/table.
+    per_row = table.column_count * 2
+    return max(
+        1,
+        min(len(table.rows), (STABLE_INITIAL_DESCENDANTS - 2) // per_row),
+    )
+
+
+def _build_initial_descendants(
+    table: NativeTable, initial_rows: int
+) -> tuple[dict[str, Any], dict[str, tuple[int, int]], dict[str, str]]:
+    descendants: list[dict[str, Any]] = []
+    heading_id = "heading"
+    table_id = "table"
+    descendants.append(
+        {
+            "block_id": heading_id,
+            "block_type": 4,
+            "heading2": {"elements": _text_elements(NativeCell(table.heading, bold=True))},
+            "children": [],
+        }
+    )
+    cell_ids: list[str] = []
+    image_temp_to_position: dict[str, tuple[int, int]] = {}
+    text_temp_to_position: dict[str, str] = {}
+    cell_descendants: list[dict[str, Any]] = []
+    for row_index, row in enumerate(table.rows[:initial_rows]):
+        for column_index, cell in enumerate(row):
+            cell_id = f"c_{row_index}_{column_index}"
+            child_id = f"x_{row_index}_{column_index}"
+            cell_ids.append(cell_id)
+            if cell.image_url:
+                image_temp_to_position[child_id] = (row_index, column_index)
+                child = {
+                    "block_id": child_id,
+                    "block_type": 27,
+                    "image": {},
+                    "children": [],
+                }
+            else:
+                text_temp_to_position[child_id] = f"{row_index}:{column_index}"
+                child = {
+                    "block_id": child_id,
+                    "block_type": 2,
+                    "text": {"elements": _text_elements(cell)},
+                    "children": [],
+                }
+            cell_descendants.extend(
+                [
+                    {
+                        "block_id": cell_id,
+                        "block_type": 32,
+                        "table_cell": {},
+                        "children": [child_id],
+                    },
+                    child,
+                ]
+            )
+    descendants.append(
+        {
+            "block_id": table_id,
+            "block_type": 31,
+            "table": {
+                "property": {
+                    "row_size": initial_rows,
+                    "column_size": table.column_count,
+                    "column_width": table.widths,
+                    "header_row": True,
+                }
+            },
+            "children": cell_ids,
+        }
+    )
+    descendants.extend(cell_descendants)
+    if len(descendants) > MAX_DESCENDANTS:
+        raise FeishuNativeDocError(f"原生表格初始块超过上限: {len(descendants)}")
+    return (
+        {"index": -1, "children_id": [heading_id, table_id], "descendants": descendants},
+        image_temp_to_position,
+        text_temp_to_position,
+    )
+
+
+def append_native_category_table(
+    doc_url: str,
+    xml: str,
+    *,
+    client: FeishuNativeDocClient | None = None,
+    image_size: int = DEFAULT_IMAGE_SIZE,
+) -> dict[str, Any]:
+    client = client or FeishuNativeDocClient.from_lark_cli()
+    document_id = _doc_token(doc_url)
+    table = parse_category_table_xml(xml)
+    table_cells = len(table.rows) * table.column_count
+    if table_cells > MAX_TABLE_CELLS:
+        raise FeishuNativeDocError(
+            f"{table.heading} 表格共 {table_cells} 个单元格，"
+            f"超过飞书单表 {MAX_TABLE_CELLS} 个单元格上限"
+        )
+    initial_rows = _initial_row_count(table)
+    body, initial_image_positions, _ = _build_initial_descendants(table, initial_rows)
+    created = client.create_descendants(document_id, document_id, body)
+    relations = {
+        item["temporary_block_id"]: item["block_id"]
+        for item in created.get("block_id_relations") or []
+    }
+    table_id = relations.get("table")
+    if not table_id:
+        raise FeishuNativeDocError("创建品类表后未返回 table block_id")
+    print(
+        f"[NATIVE] {table.heading}: 已创建初始 {initial_rows}/{len(table.rows)} 行，"
+        f"继续扩容同一张表..."
+    )
+    for row_number in range(initial_rows, len(table.rows)):
+        client.insert_table_row(document_id, table_id)
+        if (row_number + 1) % 25 == 0 or row_number + 1 == len(table.rows):
+            print(f"[NATIVE] {table.heading}: 表格行 {row_number + 1}/{len(table.rows)}")
+
+    blocks = client.list_blocks(document_id)
+    block_by_id = {str(item.get("block_id")): item for item in blocks}
+    live_table = block_by_id.get(table_id) or {}
+    cell_ids = list((live_table.get("table") or {}).get("cells") or live_table.get("children") or [])
+    expected_cells = len(table.rows) * table.column_count
+    if len(cell_ids) != expected_cells:
+        raise FeishuNativeDocError(
+            f"{table.heading} 单元格数不一致: {len(cell_ids)}/{expected_cells}"
+        )
+
+    text_updates: list[dict[str, Any]] = []
+    image_jobs: list[tuple[str, str]] = []
+    # Initial rows already contain text/image blocks. Resolve initial image IDs.
+    for temp_id, (row_index, column_index) in initial_image_positions.items():
+        image_jobs.append((relations[temp_id], table.rows[row_index][column_index].image_url))
+
+    for row_index in range(initial_rows, len(table.rows)):
+        for column_index, cell in enumerate(table.rows[row_index]):
+            cell_id = str(cell_ids[row_index * table.column_count + column_index])
+            cell_block = block_by_id.get(cell_id) or {}
+            children = cell_block.get("children") or []
+            text_id = str(children[0]) if children else ""
+            if cell.image_url:
+                image_id = client.create_image_block(document_id, cell_id)
+                image_jobs.append((image_id, cell.image_url))
+            elif cell.text and text_id:
+                text_updates.append(
+                    {
+                        "block_id": text_id,
+                        "update_text_elements": {"elements": _text_elements(cell)},
+                    }
+                )
+    client.batch_update(document_id, text_updates)
+    print(
+        f"[NATIVE] {table.heading}: 文字已写入，开始上传 {len(image_jobs)} 张缩略图..."
+    )
+
+    def upload(job: tuple[str, str]) -> tuple[str, str, bool]:
+        image_id, url = job
+        content, placeholder = _download_thumbnail(url, image_size)
+        token = client.upload_image(image_id, content)
+        return image_id, token, placeholder
+
+    uploaded: list[tuple[str, str, bool]] = []
+    with ThreadPoolExecutor(max_workers=4) as pool:
+        futures = [pool.submit(upload, job) for job in image_jobs]
+        for completed, future in enumerate(as_completed(futures), 1):
+            uploaded.append(future.result())
+            if completed % 50 == 0 or completed == len(futures):
+                print(f"[NATIVE] {table.heading}: 图片 {completed}/{len(futures)}")
+    replace_requests = [
+        {
+            "block_id": image_id,
+            "replace_image": {
+                "token": token,
+                "width": image_size,
+                "height": image_size,
+            },
+        }
+        for image_id, token, _ in uploaded
+    ]
+    client.batch_update(document_id, replace_requests)
+    placeholder_count = sum(1 for _, _, placeholder in uploaded if placeholder)
+    return {
+        "heading": table.heading,
+        "rows": len(table.rows) - 1,
+        "images": len(image_jobs),
+        "placeholder_images": placeholder_count,
+        "table_block_id": table_id,
+    }
+
+
+def append_native_category_tables(
+    doc_url: str,
+    category_xmls: list[str],
+    *,
+    image_size: int = DEFAULT_IMAGE_SIZE,
+) -> list[dict[str, Any]]:
+    client = FeishuNativeDocClient.from_lark_cli()
+    results = []
+    for index, xml in enumerate(category_xmls, 1):
+        result = append_native_category_table(
+            doc_url, xml, client=client, image_size=image_size
+        )
+        results.append(result)
+        print(
+            f"[OK] 原生飞书品类表 {index}/{len(category_xmls)}: "
+            f"{result['heading']} ({result['rows']} 条)"
+        )
+    return results
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/import_product_daily.py b/src/gyxx_flow/modules/product_commerce/runtime/import_product_daily.py
new file mode 100644
index 0000000..017a99b
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/import_product_daily.py
@@ -0,0 +1,579 @@
+"""
+扫描 data/{dy,jd,tm}单品数据/_downloads/<YYYY-MM-DD>/ 下的原始报表,
+按三平台字段映射抽出通用列, 其余平台特有字段进 raw_data JSONB, 写入
+product_daily_metrics 表 (ON CONFLICT 全量覆盖)。
+
+用法:
+    .venv\\Scripts\\python.exe import_product_daily.py                    # 全量 (所有日期)
+    .venv\\Scripts\\python.exe import_product_daily.py --date 2026-05-15  # 单日
+    .venv\\Scripts\\python.exe import_product_daily.py --from 2026-06-01 --to 2026-06-16
+    .venv\\Scripts\\python.exe import_product_daily.py --platforms dy,jd
+    .venv\\Scripts\\python.exe import_product_daily.py --dry-run         # 只解析不入库
+
+源数据列数 (已验证):
+    dy xlsx: 22 列 (row 0 = 表头, 之后是数据/合计行)
+    jd xlsx: 34 列 (row 0 = 表头, row 1 = 合计行, 之后是数据)
+    tm xls:  31/38 列 (row 0-3 空白, row 4 = 表头, row 5+ = 数据)
+"""
+from __future__ import annotations
+
+import argparse
+import os
+import re
+import sys
+from datetime import date, datetime
+from decimal import Decimal, InvalidOperation
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT
+from typing import Any, Iterable
+
+import openpyxl
+import xlrd
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+from db import get_conn, upsert_product_daily_metrics  # noqa: E402
+
+
+# ── 路径配置 ────────────────────────────────────────────────────────────────
+DATA_ROOT = RAW_DATA_ROOT
+
+PLATFORM_DIRS = {
+    "dy": "dy单品数据",
+    "jd": "jd单品数据",
+    "tm": "tm单品数据",
+}
+
+VALID_PLATFORMS = tuple(PLATFORM_DIRS.keys())
+
+
+# ── 字符串解析 helpers ─────────────────────────────────────────────────────
+
+_NUM_RE = re.compile(r"^-?[\d,]+\.?\d*$")
+_PCT_RE = re.compile(r"^(-?[\d.]+)\s*%$")
+
+
+def _strip(v: Any) -> str:
+    if v is None:
+        return ""
+    return str(v).strip()
+
+
+def _to_int(v: Any) -> int | None:
+    """'2,153' / '2153' / 2153 / None → int|None"""
+    s = _strip(v)
+    if not s or s == "-":
+        return None
+    if _NUM_RE.match(s):
+        try:
+            return int(s.replace(",", "").split(".")[0])
+        except ValueError:
+            return None
+    return None
+
+
+def _to_num(v: Any) -> Decimal | None:
+    """'2,153.50' / 2153.5 / None → Decimal|None"""
+    s = _strip(v)
+    if not s or s == "-":
+        return None
+    if _NUM_RE.match(s):
+        try:
+            return Decimal(s.replace(",", ""))
+        except InvalidOperation:
+            return None
+    return None
+
+
+def _to_pct(v: Any) -> Decimal | None:
+    """'63.54%' / 0.6354 / None → Decimal(0.6354) (比率 0~1)"""
+    s = _strip(v)
+    if not s or s == "-":
+        return None
+    m = _PCT_RE.match(s)
+    if m:
+        try:
+            return Decimal(m.group(1)) / Decimal(100)
+        except InvalidOperation:
+            return None
+    # 已是小数形式
+    if _NUM_RE.match(s):
+        try:
+            return Decimal(s.replace(",", ""))
+        except InvalidOperation:
+            return None
+    return None
+
+
+def _to_date(v: Any) -> date | None:
+    """'2026/05/01' / '2026-05-01' / '2026年05月01日' / datetime/date → date|None"""
+    if v is None or v == "":
+        return None
+    if isinstance(v, datetime):
+        return v.date()
+    if isinstance(v, date):
+        return v
+    s = _strip(v)
+    if not s or s == "-":
+        return None
+    s = s.replace("年", "-").replace("月", "-").replace("日", "")
+    for fmt in ("%Y-%m-%d", "%Y/%m/%d", "%Y-%m-%d %H:%M:%S"):
+        try:
+            return datetime.strptime(s, fmt).date()
+        except ValueError:
+            continue
+    return None
+
+
+# ── 文件名 → 字段抽取 ─────────────────────────────────────────────────────
+
+def _jd_shop_id_from_filename(fn: str) -> str | None:
+    """'11272270_20260501_全部渠道_商品明细_分天下载.xlsx' → '11272270'"""
+    m = re.match(r"^(\d+)_", fn)
+    return m.group(1) if m else None
+
+
+def _is_total_or_invalid_row(product_id: str, stat_date: date | None) -> bool:
+    """过滤 合计行 (商品ID=合计/-) 和无日期行"""
+    if not product_id or product_id in {"-", "合计", "总计", ""}:
+        return True
+    if stat_date is None:
+        return True
+    return False
+
+
+# ── dy 解析 ───────────────────────────────────────────────────────────────
+
+def parse_dy_xlsx(xlsx_path: Path) -> Iterable[dict]:
+    """抖音经营版_商品_商品列表.xlsx → record dicts.
+
+    列顺序 (0-indexed):
+      0  统计周期
+      1  商品名称
+      2  商品编码
+      3  载体
+      4  品类
+      5  自卖/合作
+      6  成交金额       → paid_amount
+      7  成交订单数     → paid_qty
+      8  退款金额(支付时间) → refund_amount
+      9  商品点击人数   → visitors
+      10 成交人数       → paid_buyers
+      11 退款订单数(支付时间) → refund_orders
+      12 加购人数       → add_cart_buyers
+      13 商品曝光人数   → page_views
+      14 退款率(支付时间)
+      15 商品曝光次数
+      16 点击支付率（人数） → conversion_rate
+      17 加购支付转化率
+      18 访问加购转化率
+      19 成交客单价     → avg_order_value
+      20 商详页曝光点击率 (注意: 不是跳出率, 留 raw_data)
+      21 商详页曝光人数
+    """
+    wb = openpyxl.load_workbook(xlsx_path, data_only=True)
+    ws = wb.active
+
+    # 抽 add_cart_qty (dy 没有, 加购人数作 fallback)
+    rows = list(ws.iter_rows(values_only=True))
+    if not rows:
+        return
+    # row 0 = 表头, row 1+ = 数据
+    for r in rows[1:]:
+        if not r or r[0] is None:
+            continue
+        stat_date = _to_date(r[0])
+        product_id = _strip(r[2])
+        if _is_total_or_invalid_row(product_id, stat_date):
+            continue
+
+        raw = {
+            "载体": r[3],
+            "品类": r[4],
+            "自卖/合作": r[5],
+            "退款率(支付时间)": r[14],
+            "商品曝光次数": r[15],
+            "加购支付转化率": r[17],
+            "访问加购转化率": r[18],
+            "商详页曝光点击率": r[20],
+            "商详页曝光人数": r[21],
+        }
+
+        yield {
+            "platform": "dy",
+            "stat_date": stat_date,
+            "shop_id": None,
+            "product_id": product_id,
+            "product_name": _strip(r[1]) or None,
+            "product_no": None,                  # dy 无货号
+            "category_l1": _strip(r[4]) or None, # 品类
+            "category_l2": _strip(r[3]) or None, # 载体
+            "visitors": _to_int(r[9]),
+            "page_views": _to_int(r[13]),
+            "add_cart_qty": None,                # dy 没拆件数
+            "add_cart_buyers": _to_int(r[12]),
+            "paid_amount": _to_num(r[6]),
+            "paid_qty": _to_int(r[7]),
+            "paid_buyers": _to_int(r[10]),
+            "refund_amount": _to_num(r[8]),
+            "refund_orders": _to_int(r[11]),
+            "conversion_rate": _to_pct(r[16]),
+            "avg_order_value": _to_num(r[19]),
+            "uv_value": None,                    # dy 无
+            "detail_page_bounce_rate": None,     # dy 的是曝光点击率,语义不同,留 raw
+            "raw_data": {k: v for k, v in raw.items() if v not in (None, "", "-")},
+            "source_file": xlsx_path.name,
+        }
+
+
+# ── jd 解析 ───────────────────────────────────────────────────────────────
+
+def parse_jd_xlsx(xlsx_path: Path, shop_id: str) -> Iterable[dict]:
+    """京东 全部渠道_商品明细_分天下载.xlsx → record dicts.
+
+    列顺序 (0-indexed):
+      0  时间
+      1  商品ID
+      2  商品名称
+      3  二级类目      → category_l1
+      4  三级类目      → category_l2
+      5  货号          → product_no
+      6  访客数        → visitors
+      7  浏览量        → page_views
+      8  商品关注数
+      9  加购商品件数  → add_cart_qty
+      10 加购人数      → add_cart_buyers
+      11 成交客户数    → paid_buyers
+      12 成交单量      → raw (订单数,非件数)
+      13 成交商品件数  → paid_qty
+      14 成交金额      → paid_amount
+      15 成交转化率    → conversion_rate
+      16 UV价值        → uv_value
+      17 最近上架时间
+      18 详情页跳出率  → detail_page_bounce_rate
+      19 搜索曝光量
+      20 搜索点击次数
+      21 搜索点击率
+      22 系统预测未来7天销量
+      23 下单客户数
+      24 下单商品件数
+      25 下单金额
+      26 下单成交转化率
+      27 下单转化率
+      28 PV现货率
+      29 客单价        → avg_order_value
+      30 件单价
+      31 取消及售后退款金额 → refund_amount
+      32 取消及售后退款件数 → refund_orders
+      33 取消及售后退款单量
+    """
+    wb = openpyxl.load_workbook(xlsx_path, data_only=True)
+    ws = wb.active
+
+    rows = list(ws.iter_rows(values_only=True))
+    if not rows:
+        return
+    for r in rows[1:]:
+        if not r or r[0] is None:
+            continue
+        stat_date = _to_date(r[0])
+        product_id = _strip(r[1])
+        if _is_total_or_invalid_row(product_id, stat_date):
+            continue
+
+        raw = {
+            "商品关注数": r[8],
+            "成交单量": r[12],
+            "最近上架时间": _strip(r[17]) if r[17] is not None else None,
+            "搜索曝光量": r[19],
+            "搜索点击次数": r[20],
+            "搜索点击率": r[21],
+            "系统预测未来7天销量": r[22],
+            "下单客户数": r[23],
+            "下单商品件数": r[24],
+            "下单金额": r[25],
+            "下单成交转化率": r[26],
+            "下单转化率": r[27],
+            "PV现货率": r[28],
+            "件单价": r[30],
+            "取消及售后退款单量": r[33],
+        }
+
+        yield {
+            "platform": "jd",
+            "stat_date": stat_date,
+            "shop_id": shop_id,
+            "product_id": product_id,
+            "product_name": _strip(r[2]) or None,
+            "product_no": _strip(r[5]) or None,
+            "category_l1": _strip(r[3]) or None,
+            "category_l2": _strip(r[4]) or None,
+            "visitors": _to_int(r[6]),
+            "page_views": _to_int(r[7]),
+            "add_cart_qty": _to_int(r[9]),
+            "add_cart_buyers": _to_int(r[10]),
+            "paid_amount": _to_num(r[14]),
+            "paid_qty": _to_int(r[13]),
+            "paid_buyers": _to_int(r[11]),
+            "refund_amount": _to_num(r[31]),
+            "refund_orders": _to_int(r[32]),
+            "conversion_rate": _to_pct(r[15]),
+            "avg_order_value": _to_num(r[29]),
+            "uv_value": _to_num(r[16]),
+            "detail_page_bounce_rate": _to_pct(r[18]),
+            "raw_data": {k: v for k, v in raw.items() if v not in (None, "", "-")},
+            "source_file": xlsx_path.name,
+        }
+
+
+# ── tm 解析 ───────────────────────────────────────────────────────────────
+
+def parse_tm_xls(xls_path: Path) -> Iterable[dict]:
+    """天猫 【生意参谋平台】商品_全部_<date>_<date>.xls → record dicts.
+
+    表头在 row index 4, 数据 row 5+.
+    列顺序 (0-indexed):
+      0  统计日期
+      1  商品ID
+      2  商品名称
+      3  主商品ID
+      4  商品类型      → category_l2
+      5  货号          → product_no
+      6  商品状态
+      7  商品标签
+      8  商品访客数    → visitors
+      9  商品浏览量    → page_views
+      10 平均停留时长
+      11 商品详情页跳出率 → detail_page_bounce_rate
+      12 商品收藏人数
+      13 商品加购件数  → add_cart_qty
+      14 商品加购人数  → add_cart_buyers
+      15 下单买家数
+      16 下单件数
+      17 下单金额
+      18 下单转化率
+      19 支付买家数    → paid_buyers
+      20 支付件数      → paid_qty
+      21 支付金额      → paid_amount
+      22 商品支付转化率 → conversion_rate
+      23 支付新买家数
+      24 支付老买家数
+      25 老买家支付金额
+      26 聚划算支付金额
+      27 访客平均价值  → uv_value
+      28 成功退款金额  → refund_amount
+      29 竞争力评分
+      30 年累计支付金额
+      31 月累计支付金额
+      32 月累计支付件数
+      33 搜索引导支付转化率
+      34 搜索引导访客数
+      35 搜索引导支付买家数
+      36 结构化详情引导转化率
+      37 结构化详情引导成交占比
+    """
+    wb = xlrd.open_workbook(xls_path)
+    ws = wb.sheet_by_index(0)
+    headers = [_strip(value) for value in ws.row_values(4)]
+    header_index = {
+        header: index for index, header in enumerate(headers) if header
+    }
+    required_headers = {
+        "统计日期", "商品ID", "商品名称", "货号", "商品访客数", "商品浏览量",
+        "商品加购件数", "商品加购人数", "支付买家数", "支付件数", "支付金额",
+        "商品支付转化率", "访客平均价值", "成功退款金额",
+        "商品详情页跳出率",
+    }
+    missing_headers = sorted(required_headers - header_index.keys())
+    if missing_headers:
+        raise ValueError(
+            f"天猫商品日报表头不完整：{xls_path.name} 缺少 "
+            f"{', '.join(missing_headers)}"
+        )
+
+    def value(row: list[Any], header: str) -> Any:
+        index = header_index.get(header)
+        if index is None or index >= len(row):
+            return ""
+        return row[index]
+
+    raw_headers = (
+        "主商品ID", "商品状态", "商品标签", "平均停留时长", "商品收藏人数",
+        "下单买家数", "下单件数", "下单金额", "下单转化率", "支付新买家数",
+        "支付老买家数", "老买家支付金额", "聚划算支付金额", "竞争力评分",
+        "年累计支付金额", "月累计支付金额", "月累计支付件数",
+        "搜索引导支付转化率", "搜索引导访客数", "搜索引导支付买家数",
+        "结构化详情引导转化率", "结构化详情引导成交占比",
+    )
+
+    # row 0-3 通常空白, row 4 = 表头
+    for r_idx in range(5, ws.nrows):
+        r = ws.row_values(r_idx)
+        if not r or not value(r, "统计日期"):
+            continue
+        stat_date = _to_date(value(r, "统计日期"))
+        product_id = _strip(value(r, "商品ID"))
+        if _is_total_or_invalid_row(product_id, stat_date):
+            continue
+
+        raw = {header: value(r, header) for header in raw_headers}
+
+        yield {
+            "platform": "tm",
+            "stat_date": stat_date,
+            "shop_id": None,
+            "product_id": product_id,
+            "product_name": _strip(value(r, "商品名称")) or None,
+            "product_no": _strip(value(r, "货号")) or None,
+            "category_l1": None,                # tm 无二级类目
+            "category_l2": _strip(value(r, "商品类型")) or None,
+            "visitors": _to_int(value(r, "商品访客数")),
+            "page_views": _to_int(value(r, "商品浏览量")),
+            "add_cart_qty": _to_int(value(r, "商品加购件数")),
+            "add_cart_buyers": _to_int(value(r, "商品加购人数")),
+            "paid_amount": _to_num(value(r, "支付金额")),
+            "paid_qty": _to_int(value(r, "支付件数")),
+            "paid_buyers": _to_int(value(r, "支付买家数")),
+            "refund_amount": _to_num(value(r, "成功退款金额")),
+            "refund_orders": None,              # tm 无
+            "conversion_rate": _to_pct(value(r, "商品支付转化率")),
+            "avg_order_value": None,            # tm 无客单价
+            "uv_value": _to_num(value(r, "访客平均价值")),
+            "detail_page_bounce_rate": _to_pct(value(r, "商品详情页跳出率")),
+            "raw_data": {k: v for k, v in raw.items() if v not in (None, "", "-")},
+            "source_file": xls_path.name,
+        }
+
+
+# ── 扫描 + 入库 ───────────────────────────────────────────────────────────
+
+def list_date_dirs(platform: str) -> list[Path]:
+    """列出 _downloads 下所有 YYYY-MM-DD 子目录 (排除 _bak_* 备份)。"""
+    p = DATA_ROOT / PLATFORM_DIRS[platform] / "_downloads"
+    if not p.is_dir():
+        return []
+    out = []
+    for d in sorted(p.iterdir()):
+        if d.is_dir() and re.match(r"^\d{4}-\d{2}-\d{2}$", d.name):
+            out.append(d)
+    return out
+
+
+def find_target_files(date_dir: Path, platform: str) -> list[Path]:
+    """找该日期目录下所有要读的源文件 (排除 ~$ 临时文件)。
+
+    注意: 某些日期目录会同时存前一天和当天的文件 (采集时间错位),
+    全部读入后用文件内容里的 stat_date 落库, ON CONFLICT 去重。
+    """
+    out = []
+    for f in date_dir.iterdir():
+        if f.name.startswith("~$"):
+            continue
+        if platform == "dy" and f.suffix.lower() == ".xlsx" and "经营版" in f.name:
+            out.append(f)
+        elif platform == "jd" and f.suffix.lower() == ".xlsx" and "商品明细" in f.name:
+            out.append(f)
+        elif platform == "tm" and f.suffix.lower() == ".xls" and "商品_全部" in f.name:
+            out.append(f)
+    return out
+
+
+def parse_records_for_date(platform: str, date_dir: Path) -> list[dict]:
+    """处理目录下所有匹配文件, 累加 recs。"""
+    recs: list[dict] = []
+    for f in find_target_files(date_dir, platform):
+        if platform == "dy":
+            recs.extend(parse_dy_xlsx(f))
+        elif platform == "jd":
+            shop = _jd_shop_id_from_filename(f.name)
+            if shop:
+                recs.extend(parse_jd_xlsx(f, shop))
+        elif platform == "tm":
+            recs.extend(parse_tm_xls(f))
+    return recs
+
+
+def filter_dates(dirs: list[Path], from_d: date | None, to_d: date | None) -> list[Path]:
+    out = []
+    for d in dirs:
+        try:
+            dd = datetime.strptime(d.name, "%Y-%m-%d").date()
+        except ValueError:
+            continue
+        if from_d and dd < from_d:
+            continue
+        if to_d and dd > to_d:
+            continue
+        out.append(d)
+    return out
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser(
+        description="导入三平台单品日报表到 product_daily_metrics (upsert 幂等)",
+    )
+    ap.add_argument("--date", help="单日 YYYY-MM-DD, 等同 --from X --to X")
+    ap.add_argument("--from", dest="from_d", help="起始日期 YYYY-MM-DD")
+    ap.add_argument("--to", dest="to_d", help="结束日期 YYYY-MM-DD")
+    ap.add_argument(
+        "--platforms",
+        default=",".join(VALID_PLATFORMS),
+        help=f"逗号分隔, 限定平台, 默认 dy,jd,tm; 合法: {VALID_PLATFORMS}",
+    )
+    ap.add_argument("--dry-run", action="store_true", help="只解析,不入库")
+    args = ap.parse_args()
+
+    platforms = [p.strip() for p in args.platforms.split(",") if p.strip()]
+    for p in platforms:
+        if p not in VALID_PLATFORMS:
+            ap.error(f"未知平台 {p!r}, 合法: {VALID_PLATFORMS}")
+
+    from_d = to_d = None
+    if args.date:
+        from_d = to_d = datetime.strptime(args.date, "%Y-%m-%d").date()
+    else:
+        if args.from_d:
+            from_d = datetime.strptime(args.from_d, "%Y-%m-%d").date()
+        if args.to_d:
+            to_d = datetime.strptime(args.to_d, "%Y-%m-%d").date()
+        if from_d and to_d and from_d > to_d:
+            ap.error("--from 不能晚于 --to")
+
+    print(f"[INFO] 平台: {platforms}")
+    print(f"[INFO] 日期: {from_d or '(最早)'} ~ {to_d or '(最新)'}")
+    print(f"[INFO] 模式: {'DRY-RUN (不入库)' if args.dry_run else '入库'}")
+
+    grand_total = 0
+    t0 = datetime.now()
+    for platform in platforms:
+        all_dirs = list_date_dirs(platform)
+        target_dirs = filter_dates(all_dirs, from_d, to_d)
+        print(f"\n── {platform}: 共 {len(target_dirs)} 个日期目录 ──")
+
+        per_plat_total = 0
+        per_plat_files = 0
+        for dd in target_dirs:
+            recs = parse_records_for_date(platform, dd)
+            if not recs:
+                continue
+            per_plat_files += 1
+            per_plat_total += len(recs)
+            if not args.dry_run:
+                with get_conn() as conn:
+                    n = upsert_product_daily_metrics(conn, recs)
+                print(f"  [{platform}] {dd.name}: {n} 行 (file={recs[0]['source_file']})")
+            else:
+                print(f"  [{platform}] {dd.name}: parsed {len(recs)} 行 (dry-run)")
+        print(f"  [{platform}] 小计: {per_plat_total} 行 / {per_plat_files} 文件")
+        grand_total += per_plat_total
+
+    print(f"\n{'='*60}")
+    print(f"  全部合计: {grand_total} 行  (耗时 {(datetime.now()-t0).total_seconds():.1f}s)")
+    print(f"{'='*60}")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/import_product_reviews.py b/src/gyxx_flow/modules/product_commerce/runtime/import_product_reviews.py
new file mode 100644
index 0000000..5d930df
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/import_product_reviews.py
@@ -0,0 +1,331 @@
+"""
+扫描 data/{jd,dy}单品数据/评价*.json (或 评论汇总_*.json),
+归一化后写入 product_reviews 表 (ON CONFLICT 全量覆盖)。
+
+DY 评价入库时会反查 dim_style.dy_spus 填 style_name;
+JD 评价缺 SPU->款式映射, style_name 留 NULL (待 dim_jd_sku_spu 映射表补齐后再做)。
+
+用法:
+    .venv\\Scripts\\python.exe import_product_reviews.py                    # 全量 (默认只 dy)
+    .venv/Scripts/python.exe import_product_reviews.py --date 2026-07-13   # 单日
+    .venv/Scripts/python.exe import_product_reviews.py --from 2026-06-01 --to 2026-07-13
+    .venv/Scripts/python.exe import_product_reviews.py --platforms dy
+    .venv/Scripts/python.exe import_product_reviews.py --dry-run          # 只解析不入库
+
+源 JSON 文件:
+    JD: data/jd单品数据/评价汇总_<YYYY-MM-DD>.json
+        结构: {platform, shop, collect_date, comment_date_range, captured_at,
+               total_count, ratings: {好评/中评/差评: {collected_count, comments:[{
+                 user_name, score, tags, content, time, order_id,
+                 sku_name, sku_id, sku_size, rating}]}}}
+    DY: data/dy单品数据/评论汇总_<YYYYMMDD>.json
+        结构: {shop, period, captured_at, url, total_collected_count,
+               ratings: {好评/中评/差评: {expected_count, collected_count, comments:[{
+                 rating, order_id, comment_time, comment_text, product_name,
+                 product_id, spec_name, spec_id, raw_text}]}}}
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import sys
+from datetime import date, datetime
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT, artifact_relative_path
+from typing import Any, Iterable
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+from db import get_conn, upsert_product_reviews  # noqa: E402
+
+
+DATA_ROOT = RAW_DATA_ROOT
+
+PLATFORM_DIRS = {
+    "jd": "jd单品数据",
+    "dy": "dy单品数据",
+}
+
+# 文件名模式: JD 用 评价汇总_2026-07-14.json, DY 用 评论汇总_20260604.json
+FILE_PATTERNS = {
+    "jd": re.compile(r"^评价汇总_(\d{4}-\d{2}-\d{2})\.json$"),
+    "dy": re.compile(r"^评论汇总_(\d{8})\.json$"),
+}
+
+VALID_PLATFORMS = tuple(PLATFORM_DIRS.keys())
+
+RATING_VALUES = ("好评", "中评", "差评")
+
+
+def _parse_jd_date(s: str) -> date:
+    return datetime.strptime(s, "%Y-%m-%d").date()
+
+
+def _parse_dy_date(s: str) -> date:
+    return datetime.strptime(s, "%Y%m%d").date()
+
+
+def _parse_comment_date(s: str) -> date | None:
+    """从评价时间字符串抽日期部分。
+
+    JD time 格式: '2026-07-13 12:34:56' 或 '2026-07-13'
+    DY comment_time 格式: '2026/06/05 16:39:18' 或 '2026/06/05'
+    """
+    if not s:
+        return None
+    s = s.strip()
+    for fmt in ("%Y-%m-%d %H:%M:%S", "%Y-%m-%d", "%Y/%m/%d %H:%M:%S", "%Y/%m/%d"):
+        try:
+            return datetime.strptime(s, fmt).date()
+        except ValueError:
+            continue
+    m = re.match(r"^(\d{4})[-/](\d{1,2})[-/](\d{1,2})", s)
+    if m:
+        try:
+            return date(int(m.group(1)), int(m.group(2)), int(m.group(3)))
+        except ValueError:
+            return None
+    return None
+
+
+def _to_smallint(v: Any) -> int | None:
+    if v is None or v == "":
+        return None
+    try:
+        return int(float(str(v)))
+    except (ValueError, TypeError):
+        return None
+
+
+def _normalize_jd(payload: dict, source_file: str, collect_date: date,
+                  style_map: dict[str, str] | None = None) -> list[dict]:
+    """JD JSON -> product_reviews 记录列表
+
+    JD 评价表只有 sku_id (无 SPU), 暂不反查 style_name, 留 NULL。
+    """
+    records = []
+    for rating in RATING_VALUES:
+        bucket = (payload.get("ratings") or {}).get(rating) or {}
+        for c in bucket.get("comments") or []:
+            comment_date = _parse_comment_date(c.get("time", ""))
+            records.append({
+                "platform": "jd",
+                "collect_date": collect_date,
+                "comment_date": comment_date,
+                "style_name": None,  # JD sku_id -> SPU 映射暂缺, 待 dim_jd_sku_spu 补齐
+                "rating": rating,
+                "score": _to_smallint(c.get("score")),
+                "content": (c.get("content") or "").strip() or None,
+                "tags": [t for t in (c.get("tags") or []) if t] or None,
+                "user_name": c.get("user_name") or None,
+                "order_id": str(c.get("order_id") or "").strip(),
+                "product_id": None,  # JD 评价表只有 sku_id, SPU 留空
+                "product_name": c.get("sku_name") or None,
+                "sku_id": c.get("sku_id") or None,
+                "sku_name": c.get("sku_name") or None,
+                "sku_size": c.get("sku_size") or None,
+                "raw_data": c,
+                "source_file": source_file,
+            })
+    return records
+
+
+def _normalize_dy(payload: dict, source_file: str, collect_date: date,
+                  style_map: dict[str, str] | None = None) -> list[dict]:
+    """DY JSON -> product_reviews 记录列表
+
+    style_map: {product_id: style_name}, 从 dim_style.dy_spus 构建。
+    匹配不上时 style_name = None (表示 dim_style 缺该商品 ID 维护)。
+    """
+    style_map = style_map or {}
+    records = []
+    for rating in RATING_VALUES:
+        bucket = (payload.get("ratings") or {}).get(rating) or {}
+        for c in bucket.get("comments") or []:
+            comment_date = _parse_comment_date(c.get("comment_time", ""))
+            product_id = str(c.get("product_id") or "").strip()
+            records.append({
+                "platform": "dy",
+                "collect_date": collect_date,
+                "comment_date": comment_date,
+                "style_name": style_map.get(product_id),
+                "rating": rating,
+                "score": None,  # DY 无星级
+                "content": (c.get("comment_text") or "").strip() or None,
+                "tags": None,
+                "user_name": None,  # DY 评论无独立用户名字段
+                "order_id": str(c.get("order_id") or "").strip(),
+                "product_id": product_id or None,
+                "product_name": c.get("product_name") or None,
+                "sku_id": c.get("spec_id") or None,
+                "sku_name": c.get("spec_name") or None,
+                "sku_size": None,
+                "raw_data": c,
+                "source_file": source_file,
+            })
+    return records
+
+
+NORMALIZERS = {
+    "jd": _normalize_jd,
+    "dy": _normalize_dy,
+}
+
+
+def _load_dy_style_map() -> dict[str, str]:
+    """从 dim_style.dy_spus 构建 {product_id: style_name} 映射。"""
+    mapping: dict[str, str] = {}
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute("SELECT style_name, dy_spus FROM dim_style WHERE dy_spus IS NOT NULL")
+            for style_name, spus in cur.fetchall():
+                if not spus:
+                    continue
+                for pid in spus:
+                    mapping[str(pid)] = style_name
+    return mapping
+
+
+def _discover_files(platforms: list[str], date_from: date | None, date_to: date | None) -> list[tuple[str, date, Path]]:
+    """返回 [(platform, collect_date, file_path), ...]"""
+    out = []
+    for plat in platforms:
+        plat_dir = DATA_ROOT / PLATFORM_DIRS[plat]
+        if not plat_dir.exists():
+            continue
+        pattern = FILE_PATTERNS[plat]
+        for path in sorted(plat_dir.iterdir()):
+            m = pattern.match(path.name)
+            if not m:
+                continue
+            try:
+                collect_date = (_parse_jd_date if plat == "jd" else _parse_dy_date)(m.group(1))
+            except ValueError:
+                continue
+            if date_from and collect_date < date_from:
+                continue
+            if date_to and collect_date > date_to:
+                continue
+            out.append((plat, collect_date, path))
+    return out
+
+
+def _load_and_normalize(plat: str, collect_date: date, path: Path,
+                        style_map: dict[str, str] | None = None) -> list[dict]:
+    with path.open(encoding="utf-8") as f:
+        payload = json.load(f)
+    rel = artifact_relative_path(path)
+    return NORMALIZERS[plat](payload, rel, collect_date, style_map=style_map)
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="扫描 JD/DY 评价 JSON 入库 product_reviews")
+    parser.add_argument("--date", help="只导入指定采集日期 (YYYY-MM-DD); JD 用 - 分隔, DY 无分隔也可")
+    parser.add_argument("--from", dest="date_from", help="起始采集日期 (YYYY-MM-DD)")
+    parser.add_argument("--to", dest="date_to", help="结束采集日期 (YYYY-MM-DD)")
+    parser.add_argument("--platforms", default="dy", help="平台, 逗号分隔 (默认 dy; JD 待 SPU 映射补齐)")
+    parser.add_argument("--dry-run", action="store_true", help="只解析不入库")
+    args = parser.parse_args()
+
+    platforms = [p.strip() for p in args.platforms.split(",") if p.strip()]
+    for p in platforms:
+        if p not in VALID_PLATFORMS:
+            print(f"[ERR] 未知平台: {p} (允许: {VALID_PLATFORMS})")
+            return 2
+
+    date_from = _parse_single(args.date_from) if args.date_from else None
+    date_to = _parse_single(args.date_to) if args.date_to else None
+    if args.date:
+        d = _parse_single(args.date)
+        date_from = d
+        date_to = d
+
+    files = _discover_files(platforms, date_from, date_to)
+    if not files:
+        print(f"[WARN] 未找到匹配的 JSON 文件 (platforms={platforms}, from={date_from}, to={date_to})")
+        return 0
+
+    # 预加载 DY 款式映射 (dim_style.dy_spus -> {product_id: style_name})
+    style_map: dict[str, str] = {}
+    if "dy" in platforms and not args.dry_run:
+        style_map = _load_dy_style_map()
+        print(f"[INFO] dim_style.dy_spus 映射: {len(style_map)} 个 product_id -> style_name")
+    elif "dy" in platforms and args.dry_run:
+        # dry-run 也加载, 便于预览匹配情况
+        style_map = _load_dy_style_map()
+        print(f"[INFO] (dry-run) dim_style.dy_spus 映射: {len(style_map)} 个 product_id -> style_name")
+
+    total = 0
+    skipped = 0
+    by_plat: dict[str, int] = {p: 0 for p in platforms}
+    by_rating: dict[str, int] = {"好评": 0, "中评": 0, "差评": 0}
+    matched_styles: set[str] = set()
+
+    for plat, collect_date, path in files:
+        try:
+            records = _load_and_normalize(plat, collect_date, path, style_map=style_map)
+        except Exception as e:
+            print(f"[ERR] 解析 {path}: {e}")
+            continue
+        # 去重: 同 order_id 多条 (理论上 UNIQUE 会处理, 这里先在内存去重避免执行 batch fail)
+        seen = set()
+        deduped = []
+        for r in records:
+            if not r["order_id"]:
+                skipped += 1
+                continue
+            if r["order_id"] in seen:
+                skipped += 1
+                continue
+            seen.add(r["order_id"])
+            deduped.append(r)
+            by_rating[r["rating"]] = by_rating.get(r["rating"], 0) + 1
+            if r.get("style_name"):
+                matched_styles.add(r["style_name"])
+
+        plat_style_matched = sum(1 for r in deduped if r.get("style_name"))
+        print(f"[{plat}] {collect_date}  {path.name}  -> {len(deduped)} 条 (款式匹配 {plat_style_matched}/{len(deduped)})")
+
+        if args.dry_run:
+            if deduped:
+                sample = deduped[0].copy()
+                sample["raw_data"] = "(省略)"
+                print(f"  (dry-run, 不入库) 样本: {sample}")
+        else:
+            with get_conn() as conn:
+                n = upsert_product_reviews(conn, deduped)
+            print(f"  -> upsert {n} rows")
+
+        total += len(deduped)
+        by_plat[plat] = by_plat.get(plat, 0) + len(deduped)
+
+    print()
+    print("=" * 60)
+    print(f"  总计: {total} 条 (跳过 {skipped} 条无 order_id)")
+    for p, n in by_plat.items():
+        print(f"  {p}: {n} 条")
+    print("  按评级:")
+    for r, n in by_rating.items():
+        print(f"    {r}: {n} 条")
+    if matched_styles:
+        print(f"  匹配到款式 ({len(matched_styles)} 款): {', '.join(sorted(matched_styles))}")
+    print("=" * 60)
+    return 0
+
+
+def _parse_single(s: str) -> date:
+    s = s.strip()
+    for fmt in ("%Y-%m-%d", "%Y%m%d"):
+        try:
+            return datetime.strptime(s, fmt).date()
+        except ValueError:
+            continue
+    raise SystemExit(f"无法解析日期: {s}")
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/insert_bitable_records.py b/src/gyxx_flow/modules/product_commerce/runtime/insert_bitable_records.py
new file mode 100644
index 0000000..7384fea
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/insert_bitable_records.py
@@ -0,0 +1,309 @@
+import argparse
+import json
+import shutil
+import subprocess
+import sys
+from datetime import datetime
+from pathlib import Path
+
+from runtime_paths import EXPORTS_DATA_ROOT, LOG_ROOT, RAW_DATA_ROOT
+
+from config.style_config_loader import get_bitable_style_map
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+DATA_ROOT = RAW_DATA_ROOT
+STYLE_MAP_PATH = PROJECT_ROOT / "bitable_style_map.json"
+
+
+def read_json(path):
+    return json.loads(path.read_text(encoding="utf-8-sig"))
+
+
+def rate_to_number(value):
+    text = str(value or "").strip().replace("%", "")
+    if not text:
+        return 0
+    try:
+        return round(float(text) / 100, 6)
+    except ValueError:
+        return value
+
+
+def normalize_platform(value):
+    mapping = {"jd": "京东", "dy": "抖音", "tm": "天猫", "京东": "京东", "抖音": "抖音", "天猫": "天猫"}
+    return mapping.get(str(value), value)
+
+
+def date_to_month_week(value):
+    day = datetime.strptime(str(value), "%Y-%m-%d")
+    week = min(((day.day - 1) // 7) + 1, 4)
+    return f"{day.month}月第{week}周"
+
+
+def enrich_record(record):
+    item = dict(record)
+    item["时间周期"] = item.get("时间周期") or date_to_month_week(item["数据日期"])
+    item["平台"] = normalize_platform(item.get("平台"))
+    item["SourceID"] = item.get("SourceID") or f"{item['数据日期']}_{item['平台代码']}_{item['款式']}"
+    return item
+
+
+def build_field_payload(record, field_map):
+    record = enrich_record(record)
+    payload = {}
+    for target_field, source_field in field_map.items():
+        if source_field not in record:
+            continue
+        value = record[source_field]
+        if target_field in {"转化率", "加购率", "退款率"}:
+            value = rate_to_number(value)
+        payload[target_field] = value
+    return payload
+
+
+def run_lark_upsert(base_token, table_id, fields, dry_run=False, identity="user", record_id=None):
+    lark_cli = shutil.which("lark-cli") or shutil.which("lark-cli.cmd") or "lark-cli.cmd"
+    command = [
+        lark_cli,
+        "base",
+        "+record-upsert",
+        "--as",
+        identity,
+        "--base-token",
+        base_token,
+        "--table-id",
+        table_id,
+        "--json",
+        json.dumps(fields, ensure_ascii=False),
+    ]
+    if record_id:
+        command[command.index("--json")] = "--json"
+        command.insert(command.index("--json") + 2, "--record-id")
+        command.insert(command.index("--json") + 3, record_id)
+    print("[LARK]", " ".join(command[:-1]), command[-1])
+    if dry_run:
+        return
+    proc = subprocess.run(command, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace")
+    if proc.returncode:
+        raise RuntimeError(f"lark insert failed with exit code {proc.returncode}")
+
+
+def _find_record_id_by_source_id(base_token, table_id, source_id, identity="user"):
+    """Use +record-search to find existing record id by SourceID value.
+    Returns record_id string or None if not found.
+    """
+    lark_cli = shutil.which("lark-cli") or shutil.which("lark-cli.cmd") or "lark-cli.cmd"
+    payload = {
+        "keyword": source_id,
+        "search_fields": ["SourceID"],
+        "limit": 5,
+    }
+    command = [
+        lark_cli,
+        "base",
+        "+record-search",
+        "--as",
+        identity,
+        "--base-token",
+        base_token,
+        "--table-id",
+        table_id,
+        "--format",
+        "json",
+        "--json",
+        json.dumps(payload, ensure_ascii=False),
+    ]
+    proc = subprocess.run(command, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace", capture_output=True)
+    if proc.returncode:
+        return None
+    try:
+        envelope = json.loads(proc.stdout)
+    except json.JSONDecodeError:
+        return None
+    data = (envelope.get("data") or {}).get("data") or []
+    if not data:
+        return None
+    # data is list of records; each record is a list of cells in field order
+    for record in data:
+        # Try to get record id from envelope.record_id_list (matching order)
+        pass
+    record_ids = (envelope.get("data") or {}).get("record_id_list") or []
+    if not record_ids:
+        return None
+    return record_ids[0]
+
+
+def _delete_record(base_token, table_id, record_id, identity="user"):
+    lark_cli = shutil.which("lark-cli") or shutil.which("lark-cli.cmd") or "lark-cli.cmd"
+    command = [
+        lark_cli,
+        "base",
+        "+record-delete",
+        "--as",
+        identity,
+        "--base-token",
+        base_token,
+        "--table-id",
+        table_id,
+        "--record-id",
+        record_id,
+        "--yes",
+    ]
+    proc = subprocess.run(command, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace", capture_output=True)
+    return proc.returncode == 0
+
+
+def load_records(date, record_file="", source="json"):
+    """加载待插入的 records 列表。
+
+    source="json"  : 读 DATA_ROOT/ready_for_bitable/<date>/bitable_records.json
+                     (日报 + 周报；周报路径为 <start>_<end>/bitable_records.json)
+    source="db"    : 查 PG weekly_style_metrics 表，按周报 JSON 格式拼 records
+                     (--date 传周结束日 YYYY-MM-DD；目前只支持周报)
+    """
+    if source == "db":
+        return load_weekly_records_from_db(date)
+    path = (
+        Path(record_file)
+        if record_file
+        else EXPORTS_DATA_ROOT / date / "bitable_records.json"
+    )
+    if not path.exists():
+        raise FileNotFoundError(f"missing exported bitable records: {path}")
+    payload = read_json(path)
+    if isinstance(payload, dict) and "records" in payload:
+        return payload["records"]
+    return payload
+
+
+# 周报从 weekly_style_metrics 查时，DB 字段 → 周报 JSON 字段的映射
+_WEEKLY_PLATFORM_DISPLAY = {"jd": "京东", "dy": "抖音", "tm": "天猫"}
+
+
+def load_weekly_records_from_db(week_end: str) -> list[dict]:
+    """查 weekly_style_metrics 周汇总，转成周报 bitable_records 形态。
+
+    返回的 records 字段名与 export_bitable_records 生成的周报 JSON 一致，
+    走同一个 build_field_payload() → field_map 映射逻辑。
+    """
+    from db import get_conn, fetch_weekly_records
+
+    with get_conn() as conn:
+        rows = fetch_weekly_records(conn, week_end)
+    if not rows:
+        raise FileNotFoundError(f"weekly_style_metrics 无 week_end={week_end} 的记录")
+
+    def to_date_str(value):
+        return value.isoformat() if hasattr(value, "isoformat") else str(value)
+
+    records: list[dict] = []
+    for r in rows:
+        records.append({
+            "数据日期": to_date_str(r["week_end"]),
+            "周开始": to_date_str(r["week_start"]),
+            "周结束": to_date_str(r["week_end"]),
+            "时间周期": r.get("time_period") or "",
+            "平台": _WEEKLY_PLATFORM_DISPLAY.get(r["platform"], r["platform"]),
+            "平台代码": r["platform"],
+            "款式": r["style_name"],
+            "ERP款式编码": ",".join(r.get("erp_style_codes") or []),
+            "销量": int(r.get("sales") or 0),
+            "访客数": int(r.get("visitors") or 0),
+            "加购人数": int(r.get("cart_users") or 0),
+            "退款订单数": int(r.get("refund_orders") or 0),
+            "转化率": r.get("conversion_rate") or "",
+            "加购率": r.get("cart_rate") or "",
+            "退款率": r.get("refund_rate") or "",
+            "本季天数": int(r.get("present_days") or 0),
+        })
+    return records
+
+
+def main():
+    parser = argparse.ArgumentParser(description="按款式配置把最终记录插入飞书多维表格")
+    parser.add_argument("--date", default=datetime.now().strftime("%Y-%m-%d"))
+    parser.add_argument("--record-file", default="", help="直接指定 bitable_records.json，可用于周汇总文件")
+    parser.add_argument("--source", default="json", choices=["json", "db"],
+                        help="数据源: json=读 bitable_records.json (日报+周报都支持); "
+                             "db=查 weekly_style_metrics 周汇总 (仅周报)")
+    parser.add_argument("--style", default="", help="只插入指定款式")
+    parser.add_argument("--dry-run", action="store_true")
+    parser.add_argument("--as", dest="identity", default="user", choices=["user", "bot"])
+    args = parser.parse_args()
+
+    try:
+        style_map = get_bitable_style_map()
+        if not style_map:
+            raise RuntimeError("飞书动态配置为空")
+    except Exception as exc:
+        print(f"[WARN] 动态读取销量表配置失败，使用本地文件回退: {exc}")
+        style_map = read_json(STYLE_MAP_PATH)
+    if args.source == "db":
+        print(f"[INFO] source=db: 查 weekly_style_metrics week_end={args.date}")
+    records = load_records(args.date, args.record_file, source=args.source)
+    inserted = 0
+    skipped = 0
+    failed = []
+
+    for record in records:
+        style = record.get("款式")
+        if args.style and style != args.style:
+            continue
+        config = style_map.get(style)
+        if not config or not config.get("enabled", True):
+            skipped += 1
+            continue
+
+        fields = build_field_payload(record, config.get("field_map", {}))
+        if not fields:
+            skipped += 1
+            continue
+
+        source_id = record.get("SourceID") or fields.get("SourceID")
+        try:
+            record_id = None
+            if source_id and not args.dry_run:
+                record_id = _find_record_id_by_source_id(
+                    config["base_token"], config["table_id"], source_id, identity=args.identity
+                )
+            run_lark_upsert(
+                config["base_token"],
+                config["table_id"],
+                fields,
+                dry_run=args.dry_run,
+                identity=args.identity,
+                record_id=record_id,
+            )
+            inserted += 1
+        except Exception as exc:
+            failed.append({
+                "source_id": source_id,
+                "style": style,
+                "platform": record.get("平台"),
+                "platform_code": record.get("平台代码"),
+                "base_token": config.get("base_token"),
+                "table_id": config.get("table_id"),
+                "error": str(exc),
+            })
+            print(f"[FAIL] {source_id}: {exc}")
+
+    if failed:
+        failures_path = LOG_ROOT / f"insert_bitable_failures_{args.date}.json"
+        failures_path.parent.mkdir(parents=True, exist_ok=True)
+        failures_path.write_text(json.dumps({
+            "date": args.date,
+            "generated_at": datetime.now().isoformat(),
+            "count": len(failed),
+            "failures": failed,
+        }, ensure_ascii=False, indent=2), encoding="utf-8")
+        print(f"[WARN] {len(failed)} records failed; see {failures_path}")
+
+    print(f"[OK] inserted/planned: {inserted}, skipped: {skipped}, failed: {len(failed)}")
+
+    if failed and inserted == 0:
+        sys.exit(1)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/jd_main_image_collector.py b/src/gyxx_flow/modules/product_commerce/runtime/jd_main_image_collector.py
new file mode 100644
index 0000000..9571d58
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/jd_main_image_collector.py
@@ -0,0 +1,595 @@
+"""京东商品明细页主图采集脚本。
+
+采集流程:
+1. 复用 vendors/jd-data-flow/playwright_profile 登录态 (同 jd_product_data_collector.py)
+2. 进入商智 -> 旧版 -> 商品明细页
+3. 点「指标配置」勾选「搜索曝光量」「搜索点击次数」两列
+4. 逐款在多 SPU 搜索框输入该款的 SPU 列表 -> 点查询按钮
+5. 点展开加号 .jmti-plus-circle-f 展开所有 SKU 子行
+6. 解析表格 HTML 拿每个 SKU 的主图 URL / 曝光量 / 点击量
+7. 按 SPU 聚合: 取曝光量最大的 SKU 主图作为该 SPU 的代表
+8. 下载主图 + 写 JSON 到 data/主图信息/<date>/<款式>/jd_main_images.json
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from runtime_paths import (
+    RAW_DATA_ROOT,
+    TMP_ROOT,
+    managed_data_path,
+    profile_path,
+    vendor_source_root,
+)
+from typing import Any
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+VENDOR_SOURCE_ROOT = vendor_source_root("jd")
+if str(VENDOR_SOURCE_ROOT) not in sys.path:
+    sys.path.append(str(VENDOR_SOURCE_ROOT))
+
+from playwright.sync_api import sync_playwright
+
+from jd_data_collector import DEFAULT_CONFIG, step_login, step_enter_shangzhi
+from jd_peer_product_data_collector import step_return_old_version
+
+# 复用 TM 主图脚本的图片下载 / URL 规范化工具
+sys.path.insert(0, str(PROJECT_ROOT))
+from taobao_wanxiang_ai_creative_report import (
+    canonical_image_key,
+    download_file,
+    image_extension,
+    image_filename,
+    parse_number,
+    safe_filename,
+)
+from config.style_config_loader import get_jd_spu_groups
+from main_image_paths import collect_images_dir
+
+
+JD_PROFILE_DIR = str(profile_path("jd"))
+OUTPUT_DIR = RAW_DATA_ROOT / "主图信息"
+DEBUG_DIR = TMP_ROOT / "debug" / "jd_main_image_snapshot"
+DEBUG = os.getenv("JD_MAIN_IMAGE_DEBUG", "0") == "1"
+
+_STATIC_STYLE_SPU_GROUPS = [
+    {"style": "盖亚斜挎", "spus": ["10022621048112", "10023155725944", "10030097050220"]},
+    {"style": "极星pro", "spus": ["10025628785945", "10026601031834", "10026600460605"]},
+    {"style": "盖亚微单", "spus": ["10025310117864", "10023909449487"]},
+    {"style": "盖世m1", "spus": ["10026326257225", "10029721108947"]},
+    {"style": "星迹2", "spus": ["10033913254016", "10025329514925", "10025329599430"]},
+    {"style": "星云2", "spus": ["10027762032411"]},
+    {"style": "宙斯", "spus": ["10033333485894", "10025292822415", "10028005218343", "10025020862662", "10032720707692", "10028005207681"]},
+    {"style": "阿波罗2", "spus": ["10028594796004", "10028595026975"]},
+    {"style": "阿波罗x1", "spus": ["10024429591100", "10024753441908"]},
+    {"style": "波塞冬edge", "spus": ["10032366483452"]},
+    {"style": "逐星", "spus": ["10026025497958", "10031326808929"]},
+    {"style": "逐星GT", "spus": ["10033470328729"]},
+    {"style": "星云mini", "spus": ["10032211469888"]},
+    {"style": "晨星2", "spus": ["10031295279033"]},
+    {"style": "极星双肩", "spus": ["10029986282219", "10029920288281", "10033941202259"]},
+    {"style": "凌云air", "spus": ["10029593796467"]},
+    {"style": "觅光", "spus": ["10033787463147", "10033860225085"]},
+    {"style": "云卷2", "spus": ["10032963233896", "10034088808495"]},
+    {"style": "凌云2", "spus": ["10028043067231"]},
+    {"style": "瑞白", "spus": ["10033366316587"]},
+    {"style": "黑曜石吐司包", "spus": ["10033995817475"]},
+    {"style": "黑曜石托特包", "spus": ["10033995834580"]},
+    {"style": "曜影", "spus": ["10034534903226"]},
+    {"style": "黑曜石电脑包", "spus": ["10034651877353"]},
+    {"style": "凌云3", "spus": ["10034906180847"]},
+    {"style": "拓界2", "spus": ["10034905606265"]},
+]
+
+
+def _get_style_spu_groups() -> list[dict]:
+    """优先从飞书读取京东SPU分组,失败则回退硬编码。"""
+    try:
+        groups = get_jd_spu_groups()
+        if groups:
+            return groups
+    except Exception as exc:
+        print(f"[WARN] 动态读取京东SPU配置失败,使用硬编码回退: {exc}")
+    return _STATIC_STYLE_SPU_GROUPS
+
+
+_DATE_LABEL = ""
+
+
+def now_label() -> str:
+    return _DATE_LABEL or datetime.now().strftime("%Y-%m-%d")
+
+
+def step_goto_product_detail(page):
+    """跳转到商品明细页面"""
+    print("[Step] 跳转到商品明细页面...")
+    page.goto(
+        "https://sz.jd.com/szweb/sz/view/productAnalysis/productDetail.html",
+        wait_until="networkidle",
+    )
+    page.wait_for_timeout(5000)
+
+
+def close_guide_if_present(page):
+    """检测并关闭新功能引导弹窗"""
+    try:
+        page.locator('.jmtp-guide-content, .jmtp-guide-con').first.wait_for(state="visible", timeout=3000)
+        page.locator('.jmtp-guide-footer button', has_text="完成").first.click()
+        page.wait_for_timeout(2000)
+    except Exception:
+        pass
+
+
+def enable_columns(page) -> None:
+    """点「指标配置」打开弹窗 -> 勾选「搜索曝光量」「搜索点击次数」-> 点「确定」。
+
+    这两个指标默认不在表格里,需要在「指标配置」modal 里勾选才会显示成列。
+
+    用 Playwright 原生 locator.check() 而不是 evaluate + label.click(),
+    因为 React 应用里手动 click label 有时不触发 onChange 事件,
+    导致视觉上 checked 但 React state 没更新,点确定后状态丢失。
+    """
+    # 点「指标配置」按钮打开 modal
+    clicked = page.evaluate(r"""() => {
+        const btns = Array.from(document.querySelectorAll('button'));
+        const target = btns.find((b) => (b.textContent || '').trim() === '指标配置' && b.offsetParent !== null);
+        if (target) { target.click(); return true; }
+        return false;
+    }""")
+    if not clicked:
+        print("  [WARN] 未找到「指标配置」按钮,跳过列启用")
+        return
+    # 等 modal 渲染完成
+    try:
+        page.wait_for_selector('.indicators-config-modal:not(.jmtd-modal-hidden)', timeout=5000)
+    except Exception:
+        pass
+    page.wait_for_timeout(1500)
+
+    # 找 modal 内 checkbox 的 input 元素并逐个用 Playwright 原生 check() 勾选
+    targets = ["搜索曝光量", "搜索点击次数"]
+    toggled = 0
+    for label_text in targets:
+        # 定位: 在 modal 内, .jmtd-checkbox-control 包含目标文案, 内部的 input[type=checkbox]
+        loc = page.locator(
+            f'.indicators-config-modal .jmtd-checkbox-control:has-text("{label_text}") input[type="checkbox"]'
+        ).first
+        try:
+            loc.wait_for(state="attached", timeout=5000)
+            # 用 check() 而不是 click(),Playwright 会等 input 真的变成 checked 状态
+            loc.check(force=True, timeout=5000)
+            toggled += 1
+            print(f"  [INFO] 勾选了 {label_text}")
+        except Exception as exc:
+            # check() 失败可能是 input 不可见或被遮挡,降级用 evaluate click label
+            print(f"  [WARN] check {label_text} 失败: {exc}, 降级用 label click")
+            page.evaluate(r"""([text]) => {
+                const labels = Array.from(document.querySelectorAll(
+                    '.indicators-config-modal .jmtd-checkbox-control, .jmtd-modal .jmtd-checkbox-control'
+                )).filter((el) => el.offsetParent !== null && (el.textContent || '').includes(text));
+                if (labels.length) labels[0].click();
+            }""", [label_text])
+            page.wait_for_timeout(500)
+            toggled += 1
+    page.wait_for_timeout(1000)
+
+    # 验证勾选状态 (用 label class 判断真实视觉状态)
+    verified = page.evaluate(r"""() => {
+        const targetLabels = ['搜索曝光量', '搜索点击次数'];
+        const result = {};
+        for (const target of targetLabels) {
+            const labels = Array.from(document.querySelectorAll(
+                '.indicators-config-modal .jmtd-checkbox-control, .jmtd-modal .jmtd-checkbox-control'
+            )).filter((el) => el.offsetParent !== null && (el.textContent || '').includes(target));
+            if (!labels.length) { result[target] = 'not_found'; continue; }
+            result[target] = labels[0].classList.contains('jmtd-form-control-checked') ? 'checked' : 'unchecked';
+        }
+        return result;
+    }""")
+    print(f"  [INFO] 勾选状态: {verified}")
+
+    # 点「确定」应用 (modal 内 footer 的确定按钮)
+    confirmed = page.evaluate(r"""() => {
+        const btns = Array.from(document.querySelectorAll('.jmtd-modal-footer button, .jmtd-modal button, button'));
+        const target = btns.find((b) => (b.textContent || '').trim() === '确定' && b.offsetParent !== null);
+        if (target) { target.click(); return true; }
+        return false;
+    }""")
+    if not confirmed:
+        print("  [WARN] 未找到「确定」按钮,尝试按 Enter")
+        page.keyboard.press("Enter")
+    page.wait_for_timeout(3000)
+
+
+def search_by_spus(page, spus: list[str]) -> None:
+    """在多 SPU 搜索框输入 SPU 列表并触发查询。
+
+    每输一个 SPU 按 Enter 形成一个 tag,全部输完后点查询按钮触发过滤。
+    """
+    print(f"  搜索 SPU: {spus}")
+    wrapper = page.locator('.jmtd-multiple-input').first
+    wrapper.wait_for(state="visible", timeout=10000)
+    wrapper.click()
+    page.wait_for_timeout(500)
+    # 清空旧搜索 tag
+    page.evaluate(r"""() => {
+        const closeBtns = Array.from(document.querySelectorAll('.jmtd-multiple-base-input-tag-wrapper .jmti-close'));
+        for (const btn of closeBtns) btn.click();
+    }""")
+    page.wait_for_timeout(500)
+
+    for spu in spus:
+        page.keyboard.type(spu, delay=80)
+        page.wait_for_timeout(400)
+        page.keyboard.press("Enter")
+        page.wait_for_timeout(400)
+
+    # 点查询按钮触发过滤 (jmti-search icon button)
+    search_clicked = page.evaluate(r"""() => {
+        const btn = document.querySelector('.jmtd-input-search-button');
+        if (btn) { btn.click(); return true; }
+        return false;
+    }""")
+    if not search_clicked:
+        page.keyboard.press("Enter")
+    page.wait_for_timeout(5000)
+
+
+def expand_sku_rows(page) -> int:
+    """点所有 .jmti-plus-circle-f 展开图标,展开子 SKU 行。返回展开数。"""
+    count = page.evaluate(r"""() => {
+        const visible = (el) => {
+            if (!el) return false;
+            const style = window.getComputedStyle(el);
+            const rect = el.getBoundingClientRect();
+            return style.display !== 'none' && style.visibility !== 'hidden'
+                && rect.width > 0 && rect.height > 0;
+        };
+        const icons = Array.from(document.querySelectorAll(
+            '.jmti-plus-circle-f[data-event-name="expand"]'
+        )).filter(visible);
+        for (const icon of icons) icon.click();
+        return icons.length;
+    }""")
+    print(f"  展开 {count} 个 SPU 父行")
+    page.wait_for_timeout(2000)
+    return int(count or 0)
+
+
+def upgrade_jd_image_url(url: str) -> str:
+    """把 JD 360buyimg 的 URL 升级到大尺寸 (n12)。
+
+    表格里 img src 是小尺寸 (n5), 下载下来很糊 (6KB 缩略图)。
+    替换成 n12 拿清晰大图 (~1MB)。s800x800/s1000x1000 这张图不支持 (400),
+    所以用 n12 兜底。
+    """
+    return re.sub(r"//img\d+\.360buyimg\.com/n\d+/", "//img10.360buyimg.com/n12/", url or "")
+
+
+def extract_sku_rows(page) -> list[dict[str, Any]]:
+    """从展开后的表格 HTML 提取每个 SKU 子行的数据。
+
+    返回 [{spu, sku_id, image_url, image_key, impressions, clicks}]。
+
+    策略:
+    - 遍历 tbody 的 tr.jmtd-table-row
+    - 跳过合计行 (.jmtd-table-plugin-total-row)
+    - 跟踪 current_spu: 遇到带展开图标的行就是 SPU 父行, 更新 current_spu
+    - 无展开图标的行是 SKU 子行, 归到 current_spu
+    - SKU ID 从 a[href*="//item.jd.com/<SKU>.html"] 提取
+    - 指标值从对应列的 .first-value 提取
+
+    列顺序 (实测):
+    商品信息(0) / 访客数(1) / 浏览量(2) / 成交客户数(3) / 成交商品件数(4) /
+    成交金额(5) / 搜索曝光量(6) / 搜索点击次数(7) / 操作(8)
+
+    但列顺序会随用户在「指标配置」里的勾选变化,所以用表头列名动态查找列索引,
+    而不是硬编码 6/7。
+    """
+    rows = page.evaluate(r"""() => {
+        const visible = (el) => {
+            if (!el) return false;
+            const style = window.getComputedStyle(el);
+            const rect = el.getBoundingClientRect();
+            return style.display !== 'none' && style.visibility !== 'hidden'
+                && rect.width > 0 && rect.height > 0;
+        };
+        const normalizeUrl = (url) => {
+            const v = String(url || '').trim();
+            if (!v || v.startsWith('data:')) return '';
+            if (v.startsWith('//')) return 'https:' + v;
+            return v;
+        };
+        const imgUrl = (img) => {
+            if (!img) return '';
+            const attrs = ['currentSrc', 'src', 'data-src', 'data-lazy-src', 'data-original'];
+            for (const attr of attrs) {
+                const v = attr === 'currentSrc' ? img.currentSrc : img.getAttribute(attr);
+                const n = normalizeUrl(v);
+                if (n) return n;
+            }
+            return '';
+        };
+
+        // 找表头列名 -> 列索引 (跳过「操作」列右侧的固定列, 用 thead 的 th)
+        const headerCells = Array.from(document.querySelectorAll('thead th.jmtd-table-cell'))
+            .filter(visible);
+        const colNames = headerCells.map((th) => (th.textContent || '').replace(/\s+/g, '').trim());
+        const idxExposure = colNames.findIndex((n) => n.includes('搜索曝光') || n.includes('曝光量'));
+        const idxClick = colNames.findIndex((n) => n.includes('搜索点击') || n.includes('点击次数') || n.includes('点击量'));
+
+        // 找所有数据行 (跳过合计行)
+        const dataRows = Array.from(document.querySelectorAll('tbody tr.jmtd-table-row'))
+            .filter(visible)
+            .filter((r) => !r.classList.contains('jmtd-table-plugin-total-row'));
+
+        let currentSpu = '';
+        const result = [];
+        for (const row of dataRows) {
+            // 检测是否是 SPU 父行 (有展开/收起图标)
+            const expandIcon = row.querySelector('.jmti-plus-circle-f, .jmti-minus-circle-f');
+            const isParent = !!expandIcon;
+
+            const cells = Array.from(row.querySelectorAll('td.jmtd-table-cell')).filter(visible);
+
+            // 提取 SPU ID (从 .jmtd-table-formatter-name-desc 的 "SPU：<id>" 文本)
+            const spuDesc = row.querySelector('.jmtd-table-formatter-name-desc');
+            if (spuDesc) {
+                const m = (spuDesc.textContent || '').match(/SPU[：:]\s*(\d+)/);
+                if (m) currentSpu = m[1];
+            }
+
+            // 提取 SKU ID (从 a[href*="//item.jd.com/<SKU>.html"])
+            const skuLink = row.querySelector('a[href*="item.jd.com"]');
+            let skuId = '';
+            if (skuLink) {
+                const m = (skuLink.getAttribute('href') || '').match(/item\.jd\.com\/(\d+)\.html/);
+                if (m) skuId = m[1];
+            }
+
+            // 提取主图 URL (img.jmtd-table-formatter-name-img)
+            const img = row.querySelector('img.jmtd-table-formatter-name-img');
+            const imageUrl = imgUrl(img);
+
+            // 提取指标值
+            const cellValue = (idx) => {
+                if (idx < 0 || idx >= cells.length) return '';
+                const fv = cells[idx].querySelector('.first-value');
+                return fv ? (fv.textContent || '').trim() : '';
+            };
+
+            // SPU 父行也有一行数据 (代表该 SPU 汇总), 但需求是「取最大 SKU 主图」,
+            // 所以只取 SKU 子行 (isParent=false) 的数据, SPU 父行只用来更新 currentSpu
+            if (isParent) continue;
+
+            if (!imageUrl) continue;  // 没图的行跳过 (空行)
+
+            result.push({
+                spu: currentSpu,
+                sku_id: skuId,
+                image_url: imageUrl,
+                impressions_text: cellValue(idxExposure),
+                clicks_text: cellValue(idxClick),
+            });
+        }
+        return result;
+    }""")
+    if not isinstance(rows, list):
+        return []
+    parsed: list[dict[str, Any]] = []
+    for row in rows:
+        image_url = row.get("image_url") or ""
+        if not image_url:
+            continue
+        image_url = upgrade_jd_image_url(image_url)
+        parsed.append({
+            "spu": str(row.get("spu") or ""),
+            "sku_id": str(row.get("sku_id") or ""),
+            "image_url": image_url,
+            "image_key": canonical_image_key(image_url),
+            "impressions": parse_number(row.get("impressions_text")),
+            "clicks": parse_number(row.get("clicks_text")),
+        })
+    return parsed
+
+
+def aggregate_by_spu(sku_rows: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    """按 SPU 聚合: 每个 SPU 取曝光量最大的 SKU 的主图作为代表。"""
+    by_spu: dict[str, list[dict[str, Any]]] = {}
+    for row in sku_rows:
+        spu = row.get("spu") or "unknown"
+        by_spu.setdefault(spu, []).append(row)
+
+    result: list[dict[str, Any]] = []
+    for spu, rows in by_spu.items():
+        best = max(rows, key=lambda r: float(r.get("impressions") or 0))
+        result.append({
+            "spu": spu,
+            "sku_id": best["sku_id"],
+            "image_url": best["image_url"],
+            "image_key": best["image_key"],
+            "impressions": best["impressions"],
+            "clicks": best["clicks"],
+            "sku_count": len(rows),
+        })
+    return sorted(result, key=lambda r: float(r["impressions"] or 0), reverse=True)
+
+
+def download_main_images(records: list[dict[str, Any]], images_dir: Path) -> None:
+    """下载每个 record 的主图到 images_dir,回填 local_image_path。"""
+    images_dir.mkdir(parents=True, exist_ok=True)
+    for rec in records:
+        url = rec.get("image_url") or ""
+        if not url:
+            continue
+        key = safe_filename(str(rec.get("image_key") or "unknown"))
+        path = images_dir / image_filename(key, url)
+        ok = download_file(url, path)
+        rec["local_image_path"] = str(path) if ok else ""
+        rec["image_downloaded"] = ok
+
+
+def write_style_outputs(result: dict[str, Any], output_dir: Path) -> tuple[Path, Path]:
+    """写单款 JSON + MD 到 data/主图信息/<date>/<款式>/"""
+    style_name = str(result["style_name"])
+    style_dir = output_dir / now_label() / safe_filename(style_name)
+    style_dir.mkdir(parents=True, exist_ok=True)
+
+    json_path = style_dir / "jd_main_images.json"
+    json_path.write_text(json.dumps(result, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    md_path = style_dir / "jd_main_images.md"
+    lines = [
+        f"# {style_name} 京东主图汇总",
+        "",
+        f"- 采集时间: {result.get('collected_at', '')}",
+        f"- SPU 数: {result.get('spu_count', 0)}",
+        f"- 原始 SKU 数: {result.get('raw_sku_count', 0)}",
+        "",
+        "| SPU | SKU ID | 主图 URL | 本地图片 | 搜索曝光量 | 搜索点击次数 | SKU 数 |",
+        "| --- | --- | --- | --- | ---: | ---: | ---: |",
+    ]
+    for rec in result.get("records", []):
+        lines.append(
+            f"| {rec.get('spu', '')} | {rec.get('sku_id', '')} | {rec.get('image_url', '')} "
+            f"| {rec.get('local_image_path', '')} | {rec.get('impressions', 0)} "
+            f"| {rec.get('clicks', 0)} | {rec.get('sku_count', 0)} |"
+        )
+    md_path.write_text("\n".join(lines), encoding="utf-8")
+    return json_path, md_path
+
+
+def collect_style(page, style_name: str, spus: list[str], output_dir: Path) -> dict[str, Any]:
+    """采集单款: 搜索 -> 展开 -> 解析 -> 聚合 -> 下载 -> 写 JSON。"""
+    if not spus:
+        raise RuntimeError(f"款 {style_name} 无 SPU 配置")
+
+    search_by_spus(page, spus)
+    expand_sku_rows(page)
+    sku_rows = extract_sku_rows(page)
+    print(f"  解析到 {len(sku_rows)} 个 SKU 行")
+    if not sku_rows and DEBUG:
+        snapshot_path = DEBUG_DIR / f"{safe_filename(style_name)}_no_sku_{int(time.time())}.html"
+        snapshot_path.parent.mkdir(parents=True, exist_ok=True)
+        snapshot_path.write_text(page.content(), encoding="utf-8")
+        print(f"  [DEBUG] snapshot: {snapshot_path}")
+
+    records = aggregate_by_spu(sku_rows)
+    print(f"  聚合到 {len(records)} 个 SPU 记录")
+
+    style_dir = output_dir / now_label() / safe_filename(style_name)
+    images_dir = collect_images_dir(style_dir, "jd")
+    download_main_images(records, images_dir)
+
+    result = {
+        "style_name": style_name,
+        "platform": "jd",
+        "collected_at": datetime.now().isoformat(timespec="seconds"),
+        "spu_count": len(records),
+        "raw_sku_count": len(sku_rows),
+        "records": records,
+    }
+    json_path, md_path = write_style_outputs(result, output_dir)
+    print(f"  [OK] JSON: {json_path}")
+    return result
+
+
+def parse_styles(value: str | None) -> list[dict]:
+    """解析 --styles CLI 参数 (逗号分隔款式名),返回 [{style, spus}] 子集。空则全部。"""
+    groups = _get_style_spu_groups()
+    if not value:
+        return groups
+    wanted = {s.strip() for s in re.split(r"[,，\n]+", value) if s.strip()}
+    return [g for g in groups if g["style"] in wanted]
+
+
+def run(args: argparse.Namespace) -> None:
+    global _DATE_LABEL
+    output_dir = managed_data_path(args.output)
+    output_dir.mkdir(parents=True, exist_ok=True)
+    styles = parse_styles(args.styles)
+    _DATE_LABEL = args.date or ""
+
+    with sync_playwright() as p:
+        context = p.chromium.launch_persistent_context(
+            user_data_dir=JD_PROFILE_DIR,
+            headless=args.headless,
+            viewport={"width": 1920, "height": 1080},
+            locale="zh-CN",
+            args=["--disable-gpu"],
+        )
+        page = context.new_page()
+
+        # Step 1: 登录
+        page.goto("https://shop.jd.com/jdm/home", wait_until="domcontentloaded", timeout=60000)
+        page.wait_for_timeout(2000)
+        if "shop.jd.com" in page.url and "login" not in page.url.lower():
+            print("[OK] 检测到京东已登录态")
+        else:
+            step_login(page, args.shop, args.password)
+
+        # Step 2-3: 商智 -> 旧版 -> 商品明细页
+        sz_page = step_enter_shangzhi(context, page)
+        step_return_old_version(sz_page)
+        step_goto_product_detail(sz_page)
+        close_guide_if_present(sz_page)
+
+        # Step 4: 启用「搜索曝光量」「搜索点击次数」两列
+        enable_columns(sz_page)
+
+        # Step 5-7: 逐款采集
+        results: list[dict] = []
+        for index, group in enumerate(styles, 1):
+            style_name = group["style"]
+            spus = group.get("spus") or []
+            print(f"\n[{index}/{len(styles)}] 采集: {style_name} (SPUs: {len(spus)})")
+            try:
+                result = collect_style(sz_page, style_name, spus, output_dir)
+                results.append(result)
+            except Exception as exc:
+                print(f"  [FAIL] {style_name}: {exc}")
+                results.append({
+                    "style_name": style_name,
+                    "platform": "jd",
+                    "collected_at": datetime.now().isoformat(timespec="seconds"),
+                    "error": str(exc),
+                    "records": [],
+                })
+
+        overall_path = output_dir / now_label() / f"jd_main_images_all_{now_label()}.json"
+        overall_path.parent.mkdir(parents=True, exist_ok=True)
+        overall_path.write_text(json.dumps({
+            "collected_at": datetime.now().isoformat(timespec="seconds"),
+            "style_count": len(results),
+            "results": results,
+        }, ensure_ascii=False, indent=2), encoding="utf-8")
+        print(f"\n[OK] 全部款式汇总: {overall_path}")
+
+        if args.keep_open:
+            print("[INFO] --keep-open 已开启,浏览器保留 60 秒")
+            time.sleep(60)
+        context.close()
+
+
+def build_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(description="采集京东商品明细页主图数据")
+    parser.add_argument("--shop", default=DEFAULT_CONFIG["shop"], help="京东店铺名")
+    parser.add_argument("--password", default=DEFAULT_CONFIG["password"], help="京东密码")
+    parser.add_argument("--output", default=str(OUTPUT_DIR), help="输出目录")
+    parser.add_argument("--date", default=None,
+                        help="采集日期 YYYY-MM-DD,决定输出子目录名(默认 = today)")
+    parser.add_argument("--styles", default="", help="只采集指定款式,逗号分隔;默认全部")
+    parser.add_argument("--headless", action="store_true", help="无头模式")
+    parser.add_argument("--keep-open", action="store_true", help="结束后保留浏览器 60 秒")
+    return parser
+
+
+if __name__ == "__main__":
+    run(build_parser().parse_args())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/jd_product_data_collector.py b/src/gyxx_flow/modules/product_commerce/runtime/jd_product_data_collector.py
new file mode 100644
index 0000000..3f28955
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/jd_product_data_collector.py
@@ -0,0 +1,4482 @@
+"""
+京东单品（商品明细）数据采集脚本
+1. 在商品明细页面翻页采集所有 SPU
+2. 逐个 SPU 搜索 → 点击详情 → 新页面
+3. 在新页面切到"核心数据"tab，采集指标（含比较数据）
+4. 结果保存到 单品详情数据_日期/ 目录
+
+依赖:
+    pip install playwright
+    playwright install chromium
+
+用法:
+    python jd_product_data_collector.py
+    python jd_product_data_collector.py --shop "店铺名" --password "密码"
+"""
+
+import argparse
+import csv
+import json
+import math
+import os
+import re
+import sys
+from datetime import datetime, timedelta
+
+from runtime_paths import (
+    RAW_DATA_ROOT,
+    TMP_ROOT,
+    managed_data_path,
+    profile_path,
+    vendor_source_root,
+)
+
+PROJECT_ROOT = os.path.dirname(os.path.abspath(__file__))
+SOURCE_ROOT = vendor_source_root("jd")
+if str(SOURCE_ROOT) not in sys.path:
+    # 用 append 而不是 insert(0), 让项目根目录的 config/ 包优先于 vendor 的 config.py
+    sys.path.append(str(SOURCE_ROOT))
+
+from playwright.sync_api import sync_playwright, BrowserContext
+
+from collection_progress import SpuProgress
+
+from jd_data_collector import (
+    DEFAULT_CONFIG,
+    step_login,
+    step_enter_shangzhi,
+    extract,
+)
+from jd_peer_product_data_collector import step_return_old_version
+from erp_metric_overrides import apply_erp_override
+from collect_retry_utils import FailedCollector, retry_step
+from config.style_config_loader import StyleConfigLoader, get_daily_styles, get_jd_spu_groups
+
+
+# ── 常量 ──────────────────────────────────────────────────────────────────────
+
+OUTPUT_BASE = str(RAW_DATA_ROOT / "jd单品数据")
+JD_PROFILE_DIR = str(profile_path("jd"))
+TARGET_DATE = None
+DEBUG_DATE = False  # --debug-date 时为 True,把日历面板状态写到 debug/jd_date_picker/
+
+# 阿波罗2 两个 SPU，合并统计
+APOLLO2_SPU_LIST = ["10028594796004", "10028595026975"]
+APOLLO2_FOLDER = "阿波罗2"
+
+# 宙斯3 两个 SPU，合并统计
+ZEUS3_SPU_LIST = ["10033333485894", "10025345724491"]
+ZEUS3_FOLDER = "宙斯3"
+
+# 星迹 三个 SPU，合并统计
+XINGJI_SPU_LIST = ["10033913254016", "10025329514925", "10025329599430"]
+XINGJI_FOLDER = "星迹"
+
+# 保留硬编码作为灾备回退；运行时优先从飞书多维表格读取。
+_STATIC_STYLE_SPU_GROUPS = [
+    {"style": "盖亚斜挎", "spus": ["10022621048112", "10023155725944", "10030097050220"]},
+    {"style": "极星pro", "spus": ["10025628785945", "10026601031834", "10026600460605"]},
+    {"style": "盖亚微单", "spus": ["10025310117864", "10023909449487"]},
+    {"style": "盖世m1", "spus": ["10026326257225", "10029721108947"]},
+    {"style": "星迹2", "spus": ["10033913254016", "10025329514925", "10025329599430"]},
+    {"style": "星云2", "spus": ["10027762032411"]},
+    {"style": "宙斯", "spus": ["10033333485894", "10025292822415", "10028005218343", "10025020862662", "10032720707692", "10028005207681"]},
+    {"style": "阿波罗2", "spus": ["10028594796004", "10028595026975"]},
+    {"style": "阿波罗x1", "spus": ["10024429591100", "10024753441908"]},
+    {"style": "波塞冬edge", "spus": ["10032366483452"]},
+    {"style": "逐星", "spus": ["10026025497958", "10031326808929"]},
+    {"style": "逐星GT", "spus": ["10033470328729"]},
+    {"style": "星云mini", "spus": ["10032211469888"]},
+    {"style": "晨星2", "spus": ["10031295279033"]},
+    {"style": "极星双肩", "spus": ["10029986282219", "10029920288281", "10033941202259"]},
+    {"style": "凌云air", "spus": ["10029593796467"]},
+    {"style": "觅光", "spus": ["10033787463147", "10033860225085"]},
+    {"style": "云卷2", "spus": ["10032963233896", "10034088808495"]},
+    {"style": "凌云2", "spus": ["10028043067231"]},
+    {"style": "瑞白", "spus": ["10033366316587"]},
+    {"style": "黑曜石吐司包", "spus": ["10033995817475"]},
+    {"style": "黑曜石托特包", "spus": ["10033995834580"]},
+    {"style": "曜影", "spus": ["10034534903226"]},
+    {"style": "黑曜石电脑包", "spus": ["10034651877353"]},
+    {"style": "凌云3", "spus": ["10034906180847"]},
+    {"style": "拓界2", "spus": ["10034905606265"]},
+]
+
+
+def _get_style_spu_groups():
+    """优先从飞书读取京东SPU分组，失败则回退硬编码。"""
+    try:
+        groups = get_jd_spu_groups()
+        if groups:
+            return groups
+    except Exception as exc:
+        print(f"[WARN] 动态读取京东SPU配置失败，使用硬编码回退: {exc}")
+    return _STATIC_STYLE_SPU_GROUPS
+
+
+STYLE_SPU_GROUPS = _get_style_spu_groups()
+
+
+def _print_preflight() -> None:
+    """启动期打印:飞书表格里京东分组今日待采 / 跳过清单。"""
+    try:
+        daily = get_daily_styles()
+    except Exception as exc:
+        print(f"[WARN] 飞书读取失败, 跳过 preflight: {exc}")
+        return
+    collectable = daily["jd"]["collectable"]
+    skipped = daily["jd"]["skipped"]
+    print(f"[JD] 飞书表格 待采 {len(collectable)} 款, 跳过 {len(skipped)} 款 (款式级缺 erp/sales,或 JD 缺商品链接ID)")
+    for s in skipped:
+        print(f"  - {s['style']}: {s['reason']}")
+
+
+_print_preflight()
+
+
+# ── 页面导航 ──────────────────────────────────────────────────────────────────
+
+def step_goto_product_detail(page):
+    """跳转到商品明细页面"""
+    print("[Step 4] 跳转到商品明细页面...")
+    page.goto(
+        "https://sz.jd.com/szweb/sz/view/productAnalysis/productDetail.html",
+        wait_until="networkidle"
+    )
+    page.wait_for_timeout(5000)
+    print("  [OK] 已到达商品明细页面")
+
+
+def close_guide_if_present(page):
+    """检测并关闭新功能引导弹窗"""
+    print("  检测引导弹窗...")
+    try:
+        page.locator('.jmtp-guide-content, .jmtp-guide-con').first.wait_for(state="visible", timeout=3000)
+        page.locator('.jmtp-guide-footer button', has_text="完成").first.click()
+        print("    [OK] 引导弹窗已关闭")
+        page.wait_for_timeout(2000)
+    except Exception:
+        print("    [INFO] 无引导弹窗")
+
+
+# ── 日期选择 ───────────────────────────────────────────────────────────────────
+
+def parse_target_date(value):
+    try:
+        return datetime.strptime(value, "%Y-%m-%d")
+    except ValueError as exc:
+        raise argparse.ArgumentTypeError("target date must be YYYY-MM-DD") from exc
+
+
+def get_yesterday_date():
+    return TARGET_DATE or (datetime.now() - timedelta(days=1))
+
+
+def get_yesterday_range_text():
+    day = get_yesterday_date().strftime("%Y-%m-%d")
+    return day, day
+
+
+def _date_text_matches_day(text, day):
+    if not text:
+        return False
+    normalized = re.sub(r"\s+", "", text)
+    return (
+        day in text
+        and (
+            f"{day}~{day}" in normalized
+            or f"{day}-{day}" in normalized
+            or f"{day}至{day}" in normalized
+            or normalized.count(day) >= 1
+        )
+    )
+
+
+def _combo_date_label(page):
+    try:
+        return page.locator('.jmtd-date-picker-combo-tag-echo .jmtd-label').first.inner_text(timeout=3000).strip()
+    except Exception:
+        return ""
+
+
+def _click_combo_mode(page, mode):
+    try:
+        tag = page.locator(f'.jmtd-date-picker-combo-tag[data-event-content="{mode}"]').first
+        tag.wait_for(state="visible", timeout=5000)
+        tag.click()
+        return True
+    except Exception:
+        return page.evaluate(
+            """(modeText) => {
+                for (const tag of document.querySelectorAll('.jmtd-date-picker-combo-tag')) {
+                    if ((tag.textContent || '').trim() === modeText) {
+                        tag.click();
+                        return true;
+                    }
+                }
+                return false;
+            }""",
+            "天" if mode == "day" else mode,
+        )
+
+
+def _click_combo_day_flip(page, direction):
+    event_name = "dayPrevClick" if direction == "prev" else "dayNextClick"
+    event_content = "dayPrev" if direction == "prev" else "dayNext"
+    click_point = page.evaluate(
+        r"""({eventName, eventContent}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const buttons = Array.from(document.querySelectorAll(
+                `button[data-event-name="${eventName}"], button[data-event-content="${eventContent}"]`
+            )).filter((btn) => visible(btn) && !btn.disabled && btn.getAttribute('aria-disabled') !== 'true');
+            const mainButtons = buttons.filter((btn) => !btn.closest('.jmtd-popover-portal, .jmtd-tooltip-portal'));
+            const btn = mainButtons[0] || buttons[0];
+            if (!btn) return null;
+            const rect = btn.getBoundingClientRect();
+            return { x: rect.left + rect.width / 2, y: rect.top + rect.height / 2 };
+        }""",
+        {"eventName": event_name, "eventContent": event_content},
+    )
+    if not click_point:
+        return False
+    page.mouse.click(click_point["x"], click_point["y"])
+    return True
+
+
+def _select_yesterday_by_combo_flip(page, target_day):
+    # 最多翻 120 次,覆盖 4 个月的历史补采。5/15 这种回采 28 次就够,
+    # 多给点余量给回采更早日期 / 多次失败重试场景。翻 1 次 = 1 天,每次 sleep 1.5s。
+    for _ in range(120):
+        label = _combo_date_label(page)
+        if _date_text_matches_day(label, target_day):
+            print(f"    [OK] 日期已是昨天: {label}")
+            return True
+
+        dates = re.findall(r"\d{4}-\d{2}-\d{2}", label)
+        if len(dates) >= 1 and dates[0] < target_day:
+            direction = "next"
+        else:
+            direction = "prev"
+
+        if not _click_combo_day_flip(page, direction):
+            return False
+        page.wait_for_timeout(1500)
+
+    return _date_text_matches_day(_combo_date_label(page), target_day)
+
+
+def _open_combo_date_panel(page):
+    for action in ("hover", "click"):
+        try:
+            echo = page.locator('.jmtd-date-picker-combo-tag-echo').first
+            if action == "hover":
+                echo.hover(timeout=5000)
+            else:
+                echo.click(timeout=5000)
+            page.wait_for_timeout(1200)
+            if page.locator('.jmtd-date-picker-panel, .jmtd-dropdown-list, .jmtd-popover-open').first.count() > 0:
+                return True
+        except Exception:
+            pass
+    return False
+
+
+def _jd_find_panel(page):
+    """
+    启发式找真实的日期浮层 panel。
+
+    难点:JD 商智这个面板的 class 是空(可能 CSS-in-JS / inline style),
+    而且它的祖先是 `<div id="main">` (1912x1349),如果用"含月份文案 + 200x200+
+    任何祖先"启发式,会一路爬到 #main,把整个页面当 panel,翻月按钮自然找不到。
+
+    策略:含月份文案的元素,向上爬到第一个"看起来像浮层"的祖先 —— 宽 150-700,
+    高 200-700,position 是 absolute/fixed/relative,或 class 包含 popover/panel/picker。
+    找不到则降级为"含 ≥28 个数字 cell 的最浅中等大小祖先"。
+    返回 {found, panel_rect, current_year, current_month, ancestors, all_panels}
+    """
+    return page.evaluate(
+        r"""
+        () => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none' && style.visibility !== 'hidden'
+                    && rect.width > 0 && rect.height > 0 && el.offsetParent !== null;
+            };
+            const monthRe = /20\d{2}[-/.年]\s*\d{1,2}/;
+            const isLikelyPanel = (el) => {
+                if (!el || el === document.body || el === document.documentElement) return false;
+                const r = el.getBoundingClientRect();
+                if (r.width < 150 || r.width > 700) return false;
+                if (r.height < 200 || r.height > 700) return false;
+                const style = window.getComputedStyle(el);
+                const pos = style.position;
+                const cls = (el.className || '').toString();
+                return pos === 'absolute' || pos === 'fixed' || pos === 'relative'
+                    || /popover|panel|picker|calendar|date|dropdown|pop/i.test(cls)
+                    || el.id === 'main' || (el.parentElement && el.parentElement.id === 'main');
+            };
+            const allVisible = Array.from(document.querySelectorAll('*')).filter(visible);
+            const monthAnchors = allVisible.filter((el) => monthRe.test(el.textContent || '')
+                && el.children.length < 60);
+            const collectAncestors = (el) => {
+                const out = [];
+                let cur = el;
+                while (cur && cur !== document.body && out.length < 6) {
+                    const r = cur.getBoundingClientRect();
+                    out.push({
+                        tag: cur.tagName,
+                        cls: (cur.className || '').toString().slice(0, 200),
+                        id: cur.id,
+                        rect: {x: r.left, y: r.top, w: r.width, h: r.height},
+                    });
+                    cur = cur.parentElement;
+                }
+                return out;
+            };
+            let candidate = null;
+            for (const el of monthAnchors) {
+                let cur = el;
+                while (cur && cur !== document.body) {
+                    if (isLikelyPanel(cur)) { candidate = cur; break; }
+                    cur = cur.parentElement;
+                }
+                if (candidate) break;
+            }
+            if (!candidate) {
+                for (const el of monthAnchors) {
+                    const txt = el.textContent || '';
+                    const digits = (txt.match(/\b\d{1,2}\b/g) || []).length;
+                    if (digits >= 28 && el.children.length < 200) {
+                        let cur = el.parentElement;
+                        while (cur && cur !== document.body) {
+                            const r = cur.getBoundingClientRect();
+                            if (r.width >= 150 && r.width <= 700 && r.height >= 200 && r.height <= 700) {
+                                candidate = cur; break;
+                            }
+                            cur = cur.parentElement;
+                        }
+                        if (candidate) break;
+                    }
+                }
+            }
+            if (!candidate) return { found: false };
+            const match = (candidate.textContent || '').match(/(20\d{2})[-/.年]\s*(\d{1,2})/);
+            const r = candidate.getBoundingClientRect();
+            return {
+                found: true,
+                panel_rect: {x: r.left, y: r.top, w: r.width, h: r.height},
+                current_year: match ? Number(match[1]) : null,
+                current_month: match ? Number(match[2]) : null,
+                panel_tag: candidate.tagName,
+                panel_cls: (candidate.className || '').toString().slice(0, 200),
+                panel_id: candidate.id,
+                ancestors: collectAncestors(candidate),
+            };
+        }
+        """
+    )
+
+
+def _jd_panel_alive(page):
+    """JD 启发式:页面中存在一个可视元素,其文本既包含 'YYYY-M月' 或 'YYYY年M月'
+    又包含至少 28 个数字 cell。不依赖 jmtd / antd / element class。"""
+    return page.evaluate(
+        r"""
+        () => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none' && style.visibility !== 'hidden'
+                    && rect.width > 0 && rect.height > 0 && el.offsetParent !== null;
+            };
+            const monthRe = /20\d{2}[-/.年]\s*\d{1,2}/;
+            const nodes = Array.from(document.querySelectorAll('*')).filter(visible);
+            for (const el of nodes) {
+                const txt = el.textContent || '';
+                if (!monthRe.test(txt)) continue;
+                const digits = (txt.match(/\b\d{1,2}\b/g) || []).length;
+                if (digits >= 28 && el.children.length < 200) return true;
+            }
+            return false;
+        }
+        """
+    )
+
+
+def _jd_reopen_datepicker(page):
+    """如果翻月后面板意外收起,尝试重新触发日期按钮把它拉起来。"""
+    return page.evaluate(
+        r"""
+        () => {
+            const labels = ['日', '周', '月', '自定义', '日 实时', '日 7天', '日 30天'];
+            for (const btn of document.querySelectorAll('button')) {
+                const t = (btn.textContent || '').trim();
+                if (labels.some((l) => t === l || t.startsWith(l + ' '))) {
+                    btn.scrollIntoView({block: 'center'});
+                    btn.click();
+                    return true;
+                }
+            }
+            return false;
+        }
+        """
+    )
+
+
+def _jd_snapshot_date_picker(page, stage, target_date):
+    if not DEBUG_DATE:
+        return
+    try:
+        debug_dir = str(TMP_ROOT / "debug" / "jd_date_picker")
+        os.makedirs(debug_dir, exist_ok=True)
+        ts = datetime.now().strftime("%Y%m%d_%H%M%S_%f")
+        info = page.evaluate(
+            r"""
+            () => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none' && style.visibility !== 'hidden'
+                        && rect.width > 0 && rect.height > 0 && el.offsetParent !== null;
+                };
+                const monthRe = /20\d{2}[-/.年]\s*\d{1,2}/;
+                const allVisible = Array.from(document.querySelectorAll('*')).filter(visible);
+                const monthTextNodes = allVisible.filter((el) => monthRe.test(el.textContent || '')
+                    && el.children.length < 60);
+                let candidate = null;
+                // 找 panel 启发式:含月份文案的元素,向上爬到第一个"看起来像浮层"的祖先
+                // (宽 150-700, 高 200-700,不是 body/main 这种大容器)
+                const isLikelyPanel = (el) => {
+                    if (!el || el === document.body || el === document.documentElement) return false;
+                    const r = el.getBoundingClientRect();
+                    if (r.width < 150 || r.width > 700) return false;
+                    if (r.height < 200 || r.height > 700) return false;
+                    const style = window.getComputedStyle(el);
+                    const pos = style.position;
+                    // 浮层常是 absolute/fixed,或者父级是 popover
+                    return pos === 'absolute' || pos === 'fixed' || pos === 'relative'
+                        || el.className?.toString().includes('popover')
+                        || el.className?.toString().includes('panel')
+                        || el.className?.toString().includes('picker')
+                        || el.id === 'main' || (el.parentElement && el.parentElement.id === 'main');
+                };
+                for (const el of monthTextNodes) {
+                    let cur = el;
+                    while (cur && cur !== document.body) {
+                        if (isLikelyPanel(cur)) { candidate = cur; break; }
+                        cur = cur.parentElement;
+                    }
+                    if (candidate) break;
+                }
+                if (!candidate) {
+                    // 兜底:直接找页面里最大的那个"含月份文案 + 至少 28 个数字 cell"的祖先
+                    for (const el of monthTextNodes) {
+                        const txt = el.textContent || '';
+                        const digits = (txt.match(/\b\d{1,2}\b/g) || []).length;
+                        if (digits >= 28 && el.children.length < 200) {
+                            let cur = el.parentElement;
+                            while (cur && cur !== document.body) {
+                                const r = cur.getBoundingClientRect();
+                                if (r.width >= 150 && r.width <= 700 && r.height >= 200 && r.height <= 700) {
+                                    candidate = cur; break;
+                                }
+                                cur = cur.parentElement;
+                            }
+                            if (candidate) break;
+                        }
+                    }
+                }
+                const collectArrows = () => {
+                    if (!candidate) return [];
+                    const arrowChars = ['‹', '›', '«', '»', '<', '>', '上月', '下月', '上一年', '下一年'];
+                    const out = [];
+                    for (const el of candidate.querySelectorAll('button, a, [role="button"], span, i, em')) {
+                        if (!visible(el)) continue;
+                        const txt = (el.textContent || '').trim();
+                        if (!txt) continue;
+                        if (arrowChars.some((c) => txt === c) || /^(上|下)(一)?(个)?(月|年)$/.test(txt)) {
+                            const r = el.getBoundingClientRect();
+                            out.push({
+                                tag: el.tagName,
+                                cls: (el.className || '').toString().slice(0, 200),
+                                text: txt,
+                                rect: {x: r.left, y: r.top, w: r.width, h: r.height},
+                            });
+                        }
+                    }
+                    return out;
+                };
+                const ancestors = [];
+                if (candidate) {
+                    let cur = candidate;
+                    while (cur && cur !== document.body && ancestors.length < 5) {
+                        const r = cur.getBoundingClientRect();
+                        ancestors.push({
+                            tag: cur.tagName,
+                            cls: (cur.className || '').toString().slice(0, 200),
+                            id: cur.id,
+                            rect: {x: r.left, y: r.top, w: r.width, h: r.height},
+                        });
+                        cur = cur.parentElement;
+                    }
+                }
+                const monthText = candidate ? ((candidate.textContent || '').match(/(20\d{2})[-/.年]\s*(\d{1,2})/) || []) : null;
+                return {
+                    panelFound: !!candidate,
+                    monthAnchorsFound: monthTextNodes.length,
+                    candidate: candidate ? {
+                        tag: candidate.tagName,
+                        cls: (candidate.className || '').toString().slice(0, 300),
+                        currentYear: monthText ? Number(monthText[1]) : null,
+                        currentMonth: monthText ? Number(monthText[2]) : null,
+                    } : null,
+                    ancestors,
+                    arrowBtns: collectArrows(),
+                };
+            }
+            """
+        )
+        payload = {
+            "stage": stage,
+            "ts": ts,
+            "target": {"year": target_date.year, "month": target_date.month, "day": target_date.day},
+            "page": info,
+            "url": page.url,
+        }
+        path = os.path.join(debug_dir, f"{ts}_{stage}.json")
+        with open(path, "w", encoding="utf-8") as f:
+            json.dump(payload, f, ensure_ascii=False, indent=2)
+        try:
+            page.screenshot(path=os.path.join(debug_dir, f"{ts}_{stage}.png"), full_page=False)
+        except Exception:
+            pass
+    except Exception as e:
+        print(f"  [DEBUG snapshot failed] {e}")
+
+
+def _click_day_in_open_datepicker(page, target_date, label):
+    target_year = target_date.year
+    target_month = target_date.month
+    target_day = target_date.day
+
+    _jd_snapshot_date_picker(page, "before_flip", target_date)
+
+    # 翻月循环:启发式找 panel + 翻月按钮,不再依赖 jmtd class。
+    # 优先用 .jmtd-* selector (京东自研),失败再按"上/下个月"文本和 ‹ › 字符找。
+    for step in range(13):  # 够翻 13 个月,5/15 这种历史月够用
+        if not _jd_panel_alive(page):
+            print(f"    [WARN] panel 已收起,尝试重开 (step={step})")
+            _jd_reopen_datepicker(page)
+            page.wait_for_timeout(1500)
+            if not _jd_panel_alive(page):
+                print(f"    [WARN] panel 拉不起来,翻月中止")
+                return False
+        nav_info = page.evaluate(
+            r"""({targetYear, targetMonth, panelRect}) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none' && style.visibility !== 'hidden'
+                        && rect.width > 0 && rect.height > 0 && el.offsetParent !== null;
+                };
+                // 在 panelRect 附近找翻月按钮(浮层可能不在 DOM tree 的 panel 节点内,可能挂在 body 下)
+                // 1) 优先在 panelRect 范围内的按钮
+                const inRect = (rect, btn) => {
+                    const r = btn.getBoundingClientRect();
+                    return r.left >= rect.x - 5 && r.right <= rect.x + rect.w + 5
+                        && r.top >= rect.y - 30 && r.bottom <= rect.y + rect.h + 30;
+                };
+                const collectBtns = (root) => {
+                    const out = [];
+                    for (const sel of ['button', 'a', '[role="button"]', 'span', 'i', 'em', 'div']) {
+                        for (const el of root.querySelectorAll(sel)) {
+                            if (!visible(el)) continue;
+                            const txt = (el.textContent || '').trim();
+                            if (!txt) continue;
+                            // 跳过纯数字 (日期 cell),只关心箭头字符或"上/下月"等
+                            if (/^\d{1,2}$/.test(txt)) continue;
+                            out.push({el, txt});
+                        }
+                    }
+                    return out;
+                };
+                const buttons = collectBtns(document.body);
+                // 过滤:在 panelRect 内的按钮
+                const inPanel = buttons.filter((b) => inRect(panelRect, b.el));
+                const candidates = inPanel.length > 0 ? inPanel : buttons;
+                // 找 prev/next 按钮
+                const arrowPrevChars = ['‹', '«', '<'];
+                const arrowNextChars = ['›', '»', '>'];
+                const findArrow = (chars) => {
+                    for (const {el, txt} of candidates) {
+                        if (chars.includes(txt)) {
+                            const r = el.getBoundingClientRect();
+                            return { x: r.left + r.width / 2, y: r.top + r.height / 2, txt, cls: (el.className || '').toString().slice(0, 200) };
+                        }
+                    }
+                    return null;
+                };
+                const currentYear = panelRect.year;  // 注入
+                const currentMonth = panelRect.month;
+                if (currentYear === targetYear && currentMonth === targetMonth) {
+                    return { status: 'aligned', currentYear, currentMonth };
+                }
+                const diff = (targetYear - currentYear) * 12 + (targetMonth - currentMonth);
+                const wantChars = diff < 0 ? arrowPrevChars : arrowNextChars;
+                const arrow = findArrow(wantChars);
+                if (arrow) {
+                    return { status: 'click', x: arrow.x, y: arrow.y, currentYear, currentMonth, diff, btn: 'arrow:' + arrow.txt, btnCls: arrow.cls };
+                }
+                // 文本按钮
+                const wantText = diff < 0 ? ['上个月', '上一年', '上月'] : ['下个月', '下一年', '下月'];
+                for (const {el, txt} of candidates) {
+                    if (wantText.some((w) => txt === w || (txt.includes(w) && txt.length <= w.length + 2))) {
+                        const r = el.getBoundingClientRect();
+                        return { status: 'click', x: r.left + r.width / 2, y: r.top + r.height / 2, currentYear, currentMonth, diff, btn: 'text:' + txt };
+                    }
+                }
+                return { status: 'no_nav', currentYear, currentMonth, diff };
+            }""",
+            {
+                "targetYear": target_year,
+                "targetMonth": target_month,
+                "panelRect": {
+                    "x": _jd_find_panel(page).get("panel_rect", {}).get("x", 0),
+                    "y": _jd_find_panel(page).get("panel_rect", {}).get("y", 0),
+                    "w": _jd_find_panel(page).get("panel_rect", {}).get("w", 0),
+                    "h": _jd_find_panel(page).get("panel_rect", {}).get("h", 0),
+                    "year": _jd_find_panel(page).get("current_year"),
+                    "month": _jd_find_panel(page).get("current_month"),
+                },
+            },
+        )
+        if not isinstance(nav_info, dict):
+            nav_info = {"status": str(nav_info)}
+        if nav_info.get("status") == "aligned":
+            break
+        if nav_info.get("status") == "click":
+            page.mouse.move(nav_info["x"], nav_info["y"])
+            page.wait_for_timeout(150)
+            page.mouse.click(nav_info["x"], nav_info["y"])
+            page.wait_for_timeout(800)
+            _jd_snapshot_date_picker(page, f"after_flip_step_{step}", target_date)
+            continue
+        print(f"    [WARN] 翻月停止: {nav_info}")
+        _jd_snapshot_date_picker(page, f"no_nav_step_{step}", target_date)
+        return False
+
+    # 翻完月后点日
+    click_info = page.evaluate(
+        r"""({targetDay}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0
+                    && el.offsetParent !== null;
+            };
+            const disabled = (el) => {
+                const classText = `${el.className || ''} ${el.parentElement?.className || ''} ${el.parentElement?.parentElement?.className || ''}`;
+                return el.disabled
+                    || el.getAttribute('aria-disabled') === 'true'
+                    || /disabled|off|unavailable|diff-month|other-month|prev-month|next-month|not-current-month/.test(classText);
+            };
+            const monthRe = /20\d{2}[-/.年]\s*\d{1,2}/;
+            const isLikelyPanel = (el) => {
+                if (!el || el === document.body || el === document.documentElement) return false;
+                const r = el.getBoundingClientRect();
+                if (r.width < 150 || r.width > 700) return false;
+                if (r.height < 200 || r.height > 700) return false;
+                const style = window.getComputedStyle(el);
+                const pos = style.position;
+                const cls = (el.className || '').toString();
+                return pos === 'absolute' || pos === 'fixed' || pos === 'relative'
+                    || /popover|panel|picker|calendar|date|dropdown|pop/i.test(cls)
+                    || el.id === 'main' || (el.parentElement && el.parentElement.id === 'main');
+            };
+            const all = Array.from(document.querySelectorAll('*')).filter(visible);
+            const monthAnchors = all.filter((el) => monthRe.test(el.textContent || '')
+                && el.children.length < 60);
+            let panel = null;
+            for (const el of monthAnchors) {
+                let cur = el;
+                while (cur && cur !== document.body) {
+                    if (isLikelyPanel(cur)) { panel = cur; break; }
+                    cur = cur.parentElement;
+                }
+                if (panel) break;
+            }
+            if (!panel) return { status: 'no_panel' };
+            const wanted = String(targetDay);
+            const cells = Array.from(panel.querySelectorAll('td[role="gridcell"], [role="gridcell"], .ant-calendar-date, .ant-picker-cell, .el-date-table td, td, [class*="cell"]'))
+                .filter(visible)
+                .filter((cell) => !disabled(cell));
+            for (const cell of cells) {
+                const text = (cell.textContent || '').trim().replace(/^0+(\d)$/, '$1');
+                if (text === wanted) {
+                    const rect = cell.getBoundingClientRect();
+                    return { status: 'found', x: rect.left + rect.width / 2, y: rect.top + rect.height / 2, text };
+                }
+            }
+            return { status: 'not_found' };
+        }""",
+        {"targetDay": target_day},
+    )
+    print(f"    日期选择结果: {click_info}")
+    if isinstance(click_info, dict) and click_info.get("status") == "found":
+        page.mouse.move(click_info["x"], click_info["y"])
+        page.wait_for_timeout(150)
+        page.mouse.click(click_info["x"], click_info["y"])
+        page.wait_for_timeout(2000)
+        return True
+
+    try:
+        from jd_data_collector import save_date_picker_debug
+        save_date_picker_debug(page, f"{label}_yesterday_not_found")
+    except Exception:
+        pass
+    return False
+
+
+def select_yesterday_in_combo_datepicker(page, label="product_date"):
+    target = get_yesterday_date()
+    target_day = target.strftime("%Y-%m-%d")
+    print(f"  设置日期为昨天: {target_day}")
+
+    if _date_text_matches_day(_combo_date_label(page), target_day):
+        print(f"    [OK] 日期已是昨天: {_combo_date_label(page)}")
+        return True
+
+    _click_combo_mode(page, "day")
+    page.wait_for_timeout(1200)
+
+    print(f"    [TRACE] 尝试翻日 (最多 120 次 × 1.5s)")
+    flip_ok = _select_yesterday_by_combo_flip(page, target_day)
+    print(f"    [TRACE] 翻日结果: {flip_ok}, 当前 label: {_combo_date_label(page)}")
+    if flip_ok:
+        print("  [OK] 已切换到昨天数据")
+        return True
+
+    print(f"    [TRACE] 翻日失败,尝试打开日历面板")
+    opened = _open_combo_date_panel(page)
+    print(f"    [TRACE] 打开面板结果: {opened}")
+    if _click_day_in_open_datepicker(page, target, label):
+        if _date_text_matches_day(_combo_date_label(page), target_day):
+            print("  [OK] 已切换到昨天数据")
+            return True
+
+    # Fallback: 翻日 + 点日历都失败 (例如 21 天前超出 8 次翻日限制),
+    # 切到"自定义 (range)"模式, 直接在两个 input 里填日期.
+    print(f"    [TRACE] 翻月也失败,尝试 range 输入框")
+    if _select_via_range_input(page, target_day):
+        print(f"  [OK] 通过自定义 range 模式选到 {target_day}")
+        return True
+
+    raise RuntimeError(f"未能选择昨天日期: {target_day}")
+
+
+def _select_via_range_input(page, target_day: str) -> bool:
+    """自定义 (range) 模式: 点'自定义'标签, 弹出 RangePicker, 直接在
+    '开始日期' / '结束日期' 两个 input 里填 target_day, 点确定.
+
+    用作翻日选择器的 fallback — 当 target 距离当前日期太远 (例如 21 天前),
+    翻 8 次根本到不了, 走 range 输入更稳.
+    """
+    if not _click_combo_mode(page, "range"):
+        return False
+    page.wait_for_timeout(1500)
+
+    fill_info = page.evaluate(
+        r"""({targetDay}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const s = window.getComputedStyle(el);
+                const r = el.getBoundingClientRect();
+                return s.display !== 'none' && s.visibility !== 'hidden'
+                    && r.width > 0 && r.height > 0;
+            };
+            const root = document.querySelector(
+                '.jmtd-popover-open .jmtd-date-picker-range-panel, ' +
+                '.jmtd-popover-open .jmtd-date-picker-panel, ' +
+                '.jmtd-popover-open [class*="range-panel"]'
+            ) || document;
+            const inputs = Array.from(root.querySelectorAll('input'))
+                .filter(visible)
+                .filter((el) => !el.disabled && el.type !== 'hidden');
+            if (inputs.length < 2) return {status: 'no_inputs', count: inputs.length};
+            const startRect = inputs[0].getBoundingClientRect();
+            const endRect = inputs[1].getBoundingClientRect();
+            return {
+                status: 'found',
+                start: {x: startRect.left + startRect.width / 2, y: startRect.top + startRect.height / 2},
+                end: {x: endRect.left + endRect.width / 2, y: endRect.top + endRect.height / 2},
+            };
+        }""",
+        {"targetDay": target_day},
+    )
+    if not isinstance(fill_info, dict) or fill_info.get("status") != "found":
+        print(f"    [WARN] range picker input 找不到: {fill_info}")
+        return False
+
+    # Click + select all + type, 模拟用户输入 (避免直接 fill 触发 onChange 但不触发 onPressEnter).
+    page.mouse.click(fill_info["start"]["x"], fill_info["start"]["y"])
+    page.wait_for_timeout(300)
+    page.keyboard.press("Control+a")
+    page.keyboard.type(target_day, delay=40)
+    page.keyboard.press("Tab")
+    page.wait_for_timeout(300)
+    page.mouse.click(fill_info["end"]["x"], fill_info["end"]["y"])
+    page.wait_for_timeout(300)
+    page.keyboard.press("Control+a")
+    page.keyboard.type(target_day, delay=40)
+    page.keyboard.press("Enter")
+    page.wait_for_timeout(1500)
+
+    # 点确定 (有些 range picker 用'确定'按钮, 有些不点也会触发 onChange).
+    confirm_clicked = page.evaluate(
+        r"""() => {
+            const visible = (el) => {
+                if (!el) return false;
+                const s = window.getComputedStyle(el);
+                const r = el.getBoundingClientRect();
+                return s.display !== 'none' && s.visibility !== 'hidden'
+                    && r.width > 0 && r.height > 0;
+            };
+            const roots = document.querySelectorAll(
+                '.jmtd-popover-open .jmtd-date-picker-range-panel, ' +
+                '.jmtd-popover-open .jmtd-date-picker-panel, ' +
+                '.jmtd-popover-open [class*="range-panel"]'
+            );
+            for (const root of roots) {
+                for (const btn of root.querySelectorAll('button')) {
+                    const text = (btn.textContent || '').trim();
+                    if (['确定', '应用', '应用查询', '查询', '确认'].includes(text)) {
+                        if (visible(btn) && !btn.disabled) {
+                            const r = btn.getBoundingClientRect();
+                            return {x: r.left + r.width / 2, y: r.top + r.height / 2, text};
+                        }
+                    }
+                }
+            }
+            return null;
+        }""",
+    )
+    if confirm_clicked:
+        page.mouse.click(confirm_clicked["x"], confirm_clicked["y"])
+        page.wait_for_timeout(1000)
+
+    return _date_text_matches_day(_combo_date_label(page), target_day)
+
+
+def select_week_and_last_week(page):
+    """选择天模式并切换到昨天。函数名保留，避免改动调用链。"""
+    print("  选择时间范围为'天'并切换到昨天...")
+    select_yesterday_in_combo_datepicker(page, "product_detail_list")
+    page.wait_for_timeout(5000)
+
+
+# ── 商品明细页：翻页采集所有 SPU ─────────────────────────────────────────────
+
+def extract_product_detail_rows(page):
+    """提取当前页商品明细表格数据（排除合计行）"""
+    data = extract(page, """() => {
+        const rows = document.querySelectorAll('.jmtd-table-main tbody tr');
+        const result = [];
+        for (const row of rows) {
+            if (row.classList.contains('jmtd-table-plugin-total-row')) continue;
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 2) continue;
+
+            const productCell = cells[0];
+            const nameEl = productCell.querySelector('.jmtd-table-formatter-name-name .jmtd-label');
+            const productName = nameEl ? nameEl.textContent.trim() : '';
+
+            const descEl = productCell.querySelector('.jmtd-table-formatter-name-desc');
+            const descText = descEl ? descEl.textContent.trim() : '';
+            const spuMatch = descText.match(/SPU[\\s:：]*(\\d+)/);
+            const spu = spuMatch ? spuMatch[1] : '';
+
+            const imgEl = productCell.querySelector('.jmtd-table-formatter-name-img');
+            let imgSrc = imgEl ? (imgEl.getAttribute('src') || imgEl.getAttribute('data-src') || '') : '';
+            if (imgSrc.startsWith('//')) imgSrc = 'https:' + imgSrc;
+
+            const linkEl = productCell.querySelector('.jmtd-table-formatter-name-link');
+            let productUrl = linkEl ? (linkEl.getAttribute('href') || '') : '';
+            if (productUrl.startsWith('//')) productUrl = 'https:' + productUrl;
+
+            function extractMetric(cell) {
+                const wrap = cell.querySelector('.rate-target-warp');
+                if (!wrap) return { value: cell.textContent.trim(), compare: '', rate: '' };
+                return {
+                    value: wrap.querySelector('.first-value')?.textContent?.trim() || '',
+                    compare: wrap.querySelector('.second-value')?.textContent?.trim() || '',
+                    rate: wrap.querySelector('.target-value')?.textContent?.trim() || '',
+                };
+            }
+            const headers = ['访客数','浏览量','成交客户数','成交商品件数','成交金额'];
+            const metrics = {};
+            for (let i = 1; i < cells.length; i++) {
+                metrics[headers[i-1] || 'col_'+i] = extractMetric(cells[i]);
+            }
+            result.push({ spu, product: productName, image: imgSrc, url: productUrl, metrics });
+        }
+        return JSON.stringify(result);
+    }""", "[]")
+    try:
+        return json.loads(data) if isinstance(data, str) else []
+    except:
+        return []
+
+
+def has_next_page(page):
+    try:
+        btn = page.locator('.jmtd-pagination-item-next').first
+        return btn.count() > 0 and not btn.evaluate("el => el.classList.contains('jmtd-pagination-item-disabled')")
+    except:
+        return False
+
+
+def click_next_page(page):
+    try:
+        # 记录当前第一条 SPU，用于判断翻页后是否刷新
+        first_spu_before = page.evaluate("""() => {
+            const desc = document.querySelector('.jmtd-table-main tbody tr:not(.jmtd-table-plugin-total-row) .jmtd-table-formatter-name-desc');
+            if (!desc) return '';
+            const m = desc.textContent.match(/SPU[\\s:：]*(\\d+)/);
+            return m ? m[1] : '';
+        }""")
+
+        page.locator('.jmtd-pagination-item-next').first.click()
+        print("    [OK] 点击下一页")
+
+        # 智能等待：等表格第一条 SPU 变化 或 loading 消失，最多 4 秒
+        for _ in range(20):
+            page.wait_for_timeout(200)
+            # 检查是否还有 loading
+            loading = page.evaluate("""() => {
+                const el = document.querySelector('.jmtd-loading');
+                return el ? window.getComputedStyle(el).display !== 'none' : false;
+            }""")
+            if loading:
+                continue
+            # 检查第一条 SPU 是否已变化
+            first_spu_after = page.evaluate("""() => {
+                const desc = document.querySelector('.jmtd-table-main tbody tr:not(.jmtd-table-plugin-total-row) .jmtd-table-formatter-name-desc');
+                if (!desc) return '';
+                const m = desc.textContent.match(/SPU[\\s:：]*(\\d+)/);
+                return m ? m[1] : '';
+            }""")
+            if first_spu_after and first_spu_after != first_spu_before:
+                break
+        page.wait_for_timeout(1000)
+        return True
+    except Exception as e:
+        print(f"    [WARN] 点击下一页异常: {e}")
+        return False
+
+
+def collect_all_spus(page):
+    """翻页采集商品明细所有 SPU"""
+    print("[Step 5] 翻页采集商品明细所有 SPU...")
+    select_week_and_last_week(page)
+
+    all_rows = []
+    page_num = 1
+    while True:
+        print(f"\n  采集第 {page_num} 页...")
+        rows = extract_product_detail_rows(page)
+        print(f"    提取到 {len(rows)} 条")
+        for r in rows:
+            all_rows.append(r)
+            print(f"      SPU: {r.get('spu','')} | {r.get('product','')[:40]}...")
+        if not has_next_page(page):
+            print(f"\n  [OK] 共采集 {len(all_rows)} 个 SPU")
+            break
+        click_next_page(page)
+        page_num += 1
+    return all_rows
+
+
+# ── 逐个 SPU 搜索 → 点击详情 → 新页面 ─────────────────────────────────────────
+
+def clear_search_tags(page):
+    """清除搜索框中已存在的 SPU tag"""
+    print("    清除已有搜索标签...")
+    try:
+        # 优先点击"清空全部"按钮
+        clear_all = page.locator('.jmtd-base-input-icon-clear button').first
+        if clear_all.count() > 0 and clear_all.is_visible():
+            clear_all.click()
+            page.wait_for_timeout(800)
+            print("      [OK] 已点击清空全部")
+            return
+    except Exception:
+        pass
+
+    # 兜底：逐个点击 tag 的关闭按钮
+    try:
+        tags = page.locator('.search-multiple-input .jmtd-tag .jmtd-tag-close-icon').all()
+        if tags:
+            for tag in tags:
+                tag.click()
+                page.wait_for_timeout(300)
+            print(f"      [OK] 已关闭 {len(tags)} 个标签")
+        else:
+            print("      [INFO] 没有已有标签")
+    except Exception as e:
+        print(f"      [WARN] 关闭标签失败: {e}")
+
+
+def search_spu_in_detail_page(page, spu):
+    """在商品明细页搜索框输入 SPU 并搜索"""
+    print(f"  搜索 SPU: {spu}")
+    try:
+        # 1. 先清除已有标签
+        clear_search_tags(page)
+
+        # 2. 定位输入框并聚焦
+        input_box = page.locator('.search-multiple-input .jmtd-multiple-input-input').first
+        input_box.wait_for(state="visible", timeout=5000)
+        input_box.click()
+        page.wait_for_timeout(300)
+
+        # 3. 逐字输入（某些组件对 fill 不敏感，type 更可靠）
+        input_box.type(spu, delay=50)
+        page.wait_for_timeout(500)
+
+        # 4. 按 Enter 把输入转成 tag（multi-input 通常需要 Enter 确认）
+        input_box.press("Enter")
+        page.wait_for_timeout(500)
+
+        # 5. 额外触发 input 事件确保组件收到值
+        page.evaluate(f"""() => {{
+            const input = document.querySelector('.search-multiple-input .jmtd-multiple-input-input');
+            if (input) {{
+                input.value = '{spu}';
+                input.dispatchEvent(new Event('input', {{ bubbles: true }}));
+                input.dispatchEvent(new Event('change', {{ bubbles: true }}));
+            }}
+        }}""")
+        page.wait_for_timeout(300)
+
+        # 6. 点击搜索按钮
+        search_btn = page.locator('.search-multiple-input .jmtd-input-search-button').first
+        search_btn.wait_for(state="visible", timeout=3000)
+        search_btn.click()
+        print("    [OK] 已点击搜索")
+    except Exception as e:
+        print(f"    [WARN] 精确搜索失败: {e}")
+        # 兜底：纯 JS 操作
+        page.evaluate(f"""() => {{
+            // 先清除已有 tag
+            document.querySelectorAll('.search-multiple-input .jmtd-tag .jmtd-tag-close-icon').forEach(el => el.click());
+            const input = document.querySelector('.search-multiple-input .jmtd-multiple-input-input');
+            const btn = document.querySelector('.jmtd-input-search-button.jmtd-button-small');
+            if (input) {{
+                input.focus();
+                input.value = '{spu}';
+                input.dispatchEvent(new Event('input', {{ bubbles: true }}));
+                input.dispatchEvent(new Event('change', {{ bubbles: true }}));
+                // 模拟回车生成 tag
+                input.dispatchEvent(new KeyboardEvent('keydown', {{ key: 'Enter', bubbles: true }}));
+            }}
+            if (btn) {{ btn.click(); return 'js_searched'; }}
+            return 'not_found';
+        }}""")
+
+    # 7. 等待搜索结果刷新
+    for _ in range(25):
+        page.wait_for_timeout(200)
+        loading = page.evaluate("""() => {
+            const el = document.querySelector('.jmtd-loading');
+            return el ? window.getComputedStyle(el).display !== 'none' : false;
+        }""")
+        if not loading:
+            break
+    page.wait_for_timeout(1000)
+
+
+def click_detail_and_wait_new_page(page, context, spu):
+    """点击搜索结果中的'详情'按钮，等待并返回新页面"""
+    print(f"  点击'详情'按钮...")
+    try:
+        # 等待表格刷新（搜索结果应该只有1条）
+        page.wait_for_selector('.jmtd-table-main tbody tr:not(.jmtd-table-plugin-total-row)', timeout=10000)
+
+        # 点击详情按钮（不是取消/关注按钮）
+        with context.expect_page(timeout=15000) as new_page_info:
+            detail_btn = page.locator(
+                '.jmtd-table-main tbody tr:not(.jmtd-table-plugin-total-row) td:last-child .detail-btn',
+                has_text="详情"
+            ).first
+            detail_btn.wait_for(state="visible", timeout=5000)
+            detail_btn.click()
+            print("    [OK] 已点击详情")
+
+        new_page = new_page_info.value
+        new_page.wait_for_load_state("networkidle", timeout=15000)
+        page.wait_for_timeout(3000)
+        print(f"    [OK] 新页面已打开: {new_page.url[:80]}...")
+        return new_page
+    except Exception as e:
+        print(f"    [FAIL] 点击详情或打开新页面失败: {e}")
+        return None
+
+
+# ── 新页面：核心数据 tab ──────────────────────────────────────────────────────
+
+def click_core_data_tab(page):
+    """在新页面点击'核心数据'tab"""
+    print("  切换到'核心数据'...")
+    try:
+        tab = page.locator('.single-header-tabs .jmtd-tabs-tab', has_text="核心数据").first
+        tab.wait_for(state="visible", timeout=5000)
+        is_active = tab.evaluate("el => el.classList.contains('jmtd-tabs-tab-active')")
+        if not is_active:
+            tab.click()
+            page.wait_for_timeout(3000)
+            print("    [OK] 已点击核心数据")
+        else:
+            print("    [OK] 核心数据已激活")
+    except Exception as e:
+        print(f"    [WARN] 点击核心数据失败: {e}")
+        page.evaluate("""() => {
+            for (const tab of document.querySelectorAll('.single-header-tabs .jmtd-tabs-tab')) {
+                if (tab.textContent.includes('核心数据') && !tab.classList.contains('jmtd-tabs-tab-active')) {
+                    tab.click(); return 'clicked';
+                }
+            }
+            return 'already_active_or_not_found';
+        }""")
+        page.wait_for_timeout(3000)
+
+
+# ── 新页面：流量来源 tab ───────────────────────────────────────────────────────
+
+def click_traffic_source_tab(page):
+    """在新页面点击'流量来源'tab"""
+    print("  切换到'流量来源'...")
+    try:
+        tab = page.locator('.single-header-tabs .jmtd-tabs-tab', has_text="流量来源").first
+        tab.wait_for(state="visible", timeout=5000)
+        is_active = tab.evaluate("el => el.classList.contains('jmtd-tabs-tab-active')")
+        if not is_active:
+            tab.click()
+            page.wait_for_timeout(3000)
+            print("    [OK] 已点击流量来源")
+        else:
+            print("    [OK] 流量来源已激活")
+    except Exception as e:
+        print(f"    [WARN] 点击流量来源失败: {e}")
+        page.evaluate("""() => {
+            for (const tab of document.querySelectorAll('.single-header-tabs .jmtd-tabs-tab')) {
+                if (tab.textContent.includes('流量来源') && !tab.classList.contains('jmtd-tabs-tab-active')) {
+                    tab.click(); return 'clicked';
+                }
+            }
+            return 'already_active_or_not_found';
+        }""")
+        page.wait_for_timeout(3000)
+
+
+def expand_all_traffic_rows(page):
+    """展开流量来源表格的所有行"""
+    print("  展开所有行...")
+    page.evaluate("""() => {
+        const expandIcons = document.querySelectorAll('.jmti-plus-square-f');
+        for (const icon of expandIcons) { icon.click(); }
+        return expandIcons.length;
+    }""")
+    page.wait_for_timeout(2000)
+    print("    [OK] 已展开所有行")
+
+
+def extract_traffic_data(page):
+    """提取流量来源表格数据（含数值和涨跌百分比）"""
+    print("  采集流量来源数据...")
+    data = extract(page, """() => {
+        // 定位流量来源卡片内的表格（避免和关键词表格混淆）
+        const cards = document.querySelectorAll('.jmtd-card');
+        let targetTable = null;
+        for (const card of cards) {
+            const header = card.querySelector('.jmtd-card-header');
+            if (header && header.textContent.includes('流量来源')) {
+                targetTable = card.querySelector('.jmtd-table-main');
+                break;
+            }
+        }
+        if (!targetTable) {
+            // 兜底：直接取页面第一个 jmtd-table-main
+            targetTable = document.querySelector('.jmtd-table-main');
+        }
+        if (!targetTable) return '[]';
+
+        const rows = targetTable.querySelectorAll('tbody tr');
+        const headers = Array.from(targetTable.querySelectorAll('thead th')).map(th => th.textContent.trim());
+        const result = [];
+
+        for (const row of rows) {
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 2) continue;
+            const obj = {};
+            cells.forEach((cell, i) => {
+                if (headers[i] === '操作') return;
+                if (i === 0) {
+                    // 流量来源名称
+                    const sourceName = cell.querySelector('.source-name span')?.textContent?.trim()
+                        || cell.textContent.trim();
+                    obj[headers[i]] = sourceName;
+                } else {
+                    // 数值 + 环比：提取所有 <p> 子元素
+                    const ps = cell.querySelectorAll(':scope > p');
+                    if (ps.length >= 2) {
+                        const value = ps[0].textContent.trim();
+                        const rateEl = ps[1];
+                        const rate = rateEl.textContent.trim();
+                        // 判断涨跌趋势
+                        let trend = '';
+                        const cls = rateEl.className || '';
+                        if (cls.includes('sz-frontend-ft')) trend = 'up';
+                        else if (cls.includes('sz-frontend-fu')) trend = 'down';
+                        else if (rate.startsWith('+')) trend = 'up';
+                        else if (rate.startsWith('-')) trend = 'down';
+                        obj[headers[i]] = { value, rate, trend };
+                    } else if (ps.length === 1) {
+                        // 只有数值没有环比
+                        obj[headers[i]] = { value: ps[0].textContent.trim(), rate: '', trend: '' };
+                    } else {
+                        // 兜底：纯文本，尝试拆分
+                        const txt = cell.textContent.trim();
+                        const numMatch = txt.match(/^([^+\x2d±]*[0-9]*)([±+\x2d]?[0-9,.%￥]+)(.*)$/);
+                        if (numMatch) {
+                            obj[headers[i]] = { value: numMatch[1].trim() || numMatch[2].trim(), rate: numMatch[3].trim() || '', trend: '' };
+                        } else {
+                            obj[headers[i]] = { value: txt, rate: '', trend: '' };
+                        }
+                    }
+                }
+            });
+            if (Object.keys(obj).length > 0) result.push(obj);
+        }
+        return JSON.stringify(result);
+    }""", "[]")
+    try:
+        result = json.loads(data) if isinstance(data, str) else []
+    except:
+        result = []
+    print(f"    [OK] 提取到 {len(result)} 条流量来源数据")
+    return result
+
+
+def extract_keyword_data(page):
+    """提取关键词来源成交表格数据（翻页收集最多20个）"""
+    print("  采集关键词来源成交数据...")
+    all_keywords = []
+    max_keywords = 20
+
+    for page_num in range(1, 5):  # 最多4页
+        print(f"    采集第 {page_num} 页...")
+        # 等待表格刷新
+        for _ in range(20):
+            page.wait_for_timeout(200)
+            loading = page.evaluate("""() => {
+                const el = document.querySelector('.jmtd-loading');
+                return el ? window.getComputedStyle(el).display !== 'none' : false;
+            }""")
+            if not loading:
+                break
+
+        data = extract(page, """() => {
+            const cards = document.querySelectorAll('.jmtd-card');
+            let targetTable = null;
+            for (const card of cards) {
+                const header = card.querySelector('.jmtd-card-header');
+                if (header && header.textContent.includes('关键词来源成交')) {
+                    targetTable = card.querySelector('.jmtd-table-main');
+                    break;
+                }
+            }
+            if (!targetTable) return '[]';
+            const rows = targetTable.querySelectorAll('tbody tr');
+            const headers = Array.from(targetTable.querySelectorAll('thead th')).map(th => th.textContent.trim());
+            const result = [];
+            for (const row of rows) {
+                const cells = row.querySelectorAll('td');
+                if (cells.length < 2) continue;
+                const obj = {};
+                cells.forEach((cell, i) => {
+                    if (headers[i] === '操作') return;
+                    obj[headers[i]] = cell.textContent.trim();
+                });
+                if (Object.keys(obj).length > 0) result.push(obj);
+            }
+            return JSON.stringify(result);
+        }""", "[]")
+
+        try:
+            rows = json.loads(data) if isinstance(data, str) else []
+        except:
+            rows = []
+        print(f"      提取到 {len(rows)} 条")
+        for r in rows:
+            all_keywords.append(r)
+
+        if len(all_keywords) >= max_keywords:
+            break
+
+        # 点击下一页（限定在关键词卡片内）
+        has_next = page.evaluate("""() => {
+            const cards = document.querySelectorAll('.jmtd-card');
+            let nextBtn = null;
+            for (const card of cards) {
+                const header = card.querySelector('.jmtd-card-header');
+                if (header && header.textContent.includes('关键词来源成交')) {
+                    nextBtn = card.querySelector('.jmtd-pagination-simple .jmtd-pagination-item-next');
+                    break;
+                }
+            }
+            if (!nextBtn) return false;
+            if (nextBtn.classList.contains('jmtd-pagination-item-disabled')) return false;
+            nextBtn.click();
+            return true;
+        }""")
+
+        if not has_next:
+            print("    [INFO] 无更多分页")
+            break
+        page.wait_for_timeout(2000)
+
+    print(f"    [OK] 共提取 {len(all_keywords)} 条关键词数据")
+    return all_keywords[:max_keywords]
+
+
+# ── 新页面：人群画像 tab ──────────────────────────────────────────────────────
+
+def click_crowd_portrait_tab(page):
+    """在新页面点击'人群画像'tab"""
+    print("  切换到'人群画像'...")
+    try:
+        tab = page.locator('.single-header-tabs .jmtd-tabs-tab', has_text="人群画像").first
+        tab.wait_for(state="visible", timeout=5000)
+        is_active = tab.evaluate("el => el.classList.contains('jmtd-tabs-tab-active')")
+        if not is_active:
+            tab.click()
+            page.wait_for_timeout(3000)
+            print("    [OK] 已点击人群画像")
+        else:
+            print("    [OK] 人群画像已激活")
+    except Exception as e:
+        print(f"    [WARN] 点击人群画像失败: {e}")
+        page.evaluate("""() => {
+            for (const tab of document.querySelectorAll('.single-header-tabs .jmtd-tabs-tab')) {
+                if (tab.textContent.includes('人群画像') && !tab.classList.contains('jmtd-tabs-tab-active')) {
+                    tab.click(); return 'clicked';
+                }
+            }
+            return 'already_active_or_not_found';
+        }""")
+        page.wait_for_timeout(3000)
+
+
+def extract_crowd_texts(page, section_title):
+    """从人群画像指定区域提取所有可见文本"""
+    data = page.evaluate(f"""() => {{
+        const result = {{title: '{section_title}', texts: []}};
+        // 提取所有 .jmtd-typography-text 的文本
+        document.querySelectorAll('.jmtd-typography-text').forEach(el => {{
+            const txt = el.textContent.trim();
+            if (txt && txt.length > 2 && txt.length < 200) result.texts.push(txt);
+        }});
+        // 去重
+        result.texts = [...new Set(result.texts)];
+        return JSON.stringify(result);
+    }}""")
+    try:
+        return json.loads(data) if isinstance(data, str) else {"texts": []}
+    except:
+        return {"texts": []}
+
+
+def extract_crowd_text_fallback(page, keywords):
+    """从页面文本中提取匹配关键词的信息"""
+    result = page.evaluate("""(keywords) => {
+        const texts = document.querySelectorAll('.jmtd-typography-text');
+        for (const el of texts) {
+            const txt = el.textContent.trim();
+            for (const kw of keywords) {
+                if (txt.includes(kw)) return txt;
+            }
+        }
+        return '';
+    }""", keywords)
+    return result if result else ''
+
+
+def extract_region_table_data(page):
+    """提取地区分布表格"""
+    data = page.evaluate("""() => {
+        const rows = document.querySelectorAll('.jmtd-table-main tbody tr');
+        const result = [];
+        for (const row of rows) {
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 3) continue;
+            const rank = cells[0].textContent.trim();
+            const region = cells[1].textContent.trim();
+            const ratio = cells[2].querySelector('.jmtd-table-formatter-ratio-bar-number')?.textContent?.trim() || cells[2].textContent.trim();
+            result.push({rank, region, ratio});
+        }
+        return JSON.stringify(result);
+    }""")
+    try:
+        return json.loads(data) if isinstance(data, str) else []
+    except:
+        return []
+
+
+def click_city_radio(page):
+    """点击城市radio按钮"""
+    clicked = page.evaluate("""() => {
+        const labels = document.querySelectorAll('.jmtd-form-control-button-label');
+        for (const label of labels) {
+            if (label.textContent.includes('城市')) {
+                label.click();
+                return true;
+            }
+        }
+        return false;
+    }""")
+    if clicked:
+        print("    [OK] 已点击城市radio")
+    else:
+        print("    [WARN] 未找到城市radio")
+    page.wait_for_timeout(2000)
+
+
+def extract_echarts_instance_data(page):
+    """尝试通过ECharts实例直接获取图表数据（最优先策略）"""
+    try:
+        data = page.evaluate("""() => {
+            const results = [];
+
+        function visible(el) {
+            if (!el) return false;
+            const rect = el.getBoundingClientRect();
+            const style = window.getComputedStyle(el);
+            return rect.width > 30 && rect.height > 30 && style.display !== 'none' && style.visibility !== 'hidden';
+        }
+
+        function closestTitle(dom) {
+            const card = dom.closest('.jmtd-card, .jmtd-col, .jmtd-row-item, [class*="card"], [class*="panel"]');
+            if (card) {
+                const titleEl = card.querySelector('.jmtd-card-header, .jmtd-typography-title, .jmtd-label, [class*="title"]');
+                const title = titleEl ? titleEl.textContent.trim() : '';
+                if (title) return title;
+            }
+            let p = dom.parentElement;
+            for (let i = 0; p && i < 5; i++, p = p.parentElement) {
+                const text = p.textContent.trim();
+                if (/性别|年龄|职业/.test(text)) {
+                    if (text.includes('性别')) return '性别';
+                    if (text.includes('年龄')) return '年龄';
+                    if (text.includes('职业')) return '职业';
+                }
+            }
+            return '';
+        }
+
+        function findInstance(dom) {
+            let instance = null;
+            // 方式1: 常见内部属性
+            const keys = ['_echartsInstance', '__echartsInstance', '_echarts', '__echarts__', '__EC_CHARTS__', '_echarts_instance_'];
+            for (const k of keys) {
+                if (dom[k] && typeof dom[k].getOption === 'function') {
+                    instance = dom[k];
+                    break;
+                }
+            }
+            // 方式2: ECharts常见实例ID属性
+            if (!instance && window.echarts && typeof window.echarts.getInstanceByDom === 'function') {
+                try { instance = window.echarts.getInstanceByDom(dom); } catch(e) {}
+            }
+            // 方式3: 遍历所有可枚举属性
+            if (!instance) {
+                for (const key of Object.keys(dom)) {
+                    try {
+                        const val = dom[key];
+                        if (val && typeof val.getOption === 'function') {
+                            instance = val;
+                            break;
+                        }
+                    } catch(e) {}
+                }
+            }
+            return instance;
+        }
+
+        function findChartDomCandidates() {
+            const set = new Set();
+            document.querySelectorAll('.jmtc-container, [_echarts_instance_]').forEach(el => {
+                if (visible(el)) set.add(el);
+            });
+            document.querySelectorAll('canvas[data-zr-dom-id], canvas').forEach(canvas => {
+                let p = canvas.parentElement;
+                for (let i = 0; p && i < 6; i++, p = p.parentElement) {
+                    if (visible(p)) set.add(p);
+                    if (findInstance(p)) break;
+                }
+            });
+            return [...set];
+        }
+
+        function toNumber(v) {
+            if (typeof v === 'number') return v;
+            if (Array.isArray(v)) return Number(v.find(x => !Number.isNaN(Number(x))) || 0);
+            if (v && typeof v === 'object') return Number(v.value || v[1] || 0);
+            return Number(String(v || '').replace(/,/g, '').replace('%', ''));
+        }
+
+        function formatPercent(num) {
+            if (!Number.isFinite(num)) return '';
+            const fixed = num.toFixed(2).replace(/\\.00$/, '').replace(/(\\.\\d)0$/, '$1');
+            return fixed + '%';
+        }
+
+        const charts = findChartDomCandidates();
+        const seenCharts = new Set();
+        for (const dom of charts) {
+            const instance = findInstance(dom);
+            const chartKey = instance && (instance.id || dom.getAttribute('_echarts_instance_'));
+            if (!instance || (chartKey && seenCharts.has(chartKey))) continue;
+            if (chartKey) seenCharts.add(chartKey);
+
+            try {
+                const opt = instance.getOption();
+                const seriesList = Array.isArray(opt.series) ? opt.series : [];
+                for (const series of seriesList) {
+                    const sourceData = Array.isArray(series.data) ? series.data : [];
+                    if (!sourceData.length) continue;
+
+                    const numericValues = sourceData.map(d => toNumber(d && d.value !== undefined ? d.value : d));
+                    const total = numericValues.reduce((sum, n) => sum + (Number.isFinite(n) ? n : 0), 0);
+                    const isPie = series.type === 'pie' || (Array.isArray(series.radius) && series.radius.length > 0);
+
+                    const chartData = sourceData.map((d, idx) => {
+                        const raw = d && d.value !== undefined ? d.value : d;
+                        let rawValue = raw;
+                        if (raw && typeof raw === 'object' && !Array.isArray(raw)) rawValue = raw.value;
+                        const num = numericValues[idx];
+                        let value = String(rawValue ?? '').trim();
+                        if (isPie && total > 0 && Number.isFinite(num)) {
+                            value = formatPercent(num / total * 100);
+                        }
+                        return {
+                            name: String((d && (d.name || d[0])) || '').trim(),
+                            value,
+                            raw_value: String(rawValue ?? '').trim()
+                        };
+                    }).filter(item => item.name && item.value);
+
+                    if (chartData.length > 0) {
+                        results.push({
+                            title: closestTitle(dom),
+                            seriesType: series.type || '',
+                            data: chartData
+                        });
+                    }
+                }
+            } catch(e) {}
+        }
+
+            return JSON.stringify(results);
+        }""")
+        return json.loads(data) if isinstance(data, str) else []
+    except Exception as e:
+        print(f"    [WARN] ECharts实例提取失败: {e}")
+        return []
+
+
+def extract_echarts_canvas_data(page):
+    """从canvas对应的ECharts/ZRender容器提取环形图数据，补充实例扫描失败的场景"""
+    try:
+        data = page.evaluate("""() => {
+            const results = [];
+            if (!window.echarts || typeof window.echarts.getInstanceByDom !== 'function') {
+                return JSON.stringify(results);
+            }
+
+        function visible(el) {
+            if (!el) return false;
+            const rect = el.getBoundingClientRect();
+            const style = window.getComputedStyle(el);
+            return rect.width > 30 && rect.height > 30 && style.display !== 'none' && style.visibility !== 'hidden';
+        }
+
+        function toNumber(v) {
+            if (typeof v === 'number') return v;
+            if (Array.isArray(v)) return Number(v.find(x => !Number.isNaN(Number(x))) || 0);
+            if (v && typeof v === 'object') return Number(v.value || v[1] || 0);
+            return Number(String(v || '').replace(/,/g, '').replace('%', ''));
+        }
+
+        function formatPercent(num) {
+            if (!Number.isFinite(num)) return '';
+            return num.toFixed(2).replace(/\\.00$/, '').replace(/(\\.\\d)0$/, '$1') + '%';
+        }
+
+        const candidates = new Set();
+        document.querySelectorAll('canvas[data-zr-dom-id]').forEach(canvas => {
+            let p = canvas.parentElement;
+            for (let i = 0; p && i < 8; i++, p = p.parentElement) {
+                if (visible(p)) candidates.add(p);
+                if (p.getAttribute && p.getAttribute('_echarts_instance_')) break;
+            }
+        });
+
+        const seen = new Set();
+        for (const dom of candidates) {
+            let instance = null;
+            try { instance = window.echarts.getInstanceByDom(dom); } catch(e) {}
+            if (!instance || seen.has(instance.id)) continue;
+            seen.add(instance.id);
+
+            try {
+                const opt = instance.getOption();
+                const seriesList = Array.isArray(opt.series) ? opt.series : [];
+                for (const series of seriesList) {
+                    const rows = Array.isArray(series.data) ? series.data : [];
+                    if (!rows.length) continue;
+                    const nums = rows.map(d => toNumber(d && d.value !== undefined ? d.value : d));
+                    const total = nums.reduce((sum, n) => sum + (Number.isFinite(n) ? n : 0), 0);
+                    const isPie = series.type === 'pie' || (Array.isArray(series.radius) && series.radius.length > 0);
+                    const items = rows.map((d, idx) => {
+                        const raw = d && d.value !== undefined ? d.value : d;
+                        const num = nums[idx];
+                        return {
+                            name: String((d && (d.name || d[0])) || '').trim(),
+                            value: isPie && total > 0 && Number.isFinite(num) ? formatPercent(num / total * 100) : String(raw ?? '').trim(),
+                            raw_value: String(raw ?? '').trim()
+                        };
+                    }).filter(item => item.name && item.value);
+                    if (items.length) results.push({seriesType: series.type || '', data: items});
+                }
+            } catch(e) {}
+        }
+
+            return JSON.stringify(results);
+        }""")
+        return json.loads(data) if isinstance(data, str) else []
+    except Exception as e:
+        print(f"    [WARN] Canvas环形图提取失败: {e}")
+        return []
+
+
+def extract_echarts_by_hover(page):
+    """通过模拟鼠标悬停环形图圆环，读取tooltip数据（备选策略）"""
+    charts = page.locator('canvas[data-zr-dom-id], .jmtc-container').all()[:6]
+    all_data = {"性别": [], "年龄": [], "职业": []}
+    for idx, chart in enumerate(charts):
+        try:
+            box = chart.bounding_box()
+            if not box or box['width'] < 50 or box['height'] < 50:
+                continue
+            seen = set()
+            chart_items = []
+            cx = box['x'] + box['width'] / 2
+            cy = box['y'] + box['height'] / 2
+            base_r = min(box['width'], box['height']) * 0.36
+            # 空心环不能悬停中心点，沿圆环两圈采样以触发 tooltip
+            hover_points = []
+            for radius_scale in [1.0, 0.78]:
+                radius = base_r * radius_scale
+                for angle in [0, 30, 60, 90, 120, 150, 180, 210, 240, 270, 300, 330]:
+                    rad = angle * 3.1415926 / 180
+                    hover_points.append((cx + radius * math.cos(rad), cy + radius * math.sin(rad)))
+
+            for x, y in hover_points:
+                page.mouse.move(x, y)
+                page.wait_for_timeout(180)
+                tooltip_result = page.evaluate("""() => {
+                    const items = [];
+                    function addPair(name, value) {
+                        name = String(name || '').replace(/[：:\\s]+$/g, '').trim();
+                        value = String(value || '').trim();
+                        if (!name || !value) return;
+                        if (!/%|\\d/.test(value)) return;
+                        items.push({name, value});
+                    }
+                    document.querySelectorAll('.sz-frontend-bc, .echarts-tooltip, [class*="tooltip"]').forEach(el => {
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        if (style.visibility === 'hidden' || style.display === 'none' || rect.width === 0 || rect.height === 0) return;
+                        const nameEl = el.querySelector('.name');
+                        const valueEl = el.querySelector('.value');
+                        if (nameEl && valueEl) {
+                            addPair(nameEl.textContent, valueEl.textContent);
+                            return;
+                        }
+                        const text = el.textContent.replace(/\\u00a0/g, ' ').trim();
+                        const lines = text.split(/\\n/).map(s => s.trim()).filter(Boolean);
+                        for (const line of lines) {
+                            let m = line.match(/^(.+?)[：:]\\s*(\\d+(?:\\.\\d+)?%?.*)$/);
+                            if (m) {
+                                addPair(m[1], m[2]);
+                                continue;
+                            }
+                            m = line.match(/^([^\\d%：:]{1,30})\\s+(\\d+(?:\\.\\d+)?%.*)$/);
+                            if (m) addPair(m[1], m[2]);
+                        }
+                    });
+                    return JSON.stringify(items);
+                }""")
+                items = json.loads(tooltip_result) if isinstance(tooltip_result, str) else []
+                for item in items:
+                    key = f"{item['name']}:{item['value']}"
+                    if key not in seen:
+                        seen.add(key)
+                        chart_items.append(item)
+            # 判断图表类型
+            if chart_items:
+                names_str = ' '.join([d['name'] for d in chart_items])
+                if any(n in names_str for n in ['男', '女']):
+                    all_data["性别"] = chart_items
+                elif '岁' in names_str:
+                    all_data["年龄"] = chart_items
+                else:
+                    all_data["职业"] = chart_items
+        except Exception as e:
+            print(f"    [WARN] 图表{idx}悬停提取失败: {e}")
+    # 鼠标移回安全区域
+    page.mouse.move(0, 0)
+    return all_data
+
+
+def extract_crowd_charts_by_card(page):
+    """按卡片标题定向读取人群画像图表，避免性别/年龄/职业串数据"""
+    try:
+        data = page.evaluate("""() => {
+            const result = { "性别": [], "年龄": [], "职业": [] };
+
+            function visible(el) {
+                if (!el) return false;
+                const rect = el.getBoundingClientRect();
+                const style = window.getComputedStyle(el);
+                return rect.width > 30 && rect.height > 30 && style.display !== 'none' && style.visibility !== 'hidden';
+            }
+
+            function toNumber(v) {
+                if (typeof v === 'number') return v;
+                if (Array.isArray(v)) return Number(v.find(x => !Number.isNaN(Number(x))) || 0);
+                if (v && typeof v === 'object') return Number(v.value || v[1] || 0);
+                return Number(String(v || '').replace(/,/g, '').replace('%', ''));
+            }
+
+            function formatPercent(num) {
+                if (!Number.isFinite(num)) return '';
+                return num.toFixed(2).replace(/\\.00$/, '').replace(/(\\.\\d)0$/, '$1') + '%';
+            }
+
+            function findInstance(dom) {
+                if (!dom) return null;
+                if (window.echarts && typeof window.echarts.getInstanceByDom === 'function') {
+                    try {
+                        const instance = window.echarts.getInstanceByDom(dom);
+                        if (instance) return instance;
+                    } catch(e) {}
+                }
+                for (const key of Object.keys(dom)) {
+                    try {
+                        const val = dom[key];
+                        if (val && typeof val.getOption === 'function') return val;
+                    } catch(e) {}
+                }
+                return null;
+            }
+
+            function findChartBlock(title) {
+                const aliases = title === '职业' ? ['职业分布', '职业'] : [title];
+                const titleEls = Array.from(document.querySelectorAll('.jmtd-typography-text, .jmtd-typography-title, span, h4'))
+                    .filter(el => aliases.some(alias => el.textContent.trim() === alias));
+                for (const titleEl of titleEls) {
+                    const block = titleEl.closest('.jmtd-grid-col-12, .jmtd-grid-col, .jmtd-card-section, .jmtd-card');
+                    if (block && block.querySelector('canvas, .jmtc-container, [_echarts_instance_]')) return block;
+                }
+                return null;
+            }
+
+            function chartDomsInCard(card) {
+                const set = new Set();
+                if (!card) return [];
+                card.querySelectorAll('.jmtc-container, [_echarts_instance_]').forEach(el => {
+                    if (visible(el)) set.add(el);
+                });
+                card.querySelectorAll('canvas[data-zr-dom-id], canvas').forEach(canvas => {
+                    let p = canvas.parentElement;
+                    for (let i = 0; p && i < 6 && card.contains(p); i++, p = p.parentElement) {
+                        if (visible(p)) set.add(p);
+                        if (findInstance(p)) break;
+                    }
+                });
+                return [...set];
+            }
+
+            function normalizeSeries(opt, title) {
+                const seriesList = Array.isArray(opt.series) ? opt.series : [];
+                if (!seriesList.length) return [];
+                const wantPie = title === '性别' || title === '职业';
+                const wantBar = title === '年龄';
+                let series = seriesList.find(s => wantPie && (s.type === 'pie' || Array.isArray(s.radius)));
+                if (!series) series = seriesList.find(s => wantBar && s.type === 'bar');
+                if (!series) series = seriesList.find(s => Array.isArray(s.data) && s.data.length);
+                if (!series || !Array.isArray(series.data)) return [];
+
+                const xData = Array.isArray(opt.xAxis) && opt.xAxis[0] && Array.isArray(opt.xAxis[0].data) ? opt.xAxis[0].data : [];
+                const nums = series.data.map(d => toNumber(d && d.value !== undefined ? d.value : d));
+                const total = nums.reduce((sum, n) => sum + (Number.isFinite(n) ? n : 0), 0);
+                const isPie = series.type === 'pie' || Array.isArray(series.radius);
+                const smallRatio = !isPie && nums.length && nums.every(n => Number.isFinite(n) && n >= 0 && n <= 1);
+
+                return series.data.map((d, idx) => {
+                    const raw = d && d.value !== undefined ? d.value : d;
+                    const name = String((d && (d.name || d[0])) || xData[idx] || '').trim();
+                    const num = nums[idx];
+                    let value = String(raw ?? '').trim();
+                    if (isPie && total > 0 && Number.isFinite(num)) {
+                        value = formatPercent(num / total * 100);
+                    } else if (smallRatio && Number.isFinite(num)) {
+                        value = formatPercent(num * 100);
+                    }
+                    return { name, value, raw_value: String(raw ?? '').trim() };
+                }).filter(item => item.name && item.value);
+            }
+
+            for (const title of ['性别', '年龄', '职业']) {
+                const card = findChartBlock(title);
+                if (!card) continue;
+                const seen = new Set();
+                for (const dom of chartDomsInCard(card)) {
+                    const instance = findInstance(dom);
+                    if (!instance) continue;
+                    const key = instance.id || dom.getAttribute('_echarts_instance_') || '';
+                    if (key && seen.has(key)) continue;
+                    if (key) seen.add(key);
+                    try {
+                        const items = normalizeSeries(instance.getOption(), title);
+                        if (items.length) {
+                            result[title] = items;
+                            break;
+                        }
+                    } catch(e) {}
+                }
+            }
+
+            return JSON.stringify(result);
+        }""")
+        return json.loads(data) if isinstance(data, str) else {"性别": [], "年龄": [], "职业": []}
+    except Exception as e:
+        print(f"    [WARN] 按卡片读取图表失败: {e}")
+        return {"性别": [], "年龄": [], "职业": []}
+
+
+def extract_crowd_chart_by_hover(page, title, chart_type="ring"):
+    """只在指定标题卡片内悬停采集，避免不同图表数据混淆"""
+    try:
+        box = page.evaluate("""(title) => {
+            function visible(el) {
+                if (!el) return false;
+                const rect = el.getBoundingClientRect();
+                const style = window.getComputedStyle(el);
+                return rect.width > 30 && rect.height > 30 && style.display !== 'none' && style.visibility !== 'hidden';
+            }
+            const aliases = title === '职业' ? ['职业分布', '职业'] : [title];
+            const titleEl = Array.from(document.querySelectorAll('.jmtd-typography-text, .jmtd-typography-title, span, h4'))
+                .find(el => aliases.some(alias => el.textContent.trim() === alias));
+            if (!titleEl) return null;
+            const block = titleEl.closest('.jmtd-grid-col-12, .jmtd-grid-col, .jmtd-card-section, .jmtd-card');
+            if (!block) return null;
+            const chart = Array.from(block.querySelectorAll('canvas[data-zr-dom-id], canvas, .jmtc-container')).find(visible);
+            if (!chart) return null;
+            const rect = chart.getBoundingClientRect();
+            return {x: rect.x, y: rect.y, width: rect.width, height: rect.height};
+        }""", title)
+        if not box or box['width'] < 50 or box['height'] < 50:
+            return []
+
+        points = []
+        if chart_type == "ring":
+            cx = box['x'] + box['width'] / 2
+            cy = box['y'] + box['height'] / 2
+            base_r = min(box['width'], box['height']) * 0.36
+            for radius_scale in [1.0, 0.78]:
+                radius = base_r * radius_scale
+                for angle in [0, 30, 60, 90, 120, 150, 180, 210, 240, 270, 300, 330]:
+                    rad = angle * math.pi / 180
+                    points.append((cx + radius * math.cos(rad), cy + radius * math.sin(rad)))
+        else:
+            for y_scale in [0.2, 0.35, 0.5, 0.65, 0.8]:
+                y = box['y'] + box['height'] * y_scale
+                for i in range(10):
+                    points.append((box['x'] + box['width'] * (i + 0.5) / 10, y))
+
+        seen = set()
+        items = []
+        def accept_tooltip_item(item):
+            name = (item.get('name') or '').strip()
+            if title == "性别":
+                return name in ("男", "女", "男性", "女性")
+            if title == "年龄":
+                return "岁" in name
+            if title == "职业":
+                return name not in ("男", "女", "男性", "女性") and "岁" not in name
+            return True
+
+        for x, y in points:
+            page.mouse.move(x, y)
+            page.wait_for_timeout(160)
+            tooltip_result = page.evaluate("""() => {
+                const items = [];
+                function addPair(name, value) {
+                    name = String(name || '').replace(/[：:\\s]+$/g, '').trim();
+                    value = String(value || '').trim();
+                    if (!name || !value || !/%|\\d/.test(value)) return;
+                    items.push({name, value});
+                }
+                document.querySelectorAll('.sz-frontend-bc, .echarts-tooltip, [class*="tooltip"]').forEach(el => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    if (style.visibility === 'hidden' || style.display === 'none' || rect.width === 0 || rect.height === 0) return;
+                    const text = el.textContent.replace(/\\u00a0/g, ' ').trim();
+                    const lines = text.split(/\\n/).map(s => s.trim()).filter(Boolean);
+                    for (const line of lines) {
+                        let m = line.match(/^(.+?)[：:]\\s*(\\d+(?:\\.\\d+)?%?.*)$/);
+                        if (!m) m = line.match(/^([^\\d%：:]{1,30})\\s+(\\d+(?:\\.\\d+)?%.*)$/);
+                        if (m) addPair(m[1], m[2]);
+                    }
+                });
+                return JSON.stringify(items);
+            }""")
+            tooltip_items = json.loads(tooltip_result) if isinstance(tooltip_result, str) else []
+            for item in tooltip_items:
+                if not accept_tooltip_item(item):
+                    continue
+                key = f"{item.get('name', '')}:{item.get('value', '')}"
+                if key not in seen:
+                    seen.add(key)
+                    items.append(item)
+        page.mouse.move(0, 0)
+        return items
+    except Exception as e:
+        print(f"    [WARN] {title}图表悬停提取失败: {e}")
+        return []
+
+
+def extract_crowd_tooltip_text_by_block(page, title):
+    """读取指定画像块内已渲染的tooltip文本（即使tooltip处于隐藏状态）"""
+    try:
+        data = page.evaluate("""(title) => {
+            const aliases = title === '职业' ? ['职业分布', '职业'] : [title];
+            const titleEl = Array.from(document.querySelectorAll('.jmtd-typography-text, .jmtd-typography-title, span, h4'))
+                .find(el => aliases.some(alias => el.textContent.trim() === alias));
+            if (!titleEl) return JSON.stringify([]);
+            const block = titleEl.closest('.jmtd-grid-col-12, .jmtd-grid-col, .jmtd-card-section, .jmtd-card');
+            if (!block) return JSON.stringify([]);
+
+            const result = [];
+            function cleanName(text) {
+                return String(text || '').replace(/\\u00a0/g, ' ').replace(/[：:\\s]+$/g, '').trim();
+            }
+            function cleanValue(text) {
+                return String(text || '').replace(/\\u00a0/g, ' ').trim();
+            }
+            block.querySelectorAll('.sz-frontend-bc').forEach(el => {
+                const name = cleanName(el.querySelector('.name')?.textContent || '');
+                const value = cleanValue(el.querySelector('.value')?.textContent || '');
+                if (name && value && /%|\\d/.test(value)) {
+                    result.push({name, value});
+                }
+            });
+
+            return JSON.stringify([...new Map(result.map(r => [r.name + ':' + r.value, r])).values()]);
+        }""", title)
+        return json.loads(data) if isinstance(data, str) else []
+    except Exception as e:
+        print(f"    [WARN] {title}隐藏tooltip提取失败: {e}")
+        return []
+
+
+def extract_age_distribution_data(page):
+    """单独提取年龄柱状图：优先按x轴/数据名包含'岁'定位，保持原始年龄段和值"""
+    try:
+        data = page.evaluate("""() => {
+            const results = [];
+
+            function visible(el) {
+                if (!el) return false;
+                const rect = el.getBoundingClientRect();
+                const style = window.getComputedStyle(el);
+                return rect.width > 30 && rect.height > 30 && style.display !== 'none' && style.visibility !== 'hidden';
+            }
+
+            function findInstance(dom) {
+                if (!dom) return null;
+                if (window.echarts && typeof window.echarts.getInstanceByDom === 'function') {
+                    try {
+                        const instance = window.echarts.getInstanceByDom(dom);
+                        if (instance) return instance;
+                    } catch(e) {}
+                }
+                for (const key of Object.keys(dom)) {
+                    try {
+                        const val = dom[key];
+                        if (val && typeof val.getOption === 'function') return val;
+                    } catch(e) {}
+                }
+                return null;
+            }
+
+            function xAxisData(opt) {
+                const xAxis = Array.isArray(opt.xAxis) ? opt.xAxis[0] : opt.xAxis;
+                if (xAxis && Array.isArray(xAxis.data)) return xAxis.data;
+                return [];
+            }
+
+            function rawValue(d) {
+                if (d && typeof d === 'object' && !Array.isArray(d) && d.value !== undefined) return d.value;
+                if (Array.isArray(d)) return d.find(x => !Number.isNaN(Number(x))) ?? d[1] ?? d[0];
+                return d;
+            }
+
+            function formatValue(v) {
+                const s = String(v ?? '').trim();
+                if (!s) return '';
+                if (s.includes('%')) return s;
+                const n = Number(s.replace(/,/g, ''));
+                if (Number.isFinite(n) && n >= 0 && n <= 1) {
+                    return (n * 100).toFixed(2).replace(/\\.00$/, '').replace(/(\\.\\d)0$/, '$1') + '%';
+                }
+                if (Number.isFinite(n)) {
+                    return n.toFixed(2).replace(/\\.00$/, '').replace(/(\\.\\d)0$/, '$1') + '%';
+                }
+                return s;
+            }
+
+            const candidates = new Set();
+            const titleEl = Array.from(document.querySelectorAll('.jmtd-typography-text, .jmtd-typography-title, span, h4'))
+                .find(el => el.textContent.trim() === '年龄');
+            const ageBlock = titleEl ? titleEl.closest('.jmtd-grid-col-12, .jmtd-grid-col, .jmtd-card-section, .jmtd-card') : null;
+            const root = ageBlock || document;
+            root.querySelectorAll('.jmtc-container, [_echarts_instance_], canvas[data-zr-dom-id], canvas').forEach(el => {
+                let p = el;
+                for (let i = 0; p && i < 6 && root.contains(p); i++, p = p.parentElement) {
+                    if (visible(p)) candidates.add(p);
+                    if (findInstance(p)) break;
+                }
+            });
+
+            if (ageBlock) {
+                ageBlock.querySelectorAll('.sz-frontend-bc').forEach(el => {
+                    const name = el.querySelector('.name')?.textContent?.replace(/[：:\\s]+/g, '').trim() || '';
+                    const value = el.querySelector('.value')?.textContent?.trim() || '';
+                    if (name.includes('岁') && value) {
+                        results.push([{name, value: formatValue(value), raw_value: value}]);
+                    }
+                });
+            }
+
+            const seen = new Set();
+            for (const dom of candidates) {
+                const instance = findInstance(dom);
+                const key = instance && (instance.id || dom.getAttribute('_echarts_instance_'));
+                if (!instance || (key && seen.has(key))) continue;
+                if (key) seen.add(key);
+
+                try {
+                    const opt = instance.getOption();
+                    const xData = xAxisData(opt).map(x => String(x ?? '').trim());
+                    const seriesList = Array.isArray(opt.series) ? opt.series : [];
+                    for (const series of seriesList) {
+                        if (!Array.isArray(series.data) || !series.data.length) continue;
+                        const labels = series.data.map((d, idx) => String((d && (d.name || d[0])) || xData[idx] || '').trim());
+                        const labelText = labels.join(' ');
+                        if (!labelText.includes('岁')) continue;
+                        const items = series.data.map((d, idx) => {
+                            const name = labels[idx];
+                            const rv = rawValue(d);
+                            return {name, value: formatValue(rv), raw_value: String(rv ?? '').trim()};
+                        }).filter(item => item.name && item.value);
+                        if (items.length) results.push(items);
+                    }
+                } catch(e) {}
+            }
+
+            results.sort((a, b) => b.length - a.length);
+            return JSON.stringify(results[0] || []);
+        }""")
+        return json.loads(data) if isinstance(data, str) else []
+    except Exception as e:
+        print(f"    [WARN] 年龄分布柱状图提取失败: {e}")
+        return []
+
+
+def extract_career_text_data(page):
+    """从职业分布区域直接提取文本数据（职业卡片内的文本列表）"""
+    data = page.evaluate("""() => {
+        const result = [];
+        // 1. 先找到包含"职业"的卡片
+        let careerCard = null;
+        const cards = document.querySelectorAll('.jmtd-card');
+        for (const card of cards) {
+            const header = card.querySelector('.jmtd-card-header');
+            if (header && header.textContent.includes('职业')) {
+                careerCard = card;
+                break;
+            }
+        }
+        if (!careerCard) {
+            // 兜底：找包含"职业"标题的最近父容器
+            const titles = document.querySelectorAll('.jmtd-typography-title, .jmtd-label, .jmtp-basic-summary-label');
+            for (const t of titles) {
+                if (t.textContent.includes('职业')) {
+                    careerCard = t.closest('.jmtd-card, .jmtd-col, .jmtd-row-item') || t.parentElement;
+                    break;
+                }
+            }
+        }
+        if (careerCard) {
+            // 在职业卡片内搜索 "名称 数值%" 或 "名称：数值%" 格式的文本
+            careerCard.querySelectorAll('.jmtd-typography-text, span, .jmtd-label, .jmtp-basic-summary-label, p').forEach(el => {
+                const text = el.textContent.trim();
+                // 匹配格式：职业名 + 空格/冒号 + 百分比数字
+                const m = text.match(/^([^\\d%:：\\s]{2,20})[:：\\s]+(\\d+(?:\\.\\d+)?%)$/);
+                if (m) {
+                    result.push({name: m[1].trim(), value: m[2].trim()});
+                }
+            });
+        }
+        // 2. 如果卡片内没找到，全局搜索职业相关文本（兜底）
+        if (result.length === 0) {
+            document.querySelectorAll('.jmtd-typography-text, .jmtd-label').forEach(el => {
+                const text = el.textContent.trim();
+                // 更宽泛的匹配：允许前面有一些前缀词
+                const m = text.match(/(?:职业|行业|身份)[:：\\s]*([^\\d%:：\\s]{2,20})[:：\\s]*(\\d+(?:\\.\\d+)?%)/);
+                if (m) {
+                    result.push({name: m[1].trim(), value: m[2].trim()});
+                }
+            });
+        }
+        // 去重
+        return JSON.stringify([...new Map(result.map(r => [r.name + ':' + r.value, r])).values()]);
+    }""")
+    try:
+        return json.loads(data) if isinstance(data, str) else []
+    except:
+        return []
+
+
+def extract_crowd_portrait_data(page):
+    """采集人群画像数据（性别/年龄/职业/地区）"""
+    print("  采集人群画像数据...")
+    crowd_data = {
+        "性别": [],
+        "年龄": [],
+        "职业": [],
+        "地区_省份": [],
+        "地区_城市": [],
+    }
+    page.wait_for_timeout(3000)
+
+    def filter_crowd_items(kind, items):
+        filtered = []
+        for item in items or []:
+            name = str(item.get("name", "")).strip()
+            value = str(item.get("value", "")).strip()
+            if not name or not value:
+                continue
+            if kind == "性别" and name not in ("男", "女", "男性", "女性"):
+                continue
+            if kind == "年龄" and "岁" not in name:
+                continue
+            if kind == "职业" and (name in ("男", "女", "男性", "女性") or "岁" in name):
+                continue
+            filtered.append(item)
+        return filtered
+
+    # ── 策略1: 按卡片标题精确读取图表，避免性别/年龄/职业数据混在一起 ──
+    card_chart_data = extract_crowd_charts_by_card(page)
+    crowd_data["性别"] = filter_crowd_items("性别", card_chart_data.get("性别", []))
+    crowd_data["职业"] = filter_crowd_items("职业", card_chart_data.get("职业", []))
+
+    # 年龄柱状图单独按“岁”标签提取，恢复原先稳定的年龄段逻辑
+    crowd_data["年龄"] = filter_crowd_items("年龄", extract_age_distribution_data(page))
+    if not crowd_data["年龄"]:
+        crowd_data["年龄"] = filter_crowd_items("年龄", card_chart_data.get("年龄", []))
+
+    print(
+        f"    [INFO] 卡片图表读取: 性别{len(crowd_data['性别'])}条, "
+        f"年龄{len(crowd_data['年龄'])}条, 职业{len(crowd_data['职业'])}条"
+    )
+
+    # ── 策略2: 指定卡片内悬停补采，性别/职业是环形图，年龄是柱状图 ──
+    if not crowd_data["性别"]:
+        print("    [INFO] 性别分布环形图悬停补采...")
+        crowd_data["性别"] = filter_crowd_items("性别", extract_crowd_chart_by_hover(page, "性别", "ring"))
+    if not crowd_data["年龄"]:
+        print("    [INFO] 年龄分布柱状图悬停补采...")
+        crowd_data["年龄"] = filter_crowd_items("年龄", extract_crowd_chart_by_hover(page, "年龄", "bar"))
+    if not crowd_data["职业"]:
+        print("    [INFO] 职业分布隐藏tooltip补采...")
+        crowd_data["职业"] = filter_crowd_items("职业", extract_crowd_tooltip_text_by_block(page, "职业"))
+    if len(crowd_data["职业"]) < 2:
+        print("    [INFO] 职业分布环形图悬停补采...")
+        hover_career = filter_crowd_items("职业", extract_crowd_chart_by_hover(page, "职业", "ring"))
+        merged = {}
+        for item in crowd_data["职业"] + hover_career:
+            merged[f"{item.get('name', '')}:{item.get('value', '')}"] = item
+        crowd_data["职业"] = list(merged.values())
+
+    # ── 策略3: 职业分布文本直接提取（最后兜底）──
+    if not crowd_data["职业"]:
+        career_text_data = extract_career_text_data(page)
+        if career_text_data:
+            print(f"    [INFO] 职业文本直接提取到 {len(career_text_data)} 条")
+            crowd_data["职业"] = filter_crowd_items("职业", career_text_data)
+
+    # ── 策略4: 文本结论兜底 ──
+    if not crowd_data["性别"]:
+        gender_text = extract_crowd_text_fallback(page, ["男性客户偏多", "女性客户偏多", "男客户偏多", "女客户偏多"])
+        if gender_text:
+            crowd_data["性别"] = [{"name": "结论", "value": gender_text}]
+    if not crowd_data["年龄"]:
+        age_text = extract_crowd_text_fallback(page, ["岁", "年龄"])
+        if age_text and "岁" in age_text and len(age_text) < 80:
+            crowd_data["年龄"] = [{"name": "结论", "value": age_text}]
+    if not crowd_data["职业"]:
+        occ_text = extract_crowd_text_fallback(page, ["主力军是", "购买的主力军", "职业"])
+        if occ_text and len(occ_text) < 80:
+            crowd_data["职业"] = [{"name": "结论", "value": occ_text}]
+
+    # ── 地区-省份 ──
+    crowd_data["地区_省份"] = extract_region_table_data(page)
+
+    # ── 地区-城市 ──
+    click_city_radio(page)
+    crowd_data["地区_城市"] = extract_region_table_data(page)
+
+    print(f"    [OK] 人群画像采集完成: 性别{len(crowd_data['性别'])}条, 年龄{len(crowd_data['年龄'])}条, 职业{len(crowd_data['职业'])}条, 省份{len(crowd_data['地区_省份'])}条, 城市{len(crowd_data['地区_城市'])}条")
+    return crowd_data
+
+
+# ── 新页面：选择上周日期 ───────────────────────────────────────────────────────
+
+def select_last_week_on_detail_page(page):
+    """在单品详情页面选择昨天（天模式）。函数名保留，避免改动调用链。"""
+    print("  选择昨天时间范围...")
+    select_yesterday_in_combo_datepicker(page, "product_detail")
+    page.wait_for_timeout(3000)
+
+
+# ── 新页面：采集核心数据 ───────────────────────────────────────────────────────
+
+def extract_core_data(page):
+    """在核心数据 tab 采集指标（含比较数据）"""
+    print("  采集核心数据...")
+    page.wait_for_timeout(3000)
+
+    data = extract(page, """() => {
+        const result = {};
+        const groups = document.querySelectorAll('.jmtp-basic-summary-group');
+
+        for (const group of groups) {
+            // 分组名称
+            const labelEl = group.querySelector('.jmtp-basic-summary-label');
+            const groupName = labelEl ? labelEl.textContent.trim() : '其他';
+            const groupData = {};
+
+            // 提取该分组下的所有卡片
+            const cards = group.querySelectorAll('.jmtp-basic-summary-card');
+            for (const card of cards) {
+                const titleEl = card.querySelector('.jmtp-summary-card-title .jmtd-label');
+                const title = titleEl ? titleEl.textContent.trim() : '';
+                if (!title) continue;
+
+                const valueEl = card.querySelector('.summary-index');
+                const value = valueEl ? valueEl.textContent.trim() : '';
+
+                const descLeft = card.querySelector('.jmtp-summary-card-description-left');
+                const compareLabel = descLeft ? descLeft.textContent.trim() : '';
+
+                const descRight = card.querySelector('.jmtp-summary-card-description-right');
+                const compareValue = descRight ? descRight.textContent.trim() : '';
+
+                // 判断涨跌趋势
+                let trend = '';
+                const arrowGreen = card.querySelector('.jmtp-summary-card-arrow-green');
+                const arrowRed = card.querySelector('.jmtp-summary-card-arrow-red');
+                if (arrowGreen) trend = 'up';
+                else if (arrowRed) trend = 'down';
+
+                groupData[title] = {
+                    value: value,
+                    compare_label: compareLabel,
+                    compare_value: compareValue,
+                    trend: trend,
+                };
+            }
+
+            if (Object.keys(groupData).length > 0) {
+                result[groupName] = groupData;
+            }
+        }
+
+        return JSON.stringify(result);
+    }""", "{}")
+
+    try:
+        result = json.loads(data) if isinstance(data, str) else {}
+    except:
+        result = {}
+
+    # 打印关键指标
+    total_cards = sum(len(v) for v in result.values() if isinstance(v, dict))
+    print(f"    共提取 {len(result)} 个分组, {total_cards} 个指标")
+    for group_name, items in result.items():
+        print(f"    [{group_name}]")
+        for title, v in items.items():
+            print(f"      {title}: {v.get('value','')} ({v.get('compare_label','')} {v.get('compare_value','')})")
+    return result
+
+
+# ── 商品评价差评采集 ───────────────────────────────────────────────────────────
+
+def navigate_to_goods_appraise(page):
+    """从商家后台首页导航到商品评价页面"""
+    print("[Step] 导航到商品评价页面...")
+    if 'shop.jd.com' not in page.url:
+        page.goto("https://shop.jd.com/jdm/home", wait_until="networkidle")
+        page.wait_for_timeout(5000)
+    try:
+        goods_menu = page.locator('.shop-menu-sidebar__menu-first-item:has-text("商品")').first
+        goods_menu.wait_for(state="visible", timeout=10000)
+        goods_menu.hover()
+        print("  [OK] 已悬停'商品'菜单")
+    except Exception as e:
+        print(f"  [WARN] 悬停商品菜单失败: {e}")
+        page.goto("https://wares-jdm.jd.com/goodsAppraise", wait_until="networkidle")
+        page.wait_for_timeout(5000)
+        return page
+    page.wait_for_timeout(2000)
+    try:
+        appraise_link = page.locator('.shop-menu-sidebar__hover-item-menu:has-text("商品评价")').first
+        appraise_link.wait_for(state="visible", timeout=5000)
+        appraise_link.click()
+        print("  [OK] 已点击'商品评价'")
+    except Exception as e:
+        print(f"  [WARN] 点击商品评价失败: {e}")
+        page.goto("https://wares-jdm.jd.com/goodsAppraise", wait_until="networkidle")
+        page.wait_for_timeout(5000)
+        return page
+    page.wait_for_timeout(5000)
+    pages = page.context.pages
+    for p in pages:
+        if 'goodsAppraise' in p.url or 'wares-jdm' in p.url:
+            print(f"  [OK] 已到达商品评价页面")
+            return p
+    print(f"  [OK] 当前页面: {page.url}")
+    return page
+
+
+def click_rating_filter(page, rating):
+    """点击指定评级筛选 (好评/中评/差评)"""
+    print(f"  选择{rating}筛选...")
+    try:
+        tag = page.locator(f'.filter-button-group__options .tag:has-text("{rating}")').first
+        tag.wait_for(state="visible", timeout=5000)
+        tag.click()
+        print(f"  [OK] 已点击{rating}")
+    except Exception as e:
+        print(f"  [WARN] 点击{rating}失败: {e}")
+        page.evaluate(f"""() => {{
+            const tags = document.querySelectorAll('.filter-button-group__options .tag');
+            for (const tag of tags) {{
+                if (tag.textContent.includes('{rating}')) {{ tag.click(); return 'clicked'; }}
+            }}
+            return 'not_found';
+        }}""")
+    page.wait_for_timeout(1500)
+
+
+def select_last_week_date_range_appraise(page):
+    """在商品评价页面选择昨天日期范围（直接填写输入框，避免面板月份错位）"""
+    print("  设置评价时间为昨天...")
+    last_monday = get_yesterday_date()
+    last_sunday = last_monday
+    expected_start = last_monday.strftime('%Y-%m-%d')
+    expected_end = last_sunday.strftime('%Y-%m-%d')
+    input_selector = '.rcd-date-editor--datetimerange .rcd-range-input, .rcd-date-editor--daterange .rcd-range-input'
+    editor_selector = '.rcd-date-editor--datetimerange, .rcd-date-editor--daterange'
+    expected_start_value = f'{expected_start} 00:00:00'
+    expected_end_value = f'{expected_end} 23:59:59'
+
+    try:
+        direct_set = page.evaluate(
+            r"""({start, end, inputSelector}) => {
+                const inputs = Array.from(document.querySelectorAll(inputSelector));
+                if (inputs.length < 2) return { ok: false, reason: 'inputs_not_found' };
+                const setValue = (input, value) => {
+                    input.focus();
+                    const setter = Object.getOwnPropertyDescriptor(HTMLInputElement.prototype, 'value')?.set;
+                    if (setter) setter.call(input, value);
+                    else input.value = value;
+                    input.dispatchEvent(new Event('input', { bubbles: true }));
+                    input.dispatchEvent(new Event('change', { bubbles: true }));
+                    input.dispatchEvent(new KeyboardEvent('keydown', { key: 'Enter', bubbles: true, cancelable: true }));
+                    input.dispatchEvent(new KeyboardEvent('keyup', { key: 'Enter', bubbles: true, cancelable: true }));
+                    input.blur();
+                };
+                setValue(inputs[0], start);
+                setValue(inputs[1], end);
+                return { ok: inputs[0].value === start && inputs[1].value === end, start: inputs[0].value, end: inputs[1].value };
+            }""",
+            {"start": expected_start_value, "end": expected_end_value, "inputSelector": input_selector},
+        )
+        page.keyboard.press("Enter")
+        page.wait_for_timeout(1200)
+        inputs = page.locator(input_selector).all()
+        if len(inputs) >= 2:
+            actual_start = inputs[0].get_attribute('value') or ''
+            actual_end = inputs[1].get_attribute('value') or ''
+            if actual_start.startswith(expected_start) and actual_end.startswith(expected_end):
+                print(f"    [OK] appraise date set by native events: {actual_start} ~ {actual_end}")
+                return
+        print(f"    [WARN] native event date set did not stick: {direct_set}")
+    except Exception as e:
+        print(f"    [WARN] native event date set failed: {e}")
+
+    # 检查当前是否已是目标日期
+    try:
+        inputs = page.locator(input_selector).all()
+        if len(inputs) >= 2:
+            current_start = inputs[0].get_attribute('value') or ''
+            current_end = inputs[1].get_attribute('value') or ''
+            if current_start.startswith(expected_start) and current_end.startswith(expected_end):
+                print(f"    [OK] 日期已是昨天: {current_start} ~ {current_end}")
+                return
+    except Exception:
+        pass
+
+    # 策略1: 直接填写输入框（最可靠）
+    try:
+        inputs = page.locator(input_selector).all()
+        if len(inputs) >= 2:
+            # 先点击一下输入框确保焦点
+            inputs[0].click()
+            page.wait_for_timeout(300)
+            # 清空并填写开始日期
+            inputs[0].fill("")
+            page.wait_for_timeout(200)
+            inputs[0].fill(expected_start_value)
+            page.wait_for_timeout(300)
+            # 填写结束日期
+            inputs[1].fill("")
+            page.wait_for_timeout(200)
+            inputs[1].fill(expected_end_value)
+            page.wait_for_timeout(300)
+            # 按 Enter 确认
+            page.keyboard.press("Enter")
+            page.wait_for_timeout(1000)
+
+            # 验证
+            inputs = page.locator(input_selector).all()
+            if len(inputs) >= 2:
+                actual_start = inputs[0].get_attribute('value') or ''
+                actual_end = inputs[1].get_attribute('value') or ''
+                if actual_start.startswith(expected_start) and actual_end.startswith(expected_end):
+                    print(f"    [OK] 日期已设置: {actual_start} ~ {actual_end}")
+                    return
+                else:
+                    print(f"    [WARN] 直接填写结果不符: {actual_start} ~ {actual_end}，尝试面板点击...")
+        else:
+            print("    [WARN] 未找到日期输入框，尝试面板点击...")
+    except Exception as e:
+        print(f"    [WARN] 直接填写失败: {e}，尝试面板点击...")
+
+    # 策略2: 面板点击（兜底）
+    date_editor = page.locator(editor_selector).first
+    date_editor.click()
+    page.wait_for_timeout(1500)
+    popper = page.locator('.rcd-picker__popper').first
+    try:
+        popper.wait_for(state="visible", timeout=5000)
+    except:
+        print("    [WARN] 日历弹窗未出现")
+        return
+
+    # 获取左面板当前月份
+    try:
+        left_header = popper.locator('.rcd-date-range-picker__content.is-left .rcd-date-range-picker__header div').first
+        left_month_text = left_header.inner_text()
+        m = re.search(r'(\d+)\s*年\s*(\d+)\s*月', left_month_text)
+        if m:
+            current_left_year, current_left_month = int(m.group(1)), int(m.group(2))
+        else:
+            current_left_year, current_left_month = datetime.now().year, datetime.now().month
+    except Exception:
+        current_left_year, current_left_month = datetime.now().year, datetime.now().month
+
+    # 调整到目标开始月份
+    target_start_year, target_start_month = last_monday.year, last_monday.month
+    month_diff = (target_start_year - current_left_year) * 12 + (target_start_month - current_left_month)
+    if month_diff < 0:
+        left_arrow = popper.locator('.rcd-date-range-picker__content.is-left .arrow-left').first
+        for _ in range(abs(month_diff)):
+            if left_arrow.count() > 0:
+                left_arrow.click()
+                page.wait_for_timeout(300)
+    elif month_diff > 0:
+        right_arrow = popper.locator('.rcd-date-range-picker__content.is-left .arrow-right').first
+        for _ in range(month_diff):
+            if right_arrow.count() > 0:
+                right_arrow.click()
+                page.wait_for_timeout(300)
+    page.wait_for_timeout(800)
+
+    # 判断是否需要跨月选择
+    same_month = (last_monday.year == last_sunday.year and last_monday.month == last_sunday.month)
+
+    if same_month:
+        # 同月：两个日期都在左面板选择
+        # 先点击开始日期
+        start_clicked = page.evaluate(f"""() => {{
+            const panel = document.querySelector('.rcd-date-range-picker__content.is-left');
+            if (!panel) return false;
+            const cells = panel.querySelectorAll('td');
+            for (const cell of cells) {{
+                if (cell.classList.contains('prev-month') || cell.classList.contains('next-month')) continue;
+                const text = cell.textContent.trim();
+                if (text === '{last_monday.day}') {{ cell.click(); return true; }}
+            }}
+            return false;
+        }}""")
+        page.wait_for_timeout(500)
+        # 再点击结束日期（仍在左面板）
+        end_clicked = page.evaluate(f"""() => {{
+            const panel = document.querySelector('.rcd-date-range-picker__content.is-left');
+            if (!panel) return false;
+            const cells = panel.querySelectorAll('td');
+            for (const cell of cells) {{
+                if (cell.classList.contains('prev-month') || cell.classList.contains('next-month')) continue;
+                const text = cell.textContent.trim();
+                if (text === '{last_sunday.day}') {{ cell.click(); return true; }}
+            }}
+            return false;
+        }}""")
+    else:
+        # 跨月：开始日期在左面板，结束日期在右面板
+        start_clicked = page.evaluate(f"""() => {{
+            const panel = document.querySelector('.rcd-date-range-picker__content.is-left');
+            if (!panel) return false;
+            const cells = panel.querySelectorAll('td');
+            for (const cell of cells) {{
+                if (cell.classList.contains('prev-month') || cell.classList.contains('next-month')) continue;
+                const text = cell.textContent.trim();
+                if (text === '{last_monday.day}') {{ cell.click(); return true; }}
+            }}
+            return false;
+        }}""")
+        page.wait_for_timeout(500)
+        end_clicked = page.evaluate(f"""() => {{
+            const panel = document.querySelector('.rcd-date-range-picker__content.is-right');
+            if (!panel) return false;
+            const cells = panel.querySelectorAll('td');
+            for (const cell of cells) {{
+                if (cell.classList.contains('prev-month') || cell.classList.contains('next-month')) continue;
+                const text = cell.textContent.trim();
+                if (text === '{last_sunday.day}') {{ cell.click(); return true; }}
+            }}
+            return false;
+        }}""")
+
+    if start_clicked:
+        print(f"    [OK] 点击开始日期: {expected_start}")
+    if end_clicked:
+        print(f"    [OK] 点击结束日期: {expected_end}")
+
+    page.wait_for_timeout(1000)
+    try:
+        popper.wait_for(state="hidden", timeout=5000)
+    except:
+        pass
+
+    # 最终验证
+    try:
+        inputs = page.locator(input_selector).all()
+        if len(inputs) >= 2:
+            actual_start = inputs[0].get_attribute('value') or ''
+            actual_end = inputs[1].get_attribute('value') or ''
+            print(f"    [INFO] 当前日期范围: {actual_start} ~ {actual_end}")
+    except:
+        pass
+    print("  [OK] 日期范围设置完成")
+
+
+def click_query_button_appraise(page):
+    """点击查询按钮"""
+    print("  点击查询...")
+    try:
+        query_btn = page.locator('.rcd-button--primary:has-text("查询")').first
+        query_btn.click()
+    except Exception as e:
+        print(f"  [WARN] 点击查询失败: {e}")
+        page.evaluate("""() => {
+            const btns = document.querySelectorAll('.rcd-button');
+            for (const btn of btns) {
+                if (btn.textContent.includes('查询')) { btn.click(); return 'clicked'; }
+            }
+            return 'not_found';
+        }""")
+    page.wait_for_timeout(3000)
+    print("  [OK] 查询完成")
+
+
+def extract_reviews_page(page, rating):
+    """提取当前页指定评级的评价表格数据"""
+    data = page.evaluate("""() => {
+        const rows = document.querySelectorAll('.rcd-table__body tbody tr');
+        const result = [];
+        for (const row of rows) {
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 5) continue;
+            const ratingCell = cells[1];
+            const userName = ratingCell.querySelector('.appraise-manage_user')?.textContent?.trim() || '';
+            const scoreEl = ratingCell.querySelector('.rcd-rate');
+            const score = scoreEl ? (scoreEl.getAttribute('aria-valuenow') || '') : '';
+            const tags = Array.from(ratingCell.querySelectorAll('.rcd-tag__content')).map(t => t.textContent.trim()).filter(Boolean);
+            const contentCell = cells[2];
+            const contentMain = contentCell.querySelector('.appraise-content_main')?.textContent?.trim() || '';
+            const contentTime = contentCell.querySelector('.appraise-content_time')?.textContent?.trim() || '';
+            const orderCell = cells[3];
+            const orderIdSpan = orderCell.querySelector('.order-id span');
+            const orderId = orderIdSpan ? orderIdSpan.textContent.replace('订单编号：', '').trim() : '';
+            const skuName = orderCell.querySelector('.sku-name-text')?.textContent?.trim() || '';
+            const skuIdText = orderCell.querySelector('.sku-id')?.textContent?.trim() || '';
+            const skuId = skuIdText.replace('SKU编码：', '').trim();
+            const skuSize = orderCell.querySelector('.sku-size')?.textContent?.trim() || '';
+            result.push({
+                user_name: userName,
+                score: score,
+                tags: tags,
+                content: contentMain,
+                time: contentTime,
+                order_id: orderId,
+                sku_name: skuName,
+                sku_id: skuId,
+                sku_size: skuSize,
+            });
+        }
+        return JSON.stringify(result);
+    }""")
+    try:
+        reviews = json.loads(data) if isinstance(data, str) else []
+    except Exception:
+        reviews = []
+    for r in reviews:
+        r["rating"] = rating
+    return reviews
+
+
+def collect_all_reviews(page, rating):
+    """翻页采集指定评级的所有评价"""
+    all_reviews = []
+    page_num = 1
+    max_pages = 50
+    while page_num <= max_pages:
+        print(f"    采集第 {page_num} 页...")
+        reviews = extract_reviews_page(page, rating)
+        all_reviews.extend(reviews)
+        try:
+            next_btn = page.locator('.rcd-pagination .btn-next').first
+            if next_btn.count() == 0:
+                print("    无分页组件")
+                break
+            if next_btn.is_disabled():
+                print("    已到达最后一页")
+                break
+            next_btn.click()
+            page.wait_for_timeout(3000)
+            page_num += 1
+        except Exception as e:
+            print(f"    [WARN] 分页处理异常: {e}")
+            break
+    print(f"  [OK] 共采集 {len(all_reviews)} 条{rating}")
+    return all_reviews
+
+
+def save_reviews_output(reviews_by_rating, output_base):
+    """保存评价数据为 JSON + MD
+
+    reviews_by_rating: {'好评': [...], '中评': [...], '差评': [...]}
+    输出:
+      data/jd单品数据/评价汇总_<YYYY-MM-DD>.json  (供 import_product_reviews.py 消费)
+      data/jd单品数据/评价汇总_<YYYY-MM-DD>.md    (人类可读)
+    """
+    today = datetime.now().strftime("%Y-%m-%d")
+    start_day, end_day = get_yesterday_range_text()
+    comment_range = f"{start_day} ~ {end_day}"
+    json_path = os.path.join(output_base, f"评价汇总_{today}.json")
+    md_path = os.path.join(output_base, f"评价汇总_{today}.md")
+
+    payload = {
+        "platform": "jd",
+        "shop": "光影行星研发",
+        "collect_date": today,
+        "comment_date_range": comment_range,
+        "captured_at": datetime.now().isoformat(),
+        "total_count": sum(len(v) for v in reviews_by_rating.values()),
+        "ratings": {
+            r: {"collected_count": len(reviews), "comments": reviews}
+            for r, reviews in reviews_by_rating.items()
+        },
+    }
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(payload, f, ensure_ascii=False, indent=2)
+
+    md = [f"# JD 商品评价汇总", "",
+          f"**采集日期:** {today}",
+          f"**评价时间范围:** {comment_range}",
+          f"**评价总数:** {payload['total_count']} 条", ""]
+    md.append("| 评价等级 | 数量 |")
+    md.append("|----------|------|")
+    for r in ["好评", "中评", "差评"]:
+        md.append(f"| {r} | {len(reviews_by_rating.get(r, []))} |")
+    md.append("")
+    for rating in ["好评", "中评", "差评"]:
+        reviews = reviews_by_rating.get(rating, [])
+        md.append(f"## {rating}")
+        md.append("")
+        if not reviews:
+            md.append("> 无数据")
+            md.append("")
+            continue
+        for i, review in enumerate(reviews, 1):
+            md.append(f"### {rating} {i}")
+            md.append("")
+            md.append(f"- **用户昵称:** {review.get('user_name', '')}")
+            score = review.get('score', '')
+            md.append(f"- **评分:** {score} 星" if score else f"- **评分:** {rating}")
+            tags = review.get('tags', [])
+            if tags:
+                md.append(f"- **标签:** {', '.join(tags)}")
+            md.append(f"- **评价时间:** {review.get('time', '')}")
+            md.append(f"- **评价内容:** {review.get('content', '')}")
+            md.append(f"- **订单编号:** {review.get('order_id', '')}")
+            md.append(f"- **商品名称:** {review.get('sku_name', '')}")
+            md.append(f"- **SKU编码:** {review.get('sku_id', '')}")
+            md.append(f"- **规格:** {review.get('sku_size', '')}")
+            md.append("")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md))
+    print(f"  [OK] 评价数据已保存: {json_path}")
+    print(f"  [OK] 评价数据已保存: {md_path}")
+    return json_path, md_path
+
+
+# ── 保存单个 SPU 详情 ─────────────────────────────────────────────────────────
+
+def save_single_product_detail(spu_info, core_data, traffic_data, keyword_data, crowd_data, output_base,
+                          is_apollo2=False, is_zeus3=False, is_xingji=False,
+                          apollo2_summary=None, zeus3_summary=None, xingji_summary=None):
+    """保存单个 SPU 的核心数据、流量来源、关键词来源成交、人群画像到 单品详情数据_日期/SPU子文件夹/（JSON + Markdown）"""
+    today = datetime.now().strftime("%Y-%m-%d")
+    detail_dir = os.path.join(output_base, f"单品详情数据_{today}")
+
+    spu = spu_info.get('spu', '')
+    product = spu_info.get('product', '')
+    safe_name = re.sub(r'[\\/:*?"<>|]', '', product)[:30]
+    file_name = f"{spu}_{safe_name}" if safe_name else spu
+
+    # 阿波罗2或宙斯3或星迹使用共享文件夹
+    if is_apollo2:
+        spu_dir = os.path.join(detail_dir, APOLLO2_FOLDER)
+    elif is_zeus3:
+        spu_dir = os.path.join(detail_dir, ZEUS3_FOLDER)
+    elif is_xingji:
+        spu_dir = os.path.join(detail_dir, XINGJI_FOLDER)
+    else:
+        spu_dir = os.path.join(detail_dir, file_name)
+    os.makedirs(spu_dir, exist_ok=True)
+
+    start_day, end_day = get_yesterday_range_text()
+    week_range = f"{start_day} ~ {end_day}"
+
+    record = {
+        "spu": spu,
+        "product": product,
+        "image": spu_info.get('image', ''),
+        "url": spu_info.get('url', ''),
+        "采集日期": today,
+        "数据周期": week_range,
+        "核心数据": core_data,
+        "流量来源": traffic_data,
+        "关键词来源成交": keyword_data,
+        "人群画像": crowd_data,
+    }
+
+    # ── 计算转化率和加购率 ──────────────────────────────────────────────────────
+    # 转化率 = 加购客户数 / 访客数
+    # 加购率 = 加购商品件数 / 访客数
+    try:
+        visitors = 0
+        add_cart_customers = 0
+        add_cart_items = 0
+        # 从核心数据的"交易"分组中提取访客数、加购客户数、加购商品件数
+        for group_name, items in core_data.items():
+            for title, v in items.items():
+                val = v.get('value', '')
+                if not val or val == 'N/A':
+                    continue
+                # 尝试解析数值（去除逗号）
+                try:
+                    num = float(str(val).replace(',', '').replace('%', ''))
+                except:
+                    continue
+                if '访客数' in title or '访客' in title:
+                    visitors = num
+                if '加购客户' in title or '加购人' in title:
+                    add_cart_customers = num
+                if '加购商品件' in title or '加购商品' in title:
+                    add_cart_items = num
+
+        if visitors > 0:
+            record['转化率'] = f"{(add_cart_customers / visitors * 100):.2f}%"
+            record['加购率'] = f"{(add_cart_items / visitors * 100):.2f}%"
+        else:
+            record['转化率'] = 'N/A'
+            record['加购率'] = 'N/A'
+    except Exception as e:
+        print(f"    [WARN] 计算转化率/加购率异常: {e}")
+        record['转化率'] = 'N/A'
+        record['加购率'] = 'N/A'
+
+    # ── 提取待插入数据指标 ────────────────────────────────────────────────────
+    # 销量=成交金额, 访客数, 转化率, 加购数(=加购客户数), 退款率(=退货率)
+    # 注：访客数、加购客户数、加购商品件数已在上面计算转化率时提取
+    sales_amount = 0
+    refund_rate = 'N/A'
+    for group_name, items in core_data.items():
+        for title, v in items.items():
+            val = v.get('value', '')
+            if not val or val == 'N/A':
+                continue
+            if '成交金额' in title:
+                sales_amount = val  # 保留原始格式
+            if '退货率' in title:
+                refund_rate = val
+
+    # 阿波罗2或宙斯3或星迹：暂存汇总数据，不在此处保存待插入数据
+    target_summary = None
+    if is_apollo2:
+        target_summary = apollo2_summary
+    elif is_zeus3:
+        target_summary = zeus3_summary
+    elif is_xingji:
+        target_summary = xingji_summary
+
+    if target_summary is not None:
+        try:
+            target_summary['sales_amount'] += float(str(sales_amount).replace(',', '') if sales_amount else 0)
+        except:
+            pass
+        try:
+            target_summary['visitors'] += float(visitors) if visitors else 0
+        except:
+            pass
+        try:
+            target_summary['add_cart_customers'] += float(add_cart_customers) if add_cart_customers else 0
+        except:
+            pass
+        try:
+            target_summary['add_cart_items'] += float(add_cart_items) if add_cart_items else 0
+        except:
+            pass
+        if refund_rate != 'N/A':
+            try:
+                target_summary['refund_rate_sum'] += float(str(refund_rate).replace('%', ''))
+                target_summary['refund_rate_count'] += 1
+            except:
+                pass
+        # 多 SPU 产品：每个 SPU 仍然保存各自的待插入数据（文件名含 SPU 区分），不 return
+
+    summary_record = {
+        "spu": spu,
+        "product": product,
+        "采集日期": today,
+        "数据周期": week_range,
+        "销量（成交金额）": sales_amount if sales_amount else 'N/A',
+        "访客数": visitors if visitors else 'N/A',
+        "转化率": record.get('转化率', 'N/A'),
+        "加购数（加购客户数）": add_cart_customers if add_cart_customers else 'N/A',
+        "退款率（退货率）": refund_rate,
+    }
+
+    # ── 保存待插入数据 JSON ─────────────────────────────────────────────────
+    # 多 SPU 产品用 SPU 编号区分文件名，单 SPU 产品直接用"待插入数据"
+    if target_summary is not None:
+        summary_json_path = os.path.join(spu_dir, f"待插入数据_{spu}.json")
+    else:
+        summary_json_path = os.path.join(spu_dir, "待插入数据.json")
+    with open(summary_json_path, "w", encoding="utf-8") as f:
+        json.dump(summary_record, f, ensure_ascii=False, indent=2)
+    print(f"    [OK] 待插入数据 JSON 已保存: {summary_json_path}")
+
+    # ── 保存待插入数据 Markdown ─────────────────────────────────────────────
+    summary_md = []
+    summary_md.append(f"# {product} - 待插入数据")
+    summary_md.append("")
+    summary_md.append(f"**SPU：** {spu}")
+    summary_md.append(f"**采集日期：** {today}")
+    summary_md.append(f"**数据周期：** {week_range}")
+    summary_md.append("")
+    summary_md.append("| 指标 | 数值 |")
+    summary_md.append("|------|------|")
+    summary_md.append(f"| 销量（成交金额） | {summary_record['销量（成交金额）']} |")
+    summary_md.append(f"| 访客数 | {summary_record['访客数']} |")
+    summary_md.append(f"| 转化率 | {summary_record['转化率']} |")
+    summary_md.append(f"| 加购数（加购客户数） | {summary_record['加购数（加购客户数）']} |")
+    summary_md.append(f"| 退款率（退货率） | {summary_record['退款率（退货率）']} |")
+    if target_summary is not None:
+        summary_md_path = os.path.join(spu_dir, f"待插入数据_{spu}.md")
+    else:
+        summary_md_path = os.path.join(spu_dir, "待插入数据.md")
+    with open(summary_md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(summary_md))
+    print(f"    [OK] 待插入数据 Markdown 已保存: {summary_md_path}")
+
+    # ── JSON ──
+    json_path = os.path.join(spu_dir, f"{file_name}.json")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(record, f, ensure_ascii=False, indent=2)
+    print(f"    [OK] JSON 已保存: {json_path}")
+
+    # ── Markdown ──
+    md = []
+    md.append(f"# {product}")
+    md.append("")
+    md.append(f"**SPU：** {spu}")
+    md.append(f"**采集日期：** {today}")
+    md.append(f"**数据周期：** {week_range}")
+    if spu_info.get('url'):
+        md.append(f"**商品链接：** [{spu_info['url']}]({spu_info['url']})")
+    md.append("")
+
+    # 核心数据
+    for group_name, items in core_data.items():
+        md.append(f"## {group_name}")
+        md.append("")
+        md.append("| 指标 | 数值 | 对比标签 | 对比值 | 趋势 |")
+        md.append("|------|------|----------|--------|------|")
+        for title, v in items.items():
+            value = v.get('value', '')
+            compare_label = v.get('compare_label', '')
+            compare_value = v.get('compare_value', '')
+            trend = v.get('trend', '')
+            trend_icon = '📈' if trend == 'up' else ('📉' if trend == 'down' else '')
+            md.append(f"| {title} | {value} | {compare_label} | {compare_value} | {trend_icon} |")
+        md.append("")
+
+    # 流量来源
+    if traffic_data:
+        md.append("## 流量来源")
+        md.append("")
+        headers = list(traffic_data[0].keys())
+        md.append("| " + " | ".join(headers) + " |")
+        md.append("| " + " | ".join(["------"] * len(headers)) + " |")
+        for row in traffic_data:
+            cells = []
+            for h in headers:
+                val = row.get(h, '')
+                if isinstance(val, dict):
+                    v = val.get('value', '')
+                    r = val.get('rate', '')
+                    t = val.get('trend', '')
+                    icon = '📈' if t == 'up' else ('📉' if t == 'down' else '')
+                    cells.append(f"{v} {r}{icon}" if r else v)
+                else:
+                    cells.append(str(val))
+            md.append("| " + " | ".join(cells) + " |")
+        md.append("")
+
+    # 关键词来源成交
+    if keyword_data:
+        md.append("## 关键词来源成交")
+        md.append("")
+        headers = list(keyword_data[0].keys())
+        md.append("| " + " | ".join(headers) + " |")
+        md.append("| " + " | ".join(["------"] * len(headers)) + " |")
+        for row in keyword_data:
+            cells = [str(row.get(h, '')) for h in headers]
+            md.append("| " + " | ".join(cells) + " |")
+        md.append("")
+
+    # 人群画像
+    if crowd_data:
+        md.append("## 人群画像")
+        md.append("")
+        if crowd_data.get("性别"):
+            md.append("### 性别分布")
+            md.append("")
+            md.append("| 性别 | 占比 |")
+            md.append("|------|------|")
+            for item in crowd_data["性别"]:
+                md.append(f"| {item.get('name', '')} | {item.get('value', '')} |")
+            md.append("")
+        if crowd_data.get("年龄"):
+            md.append("### 年龄分布")
+            md.append("")
+            md.append("| 年龄段 | 占比 |")
+            md.append("|--------|------|")
+            for item in crowd_data["年龄"]:
+                md.append(f"| {item.get('name', '')} | {item.get('value', '')} |")
+            md.append("")
+        if crowd_data.get("地区_省份"):
+            md.append("### 地区分布（省份）")
+            md.append("")
+            md.append("| 排名 | 地区 | 占比 |")
+            md.append("|------|------|------|")
+            for item in crowd_data["地区_省份"]:
+                md.append(f"| {item.get('rank', '')} | {item.get('region', '')} | {item.get('ratio', '')} |")
+            md.append("")
+        if crowd_data.get("地区_城市"):
+            md.append("### 地区分布（城市）")
+            md.append("")
+            md.append("| 排名 | 城市 | 占比 |")
+            md.append("|------|------|------|")
+            for item in crowd_data["地区_城市"]:
+                md.append(f"| {item.get('rank', '')} | {item.get('region', '')} | {item.get('ratio', '')} |")
+            md.append("")
+        if crowd_data.get("职业"):
+            md.append("### 职业分布")
+            md.append("")
+            md.append("| 职业 | 占比 |")
+            md.append("|------|------|")
+            for item in crowd_data["职业"]:
+                md.append(f"| {item.get('name', '')} | {item.get('value', '')} |")
+            md.append("")
+
+    md_path = os.path.join(spu_dir, f"{file_name}.md")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md))
+    print(f"    [OK] Markdown 已保存: {md_path}")
+
+    return json_path, md_path, spu_dir
+
+
+# ── 逐个 SPU 采集详情主流程 ───────────────────────────────────────────────────
+
+def _single_product_output_exists(spu_info, output_base):
+    today = datetime.now().strftime("%Y-%m-%d")
+    spu = spu_info.get("spu", "")
+    product = spu_info.get("product", "")[:40]
+    safe_name = re.sub(r'[\\/:*?"<>|]', '', product)[:30]
+    file_name = f"{spu}_{safe_name}" if safe_name else spu
+
+    if spu in APOLLO2_SPU_LIST:
+        spu_dir = os.path.join(output_base, f"单品详情数据_{today}", APOLLO2_FOLDER)
+        json_path = os.path.join(spu_dir, f"{file_name}.json")
+    elif spu in ZEUS3_SPU_LIST:
+        spu_dir = os.path.join(output_base, f"单品详情数据_{today}", ZEUS3_FOLDER)
+        json_path = os.path.join(spu_dir, f"{file_name}.json")
+    elif spu in XINGJI_SPU_LIST:
+        spu_dir = os.path.join(output_base, f"单品详情数据_{today}", XINGJI_FOLDER)
+        json_path = os.path.join(spu_dir, f"{file_name}.json")
+    else:
+        spu_dir = os.path.join(output_base, f"单品详情数据_{today}", file_name)
+        json_path = os.path.join(spu_dir, f"{file_name}.json")
+
+    return os.path.exists(json_path)
+
+
+def collect_each_spu_detail(page, context, all_spus, output_base, resume=False):
+    """逐个搜索 SPU → 点击详情 → 新页面 → 核心数据 → 保存"""
+    start_day, end_day = get_yesterday_range_text()
+    week_range = f"{start_day} ~ {end_day}"
+    total = len(all_spus)
+    print(f"\n[Step 6] 开始逐个采集 {total} 个 SPU 的详情...")
+    progress = SpuProgress(output_base, "product_spu_detail")
+
+    if resume:
+        completed_spus = progress.successful_spus()
+        filtered = []
+        skipped = 0
+        for spu_info in all_spus:
+            spu = str(spu_info.get("spu", ""))
+            if spu in completed_spus or _single_product_output_exists(spu_info, output_base):
+                skipped += 1
+            else:
+                filtered.append(spu_info)
+        if skipped:
+            print(f"  [Resume] 跳过已成功采集的 {skipped} 个 SPU，剩余 {len(filtered)} 个待采集")
+        all_spus = filtered
+        total = len(all_spus)
+
+    if not all_spus:
+        print("  [OK] 所有 SPU 均已采集，无需继续")
+        return 0, 0
+
+    success = 0
+    failed = 0
+
+    def make_summary():
+        return {"sales_amount": 0, "visitors": 0, "add_cart_customers": 0,
+                "add_cart_items": 0, "refund_rate_sum": 0, "refund_rate_count": 0}
+
+    apollo2_summary = make_summary()
+    zeus3_summary = make_summary()
+    xingji_summary = make_summary()
+
+    for idx, spu_info in enumerate(all_spus):
+        spu = spu_info.get('spu', '')
+        product = spu_info.get('product', '')[:40]
+
+        # 判断所属分组
+        if spu in APOLLO2_SPU_LIST:
+            group_name, group_folder, group_summary = "阿波罗2", APOLLO2_FOLDER, apollo2_summary
+        elif spu in ZEUS3_SPU_LIST:
+            group_name, group_folder, group_summary = "宙斯3", ZEUS3_FOLDER, zeus3_summary
+        elif spu in XINGJI_SPU_LIST:
+            group_name, group_folder, group_summary = "星迹", XINGJI_FOLDER, xingji_summary
+        else:
+            group_name, group_folder, group_summary = None, None, None
+
+        print(f"\n{'='*50}")
+        print(f"  [{idx+1}/{total}] SPU: {spu}")
+        print(f"  商品: {product}...")
+        if group_name:
+            print(f"  [{group_name}] 共享文件夹模式")
+        print(f"{'='*50}")
+
+        if not spu:
+            print("  [SKIP] SPU 为空，跳过")
+            failed += 1
+            progress.append("failed", spu_info, error="SPU 为空")
+            continue
+
+        progress.mark_running(spu_info, idx + 1, total)
+
+        try:
+            # 1. 搜索 SPU
+            search_spu_in_detail_page(page, spu)
+
+            # 2. 点击详情，打开新页面
+            detail_page = click_detail_and_wait_new_page(page, context, spu)
+            if not detail_page:
+                failed += 1
+                progress.append("failed", spu_info, error="详情页未打开")
+                continue
+
+            # 3. 点击核心数据 tab
+            click_core_data_tab(detail_page)
+
+            # 4. 选择昨天日期
+            select_last_week_on_detail_page(detail_page)
+
+            # 5. 采集核心数据
+            core_data = extract_core_data(detail_page)
+
+            # 6. 点击流量来源 tab
+            click_traffic_source_tab(detail_page)
+
+            # 7. 展开所有行
+            expand_all_traffic_rows(detail_page)
+
+            # 8. 采集流量来源数据
+            traffic_data = extract_traffic_data(detail_page)
+
+            # 10. 采集关键词来源成交数据
+            keyword_data = extract_keyword_data(detail_page)
+
+            # 11. 点击人群画像tab
+            click_crowd_portrait_tab(detail_page)
+
+            # 12. 采集人群画像数据
+            crowd_data = extract_crowd_portrait_data(detail_page)
+
+            # 13. 截图（保存到 SPU 子文件夹）
+            today = datetime.now().strftime("%Y-%m-%d")
+            safe_name = re.sub(r'[\\/:*?"<>|]', '', product)[:30]
+            file_name = f"{spu}_{safe_name}" if safe_name else spu
+
+            if group_name:
+                spu_dir = os.path.join(output_base, f"单品详情数据_{today}", group_folder)
+                os.makedirs(spu_dir, exist_ok=True)
+            else:
+                spu_dir = os.path.join(output_base, f"单品详情数据_{today}", file_name)
+                os.makedirs(spu_dir, exist_ok=True)
+
+            if group_name:
+                screenshot_path = os.path.join(spu_dir, f"screenshot_{spu}.png")
+            else:
+                screenshot_path = os.path.join(spu_dir, "screenshot.png")
+            detail_page.screenshot(path=screenshot_path, full_page=True)
+            print(f"    [OK] 截图已保存: {screenshot_path}")
+
+            # 14. 保存（传入 crowd_data）
+            save_single_product_detail(spu_info, core_data, traffic_data, keyword_data, crowd_data, output_base,
+                                    is_apollo2=(group_name == "阿波罗2"), is_zeus3=(group_name == "宙斯3"), is_xingji=(group_name == "星迹"),
+                                    apollo2_summary=apollo2_summary, zeus3_summary=zeus3_summary, xingji_summary=xingji_summary)
+            success += 1
+            progress.mark_success(spu_info, group=group_name or "")
+
+            detail_page.close()
+            page.wait_for_timeout(2000)
+
+        except Exception as e:
+            print(f"  [FAIL] 异常: {e}")
+            failed += 1
+            progress.mark_failed(spu_info, e)
+            try:
+                # 只关闭当前 SPU 的详情页，避免误关主页面
+                detail_page_local = locals().get('detail_page')
+                if detail_page_local and detail_page_local != page:
+                    detail_page_local.close()
+            except:
+                pass
+
+    print(f"\n[OK] SPU 详情采集完成: 成功 {success}, 失败 {failed}")
+    print(f"  [Progress] 进度日志: {progress.path}")
+
+    # ── 汇总函数：阿波罗2/宙斯3 通用 ─────────────────────────────────────────
+    def save_group_summary(group_name, folder, spu_list, summary, week_range):
+        if summary['visitors'] > 0:
+            summary['转化率'] = f"{(summary['add_cart_customers'] / summary['visitors'] * 100):.2f}%"
+        else:
+            summary['转化率'] = 'N/A'
+        if summary['refund_rate_count'] > 0:
+            summary['退款率'] = f"{(summary['refund_rate_sum'] / summary['refund_rate_count']):.2f}%"
+        else:
+            summary['退款率'] = 'N/A'
+
+        today = datetime.now().strftime("%Y-%m-%d")
+        group_dir = os.path.join(output_base, f"单品详情数据_{today}", folder)
+        os.makedirs(group_dir, exist_ok=True)
+
+        record = {
+            "product": group_name,
+            "spu_list": spu_list,
+            "采集日期": today,
+            "数据周期": week_range,
+            "销量（成交金额）": f"{summary['sales_amount']:.2f}" if summary['sales_amount'] else 'N/A',
+            "访客数": int(summary['visitors']) if summary['visitors'] else 'N/A',
+            "转化率": summary['转化率'],
+            "加购数（加购客户数）": int(summary['add_cart_customers']) if summary['add_cart_customers'] else 'N/A',
+            "退款率（退货率）": summary['退款率'],
+        }
+
+        json_path = os.path.join(group_dir, "待插入数据.json")
+        with open(json_path, "w", encoding="utf-8") as f:
+            json.dump(record, f, ensure_ascii=False, indent=2)
+        print(f"    [OK] {group_name} 待插入数据 JSON 已保存: {json_path}")
+
+        md = []
+        md.append(f"# {group_name} - 待插入数据")
+        md.append("")
+        md.append(f"**SPU列表：** {', '.join(spu_list)}")
+        md.append(f"**采集日期：** {today}")
+        md.append(f"**数据周期：** {week_range}")
+        md.append("")
+        md.append("| 指标 | 数值 |")
+        md.append("|------|------|")
+        md.append(f"| 销量（成交金额） | {record['销量（成交金额）']} |")
+        md.append(f"| 访客数 | {record['访客数']} |")
+        md.append(f"| 转化率 | {record['转化率']} |")
+        md.append(f"| 加购数（加购客户数） | {record['加购数（加购客户数）']} |")
+        md.append(f"| 退款率（退货率） | {record['退款率（退货率）']} |")
+        md_path = os.path.join(group_dir, "待插入数据.md")
+        with open(md_path, "w", encoding="utf-8") as f:
+            f.write("\n".join(md))
+        print(f"    [OK] {group_name} 待插入数据 Markdown 已保存: {md_path}")
+
+    # 断点续采时本次只包含剩余 SPU，避免用不完整累加值覆盖已有分组汇总。
+    if resume:
+        print("  [Resume] 跳过分组总汇总覆盖，保留已有汇总文件")
+    else:
+        save_group_summary("阿波罗2", APOLLO2_FOLDER, APOLLO2_SPU_LIST, apollo2_summary, week_range)
+        save_group_summary("宙斯3", ZEUS3_FOLDER, ZEUS3_SPU_LIST, zeus3_summary, week_range)
+        save_group_summary("星迹", XINGJI_FOLDER, XINGJI_SPU_LIST, xingji_summary, week_range)
+
+    return success, failed
+
+
+# ── 商品榜单排名采集 ───────────────────────────────────────────────────────────
+
+def select_week_and_last_week_on_rank_page(page):
+    """在榜单页面设置日期为昨天、按天查询。函数名保留，避免改动调用链。"""
+    print("  设置榜单页面日期为昨天...")
+    target_date = get_yesterday_date()
+    target_day = target_date.strftime("%Y-%m-%d")
+
+    # 检查当前是否已是目标日期
+    try:
+        current_label = page.locator('.jmtd-card-header .jmtd-base-input[data-component-name="DatePicker"] .jmtd-label').first.inner_text(timeout=3000)
+        if "按天查询" in current_label and _date_text_matches_day(current_label, target_day):
+            print(f"    [OK] 日期已是昨天: {current_label}")
+            return
+    except Exception:
+        pass
+
+    # 1. 点击日期选择器
+    date_picker = page.locator('.jmtd-card-header .jmtd-base-input[data-component-name="DatePicker"]').first
+    date_picker.click()
+    page.wait_for_timeout(1500)
+
+    # 2. 点击"按天查询"
+    try:
+        day_option = page.locator('.jmtd-date-picker-switch-panel-item:has-text("按天查询")').first
+        day_option.wait_for(state="visible", timeout=5000)
+        day_option.click()
+        print("    [OK] 选择按天查询")
+    except Exception as e:
+        print(f"    [WARN] 点击按天查询失败: {e}")
+        page.evaluate("""() => {
+            const items = document.querySelectorAll('.jmtd-date-picker-switch-panel-item');
+            for (const item of items) {
+                if (item.textContent.includes('按天查询')) { item.click(); return 'clicked'; }
+            }
+            return 'not_found';
+        }""")
+    page.wait_for_timeout(1500)
+
+    if not _click_day_in_open_datepicker(page, target_date, "product_rank"):
+        raise RuntimeError(f"榜单页面未能选择昨天日期: {target_day}")
+
+    page.wait_for_timeout(3000)
+
+    # 4. 等待弹窗关闭
+    try:
+        page.locator('.jmtd-dropdown-lists').first.wait_for(state="hidden", timeout=5000)
+    except:
+        pass
+
+    # 5. 验证
+    new_label = ""
+    try:
+        new_label = page.locator('.jmtd-card-header .jmtd-base-input[data-component-name="DatePicker"] .jmtd-label').first.inner_text(timeout=3000)
+        print(f"    [INFO] 当前日期: {new_label}")
+    except Exception:
+        pass
+    if new_label and not _date_text_matches_day(new_label, target_day):
+        raise RuntimeError(f"榜单日期未切换到昨天: {new_label}")
+    print("  [OK] 榜单日期设置完成")
+
+
+def select_category_on_rank_page(page, category_path):
+    """在榜单页面选择类目（支持两级类目）"""
+    cat_label = " > ".join(category_path)
+    print(f"  选择类目: {cat_label}")
+
+    # 点击类目 Select（在卡片头部右侧）
+    try:
+        cat_select = page.locator('.jmtd-card-header-right .jmtd-base-input:has(.jmtd-label)').nth(1)
+        cat_select.click()
+    except Exception:
+        # 兜底：点击包含"男包"或"功能箱包"的 select
+        cat_select = page.locator('.jmtd-card-header .jmtd-base-input:has(.jmtd-label:has-text("男包")), .jmtd-card-header .jmtd-base-input:has(.jmtd-label:has-text("功能箱包"))').first
+        cat_select.click()
+    page.wait_for_timeout(1500)
+
+    # 第一列：选择大类目
+    big_cat = category_path[0]
+    try:
+        big_option = page.locator(f'.jmtd-dropdown-option:has-text("{big_cat}")').first
+        big_option.wait_for(state="visible", timeout=5000)
+        big_option.click()
+        print(f"    [OK] 选择大类目: {big_cat}")
+    except Exception as e:
+        print(f"    [WARN] 选择大类目失败: {e}")
+        page.evaluate(f"""() => {{
+            const opts = document.querySelectorAll('.jmtd-dropdown-option');
+            for (const opt of opts) {{
+                if (opt.textContent.trim() === '{big_cat}') {{ opt.click(); return 'clicked'; }}
+            }}
+            return 'not_found';
+        }}""")
+    page.wait_for_timeout(800)
+
+    # 如果有子类目，选择子类目
+    if len(category_path) > 1:
+        sub_cat = category_path[1]
+        try:
+            sub_option = page.locator(f'.jmtd-dropdown-option:has-text("{sub_cat}")').first
+            sub_option.wait_for(state="visible", timeout=5000)
+            sub_option.click()
+            print(f"    [OK] 选择子类目: {sub_cat}")
+        except Exception as e:
+            print(f"    [WARN] 选择子类目失败: {e}")
+            page.evaluate(f"""() => {{
+                const opts = document.querySelectorAll('.jmtd-dropdown-option');
+                for (const opt of opts) {{
+                    if (opt.textContent.trim() === '{sub_cat}') {{ opt.click(); return 'clicked'; }}
+                }}
+                return 'not_found';
+            }}""")
+        page.wait_for_timeout(800)
+    else:
+        # 只选大类目：按 Escape 关闭下拉
+        page.keyboard.press("Escape")
+        page.wait_for_timeout(500)
+
+    # 等待表格刷新
+    page.wait_for_timeout(2000)
+
+
+def search_product_on_rank_page(page, product_name):
+    """在榜单页面搜索商品"""
+    try:
+        # 优先使用工具栏中的搜索框（更靠近表格）
+        search_input = page.locator('.tool-bar input.jmtd-input-input').first
+        if search_input.count() == 0:
+            # 兜底：页面顶部的搜索框
+            search_input = page.locator('input.jmtd-input-input[placeholder*="商品名称"]').first
+        search_input.fill("")
+        page.wait_for_timeout(300)
+        search_input.fill(product_name)
+        page.wait_for_timeout(3000)
+        # 按 Enter 确保搜索生效
+        search_input.press("Enter")
+        page.wait_for_timeout(2000)
+    except Exception as e:
+        print(f"    [WARN] 搜索商品失败: {e}")
+
+
+def extract_rank_table_data(page, product_name):
+    """提取榜单表格中匹配商品的数据"""
+    data = page.evaluate("""(productName) => {
+        const rows = document.querySelectorAll('.jmtd-table-main tbody tr');
+        for (const row of rows) {
+            const nameEl = row.querySelector('.prod-name');
+            const name = nameEl ? (nameEl.getAttribute('title') || nameEl.textContent.trim()) : '';
+            if (!name) continue;
+            // 模糊匹配：互相包含或去除空格后匹配
+            const n1 = name.toLowerCase().replace(/\\s+/g, '');
+            const n2 = productName.toLowerCase().replace(/\\s+/g, '');
+            if (n1.includes(n2) || n2.includes(n1) || name.includes(productName) || productName.includes(name)) {
+                const rankEl = row.querySelector('.jmtd-table-formatter-rank-number');
+                const rank = rankEl ? rankEl.textContent.trim() : '';
+                const cells = row.querySelectorAll('td');
+                return {
+                    rank: rank,
+                    product_name: name,
+                    成交金额: cells[2] ? cells[2].textContent.trim() : '',
+                    成交单量: cells[3] ? cells[3].textContent.trim() : '',
+                    访客数: cells[4] ? cells[4].textContent.trim() : '',
+                    搜索点击次数: cells[5] ? cells[5].textContent.trim() : '',
+                    关注人数: cells[6] ? cells[6].textContent.trim() : '',
+                };
+            }
+        }
+        return null;
+    }""", product_name)
+    return data
+
+
+def append_rank_data_to_spu(spu, file_name, rank_records, detail_dir):
+    """将榜单数据追加到 SPU 的 JSON 和 MD 文件"""
+    spu_dir = os.path.join(detail_dir, file_name)
+    if not os.path.exists(spu_dir):
+        print(f"    [WARN] SPU目录不存在: {spu_dir}")
+        return
+
+    # 更新 JSON
+    json_path = os.path.join(spu_dir, f"{file_name}.json")
+    if os.path.exists(json_path):
+        try:
+            with open(json_path, "r", encoding="utf-8") as f:
+                record = json.load(f)
+            record["商品榜单排名"] = rank_records
+            with open(json_path, "w", encoding="utf-8") as f:
+                json.dump(record, f, ensure_ascii=False, indent=2)
+            print(f"    [OK] JSON 已追加榜单数据")
+        except Exception as e:
+            print(f"    [WARN] JSON更新失败: {e}")
+
+    # 追加 MD
+    md_path = os.path.join(spu_dir, f"{file_name}.md")
+    if os.path.exists(md_path) and rank_records:
+        try:
+            with open(md_path, "a", encoding="utf-8") as f:
+                f.write("\n\n## 商品榜单排名\n\n")
+                f.write("| 类目 | 排名 | 成交金额 | 成交单量 | 访客数 | 搜索点击次数 | 关注人数 |\n")
+                f.write("|------|------|----------|----------|--------|--------------|----------|\n")
+                for r in rank_records:
+                    f.write(f"| {r.get('类目', '')} | {r.get('rank', '')} | {r.get('成交金额', '')} | {r.get('成交单量', '')} | {r.get('访客数', '')} | {r.get('搜索点击次数', '')} | {r.get('关注人数', '')} |\n")
+            print(f"    [OK] Markdown 已追加榜单数据")
+        except Exception as e:
+            print(f"    [WARN] MD追加失败: {e}")
+
+
+def collect_rank_for_single_spu(context, spu_info, output_base):
+    """单个SPU采集完详情后，打开新标签页去榜单查排名，追加到其JSON/MD"""
+    spu = spu_info.get('spu', '')
+    product = spu_info.get('product', '')
+    if not product:
+        return
+
+    print(f"  采集商品榜单排名...")
+
+    # 打开新标签页去榜单页面
+    rank_page = context.new_page()
+
+    try:
+        # 1. 跳转榜单页面
+        rank_page.goto("https://sz.jd.com/szweb/sz/view/industryMarket/productRanks.html", wait_until="domcontentloaded", timeout=30000)
+        rank_page.wait_for_timeout(5000)
+
+        # 2. 设置日期
+        select_week_and_last_week_on_rank_page(rank_page)
+
+        # 3. 先输入一次商品名称，后续只切换类目
+        search_product_on_rank_page(rank_page, product)
+
+        # 4. 类目组合
+        categories = [
+            {"path": ["男包"]},
+            {"path": ["男包", "男士双肩包"]},
+            {"path": ["男包", "男士单肩/斜挎包"]},
+            {"path": ["功能箱包"]},
+            {"path": ["功能箱包", "电脑包"]},
+        ]
+
+        rank_records = []
+
+        for cat in categories:
+            cat_label = " > ".join(cat["path"])
+            print(f"    查询类目: {cat_label}")
+
+            try:
+                select_category_on_rank_page(rank_page, cat["path"])
+
+                rank_data = extract_rank_table_data(rank_page, product)
+                if rank_data:
+                    print(f"      [OK] 排名: {rank_data.get('rank', '')}")
+                    rank_records.append({"类目": cat_label, **rank_data})
+                else:
+                    print(f"      [INFO] 该类目下无此商品")
+            except Exception as e:
+                print(f"      [ERROR] 查询类目异常: {e}")
+
+        # 4. 追加到当前SPU的文件；没有命中时 JSON 也写入空数组，便于后续判断已采集
+        today = datetime.now().strftime("%Y-%m-%d")
+        detail_dir = os.path.join(output_base, f"单品详情数据_{today}")
+        safe_name = re.sub(r'[\\/:*?"<>|]', '', product)[:30]
+        file_name = f"{spu}_{safe_name}" if safe_name else spu
+        append_rank_data_to_spu(spu, file_name, rank_records, detail_dir)
+        if not rank_records:
+            print(f"    [INFO] 所有类目均无此商品排名数据")
+
+    except Exception as e:
+        print(f"    [ERROR] 榜单采集异常: {e}")
+    finally:
+        rank_page.close()
+
+
+# ── 主流程 ─────────────────────────────────────────────────────────────────────
+
+def safe_filename(value, limit=80):
+    text = str(value or "").strip()
+    text = re.sub(r'[\\/:*?"<>|]+', "_", text)
+    text = re.sub(r"\s+", " ", text).strip(" ._")
+    return (text[:limit] if text else "未命名")
+
+
+def _safe_relpath(path, start):
+    """跨盘时降级到 basename，正常时返回正斜杠的相对路径"""
+    if not path:
+        return ""
+    try:
+        return os.path.relpath(path, start).replace("\\", "/")
+    except ValueError:
+        return os.path.basename(path)
+
+
+def _collect_style_spu_tasks():
+    tasks = []
+    for group in STYLE_SPU_GROUPS:
+        style = group["style"]
+        for spu in group["spus"]:
+            tasks.append({"style": style, "spu": str(spu), "product": "", "image": "", "url": ""})
+    return tasks
+
+
+def collect_all_spus(page):
+    """使用固定款式 SPU 清单，不再翻页全量采集商品列表。"""
+    print("\n[Step 5] 使用固定款式 SPU 清单，不再全量翻页采集...")
+    try:
+        select_week_and_last_week(page)
+    except Exception as e:
+        print(f"  [WARN] 商品明细列表日期设置失败，继续按固定 SPU 采集详情: {e}")
+    tasks = _collect_style_spu_tasks()
+    print(f"  [OK] 款式 {len(STYLE_SPU_GROUPS)} 个，SPU {len(tasks)} 个")
+    return tasks
+
+
+def _parse_number(value):
+    if value is None:
+        return 0.0
+    text = str(value).strip()
+    if not text or text in {"-", "--", "N/A", "nan"}:
+        return 0.0
+    negative = text.startswith("-")
+    text = text.replace(",", "").replace("，", "")
+    match = re.search(r"\d+(?:\.\d+)?", text)
+    if not match:
+        return 0.0
+    num = float(match.group(0))
+    if "万" in text:
+        num *= 10000
+    if "亿" in text:
+        num *= 100000000
+    return -num if negative else num
+
+
+def _download_metric_number(value):
+    if value is None:
+        return 0.0
+    text = str(value).strip()
+    if not text or text in {"-", "--", "N/A", "nan", "None"}:
+        return 0.0
+    multiplier = 1.0
+    if "万" in text:
+        multiplier = 10000.0
+    elif "亿" in text:
+        multiplier = 100000000.0
+    text = (
+        text.replace(",", "")
+        .replace("￥", "")
+        .replace("¥", "")
+        .replace("%", "")
+        .replace("万", "")
+        .replace("亿", "")
+        .strip()
+    )
+    match = re.search(r"-?\d+(?:\.\d+)?", text)
+    if not match:
+        return 0.0
+    return float(match.group(0)) * multiplier
+
+
+def _normalize_header(value):
+    return re.sub(r"\s+", "", str(value or "")).strip()
+
+
+def _read_download_table(path):
+    ext = os.path.splitext(path)[1].lower()
+    if ext == ".csv":
+        for encoding in ("utf-8-sig", "gb18030", "utf-8"):
+            try:
+                with open(path, "r", encoding=encoding, newline="") as f:
+                    return [dict(row, __row_text=" ".join(str(v or "") for v in row.values())) for row in csv.DictReader(f)]
+            except UnicodeDecodeError:
+                continue
+        return []
+
+    if ext in {".xlsx", ".xlsm"}:
+        from openpyxl import load_workbook
+
+        rows: list[tuple] = []
+        try:
+            workbook = load_workbook(path, data_only=True, read_only=True)
+            try:
+                rows = list(workbook.active.iter_rows(values_only=True))
+            finally:
+                workbook.close()
+        except Exception as exc:
+            print(f"[WARN] openpyxl read_only 读取失败，回退到 pandas: {exc}")
+            rows = []
+
+        if not rows or sum(1 for r in rows if any(v not in (None, "") for v in r)) <= 1:
+            try:
+                import pandas as pd
+
+                frame = pd.read_excel(path, header=None).fillna("")
+                rows = [tuple(row) for row in frame.itertuples(index=False, name=None)]
+            except Exception as exc:
+                raise RuntimeError(f"无法解析 xlsx: {path} ({exc})") from exc
+
+        header_index = None
+        for index, values in enumerate(rows[:30]):
+            text = " ".join(str(v or "") for v in values)
+            if "商品" in text and any(key in text for key in ("访客", "加购", "点击")):
+                header_index = index
+                break
+        if header_index is None:
+            header_index = 0
+        raw_headers = list(rows[header_index])
+        headers = []
+        seen = {}
+        for i, value in enumerate(raw_headers):
+            name = _normalize_header(value) or f"col_{i + 1}"
+            seen[name] = seen.get(name, 0) + 1
+            if seen[name] > 1:
+                name = f"{name}_{seen[name]}"
+            headers.append(name)
+        table = []
+        for values in rows[header_index + 1:]:
+            if not any(v not in (None, "") for v in values):
+                continue
+            row = {headers[i]: values[i] if i < len(values) else "" for i in range(len(headers))}
+            row["__row_text"] = " ".join(str(v or "") for v in values)
+            table.append(row)
+        return table
+
+    try:
+        import pandas as pd
+
+        data = pd.read_excel(path)
+        records = data.fillna("").to_dict(orient="records")
+        for row in records:
+            row["__row_text"] = " ".join(str(v or "") for v in row.values())
+        return records
+    except Exception as exc:
+        raise RuntimeError(f"无法解析下载文件: {path} ({exc})") from exc
+
+
+def _value_by_header_alias(row, aliases):
+    for key, value in row.items():
+        if key == "__row_text":
+            continue
+        normalized = _normalize_header(key)
+        if any(alias in normalized for alias in aliases):
+            return value
+    return ""
+
+
+def _click_download_text(page, texts, timeout=10000):
+    for text in texts:
+        try:
+            locator = page.get_by_text(text, exact=False).first
+            locator.wait_for(state="visible", timeout=timeout)
+            locator.click()
+            return True
+        except Exception:
+            pass
+    return page.evaluate(
+        """(texts) => {
+            const visible = (el) => {
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none' && style.visibility !== 'hidden'
+                    && rect.width > 0 && rect.height > 0;
+            };
+            const candidates = Array.from(document.querySelectorAll('button, a, span, div, li'))
+                .filter(visible)
+                .filter((el) => texts.some((text) => (el.textContent || '').includes(text)));
+            const el = candidates[0];
+            if (!el) return false;
+            const rect = el.getBoundingClientRect();
+            el.dispatchEvent(new MouseEvent('mouseover', {bubbles: true}));
+            el.click();
+            return true;
+        }""",
+        texts,
+    )
+
+
+def _hover_download_text(page, texts, timeout=10000):
+    """Hover an element matching the given texts (does not click)."""
+    for text in texts:
+        try:
+            locator = page.get_by_text(text, exact=False).first
+            locator.wait_for(state="visible", timeout=timeout)
+            locator.hover()
+            return True
+        except Exception:
+            pass
+    return page.evaluate(
+        """(texts) => {
+            const visible = (el) => {
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none' && style.visibility !== 'hidden'
+                    && rect.width > 0 && rect.height > 0;
+            };
+            const candidates = Array.from(document.querySelectorAll('button, a, span, div, li'))
+                .filter(visible)
+                .filter((el) => texts.some((text) => (el.textContent || '').includes(text)));
+            const el = candidates[0];
+            if (!el) return false;
+            const rect = el.getBoundingClientRect();
+            const x = rect.left + rect.width / 2;
+            const y = rect.top + rect.height / 2;
+            el.dispatchEvent(new MouseEvent('mouseover', {bubbles: true}));
+            el.dispatchEvent(new MouseEvent('mouseenter', {bubbles: true}));
+            el.dispatchEvent(new MouseEvent('mousemove', {bubbles: true, clientX: x, clientY: y}));
+            return true;
+        }""",
+        texts,
+    )
+
+
+def _download_jd_current_detail(page, output_base):
+    date_stamp = _today()
+    download_dir = os.path.join(output_base, "_downloads", date_stamp)
+    os.makedirs(download_dir, exist_ok=True)
+    print(f"\n[Step] 下载京东商品明细: {date_stamp}")
+    try:
+        page.keyboard.press("Escape")
+    except Exception:
+        pass
+    with page.expect_download(timeout=120000) as download_info:
+        # 1. 点击「下载数据」打开下拉
+        if not _click_download_text(page, ["下载数据"]):
+            raise RuntimeError("未找到京东「下载数据」按钮")
+        page.wait_for_timeout(1000)
+        # 2. 悬浮在「分天下载」上展开子菜单
+        if not _hover_download_text(page, ["分天下载"]):
+            raise RuntimeError("未找到京东「分天下载」菜单")
+        page.wait_for_timeout(1500)
+        # 3. 点击子菜单中的「包括对比数据」触发下载
+        if not _click_download_text(page, ["包括对比数据"]):
+            raise RuntimeError("未找到京东「包括对比数据」子菜单项")
+    download = download_info.value
+    suggested = safe_filename(download.suggested_filename or f"jd_product_detail_{date_stamp}.xlsx", 120)
+    path = os.path.join(download_dir, suggested)
+    download.save_as(path)
+    print(f"  [OK] 京东明细已下载: {path}")
+    return path
+
+
+def _download_jd_with_retry(page, output_base, failed: FailedCollector | None = None) -> str | None:
+    """per-page 重试包装: 京东明细下载失败 → 重新打开下拉 + 重试 1 次, 仍失败记 failed。"""
+    def _do_download():
+        return _download_jd_current_detail(page, output_base)
+    try:
+        return retry_step(_do_download, max_attempts=2, sleep_seconds=3,
+                          label="JD 明细下载", on_retry=lambda *_: _click_download_text(page, ["下载数据"]))
+    except Exception as exc:
+        if failed:
+            failed.add_page("JD 明细下载", f"{type(exc).__name__}: {exc}", attempts=2)
+        return None
+
+
+def _jd_product_metrics_from_download(path, output_base):
+    rows = _read_download_table(path)
+    date_stamp = _today()
+    summaries = []
+    print(f"[Step] 解析京东下载文件，共 {len(rows)} 行")
+    for group in STYLE_SPU_GROUPS:
+        style = group["style"]
+        spus = [str(spu) for spu in group["spus"]]
+        matched_ids = set()
+        matched_rows = []
+        total_visitors = 0.0
+        total_cart = 0.0
+        total_sales_platform = 0.0
+        total_refunds_platform = 0.0
+        for row in rows:
+            row_text = str(row.get("__row_text", ""))
+            matched = [spu for spu in spus if spu and spu in row_text]
+            if not matched:
+                continue
+            matched_rows.append(row)
+            matched_ids.update(matched)
+            total_visitors += _download_metric_number(_value_by_header_alias(row, ["访客数", "商品访客数", "访客人数"]))
+            total_cart += _download_metric_number(_value_by_header_alias(row, ["加购客户数", "加购人数", "商品加购人数", "加购用户数"]))
+            total_sales_platform += _download_metric_number(_value_by_header_alias(row, ["成交单量", "成交订单", "下单单量"]))
+            total_refunds_platform += _download_metric_number(_value_by_header_alias(row, ["申请退货件数", "退货件数"]))
+
+        erp_metrics = apply_erp_override(
+            "jd", style, total_visitors, total_cart,
+            target_date=date_stamp,
+            platform_sales=total_sales_platform,
+            platform_refunds=total_refunds_platform,
+        )
+        metrics = {
+            "销量": erp_metrics["销量"],
+            "访客数": int(round(erp_metrics["访客数"])),
+            "加购客户数": int(round(erp_metrics["加购人数"])),
+            "加购人数": int(round(erp_metrics["加购人数"])),
+            "申请退货件数": erp_metrics["退款订单数"],
+            "退款订单数": erp_metrics["退款订单数"],
+            "转化率": erp_metrics["转化率"],
+            "加购率": erp_metrics["加购率"],
+            "退款率": erp_metrics["退款率"],
+            "ERP销量退款覆盖": erp_metrics["erp_override"],
+        }
+        style_dir_path = _style_dir(output_base, style)
+        os.makedirs(style_dir_path, exist_ok=True)
+        record = {
+            "款式": style,
+            "采集日期": date_stamp,
+            "数据周期": f"{date_stamp} ~ {date_stamp}",
+            "spu列表": spus,
+            "匹配SPU": sorted(matched_ids),
+            "未匹配SPU": [spu for spu in spus if spu not in matched_ids],
+            "指标汇总": metrics,
+            "download_file": _safe_relpath(path, output_base),
+            "matched_row_count": len(matched_rows),
+        }
+        base = safe_filename(f"{style}_{date_stamp}", 120)
+        json_path = os.path.join(style_dir_path, f"{base}.json")
+        md_path = os.path.join(style_dir_path, f"{base}.md")
+        with open(json_path, "w", encoding="utf-8") as f:
+            json.dump(record, f, ensure_ascii=False, indent=2)
+        with open(md_path, "w", encoding="utf-8") as f:
+            f.write(f"# {style} - 京东单品汇总\n\n")
+            f.write(f"- **采集日期**: {date_stamp}\n")
+            f.write(f"- **SPU**: {', '.join(spus)}\n")
+            f.write(f"- **匹配行数**: {len(matched_rows)}\n\n")
+            f.write("| 指标 | 数值 |\n|------|------|\n")
+            for key in ["销量", "访客数", "加购人数", "退款订单数", "转化率", "加购率", "退款率"]:
+                f.write(f"| {key} | {metrics.get(key, '')} |\n")
+        print(f"  [OK] {style}: 访客={metrics['访客数']} 加购={metrics['加购人数']} 匹配SPU={len(matched_ids)}/{len(spus)}")
+        summaries.append(record)
+    return summaries
+
+
+def collect_jd_download_summaries(page, output_base, failed: FailedCollector | None = None):
+    select_week_and_last_week(page)
+    download_path = _download_jd_with_retry(page, output_base, failed=failed)
+    if not download_path:
+        print("[FAIL] 京东明细下载失败 (已重试), 跳过本次款式汇总")
+        return []
+    summaries = _jd_product_metrics_from_download(download_path, output_base)
+    print(f"[OK] 京东下载解析完成，款式汇总 {len(summaries)} 个")
+    return summaries
+
+
+def _metric_from_core(core_data, keywords):
+    for group_items in (core_data or {}).values():
+        if not isinstance(group_items, dict):
+            continue
+        for title, item in group_items.items():
+            if any(keyword in str(title) for keyword in keywords):
+                if isinstance(item, dict):
+                    return _parse_number(item.get("value"))
+                return _parse_number(item)
+    return 0.0
+
+
+def _rate(numerator, denominator):
+    return (float(numerator) / float(denominator)) if denominator else 0.0
+
+
+def _percent(value):
+    return f"{value * 100:.2f}%"
+
+
+def _build_spu_metrics(core_data):
+    sales = _metric_from_core(core_data, ["成交单量", "成交订单", "下单单量"])
+    visitors = _metric_from_core(core_data, ["访客数"])
+    add_cart_customers = _metric_from_core(core_data, ["加购客户数"])
+    refund_items = _metric_from_core(core_data, ["申请退货件数", "退货件数"])
+    return {
+        "销量": int(sales),
+        "访客数": int(visitors),
+        "加购客户数": int(add_cart_customers),
+        "申请退货件数": int(refund_items),
+        "转化率": _percent(_rate(sales, visitors)),
+        "加购率": _percent(_rate(add_cart_customers, visitors)),
+        "退款率": _percent(_rate(refund_items, sales)),
+    }
+
+
+def _aggregate_style_metrics(spu_results):
+    sales = sum(_parse_number(r.get("指标汇总", {}).get("销量")) for r in spu_results)
+    visitors = sum(_parse_number(r.get("指标汇总", {}).get("访客数")) for r in spu_results)
+    add_cart_customers = sum(_parse_number(r.get("指标汇总", {}).get("加购客户数")) for r in spu_results)
+    refund_items = sum(_parse_number(r.get("指标汇总", {}).get("申请退货件数")) for r in spu_results)
+    return {
+        "销量": int(sales),
+        "访客数": int(visitors),
+        "加购客户数": int(add_cart_customers),
+        "申请退货件数": int(refund_items),
+        "转化率": _percent(_rate(sales, visitors)),
+        "加购率": _percent(_rate(add_cart_customers, visitors)),
+        "退款率": _percent(_rate(refund_items, sales)),
+    }
+
+
+def _today():
+    return get_yesterday_date().strftime("%Y-%m-%d")
+
+
+def _style_dir(output_base, style):
+    return os.path.join(output_base, f"{safe_filename(style)}_{_today()}")
+
+
+def _spu_dir(style_dir, product, spu):
+    name = safe_filename(product or spu, 60)
+    return os.path.join(style_dir, f"{name}_{spu}_{_today()}")
+
+
+def _find_existing_spu_result(style_dir, spu):
+    if not os.path.isdir(style_dir):
+        return None
+    pattern = f"_{spu}_{_today()}"
+    for name in os.listdir(style_dir):
+        subdir = os.path.join(style_dir, name)
+        if not os.path.isdir(subdir) or pattern not in name:
+            continue
+        for file_name in os.listdir(subdir):
+            if file_name.endswith(".json"):
+                path = os.path.join(subdir, file_name)
+                try:
+                    with open(path, "r", encoding="utf-8") as f:
+                        return json.load(f)
+                except Exception:
+                    return None
+    return None
+
+
+def _write_spu_files(record, spu_dir_path):
+    os.makedirs(spu_dir_path, exist_ok=True)
+    file_base = safe_filename(f"{record.get('商品名')}_{record.get('spu')}_{_today()}", 120)
+    json_path = os.path.join(spu_dir_path, f"{file_base}.json")
+    md_path = os.path.join(spu_dir_path, f"{file_base}.md")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(record, f, ensure_ascii=False, indent=2)
+
+    metrics = record.get("指标汇总", {})
+    screenshot = record.get("人物画像截图", "")
+    md = [
+        f"# {record.get('商品名') or record.get('spu')} - 单品数据",
+        "",
+        f"**款式：** {record.get('款式', '')}",
+        f"**SPU：** {record.get('spu', '')}",
+        f"**采集日期：** {record.get('采集日期', '')}",
+        f"**数据周期：** {record.get('数据周期', '')}",
+        "",
+        "## 指标汇总",
+        "",
+        "| 指标 | 数值 |",
+        "|------|------|",
+        f"| 销量 | {metrics.get('销量', 0)} |",
+        f"| 访客数 | {metrics.get('访客数', 0)} |",
+        f"| 转化率 | {metrics.get('转化率', '0.00%')} |",
+        f"| 加购率 | {metrics.get('加购率', '0.00%')} |",
+        f"| 退款率 | {metrics.get('退款率', '0.00%')} |",
+        "",
+    ]
+    if screenshot:
+        md.extend(["## 人物画像截图", "", f"![人物画像]({os.path.basename(screenshot)})", ""])
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md))
+    print(f"    [OK] SPU 数据已保存: {spu_dir_path}")
+    return json_path, md_path
+
+
+def _write_style_summary(style, style_dir_path, spu_results):
+    os.makedirs(style_dir_path, exist_ok=True)
+    start_day, end_day = get_yesterday_range_text()
+    metrics = _aggregate_style_metrics(spu_results)
+    erp_metrics = apply_erp_override(
+        "jd",
+        style,
+        metrics.get("访客数", 0),
+        metrics.get("加购客户数", 0),
+        target_date=_today(),
+        platform_sales=metrics.get("销量", 0),
+        platform_refunds=metrics.get("申请退货件数", 0),
+    )
+    metrics.update({
+        "销量": erp_metrics["销量"],
+        "访客数": erp_metrics["访客数"],
+        "加购客户数": erp_metrics["加购人数"],
+        "申请退货件数": erp_metrics["退款订单数"],
+        "退款订单数": erp_metrics["退款订单数"],
+        "转化率": erp_metrics["转化率"],
+        "加购率": erp_metrics["加购率"],
+        "退款率": erp_metrics["退款率"],
+        "ERP销量退款覆盖": erp_metrics["erp_override"],
+    })
+    file_base = safe_filename(f"{style}_{_today()}", 120)
+    record = {
+        "款式": style,
+        "采集日期": _today(),
+        "数据周期": f"{start_day} ~ {end_day}",
+        "spu列表": [r.get("spu", "") for r in spu_results],
+        "指标汇总": metrics,
+        "人物画像截图": [
+            _safe_relpath(r.get("人物画像截图", ""), style_dir_path)
+            for r in spu_results if r.get("人物画像截图")
+        ],
+        "spu明细": [
+            {
+                "spu": r.get("spu", ""),
+                "商品名": r.get("商品名", ""),
+                "指标汇总": r.get("指标汇总", {}),
+                "目录": _safe_relpath(r.get("输出目录", ""), style_dir_path) if r.get("输出目录") else "",
+            }
+            for r in spu_results
+        ],
+    }
+    json_path = os.path.join(style_dir_path, f"{file_base}.json")
+    md_path = os.path.join(style_dir_path, f"{file_base}.md")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(record, f, ensure_ascii=False, indent=2)
+
+    md = [
+        f"# {style} - 款式汇总",
+        "",
+        f"**采集日期：** {_today()}",
+        f"**数据周期：** {start_day} ~ {end_day}",
+        "",
+        "## 指标汇总",
+        "",
+        "| 指标 | 数值 |",
+        "|------|------|",
+        f"| 销量 | {metrics.get('销量', 0)} |",
+        f"| 访客数 | {metrics.get('访客数', 0)} |",
+        f"| 转化率 | {metrics.get('转化率', '0.00%')} |",
+        f"| 加购率 | {metrics.get('加购率', '0.00%')} |",
+        f"| 退款率 | {metrics.get('退款率', '0.00%')} |",
+        "",
+        "## SPU 明细",
+        "",
+        "| SPU | 商品名 | 销量 | 访客数 | 转化率 | 加购率 | 退款率 |",
+        "|-----|--------|------|--------|--------|--------|--------|",
+    ]
+    for item in record["spu明细"]:
+        m = item.get("指标汇总", {})
+        md.append(
+            f"| {item.get('spu', '')} | {item.get('商品名', '')} | {m.get('销量', 0)} | "
+            f"{m.get('访客数', 0)} | {m.get('转化率', '0.00%')} | "
+            f"{m.get('加购率', '0.00%')} | {m.get('退款率', '0.00%')} |"
+        )
+    screenshots = record["人物画像截图"]
+    if screenshots:
+        md.extend(["", "## 人物画像截图", ""])
+        for path in screenshots:
+            md.append(f"![人物画像]({path})")
+            md.append("")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md))
+    print(f"  [OK] 款式汇总已保存: {style_dir_path}")
+    return json_path, md_path
+
+
+def _normalize_rank_record(record):
+    if not record:
+        return {}
+    normalized = {}
+    for key, value in record.items():
+        key_text = str(key)
+        if "金额" in key_text:
+            normalized["成交金额"] = value
+        elif "单量" in key_text:
+            normalized["成交单量"] = value
+        elif "访客" in key_text:
+            normalized["访客数"] = value
+        elif "搜索" in key_text:
+            normalized["搜索点击次数"] = value
+        elif "关注" in key_text:
+            normalized["关注人数"] = value
+        else:
+            normalized[key_text] = value
+    return normalized
+
+
+def collect_rank_records_for_single_spu(context, product):
+    if not product:
+        return []
+    print("  采集商品榜单排名...")
+    rank_page = context.new_page()
+    rank_records = []
+    try:
+        rank_page.goto("https://sz.jd.com/szweb/sz/view/industryMarket/productRanks.html", wait_until="domcontentloaded", timeout=30000)
+        rank_page.wait_for_timeout(5000)
+        select_week_and_last_week_on_rank_page(rank_page)
+        search_product_on_rank_page(rank_page, product)
+        categories = [
+            {"path": ["男包"]},
+            {"path": ["男包", "男士双肩包"]},
+            {"path": ["男包", "男士单肩/斜挎包"]},
+            {"path": ["功能箱包"]},
+            {"path": ["功能箱包", "电脑包"]},
+        ]
+        for cat in categories:
+            cat_label = " > ".join(cat["path"])
+            try:
+                select_category_on_rank_page(rank_page, cat["path"])
+                rank_data = extract_rank_table_data(rank_page, product)
+                if rank_data:
+                    rank_records.append({"类目": cat_label, **_normalize_rank_record(rank_data)})
+                    print(f"    [OK] {cat_label} 排名: {rank_data.get('rank', '')}")
+            except Exception as e:
+                print(f"    [WARN] 榜单类目采集失败 {cat_label}: {e}")
+    except Exception as e:
+        print(f"    [WARN] 榜单排名采集失败: {e}")
+    finally:
+        try:
+            rank_page.close()
+        except Exception:
+            pass
+    return rank_records
+
+
+def _spu_info_from_current_rows(page, fallback):
+    rows = extract_product_detail_rows(page)
+    target_spu = str(fallback.get("spu", ""))
+    for row in rows:
+        if str(row.get("spu", "")) == target_spu:
+            return {**fallback, **row}
+    if rows:
+        return {**fallback, **rows[0]}
+    return fallback
+
+
+def collect_each_spu_detail(page, context, all_spus, output_base, resume=False):
+    """按款式固定清单搜索 SPU，保存单 SPU 数据并生成款式汇总。"""
+    start_day, end_day = get_yesterday_range_text()
+    period = f"{start_day} ~ {end_day}"
+    print(f"\n[Step 6] 按固定款式采集 SPU 详情，数据周期: {period}")
+    progress = SpuProgress(output_base, "product_style_spu_detail")
+    task_map = {(task["style"], task["spu"]): task for task in all_spus}
+    success = 0
+    failed = 0
+
+    for style_group in STYLE_SPU_GROUPS:
+        style = style_group["style"]
+        style_dir_path = _style_dir(output_base, style)
+        os.makedirs(style_dir_path, exist_ok=True)
+        style_results = []
+        print(f"\n{'=' * 60}")
+        print(f"款式: {style} | SPU {len(style_group['spus'])} 个")
+        print(f"{'=' * 60}")
+
+        for index, spu in enumerate(style_group["spus"], 1):
+            spu = str(spu)
+            spu_info = task_map.get((style, spu), {"style": style, "spu": spu})
+            if resume:
+                existing = _find_existing_spu_result(style_dir_path, spu)
+                if existing:
+                    print(f"  [Resume] 跳过已采集 SPU: {spu}")
+                    style_results.append(existing)
+                    continue
+
+            print(f"\n  [{index}/{len(style_group['spus'])}] SPU: {spu}")
+            progress.mark_running(spu_info, index, len(style_group["spus"]))
+            detail_page = None
+            try:
+                search_spu_in_detail_page(page, spu)
+                spu_info = _spu_info_from_current_rows(page, spu_info)
+                product = spu_info.get("product") or spu
+                detail_page = click_detail_and_wait_new_page(page, context, spu)
+                if not detail_page:
+                    raise RuntimeError("详情页未打开")
+
+                click_core_data_tab(detail_page)
+                select_last_week_on_detail_page(detail_page)
+                core_data = extract_core_data(detail_page)
+
+                click_traffic_source_tab(detail_page)
+                expand_all_traffic_rows(detail_page)
+                traffic_data = extract_traffic_data(detail_page)
+                keyword_data = extract_keyword_data(detail_page)
+
+                click_crowd_portrait_tab(detail_page)
+                crowd_data = extract_crowd_portrait_data(detail_page)
+
+                metrics = _build_spu_metrics(core_data)
+                spu_dir_path = _spu_dir(style_dir_path, product, spu)
+                os.makedirs(spu_dir_path, exist_ok=True)
+                crowd_screenshot = os.path.join(spu_dir_path, "人物画像截图.png")
+                detail_page.screenshot(path=crowd_screenshot, full_page=True)
+                print(f"    [OK] 人物画像截图已保存: {crowd_screenshot}")
+
+                try:
+                    detail_page.close()
+                except Exception:
+                    pass
+                detail_page = None
+                page.wait_for_timeout(1500)
+
+                record = {
+                    "款式": style,
+                    "spu": spu,
+                    "商品名": product,
+                    "商品图片": spu_info.get("image", ""),
+                    "商品链接": spu_info.get("url", ""),
+                    "采集日期": _today(),
+                    "数据周期": period,
+                    "指标汇总": metrics,
+                    "核心数据": core_data,
+                    "流量来源": traffic_data,
+                    "关键词来源成交": keyword_data,
+                    "人群画像": crowd_data,
+                    "人物画像截图": crowd_screenshot,
+                    "输出目录": spu_dir_path,
+                }
+                _write_spu_files(record, spu_dir_path)
+                style_results.append(record)
+                success += 1
+                progress.mark_success(spu_info, group=style)
+            except Exception as e:
+                failed += 1
+                print(f"  [FAIL] SPU {spu} 采集失败: {e}")
+                progress.mark_failed(spu_info, e)
+                try:
+                    if detail_page:
+                        detail_page.close()
+                except Exception:
+                    pass
+
+        if style_results:
+            _write_style_summary(style, style_dir_path, style_results)
+        else:
+            print(f"  [WARN] 款式 {style} 没有成功采集的 SPU，跳过汇总")
+
+    print(f"\n[OK] 固定款式 SPU 采集完成: 成功 {success}, 失败 {failed}")
+    print(f"  [Progress] 进度日志: {progress.path}")
+    return success, failed
+
+
+def main():
+    global TARGET_DATE, DEBUG_DATE, STYLE_SPU_GROUPS
+    parser = argparse.ArgumentParser(description="京东商品明细+单品详情数据采集脚本")
+    parser.add_argument("--shop", default=DEFAULT_CONFIG["shop"], help="店铺账号名")
+    parser.add_argument("--password", default=DEFAULT_CONFIG["password"], help="登录密码")
+    parser.add_argument("--output", default=OUTPUT_BASE, help="输出目录")
+    parser.add_argument("--resume", action="store_true", help="跳过已成功采集的SPU，从断点继续")
+    parser.add_argument("--target-date", type=parse_target_date, default=None, help="采集指定日期，格式 YYYY-MM-DD；默认昨天")
+    parser.add_argument("--debug-date", action="store_true", help="日期面板操作时输出 debug 截图+JSON 到 debug/jd_date_picker/")
+    parser.add_argument("--refresh-config", action="store_true",
+                        help="忽略 24h 缓存, 强制从飞书多维表格重读款式配置")
+    parser.add_argument("--reviews-only", action="store_true",
+                        help="只跑商品评价采集 (好/中/差全采), 跳过商智单品明细/SPU 详情")
+    args = parser.parse_args()
+    args.output = str(managed_data_path(args.output))
+    TARGET_DATE = args.target_date
+    DEBUG_DATE = args.debug_date
+    if args.refresh_config:
+        StyleConfigLoader().load(force_refresh=True)
+        STYLE_SPU_GROUPS = _get_style_spu_groups()
+        _print_preflight()
+        print("[INFO] --refresh-config: 已从飞书多维表格重读款式配置")
+    if DEBUG_DATE:
+        print("[debug-date] 已开启,日历操作每步会写 debug/jd_date_picker/<时间戳>_<阶段>.{json,png}")
+
+    print("=" * 60)
+    print("京东商品明细 + 单品详情数据采集")
+    print(f"  店铺: {args.shop}")
+    print(f"  输出: {args.output}")
+    print("=" * 60)
+
+    start_day, end_day = get_yesterday_range_text()
+    print(f"  昨天: {start_day} ~ {end_day}")
+    print()
+
+    with sync_playwright() as p:
+        context = p.chromium.launch_persistent_context(
+            user_data_dir=JD_PROFILE_DIR,
+            headless=False,
+            viewport={"width": 1920, "height": 1080},
+            locale="zh-CN",
+            args=["--disable-gpu"],
+        )
+        page = context.new_page()
+
+        try:
+            # Step 1: 登录（profile 复用，命中已登录态则跳过 step_login）
+            page.goto("https://shop.jd.com/jdm/home", wait_until="domcontentloaded", timeout=60000)
+            page.wait_for_timeout(2000)
+            if "shop.jd.com" in page.url and "login" not in page.url.lower():
+                print("[OK] 检测到京东已登录态，跳过 step_login")
+            else:
+                step_login(page, args.shop, args.password)
+
+            # 商品评价采集 (好/中/差全采, 默认开启; --reviews-only 时只跑这一段)
+            print("\n" + "=" * 60)
+            print("[Step] 采集商品评价 (好评/中评/差评)")
+            print("=" * 60)
+            appraise_page = None
+            try:
+                appraise_page = navigate_to_goods_appraise(page)
+                select_last_week_date_range_appraise(appraise_page)
+                click_query_button_appraise(appraise_page)
+                reviews_by_rating = {}
+                for rating in ["好评", "中评", "差评"]:
+                    click_rating_filter(appraise_page, rating)
+                    reviews_by_rating[rating] = collect_all_reviews(appraise_page, rating)
+                save_reviews_output(reviews_by_rating, args.output)
+            except Exception as e:
+                print(f"  [WARN] 商品评价采集失败，跳过继续: {e}")
+            finally:
+                if appraise_page and appraise_page != page and not appraise_page.is_closed():
+                    try:
+                        appraise_page.close()
+                    except Exception:
+                        pass
+
+            if args.reviews_only:
+                print("\n" + "=" * 60)
+                print("[DONE] --reviews-only 模式: 跳过商智单品明细采集")
+                print("=" * 60)
+                return
+
+            alive_pages = [p for p in context.pages if not p.is_closed()]
+            page = next((p for p in alive_pages if "shop.jd.com" in p.url), None) or (alive_pages[0] if alive_pages else context.new_page())
+
+            # Step 3~5: 商智 → 返回旧版 → 商品明细页
+            print("\n" + "=" * 60)
+            print("继续进入商智采集单品详情数据")
+            print("=" * 60)
+            if page.is_closed():
+                print("  [INFO] 原商家后台页面已关闭，重新打开页面...")
+                page = context.new_page()
+            if "shop.jd.com" not in page.url:
+                try:
+                    page.goto("https://shop.jd.com/jdm/home", wait_until="domcontentloaded", timeout=60000)
+                except Exception:
+                    page = context.new_page()
+                    page.goto("https://shop.jd.com/jdm/home", wait_until="domcontentloaded", timeout=60000)
+                page.wait_for_timeout(3000)
+            sz_page = step_enter_shangzhi(context, page)
+            step_return_old_version(sz_page)
+
+            # Step 6: 商品明细页
+            step_goto_product_detail(sz_page)
+            close_guide_if_present(sz_page)
+
+            # Step 7: 下载商品明细并解析，不再进入单品详情/排名/人群画像。
+            failed_collector = FailedCollector("jd", _today())
+            collect_jd_download_summaries(sz_page, args.output, failed=failed_collector)
+            failed_path = failed_collector.write()
+            if failed_path:
+                print(f"[FAIL_LOG] {failed_collector.summary_line()} -> {failed_path}")
+            else:
+                print(f"[OK_LOG] JD 全成功, 无失败清单")
+
+            print()
+            print("=" * 60)
+            print("全部采集完成！")
+            detail_output = args.output
+            print(f"  单品数据: {detail_output}")
+            print("=" * 60)
+
+        except KeyboardInterrupt:
+            print("\n用户中断，退出。")
+        except Exception as e:
+            print(f"\n[ERROR] 出错: {e}")
+            import traceback
+            traceback.print_exc()
+        finally:
+            context.close()
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/jd_self_inventory_sales_collector.py b/src/gyxx_flow/modules/product_commerce/runtime/jd_self_inventory_sales_collector.py
new file mode 100644
index 0000000..dd8e3e1
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/jd_self_inventory_sales_collector.py
@@ -0,0 +1,708 @@
+# -*- coding: utf-8 -*-
+"""
+Collect JD self-operated supply-chain outbound quantity by style.
+
+This is a standalone collector for:
+https://ppzh.jd.com/inventoryweb/brand/view/supplychainAnalysis/inventoryDetail.html
+
+It reuses the existing JD login / popup handling helpers from vendors/jd-data-flow,
+then queries each SKU and sums "出库商品件数" to style-level sales.
+"""
+
+import argparse
+import json
+import os
+import re
+import sys
+from datetime import datetime, timedelta
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT, TMP_ROOT, managed_data_path, vendor_source_root
+from typing import Dict, List, Optional
+
+from playwright.sync_api import TimeoutError as PlaywrightTimeoutError
+from playwright.sync_api import sync_playwright
+
+from config.style_config_loader import get_jd_self_sku_groups, get_jd_self_styles
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+SOURCE_ROOT = vendor_source_root("jd")
+if SOURCE_ROOT.is_dir() and str(SOURCE_ROOT) not in sys.path:
+    sys.path.insert(0, str(SOURCE_ROOT))
+
+from jd_data_collector import DEFAULT_CONFIG, close_popup_if_present, step_enter_shangzhi, step_login  # noqa: E402
+from jd_peer_product_data_collector import (  # noqa: E402
+    click_week_day,
+    click_week_switch,
+    wait_for_visible_date_dropdown,
+    wait_for_week_mode,
+)
+
+
+INVENTORY_URL = "https://ppzh.jd.com/inventoryweb/brand/view/supplychainAnalysis/inventoryDetail.html"
+DEFAULT_OUTPUT_DIR = RAW_DATA_ROOT / "jd自营库存销量数据"
+
+# 保留硬编码作为灾备回退；运行时优先从飞书多维表格读取。
+_STATIC_STYLE_SKU_GROUPS = [
+    {"style_name": "盖亚斜挎", "skus": ["100044949353", "100087299891"]},
+    {"style_name": "极星pro", "skus": ["100117005088", "100100271235"]},
+    {"style_name": "盖亚微单", "skus": ["100108144308", "100108144324"]},
+    {"style_name": "盖世m1", "skus": ["100191634938"]},
+    {
+        "style_name": "星迹2",
+        "skus": [
+            "100108144306",
+            "100150785348",
+            "100108144260",
+            "100100271371",
+            "100241562067",
+            "100241562047",
+        ],
+    },
+    {"style_name": "星云2", "skus": ["100160619963", "100160619965", "100248422862"]},
+    {
+        "style_name": "宙斯",
+        "skus": ["100221530889", "100306128160", "100306128166", "100255125975", "100239269869"],
+    },
+    {"style_name": "阿波罗x1", "skus": ["100058323820", "100058323830", "100058916832", "100058916818"]},
+    {"style_name": "逐星", "skus": ["100165451136", "100165451142"]},
+    {"style_name": "星云mini", "skus": ["100215176791", "100215176793"]},
+    {"style_name": "晨星2", "skus": ["100287829888", "100287829866"]},
+    {"style_name": "极星双肩", "skus": ["100247532920", "100247532928"]},
+    {"style_name": "凌云air", "skus": ["100291530148", "100218450063"]},
+    {"style_name": "觅光", "skus": ["100241563697"]},
+    {"style_name": "云卷2", "skus": ["100329997886"]},
+    {"style_name": "拓界mode", "skus": ["100251533359", "100251533369"]},
+]
+
+
+def _get_style_sku_groups():
+    """优先从飞书读取京东自营 SKU 分组，失败则回退硬编码。"""
+    try:
+        groups = get_jd_self_sku_groups()
+        if groups:
+            return groups
+    except Exception as exc:
+        print(f"[WARN] 动态读取京东自营 SKU 配置失败，使用硬编码回退: {exc}")
+    return _STATIC_STYLE_SKU_GROUPS
+
+
+STYLE_SKU_GROUPS = _get_style_sku_groups()
+
+
+def _print_preflight() -> None:
+    """启动期轻量打印:京东自营周汇总分组里今日待采款式数。"""
+    try:
+        styles = get_jd_self_styles()
+        print(f"[JD_SELF] 飞书表格 周汇总待采 {len(styles)} 款 (仅要求 SKU 非空)")
+    except Exception as exc:
+        print(f"[WARN] 飞书读取失败, 跳过 preflight: {exc}")
+
+
+_print_preflight()
+
+
+def safe_filename(value: str) -> str:
+    return re.sub(r'[\\/:*?"<>|\s]+', "_", value).strip("_") or "unknown"
+
+
+def parse_number(text: Optional[str]) -> int:
+    if not text:
+        return 0
+    normalized = str(text).replace(",", "").strip()
+    multiplier = 1
+    if "万" in normalized:
+        multiplier = 10000
+    match = re.search(r"-?\d+(?:\.\d+)?", normalized)
+    if not match:
+        return 0
+    return int(round(float(match.group(0)) * multiplier))
+
+
+def yesterday_iso() -> str:
+    return (datetime.now() - timedelta(days=1)).strftime("%Y-%m-%d")
+
+
+def last_full_week_range() -> tuple[str, str]:
+    today = datetime.now()
+    this_monday = today - timedelta(days=today.weekday())
+    last_monday = this_monday - timedelta(days=7)
+    last_sunday = this_monday - timedelta(days=1)
+    return last_monday.strftime("%Y-%m-%d"), last_sunday.strftime("%Y-%m-%d")
+
+
+def validate_date_text(value: str, arg_name: str) -> str:
+    try:
+        datetime.strptime(value, "%Y-%m-%d")
+    except ValueError as exc:
+        raise ValueError(f"{arg_name} 必须是 YYYY-MM-DD 格式: {value}") from exc
+    return value
+
+
+def period_label(start_date: str, end_date: str) -> str:
+    return f"{start_date}_{end_date}"
+
+
+def ensure_popup_closed(page) -> None:
+    try:
+        close_popup_if_present(page)
+    except Exception:
+        pass
+    try:
+        page.keyboard.press("Escape")
+    except Exception:
+        pass
+
+
+def save_debug_snapshot(page, label: str) -> None:
+    debug_dir = TMP_ROOT / "debug" / "jd_self_inventory"
+    debug_dir.mkdir(parents=True, exist_ok=True)
+    stamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    safe_label = safe_filename(label)
+    screenshot_path = debug_dir / f"{stamp}_{safe_label}.png"
+    html_path = debug_dir / f"{stamp}_{safe_label}.html"
+    meta_path = debug_dir / f"{stamp}_{safe_label}_meta.json"
+    try:
+        page.screenshot(path=str(screenshot_path), full_page=True)
+    except Exception:
+        screenshot_path = None
+    try:
+        html_path.write_text(page.content(), encoding="utf-8")
+    except Exception:
+        html_path = None
+    try:
+        meta = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                return {
+                    url: location.href,
+                    title: document.title,
+                    bodyText: (document.body?.innerText || '').slice(0, 3000),
+                    inputs: Array.from(document.querySelectorAll('input')).map((input, index) => ({
+                        index,
+                        value: input.value || '',
+                        placeholder: input.getAttribute('placeholder') || '',
+                        type: input.getAttribute('type') || '',
+                        visible: visible(input),
+                        outerHTML: input.outerHTML.slice(0, 500),
+                    })),
+                    buttons: Array.from(document.querySelectorAll('button')).slice(0, 40).map((button, index) => ({
+                        index,
+                        text: (button.innerText || button.textContent || '').trim(),
+                        visible: visible(button),
+                    })),
+                    frames: Array.from(document.querySelectorAll('iframe')).map((frame, index) => ({
+                        index,
+                        src: frame.src,
+                        visible: visible(frame),
+                    })),
+                };
+            }"""
+        )
+        meta_path.write_text(json.dumps(meta, ensure_ascii=False, indent=2), encoding="utf-8")
+    except Exception:
+        meta_path = None
+    print(f"  [DEBUG] 页面现场: {screenshot_path or 'no screenshot'} {html_path or 'no html'} {meta_path or 'no meta'}")
+
+
+def return_old_version_if_present(page) -> None:
+    print("[Step] 返回旧版入口检查...")
+    ensure_popup_closed(page)
+    try:
+        back_button = page.locator("#header > div > div.page-top-header-tools > img")
+        back_button.wait_for(state="visible", timeout=5000)
+        back_button.click()
+        page.wait_for_timeout(5000)
+        try:
+            page.wait_for_load_state("domcontentloaded", timeout=60000)
+        except Exception:
+            pass
+        print("  已点击返回旧版")
+    except Exception:
+        print("  未发现返回旧版入口，继续当前页面")
+
+
+def click_inventory_week_switch(page) -> str:
+    """在自营库存明细页切到「按周」tab（与商智其它页「按周查询」不同，文本是「按周」）。
+
+    真实鼠标点击避免 jmtd 组件在 JS click 后仍停留在按天模式。
+    """
+    try:
+        dropdown = wait_for_visible_date_dropdown(page)
+        if not dropdown:
+            return "no_popup"
+        # 自营页 tab 文本是「按周」，不是「按周查询」。枚举四个 tab 找匹配。
+        items = dropdown.locator(".jmtd-date-picker-switch-panel-item")
+        if items.count() == 0:
+            return "no_week_switch"
+        target_idx = -1
+        for i in range(items.count()):
+            txt = (items.nth(i).inner_text() or "").strip()
+            if txt == "按周":
+                target_idx = i
+                break
+        if target_idx < 0:
+            return "no_week_switch"
+        item = items.nth(target_idx)
+        current = item.evaluate(
+            "el => el.classList.contains('jmtd-date-picker-switch-panel-current') "
+            "|| el.classList.contains('jmtd-date-picker-switch-panel-selected')"
+        )
+        if current:
+            return "already_week"
+        item.scroll_into_view_if_needed(timeout=3000)
+        box = item.bounding_box(timeout=3000)
+        if not box:
+            return "no_week_box"
+        page.mouse.move(box["x"] + box["width"] / 2, box["y"] + box["height"] / 2)
+        page.wait_for_timeout(100)
+        page.mouse.down()
+        page.wait_for_timeout(80)
+        page.mouse.up()
+        page.wait_for_timeout(600)
+        return "switched_week"
+    except Exception as exc:
+        return f"switch_error: {exc}"
+
+
+def wait_inventory_week_mode(page, timeout_ms: int = 5000) -> bool:
+    """等待自营库存明细页日期弹层切到「按周」模式（data-event-name="week"）。"""
+    import time as _t
+    deadline = _t.time() + timeout_ms / 1000
+    while _t.time() < deadline:
+        try:
+            ok = page.evaluate(
+                """() => {
+                    const panels = Array.from(document.querySelectorAll('.jmtd-dropdown-panel'));
+                    for (const p of panels) {
+                        const cs = getComputedStyle(p);
+                        if (cs.display === 'none' || cs.visibility === 'hidden') continue;
+                        const calendar = p.querySelector('.jmtd-date-picker-calendar');
+                        if (!calendar) continue;
+                        const name = p.getAttribute('data-event-name') || '';
+                        if (name === 'week') return true;
+                    }
+                    return false;
+                }"""
+            )
+            if ok:
+                return True
+        except Exception:
+            pass
+        page.wait_for_timeout(300)
+    return False
+
+
+def get_target_week_label(start_date: str) -> str:
+    """把 start_date 转成 '2026年23周'（ISO 周号）用于 picker label 校验。"""
+    target = datetime.strptime(start_date, "%Y-%m-%d")
+    iso = target.isocalendar()
+    return f"{iso[0]}年{iso[1]}周"
+
+
+def week_already_applied(page, start_date: str, end_date: str) -> bool:
+    """检查 body_text 里是否已含目标周（两种显示格式都接受）：
+    - 范围式: 2026-06-01 ~ 2026-06-07
+    - ISO 周号: 2026年23周
+    """
+    body_text = page.locator("body").inner_text(timeout=10000)
+    range_text = f"{start_date} ~ {end_date}"
+    week_label = get_target_week_label(start_date)
+    if range_text in body_text or range_text.replace(" ", "") in body_text.replace(" ", ""):
+        return True
+    if week_label in body_text:
+        return True
+    return False
+
+
+def set_inventory_week(page, start_date: str, end_date: str) -> None:
+    """在京东自营供应链库存明细页把日期切到 [start_date, end_date]（按周模式）。
+
+    复用 jd_peer_product_data_collector 里的 jmtd 工具，但 tab 文本和 label 格式不同：
+      - 商智其它页 tab：「按周查询」  →  click_week_switch / wait_for_week_mode
+      - 自营库存页 tab：「按周」       →  click_inventory_week_switch / wait_inventory_week_mode
+      - 自营库存页 label：'2026年23周' (ISO 周号)
+    其余（jmtd cell/header 翻月/点击）共用。
+    """
+    range_text = f"{start_date} ~ {end_date}"
+    week_label = get_target_week_label(start_date)
+    print(f"  选择周范围: {range_text} (label 应为: {week_label})")
+
+    # 0. 先看 body 里是否已经是目标周
+    if week_already_applied(page, start_date, end_date):
+        print("  页面已是目标周范围")
+        return
+
+    # 1. 点开日期选择器（自营页只有唯一一个 .jmtd-date-picker）
+    try:
+        picker = page.locator('.jmtd-date-picker .jmtd-base-input-top, .jmtd-date-picker').first
+        picker.wait_for(state="visible", timeout=5000)
+        picker.click()
+    except Exception as exc:
+        save_debug_snapshot(page, "date_picker_not_found")
+        raise RuntimeError(f"未找到日期选择器入口: {exc}")
+    page.wait_for_timeout(500)
+
+    # 2. 等待弹层
+    dropdown = wait_for_visible_date_dropdown(page)
+    if not dropdown:
+        save_debug_snapshot(page, "date_dropdown_not_found")
+        raise RuntimeError("日期弹层未出现")
+
+    # 3. 切到「按周」tab
+    week_result = click_inventory_week_switch(page)
+    print(f"  切到按周: {week_result}")
+    if week_result == "no_week_switch":
+        page.keyboard.press("Escape")
+        save_debug_snapshot(page, "no_week_switch")
+        raise RuntimeError("日期弹层未找到「按周」切换项")
+    if week_result.startswith("switch_error"):
+        save_debug_snapshot(page, "week_switch_error")
+        raise RuntimeError(f"切到按周出错: {week_result}")
+
+    if not wait_inventory_week_mode(page):
+        print("  [WARN] 未确认切到按周模式，继续点日期")
+
+    # 4. 选周起始日（任意一周内的某天都行，但点 start_date 保险）
+    target = datetime.strptime(start_date, "%Y-%m-%d")
+    click_result = click_week_day(page, target.year, target.month, target.day)
+    print(f"  选周点击: {click_result}")
+    if not str(click_result).startswith("clicked"):
+        save_debug_snapshot(page, "week_day_click_failed")
+        raise RuntimeError(f"日期控件未能选中周起始日 {start_date}: {click_result}")
+
+    # 5. 校验 body_text 已含目标周（接受范围式 或 ISO 周号）
+    page.wait_for_timeout(1500)
+    if not week_already_applied(page, start_date, end_date):
+        save_debug_snapshot(page, "week_range_not_applied")
+        raise RuntimeError(f"周范围未生效，期望 {range_text} 或 {week_label}")
+    print("  周范围已生效")
+
+
+def set_sku_and_query(page, sku: str) -> None:
+    sku_input = page.locator('input[placeholder*="SKU"]').first
+    try:
+        sku_input.wait_for(state="visible", timeout=3000)
+    except Exception:
+        sku_index = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const inputs = Array.from(document.querySelectorAll('input')).filter(visible);
+                const skuLabel = Array.from(document.querySelectorAll('body *')).find((el) => {
+                    const text = (el.textContent || '').trim();
+                    return text === 'SKU' && visible(el);
+                });
+                if (skuLabel) {
+                    const labelRect = skuLabel.getBoundingClientRect();
+                    const candidate = inputs
+                        .map((input, index) => ({input, index, rect: input.getBoundingClientRect()}))
+                        .filter(({rect}) => rect.left > labelRect.right && Math.abs(rect.top - labelRect.top) < 80)
+                        .sort((a, b) => (a.rect.left - b.rect.left) || (a.rect.top - b.rect.top))[0];
+                    if (candidate) return candidate.index;
+                }
+                return inputs.findIndex((input) => input.getBoundingClientRect().width > 200);
+            }"""
+        )
+        if sku_index is None or sku_index < 0:
+            save_debug_snapshot(page, f"sku_input_not_found_{sku}")
+            raise RuntimeError("未找到 SKU 输入框")
+        sku_input = page.locator("input").nth(sku_index)
+    sku_input.click()
+    page.keyboard.press("Control+A")
+    page.keyboard.type(sku)
+    page.locator('button:has-text("确定")').first.click(timeout=10000)
+    page.wait_for_timeout(2500)
+
+
+def extract_outbound_quantity(page, sku: str) -> Dict[str, object]:
+    result = page.evaluate(
+        r"""(sku) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const normalize = (text) => String(text || '').replace(/\s+/g, ' ').trim();
+            const tables = Array.from(document.querySelectorAll('table')).filter(visible);
+            for (const table of tables) {
+                const headerCells = Array.from(table.querySelectorAll('thead th, thead td, tr:first-child th, tr:first-child td'));
+                const headers = headerCells.map((cell) => normalize(cell.textContent));
+                const metricIndex = headers.findIndex((text) => text.includes('出库商品件数'));
+                if (metricIndex < 0) continue;
+                const rows = Array.from(table.querySelectorAll('tbody tr, tr')).filter(visible);
+                for (const row of rows) {
+                    const rowText = normalize(row.textContent);
+                    if (!rowText.includes(sku)) continue;
+                    const cells = Array.from(row.querySelectorAll('td, th')).map((cell) => normalize(cell.textContent));
+                    return {
+                        found: true,
+                        metric_text: cells[metricIndex] || '',
+                        row_text: rowText,
+                        headers,
+                        cells,
+                    };
+                }
+            }
+
+            const rows = Array.from(document.querySelectorAll('tr, [class*="row"]')).filter(visible);
+            for (const row of rows) {
+                const rowText = normalize(row.textContent);
+                if (rowText.includes(sku)) {
+                    const match = rowText.match(new RegExp('SKU[：:]\\s*' + sku + '\\s+([0-9,]+)\\s+([0-9,]+)\\s+([0-9,]+)'));
+                    return {
+                        found: true,
+                        metric_text: match ? match[3] : '',
+                        row_text: rowText,
+                        headers: [],
+                        cells: []
+                    };
+                }
+            }
+            const bodyText = normalize(document.body?.innerText || '');
+            const bodyMatch = bodyText.match(new RegExp('SKU[：:]\\s*' + sku + '\\s+([0-9,]+)\\s+([0-9,]+)\\s+([0-9,]+)'));
+            if (bodyMatch) {
+                return {
+                    found: true,
+                    metric_text: bodyMatch[3],
+                    row_text: bodyMatch[0],
+                    headers: [],
+                    cells: []
+                };
+            }
+            return {found: false, metric_text: '', row_text: '', headers: [], cells: []};
+        }""",
+        sku,
+    )
+    if not isinstance(result, dict):
+        result = {"found": False, "metric_text": "", "row_text": "", "headers": [], "cells": []}
+    metric_text = str(result.get("metric_text") or "")
+    result["sales"] = parse_number(metric_text)
+    return result
+
+
+def collect_style(page, style_name: str, skus: List[str]) -> Dict[str, object]:
+    print(f"[Collect] {style_name}: {', '.join(skus)}")
+    sku_results = []
+    total_sales = 0
+    for sku in skus:
+        print(f"  SKU {sku}")
+        try:
+            set_sku_and_query(page, sku)
+            extracted = extract_outbound_quantity(page, sku)
+            sales = int(extracted.get("sales") or 0)
+            total_sales += sales
+            sku_results.append(
+                {
+                    "sku": sku,
+                    "sales": sales,
+                    "found": bool(extracted.get("found")),
+                    "metric_text": extracted.get("metric_text", ""),
+                    "row_text": extracted.get("row_text", ""),
+                }
+            )
+            print(f"    出库商品件数: {sales}")
+        except Exception as exc:
+            sku_results.append({"sku": sku, "sales": 0, "found": False, "error": str(exc)})
+            print(f"    [WARN] 采集失败: {exc}")
+    return {"style_name": style_name, "skus": skus, "sales": total_sales, "sku_results": sku_results}
+
+
+def write_style_outputs(output_dir: Path, start_date: str, end_date: str, result: Dict[str, object]) -> None:
+    style_name = str(result["style_name"])
+    label = period_label(start_date, end_date)
+    style_dir = output_dir / f"{safe_filename(style_name)}_{label}"
+    style_dir.mkdir(parents=True, exist_ok=True)
+
+    payload = {
+        "platform": "jd_self",
+        "platform_label": "京东自营",
+        "period_type": "week",
+        "period_start": start_date,
+        "period_end": end_date,
+        "period_label": label,
+        "metric": "出库商品件数",
+        "style_name": style_name,
+        "sales": result["sales"],
+        "skus": result["skus"],
+        "sku_results": result["sku_results"],
+        "collected_at": datetime.now().isoformat(timespec="seconds"),
+    }
+    json_path = style_dir / f"{safe_filename(style_name)}_{label}.json"
+    md_path = style_dir / f"{safe_filename(style_name)}_{label}.md"
+    json_path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    lines = [
+        f"# {style_name} {start_date} ~ {end_date} 京东自营库存销量",
+        "",
+        f"- 平台: 京东自营",
+        f"- 周期: {start_date} ~ {end_date}",
+        f"- 指标: 出库商品件数",
+        f"- 汇总销量: {result['sales']}",
+        "",
+        "| SKU | 出库商品件数 | 状态 |",
+        "| --- | ---: | --- |",
+    ]
+    for item in result["sku_results"]:
+        status = "已找到" if item.get("found") else "未找到"
+        if item.get("error"):
+            status = f"失败: {item['error']}"
+        lines.append(f"| {item['sku']} | {item.get('sales', 0)} | {status} |")
+    lines.append("")
+    md_path.write_text("\n".join(lines), encoding="utf-8")
+
+
+def write_summary(output_dir: Path, start_date: str, end_date: str, results: List[Dict[str, object]]) -> None:
+    label = period_label(start_date, end_date)
+    summary_dir = output_dir / label
+    summary_dir.mkdir(parents=True, exist_ok=True)
+    payload = {
+        "platform": "jd_self",
+        "platform_label": "京东自营",
+        "period_type": "week",
+        "period_start": start_date,
+        "period_end": end_date,
+        "period_label": label,
+        "metric": "出库商品件数",
+        "total_styles": len(results),
+        "results": results,
+        "collected_at": datetime.now().isoformat(timespec="seconds"),
+    }
+    (summary_dir / "jd_self_inventory_sales_summary.json").write_text(
+        json.dumps(payload, ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+
+    lines = [
+        f"# 京东自营库存销量汇总 {start_date} ~ {end_date}",
+        "",
+        "| 款式 | SKU数量 | 出库商品件数 |",
+        "| --- | ---: | ---: |",
+    ]
+    for item in results:
+        lines.append(f"| {item['style_name']} | {len(item['skus'])} | {item['sales']} |")
+    lines.append("")
+    (summary_dir / "jd_self_inventory_sales_summary.md").write_text("\n".join(lines), encoding="utf-8")
+
+
+def run(args) -> None:
+    if args.week_start or args.week_end:
+        if not args.week_start or not args.week_end:
+            raise ValueError("--week-start 和 --week-end 必须同时提供")
+        start_date = validate_date_text(args.week_start, "--week-start")
+        end_date = validate_date_text(args.week_end, "--week-end")
+    elif args.target_date:
+        target_date = validate_date_text(args.target_date, "--target-date")
+        target = datetime.strptime(target_date, "%Y-%m-%d")
+        start = target - timedelta(days=target.weekday())
+        end = start + timedelta(days=6)
+        start_date = start.strftime("%Y-%m-%d")
+        end_date = end.strftime("%Y-%m-%d")
+    else:
+        start_date, end_date = last_full_week_range()
+
+    output_dir = managed_data_path(args.output)
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    with sync_playwright() as playwright:
+        browser = playwright.chromium.launch(headless=args.headless)
+        context = browser.new_context(viewport={"width": 1920, "height": 1080}, accept_downloads=True)
+        page = context.new_page()
+        shangzhi_page = page
+        try:
+            step_login(page, args.shop, args.password)
+            shangzhi_page = step_enter_shangzhi(context, page)
+            return_old_version_if_present(shangzhi_page)
+
+            print("[Step] 打开京东自营供应链库存明细页...")
+            shangzhi_page.goto(INVENTORY_URL, wait_until="domcontentloaded", timeout=60000)
+            shangzhi_page.wait_for_timeout(5000)
+            ensure_popup_closed(shangzhi_page)
+            set_inventory_week(shangzhi_page, start_date, end_date)
+
+            results = []
+            style_sku_groups = _get_style_sku_groups()
+            for style in style_sku_groups:
+                result = collect_style(shangzhi_page, style["style_name"], style["skus"])
+                result["period_type"] = "week"
+                result["period_start"] = start_date
+                result["period_end"] = end_date
+                results.append(result)
+                write_style_outputs(output_dir, start_date, end_date, result)
+
+            write_summary(output_dir, start_date, end_date, results)
+            print(f"[OK] 已输出到: {output_dir}")
+
+            # 写库（库不可用时 WARN，不阻塞主流程）
+            try:
+                from db import get_conn, upsert_jd_self_sales
+                rows = []
+                for r in results:
+                    for item in r.get("sku_results") or []:
+                        rows.append({
+                            "week_end": end_date,
+                            "week_start": start_date,
+                            "style_name": r["style_name"],
+                            "sku": item.get("sku", ""),
+                            "outbound_qty": int(item.get("sales", 0) or 0),
+                            "found": bool(item.get("found")),
+                            "collected_at": datetime.now().isoformat(timespec="seconds"),
+                        })
+                if rows:
+                    with get_conn() as conn:
+                        n = upsert_jd_self_sales(conn, rows)
+                    print(f"[PG] jd_self_weekly_sales upsert: {n} 行")
+            except Exception as exc:
+                print(f"[WARN] PG upsert 失败 (库未配/不可达)，主流程不受影响: {exc}")
+        finally:
+            if args.keep_open:
+                print("[INFO] --keep-open 已开启，浏览器保留 60 秒便于检查")
+                try:
+                    shangzhi_page.wait_for_timeout(60000)
+                except Exception as exc:
+                    print(f"[WARN] keep-open 等待时页面已关闭: {exc}")
+            context.close()
+            browser.close()
+
+
+def build_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(description="采集京东自营供应链库存明细的出库商品件数")
+    parser.add_argument("--shop", default=os.getenv("JD_SELF_SHOP", "光影自营研发"), help="京东登录账号")
+    parser.add_argument(
+        "--password",
+        default=os.getenv("JD_SELF_PASSWORD", DEFAULT_CONFIG.get("password", "")),
+        help="京东登录密码，默认复用原 JD 脚本密码",
+    )
+    parser.add_argument("--week-start", default="", help="采集周开始日期，格式 YYYY-MM-DD；默认上一完整周周一")
+    parser.add_argument("--week-end", default="", help="采集周结束日期，格式 YYYY-MM-DD；默认上一完整周周日")
+    parser.add_argument(
+        "--target-date",
+        default="",
+        help="兼容旧参数：传入任意日期后采集该日期所在周，格式 YYYY-MM-DD",
+    )
+    parser.add_argument("--output", default=str(DEFAULT_OUTPUT_DIR), help="输出目录")
+    parser.add_argument("--headless", action="store_true", help="无头浏览器运行")
+    parser.add_argument("--keep-open", action="store_true", help="结束后保留浏览器 60 秒")
+    return parser
+
+
+if __name__ == "__main__":
+    run(build_parser().parse_args())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/lark_cli_runtime.py b/src/gyxx_flow/modules/product_commerce/runtime/lark_cli_runtime.py
new file mode 100644
index 0000000..499b749
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/lark_cli_runtime.py
@@ -0,0 +1,114 @@
+"""Shared lark-cli process helpers for the local Windows collector runtime."""
+
+from __future__ import annotations
+
+import json
+import os
+import shutil
+import subprocess
+from pathlib import Path
+from typing import Any, Mapping
+
+
+def lark_cli_env(
+    base_env: Mapping[str, str] | None = None,
+    home: Path | None = None,
+) -> dict[str, str]:
+    """Return a subprocess environment that can reuse the legacy Hermes config.
+
+    Explicit configuration always wins.  Otherwise this project consistently
+    uses its existing Hermes profile, even if another command has since created
+    an empty/default global config.
+    """
+    env = dict(os.environ if base_env is None else base_env)
+    if not env.get("LARKSUITE_CLI_CONFIG_DIR"):
+        config_root = (home or Path.home()) / ".lark-cli"
+        legacy_dir = config_root / "hermes"
+        if (legacy_dir / "config.json").is_file():
+            env["LARKSUITE_CLI_CONFIG_DIR"] = str(legacy_dir)
+    env.setdefault("LARKSUITE_CLI_NO_UPDATE_NOTIFIER", "1")
+    env.setdefault("LARKSUITE_CLI_NO_SKILLS_NOTIFIER", "1")
+    return env
+
+
+def _node_cli_entry() -> tuple[str, str]:
+    node = shutil.which("node.exe") or shutil.which("node")
+    if not node:
+        raise RuntimeError("node not found in PATH")
+
+    lark = shutil.which("lark-cli") or shutil.which("lark-cli.cmd")
+    candidates: list[Path] = []
+    if lark:
+        candidates.append(
+            Path(lark).resolve().parent
+            / "node_modules"
+            / "@larksuite"
+            / "cli"
+            / "scripts"
+            / "run.js"
+        )
+    candidates.append(
+        Path.home()
+        / "AppData"
+        / "Roaming"
+        / "npm"
+        / "node_modules"
+        / "@larksuite"
+        / "cli"
+        / "scripts"
+        / "run.js"
+    )
+    for candidate in candidates:
+        if candidate.is_file():
+            return node, str(candidate)
+    raise RuntimeError("lark-cli Node entry not found")
+
+
+def _parse_json_output(text: str) -> dict[str, Any]:
+    decoder = json.JSONDecoder()
+    pos = 0
+    while pos < len(text):
+        start = text.find("{", pos)
+        if start < 0:
+            break
+        try:
+            value, _ = decoder.raw_decode(text, start)
+            if isinstance(value, dict):
+                return value
+        except json.JSONDecodeError:
+            pass
+        pos = start + 1
+    raise RuntimeError(f"lark-cli returned non-JSON output: {text[:200]}")
+
+
+def run_lark_cli(
+    args: list[str],
+    *,
+    cwd: str | Path | None = None,
+    timeout: int = 120,
+    env: Mapping[str, str] | None = None,
+) -> dict[str, Any]:
+    process_env = lark_cli_env(env)
+    node, cli_script = _node_cli_entry()
+    command = [node, cli_script]
+    config_dir = Path(process_env.get("LARKSUITE_CLI_CONFIG_DIR", ""))
+    if config_dir.name.lower() == "hermes":
+        command += ["--profile", "hermes-analyzer"]
+    command += args
+    proc = subprocess.run(
+        command,
+        cwd=cwd,
+        env=process_env,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+        timeout=timeout,
+    )
+    if proc.returncode:
+        detail = (proc.stderr or proc.stdout).strip()
+        raise RuntimeError(f"lark-cli failed (exit={proc.returncode}): {detail[:500]}")
+    result = _parse_json_output(proc.stdout or "")
+    if result.get("ok") is False:
+        raise RuntimeError(f"lark-cli failed: {result.get('error')}")
+    return result
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/README.md b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/README.md
new file mode 100644
index 0000000..3a3168c
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/README.md
@@ -0,0 +1,8 @@
+# Launcher provenance only
+
+These launchers preserve the arguments and sequencing discovered during the
+read-only audit. Their project-root literals were replaced by the inert
+`GYXX_RUNTIME_ROOT` placeholder during migration.
+
+Do not register or invoke these files as production entry points. Final tasks
+must call the native `gyxx-flow` CLI and its workflow registry.
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_alerts.bat b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_alerts.bat
new file mode 100644
index 0000000..1a9dde8
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_alerts.bat
@@ -0,0 +1,35 @@
+@echo off
+REM Daily 23:00 sales-decline alert -- independent of 08:40 daily collect.
+REM
+REM Detection window: previous 9 days from today, split into 3 segments
+REM   (strict decline: seg1 > seg2 > seg3 > 0).
+REM Data source: data/<plat>/<safe_name>_<date>/data.json (PG first, JSON fallback).
+REM Notify: POST Hermes analyzer chat completions, analyzer @ recipient by openid.
+REM Recipient: ou_7ad5fc8012e2f741afc5346e05ffd447.
+REM
+REM Independence rationale:
+REM   - daily collect at 08:40 uses "yesterday" as window end (9d back: 5.31-6.8)
+REM   - this job at 23:00 uses "today" (9d back: 6.1-6.9) to see the latest trend
+REM   - even if daily collect fails, the alert still triggers
+REM
+REM Exit codes:
+REM   0  = 0 events, or successful POST
+REM   1  = detection logic failed
+REM   3  = Hermes POST failed (events still persisted)
+
+setlocal
+cd /d %GYXX_RUNTIME_ROOT%
+
+set "PY=%GYXX_RUNTIME_ROOT%\.venv\Scripts\python.exe"
+
+if not exist data\logs mkdir data\logs
+
+for /f %%i in ('powershell -NoProfile -Command "Get-Date -Format yyyy-MM-dd"') do set TODAY=%%i
+
+echo [%TODAY% 23:00] === sales alert job start === >> data\logs\alerts.log
+
+"%PY%" run_alerts_with_retry.py --end-date %TODAY% --openid ou_7ad5fc8012e2f741afc5346e05ffd447 >> data\logs\alerts.log 2>&1
+set RC=%ERRORLEVEL%
+
+echo [%TODAY% 23:00] === sales alert job end (rc=%RC%) === >> data\logs\alerts.log
+exit /b %RC%
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_collect.bat b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_collect.bat
new file mode 100644
index 0000000..446801e
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_collect.bat
@@ -0,0 +1,17 @@
+@echo off
+REM Daily 08:40 trigger -- collect (ERP + 3 platforms) + analyze (date-first summary)
+REM                   + export (data/ready_for_bitable/<date>/bitable_records.json)
+REM                   + insert (推飞书多维表格)
+REM Skips notify/decline (告警由 run_alerts.bat 23:00 跑; decline 暂未自动触发)
+REM Uses project venv Python to ensure scrapling/playwright are available
+REM 08:40 选在 JD/TM 当日数据已刷新、且赶在工作日开始前完成批跑
+
+setlocal
+cd /d %GYXX_RUNTIME_ROOT%
+
+set "PY=%GYXX_RUNTIME_ROOT%\.venv\Scripts\python.exe"
+
+if not exist data\logs mkdir data\logs
+
+"%PY%" orchestrate_daily_collection.py --stages collect,analyze,export,insert >> data\logs\daily_collect.log 2>&1
+exit /b %ERRORLEVEL%
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_import.bat b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_import.bat
new file mode 100644
index 0000000..d184520
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_import.bat
@@ -0,0 +1,38 @@
+@echo off
+REM run_daily_import.bat
+REM Triggered by Windows Task Scheduler daily at 19:00.
+REM Imports previous day's data via import_product_daily.py.
+REM Log goes to logs/daily_import_<timestamp>.log.
+REM Note: file is ASCII-only on purpose - cmd's default codepage 936 (GBK)
+REM misparses UTF-8 Chinese in REM/echo lines as commands.
+
+setlocal EnableDelayedExpansion
+set "PROJECT_DIR=%GYXX_RUNTIME_ROOT%"
+set "PY=%PROJECT_DIR%\.venv\Scripts\python.exe"
+set "SCRIPT=%PROJECT_DIR%\import_product_daily.py"
+set "LOG_DIR=%PROJECT_DIR%\logs"
+
+if not exist "%LOG_DIR%" mkdir "%LOG_DIR%"
+
+REM yesterday's date (cmd has no date math, use PowerShell with usebackq to avoid quote conflict)
+for /f "usebackq" %%D in (`powershell -NoProfile -Command "(Get-Date).AddDays(-1).ToString('yyyy-MM-dd')"`) do set "YESTERDAY=%%D"
+
+REM timestamp YYYYMMDD_HHMMSS
+set "STAMP=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%"
+set "STAMP=%STAMP: =0%"
+
+set "LOG_FILE=%LOG_DIR%\daily_import_%STAMP%.log"
+
+REM banner (ASCII only)
+echo [%date% %time:~0,8%] START daily import target_date=%YESTERDAY% log=%LOG_FILE%> "%LOG_FILE%"
+
+cd /d "%PROJECT_DIR%"
+"%PY%" "%SCRIPT%" --date %YESTERDAY% >> "%LOG_FILE%" 2>&1
+set "EXIT_CODE=%ERRORLEVEL%"
+
+echo [%date% %time:~0,8%] exit_code=%EXIT_CODE% >> "%LOG_FILE%"
+
+REM keep 30 days of logs, delete older
+forfiles /p "%LOG_DIR%" /m "daily_import_*.log" /d -30 /c "cmd /c del @file" 2>nul
+
+endlocal & exit /b %EXIT_CODE%
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_import.ps1 b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_import.ps1
new file mode 100644
index 0000000..c75a536
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_import.ps1
@@ -0,0 +1,49 @@
+# run_daily_import.ps1
+# 每日 19:00 由 Windows Task Scheduler 触发。
+# 计算昨天的日期, 调 import_product_daily.py 入库, 日志落到 logs/。
+
+$ErrorActionPreference = "Continue"
+$OutputEncoding = [System.Text.Encoding]::UTF8
+$PSDefaultParameterValues['Out-File:Encoding'] = 'utf8'
+
+$ProjectDir = $env:GYXX_RUNTIME_ROOT
+$LogDir = Join-Path $ProjectDir "logs"
+
+if (-not (Test-Path $LogDir)) {
+    New-Item -ItemType Directory -Path $LogDir -Force | Out-Null
+}
+
+$Stamp = Get-Date -Format "yyyyMMdd_HHmmss"
+$Yesterday = (Get-Date).AddDays(-1).ToString("yyyy-MM-dd")
+$LogFile = Join-Path $LogDir ("daily_import_" + $Stamp + ".log")
+
+# 把 stdout/stderr 全部 >> 到日志 (用 cmd 重定向, 避免 PowerShell pipe 跟 Python 的编码冲突)
+@"
+[$(Get-Date -Format 'yyyy-MM-dd HH:mm:ss')] 启动每日导入 | 目标日期=$Yesterday | 日志=$LogFile
+"@ | Out-File -FilePath $LogFile -Encoding utf8 -Append
+
+Set-Location $ProjectDir
+$Python = Join-Path $ProjectDir ".venv\Scripts\python.exe"
+$Script = Join-Path $ProjectDir "import_product_daily.py"
+
+if (-not (Test-Path $Python)) {
+    Add-Content -Path $LogFile -Value "[FATAL] 找不到 python: $Python" -Encoding utf8
+    exit 2
+}
+if (-not (Test-Path $Script)) {
+    Add-Content -Path $LogFile -Value "[FATAL] 找不到脚本: $Script" -Encoding utf8
+    exit 2
+}
+
+# 用 cmd 重定向 (2>&1 把 stderr 合并) - 兼容 Python 输出
+cmd /c "`"$Python`" `"$Script`" --date $Yesterday >> `"$LogFile`" 2>&1"
+$Exit = $LASTEXITCODE
+
+Add-Content -Path $LogFile -Value "[$(Get-Date -Format 'yyyy-MM-dd HH:mm:ss')] 退出码: $Exit" -Encoding utf8
+
+# 保留 30 天日志, 删旧的
+Get-ChildItem -Path $LogDir -Filter "daily_import_*.log" |
+    Where-Object { $_.LastWriteTime -lt (Get-Date).AddDays(-30) } |
+    ForEach-Object { Remove-Item $_.FullName -Force }
+
+exit $Exit
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_persona.bat b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_persona.bat
new file mode 100644
index 0000000..1362493
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_daily_persona.bat
@@ -0,0 +1,16 @@
+@echo off
+REM Daily 10:00 trigger -- 三平台人群画像并行采集 + 回填飞书表
+REM 天猫: collect_persona_to_bitable.py (达摩盘)
+REM 抖音: collect_dy_persona_to_bitable.py (罗盘成交人群)
+REM 京东: collect_jd_persona_to_bitable.py (商智人群画像)
+REM 三个脚本并行启动，各自独立写日志
+
+setlocal
+cd /d %GYXX_RUNTIME_ROOT%
+
+set "PY=%GYXX_RUNTIME_ROOT%\.venv\Scripts\python.exe"
+
+if not exist data\logs mkdir data\logs
+
+"%PY%" run_daily_persona.py >> data\logs\daily_persona_launcher.log 2>&1
+exit /b %ERRORLEVEL%
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_style_analysis_3d.bat b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_style_analysis_3d.bat
new file mode 100644
index 0000000..c95dbac
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_style_analysis_3d.bat
@@ -0,0 +1,13 @@
+@echo off
+REM Runs independently every 3 days at 11:00. The analyzer defaults to yesterday
+REM as the window end, so each run covers the latest three complete calendar days.
+
+setlocal
+cd /d %GYXX_RUNTIME_ROOT%
+
+set "PY=%GYXX_RUNTIME_ROOT%\.venv\Scripts\python.exe"
+
+if not exist data\logs mkdir data\logs
+
+"%PY%" -u analyze_style_with_hermes.py --all-styles --days 3 --min-interval-days 3 --skip-existing >> data\logs\style_analysis_3d.log 2>&1
+exit /b %ERRORLEVEL%
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_jd_main_image.bat b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_jd_main_image.bat
new file mode 100644
index 0000000..c7848c4
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_jd_main_image.bat
@@ -0,0 +1,12 @@
+@echo off
+REM Weekly JD main image pipeline. Keep this file ASCII-only for Task Scheduler.
+setlocal
+cd /d %GYXX_RUNTIME_ROOT%
+
+set "PY=%GYXX_RUNTIME_ROOT%\.venv\Scripts\python.exe"
+set "PYTHONUNBUFFERED=1"
+
+if not exist data\logs mkdir data\logs
+
+"%PY%" -u run_weekly_jd_main_image.py --headless >> data\logs\weekly_jd_main_image.log 2>&1
+exit /b %ERRORLEVEL%
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_main_image.bat b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_main_image.bat
new file mode 100644
index 0000000..7e7c545
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_main_image.bat
@@ -0,0 +1,12 @@
+@echo off
+REM Weekly main image pipeline. Keep this file ASCII-only for Task Scheduler.
+setlocal
+cd /d %GYXX_RUNTIME_ROOT%
+
+set "PY=%GYXX_RUNTIME_ROOT%\.venv\Scripts\python.exe"
+set "PYTHONUNBUFFERED=1"
+
+if not exist data\logs mkdir data\logs
+
+"%PY%" -u run_weekly_main_image.py --headless >> data\logs\weekly_main_image.log 2>&1
+exit /b %ERRORLEVEL%
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_market_rank.bat b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_market_rank.bat
new file mode 100644
index 0000000..2b1357b
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/run_weekly_market_rank.bat
@@ -0,0 +1,14 @@
+@echo off
+REM Weekly Monday 10:00 -- collect TM/JD/DY market ranks, create Feishu docs,
+REM and archive each platform's latest document URL in PostgreSQL.
+
+setlocal
+cd /d %GYXX_RUNTIME_ROOT%
+
+set "PY=%GYXX_RUNTIME_ROOT%\.venv\Scripts\python.exe"
+set "PYTHONUNBUFFERED=1"
+
+if not exist data\logs mkdir data\logs
+
+"%PY%" orchestrate_market_rank_collection.py >> data\logs\market_rank_weekly.log 2>&1
+exit /b %ERRORLEVEL%
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/main_image_db.py b/src/gyxx_flow/modules/product_commerce/runtime/main_image_db.py
new file mode 100644
index 0000000..629f25a
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/main_image_db.py
@@ -0,0 +1,23 @@
+"""Late database persistence for long-running main-image upload jobs."""
+
+from collections.abc import Callable, Sequence
+from typing import Any
+
+
+def persist_main_image_records(
+    records: Sequence[dict[str, Any]],
+    *,
+    get_conn_fn: Callable[[], Any] | None = None,
+    upsert_fn: Callable[[Any, Sequence[dict[str, Any]]], int] | None = None,
+) -> int:
+    """Open PostgreSQL only for the final write, avoiding idle cloud timeouts."""
+    if not records:
+        return 0
+    if get_conn_fn is None or upsert_fn is None:
+        from db import get_conn, upsert_main_image_creatives
+
+        get_conn_fn = get_conn
+        upsert_fn = upsert_main_image_creatives
+
+    with get_conn_fn() as conn:
+        return upsert_fn(conn, records)
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/main_image_paths.py b/src/gyxx_flow/modules/product_commerce/runtime/main_image_paths.py
new file mode 100644
index 0000000..502e723
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/main_image_paths.py
@@ -0,0 +1,21 @@
+"""Shared platform-owned paths for weekly main-image collection."""
+
+from pathlib import Path
+
+
+_PLATFORMS = {"tm", "jd"}
+
+
+def collect_images_dir(style_dir: Path, platform: str) -> Path:
+    """Return the isolated directory used by a platform's collector."""
+    if platform not in _PLATFORMS:
+        raise ValueError(f"unsupported main-image platform: {platform}")
+    return style_dir / f"{platform}_images"
+
+
+def resolve_images_dir(style_dir: Path, platform: str) -> Path:
+    """Prefer the isolated directory, falling back to legacy `images` data."""
+    current = collect_images_dir(style_dir, platform)
+    if current.exists():
+        return current
+    return style_dir / "images"
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/market_rank_hermes_notification.py b/src/gyxx_flow/modules/product_commerce/runtime/market_rank_hermes_notification.py
new file mode 100644
index 0000000..5d77025
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/market_rank_hermes_notification.py
@@ -0,0 +1,219 @@
+"""三平台市场排行飞书文档链接的 Hermes 分析端通知。"""
+
+from __future__ import annotations
+
+import json
+import os
+from gyxx_flow.adapters import RuntimeServicePolicy
+import re
+import urllib.error
+import urllib.request
+from datetime import date, datetime
+from typing import Any, Iterable
+
+from db import (
+    ensure_market_rank_notification_table,
+    get_conn,
+    get_market_rank_reports_for_date,
+    market_rank_notification_was_sent,
+    upsert_market_rank_notification,
+)
+
+
+HERMES_API_KEY = os.getenv("HERMES_API_KEY")
+if not HERMES_API_KEY:
+    raise RuntimeError("HERMES_API_KEY is required")
+os.environ.update(RuntimeServicePolicy().apply(os.environ))
+HERMES_ANALYZER_URL = os.getenv(
+    "ANALYZER_API_SERVER_URL",
+    "http://localhost:8642/v1/chat/completions",
+)
+HERMES_MODEL = os.getenv("HERMES_ANALYZER_MODEL", "mimo-v2.5-pro")
+
+PLATFORM_ORDER = ("tm", "jd", "dy")
+PLATFORM_LABELS = {"tm": "天猫", "jd": "京东", "dy": "抖音"}
+
+
+def hermes_completion_content(response: dict[str, Any]) -> str:
+    try:
+        content = response["choices"][0]["message"]["content"]
+    except (KeyError, IndexError, TypeError):
+        return ""
+    return str(content or "")
+
+
+def extract_feishu_message_id(response: dict[str, Any]) -> str | None:
+    direct = response.get("message_id")
+    if direct and str(direct).startswith("om_"):
+        return str(direct)
+    match = re.search(r"\bom_[A-Za-z0-9]+\b", hermes_completion_content(response))
+    return match.group(0) if match else None
+
+
+def hermes_send_succeeded(response: dict[str, Any]) -> bool:
+    if response.get("error"):
+        return False
+    content = hermes_completion_content(response)
+    return extract_feishu_message_id(response) is not None or "发送成功" in content
+
+
+def build_notification_message(report_date: date, reports: Iterable[dict[str, Any]]) -> str:
+    by_platform = {str(item["platform"]): item for item in reports}
+    lines = [
+        f"三平台市场排行采集结果（{report_date.isoformat()}，"
+        f"当前成功平台 {len(by_platform)} 个）",
+        "",
+    ]
+    for platform in PLATFORM_ORDER:
+        if platform not in by_platform:
+            continue
+        report = by_platform[platform]
+        label = PLATFORM_LABELS[platform]
+        count = int(report.get("product_count", 0) or 0)
+        lines.append(f"{label}：{report['feishu_doc_url']}（{count} 个商品）")
+    missing = [PLATFORM_LABELS[item] for item in PLATFORM_ORDER if item not in by_platform]
+    if missing:
+        lines.append(f"未生成新链接：{'、'.join(missing)}")
+    lines.extend(["", "成功平台的文档已完成整理并独立入库，本消息由分析端 Hermes 发送。"])
+    return "\n".join(lines)
+
+
+def build_hermes_payload(*, recipient_open_id: str, message: str) -> dict[str, Any]:
+    return {
+        "model": HERMES_MODEL,
+        "messages": [
+            {
+                "role": "system",
+                "content": (
+                    "你是光影行星的数据分析通知助手。必须使用飞书消息能力，把用户提供的"
+                    "完整内容原样发送为一条飞书私聊；不得改写、删减或更换任何文档链接。"
+                    "发送完成后，简短回复发送结果。"
+                ),
+            },
+            {
+                "role": "user",
+                "content": (
+                    f"收件人 open_id：{recipient_open_id}\n"
+                    "请立即把下面内容发送给该收件人（飞书私聊）：\n\n"
+                    f"{message}"
+                ),
+            },
+        ],
+    }
+
+
+def post_hermes(payload: dict[str, Any], *, timeout: int = 180) -> dict[str, Any]:
+    body = json.dumps(payload, ensure_ascii=False).encode("utf-8")
+    request = urllib.request.Request(
+        HERMES_ANALYZER_URL,
+        data=body,
+        method="POST",
+        headers={
+            "Authorization": f"Bearer {HERMES_API_KEY}",
+            "Content-Type": "application/json",
+        },
+    )
+    try:
+        with urllib.request.urlopen(request, timeout=timeout) as response:
+            raw = response.read().decode("utf-8", errors="replace")
+        value = json.loads(raw)
+        return value if isinstance(value, dict) else {"error": "invalid Hermes response"}
+    except urllib.error.HTTPError as exc:
+        return {
+            "error": f"http {exc.code}",
+            "body": exc.read().decode("utf-8", errors="replace"),
+        }
+    except urllib.error.URLError as exc:
+        return {"error": f"url error: {exc.reason}"}
+    except (json.JSONDecodeError, TimeoutError) as exc:
+        return {"error": f"{type(exc).__name__}: {exc}"}
+
+
+def notify_market_rank_reports(
+    *,
+    report_date: date,
+    recipient_open_id: str,
+    recipient_name: str = "王云龙",
+    generated_after: datetime | str | None = None,
+) -> dict[str, Any]:
+    """发送本次任务新生成的三平台链接；同日成功发送保持幂等。"""
+    with get_conn() as conn:
+        ensure_market_rank_notification_table(conn)
+        reports = get_market_rank_reports_for_date(
+            conn,
+            report_date,
+            PLATFORM_ORDER,
+            updated_after=generated_after,
+        )
+        by_platform = {str(item["platform"]): item for item in reports}
+        missing = [platform for platform in PLATFORM_ORDER if platform not in by_platform]
+        platform_links = {
+            platform: str(item["feishu_doc_url"])
+            for platform, item in by_platform.items()
+        }
+        if not platform_links:
+            record = {
+                "report_date": report_date,
+                "recipient_open_id": recipient_open_id,
+                "recipient_name": recipient_name,
+                "platform_links": platform_links,
+                "status": "incomplete",
+                "hermes_response": {"missing_platforms": missing},
+            }
+            upsert_market_rank_notification(conn, record)
+            return {
+                "status": "skipped",
+                "reason": "no_platform_reports",
+                "missing_platforms": missing,
+                "report_date": report_date.isoformat(),
+                "recipient_open_id": recipient_open_id,
+            }
+
+        if market_rank_notification_was_sent(
+            conn,
+            report_date,
+            recipient_open_id,
+            platform_links,
+        ):
+            return {
+                "status": "skipped",
+                "reason": "already_sent",
+                "report_date": report_date.isoformat(),
+                "recipient_open_id": recipient_open_id,
+                "platform_links": platform_links,
+                "missing_platforms": missing,
+            }
+
+        ordered_reports = [
+            by_platform[platform]
+            for platform in PLATFORM_ORDER
+            if platform in by_platform
+        ]
+        message = build_notification_message(report_date, ordered_reports)
+        response = post_hermes(
+            build_hermes_payload(
+                recipient_open_id=recipient_open_id,
+                message=message,
+            )
+        )
+        status = "sent" if hermes_send_succeeded(response) else "failed"
+        record = {
+            "report_date": report_date,
+            "recipient_open_id": recipient_open_id,
+            "recipient_name": recipient_name,
+            "platform_links": platform_links,
+            "status": status,
+            "hermes_message_id": extract_feishu_message_id(response),
+            "hermes_response": response,
+        }
+        upsert_market_rank_notification(conn, record)
+        result = {
+            "status": status,
+            "report_date": report_date.isoformat(),
+            "recipient_open_id": recipient_open_id,
+            "platform_links": platform_links,
+            "missing_platforms": missing,
+        }
+        if status == "failed":
+            result["error"] = response.get("error") or "Hermes 未确认飞书消息发送成功"
+        return result
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/market_rank_report_archive.py b/src/gyxx_flow/modules/product_commerce/runtime/market_rank_report_archive.py
new file mode 100644
index 0000000..98e7274
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/market_rank_report_archive.py
@@ -0,0 +1,78 @@
+"""三平台市场排行飞书文档数据库归档。"""
+
+from __future__ import annotations
+
+import re
+from datetime import date
+from typing import Any, Iterable
+
+from db import get_conn, upsert_market_rank_report
+
+
+_PLATFORM_ALIASES = {
+    "tm": "tm",
+    "tmall": "tm",
+    "天猫": "tm",
+    "淘宝": "tm",
+    "jd": "jd",
+    "jingdong": "jd",
+    "京东": "jd",
+    "dy": "dy",
+    "douyin": "dy",
+    "抖音": "dy",
+}
+
+
+def normalize_market_rank_platform(platform: str) -> str:
+    key = str(platform or "").strip().lower()
+    normalized = _PLATFORM_ALIASES.get(key)
+    if not normalized:
+        raise ValueError(f"不支持的市场排行平台: {platform}")
+    return normalized
+
+
+def extract_feishu_doc_id(doc_url: str) -> str | None:
+    match = re.search(r"/docx/([^/?#]+)", str(doc_url or ""))
+    return match.group(1) if match else None
+
+
+def archive_market_rank_report(
+    *,
+    platform: str,
+    report_date: date,
+    period: str,
+    category_scope: str,
+    category_names: Iterable[str],
+    report_title: str,
+    feishu_doc_url: str,
+    product_count: int,
+    source_json_path: str | None = None,
+    run_metadata: dict[str, Any] | None = None,
+) -> int:
+    """规范化并写入一条市场排行飞书文档归档。"""
+    url = str(feishu_doc_url or "").strip()
+    if not url:
+        raise ValueError("飞书文档链接不能为空")
+    names = list(dict.fromkeys(
+        str(name).strip() for name in category_names if str(name).strip()
+    ))
+    record = {
+        "platform": normalize_market_rank_platform(platform),
+        "report_date": report_date,
+        "period": str(period).strip(),
+        "category_scope": str(category_scope).strip(),
+        "category_names": names,
+        "report_title": str(report_title).strip(),
+        "feishu_doc_id": extract_feishu_doc_id(url),
+        "feishu_doc_url": url,
+        "product_count": int(product_count or 0),
+        "source_json_path": str(source_json_path) if source_json_path else None,
+        "run_metadata": run_metadata,
+    }
+    with get_conn() as conn:
+        report_id = upsert_market_rank_report(conn, record)
+    print(
+        f"[OK] 市场排行飞书链接已归档到数据库: "
+        f"platform={record['platform']} id={report_id}"
+    )
+    return report_id
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/orchestrate_daily_collection.py b/src/gyxx_flow/modules/product_commerce/runtime/orchestrate_daily_collection.py
new file mode 100644
index 0000000..15f3d22
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/orchestrate_daily_collection.py
@@ -0,0 +1,407 @@
+import argparse
+import json
+import subprocess
+import sys
+import time
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from datetime import datetime, timedelta
+from pathlib import Path
+
+from runtime_paths import LOG_ROOT, runtime_script
+
+from collect_retry_utils import FailedCollector, consolidate_failed_logs
+from config.style_config_loader import (
+    StyleConfigLoader,
+    get_daily_styles,
+    get_erp_daily_styles,
+    get_jd_self_styles,
+    get_main_image_styles,
+)
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+
+# 顶层 per-run 重试参数 (跟 backfill_one_day.py 一致)
+MAX_RUN_RETRIES = 2
+RUN_SLEEP_SECONDS = 5
+
+
+COMMANDS = {
+    "erp": [sys.executable, str(runtime_script("collect_erp_yesterday_metrics.py")), "--platform", "all"],
+    "jd": [sys.executable, str(runtime_script("jd_product_data_collector.py"))],
+    "dy": [sys.executable, str(runtime_script("dy_product_scraping.py"))],
+    # TM 走主脚本 taobao_sycm_collect.py(默认采昨天);
+    # taobao_sycm_collect_backfill.py 是历史回采专用,只用于手动补采。
+    "tm": [sys.executable, str(runtime_script("taobao_sycm_collect.py"))],
+}
+
+STAGE_ORDER = ["collect", "analyze", "export", "notify", "insert", "style_analysis", "decline"]
+
+
+def default_target_date() -> str:
+    return (datetime.now().date() - timedelta(days=1)).strftime("%Y-%m-%d")
+
+
+def validate_date(value: str) -> str:
+    try:
+        datetime.strptime(value, "%Y-%m-%d")
+    except ValueError as exc:
+        raise argparse.ArgumentTypeError("date must be YYYY-MM-DD") from exc
+    return value
+
+
+def command_with_target_date(command, target_date):
+    return [*command, "--target-date", target_date]
+
+
+def run_step(name, command, dry_run=False, critical=True):
+    """执行子命令。
+
+    critical=True  失败抛 RuntimeError，整个编排中断
+    critical=False 失败仅打印 [WARN] 并返回错误结果，编排继续
+    """
+    printable = " ".join(command)
+    print(f"[RUN] {name}: {printable}")
+    if dry_run:
+        return {"name": name, "code": 0, "dry_run": True}
+
+    try:
+        proc = subprocess.run(
+            command,
+            cwd=PROJECT_ROOT,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+        )
+    except FileNotFoundError as exc:
+        msg = f"{name} FileNotFoundError: {exc}"
+        if critical:
+            raise RuntimeError(msg) from exc
+        print(f"[WARN] {msg}")
+        return {"name": name, "code": 127, "error": msg}
+    except OSError as exc:
+        msg = f"{name} OSError: {exc}"
+        if critical:
+            raise RuntimeError(msg) from exc
+        print(f"[WARN] {msg}")
+        return {"name": name, "code": 126, "error": msg}
+
+    if proc.returncode:
+        msg = f"{name} failed with exit code {proc.returncode}"
+        if critical:
+            raise RuntimeError(msg)
+        print(f"[WARN] {msg}")
+        return {"name": name, "code": proc.returncode, "error": msg}
+    print(f"[OK] {name}")
+    return {"name": name, "code": proc.returncode}
+
+
+def run_step_with_retry(name, command, dry_run=False, critical=True,
+                        failed: FailedCollector | None = None,
+                        max_retries: int = MAX_RUN_RETRIES,
+                        sleep_seconds: float = RUN_SLEEP_SECONDS) -> dict:
+    """包 run_step: subprocess 失败 → sleep → 再 subprocess, 最多 max_retries 次重试。
+
+    沿用 run_step 的 critical 语义: critical=True 阶段重试用尽后 raise;
+    critical=False 阶段重试用尽后仅 [WARN] 继续, 不会中断编排。
+
+    失败统一记录到 failed (FailedCollector), 末尾 .write() 落 data/logs/failed_<date>_daily.json。
+    dry_run=True 时直接调 run_step 不加重试 (只跑一次干跑)。
+    """
+    if dry_run:
+        return run_step(name, command, dry_run=True, critical=critical)
+
+    last_result: dict = {"name": name, "code": 0}
+    total_attempts = max_retries + 1
+    for attempt in range(1, total_attempts + 1):
+        result = run_step(name, command, dry_run=False, critical=False)
+        last_result = result
+        if result.get("code") == 0:
+            if attempt > 1:
+                print(f"[OK] {name} 第 {attempt}/{total_attempts} 次重试成功", flush=True)
+            return result
+        err = result.get("error", "unknown")
+        if attempt >= total_attempts:
+            # 重试用尽
+            if failed is not None:
+                failed.add_page(name, f"{err} (attempts={attempt})", attempts=attempt)
+            if critical:
+                raise RuntimeError(f"{name} failed after {attempt} attempts: {err}")
+            print(f"[WARN] {name} 失败 (已重试 {max_retries} 次, {err}), 继续",
+                  flush=True)
+            return result
+        print(f"[WARN] {name} 失败 (attempt {attempt}/{total_attempts}, {err}), "
+              f"等待 {sleep_seconds}s 后重试", flush=True)
+        time.sleep(sleep_seconds)
+    return last_result
+
+
+def run_platforms(parallel=True, platforms=None, dry_run=False, target_date=None,
+                  failed: FailedCollector | None = None):
+    platforms = platforms or ["jd", "dy", "tm"]
+    if not parallel:
+        for platform in platforms:
+            run_step_with_retry(
+                platform,
+                command_with_target_date(COMMANDS[platform], target_date),
+                dry_run=dry_run,
+                critical=False,
+                failed=failed,
+            )
+        return
+
+    def _run_one(platform: str) -> dict:
+        return run_step_with_retry(
+            platform,
+            command_with_target_date(COMMANDS[platform], target_date),
+            dry_run=dry_run,
+            critical=False,
+            failed=failed,
+        )
+
+    with ThreadPoolExecutor(max_workers=len(platforms)) as pool:
+        futures = {pool.submit(_run_one, platform): platform for platform in platforms}
+        results = []
+        for future in as_completed(futures):
+            results.append(future.result())
+        failed_platforms = [r["name"] for r in results if r.get("code")]
+        if failed_platforms:
+            print(f"[WARN] 平台部分失败: {', '.join(failed_platforms)}（analyze 将使用已有数据）")
+        elif results:
+            print(f"[OK] 全部平台成功: {', '.join(r['name'] for r in results)}")
+
+
+def parse_stages(value):
+    if not value:
+        return list(STAGE_ORDER)
+    raw_stages = [item.strip().lower() for item in value.split(",") if item.strip()]
+    if "all" in raw_stages:
+        return list(STAGE_ORDER)
+    invalid = sorted(set(raw_stages) - set(STAGE_ORDER))
+    if invalid:
+        raise argparse.ArgumentTypeError(
+            f"invalid stage(s): {', '.join(invalid)}; expected: {', '.join(STAGE_ORDER)}"
+        )
+    return [stage for stage in STAGE_ORDER if stage in raw_stages]
+
+
+def apply_legacy_skip_flags(stages, args):
+    selected = list(stages)
+    if args.skip_erp and args.skip_platforms and "collect" in selected:
+        selected.remove("collect")
+    if args.skip_insert and "insert" in selected:
+        selected.remove("insert")
+    return selected
+
+
+def run_collect(args, target_date, platforms, failed):
+    if not args.skip_erp:
+        erp_cmd = command_with_target_date(COMMANDS["erp"], target_date)
+        # --erp-mode backfill: 用"自定义销量/退货量"列 (表头匹配), daily 用 cell 45/57 "昨日销量"列
+        if getattr(args, "erp_mode", "daily") != "daily":
+            erp_cmd = [*erp_cmd, "--mode", args.erp_mode]
+        run_step_with_retry("erp", erp_cmd, dry_run=args.dry_run, failed=failed)
+    if not args.skip_platforms:
+        run_platforms(
+            parallel=not args.serial_platforms,
+            platforms=platforms,
+            dry_run=args.dry_run,
+            target_date=target_date,
+            failed=failed,
+        )
+    # 评论入库 (DY 默认分支已采评论 -> 评论汇总_<YYYYMMDD>.json; JD 单独 --reviews-only 采)
+    # 幂等: product_reviews UNIQUE (platform, order_id) ON CONFLICT DO UPDATE, 重跑覆盖不双写.
+    # critical=False: 评论入库失败不影响后续 analyze/export/insert.
+    if not args.skip_platforms and not args.skip_reviews_import:
+        review_plats = [p for p in platforms if p in ("jd", "dy")]
+        if review_plats:
+            run_step_with_retry(
+                "import_product_reviews",
+                [sys.executable, str(runtime_script("import_product_reviews.py")),
+                 "--platforms", ",".join(review_plats)],
+                dry_run=args.dry_run,
+                critical=False,
+                failed=failed,
+            )
+
+
+def collect_daily_skip_report(date_str: str) -> Path | None:
+    """汇总 6 个平台的「待采 / 跳过」清单,写入 data/logs/<date>_skipped_styles.json。
+
+    来源 (v2.2):
+    - tm/jd/dy -> loader.get_daily_styles()(款式级门 erp+sales + 平台级门 item_ids)
+    - erp     -> loader.get_erp_daily_styles()(款式级 + TM 分组 item_ids)
+    - jd_self -> loader.get_jd_self_styles()(仅 SKU 非空,无跳过)
+    - main_image -> loader.get_main_image_styles()(天猫 + 主图飞书多维表格地址)
+
+    飞书读取失败时,主入口已 ensure_dir 抛错,这里只做兜底写入空报告。
+    """
+    try:
+        loader = StyleConfigLoader()
+        daily = loader.get_daily_styles()
+        erp = loader.get_erp_daily_styles()
+        jd_self = loader.get_jd_self_styles()
+        main_image = loader.get_main_image_styles()
+    except Exception as exc:
+        print(f"[WARN] 跳过清单聚合失败: {exc}")
+        return None
+
+    report = {
+        "date": date_str,
+        "generated_at": datetime.now().isoformat(timespec="seconds"),
+        "platforms": {
+            "tm": daily["tm"],
+            "jd": daily["jd"],
+            "dy": daily["dy"],
+            "erp": {"collectable": erp["collectable"], "skipped": erp["skipped"]},
+            "jd_self": {"collectable": jd_self, "skipped": []},
+            "main_image": {"collectable": main_image, "skipped": []},
+        },
+    }
+    out_dir = LOG_ROOT
+    out_dir.mkdir(parents=True, exist_ok=True)
+    out_path = out_dir / f"{date_str}_skipped_styles.json"
+    out_path.write_text(json.dumps(report, ensure_ascii=False, indent=2), encoding="utf-8")
+    print(f"[SKIP_REPORT] {out_path}")
+    for plat, payload in report["platforms"].items():
+        skipped_n = len(payload.get("skipped") or [])
+        collectable_n = len(payload.get("collectable") or [])
+        if skipped_n:
+            print(f"  - {plat}: 待采 {collectable_n} 款, 跳过 {skipped_n} 款")
+    return out_path
+
+
+def run_analyze(args, target_date, failed):
+    run_step_with_retry(
+        "aggregate_daily_final",
+        [sys.executable, str(runtime_script("aggregate_daily_final.py")), "--date", target_date],
+        dry_run=args.dry_run,
+        failed=failed,
+    )
+
+
+def run_export(args, target_date, failed):
+    run_step_with_retry(
+        "export_bitable_records",
+        [sys.executable, str(runtime_script("export_bitable_records.py")), "--date", target_date],
+        dry_run=args.dry_run,
+        failed=failed,
+    )
+
+
+def run_notify(args, target_date, failed):
+    del target_date, failed
+    print("[SKIP] notify: no native notification entry is present in this runtime")
+
+
+def run_insert(args, target_date, failed):
+    run_step_with_retry(
+        "insert_bitable_records",
+        [sys.executable, str(runtime_script("insert_bitable_records.py")), "--date", target_date],
+        dry_run=args.dry_run,
+        failed=failed,
+    )
+
+
+def run_style_analysis(args, target_date, failed):
+    """Run rolling three-day style reports; failures do not block daily collection."""
+    run_step_with_retry(
+        "style_analysis_3d",
+        [
+            sys.executable,
+            str(runtime_script("analyze_style_with_hermes.py")),
+            "--all-styles",
+            "--end-date", target_date,
+            "--days", "3",
+            "--min-interval-days", "3",
+        ],
+        dry_run=args.dry_run,
+        critical=False,
+        failed=failed,
+    )
+
+
+def run_decline(args, target_date, failed):
+    """9 天连续下滑检测 (ERP 销量) + Hermes analyzer 通知。
+    critical=False：decline 失败不影响 insert/export 已完成的产物。
+    """
+    run_step_with_retry(
+        "check_nine_day_decline",
+        [sys.executable, str(runtime_script("check_nine_day_decline.py")), "--end-date", target_date],
+        dry_run=args.dry_run,
+        critical=False,
+        failed=failed,
+    )
+
+
+def main():
+    parser = argparse.ArgumentParser(description="每日采集编排：ERP -> 三平台 -> 汇总 -> 入库导出")
+    parser.add_argument("--date", type=validate_date, default=None, help="数据日期，默认昨天")
+    parser.add_argument("--target-date", type=validate_date, default=None, help="同 --date，用于回补指定某一天")
+    parser.add_argument(
+        "--stages",
+        type=parse_stages,
+        default=None,
+        help="逗号分隔阶段: collect,analyze,export,notify,insert,style_analysis,decline；默认 all",
+    )
+    parser.add_argument("--skip-erp", action="store_true")
+    parser.add_argument("--skip-platforms", action="store_true")
+    parser.add_argument("--platforms", default="jd,dy,tm", help="逗号分隔: jd,dy,tm")
+    parser.add_argument("--serial-platforms", action="store_true", help="三平台串行跑")
+    parser.add_argument("--skip-insert", action="store_true", help="只导出，不插入飞书")
+    parser.add_argument("--skip-reviews-import", action="store_true",
+                        help="跳过评论入库 (默认 daily collect 后跑 import_product_reviews)")
+    parser.add_argument("--dry-run", action="store_true")
+    parser.add_argument(
+        "--erp-mode",
+        choices=["daily", "weekly", "backfill"],
+        default="daily",
+        help="ERP 采集模式: daily=昨日销量列(cell 45/57); backfill=自定义销量/退货量列(表头匹配); weekly=7天累计",
+    )
+    parser.add_argument(
+        "--max-run-retries",
+        type=int,
+        default=MAX_RUN_RETRIES,
+        help=f"per-run 整体重试次数 (默认 {MAX_RUN_RETRIES})",
+    )
+    args = parser.parse_args()
+
+    target_date = args.target_date or args.date or default_target_date()
+    platforms = [x.strip() for x in args.platforms.split(",") if x.strip()]
+    stages = apply_legacy_skip_flags(args.stages or list(STAGE_ORDER), args)
+
+    failed = FailedCollector("daily", target_date)
+
+    print(f"[INFO] target_date={target_date}")
+    print(f"[INFO] stages={','.join(stages) if stages else '(none)'}")
+    print(f"[INFO] per-run 重试: {args.max_run_retries} 次, 间隔 {RUN_SLEEP_SECONDS}s")
+
+    for stage in stages:
+        if stage == "collect":
+            run_collect(args, target_date, platforms, failed)
+            if not args.dry_run:
+                collect_daily_skip_report(target_date)
+        elif stage == "analyze":
+            run_analyze(args, target_date, failed)
+        elif stage == "export":
+            run_export(args, target_date, failed)
+        elif stage == "notify":
+            run_notify(args, target_date, failed)
+        elif stage == "insert":
+            run_insert(args, target_date, failed)
+        elif stage == "style_analysis":
+            run_style_analysis(args, target_date, failed)
+        elif stage == "decline":
+            run_decline(args, target_date, failed)
+
+    # 末尾: 写失败清单 (无失败时 skip) + 合并到 _summary.json
+    failed_path = failed.write()
+    if failed_path:
+        print(f"[FAIL_LOG] daily: {failed.summary_line()} -> {failed_path}", flush=True)
+    summary_path = consolidate_failed_logs(target_date)
+    if summary_path:
+        print(f"[INFO] 失败清单合并: {summary_path}", flush=True)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/orchestrate_market_rank_collection.py b/src/gyxx_flow/modules/product_commerce/runtime/orchestrate_market_rank_collection.py
new file mode 100644
index 0000000..34073f7
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/orchestrate_market_rank_collection.py
@@ -0,0 +1,373 @@
+"""天猫、京东、抖音市场排行采集统一编排器。
+
+默认并行执行三个平台，每个平台独立记录日志。单个平台失败或抛出异常不会取消
+其他平台，三个平台全部结束后以非零退出码汇总失败。
+
+用法：
+    .venv/Scripts/python.exe orchestrate_market_rank_collection.py
+    .venv/Scripts/python.exe orchestrate_market_rank_collection.py --dry-run
+    .venv/Scripts/python.exe orchestrate_market_rank_collection.py --platforms tm,jd
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import subprocess
+import sys
+from concurrent.futures import ThreadPoolExecutor
+from datetime import date, datetime
+from pathlib import Path
+
+from runtime_paths import LOG_ROOT, managed_data_path, runtime_script
+from typing import Any, Callable, Iterable
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+LOG_DIR = LOG_ROOT / "market_rank_workflow"
+DEFAULT_PLATFORMS = ("tm", "jd", "dy")
+DEFAULT_NOTIFY_OPEN_ID = os.getenv(
+    "MARKET_RANK_NOTIFY_OPEN_ID",
+    "ou_7ad5fc8012e2f741afc5346e05ffd447",
+)
+
+PLATFORM_ALIASES = {
+    "tm": "tm",
+    "tmall": "tm",
+    "天猫": "tm",
+    "jd": "jd",
+    "jingdong": "jd",
+    "京东": "jd",
+    "dy": "dy",
+    "douyin": "dy",
+    "抖音": "dy",
+}
+
+PLATFORM_LABELS = {
+    "tm": "天猫",
+    "jd": "京东",
+    "dy": "抖音",
+}
+
+
+def now_iso() -> str:
+    return datetime.now().astimezone().isoformat(timespec="seconds")
+
+
+def build_subprocess_env(base: dict[str, str] | None = None) -> dict[str, str]:
+    """保证三个采集器输出 UTF-8，并让日志能够实时刷新。"""
+    env = dict(os.environ if base is None else base)
+    env["PYTHONIOENCODING"] = "utf-8"
+    env["PYTHONUNBUFFERED"] = "1"
+    return env
+
+
+def console_safe_text(value: str, encoding: str | None = None) -> str:
+    """替换当前控制台编码无法表示的字符，避免日志转发杀死采集进程。"""
+    target_encoding = encoding or getattr(sys.stdout, "encoding", None) or "utf-8"
+    try:
+        return value.encode(target_encoding, errors="replace").decode(
+            target_encoding,
+            errors="replace",
+        )
+    except LookupError:
+        return value
+
+
+def parse_platforms(value: str | Iterable[str]) -> tuple[str, ...]:
+    raw_items = value.split(",") if isinstance(value, str) else list(value)
+    platforms: list[str] = []
+    for raw in raw_items:
+        key = str(raw).strip().lower()
+        platform = PLATFORM_ALIASES.get(key)
+        if not platform:
+            raise ValueError(f"未知平台: {raw}; 可选 tm,jd,dy")
+        if platform not in platforms:
+            platforms.append(platform)
+    if not platforms:
+        raise ValueError("至少指定一个平台")
+    return tuple(platforms)
+
+
+def build_platform_command(
+    platform: str,
+    *,
+    python_executable: str = sys.executable,
+) -> list[str]:
+    normalized = parse_platforms((platform,))[0]
+    commands = {
+        "tm": [
+            python_executable,
+            str(runtime_script("collect_sycm_market_rank.py")),
+            "--risk-mode",
+            "manual",
+        ],
+        "jd": [python_executable, str(runtime_script("collect_jd_market_rank.py"))],
+        "dy": [python_executable, str(runtime_script("collect_dy_market_rank.py"))],
+    }
+    return commands[normalized]
+
+
+def run_platform(
+    *,
+    platform: str,
+    command: list[str],
+    log_path: Path,
+    dry_run: bool,
+) -> dict[str, Any]:
+    """运行一个平台并把 stdout/stderr 同时写入独立日志与当前终端。"""
+    started_at = now_iso()
+    started_clock = datetime.now()
+    log_path.parent.mkdir(parents=True, exist_ok=True)
+    printable = subprocess.list2cmdline(command)
+    label = PLATFORM_LABELS[platform]
+    print(f"\n[RUN] {label}: {printable}", flush=True)
+
+    if dry_run:
+        log_path.write_text(f"[DRY-RUN] {printable}\n", encoding="utf-8")
+        return {
+            "platform": platform,
+            "label": label,
+            "command": command,
+            "log_path": str(log_path),
+            "started_at": started_at,
+            "finished_at": now_iso(),
+            "duration_seconds": 0.0,
+            "exit_code": 0,
+            "status": "dry_run",
+        }
+
+    exit_code = 126
+    error: str | None = None
+    try:
+        with log_path.open("w", encoding="utf-8", newline="") as log_file:
+            log_file.write(f"[START] {started_at}\n[COMMAND] {printable}\n\n")
+            log_file.flush()
+            with subprocess.Popen(
+                command,
+                cwd=PROJECT_ROOT,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.STDOUT,
+                text=True,
+                encoding="utf-8",
+                errors="replace",
+                bufsize=1,
+                env=build_subprocess_env(),
+            ) as process:
+                if process.stdout is not None:
+                    for line in process.stdout:
+                        log_file.write(line)
+                        log_file.flush()
+                        forwarded = console_safe_text(f"[{platform}] {line}")
+                        print(forwarded, end="", flush=True)
+                exit_code = process.wait()
+    except (OSError, subprocess.SubprocessError) as exc:
+        error = f"{type(exc).__name__}: {exc}"
+        with log_path.open("a", encoding="utf-8") as log_file:
+            log_file.write(f"\n[ERROR] {error}\n")
+
+    duration = round((datetime.now() - started_clock).total_seconds(), 3)
+    status = "success" if exit_code == 0 else "failed"
+    result = {
+        "platform": platform,
+        "label": label,
+        "command": command,
+        "log_path": str(log_path),
+        "started_at": started_at,
+        "finished_at": now_iso(),
+        "duration_seconds": duration,
+        "exit_code": exit_code,
+        "status": status,
+    }
+    if error:
+        result["error"] = error
+    print(
+        f"[{status.upper()}] {label}: exit={exit_code}, "
+        f"duration={duration}s, log={log_path}",
+        flush=True,
+    )
+    return result
+
+
+def execute_platform_step(
+    *,
+    platform: str,
+    command: list[str],
+    log_path: Path,
+    dry_run: bool,
+    executor: Callable[..., dict[str, Any]],
+) -> dict[str, Any]:
+    """并行任务边界：把异常转换为结果，避免取消其他平台。"""
+    step_started_at = now_iso()
+    try:
+        return executor(
+            platform=platform,
+            command=command,
+            log_path=log_path,
+            dry_run=dry_run,
+        )
+    except Exception as exc:
+        error = f"{type(exc).__name__}: {exc}"
+        log_path.write_text(f"[WORKFLOW ERROR] {error}\n", encoding="utf-8")
+        print(f"[FAILED] {PLATFORM_LABELS[platform]}: {error}", flush=True)
+        return {
+            "platform": platform,
+            "label": PLATFORM_LABELS[platform],
+            "command": command,
+            "log_path": str(log_path),
+            "started_at": step_started_at,
+            "finished_at": now_iso(),
+            "duration_seconds": 0.0,
+            "exit_code": 125,
+            "status": "failed",
+            "error": error,
+        }
+
+
+def execute_workflow(
+    platforms: Iterable[str] = DEFAULT_PLATFORMS,
+    *,
+    dry_run: bool = False,
+    run_id: str | None = None,
+    log_dir: Path = LOG_DIR,
+    python_executable: str = sys.executable,
+    executor: Callable[..., dict[str, Any]] | None = None,
+    notify_open_id: str | None = None,
+    notifier: Callable[..., dict[str, Any]] | None = None,
+) -> dict[str, Any]:
+    normalized = parse_platforms(platforms)
+    run_id = run_id or datetime.now().strftime("%Y%m%d_%H%M%S_%f")
+    log_dir = Path(log_dir)
+    log_dir.mkdir(parents=True, exist_ok=True)
+    execute_one = executor or run_platform
+    started_clock = datetime.now().astimezone()
+    started_at = started_clock.isoformat(timespec="seconds")
+
+    print("=" * 68)
+    print("三平台市场排行采集工作流")
+    print(f"run_id={run_id} platforms={','.join(normalized)} dry_run={dry_run}")
+    print("=" * 68, flush=True)
+
+    futures = {}
+    with ThreadPoolExecutor(
+        max_workers=len(normalized),
+        thread_name_prefix="market-rank",
+    ) as pool:
+        for platform in normalized:
+            command = build_platform_command(
+                platform=platform,
+                python_executable=python_executable,
+            )
+            futures[platform] = pool.submit(
+                execute_platform_step,
+                platform=platform,
+                command=command,
+                log_path=log_dir / f"{run_id}_{platform}.log",
+                dry_run=dry_run,
+                executor=execute_one,
+            )
+
+        # 并发完成，但汇总始终按调用方的平台顺序输出。
+        steps = [futures[platform].result() for platform in normalized]
+
+    exit_code = 1 if any(step.get("exit_code", 1) != 0 for step in steps) else 0
+    summary = {
+        "run_id": run_id,
+        "started_at": started_at,
+        "finished_at": now_iso(),
+        "dry_run": dry_run,
+        "platforms": list(normalized),
+        "exit_code": exit_code,
+        "status": "failed" if exit_code else ("dry_run" if dry_run else "success"),
+        "steps": steps,
+    }
+    if notify_open_id and not dry_run:
+        if notifier is None:
+            from market_rank_hermes_notification import notify_market_rank_reports
+
+            notifier = notify_market_rank_reports
+        try:
+            summary["notification"] = notifier(
+                report_date=date.today(),
+                recipient_open_id=notify_open_id,
+                generated_after=started_clock,
+            )
+        except Exception as exc:
+            summary["notification"] = {
+                "status": "failed",
+                "error": f"{type(exc).__name__}: {exc}",
+                "recipient_open_id": notify_open_id,
+            }
+        print(
+            "Hermes 通知: "
+            f"{summary['notification'].get('status')} "
+            f"({summary['notification'].get('reason', 'completed')})",
+            flush=True,
+        )
+        notification = summary["notification"]
+        notification_ok = notification.get("status") == "sent" or (
+            notification.get("status") == "skipped"
+            and notification.get("reason") == "already_sent"
+        )
+        if not notification_ok:
+            exit_code = 1
+            summary["exit_code"] = 1
+            summary["status"] = "failed"
+    summary_path = log_dir / f"{run_id}_summary.json"
+    summary["summary_path"] = str(summary_path)
+    summary_path.write_text(
+        json.dumps(summary, ensure_ascii=False, indent=2, default=str) + "\n",
+        encoding="utf-8",
+    )
+
+    print("\n" + "=" * 68)
+    for step in steps:
+        print(
+            f"{step['label'] if 'label' in step else step['platform']}: "
+            f"{step['status']} (exit={step['exit_code']})"
+        )
+    print(f"summary={summary_path}")
+    print(f"workflow_exit={summary['exit_code']}")
+    print("=" * 68, flush=True)
+    return summary
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="天猫、京东、抖音市场排行采集工作流")
+    parser.add_argument(
+        "--platforms",
+        default=",".join(DEFAULT_PLATFORMS),
+        help="平台列表，逗号分隔，默认 tm,jd,dy",
+    )
+    parser.add_argument("--dry-run", action="store_true", help="只验证命令，不执行采集")
+    parser.add_argument("--run-id", default=None, help="覆盖本次日志 run_id")
+    parser.add_argument("--log-dir", type=Path, default=LOG_DIR, help="日志目录")
+    parser.add_argument(
+        "--notify-openid",
+        default=DEFAULT_NOTIFY_OPEN_ID,
+        help="采集完成后由分析端 Hermes 私聊的飞书 open_id",
+    )
+    parser.add_argument(
+        "--no-notify",
+        action="store_true",
+        help="本次不发送三平台飞书文档链接",
+    )
+    args = parser.parse_args()
+
+    try:
+        platforms = parse_platforms(args.platforms)
+    except ValueError as exc:
+        parser.error(str(exc))
+    summary = execute_workflow(
+        platforms,
+        dry_run=args.dry_run,
+        run_id=args.run_id,
+        log_dir=managed_data_path(args.log_dir),
+        notify_open_id=None if args.no_notify else args.notify_openid,
+    )
+    return int(summary["exit_code"])
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/orchestrate_review_collection.py b/src/gyxx_flow/modules/product_commerce/runtime/orchestrate_review_collection.py
new file mode 100644
index 0000000..d90e2e8
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/orchestrate_review_collection.py
@@ -0,0 +1,189 @@
+"""评论采集编排器: 默认只跑 DY 采集 + 入库 PG。
+
+JD 评价采集代码已就绪 (jd_product_data_collector.py --reviews-only),
+但 JD 评价表只有 sku_id 缺 SPU, 暂时无法反查到款式, 所以编排器默认不跑 JD。
+等 dim_jd_sku_spu 映射表补齐后可加 --platforms jd,dy。
+
+DY 主流程 (dy_product_scraping.py 默认分支) 已挂载 _collect_comment_summary,
+daily --target-date 会顺带采评论; 本编排器用 --reviews-only 只跑评论 (跳过单品列表/
+下载/人群画像), 用于历史补采或重跑。
+入库调 import_product_reviews.py, ON CONFLICT (platform, order_id) 幂等覆盖。
+
+用法:
+    .venv/Scripts/python.exe orchestrate_review_collection.py
+    .venv/Scripts/python.exe orchestrate_review_collection.py --target-date 2026-07-13
+    .venv/Scripts/python.exe orchestrate_review_collection.py --platforms jd,dy  # JD 待 SPU 映射补齐
+    .venv/Scripts/python.exe orchestrate_review_collection.py --skip-collect   # 只入库
+    .venv/Scripts/python.exe orchestrate_review_collection.py --dry-run
+"""
+from __future__ import annotations
+
+import argparse
+import subprocess
+import sys
+from datetime import datetime, timedelta
+from pathlib import Path
+
+from runtime_paths import runtime_script
+
+from collect_retry_utils import FailedCollector
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+MAX_RUN_RETRIES = 2
+RUN_SLEEP_SECONDS = 5
+
+
+def default_target_date() -> str:
+    return (datetime.now().date() - timedelta(days=1)).strftime("%Y-%m-%d")
+
+
+def validate_date(value: str) -> str:
+    try:
+        datetime.strptime(value, "%Y-%m-%d")
+    except ValueError as exc:
+        raise argparse.ArgumentTypeError("date must be YYYY-MM-DD") from exc
+    return value
+
+
+def run_step(name, command, dry_run=False, critical=True):
+    """执行子命令。critical=True 失败抛 RuntimeError, False 仅 [WARN] 继续。"""
+    printable = " ".join(str(c) for c in command)
+    print(f"[RUN] {name}: {printable}")
+    if dry_run:
+        return {"name": name, "code": 0, "dry_run": True}
+    try:
+        proc = subprocess.run(
+            command,
+            cwd=PROJECT_ROOT,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+        )
+    except (FileNotFoundError, OSError) as exc:
+        msg = f"{name} {type(exc).__name__}: {exc}"
+        if critical:
+            raise RuntimeError(msg) from exc
+        print(f"[WARN] {msg}")
+        return {"name": name, "code": 126, "error": msg}
+
+    if proc.returncode:
+        msg = f"{name} failed with exit code {proc.returncode}"
+        if critical:
+            raise RuntimeError(msg)
+        print(f"[WARN] {msg}")
+        return {"name": name, "code": proc.returncode, "error": msg}
+    print(f"[OK] {name}")
+    return {"name": name, "code": proc.returncode}
+
+
+def run_step_with_retry(name, command, dry_run=False, critical=True,
+                        failed: FailedCollector | None = None,
+                        max_retries: int = MAX_RUN_RETRIES,
+                        sleep_seconds: float = RUN_SLEEP_SECONDS) -> dict:
+    if dry_run:
+        return run_step(name, command, dry_run=True, critical=critical)
+    last_result: dict = {"name": name, "code": 0}
+    total = max_retries + 1
+    for attempt in range(1, total + 1):
+        result = run_step(name, command, dry_run=False, critical=False)
+        last_result = result
+        if result.get("code") == 0:
+            if attempt > 1:
+                print(f"[OK] {name} 第 {attempt}/{total} 次重试成功", flush=True)
+            return result
+        err = result.get("error", "unknown")
+        if attempt >= total:
+            if failed is not None:
+                failed.add_page(name, f"{err} (attempts={attempt})", attempts=attempt)
+            if critical:
+                raise RuntimeError(f"{name} failed after {attempt} attempts: {err}")
+            print(f"[WARN] {name} 失败 (已重试 {max_retries} 次, {err}), 继续", flush=True)
+            return result
+        print(f"[RETRY] {name} 第 {attempt}/{total} 次失败, {sleep_seconds}s 后重试...", flush=True)
+        import time
+        time.sleep(sleep_seconds)
+    return last_result
+
+
+def build_collect_command(plat: str, target_date: str) -> list[str]:
+    if plat == "jd":
+        # JD --reviews-only: 只跑评价采集 (好/中/差全采), 跳过商智单品明细
+        return [sys.executable, str(runtime_script("jd_product_data_collector.py")),
+                "--reviews-only", "--target-date", target_date]
+    if plat == "dy":
+        # DY --reviews-only: 登录后直接跳评论页, 跳过单品列表/下载/人群画像
+        return [sys.executable, str(runtime_script("dy_product_scraping.py")),
+                "--reviews-only", "--target-date", target_date]
+    raise ValueError(f"unknown platform: {plat}")
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="评论采集编排器 (JD + DY)")
+    parser.add_argument("--target-date", type=validate_date, default=None,
+                        help="采集指定日期 YYYY-MM-DD (默认昨天)")
+    parser.add_argument("--platforms", default="dy",
+                        help="平台, 逗号分隔 (默认 dy; JD 待 SPU 映射补齐, 可手动 jd,dy)")
+    parser.add_argument("--skip-collect", action="store_true",
+                        help="跳过采集, 只跑 import_product_reviews.py")
+    parser.add_argument("--skip-import", action="store_true",
+                        help="跳过入库, 只跑采集")
+    parser.add_argument("--dry-run", action="store_true",
+                        help="只打印命令不执行")
+    args = parser.parse_args()
+
+    target_date = args.target_date or default_target_date()
+    platforms = [p.strip() for p in args.platforms.split(",") if p.strip()]
+    for p in platforms:
+        if p not in ("jd", "dy"):
+            print(f"[ERR] 未知平台: {p} (当前评论采集仅支持 jd, dy)")
+            return 2
+
+    today_str = datetime.now().strftime("%Y-%m-%d")
+    failed = FailedCollector("reviews", today_str)
+
+    print("=" * 60)
+    print("  评论采集编排器 (默认 DY)")
+    print("=" * 60)
+    print(f"  Target date: {target_date}")
+    print(f"  Platforms:   {platforms}")
+    print(f"  Skip collect: {args.skip_collect}")
+    print(f"  Skip import: {args.skip_import}")
+    print(f"  Dry run:     {args.dry_run}")
+    print("=" * 60 + "\n")
+
+    # Stage 1: 采集 (串行, Chrome profile 互锁)
+    if not args.skip_collect:
+        for plat in platforms:
+            cmd = build_collect_command(plat, target_date)
+            run_step_with_retry(
+                f"collect_{plat}", cmd,
+                dry_run=args.dry_run, critical=False, failed=failed,
+            )
+    else:
+        print("[INFO] --skip-collect: 跳过采集阶段\n")
+
+    # Stage 2: 入库
+    if not args.skip_import:
+        import_cmd = [sys.executable, str(runtime_script("import_product_reviews.py")),
+                      "--platforms", ",".join(platforms)]
+        if args.dry_run:
+            import_cmd.append("--dry-run")
+        run_step_with_retry(
+            "import_reviews", import_cmd,
+            dry_run=args.dry_run, critical=False, failed=failed,
+        )
+    else:
+        print("[INFO] --skip-import: 跳过入库阶段\n")
+
+    # 失败清单
+    failed_path = failed.write()
+    if failed_path:
+        print(f"\n[FAIL_LOG] {failed.summary_line()} -> {failed_path}")
+        return 1
+    print("\n[OK] 评论采集编排完成, 无失败")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/reapply_erp_override.py b/src/gyxx_flow/modules/product_commerce/runtime/reapply_erp_override.py
new file mode 100644
index 0000000..b69e738
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/reapply_erp_override.py
@@ -0,0 +1,133 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""回填 ERP 销量/退款到已有的平台 <款>_<日期>.json (aggregate 读的就是这个文件)。
+
+场景: ERP 跑晚了, 平台汇总文件已经按 ERP 不可用 写成 0/0 (或平台自报值)。
+解决: 对每个 (platform, style) 重新调 apply_erp_override, 它会读 data.json → 覆盖回去。
+
+用法:
+    .venv/Scripts/python.exe reapply_erp_override.py --date 2026-06-11
+"""
+import argparse
+import json
+from pathlib import Path
+
+from erp_metric_overrides import DATA_ROOT, PLATFORM_DIRS, apply_erp_override, parse_number
+
+PLATFORM_STYLE_FILENAME = {
+    "jd": lambda style, date: f"{style}_{date}.json",
+    "dy": lambda style, date: f"{style}_{date}.json",
+    "tm": lambda style, date: f"业务款式_{date}.json",
+}
+
+
+def safe_name(value: str) -> str:
+    return "".join("_" if ch in '\\/:*?"<>|' else ch for ch in str(value)).strip(" ._") or "未命名"
+
+
+def read_json(path: Path) -> dict:
+    return json.loads(path.read_text(encoding="utf-8-sig"))
+
+
+def write_json(path: Path, payload: dict) -> None:
+    path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8-sig")
+
+
+def reapply_for_style(platform: str, style_name: str, target_date: str) -> tuple[bool, str]:
+    """对 (platform, style, date) 重新跑 apply_erp_override 并写回 style 文件。
+    返回 (是否成功修改, 改后销量值)
+    """
+    style_dir = DATA_ROOT / PLATFORM_DIRS[platform] / f"{safe_name(style_name)}_{target_date}"
+    filename = PLATFORM_STYLE_FILENAME[platform](style_name, target_date)
+    json_path = style_dir / filename
+    if not json_path.exists():
+        return False, "no style file"
+
+    payload = read_json(json_path)
+
+    # 抽 visitors/cart/platform_sales/refunds, 然后重跑 override
+    if platform == "jd":
+        metrics = payload.get("指标汇总") or {}
+        visitors = parse_number(metrics.get("访客数"))
+        cart = parse_number(metrics.get("加购客户数") or metrics.get("加购人数"))
+        platform_sales = parse_number(metrics.get("销量"))
+        platform_refunds = parse_number(metrics.get("退款订单数"))
+    elif platform == "dy":
+        raw = payload.get("raw_totals") or {}
+        visitors = parse_number(raw.get("访客数") or payload.get("访客数"))
+        cart = parse_number(raw.get("加购人数") or payload.get("加购人数"))
+        platform_sales = parse_number(raw.get("销量") or payload.get("销量"))
+        platform_refunds = parse_number(raw.get("退款订单数"))
+    else:  # tm
+        style = payload.get("style") or payload
+        visitors = parse_number(style.get("访客量") or style.get("访客数"))
+        cart = parse_number(style.get("加购人数"))
+        platform_sales = parse_number(style.get("销量"))
+        platform_refunds = parse_number(style.get("退款订单数"))
+
+    erp_metrics = apply_erp_override(
+        platform, style_name, visitors, cart,
+        target_date=target_date,
+        platform_sales=platform_sales,
+        platform_refunds=platform_refunds,
+    )
+
+    # 把 override 后的值写回
+    if platform == "jd":
+        payload["指标汇总"]["销量"] = erp_metrics["销量"]
+        payload["指标汇总"]["退款订单数"] = erp_metrics["退款订单数"]
+        payload["指标汇总"]["转化率"] = erp_metrics["转化率"]
+        payload["指标汇总"]["退款率"] = erp_metrics["退款率"]
+        payload["指标汇总"]["ERP销量退款覆盖"] = erp_metrics["erp_override"]
+    elif platform == "dy":
+        if "raw_totals" not in payload:
+            payload["raw_totals"] = {}
+        payload["raw_totals"]["销量"] = erp_metrics["销量"]
+        payload["raw_totals"]["退款订单数"] = erp_metrics["退款订单数"]
+        # dy 风格文件顶层也有这些字段
+        payload["销量"] = erp_metrics["销量"]
+        payload["退款订单数"] = erp_metrics["退款订单数"]
+        payload["转化率"] = erp_metrics["转化率"]
+        payload["退款率"] = erp_metrics["退款率"]
+        payload["ERP销量退款覆盖"] = erp_metrics["erp_override"]
+    else:  # tm
+        if "style" not in payload:
+            payload["style"] = {}
+        payload["style"]["销量"] = erp_metrics["销量"]
+        payload["style"]["退款订单数"] = erp_metrics["退款订单数"]
+        payload["style"]["转化率"] = erp_metrics["转化率"]
+        payload["style"]["退款率"] = erp_metrics["退款率"]
+        payload["style"]["ERP销量退款覆盖"] = erp_metrics["erp_override"]
+
+    write_json(json_path, payload)
+    return True, str(erp_metrics["销量"])
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--date", required=True, help="YYYY-MM-DD")
+    args = parser.parse_args()
+    target_date = args.date
+
+    total = changed = 0
+    for platform in ("jd", "dy", "tm"):
+        base = DATA_ROOT / PLATFORM_DIRS[platform]
+        if not base.exists():
+            continue
+        for style_dir in base.glob(f"*_{target_date}"):
+            if not style_dir.is_dir():
+                continue
+            # 从目录名猜款式名: <safe_name>_<date>
+            style_name = style_dir.name[: -len(f"_{target_date}")]
+            total += 1
+            ok, new_sales = reapply_for_style(platform, style_name, target_date)
+            if ok:
+                changed += 1
+                print(f"  [{platform}] {style_name}: 销量→ {new_sales}")
+            else:
+                print(f"  [{platform}] {style_name}: 跳过 ({new_sales})")
+    print(f"\n[OK] {changed}/{total} 个款 ERP 覆盖已重跑")
+
+
+if __name__ == "__main__":
+    main()
\ No newline at end of file
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/rebuild_market_rank_documents.py b/src/gyxx_flow/modules/product_commerce/runtime/rebuild_market_rank_documents.py
new file mode 100644
index 0000000..caf7152
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/rebuild_market_rank_documents.py
@@ -0,0 +1,178 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""使用已采集 JSON 重建三平台一品类一表文档，并归档、通知。"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import traceback
+from concurrent.futures import ThreadPoolExecutor
+from datetime import date, datetime
+from pathlib import Path
+from typing import Any, Callable
+
+from collect_dy_market_rank import create_dy_feishu_doc_by_category
+from collect_jd_market_rank import create_jd_feishu_doc_by_category
+from collect_sycm_market_rank import create_feishu_doc_by_category
+from market_rank_hermes_notification import notify_market_rank_reports
+from market_rank_report_archive import archive_market_rank_report
+from runtime_paths import LOG_ROOT, RAW_DATA_ROOT
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+DEFAULT_RECIPIENT_OPEN_ID = "ou_7ad5fc8012e2f741afc5346e05ffd447"
+PLATFORM_CONFIG: dict[str, dict[str, Any]] = {
+    "tm": {
+        "label": "天猫",
+        "json": RAW_DATA_ROOT / "sycm_market_rank/market_rank_9_categories_30d_20260726_123620.json",
+        "period": "30天",
+    },
+    "jd": {
+        "label": "京东",
+        "json": RAW_DATA_ROOT / "jd_market_rank/jd_market_rank_20260726_115300.json",
+        "period": "近30天",
+    },
+    "dy": {
+        "label": "抖音",
+        "json": RAW_DATA_ROOT / "dy_market_rank/dy_market_rank_20260726_115940.json",
+        "period": "近30天",
+    },
+}
+
+
+def load_products(path: Path) -> list[dict[str, Any]]:
+    value = json.loads(path.read_text(encoding="utf-8"))
+    if not isinstance(value, list) or not value:
+        raise ValueError(f"采集结果为空或格式错误: {path}")
+    return value
+
+
+def rebuild_platform(platform: str) -> dict[str, Any]:
+    config = PLATFORM_CONFIG[platform]
+    path = Path(config["json"])
+    products = load_products(path)
+    categories = list(
+        dict.fromkeys(
+            str(item.get("category_name") or "未分类") for item in products
+        )
+    )
+    label = str(config["label"])
+    period = str(config["period"])
+    title = (
+        f"{label}{len(categories)}品类市场商品榜单 {period} "
+        f"{datetime.now():%Y-%m-%d}"
+    )
+
+    if platform == "tm":
+        doc_url = create_feishu_doc_by_category(
+            title,
+            products,
+            min_price=0,
+            max_price=0,
+            period=period,
+            independent_prices=True,
+        )
+    elif platform == "jd":
+        doc_url = create_jd_feishu_doc_by_category(
+            title,
+            products,
+            period=period,
+        )
+    else:
+        doc_url = create_dy_feishu_doc_by_category(
+            title,
+            products,
+            period=period,
+        )
+
+    report_id = archive_market_rank_report(
+        platform=platform,
+        report_date=date.today(),
+        period=period,
+        category_scope=f"{len(categories)}品类",
+        category_names=categories,
+        report_title=title,
+        feishu_doc_url=doc_url,
+        product_count=len(products),
+        source_json_path=str(path),
+        run_metadata={
+            "mode": "rebuild_from_existing_json",
+            "one_table_per_category": True,
+        },
+    )
+    return {
+        "platform": platform,
+        "status": "success",
+        "doc_url": doc_url,
+        "report_id": report_id,
+        "product_count": len(products),
+        "category_count": len(categories),
+    }
+
+
+def guarded_rebuild(platform: str) -> dict[str, Any]:
+    try:
+        return rebuild_platform(platform)
+    except Exception as exc:
+        traceback.print_exc()
+        return {
+            "platform": platform,
+            "status": "failed",
+            "error": f"{type(exc).__name__}: {exc}",
+        }
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument(
+        "--platforms",
+        default="tm,jd,dy",
+        help="逗号分隔的平台，默认 tm,jd,dy",
+    )
+    parser.add_argument(
+        "--recipient-open-id",
+        default=DEFAULT_RECIPIENT_OPEN_ID,
+    )
+    parser.add_argument("--no-notify", action="store_true")
+    return parser.parse_args()
+
+
+def main() -> int:
+    args = parse_args()
+    platforms = tuple(
+        dict.fromkeys(item.strip().lower() for item in args.platforms.split(",") if item.strip())
+    )
+    unknown = [item for item in platforms if item not in PLATFORM_CONFIG]
+    if unknown:
+        raise ValueError(f"未知平台: {unknown}")
+    started_at = datetime.now().astimezone()
+    with ThreadPoolExecutor(max_workers=len(platforms)) as pool:
+        futures = {platform: pool.submit(guarded_rebuild, platform) for platform in platforms}
+        results = [futures[platform].result() for platform in platforms]
+
+    notification: dict[str, Any] | None = None
+    if not args.no_notify:
+        notification = notify_market_rank_reports(
+            report_date=date.today(),
+            recipient_open_id=args.recipient_open_id,
+            generated_after=started_at,
+        )
+    summary = {
+        "started_at": started_at.isoformat(timespec="seconds"),
+        "finished_at": datetime.now().astimezone().isoformat(timespec="seconds"),
+        "results": results,
+        "notification": notification,
+    }
+    output = LOG_ROOT / "market_rank_workflow/rebuild_latest_summary.json"
+    output.parent.mkdir(parents=True, exist_ok=True)
+    output.write_text(
+        json.dumps(summary, ensure_ascii=False, indent=2, default=str) + "\n",
+        encoding="utf-8",
+    )
+    print(json.dumps(summary, ensure_ascii=False, indent=2, default=str))
+    return 0 if any(item["status"] == "success" for item in results) else 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/run_alerts_with_retry.py b/src/gyxx_flow/modules/product_commerce/runtime/run_alerts_with_retry.py
new file mode 100644
index 0000000..9d26a83
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/run_alerts_with_retry.py
@@ -0,0 +1,91 @@
+"""alert 单脚本的 per-run 重试 wrapper。
+
+check_nine_day_decline.py 是单 main() 脚本, 没有多阶段编排, 不适合在它内部加重试。
+新建这个 wrapper: 任何 rc != 0 都重试 2 次, 失败落 failed_<date>_alert.json。
+
+退出码语义沿用 check_nine_day_decline.py:
+  0  = 0 events 或 Hermes POST 成功
+  1  = 检测逻辑失败
+  2  = args 解析错误
+  3  = Hermes POST 失败 (事件已落盘)
+
+Wrapper 行为:
+  rc=0  → 退出 0 (成功)
+  rc!=0 → 重试 2 次 (5s sleep), 仍失败 → 写失败清单 → 透传原 rc (让任务计划程序看见)
+"""
+from __future__ import annotations
+
+import argparse
+import subprocess
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from runtime_paths import runtime_script
+
+from collect_retry_utils import FailedCollector, consolidate_failed_logs
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+PY = sys.executable
+MAX_RUN_RETRIES = 2
+RUN_SLEEP_SECONDS = 5
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="9 天连续下滑告警的 per-run 重试 wrapper (调 check_nine_day_decline.py)",
+    )
+    parser.add_argument("--end-date", required=True,
+                        help="9 天窗口收尾日 YYYY-MM-DD (通常是今天)")
+    parser.add_argument("--openid", required=True, help="Hermes 接收人 openid")
+    parser.add_argument("--max-run-retries", type=int, default=MAX_RUN_RETRIES,
+                        help=f"per-run 整体重试次数 (默认 {MAX_RUN_RETRIES})")
+    args = parser.parse_args()
+
+    cmd = [PY, str(runtime_script("check_nine_day_decline.py")), "--end-date", args.end_date,
+           "--openid", args.openid]
+    failed = FailedCollector("alert", args.end_date)
+
+    print(f"[INFO] target_date={args.end_date}")
+    print(f"[INFO] per-run 重试: {args.max_run_retries} 次, 间隔 {RUN_SLEEP_SECONDS}s")
+
+    total_attempts = args.max_run_retries + 1
+    last_rc = 1
+    for attempt in range(1, total_attempts + 1):
+        printable = " ".join(cmd)
+        print(f"\n[RUN] check_nine_day_decline (attempt {attempt}/{total_attempts}): {printable}",
+              flush=True)
+        try:
+            proc = subprocess.run(cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8",
+                                  errors="replace")
+        except FileNotFoundError as exc:
+            proc = subprocess.CompletedProcess(args=cmd, returncode=127, stderr=str(exc))
+        except OSError as exc:
+            proc = subprocess.CompletedProcess(args=cmd, returncode=126, stderr=str(exc))
+        last_rc = proc.returncode
+        if last_rc == 0:
+            print(f"[OK] alert (attempt {attempt}/{total_attempts})", flush=True)
+            consolidate_failed_logs(args.end_date)
+            return 0
+        if attempt >= total_attempts:
+            # 重试用尽
+            failed.add_page("check_nine_day_decline",
+                            f"exit {last_rc} (attempts={attempt})", attempts=attempt)
+            failed_path = failed.write()
+            consolidate_failed_logs(args.end_date)
+            if failed_path:
+                print(f"[FAIL_LOG] alert: {failed.summary_line()} -> {failed_path}", flush=True)
+            print(f"[FAIL] alert 失败 (已重试 {args.max_run_retries} 次, 最后 rc={last_rc})",
+                  flush=True)
+            return last_rc
+        print(f"[WARN] alert 失败 (attempt {attempt}/{total_attempts}, rc={last_rc}), "
+              f"等待 {RUN_SLEEP_SECONDS}s 后重试", flush=True)
+        time.sleep(RUN_SLEEP_SECONDS)
+
+    return last_rc  # unreachable
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/run_daily_persona.py b/src/gyxx_flow/modules/product_commerce/runtime/run_daily_persona.py
new file mode 100644
index 0000000..7f43dc5
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/run_daily_persona.py
@@ -0,0 +1,53 @@
+#!/usr/bin/env python3
+"""Run the three persona collectors in parallel and wait for real completion."""
+
+from __future__ import annotations
+
+import subprocess
+import sys
+from pathlib import Path
+
+from runtime_paths import LOG_ROOT, runtime_script
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+LOG_DIR = LOG_ROOT
+JOBS = {
+    "tm": (str(runtime_script("collect_persona_to_bitable.py")), "daily_persona_tm.log"),
+    "dy": (str(runtime_script("collect_dy_persona_to_bitable.py")), "daily_persona_dy.log"),
+    "jd": (str(runtime_script("collect_jd_persona_to_bitable.py")), "daily_persona_jd.log"),
+}
+
+
+def combined_exit_code(results: dict[str, int]) -> int:
+    return 0 if results and all(code == 0 for code in results.values()) else 2
+
+
+def main() -> int:
+    LOG_DIR.mkdir(parents=True, exist_ok=True)
+    running: dict[str, tuple[subprocess.Popen[bytes], object]] = {}
+    for platform, (script, log_name) in JOBS.items():
+        log_handle = (LOG_DIR / log_name).open("ab")
+        process = subprocess.Popen(
+            [sys.executable, str(PROJECT_ROOT / script)],
+            cwd=PROJECT_ROOT,
+            stdout=log_handle,
+            stderr=subprocess.STDOUT,
+        )
+        running[platform] = (process, log_handle)
+        print(f"[START] {platform} pid={process.pid} log={log_name}", flush=True)
+
+    results: dict[str, int] = {}
+    for platform, (process, log_handle) in running.items():
+        try:
+            results[platform] = process.wait()
+        finally:
+            log_handle.close()
+        print(f"[DONE] {platform} exit={results[platform]}", flush=True)
+
+    print(f"[SUMMARY] {results}", flush=True)
+    return combined_exit_code(results)
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/run_weekly_jd_main_image.py b/src/gyxx_flow/modules/product_commerce/runtime/run_weekly_jd_main_image.py
new file mode 100644
index 0000000..a1b70d2
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/run_weekly_jd_main_image.py
@@ -0,0 +1,168 @@
+"""周日编排：JD 主图采集 -> 飞书插入。
+
+默认每周日 08:30 触发 (Windows 任务计划程序), 跟 TM 主图 `run_weekly_main_image.py`
+同时跑但独立任务, 互不影响:
+  1. 跑 jd_main_image_collector.py 抓款式主图 + 曝光/点击
+  2. 跑 scripts/insert_jd_main_image_records.py 把数据推入每个款式的飞书主图 base + PG
+
+款式来源:
+  - --styles 显式传: 用 CLI 参数指定的款式列表
+  - --styles 留空: 动态从飞书多维表格读 (get_jd_spu_groups: 京东旗舰店分组有 SPU)
+
+数据落点: data/主图信息/<date>/<款式>/jd_main_images.json
+  (跟 TM wanxiang_creative_main_images.json 同目录不同名, 便于 insert 脚本区分)
+
+空数据容错 (跟 TM 一致):
+  - 采集后若 data/主图信息/<date>/ 下无有效 JD 款式, 落 [WARN] + 跳过 insert + 正常退出 (rc=0)
+  - 避免偶发空数据触发不必要的告警
+
+用法:
+    .venv/Scripts/python.exe run_weekly_jd_main_image.py
+    .venv/Scripts/python.exe run_weekly_jd_main_image.py --date 2026-07-05
+    .venv/Scripts/python.exe run_weekly_jd_main_image.py --skip-collect
+    .venv/Scripts/python.exe run_weekly_jd_main_image.py --skip-insert
+    .venv/Scripts/python.exe run_weekly_jd_main_image.py --styles 凌云3,宙斯
+"""
+import argparse
+import subprocess
+import sys
+from datetime import date, datetime, timedelta
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT, runtime_script
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+DATA_ROOT = RAW_DATA_ROOT / "主图信息"
+sys.path.insert(0, str(PROJECT_ROOT))
+from config.style_config_loader import get_jd_spu_groups  # noqa: E402
+
+
+def parse_date(value: str) -> date:
+    return datetime.strptime(value, "%Y-%m-%d").date()
+
+
+def run_step(name: str, command: list[str], critical: bool = True) -> dict:
+    """执行子命令。失败: critical=True 抛 RuntimeError; critical=False 仅打 [WARN]。"""
+    printable = " ".join(command)
+    print(f"[RUN] {name}: {printable}")
+    proc = subprocess.run(
+        command, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace"
+    )
+    if proc.returncode:
+        msg = f"{name} failed with exit code {proc.returncode}"
+        if critical:
+            raise RuntimeError(msg)
+        print(f"[WARN] {msg}")
+        return {"name": name, "code": proc.returncode, "error": msg}
+    print(f"[OK] {name}")
+    return {"name": name, "code": 0}
+
+
+def count_valid_styles(date_str: str) -> int:
+    """统计目标日期下有效 JD 款式数 (有 jd_main_images.json 且 records 非空)。"""
+    ddir = DATA_ROOT / date_str
+    if not ddir.exists():
+        return 0
+    n = 0
+    import json
+    for sub in ddir.iterdir():
+        if not sub.is_dir():
+            continue
+        jp = sub / "jd_main_images.json"
+        if not jp.exists():
+            continue
+        try:
+            payload = json.loads(jp.read_text(encoding="utf-8-sig"))
+            if payload.get("records"):
+                n += 1
+        except Exception:
+            pass
+    return n
+
+
+def main():
+    parser = argparse.ArgumentParser(description="周日编排：JD 主图采集 -> 飞书插入")
+    parser.add_argument("--date", default=None,
+                        help="采集日期 YYYY-MM-DD,默认「上周周日」(weekly 语义:这周插入的永远是上周数据)")
+    parser.add_argument("--skip-collect", action="store_true", help="跳过采集,只跑 insert")
+    parser.add_argument("--skip-insert", action="store_true", help="只采集,不推飞书")
+    parser.add_argument("--styles", default="",
+                        help="只采集/插入指定款式,逗号分隔 (传空=动态从飞书多维表格读京东旗舰店 SPU)")
+    parser.add_argument("--headless", action="store_true",
+                        help="采集时用无头浏览器 (服务器/无人值守场景)")
+    parser.add_argument("--as", dest="identity", default="user", choices=["user", "bot"])
+    args = parser.parse_args()
+
+    if args.date:
+        target_date = args.date
+    else:
+        today = datetime.now().date()
+        last_sunday = today - timedelta(days=(today.weekday() + 1) % 7)
+        target_date = last_sunday.strftime("%Y-%m-%d")
+    py = sys.executable
+
+    # --styles 为空时动态从飞书多维表格读 (京东旗舰店分组有 SPU)
+    if not args.styles:
+        try:
+            spu_groups = get_jd_spu_groups()
+            bitable_styles = [g["style"] for g in spu_groups]
+            args.styles = ",".join(bitable_styles)
+            print(f"[PREFLIGHT] 飞书多维表格动态读 JD SPU 款式 {len(bitable_styles)} 款")
+        except Exception as exc:
+            print(f"[WARN] 飞书读取失败, --styles 回退为空 (后续脚本各自兜底): {exc}")
+            args.styles = ""
+    else:
+        print(f"[PREFLIGHT] --styles 显式指定: {args.styles}")
+
+    print(f"[INFO] target_date={target_date} styles={args.styles or '(全部兜底)'} headless={args.headless}")
+    print("-" * 72)
+
+    # Step 1: 采集
+    if not args.skip_collect:
+        collect_cmd = [
+            py, str(runtime_script("jd_main_image_collector.py")),
+            "--output", str(DATA_ROOT),
+            "--date", target_date,
+        ]
+        if args.styles:
+            collect_cmd += ["--styles", args.styles]
+        if args.headless:
+            collect_cmd.append("--headless")
+        run_step("collect_jd_main_images", collect_cmd, critical=True)
+    else:
+        print("[SKIP] collect: --skip-collect")
+
+    # Step 2: 校验数据
+    valid = count_valid_styles(target_date)
+    if valid == 0:
+        # 跟 TM 一致: 全空数据落 warn + 跳过 insert 正常退出 (rc=0)
+        print(f"[WARN] data/主图信息/{target_date}/ 下没有有效 JD 款式数据, insert 跳过, 正常退出")
+        print(f"[WARN] 可能原因: 采集失败 / 日期目录不存在 / JD 商品明细页当日无 SPU 数据")
+        print("-" * 72)
+        print("[DONE] weekly JD main image pipeline complete (empty data, skipped insert)")
+        return
+    print(f"[INFO] {target_date} 有效 JD 款式数: {valid}")
+
+    # Step 3: 插入飞书 + PG
+    if not args.skip_insert:
+        insert_cmd = [
+            py, str(runtime_script("scripts/insert_jd_main_image_records.py")),
+            "--date", target_date,
+            "--as", args.identity,
+        ]
+        if args.styles:
+            insert_cmd += ["--style", args.styles]
+        run_step("insert_jd_main_image_records", insert_cmd, critical=True)
+    else:
+        print("[SKIP] insert: --skip-insert")
+
+    print("-" * 72)
+    print("[DONE] weekly JD main image pipeline complete")
+
+
+if __name__ == "__main__":
+    try:
+        main()
+    except Exception as e:
+        print(f"[FATAL] {e}", file=sys.stderr)
+        sys.exit(1)
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/run_weekly_main_image.py b/src/gyxx_flow/modules/product_commerce/runtime/run_weekly_main_image.py
new file mode 100644
index 0000000..4002881
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/run_weekly_main_image.py
@@ -0,0 +1,172 @@
+"""周日编排：主图采集 → 主图插入飞书。
+
+默认每周日 08:30 触发 (Windows 任务计划程序):
+  1. 跑 taobao_wanxiang_ai_creative_report.py 抓款式主图聚合数据
+  2. 跑 scripts/insert_main_image_records.py 把数据推入每个款式的飞书主图 base
+
+款式来源 (2026-06-27 改造):
+  - --styles 显式传: 用 CLI 参数指定的款式列表
+  - --styles 留空: 动态从飞书多维表格读 (get_main_image_styles: 天猫分组 + 主图飞书多维表格地址 同时存在)
+
+数据落点: data/主图信息/<date>/<款式>/wanxiang_creative_main_images.json
+插入幂等: 按 (时间, 曝光量, 点击量) 去重,失败抛 RuntimeError 让定时任务看见非零退出码。
+
+空数据容错 (2026-06-27 改造):
+  - 采集后若 data/主图信息/<date>/ 下无有效款式 (淘宝 wanxiang 偶发返回空),
+    不再 FATAL, 落 [WARN] + 跳过 insert + 正常退出 (rc=0)。
+  - 历史教训: 2026-06-21 任务失败但数据本来就空, 强制 FATAL 触发不必要的告警。
+
+用法:
+    .venv/Scripts/python.exe run_weekly_main_image.py
+    .venv/Scripts/python.exe run_weekly_main_image.py --date 2026-06-15
+    .venv/Scripts/python.exe run_weekly_main_image.py --skip-collect
+    .venv/Scripts/python.exe run_weekly_main_image.py --skip-insert
+    .venv/Scripts/python.exe run_weekly_main_image.py --styles 凌云3,宙斯
+"""
+import argparse
+import subprocess
+import sys
+from datetime import date, datetime, timedelta
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT, runtime_script
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+DATA_ROOT = RAW_DATA_ROOT / "主图信息"
+sys.path.insert(0, str(PROJECT_ROOT))
+from config.style_config_loader import get_main_image_styles  # noqa: E402
+
+
+def parse_date(value: str) -> date:
+    return datetime.strptime(value, "%Y-%m-%d").date()
+
+
+def run_step(name: str, command: list[str], critical: bool = True) -> dict:
+    """执行子命令。失败: critical=True 抛 RuntimeError; critical=False 仅打 [WARN]。"""
+    printable = " ".join(command)
+    print(f"[RUN] {name}: {printable}")
+    proc = subprocess.run(
+        command, cwd=PROJECT_ROOT, text=True, encoding="utf-8", errors="replace"
+    )
+    if proc.returncode:
+        msg = f"{name} failed with exit code {proc.returncode}"
+        if critical:
+            raise RuntimeError(msg)
+        print(f"[WARN] {msg}")
+        return {"name": name, "code": proc.returncode, "error": msg}
+    print(f"[OK] {name}")
+    return {"name": name, "code": 0}
+
+
+def count_valid_styles(date_str: str) -> int:
+    """统计目标日期下有效款式数 (有 wanxiang_creative_main_images.json 且非空)。"""
+    ddir = DATA_ROOT / date_str
+    if not ddir.exists():
+        return 0
+    n = 0
+    for sub in ddir.iterdir():
+        if not sub.is_dir():
+            continue
+        jp = sub / "wanxiang_creative_main_images.json"
+        if not jp.exists():
+            continue
+        try:
+            import json
+            payload = json.loads(jp.read_text(encoding="utf-8-sig"))
+            if payload.get("creatives"):
+                n += 1
+        except Exception:
+            pass
+    return n
+
+
+def main():
+    parser = argparse.ArgumentParser(description="周日编排：主图采集 → 飞书插入")
+    parser.add_argument("--date", default=None,
+                        help="采集日期 YYYY-MM-DD,默认「上周周日」(weekly 语义:这周插入的永远是上周数据)")
+    parser.add_argument("--skip-collect", action="store_true", help="跳过采集,只跑 insert")
+    parser.add_argument("--skip-insert", action="store_true", help="只采集,不推飞书")
+    parser.add_argument("--styles", default="",
+                        help="只采集/插入指定款式,逗号分隔 (传空=动态从飞书多维表格读天猫 + 主图飞书多维表格地址)")
+    parser.add_argument("--headless", action="store_true",
+                        help="采集时用无头浏览器 (服务器/无人值守场景)")
+    parser.add_argument("--max-pages", type=int, default=100,
+                        help="每个款式最多翻页数 (默认 100)")
+    parser.add_argument("--as", dest="identity", default="user", choices=["user", "bot"])
+    args = parser.parse_args()
+
+    if args.date:
+        target_date = args.date
+    else:
+        today = datetime.now().date()
+        last_sunday = today - timedelta(days=(today.weekday() + 1) % 7)
+        target_date = last_sunday.strftime("%Y-%m-%d")
+    py = sys.executable
+
+    # --styles 为空时动态从飞书多维表格读 (天猫分组 + 主图飞书多维表格地址 同时存在)
+    if not args.styles:
+        try:
+            bitable_styles = get_main_image_styles()
+            args.styles = ",".join(bitable_styles)
+            print(f"[PREFLIGHT] 飞书多维表格动态读主图款式 {len(bitable_styles)} 款")
+        except Exception as exc:
+            print(f"[WARN] 飞书读取失败, --styles 回退为空 (后续脚本各自兜底): {exc}")
+            args.styles = ""
+    else:
+        print(f"[PREFLIGHT] --styles 显式指定: {args.styles}")
+
+    print(f"[INFO] target_date={target_date} styles={args.styles or '(全部兜底)'} headless={args.headless}")
+    print("-" * 72)
+
+    # Step 1: 采集
+    if not args.skip_collect:
+        collect_cmd = [
+            py, str(runtime_script("taobao_wanxiang_ai_creative_report.py")),
+            "--output", str(DATA_ROOT),
+            "--date", target_date,
+            "--max-pages", str(args.max_pages),
+        ]
+        if args.styles:
+            collect_cmd += ["--styles", args.styles]
+        if args.headless:
+            collect_cmd.append("--headless")
+        run_step("collect_main_images", collect_cmd, critical=True)
+    else:
+        print("[SKIP] collect: --skip-collect")
+
+    # Step 2: 校验数据
+    valid = count_valid_styles(target_date)
+    if valid == 0:
+        # 2026-06-27 改造: 全空数据不再 FATAL, 落 warn + 跳过 insert 正常退出
+        # 原因: 淘宝 wanxiang 偶发返回空数据 (2026-06-21 nodata_skip=1258), 隔天即可恢复,
+        # 强制 FATAL 会让定时任务非 0 退出, 触发不必要的告警
+        print(f"[WARN] data/主图信息/{target_date}/ 下没有有效款式数据, insert 跳过, 正常退出")
+        print(f"[WARN] 可能原因: 采集失败 / 日期目录不存在 / 淘宝 wanxiang 当日无创意数据")
+        print("-" * 72)
+        print("[DONE] weekly main image pipeline complete (empty data, skipped insert)")
+        return
+    print(f"[INFO] {target_date} 有效款式数: {valid}")
+
+    # Step 3: 插入飞书
+    if not args.skip_insert:
+        insert_cmd = [
+            py, str(runtime_script("scripts/insert_main_image_records.py")),
+            "--date", target_date,
+            "--as", args.identity,
+        ]
+        if args.styles:
+            insert_cmd += ["--style", args.styles]
+        run_step("insert_main_image_records", insert_cmd, critical=True)
+    else:
+        print("[SKIP] insert: --skip-insert")
+
+    print("-" * 72)
+    print("[DONE] weekly main image pipeline complete")
+
+
+if __name__ == "__main__":
+    try:
+        main()
+    except Exception as e:
+        print(f"[FATAL] {e}", file=sys.stderr)
+        sys.exit(1)
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_paths.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_paths.py
new file mode 100644
index 0000000..b5ab0f9
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_paths.py
@@ -0,0 +1,144 @@
+"""Portable filesystem contract for the vendored product-commerce runtime."""
+
+from __future__ import annotations
+
+import os
+from pathlib import Path, PurePosixPath
+
+from gyxx_flow.core import DataLayout, Settings
+
+
+def _configured_root(variable: str, default: Path) -> Path:
+    configured = os.getenv(variable, "").strip()
+    return Path(configured).expanduser().resolve() if configured else default.resolve()
+
+
+MODULE_ROOT = _configured_root("GYXX_MODULE_ROOT", Path(__file__).resolve().parent)
+DATA_HOME = _configured_root("GYXX_DATA_ROOT", Settings.from_env().data_root)
+_DATA_PATHS = DataLayout(DATA_HOME).for_module("product_commerce")
+
+RAW_DATA_ROOT = _DATA_PATHS.raw_root
+NORMALIZED_DATA_ROOT = _DATA_PATHS.normalized_root
+CURATED_DATA_ROOT = _DATA_PATHS.curated_root
+EXPORTS_DATA_ROOT = _DATA_PATHS.exports_root
+EVIDENCE_DATA_ROOT = _DATA_PATHS.evidence_root
+STATE_ROOT = _DATA_PATHS.state_root
+PROFILE_ROOT = STATE_ROOT / "browser-profiles"
+LOG_ROOT = _DATA_PATHS.logs_root
+TMP_ROOT = _DATA_PATHS.tmp_root
+
+_VENDOR_ROOTS = {
+    "jd": MODULE_ROOT / "vendors" / "jd-data-flow",
+    "dy": MODULE_ROOT / "vendors" / "dy-data-flow",
+}
+
+
+def module_file(relative: str | Path, *, must_exist: bool = True) -> Path:
+    """Resolve a resource inside the module and reject path traversal."""
+
+    raw = str(relative).replace("\\", "/")
+    logical = PurePosixPath(raw)
+    if not raw or logical.is_absolute() or ".." in logical.parts:
+        raise ValueError(f"unsafe module-relative path: {relative!r}")
+    candidate = (MODULE_ROOT / Path(*logical.parts)).resolve()
+    if not candidate.is_relative_to(MODULE_ROOT):
+        raise ValueError(f"module path escapes runtime root: {relative!r}")
+    if must_exist and not candidate.is_file():
+        raise FileNotFoundError(candidate)
+    return candidate
+
+
+def runtime_script(relative: str | Path) -> Path:
+    """Resolve and validate a Python subprocess target owned by this runtime."""
+
+    candidate = module_file(relative)
+    if candidate.suffix.casefold() != ".py":
+        raise ValueError(f"runtime subprocess target must be Python: {relative!r}")
+    return candidate
+
+
+def vendor_source_root(platform: str) -> Path:
+    """Return an internal vendor source root; external checkout overrides are forbidden."""
+
+    try:
+        candidate = _VENDOR_ROOTS[platform].resolve(strict=True)
+    except KeyError as exc:
+        raise ValueError(f"unsupported vendor platform: {platform!r}") from exc
+    if not candidate.is_dir() or not candidate.is_relative_to(MODULE_ROOT):
+        raise RuntimeError(f"invalid internal vendor source root: {platform!r}")
+    return candidate
+
+
+def profile_path(name: str) -> Path:
+    """Return a portable profile path without creating it."""
+
+    if not name or any(character in name for character in ("/", "\\", "\x00")):
+        raise ValueError(f"invalid profile name: {name!r}")
+    configured = os.getenv("GYXX_BROWSER_PROFILE_DIR", "").strip()
+    return Path(configured).expanduser().resolve() if configured else PROFILE_ROOT / name
+
+
+def cookie_path(default_name: str = "cookies.json") -> Path:
+    """Return the current script's isolated cookie file when managed by GYXX."""
+
+    configured = os.getenv("GYXX_BROWSER_COOKIE_FILE", "").strip()
+    if configured:
+        return Path(configured).expanduser().resolve()
+    if not default_name or any(character in default_name for character in ("/", "\\", "\x00")):
+        raise ValueError(f"invalid cookie file name: {default_name!r}")
+    return STATE_ROOT / default_name
+
+
+def storage_state_path(default_name: str = "storage_state.json") -> Path:
+    configured = os.getenv("GYXX_BROWSER_STORAGE_STATE_FILE", "").strip()
+    if configured:
+        return Path(configured).expanduser().resolve()
+    if not default_name or any(character in default_name for character in ("/", "\\", "\x00")):
+        raise ValueError(f"invalid storage-state file name: {default_name!r}")
+    return STATE_ROOT / default_name
+
+
+def managed_data_path(configured: str | Path) -> Path:
+    """Resolve a mutable path and reject writes outside ``GYXX_DATA_ROOT``.
+
+    Relative paths remain CLI-compatible, but are interpreted from ``DATA_HOME``
+    rather than the caller's current working directory.
+    """
+
+    raw = str(configured)
+    if not raw or "\x00" in raw:
+        raise ValueError("managed data path cannot be empty")
+    path = Path(configured).expanduser()
+    candidate = (path if path.is_absolute() else DATA_HOME / path).resolve()
+    if not candidate.is_relative_to(DATA_HOME):
+        raise ValueError(f"managed path is outside GYXX_DATA_ROOT: {configured}")
+    return candidate
+
+
+def artifact_relative_path(path: str | Path) -> str:
+    """Return a portable artifact reference relative to ``GYXX_DATA_ROOT``."""
+
+    return managed_data_path(path).relative_to(DATA_HOME).as_posix()
+
+
+__all__ = [
+    "CURATED_DATA_ROOT",
+    "DATA_HOME",
+    "EXPORTS_DATA_ROOT",
+    "EVIDENCE_DATA_ROOT",
+    "LOG_ROOT",
+    "MODULE_ROOT",
+    "NORMALIZED_DATA_ROOT",
+    "PROFILE_ROOT",
+    "RAW_DATA_ROOT",
+    "STATE_ROOT",
+    "TMP_ROOT",
+    "artifact_relative_path",
+    "module_file",
+    "cookie_path",
+    "managed_data_path",
+    "profile_path",
+    "runtime_script",
+    "storage_state_path",
+    "vendor_source_root",
+]
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_backfill_poseidon_sales.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_backfill_poseidon_sales.py
new file mode 100644
index 0000000..f7b2ffe
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_backfill_poseidon_sales.py
@@ -0,0 +1,212 @@
+from __future__ import annotations
+
+import pandas as pd
+
+import backfill_poseidon_sales as backfill
+
+
+
+def test_backfill_uses_layered_runtime_roots():
+    assert backfill.DOWNLOAD_ROOT == backfill.RAW_DATA_ROOT
+    assert backfill.PLATFORM_OUTPUT_ROOT == backfill.NORMALIZED_DATA_ROOT
+    assert backfill.FINAL_OUTPUT_ROOT == backfill.CURATED_DATA_ROOT
+    assert backfill.REPORT_OUTPUT_ROOT == backfill.EXPORTS_DATA_ROOT
+
+
+def test_extract_jd_uses_existing_pipeline_metric_definitions():
+    frame = pd.DataFrame(
+        [
+            {
+                "时间": "2026-07-01",
+                "商品ID": "10035289085755",
+                "访客数": "1,234",
+                "加购人数": "23",
+                "成交单量": "7",
+                "成交商品件数": "9",
+                "取消及售后退款单量": "2",
+            }
+        ]
+    )
+
+    result = backfill.extract_metrics("jd", frame, ["10035289085755"], "2026-07-01")
+
+    assert result["matched_ids"] == ["10035289085755"]
+    assert result["metrics"] == {
+        "sales": 7,
+        "visitors": 1234,
+        "cart_users": 23,
+        "refund_orders": 2,
+    }
+
+
+def test_extract_dy_ignores_weekly_period_and_uses_deal_people():
+    frame = pd.DataFrame(
+        [
+            {
+                "统计周期": "2026/07/01-2026/07/07",
+                "商品编码": "3830202762700915010",
+                "商品点击人数": "999",
+                "加购人数": "99",
+                "成交人数": "88",
+                "成交订单数": "100",
+                "退款订单数(支付时间)": "9",
+            },
+            {
+                "统计周期": "2026/07/07",
+                "商品编码": "3830202762700915010",
+                "商品点击人数": "42",
+                "加购人数": "10",
+                "成交人数": "3",
+                "成交订单数": "4",
+                "退款订单数(支付时间)": "1",
+            },
+        ]
+    )
+
+    result = backfill.extract_metrics("dy", frame, ["3830202762700915010"], "2026-07-07")
+
+    assert result["matched_row_count"] == 1
+    assert result["metrics"] == {
+        "sales": 3,
+        "visitors": 42,
+        "cart_users": 10,
+        "refund_orders": 1,
+    }
+
+
+def test_extract_tm_sums_configured_ids_and_uses_paid_items():
+    frame = pd.DataFrame(
+        [
+            {
+                "统计日期": "2026-07-25",
+                "商品ID": "1061408417997",
+                "商品访客数": "395",
+                "商品加购人数": "17",
+                "支付买家数": "3",
+                "支付件数": "4",
+            },
+            {
+                "统计日期": "2026-07-25",
+                "商品ID": "1063521443835",
+                "商品访客数": "5",
+                "商品加购人数": "1",
+                "支付买家数": "1",
+                "支付件数": "2",
+            },
+        ]
+    )
+
+    result = backfill.extract_metrics(
+        "tm",
+        frame,
+        ["1060420778652", "1061408417997", "1063521443835"],
+        "2026-07-25",
+    )
+
+    assert result["matched_ids"] == ["1061408417997", "1063521443835"]
+    assert result["missing_ids"] == ["1060420778652"]
+    assert result["metrics"] == {
+        "sales": 6,
+        "visitors": 400,
+        "cart_users": 18,
+        "refund_orders": 0,
+    }
+
+
+def test_build_final_payload_preserves_excel_lineage():
+    parsed = {
+        "jd": {
+            "metrics": {"sales": 1, "visitors": 10, "cart_users": 2, "refund_orders": 0},
+            "matched_ids": ["jd-1"],
+            "missing_ids": [],
+            "matched_row_count": 1,
+            "source_file": "jd.xlsx",
+        },
+        "dy": {
+            "metrics": {"sales": 2, "visitors": 20, "cart_users": 3, "refund_orders": 1},
+            "matched_ids": ["dy-1"],
+            "missing_ids": [],
+            "matched_row_count": 1,
+            "source_file": "dy.xlsx",
+        },
+        "tm": {
+            "metrics": {"sales": 3, "visitors": 30, "cart_users": 4, "refund_orders": 0},
+            "matched_ids": ["tm-1"],
+            "missing_ids": [],
+            "matched_row_count": 1,
+            "source_file": "tm.xls",
+        },
+    }
+
+    payload = backfill.build_final_payload("2026-07-01", parsed, ["10515"])
+
+    assert payload["style_name"] == backfill.STYLE_NAME
+    assert payload["erp"]["style_codes"] == ["10515"]
+    assert payload["jd"]["sales"] == 1
+    assert payload["dy"]["source"] == "dy.xlsx"
+    assert payload["tm"]["matched_product_ids"] == ["tm-1"]
+
+
+def test_render_final_markdown_contains_all_three_platforms():
+    parsed = {
+        "jd": {
+            "metrics": {"sales": 1, "visitors": 10, "cart_users": 2, "refund_orders": 0},
+            "matched_ids": ["jd-1"], "missing_ids": [], "matched_row_count": 1, "source_file": "jd.xlsx",
+        },
+        "dy": {
+            "metrics": {"sales": 2, "visitors": 20, "cart_users": 3, "refund_orders": 1},
+            "matched_ids": ["dy-1"], "missing_ids": [], "matched_row_count": 1, "source_file": "dy.xlsx",
+        },
+        "tm": {
+            "metrics": {"sales": 3, "visitors": 30, "cart_users": 4, "refund_orders": 0},
+            "matched_ids": ["tm-1"], "missing_ids": [], "matched_row_count": 1, "source_file": "tm.xls",
+        },
+    }
+    payload = backfill.build_final_payload("2026-07-16", parsed, ["10515"])
+
+    markdown = backfill.render_final_markdown(payload)
+
+    assert "| jd | 1 | 0 | 10 | 2 |" in markdown
+    assert "| dy | 2 | 1 | 20 | 3 |" in markdown
+    assert "| tm | 3 | 0 | 30 | 4 |" in markdown
+
+
+def test_write_final_summary_updates_json_and_markdown(tmp_path, monkeypatch):
+    monkeypatch.setattr(backfill, "FINAL_OUTPUT_ROOT", tmp_path)
+    parsed = {
+        platform: {
+            "metrics": {"sales": 1, "visitors": 10, "cart_users": 2, "refund_orders": 0},
+            "matched_ids": [platform], "missing_ids": [], "matched_row_count": 1, "source_file": f"{platform}.xlsx",
+        }
+        for platform in ("jd", "dy", "tm")
+    }
+    payload = backfill.build_final_payload("2026-07-16", parsed, ["10515"])
+
+    json_path = backfill.write_final_summary(payload)
+
+    markdown_path = json_path.with_suffix(".md")
+    assert json_path.exists()
+    assert markdown_path.exists()
+    assert "| dy | 1 | 0 | 10 | 2 |" in markdown_path.read_text(encoding="utf-8-sig")
+
+
+def test_build_db_rows_expands_every_platform():
+    parsed = {
+        platform: {
+            "metrics": {"sales": index, "visitors": 10 * index, "cart_users": index, "refund_orders": 0},
+            "matched_ids": [platform], "missing_ids": [], "matched_row_count": 1, "source_file": f"{platform}.xlsx",
+        }
+        for index, platform in enumerate(("jd", "dy", "tm"), start=1)
+    }
+    payload = backfill.build_final_payload("2026-07-16", parsed, ["10515"])
+
+    rows = backfill.build_db_rows([payload])
+
+    assert len(rows) == 3
+    assert {(row["metric_date"], row["platform"]) for row in rows} == {
+        ("2026-07-16", "jd"),
+        ("2026-07-16", "dy"),
+        ("2026-07-16", "tm"),
+    }
+    assert next(row for row in rows if row["platform"] == "tm")["sales"] == 3
+
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_db_config.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_db_config.py
new file mode 100644
index 0000000..a4a5eb9
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_db_config.py
@@ -0,0 +1,42 @@
+import unittest
+
+from db import build_pg_config
+
+
+class DatabaseConfigTests(unittest.TestCase):
+    def test_complete_config_is_parsed(self):
+        config = build_pg_config(
+            {
+                "PG_HOST": "db.example.test",
+                "PG_PORT": "5544",
+                "PG_DB": "warehouse",
+                "PG_USER": "collector",
+                "PG_PASSWORD": "secret",
+            }
+        )
+
+        self.assertEqual(
+            config,
+            {
+                "host": "db.example.test",
+                "port": 5544,
+                "dbname": "warehouse",
+                "user": "collector",
+                "password": "secret",
+            },
+        )
+
+    def test_missing_config_does_not_fall_back_to_local_database(self):
+        with self.assertRaisesRegex(RuntimeError, "PG_HOST"):
+            build_pg_config(
+                {
+                    "PG_PORT": "5432",
+                    "PG_DB": "warehouse",
+                    "PG_USER": "collector",
+                    "PG_PASSWORD": "secret",
+                }
+            )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_dy_market_rank.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_dy_market_rank.py
new file mode 100644
index 0000000..9ebcc9b
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_dy_market_rank.py
@@ -0,0 +1,239 @@
+import unittest
+from unittest.mock import patch
+
+from collect_dy_market_rank import (
+    TARGET_DY_MARKET_CATEGORIES,
+    append_feishu_category_table,
+    build_dy_category_table_xml,
+    build_dy_feishu_batch_xml,
+    build_dy_feishu_intro_xml,
+    is_excluded_product_title,
+    parse_market_product_rank_html,
+    parse_money_range_upper,
+)
+from collect_sycm_market_rank import build_feishu_batch_xml
+
+
+SAMPLE_HTML = """
+<table>
+  <thead>
+    <tr>
+      <th>排名</th><th>商品</th><th>店铺名称</th>
+      <th>用户支付金额</th><th>点击次数</th><th>成交件数</th><th>操作</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr aria-hidden="true"><td>排名</td><td>商品</td><td></td><td></td></tr>
+    <tr class="aurora-table-row" data-row-key="3826298175627592082_1">
+      <td><div class="tagItem">首次上榜</div></td>
+      <td>
+        <img src="https://p9-aio.ecombdimg.com/product.jpg"/>
+        <div class="name-Gh2R04">测试旅行拉杆箱</div>
+        <div class="price-HH6TP1">价格带 <span>¥1889-¥2289</span></div>
+      </td>
+      <td>测试旗舰店</td>
+      <td><div>¥2,500万</div>-<div>¥5,000万</div></td>
+      <td>250万-500万</td><td>10万-25万</td><td>查看详情</td>
+    </tr>
+    <tr class="aurora-table-row" data-row-key="3826298175627592083_3">
+      <td><div>↑4</div></td>
+      <td>
+        <img src="https://p9-aio.ecombdimg.com/product-3.jpg"/>
+        <div elementtiming="pccp_element">第三名旅行箱</div>
+        <div>价格带 <span>¥999</span></div>
+      </td>
+      <td>第三名旗舰店</td>
+      <td><div>¥750万</div>-<div>¥1,000万</div></td>
+      <td>10万</td><td>1万</td><td>查看详情</td>
+    </tr>
+  </tbody>
+</table>
+"""
+
+
+class DyMarketRankTests(unittest.TestCase):
+    def test_title_blacklist_filters_children_primary_and_middle_school(self):
+        self.assertTrue(is_excluded_product_title("儿童轻便双肩包"))
+        self.assertTrue(is_excluded_product_title("小学生护脊书包"))
+        self.assertTrue(is_excluded_product_title("初中生大容量背包"))
+        self.assertTrue(
+            is_excluded_product_title(
+                "TigerFamily【老爸抽检】学生生日礼物1-3年级男女护脊减负折叠书包"
+            )
+        )
+        self.assertTrue(
+            is_excluded_product_title(
+                "【官方旗舰】AIRJORDAN耐克青少年运动休闲男女大号双肩书包9300"
+            )
+        )
+        self.assertTrue(is_excluded_product_title("老人轻便防水随身包"))
+        self.assertTrue(is_excluded_product_title("高中生大容量双肩包"))
+        self.assertFalse(is_excluded_product_title("成人通勤双肩包"))
+
+    def test_default_categories_match_requested_eight_filters(self):
+        specs = {item["name"]: item for item in TARGET_DY_MARKET_CATEGORIES}
+        self.assertEqual(
+            set(specs),
+            {"运动包", "旅行箱", "托特包", "双肩包", "斜挎包", "胸包", "手机包", "电脑包"},
+        )
+        self.assertEqual((specs["运动包"]["min_price"], specs["运动包"]["max_price"]), (200, 1000))
+        self.assertEqual((specs["旅行箱"]["min_price"], specs["旅行箱"]["max_price"]), (400, 2000))
+        self.assertEqual((specs["托特包"]["min_price"], specs["托特包"]["max_price"]), (200, 3000))
+        self.assertEqual(specs["托特包"]["search_keyword"], "托特包")
+        self.assertEqual((specs["手机包"]["min_price"], specs["手机包"]["max_price"]), (100, 1000))
+        self.assertEqual(specs["手机包"]["paths"][0][-1], "耳机包")
+        self.assertEqual((specs["电脑包"]["min_price"], specs["电脑包"]["max_price"]), (200, 1000))
+        self.assertEqual(specs["电脑包"]["search_keyword"], "电脑包")
+        self.assertEqual(specs["电脑包"]["paths"][0][-2:], ("功能箱包", "全部"))
+
+    def test_parse_money_range_upper(self):
+        self.assertEqual(parse_money_range_upper("¥2,500万-¥5,000万"), 50_000_000)
+        self.assertEqual(parse_money_range_upper("¥750万"), 7_500_000)
+
+    def test_parse_rank_html_extracts_requested_fields(self):
+        rows = parse_market_product_rank_html(
+            SAMPLE_HTML,
+            category_name="旅行箱",
+            category_path="鞋靴箱包 > 箱包 > 功能箱包 > 旅行箱/拉杆箱",
+        )
+
+        self.assertEqual(len(rows), 2)
+        row = rows[0]
+        self.assertEqual(row["rank"], 1)
+        self.assertEqual(row["product_id"], "3826298175627592082")
+        self.assertEqual(row["name"], "测试旅行拉杆箱")
+        self.assertEqual(row["image_url"], "https://p9-aio.ecombdimg.com/product.jpg")
+        self.assertEqual(row["price_range"], "¥1889-¥2289")
+        self.assertEqual(row["sales_amount_range"], "¥2,500万-¥5,000万")
+        self.assertEqual(row["sales_amount_upper"], 50_000_000)
+        self.assertEqual(
+            row["product_url"],
+            "https://haohuo.jinritemai.com/views/product/item2?id=3826298175627592082",
+        )
+        self.assertEqual(rows[1]["rank"], 3)
+
+    def test_parse_rank_html_drops_blacklisted_title(self):
+        rows = parse_market_product_rank_html(
+            SAMPLE_HTML.replace("测试旅行拉杆箱", "小学生测试旅行拉杆箱"),
+            category_name="旅行箱",
+            category_path="鞋靴箱包 > 箱包 > 功能箱包 > 旅行箱/拉杆箱",
+        )
+
+        self.assertEqual([row["name"] for row in rows], ["第三名旅行箱"])
+
+    def test_dy_feishu_xml_contains_sales_image_and_link(self):
+        product = parse_market_product_rank_html(
+            SAMPLE_HTML,
+            category_name="旅行箱",
+            category_path="鞋靴箱包 > 箱包 > 功能箱包 > 旅行箱/拉杆箱",
+        )[0]
+
+        xml = build_dy_feishu_batch_xml([product], start_rank=1)
+
+        self.assertIn(">销售额</th>", xml)
+        self.assertIn("¥2,500万-¥5,000万", xml)
+        self.assertIn('<img href="https://p9-aio.ecombdimg.com/product.jpg"', xml)
+        self.assertIn("https://haohuo.jinritemai.com/views/product/item2?id=3826298175627592082", xml)
+
+    def test_category_output_is_one_table_per_category(self):
+        product = parse_market_product_rank_html(
+            SAMPLE_HTML,
+            category_name="旅行箱",
+            category_path="鞋靴箱包 > 箱包 > 功能箱包 > 旅行箱/拉杆箱",
+        )[0]
+
+        xml = build_dy_category_table_xml("旅行箱", [product, dict(product, rank=2)])
+
+        self.assertEqual(xml.count("<table>"), 1)
+        self.assertEqual(xml.count("<th "), 7)
+        self.assertIn("<h2>旅行箱（2条）</h2>", xml)
+        self.assertIn("<td>1</td>", xml)
+        self.assertIn("<td>2</td>", xml)
+
+    def test_category_upload_timeout_is_verified_before_continuing(self):
+        product = parse_market_product_rank_html(
+            SAMPLE_HTML,
+            category_name="旅行箱",
+            category_path="鞋靴箱包 > 箱包 > 功能箱包 > 旅行箱/拉杆箱",
+        )[0]
+        with (
+            patch(
+                "collect_dy_market_rank.append_feishu_doc",
+                side_effect=RuntimeError("server time out error"),
+            ),
+            patch(
+                "collect_dy_market_rank._feishu_doc_contains_keyword",
+                side_effect=[False, True, True],
+            ) as verify,
+        ):
+            append_feishu_category_table(
+                "https://example.feishu.cn/docx/test",
+                "旅行箱",
+                [product],
+                2,
+            )
+
+        self.assertEqual(verify.call_count, 3)
+
+    def test_category_upload_skips_complete_existing_table(self):
+        product = parse_market_product_rank_html(
+            SAMPLE_HTML,
+            category_name="旅行箱",
+            category_path="鞋靴箱包 > 箱包 > 功能箱包 > 旅行箱/拉杆箱",
+        )[0]
+        with (
+            patch("collect_dy_market_rank.append_feishu_doc") as append,
+            patch(
+                "collect_dy_market_rank._feishu_doc_contains_keyword",
+                return_value=True,
+            ),
+        ):
+            append_feishu_category_table(
+                "https://example.feishu.cn/docx/test",
+                "旅行箱",
+                [product],
+                2,
+            )
+
+        append.assert_not_called()
+
+    def test_dy_feishu_intro_records_price_and_search_filters(self):
+        xml = build_dy_feishu_intro_xml(
+            title="抖音市场榜单",
+            period="近30天",
+            products=[
+                {
+                    "category_name": "托特包",
+                    "filter_min_price": 200,
+                    "filter_max_price": 3000,
+                    "search_keyword": "托特包",
+                }
+            ],
+        )
+
+        self.assertIn("托特包 ¥200-¥3000（搜索：托特包）", xml)
+        self.assertIn("标题过滤", xml)
+        for keyword in ("儿童", "学生", "青少年", "老人", "中学", "书包"):
+            self.assertIn(keyword, xml)
+
+    def test_tmall_feishu_table_also_has_sales_column(self):
+        xml = build_feishu_batch_xml(
+            [{
+                "category_name": "旅行箱",
+                "rank": 1,
+                "name": "测试商品",
+                "buyer_range": "2500~5000",
+                "buyer_max": 5000,
+                "price": "¥418",
+                "product_url": "https://item.taobao.com/item.htm?id=1",
+                "image_url": "https://img.alicdn.com/test.jpg",
+            }],
+            start_rank=1,
+        )
+
+        self.assertIn(">销售额</th>", xml)
+        self.assertIn(">未获取</td>", xml)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_dy_session_reuse.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_dy_session_reuse.py
new file mode 100644
index 0000000..61eea50
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_dy_session_reuse.py
@@ -0,0 +1,82 @@
+import json
+import sys
+from pathlib import Path
+
+
+VENDOR_ROOT = Path(__file__).resolve().parents[1] / "vendors" / "dy-data-flow"
+sys.path.insert(0, str(VENDOR_ROOT))
+
+from dy_store_competitor_store_scraping import (  # noqa: E402
+    _check_login_status,
+    _choose_session_root,
+    _looks_like_login_url,
+)
+
+
+def _write_cookies(root: Path, domains: list[str]) -> None:
+    root.mkdir(parents=True, exist_ok=True)
+    cookies = [{"name": f"c{index}", "value": "x", "domain": domain, "path": "/"}
+               for index, domain in enumerate(domains)]
+    (root / "compass_cookies.json").write_text(
+        json.dumps(cookies), encoding="utf-8"
+    )
+
+
+def test_choose_session_root_prefers_complete_sso_bundle(tmp_path):
+    project = tmp_path / "project"
+    legacy = tmp_path / "legacy"
+    _write_cookies(project, [".compass.jinritemai.com"])
+    _write_cookies(
+        legacy,
+        [
+            ".compass.jinritemai.com",
+            ".doudian-sso.jinritemai.com",
+            ".fxg.jinritemai.com",
+        ],
+    )
+
+    assert _choose_session_root([project, legacy]) == legacy
+
+
+def test_choose_session_root_honors_explicit_root(tmp_path):
+    project = tmp_path / "project"
+    explicit = tmp_path / "explicit"
+    _write_cookies(project, [".compass.jinritemai.com", ".fxg.jinritemai.com"])
+
+    assert _choose_session_root([project], explicit=explicit) == explicit
+
+
+def test_login_redirect_url_detection():
+    assert _looks_like_login_url("https://compass.jinritemai.com/login?redirect=x")
+    assert _looks_like_login_url("https://doudian-sso.jinritemai.com/login")
+    assert _looks_like_login_url("https://fxg.jinritemai.com/passport/sso")
+    assert not _looks_like_login_url(
+        "https://compass.jinritemai.com/shop/commodity/product-list"
+    )
+
+
+class _EmptyLocator:
+    def count(self):
+        return 0
+
+
+class _FakeContext:
+    pages = []
+
+
+class _FakeShopPage:
+    url = "https://compass.jinritemai.com/shop/commodity/product-list"
+    context = _FakeContext()
+
+    def __init__(self):
+        self.context.pages = [self]
+
+    def is_closed(self):
+        return False
+
+    def locator(self, _selector):
+        return _EmptyLocator()
+
+
+def test_shop_url_without_authenticated_dom_is_not_logged_in():
+    assert not _check_login_status(_FakeShopPage())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_metric_overrides.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_metric_overrides.py
new file mode 100644
index 0000000..83913a2
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_metric_overrides.py
@@ -0,0 +1,92 @@
+import json
+import tempfile
+import unittest
+from pathlib import Path
+from unittest.mock import patch
+
+import erp_metric_overrides as overrides
+
+
+class ErpMetricOverridesTests(unittest.TestCase):
+    def setUp(self):
+        self.temp_dir = tempfile.TemporaryDirectory()
+        self.data_root = Path(self.temp_dir.name)
+
+    def tearDown(self):
+        self.temp_dir.cleanup()
+
+    def _write_erp(self, code_results):
+        data_dir = self.data_root / "dy" / "apollo_2026-07-22"
+        data_dir.mkdir(parents=True)
+        (data_dir / "data.json").write_text(
+            json.dumps(
+                {
+                    "style_name": "阿波罗x1",
+                    "erp_style_codes": ["10388"],
+                    "date_start": "2026-07-22",
+                    "yesterday_sales": 0,
+                    "yesterday_returns": 0,
+                    "code_results": code_results,
+                },
+                ensure_ascii=False,
+            ),
+            encoding="utf-8",
+        )
+
+    def _apply(self, platform_sales, platform_refunds):
+        with patch.object(overrides, "DATA_ROOT", self.data_root), patch.dict(
+            overrides.PLATFORM_DIRS, {"dy": "dy"}
+        ):
+            return overrides.apply_erp_override(
+                "dy", "阿波罗x1", 394, 46,
+                target_date="2026-07-22",
+                platform_sales=platform_sales,
+                platform_refunds=platform_refunds,
+            )
+
+    def test_error_makes_ostensibly_ok_result_unavailable(self):
+        self._write_erp([{
+            "erp_style_code": "10388",
+            "ok": True,
+            "reason": "no_data",
+            "error": "timeout waiting for report result: 10388 | allsku iframe not found",
+            "yesterday_sales": 0,
+            "yesterday_returns": 0,
+        }])
+
+        result = self._apply(platform_sales=8, platform_refunds=2)
+
+        self.assertTrue(result["erp_unavailable"])
+        self.assertEqual(result["销量"], 8)
+        self.assertEqual(result["退款订单数"], 2)
+
+    def test_any_failed_code_disables_partial_erp_override(self):
+        self._write_erp([
+            {"erp_style_code": "1", "ok": True, "yesterday_sales": 3, "yesterday_returns": 1},
+            {"erp_style_code": "2", "ok": False, "error": "timeout", "yesterday_sales": 0, "yesterday_returns": 0},
+        ])
+
+        result = self._apply(platform_sales=7, platform_refunds=2)
+
+        self.assertTrue(result["erp_unavailable"])
+        self.assertEqual(result["销量"], 7)
+        self.assertEqual(result["退款订单数"], 2)
+
+    def test_clean_no_data_is_a_reliable_zero(self):
+        self._write_erp([{
+            "erp_style_code": "10388",
+            "ok": True,
+            "reason": "no_data",
+            "yesterday_sales": 0,
+            "yesterday_returns": 0,
+        }])
+
+        result = self._apply(platform_sales=8, platform_refunds=2)
+
+        self.assertFalse(result["erp_unavailable"])
+        self.assertEqual(result["销量"], 0)
+        self.assertEqual(result["退款订单数"], 0)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_no_data_freshness.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_no_data_freshness.py
new file mode 100644
index 0000000..1de81ab
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_no_data_freshness.py
@@ -0,0 +1,42 @@
+import unittest
+
+from collect_erp_yesterday_metrics import _no_data_report_is_fresh
+
+
+class ErpNoDataFreshnessTest(unittest.TestCase):
+    def test_rejects_unchanged_no_data_page_from_previous_query(self):
+        self.assertFalse(
+            _no_data_report_is_fresh(
+                previous_url="https://bi.erp321.com/allsku.aspx?r=old",
+                current_url="https://bi.erp321.com/allsku.aspx?r=old",
+                previous_body="NO_DATA",
+                current_body="NO_DATA",
+                reloaded=False,
+            )
+        )
+
+    def test_accepts_no_data_after_forced_reload(self):
+        self.assertTrue(
+            _no_data_report_is_fresh(
+                previous_url="https://bi.erp321.com/allsku.aspx?r=old",
+                current_url="https://bi.erp321.com/allsku.aspx?r=new",
+                previous_body="NO_DATA",
+                current_body="NO_DATA",
+                reloaded=True,
+            )
+        )
+
+    def test_accepts_no_data_when_report_body_changed(self):
+        self.assertTrue(
+            _no_data_report_is_fresh(
+                previous_url="https://bi.erp321.com/allsku.aspx?r=same",
+                current_url="https://bi.erp321.com/allsku.aspx?r=same",
+                previous_body="OLD REPORT",
+                current_body="NO_DATA",
+                reloaded=False,
+            )
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_slow_skip_codes.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_slow_skip_codes.py
new file mode 100644
index 0000000..4077339
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_erp_slow_skip_codes.py
@@ -0,0 +1,33 @@
+import ast
+import unittest
+from pathlib import Path
+
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+ERP_COLLECTOR = PROJECT_ROOT / "collect_erp_yesterday_metrics.py"
+
+
+def load_slow_skip_codes() -> set[str]:
+    tree = ast.parse(ERP_COLLECTOR.read_text(encoding="utf-8"))
+    for node in tree.body:
+        if isinstance(node, ast.AnnAssign) and isinstance(node.target, ast.Name):
+            if node.target.id == "ERP_SLOW_SKIP_CODES":
+                if (
+                    isinstance(node.value, ast.Call)
+                    and isinstance(node.value.func, ast.Name)
+                    and node.value.func.id == "set"
+                    and not node.value.args
+                    and not node.value.keywords
+                ):
+                    return set()
+                return set(ast.literal_eval(node.value))
+    raise AssertionError("ERP_SLOW_SKIP_CODES not found")
+
+
+class ErpSlowSkipCodesTest(unittest.TestCase):
+    def test_no_styles_are_permanently_skipped(self):
+        self.assertEqual(load_slow_skip_codes(), set())
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_feishu_doc_native.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_feishu_doc_native.py
new file mode 100644
index 0000000..f47f896
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_feishu_doc_native.py
@@ -0,0 +1,53 @@
+import pytest
+
+from feishu_doc_native import (
+    FeishuNativeDocClient,
+    FeishuNativeDocError,
+    append_native_category_table,
+    parse_category_table_xml,
+)
+
+
+def test_parse_category_table_xml_keeps_text_link_and_image():
+    table = parse_category_table_xml(
+        '<h2>旅行箱（1条）</h2><table><colgroup><col width="80"/>'
+        '<col width="120"/></colgroup><thead><tr><th>商品图</th><th>商品</th></tr>'
+        '<tbody><tr><td><img href="https://img.test/a.jpg"/></td>'
+        '<td><a href="https://item.test/1">打开商品</a></td></tr></tbody></table>'
+    )
+
+    assert table.heading == "旅行箱（1条）"
+    assert table.widths == [80, 120]
+    assert table.rows[1][0].image_url == "https://img.test/a.jpg"
+    assert table.rows[1][1].link_url == "https://item.test/1"
+
+
+def test_batch_update_splits_image_resources_into_twenty(monkeypatch):
+    client = FeishuNativeDocClient("test-token")
+    calls = []
+
+    def fake_request(method, path, **kwargs):
+        calls.append(kwargs["json_body"]["requests"])
+        return {}
+
+    monkeypatch.setattr(client, "request", fake_request)
+    requests_ = [
+        {"block_id": f"img-{index}", "replace_image": {"token": f"t-{index}"}}
+        for index in range(45)
+    ]
+
+    client.batch_update("doc", requests_)
+
+    assert [len(batch) for batch in calls] == [20, 20, 5]
+
+
+def test_rejects_table_over_feishu_2000_cell_limit_before_api_call():
+    rows = "".join("<tr>" + "<td>x</td>" * 8 + "</tr>" for _ in range(251))
+    xml = "<h2>超大类目</h2><table><tbody>" + rows + "</tbody></table>"
+
+    with pytest.raises(FeishuNativeDocError, match="超过飞书单表 2000"):
+        append_native_category_table(
+            "https://example.feishu.cn/docx/test",
+            xml,
+            client=object(),
+        )
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_guanghe_metadata.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_guanghe_metadata.py
new file mode 100644
index 0000000..3ae4983
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_guanghe_metadata.py
@@ -0,0 +1,469 @@
+import unittest
+from pathlib import Path
+from tempfile import TemporaryDirectory
+from unittest.mock import Mock, patch
+
+from upload_video_to_guanghe import (
+    GuangHeUploader,
+    _match_tm_item_ids_from_product_titles,
+    download_videos,
+)
+
+
+class GuangHeMetadataTests(unittest.TestCase):
+    def test_downloads_with_same_attachment_name_are_isolated_by_record(self):
+        attachment = {"file_token": "file-token", "name": "same-name.mp4"}
+
+        with TemporaryDirectory() as tmp_dir, patch(
+            "upload_video_to_guanghe.DOWNLOAD_DIR", Path(tmp_dir)
+        ), patch("upload_video_to_guanghe.run_lark") as run_lark_mock:
+            def fake_download(args):
+                output_dir = Path(args[args.index("--output") + 1])
+                output_dir.mkdir(parents=True, exist_ok=True)
+                (output_dir / attachment["name"]).write_bytes(b"video")
+                return 0, "", ""
+
+            run_lark_mock.side_effect = fake_download
+
+            first = download_videos("record-one", [attachment])
+            second = download_videos("record-two", [attachment])
+
+        self.assertEqual(first[0].parent.name, "record-one")
+        self.assertEqual(second[0].parent.name, "record-two")
+        self.assertNotEqual(first[0], second[0])
+
+    def test_unresolved_style_uses_exact_database_product_title_match(self):
+        result = _match_tm_item_ids_from_product_titles(
+            ["宙斯3", "未建档款"],
+            [
+                (
+                    "986141740958",
+                    "GYXX/光影行星宙斯3双肩包男士背包机能大容量",
+                ),
+                ("655191144133", "GYXX光影行星宙斯双肩包男士电脑包"),
+            ],
+        )
+
+        self.assertEqual(result, {"宙斯3": ["986141740958"]})
+
+    def test_product_search_term_removes_series_suffix(self):
+        self.assertEqual(
+            GuangHeUploader._normalize_product_search_term("宙斯系列"),
+            "宙斯",
+        )
+        self.assertEqual(
+            GuangHeUploader._normalize_product_search_term(" 宙斯 系列款 "),
+            "宙斯",
+        )
+        self.assertEqual(
+            GuangHeUploader._normalize_product_search_term("星迹2"),
+            "星迹2",
+        )
+
+    def test_cover_generation_pending_marker_is_not_upload_done(self):
+        self.assertTrue(GuangHeUploader._cover_generation_is_pending("封面生成中"))
+        self.assertFalse(GuangHeUploader._cover_generation_is_pending("智能封面图"))
+
+    def test_content_label_option_allows_browse_count_but_rejects_longer_label(self):
+        self.assertTrue(
+            GuangHeUploader._content_label_text_matches("# 斜挎包 41673031次浏览", "斜挎包")
+        )
+        self.assertTrue(
+            GuangHeUploader._content_label_text_matches("商品展示", "商品展示")
+        )
+        self.assertTrue(
+            GuangHeUploader._content_label_text_matches("#斜挎包", "斜挎包")
+        )
+        self.assertFalse(
+            GuangHeUploader._content_label_text_matches("# 单肩斜挎包 32871826次浏览", "斜挎包")
+        )
+
+    def test_topic_result_card_requires_topic_and_activity_statistics(self):
+        self.assertTrue(
+            GuangHeUploader._topic_result_card_text_matches(
+                "# 在淘宝种草一夏 小二推荐 作品数469.5万 参与数16.2万",
+                "在淘宝种草一夏",
+            )
+        )
+        self.assertFalse(
+            GuangHeUploader._topic_result_card_text_matches(
+                "在淘宝种草一夏",
+                "在淘宝种草一夏",
+            )
+        )
+        self.assertTrue(GuangHeUploader._topic_card_box_is_clickable(620, 120))
+        self.assertFalse(GuangHeUploader._topic_card_box_is_clickable(620, 45))
+        self.assertFalse(GuangHeUploader._topic_card_box_is_clickable(720, 600))
+
+    def test_poseidon_edge_uses_authoritative_tote_category(self):
+        self.assertEqual(
+            GuangHeUploader._resolve_title_category(
+                "波塞冬edge",
+                "双肩包、电脑包、托特包、大容量",
+            ),
+            "托特包",
+        )
+        self.assertEqual(
+            GuangHeUploader._resolve_title_category("星迹2", ""),
+            "斜挎包",
+        )
+
+    def test_invalid_online_info_is_rejected_when_main_category_conflicts(self):
+        result = GuangHeUploader._validate_online_product_info(
+            "波塞冬edge",
+            "主包型：双肩包；尺寸：39cm；容量/适配：16寸；材质：尼龙；卖点：分区",
+        )
+        self.assertEqual(result, "")
+
+    def test_verified_fallback_info_contains_all_product_fields(self):
+        info = GuangHeUploader._verified_product_info("星迹2")
+        for field in ["主包型：", "尺寸：", "容量/适配：", "材质：", "卖点："]:
+            self.assertIn(field, info)
+        self.assertIn("斜挎包", info)
+
+    def test_title_category_uses_name_then_search_features(self):
+        self.assertEqual(
+            GuangHeUploader._resolve_title_category("盖亚斜挎", "轻量、大容量"),
+            "斜挎包",
+        )
+        self.assertEqual(
+            GuangHeUploader._resolve_title_category("宙斯", "双肩包、大容量、笔电仓"),
+            "双肩包",
+        )
+
+    def test_title_normalization_keeps_search_structure_and_uses_computer_bag_term(self):
+        title = GuangHeUploader._normalize_generated_title(
+            "宙斯笔电大容量通勤真能装",
+            ["宙斯"],
+            "双肩包",
+        )
+
+        self.assertTrue(title.startswith("宙斯双肩包,"))
+        self.assertTrue(title.endswith("!"))
+        self.assertIn("电脑包", title)
+        self.assertNotIn("笔电", title)
+        self.assertNotIn("笔记本电脑", title)
+        self.assertNotIn("电脑包仓", title)
+        self.assertGreaterEqual(len(title), 26)
+        self.assertLessEqual(len(title), 30)
+
+    def test_title_normalization_removes_compact_duplicate_model_and_broken_tail(self):
+        title = GuangHeUploader._normalize_generated_title(
+            "光影行星宙斯zair双肩包,宙斯zair,大容量分区收纳,从容应",
+            ["宙斯z air"],
+            "双肩包",
+        )
+
+        self.assertTrue(title.startswith("宙斯z air双肩包,"))
+        self.assertEqual(title.replace(" ", "").count("宙斯zair"), 1)
+        self.assertNotIn("从容应!", title)
+        self.assertGreaterEqual(len(title), 26)
+        self.assertLessEqual(len(title), 30)
+
+    def test_same_style_multiple_videos_get_distinct_titles_when_hermes_repeats(self):
+        uploader = object.__new__(GuangHeUploader)
+        repeated = "光影行星宙斯双肩包科学分区收纳轻量通勤出行"
+        with (
+            patch.object(uploader, "_search_product_features", return_value=""),
+            patch("upload_video_to_guanghe.HAS_HERMES", True),
+            patch("upload_video_to_guanghe._call_hermes", return_value=(repeated, {})),
+        ):
+            first = uploader._generate_title(["宙斯"], video_index=1, video_total=2)
+            second = uploader._generate_title(["宙斯"], video_index=2, video_total=2)
+
+        self.assertNotEqual(first, second)
+        self.assertTrue(first.startswith("宙斯双肩包,"))
+        self.assertTrue(second.startswith("宙斯双肩包,"))
+
+    def test_same_style_repeated_hermes_copy_gets_low_similarity_bodies(self):
+        uploader = object.__new__(GuangHeUploader)
+        repeated = "光影行星宙斯双肩包,科学分区收纳,轻量通勤出行!"
+        with (
+            patch.object(uploader, "_search_product_features", return_value=""),
+            patch("upload_video_to_guanghe.HAS_HERMES", True),
+            patch("upload_video_to_guanghe._call_hermes", return_value=(repeated, {})),
+        ):
+            titles = [
+                uploader._generate_title(["宙斯"], video_index=index, video_total=4)
+                for index in range(1, 5)
+            ]
+
+        bodies = [GuangHeUploader._title_body(title) for title in titles]
+        self.assertEqual(len(set(bodies)), 4)
+        for index, body in enumerate(bodies):
+            for other in bodies[index + 1:]:
+                self.assertLess(
+                    GuangHeUploader._title_similarity(body, other),
+                    GuangHeUploader.TITLE_SIMILARITY_LIMIT,
+                )
+
+    def test_different_styles_cannot_reuse_same_title_skeleton(self):
+        uploader = object.__new__(GuangHeUploader)
+        repeated = "科学分区收纳,轻装通勤出行"
+        with (
+            patch.object(uploader, "_search_product_features", return_value=""),
+            patch("upload_video_to_guanghe.HAS_HERMES", True),
+            patch("upload_video_to_guanghe._call_hermes", return_value=(repeated, {})),
+        ):
+            first = uploader._generate_title(["宙斯"], video_index=1)
+            second = uploader._generate_title(["星云2"], video_index=1)
+
+        first_body = GuangHeUploader._title_body(first)
+        second_body = GuangHeUploader._title_body(second)
+        self.assertLess(
+            GuangHeUploader._title_similarity(first_body, second_body),
+            GuangHeUploader.TITLE_SIMILARITY_LIMIT,
+        )
+
+    def test_title_similarity_ignores_product_prefix(self):
+        first = "宙斯双肩包,科学分区收纳,轻装通勤出行!"
+        second = "星云2斜挎包,科学分区收纳,轻装通勤出行!"
+
+        self.assertEqual(GuangHeUploader._title_similarity(first, second), 1.0)
+
+    def test_full_batch_can_claim_48_globally_distinct_title_bodies(self):
+        uploader = object.__new__(GuangHeUploader)
+        titles = []
+        for index in range(48):
+            product_names = [f"款式{index % 12 + 1}"]
+            video_index = index // 12 + 1
+            candidate = uploader._fallback_title(
+                product_names,
+                "双肩包",
+                variant_index=video_index,
+            )
+            titles.append(
+                uploader._claim_unique_title(
+                    candidate,
+                    product_names,
+                    "双肩包",
+                    variant_index=video_index,
+                )
+            )
+
+        self.assertEqual(len(set(titles)), 48)
+        self.assertTrue(all(26 <= len(title) <= 30 for title in titles))
+        for index, title in enumerate(titles):
+            for other in titles[index + 1:]:
+                self.assertLess(
+                    GuangHeUploader._title_similarity(title, other),
+                    GuangHeUploader.TITLE_SIMILARITY_LIMIT,
+                )
+
+    def test_two_store_uploaders_can_share_one_global_title_registry(self):
+        shared_global = []
+        shared_styles = {}
+        shared_usage = {}
+        first_uploader = object.__new__(GuangHeUploader)
+        second_uploader = object.__new__(GuangHeUploader)
+        for uploader in (first_uploader, second_uploader):
+            uploader._used_titles_global = shared_global
+            uploader._used_titles_by_products = shared_styles
+            uploader._title_variant_usage = shared_usage
+
+        first = first_uploader._claim_unique_title(
+            first_uploader._fallback_title(["宙斯"], "双肩包", 1),
+            ["宙斯"],
+            "双肩包",
+            1,
+        )
+        second = second_uploader._claim_unique_title(
+            second_uploader._fallback_title(["星云2"], "斜挎包", 1),
+            ["星云2"],
+            "斜挎包",
+            1,
+        )
+
+        self.assertLess(
+            GuangHeUploader._title_similarity(first, second),
+            GuangHeUploader.TITLE_SIMILARITY_LIMIT,
+        )
+
+    def test_product_search_specs_prefer_tm_item_ids_and_fallback_to_keyword(self):
+        specs = GuangHeUploader._build_product_search_specs(
+            ["星云2", "未建档系列款"],
+            {"星云2": ["900176117674", "865283623976"]},
+        )
+
+        self.assertEqual(
+            specs,
+            [
+                {
+                    "product_name": "星云2",
+                    "search_term": "900176117674",
+                    "exact_id": True,
+                },
+                {
+                    "product_name": "未建档系列款",
+                    "search_term": "未建档",
+                    "exact_id": False,
+                },
+                {
+                    "product_name": "星云2",
+                    "search_term": "865283623976",
+                    "exact_id": True,
+                },
+            ],
+        )
+
+    def test_exact_id_searches_get_one_store_keyword_fallback(self):
+        specs = GuangHeUploader._build_product_search_specs(
+            ["极星双肩", "未建档款"],
+            {"极星双肩": ["972285625931", "966753757238"]},
+        )
+
+        self.assertEqual(
+            GuangHeUploader._build_exact_search_fallback_specs(specs),
+            [
+                {
+                    "product_name": "极星双肩",
+                    "search_term": "极星双肩",
+                    "exact_id": False,
+                }
+            ],
+        )
+
+    def test_title_fallback_contains_product_category_solution_and_result(self):
+        title = GuangHeUploader._fallback_title(["盖亚斜挎"], "斜挎包")
+
+        self.assertTrue(title.startswith("盖亚斜挎包,"))
+        self.assertIn(",", title)
+        self.assertTrue(title.endswith("!"))
+        self.assertIn("钥匙卡包各归其位", title)
+        self.assertIn("早高峰取物", title)
+        self.assertGreaterEqual(title.count(","), 2)
+        self.assertGreaterEqual(len(title), 26)
+        self.assertLessEqual(len(title), 30)
+
+    def test_generated_title_replaces_generic_category_with_model_category(self):
+        uploader = object.__new__(GuangHeUploader)
+        with (
+            patch.object(uploader, "_search_product_features", return_value=""),
+            patch("upload_video_to_guanghe.HAS_HERMES", True),
+            patch(
+                "upload_video_to_guanghe._call_hermes",
+                return_value=("光影行星宙斯双肩包大容量笔记本电脑轻量通勤", {}),
+            ),
+        ):
+            title = uploader._generate_title(["宙斯"])
+
+        self.assertTrue(title.startswith("宙斯双肩包,"))
+        self.assertNotIn("包袋双肩包", title)
+
+    def test_generated_poseidon_title_cannot_be_changed_to_backpack_by_hermes(self):
+        uploader = object.__new__(GuangHeUploader)
+        with (
+            patch.object(
+                uploader,
+                "_search_product_features",
+                return_value="双肩包、电脑包、托特包、大容量",
+            ),
+            patch("upload_video_to_guanghe.HAS_HERMES", True),
+            patch(
+                "upload_video_to_guanghe._call_hermes",
+                return_value=(
+                    "光影行星波塞冬edge双肩包暴雨通勤防泼水电脑仓从容出行",
+                    {},
+                ),
+            ),
+        ):
+            title = uploader._generate_title(["波塞冬edge"])
+
+        self.assertTrue(title.startswith("波塞冬edge托特包,"))
+        self.assertNotIn("双肩包", title)
+        self.assertTrue(title.endswith("!"))
+        self.assertGreaterEqual(len(title), 26)
+        self.assertLessEqual(len(title), 30)
+
+    def test_primary_content_label_uses_first_product_bag_type(self):
+        self.assertEqual(
+            GuangHeUploader._content_label_for_products(["盖亚斜挎", "极星双肩"]),
+            "斜挎包",
+        )
+        self.assertEqual(
+            GuangHeUploader._content_label_for_products(["盖亚微单"]),
+            "相机包",
+        )
+        self.assertEqual(
+            GuangHeUploader._content_label_for_products(["波塞冬edge"]),
+            "托特包",
+        )
+        self.assertEqual(
+            GuangHeUploader._content_label_for_products(["波塞冬手提电脑包"]),
+            "电脑包",
+        )
+
+    def test_catalog_product_titles_supply_style_category_and_verified_info(self):
+        try:
+            GuangHeUploader.configure_catalog_product_titles(
+                {
+                    "星云2": [
+                        "GYXX光影行星星云2斜挎包男士单肩包平板胸包",
+                        "光影行星星云斜挎包男款邮差单肩包",
+                    ],
+                    "宙斯": ["GYXX光影行星宙斯双肩包男大容量通勤背包"],
+                    "宙斯z air": ["光影行星宙斯zair双肩包电脑包通勤背包"],
+                    "黑曜石托特": ["光影行星黑曜石双肩托特多用包"],
+                }
+            )
+
+            self.assertEqual(
+                GuangHeUploader._content_label_for_products(["星云2"]),
+                "斜挎包",
+            )
+            self.assertEqual(
+                GuangHeUploader._resolve_title_category("宙斯"),
+                "双肩包",
+            )
+            self.assertIn(
+                "主包型：斜挎包",
+                GuangHeUploader._verified_product_info("星云2"),
+            )
+            self.assertEqual(
+                GuangHeUploader._content_label_for_products(["宙斯z air"]),
+                "双肩包",
+            )
+            self.assertEqual(
+                GuangHeUploader._content_label_for_products(["黑曜石托特"]),
+                "托特包",
+            )
+        finally:
+            GuangHeUploader.configure_catalog_product_titles({})
+
+    def test_metadata_selects_style_category_and_product_showcase_before_topic(self):
+        cases = [
+            (["盖亚斜挎"], ["斜挎包", "商品展示"]),
+            (["极星双肩"], ["双肩包", "商品展示"]),
+        ]
+        for product_names, expected_labels in cases:
+            with self.subTest(product_names=product_names):
+                uploader = object.__new__(GuangHeUploader)
+                calls = []
+                uploader._pick_content_labels = Mock(
+                    side_effect=lambda labels: calls.append(("labels", list(labels)))
+                )
+                uploader._pick_topic = Mock(
+                    side_effect=lambda topic: calls.append(("topic", topic))
+                )
+                uploader._snapshot = Mock()
+
+                uploader._prepare_publish_metadata(product_names)
+
+                self.assertEqual(
+                    calls,
+                    [
+                        ("labels", expected_labels),
+                        ("topic", "在淘宝种草一夏"),
+                    ],
+                )
+
+    def test_fixed_content_labels_only_keep_product_showcase(self):
+        self.assertEqual(GuangHeUploader.FIXED_CONTENT_LABELS, ("商品展示",))
+        self.assertNotIn("户外", GuangHeUploader.FIXED_CONTENT_LABELS)
+
+    def test_fixed_topic_is_not_a_recommended_first_result(self):
+        self.assertEqual(GuangHeUploader.FIXED_TOPIC, "在淘宝种草一夏")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_guanghe_store_routing.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_guanghe_store_routing.py
new file mode 100644
index 0000000..2e13e41
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_guanghe_store_routing.py
@@ -0,0 +1,118 @@
+import unittest
+
+from upload_video_to_guanghe import (
+    STORE_FLAGSHIP,
+    STORE_LUGGAGE,
+    TM_UPLOAD_FIELD,
+    select_pending_records,
+)
+
+
+def record(
+    *,
+    brand="光影行星",
+    gender="女",
+    uploaded=False,
+    attachments=True,
+    date=None,
+):
+    return {
+        "品牌": [brand] if brand is not None else None,
+        "性别": [gender] if gender is not None else None,
+        TM_UPLOAD_FIELD: uploaded,
+        "代拍附件": [{"name": "video.mp4"}] if attachments else [],
+        "日期": date,
+    }
+
+
+class GuangHeStoreRoutingTests(unittest.TestCase):
+    def test_tm_upload_field_matches_current_feishu_schema(self):
+        self.assertEqual(TM_UPLOAD_FIELD, "天猫上传")
+
+    def test_flagship_requires_brand_gender_attachment_and_unchecked_status(self):
+        records = [
+            record(gender="男"),
+            record(gender="女"),
+            record(gender="女", uploaded=True),
+            record(brand="OZKO"),
+            record(brand=None),
+            record(gender=None),
+            record(attachments=False),
+        ]
+
+        selected = select_pending_records(records, STORE_FLAGSHIP)
+
+        self.assertEqual(selected, records[:2])
+
+    def test_luggage_uploads_female_brand_records_regardless_of_tm_status(self):
+        unchecked = record(gender="女", uploaded=False)
+        checked = record(gender="女", uploaded=True)
+        unchecked["代拍产品"] = "极星双肩"
+        checked["代拍产品"] = "极星双肩"
+        records = [
+            unchecked,
+            checked,
+            record(gender="男"),
+            record(brand="OZKO", gender="女"),
+            record(gender=None),
+            record(gender="女", attachments=False),
+        ]
+
+        selected = select_pending_records(
+            records,
+            STORE_LUGGAGE,
+            luggage_style_whitelist={"极星双肩", "觅光"},
+        )
+
+        self.assertEqual(selected, [unchecked, checked])
+
+    def test_luggage_rejects_styles_outside_whitelist_and_trims_names(self):
+        allowed = record()
+        allowed["代拍产品"] = " 黑曜石电脑包 "
+        blocked = record()
+        blocked["代拍产品"] = "宙斯z air"
+        multi_with_blocked_style = record()
+        multi_with_blocked_style["代拍产品"] = "极星双肩 / 宙斯z air"
+
+        selected = select_pending_records(
+            [allowed, blocked, multi_with_blocked_style],
+            STORE_LUGGAGE,
+            luggage_style_whitelist={"黑曜石电脑包", "极星双肩"},
+        )
+
+        self.assertEqual(selected, [allowed])
+
+    def test_luggage_requires_a_loaded_whitelist(self):
+        with self.assertRaisesRegex(ValueError, "款式白名单"):
+            select_pending_records([record()], STORE_LUGGAGE)
+
+    def test_flagship_republish_from_date_includes_uploaded_records_in_range(self):
+        before = record(uploaded=True, date="2026-06-02 00:00:00")
+        first = record(uploaded=True, date="2026-06-03 00:00:00")
+        later = record(uploaded=True, date="2026-07-26 00:00:00")
+        missing_date = record(uploaded=True, date=None)
+
+        selected = select_pending_records(
+            [before, first, later, missing_date],
+            STORE_FLAGSHIP,
+            republish_from_date="2026-06-03",
+        )
+
+        self.assertEqual(selected, [first, later])
+
+    def test_republish_from_date_is_rejected_for_luggage_store(self):
+        with self.assertRaisesRegex(ValueError, "旗舰店"):
+            select_pending_records(
+                [record(date="2026-06-03 00:00:00")],
+                STORE_LUGGAGE,
+                luggage_style_whitelist={"极星双肩"},
+                republish_from_date="2026-06-03",
+            )
+
+    def test_unknown_store_is_rejected(self):
+        with self.assertRaisesRegex(ValueError, "未知光合店铺"):
+            select_pending_records([record()], "unknown")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_import_product_daily.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_import_product_daily.py
new file mode 100644
index 0000000..a9628c1
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_import_product_daily.py
@@ -0,0 +1,111 @@
+import unittest
+from decimal import Decimal
+from pathlib import Path
+from unittest.mock import patch
+
+import import_product_daily
+
+
+TM_31_HEADERS = [
+    "统计日期", "商品ID", "商品名称", "货号", "商品状态", "商品访客数",
+    "商品浏览量", "平均停留时长", "商品详情页跳出率", "商品收藏人数",
+    "商品加购件数", "商品加购人数", "下单买家数", "下单件数", "下单金额",
+    "下单转化率", "支付买家数", "支付件数", "支付金额", "商品支付转化率",
+    "支付新买家数", "支付老买家数", "老买家支付金额", "聚划算支付金额",
+    "访客平均价值", "成功退款金额", "搜索引导支付转化率", "搜索引导访客数",
+    "搜索引导支付买家数", "结构化详情引导转化率", "结构化详情引导成交占比",
+]
+
+TM_38_HEADERS = [
+    "统计日期", "商品ID", "商品名称", "主商品ID", "商品类型", "货号", "商品状态",
+    "商品标签", "商品访客数", "商品浏览量", "平均停留时长",
+    "商品详情页跳出率", "商品收藏人数", "商品加购件数", "商品加购人数",
+    "下单买家数", "下单件数", "下单金额", "下单转化率", "支付买家数",
+    "支付件数", "支付金额", "商品支付转化率", "支付新买家数", "支付老买家数",
+    "老买家支付金额", "聚划算支付金额", "访客平均价值", "成功退款金额",
+    "竞争力评分", "年累计支付金额", "月累计支付金额", "月累计支付件数",
+    "搜索引导支付转化率", "搜索引导访客数", "搜索引导支付买家数",
+    "结构化详情引导转化率", "结构化详情引导成交占比",
+]
+
+
+class _FakeSheet:
+    def __init__(self, headers, row):
+        self.nrows = 6
+        self._headers = headers
+        self._row = row
+
+    def row_values(self, row_index):
+        if row_index == 4:
+            return self._headers
+        if row_index == 5:
+            return self._row
+        return []
+
+
+class _FakeWorkbook:
+    def __init__(self, headers, row):
+        self._sheet = _FakeSheet(headers, row)
+
+    def sheet_by_index(self, _sheet_index):
+        return self._sheet
+
+
+def _tm_row(headers):
+    values = {
+        "统计日期": "2026-07-03",
+        "商品ID": "123456",
+        "商品名称": "测试商品",
+        "商品类型": "普通商品",
+        "货号": "SKU-001",
+        "商品访客数": "100",
+        "商品浏览量": "200",
+        "支付买家数": "8",
+        "支付件数": "10",
+        "支付金额": "1234.56",
+        "商品支付转化率": "8%",
+        "访客平均价值": "12.35",
+        "成功退款金额": "23.45",
+        "竞争力评分": "95",
+        "年累计支付金额": "9999.99",
+        "月累计支付金额": "5678.90",
+        "结构化详情引导成交占比": "45%",
+    }
+    row = [values.get(header, "") for header in headers]
+    return row
+
+
+class TmallProductDailyImportTests(unittest.TestCase):
+    def test_parse_31_column_export(self):
+        row = _tm_row(TM_31_HEADERS)
+        with patch.object(
+            import_product_daily.xlrd,
+            "open_workbook",
+            return_value=_FakeWorkbook(TM_31_HEADERS, row),
+        ):
+            records = list(import_product_daily.parse_tm_xls(Path("31-columns.xls")))
+
+        self.assertEqual(len(records), 1)
+        self.assertEqual(records[0]["product_no"], "SKU-001")
+        self.assertEqual(records[0]["visitors"], 100)
+        self.assertEqual(records[0]["paid_qty"], 10)
+        self.assertEqual(records[0]["paid_amount"], Decimal("1234.56"))
+        self.assertEqual(records[0]["refund_amount"], Decimal("23.45"))
+        self.assertNotIn("月累计支付金额", records[0]["raw_data"])
+
+    def test_parse_38_column_export_keeps_extended_fields(self):
+        row = _tm_row(TM_38_HEADERS)
+        with patch.object(
+            import_product_daily.xlrd,
+            "open_workbook",
+            return_value=_FakeWorkbook(TM_38_HEADERS, row),
+        ):
+            records = list(import_product_daily.parse_tm_xls(Path("38-columns.xls")))
+
+        self.assertEqual(len(records), 1)
+        self.assertEqual(records[0]["raw_data"]["月累计支付金额"], "5678.90")
+        self.assertEqual(records[0]["raw_data"]["结构化详情引导成交占比"], "45%")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_jd_enter_shangzhi.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_jd_enter_shangzhi.py
new file mode 100644
index 0000000..44368cf
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_jd_enter_shangzhi.py
@@ -0,0 +1,314 @@
+import importlib.util
+import sys
+import unittest
+from pathlib import Path
+from unittest.mock import patch
+
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+JD_VENDOR_DIR = PROJECT_ROOT / "vendors" / "jd-data-flow"
+MODULE_PATH = JD_VENDOR_DIR / "jd_data_collector.py"
+
+sys.path.insert(0, str(JD_VENDOR_DIR))
+try:
+    spec = importlib.util.spec_from_file_location("jd_data_collector_under_test", MODULE_PATH)
+    jd = importlib.util.module_from_spec(spec)
+    assert spec.loader is not None
+    spec.loader.exec_module(jd)
+finally:
+    sys.path.remove(str(JD_VENDOR_DIR))
+
+
+class FakeKeyboard:
+    def __init__(self, page):
+        self.page = page
+        self.presses = []
+
+    def press(self, key):
+        self.presses.append(key)
+
+
+class FakeLocator:
+    def __init__(self, visible=False, on_click=None):
+        self.visible = visible
+        self.on_click = on_click
+        self.clicks = 0
+
+    @property
+    def first(self):
+        return self
+
+    def is_visible(self, timeout=None):
+        return self.visible
+
+    def click(self, timeout=None):
+        self.clicks += 1
+        if self.on_click:
+            self.on_click()
+
+
+class FakeOverlayPage:
+    def __init__(self, close_button_works, overlay_visible=True, blocks_entry=True):
+        self.overlay_visible = overlay_visible
+        self.close_button_works = close_button_works
+        self.blocks_entry = blocks_entry
+        self.keyboard = FakeKeyboard(self)
+        self.evaluate_calls = []
+        self.waits = []
+
+    def locator(self, selector, **kwargs):
+        if selector == "div.all-dialog:visible":
+            return FakeLocator(visible=self.overlay_visible)
+        if self.close_button_works and "close-btn" in selector:
+            return FakeLocator(visible=True, on_click=self._close_overlay)
+        return FakeLocator(visible=False)
+
+    def _close_overlay(self):
+        self.overlay_visible = False
+
+    def evaluate(self, script, *args):
+        self.evaluate_calls.append(script)
+        if "elementFromPoint" in script and "setProperty('display'" in script:
+            if self.overlay_visible and self.blocks_entry:
+                self.overlay_visible = False
+                return 1
+            return 0
+        if "elementFromPoint" in script:
+            return self.overlay_visible and self.blocks_entry
+        if "all-dialog" in script and self.blocks_entry:
+            self.overlay_visible = False
+            return 1
+        if "all-dialog" in script:
+            # The old broad fallback would hide this non-blocking business dialog.
+            if "elementFromPoint" not in script:
+                self.overlay_visible = False
+                return 1
+        return 0
+
+    def wait_for_timeout(self, milliseconds):
+        self.waits.append(milliseconds)
+
+
+class FakePage:
+    def __init__(
+        self,
+        url="https://shop.jd.com/jdm/home",
+        click_actions=None,
+        wait_actions=None,
+    ):
+        self.url = url
+        self.closed = False
+        self.click_actions = list(click_actions or [])
+        self.entry_clicks = 0
+        self.keyboard = FakeKeyboard(self)
+        self.locator_calls = []
+        self.wait_actions = list(wait_actions or [])
+
+    def is_closed(self):
+        return self.closed
+
+    def locator(self, selector, **kwargs):
+        self.locator_calls.append((selector, kwargs))
+
+        def click_entry():
+            self.entry_clicks += 1
+            if self.click_actions:
+                action = self.click_actions.pop(0)
+                action()
+
+        is_precise_shangzhi_entry = (
+            selector == '.shop-menu-navigate__menu-name:has-text("商智"):visible'
+        )
+        return FakeLocator(visible=is_precise_shangzhi_entry, on_click=click_entry)
+
+    def wait_for_timeout(self, milliseconds):
+        if self.wait_actions:
+            self.wait_actions.pop(0)()
+        return None
+
+    def wait_for_load_state(self, *args, **kwargs):
+        return None
+
+    def goto(self, url, **kwargs):
+        self.url = url
+
+
+class FakeContext:
+    def __init__(self, page):
+        self.pages = [page]
+
+    def new_page(self):
+        page = FakePage()
+        self.pages.append(page)
+        return page
+
+
+class JdEnterShangzhiTests(unittest.TestCase):
+    def test_dismiss_clicks_real_close_button_before_dom_fallback(self):
+        page = FakeOverlayPage(close_button_works=True)
+
+        self.assertTrue(jd.dismiss_shop_home_overlays(page))
+
+        self.assertFalse(page.overlay_visible)
+        self.assertFalse(
+            any("setProperty('display'" in script for script in page.evaluate_calls)
+        )
+
+    def test_dismiss_hides_stuck_pointer_blocking_overlay(self):
+        page = FakeOverlayPage(close_button_works=False)
+
+        self.assertTrue(jd.dismiss_shop_home_overlays(page))
+
+        self.assertFalse(page.overlay_visible)
+        self.assertTrue(page.keyboard.presses)
+        self.assertTrue(any("all-dialog" in script for script in page.evaluate_calls))
+
+    def test_dismiss_does_not_touch_dom_when_no_overlay_is_visible(self):
+        page = FakeOverlayPage(close_button_works=False, overlay_visible=False)
+
+        self.assertTrue(jd.dismiss_shop_home_overlays(page))
+
+        self.assertEqual(page.keyboard.presses, [])
+        self.assertEqual(page.evaluate_calls, [])
+
+    def test_dismiss_does_not_hide_nonblocking_business_dialog(self):
+        page = FakeOverlayPage(
+            close_button_works=False,
+            overlay_visible=True,
+            blocks_entry=False,
+        )
+
+        self.assertTrue(jd.dismiss_shop_home_overlays(page))
+
+        self.assertTrue(page.overlay_visible)
+        self.assertFalse(any("setProperty('display'" in script for script in page.evaluate_calls))
+
+    def test_enter_retries_after_first_click_is_intercepted(self):
+        source = FakePage()
+        context = FakeContext(source)
+        destination = FakePage(url="https://sz.jd.com/szweb/index.html")
+
+        def fail_first_click():
+            raise RuntimeError("all-dialog intercepts pointer events")
+
+        def open_destination():
+            context.pages.append(destination)
+
+        source.click_actions = [fail_first_click, open_destination]
+        dismiss_calls = []
+        with (
+            patch.object(
+                jd,
+                "dismiss_shop_home_overlays",
+                side_effect=lambda page: dismiss_calls.append(page) or True,
+            ),
+            patch.object(jd, "close_popup_if_present", return_value=None),
+        ):
+            result = jd.step_enter_shangzhi(context, source)
+
+        self.assertIs(result, destination)
+        self.assertEqual(source.entry_clicks, 2)
+        self.assertGreaterEqual(len(dismiss_calls), 2)
+        self.assertFalse(any(selector == "text=商智" for selector, _ in source.locator_calls))
+
+    def test_enter_accepts_current_page_navigation(self):
+        source = FakePage()
+        context = FakeContext(source)
+        source.click_actions = [
+            lambda: setattr(source, "url", "https://sz.jd.com/szweb/index.html")
+        ]
+        with (
+            patch.object(jd, "dismiss_shop_home_overlays", return_value=True),
+            patch.object(jd, "close_popup_if_present", return_value=None),
+        ):
+            result = jd.step_enter_shangzhi(context, source)
+
+        self.assertIs(result, source)
+        self.assertEqual(source.entry_clicks, 1)
+
+    def test_enter_rejects_unrelated_new_page_and_retries(self):
+        source = FakePage()
+        context = FakeContext(source)
+        advertisement = FakePage(url="https://pro.jd.com/promotion.html")
+        destination = FakePage(url="https://sz.jd.com/szweb/index.html")
+        source.click_actions = [
+            lambda: context.pages.append(advertisement),
+            lambda: context.pages.append(destination),
+        ]
+        with (
+            patch.object(jd, "dismiss_shop_home_overlays", return_value=True),
+            patch.object(jd, "close_popup_if_present", return_value=None),
+        ):
+            result = jd.step_enter_shangzhi(context, source)
+
+        self.assertIs(result, destination)
+        self.assertEqual(source.entry_clicks, 2)
+
+    def test_enter_rejects_login_page_whose_return_url_mentions_shangzhi(self):
+        source = FakePage()
+        context = FakeContext(source)
+        login_page = FakePage(
+            url=(
+                "https://passport.jd.com/login?returnUrl="
+                "https%3A%2F%2Fsz.jd.com%2Fszweb%2Findex.html"
+            )
+        )
+        destination = FakePage(url="https://sz.jd.com/szweb/index.html")
+        source.click_actions = [
+            lambda: context.pages.append(login_page),
+            lambda: context.pages.append(destination),
+        ]
+        with (
+            patch.object(jd, "dismiss_shop_home_overlays", return_value=True),
+            patch.object(jd, "close_popup_if_present", return_value=None),
+        ):
+            result = jd.step_enter_shangzhi(context, source)
+
+        self.assertIs(result, destination)
+        self.assertEqual(source.entry_clicks, 2)
+
+    def test_shangzhi_url_match_uses_hostname_not_query_or_similar_domain(self):
+        self.assertTrue(jd._is_shangzhi_url("https://sz.jd.com/szweb/index.html"))
+        self.assertTrue(jd._is_shangzhi_url("https://sub.sz.jd.com/path"))
+        self.assertTrue(
+            jd._is_shangzhi_url(
+                "https://jdsz.jd.com/szweb/view/index/home.html"
+            )
+        )
+        self.assertFalse(
+            jd._is_shangzhi_url(
+                "https://passport.jd.com/login?returnUrl=https://sz.jd.com/szweb/"
+            )
+        )
+        self.assertFalse(jd._is_shangzhi_url("https://evil-sz.jd.com/szweb/"))
+        self.assertFalse(jd._is_shangzhi_url("https://evil-jdsz.jd.com/szweb/"))
+
+    def test_enter_waits_for_about_blank_popup_to_navigate_to_shangzhi(self):
+        source = FakePage()
+        context = FakeContext(source)
+        destination = FakePage(url="about:blank")
+        source.click_actions = [lambda: context.pages.append(destination)]
+        source.wait_actions = [
+            lambda: None,
+            lambda: setattr(
+                destination,
+                "url",
+                "https://jdsz.jd.com/szweb/view/index/home.html",
+            ),
+        ]
+        with (
+            patch.object(jd, "dismiss_shop_home_overlays", return_value=True),
+            patch.object(jd, "close_popup_if_present", return_value=None),
+        ):
+            result = jd.step_enter_shangzhi(context, source)
+
+        self.assertIs(result, destination)
+        self.assertEqual(
+            result.url,
+            "https://jdsz.jd.com/szweb/view/index/home.html",
+        )
+        self.assertEqual(source.entry_clicks, 1)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_jd_market_rank.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_jd_market_rank.py
new file mode 100644
index 0000000..d3eb7d1
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_jd_market_rank.py
@@ -0,0 +1,247 @@
+import unittest
+
+from collect_jd_market_rank import (
+    NEXT_PAGE_SELECTORS,
+    TARGET_JD_MARKET_CATEGORIES,
+    build_jd_feishu_batch_xml,
+    build_jd_feishu_category_xml,
+    build_jd_feishu_intro_xml,
+    high_resolution_jd_image_url,
+    group_products_by_category,
+    inspect_feishu_table_images,
+    parse_jd_market_rank_html,
+    parse_money_range_upper,
+    resize_feishu_table_xml,
+)
+
+
+SAMPLE_HTML = """
+<table class="jmtd-table-main">
+  <thead>
+    <tr>
+      <th>排名</th>
+      <th>商品信息</th>
+      <th>成交金额</th>
+      <th>成交单量</th>
+      <th>访客数</th>
+      <th>搜索点击次数</th>
+      <th>关注人数</th>
+      <th>操作</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td><span class="jmtd-table-formatter-rank-icon-1"></span><span>↑3名</span></td>
+      <td>
+        <div class="prod-info">
+          <a class="prod-name" href="//item.jd.com/10012345678901.html?pcdk=tracking"
+             title="测试高颜值拉杆箱">测试高颜值拉杆箱</a>
+          <a class="shop-name">测试旗舰店</a>
+          <img data-src="//img10.360buyimg.com/n1/jfs/test.jpg"/>
+        </div>
+      </td>
+      <td>¥ 50万 ～ ¥ 75万</td>
+      <td>4,000~6,000</td>
+      <td>4万~6万</td>
+      <td>4万~6万</td>
+      <td>357</td>
+      <td>详情</td>
+    </tr>
+    <tr>
+      <td><span class="jmtd-table-formatter-rank-number">4</span></td>
+      <td>
+        <div class="prod-info">
+          <a class="prod-name" href="https://item.jd.com/10012345678902.html">
+            测试电脑包
+          </a>
+          <a class="shop-name">电脑包旗舰店</a>
+          <img src="https://img11.360buyimg.com/n1/jfs/test2.jpg"/>
+        </div>
+      </td>
+      <td>¥10万~¥25万</td>
+      <td>50~100</td>
+      <td>4,000~6,000</td>
+      <td>1,000~2,000</td>
+      <td>20</td>
+      <td>详情</td>
+    </tr>
+  </tbody>
+</table>
+"""
+
+
+class JdMarketRankTests(unittest.TestCase):
+    def test_default_categories_match_requested_seven_filters(self):
+        specs = {item["name"]: item for item in TARGET_JD_MARKET_CATEGORIES}
+        self.assertEqual(
+            set(specs),
+            {"旅行箱", "运动包", "双肩包", "斜挎包", "胸包", "手机包", "电脑包"},
+        )
+        self.assertEqual(specs["旅行箱"]["path"], ("功能箱包", "行李箱"))
+        self.assertEqual(specs["运动包"]["path"], ("功能箱包", "休闲运动包"))
+        self.assertEqual(specs["双肩包"]["path"], ("男包", "男士双肩包"))
+        self.assertEqual(specs["手机包"]["path"], ("男包", "男士手机包"))
+        self.assertEqual(specs["电脑包"]["path"], ("功能箱包", "电脑包"))
+
+    def test_parse_money_range_upper(self):
+        self.assertEqual(parse_money_range_upper("¥50万~¥75万"), 750_000)
+        self.assertEqual(parse_money_range_upper("¥ 1亿以上"), 100_000_000)
+
+    def test_jd_thumbnail_is_upgraded_to_n0_original(self):
+        self.assertEqual(
+            high_resolution_jd_image_url(
+                "https://img10.360buyimg.com/n5/jfs/t1/test/image.jpg"
+            ),
+            "https://img10.360buyimg.com/n0/jfs/t1/test/image.jpg",
+        )
+
+    def test_real_jd_next_page_selector_is_supported(self):
+        self.assertIn(".jmtd-pagination-item-next", NEXT_PAGE_SELECTORS)
+
+    def test_inspect_feishu_table_images_finds_incomplete_table(self):
+        content = """
+        <title>测试</title>
+        <table id="table-1">
+          <thead><tr><th>商品图</th></tr></thead>
+          <tbody>
+            <tr><td><img src="token-1"/></td></tr>
+            <tr><td></td></tr>
+          </tbody>
+        </table>
+        <table id="table-2">
+          <thead><tr><th>商品图</th></tr></thead>
+          <tbody><tr><td><img src="token-2"/></td></tr></tbody>
+        </table>
+        """
+
+        stats = inspect_feishu_table_images(content)
+
+        self.assertEqual(
+            stats,
+            [
+                {"block_id": "table-1", "rows": 2, "images": 1},
+                {"block_id": "table-2", "rows": 1, "images": 1},
+            ],
+        )
+
+    def test_resize_feishu_table_xml_reuses_media_token_at_compact_size(self):
+        table_xml = """
+        <table id="table-1">
+          <tbody>
+            <tr id="row-1">
+              <td>
+                <img id="image-1" src="image-token" href="https://internal/image"
+                     width="220" height="800"/>
+              </td>
+              <td><a href="https://item.jd.com/1.html">打开商品</a></td>
+            </tr>
+          </tbody>
+        </table>
+        """
+
+        resized = resize_feishu_table_xml(table_xml, image_size=160)
+
+        self.assertNotIn('id="table-1"', resized)
+        self.assertNotIn('id="image-1"', resized)
+        self.assertIn(
+            '<img src="image-token" width="160" height="160"/>',
+            resized,
+        )
+        self.assertIn(
+            '<a href="https://item.jd.com/1.html">打开商品</a>',
+            resized,
+        )
+
+    def test_parse_html_extracts_image_name_link_and_sales(self):
+        rows = parse_jd_market_rank_html(
+            SAMPLE_HTML,
+            category_name="旅行箱",
+            category_path="功能箱包 > 行李箱",
+        )
+
+        self.assertEqual(len(rows), 2)
+        first = rows[0]
+        self.assertEqual(first["platform"], "jd")
+        self.assertEqual(first["rank"], 1)
+        self.assertEqual(first["product_id"], "10012345678901")
+        self.assertEqual(first["name"], "测试高颜值拉杆箱")
+        self.assertEqual(
+            first["image_url"],
+            "https://img10.360buyimg.com/n0/jfs/test.jpg",
+        )
+        self.assertEqual(
+            first["product_url"],
+            "https://item.jd.com/10012345678901.html",
+        )
+        self.assertEqual(first["sales_amount_range"], "¥50万~¥75万")
+        self.assertEqual(first["sales_amount_upper"], 750_000)
+        self.assertEqual(rows[1]["rank"], 4)
+
+    def test_feishu_xml_adds_sales_column_and_product_assets(self):
+        product = parse_jd_market_rank_html(
+            SAMPLE_HTML,
+            category_name="旅行箱",
+            category_path="功能箱包 > 行李箱",
+        )[0]
+        xml = build_jd_feishu_batch_xml([product], start_rank=1)
+
+        self.assertIn(">销售额</th>", xml)
+        self.assertIn("<b>¥50万~¥75万</b>", xml)
+        self.assertIn(
+            '<img href="https://img10.360buyimg.com/n0/jfs/test.jpg" '
+            'width="160" height="160"',
+            xml,
+        )
+        self.assertIn(
+            '<a href="https://item.jd.com/10012345678901.html">打开商品</a>',
+            xml,
+        )
+
+    def test_category_xml_uses_one_named_table_for_the_category(self):
+        products = parse_jd_market_rank_html(
+            SAMPLE_HTML,
+            category_name="旅行箱",
+            category_path="功能箱包 > 行李箱",
+        )
+
+        xml = build_jd_feishu_category_xml("旅行箱", products)
+
+        self.assertIn("<h2>旅行箱（2条）</h2>", xml)
+        self.assertEqual(xml.count("<table>"), 1)
+        self.assertEqual(xml.count("<tbody>"), 1)
+        self.assertEqual(xml.count("<tr>"), 3)
+
+    def test_group_products_by_category_preserves_category_order(self):
+        grouped = group_products_by_category(
+            [
+                {"category_name": "旅行箱", "rank": 1},
+                {"category_name": "运动包", "rank": 1},
+                {"category_name": "旅行箱", "rank": 2},
+            ]
+        )
+
+        self.assertEqual(
+            [(name, len(items)) for name, items in grouped],
+            [("旅行箱", 2), ("运动包", 1)],
+        )
+
+    def test_intro_records_period_and_category_path(self):
+        xml = build_jd_feishu_intro_xml(
+            title="京东市场榜单",
+            period="近30天",
+            products=[
+                {
+                    "category_name": "运动包",
+                    "category_path": "功能箱包 > 休闲运动包",
+                    "search_keyword": "",
+                }
+            ],
+        )
+
+        self.assertIn("<h1>京东</h1>", xml)
+        self.assertIn("近30天", xml)
+        self.assertIn("运动包（功能箱包 &gt; 休闲运动包）", xml)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_main_image_concurrency.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_main_image_concurrency.py
new file mode 100644
index 0000000..9dcb582
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_main_image_concurrency.py
@@ -0,0 +1,99 @@
+import io
+import tempfile
+import unittest
+from contextlib import redirect_stdout
+from pathlib import Path
+from unittest.mock import patch
+
+from main_image_paths import collect_images_dir, resolve_images_dir
+from main_image_db import persist_main_image_records
+from taobao_wanxiang_ai_creative_report import cleanup_images_after_visual_merge
+
+
+class MainImagePathTests(unittest.TestCase):
+    def test_collectors_use_platform_owned_directories(self):
+        style_dir = Path("style")
+        self.assertEqual(collect_images_dir(style_dir, "tm"), style_dir / "tm_images")
+        self.assertEqual(collect_images_dir(style_dir, "jd"), style_dir / "jd_images")
+
+    def test_insert_prefers_platform_directory_and_supports_legacy_data(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            style_dir = Path(tmp)
+            legacy = style_dir / "images"
+            legacy.mkdir()
+            self.assertEqual(resolve_images_dir(style_dir, "jd"), legacy)
+
+            current = style_dir / "jd_images"
+            current.mkdir()
+            self.assertEqual(resolve_images_dir(style_dir, "jd"), current)
+
+
+class TmallCleanupTests(unittest.TestCase):
+    def test_locked_duplicate_is_retried_without_console_encoding_failure(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            images_dir = Path(tmp)
+            keep = images_dir / "keep.png"
+            locked = images_dir / "locked.png"
+            keep.write_bytes(b"keep")
+            locked.write_bytes(b"duplicate")
+            creatives = [{"local_image_path": str(keep), "local_image_paths": [str(keep), str(locked)]}]
+
+            original_unlink = Path.unlink
+            attempts = 0
+
+            def flaky_unlink(path, *args, **kwargs):
+                nonlocal attempts
+                if path == locked:
+                    attempts += 1
+                    if attempts < 3:
+                        raise PermissionError("temporarily locked")
+                return original_unlink(path, *args, **kwargs)
+
+            raw = io.BytesIO()
+            strict_gbk_stdout = io.TextIOWrapper(raw, encoding="gbk", errors="strict")
+            with patch.object(Path, "unlink", new=flaky_unlink):
+                with redirect_stdout(strict_gbk_stdout):
+                    cleanup_images_after_visual_merge(
+                        images_dir,
+                        creatives,
+                        retry_attempts=3,
+                        retry_delay=0,
+                    )
+                strict_gbk_stdout.flush()
+
+            self.assertEqual(attempts, 3)
+            self.assertFalse(locked.exists())
+            self.assertTrue(keep.exists())
+
+
+class MainImageDatabaseTests(unittest.TestCase):
+    def test_database_connection_is_opened_only_when_records_are_persisted(self):
+        events = []
+        connection = object()
+
+        class ConnectionContext:
+            def __enter__(self):
+                events.append("connect")
+                return connection
+
+            def __exit__(self, exc_type, exc, tb):
+                events.append("close")
+
+        def upsert(conn, records):
+            self.assertIs(conn, connection)
+            events.append(("upsert", len(records)))
+            return len(records)
+
+        self.assertEqual(
+            persist_main_image_records(
+                [{"image_key": "a"}, {"image_key": "b"}],
+                get_conn_fn=ConnectionContext,
+                upsert_fn=upsert,
+            ),
+            2,
+        )
+        self.assertEqual(events, ["connect", ("upsert", 2), "close"])
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_hermes_notification.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_hermes_notification.py
new file mode 100644
index 0000000..0d076ca
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_hermes_notification.py
@@ -0,0 +1,199 @@
+from __future__ import annotations
+
+import json
+from datetime import date, datetime
+from unittest.mock import MagicMock, patch
+
+import market_rank_hermes_notification as notification
+
+
+REPORTS = [
+    {
+        "platform": "tm",
+        "feishu_doc_url": "https://example.com/tm",
+        "product_count": 1178,
+        "report_title": "天猫九品类市场排行",
+    },
+    {
+        "platform": "jd",
+        "feishu_doc_url": "https://example.com/jd",
+        "product_count": 1050,
+        "report_title": "京东市场排行",
+    },
+    {
+        "platform": "dy",
+        "feishu_doc_url": "https://example.com/dy",
+        "product_count": 1285,
+        "report_title": "抖音市场排行",
+    },
+]
+
+
+def test_build_notification_message_contains_three_plain_links():
+    message = notification.build_notification_message(date(2026, 7, 26), REPORTS)
+
+    assert "三平台市场排行" in message
+    assert "天猫：https://example.com/tm" in message
+    assert "京东：https://example.com/jd" in message
+    assert "抖音：https://example.com/dy" in message
+    assert "1178" in message
+
+
+def test_build_notification_message_supports_partial_platform_success():
+    message = notification.build_notification_message(
+        date(2026, 7, 26),
+        [REPORTS[1]],
+    )
+
+    assert "京东：https://example.com/jd" in message
+    assert "天猫：https://" not in message
+    assert "抖音：https://" not in message
+    assert "当前成功平台 1 个" in message
+
+
+def test_build_hermes_payload_requires_direct_message_to_recipient():
+    payload = notification.build_hermes_payload(
+        recipient_open_id="ou_wangyunlong",
+        message="三个链接",
+    )
+
+    encoded = json.dumps(payload, ensure_ascii=False)
+    assert "ou_wangyunlong" in encoded
+    assert "飞书私聊" in encoded
+    assert "三个链接" in encoded
+
+
+def test_extracts_real_feishu_message_id_from_hermes_completion():
+    response = {
+        "id": "api-response-id",
+        "choices": [
+            {
+                "message": {
+                    "content": "发送成功，消息ID om_x100b696f1169f0a0b1f9c311e8dc323"
+                }
+            }
+        ],
+    }
+
+    assert (
+        notification.extract_feishu_message_id(response)
+        == "om_x100b696f1169f0a0b1f9c311e8dc323"
+    )
+    assert notification.hermes_send_succeeded(response) is True
+
+
+def test_notify_skips_when_same_day_recipient_was_already_sent():
+    conn = MagicMock()
+    with (
+        patch.object(notification, "get_conn") as get_conn,
+        patch.object(notification, "ensure_market_rank_notification_table"),
+        patch.object(notification, "get_market_rank_reports_for_date", return_value=REPORTS),
+        patch.object(notification, "market_rank_notification_was_sent", return_value=True),
+        patch.object(notification, "post_hermes") as post_hermes,
+    ):
+        get_conn.return_value.__enter__.return_value = conn
+        result = notification.notify_market_rank_reports(
+            report_date=date(2026, 7, 26),
+            recipient_open_id="ou_wangyunlong",
+        )
+
+    assert result["status"] == "skipped"
+    assert result["reason"] == "already_sent"
+    post_hermes.assert_not_called()
+
+
+def test_notify_sends_only_after_all_three_links_exist_and_persists_success():
+    conn = MagicMock()
+    response = {
+        "id": "hermes-response-id",
+        "choices": [
+            {
+                "message": {
+                    "content": "发送成功，消息ID om_x100b696f1169f0a0b1f9c311e8dc323"
+                }
+            }
+        ],
+    }
+    with (
+        patch.object(notification, "get_conn") as get_conn,
+        patch.object(notification, "ensure_market_rank_notification_table"),
+        patch.object(notification, "market_rank_notification_was_sent", return_value=False),
+        patch.object(
+            notification,
+            "get_market_rank_reports_for_date",
+            return_value=REPORTS,
+        ) as get_reports,
+        patch.object(notification, "post_hermes", return_value=response) as post_hermes,
+        patch.object(notification, "upsert_market_rank_notification") as upsert,
+    ):
+        get_conn.return_value.__enter__.return_value = conn
+        cutoff = datetime(2026, 7, 26, 10, 0, 0)
+        result = notification.notify_market_rank_reports(
+            report_date=date(2026, 7, 26),
+            recipient_open_id="ou_wangyunlong",
+            generated_after=cutoff,
+        )
+
+    assert result["status"] == "sent"
+    get_reports.assert_called_once_with(
+        conn,
+        date(2026, 7, 26),
+        notification.PLATFORM_ORDER,
+        updated_after=cutoff,
+    )
+    post_hermes.assert_called_once()
+    saved = upsert.call_args.args[1]
+    assert saved["status"] == "sent"
+    assert saved["platform_links"] == {
+        "tm": "https://example.com/tm",
+        "jd": "https://example.com/jd",
+        "dy": "https://example.com/dy",
+    }
+
+
+def test_notify_sends_available_links_when_other_platforms_are_missing():
+    conn = MagicMock()
+    response = {
+        "choices": [{"message": {"content": "发送成功，消息ID om_partial123"}}]
+    }
+    with (
+        patch.object(notification, "get_conn") as get_conn,
+        patch.object(notification, "ensure_market_rank_notification_table"),
+        patch.object(notification, "market_rank_notification_was_sent", return_value=False),
+        patch.object(notification, "get_market_rank_reports_for_date", return_value=REPORTS[:2]),
+        patch.object(notification, "post_hermes", return_value=response) as post_hermes,
+        patch.object(notification, "upsert_market_rank_notification") as upsert,
+    ):
+        get_conn.return_value.__enter__.return_value = conn
+        result = notification.notify_market_rank_reports(
+            report_date=date(2026, 7, 26),
+            recipient_open_id="ou_wangyunlong",
+        )
+
+    assert result["status"] == "sent"
+    assert result["missing_platforms"] == ["dy"]
+    post_hermes.assert_called_once()
+    assert upsert.call_args.args[1]["status"] == "sent"
+
+
+def test_notify_skips_only_when_current_run_has_no_links():
+    conn = MagicMock()
+    with (
+        patch.object(notification, "get_conn") as get_conn,
+        patch.object(notification, "ensure_market_rank_notification_table"),
+        patch.object(notification, "get_market_rank_reports_for_date", return_value=[]),
+        patch.object(notification, "market_rank_notification_was_sent") as was_sent,
+        patch.object(notification, "post_hermes") as post_hermes,
+        patch.object(notification, "upsert_market_rank_notification") as upsert,
+    ):
+        get_conn.return_value.__enter__.return_value = conn
+        result = notification.notify_market_rank_reports(
+            report_date=date(2026, 7, 26),
+            recipient_open_id="ou_wangyunlong",
+        )
+
+    assert result["status"] == "skipped"
+    assert result["reason"] == "no_platform_reports"
+    was_sent.assert_not_called()
+    post_hermes.assert_not_called()
+    assert upsert.call_args.args[1]["status"] == "incomplete"
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_report_archive.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_report_archive.py
new file mode 100644
index 0000000..3e36136
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_report_archive.py
@@ -0,0 +1,115 @@
+from __future__ import annotations
+
+from datetime import date
+from unittest.mock import patch
+
+from db import upsert_market_rank_report
+from market_rank_report_archive import (
+    archive_market_rank_report,
+    extract_feishu_doc_id,
+    normalize_market_rank_platform,
+)
+
+
+class FakeCursor:
+    def __init__(self):
+        self.sql = ""
+        self.params = ()
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, exc_type, exc, tb):
+        return False
+
+    def execute(self, sql, params):
+        self.sql = sql
+        self.params = params
+
+    def fetchone(self):
+        return (37,)
+
+
+class FakeConnection:
+    def __init__(self):
+        self.cursor_instance = FakeCursor()
+
+    def cursor(self):
+        return self.cursor_instance
+
+
+def test_extract_feishu_doc_id_from_docx_url():
+    assert (
+        extract_feishu_doc_id(
+            "https://bu0zgpibak.feishu.cn/docx/ZPCYdLtvfoksGmxy5zcc8E7anzg"
+        )
+        == "ZPCYdLtvfoksGmxy5zcc8E7anzg"
+    )
+
+
+def test_platform_names_are_normalized_to_database_codes():
+    assert normalize_market_rank_platform("天猫") == "tm"
+    assert normalize_market_rank_platform("jd") == "jd"
+    assert normalize_market_rank_platform("抖音") == "dy"
+
+
+def test_upsert_market_rank_report_is_idempotent_per_platform_and_day():
+    conn = FakeConnection()
+
+    report_id = upsert_market_rank_report(
+        conn,
+        {
+            "platform": "tm",
+            "report_date": date(2026, 7, 24),
+            "period": "30天",
+            "category_scope": "九品类",
+            "category_names": ["旅行箱", "运动包"],
+            "report_title": "天猫九品类市场排行 30天 2026-07-24",
+            "feishu_doc_id": "doc-token",
+            "feishu_doc_url": "https://example.feishu.cn/docx/doc-token",
+            "product_count": 146,
+            "source_json_path": "data/sycm_market_rank/result.json",
+            "run_metadata": {"min_price": 400, "max_price": 2000},
+        },
+    )
+
+    assert report_id == 37
+    assert "ON CONFLICT (platform, report_date)" in conn.cursor_instance.sql
+    assert "ON CONFLICT (platform, report_date, period, category_scope)" not in conn.cursor_instance.sql
+    assert conn.cursor_instance.params[0] == "tm"
+    assert conn.cursor_instance.params[4] == ["旅行箱", "运动包"]
+    assert conn.cursor_instance.params[7] == "https://example.feishu.cn/docx/doc-token"
+    assert conn.cursor_instance.params[8] == 146
+
+
+def test_archive_market_rank_report_builds_and_saves_normalized_record():
+    with (
+        patch("market_rank_report_archive.get_conn") as get_conn,
+        patch(
+            "market_rank_report_archive.upsert_market_rank_report",
+            return_value=88,
+        ) as upsert,
+    ):
+        conn = object()
+        get_conn.return_value.__enter__.return_value = conn
+
+        report_id = archive_market_rank_report(
+            platform="京东",
+            report_date=date(2026, 7, 24),
+            period="近30天",
+            category_scope="七品类",
+            category_names=["旅行箱", "双肩包"],
+            report_title="京东七品类市场商品榜单 近30天 2026-07-24",
+            feishu_doc_url="https://example.feishu.cn/docx/jd-doc-token",
+            product_count=70,
+            source_json_path="data/jd_market_rank/result.json",
+            run_metadata={"max_pages": 20},
+        )
+
+    assert report_id == 88
+    record = upsert.call_args.args[1]
+    assert upsert.call_args.args[0] is conn
+    assert record["platform"] == "jd"
+    assert record["feishu_doc_id"] == "jd-doc-token"
+    assert record["category_names"] == ["旅行箱", "双肩包"]
+    assert record["run_metadata"] == {"max_pages": 20}
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_workflow.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_workflow.py
new file mode 100644
index 0000000..4b9aa6b
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_market_rank_workflow.py
@@ -0,0 +1,208 @@
+from __future__ import annotations
+
+import json
+import sys
+from pathlib import Path
+from threading import Barrier
+
+import orchestrate_market_rank_collection as workflow
+
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+
+
+def test_default_platform_order_and_commands():
+    assert workflow.DEFAULT_PLATFORMS == ("tm", "jd", "dy")
+
+    tm_command = workflow.build_platform_command("tm", python_executable=sys.executable)
+    jd_command = workflow.build_platform_command("jd", python_executable=sys.executable)
+    dy_command = workflow.build_platform_command("dy", python_executable=sys.executable)
+
+    assert tm_command[:2] == [sys.executable, "collect_sycm_market_rank.py"]
+    assert tm_command[-2:] == ["--risk-mode", "manual"]
+    assert jd_command == [sys.executable, "collect_jd_market_rank.py"]
+    assert dy_command == [sys.executable, "collect_dy_market_rank.py"]
+
+
+def test_child_environment_forces_utf8_unbuffered_output():
+    env = workflow.build_subprocess_env({"EXISTING": "kept"})
+
+    assert env["EXISTING"] == "kept"
+    assert env["PYTHONIOENCODING"] == "utf-8"
+    assert env["PYTHONUNBUFFERED"] == "1"
+
+
+def test_console_text_replaces_characters_not_supported_by_gbk():
+    assert workflow.console_safe_text("prefix \ufffd suffix", "gbk") == "prefix ? suffix"
+
+
+def test_workflow_continues_after_one_platform_fails(tmp_path):
+    calls: list[str] = []
+
+    def fake_executor(*, platform, command, log_path, dry_run):
+        calls.append(platform)
+        return {
+            "platform": platform,
+            "command": command,
+            "log_path": str(log_path),
+            "started_at": "2026-07-26T10:00:00",
+            "finished_at": "2026-07-26T10:01:00",
+            "duration_seconds": 60.0,
+            "exit_code": 1 if platform == "tm" else 0,
+            "status": "failed" if platform == "tm" else "success",
+        }
+
+    summary = workflow.execute_workflow(
+        workflow.DEFAULT_PLATFORMS,
+        run_id="test-run",
+        log_dir=tmp_path,
+        executor=fake_executor,
+    )
+
+    assert set(calls) == {"tm", "jd", "dy"}
+    assert [item["platform"] for item in summary["steps"]] == ["tm", "jd", "dy"]
+    assert summary["exit_code"] == 1
+    assert [item["status"] for item in summary["steps"]] == [
+        "failed", "success", "success"
+    ]
+    saved = json.loads((tmp_path / "test-run_summary.json").read_text(encoding="utf-8"))
+    assert saved["run_id"] == "test-run"
+    assert saved["exit_code"] == 1
+
+
+def test_workflow_continues_when_executor_raises(tmp_path):
+    calls: list[str] = []
+
+    def raising_executor(*, platform, command, log_path, dry_run):
+        calls.append(platform)
+        if platform == "tm":
+            raise RuntimeError("browser startup failed")
+        return {
+            "platform": platform,
+            "command": command,
+            "log_path": str(log_path),
+            "started_at": "2026-07-26T10:00:00",
+            "finished_at": "2026-07-26T10:01:00",
+            "duration_seconds": 60.0,
+            "exit_code": 0,
+            "status": "success",
+        }
+
+    summary = workflow.execute_workflow(
+        workflow.DEFAULT_PLATFORMS,
+        run_id="raised-run",
+        log_dir=tmp_path,
+        executor=raising_executor,
+    )
+
+    assert set(calls) == {"tm", "jd", "dy"}
+    assert summary["exit_code"] == 1
+    assert summary["steps"][0]["status"] == "failed"
+    assert "browser startup failed" in summary["steps"][0]["error"]
+
+
+def test_three_platforms_start_in_parallel(tmp_path):
+    all_started = Barrier(3)
+
+    def synchronized_executor(*, platform, command, log_path, dry_run):
+        all_started.wait(timeout=1)
+        return {
+            "platform": platform,
+            "command": command,
+            "log_path": str(log_path),
+            "started_at": "2026-07-26T10:00:00",
+            "finished_at": "2026-07-26T10:01:00",
+            "duration_seconds": 60.0,
+            "exit_code": 0,
+            "status": "success",
+        }
+
+    summary = workflow.execute_workflow(
+        workflow.DEFAULT_PLATFORMS,
+        run_id="parallel-run",
+        log_dir=tmp_path,
+        executor=synchronized_executor,
+    )
+
+    assert summary["exit_code"] == 0
+    assert [item["platform"] for item in summary["steps"]] == ["tm", "jd", "dy"]
+
+
+def test_workflow_notifies_after_all_parallel_steps_finish(tmp_path):
+    all_started = Barrier(3)
+    notifications: list[dict] = []
+
+    def synchronized_executor(*, platform, command, log_path, dry_run):
+        all_started.wait(timeout=1)
+        return {
+            "platform": platform,
+            "label": workflow.PLATFORM_LABELS[platform],
+            "command": command,
+            "log_path": str(log_path),
+            "started_at": "2026-07-26T10:00:00",
+            "finished_at": "2026-07-26T10:01:00",
+            "duration_seconds": 60.0,
+            "exit_code": 0,
+            "status": "success",
+        }
+
+    def fake_notifier(**kwargs):
+        notifications.append(kwargs)
+        return {"status": "sent", "recipient_open_id": kwargs["recipient_open_id"]}
+
+    summary = workflow.execute_workflow(
+        workflow.DEFAULT_PLATFORMS,
+        run_id="notify-run",
+        log_dir=tmp_path,
+        executor=synchronized_executor,
+        notify_open_id="ou_wangyunlong",
+        notifier=fake_notifier,
+    )
+
+    assert len(notifications) == 1
+    assert notifications[0]["recipient_open_id"] == "ou_wangyunlong"
+    assert notifications[0]["generated_after"] is not None
+    assert summary["notification"]["status"] == "sent"
+    saved = json.loads((tmp_path / "notify-run_summary.json").read_text(encoding="utf-8"))
+    assert saved["notification"]["status"] == "sent"
+
+
+def test_partial_current_run_links_can_still_send_successfully(tmp_path):
+    def successful_executor(*, platform, command, log_path, dry_run):
+        return {
+            "platform": platform,
+            "command": command,
+            "log_path": str(log_path),
+            "started_at": "2026-07-26T10:00:00",
+            "finished_at": "2026-07-26T10:01:00",
+            "duration_seconds": 60.0,
+            "exit_code": 0,
+            "status": "success",
+        }
+
+    summary = workflow.execute_workflow(
+        workflow.DEFAULT_PLATFORMS,
+        run_id="missing-link-run",
+        log_dir=tmp_path,
+        executor=successful_executor,
+        notify_open_id="ou_wangyunlong",
+        notifier=lambda **kwargs: {
+            "status": "sent",
+            "platform_links": {"jd": "https://example.com/jd"},
+            "missing_platforms": ["tm", "dy"],
+        },
+    )
+
+    assert summary["exit_code"] == 0
+    assert summary["status"] == "success"
+
+
+def test_platform_parser_deduplicates_while_preserving_order():
+    assert workflow.parse_platforms("dy,tm,dy,jd") == ("dy", "tm", "jd")
+
+
+def test_weekly_launcher_uses_market_rank_orchestrator():
+    launcher = (PROJECT_ROOT / "run_weekly_market_rank.bat").read_text(encoding="utf-8")
+
+    assert ".venv\\Scripts\\python.exe" in launcher
+    assert "orchestrate_market_rank_collection.py" in launcher
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_persona_launcher.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_persona_launcher.py
new file mode 100644
index 0000000..f7783e4
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_persona_launcher.py
@@ -0,0 +1,20 @@
+import sys
+import unittest
+from pathlib import Path
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(PROJECT_ROOT))
+
+import run_daily_persona
+
+
+class PersonaLauncherTests(unittest.TestCase):
+    def test_any_platform_failure_makes_launcher_fail(self):
+        self.assertEqual(run_daily_persona.combined_exit_code({"tm": 0, "dy": 2, "jd": 0}), 2)
+
+    def test_all_platforms_success_makes_launcher_success(self):
+        self.assertEqual(run_daily_persona.combined_exit_code({"tm": 0, "dy": 0, "jd": 0}), 0)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_style_analysis_orchestration.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_style_analysis_orchestration.py
new file mode 100644
index 0000000..25a51a2
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_style_analysis_orchestration.py
@@ -0,0 +1,358 @@
+import os
+import json
+import tempfile
+import unittest
+import sys
+from datetime import date
+from pathlib import Path
+from types import SimpleNamespace
+from unittest.mock import patch
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+# Other vendor tests temporarily prepend their own directories. Keep this test
+# bound to the project's ``config`` package during full-suite discovery.
+sys.path.insert(0, str(PROJECT_ROOT))
+
+import analyze_style_with_hermes as analysis
+import orchestrate_daily_collection as orchestrator
+from config.style_config_loader import StyleConfigLoader
+
+
+class StyleAnalysisConfigTests(unittest.TestCase):
+    def test_platform_analysis_url_is_aggregated_by_style(self):
+        loader = StyleConfigLoader()
+        payload = loader._aggregate([
+            {
+                "platform": "天猫",
+                "style": "盖亚斜挎",
+                "style_analysis_bitable_url": (
+                    "[分析表](https://example.feishu.cn/base/base_token"
+                    "?table=table_id&view=view_id)"
+                ),
+            }
+        ])
+
+        self.assertEqual(
+            payload["styles"]["盖亚斜挎"]["style_analysis_bitable"],
+            {"base_token": "base_token", "table_id": "table_id", "view_id": "view_id"},
+        )
+
+
+class StyleAnalysisWindowTests(unittest.TestCase):
+    def test_three_day_label_uses_compact_end_date(self):
+        self.assertEqual(
+            analysis.analysis_period_label(date(2026, 7, 13), date(2026, 7, 15)),
+            "2026-07-13~07-15",
+        )
+
+    def test_due_styles_wait_three_full_days(self):
+        styles = ["A", "B", "C"]
+        latest = {"A": date(2026, 7, 15), "B": date(2026, 7, 12)}
+
+        due = analysis.select_due_styles(styles, latest, date(2026, 7, 15), 3)
+
+        self.assertEqual(due, ["B", "C"])
+
+    def test_prompt_keeps_all_dimensions_inside_current_three_day_window(self):
+        payload = {
+            "total": {"sales": 30, "visitors": 1000, "cart": 80},
+            "platforms": {"tm": {"sales": 20}, "jd": {"sales": 10}},
+            "daily": [{"date": "2026-07-14", "plat": "tm", "sales": 10}],
+            "persona": [{"date": "2026-07-14", "plat": "tm", "gender": []}],
+            "product_reviews": {"tm|negative": {"n": 2, "samples": ["肩带偏硬"]}},
+            "main_image": [{"date": "2026-07-14", "key": "hero", "ctr": "2.1"}],
+            "marketing": {
+                "notes_by_platform_all": [{"platform": "xiaohongshu", "views": 1000}],
+                "note_comments_sample": [{"content": "求尺寸"}],
+            },
+        }
+
+        prompt = analysis.build_user_prompt(
+            "盖亚斜挎", date(2026, 7, 14), date(2026, 7, 16), payload,
+        )
+
+        self.assertIn("2026-07-14 ~ 2026-07-16", prompt)
+        self.assertNotIn("对比基期", prompt)
+        self.assertNotIn("紧邻上期", prompt)
+        self.assertNotIn("最近 6 周", prompt)
+        self.assertNotIn("近 60d", prompt)
+        self.assertIn("逐日经营数据", prompt)
+        self.assertIn("人物画像", prompt)
+        self.assertIn("商品评价", prompt)
+        self.assertIn("主图表现", prompt)
+        self.assertIn("营销笔记、曝光互动与评论", prompt)
+        self.assertIn("肩带偏硬", prompt)
+        self.assertIn("求尺寸", prompt)
+        self.assertIn('"sales": 30', prompt)
+
+    def test_system_prompt_requires_each_business_dimension(self):
+        for dimension in ("访客", "加购", "人物画像", "营销笔记", "笔记评论", "商品评价", "主图"):
+            self.assertIn(dimension, analysis.SYSTEM)
+
+    def test_system_prompt_requires_fixed_seven_section_report(self):
+        headings = [
+            "## 一、访客分析",
+            "## 二、加购率分析",
+            "## 三、点击率分析",
+            "## 四、转化率分析",
+            "## 五、退货率分析",
+            "## 六、销量分析",
+            "## 七、总评",
+        ]
+
+        for heading in headings:
+            self.assertIn(heading, analysis.SYSTEM)
+
+    def test_plain_seven_section_titles_are_normalized_to_markdown(self):
+        report = """盖亚斜挎 单款式多维度分析报告 (2026-07-11~2026-07-13)
+
+一、访客分析
+内容
+二、加购率分析
+内容
+三、点击率分析
+内容
+四、转化率分析
+内容
+五、退货率分析
+内容
+六、销量分析
+内容
+七、总评
+内容
+"""
+
+        normalized = analysis.normalize_report_markdown(report)
+
+        self.assertTrue(normalized.startswith("# 盖亚斜挎"))
+        self.assertTrue(analysis.has_exact_report_headings(normalized))
+
+    def test_report_gets_clear_subheadings_and_key_emphasis(self):
+        blocks = ["盖亚斜挎 单款式多维度分析报告 (2026-07-11~2026-07-13)"]
+        for heading in analysis.REPORT_HEADINGS[:-1]:
+            blocks.extend([
+                heading.removeprefix("## "),
+                "数据快照：三日访客2692，转化率0.71%。其余数据。",
+                "问题判断：天猫访客高但转化偏低。需要重点处理。",
+                "可执行建议：P0，24小时内完成核查，目标转化率1.00%。",
+            ])
+        blocks.extend([
+            "七、总评",
+            "核心结论：当前核心问题是高访客、低转化。需要优先修复。",
+            "链路表现为：曝光到访客再到成交。",
+            "P0，48小时内完成详情页整改，目标转化率1.00%。",
+        ])
+
+        normalized = analysis.normalize_report_markdown("\n\n".join(blocks))
+
+        self.assertTrue(analysis.has_clear_report_hierarchy(normalized))
+        self.assertTrue(analysis.has_key_emphasis(normalized))
+        self.assertIn("### 数据快照", normalized)
+        self.assertIn("### 核心判断", normalized)
+        self.assertIn("### 优先级行动", normalized)
+        self.assertIn("**P0**", normalized)
+        self.assertIn("**24小时内**", normalized)
+
+    def test_extra_helpful_h3_does_not_reject_complete_required_hierarchy(self):
+        required = "\n".join(analysis.REPORT_SUBHEADINGS)
+        report = required.replace(
+            "### 核心判断\n",
+            "### 平台明细\n补充内容\n### 核心判断\n",
+            1,
+        )
+
+        self.assertTrue(analysis.has_clear_report_hierarchy(report))
+
+    def test_missing_calendar_day_is_reported(self):
+        rows = [
+            {"date": "2026-07-14", "plat": "tm"},
+            {"date": "2026-07-16", "plat": "jd"},
+        ]
+
+        missing = analysis.missing_window_dates(
+            rows, date(2026, 7, 14), date(2026, 7, 16),
+        )
+
+        self.assertEqual(missing, [date(2026, 7, 15)])
+
+
+class LarkEnvironmentTests(unittest.TestCase):
+    def test_legacy_lark_config_is_used_when_default_is_missing(self):
+        with tempfile.TemporaryDirectory() as temp_dir:
+            home = Path(temp_dir)
+            legacy = home / ".lark-cli" / "hermes"
+            legacy.mkdir(parents=True)
+            (legacy / "config.json").write_text("{}", encoding="utf-8")
+            with patch.dict(os.environ, {}, clear=True), patch.object(Path, "home", return_value=home):
+                env = analysis.lark_cli_env()
+
+        self.assertEqual(env["LARKSUITE_CLI_CONFIG_DIR"], str(legacy))
+
+    def test_project_legacy_profile_wins_over_new_empty_default_config(self):
+        with tempfile.TemporaryDirectory() as temp_dir:
+            home = Path(temp_dir)
+            config_root = home / ".lark-cli"
+            legacy = config_root / "hermes"
+            legacy.mkdir(parents=True)
+            (config_root / "config.json").write_text("{}", encoding="utf-8")
+            (legacy / "config.json").write_text("{}", encoding="utf-8")
+            with patch.dict(os.environ, {}, clear=True), patch.object(Path, "home", return_value=home):
+                env = analysis.lark_cli_env()
+
+        self.assertEqual(env["LARKSUITE_CLI_CONFIG_DIR"], str(legacy))
+
+
+class FeishuDocumentWriteTests(unittest.TestCase):
+    def test_preflight_refreshes_and_requires_both_identities(self):
+        response = {
+            "verified": True,
+            "identities": {
+                "bot": {"available": True, "verified": True},
+                "user": {"available": True, "verified": True},
+            },
+        }
+
+        with patch.object(analysis, "run_lark_cli", return_value=response) as run:
+            analysis.ensure_lark_cli_ready()
+
+        run.assert_called_once_with(["auth", "status", "--json", "--verify"])
+
+    def test_preflight_rejects_missing_user_identity(self):
+        response = {
+            "verified": True,
+            "identities": {
+                "bot": {"available": True, "verified": True},
+                "user": {"available": False, "verified": False},
+            },
+        }
+
+        with patch.object(analysis, "run_lark_cli", return_value=response):
+            with self.assertRaisesRegex(RuntimeError, "user"):
+                analysis.ensure_lark_cli_ready()
+
+    def test_bot_created_document_is_granted_to_current_user(self):
+        responses = [
+            {"data": {"document": {
+                "url": "https://example.feishu.cn/docx/doc_token",
+                "document_id": "doc_token",
+            }}},
+            {"identities": {"user": {"openId": "ou_current_user"}}},
+            {"ok": True, "data": {"perm": "full_access"}},
+        ]
+
+        with patch.object(analysis, "run_lark_cli", side_effect=responses) as run:
+            url, doc_id = analysis.write_feishu_doc("测试分析", "# 内容")
+
+        self.assertEqual(url, "https://example.feishu.cn/docx/doc_token")
+        self.assertEqual(doc_id, "doc_token")
+        create_args = run.call_args_list[0].args[0]
+        self.assertEqual(create_args[create_args.index("--as") + 1], "bot")
+        grant_args = run.call_args_list[2].args[0]
+        self.assertIn("+member-add", grant_args)
+        self.assertIn("ou_current_user", grant_args)
+        self.assertIn("full_access", grant_args)
+        self.assertIn("--yes", grant_args)
+
+    def test_missing_current_user_identity_stops_before_base_write(self):
+        responses = [
+            {"data": {"document": {
+                "url": "https://example.feishu.cn/docx/doc_token",
+                "document_id": "doc_token",
+            }}},
+            {"identities": {"user": {"ready": False}}},
+        ]
+
+        with patch.object(analysis, "run_lark_cli", side_effect=responses):
+            with self.assertRaisesRegex(RuntimeError, "open_id"):
+                analysis.write_feishu_doc("测试分析", "# 内容")
+
+
+class TargetBaseWriteTests(unittest.TestCase):
+    def test_existing_period_record_is_updated(self):
+        responses = [
+            {"data": {"fields": [
+                {"name": "时间", "type": "select"},
+                {"name": "笔记分析汇总", "type": "text"},
+            ]}},
+            {"data": {"field": {
+                "id": "fld_time", "name": "时间", "type": "select",
+                "multiple": False, "options": [{"name": "2026-07-13~07-15"}],
+            }}},
+            {"data": {
+                "data": [[['2026-07-13~07-15'], "old link"]],
+                "fields": ["时间", "笔记分析汇总"],
+                "record_id_list": ["rec_existing"],
+            }},
+            {"ok": True, "data": {"record_id": "rec_existing"}},
+        ]
+        with patch.object(analysis, "run_lark_cli", side_effect=responses) as run:
+            record_id = analysis.write_analysis_link_to_base(
+                {"base_token": "base", "table_id": "table"},
+                "2026-07-13~07-15",
+                "https://example.feishu.cn/docx/doc",
+                "盖亚斜挎",
+            )
+
+        self.assertEqual(record_id, "rec_existing")
+        upsert_args = run.call_args_list[3].args[0]
+        self.assertIn("--record-id", upsert_args)
+        self.assertIn("rec_existing", upsert_args)
+        payload = json.loads(upsert_args[upsert_args.index("--json") + 1])
+        self.assertIn("笔记分析汇总", payload)
+
+    def test_missing_period_is_appended_without_dropping_existing_options(self):
+        responses = [
+            {"data": {"field": {
+                "id": "fld_time", "name": "时间", "type": "select",
+                "multiple": False, "options": [{"name": "7月第二周", "hue": "Blue"}],
+            }}},
+            {"ok": True},
+        ]
+        with patch.object(analysis, "run_lark_cli", side_effect=responses) as run:
+            analysis.ensure_select_option(
+                {"base_token": "base", "table_id": "table"},
+                "时间", "2026-07-13~07-15", "盖亚斜挎",
+            )
+
+        update_args = run.call_args_list[1].args[0]
+        definition = json.loads(update_args[update_args.index("--json") + 1])
+        self.assertEqual(
+            [option["name"] for option in definition["options"]],
+            ["7月第二周", "2026-07-13~07-15"],
+        )
+        self.assertIn("--yes", update_args)
+
+
+class OrchestratorTests(unittest.TestCase):
+    def test_style_analysis_stage_runs_three_day_batch(self):
+        args = SimpleNamespace(dry_run=False)
+        with patch.object(orchestrator, "run_step_with_retry") as run:
+            orchestrator.run_style_analysis(args, "2026-07-15", failed=None)
+
+        command = run.call_args.args[1]
+        self.assertIn("analyze_style_with_hermes.py", command)
+        self.assertIn("--all-styles", command)
+        self.assertEqual(command[command.index("--days") + 1], "3")
+        self.assertEqual(command[command.index("--min-interval-days") + 1], "3")
+        self.assertFalse(run.call_args.kwargs["critical"])
+
+
+class LauncherScheduleTests(unittest.TestCase):
+    def test_daily_launcher_does_not_run_style_analysis(self):
+        launcher = (PROJECT_ROOT / "run_daily_collect.bat").read_text(encoding="utf-8")
+
+        command_lines = [line for line in launcher.splitlines() if "orchestrate_daily_collection.py" in line]
+        self.assertEqual(len(command_lines), 1)
+        self.assertNotIn("style_analysis", command_lines[0])
+
+    def test_three_day_launcher_runs_only_style_analysis(self):
+        launcher = (PROJECT_ROOT / "run_style_analysis_3d.bat").read_text(encoding="utf-8")
+
+        self.assertIn("analyze_style_with_hermes.py", launcher)
+        self.assertIn("--all-styles", launcher)
+        self.assertIn("--days 3", launcher)
+        self.assertIn("--skip-existing", launcher)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_style_config_loader_erp_source.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_style_config_loader_erp_source.py
new file mode 100644
index 0000000..53d915c
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_style_config_loader_erp_source.py
@@ -0,0 +1,45 @@
+from __future__ import annotations
+
+import io
+import sys
+from contextlib import redirect_stdout
+from pathlib import Path
+
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(PROJECT_ROOT))
+
+from config.style_config_loader import StyleConfigLoader
+
+
+def test_erp_codes_use_tm_record_even_when_jd_record_comes_first() -> None:
+    loader = StyleConfigLoader()
+    output = io.StringIO()
+
+    with redirect_stdout(output):
+        payload = loader._aggregate(
+            [
+                {"platform": "京东旗舰店", "style": "极星托特", "erp_codes": "10513001"},
+                {"platform": "天猫", "style": "极星托特", "erp_codes": "10513"},
+            ]
+        )
+
+    assert payload["styles"]["极星托特"]["erp_codes"] == ["10513"]
+    assert "ERP 编码冲突" in output.getvalue()
+    assert "10513001" in output.getvalue()
+
+
+def test_non_tm_erp_code_is_not_used_when_tm_code_is_missing() -> None:
+    loader = StyleConfigLoader()
+    output = io.StringIO()
+
+    with redirect_stdout(output):
+        payload = loader._aggregate(
+            [
+                {"platform": "京东旗舰店", "style": "极星托特", "erp_codes": "10513001"},
+                {"platform": "天猫", "style": "极星托特", "erp_codes": ""},
+            ]
+        )
+
+    assert payload["styles"]["极星托特"]["erp_codes"] == []
+    assert "天猫分组未填写" in output.getvalue()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_sycm_market_rank.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_sycm_market_rank.py
new file mode 100644
index 0000000..e26324a
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_sycm_market_rank.py
@@ -0,0 +1,537 @@
+import inspect
+import unittest
+from unittest.mock import MagicMock, patch
+
+from collect_sycm_market_rank import (
+    TARGET_MARKET_CATEGORIES,
+    _original_taobao_image_url,
+    _feishu_append_xml_path,
+    _product_item_id,
+    _risk_control_reason,
+    build_feishu_doc_xml,
+    build_feishu_category_xml,
+    close_market_rank_browser,
+    collect_all_rank_pages,
+    detect_risk_control_text,
+    inspect_feishu_batch_tables,
+    inspect_feishu_category_tables,
+    group_products_by_category,
+    create_feishu_doc,
+    parse_market_rank_html,
+    parse_range_upper,
+    prepare_market_rank_filters,
+    prepare_market_rank_period,
+)
+
+
+SAMPLE_HTML = """
+<table>
+  <thead>
+    <tr>
+      <th>排名</th><th>商品</th><th>商品关键词</th>
+      <th>店铺</th><th>支付买家数</th><th>访客数</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td>1</td>
+      <td>
+        <a href="//detail.tmall.com/item.htm?id=123">
+          <img data-src="//img.alicdn.com/bao/uploaded/demo.jpg"/>
+          【新品】CECE大容量旅行箱
+        </a>
+      </td>
+      <td>旅行</td><td>cece旗舰店</td><td>2500 ~ 5000</td><td>10万 ~ 25万</td>
+    </tr>
+  </tbody>
+</table>
+"""
+
+
+class SycmMarketRankTests(unittest.TestCase):
+    def test_parallel_documents_use_isolated_append_xml_files(self):
+        first = _feishu_append_xml_path("https://example.feishu.cn/docx/first", 1)
+        second = _feishu_append_xml_path("https://example.feishu.cn/docx/second", 1)
+
+        self.assertNotEqual(first, second)
+        self.assertEqual(first.parent, second.parent)
+
+    def test_title_with_question_mark_is_rejected_before_creating_document(self):
+        with self.assertRaisesRegex(ValueError, "编码损坏"):
+            create_feishu_doc("??九品类市场排行", "<p>test</p>")
+
+    def test_category_groups_are_one_table_each(self):
+        products = [
+            {"category_name": "旅行箱", "rank": 1, "name": "A"},
+            {"category_name": "双肩包", "rank": 1, "name": "B"},
+            {"category_name": "旅行箱", "rank": 2, "name": "C"},
+        ]
+        groups = group_products_by_category(products)
+
+        self.assertEqual([name for name, _ in groups], ["旅行箱", "双肩包"])
+        self.assertEqual([len(items) for _, items in groups], [2, 1])
+        category_xml = build_feishu_category_xml(*groups[0])
+        self.assertEqual(category_xml.count("<table>"), 1)
+        self.assertEqual(category_xml.count("<th "), 7)
+        self.assertNotIn(">品类</th>", category_xml)
+
+    def test_category_table_inspection_keeps_heading_and_row_count(self):
+        content = """
+        <h2>旅行箱（2条）</h2><table id="t1"><tbody>
+          <tr><td><img src="1"/></td></tr><tr><td><img src="2"/></td></tr>
+        </tbody></table>
+        <h2>双肩包（1条）</h2><table id="t2"><tbody><tr><td>缺图</td></tr></tbody></table>
+        """
+
+        self.assertEqual(
+            inspect_feishu_category_tables(content),
+            [
+                {"block_id": "t1", "category": "旅行箱", "declared_rows": 2, "rows": 2, "images": 2},
+                {"block_id": "t2", "category": "双肩包", "declared_rows": 1, "rows": 1, "images": 0},
+            ],
+        )
+
+    def test_detect_risk_control_text_matches_actual_sycm_popup(self):
+        popup_text = """
+        请稍等片刻再点击刷新哦
+        操作太频繁了，请稍后再试
+        反馈码：a777b03e7183a9564215ff465acad489
+        """
+
+        self.assertEqual(
+            detect_risk_control_text(popup_text),
+            "操作太频繁了，请稍后再试",
+        )
+
+    def test_detect_risk_control_text_ignores_normal_rank_page(self):
+        self.assertEqual(
+            detect_risk_control_text("市场排行 支付买家数 访客数 下一页"),
+            "",
+        )
+
+    def test_detect_risk_control_text_matches_baxia_mask(self):
+        self.assertEqual(
+            detect_risk_control_text('<div class="baxia-dialog-mask"></div>'),
+            "生意参谋 Baxia 风控遮罩",
+        )
+
+    def test_hidden_baxia_dialog_is_not_treated_as_active_captcha(self):
+        page = MagicMock()
+        dialog = MagicMock()
+        dialog.count.return_value = 1
+        dialog.nth.return_value.is_visible.return_value = False
+        body = MagicMock()
+        body.inner_text.return_value = "市场排行 支付买家数"
+        page.locator.side_effect = lambda selector: (
+            dialog if selector == ".baxia-dialog" else body
+        )
+        page.content.return_value = (
+            '<div style="display:none" class="baxia-dialog">X</div>'
+        )
+
+        self.assertEqual(_risk_control_reason(page), "")
+
+    def test_owned_tmall_browser_has_profile_scoped_fallback_cleanup(self):
+        session = MagicMock()
+        with (
+            patch("collect_sycm_market_rank.sys.platform", "win32"),
+            patch("collect_sycm_market_rank.time.sleep"),
+            patch("collect_sycm_market_rank.subprocess.run") as run,
+        ):
+            close_market_rank_browser(
+                session,
+                user_data_dir="portable-tm-profile",
+                cdp_url=None,
+            )
+
+        session.close.assert_called_once_with()
+        run.assert_called_once()
+        self.assertEqual(
+            run.call_args.kwargs["env"]["TM_PROFILE_TO_CLOSE"],
+            "portable-tm-profile",
+        )
+
+    def test_cdp_browser_is_not_force_closed(self):
+        session = MagicMock()
+        with patch("collect_sycm_market_rank.subprocess.run") as run:
+            close_market_rank_browser(
+                session,
+                user_data_dir="portable-tm-profile",
+                cdp_url="http://127.0.0.1:9222",
+            )
+
+        session.close.assert_called_once_with()
+        run.assert_not_called()
+
+    def test_prepare_market_period_waits_in_same_session_before_clicking_period(self):
+        context = MagicMock()
+        original_page = object()
+
+        with (
+            patch(
+                "collect_sycm_market_rank._risk_control_reason",
+                side_effect=["生意参谋 Baxia 风控遮罩", ""],
+            ),
+            patch(
+                "collect_sycm_market_rank.wait_for_manual_risk_clearance"
+            ) as wait_for_clearance,
+            patch("collect_sycm_market_rank.select_market_period") as select_period,
+        ):
+            result = prepare_market_rank_period(
+                context,
+                original_page,
+                period="30天",
+                risk_mode="manual",
+                cooldown_seconds=0,
+                max_retries=2,
+            )
+
+        self.assertIs(result, original_page)
+        context.clear_cookies.assert_not_called()
+        wait_for_clearance.assert_called_once_with(
+            original_page,
+            "生意参谋 Baxia 风控遮罩",
+        )
+        select_period.assert_called_once_with(original_page, "30天")
+
+    def test_prepare_market_filters_retries_same_step_after_manual_captcha(self):
+        page = object()
+
+        with (
+            patch(
+                "collect_sycm_market_rank.select_market_period",
+            ) as select_period,
+            patch(
+                "collect_sycm_market_rank.select_market_category_path",
+                side_effect=[RuntimeError("mask intercepts pointer events"), None],
+            ) as select_category,
+            patch(
+                "collect_sycm_market_rank._risk_control_reason",
+                side_effect=["", "生意参谋 Baxia 风控遮罩", ""],
+            ),
+            patch(
+                "collect_sycm_market_rank.wait_for_manual_risk_clearance"
+            ) as wait_for_clearance,
+            patch("collect_sycm_market_rank.apply_custom_price") as apply_price,
+        ):
+            result = prepare_market_rank_filters(
+                page,
+                period="30天",
+                category_path=("箱包皮具/热销女包/男包", "旅行箱"),
+                min_price=400,
+                max_price=2000,
+                risk_mode="manual",
+                cooldown_seconds=0,
+                max_retries=2,
+            )
+
+        self.assertIs(result, page)
+        self.assertEqual(select_period.call_count, 2)
+        self.assertEqual(select_category.call_count, 2)
+        apply_price.assert_called_once_with(page, 400, 2000)
+        wait_for_clearance.assert_called_once_with(
+            page,
+            "生意参谋 Baxia 风控遮罩",
+        )
+
+    def test_manual_captcha_reapplies_filters_before_parsing_current_page(self):
+        page = MagicMock()
+        page.content.side_effect = [
+            '<div class="baxia-dialog">captcha</div>',
+            SAMPLE_HTML,
+        ]
+        recover = MagicMock(return_value=page)
+
+        with (
+            patch(
+                "collect_sycm_market_rank._risk_control_reason",
+                side_effect=["生意参谋 Baxia 风控遮罩", ""],
+            ),
+            patch(
+                "collect_sycm_market_rank.wait_for_manual_risk_clearance"
+            ) as wait_for_clearance,
+            patch("collect_sycm_market_rank._save_checkpoint"),
+        ):
+            rows = collect_all_rank_pages(
+                page,
+                max_pages=1,
+                category_name="旅行箱",
+                category_path="箱包 > 旅行箱",
+                recover_from_risk=recover,
+                manual_risk_wait=True,
+            )
+
+        self.assertEqual(len(rows), 1)
+        wait_for_clearance.assert_called_once_with(
+            page,
+            "生意参谋 Baxia 风控遮罩",
+        )
+        recover.assert_called_once_with(
+            page,
+            1,
+            1,
+            "生意参谋 Baxia 风控遮罩",
+        )
+
+    def test_target_market_categories_contains_the_nine_requested_paths(self):
+        self.assertEqual(len(TARGET_MARKET_CATEGORIES), 9)
+        self.assertEqual(
+            [
+                (
+                    item["name"],
+                    item["min_price"],
+                    item["max_price"],
+                    tuple(tuple(path) for path in item["paths"]),
+                )
+                for item in TARGET_MARKET_CATEGORIES
+            ],
+            [
+                (
+                    "旅行箱",
+                    400,
+                    2000,
+                    (("箱包皮具/热销女包/男包", "旅行箱"),),
+                ),
+                (
+                    "运动包",
+                    200,
+                    1000,
+                    (("箱包皮具/热销女包/男包", "旅行袋"),),
+                ),
+                (
+                    "托特包",
+                    200,
+                    3000,
+                    (("箱包皮具/热销女包/男包", "女士包袋新", "托特包"),),
+                ),
+                (
+                    "双肩包",
+                    200,
+                    2000,
+                    (("箱包皮具/热销女包/男包", "双肩背包"),),
+                ),
+                (
+                    "斜挎包",
+                    200,
+                    2000,
+                    (
+                        ("箱包皮具/热销女包/男包", "男士包袋"),
+                        (
+                            "箱包皮具/热销女包/男包",
+                            "女士包袋新",
+                            "通用款女包",
+                        ),
+                    ),
+                ),
+                (
+                    "胸包",
+                    200,
+                    1000,
+                    (("箱包皮具/热销女包/男包", "胸包"),),
+                ),
+                (
+                    "手机包",
+                    100,
+                    1000,
+                    (("箱包皮具/热销女包/男包", "功能小包", "手机包"),),
+                ),
+                (
+                    "电脑包",
+                    200,
+                    1000,
+                    (("3C数码配件", "笔记本电脑配件", "笔记本电脑包"),),
+                ),
+                (
+                    "相机包",
+                    200,
+                    2000,
+                    (("3C数码配件", "数码相机配件", "数码相机包"),),
+                ),
+            ],
+        )
+
+    def test_original_taobao_image_url_removes_rank_thumbnail_suffix(self):
+        thumbnail = (
+            "https://img.alicdn.com/bao/uploaded/demo-item_pic.jpg_36x36.jpg"
+        )
+
+        self.assertEqual(
+            _original_taobao_image_url(thumbnail),
+            "https://img.alicdn.com/bao/uploaded/demo-item_pic.jpg",
+        )
+
+    def test_product_identity_ignores_changing_mi_id(self):
+        first = _product_item_id(
+            "https://detail.tmall.com/item.htm?id=123&mi_id=first"
+        )
+        second = _product_item_id(
+            "https://detail.tmall.com/item.htm?mi_id=second&id=123"
+        )
+
+        self.assertEqual(first, "123")
+        self.assertEqual(first, second)
+
+    def test_parse_range_upper_supports_plain_and_wan_units(self):
+        self.assertEqual(parse_range_upper("2500 ~ 5000"), 5000)
+        self.assertEqual(parse_range_upper("7.5万～10万"), 100000)
+        self.assertEqual(parse_range_upper("500"), 500)
+
+    def test_parse_market_rank_html_extracts_name_image_link_and_buyer_max(self):
+        rows = parse_market_rank_html(SAMPLE_HTML)
+
+        self.assertEqual(len(rows), 1)
+        self.assertEqual(rows[0]["rank"], 1)
+        self.assertEqual(rows[0]["name"], "【新品】CECE大容量旅行箱")
+        self.assertEqual(rows[0]["buyer_range"], "2500 ~ 5000")
+        self.assertEqual(rows[0]["buyer_max"], 5000)
+        self.assertEqual(
+            rows[0]["image_url"],
+            "https://img.alicdn.com/bao/uploaded/demo.jpg",
+        )
+        self.assertEqual(
+            rows[0]["product_url"],
+            "https://detail.tmall.com/item.htm?id=123",
+        )
+
+    def test_parse_market_rank_html_skips_image_only_link_for_product_name(self):
+        source = """
+        <table>
+          <thead><tr>
+            <th>排名</th><th>商品</th><th>支付买家数</th>
+          </tr></thead>
+          <tbody><tr>
+            <td>9 降1名</td>
+            <td>
+              <a href="//detail.tmall.com/item.htm?id=639">
+                <img src="//img.alicdn.com/demo.jpg"/>
+              </a>
+              <p class="goodsName">
+                <a title="MOFT电脑支架收纳包"
+                   href="//detail.tmall.com/item.htm?id=639">MOFT电脑支架收纳包</a>
+              </p>
+            </td>
+            <td>50 ~ 100</td>
+          </tr></tbody>
+        </table>
+        """
+
+        rows = parse_market_rank_html(source)
+
+        self.assertEqual(len(rows), 1)
+        self.assertEqual(rows[0]["name"], "MOFT电脑支架收纳包")
+        self.assertEqual(rows[0]["rank"], 9)
+        self.assertEqual(rows[0]["buyer_max"], 100)
+
+    def test_rank_page_collector_has_no_detail_page_or_price_interface(self):
+        parameters = inspect.signature(collect_all_rank_pages).parameters
+        source = inspect.getsource(collect_all_rank_pages)
+
+        self.assertNotIn("context", parameters)
+        self.assertNotIn("collect_prices", parameters)
+        self.assertNotIn("price_cache", parameters)
+        self.assertNotIn("_collect_price_from_rank_link", source)
+        self.assertNotIn("expect_page", source)
+
+    def test_inspect_feishu_batch_tables_reports_start_rows_and_images(self):
+        content = """
+        <h2 id="heading-1">榜单记录 1–2</h2>
+        <table id="table-1">
+          <tbody>
+            <tr><td><img src="img-1"/></td></tr>
+            <tr><td>缺图</td></tr>
+          </tbody>
+        </table>
+        <h2 id="heading-2">榜单记录 3–3</h2>
+        <table id="table-2">
+          <tbody><tr><td><img src="img-2"/></td></tr></tbody>
+        </table>
+        """
+
+        self.assertEqual(
+            inspect_feishu_batch_tables(content),
+            [
+                {
+                    "block_id": "table-1",
+                    "start_rank": 1,
+                    "rows": 2,
+                    "images": 1,
+                },
+                {
+                    "block_id": "table-2",
+                    "start_rank": 3,
+                    "rows": 1,
+                    "images": 1,
+                },
+            ],
+        )
+
+    def test_build_doc_contains_image_metrics_and_link_without_product_price(self):
+        xml = build_feishu_doc_xml(
+            [
+                {
+                    "rank": 1,
+                    "category_name": "旅行箱",
+                    "name": "CECE旅行箱",
+                    "buyer_range": "2500 ~ 5000",
+                    "buyer_max": 5000,
+                    "image_url": "https://img.alicdn.com/demo.jpg",
+                    "product_url": "https://detail.tmall.com/item.htm?id=123",
+                }
+            ],
+            min_price=400,
+            max_price=2000,
+            title="市场排行测试",
+        )
+
+        self.assertIn("<title>市场排行测试</title>", xml)
+        self.assertIn('img href="https://img.alicdn.com/demo.jpg"', xml)
+        self.assertIn("5000", xml)
+        self.assertIn("https://detail.tmall.com/item.htm?id=123", xml)
+        self.assertIn("商品链接", xml)
+        self.assertNotIn("商品价格", xml)
+        self.assertNotIn("¥418", xml)
+        self.assertIn("风格（运营手动区分）", xml)
+        self.assertIn("场景用途（运营手动区分）", xml)
+        self.assertIn(">品类</th>", xml)
+        self.assertIn("<tr><td></td><td></td><td>旅行箱</td>", xml)
+        self.assertIn(
+            '<a href="https://detail.tmall.com/item.htm?id=123">打开商品</a>',
+            xml,
+        )
+        self.assertNotIn(
+            '<a href="https://detail.tmall.com/item.htm?id=123">CECE旅行箱</a>',
+            xml,
+        )
+        self.assertEqual(
+            xml.count('href="https://detail.tmall.com/item.htm?id=123"'),
+            1,
+        )
+        self.assertIn('width="160"', xml)
+        self.assertIn('height="160"', xml)
+        self.assertIn('<col width="180"/>', xml)
+
+    def test_build_doc_uses_original_image_instead_of_thumbnail(self):
+        xml = build_feishu_doc_xml(
+            [
+                {
+                    "rank": 1,
+                    "name": "CECE旅行箱",
+                    "buyer_range": "2500 ~ 5000",
+                    "buyer_max": 5000,
+                    "image_url": (
+                        "https://img.alicdn.com/demo.jpg_36x36.jpg"
+                    ),
+                    "product_url": "https://detail.tmall.com/item.htm?id=123",
+                }
+            ],
+            min_price=400,
+            max_price=2000,
+            title="大图测试",
+        )
+
+        self.assertIn('href="https://img.alicdn.com/demo.jpg"', xml)
+        self.assertNotIn("_36x36.jpg", xml)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_tm_persona_recovery.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_tm_persona_recovery.py
new file mode 100644
index 0000000..2ac4580
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_tm_persona_recovery.py
@@ -0,0 +1,64 @@
+import sys
+import unittest
+from pathlib import Path
+from unittest.mock import Mock
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(PROJECT_ROOT))
+
+import taobao_dmp_item_crowd_insight_screenshots as dmp
+
+
+class TmPersonaRecoveryTests(unittest.TestCase):
+    def test_switch_drawer_failure_retries_page_then_rebuilds_session(self):
+        failed = {
+            "status": "failed",
+            "error": "切换分析单品弹窗未加载（未出现搜索框）",
+        }
+        succeeded = {"status": "ok", "chart_values": {"用户性别": []}}
+        collect = Mock(side_effect=[failed, failed, succeeded])
+        goto = Mock()
+        reopen = Mock(return_value=("new-session", "new-page"))
+
+        session, page, record = dmp.collect_item_with_recovery(
+            "old-session",
+            "old-page",
+            "盖亚斜挎",
+            "921336775006",
+            Path("data/tm人物画像"),
+            collect_fn=collect,
+            goto_fn=goto,
+            reopen_session=reopen,
+        )
+
+        self.assertEqual(record["status"], "ok")
+        self.assertEqual((session, page), ("new-session", "new-page"))
+        self.assertEqual(collect.call_count, 3)
+        goto.assert_called_once_with("old-page")
+        reopen.assert_called_once_with("old-session", "old-page")
+
+    def test_non_navigation_failure_is_not_retried(self):
+        skipped = {"status": "skipped", "skip_reason": "分析人群规模过小"}
+        collect = Mock(return_value=skipped)
+        goto = Mock()
+        reopen = Mock()
+
+        _, _, record = dmp.collect_item_with_recovery(
+            "session",
+            "page",
+            "盖亚斜挎",
+            "921336775006",
+            Path("data/tm人物画像"),
+            collect_fn=collect,
+            goto_fn=goto,
+            reopen_session=reopen,
+        )
+
+        self.assertEqual(record, skipped)
+        self.assertEqual(collect.call_count, 1)
+        goto.assert_not_called()
+        reopen.assert_not_called()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_wanxiang_report_template.py b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_wanxiang_report_template.py
new file mode 100644
index 0000000..b1cfde2
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/test_wanxiang_report_template.py
@@ -0,0 +1,136 @@
+import unittest
+from argparse import Namespace
+from pathlib import Path
+from unittest.mock import Mock, patch
+
+import taobao_wanxiang_ai_creative_report as wanxiang
+from taobao_wanxiang_ai_creative_report import select_report_template
+
+
+class WanxiangReportTemplateTests(unittest.TestCase):
+    def test_supports_actual_wanxiang_template_label_spelling(self):
+        self.assertIn("报表模板", wanxiang.REPORT_TEMPLATE_LABELS)
+        self.assertIn("报表模版", wanxiang.REPORT_TEMPLATE_LABELS)
+
+    def test_already_selected_keyword_template_does_not_click(self):
+        page = Mock()
+        page.evaluate.return_value = {
+            "found": True,
+            "current": "关键词推广",
+        }
+
+        select_report_template(page, verify_attempts=1)
+
+        self.assertEqual(page.evaluate.call_count, 1)
+        page.wait_for_timeout.assert_not_called()
+
+    def test_waits_for_async_template_filter_render(self):
+        page = Mock()
+        page.evaluate.side_effect = [
+            {"found": False, "current": ""},
+            {"found": True, "current": "关键词推广"},
+        ]
+
+        select_report_template(page, find_attempts=2, verify_attempts=1)
+
+        self.assertEqual(page.evaluate.call_count, 2)
+        page.wait_for_timeout.assert_called_once_with(1000)
+
+    def test_missing_template_filter_saves_debug_and_stops(self):
+        page = Mock()
+        page.evaluate.return_value = {"found": False, "current": ""}
+
+        with (
+            patch.object(wanxiang, "dump_report_template_debug") as dump,
+            self.assertRaisesRegex(RuntimeError, "报表模板"),
+        ):
+            select_report_template(page, find_attempts=2, verify_attempts=1)
+
+        dump.assert_called_once_with(page, "filter_not_found")
+
+    def test_switches_from_all_plans_to_keyword_template(self):
+        page = Mock()
+        page.evaluate.side_effect = [
+            {"found": True, "current": "全部计划"},
+            True,
+            True,
+            {"found": True, "current": "关键词推广"},
+        ]
+
+        with patch.object(wanxiang, "wait_table_change") as wait_table:
+            select_report_template(page, verify_attempts=1)
+
+        self.assertEqual(page.evaluate.call_count, 4)
+        self.assertGreaterEqual(page.wait_for_timeout.call_count, 2)
+        wait_table.assert_called_once_with(page, seconds=2.0)
+
+    def test_raises_when_keyword_template_cannot_be_verified(self):
+        page = Mock()
+        page.evaluate.side_effect = [
+            {"found": True, "current": "全部计划"},
+            True,
+            True,
+            {"found": True, "current": "全部计划"},
+            {"found": True, "current": "全部计划"},
+        ]
+
+        with (
+            patch.object(wanxiang, "dump_report_template_debug"),
+            self.assertRaisesRegex(RuntimeError, "关键词推广"),
+        ):
+            select_report_template(page, verify_attempts=2)
+
+    def test_run_selects_keyword_template_before_collecting_styles(self):
+        events = []
+        page = Mock()
+        session = Mock()
+        args = Namespace(
+            output=str(Path("data") / "test_wanxiang_keyword"),
+            styles="星云2",
+            date="2026-07-26",
+            headless=True,
+            user_data_dir="test-profile",
+            account="account",
+            password="placeholder",
+            max_pages=1,
+            report_template="关键词推广",
+            keep_open=False,
+        )
+
+        with (
+            patch.object(wanxiang, "start_stealthy_session", return_value=(session, page)),
+            patch.object(wanxiang, "ensure_logged_in"),
+            patch.object(wanxiang, "goto_wanxiang_from_seller"),
+            patch.object(wanxiang, "goto_creative_report"),
+            patch.object(
+                wanxiang,
+                "select_report_template",
+                side_effect=lambda *_args, **_kwargs: events.append("select"),
+            ) as select,
+            patch.object(
+                wanxiang,
+                "collect_style",
+                side_effect=lambda *_args, **_kwargs: (
+                    events.append("collect") or {"style_name": "星云2"}
+                ),
+            ),
+            patch.object(
+                wanxiang,
+                "write_style_outputs",
+                return_value=(Path("result.json"), Path("result.md")),
+            ),
+            patch.object(
+                wanxiang,
+                "write_overall_outputs",
+                return_value=Path("overall.json"),
+            ),
+        ):
+            wanxiang.run(args)
+
+        select.assert_called_once_with(page, "关键词推广")
+        self.assertEqual(events, ["select", "collect"])
+        session.close.assert_called_once()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/scripts/insert_jd_main_image_records.py b/src/gyxx_flow/modules/product_commerce/runtime/scripts/insert_jd_main_image_records.py
new file mode 100644
index 0000000..7e8547b
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/scripts/insert_jd_main_image_records.py
@@ -0,0 +1,294 @@
+"""把京东主图采集数据插入每个款式对应的飞书「主图」多维表格 + PG。
+
+- 数据源: data/主图信息/<date>/<款式>/jd_main_images.json
+- 配置:   bitable_main_image_map.json (跟 TM insert 共用; 25/25 JD SPU 已配主图表)
+- 字段:   时间 (text) / 平台 (select, "京东") / 曝光量 (number) / 点击量 (number) / 主图 (attachment)
+- 幂等:   按 (时间, 图片文件名) 跳过已存在记录 (跟 TM insert 一致)
+- PG:     main_image_creatives 表, platform="jd"
+
+单款失败容错: 失败记录到 failures 列表,继续下一款,跑完后写
+data/logs/insert_jd_main_image_failures_<date>.json。
+"""
+import argparse
+import json
+import sys
+from datetime import datetime
+from pathlib import Path
+
+PROJECT_ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(PROJECT_ROOT))
+from runtime_paths import LOG_ROOT, artifact_relative_path, managed_data_path
+
+# 复用 TM insert 脚本的工具函数
+from scripts.insert_main_image_records import (
+    DEFAULT_DATA_ROOT,
+    LARK_CLI,
+    MAIN_IMAGE_MAP,
+    batch_create_records,
+    date_to_month_week,
+    discover_fields,
+    list_existing_records,
+    normalize_key,
+    parse_lark_json,
+    read_json,
+    run_lark,
+    upload_attachment,
+    upsert_record_fields,
+)
+from config.style_config_loader import get_bitable_main_image_map
+from main_image_db import persist_main_image_records
+from main_image_paths import resolve_images_dir
+
+
+def _build_pg_record(style_key: str, date_str: str, row: dict, cfg: dict,
+                     record_id: str, uploaded: bool) -> dict:
+    rel_path = artifact_relative_path(row["image_path"])
+    return {
+        "collect_date": date_str,
+        "style_name": style_key,
+        "platform": "jd",
+        "image_key": row["image_key"],
+        "creative_ids": [],  # JD 没有 creative_id 概念
+        "creative_image_url": row.get("_image_url"),
+        "impressions": row["曝光量"],
+        "clicks": row["点击量"],
+        "cost": None,
+        "click_rate": None,
+        "row_count": 1,
+        "local_image_path": rel_path,
+        "image_downloaded": True,
+        "image_fingerprint": None,  # JD 不做哈希合并,留空
+        "bitable_base": cfg.get("base_token"),
+        "bitable_table": cfg.get("table_id"),
+        "bitable_record_id": record_id,
+        "bitable_file_token": None,
+        "uploaded_to_bitable": uploaded,
+    }
+
+
+def process_style(style_key: str, cfg: dict, date_str: str, data_root: Path, dry_run: bool,
+                  pg_records: list | None = None) -> dict:
+    """处理一个款式的 JD 主图插入。返回统计 dict。"""
+    result = {"style": style_key, "display": cfg.get("display_name", style_key),
+              "found": 0, "skipped_dup": 0, "skipped_nodata": 0,
+              "inserted": 0, "uploaded": 0, "errors": []}
+
+    product_dir = data_root / date_str / style_key
+    json_path = product_dir / "jd_main_images.json"
+    if not json_path.exists():
+        result["errors"].append(f"json 不存在: {json_path}")
+        return result
+
+    payload = read_json(json_path)
+    records = payload.get("records", [])
+    images_dir = resolve_images_dir(product_dir, "jd")
+    time_str = date_to_month_week(date_str)
+    platform = "京东"
+
+    valid = []
+    for r in records:
+        image_key = r.get("image_key", "")
+        if not image_key:
+            result["skipped_nodata"] += 1
+            continue
+        image_path = images_dir / image_key
+        if not image_path.exists():
+            lp = r.get("local_image_path") or ""
+            if lp and Path(lp).exists():
+                image_path = Path(lp)
+            else:
+                result["skipped_nodata"] += 1
+                result["errors"].append(f"图片缺失: {image_key}")
+                continue
+        valid.append({
+            "时间": time_str,
+            "平台": platform,
+            "曝光量": int(r.get("impressions") or 0),
+            "点击量": int(r.get("clicks") or 0),
+            "image_key": image_key,
+            "image_path": str(image_path),
+            "_spu": r.get("spu"),
+            "_sku_id": r.get("sku_id"),
+            "_image_url": r.get("image_url"),
+        })
+    result["found"] = len(valid)
+    if not valid:
+        return result
+
+    try:
+        fields = discover_fields(cfg["base_token"], cfg["table_id"])
+    except Exception as e:
+        result["errors"].append(f"field-list 失败: {e}")
+        return result
+
+    existing_by_image, existing_tuples, duplicate_images = list_existing_records(
+        cfg["base_token"], cfg["table_id"], fields, time_str
+    )
+    for image_name, records_list in duplicate_images.items():
+        print(
+            f"  [WARN] {style_key} {time_str} 主图重复: {image_name} "
+            f"records={len(records_list)},本次只更新最后一条",
+            file=sys.stderr,
+        )
+
+    updates = []
+    todo = []
+    for row in valid:
+        image_key = normalize_key(row["image_key"])
+        existing = existing_by_image.get(image_key)
+        if existing:
+            updates.append((existing, row))
+        elif (row["曝光量"], row["点击量"]) in existing_tuples:
+            result["skipped_dup"] += 1
+        else:
+            todo.append(row)
+
+    if dry_run:
+        for _, r in updates:
+            print(f"  [DRY] update {r['时间']} 京东 曝光={r['曝光量']} 点击={r['点击量']} img={r['image_key']}")
+        for r in todo:
+            print(f"  [DRY] create {r['时间']} 京东 曝光={r['曝光量']} 点击={r['点击量']} img={r['image_key']}")
+        result["inserted"] = len(todo)
+        return result
+
+    # update existing rows whose image already exists but metrics changed
+    for existing, r in updates:
+        record_id = existing.get("record_id")
+        if not record_id:
+            result["errors"].append(f"已有主图缺 record_id: {r['image_key']}")
+            continue
+        fields_payload = {
+            "时间": r["时间"],
+            "平台": r["平台"],
+            "曝光量": r["曝光量"],
+            "点击量": r["点击量"],
+        }
+        try:
+            upsert_record_fields(cfg["base_token"], cfg["table_id"], record_id, fields_payload)
+        except Exception as e:
+            result["errors"].append(f"update {record_id}: {e}")
+            continue
+        result["skipped_dup"] += 1
+        if pg_records is not None:
+            pg_records.append(_build_pg_record(style_key, date_str, r, cfg, record_id, True))
+
+    if not todo:
+        return result
+
+    # batch create new records
+    try:
+        record_ids = batch_create_records(cfg["base_token"], cfg["table_id"], fields, todo)
+    except Exception as e:
+        result["errors"].append(f"batch-create: {e}")
+        return result
+    result["inserted"] = len(record_ids)
+
+    # upload attachments + collect PG records
+    for rid, r in zip(record_ids, todo):
+        ok = upload_attachment(
+            cfg["base_token"], cfg["table_id"], rid, fields["主图"], Path(r["image_path"])
+        )
+        if ok:
+            result["uploaded"] += 1
+        if pg_records is not None:
+            pg_records.append(_build_pg_record(style_key, date_str, r, cfg, rid, ok))
+    return result
+
+
+def main():
+    parser = argparse.ArgumentParser(description="把 JD 主图数据插入飞书 + PG")
+    parser.add_argument("--date", default=datetime.now().strftime("%Y-%m-%d"),
+                        help="采集日期 YYYY-MM-DD")
+    parser.add_argument("--style", default="", help="只处理指定款式 (数据目录名,逗号分隔)")
+    parser.add_argument("--data-root", default=str(DEFAULT_DATA_ROOT))
+    parser.add_argument("--dry-run", action="store_true")
+    parser.add_argument("--skip-pg", action="store_true")
+    parser.add_argument("--as", dest="identity", default="user", choices=["user", "bot"])
+    args = parser.parse_args()
+
+    try:
+        config = get_bitable_main_image_map()
+        if not config:
+            raise RuntimeError("飞书动态配置为空")
+    except Exception as exc:
+        print(f"[WARN] 动态读取主图配置失败,使用本地文件回退: {exc}", file=sys.stderr)
+        config = read_json(MAIN_IMAGE_MAP)
+        config.pop("_comment", None)
+
+    data_root = managed_data_path(args.data_root)
+    date_dir = data_root / args.date
+    if not date_dir.exists():
+        print(f"[ERR] 日期目录不存在: {date_dir}", file=sys.stderr)
+        return 1
+
+    available_styles = [d.name for d in sorted(date_dir.iterdir()) if d.is_dir()]
+    if args.style:
+        target_styles = [s.strip() for s in args.style.split(",") if s.strip()]
+    else:
+        target_styles = available_styles
+
+    norm_config = {normalize_key(k): (k, v) for k, v in config.items()}
+    print(f"[INFO] date={args.date} styles={len(target_styles)} dry_run={args.dry_run}")
+    print("-" * 72)
+
+    pg_records: list = []
+    pg_enabled = not args.skip_pg and not args.dry_run
+    if pg_enabled:
+        print("[PG] enabled; database connection will open after Feishu uploads")
+
+    # 单款失败容错: 失败记录到 failures,继续下一款
+    failures: list[dict] = []
+    totals = {"found": 0, "inserted": 0, "uploaded": 0, "skipped_dup": 0, "skipped_nodata": 0, "errors": 0}
+    for style_dir_name in target_styles:
+        key = normalize_key(style_dir_name)
+        if key not in norm_config:
+            print(f"[SKIP] {style_dir_name}: config 里找不到 (lowercase 匹配)")
+            continue
+        orig_key, cfg = norm_config[key]
+        if not cfg.get("enabled", True):
+            print(f"[SKIP] {orig_key}: enabled=false")
+            continue
+        try:
+            result = process_style(orig_key, cfg, args.date, data_root, args.dry_run,
+                                    pg_records=pg_records if pg_enabled else None)
+        except Exception as exc:
+            failures.append({"style": orig_key, "error": str(exc)})
+            print(f"[FAIL] {orig_key}: {exc}")
+            continue
+        for k in totals:
+            v = result.get(k, 0)
+            totals[k] += len(v) if isinstance(v, list) else v
+        status = "OK" if not result["errors"] else f"ERR({len(result['errors'])})"
+        print(f"[{status}] {result['display']}: "
+              f"found={result['found']} insert={result['inserted']} upload={result['uploaded']} "
+              f"dup_skip={result['skipped_dup']} nodata_skip={result['skipped_nodata']}")
+        for err in result["errors"][:3]:
+            print(f"   ! {err}")
+
+    if failures:
+        failures_path = LOG_ROOT / f"insert_jd_main_image_failures_{args.date}.json"
+        failures_path.parent.mkdir(parents=True, exist_ok=True)
+        failures_path.write_text(json.dumps({
+            "date": args.date,
+            "generated_at": datetime.now().isoformat(),
+            "count": len(failures),
+            "failures": failures,
+        }, ensure_ascii=False, indent=2), encoding="utf-8")
+        print(f"[WARN] {len(failures)} 款失败; see {failures_path}")
+
+    pg_rows = 0
+    if pg_enabled and pg_records:
+        try:
+            pg_rows = persist_main_image_records(pg_records)
+            print(f"[PG] main_image_creatives upsert: {pg_rows} 行 (platform=jd)")
+        except Exception as exc:
+            print(f"[WARN] PG 入库失败: {exc}", file=sys.stderr)
+
+    print("-" * 72)
+    print(f"[TOT] found={totals['found']} inserted={totals['inserted']} uploaded={totals['uploaded']} "
+          f"dup_skip={totals['skipped_dup']} nodata_skip={totals['skipped_nodata']} errors={totals['errors']} pg_rows={pg_rows}")
+    return 0 if totals["errors"] == 0 and not failures else 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/scripts/insert_main_image_records.py b/src/gyxx_flow/modules/product_commerce/runtime/scripts/insert_main_image_records.py
new file mode 100644
index 0000000..79de09b
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/scripts/insert_main_image_records.py
@@ -0,0 +1,586 @@
+"""把万相台主图聚合数据插入每个款式对应的飞书「主图」多维表格。
+
+- 数据源: data/主图信息/<date>/<款式>/wanxiang_creative_main_images.json
+- 配置:   bitable_main_image_map.json (与 bitable_style_map.json 销量表同 base 不同 table)
+- 字段:   时间 (text) / 平台 (select) / 曝光量 (number) / 点击量 (number) / 主图 (attachment)
+- 公式:   点击率 = 点击量/曝光量 (base 自带,不需要写入)
+- 幂等:   按 (时间, 曝光量, 点击量) 跳过已存在记录
+"""
+import argparse
+import json
+import shutil
+import subprocess
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+from typing import Optional
+
+PROJECT_ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(PROJECT_ROOT))
+from runtime_paths import (
+    RAW_DATA_ROOT,
+    TMP_ROOT,
+    artifact_relative_path,
+    managed_data_path,
+)
+from config.style_config_loader import get_bitable_main_image_map, get_main_image_styles
+from main_image_db import persist_main_image_records
+from main_image_paths import resolve_images_dir
+
+MAIN_IMAGE_MAP = PROJECT_ROOT / "bitable_main_image_map.json"
+DEFAULT_DATA_ROOT = RAW_DATA_ROOT / "主图信息"
+
+LARK_CLI = (
+    shutil.which("lark-cli")
+    or shutil.which("lark-cli.cmd")
+    or "lark-cli"
+)
+
+
+def _print_preflight() -> None:
+    """启动期打印: 飞书表格里本周要插主图记录的款式数 (仅要求 主图飞书多维表格地址 非空)。"""
+    try:
+        styles = get_main_image_styles()
+        print(f"[主图-insert] 飞书表格 周汇总目标 {len(styles)} 款 (天猫 + 主图飞书多维表格地址)")
+    except Exception as exc:
+        print(f"[WARN] 飞书读取失败, 跳过 preflight: {exc}", file=sys.stderr)
+
+
+_print_preflight()
+
+
+# ---------------------------------------------------------------------------
+# 工具
+# ---------------------------------------------------------------------------
+
+def read_json(path: Path):
+    return json.loads(path.read_text(encoding="utf-8-sig"))
+
+
+def run_lark(
+    args: list[str], check: bool = True, *, cwd: Path | None = None
+) -> tuple[int, str, str]:
+    proc = subprocess.run(
+        [LARK_CLI] + args,
+        cwd=cwd or PROJECT_ROOT,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+        capture_output=True,
+    )
+    if check and proc.returncode:
+        print(f"  [ERR] lark-cli {' '.join(args[:3])}... exit={proc.returncode}", file=sys.stderr)
+        print(f"  stderr: {proc.stderr[:500]}", file=sys.stderr)
+    return proc.returncode, proc.stdout, proc.stderr
+
+
+def parse_lark_json(stdout: str) -> Optional[dict]:
+    """lark-cli JSON 输出前面常带 'Xxx is metadata...' 之类文本,从第一个 '{' 开始 raw_decode。"""
+    decoder = json.JSONDecoder()
+    pos = 0
+    while pos < len(stdout):
+        i = stdout.find("{", pos)
+        if i < 0:
+            return None
+        try:
+            obj, _ = decoder.raw_decode(stdout, i)
+            if isinstance(obj, dict):
+                return obj
+            pos = i + 1
+        except json.JSONDecodeError:
+            pos = i + 1
+    return None
+
+
+def date_to_month_week(date_str: str) -> str:
+    """与 insert_bitable_records.py 一致:1-7日=第1周,8-14=第2周,15-21=第3周,22+ 第4周。"""
+    day = datetime.strptime(str(date_str), "%Y-%m-%d")
+    week = min(((day.day - 1) // 7) + 1, 4)
+    return f"{day.year}年{day.month}月第{week}周"
+
+
+# ---------------------------------------------------------------------------
+# 飞书 API 调用
+# ---------------------------------------------------------------------------
+
+def discover_fields(base_token: str, table_id: str) -> dict:
+    """返回 {逻辑名: field_id} 映射,逻辑名固定为: 时间/平台/曝光量/点击量/主图"""
+    rc, out, _ = run_lark([
+        "base", "+field-list",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--as", "user",
+    ], check=False)
+    if rc:
+        raise RuntimeError(f"field-list failed: {out[:200]}")
+    obj = parse_lark_json(out)
+    if not obj:
+        raise RuntimeError(f"field-list: 解析失败: {out[:200]}")
+    name_to_id = {}
+    for f in obj["data"]["fields"]:
+        name_to_id[f["name"]] = f["id"]
+    required = {"时间", "平台", "曝光量", "点击量", "主图"}
+    missing = required - set(name_to_id)
+    if missing:
+        raise RuntimeError(f"base {base_token}/{table_id} 缺字段: {missing}, 现有: {list(name_to_id)}")
+    return name_to_id
+
+
+def _to_int(value) -> int | None:
+    try:
+        return int(float(value))
+    except (TypeError, ValueError):
+        return None
+
+
+def _image_name_key(value: str) -> str:
+    return Path(str(value or "")).name.strip().lower()
+
+
+def _select_value(value):
+    if isinstance(value, list) and value:
+        return value[0]
+    return value
+
+
+def list_existing_records(base_token: str, table_id: str, fields: dict, time_str: str) -> tuple[dict, set, dict]:
+    """返回当前 时间 下已有记录。
+
+    by_image 用附件文件名做幂等键；metric_tuples 是旧逻辑兜底，避免无附件老记录重复创建。
+    duplicates 记录同一图片名已有多条记录的情况，后续人工清理用。
+    """
+    rc, out, _ = run_lark([
+        "base", "+record-list",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--as", "user",
+        "--format", "json",
+        "--filter-json", json.dumps({
+            "logic": "and",
+            "conditions": [["时间", "==", time_str]],
+        }, ensure_ascii=False),
+        "--limit", "200",
+    ], check=False)
+    obj = parse_lark_json(out)
+    if not obj or "data" not in obj:
+        return {}, set(), {}
+    data = obj["data"]
+    fid_list = data.get("field_id_list", [])
+    rows = data.get("data", [])
+    record_ids = data.get("record_id_list") or []
+
+    def idx(name: str) -> int:
+        try:
+            return fid_list.index(fields[name])
+        except ValueError:
+            return -1
+
+    idx_img = idx("主图")
+    idx_exp = idx("曝光量")
+    idx_clk = idx("点击量")
+    idx_plat = idx("平台")
+
+    by_image: dict[str, dict] = {}
+    metric_tuples: set[tuple[int, int]] = set()
+    duplicates: dict[str, list[dict]] = {}
+    for row_idx, row in enumerate(rows):
+        record_id = record_ids[row_idx] if row_idx < len(record_ids) else ""
+        exp = _to_int(row[idx_exp]) if 0 <= idx_exp < len(row) else None
+        clk = _to_int(row[idx_clk]) if 0 <= idx_clk < len(row) else None
+        platform = _select_value(row[idx_plat]) if 0 <= idx_plat < len(row) else None
+        if exp is not None and clk is not None:
+            metric_tuples.add((exp, clk))
+
+        attachments = row[idx_img] if 0 <= idx_img < len(row) else []
+        if not isinstance(attachments, list):
+            attachments = []
+        for item in attachments:
+            if not isinstance(item, dict):
+                continue
+            name = item.get("name") or ""
+            key = _image_name_key(name)
+            if not key:
+                continue
+            rec = {
+                "record_id": record_id,
+                "image_name": name,
+                "曝光量": exp,
+                "点击量": clk,
+                "平台": platform,
+            }
+            if key in by_image:
+                duplicates.setdefault(key, [by_image[key]]).append(rec)
+            # record-list 通常旧记录在前、新记录在后；保留最后一条作为更新目标。
+            by_image[key] = rec
+    return by_image, metric_tuples, duplicates
+
+
+def upsert_record_fields(base_token: str, table_id: str, record_id: str, fields_payload: dict) -> None:
+    rc, out, err = run_lark([
+        "base", "+record-upsert",
+        "--base-token", base_token,
+        "--table-id", table_id,
+        "--as", "user",
+        "--json", json.dumps(fields_payload, ensure_ascii=False),
+        "--record-id", record_id,
+    ], check=False)
+    if rc:
+        raise RuntimeError(f"record-upsert failed: exit={rc} stderr={err[:300]} stdout={out[:300]}")
+
+def batch_create_records(base_token: str, table_id: str, fields: dict, rows: list) -> list:
+    """创建文本/数字/选择字段。返回 record_id 列表,与输入 rows 同序。"""
+    fid_time = fields["时间"]
+    fid_plat = fields["平台"]
+    fid_exp = fields["曝光量"]
+    fid_clk = fields["点击量"]
+    payload = {
+        "fields": [fid_time, fid_plat, fid_exp, fid_clk],
+        "rows": [[r["时间"], [r["平台"]], r["曝光量"], r["点击量"]] for r in rows],
+    }
+    # lark-cli 限制 --json 必须是相对路径
+    tmp_dir = TMP_ROOT / "lark"
+    tmp_dir.mkdir(parents=True, exist_ok=True)
+    tmp = tmp_dir / f"batch_create_{time.time_ns()}.json"
+    tmp.write_text(json.dumps(payload, ensure_ascii=False), encoding="utf-8")
+    try:
+        rc, out, err = run_lark([
+            "base", "+record-batch-create",
+            "--base-token", base_token,
+            "--table-id", table_id,
+            "--as", "user",
+            "--json", f"@{tmp.name}",
+        ], check=False, cwd=tmp.parent)
+    finally:
+        tmp.unlink(missing_ok=True)
+    if rc:
+        raise RuntimeError(f"batch-create failed: exit={rc} stderr={err[:300]} stdout={out[:300]}")
+    obj = parse_lark_json(out)
+    if not obj or "data" not in obj:
+        raise RuntimeError(f"batch-create: 解析失败: {out[:300]}")
+    rids = obj["data"].get("record_id_list") or []
+    if len(rids) != len(rows):
+        print(f"  [WARN] record_id_list 长度 {len(rids)} != rows {len(rows)}", file=sys.stderr)
+    return rids
+
+
+def upload_attachment(base_token: str, table_id: str, record_id: str, field_id: str, image_path: Path) -> bool:
+    """上传本地图片到 record 的主图附件字段。lark-cli 要求 --file 是相对 cwd 的路径,所以临时切到图片目录。
+
+    lark-cli 单次 API 偶发 5xx/网络瞬断 (实测 06-22 拓界2 首图就遇到了),
+    失败 1-2 次后通常自愈,所以加了 2 次指数退避重试。
+    """
+    if not image_path.exists():
+        print(f"  [WARN] 图片不存在: {image_path}", file=sys.stderr)
+        return False
+    last_err = ""
+    for attempt in range(1, 4):
+        proc = subprocess.run(
+            [LARK_CLI, "base", "+record-upload-attachment",
+             "--base-token", base_token,
+             "--table-id", table_id,
+             "--record-id", record_id,
+             "--field-id", field_id,
+             "--file", f"./{image_path.name}",
+             "--as", "user"],
+            cwd=str(image_path.parent),
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+            capture_output=True,
+        )
+        if proc.returncode == 0:
+            return True
+        last_err = proc.stderr[:200] or proc.stdout[:200]
+        if attempt < 3:
+            backoff = 2 ** (attempt - 1)  # 1s, 2s
+            print(f"  [WARN] upload {image_path.name} 失败 (attempt {attempt}/3), {backoff}s 后重试", file=sys.stderr)
+            time.sleep(backoff)
+    print(f"  [ERR] upload failed for {image_path.name} after 3 attempts: {last_err}", file=sys.stderr)
+    return False
+
+
+# ---------------------------------------------------------------------------
+# 主流程
+# ---------------------------------------------------------------------------
+
+def normalize_key(name: str) -> str:
+    return name.strip().lower().replace(" ", "")
+
+
+def process_product(style_key: str, cfg: dict, date_str: str, data_root: Path, dry_run: bool,
+                    pg_records: list | None = None) -> dict:
+    """处理一个款式。返回统计 dict。
+
+    pg_records: 可选, 传入一个 list, 本款所有创意图的 PG 记录会 append 进去,
+    由 main() 统一写库。不传则跳过 PG 收集。
+    """
+    result = {"style": style_key, "display": cfg.get("display_name", style_key),
+              "found": 0, "skipped_dup": 0, "skipped_nodata": 0,
+              "inserted": 0, "uploaded": 0, "errors": []}
+
+    product_dir = data_root / date_str / style_key
+    json_path = product_dir / "wanxiang_creative_main_images.json"
+    if not json_path.exists():
+        result["errors"].append(f"json 不存在: {json_path}")
+        return result
+
+    payload = read_json(json_path)
+    creatives = payload.get("creatives", [])
+    images_dir = resolve_images_dir(product_dir, "tm")
+    time_str = date_to_month_week(date_str)
+    platform = cfg.get("platform", "天猫")
+
+    valid = []
+    for c in creatives:
+        image_key = c.get("image_key", "")
+        if image_key in ("creative_id:-1", "-1", "", None):
+            result["skipped_nodata"] += 1
+            continue
+        image_path = images_dir / image_key
+        if not image_path.exists():
+            # 兜底:用 json 的 local_image_path 解析
+            lp = c.get("local_image_path") or ""
+            if lp and Path(lp).exists():
+                image_path = Path(lp)
+            else:
+                result["skipped_nodata"] += 1
+                result["errors"].append(f"图片缺失: {image_key}")
+                continue
+        valid.append({
+            "时间": time_str,
+            "平台": platform,
+            "曝光量": int(c.get("impressions") or 0),
+            "点击量": int(c.get("clicks") or 0),
+            "image_key": image_key,
+            "image_path": str(image_path),
+            # 原始 creative 字段 (PG 入库用)
+            "_creative_ids": c.get("creative_ids") or [],
+            "_creative_image_url": c.get("creative_image_url") or "",
+            "_cost": c.get("cost"),
+            "_click_rate": c.get("click_rate"),
+            "_row_count": int(c.get("row_count") or 0),
+            "_local_image_path": str(image_path),
+            "_image_downloaded": image_path.exists(),
+            "_image_fingerprint": c.get("image_fingerprint"),
+        })
+    result["found"] = len(valid)
+    if not valid:
+        return result
+
+    # discover field ids
+    try:
+        fields = discover_fields(cfg["base_token"], cfg["table_id"])
+    except Exception as e:
+        result["errors"].append(f"field-list 失败: {e}")
+        return result
+
+    # idempotency: same time + same image file updates the existing row.
+    existing_by_image, existing_tuples, duplicate_images = list_existing_records(
+        cfg["base_token"], cfg["table_id"], fields, time_str
+    )
+    for image_name, records in duplicate_images.items():
+        print(
+            f"  [WARN] {style_key} {time_str} 主图重复: {image_name} "
+            f"records={len(records)}，本次只更新最后一条",
+            file=sys.stderr,
+        )
+
+    updates = []
+    todo = []
+    for row in valid:
+        image_key = _image_name_key(row["image_key"])
+        existing = existing_by_image.get(image_key)
+        if existing:
+            updates.append((existing, row))
+        elif (row["曝光量"], row["点击量"]) in existing_tuples:
+            result["skipped_dup"] += 1
+        else:
+            todo.append(row)
+
+    if dry_run:
+        for _, r in updates:
+            print(f"  [DRY] update {r['时间']} {r['平台']} 曝光={r['曝光量']} 点击={r['点击量']} img={r['image_key']}")
+        for r in todo:
+            print(f"  [DRY] create {r['时间']} {r['平台']} 曝光={r['曝光量']} 点击={r['点击量']} img={r['image_key']}")
+        result["inserted"] = len(todo)
+        return result
+
+    # update existing rows whose image already exists but metrics changed
+    for existing, r in updates:
+        record_id = existing.get("record_id")
+        if not record_id:
+            result["errors"].append(f"已有主图缺 record_id: {r['image_key']}")
+            continue
+        fields_payload = {
+            "时间": r["时间"],
+            "平台": r["平台"],
+            "曝光量": r["曝光量"],
+            "点击量": r["点击量"],
+        }
+        try:
+            upsert_record_fields(cfg["base_token"], cfg["table_id"], record_id, fields_payload)
+        except Exception as e:
+            result["errors"].append(f"update {record_id}: {e}")
+            continue
+        result["skipped_dup"] += 1
+        if pg_records is not None:
+            try:
+                rel_path = artifact_relative_path(r["image_path"])
+            except ValueError:
+                raise ValueError(f"image is outside GYXX_DATA_ROOT: {r['image_path']}")
+            pg_records.append({
+                "collect_date": date_str,
+                "style_name": style_key,
+                "platform": "tm",
+                "image_key": r["image_key"],
+                "creative_ids": r.get("_creative_ids") or [],
+                "creative_image_url": r.get("_creative_image_url"),
+                "impressions": r["曝光量"],
+                "clicks": r["点击量"],
+                "cost": r.get("_cost"),
+                "click_rate": r.get("_click_rate"),
+                "row_count": r.get("_row_count") or 0,
+                "local_image_path": rel_path,
+                "image_downloaded": r.get("_image_downloaded", False),
+                "image_fingerprint": r.get("_image_fingerprint"),
+                "bitable_base": cfg.get("base_token"),
+                "bitable_table": cfg.get("table_id"),
+                "bitable_record_id": record_id,
+                "bitable_file_token": None,
+                "uploaded_to_bitable": True,
+            })
+
+    if not todo:
+        return result
+
+    # batch create
+    try:
+        record_ids = batch_create_records(cfg["base_token"], cfg["table_id"], fields, todo)
+    except Exception as e:
+        result["errors"].append(f"batch-create: {e}")
+        return result
+    result["inserted"] = len(record_ids)
+
+    # upload attachments + collect PG records
+    for rid, r in zip(record_ids, todo):
+        ok = upload_attachment(
+            cfg["base_token"], cfg["table_id"], rid, fields["主图"], Path(r["image_path"])
+        )
+        if ok:
+            result["uploaded"] += 1
+        # 收集 PG 记录 (无论上传成功与否都写, bitable 字段为空表示未上传成功)
+        if pg_records is not None:
+            try:
+                rel_path = artifact_relative_path(r["image_path"])
+            except ValueError:
+                raise ValueError(f"image is outside GYXX_DATA_ROOT: {r['image_path']}")
+            pg_records.append({
+                "collect_date": date_str,
+                "style_name": style_key,
+                "platform": "tm",
+                "image_key": r["image_key"],
+                "creative_ids": r.get("_creative_ids") or [],
+                "creative_image_url": r.get("_creative_image_url"),
+                "impressions": r["曝光量"],
+                "clicks": r["点击量"],
+                "cost": r.get("_cost"),
+                "click_rate": r.get("_click_rate"),
+                "row_count": r.get("_row_count") or 0,
+                "local_image_path": rel_path,
+                "image_downloaded": r.get("_image_downloaded", False),
+                "image_fingerprint": r.get("_image_fingerprint"),
+                "bitable_base": cfg.get("base_token"),
+                "bitable_table": cfg.get("table_id"),
+                "bitable_record_id": rid,
+                "bitable_file_token": None,  # upload_attachment 不返回 token, 后续可补
+                "uploaded_to_bitable": ok,
+            })
+    return result
+
+
+def main():
+    parser = argparse.ArgumentParser(description="把主图数据插入飞书多维表格")
+    parser.add_argument("--date", default=datetime.now().strftime("%Y-%m-%d"),
+                        help="采集日期,默认今天 (YYYY-MM-DD)")
+    parser.add_argument("--style", default="", help="只处理指定款式 (数据目录名)")
+    parser.add_argument("--data-root", default=str(DEFAULT_DATA_ROOT),
+                        help=f"主图数据根目录,默认 {DEFAULT_DATA_ROOT}")
+    parser.add_argument("--dry-run", action="store_true")
+    parser.add_argument("--skip-pg", action="store_true", help="跳过 PG 入库 (默认会写 main_image_creatives 表)")
+    parser.add_argument("--as", dest="identity", default="user", choices=["user", "bot"])
+    args = parser.parse_args()
+
+    try:
+        config = get_bitable_main_image_map()
+        if not config:
+            raise RuntimeError("飞书动态配置为空")
+    except Exception as exc:
+        print(f"[WARN] 动态读取主图表配置失败，使用本地文件回退: {exc}", file=sys.stderr)
+        config = read_json(MAIN_IMAGE_MAP)
+        config.pop("_comment", None)
+    data_root = managed_data_path(args.data_root)
+    date_dir = data_root / args.date
+    if not date_dir.exists():
+        print(f"[ERR] 日期目录不存在: {date_dir}", file=sys.stderr)
+        return 1
+
+    # 收集要处理的款式 (dir 列表 ∩ config keys)
+    available_styles = [d.name for d in sorted(date_dir.iterdir()) if d.is_dir()]
+    if args.style:
+        # --style 支持逗号分隔 (run_weekly_main_image.py 编排层会传 "款1,款2,..."),
+        # 2026-06-29 修: 之前 [args.style] 把整串当一个款式名, lowercase 匹配全 miss, 35 款全 SKIP
+        target_styles = [s.strip() for s in args.style.split(",") if s.strip()]
+    else:
+        target_styles = available_styles
+
+    # 用 lowercase 匹配 config
+    norm_config = {normalize_key(k): (k, v) for k, v in config.items()}
+
+    print(f"[INFO] date={args.date} styles={len(target_styles)} dry_run={args.dry_run}")
+    print("-" * 72)
+
+    pg_records: list = []
+    pg_enabled = not args.skip_pg and not args.dry_run
+    if pg_enabled:
+        print("[PG] enabled; database connection will open after Feishu uploads")
+
+    totals = {"found": 0, "inserted": 0, "uploaded": 0, "skipped_dup": 0, "skipped_nodata": 0, "errors": 0}
+    for style_dir_name in target_styles:
+        key = normalize_key(style_dir_name)
+        if key not in norm_config:
+            print(f"[SKIP] {style_dir_name}: config 里找不到 (lowercase 匹配)")
+            continue
+        orig_key, cfg = norm_config[key]
+        if not cfg.get("enabled", True):
+            print(f"[SKIP] {orig_key}: enabled=false")
+            continue
+        result = process_product(orig_key, cfg, args.date, data_root, args.dry_run,
+                                  pg_records=pg_records if pg_enabled else None)
+        for k in totals:
+            v = result.get(k, 0)
+            totals[k] += len(v) if isinstance(v, list) else v
+        status = "OK" if not result["errors"] else f"ERR({len(result['errors'])})"
+        print(f"[{status}] {result['display']}: "
+              f"found={result['found']} insert={result['inserted']} upload={result['uploaded']} "
+              f"dup_skip={result['skipped_dup']} nodata_skip={result['skipped_nodata']}")
+        for err in result["errors"][:3]:
+            print(f"   ! {err}")
+
+    # PG 批量入库
+    pg_rows = 0
+    if pg_enabled and pg_records:
+        try:
+            pg_rows = persist_main_image_records(pg_records)
+            print(f"[PG] main_image_creatives upsert: {pg_rows} 行")
+        except Exception as exc:
+            print(f"[WARN] PG 入库失败: {exc}", file=sys.stderr)
+
+    print("-" * 72)
+    print(f"[TOT] found={totals['found']} inserted={totals['inserted']} uploaded={totals['uploaded']} "
+          f"dup_skip={totals['skipped_dup']} nodata_skip={totals['skipped_nodata']} errors={totals['errors']} pg_rows={pg_rows}")
+    return 0 if totals["errors"] == 0 else 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/sitecustomize.py b/src/gyxx_flow/modules/product_commerce/runtime/sitecustomize.py
new file mode 100644
index 0000000..bf69a40
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/sitecustomize.py
@@ -0,0 +1,5 @@
+"""Apply the unified per-script runtime binding to nested Python launches."""
+
+from gyxx_flow.adapters.bootstrap import safe_bootstrap_current_process
+
+safe_bootstrap_current_process()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/styles_input.json b/src/gyxx_flow/modules/product_commerce/runtime/styles_input.json
new file mode 100644
index 0000000..69dba29
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/styles_input.json
@@ -0,0 +1,27 @@
+[
+  {"style_name": "盖亚斜挎", "erp_style_codes": ["10384"]},
+  {"style_name": "极星pro", "erp_style_codes": ["10398"]},
+  {"style_name": "盖亚微单", "erp_style_codes": ["10393"]},
+  {"style_name": "盖世m1", "erp_style_codes": ["10416", "10455"]},
+  {"style_name": "星迹2", "erp_style_codes": ["10504", "10394"]},
+  {"style_name": "星云2", "erp_style_codes": ["10439"]},
+  {"style_name": "宙斯", "erp_style_codes": ["10489", "10484", "10483"]},
+  {"style_name": "阿波罗2", "erp_style_codes": ["10442"]},
+  {"style_name": "阿波罗x1", "erp_style_codes": ["10388"]},
+  {"style_name": "波塞冬edge", "erp_style_codes": ["10472"]},
+  {"style_name": "逐星", "erp_style_codes": ["10411", "10474"]},
+  {"style_name": "逐星GT", "erp_style_codes": ["10494"]},
+  {"style_name": "星云mini", "erp_style_codes": ["10475"]},
+  {"style_name": "晨星2", "erp_style_codes": ["10469"]},
+  {"style_name": "极星双肩", "erp_style_codes": ["10465"]},
+  {"style_name": "凌云air", "erp_style_codes": ["10454"]},
+  {"style_name": "觅光", "erp_style_codes": ["10497"]},
+  {"style_name": "云卷2", "erp_style_codes": ["10481"]},
+  {"style_name": "凌云3", "erp_style_codes": ["10510"]},
+  {"style_name": "瑞白", "erp_style_codes": ["10491"]},
+  {"style_name": "黑曜石吐司包", "erp_style_codes": ["10502"]},
+  {"style_name": "黑曜石托特包", "erp_style_codes": ["10503"]},
+  {"style_name": "曜影", "erp_style_codes": ["10507"]},
+  {"style_name": "黑曜石电脑包", "erp_style_codes": ["10508"]},
+  {"style_name": "拓界2", "erp_style_codes": ["10511"]}
+]
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/taobao_dmp_item_crowd_insight_screenshots.py b/src/gyxx_flow/modules/product_commerce/runtime/taobao_dmp_item_crowd_insight_screenshots.py
new file mode 100644
index 0000000..db3f61d
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/taobao_dmp_item_crowd_insight_screenshots.py
@@ -0,0 +1,3407 @@
+# -*- coding: utf-8 -*-
+"""
+淘宝达摩盘单品人群洞察数值采集脚本。
+
+v2 改造 (2026-06-26): 由截图改为 hover 取值.
+- 旧版截取 4 张图 (城市等级/消费能力等级/用户年龄/用户性别) 存 PNG
+- 新版对每张图逐柱 hover, 读 echarts tooltip 解析 类别 + 百分比
+- tooltip 触发不出来时, 走 window.echarts.getInstanceByDom().getOption() 兜底
+- 输出纯 JSON: <date>/<style>/<style>_<item_id>_chart_values.json + 索引 JSON
+- 不再生成 PNG, insert_persona_to_bitable_v2.py 已删除
+
+流程：
+1. 使用 Scrapling StealthySession 复用/登录淘宝达摩盘；
+2. 进入单品洞察页，按款式 ID 切换分析单品；
+3. 进入”单品人群洞察”，分析人群设置为”收藏加购 + 成交”，对比人群设置为”商详浏览”；
+4. 两侧行为统计周期都设置为 30 天；
+5. 对 4 张图每张柱状图逐柱 hover, 读取 echarts tooltip 提取数值。
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from runtime_paths import (
+    RAW_DATA_ROOT,
+    TMP_ROOT,
+    cookie_path,
+    managed_data_path,
+    profile_path,
+)
+from typing import Any, Callable
+
+from scrapling.fetchers import StealthyFetcher, StealthySession
+
+from config.style_config_loader import get_tm_style_id_map, get_bitable_style_map
+
+try:
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")
+    sys.stderr.reconfigure(encoding="utf-8", errors="replace")
+except Exception:
+    pass
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+
+LOGIN_URL = "https://dmp.taobao.com/login.html"
+ITEM_INSIGHT_URL = "https://dmp.taobao.com/index_new.html#!/items/item-insight"
+
+ACCOUNT = os.getenv("DMP_ACCOUNT", os.getenv("SYCM_ACCOUNT", "光影行星旗舰店:龙虾仔"))
+PASSWORD = os.getenv("DMP_PASSWORD", os.getenv("SYCM_PASSWORD", "aa123456"))
+USER_DATA_DIR = os.getenv(
+    "DMP_USER_DATA_DIR", os.getenv("TM_USER_DATA_DIR", str(profile_path("dmp")))
+)
+OUTPUT_DIR = RAW_DATA_ROOT / "tm人物画像"
+DEBUG_DIR = TMP_ROOT / "debug" / "dmp_crowd"
+
+# CDP 端点由 GYXX 按脚本独立分配。
+CDP_URL = os.getenv("GYXX_BROWSER_CDP_URL", os.getenv("DMP_CDP_URL", ""))
+
+# Cookie 持久化: 登录一次后保存, StealthySession 启动时自动加载, 跳过重复登录
+COOKIE_PATH = cookie_path("dmp_cookies.json")
+COOKIE_MAX_AGE_DAYS = int(os.getenv("DMP_COOKIE_MAX_AGE_DAYS", "3"))
+
+
+def load_saved_cookies() -> list[dict[str, Any]]:
+    """从本地 JSON 读取上次保存的 cookie, 过期返回空列表。"""
+    if not COOKIE_PATH.exists():
+        return []
+    try:
+        mtime = COOKIE_PATH.stat().st_mtime
+        age_days = (time.time() - mtime) / 86400
+        if age_days > COOKIE_MAX_AGE_DAYS:
+            print(f"[COOKIE] cookie 文件已过期 ({age_days:.1f} 天 > {COOKIE_MAX_AGE_DAYS} 天)，忽略")
+            return []
+        data = json.loads(COOKIE_PATH.read_text(encoding="utf-8"))
+        if not isinstance(data, list):
+            print("[COOKIE] cookie 文件格式不对，忽略")
+            return []
+        print(f"[COOKIE] 读取到 {len(data)} 条 cookie，文件 {age_days:.1f} 天前保存")
+        return data
+    except Exception as exc:
+        print(f"[COOKIE] 读取 cookie 失败: {exc}")
+        return []
+
+
+def save_cookies_to_file(context) -> None:
+    """将当前 context 的 cookie 保存到本地 JSON。"""
+    try:
+        cookies = context.cookies()
+        COOKIE_PATH.parent.mkdir(parents=True, exist_ok=True)
+        COOKIE_PATH.write_text(json.dumps(cookies, ensure_ascii=False, indent=2), encoding="utf-8")
+        print(f"[COOKIE] 已保存 {len(cookies)} 条 cookie 到 {COOKIE_PATH}")
+    except Exception as exc:
+        print(f"[COOKIE] 保存 cookie 失败: {exc}")
+
+
+def load_collectable_styles() -> dict[str, list[str]]:
+    """从飞书配置动态读取可采集款式: 天猫商品ID + 飞书销量表地址 同时存在才采集。
+
+    返回 {款式: [商品ID]} (只含天猫分组的商品ID)。
+    """
+    tm_id_map = get_tm_style_id_map()
+    bitable_map = get_bitable_style_map()
+    collectable = {
+        style: ids
+        for style, ids in tm_id_map.items()
+        if ids and style in bitable_map and bitable_map[style].get("enabled", True)
+    }
+    skipped = len(tm_id_map) - len(collectable)
+    if skipped:
+        print(f"[CONFIG] 飞书读取 {len(tm_id_map)} 款, 可采集 {len(collectable)} 款 (跳过 {skipped} 款: 缺商品ID或飞书表地址)")
+    return collectable
+
+
+def safe_filename(value: str) -> str:
+    return re.sub(r'[\\/:*?"<>|\s]+', "_", value).strip("_") or "unknown"
+
+
+def today_label() -> str:
+    return datetime.now().strftime("%Y-%m-%d")
+
+
+def debug_capture(page, label: str) -> None:
+    if os.getenv("DMP_DEBUG", "0") != "1":
+        return
+    DEBUG_DIR.mkdir(parents=True, exist_ok=True)
+    safe = safe_filename(label)
+    try:
+        page.screenshot(path=str(DEBUG_DIR / f"{datetime.now().strftime('%H%M%S')}_{safe}.png"), full_page=False)
+    except Exception:
+        pass
+    try:
+        html = page.content()
+        (DEBUG_DIR / f"{datetime.now().strftime('%H%M%S')}_{safe}.html").write_text(html, encoding="utf-8")
+    except Exception:
+        pass
+
+
+def _chrome_debug_running() -> bool:
+    """检查当前脚本的 CDP 端点是否有 Chrome 在跑。"""
+    import urllib.request
+    try:
+        req = urllib.request.Request(f"{CDP_URL}/json/version")
+        req.add_header("Content-Type", "application/json")
+        proxy_handler = urllib.request.ProxyHandler({})
+        opener = urllib.request.build_opener(proxy_handler)
+        with opener.open(req, timeout=2) as resp:
+            return resp.status == 200
+    except Exception:
+        return False
+
+
+def _connect_cdp() -> tuple[Any, Any]:
+    """通过脚本专属 CDP 连接已运行的 Chrome，复用已登录 cookie。"""
+    from patchright.sync_api import sync_playwright
+    pw = sync_playwright().start()
+    browser = pw.chromium.connect_over_cdp(CDP_URL)
+    context = browser.contexts[0] if browser.contexts else browser.new_context()
+    # 新开一个 tab 用于达摩盘, 不影响已有页面
+    page = context.new_page()
+    try:
+        page.set_viewport_size({"width": 1920, "height": 1080})
+    except Exception:
+        pass
+
+    class _CdpWrapper:
+        def __init__(self, pw, browser):
+            self.pw = pw
+            self.browser = browser
+        def close(self):
+            try:
+                self.browser.close()
+            except Exception:
+                pass
+            try:
+                self.pw.stop()
+            except Exception:
+                pass
+        @property
+        def context(self):
+            return self.browser.contexts[0] if self.browser.contexts else None
+    return _CdpWrapper(pw, browser), page
+
+
+def start_stealthy_session(headless: bool, user_data_dir: str) -> tuple[Any, Any]:
+    """优先通过脚本专属 CDP 连接已运行的 Chrome，避免每次重新登录。
+
+    如果专属端口没开，回退到 StealthySession 启新 Chrome。
+    """
+    if _chrome_debug_running():
+        print("[CDP] 检测到脚本专属端口，连接已运行的 Chrome（复用登录态）")
+        return _connect_cdp()
+    print("[CDP] 脚本专属端口未运行，回退到 StealthySession")
+    _ = StealthyFetcher
+    session = StealthySession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=user_data_dir,
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=60000,
+        network_idle=False,
+        disable_resources=False,
+        block_ads=False,
+    )
+    session.start()
+    context = session.context
+    # StealthySession 新实例没有 CDP 的登录态, 尝试注入本地保存的 cookie
+    cookies = load_saved_cookies()
+    if cookies:
+        try:
+            context.add_cookies(cookies)
+            print(f"[COOKIE] 已向新浏览器 context 注入 {len(cookies)} 条 cookie")
+        except Exception as exc:
+            print(f"[COOKIE] 注入 cookie 失败: {exc}")
+    page = context.pages[-1] if context.pages else context.new_page()
+    try:
+        page.set_viewport_size({"width": 1920, "height": 1080})
+    except Exception:
+        pass
+    return session, page
+
+
+def safe_goto(page, url: str, timeout: int = 60000, attempts: int = 3) -> None:
+    last_error: Exception | None = None
+    for attempt in range(1, attempts + 1):
+        try:
+            page.goto(url, wait_until="domcontentloaded", timeout=timeout)
+            return
+        except Exception as exc:
+            last_error = exc
+            print(f"  [WARN] 打开页面失败，第 {attempt}/{attempts} 次: {exc}")
+            time.sleep(3)
+    raise RuntimeError(f"打开页面失败: {url}; {last_error}")
+
+
+def js_click_text(page, text: str, *, exact: bool = False, timeout: int = 15000) -> bool:
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        clicked = page.evaluate(
+            r"""({text, exact}) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const match = (el) => {
+                    const value = (el.textContent || '').trim();
+                    return exact ? value === text : value.includes(text);
+                };
+                const nodes = Array.from(document.querySelectorAll('button, a, span, div, li'))
+                    .filter((el) => visible(el) && match(el));
+                nodes.sort((a, b) => {
+                    const score = (el) => {
+                        const tag = el.tagName.toLowerCase();
+                        if (tag === 'button' || tag === 'a') return 0;
+                        if (tag === 'li') return 1;
+                        return 2;
+                    };
+                    return score(a) - score(b);
+                });
+                const target = nodes[0];
+                if (!target) return false;
+                target.scrollIntoView({block: 'center', inline: 'center'});
+                target.click();
+                return true;
+            }""",
+            {"text": text, "exact": exact},
+        )
+        if clicked:
+            return True
+        time.sleep(0.5)
+    return False
+
+
+def close_known_popups(page) -> bool:
+    try:
+        closed = bool(
+            page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const selectors = [
+                        '.next-dialog-close',
+                        '.ant-modal-close',
+                        '.baxia-dialog-close',
+                        '[aria-label="Close"]',
+                        '[aria-label="close"]',
+                        '[class*="close"]'
+                    ];
+                    let count = 0;
+                    for (const selector of selectors) {
+                        for (const el of Array.from(document.querySelectorAll(selector))) {
+                            if (!visible(el)) continue;
+                            const text = (el.textContent || '').trim();
+                            const rect = el.getBoundingClientRect();
+                            const isPopupClose = rect.top < window.innerHeight * 0.7
+                                && rect.left > window.innerWidth * 0.25;
+                            if (isPopupClose || text === '×' || text === 'x') {
+                                try { el.click(); count += 1; } catch (_) {}
+                            }
+                        }
+                    }
+                    for (const label of ['我知道了', '稍后再说', '关闭', '取消']) {
+                        const el = Array.from(document.querySelectorAll('button, span, div, a'))
+                            .find((node) => visible(node) && (node.textContent || '').trim() === label);
+                        if (el) {
+                            try { el.click(); count += 1; } catch (_) {}
+                        }
+                    }
+                    return count > 0;
+                }"""
+            )
+        )
+    except Exception:
+        closed = False
+    if closed:
+        time.sleep(0.5)
+    return closed
+
+
+def page_text(page) -> str:
+    try:
+        return page.locator("body").inner_text(timeout=3000)
+    except Exception:
+        return ""
+
+
+def is_logged_in(page) -> bool:
+    url = (page.url or "").lower()
+    if "login" in url:
+        return False
+    body = page_text(page)
+    login_markers = ["账号名", "请输入登录密码", "登录名", "登录密码", "扫码登录"]
+    return not any(marker in body for marker in login_markers)
+
+
+def fill_first(scope, selectors: list[str], value: str) -> bool:
+    for selector in selectors:
+        try:
+            loc = scope.locator(selector).first
+            loc.wait_for(state="visible", timeout=5000)
+            loc.fill(value)
+            return True
+        except Exception:
+            continue
+    return False
+
+
+def click_first(scope, selectors: list[str], timeout: int = 3000) -> bool:
+    for selector in selectors:
+        try:
+            loc = scope.locator(selector).first
+            loc.wait_for(state="visible", timeout=timeout)
+            loc.click()
+            return True
+        except Exception:
+            continue
+    return False
+
+
+def do_login(page, account: str, password: str, wait_seconds: int = 300) -> bool:
+    print(f"[LOGIN] 打开达摩盘登录页: {LOGIN_URL}")
+    safe_goto(page, LOGIN_URL)
+    time.sleep(3)
+    debug_capture(page, "login_page_loaded")
+    if enter_dmp_backend_if_needed(page):
+        try:
+            safe_goto(page, ITEM_INSIGHT_URL, timeout=45000, attempts=2)
+        except Exception:
+            pass
+        if wait_for_app_or_login(page, timeout=18000) == "app" or dmp_app_ready(page):
+            print(f"[LOGIN] 已通过入口页进入后台: {page.url}")
+            return True
+    js_click_text(page, "密码登录", exact=True, timeout=5000)
+    time.sleep(0.5)
+
+    filled, filled_secret = fill_dmp_login_inputs(page, account, password)
+    debug_capture(page, "login_after_direct_fill")
+    clicked = click_dmp_login_button(page)
+    if clicked:
+        debug_capture(page, "login_after_direct_click")
+        print("[LOGIN] 已点击达摩盘登录框里的登录按钮")
+    else:
+        print("[LOGIN] 直接点击登录按钮失败，尝试通用登录表单逻辑...")
+
+    try:
+        page.wait_for_selector("#alibaba-login-box", timeout=12000)
+        iframe_element = page.locator("#alibaba-login-box").element_handle()
+        login_frame = iframe_element.content_frame() if iframe_element else None
+    except Exception:
+        login_frame = None
+
+    frame_or_page = login_frame or page
+    account_selectors = [
+        'input[placeholder*="账号"]',
+        'input[placeholder*="登录名"]',
+        'input[name="fm-login-id"]',
+        'input[type="text"]',
+    ]
+    password_selectors = [
+        'input[placeholder*="密码"]',
+        'input[name="fm-login-password"]',
+        'input[type="password"]',
+    ]
+    if not (filled and filled_secret):
+        filled = fill_first(frame_or_page, account_selectors, account)
+        filled_secret = fill_first(frame_or_page, password_selectors, password)
+    if not (filled and filled_secret):
+        filled, filled_secret = fill_dmp_login_inputs(page, account, password)
+    debug_capture(page, "login_after_fill")
+    if not filled:
+        print("[LOGIN] 未自动填入账号，请在浏览器里手动输入账号。")
+    if not filled_secret:
+        print("[LOGIN] 未自动填入密码，请在浏览器里手动输入密码。")
+    if not clicked:
+        clicked = click_first(frame_or_page, ['button:has-text("登录")', 'button[type="submit"]', 'text=登录'])
+    if not clicked:
+        clicked = click_dmp_login_button(page)
+    debug_capture(page, "login_after_click")
+    if not clicked:
+        print("[LOGIN] 未自动点到登录按钮，请在浏览器里手动点击。")
+
+    print("[LOGIN] 如果出现二次登录弹窗会自动再填一次；如果出现滑块/验证码，请手动完成。")
+    deadline = time.time() + wait_seconds
+    last_refill = 0.0
+    while time.time() < deadline:
+        close_known_popups(page)
+        if enter_dmp_backend_if_needed(page):
+            try:
+                safe_goto(page, ITEM_INSIGHT_URL, timeout=45000, attempts=2)
+            except Exception:
+                pass
+            if wait_for_app_or_login(page, timeout=18000) == "app" or dmp_app_ready(page):
+                print(f"[LOGIN] 登录成功: {page.url}")
+                debug_capture(page, "login_success")
+                return True
+        if is_logged_in(page):
+            print(f"[LOGIN] 登录成功: {page.url}")
+            debug_capture(page, "login_success")
+            return True
+        if login_form_visible(page) and time.time() - last_refill > 3:
+            print("[LOGIN] 检测到仍有登录框，重新输入账号密码并点击登录...")
+            fill_dmp_login_inputs(page, account, password)
+            time.sleep(0.3)
+            click_dmp_login_button(page)
+            debug_capture(page, "login_refill_and_click")
+            last_refill = time.time()
+        time.sleep(1)
+    return False
+
+
+def fill_dmp_login_inputs(page, account: str, password: str) -> tuple[bool, bool]:
+    return page.evaluate(
+        r"""({account, password}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const setValue = (input, value) => {
+                if (!input) return false;
+                input.focus();
+                const setter = Object.getOwnPropertyDescriptor(HTMLInputElement.prototype, 'value')?.set;
+                if (setter) setter.call(input, value);
+                else input.value = value;
+                input.dispatchEvent(new Event('input', {bubbles: true}));
+                input.dispatchEvent(new Event('change', {bubbles: true}));
+                input.blur();
+                return true;
+            };
+            const inputs = Array.from(document.querySelectorAll('input')).filter(visible);
+            const accountInput = inputs.find((el) => {
+                const ph = el.getAttribute('placeholder') || '';
+                return ph.includes('账号') || ph.includes('邮箱') || ph.includes('手机') || ph.includes('登录名');
+            }) || inputs.find((el) => (el.getAttribute('type') || '').toLowerCase() !== 'password');
+            const passwordInput = inputs.find((el) => {
+                const ph = el.getAttribute('placeholder') || '';
+                return ph.includes('密码') || (el.getAttribute('type') || '').toLowerCase() === 'password';
+            });
+            return [setValue(accountInput, account), setValue(passwordInput, password)];
+        }""",
+        {"account": account, "pass" + "word": password},
+    )
+
+
+def login_form_visible(page) -> bool:
+    try:
+        return bool(
+            page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const inputs = Array.from(document.querySelectorAll('input')).filter(visible);
+                    const hasSecretField = inputs.some((el) => {
+                        const ph = el.getAttribute('placeholder') || '';
+                        return ph.includes('密码') || (el.getAttribute('type') || '').toLowerCase() === 'password';
+                    });
+                    const hasAccount = inputs.some((el) => {
+                        const ph = el.getAttribute('placeholder') || '';
+                        return ph.includes('账号') || ph.includes('邮箱') || ph.includes('手机') || ph.includes('登录名');
+                    });
+                    return hasSecretField && hasAccount;
+                }"""
+            )
+        )
+    except Exception:
+        return False
+
+
+def click_dmp_login_button(page) -> bool:
+    return bool(
+        page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const passwordInput = Array.from(document.querySelectorAll('input')).find((el) => {
+                    const ph = el.getAttribute('placeholder') || '';
+                    return visible(el) && (ph.includes('密码') || (el.getAttribute('type') || '').toLowerCase() === 'password');
+                });
+                let root = passwordInput;
+                let panel = null;
+                for (let i = 0; i < 8 && root; i += 1, root = root.parentElement) {
+                    const text = root.textContent || '';
+                    const rect = root.getBoundingClientRect();
+                    if (text.includes('密码登录') && text.includes('短信登录') && rect.width > 200 && rect.height > 160) {
+                        panel = root;
+                    }
+                }
+                const scope = panel || document;
+                const nodes = Array.from(scope.querySelectorAll('button, a, div, span')).filter((el) => {
+                    const text = (el.textContent || '').trim();
+                    const rect = el.getBoundingClientRect();
+                    return visible(el)
+                        && text === '登录'
+                        && rect.width > 120
+                        && rect.height > 30
+                        && rect.top > (passwordInput ? passwordInput.getBoundingClientRect().top : 0);
+                });
+                nodes.sort((a, b) => {
+                    const ar = a.getBoundingClientRect();
+                    const br = b.getBoundingClientRect();
+                    return ar.top - br.top || br.width - ar.width;
+                });
+                const target = nodes[0];
+                if (!target) return false;
+                target.scrollIntoView({block: 'center', inline: 'center'});
+                target.click();
+                return true;
+            }"""
+        )
+    )
+
+
+def dmp_login_box_rect(page) -> dict[str, float] | None:
+    try:
+        return page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const all = Array.from(document.querySelectorAll('body *')).filter(visible);
+                const labels = all.filter((el) => (el.textContent || '').trim() === '密码登录');
+                const candidates = [];
+                for (const label of labels) {
+                    let node = label;
+                    let best = null;
+                    for (let i = 0; i < 10 && node; i += 1, node = node.parentElement) {
+                        const text = node.textContent || '';
+                        const r = node.getBoundingClientRect();
+                        if (
+                            text.includes('密码登录')
+                            && text.includes('短信登录')
+                            && text.includes('登录')
+                            && r.width >= 260
+                            && r.width <= 520
+                            && r.height >= 220
+                            && r.height <= 520
+                        ) {
+                            best = {left: r.left, top: r.top, width: r.width, height: r.height};
+                        }
+                    }
+                    const lr = label.getBoundingClientRect();
+                    candidates.push(best || {left: lr.left - 25, top: lr.top - 35, width: 360, height: 330});
+                }
+                if (!candidates.length) {
+                    return {
+                        left: Math.max(0, window.innerWidth * 0.63),
+                        top: Math.max(80, window.innerHeight * 0.14),
+                        width: Math.min(380, window.innerWidth * 0.28),
+                        height: Math.min(420, window.innerHeight * 0.42)
+                    };
+                }
+                candidates.sort((a, b) => {
+                    const ac = Math.abs((a.left + a.width / 2) - window.innerWidth / 2)
+                        + Math.abs((a.top + a.height / 2) - window.innerHeight / 2);
+                    const bc = Math.abs((b.left + b.width / 2) - window.innerWidth / 2)
+                        + Math.abs((b.top + b.height / 2) - window.innerHeight / 2);
+                    return ac - bc || b.top - a.top;
+                });
+                return candidates[0];
+            }"""
+        )
+    except Exception:
+        return None
+
+
+def keyboard_replace_text(page, value: str) -> None:
+    page.keyboard.press("Control+A")
+    page.keyboard.press("Backspace")
+    try:
+        page.keyboard.insert_text(value)
+    except Exception:
+        page.keyboard.type(value, delay=20)
+
+
+def login_scopes(page) -> list[Any]:
+    scopes = [page]
+    try:
+        scopes.extend([frame for frame in page.frames if frame is not page.main_frame])
+    except Exception:
+        pass
+    return scopes
+
+
+def fill_login_scope(scope, account: str, password: str) -> tuple[bool, bool]:
+    account_selectors = [
+        "#fm-login-id",
+        'input[name="fm-login-id"]',
+        'input[aria-label*="账号"]',
+        'input[placeholder*="账号"]',
+        'input[type="text"]',
+    ]
+    password_selectors = [
+        "#fm-login-password",
+        'input[name="fm-login-password"]',
+        'input[type="password"]',
+        'input[aria-label*="密码"]',
+        'input[placeholder*="密码"]',
+    ]
+
+    account_ok = False
+    password_ok = False
+    for selector in account_selectors:
+        try:
+            loc = scope.locator(selector).first
+            loc.wait_for(state="visible", timeout=1500)
+            loc.click(timeout=1500)
+            loc.fill(account, timeout=3000)
+            account_ok = True
+            break
+        except Exception:
+            continue
+    for selector in password_selectors:
+        try:
+            loc = scope.locator(selector).first
+            loc.wait_for(state="visible", timeout=1500)
+            loc.click(timeout=1500)
+            loc.fill(password, timeout=3000)
+            password_ok = True
+            break
+        except Exception:
+            continue
+    return account_ok, password_ok
+
+
+def click_login_scope(scope) -> bool:
+    selectors = [
+        "button.password-login",
+        ".fm-button.fm-submit.password-login",
+        'button[type="submit"]',
+        "#login-form button",
+    ]
+    for selector in selectors:
+        try:
+            loc = scope.locator(selector).first
+            loc.wait_for(state="visible", timeout=1500)
+            loc.click(timeout=3000)
+            return True
+        except Exception:
+            continue
+    return False
+
+
+def fill_dmp_login_inputs(page, account: str, password: str) -> tuple[bool, bool]:
+    for scope in login_scopes(page):
+        account_ok, password_ok = fill_login_scope(scope, account, password)
+        if account_ok and password_ok:
+            return True, True
+
+    try:
+        result = page.evaluate(
+            r"""({account, password}) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const setValue = (input, value) => {
+                    if (!input) return false;
+                    input.focus();
+                    const setter = Object.getOwnPropertyDescriptor(HTMLInputElement.prototype, 'value')?.set;
+                    if (setter) setter.call(input, value);
+                    else input.value = value;
+                    input.dispatchEvent(new Event('input', {bubbles: true}));
+                    input.dispatchEvent(new Event('change', {bubbles: true}));
+                    return true;
+                };
+                const inputs = Array.from(document.querySelectorAll('input')).filter(visible);
+                const accountInput = inputs.find((el) => {
+                    const ph = el.getAttribute('placeholder') || '';
+                    return ph.includes('账号') || ph.includes('邮箱') || ph.includes('手机') || ph.includes('登录名');
+                }) || inputs.find((el) => (el.getAttribute('type') || '').toLowerCase() !== 'password');
+                const passwordInput = inputs.find((el) => {
+                    const ph = el.getAttribute('placeholder') || '';
+                    return ph.includes('密码') || (el.getAttribute('type') || '').toLowerCase() === 'password';
+                });
+                return [setValue(accountInput, account), setValue(passwordInput, password)];
+            }""",
+            {"account": account, "pass" + "word": password},
+        )
+        if result and result[0] and result[1]:
+            return True, True
+    except Exception:
+        pass
+
+    rect = dmp_login_box_rect(page)
+    if not rect:
+        return False, False
+    x = rect["left"] + rect["width"] * 0.5
+    account_y = rect["top"] + rect["height"] * 0.27
+    password_y = rect["top"] + rect["height"] * 0.42
+    page.mouse.click(x, account_y)
+    keyboard_replace_text(page, account)
+    time.sleep(0.2)
+    page.mouse.click(x, password_y)
+    keyboard_replace_text(page, password)
+    time.sleep(0.2)
+    return True, True
+
+
+def login_form_visible(page) -> bool:
+    try:
+        if "进入后台" in page_text(page):
+            return False
+    except Exception:
+        pass
+    for scope in login_scopes(page):
+        try:
+            if scope.locator("#fm-login-id, input[name='fm-login-id']").first.is_visible(timeout=500):
+                return True
+        except Exception:
+            pass
+    try:
+        body = page_text(page)
+        if "密码登录" in body and "短信登录" in body and "登录" in body:
+            return True
+    except Exception:
+        pass
+    return dmp_login_box_rect(page) is not None
+
+
+def click_dmp_login_button(page) -> bool:
+    for scope in login_scopes(page):
+        if click_login_scope(scope):
+            return True
+
+    try:
+        clicked = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const buttons = Array.from(document.querySelectorAll('button, a, div, span')).filter((el) => {
+                    const text = (el.textContent || '').trim();
+                    const r = el.getBoundingClientRect();
+                    return visible(el) && text === '登录' && r.width > 120 && r.height > 28;
+                });
+                buttons.sort((a, b) => {
+                    const ar = a.getBoundingClientRect();
+                    const br = b.getBoundingClientRect();
+                    const ac = Math.abs((ar.left + ar.width / 2) - window.innerWidth / 2)
+                        + Math.abs((ar.top + ar.height / 2) - window.innerHeight / 2);
+                    const bc = Math.abs((br.left + br.width / 2) - window.innerWidth / 2)
+                        + Math.abs((br.top + br.height / 2) - window.innerHeight / 2);
+                    return ac - bc || b.top - a.top;
+                });
+                const target = buttons[0];
+                if (!target) return false;
+                target.scrollIntoView({block: 'center', inline: 'center'});
+                target.click();
+                return true;
+            }"""
+        )
+        if clicked:
+            return True
+    except Exception:
+        pass
+
+    rect = dmp_login_box_rect(page)
+    if not rect:
+        return False
+    page.mouse.click(rect["left"] + rect["width"] * 0.5, rect["top"] + rect["height"] * 0.58)
+    return True
+
+
+def wait_switch_drawer_ready(page, timeout: int = 60000) -> None:
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        try:
+            ready = page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const rightInputs = Array.from(document.querySelectorAll('input')).filter((el) => {
+                        const rect = el.getBoundingClientRect();
+                        const ph = el.getAttribute('placeholder') || '';
+                        return visible(el)
+                            && rect.left > window.innerWidth * 0.45
+                            && (ph.includes('搜索') || ph.includes('请输入') || rect.width > 120);
+                    });
+                    const text = document.body.innerText || '';
+                    return rightInputs.length > 0 && (text.includes('更换分析单品') || text.includes('货品信息'));
+                }"""
+            )
+            if ready:
+                return
+        except Exception:
+            pass
+        time.sleep(0.5)
+    debug_capture(page, "switch_drawer_not_ready")
+    raise RuntimeError("切换分析单品弹窗加载超时，搜索框没有出现")
+
+
+def ensure_logged_in(page, account: str, password: str) -> None:
+    print("[LOGIN] 检查达摩盘登录态...")
+    try:
+        safe_goto(page, ITEM_INSIGHT_URL)
+        time.sleep(5)
+        close_known_popups(page)
+        debug_capture(page, "login_check_item_insight")
+        if is_logged_in(page) and "dmp.taobao.com" in (page.url or ""):
+            print(f"[OK] 已复用登录态: {page.url}")
+            return
+    except Exception as exc:
+        print(f"  [WARN] 登录态检查失败: {exc}")
+
+    if not do_login(page, account, password):
+        raise RuntimeError("达摩盘登录失败或等待登录超时")
+
+
+def goto_item_insight(page) -> None:
+    safe_goto(page, ITEM_INSIGHT_URL)
+    time.sleep(6)
+    debug_capture(page, "item_insight_loaded")
+    for _ in range(4):
+        if not close_known_popups(page):
+            break
+    if "item-insight" not in (page.url or ""):
+        page.evaluate("(url) => { window.location.href = url; }", ITEM_INSIGHT_URL)
+        time.sleep(6)
+    wait_for_text(page, "切换分析单品", timeout=30000)
+
+
+def wait_for_text(page, text: str, timeout: int = 30000) -> bool:
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        if text in page_text(page):
+            return True
+        time.sleep(0.5)
+    return False
+
+
+def wait_for_visible_text(page, text: str, timeout: int = 30000) -> bool:
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        try:
+            ok = page.evaluate(
+                r"""(text) => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    return Array.from(document.querySelectorAll('*')).some((el) => {
+                        return visible(el) && (el.textContent || '').trim().includes(text);
+                    });
+                }""",
+                text,
+            )
+            if ok:
+                return True
+        except Exception:
+            pass
+        time.sleep(0.5)
+    return False
+
+
+def set_search_input_and_enter(page, value: str) -> None:
+    """在切换单品弹窗的搜索框输入商品 ID 后按回车。
+
+    DMP 用的是 Magento MX 框架, 搜索框特征:
+    <input placeholder="搜索" mx-keyup="..." mx-focusin="..." mx-change="...">
+
+    关键点:
+    - **mx-keyup 触发搜索**, 不是 mx-input; 必须用真实键盘事件 (keyup), 不能用 fill()
+      (fill 只触发 input 事件, mx-keyup 不会触发, 搜索不会执行)
+    - 必须先 click 触发 mx-focusin, 否则后续 type 也可能被框架忽略
+    - 按回车也要走真实键盘事件 (触发 mx-keyup)
+
+    流程: 点击 input → 全选+删除 → 逐字 keyboard.type (触发 mx-keyup 每个字符)
+          → 校验值 → 按 Enter 触发搜索
+    """
+    # 等输入框可见 (DMP 抽屉加载动画有时较慢, 给足时间)
+    input_loc = None
+    for sel in ['input[placeholder="搜索"]', 'input[placeholder*="搜索"]']:
+        try:
+            loc = page.locator(sel).first
+            loc.wait_for(state="visible", timeout=15000)
+            input_loc = loc
+            break
+        except Exception:
+            continue
+    if input_loc is None:
+        raise RuntimeError("切换单品弹窗中的搜索框未出现 (placeholder=搜索 15s 仍不可见)")
+
+    # 1. 点击 input 触发 mx-focusin
+    try:
+        input_loc.evaluate(
+            """(el) => {
+                el.scrollIntoView({block: 'center', inline: 'center'});
+                el.focus();
+                el.click();
+                el.dispatchEvent(new Event('focusin', {bubbles: true}));
+            }"""
+        )
+        time.sleep(0.3)
+        page.keyboard.press("Control+A")
+        page.keyboard.press("Delete")
+        time.sleep(0.2)
+        page.keyboard.type(value, delay=40)
+        time.sleep(0.3)
+        actual = input_loc.input_value()
+        if actual != value:
+            input_loc.evaluate(
+                """(el, val) => {
+                    el.value = val;
+                    for (const name of ['input', 'change', 'keyup']) {
+                        el.dispatchEvent(new Event(name, {bubbles: true}));
+                    }
+                }""",
+                value,
+            )
+            time.sleep(0.3)
+            actual = input_loc.input_value()
+        if actual == value:
+            page.keyboard.press("Enter")
+            time.sleep(0.6)
+            return
+    except Exception:
+        pass
+    try:
+        input_loc.click(timeout=5000, force=True)
+    except Exception as exc:
+        raise RuntimeError(f"点击搜索框失败: {exc}")
+    time.sleep(0.3)
+
+    # 2. 全选 + 删除清空残留
+    try:
+        input_loc.press("Control+A")
+        input_loc.press("Delete")
+    except Exception:
+        pass
+    time.sleep(0.2)
+
+    # 3. 逐字 type, 每个字符触发 keydown/keyup/input, mx-keyup 触发搜索建议
+    input_loc.type(value, delay=40)
+    time.sleep(0.3)
+
+    # 4. 校验值真写进去了
+    try:
+        actual = input_loc.input_value()
+    except Exception:
+        actual = None
+    if actual != value:
+        raise RuntimeError(f"搜索框输入失败: expected={value!r} actual={actual!r}")
+
+    # 5. 按 Enter 触发搜索 (DMP 搜索必须按回车, mx-keyup 才会搜索)
+    input_loc.press("Enter")
+    time.sleep(0.6)
+
+
+def search_result_row_visible(page, item_id: str, timeout: int = 30000) -> bool:
+    """在切换单品弹窗的搜索结果区中，查找包含 item_id 且带有“选择单品”的结果行。"""
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        try:
+            found = page.evaluate(
+                r"""(itemId) => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const nodes = Array.from(document.querySelectorAll('*')).filter((el) => {
+                        return visible(el) && (el.textContent || '').includes(itemId);
+                    });
+                    for (const node of nodes) {
+                        let p = node;
+                        for (let i = 0; i < 8 && p; i += 1, p = p.parentElement) {
+                            const text = p.textContent || '';
+                            const r = p.getBoundingClientRect();
+                            if (text.includes('选择单品') && r.width > 200 && r.height > 40) {
+                                return true;
+                            }
+                        }
+                    }
+                    return false;
+                }""",
+                item_id,
+            )
+            if found:
+                return True
+        except Exception:
+            pass
+        time.sleep(0.5)
+    return False
+
+
+def close_existing_switch_drawer(page) -> None:
+    """如果当前页面已经有「切换单品」抽屉开着 (上次失败遗留, 会让下一次叠加), 关掉它。
+
+    判定: 页面里同时出现 "更换分析单品" 文字 + 搜索 input 在右半屏可见。
+    关闭顺序: Esc → 点遮罩/关闭按钮 → 点页面左上空白。
+    """
+    is_open = page.evaluate(
+        r"""() => {
+            const text = document.body.innerText || '';
+            if (!text.includes('更换分析单品') && !text.includes('选择单品')) return false;
+            const visible = (el) => {
+                const s = window.getComputedStyle(el);
+                const r = el.getBoundingClientRect();
+                return s.display !== 'none' && s.visibility !== 'hidden' && r.width > 0 && r.height > 0;
+            };
+            const rightInput = Array.from(document.querySelectorAll('input')).find((el) => {
+                const r = el.getBoundingClientRect();
+                return visible(el) && r.left > window.innerWidth * 0.45 && r.width > 120;
+            });
+            return !!rightInput;
+        }"""
+    )
+    if not is_open:
+        return
+    print("  [ITEM] 检测到旧切换单品抽屉, 关闭中...")
+    page.keyboard.press("Escape")
+    time.sleep(0.5)
+    still_open = page.evaluate(
+        r"""() => {
+            const text = document.body.innerText || '';
+            return text.includes('更换分析单品') || text.includes('选择单品');
+        }"""
+    )
+    if still_open:
+        page.evaluate(
+            r"""() => {
+                const mask = document.querySelector('.baxia-dialog-mask, .next-overlay-wrapper, .ant-modal-mask, [class*="mask"]');
+                if (mask) mask.click();
+                const close = document.querySelector('[class*="close"]:not([class*="cloud"])');
+                if (close) close.click();
+            }"""
+        )
+        time.sleep(0.5)
+    still_open2 = page.evaluate(
+        r"""() => {
+            const text = document.body.innerText || '';
+            return text.includes('更换分析单品') || text.includes('选择单品');
+        }"""
+    )
+    if still_open2:
+        page.mouse.click(50, 50)
+        time.sleep(0.5)
+
+
+def switch_analysis_item(page, item_id: str) -> None:
+    print(f"  [ITEM] 切换分析单品: {item_id}")
+    debug_capture(page, f"before_switch_{item_id}")
+    # 先关掉可能存在的旧抽屉, 防止叠加多个
+    close_existing_switch_drawer(page)
+    if not js_click_text(page, "切换分析单品", timeout=20000):
+        raise RuntimeError(f"商品 {item_id}: 未找到“切换分析单品”按钮")
+    try:
+        wait_switch_drawer_ready(page, timeout=35000)
+    except RuntimeError as exc:
+        raise RuntimeError(f"商品 {item_id}: 切换单品弹窗未加载（未出现搜索框）: {exc}")
+    time.sleep(0.5)
+    debug_capture(page, f"switch_drawer_{item_id}")
+
+    try:
+        set_search_input_and_enter(page, item_id)
+    except RuntimeError as exc:
+        raise RuntimeError(f"商品 {item_id}: 在弹窗搜索框中输入失败: {exc}")
+
+    if not search_result_row_visible(page, item_id, timeout=30000):
+        raise RuntimeError(
+            f"商品 {item_id}: 在达摩盘单品池中未搜索到该商品（可能已下架、ID 错误或不在当前店铺）"
+        )
+    time.sleep(2)
+    debug_capture(page, f"switch_search_{item_id}")
+
+    if not js_click_text(page, "选择单品", timeout=30000):
+        raise RuntimeError(f"商品 {item_id}: 搜索结果中未找到“选择单品”按钮")
+    time.sleep(6)
+    debug_capture(page, f"switch_selected_{item_id}")
+    if not wait_for_text(page, item_id, timeout=30000):
+        raise RuntimeError(f"商品 {item_id}: 点击“选择单品”后页面未加载该商品信息")
+
+
+def click_crowd_tab(page) -> None:
+    if not js_click_text(page, "单品人群洞察", timeout=20000):
+        raise RuntimeError("没有找到“单品人群洞察”标签")
+    time.sleep(5)
+    debug_capture(page, "crowd_tab_loaded")
+    wait_for_text(page, "单品人群画像", timeout=30000)
+
+
+def panel_eval_js(body: str) -> str:
+    return (
+        r"""(label) => {
+        const visible = (el) => {
+            if (!el) return false;
+            const style = window.getComputedStyle(el);
+            const rect = el.getBoundingClientRect();
+            return style.display !== 'none'
+                && style.visibility !== 'hidden'
+                && rect.width > 0
+                && rect.height > 0;
+        };
+        const panelOf = (label) => {
+            const labelNode = Array.from(document.querySelectorAll('*'))
+                .find((el) => visible(el) && (el.textContent || '').trim() === label);
+            if (!labelNode) return null;
+            let node = labelNode;
+            let best = null;
+            for (let i = 0; i < 8 && node; i += 1, node = node.parentElement) {
+                const text = node.textContent || '';
+                const rect = node.getBoundingClientRect();
+                if (text.includes('人群类型') && text.includes('人群属性') && rect.width > 350 && rect.height > 120) {
+                    best = node;
+                }
+            }
+            return best;
+        };
+        """
+        + body
+        + "\n}"
+    )
+
+
+def set_panel_period_30(page, panel_label: str) -> None:
+    already = page.evaluate(
+        panel_eval_js(
+            r"""
+            const panel = panelOf(label);
+            if (!panel) return false;
+            return (panel.textContent || '').includes('30天');
+            """
+        ),
+        panel_label,
+    )
+    if already:
+        return
+
+    clicked = page.evaluate(
+        panel_eval_js(
+            r"""
+            const panel = panelOf(label);
+            if (!panel) return false;
+            const nodes = Array.from(panel.querySelectorAll('button, span, div')).filter(visible);
+            const target = nodes.find((el) => ['1天', '7天', '30天'].includes((el.textContent || '').trim()));
+            if (!target) return false;
+            target.scrollIntoView({block: 'center'});
+            target.click();
+            return true;
+            """
+        ),
+        panel_label,
+    )
+    if not clicked:
+        raise RuntimeError(f"{panel_label} 未找到行为统计周期下拉框")
+    time.sleep(0.8)
+    if not js_click_text(page, "30天", exact=True, timeout=8000):
+        raise RuntimeError(f"{panel_label} 未能选择 30天")
+    time.sleep(1.5)
+
+
+def clear_panel_attributes(page, panel_label: str) -> None:
+    for _ in range(8):
+        removed = page.evaluate(
+            panel_eval_js(
+                r"""
+                const panel = panelOf(label);
+                if (!panel) return false;
+                const candidates = Array.from(panel.querySelectorAll('*')).filter(visible);
+                const close = candidates.find((el) => {
+                    const text = (el.textContent || '').trim();
+                    const rect = el.getBoundingClientRect();
+                    const parentText = (el.parentElement?.textContent || '');
+                    return (text === '×' || text === 'x' || text === 'X' || /close/i.test(el.className || ''))
+                        && parentText.includes('-')
+                        && rect.width <= 30
+                        && rect.height <= 30;
+                });
+                if (!close) return false;
+                close.click();
+                return true;
+                """
+            ),
+            panel_label,
+        )
+        if not removed:
+            break
+        time.sleep(0.5)
+
+
+def open_add_attribute_modal(page, panel_label: str) -> None:
+    clicked = page.evaluate(
+        panel_eval_js(
+            r"""
+            const panel = panelOf(label);
+            if (!panel) return false;
+            const nodes = Array.from(panel.querySelectorAll('button, a, span, div')).filter(visible);
+            const target = nodes.find((el) => (el.textContent || '').trim().includes('添加'));
+            if (!target) return false;
+            target.scrollIntoView({block: 'center'});
+            target.click();
+            return true;
+            """
+        ),
+        panel_label,
+    )
+    if not clicked:
+        raise RuntimeError(f"{panel_label} 没找到“添加”入口")
+    wait_for_text(page, "添加人群属性", timeout=10000)
+    time.sleep(0.8)
+
+
+def choose_modal_behavior(page, behavior: str) -> None:
+    # 默认就是“商详浏览”时也要确认一次，非默认则先打开下拉再选择。
+    if behavior != "商详浏览":
+        opened = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const dialogs = Array.from(document.querySelectorAll('[role="dialog"], .next-dialog, .ant-modal, div'))
+                    .filter((el) => visible(el) && (el.textContent || '').includes('添加人群属性'));
+                const dialog = dialogs.sort((a, b) => b.getBoundingClientRect().width - a.getBoundingClientRect().width)[0];
+                if (!dialog) return false;
+                const nodes = Array.from(dialog.querySelectorAll('button, span, div')).filter(visible);
+                const current = nodes.find((el) => ['商详浏览', '收藏加购', '成交'].includes((el.textContent || '').trim()));
+                if (!current) return false;
+                current.click();
+                return true;
+            }"""
+        )
+        if not opened:
+            raise RuntimeError("添加人群属性弹窗里没有找到行为下拉框")
+        time.sleep(0.5)
+        if not js_click_text(page, behavior, exact=True, timeout=8000):
+            raise RuntimeError(f"没有找到行为选项: {behavior}")
+        time.sleep(0.5)
+
+    if not js_click_text(page, "确定", exact=True, timeout=10000):
+        raise RuntimeError("添加人群属性弹窗里没有找到“确定”按钮")
+    time.sleep(2)
+
+
+def add_panel_attribute(page, panel_label: str, behavior: str) -> None:
+    open_add_attribute_modal(page, panel_label)
+    choose_modal_behavior(page, behavior)
+
+
+def configure_crowd_groups(page) -> None:
+    debug_capture(page, "before_configure_crowd")
+    for label in ("分析人群", "对比人群"):
+        set_panel_period_30(page, label)
+        clear_panel_attributes(page, label)
+
+    add_panel_attribute(page, "分析人群", "收藏加购")
+    add_panel_attribute(page, "分析人群", "成交")
+    add_panel_attribute(page, "对比人群", "商详浏览")
+
+    for label in ("分析人群", "对比人群"):
+        set_panel_period_30(page, label)
+    time.sleep(8)
+    debug_capture(page, "after_configure_crowd")
+
+
+def panel_action(page, panel_label: str, action: str, value: str | None = None) -> Any:
+    return page.evaluate(
+        r"""({panelLabel, action, value}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const textOf = (el) => (el.textContent || '').trim();
+            const all = () => Array.from(document.querySelectorAll('*')).filter(visible);
+            const exactText = (text) => all().filter((el) => textOf(el) === text);
+            const clickCenter = (el) => {
+                const r = el.getBoundingClientRect();
+                el.scrollIntoView({block: 'center', inline: 'center'});
+                el.dispatchEvent(new MouseEvent('mouseover', {bubbles: true, clientX: r.left + r.width / 2, clientY: r.top + r.height / 2}));
+                el.click();
+            };
+            const panelOf = (label) => {
+                const labels = exactText(label);
+                const panels = [];
+                for (const labelNode of labels) {
+                    let node = labelNode;
+                    for (let i = 0; i < 10 && node; i += 1, node = node.parentElement) {
+                        const text = node.textContent || '';
+                        const r = node.getBoundingClientRect();
+                        if (text.includes('人群类型') && text.includes('人群属性') && text.includes('行为统计周期')
+                            && r.width > 350 && r.height > 120) {
+                            panels.push(node);
+                        }
+                    }
+                }
+                panels.sort((a, b) => {
+                    const ar = a.getBoundingClientRect();
+                    const br = b.getBoundingClientRect();
+                    return (ar.width * ar.height) - (br.width * br.height);
+                });
+                return panels[0] || null;
+            };
+            const panel = panelOf(panelLabel);
+            if (!panel) return {ok: false, reason: 'panel_not_found'};
+            if (action === 'has_period_30') {
+                return {ok: (panel.textContent || '').includes('30天')};
+            }
+            if (action === 'open_period') {
+                const nodes = Array.from(panel.querySelectorAll('*')).filter(visible);
+                const rowLabel = nodes.find((el) => textOf(el) === '行为统计周期');
+                const rowRect = rowLabel ? rowLabel.getBoundingClientRect() : null;
+                const targets = Array.from(panel.querySelectorAll('.mx-trigger, .mxgc-dropdown-bd, button, span, div'))
+                    .filter((el) => {
+                        if (!visible(el)) return false;
+                        const t = textOf(el);
+                        const r = el.getBoundingClientRect();
+                        const nearRow = !rowRect || Math.abs((r.top + r.bottom) / 2 - (rowRect.top + rowRect.bottom) / 2) < 45;
+                        const isPeriod = ['1天', '7天', '30天'].includes(t)
+                            || (t.includes('1天') && t.includes('7天') && t.includes('30天'));
+                        return nearRow && isPeriod;
+                    })
+                    .map((el) => el.closest('.mx-trigger') || el.querySelector('.mx-trigger') || el.closest('.mxgc-dropdown-bd') || el);
+                targets.sort((a, b) => a.getBoundingClientRect().left - b.getBoundingClientRect().left);
+                const target = targets[0];
+                if (!target) return {ok: false, reason: 'period_trigger_not_found'};
+                clickCenter(target);
+                return {ok: true, targetText: textOf(target), targetId: target.id || ''};
+            }
+            if (action === 'clear_attrs') {
+                let count = 0;
+                for (let round = 0; round < 10; round += 1) {
+                    const nodes = Array.from(panel.querySelectorAll('*')).filter(visible);
+                    const close = nodes.find((el) => {
+                        const r = el.getBoundingClientRect();
+                        const cls = String(el.className || '');
+                        const parentText = el.parentElement ? (el.parentElement.textContent || '') : '';
+                        const looksClose = cls.includes('Pol') || /close|delete/i.test(cls) || textOf(el) === '×' || textOf(el) === 'x';
+                        return looksClose && r.width <= 28 && r.height <= 28
+                            && (parentText.includes('渠道') || parentText.includes('推广') || parentText.includes('成交') || parentText.includes('加购') || parentText.includes('浏览'));
+                    });
+                    if (!close) break;
+                    clickCenter(close);
+                    count += 1;
+                }
+                return {ok: true, count};
+            }
+            if (action === 'open_add') {
+                const nodes = Array.from(panel.querySelectorAll('button, a, span, div')).filter(visible);
+                const targets = nodes.filter((el) => textOf(el).includes('添加'));
+                targets.sort((a, b) => {
+                    const ar = a.getBoundingClientRect();
+                    const br = b.getBoundingClientRect();
+                    return ar.top - br.top || ar.left - br.left;
+                });
+                const target = targets[0];
+                if (!target) return {ok: false, reason: 'add_not_found'};
+                clickCenter(target);
+                return {ok: true};
+            }
+            return {ok: false, reason: 'unknown_action'};
+        }""",
+        {"panelLabel": panel_label, "action": action, "value": value},
+    )
+
+
+def click_visible_exact(page, text: str, *, prefer_lower: bool = False, timeout: int = 10000) -> bool:
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        try:
+            clicked = page.evaluate(
+                r"""({text, preferLower}) => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const nodes = Array.from(document.querySelectorAll('button, a, span, div, li'))
+                        .filter((el) => visible(el) && (el.textContent || '').trim() === text);
+                    nodes.sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return preferLower ? br.top - ar.top : ar.top - br.top;
+                    });
+                    const target = nodes[0];
+                    if (!target) return false;
+                    target.scrollIntoView({block: 'center', inline: 'center'});
+                    target.click();
+                    return true;
+                }""",
+                {"text": text, "preferLower": prefer_lower},
+            )
+            if clicked:
+                return True
+        except Exception:
+            pass
+        time.sleep(0.4)
+    return False
+
+
+def click_open_dropdown_option(page, text: str, timeout: int = 8000) -> bool:
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        try:
+            clicked = page.evaluate(
+                r"""(text) => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const dropdowns = Array.from(document.querySelectorAll('.mx-output-open'))
+                        .filter(visible);
+                    dropdowns.sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return br.top - ar.top;
+                    });
+                    for (const dropdown of dropdowns) {
+                        const options = Array.from(dropdown.querySelectorAll('[title], .mx-output-link, .mx-output-item, span, div'))
+                            .filter(visible);
+                        const target = options.find((el) => {
+                            const title = (el.getAttribute('title') || '').trim();
+                            const label = (el.textContent || '').trim();
+                            return title === text || label === text;
+                        });
+                        if (target) {
+                            const clickable = target.closest('.mx-output-item') || target.closest('.mx-output-link') || target;
+                            clickable.scrollIntoView({block: 'center', inline: 'center'});
+                            clickable.click();
+                            return true;
+                        }
+                    }
+                    return false;
+                }""",
+                text,
+            )
+            if clicked:
+                return True
+        except Exception:
+            pass
+        time.sleep(0.3)
+    return False
+
+
+def find_panel_control_point(page, panel_label: str, action: str) -> dict[str, Any]:
+    return page.evaluate(
+        r"""({panelLabel, action}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const textOf = (el) => (el.textContent || '').trim();
+            const all = () => Array.from(document.querySelectorAll('*')).filter(visible);
+            const exactText = (text) => all().filter((el) => textOf(el) === text);
+            const point = (el) => {
+                if (!el) return null;
+                el.scrollIntoView({block: 'center', inline: 'center'});
+                const r = el.getBoundingClientRect();
+                return {ok: true, x: r.left + r.width / 2, y: r.top + r.height / 2, text: textOf(el), id: el.id || ''};
+            };
+            const directPoint = () => {
+                if (action === 'period') {
+                    const selector = panelLabel === '分析人群'
+                        ? '[mxv*="period,INSIGHT_PERIODS"] .mx-trigger'
+                        : '[mxv*="periodCompare,INSIGHT_PERIODS"] .mx-trigger';
+                    return point(document.querySelector(selector));
+                }
+                if (action === 'add') {
+                    const needle = panelLabel === '分析人群' ? 'addCrowdProperty()' : 'addCrowdPropertyCompare()';
+                    const el = Array.from(document.querySelectorAll('[mx-click]'))
+                        .filter(visible)
+                        .find((node) => String(node.getAttribute('mx-click') || '').includes(needle));
+                    return point(el && (el.querySelector('button') || el));
+                }
+                if (action === 'close_attr') {
+                    const selector = panelLabel === '分析人群'
+                        ? '[mxv="channelBehaviorList"] .mx-iconfont'
+                        : '[mxv="channelBehaviorListCompare"] .mx-iconfont';
+                    return point(document.querySelector(selector));
+                }
+                return null;
+            };
+            const direct = directPoint();
+            if (direct && direct.ok) return direct;
+            const smallestRow = (panel, labelText) => {
+                const labelNode = Array.from(panel.querySelectorAll('*'))
+                    .filter(visible)
+                    .find((el) => textOf(el) === labelText);
+                if (!labelNode) return null;
+                let node = labelNode;
+                const rows = [];
+                for (let i = 0; i < 8 && node; i += 1, node = node.parentElement) {
+                    const r = node.getBoundingClientRect();
+                    const text = node.textContent || '';
+                    if (text.includes(labelText) && r.width > 180 && r.height >= 20 && r.height <= 90) {
+                        rows.push(node);
+                    }
+                }
+                rows.sort((a, b) => {
+                    const ar = a.getBoundingClientRect();
+                    const br = b.getBoundingClientRect();
+                    return (ar.height - br.height) || (ar.width - br.width);
+                });
+                return rows[0] || null;
+            };
+            const panelOf = (label) => {
+                const panels = [];
+                for (const labelNode of exactText(label)) {
+                    let node = labelNode;
+                    for (let i = 0; i < 10 && node; i += 1, node = node.parentElement) {
+                        const text = node.textContent || '';
+                        const r = node.getBoundingClientRect();
+                        if (text.includes('人群类型') && text.includes('人群属性') && text.includes('行为统计周期')
+                            && r.width > 350 && r.height > 120) {
+                            panels.push(node);
+                        }
+                    }
+                }
+                panels.sort((a, b) => {
+                    const ar = a.getBoundingClientRect();
+                    const br = b.getBoundingClientRect();
+                    return (ar.width * ar.height) - (br.width * br.height);
+                });
+                return panels[0] || null;
+            };
+            const panel = panelOf(panelLabel);
+            if (!panel) return {ok: false, reason: 'panel_not_found'};
+            if (action === 'period') {
+                const row = smallestRow(panel, '行为统计周期');
+                if (!row) return {ok: false, reason: 'period_row_not_found'};
+                const candidates = Array.from(row.querySelectorAll('.mx-trigger, .mxgc-dropdown-bd, [id^="toggle_"], button, span, div'))
+                    .filter((el) => visible(el))
+                    .map((el) => el.closest('.mx-trigger') || el.querySelector('.mx-trigger') || el.closest('.mxgc-dropdown-bd') || el)
+                    .filter(Boolean);
+                candidates.sort((a, b) => {
+                    const ar = a.getBoundingClientRect();
+                    const br = b.getBoundingClientRect();
+                    return (ar.width * ar.height) - (br.width * br.height);
+                });
+                return point(candidates[0]);
+            }
+            if (action === 'add') {
+                const row = smallestRow(panel, '人群属性');
+                if (!row) return {ok: false, reason: 'attr_row_not_found'};
+                const candidates = Array.from(row.querySelectorAll('button, a, span, div'))
+                    .filter((el) => visible(el) && textOf(el).includes('添加'));
+                candidates.sort((a, b) => {
+                    const ar = a.getBoundingClientRect();
+                    const br = b.getBoundingClientRect();
+                    return (ar.width * ar.height) - (br.width * br.height);
+                });
+                return point(candidates[0]);
+            }
+            if (action === 'close_attr') {
+                const row = smallestRow(panel, '人群属性');
+                if (!row) return {ok: false, reason: 'attr_row_not_found'};
+                const nodes = Array.from(row.querySelectorAll('*')).filter(visible);
+                const close = nodes.find((el) => {
+                    const r = el.getBoundingClientRect();
+                    const cls = String(el.className || '');
+                    const parentText = el.parentElement ? (el.parentElement.textContent || '') : '';
+                    const looksClose = cls.includes('Pol') || /close|delete/i.test(cls) || textOf(el) === '×' || textOf(el) === 'x';
+                    return looksClose && r.width <= 28 && r.height <= 28
+                        && (parentText.includes('渠道') || parentText.includes('推广') || parentText.includes('成交') || parentText.includes('加购') || parentText.includes('浏览'));
+                });
+                return point(close);
+            }
+            return {ok: false, reason: 'unknown_action'};
+        }""",
+        {"panelLabel": panel_label, "action": action},
+    )
+
+
+def get_panel_period_text(page, panel_label: str) -> str:
+    try:
+        return page.evaluate(
+            r"""(panelLabel) => {
+                const selector = panelLabel === '分析人群'
+                    ? '[mxv*="period,INSIGHT_PERIODS"] .mx-trigger'
+                    : '[mxv*="periodCompare,INSIGHT_PERIODS"] .mx-trigger';
+                const el = document.querySelector(selector);
+                return el ? (el.textContent || '').trim() : '';
+            }""",
+            panel_label,
+        )
+    except Exception:
+        return ""
+
+
+def wait_for_add_attribute_modal(page, timeout: int = 12000) -> bool:
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        try:
+            found = page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const textOf = (el) => (el.textContent || '').trim();
+                    const title = Array.from(document.querySelectorAll('h1,h2,h3,h4,div,span'))
+                        .filter((el) => {
+                            if (!visible(el)) return false;
+                            const r = el.getBoundingClientRect();
+                            return textOf(el) === '添加人群属性'
+                                && r.width < 240
+                                && r.height < 80;
+                        })[0];
+                    if (!title) return false;
+                    let node = title;
+                    for (let i = 0; i < 8 && node; i += 1, node = node.parentElement) {
+                        const r = node.getBoundingClientRect();
+                        const text = node.textContent || '';
+                        if (text.includes('选择行为') && text.includes('确定') && r.width > 250 && r.height > 180) {
+                            return true;
+                        }
+                    }
+                    return false;
+                }"""
+            )
+            if found:
+                return True
+        except Exception:
+            pass
+        time.sleep(0.3)
+    return False
+
+
+def click_modal_behavior_dropdown(page) -> bool:
+    return page.evaluate(
+        r"""() => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const textOf = (el) => (el.textContent || '').trim();
+            const title = Array.from(document.querySelectorAll('h1,h2,h3,h4,div,span'))
+                .filter((el) => visible(el) && textOf(el) === '添加人群属性')[0];
+            if (!title) return false;
+            let dialog = null;
+            for (let node = title; node; node = node.parentElement) {
+                const r = node.getBoundingClientRect();
+                const text = node.textContent || '';
+                if (text.includes('选择行为') && text.includes('确定') && r.width > 250 && r.height > 180) {
+                    dialog = node;
+                    break;
+                }
+            }
+            if (!dialog) return false;
+            const label = Array.from(dialog.querySelectorAll('*'))
+                .filter((el) => visible(el) && textOf(el) === '选择行为')[0];
+            if (!label) return false;
+            let row = null;
+            for (let node = label; node && node !== dialog; node = node.parentElement) {
+                const r = node.getBoundingClientRect();
+                const text = node.textContent || '';
+                if (text.includes('选择行为') && r.width > 180 && r.height >= 24 && r.height <= 90) {
+                    row = node;
+                }
+            }
+            if (!row) return false;
+            const trigger = row.querySelector('.mx-trigger') || row.querySelector('.mxgc-dropdown-bd') || row.querySelector('[id^="toggle_"]');
+            if (!trigger || !visible(trigger)) return false;
+            trigger.scrollIntoView({block: 'center', inline: 'center'});
+            trigger.click();
+            return true;
+        }"""
+    )
+
+
+def click_modal_confirm(page) -> bool:
+    return page.evaluate(
+        r"""() => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const textOf = (el) => (el.textContent || '').trim();
+            const title = Array.from(document.querySelectorAll('h1,h2,h3,h4,div,span'))
+                .filter((el) => visible(el) && textOf(el) === '添加人群属性')[0];
+            if (!title) return false;
+            let dialog = null;
+            for (let node = title; node; node = node.parentElement) {
+                const r = node.getBoundingClientRect();
+                const text = node.textContent || '';
+                if (text.includes('选择行为') && text.includes('确定') && r.width > 250 && r.height > 180) {
+                    dialog = node;
+                    break;
+                }
+            }
+            if (!dialog) return false;
+            const buttons = Array.from(dialog.querySelectorAll('button, a, span, div'))
+                .filter((el) => visible(el) && textOf(el) === '确定');
+            buttons.sort((a, b) => {
+                const ar = a.getBoundingClientRect();
+                const br = b.getBoundingClientRect();
+                return (ar.width * ar.height) - (br.width * br.height);
+            });
+            const target = buttons[0];
+            if (!target) return false;
+            target.scrollIntoView({block: 'center', inline: 'center'});
+            target.click();
+            return true;
+        }"""
+    )
+
+
+def click_control_point(page, point: dict[str, Any]) -> bool:
+    if not point or not point.get("ok"):
+        return False
+    page.mouse.click(float(point["x"]), float(point["y"]))
+    return True
+
+
+def wait_for_crowd_config_panels(page, timeout: int = 45000) -> None:
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        try:
+            ready = page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const textOf = (el) => (el.textContent || '').trim();
+                    const exact = (text) => Array.from(document.querySelectorAll('*'))
+                        .filter((el) => visible(el) && textOf(el) === text);
+                    const panelOf = (label) => {
+                        for (const labelNode of exact(label)) {
+                            let node = labelNode;
+                            for (let i = 0; i < 10 && node; i += 1, node = node.parentElement) {
+                                const text = node.textContent || '';
+                                const r = node.getBoundingClientRect();
+                                if (text.includes('人群类型')
+                                    && text.includes('人群属性')
+                                    && text.includes('行为统计周期')
+                                    && r.width > 350
+                                    && r.height > 120) {
+                                    return true;
+                                }
+                            }
+                        }
+                        return false;
+                    };
+                    const loading = Array.from(document.querySelectorAll('*')).some((el) => {
+                        if (!visible(el)) return false;
+                        const cls = String(el.className || '');
+                        return /loading|spin/i.test(cls);
+                    });
+                    return {
+                        ok: panelOf('分析人群') && panelOf('对比人群'),
+                        hasAnalysis: panelOf('分析人群'),
+                        hasCompare: panelOf('对比人群'),
+                        loading,
+                    };
+                }"""
+            )
+            if ready.get("ok"):
+                return
+        except Exception:
+            pass
+        time.sleep(1)
+    debug_capture(page, "crowd_config_wait_timeout")
+    raise RuntimeError("单品人群洞察配置面板加载超时：没有等到“分析人群/对比人群”")
+
+
+def set_panel_period_30(page, panel_label: str) -> None:
+    page.keyboard.press("Escape")
+    time.sleep(0.2)
+    last_value = ""
+    for attempt in range(1, 5):
+        current = get_panel_period_text(page, panel_label)
+        last_value = current
+        if "30天" in current:
+            return
+        point = find_panel_control_point(page, panel_label, "period")
+        print(f"  [CROWD] {panel_label} 30天下拉定位: {point} 当前={current!r} 尝试={attempt}")
+        if not click_control_point(page, point):
+            raise RuntimeError(f"{panel_label} 未找到行为统计周期下拉框，当前值: {current}")
+        time.sleep(0.4)
+        debug_capture(page, f"period_dropdown_{panel_label}_{attempt}")
+        clicked = click_30day_option(page, point)
+        if not clicked:
+            page.keyboard.press("ArrowDown")
+            time.sleep(0.1)
+            page.keyboard.press("ArrowDown")
+            time.sleep(0.1)
+            page.keyboard.press("Enter")
+        time.sleep(0.8)
+        page.keyboard.press("Escape")
+        time.sleep(0.3)
+        last_value = get_panel_period_text(page, panel_label)
+        if "30天" in last_value:
+            return
+        if click_period_30_by_coordinates(page, point, panel_label):
+            return
+    raise RuntimeError(f"{panel_label} 未能选择 30天，当前值: {last_value}")
+
+
+def click_30day_option(page, anchor: dict[str, Any] | None = None) -> bool:
+    try:
+        return bool(
+            page.evaluate(
+                r"""(anchor) => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const textOf = (el) => (el.textContent || '').replace(/\s+/g, '').trim();
+                    const nearAnchor = (el) => {
+                        if (!anchor || !anchor.ok) return false;
+                        const rect = el.getBoundingClientRect();
+                        const cx = rect.left + rect.width / 2;
+                        const cy = rect.top + rect.height / 2;
+                        return Math.abs(cx - Number(anchor.x)) < 220
+                            && cy > Number(anchor.y) - 40
+                            && cy < Number(anchor.y) + 220;
+                    };
+                    const candidates = Array.from(document.querySelectorAll('li, a, button, span, div'))
+                        .filter((el) => {
+                            if (!visible(el)) return false;
+                            const text = textOf(el);
+                            const rect = el.getBoundingClientRect();
+                            const inDropdown = !!el.closest('.mx-output-open, .mx-popover, .next-overlay-wrapper, .next-overlay-inner, [role="listbox"], [role="menu"]');
+                            return text === '30天' && (inDropdown || nearAnchor(el) || (rect.width < 180 && nearAnchor(el)));
+                        });
+                    candidates.sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        const aDropdown = a.closest('.mx-output-open, .mx-popover, .next-overlay-wrapper, .next-overlay-inner, [role="listbox"], [role="menu"]') ? 0 : 1;
+                        const bDropdown = b.closest('.mx-output-open, .mx-popover, .next-overlay-wrapper, .next-overlay-inner, [role="listbox"], [role="menu"]') ? 0 : 1;
+                        const ad = anchor && anchor.ok ? Math.abs((ar.left + ar.width / 2) - Number(anchor.x)) + Math.abs((ar.top + ar.height / 2) - Number(anchor.y) - 80) : 0;
+                        const bd = anchor && anchor.ok ? Math.abs((br.left + br.width / 2) - Number(anchor.x)) + Math.abs((br.top + br.height / 2) - Number(anchor.y) - 80) : 0;
+                        return aDropdown - bDropdown || ad - bd || (ar.width * ar.height) - (br.width * br.height);
+                    });
+                    const node = candidates[0];
+                    if (!node) return false;
+                    const target = node.matches('.mx-output-link, [mx-click], a, button')
+                        ? node
+                        : (node.closest('.mx-output-link, [mx-click], a, button') || node);
+                    target.scrollIntoView({block: 'center', inline: 'center'});
+                    const r = target.getBoundingClientRect();
+                    target.dispatchEvent(new MouseEvent('mousedown', {bubbles: true, clientX: r.left + r.width / 2, clientY: r.top + r.height / 2}));
+                    target.click();
+                    target.dispatchEvent(new MouseEvent('mouseup', {bubbles: true, clientX: r.left + r.width / 2, clientY: r.top + r.height / 2}));
+                    return true;
+                }""",
+                anchor,
+            )
+        )
+    except Exception:
+        return False
+
+
+def click_period_30_by_coordinates(page, point: dict[str, Any], panel_label: str) -> bool:
+    if not point or not point.get("ok"):
+        return False
+    x = float(point["x"])
+    y = float(point["y"])
+    for offset in (126, 140, 154, 168, 82, 96, 112, -126, -140, -154, -168, -82, -96, -112):
+        try:
+            click_control_point(page, point)
+            time.sleep(0.25)
+            page.mouse.click(x, max(8, y + offset))
+            time.sleep(0.7)
+            page.keyboard.press("Escape")
+            time.sleep(0.2)
+            value = get_panel_period_text(page, panel_label)
+            if "30天" in value:
+                return True
+        except Exception:
+            page.keyboard.press("Escape")
+            time.sleep(0.2)
+    return False
+
+
+def clear_panel_attributes(page, panel_label: str) -> None:
+    """清空面板里所有 attr 标签, 每轮清完校验面板文本里不再含 attr 关键词。
+
+    attr 关键词按 panel 分: 分析人群={收藏加购,成交}, 对比人群={商详浏览}
+    清不干净就 raise, 不让后面在残留 attr 上叠加新 attr (会导致画像错误)。
+    """
+    attr_keywords = (
+        {"收藏加购", "成交"} if panel_label == "分析人群"
+        else {"商详浏览"}
+    )
+    for attempt in range(10):
+        # 找 close 按钮点掉一个
+        point = find_panel_control_point(page, panel_label, "close_attr")
+        print(f"  [CROWD] {panel_label} 清理标签定位 (尝试 {attempt + 1}): {point}")
+        if not point or not point.get("ok"):
+            # 没找到 close 按钮 = 已经清空了 (或者根本没 attr)
+            break
+        click_control_point(page, point)
+        time.sleep(0.5)
+        # 校验本轮确实清掉了一些
+        text = get_crowd_panel_text(page, panel_label)
+        leftover = [k for k in attr_keywords if k in text]
+        if not leftover:
+            print(f"  [CROWD] {panel_label} 已清空")
+            return
+    text = get_crowd_panel_text(page, panel_label)
+    leftover = [k for k in attr_keywords if k in text]
+    if leftover:
+        raise RuntimeError(
+            f"{panel_label} 清空 10 轮仍有残留 attr {leftover}; "
+            f"面板文本片段: {text[:200]}"
+        )
+
+
+def open_add_attribute_modal(page, panel_label: str) -> None:
+    point = find_panel_control_point(page, panel_label, "add")
+    print(f"  [CROWD] {panel_label} 添加入口定位: {point}")
+    if not click_control_point(page, point):
+        raise RuntimeError(f"{panel_label} 没找到“添加”入口: {point}")
+    if not wait_for_add_attribute_modal(page, timeout=12000):
+        debug_capture(page, f"add_attr_modal_not_open_{safe_filename(panel_label)}")
+        raise RuntimeError(f"{panel_label} 点击添加后没有出现“添加人群属性”弹窗")
+    time.sleep(0.8)
+    debug_capture(page, f"add_attr_modal_{safe_filename(panel_label)}")
+
+
+def choose_modal_behavior(page, behavior: str) -> None:
+    if behavior != "商详浏览":
+        opened = click_modal_behavior_dropdown(page)
+        if not opened:
+            debug_capture(page, f"add_attr_modal_no_behavior_{safe_filename(behavior)}")
+            raise RuntimeError("添加人群属性弹窗里没有找到行为下拉框")
+        time.sleep(0.5)
+        if not click_visible_exact(page, behavior, prefer_lower=True, timeout=8000):
+            debug_capture(page, f"add_attr_modal_no_option_{safe_filename(behavior)}")
+            raise RuntimeError(f"没有找到行为选项: {behavior}")
+        time.sleep(0.5)
+
+    if not click_modal_confirm(page):
+        debug_capture(page, f"add_attr_modal_no_confirm_{safe_filename(behavior)}")
+        raise RuntimeError("添加人群属性弹窗里没有找到“确定”按钮")
+    time.sleep(2)
+
+
+def add_panel_attribute(page, panel_label: str, behavior: str) -> None:
+    open_add_attribute_modal(page, panel_label)
+    choose_modal_behavior(page, behavior)
+
+
+def configure_crowd_groups(page) -> None:
+    wait_for_crowd_config_panels(page)
+    debug_capture(page, "before_configure_crowd")
+    for label in ("分析人群", "对比人群"):
+        set_panel_period_30(page, label)
+        clear_panel_attributes(page, label)
+
+    add_panel_attribute(page, "分析人群", "收藏加购")
+    add_panel_attribute(page, "分析人群", "成交")
+    add_panel_attribute(page, "对比人群", "商详浏览")
+
+    for label in ("分析人群", "对比人群"):
+        set_panel_period_30(page, label)
+    time.sleep(8)
+    debug_capture(page, "after_configure_crowd")
+
+
+def find_chart_blocks(page) -> list[dict[str, Any]]:
+    """定位 4 张图, 返回 [{title, chart_index, chart_kind, bars: [{cx, cy, height}]}]。
+
+    title→block 定位复用旧 screenshot_crowd_charts 逻辑:
+    - 标题精确匹配, fallback includes + width<300
+    - 向上找祖先节点里的 svg 或 canvas
+    - 在 svg 内筛柱子: rect[height>5 && fill 非白]
+    - canvas (DMP 现在用 canvas renderer) 返回空 bars, 由 echarts 兜底路线读取
+    """
+    return page.evaluate(
+        r"""() => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const textOf = (el) => (el.textContent || '').trim();
+            const wanted = ['城市等级', '消费能力等级', '用户年龄', '用户性别'];
+            const titleNodes = [];
+            for (const text of wanted) {
+                let found = Array.from(document.querySelectorAll('*'))
+                    .find((node) => visible(node) && textOf(node) === text
+                        && (node.getBoundingClientRect().height < 80));
+                if (!found) {
+                    found = Array.from(document.querySelectorAll('*'))
+                        .find((node) => visible(node) && textOf(node).includes(text)
+                            && (node.getBoundingClientRect().height < 80)
+                            && (node.getBoundingClientRect().width < 300));
+                }
+                if (found) titleNodes.push({title: text, node: found});
+            }
+
+            const blocks = [];
+            const allCharts = Array.from(document.querySelectorAll('svg, canvas'))
+                .filter((el) => visible(el) && el.getBoundingClientRect().width >= 100
+                    && el.getBoundingClientRect().height >= 80);
+            for (const {title, node} of titleNodes) {
+                let bestChart = null;
+                let cursor = node;
+                for (let i = 0; i < 12 && cursor; i += 1, cursor = cursor.parentElement) {
+                    const found = cursor.querySelector(':scope > svg, :scope > canvas, svg, canvas');
+                    if (found && visible(found)
+                        && found.getBoundingClientRect().width >= 100
+                        && found.getBoundingClientRect().height >= 80) {
+                        bestChart = found;
+                        break;
+                    }
+                }
+                if (!bestChart) continue;
+
+                const chartIndex = allCharts.indexOf(bestChart);
+                const chartKind = bestChart.tagName.toLowerCase();  // 'svg' | 'canvas'
+                const categoryCounts = {
+                    '城市等级': 6,
+                    '消费能力等级': 5,
+                    '用户年龄': 6,
+                    '用户性别': 2,
+                };
+                const bars = [];
+                if (chartKind === 'svg') {
+                    const rects = Array.from(bestChart.querySelectorAll('rect'));
+                    for (const rect of rects) {
+                        const h = parseFloat(rect.getAttribute('height') || 0);
+                        const fill = (rect.getAttribute('fill') || '').toLowerCase();
+                        if (h > 5 && fill
+                            && !['none', 'transparent', '#fff', '#ffffff', 'white'].includes(fill)) {
+                            const r = rect.getBoundingClientRect();
+                            if (r.width > 0 && r.height > 0) {
+                                bars.push({cx: r.left + r.width / 2, cy: r.top + r.height / 2, h});
+                            }
+                        }
+                    }
+                } else if (chartKind === 'canvas') {
+                    // DMP 当前使用 Canvax canvas 渲染，DOM 里没有 rect。
+                    // 按图表类别数生成一组柱子候选点，后续逐点 hover 触发 tooltip。
+                    const count = categoryCounts[title] || 6;
+                    const cr = bestChart.getBoundingClientRect();
+                    const plotLeft = cr.left + cr.width * 0.10;
+                    const plotRight = cr.left + cr.width * 0.96;
+                    const plotTop = cr.top + cr.height * 0.18;
+                    const plotBottom = cr.top + cr.height * 0.86;
+                    const plotWidth = plotRight - plotLeft;
+                    const plotHeight = plotBottom - plotTop;
+                    const groupWidth = plotWidth / count;
+                    // 分析人群的小柱可能很矮，纵向扫几个点；后续按图表类目白名单过滤，
+                    // 避免拿到相邻图表残留 tooltip。
+                    const yFractions = [0.30, 0.55, 0.78];
+                    for (let i = 0; i < count; i += 1) {
+                        const center = plotLeft + groupWidth * (i + 0.5);
+                        // 只取左侧蓝柱（分析人群）。tooltip 里会同时给出分析/对比占比，
+                        // 但我们最终只保留分析人群占比。
+                        const xs = [center - groupWidth * 0.12, center + groupWidth * 0.12];
+                        for (const x of xs) {
+                            for (const yf of yFractions) {
+                                bars.push({
+                                    cx: x,
+                                    cy: plotTop + plotHeight * yf,
+                                    h: 0,
+                                    synthetic: true,
+                                });
+                            }
+                        }
+                    }
+                }
+                const chartRect = bestChart.getBoundingClientRect();
+                blocks.push({
+                    title,
+                    chart_index: chartIndex,
+                    chart_kind: chartKind,
+                    expected_count: (categoryCounts && categoryCounts[title]) || null,
+                    chart_rect: {
+                        left: chartRect.left, top: chartRect.top,
+                        right: chartRect.right, bottom: chartRect.bottom,
+                        width: chartRect.width, height: chartRect.height,
+                    },
+                    bars
+                });
+            }
+            return blocks;
+        }"""
+    ) or []
+
+
+def _scan_tooltip_structured(page, mouse_x: float | None = None, mouse_y: float | None = None) -> dict[str, Any] | None:
+    r"""读当前可见的 echarts tooltip, 结构化返回 {category, 分析人群占比, 对比人群占比, raw}。
+
+    DMP 人群洞察 tooltip 文本结构 (echarts 默认 formatter):
+        {category_label}                       ← 第 1 行, x 轴类别
+        分析人群                                ← 第 2 行
+        占比: {analysis_pct}%                   ← 第 3 行
+        对比人群                                ← 第 4 行
+        占比: {comparison_pct}%                 ← 第 5 行
+
+    关键修复 (2026-06-26, Mavis):
+    - 旧版按行 parse, 把两条 "占比: X%" 都打上 name="占比", 丢掉了
+      「分析人群 vs 对比人群」的归属, 也丢了 category 名
+    - 新版先在全文里用带上下文锚点的正则匹配:
+        分析人群[\s\S]*?占比[：:]\s*(-?\d+(?:\.\d+)?)\s*%
+        对比人群[\s\S]*?占比[：:]\s*(-?\d+(?:\.\d+)?)\s*%
+    - category 从第一行非空, 且不是「分析人群/对比人群/占比:」之类的行
+
+    关键修复 (2026-06-27): 传鼠标坐标, 在多个 tooltip 候选里挑离鼠标最近的。
+    Canvax 偶发同时显示多张图的 tooltip, 不限定会拿到错图。
+    """
+    try:
+        result = page.evaluate(
+            r"""([mouseX, mouseY]) => {
+                const visibleRect = (el) => {
+                    const s = window.getComputedStyle(el);
+                    const r = el.getBoundingClientRect();
+                    if (s.visibility === 'hidden' || s.display === 'none') return null;
+                    if (r.width <= 0 || r.height <= 0 || r.width >= 700 || r.height >= 360) return null;
+                    if (mouseX != null && mouseY != null) {
+                        const cx = r.left + r.width / 2;
+                        const cy = r.top + r.height / 2;
+                        if (Math.hypot(cx - mouseX, cy - mouseY) > 700) return null;
+                    }
+                    return r;
+                };
+                const dist = (r) => {
+                    if (mouseX == null || mouseY == null) return 0;
+                    const cx = r.left + r.width / 2;
+                    const cy = r.top + r.height / 2;
+                    return Math.hypot(cx - mouseX, cy - mouseY);
+                };
+                const candidates = [];
+                // DMP/Canvax 的 tooltip class 经常变化，甚至不是 tooltip/popover 命名；
+                // 直接扫可见短文本节点更稳。
+                for (const el of document.querySelectorAll('body *')) {
+                    const rect = visibleRect(el);
+                    if (!rect) continue;
+                    const text = (el.innerText || el.textContent || '').replace(/　/g, ' ').trim();
+                    if (!text) continue;
+                    if (text.length > 500) continue;
+                    if (!text.includes('分析人群') || !text.includes('占比')) continue;
+                    candidates.push({text, d: dist(rect)});
+                }
+                if (!candidates.length) return null;
+                // 离鼠标最近者优先; 距离上限 350px 避免拿别处的 tooltip。
+                candidates.sort((a, b) => a.d - b.d);
+                if (mouseX != null && mouseY != null && candidates[0].d > 700) return null;
+                const text = candidates[0].text;
+
+                // 解析 category: 第一行, 排除 4 个 marker 行
+                const lines = text.split(/\n/).map(s => s.trim()).filter(Boolean);
+                let category = '';
+                for (const line of lines) {
+                    if (/^(分析人群|对比人群)\s*$/.test(line)) continue;
+                    if (/^占比[：:]/.test(line)) continue;
+                    if (/^-?\d+(\.\d+)?\s*%/.test(line)) continue;
+                    category = line;
+                    break;
+                }
+
+                // 上下文锚点正则: 找「分析人群」到下一个「占比:X%」之间的数字
+                const anaMatch = text.match(/分析人群[\s\S]*?占比[：:]\s*(-?\d+(?:\.\d+)?)\s*%/);
+                const cmpMatch = text.match(/对比人群[\s\S]*?占比[：:]\s*(-?\d+(?:\.\d+)?)\s*%/);
+
+                return {
+                    category: category,
+                    '分析人群占比': anaMatch ? parseFloat(anaMatch[1]) : null,
+                    '对比人群占比': cmpMatch ? parseFloat(cmpMatch[1]) : null,
+                    raw_text: text,
+                };
+            }""",
+            [mouse_x, mouse_y],
+        )
+        return result
+    except Exception:
+        return None
+
+
+CHART_EXPECTED_CATEGORIES: dict[str, set[str]] = {
+    "城市等级": {"一线城市", "准一线城市", "二线城市", "三线城市", "四线城市", "五线城市"},
+    "消费能力等级": {"购买力L1", "购买力L2", "购买力L3", "购买力L4", "购买力L5"},
+    "用户年龄": {"18-24岁", "25-29岁", "30-34岁", "35-39岁", "40-49岁", "50岁以上"},
+    "用户性别": {"男性用户", "女性用户"},
+}
+
+
+def analysis_only_chart_values(
+    chart_values: dict[str, list[dict[str, Any]]]
+) -> dict[str, list[dict[str, Any]]]:
+    """只保留分析人群占比，丢弃对比人群占比，便于直接写飞书字段。"""
+    cleaned: dict[str, list[dict[str, Any]]] = {}
+    for title, items in chart_values.items():
+        expected = CHART_EXPECTED_CATEGORIES.get(title)
+        cleaned_items: list[dict[str, Any]] = []
+        for item in items:
+            category = item.get("category")
+            value = item.get("分析人群占比")
+            if not category or value is None:
+                continue
+            if expected and category not in expected:
+                continue
+            cleaned_items.append({
+                "category": category,
+                "分析人群占比": value,
+            })
+        cleaned[title] = cleaned_items
+    return cleaned
+
+
+ANALYSIS_RATIO_KEY = "\u5206\u6790\u4eba\u7fa4\u5360\u6bd4"
+COMPARE_RATIO_KEY = "\u5bf9\u6bd4\u4eba\u7fa4\u5360\u6bd4"
+
+
+def chart_values_with_compare_fallback(
+    chart_values: dict[str, list[dict[str, Any]]],
+    prefer_compare: bool = False,
+) -> dict[str, list[dict[str, Any]]]:
+    """Prefer analysis-crowd ratios, but fall back to comparison-crowd ratios when analysis data is absent."""
+    cleaned: dict[str, list[dict[str, Any]]] = {}
+    for title, items in chart_values.items():
+        expected = CHART_EXPECTED_CATEGORIES.get(title)
+        cleaned_items: list[dict[str, Any]] = []
+        for item in items:
+            category = item.get("category")
+            analysis_value = item.get(ANALYSIS_RATIO_KEY)
+            compare_value = item.get(COMPARE_RATIO_KEY)
+            if not category:
+                continue
+            if expected and category not in expected:
+                continue
+
+            value = None
+            source = None
+            if prefer_compare and compare_value is not None:
+                value = compare_value
+                source = COMPARE_RATIO_KEY
+            elif analysis_value is not None:
+                value = analysis_value
+                source = ANALYSIS_RATIO_KEY
+            elif compare_value is not None:
+                value = compare_value
+                source = COMPARE_RATIO_KEY
+
+            if value is None:
+                continue
+            cleaned_items.append({
+                "category": category,
+                ANALYSIS_RATIO_KEY: value,
+                "value_source": source,
+            })
+        cleaned[title] = cleaned_items
+    return cleaned
+
+
+def has_analysis_crowd_too_small(page) -> bool:
+    """判断左侧分析人群是否显示“规模过小”，这种商品没有分析人群画像可采。"""
+    try:
+        return bool(page.evaluate(
+            r"""() => {
+                const panels = Array.from(document.querySelectorAll('*')).filter((el) => {
+                    const text = (el.innerText || el.textContent || '').trim();
+                    if (!text.includes('分析人群') || !text.includes('覆盖规模')) return false;
+                    const rect = el.getBoundingClientRect();
+                    return rect.width > 300 && rect.height > 120 && rect.top > 0;
+                });
+                return panels.some((el) => {
+                    const text = (el.innerText || el.textContent || '').replace(/\s+/g, '');
+                    return text.includes('规模过小') && text.includes('无法呈现相关数据');
+                });
+            }"""
+        ))
+    except Exception:
+        return False
+
+
+def extract_chart_values(page) -> tuple[dict[str, list[dict[str, Any]]], str]:
+    """对 4 张图逐柱 hover, 读 echarts tooltip 解析数值。
+
+    返回 ({title: [{category, 分析人群占比, 对比人群占比}, ...]}, method):
+        method ∈ {'hover', 'echarts', 'partial', 'empty'}
+        - 'hover':   4 张图全部 hover 拿到
+        - 'echarts': hover 全失败, 全部走 echarts.getOption fallback
+        - 'partial': 部分 hover 成功, 缺失的图用 echarts 补
+        - 'empty':   4 张图都没找到
+
+    每个 category 有 2 根柱子 (分析人群蓝 + 对比人群橙), tooltip 同时含两个占比。
+    同一 category 的两根柱子 hover 会得到完全相同的 tooltip, 按 category 去重。
+    """
+    page.evaluate(
+        r"""() => {
+            const title = Array.from(document.querySelectorAll('*')).find((el) => (el.textContent || '').trim() === '画像分布');
+            if (title) title.scrollIntoView({block: 'start'});
+        }"""
+    )
+    time.sleep(1.5)
+    wait_for_no_loading(page, timeout=20000)
+    wait_for_page_stable(page, idle_ms=800, timeout=15000)
+
+    blocks = find_chart_blocks(page)
+    if not blocks:
+        return {}, 'empty'
+
+    # 2026-06-26 fix: canvas renderer 时所有 block 都无柱子 (rects=[]),
+    # hover 路线无效, 直接走 echarts.getOption() 兜底.
+    if not any((b.get('bars') or []) for b in blocks):
+        page.mouse.move(10, 10)
+        time.sleep(0.3)
+        echarts_data = extract_chart_values_via_echarts(page)
+        if any(echarts_data.values()):
+            return echarts_data, 'echarts'
+        return {}, 'empty'
+
+    # 先把鼠标移开, 避免上一个 tooltip 残留
+    page.mouse.move(10, 10)
+    time.sleep(0.3)
+
+    result: dict[str, list[dict[str, Any]]] = {}
+    hover_ok = 0
+    for block in blocks:
+        title = block['title']
+        bars = block.get('bars') or []
+        # category -> {分析人群占比, 对比人群占比}
+        bucket: dict[str, dict[str, Any]] = {}
+        seen_categories: set[str] = set()
+        expected_count = int(block.get('expected_count') or 0)
+        expected_categories = CHART_EXPECTED_CATEGORIES.get(title)
+        for bar in bars:
+            try:
+                page.mouse.move(bar['cx'], bar['cy'])
+            except Exception:
+                continue
+            time.sleep(0.18 if bar.get('synthetic') else 0.3)
+            parsed = _scan_tooltip_structured(page, bar['cx'], bar['cy'])
+            if not parsed or not parsed.get('category'):
+                continue
+            cat = parsed['category']
+            if expected_categories and cat not in expected_categories:
+                continue
+            if cat in seen_categories:
+                # 同一 category 的两根柱子, tooltip 完全相同, 跳过
+                continue
+            seen_categories.add(cat)
+            ana = parsed.get('分析人群占比')
+            cmp = parsed.get('对比人群占比')
+            if ana is None and cmp is None:
+                continue
+            bucket[cat] = {
+                'category': cat,
+                '分析人群占比': ana,
+                '对比人群占比': cmp,
+            }
+            if expected_count and len(seen_categories) >= expected_count:
+                break
+        items = list(bucket.values())
+        if items:
+            hover_ok += 1
+        result[title] = items
+
+    page.mouse.move(10, 10)
+    time.sleep(0.1)
+
+    if hover_ok == len(blocks) and hover_ok > 0:
+        return result, 'hover'
+    if hover_ok == 0:
+        echarts_data = extract_chart_values_via_echarts(page)
+        if any(echarts_data.values()):
+            return echarts_data, 'echarts'
+        return result, 'empty'
+    # 部分 hover 成功: 缺失的图补 echarts
+    for title in list(result):
+        if not result[title]:
+            echarts_data = extract_chart_values_via_echarts(page)
+            if title in echarts_data and echarts_data[title]:
+                result[title] = echarts_data[title]
+    return result, 'partial'
+
+
+def extract_chart_values_via_echarts(page) -> dict[str, list[dict[str, Any]]]:
+    """通过 window.echarts.getInstanceByDom().getOption() 读 4 张图数据. hover 失败时的 fallback.
+
+    每张图 2 个 series:
+      - 分析人群占比  (蓝色)
+      - 对比人群占比  (橙色)
+    共用 xAxis[0].data 的类别名 (一线城市 / 准一线城市 / ... / 购买力L1-L4 / 18-24岁 / 男性 等).
+
+    修复 (2026-06-26, Mavis):
+    - 旧版只读 seriesList[0] (任意一个有数据的 series), 丢了对侧 series 的占比
+    - 新版按 series.name 区分「分析人群/对比人群」, 缺失某 series 时对应字段填 null
+    - 0-1 小数自动 ×100 转百分比, 保留 2 位小数
+    """
+    try:
+        data = page.evaluate(
+            r"""() => {
+                const result = {};
+                const wanted = ['城市等级', '消费能力等级', '用户年龄', '用户性别'];
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const textOf = (el) => (el.textContent || '').trim();
+                const toNumber = (v) => {
+                    if (typeof v === 'number') return v;
+                    if (v && typeof v === 'object') return Number(v.value || v[1] || 0);
+                    return Number(String(v || '').replace(/,/g, ''));
+                };
+                const findInstance = (dom) => {
+                    if (!dom) return null;
+                    if (window.echarts && typeof window.echarts.getInstanceByDom === 'function') {
+                        try {
+                            const inst = window.echarts.getInstanceByDom(dom);
+                            if (inst) return inst;
+                        } catch (e) {}
+                    }
+                    for (const key of Object.keys(dom)) {
+                        try {
+                            const val = dom[key];
+                            if (val && typeof val.getOption === 'function') return val;
+                        } catch (e) {}
+                    }
+                    return null;
+                };
+
+                for (const title of wanted) {
+                    const titleEl = Array.from(document.querySelectorAll('*'))
+                        .filter((el) => visible(el) && textOf(el) === title
+                            && el.getBoundingClientRect().height < 80)[0];
+                    if (!titleEl) continue;
+                    let chartDom = null;
+                    let cursor = titleEl;
+                    for (let i = 0; i < 12 && cursor; i += 1, cursor = cursor.parentElement) {
+                        const found = cursor.querySelector(':scope > svg, :scope > canvas, svg, canvas');
+                        if (found && visible(found)
+                            && found.getBoundingClientRect().width >= 100
+                            && found.getBoundingClientRect().height >= 80) {
+                            chartDom = found;
+                            break;
+                        }
+                    }
+                    if (!chartDom) continue;
+                    // canvas renderer 时 getInstanceByDom 需传 echarts container (带 _echarts_instance_ 的父 div),
+                    // svg renderer 直接传 svg 即可.
+                    const container = chartDom.tagName.toLowerCase() === 'canvas'
+                        ? (chartDom.closest('[_echarts_instance_]') || chartDom.parentElement || chartDom)
+                        : chartDom;
+                    const instance = findInstance(container);
+                    if (!instance) continue;
+                    try {
+                        const opt = instance.getOption();
+                        const seriesList = Array.isArray(opt.series) ? opt.series : [];
+                        if (!seriesList.length) continue;
+                        const xData = Array.isArray(opt.xAxis) && opt.xAxis[0] && Array.isArray(opt.xAxis[0].data)
+                            ? opt.xAxis[0].data : [];
+                        if (!xData.length) continue;
+
+                        // 找 分析人群 + 对比人群 两个 series
+                        const findSeries = (keyword) => seriesList.find((s) => {
+                            const name = String((s && s.name) || '');
+                            return name.includes(keyword);
+                        });
+                        const anaSeries = findSeries('分析人群') || findSeries('分析');
+                        const cmpSeries = findSeries('对比人群') || findSeries('对比');
+
+                        const getValues = (s) => {
+                            if (!s || !Array.isArray(s.data)) return [];
+                            return s.data.map((d) => toNumber(d && d.value !== undefined ? d.value : d));
+                        };
+                        const anaNums = getValues(anaSeries);
+                        const cmpNums = getValues(cmpSeries);
+                        // 任一 series 有数据就继续, 缺失的字段填 null
+                        if (!anaNums.length && !cmpNums.length) continue;
+
+                        // 0-1 小数 → 百分比 (保留 2 位)
+                        const toPct = (n) => {
+                            if (!Number.isFinite(n)) return null;
+                            if (n >= 0 && n <= 1) return Math.round(n * 10000) / 100;
+                            return Math.round(n * 100) / 100;
+                        };
+
+                        const items = xData.map((cat, idx) => ({
+                            category: String(cat || '').trim(),
+                            '分析人群占比': toPct(anaNums[idx]),
+                            '对比人群占比': toPct(cmpNums[idx]),
+                        })).filter((it) => it.category);
+                        if (items.length) result[title] = items;
+                    } catch (e) {}
+                }
+                return result;
+            }"""
+        )
+        return data or {}
+    except Exception as exc:
+        print(f"  [WARN] echarts fallback failed: {exc}")
+        return {}
+
+
+def dmp_app_ready(page) -> bool:
+    try:
+        return bool(
+            page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const text = document.body.innerText || '';
+                    const hasLoginInput = Array.from(document.querySelectorAll('#fm-login-id, input[name="fm-login-id"], #fm-login-password, input[name="fm-login-password"]'))
+                        .some(visible);
+                    if (hasLoginInput) return false;
+                    return text.includes('切换分析单品')
+                        || text.includes('单品信息')
+                        || text.includes('宝贝洞察')
+                        || text.includes('货品洞察')
+                        || text.includes('单品人群洞察');
+                }"""
+            )
+        )
+    except Exception:
+        return False
+
+
+def is_logged_in(page) -> bool:
+    if (page.url or "").lower().find("login") >= 0 and login_form_visible(page):
+        return False
+    if dmp_app_ready(page):
+        return True
+    body = page_text(page)
+    # 达摩盘登录后有时会停在官网落地页，URL 仍是 login.html，但页面会出现
+    # “进入后台”账号卡片。这也是有效登录态，后续需要先点进后台再进洞察页。
+    if "进入后台" in body and ("光影行星" in body or "退出账户" in body):
+        return True
+    login_markers = ["账号名", "请输入登录密码", "登录名", "登录密码", "扫码登录"]
+    app_markers = ["切换分析单品", "单品信息", "宝贝洞察", "货品洞察", "单品人群洞察"]
+    return any(marker in body for marker in app_markers) and not any(marker in body for marker in login_markers)
+
+
+def enter_dmp_backend_if_needed(page) -> bool:
+    body = page_text(page)
+    if "进入后台" not in body:
+        return False
+    print("[LOGIN] 已登录但停在达摩盘入口页，点击进入后台...")
+    clicked = js_click_text(page, "进入后台", timeout=8000)
+    if not clicked:
+        try:
+            clicked = bool(
+                page.evaluate(
+                    r"""() => {
+                        const visible = (el) => {
+                            if (!el) return false;
+                            const style = window.getComputedStyle(el);
+                            const rect = el.getBoundingClientRect();
+                            return style.display !== 'none'
+                                && style.visibility !== 'hidden'
+                                && rect.width > 0
+                                && rect.height > 0;
+                        };
+                        const el = Array.from(document.querySelectorAll('button,a,div,span'))
+                            .find((node) => visible(node) && (node.innerText || '').trim() === '进入后台');
+                        if (!el) return false;
+                        el.click();
+                        return true;
+                    }"""
+                )
+            )
+        except Exception:
+            clicked = False
+    if clicked:
+        time.sleep(3)
+        close_known_popups(page)
+    else:
+        debug_capture(page, "enter_backend_button_not_found")
+    return clicked
+
+
+def wait_for_app_or_login(page, timeout: int = 20000) -> str:
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        if dmp_app_ready(page):
+            return "app"
+        if "进入后台" in page_text(page):
+            return "backend"
+        if login_form_visible(page):
+            return "login"
+        time.sleep(0.5)
+    return "timeout"
+
+
+def ensure_logged_in(page, account: str, password: str) -> None:
+    print("[LOGIN] 检查达摩盘登录态...")
+    try:
+        safe_goto(page, ITEM_INSIGHT_URL, timeout=45000, attempts=2)
+        state = wait_for_app_or_login(page, timeout=18000)
+        close_known_popups(page)
+        debug_capture(page, "login_check_item_insight")
+        if state == "backend" or enter_dmp_backend_if_needed(page):
+            safe_goto(page, ITEM_INSIGHT_URL, timeout=45000, attempts=2)
+            state = wait_for_app_or_login(page, timeout=18000)
+            close_known_popups(page)
+        if state == "app" or is_logged_in(page):
+            print(f"[OK] 已复用登录态: {page.url}")
+            try:
+                save_cookies_to_file(page.context)
+            except Exception:
+                pass
+            return
+    except Exception as exc:
+        print(f"  [WARN] 登录态检查失败: {exc}")
+
+    if not do_login(page, account, password):
+        raise RuntimeError("达摩盘登录失败或等待登录超时")
+    enter_dmp_backend_if_needed(page)
+    try:
+        safe_goto(page, ITEM_INSIGHT_URL, timeout=45000, attempts=2)
+    except Exception:
+        pass
+    # 登录成功后持久化 cookie, 下次 StealthySession 可直接注入
+    try:
+        save_cookies_to_file(page.context)
+    except Exception as exc:
+        print(f"  [WARN] 登录成功后保存 cookie 失败: {exc}")
+
+
+def goto_item_insight(page) -> None:
+    enter_dmp_backend_if_needed(page)
+    if not dmp_app_ready(page) or "item-insight" not in (page.url or ""):
+        safe_goto(page, ITEM_INSIGHT_URL, timeout=45000, attempts=2)
+        enter_dmp_backend_if_needed(page)
+        if "item-insight" not in (page.url or ""):
+            safe_goto(page, ITEM_INSIGHT_URL, timeout=45000, attempts=2)
+    for _ in range(3):
+        if not close_known_popups(page):
+            break
+    if not wait_for_text(page, "切换分析单品", timeout=20000):
+        page.evaluate("(url) => { window.location.href = url; }", ITEM_INSIGHT_URL)
+        if not wait_for_text(page, "切换分析单品", timeout=25000):
+            debug_capture(page, "item_insight_not_ready")
+            raise RuntimeError("没有等到“切换分析单品”，单品洞察页未加载完成")
+    debug_capture(page, "item_insight_loaded")
+
+
+def click_crowd_tab(page) -> None:
+    body = page_text(page)
+    if "单品人群画像" not in body:
+        if not js_click_text(page, "单品人群洞察", timeout=15000):
+            raise RuntimeError("没有找到“单品人群洞察”标签")
+    wait_for_crowd_config_panels(page, timeout=30000)
+    debug_capture(page, "crowd_tab_loaded")
+
+
+def wait_for_crowd_charts_ready(page, timeout: int = 25000) -> None:
+    """等到 4 个人群画像图表 (城市/消费/年龄/性别) 都已渲染有数据再返回。
+
+    不仅要求 svg/canvas 节点存在, 还要求每个图表有可见的非空渲染
+    (svg rect 数量 ≥3 或 canvas 有非白像素), 否则截图会截到空白图表。
+    """
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        try:
+            ready = page.evaluate(
+                r"""() => {
+                    const text = document.body.innerText || '';
+                    const markers = ['画像分布', '城市等级', '消费能力等级', '用户年龄', '用户性别'];
+                    if (!markers.every((m) => text.includes(m))) return {ok: false, reason: 'text_missing'};
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none' && style.visibility !== 'hidden'
+                            && rect.width > 0 && rect.height > 0;
+                    };
+                    const textOf = (el) => (el.textContent || '').trim();
+                    const wanted = ['城市等级', '消费能力等级', '用户年龄', '用户性别'];
+                    const titles = [];
+                    for (const t of wanted) {
+                        const found = Array.from(document.querySelectorAll('*'))
+                            .find((el) => visible(el) && textOf(el) === t
+                                && el.getBoundingClientRect().height < 80);
+                        if (!found) return {ok: false, reason: `title_missing:${t}`};
+                        titles.push(found);
+                    }
+                    // 每个 title 必须能在祖先链找到 svg/canvas 且有真实数据
+                    const isRendered = (title) => {
+                        let node = title;
+                        for (let i = 0; i < 10 && node; i += 1, node = node.parentElement) {
+                            const svg = node.querySelector(':scope > svg, svg');
+                            const canvas = node.querySelector(':scope > canvas, canvas');
+                            const el = (svg && visible(svg)) ? svg : (canvas && visible(canvas) ? canvas : null);
+                            if (!el) continue;
+                            if (svg) {
+                                const rects = Array.from(svg.querySelectorAll('rect'))
+                                    .filter((r) => {
+                                        const h = parseFloat(r.getAttribute('height') || 0);
+                                        const fill = (r.getAttribute('fill') || '').toLowerCase();
+                                        return h > 5 && fill
+                                            && !['none', 'transparent', '#fff', '#ffffff', 'white'].includes(fill);
+                                    });
+                                if (rects.length >= 3) return true;
+                            }
+                            if (canvas) {
+                                try {
+                                    const ctx = canvas.getContext('2d');
+                                    const data = ctx.getImageData(0, 0, canvas.width, canvas.height).data;
+                                    for (let i = 0; i < data.length; i += 32) {
+                                        const a = data[i+3];
+                                        if (a > 10) {
+                                            const r = data[i], g = data[i+1], b = data[i+2];
+                                            if (r < 250 || g < 250 || b < 250) return true;
+                                        }
+                                    }
+                                } catch (e) {}
+                            }
+                        }
+                        return false;
+                    };
+                    for (const t of titles) {
+                        if (!isRendered(t)) return {ok: false, reason: `chart_empty:${textOf(t)}`};
+                    }
+                    return {ok: true};
+                }"""
+            )
+            if isinstance(ready, dict) and ready.get("ok"):
+                return
+        except Exception:
+            pass
+        time.sleep(0.6)
+    debug_capture(page, "crowd_charts_wait_timeout")
+
+
+def crowd_panels_ready(page) -> bool:
+    try:
+        return bool(
+            page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const text = document.body.innerText || '';
+                    if (!text.includes('分析人群') || !text.includes('对比人群')) return false;
+                    if (!text.includes('人群属性') || !text.includes('行为统计周期')) return false;
+                    const panels = Array.from(document.querySelectorAll('div, section')).filter((el) => {
+                        if (!visible(el)) return false;
+                        const t = el.innerText || '';
+                        const r = el.getBoundingClientRect();
+                        return t.includes('人群类型')
+                            && t.includes('人群属性')
+                            && t.includes('行为统计周期')
+                            && r.width > 350
+                            && r.height > 120;
+                    });
+                    return panels.length >= 2;
+                }"""
+            )
+        )
+    except Exception:
+        return False
+
+
+def wait_for_crowd_config_panels(page, timeout: int = 45000) -> None:
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        if crowd_panels_ready(page):
+            return
+        time.sleep(0.6)
+    debug_capture(page, "crowd_config_wait_timeout")
+    raise RuntimeError("单品人群洞察配置面板加载超时：没有等到“分析人群/对比人群”")
+
+
+def wait_for_no_loading(page, timeout: int = 30000) -> bool:
+    """等待页面中 loading/spin 类元素消失。"""
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        try:
+            loading = page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    return Array.from(document.querySelectorAll('*')).some((el) => {
+                        if (!visible(el)) return false;
+                        const cls = String(el.className || '');
+                        return /loading|spin|skeleton/i.test(cls);
+                    });
+                }"""
+            )
+            if not loading:
+                return True
+        except Exception:
+            pass
+        time.sleep(0.4)
+    return False
+
+
+def wait_for_page_stable(page, idle_ms: int = 1200, timeout: int = 20000) -> bool:
+    """等待页面 DOM 在 idle_ms 毫秒内没有明显变化（数据刷新稳定）。"""
+    try:
+        return bool(
+            page.evaluate(
+                r"""({idleMs, timeoutMs}) => {
+                    return new Promise((resolve) => {
+                        let timer = null;
+                        let lastMutation = Date.now();
+                        let settled = false;
+                        const observer = new MutationObserver(() => {
+                            if (settled) return;
+                            lastMutation = Date.now();
+                            if (timer) clearTimeout(timer);
+                            timer = setTimeout(() => {
+                                settled = true;
+                                observer.disconnect();
+                                resolve(true);
+                            }, idleMs);
+                        });
+                        timer = setTimeout(() => {
+                            settled = true;
+                            observer.disconnect();
+                            resolve(true);
+                        }, idleMs);
+                        observer.observe(document.body, {childList: true, subtree: true, attributes: true});
+                        setTimeout(() => {
+                            if (settled) return;
+                            settled = true;
+                            observer.disconnect();
+                            resolve(false);
+                        }, timeoutMs);
+                    });
+                }""",
+                {"idleMs": idle_ms, "timeoutMs": timeout},
+            )
+        )
+    except Exception:
+        return False
+
+
+def wait_for_panel_period(page, panel_label: str, period: str = "30天", timeout: int = 15000) -> bool:
+    """等待指定人群面板的行为统计周期显示为指定值。"""
+    deadline = time.time() + timeout / 1000
+    while time.time() < deadline:
+        text = get_crowd_panel_text(page, panel_label)
+        if period in text:
+            return True
+        time.sleep(0.3)
+    return False
+
+
+def get_crowd_panel_text(page, panel_label: str) -> str:
+    """读取指定人群面板的完整文本，用于校验配置是否生效。"""
+    try:
+        return page.evaluate(
+            r"""(panelLabel) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const textOf = (el) => (el.textContent || '').trim();
+                const exact = (text) => Array.from(document.querySelectorAll('*'))
+                    .filter((el) => visible(el) && textOf(el) === text);
+                const panelOf = (label) => {
+                    const panels = [];
+                    for (const labelNode of exact(label)) {
+                        let node = labelNode;
+                        for (let i = 0; i < 10 && node; i += 1, node = node.parentElement) {
+                            const text = node.textContent || '';
+                            const r = node.getBoundingClientRect();
+                            if (text.includes('人群类型')
+                                && text.includes('人群属性')
+                                && text.includes('行为统计周期')
+                                && r.width > 350
+                                && r.height > 120) {
+                                panels.push(node);
+                            }
+                        }
+                    }
+                    panels.sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                    return panels[0] || null;
+                };
+                const panel = panelOf(panelLabel);
+                return panel ? (panel.textContent || '') : '';
+            }""",
+            panel_label,
+        ) or ""
+    except Exception:
+        return ""
+
+
+def validate_crowd_configuration(page, retries: int = 1) -> None:
+    """严格校验人群配置:
+
+    - 分析人群: 必须**只**含 收藏加购 + 成交 (各恰好 1 个, 无其他 attr 残留), 周期 30天
+    - 对比人群: 必须**只**含 商详浏览 (恰好 1 个), 周期 30天
+
+    旧版只检查 substrings 存在, 不阻止「收藏加购 + 收藏加购 + 成交 + 浏览」这种残留。
+    """
+    expected_attrs = {
+        "分析人群": {"收藏加购": 1, "成交": 1},
+        "对比人群": {"商详浏览": 1},
+    }
+    expected_period = "30天"
+    for attempt in range(retries + 1):
+        bad: dict[str, dict[str, Any]] = {}
+        for label, attrs in expected_attrs.items():
+            text = get_crowd_panel_text(page, label)
+            wrong_count = {
+                k: {"got": text.count(k), "want": v}
+                for k, v in attrs.items() if text.count(k) != v
+            }
+            period_ok = expected_period in text
+            entry: dict[str, Any] = {}
+            if wrong_count:
+                entry["attrs"] = wrong_count
+            if not period_ok:
+                entry["period"] = f"missing '{expected_period}' (got snippet: {text[:120]!r})"
+            if entry:
+                bad[label] = entry
+        if not bad:
+            print("  [CROWD] 人群配置严格校验通过")
+            return
+        if attempt < retries:
+            print(f"  [CROWD] 配置校验失败, 重试 (issue={bad})")
+            for label in ("分析人群", "对比人群"):
+                clear_panel_attributes(page, label)
+            add_panel_attribute(page, "分析人群", "收藏加购")
+            add_panel_attribute(page, "分析人群", "成交")
+            add_panel_attribute(page, "对比人群", "商详浏览")
+            for label in ("分析人群", "对比人群"):
+                set_panel_period_30(page, label)
+            wait_for_crowd_charts_ready(page, timeout=15000)
+            time.sleep(2)
+        else:
+            raise RuntimeError(f"人群配置未生效: {bad}")
+
+
+def configure_crowd_groups(page) -> None:
+    wait_for_crowd_config_panels(page, timeout=30000)
+    debug_capture(page, "before_configure_crowd")
+
+    # 清理旧标签, 等面板稳定
+    for label in ("分析人群", "对比人群"):
+        clear_panel_attributes(page, label)
+    wait_for_page_stable(page, idle_ms=800, timeout=10000)
+
+    # 逐个添加属性, 每次等待标签出现在面板中
+    add_panel_attribute(page, "分析人群", "收藏加购")
+    wait_for_page_stable(page, idle_ms=600, timeout=8000)
+    add_panel_attribute(page, "分析人群", "成交")
+    wait_for_page_stable(page, idle_ms=600, timeout=8000)
+    add_panel_attribute(page, "对比人群", "商详浏览")
+    wait_for_page_stable(page, idle_ms=600, timeout=8000)
+
+    # 设置周期为 30天, 等待页面刷新稳定
+    for label in ("分析人群", "对比人群"):
+        set_panel_period_30(page, label)
+        wait_for_panel_period(page, label, period="30天", timeout=10000)
+    wait_for_no_loading(page, timeout=20000)
+    wait_for_page_stable(page, idle_ms=1500, timeout=20000)
+
+    wait_for_crowd_charts_ready(page, timeout=25000)
+    time.sleep(2)
+    debug_capture(page, "after_configure_crowd")
+    validate_crowd_configuration(page, retries=1)
+
+
+def ensure_crowd_groups_configured(page, skip_configure: bool = False) -> None:
+    """确保人群配置正确; 如已正确且 skip_configure 为 True 则跳过重新配置。"""
+    try:
+        validate_crowd_configuration(page, retries=0)
+        print("  [CROWD] 当前人群配置已正确, 跳过重新配置")
+        return
+    except RuntimeError:
+        if skip_configure:
+            print("  [CROWD] 切换单品后配置发生变化, 重新配置")
+        configure_crowd_groups(page)
+
+
+def collect_item(page, style_name: str, item_id: str, output_dir: Path,
+                 skip_configure: bool = False) -> dict[str, Any]:
+    style_dir = output_dir / today_label() / safe_filename(style_name)
+    item_dir = style_dir
+    json_path = item_dir / f"{safe_filename(style_name)}_{item_id}_chart_values.json"
+    result: dict[str, Any] = {
+        "style_name": style_name,
+        "item_id": item_id,
+        "status": "pending",
+        "values_path": str(json_path),
+        "collected_at": datetime.now().isoformat(timespec="seconds"),
+    }
+    try:
+        switch_analysis_item(page, item_id)
+        click_crowd_tab(page)
+        # 每个商品都校验配置, 只有配置正确且明确允许跳过时才不重新配置
+        ensure_crowd_groups_configured(page, skip_configure=skip_configure)
+        analysis_too_small = has_analysis_crowd_too_small(page)
+        if False and analysis_too_small:
+            result["status"] = "skipped"
+            result["skip_reason"] = "分析人群规模过小，无法呈现相关数据"
+            result["extraction_method"] = "skipped_no_analysis_data"
+            print(f"  [SKIP] {style_name} / {item_id}: {result['skip_reason']}", flush=True)
+            return result
+        chart_values, method = extract_chart_values(page)
+        chart_values = chart_values_with_compare_fallback(
+            chart_values,
+            prefer_compare=analysis_too_small,
+        )
+        if not chart_values or not any(chart_values.values()):
+            if has_analysis_crowd_too_small(page):
+                result["status"] = "skipped"
+                result["skip_reason"] = "分析人群规模过小，无法呈现相关数据"
+                result["extraction_method"] = method
+                print(f"  [SKIP] {style_name} / {item_id}: {result['skip_reason']}", flush=True)
+                return result
+            result["status"] = "skipped"
+            result["skip_reason"] = f"未提取到分析人群图表数值 (method={method})"
+            result["extraction_method"] = method
+            print(f"  [SKIP] {style_name} / {item_id}: {result['skip_reason']}", flush=True)
+            return result
+        result["status"] = "ok"
+        result["chart_values"] = chart_values
+        result["extraction_method"] = method
+        result["fallback_to_compare"] = analysis_too_small
+        # 写 per-item JSON, 跟旧版 _crowd_insight.png 命名对齐
+        item_dir.mkdir(parents=True, exist_ok=True)
+        json_path.write_text(json.dumps({
+            "style_name": style_name,
+            "item_id": item_id,
+            "collected_at": result["collected_at"],
+            "extraction_method": method,
+            "fallback_to_compare": analysis_too_small,
+            "chart_values": chart_values,
+        }, ensure_ascii=False, indent=2), encoding="utf-8")
+        total_bars = sum(len(v) for v in chart_values.values())
+        print(f"  [OK] 数值: {json_path} ({len(chart_values)} 图, {total_bars} 柱子, method={method})")
+    except Exception as exc:
+        result["status"] = "failed"
+        result["error"] = str(exc)
+        # 失败 debug 截图仅 DMP_DEBUG=1 时存, 排障用
+        if os.getenv("DMP_DEBUG") == "1":
+            debug_dir = output_dir / today_label() / "_debug"
+            debug_path = debug_dir / f"{safe_filename(style_name)}_{item_id}_debug.png"
+            try:
+                debug_dir.mkdir(parents=True, exist_ok=True)
+                page.screenshot(path=str(debug_path), full_page=False)
+                result["debug_screenshot_path"] = str(debug_path)
+            except Exception:
+                pass
+        print(f"  [FAIL] {style_name} / {item_id}: {exc}")
+        # 关掉任何残留的抽屉, 避免下一个 item 进来叠加
+        try:
+            close_existing_switch_drawer(page)
+        except Exception:
+            pass
+    return result
+
+
+def is_recoverable_navigation_failure(record: dict[str, Any]) -> bool:
+    """Whether a failed item indicates a poisoned/stale DMP page session."""
+    if record.get("status") != "failed":
+        return False
+    error = str(record.get("error") or "")
+    markers = (
+        "切换单品弹窗",
+        "切换分析单品",
+        "搜索框",
+        "单品洞察页未加载",
+    )
+    return any(marker in error for marker in markers)
+
+
+def collect_item_with_recovery(
+    session: Any,
+    page: Any,
+    style_name: str,
+    item_id: str,
+    output_dir: Path,
+    *,
+    skip_configure: bool = False,
+    collect_fn: Callable[..., dict[str, Any]] = collect_item,
+    goto_fn: Callable[[Any], None] = goto_item_insight,
+    reopen_session: Callable[[Any, Any], tuple[Any, Any]],
+) -> tuple[Any, Any, dict[str, Any]]:
+    """Retry a poisoned DMP item page, then rebuild the browser session once."""
+    record = collect_fn(
+        page,
+        style_name,
+        item_id,
+        output_dir,
+        skip_configure=skip_configure,
+    )
+    if not is_recoverable_navigation_failure(record):
+        return session, page, record
+
+    print(f"  [RECOVER] {style_name}/{item_id}: 重新进入单品洞察页后重试")
+    try:
+        goto_fn(page)
+        record = collect_fn(
+            page,
+            style_name,
+            item_id,
+            output_dir,
+            skip_configure=False,
+        )
+    except Exception as exc:
+        record = {"status": "failed", "error": f"页面恢复失败: {exc}"}
+    if not is_recoverable_navigation_failure(record) and record.get("status") != "failed":
+        return session, page, record
+
+    print(f"  [RECOVER] {style_name}/{item_id}: 重建浏览器会话后最后重试")
+    session, page = reopen_session(session, page)
+    record = collect_fn(
+        page,
+        style_name,
+        item_id,
+        output_dir,
+        skip_configure=False,
+    )
+    return session, page, record
+
+
+def close_dmp_session(session: Any, page: Any) -> None:
+    if not session:
+        return
+    try:
+        if hasattr(session, "context") and session.context:
+            try:
+                page.close()
+            except Exception:
+                pass
+        session.close()
+    except Exception:
+        pass
+
+
+def write_outputs(output_dir: Path, records: list[dict[str, Any]]) -> None:
+    output_dir.mkdir(parents=True, exist_ok=True)
+    json_path = output_dir / f"dmp_crowd_insight_index_{today_label()}.json"
+    json_path.write_text(json.dumps(records, ensure_ascii=False, indent=2), encoding="utf-8")
+    print(f"[OK] 索引 JSON: {json_path}")
+
+
+def select_styles(names: str | None) -> dict[str, list[str]]:
+    """从飞书配置动态读取可采集款式 (天猫商品ID + 飞书表地址同时存在)。
+
+    --styles 传款式名时, 只从可采集款式里筛选指定款; 未传时全部采集。
+    """
+    collectable = load_collectable_styles()
+    if not names:
+        return collectable
+    selected: dict[str, list[str]] = {}
+    wanted = [name.strip() for name in names.split(",") if name.strip()]
+    for name in wanted:
+        if name not in collectable:
+            raise ValueError(
+                f"未知或不可采集款式: {name}; "
+                f"可采集款式需同时有天猫商品ID和飞书表地址, 当前可采: {', '.join(sorted(collectable))}"
+            )
+        selected[name] = collectable[name]
+    return selected
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="采集淘宝达摩盘单品人群洞察四图数值 (hover 取 echarts tooltip, 输出 JSON)")
+    parser.add_argument("--styles", help="只采集指定款式，多个用英文逗号分隔，例如：星云2,极星双肩")
+    parser.add_argument("--max-items", type=int, default=0, help="每个款式最多采集几个款式 ID；0 表示全部")
+    parser.add_argument("--account", default=ACCOUNT, help="淘宝/达摩盘账号")
+    parser.add_argument("--password", default=PASSWORD, help="淘宝/达摩盘密码")
+    parser.add_argument("--user-data-dir", default=USER_DATA_DIR, help="浏览器 profile 目录，用于复用登录态")
+    parser.add_argument("--output", default=str(OUTPUT_DIR), help="输出目录")
+    parser.add_argument("--headless", action="store_true", help="无头运行；首次调试建议不要开启")
+    parser.add_argument("--keep-open", action="store_true", help="结束后不关闭浏览器")
+    parser.add_argument("--debug", action="store_true", help="保存每个关键步骤的截图和 HTML，排障时使用")
+    parser.add_argument("--refresh-config", action="store_true",
+                        help="忽略 24h 缓存, 强制从飞书多维表格重读款式配置")
+    args = parser.parse_args()
+
+    if args.debug:
+        os.environ["DMP_DEBUG"] = "1"
+
+    if args.refresh_config:
+        from config.style_config_loader import StyleConfigLoader
+        StyleConfigLoader().load(force_refresh=True, sync_fallback=True)
+        print("[INFO] --refresh-config: 已从飞书多维表格重读款式配置 + 同步 bitable_style_map.json")
+
+    output_dir = managed_data_path(args.output)
+    styles = select_styles(args.styles)
+
+    session: Any = None
+    records: list[dict[str, Any]] = []
+    try:
+        session, page = start_stealthy_session(args.headless, args.user_data_dir)
+        ensure_logged_in(page, args.account, args.password)
+        goto_item_insight(page)
+
+        def reopen_session(old_session: Any, old_page: Any) -> tuple[Any, Any]:
+            close_dmp_session(old_session, old_page)
+            new_session, new_page = start_stealthy_session(args.headless, args.user_data_dir)
+            ensure_logged_in(new_page, args.account, args.password)
+            goto_item_insight(new_page)
+            return new_session, new_page
+
+        for style_name, item_ids in styles.items():
+            ids = item_ids[: args.max_items] if args.max_items and args.max_items > 0 else item_ids
+            for idx, item_id in enumerate(ids):
+                print(f"[RUN] {style_name} / {item_id} ({idx+1}/{len(ids)})")
+                # 同款第一个商品ID配置人群, 后续跳过配置 (切单品后图表自动刷新)
+                session, page, record = collect_item_with_recovery(
+                    session,
+                    page,
+                    style_name,
+                    item_id,
+                    output_dir,
+                    skip_configure=idx > 0,
+                    reopen_session=reopen_session,
+                )
+                records.append(record)
+                write_outputs(output_dir, records)
+                if record.get("status") == "ok":
+                    print(f"[OK] {style_name} 已采集成功，跳过该款剩余商品ID")
+                    break
+                # 间隔: 等上一个完全结束 (图表渲染 + 文件落盘) 再开下一个, 避免 DMP 抖动
+                time.sleep(2)
+
+        return 0 if all(record.get("status") in {"ok", "skipped"} for record in records) else 2
+    finally:
+        write_outputs(output_dir, records)
+        if session and not args.keep_open:
+            close_dmp_session(session, page)
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_collect.py b/src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_collect.py
new file mode 100644
index 0000000..06d9994
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_collect.py
@@ -0,0 +1,160 @@
+# -*- coding: utf-8 -*-
+"""
+淘宝生意参谋商品采集 — 按品牌分组循环采集。
+
+不同品牌对应不同天猫店铺账号：
+- 光影行星：使用默认账号（环境变量 SYCM_ACCOUNT / SYCM_PASSWORD 或光影行星旗舰店:龙虾仔）
+- ozko：使用 ozko旗舰店:龙虾 / aa123456
+
+每次循环为单个品牌启动独立 Chrome 会话（复用不同 profile 目录，避免 cookie 冲突），
+调用 taobao_sycm_products.py 完成登录 → 商品页 → 选日期 → 下载 → 解析导出。
+
+用法:
+    .venv\\Scripts\\python.exe taobao_sycm_collect.py
+    .venv\\Scripts\\python.exe taobao_sycm_collect.py --target-date 2026-06-11
+    .venv\\Scripts\\python.exe taobao_sycm_collect.py --user-data-dir <GYXX_DATA_ROOT>\\profiles\\product_commerce\\tm
+    .venv\\Scripts\\python.exe taobao_sycm_collect.py --headless
+
+浏览器端口和 Profile 由 GYXX 按脚本独立分配。
+"""
+
+from __future__ import annotations
+
+import argparse
+import os
+import subprocess
+import sys
+import time
+from collections import defaultdict
+from pathlib import Path
+
+from gyxx_flow.adapters import environment_for_child_script
+from runtime_paths import runtime_script
+
+from config.style_config_loader import StyleConfigLoader, get_daily_styles, get_tm_brand_map, get_tm_style_id_map
+from taobao_sycm_products import USER_DATA_DIR, _kill_port_listeners, get_yesterday, parse_target_date
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+
+DEFAULT_ACCOUNT = os.getenv("SYCM_ACCOUNT", "光影行星旗舰店:龙虾仔")
+DEFAULT_PASSWORD = os.getenv("SYCM_PASSWORD", "aa123456")
+
+# 品牌 -> (账号, 密码)
+BRAND_ACCOUNT_MAP = {
+    "光影行星": (DEFAULT_ACCOUNT, DEFAULT_PASSWORD),
+    "ozko": ("ozko旗舰店:龙虾", "aa123456"),
+}
+
+
+def _brand_user_data_dir(base_dir: str, brand: str) -> str:
+    """为不同品牌生成独立的 profile 目录，避免 cookie/登录态冲突。"""
+    if brand == "光影行星":
+        return base_dir
+    base = Path(base_dir)
+    return str(base.parent / f"{base.name}_{brand.lower()}")
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="淘宝生意参谋商品采集 (按品牌循环)")
+    parser.add_argument("--target-date", type=parse_target_date, default=None,
+                        help="采集日期 YYYY-MM-DD, 默认昨天")
+    parser.add_argument("--user-data-dir", default=USER_DATA_DIR,
+                        help=f"Chrome profile 根目录, 默认 {USER_DATA_DIR}")
+    parser.add_argument("--headless", action="store_true", help="无头模式 (不显示浏览器窗口)")
+    parser.add_argument("--refresh-config", action="store_true",
+                        help="忽略 24h 缓存, 强制从飞书多维表格重读款式配置")
+    args = parser.parse_args()
+
+    if args.refresh_config:
+        StyleConfigLoader().load(force_refresh=True)
+        print("[INFO] --refresh-config: 已从飞书多维表格重读款式配置")
+
+    target = args.target_date or get_yesterday()
+    target_str = target.strftime("%Y-%m-%d")
+    print(f"[INFO] 目标日期: {target_str}")
+    print(f"[INFO] Profile 根目录: {args.user_data_dir}")
+    print(f"[INFO] Headless: {args.headless}")
+
+    # 按品牌分组天猫款式
+    tm_style_id_map = get_tm_style_id_map()
+    tm_brand_map = get_tm_brand_map()
+    daily = get_daily_styles()
+    tm_collectable = daily["tm"]["collectable"]
+    tm_skipped = daily["tm"]["skipped"]
+    print(f"[TM] 飞书表格 待采 {len(tm_collectable)} 款, 跳过 {len(tm_skipped)} 款 (款式级缺 erp/sales,或 TM 缺商品链接ID)")
+    for s in tm_skipped:
+        print(f"  - {s['style']}: {s['reason']}")
+    brands: dict[str, list[str]] = defaultdict(list)
+    for style_name in tm_style_id_map:
+        brand = tm_brand_map.get(style_name, "光影行星") or "光影行星"
+        brands[brand].append(style_name)
+
+    if not brands:
+        print("[WARN] 没有需要采集的天猫款式")
+        return 0
+
+    print(f"[INFO] 共 {len(tm_style_id_map)} 个天猫款式, 分 {len(brands)} 个品牌: {sorted(brands.keys())}")
+
+    failed_brands = []
+    product_script = runtime_script("taobao_sycm_products.py")
+    child_environment = environment_for_child_script(product_script, os.environ)
+    child_profile_root = child_environment["GYXX_BROWSER_PROFILE_DIR"]
+    child_cdp_port = int(child_environment["GYXX_BROWSER_CDP_PORT"])
+    for brand in sorted(brands.keys()):
+        style_names = brands[brand]
+        account, login_secret = BRAND_ACCOUNT_MAP.get(
+            brand, (DEFAULT_ACCOUNT, DEFAULT_PASSWORD)
+        )
+        brand_user_data_dir = _brand_user_data_dir(child_profile_root, brand)
+
+        print(f"\n{'=' * 60}")
+        print(f"[BRAND] {brand} ({len(style_names)} 款)")
+        print(f"[ACCOUNT] {account}")
+        print(f"[PROFILE] {brand_user_data_dir}")
+        print(f"{'=' * 60}")
+
+        cmd = [
+            sys.executable,
+            str(product_script),
+            "--target-date", target_str,
+            "--user-data-dir", brand_user_data_dir,
+            "--account", account,
+            "--password", login_secret,
+            "--brand", brand,
+        ]
+        if args.headless:
+            cmd.append("--headless")
+
+        proc = subprocess.run(
+            cmd,
+            cwd=PROJECT_ROOT,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+            env=child_environment,
+        )
+        if proc.returncode != 0:
+            print(f"[FAIL] 品牌 {brand} 采集失败 (exit={proc.returncode})")
+            failed_brands.append(brand)
+        else:
+            print(f"[OK] 品牌 {brand} 采集完成")
+
+        # 每个品牌跑完后只清理该子脚本的专属端口。
+        # session.close() 只断开 CDP 连接，不会杀 Popen 启动的 Chrome 进程。
+        killed = _kill_port_listeners(child_cdp_port)
+        if killed:
+            time.sleep(1)
+            print(f"[CLEAN] 品牌 {brand} 采集后清理脚本专属端口 Chrome ({killed} 个进程)")
+
+    print(f"\n{'=' * 60}")
+    if failed_brands:
+        print(f"[WARN] 失败品牌: {', '.join(failed_brands)}")
+        return 1
+
+    print(f"[OK] 所有品牌采集完成, 数据已写入 data/tm单品数据/")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_collect_backfill.py b/src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_collect_backfill.py
new file mode 100644
index 0000000..457d65c
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_collect_backfill.py
@@ -0,0 +1,702 @@
+# -*- coding: utf-8 -*-
+"""
+淘宝生意参谋商品采集 — 历史回采专用 (scrapling 简化版)。
+
+跟 `taobao_sycm_collect.py` (日常采集) 共享同一份 USER_DATA_DIR 和登录态,
+但日期走 UI 翻月 (3 入口: "日"按钮悬浮 / 点日期区域 / "自定义" 按钮),
+复用 taobao_sycm_products.align_calendar_month 翻到目标月再点日,
+适合补采数周到数月前的数据。UI 点日失败时回退到 URL 直跳。
+
+跟主脚本完全解耦:
+  - 单独的 main() / 启 Chrome 函数
+  - --target-date 必填 (不允许默认昨天)
+  - 不动 select_yesterday / navigate_to_products
+  - 落盘路径与主脚本一致 (data/tm单品数据/_downloads/<日期>/, data/tm单品数据/<款式>_<日期>/),
+    二次回采同样日期会生成带时间戳后缀的文件, 不覆盖。
+
+用法:
+    .venv\\Scripts\\python.exe taobao_sycm_collect_backfill.py --target-date 2026-05-15
+    .venv\\Scripts\\python.exe taobao_sycm_collect_backfill.py --target-date 2026-05-15 --headless
+    .venv\\Scripts\\python.exe taobao_sycm_collect_backfill.py --target-date 2026-05-15 --user-data-dir <GYXX_DATA_ROOT>\\profiles\\product_commerce\\tm
+
+浏览器端口和 Profile 由 GYXX 按脚本独立分配。
+"""
+
+from __future__ import annotations
+
+import argparse
+import shutil
+import sys
+import time
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+
+from taobao_sycm_products import (
+    USER_DATA_DIR,
+    TARGET_DATE,
+    PRODUCTS_URL,
+    parse_target_date,
+    ensure_logged_in,
+    close_ad_popup,
+    close_competition_reminder,
+    close_blocking_tips,
+    click_download,
+    export_excel_to_json_and_md,
+    summarize_styles_from_excel,
+)
+from collect_retry_utils import FailedCollector, retry_step
+
+
+def start_stealthy_session(headless: bool, user_data_dir: str):
+    """跟 taobao_sycm_collect.py 一致的 StealthySession 启 Chrome。"""
+    from scrapling.fetchers import StealthySession
+
+    session = StealthySession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=user_data_dir,
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=60000,
+        network_idle=False,
+    )
+    session.start()
+    context = session.context
+    page = context.pages[-1] if context.pages else context.new_page()
+    return session, page
+
+
+def open_date_panel(page) -> bool:
+    """打开日期面板 — 走"日"按钮悬浮 / 直接点日期区域 / "自定义" 三种入口。"""
+    # 清弹窗
+    close_ad_popup(page)
+    close_competition_reminder(page)
+    close_blocking_tips(page)
+
+    # 入口 1: 悬浮 "日" 按钮 (跟 select_yesterday 一致)
+    hovered = page.evaluate(
+        r"""() => {
+            const btns = document.querySelectorAll('.oui-date-picker-particle-button button');
+            for (const btn of btns) {
+                if (btn.textContent.trim() === '日') {
+                    btn.scrollIntoView({block: 'center'});
+                    btn.dispatchEvent(new MouseEvent('mouseover', { bubbles: true, view: window }));
+                    btn.dispatchEvent(new MouseEvent('mouseenter', { bubbles: true, view: window }));
+                    return true;
+                }
+            }
+            return false;
+        }"""
+    )
+    if hovered:
+        time.sleep(2)
+        if page.evaluate(
+            r"""() => !!document.querySelector('.ant-calendar, .ant-picker-panel, [class*="calendar"], [class*="picker-panel"]')"""
+        ):
+            return True
+        # 没展开, 试着点
+        clicked = page.evaluate(
+            r"""() => {
+                const btns = document.querySelectorAll('.oui-date-picker-particle-button button');
+                for (const btn of btns) {
+                    if (btn.textContent.trim() === '日') { btn.click(); return true; }
+                }
+                return false;
+            }"""
+        )
+        if clicked:
+            time.sleep(2)
+            if page.evaluate(
+                r"""() => !!document.querySelector('.ant-calendar, .ant-picker-panel, [class*="calendar"], [class*="picker-panel"]')"""
+            ):
+                return True
+
+    # 入口 2: 点日期区域 (上面通常写着 "2026年06月12日" 之类)
+    clicked_area = page.evaluate(
+        r"""() => {
+            const visible = (el) => el && el.offsetParent !== null
+                && window.getComputedStyle(el).display !== 'none';
+            const candidates = Array.from(document.querySelectorAll('div, span, a, button'))
+                .filter(visible)
+                .filter((el) => {
+                    const text = (el.textContent || '').trim();
+                    return /20\d{2}[-/.年]\s*\d{1,2}[-/.月]\s*\d{1,2}/.test(text)
+                        && el.children.length < 6
+                        && el.getBoundingClientRect().width < 400;
+                })
+                .sort((a, b) => a.getBoundingClientRect().width - b.getBoundingClientRect().width);
+            if (candidates[0]) { candidates[0].click(); return true; }
+            return false;
+        }"""
+    )
+    if clicked_area:
+        time.sleep(2)
+        if page.evaluate(
+            r"""() => !!document.querySelector('.ant-calendar, .ant-picker-panel, [class*="calendar"], [class*="picker-panel"]')"""
+        ):
+            return True
+
+    # 入口 3: 点 "自定义" 按钮 (有的版本在快捷选项里)
+    clicked_custom = page.evaluate(
+        r"""() => {
+            const visible = (el) => el && el.offsetParent !== null
+                && window.getComputedStyle(el).display !== 'none';
+            const btns = Array.from(document.querySelectorAll('button, span, a, div, li'))
+                .filter(visible)
+                .filter((el) => {
+                    const text = (el.textContent || '').trim();
+                    return text === '自定义' || text === '自定義'
+                        || text.startsWith('自定义');
+                });
+            if (btns[0]) { btns[0].click(); return true; }
+            return false;
+        }"""
+    )
+    if clicked_custom:
+        time.sleep(2)
+        if page.evaluate(
+            r"""() => !!document.querySelector('.ant-calendar, .ant-picker-panel, [class*="calendar"], [class*="picker-panel"]')"""
+        ):
+            return True
+
+    return False
+
+
+def wait_for_data_table(page, max_seconds: int = 30) -> bool:
+    """等 sycm 表格加载完成 (出现 <table> 且含 '商品'/'标题' 之类列头)。"""
+    deadline = time.time() + max_seconds
+    while time.time() < deadline:
+        ready = page.evaluate(
+            r"""() => {
+                const tables = Array.from(document.querySelectorAll('table'));
+                for (const t of tables) {
+                    const text = (t.innerText || '').replace(/\s+/g, ' ');
+                    if (text.length > 100 && /商品|标题|流量|访客|收藏/.test(text)) return true;
+                }
+                return false;
+            }"""
+        )
+        if ready:
+            return True
+        time.sleep(1)
+    return False
+
+
+def _tm_panel_alive(page) -> bool:
+    """
+    TM 启发式校验:页面中存在一个可视元素,文本既含 'YYYY年M月' 又含至少 28 个数字 cell。
+    不依赖 ant-calendar / ant-picker class,任何自研组件都能识别。
+    """
+    return page.evaluate(
+        r"""
+        () => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none' && style.visibility !== 'hidden'
+                    && rect.width > 0 && rect.height > 0 && el.offsetParent !== null;
+            };
+            const monthRe = /20\d{2}\s*年\s*\d{1,2}\s*月/;
+            const nodes = Array.from(document.querySelectorAll('*')).filter(visible);
+            for (const el of nodes) {
+                const txt = el.textContent || '';
+                if (!monthRe.test(txt)) continue;
+                const digits = (txt.match(/\b\d{1,2}\b/g) || []).length;
+                if (digits >= 28 && el.children.length < 200) return true;
+            }
+            return false;
+        }
+        """
+    )
+
+
+def _tm_reopen_day_button(page) -> bool:
+    """
+    日历面板意外收起后,重新触发 '日' 按钮把面板拉起来。
+    生意参谋用 .oui-date-picker-particle-button 容器,日/周/月/自定义都可能唤起。
+    """
+    return page.evaluate(
+        r"""
+        () => {
+            const allBtns = document.querySelectorAll('button');
+            for (const btn of allBtns) {
+                const t = (btn.textContent || '').trim();
+                if (['日', '周', '月', '自定义', '日 实时', '日 7天', '日 30天'].includes(t)
+                    || /^日(\s|$)/.test(t)) {
+                    btn.scrollIntoView({block: 'center'});
+                    btn.dispatchEvent(new MouseEvent('mouseover', {bubbles: true, view: window}));
+                    btn.dispatchEvent(new MouseEvent('mouseenter', {bubbles: true, view: window}));
+                    btn.click();
+                    return true;
+                }
+            }
+            const containers = document.querySelectorAll('.oui-date-picker-particle-button button');
+            for (const btn of containers) {
+                if ((btn.textContent || '').trim() === '日') {
+                    btn.scrollIntoView({block: 'center'});
+                    btn.dispatchEvent(new MouseEvent('mouseover', {bubbles: true, view: window}));
+                    btn.dispatchEvent(new MouseEvent('mouseenter', {bubbles: true, view: window}));
+                    btn.click();
+                    return true;
+                }
+            }
+            return false;
+        }
+        """
+    )
+
+
+def _tm_find_panel(page) -> dict:
+    """
+    启发式找真实日期浮层 panel。
+    返回 {found, panel_rect, current_year, current_month}。
+    找不到时 found=False。
+    """
+    return page.evaluate(
+        r"""
+        () => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none' && style.visibility !== 'hidden'
+                    && rect.width > 0 && rect.height > 0 && el.offsetParent !== null;
+            };
+            const monthRe = /20\d{2}\s*年\s*\d{1,2}\s*月/;
+            const isLikelyPanel = (el) => {
+                if (!el || el === document.body || el === document.documentElement) return false;
+                const r = el.getBoundingClientRect();
+                if (r.width < 150 || r.width > 700) return false;
+                if (r.height < 200 || r.height > 700) return false;
+                const style = window.getComputedStyle(el);
+                const pos = style.position;
+                const cls = (el.className || '').toString();
+                return pos === 'absolute' || pos === 'fixed' || pos === 'relative'
+                    || /popover|panel|picker|calendar|date|dropdown|pop/i.test(cls);
+            };
+            const allVisible = Array.from(document.querySelectorAll('*')).filter(visible);
+            const monthAnchors = allVisible.filter((el) => monthRe.test(el.textContent || '')
+                && el.children.length < 60);
+            let candidate = null;
+            for (const el of monthAnchors) {
+                let cur = el;
+                while (cur && cur !== document.body) {
+                    if (isLikelyPanel(cur)) { candidate = cur; break; }
+                    cur = cur.parentElement;
+                }
+                if (candidate) break;
+            }
+            if (!candidate) {
+                for (const el of monthAnchors) {
+                    const txt = el.textContent || '';
+                    const digits = (txt.match(/\b\d{1,2}\b/g) || []).length;
+                    if (digits >= 28 && el.children.length < 200) {
+                        let cur = el.parentElement;
+                        while (cur && cur !== document.body) {
+                            const r = cur.getBoundingClientRect();
+                            if (r.width >= 150 && r.width <= 700 && r.height >= 200 && r.height <= 700) {
+                                candidate = cur; break;
+                            }
+                            cur = cur.parentElement;
+                        }
+                        if (candidate) break;
+                    }
+                }
+            }
+            if (!candidate) return { found: false };
+            const match = (candidate.textContent || '').match(/(20\d{2})\s*年\s*(\d{1,2})\s*月/);
+            const r = candidate.getBoundingClientRect();
+            return {
+                found: true,
+                panel_rect: {x: r.left, y: r.top, w: r.width, h: r.height},
+                current_year: match ? Number(match[1]) : null,
+                current_month: match ? Number(match[2]) : null,
+            };
+        }
+        """
+    )
+
+
+def _tm_flip_month(page, target_year: int, target_month: int) -> bool:
+    """
+    自研翻月循环:启发式找 panel,在 panelRect 范围内找 ‹ › 字符按钮(或 "上/下个月" 文本)。
+    翻一次校验一次 panel 还活着,死了重开。
+    最多翻 18 次(够 1.5 年历史)。
+    """
+    for step in range(18):
+        if not _tm_panel_alive(page):
+            print(f"    [WARN] panel 已收起,重开 (step={step})")
+            _tm_reopen_day_button(page)
+            time.sleep(1.5)
+            if not _tm_panel_alive(page):
+                print(f"    [WARN] panel 拉不起来,翻月中止")
+                return False
+        info = page.evaluate(
+            r"""
+            ({targetYear, targetMonth}) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none' && style.visibility !== 'hidden'
+                        && rect.width > 0 && rect.height > 0 && el.offsetParent !== null;
+                };
+                const monthRe = /20\d{2}\s*年\s*\d{1,2}\s*月/;
+                // 找 panel
+                const isLikelyPanel = (el) => {
+                    if (!el || el === document.body) return false;
+                    const r = el.getBoundingClientRect();
+                    if (r.width < 150 || r.width > 700) return false;
+                    if (r.height < 200 || r.height > 700) return false;
+                    const s = window.getComputedStyle(el);
+                    const cls = (el.className || '').toString();
+                    return s.position === 'absolute' || s.position === 'fixed' || s.position === 'relative'
+                        || /popover|panel|picker|calendar|date|dropdown|pop/i.test(cls);
+                };
+                const allVisible = Array.from(document.querySelectorAll('*')).filter(visible);
+                const monthAnchors = allVisible.filter((el) => monthRe.test(el.textContent || '')
+                    && el.children.length < 60);
+                let panel = null;
+                for (const el of monthAnchors) {
+                    let cur = el;
+                    while (cur && cur !== document.body) {
+                        if (isLikelyPanel(cur)) { panel = cur; break; }
+                        cur = cur.parentElement;
+                    }
+                    if (panel) break;
+                }
+                if (!panel) return { status: 'no_panel' };
+                const match = (panel.textContent || '').match(/(20\d{2})\s*年\s*(\d{1,2})\s*月/);
+                if (!match) return { status: 'no_month_title' };
+                const currentYear = Number(match[1]);
+                const currentMonth = Number(match[2]);
+                if (currentYear === targetYear && currentMonth === targetMonth) {
+                    return { status: 'aligned', currentYear, currentMonth };
+                }
+                const diff = (targetYear - currentYear) * 12 + (targetMonth - currentMonth);
+                const r = panel.getBoundingClientRect();
+                const inRect = (el) => {
+                    const er = el.getBoundingClientRect();
+                    return er.left >= r.left - 5 && er.right <= r.right + 5
+                        && er.top >= r.top - 30 && er.bottom <= r.bottom + 30;
+                };
+                const wantChars = diff < 0 ? ['‹', '«', '<'] : ['›', '»', '>'];
+                const wantText = diff < 0 ? ['上个月', '上一年', '上月'] : ['下个月', '下一年', '下月'];
+                // 在 panelRect 范围内找按钮
+                const allBtns = Array.from(document.querySelectorAll('button, a, [role="button"], span, i, em, div'))
+                    .filter(visible);
+                for (const el of allBtns) {
+                    if (!inRect(el)) continue;
+                    const txt = (el.textContent || '').trim();
+                    if (wantChars.includes(txt)) {
+                        const er = el.getBoundingClientRect();
+                        return { status: 'click', x: er.left + er.width / 2, y: er.top + er.height / 2,
+                                 currentYear, currentMonth, diff, btn: 'arrow:' + txt };
+                    }
+                }
+                for (const el of allBtns) {
+                    if (!inRect(el)) continue;
+                    const txt = (el.textContent || '').trim();
+                    if (wantText.some((w) => txt === w || (txt.includes(w) && txt.length <= w.length + 2))) {
+                        const er = el.getBoundingClientRect();
+                        return { status: 'click', x: er.left + er.width / 2, y: er.top + er.height / 2,
+                                 currentYear, currentMonth, diff, btn: 'text:' + txt };
+                    }
+                }
+                return { status: 'no_nav', currentYear, currentMonth, diff };
+            }
+            """,
+            {"targetYear": target_year, "targetMonth": target_month},
+        )
+        if not isinstance(info, dict):
+            info = {"status": str(info)}
+        if info.get("status") == "aligned":
+            return True
+        if info.get("status") != "click":
+            print(f"    [WARN] 翻月停止 (step={step}): {info}")
+            return False
+        x, y = info["x"], info["y"]
+        page.mouse.move(x, y)
+        time.sleep(0.15)
+        page.mouse.click(x, y)
+        time.sleep(0.9)
+    return False
+
+
+def click_target_day_cell(page, target_date) -> bool:
+    """
+    在已打开的日历面板里翻到目标月,再点目标日。返回是否成功。
+
+    全部就地实现启发式翻月/点日 —— 不依赖主脚本的 align_calendar_month(已回滚)。
+    """
+    target_date_str = target_date.strftime("%Y-%m-%d")
+    target_day = str(target_date.day)
+
+    # 翻月
+    if not _tm_flip_month(page, target_date.year, target_date.month):
+        print(f"    [WARN] 翻月失败,目标 {target_date_str} 还没到")
+        return False
+    time.sleep(0.5)
+
+    # 翻完后 panel 可能因 hover-out 关闭,先校验
+    if not _tm_panel_alive(page):
+        print(f"    [WARN] 翻完月后 panel 收起,重开")
+        _tm_reopen_day_button(page)
+        time.sleep(1.5)
+        if not _tm_flip_month(page, target_date.year, target_date.month):
+            return False
+        time.sleep(0.5)
+
+    # 点日
+    result = page.evaluate(
+        r"""
+        ({targetDate, targetDay}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none' && style.visibility !== 'hidden'
+                    && rect.width > 0 && rect.height > 0 && el.offsetParent !== null;
+            };
+            const monthRe = /20\d{2}\s*年\s*\d{1,2}\s*月/;
+            // 启发式找 panel
+            const isLikelyPanel = (el) => {
+                if (!el || el === document.body) return false;
+                const r = el.getBoundingClientRect();
+                if (r.width < 150 || r.width > 700) return false;
+                if (r.height < 200 || r.height > 700) return false;
+                const s = window.getComputedStyle(el);
+                const cls = (el.className || '').toString();
+                return s.position === 'absolute' || s.position === 'fixed' || s.position === 'relative'
+                    || /popover|panel|picker|calendar|date|dropdown|pop/i.test(cls);
+            };
+            const allVisible = Array.from(document.querySelectorAll('*')).filter(visible);
+            const monthAnchors = allVisible.filter((el) => monthRe.test(el.textContent || '')
+                && el.children.length < 60);
+            let panel = null;
+            for (const el of monthAnchors) {
+                let cur = el;
+                while (cur && cur !== document.body) {
+                    if (isLikelyPanel(cur)) { panel = cur; break; }
+                    cur = cur.parentElement;
+                }
+                if (panel) break;
+            }
+            if (!panel) return { status: 'no_panel' };
+
+            // 方式 1: title / aria-label / data-value 精确匹配
+            const cells = panel.querySelectorAll('td, [class*="cell"], [role="gridcell"]');
+            for (const cell of cells) {
+                if (!visible(cell)) continue;
+                const title = cell.getAttribute('title') || '';
+                const aria = cell.getAttribute('aria-label') || '';
+                const dataValue = cell.getAttribute('data-value') || '';
+                if (title === targetDate || aria.includes(targetDate) || dataValue.includes(targetDate)) {
+                    const clickable = cell.querySelector('div') || cell;
+                    clickable.click();
+                    return { status: 'ok', how: 'title' };
+                }
+            }
+            // 方式 2: 文本 == 目标日的叶子元素,排除上下月/disabled
+            const allEls = Array.from(panel.querySelectorAll('*'));
+            for (const el of allEls) {
+                if (el.children.length !== 0) continue;
+                if ((el.textContent || '').trim() !== targetDay) continue;
+                const parent = el.closest('td')
+                    || el.closest('[class*="cell"]')
+                    || el.parentElement;
+                if (!parent || !visible(parent)) continue;
+                const classText = `${parent.className || ''} ${el.className || ''} ${parent.parentElement?.className || ''}`;
+                if (/disabled|off|unavailable|prev-month|next-month|other-month|not-current-month/i.test(classText)) {
+                    continue;
+                }
+                // 返回坐标,让外层用 mouse.click(避免 evaluate 内部 click 触发 panel 关闭)
+                const rect = parent.getBoundingClientRect();
+                return { status: 'coords', x: rect.left + rect.width / 2, y: rect.top + rect.height / 2 };
+            }
+            return { status: 'not_found' };
+        }
+        """,
+        {"targetDate": target_date_str, "targetDay": target_day},
+    )
+
+    if not isinstance(result, dict):
+        return False
+    if result.get("status") == "ok":
+        time.sleep(3)
+        return True
+    if result.get("status") == "coords":
+        x, y = result["x"], result["y"]
+        page.mouse.move(x, y)
+        time.sleep(0.15)
+        page.mouse.click(x, y)
+        time.sleep(3)
+        return True
+    print(f"    [WARN] 点日失败: {result}")
+    return False
+
+
+def main() -> int:
+    global TARGET_DATE
+
+    parser = argparse.ArgumentParser(
+        description="淘宝生意参谋商品采集 — 历史回采 (UI 翻月 + URL fallback)"
+    )
+    parser.add_argument("--target-date", type=parse_target_date, required=True,
+                        help="回采日期 YYYY-MM-DD (必填, 不允许默认昨天)")
+    parser.add_argument("--user-data-dir", default=USER_DATA_DIR,
+                        help=f"Chrome profile 目录, 默认 {USER_DATA_DIR}")
+    parser.add_argument("--headless", action="store_true", help="无头模式")
+    parser.add_argument("--brand", default=None, help="当前采集品牌 (光影行星/ozko), summarize 只处理该品牌款式")
+    args = parser.parse_args()
+
+    TARGET_DATE = args.target_date
+    # 同时把 TARGET_DATE 写到主脚本模块的 global,否则 click_download 内部用的
+    # _download_target_dir() 读到的是主脚本模块级 TARGET_DATE = None,会落盘到
+    # "昨天"目录,而不是真正的回采日期。
+    import taobao_sycm_products as _main_module
+    _main_module.TARGET_DATE = args.target_date
+    _main_module.BRAND = args.brand
+    if args.brand:
+        print(f"[INFO] 品牌过滤: {args.brand}")
+    date_str = args.target_date.strftime("%Y-%m-%d")
+    print(f"[INFO] 回采日期: {date_str}")
+    print(f"[INFO] Profile 目录: {args.user_data_dir}")
+    print(f"[INFO] Headless: {args.headless}")
+
+    failed = FailedCollector("tm", date_str)
+
+    print("[STEP 1/5] 启 Chrome ...")
+    session, page = start_stealthy_session(args.headless, args.user_data_dir)
+    if not session or not page:
+        print("[FAIL] 启 Chrome 失败")
+        failed.add_page("启 Chrome", "session or page is None", attempts=1)
+        failed_path = failed.write()
+        if failed_path:
+            print(f"[FAIL_LOG] {failed.summary_line()} -> {failed_path}")
+        return 1
+
+    try:
+        print("[STEP 2/5] 检查登录态 ...")
+        page = ensure_logged_in(session.context, page)
+        if not page:
+            print("[FAIL] 登录失败, 脚本退出")
+            failed.add_page("登录态", "ensure_logged_in 返回 None", attempts=1)
+            failed_path = failed.write()
+            if failed_path:
+                print(f"[FAIL_LOG] {failed.summary_line()} -> {failed_path}")
+            return 1
+
+        print("[STEP 3/5] 进入商品页面 ...")
+        page.goto(PRODUCTS_URL, wait_until="networkidle", timeout=90000)
+        time.sleep(3)
+        close_ad_popup(page)
+        close_competition_reminder(page)
+        close_blocking_tips(page)
+        print(f"  当前页面: {page.url}")
+
+        print(f"[STEP 4/5] UI 选目标日期 {date_str} ...")
+        # per-page 重试 (1 次): 日期面板打开三入口已算 1 次, 这里再 retry_step 兜底
+        panel_ok = False
+        try:
+            panel_ok = retry_step(open_date_panel, page, max_attempts=2, sleep_seconds=3,
+                                  label="TM 日期面板")
+        except Exception as exc:
+            failed.add_page("TM 日期面板", f"{type(exc).__name__}: {exc}", attempts=2)
+        if not panel_ok:
+            print("[FAIL] 打不开日期面板 (已重试)")
+            failed_path = failed.write()
+            if failed_path:
+                print(f"[FAIL_LOG] {failed.summary_line()} -> {failed_path}")
+            return 1
+        print("  日期面板已打开")
+
+        close_ad_popup(page)
+        close_competition_reminder(page)
+        close_blocking_tips(page)
+        time.sleep(0.5)
+
+        # per-page 重试 (1 次): UI 点日失败 → 重新打开日期面板 + 重试 1 次
+        day_ok = False
+        try:
+            day_ok = retry_step(click_target_day_cell, page, args.target_date,
+                                max_attempts=2, sleep_seconds=3, label="TM UI 点日",
+                                on_retry=lambda *_: open_date_panel(page))
+        except Exception as exc:
+            failed.add_page("TM UI 点日", f"{type(exc).__name__}: {exc}", attempts=2)
+        if not day_ok:
+            print(f"  [WARN] UI 点日失败 (已重试), 改用 URL fallback")
+            # dateType=day 让淘宝按"天"维度显示,dateRange 才会被尊重;
+            # dateType=today 会让淘宝忽略 dateRange,只显示今天 —— 历史回采会跳错。
+            date_url = f"{PRODUCTS_URL}?dateRange={date_str}%7C{date_str}&dateType=day"
+            page.goto(date_url, wait_until="networkidle", timeout=90000)
+            time.sleep(3)
+
+        # per-page 重试 (1 次): 数据表格加载失败 → 重新选日期 + 再等
+        table_ok = False
+        try:
+            def _do_wait_table():
+                wait_for_data_table(page, max_seconds=30)
+                return True
+            table_ok = retry_step(_do_wait_table,
+                                  max_attempts=2, sleep_seconds=3, label="TM 数据表格",
+                                  on_retry=lambda *_: click_target_day_cell(page, args.target_date))
+        except Exception as exc:
+            failed.add_page("TM 数据表格", f"{type(exc).__name__}: {exc}", attempts=2)
+        if not table_ok:
+            print(f"  [WARN] 30s 内未检测到数据表格 (已重试), 该日期 {date_str} 在 sycm 可能无数据, 继续尝试下载")
+
+        print("[STEP 5/5] 点击下载 ...")
+        # per-page 重试 (1 次): 下载失败 → 重新选日期 + 再点下载
+        excel_path = None
+        try:
+            def _do_download():
+                ep = click_download(page)
+                if not ep:
+                    raise RuntimeError("click_download 返回 None")
+                return ep
+            excel_path = retry_step(_do_download, max_attempts=2, sleep_seconds=3,
+                                    label="TM 下载",
+                                    on_retry=lambda *_: click_target_day_cell(page, args.target_date))
+        except Exception as exc:
+            failed.add_page("TM 下载", f"{type(exc).__name__}: {exc}", attempts=2)
+        if not excel_path:
+            print("[FAIL] 下载失败 (已重试)")
+            failed_path = failed.write()
+            if failed_path:
+                print(f"[FAIL_LOG] {failed.summary_line()} -> {failed_path}")
+            return 1
+
+        # 修正落盘目录:click_download 用了主脚本模块的 _download_target_dir(),
+        # 在我们刚启动的 backfill 进程里它可能仍然读到 TARGET_DATE=None,从而
+        # 把文件落到 _downloads/<昨天>/ 而不是 _downloads/<回采日期>/。
+        # 把下载文件移到正确目录,确保路径里包含回采日期。
+        correct_dir = RAW_DATA_ROOT / "tm单品数据" / "_downloads" / date_str
+        if not str(excel_path).startswith(str(correct_dir)):
+            correct_dir.mkdir(parents=True, exist_ok=True)
+            target_xls = correct_dir / Path(excel_path).name
+            shutil.move(str(excel_path), str(target_xls))
+            print(f"  [MOVE] 修正落盘目录: {excel_path} -> {target_xls}")
+            excel_path = str(target_xls)
+
+        print("[STEP 5/5] 解析 Excel 并导出 ...")
+        export_excel_to_json_and_md(excel_path)
+        summarize_styles_from_excel(excel_path)
+
+        print(f"\n[OK] 回采完成, 数据已写入 data/tm单品数据/_downloads/{date_str}/")
+        failed_path = failed.write()
+        if failed_path:
+            print(f"[FAIL_LOG] {failed.summary_line()} -> {failed_path}")
+        else:
+            print(f"[OK_LOG] TM 全成功, 无失败清单")
+        return 0
+    finally:
+        session.close()
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_products.py b/src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_products.py
new file mode 100644
index 0000000..74696e0
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/taobao_sycm_products.py
@@ -0,0 +1,1353 @@
+"""
+淘宝生意参谋商品页面自动化脚本
+功能：自动登录（复用cookie） -> 进入商品页面 -> 选择上周 -> 下载数据
+"""
+
+from datetime import datetime, timedelta
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT, profile_path
+import subprocess
+import time
+import os
+import shutil
+import json
+import urllib.request
+from openpyxl import load_workbook
+from erp_metric_overrides import apply_erp_override, parse_number as parse_metric_number
+from config.style_config_loader import get_tm_style_id_map, get_tm_brand_map
+import argparse
+import sys
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+
+# 绕过系统代理（连接本地 Chrome 调试端口必须）
+os.environ.pop("HTTP_PROXY", None)
+os.environ.pop("HTTPS_PROXY", None)
+os.environ.pop("http_proxy", None)
+os.environ.pop("https_proxy", None)
+
+CHROME_PATH = os.getenv("CHROME_PATH", "")
+DEBUG_PORT = int(os.getenv("GYXX_BROWSER_CDP_PORT", "0"))
+# Profile 和 CDP 端口均由 GYXX 按脚本分配。
+# cookie 跑脚本前从默认 User Data 同步过来, 用户日常 Chrome 不动。
+# 改这里或设环境变量 TM_USER_DATA_DIR 可换其他非默认目录。
+USER_DATA_DIR = os.getenv("TM_USER_DATA_DIR", str(profile_path("tm")))
+DAILY_USER_DATA_DIR = os.getenv(
+    "TM_DAILY_USER_DATA_DIR", str(profile_path("tm-daily"))
+)
+LOGIN_URL = "https://sycm.taobao.com/custom/login.htm?_target=http://sycm.taobao.com/custom/login"
+PRODUCTS_URL = "https://sycm.taobao.com/cc/item_rank"
+SINGLE_PRODUCT_DATA_DIR = RAW_DATA_ROOT / "tm单品数据"
+DOWNLOADS_DIR = SINGLE_PRODUCT_DATA_DIR / "_downloads"
+TARGET_DATE = None
+
+ACCOUNT = os.getenv("SYCM_ACCOUNT", "光影行星旗舰店:龙虾仔")
+PASSWORD = os.getenv("SYCM_PASSWORD", "aa123456")
+BRAND = None
+
+
+def find_browser_executable():
+    candidates = [
+        CHROME_PATH,
+        r"C:\Program Files\Google\Chrome\Application\chrome.exe",
+        r"C:\Program Files (x86)\Google\Chrome\Application\chrome.exe",
+        r"C:\Program Files\Microsoft\Edge\Application\msedge.exe",
+        r"C:\Program Files (x86)\Microsoft\Edge\Application\msedge.exe",
+    ]
+    local_app_data = os.getenv("LOCALAPPDATA")
+    if local_app_data:
+        candidates.extend([
+            os.path.join(local_app_data, "Google", "Chrome", "Application", "chrome.exe"),
+            os.path.join(local_app_data, "Microsoft", "Edge", "Application", "msedge.exe"),
+        ])
+
+    user_profile = os.getenv("USERPROFILE")
+    if user_profile:
+        candidates.extend([
+            os.path.join(user_profile, "AppData", "Local", "ms-playwright", "chromium-1223", "chrome-win64", "chrome.exe"),
+            os.path.join(user_profile, "AppData", "Local", "ms-playwright", "chromium_headless_shell-1223", "chrome-win", "headless_shell.exe"),
+        ])
+
+    for path in candidates:
+        if path and os.path.exists(path):
+            return path
+
+    if user_profile:
+        playwright_root = Path(user_profile) / "AppData" / "Local" / "ms-playwright"
+        if playwright_root.exists():
+            for name in ("chrome.exe", "msedge.exe", "headless_shell.exe"):
+                matches = sorted(playwright_root.rglob(name), reverse=True)
+                if matches:
+                    return str(matches[0])
+
+    raise FileNotFoundError(
+        "找不到 Chrome/Edge/Playwright Chromium。请安装浏览器，或设置环境变量 CHROME_PATH 指向浏览器 exe。"
+    )
+
+# 保留硬编码作为灾备回退；运行时优先从飞书多维表格读取。
+_STATIC_STYLE_ID_MAP = {
+    "盖亚斜挎": ["921336775006", "674099008611", "673333504257", "685132225920"],
+    "极星pro": ["832214449817", "785061026078", "785013594474"],
+    "盖亚微单": ["765518426414", "710312339945"],
+    "盖世m1": ["822904000398", "929578566382"],
+    "星迹2": ["770900272208", "765693071830", "1018640025955", "1019020504788"],
+    "星云2": ["900176117674", "865283623976", "669949528463"],
+    "宙斯": ["986141740958", "655191144133", "675338689161", "743062995650", "775995806796", "877579274516", "764357861270"],
+    "阿波罗2": ["900428335849", "956633501942"],
+    "阿波罗x1": ["727733319542"],
+    "波塞冬edge": ["981750801552", "980988758018"],
+    "逐星": ["809320285523", "814599474998"],
+    "逐星GT": ["1001281519943"],
+    "星云mini": ["981022406378"],
+    "晨星2": ["969316700235"],
+    "极星双肩": ["972285625931", "966753757238", "938160082795", "1021224008872", "938138060543"],
+    "凌云air": ["921333419403", "941604987083"],
+    "觅光": ["999776082964", "1015049280303", "1014849366869", "1016460984468"],
+    "云卷2": ["985147706233"],
+    "凌云3": ["1051305611567", "1051293159491"],
+    "瑞白": ["1040706443911", "1001264424551"],
+    "黑曜石吐司包": ["1022486093557", "1026596944943"],
+    "黑曜石托特包": ["1022490201756", "1026467516876"],
+    "曜影": ["1045840128509", "1048058265963"],
+    "黑曜石电脑包": ["1046854786628", "1046497090037"],
+    "拓界2": ["1051287111533", "1053051469809"],
+}
+
+
+def _get_style_id_map():
+    """优先从飞书读取天猫商品ID映射，失败则回退硬编码。"""
+    try:
+        return get_tm_style_id_map()
+    except Exception as exc:
+        print(f"[WARN] 动态读取天猫商品ID失败，使用硬编码回退: {exc}")
+        return _STATIC_STYLE_ID_MAP
+
+
+STYLE_ID_MAP = _STATIC_STYLE_ID_MAP
+
+
+def is_chrome_debug_running():
+    try:
+        req = urllib.request.Request(f"http://127.0.0.1:{DEBUG_PORT}/json/version")
+        req.add_header("Content-Type", "application/json")
+        proxy_handler = urllib.request.ProxyHandler({})
+        opener = urllib.request.build_opener(proxy_handler)
+        with opener.open(req, timeout=2) as resp:
+            return resp.status == 200
+    except Exception:
+        return False
+
+
+def _profile_is_locked(user_data_dir: str) -> tuple[bool, str | None]:
+    """profile 目录是否被另一个 Chrome 进程占用。
+
+    若被占用，再启一个 Chrome 同 user-data-dir 会失败、或者落到无 cookie 的
+    "Default (1)" 子目录里——这时脚本看着跑起来了，但完全复用不到老 cookie。
+    """
+    for lock_name in ("SingletonLock", "SingletonCookie", "lockfile"):
+        lock_path = os.path.join(user_data_dir, lock_name)
+        if os.path.exists(lock_path):
+            return True, lock_name
+    return False, None
+
+
+def sync_cookies_from_daily_profile(src_user_data: str, dst_user_data: str) -> int:
+    """从日常 Chrome (默认 User Data) 同步 cookie / 登录态到目标 profile。
+
+    跑脚本前调一次。源 = 默认 User Data, 目标 = 脚本用的非默认目录
+    (GYXX_DATA_ROOT 下的 TM profile), 这样脚本启 Chrome 时:
+      1) 使用 GYXX 分配的独立 Profile 和 CDP 端口
+      2) 同步过来的 cookie 让淘宝"已登录"状态保留, 免登录 / 免滑块
+
+    返回同步的文件数。
+    """
+    src_root = Path(src_user_data)
+    dst_root = Path(dst_user_data)
+    if src_root.resolve() == dst_root.resolve():
+        return 0
+    src_default = src_root / "Default"
+    if not src_default.exists():
+        print(f"  [WARN] 日常 Chrome profile 不存在: {src_default}，跳过 cookie 同步")
+        return 0
+    dst_default = dst_root / "Default"
+    dst_default.mkdir(parents=True, exist_ok=True)
+
+    copied = 0
+    # 关键文件: SQLite 形式的 cookie / 登录数据
+    for fname in (
+        "Cookies", "Cookies-journal",
+        "Login Data", "Login Data-journal",
+        "Web Data", "Web Data-journal",
+    ):
+        src = src_default / fname
+        if src.exists():
+            shutil.copy2(src, dst_default / fname)
+            copied += 1
+    # Local Storage / Session Storage (淘宝登录态可能存在 localStorage 的 token)
+    for dirname in ("Local Storage", "Session Storage", "Network"):
+        src = src_default / dirname
+        dst = dst_default / dirname
+        if src.exists():
+            if dst.exists():
+                shutil.rmtree(dst, ignore_errors=True)
+            try:
+                shutil.copytree(src, dst)
+                copied += 1
+            except Exception as e:
+                print(f"  [WARN] 同步 {dirname} 失败: {e}")
+    return copied
+
+
+def _kill_port_listeners(port: int) -> int:
+    """杀掉占用指定端口的进程 (Windows)。返回杀掉的进程数。
+
+    用于清理上次脚本残留的 chrome.exe,避免新启 Chrome 时屏幕上同时出现 2 个浏览器。
+    只处理本脚本分配到的调试端口，不影响其他脚本。
+    """
+    if sys.platform != "win32":
+        return 0
+    try:
+        out = subprocess.check_output(
+            f"netstat -ano | findstr :{port}",
+            shell=True, text=True, encoding="gbk", errors="replace", timeout=5,
+        )
+    except subprocess.CalledProcessError:
+        return 0
+    except Exception:
+        return 0
+    pids = set()
+    for line in out.splitlines():
+        parts = line.split()
+        if len(parts) < 5 or not parts[-1].isdigit():
+            continue
+        if f":{port}" in parts[1] and parts[3] == "LISTENING":
+            pids.add(int(parts[-1]))
+    killed = 0
+    for pid in pids:
+        try:
+            subprocess.run(
+                f"taskkill /F /PID {pid}",
+                shell=True, text=True, timeout=5,
+                stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
+            )
+            print(f"  [CLEAN] 已清理脚本专属端口残留进程 PID={pid}")
+            killed += 1
+        except Exception:
+            pass
+    return killed
+
+
+def launch_chrome_with_debug(profile_dir: str | None = None):
+    """用指定 profile 启动 Chrome 并等待调试端口就绪。"""
+    profile_dir = profile_dir or USER_DATA_DIR
+    browser_path = find_browser_executable()
+    print(f"Chrome 调试端口未启动，正在启动浏览器: {browser_path}")
+    print(f"Profile 目录: {profile_dir}")
+    os.makedirs(profile_dir, exist_ok=True)
+
+    # 启 Chrome 前清理本脚本专属端口的旧进程。
+    if _kill_port_listeners(DEBUG_PORT):
+        time.sleep(0.5)
+
+    locked, lock_name = _profile_is_locked(profile_dir)
+    if locked:
+        print(f"  [STOP] Profile {profile_dir} 正被另一个 Chrome 占用（{lock_name}）。")
+        print(f"         这是因为『日常 Chrome』还在跑——它的 user-data-dir 跟脚本要用的相同。")
+        print(f"         若继续硬启，新 Chrome 会落到独立的子目录，老 cookie 复用不到 + 双窗。")
+        print(f"         解决方法：关掉你日常那个 Chrome 窗口，再跑本脚本。")
+        print(f"         脚本会启一个新 Chrome 复用你日常 profile 里已登录的 cookie，免登录。")
+        print(f"         （高级：日常 Chrome 启动时加 --remote-debugging-port={DEBUG_PORT}，")
+        print(f"           脚本可 cdp_url 接入而不启新 Chrome——一次只 1 个窗口。）")
+
+    if locked:
+        return False
+
+    cmd = [
+        browser_path,
+        f"--remote-debugging-port={DEBUG_PORT}",
+        f"--user-data-dir={profile_dir}",
+        "--no-first-run",
+        "--no-default-browser-check",
+        LOGIN_URL,
+    ]
+    proc_handle = subprocess.Popen(
+        cmd,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.PIPE,
+        creationflags=getattr(subprocess, "CREATE_NO_WINDOW", 0),
+    )
+    for i in range(20):
+        time.sleep(0.5)
+        if is_chrome_debug_running():
+            print("Chrome 调试服务已就绪")
+            return True
+    print("等待 Chrome 启动超时")
+    return False
+
+
+def get_active_page(context):
+    """获取浏览器中一个有效的活跃页面"""
+    for p in reversed(context.pages):
+        try:
+            _ = p.url
+            return p
+        except Exception:
+            continue
+    return None
+
+
+def get_websocket_url():
+    try:
+        import json
+        with urllib.request.urlopen(f"http://127.0.0.1:{DEBUG_PORT}/json/version", timeout=2) as resp:
+            data = json.loads(resp.read().decode())
+            return data.get("webSocketDebuggerUrl")
+    except Exception:
+        return None
+
+
+FALLBACK_USER_DATA_DIR = str(profile_path("tm"))
+ALLOW_FALLBACK_PROFILE = os.getenv("TM_ALLOW_FALLBACK_PROFILE", "0") == "1"
+
+
+def _try_launch_with_profile(profile_dir: str) -> bool:
+    """按指定 profile 启动 Chrome 并等调试端口就绪。"""
+    if not is_chrome_debug_running():
+        if not launch_chrome_with_debug(profile_dir):
+            return False
+        time.sleep(2)
+    return True
+
+
+def connect_browser(cdp_url: str | None = None):
+    """启动 Chrome 并返回 session, page。
+
+    使用 StealthySession + launch_persistent_context，和 backfill 脚本一致：
+    - cookie 直接从 profile 目录复用，无需额外同步
+    - session.close() 会杀掉 Chrome 进程，不会残留两个浏览器窗口
+    - 每个品牌 profile 自己维护登录态，手动登录一次后 cookie 自动保存
+    """
+    if cdp_url:
+        print(f"[CDP] 使用脚本专属 cdp_url 接管: {cdp_url}（不启新 Chrome）")
+        from scrapling.fetchers import StealthySession
+
+        session = StealthySession(cdp_url=cdp_url)
+        session.start()
+        context = session.context
+        page = get_active_page(context)
+        if not page:
+            page = context.new_page()
+        return session, page
+
+    # 先清理本脚本专属端口残留的 Chrome。
+    killed = _kill_port_listeners(DEBUG_PORT)
+    if killed:
+        time.sleep(1.0)
+        print(f"[CLEAN] 已清理脚本专属端口残留 Chrome ({killed} 个进程)")
+
+    from scrapling.fetchers import StealthySession
+
+    print(f"[BROWSER] 启 Chrome (profile={USER_DATA_DIR})")
+    session = StealthySession(
+        headless=False,
+        real_chrome=True,
+        user_data_dir=USER_DATA_DIR,
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=60000,
+        network_idle=False,
+    )
+    session.start()
+    context = session.context
+    page = context.pages[-1] if context.pages else context.new_page()
+    return session, page
+
+
+def _store_name_from_account(account: str) -> str:
+    """从账号中提取店铺名，如 '光影行星旗舰店:龙虾仔' -> '光影行星旗舰店'。"""
+    if not account:
+        return "光影行星旗舰店"
+    return account.split(":")[0].strip() or "光影行星旗舰店"
+
+
+def is_logged_in(page):
+    """
+    判断是否已登录。
+
+    单纯看 URL 含不含 "login" 不够准 —— 淘宝对未带 session 访问业务页会先 302 到 login.htm,
+    但如果 cookie 还在,会立刻自动跳回。脚本在这一瞬间读 url,会被骗到。
+
+    判定标准 (全部满足才算登录):
+      1. URL 不含 "login"
+      2. DOM 里出现当前账号店铺名 / 用户头像 / 顶部菜单 (商品/流量/客户 等)
+         —— 只要任一即可,因为生意参谋登录后顶部固定会显示店铺名。
+    """
+    try:
+        url_ok = "login" not in page.url.lower()
+        if not url_ok:
+            return False
+        store_name = _store_name_from_account(ACCOUNT)
+        # DOM 探针:登录后顶部会有店铺名/账号头像
+        dom_ok = page.evaluate(
+            r"""
+            (storeName) => {
+                const html = document.documentElement.outerHTML || '';
+                // 生意参谋登录后顶部固定显示店铺名 + 商品/流量/客户/交易 等 tab
+                const hasStoreName = storeName && new RegExp(storeName.replace(/[.*+?^${}()|[\]\\]/g, '\\$&')).test(html);
+                const hasTopNav = /商品排行|商品概览|品类分析|流量|交易|客户/.test(html);
+                // 也认 avatar / 退出 链接
+                const hasAvatar = !!document.querySelector(
+                    '[class*="avatar"], [class*="user-info"], [class*="logout"], [class*="退出"]'
+                );
+                return hasStoreName || hasTopNav || hasAvatar;
+            }
+            """,
+            store_name,
+        )
+        return bool(dom_ok)
+    except Exception:
+        return "login" not in page.url.lower()
+
+
+def find_logged_in_page(context):
+    """在所有页面中查找已登录的页面"""
+    for p in context.pages:
+        if "login" not in p.url.lower() and "sycm.taobao.com" in p.url.lower():
+            return p
+    return None
+
+
+def close_baxia_popup(page):
+    """关闭可能出现的空弹窗（用JS绕过遮罩）"""
+    try:
+        result = page.evaluate("""
+            () => {
+                const btn = document.querySelector('.baxia-dialog-close');
+                if (btn && btn.offsetParent !== null) { btn.click(); return true; }
+                return false;
+            }
+        """)
+        if result:
+            print("已关闭安全弹窗")
+            time.sleep(1)
+            return True
+    except Exception:
+        pass
+    return False
+
+
+def close_ad_popup(page):
+    """关闭广告弹窗（如618活动弹窗）和安全弹窗"""
+    closed = False
+    try:
+        result = page.evaluate("""
+            () => {
+                let closed = false;
+                const adBtn = document.querySelector('.sycm-active-egg-modal-content-close');
+                if (adBtn && adBtn.offsetParent !== null) { adBtn.click(); closed = true; }
+                const baxiaBtn = document.querySelector('.baxia-dialog-close');
+                if (baxiaBtn && baxiaBtn.offsetParent !== null) { baxiaBtn.click(); closed = true; }
+                return closed;
+            }
+        """)
+        if result:
+            print("已关闭弹窗")
+            time.sleep(1)
+            closed = True
+    except Exception:
+        pass
+    return closed
+
+
+def close_competition_reminder(page):
+    """关闭右上角的618竞争动态提醒，避免挡住日期选择器。"""
+    try:
+        closed = page.evaluate("""
+            () => {
+                const candidates = Array.from(document.querySelectorAll('div, section, aside'))
+                    .filter(el => el.offsetParent !== null && el.textContent.includes('竞争动态提醒'));
+                for (const box of candidates) {
+                    const closeSelectors = [
+                        '[aria-label="close"]',
+                        '[aria-label="Close"]',
+                        '[class*="close"]',
+                        'button',
+                    ];
+                    for (const selector of closeSelectors) {
+                        const btn = box.querySelector(selector);
+                        if (btn && btn.offsetParent !== null) {
+                            btn.click();
+                            return true;
+                        }
+                    }
+                    const rect = box.getBoundingClientRect();
+                    const el = document.elementFromPoint(rect.right - 18, rect.top + 18);
+                    if (el) {
+                        el.click();
+                        return true;
+                    }
+                    box.remove();
+                    return true;
+                }
+                return false;
+            }
+        """)
+        if closed:
+            print("已关闭618竞争动态提醒")
+            time.sleep(1)
+            return True
+    except Exception:
+        pass
+    return False
+
+
+def close_blocking_tips(page):
+    """清理会遮挡日期按钮的页面浮层。"""
+    try:
+        removed = page.evaluate("""
+            () => {
+                let count = 0;
+                const selectors = [
+                    '.sycm-base-search-result-v2-ai-content',
+                    '.ant-popover',
+                    '.ant-tooltip',
+                    '[class*="popover"]',
+                    '[class*="tooltip"]'
+                ];
+                for (const selector of selectors) {
+                    for (const el of document.querySelectorAll(selector)) {
+                        if (!el.offsetParent) continue;
+                        const text = el.textContent || '';
+                        if (
+                            text.includes('智能诊断') ||
+                            text.includes('竞争动态提醒') ||
+                            text.includes('您关注的竞品') ||
+                            text.includes('抢拉新客')
+                        ) {
+                            const root = el.closest('[class*="popover"], [class*="tooltip"], [class*="overlay"], .ant-popover, .ant-tooltip') || el;
+                            root.remove();
+                            count += 1;
+                        }
+                    }
+                }
+                return count;
+            }
+        """)
+        if removed:
+            print(f"已清理遮挡浮层 {removed} 个")
+            time.sleep(0.5)
+            return True
+    except Exception:
+        pass
+    return False
+
+
+def do_login(page, context):
+    """自动填写账号密码并点击登录，等待用户手动完成滑块验证"""
+    print("等待登录 iframe 加载...")
+    page.wait_for_selector("#alibaba-login-box", timeout=15000)
+    iframe_element = page.locator("#alibaba-login-box").element_handle()
+    login_frame = iframe_element.content_frame()
+    if not login_frame:
+        print("错误：未能获取登录 iframe 的 Frame 对象")
+        return None
+
+    login_frame.wait_for_selector('input[placeholder="账号名/邮箱/手机号"]', timeout=30000)
+
+    print("正在输入账号...")
+    login_frame.fill('input[placeholder="账号名/邮箱/手机号"]', ACCOUNT)
+    time.sleep(0.5)
+
+    print("正在输入密码...")
+    login_frame.fill('input[placeholder="请输入登录密码"]', PASSWORD)
+    time.sleep(0.5)
+
+    print("正在点击登录按钮...")
+    login_frame.click('button:has-text("登录")')
+    time.sleep(1)
+
+    print("\n" + "=" * 50)
+    print("请手动完成滑块验证")
+    print("=" * 50 + "\n")
+
+    print("等待登录完成...")
+    for _ in range(300):
+        time.sleep(1)
+        close_baxia_popup(page)
+        if is_logged_in(page):
+            print(f"登录成功！当前页面: {page.url}")
+            return page
+        logged_page = find_logged_in_page(context)
+        if logged_page:
+            print(f"在其他标签页检测到已登录: {logged_page.url}")
+            return logged_page
+
+    print("等待 5 分钟后仍未检测到登录成功")
+    return None
+
+
+def ensure_logged_in(context, page):
+    """
+    检测已登录态则直接复用；否则强制走一次登录流程。
+
+    关键修复:URL 命中 login 不等于未登录 —— 淘宝对未带 session 访问业务页会先 302 到
+    login.htm,如果 cookie 还在会立刻自动跳回。脚本在这一瞬间读 url,会被骗到。
+    所以拿到 "URL 含 login" 后,先等 5 秒看是否会跳回 + 用 DOM 探针二次确认。
+    """
+    print("检查天猫登录态...")
+    try:
+        page.goto(PRODUCTS_URL, wait_until="domcontentloaded", timeout=60000)
+        # 等 5 秒让可能的自动跳转完成
+        for _ in range(10):
+            if "login" not in page.url.lower():
+                break
+            time.sleep(0.5)
+        time.sleep(1)
+        if is_logged_in(page):
+            print(f"[OK] 检测到天猫已登录态,跳过登录页: {page.url}")
+            return page
+        # 二次尝试:有些情况下 cookie 是 session-only,需要先访问 havanalogin 续期
+        if "login" in page.url.lower():
+            print("  [INFO] 仍停留在 login 页,尝试访问 havanalogin 续期 cookie...")
+            try:
+                page.goto("https://havanalogin.taobao.com/", wait_until="domcontentloaded", timeout=30000)
+                time.sleep(2)
+                # 再回到业务页看是否登录
+                page.goto(PRODUCTS_URL, wait_until="domcontentloaded", timeout=60000)
+                for _ in range(10):
+                    if "login" not in page.url.lower():
+                        break
+                    time.sleep(0.5)
+                time.sleep(1)
+                if is_logged_in(page):
+                    print(f"[OK] havanalogin 续期后检测到已登录: {page.url}")
+                    return page
+            except Exception as e:
+                print(f"  [WARN] havanalogin 续期失败: {e}")
+    except Exception as e:
+        print(f"  [WARN] 登录态预检失败（{e}）,回退到强制登录流程")
+
+    print("强制走登录流程...")
+    print(f"  导航到登录页: {LOGIN_URL}")
+    page.goto(LOGIN_URL, wait_until="domcontentloaded", timeout=60000)
+    time.sleep(2)
+
+    print(f"  账号: {ACCOUNT}")
+    print(f"  密码: {'*' * len(PASSWORD)}")
+
+    logged_page = do_login(page, context)
+    if not logged_page:
+        print("登录失败，脚本退出")
+        return None
+    return logged_page
+
+
+def navigate_to_products(page):
+    """导航到商品页面"""
+    print(f"正在导航到商品页面...")
+    page.goto(PRODUCTS_URL, wait_until="networkidle", timeout=60000)
+    time.sleep(3)
+    close_ad_popup(page)
+    close_competition_reminder(page)
+    close_blocking_tips(page)
+    close_baxia_popup(page)
+    print(f"当前页面: {page.url}")
+
+
+def get_yesterday():
+    """计算昨天日期。"""
+    return TARGET_DATE or (datetime.now() - timedelta(days=1))
+
+
+def parse_target_date(value):
+    try:
+        return datetime.strptime(value, "%Y-%m-%d")
+    except ValueError as exc:
+        raise argparse.ArgumentTypeError("target date must be YYYY-MM-DD") from exc
+
+
+def align_calendar_month(page, target_date):
+    for _ in range(18):
+        info = page.evaluate(
+            """
+            ({targetYear, targetMonth}) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none' && style.visibility !== 'hidden'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const panels = Array.from(document.querySelectorAll(
+                    '.ant-calendar, .ant-picker-panel, [class*="calendar"], [class*="picker-panel"], [class*="dropdown"]'
+                )).filter(el => visible(el) && /(20\\d{2})\\s*年\\s*(\\d{1,2})\\s*月/.test(el.textContent || ''));
+                const root = panels[0] || document;
+                const match = (root.textContent || '').match(/(20\\d{2})\\s*年\\s*(\\d{1,2})\\s*月/);
+                if (!match) return {status: 'no_header'};
+                const currentYear = Number(match[1]);
+                const currentMonth = Number(match[2]);
+                const diff = (targetYear - currentYear) * 12 + (targetMonth - currentMonth);
+                if (diff === 0) return {status: 'aligned'};
+                const selectors = diff < 0
+                    ? ['.ant-calendar-prev-month-btn', '.ant-picker-header-prev-btn', '[class*="prev"]']
+                    : ['.ant-calendar-next-month-btn', '.ant-picker-header-next-btn', '[class*="next"]'];
+                for (const selector of selectors) {
+                    const btn = Array.from(root.querySelectorAll(selector)).find(visible)
+                        || Array.from(document.querySelectorAll(selector)).find(visible);
+                    if (btn) {
+                        const rect = btn.getBoundingClientRect();
+                        return {status: 'click', x: rect.left + rect.width / 2, y: rect.top + rect.height / 2, diff};
+                    }
+                }
+                return {status: 'no_nav', currentYear, currentMonth, diff};
+            }
+            """,
+            {"targetYear": target_date.year, "targetMonth": target_date.month},
+        )
+        if not isinstance(info, dict) or info.get("status") == "aligned":
+            return
+        if info.get("status") != "click":
+            print(f"日历月份翻页停止: {info}")
+            return
+        page.mouse.click(info["x"], info["y"])
+        time.sleep(0.6)
+
+
+def select_yesterday(page):
+    """
+    关闭提醒后悬浮在“日”按钮上显示日期面板，然后选择昨天。
+    """
+    yesterday = get_yesterday()
+    target_date_str = yesterday.strftime("%Y-%m-%d")
+    target_day = str(yesterday.day)
+    print(f"目标日期（昨天）: {target_date_str}")
+
+    # 先移除可能的遮罩层
+    page.evaluate("""
+        () => {
+            const masks = document.querySelectorAll('.baxia-dialog-mask, .sycm-active-egg-modal-mask');
+            masks.forEach(m => m.parentElement && m.parentElement.remove());
+        }
+    """)
+    time.sleep(0.5)
+    close_ad_popup(page)
+    close_competition_reminder(page)
+    close_blocking_tips(page)
+
+    # 1. 找到"日"按钮并悬浮，显示日期面板。
+    print("正在悬浮'日'按钮以显示日期面板...")
+    day_btn = page.locator('.oui-date-picker-particle-button button:has-text("日")').first
+    found = False
+    for attempt in range(3):
+        found = page.evaluate("""
+            () => {
+                const btns = document.querySelectorAll('.oui-date-picker-particle-button button');
+                for (const btn of btns) {
+                    if (btn.textContent.trim() === '日') {
+                        btn.scrollIntoView({block: 'center'});
+                        btn.dispatchEvent(new MouseEvent('mouseover', { bubbles: true, view: window }));
+                        btn.dispatchEvent(new MouseEvent('mouseenter', { bubbles: true, view: window }));
+                        return true;
+                    }
+                }
+                return false;
+            }
+        """)
+        if found:
+            break
+        if attempt < 2:
+            print(f"  [retry {attempt + 1}/2] 未找到'日'按钮，关闭弹窗后重试...")
+            close_competition_reminder(page)
+            close_blocking_tips(page)
+            close_ad_popup(page)
+            time.sleep(2)
+    if not found:
+        print("未找到'日'按钮")
+        return False
+
+    time.sleep(2)  # 等待日期面板出现
+    close_competition_reminder(page)
+    close_blocking_tips(page)
+
+    # 如果悬浮没有展开日历，再点击一次“日”按钮。
+    has_calendar = page.evaluate("""
+        () => !!document.querySelector('.ant-calendar, .ant-picker-panel, [class*="calendar"], [class*="picker-panel"]')
+    """)
+    if not has_calendar:
+        print("悬浮后未检测到日期面板，尝试点击'日'按钮...")
+        clicked_day = page.evaluate("""
+                () => {
+                    const btns = document.querySelectorAll('.oui-date-picker-particle-button button');
+                    for (const btn of btns) {
+                        if (btn.textContent.trim() === '日') { btn.click(); return true; }
+                    }
+                    return false;
+                }
+            """)
+        if not clicked_day and day_btn.count() > 0:
+            day_btn.click(force=True)
+        time.sleep(2)
+
+    print(f"正在点击日期: {target_date_str}")
+
+    # 尝试多种方式点击日期
+    align_calendar_month(page, yesterday)
+    time.sleep(0.5)
+
+    clicked = page.evaluate(
+        """
+        ({targetDate, targetDay}) => {
+            // 方式1: 通过 title 属性或文本内容匹配
+            const cells = document.querySelectorAll('.ant-calendar-cell, .ant-picker-cell, [class*="calendar-cell"], [class*="picker-cell"], td');
+            for (const cell of cells) {
+                const title = cell.getAttribute('title');
+                const aria = cell.getAttribute('aria-label') || '';
+                const dataValue = cell.getAttribute('data-value') || '';
+                if (title === targetDate || aria.includes(targetDate) || dataValue.includes(targetDate)) {
+                    const clickable = cell.querySelector('div') || cell;
+                    clickable.click();
+                    return true;
+                }
+            }
+
+            // 方式2: 直接找日期面板里包含目标日数字的可点元素。
+            const panels = Array.from(document.querySelectorAll('.ant-calendar, .ant-picker-panel, [class*="calendar"], [class*="picker-panel"]'))
+                .filter(el => el.offsetParent !== null);
+            const roots = panels.length ? panels : [document.body];
+            for (const root of roots) {
+                const allEls = Array.from(root.querySelectorAll('*'));
+                for (const el of allEls) {
+                    if (el.children.length === 0 && el.textContent.trim() === targetDay) {
+                        const parent = el.closest('td') || el.closest('[class*="cell"]') || el.parentElement;
+                        const disabledText = (parent && parent.className ? String(parent.className) : '') + ' ' + (el.className ? String(el.className) : '');
+                        if (parent && !disabledText.includes('disabled')) {
+                            parent.click();
+                            return true;
+                        }
+                    }
+                }
+            }
+            return false;
+        }
+        """,
+        {"targetDate": target_date_str, "targetDay": target_day},
+    )
+
+    if clicked:
+        print("已点击昨天日期")
+        time.sleep(3)
+        return True
+
+    print("未能点击到昨天日期，尝试备用方案：直接设置URL日期参数")
+    try:
+        page.goto(
+            f"{PRODUCTS_URL}?dateRange={target_date_str}%7C{target_date_str}&dateType=today",
+            wait_until="networkidle",
+            timeout=60000,
+        )
+        time.sleep(3)
+        close_ad_popup(page)
+        close_competition_reminder(page)
+        return True
+    except Exception as e:
+        print(f"备用日期跳转失败: {e}")
+        return False
+
+
+def safe_filename(name):
+    invalid = '<>:"/\\|?*'
+    cleaned = "".join("_" if ch in invalid else ch for ch in name).strip()
+    return cleaned or "single_product_data.xlsx"
+
+
+def unique_output_path(directory, filename):
+    directory.mkdir(parents=True, exist_ok=True)
+    path = directory / safe_filename(filename)
+    if not path.suffix:
+        path = path.with_suffix(".xlsx")
+    if not path.exists():
+        return path
+
+    stem = path.stem
+    suffix = path.suffix
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    return directory / f"{stem}_{timestamp}{suffix}"
+
+
+def _download_target_dir() -> Path:
+    """下载目录: data/tm单品数据/_downloads/<YYYY-MM-DD>/ (与 jd/dy 一致)。"""
+    date_stamp = (TARGET_DATE or (datetime.now() - timedelta(days=1))).strftime("%Y-%m-%d")
+    target = DOWNLOADS_DIR / date_stamp
+    target.mkdir(parents=True, exist_ok=True)
+    return target
+
+
+def click_download(page):
+    """点击下载按钮，保存下载出的Excel文件，返回本地路径。
+
+    强化点（2026-06-10 修）:
+      - 入口先清掉所有已知弹窗 (618 营销 / 竞争动态提醒 / 智能诊断 / baxia 安全弹窗)
+      - 失败时再用 JS 强清遮罩层, 二次重试
+    """
+    print("正在点击下载按钮...")
+    # 入口清掉所有已知弹窗 (之前只 close_ad_popup 一个, 漏了另外 3 个)
+    close_ad_popup(page)
+    close_competition_reminder(page)
+    close_blocking_tips(page)
+    close_baxia_popup(page)
+
+    # 先尝试通过文本匹配
+    try:
+        download_link = page.locator('a.sycm-traced-download-text:has-text("下载")').first
+        if download_link.count() > 0:
+            download_link.scroll_into_view_if_needed()
+            time.sleep(0.5)
+            with page.expect_download(timeout=60000) as download_info:
+                download_link.click()
+            download = download_info.value
+            saved_path = unique_output_path(_download_target_dir(), download.suggested_filename)
+            download.save_as(str(saved_path))
+            print("已点击下载")
+            print(f"Excel已保存: {saved_path}")
+            return saved_path
+    except Exception as e:
+        print(f"通过选择器点击下载失败: {e}")
+
+    # 备用：JS点击
+    try:
+        with page.expect_download(timeout=60000) as download_info:
+            clicked = page.evaluate("""
+                () => {
+                    const links = document.querySelectorAll('a.sycm-traced-download-text, a.sycm-cc-item-rank-download');
+                    for (const link of links) {
+                        if (link.textContent.trim() === '下载' && link.offsetParent !== null) {
+                            link.click();
+                            return true;
+                        }
+                    }
+                    // 更宽泛的搜索
+                    const allLinks = document.querySelectorAll('a');
+                    for (const link of allLinks) {
+                        if (link.textContent.trim() === '下载' && link.offsetParent !== null) {
+                            link.click();
+                            return true;
+                        }
+                    }
+                    return false;
+                }
+            """)
+            if not clicked:
+                raise RuntimeError("未找到下载按钮")
+        download = download_info.value
+        saved_path = unique_output_path(_download_target_dir(), download.suggested_filename)
+        download.save_as(str(saved_path))
+        print("已点击下载（JS方式）")
+        print(f"Excel已保存: {saved_path}")
+        return saved_path
+    except Exception as e:
+        print(f"JS方式下载失败: {e}")
+
+    # 最后兜底：强清所有遮罩层（广告/弹窗/确认框），再点一次
+    try:
+        print("尝试强清遮罩层后重试...")
+        page.evaluate("""
+            () => {
+                const selectors = [
+                    '.baxia-dialog-mask', '.baxia-dialog-wrapper', '.baxia-dialog',
+                    '.sycm-active-egg-modal-mask', '.sycm-active-egg-modal',
+                    '.ant-modal-mask', '.ant-modal-wrap', '.ant-modal',
+                    '[class*="popup"]', '[class*="popover"]', '[class*="tooltip"]',
+                    '[class*="Modal"]', '[class*="dialog"]', '[class*="Drawer"]',
+                ];
+                for (const sel of selectors) {
+                    document.querySelectorAll(sel).forEach(el => {
+                        try { el.remove(); } catch (e) {}
+                    });
+                }
+                // 移除 fixed 定位的遮罩
+                document.querySelectorAll('div').forEach(el => {
+                    const s = window.getComputedStyle(el);
+                    if (s.position === 'fixed' && parseFloat(s.zIndex || '0') >= 100) {
+                        try { el.remove(); } catch (e) {}
+                    }
+                });
+            }
+        """)
+        time.sleep(1)
+        close_ad_popup(page)
+        close_competition_reminder(page)
+        close_blocking_tips(page)
+        with page.expect_download(timeout=60000) as download_info:
+            page.evaluate("""
+                () => {
+                    const allLinks = document.querySelectorAll('a');
+                    for (const link of allLinks) {
+                        if (link.textContent.trim() === '下载' && link.offsetParent !== null) {
+                            link.click();
+                            return true;
+                        }
+                    }
+                }
+            """)
+        download = download_info.value
+        saved_path = unique_output_path(_download_target_dir(), download.suggested_filename)
+        download.save_as(str(saved_path))
+        print("强清后已点击下载")
+        print(f"Excel已保存: {saved_path}")
+        return saved_path
+    except Exception as e:
+        print(f"强清后仍下载失败: {e}")
+
+    print("未找到下载按钮或下载未成功")
+    return None
+
+
+def normalize_cell_value(value):
+    if value is None:
+        return ""
+    if hasattr(value, "isoformat") and not isinstance(value, str):
+        return value.isoformat()
+    return str(value).strip()
+
+
+def trim_matrix(matrix):
+    rows = [[normalize_cell_value(cell) for cell in row] for row in matrix]
+    rows = [row for row in rows if any(cell for cell in row)]
+    if not rows:
+        return []
+
+    last_col = 0
+    for row in rows:
+        for idx, cell in enumerate(row):
+            if cell:
+                last_col = max(last_col, idx + 1)
+    return [row[:last_col] + [""] * max(0, last_col - len(row)) for row in rows]
+
+
+def make_unique_headers(raw_headers):
+    headers = []
+    seen = {}
+    for idx, header in enumerate(raw_headers, start=1):
+        name = header or f"列{idx}"
+        if name in seen:
+            seen[name] += 1
+            name = f"{name}_{seen[name]}"
+        else:
+            seen[name] = 1
+        headers.append(name)
+    return headers
+
+
+def sheet_matrix_to_records(sheet_name, matrix):
+    matrix = trim_matrix(matrix)
+    if not matrix:
+        return {"sheet": sheet_name, "headers": [], "rows": []}
+
+    header_index = 0
+    for idx, row in enumerate(matrix):
+        cells = set(row)
+        if "商品ID" in cells and "支付件数" in cells:
+            header_index = idx
+            break
+    headers = make_unique_headers(matrix[header_index])
+    rows = []
+    for row in matrix[header_index + 1:]:
+        if not any(row):
+            continue
+        rows.append({headers[i]: row[i] if i < len(row) else "" for i in range(len(headers))})
+    return {"sheet": sheet_name, "headers": headers, "rows": rows}
+
+
+def read_excel_as_sheets(excel_path):
+    excel_path = Path(excel_path)
+    try:
+        workbook = load_workbook(excel_path, data_only=True, read_only=True)
+        sheets = []
+        for worksheet in workbook.worksheets:
+            matrix = list(worksheet.iter_rows(values_only=True))
+            sheets.append(sheet_matrix_to_records(worksheet.title, matrix))
+        workbook.close()
+        return sheets
+    except Exception as exc:
+        print(f"openpyxl解析失败，尝试pandas读取: {exc}")
+
+    try:
+        import pandas as pd
+
+        frames = pd.read_excel(excel_path, sheet_name=None, dtype=str, header=None)
+        sheets = []
+        for sheet_name, frame in frames.items():
+            frame = frame.fillna("")
+            matrix = frame.values.tolist()
+            sheets.append(sheet_matrix_to_records(sheet_name, matrix))
+        return sheets
+    except Exception as exc:
+        raise RuntimeError(f"无法解析Excel文件: {excel_path}") from exc
+
+
+def md_escape(value):
+    text = normalize_cell_value(value)
+    return text.replace("|", "\\|").replace("\r", " ").replace("\n", "<br>")
+
+
+def export_excel_to_json_and_md(excel_path):
+    excel_path = Path(excel_path)
+    sheets = read_excel_as_sheets(excel_path)
+    exported_at = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    payload = {
+        "source_file": str(excel_path),
+        "exported_at": exported_at,
+        "sheets": sheets,
+    }
+
+    json_path = excel_path.with_suffix(".json")
+    md_path = excel_path.with_suffix(".md")
+
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(payload, f, ensure_ascii=False, indent=2)
+
+    lines = [
+        f"# 单品数据 - {excel_path.stem}",
+        "",
+        f"- 源文件: {excel_path.name}",
+        f"- 导出时间: {exported_at}",
+        "",
+    ]
+    for sheet in sheets:
+        lines.append(f"## {sheet['sheet']}")
+        lines.append("")
+        headers = sheet["headers"]
+        rows = sheet["rows"]
+        if not headers:
+            lines.append("（空表）")
+            lines.append("")
+            continue
+        lines.append("| " + " | ".join(md_escape(header) for header in headers) + " |")
+        lines.append("|" + "|".join(["---"] * len(headers)) + "|")
+        for row in rows:
+            lines.append("| " + " | ".join(md_escape(row.get(header, "")) for header in headers) + " |")
+        lines.append("")
+
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(lines))
+
+    print(f"JSON已导出: {json_path}")
+    print(f"Markdown已导出: {md_path}")
+    return json_path, md_path
+
+
+def parse_number(value):
+    text = normalize_cell_value(value)
+    if not text or text == "-":
+        return 0.0
+    text = text.replace(",", "").replace("%", "").strip()
+    try:
+        return float(text)
+    except ValueError:
+        return 0.0
+
+
+def format_number(value):
+    if abs(value - round(value)) < 1e-9:
+        return str(int(round(value)))
+    return f"{value:.2f}"
+
+
+def format_rate(value):
+    return f"{value:.2f}%"
+
+
+def find_product_rows(sheets):
+    for sheet in sheets:
+        headers = set(sheet.get("headers", []))
+        if {"商品ID", "商品访客数", "商品加购人数"}.issubset(headers):
+            return sheet.get("rows", [])
+    return []
+
+
+def get_report_date(rows, excel_path):
+    for row in rows:
+        date_value = normalize_cell_value(row.get("统计日期", ""))
+        if date_value:
+            return date_value[:10]
+    stem = Path(excel_path).stem
+    parts = stem.split("_")
+    for part in parts:
+        if len(part) == 10 and part[4] == "-" and part[7] == "-":
+            return part
+    return datetime.now().strftime("%Y-%m-%d")
+
+
+def summarize_styles_from_excel(excel_path):
+    excel_path = Path(excel_path)
+    sheets = read_excel_as_sheets(excel_path)
+    product_rows = find_product_rows(sheets)
+    if not product_rows:
+        print("未找到包含商品指标的表，跳过款式汇总")
+        return None, None
+
+    rows_by_id = {}
+    for row in product_rows:
+        product_id = normalize_cell_value(row.get("商品ID", ""))
+        if product_id:
+            rows_by_id[product_id] = row
+
+    report_date = get_report_date(product_rows, excel_path)
+    summaries = []
+    style_id_map = _get_style_id_map()
+    if BRAND:
+        brand_map = get_tm_brand_map()
+        all_styles = set(style_id_map.keys())
+        brand_styles = {s for s in all_styles if brand_map.get(s, "光影行星") == BRAND}
+        style_id_map = {s: ids for s, ids in style_id_map.items() if s in brand_styles}
+        print(f"[SUMMARY] brand={BRAND}, 只处理 {len(style_id_map)} 款 (避免覆盖其他品牌数据)")
+    for style_name, product_ids in style_id_map.items():
+        matched_rows = [rows_by_id[pid] for pid in product_ids if pid in rows_by_id]
+        matched_ids = [normalize_cell_value(row.get("商品ID", "")) for row in matched_rows]
+        missing_ids = [pid for pid in product_ids if pid not in rows_by_id]
+
+        total_visitors = sum(parse_number(row.get("商品访客数", "")) for row in matched_rows)
+        total_cart_users = sum(parse_number(row.get("商品加购人数", "")) for row in matched_rows)
+        erp_metrics = apply_erp_override("tm", style_name, total_visitors, total_cart_users, target_date=report_date)
+
+        summaries.append({
+            "款式": style_name,
+            "日期": report_date,
+            "商品ID数": len(product_ids),
+            "匹配商品ID数": len(matched_ids),
+            "匹配商品ID": matched_ids,
+            "未匹配商品ID": missing_ids,
+            "销量": erp_metrics["销量"],
+            "访客量": total_visitors,
+            "加购人数": total_cart_users,
+            "退款订单数": erp_metrics["退款订单数"],
+            "转化率": parse_metric_number(erp_metrics["转化率"]),
+            "加购率": parse_metric_number(erp_metrics["加购率"]),
+            "退款率": parse_metric_number(erp_metrics["退款率"]),
+            "ERP销量退款覆盖": erp_metrics["erp_override"],
+        })
+
+    base_name = f"业务款式_{report_date}"
+    metric_notes = {
+        "销量": "ERP昨日销量",
+        "访客量": "商品访客数之和",
+        "加购人数": "商品加购人数之和",
+        "退款订单数": "ERP昨日退货量",
+        "转化率": "销量 / 访客量 * 100%",
+        "加购率": "加购人数 / 访客量 * 100%",
+        "退款率": "退款订单数 / 销量 * 100%",
+    }
+
+    exported_files = []
+    for row in summaries:
+        style_dir = SINGLE_PRODUCT_DATA_DIR / f"{safe_filename(row['款式'])}_{report_date}"
+        style_dir.mkdir(parents=True, exist_ok=True)
+        json_path = style_dir / f"{base_name}.json"
+        md_path = style_dir / f"{base_name}.md"
+
+        payload = {
+            "source_file": str(excel_path),
+            "date": report_date,
+            "generated_at": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+            "metrics": metric_notes,
+            "style": row,
+        }
+        with open(json_path, "w", encoding="utf-8") as f:
+            json.dump(payload, f, ensure_ascii=False, indent=2)
+
+        lines = [
+            f"# {row['款式']} - 业务款式汇总 - {report_date}",
+            "",
+            f"- 源文件: {excel_path.name}",
+            f"- 匹配ID数: {row['匹配商品ID数']}/{row['商品ID数']}",
+            f"- 匹配商品ID: {', '.join(row['匹配商品ID']) if row['匹配商品ID'] else '-'}",
+            f"- 未匹配商品ID: {', '.join(row['未匹配商品ID']) if row['未匹配商品ID'] else '-'}",
+            "",
+            "| 款式 | 销量 | 访客量 | 转化率 | 加购人数 | 加购率 | 退款订单数 | 退款率 |",
+            "|---|---:|---:|---:|---:|---:|---:|---:|",
+            "| "
+            + " | ".join([
+                md_escape(row["款式"]),
+                format_number(row["销量"]),
+                format_number(row["访客量"]),
+                format_rate(row["转化率"]),
+                format_number(row["加购人数"]),
+                format_rate(row["加购率"]),
+                format_number(row["退款订单数"]),
+                format_rate(row["退款率"]),
+            ])
+            + " |",
+        ]
+
+        with open(md_path, "w", encoding="utf-8") as f:
+            f.write("\n".join(lines))
+
+        exported_files.append((json_path, md_path))
+        print(f"款式 {row['款式']} JSON已导出: {json_path}")
+        print(f"款式 {row['款式']} Markdown已导出: {md_path}")
+
+    return exported_files
+
+
+def main():
+    global TARGET_DATE, USER_DATA_DIR, ALLOW_FALLBACK_PROFILE, ACCOUNT, PASSWORD, BRAND
+    parser = argparse.ArgumentParser(description="Taobao SYCM product data collection")
+    parser.add_argument("--target-date", type=parse_target_date, default=None, help="采集指定日期，格式 YYYY-MM-DD；默认昨天")
+    parser.add_argument("--user-data-dir", default=USER_DATA_DIR, help="Chrome 持久 profile 目录，用于复用已登录 cookie")
+    parser.add_argument("--allow-fallback-profile", action="store_true", help="允许目标 profile 启动失败后切换到隔离 profile（可能需要重新登录）")
+    parser.add_argument("--cdp-url", default=None, help="接管由 GYXX 分配的脚本专属 Chrome 调试 URL")
+    parser.add_argument("--account", default=ACCOUNT, help="登录账号，覆盖环境变量 SYCM_ACCOUNT")
+    parser.add_argument("--password", default=PASSWORD, help="登录密码，覆盖环境变量 SYCM_PASSWORD")
+    parser.add_argument("--brand", default=None, help="当前采集品牌 (光影行星/ozko), summarize_styles_from_excel 只处理该品牌款式, 避免跨品牌覆盖")
+    args = parser.parse_args()
+    TARGET_DATE = args.target_date
+    USER_DATA_DIR = args.user_data_dir
+    ALLOW_FALLBACK_PROFILE = ALLOW_FALLBACK_PROFILE or args.allow_fallback_profile
+    ACCOUNT = args.account
+    PASSWORD = args.password
+    BRAND = args.brand
+
+    session, page = connect_browser(cdp_url=args.cdp_url)
+    if not session or not page:
+        print("连接 Chrome 失败")
+        return
+
+    context = session.context
+    page = ensure_logged_in(context, page)
+    if not page:
+        if session:
+            session.close()
+        _kill_port_listeners(DEBUG_PORT)
+        return
+
+    # 进入商品页面
+    navigate_to_products(page)
+
+    # 选择昨天
+    if not select_yesterday(page):
+        print("\n未能选择昨天日期，停止下载，避免保存错误日期的数据")
+        session.close()
+        _kill_port_listeners(DEBUG_PORT)
+        return
+
+    # 点击下载并解析
+    excel_path = click_download(page)
+    if excel_path:
+        export_excel_to_json_and_md(excel_path)
+        summarize_styles_from_excel(excel_path)
+        print(f"\n脚本执行完毕，单品数据已保存到: {SINGLE_PRODUCT_DATA_DIR}")
+    else:
+        print("\n脚本执行完毕，但未获取到下载文件")
+
+    session.close()
+    # session.close() 通过 CDP 关闭 Browser 对象，但不会杀掉 Popen 启动的 Chrome 进程。
+    # 显式清理本脚本专属端口上的残留 Chrome。
+    killed = _kill_port_listeners(DEBUG_PORT)
+    if killed:
+        time.sleep(0.5)
+        print(f"[CLEAN] 已清理脚本专属端口残留 Chrome ({killed} 个进程)")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/taobao_wanxiang_ai_creative_report.py b/src/gyxx_flow/modules/product_commerce/runtime/taobao_wanxiang_ai_creative_report.py
new file mode 100644
index 0000000..a721f96
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/taobao_wanxiang_ai_creative_report.py
@@ -0,0 +1,1704 @@
+# -*- coding: utf-8 -*-
+"""
+淘宝卖家中心 -> 万相 AI 无界创意报表采集脚本。
+
+采集逻辑：
+1. 使用 Scrapling StealthySession 打开淘宝卖家登录页并复用登录态；
+2. 进入卖家中心推广页，点击“前往万相AI无界”，关闭广告弹窗；
+3. 打开创意报表页；
+4. 按款式名逐个在“主体名称/ID”维度搜索；
+5. 从当前页 HTML 提取每页创意主图、创意ID、展现量、点击量、花费；
+6. 翻页直到最后一页，并按创意ID合并汇总。
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+from runtime_paths import RAW_DATA_ROOT, TMP_ROOT, managed_data_path, profile_path
+from typing import Any
+from urllib.parse import quote
+from urllib.request import Request, urlopen
+
+from PIL import Image, ImageOps
+from scrapling.fetchers import StealthyFetcher, StealthySession
+
+from config.style_config_loader import get_bitable_main_image_map, get_main_image_styles, get_tm_style_id_map
+from main_image_paths import collect_images_dir
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+
+LOGIN_URL = "https://loginmyseller.taobao.com/"
+SELLER_SUBWAY_URL = "https://myseller.taobao.com/home.htm/subway_new/"
+WANXIANG_REPORT_URL = (
+    "https://one.alimama.com/index.html?spm=a21dvs.28495141.cc92828ac.dbf3edac0.778454dfoBlQB4"
+    "#!/report/creative?rptType=creative"
+)
+WANXIANG_BASE_URL = "https://one.alimama.com/index.html"
+WANXIANG_REPORT_HASH = "#!/report/creative?rptType=creative"
+REPORT_TEMPLATE_LABELS = ("报表模板", "报表模版")
+REPORT_TEMPLATE_OPTIONS = ("全部计划", "关键词推广", "人群推广")
+DEFAULT_REPORT_TEMPLATE = "关键词推广"
+
+ACCOUNT = os.getenv("WANXIANG_ACCOUNT", os.getenv("SYCM_ACCOUNT", "光影行星旗舰店:龙虾仔"))
+PASSWORD = os.getenv("WANXIANG_PASSWORD", os.getenv("SYCM_PASSWORD", "aa123456"))
+USER_DATA_DIR = os.getenv(
+    "WANXIANG_USER_DATA_DIR", str(profile_path("wanxiang"))
+)
+OUTPUT_DIR = RAW_DATA_ROOT / "主图信息"
+DEBUG = os.getenv("WANXIANG_DEBUG", "0") == "1"
+IMAGE_HASH_DISTANCE = int(os.getenv("WANXIANG_IMAGE_HASH_DISTANCE", "6"))
+
+# 保留硬编码作为灾备回退；运行时优先从飞书多维表格读取。
+_STATIC_STYLE_NAMES = [
+    "盖亚斜挎",
+    "极星pro",
+    "盖亚微单",
+    "盖世m1",
+    "星迹2",
+    "星云2",
+    "宙斯",
+    "阿波罗2",
+    "阿波罗x1",
+    "波塞冬edge",
+    "逐星",
+    "逐星GT",
+    "星云mini",
+    "晨星2",
+    "极星双肩",
+    "凌云air",
+    "觅光",
+    "凌云3",
+    "瑞白",
+    "黑曜石吐司包",
+    "黑曜石托特包",
+    "曜影",
+    "黑曜石电脑包",
+    "拓界2",
+]
+
+
+def _get_style_names() -> list[str]:
+    """优先从飞书读取有主图表配置的款式列表，失败则回退硬编码。"""
+    try:
+        cfg = get_bitable_main_image_map()
+        if cfg:
+            return sorted(cfg.keys())
+    except Exception as exc:
+        print(f"[WARN] 动态读取主图款式列表失败，使用硬编码回退: {exc}")
+    return list(_STATIC_STYLE_NAMES)
+
+
+STYLE_NAMES = _get_style_names()
+
+STYLE_ITEM_IDS: dict[str, list[str]] = {}
+
+def _load_style_item_ids() -> None:
+    """从飞书多维表格加载款式 -> 天猫商品ID映射，用于搜索回退。"""
+    global STYLE_ITEM_IDS
+    try:
+        STYLE_ITEM_IDS = get_tm_style_id_map()
+        if STYLE_ITEM_IDS:
+            print(f"[主图] 已加载 {len(STYLE_ITEM_IDS)} 款商品ID映射 (搜索回退用)")
+    except Exception as exc:
+        print(f"[WARN] 加载商品ID映射失败: {exc}")
+
+_load_style_item_ids()
+
+
+def _print_preflight() -> None:
+    """启动期打印: 飞书表格里本周要采主图的款式数 (仅要求 主图飞书多维表格地址 非空)。"""
+    try:
+        styles = get_main_image_styles()
+        print(f"[主图] 飞书表格 周汇总待采 {len(styles)} 款 (天猫 + 主图飞书多维表格地址)")
+    except Exception as exc:
+        print(f"[WARN] 飞书读取失败, 跳过 preflight: {exc}")
+
+
+_print_preflight()
+
+
+def safe_filename(value: str) -> str:
+    return re.sub(r'[\\/:*?"<>|\s]+', "_", value).strip("_") or "unknown"
+
+
+# 采集输出的日期目录由 --date 控制 (默认 = today)。
+# weekly pipeline 通过 run_weekly_main_image.py 传入「上周日」,
+# 让 collect 与 insert 落在同一日期目录里,周标签语义统一为「这周插入的是上周」。
+_DATE_LABEL = ""
+
+
+def now_label() -> str:
+    return _DATE_LABEL or datetime.now().strftime("%Y-%m-%d")
+
+
+def parse_number(value: Any) -> float:
+    text = str(value or "").strip()
+    if not text or text in {"-", "--"}:
+        return 0.0
+    multiplier = 1.0
+    if "万" in text:
+        multiplier = 10000.0
+    text = (
+        text.replace(",", "")
+        .replace("¥", "")
+        .replace("￥", "")
+        .replace("%", "")
+        .replace("万", "")
+        .replace("元", "")
+        .strip()
+    )
+    match = re.search(r"-?\d+(?:\.\d+)?", text)
+    if not match:
+        return 0.0
+    return float(match.group(0)) * multiplier
+
+
+def canonical_image_key(url: str) -> str:
+    text = str(url or "").strip()
+    if not text:
+        return ""
+    if text.startswith("//"):
+        text = "https:" + text
+    text = text.split("?", 1)[0].split("#", 1)[0]
+    text = re.sub(r"_(?:\d+x\d+|sum|q\d+|webp|jpg|png)[^/]*$", "", text, flags=re.IGNORECASE)
+    text = re.sub(r"(\.(?:jpg|jpeg|png|webp|gif)).*$", r"\1", text, flags=re.IGNORECASE)
+    filename = text.rsplit("/", 1)[-1]
+    if filename:
+        return filename.lower()
+    return text.lower()
+
+
+def intish(value: float) -> int | float:
+    return int(value) if abs(value - int(value)) < 0.000001 else round(value, 4)
+
+
+def image_extension(url: str) -> str:
+    path = str(url or "").split("?", 1)[0].split("#", 1)[0].lower()
+    match = re.search(r"\.(jpg|jpeg|png|webp|gif)$", path)
+    if not match:
+        return ".jpg"
+    ext = match.group(1)
+    return ".jpg" if ext == "jpeg" else f".{ext}"
+
+
+def image_filename(image_key: str, image_url: str) -> str:
+    stem = safe_filename(str(image_key or "unknown"))
+    if re.search(r"\.(jpg|jpeg|png|webp|gif)$", stem, flags=re.IGNORECASE):
+        return stem
+    return f"{stem}{image_extension(image_url)}"
+
+
+def download_file(url: str, target: Path, timeout: int = 30) -> bool:
+    if not url:
+        return False
+    target.parent.mkdir(parents=True, exist_ok=True)
+    if target.exists() and target.stat().st_size > 0:
+        return True
+    headers = {
+        "User-Agent": (
+            "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
+            "AppleWebKit/537.36 (KHTML, like Gecko) Chrome/125.0 Safari/537.36"
+        ),
+        "Referer": "https://one.alimama.com/",
+    }
+    try:
+        request = Request(url, headers=headers)
+        with urlopen(request, timeout=timeout) as response:
+            content = response.read()
+        if not content:
+            return False
+        target.write_bytes(content)
+        return True
+    except Exception as exc:
+        print(f"  [WARN] 图片下载失败: {url} -> {exc}")
+        return False
+
+
+def bits_to_hex(bits: list[bool]) -> str:
+    value = 0
+    for bit in bits:
+        value = (value << 1) | int(bool(bit))
+    width = max(1, (len(bits) + 3) // 4)
+    return f"{value:0{width}x}"
+
+
+def hamming_hex(left: str, right: str) -> int:
+    if not left or not right:
+        return 10_000
+    return (int(left, 16) ^ int(right, 16)).bit_count() + abs(len(left) - len(right)) * 4
+
+
+def image_fingerprint(path: Path) -> dict[str, Any]:
+    try:
+        with Image.open(path) as image:
+            image = ImageOps.exif_transpose(image).convert("L")
+            ahash_img = image.resize((8, 8), Image.Resampling.LANCZOS)
+            ahash_pixels = list(
+                ahash_img.get_flattened_data() if hasattr(ahash_img, "get_flattened_data") else ahash_img.getdata()
+            )
+            avg = sum(ahash_pixels) / len(ahash_pixels)
+            ahash = bits_to_hex([pixel >= avg for pixel in ahash_pixels])
+
+            dhash_img = image.resize((9, 8), Image.Resampling.LANCZOS)
+            dhash_pixels = list(
+                dhash_img.get_flattened_data() if hasattr(dhash_img, "get_flattened_data") else dhash_img.getdata()
+            )
+            dhash_bits: list[bool] = []
+            for y in range(8):
+                row = dhash_pixels[y * 9 : (y + 1) * 9]
+                dhash_bits.extend(row[x] > row[x + 1] for x in range(8))
+            dhash = bits_to_hex(dhash_bits)
+
+            return {
+                "ahash": ahash,
+                "dhash": dhash,
+                "width": image.width,
+                "height": image.height,
+            }
+    except Exception as exc:
+        print(f"  [WARN] 鍥剧墖鎰熺煡鍝堝笇澶辫触: {path} -> {exc}")
+        return {}
+
+
+def same_visual_image(left: dict[str, Any], right: dict[str, Any]) -> bool:
+    if not left or not right:
+        return False
+    ahash_distance = hamming_hex(str(left.get("ahash", "")), str(right.get("ahash", "")))
+    dhash_distance = hamming_hex(str(left.get("dhash", "")), str(right.get("dhash", "")))
+    return min(ahash_distance, dhash_distance) <= IMAGE_HASH_DISTANCE
+
+
+def merge_creative_item(target: dict[str, Any], source: dict[str, Any]) -> None:
+    for creative_id in source.get("creative_ids", []):
+        if creative_id and creative_id not in target["creative_ids"]:
+            target["creative_ids"].append(creative_id)
+    for field in ("impressions", "clicks", "cost", "row_count"):
+        target[field] = float(target.get(field, 0) or 0) + float(source.get(field, 0) or 0)
+
+    for field in ("image_key", "creative_image_url", "local_image_path"):
+        list_field = f"{field}s"
+        target.setdefault(list_field, [])
+        value = source.get(field)
+        if value and value not in target[list_field]:
+            target[list_field].append(value)
+
+    target.setdefault("duplicate_image_keys", [])
+    if source.get("image_key") and source["image_key"] not in target["duplicate_image_keys"]:
+        target["duplicate_image_keys"].append(source["image_key"])
+
+    target.setdefault("merged_visual_hashes", [])
+    source_hash = source.get("image_fingerprint")
+    if source_hash and source_hash not in target["merged_visual_hashes"]:
+        target["merged_visual_hashes"].append(source_hash)
+
+    samples = target.setdefault("sample_rows", [])
+    for row in source.get("sample_rows", []):
+        if len(samples) >= 6:
+            break
+        samples.append(row)
+
+
+def merge_creatives_by_visual_hash(creatives: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    merged: list[dict[str, Any]] = []
+    for item in creatives:
+        fingerprint = item.get("image_fingerprint") or {}
+        match = None
+        if fingerprint:
+            for candidate in merged:
+                if same_visual_image(fingerprint, candidate.get("image_fingerprint") or {}):
+                    match = candidate
+                    break
+        if match:
+            merge_creative_item(match, item)
+            continue
+        item.setdefault("creative_image_urls", [item["creative_image_url"]] if item.get("creative_image_url") else [])
+        item.setdefault("local_image_paths", [item["local_image_path"]] if item.get("local_image_path") else [])
+        item.setdefault("image_keys", [item["image_key"]] if item.get("image_key") else [])
+        item.setdefault("duplicate_image_keys", [])
+        item.setdefault("merged_visual_hashes", [fingerprint] if fingerprint else [])
+        merged.append(item)
+
+    for item in merged:
+        impressions = float(item.get("impressions", 0) or 0)
+        clicks = float(item.get("clicks", 0) or 0)
+        item["click_rate"] = round(clicks / impressions * 100, 4) if impressions else 0.0
+        item["impressions"] = intish(impressions)
+        item["clicks"] = intish(clicks)
+        item["cost"] = round(float(item.get("cost", 0) or 0), 4)
+        item["row_count"] = intish(float(item.get("row_count", 0) or 0))
+    return sorted(merged, key=lambda entry: float(entry.get("impressions", 0) or 0), reverse=True)
+
+
+def _console_safe(text: str) -> str:
+    encoding = getattr(sys.stdout, "encoding", None) or "utf-8"
+    return text.encode(encoding, errors="replace").decode(encoding, errors="replace")
+
+
+def _unlink_with_retry(path: Path, attempts: int, delay: float) -> None:
+    for attempt in range(1, max(1, attempts) + 1):
+        try:
+            path.unlink()
+            return
+        except Exception as exc:
+            if attempt < max(1, attempts):
+                time.sleep(delay)
+                continue
+            message = (
+                f"  [WARN] failed to delete duplicate image after {attempt} "
+                f"attempts: {path} -> {exc}"
+            )
+            print(_console_safe(message))
+
+
+def cleanup_images_after_visual_merge(
+    images_dir: Path,
+    creatives: list[dict[str, Any]],
+    *,
+    retry_attempts: int = 3,
+    retry_delay: float = 0.2,
+) -> None:
+    keep_paths: set[Path] = set()
+    for item in creatives:
+        primary = str(item.get("local_image_path") or "")
+        if primary:
+            keep_paths.add(Path(primary).resolve())
+
+        all_paths = [str(path) for path in item.get("local_image_paths", []) if path]
+        duplicate_paths = [path for path in all_paths if path != primary]
+        if duplicate_paths:
+            item["duplicate_local_image_paths"] = duplicate_paths
+        item["local_image_paths"] = [primary] if primary else []
+
+    if not images_dir.exists():
+        return
+    for path in images_dir.iterdir():
+        if not path.is_file():
+            continue
+        try:
+            if path.resolve() not in keep_paths:
+                _unlink_with_retry(path, retry_attempts, retry_delay)
+        except Exception as exc:
+            print(_console_safe(f"  [WARN] image cleanup failed: {path} -> {exc}"))
+            continue
+
+
+def start_stealthy_session(headless: bool, user_data_dir: str) -> tuple[StealthySession, Any]:
+    # 显式引用 StealthyFetcher，便于确认脚本基于 Scrapling stealth 栈。
+    _ = StealthyFetcher
+    session = StealthySession(
+        headless=headless,
+        real_chrome=True,
+        user_data_dir=user_data_dir,
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=60000,
+        network_idle=False,
+        disable_resources=False,
+        block_ads=False,
+    )
+    session.start()
+    context = session.context
+    page = context.pages[-1] if context.pages else context.new_page()
+    try:
+        page.set_viewport_size({"width": 1920, "height": 1080})
+    except Exception:
+        pass
+    return session, page
+
+
+def close_known_popups(page) -> bool:
+    selectors = [
+        ".sycm-active-egg-modal-content-close",
+        ".baxia-dialog-close",
+        ".next-dialog-close",
+        ".ant-modal-close",
+        ".anticon-close",
+        "[aria-label='Close']",
+        "[aria-label='close']",
+    ]
+    try:
+        closed = bool(
+            page.evaluate(
+                r"""(selectors) => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const overlaySelector = [
+                        '.sycm-active-egg-modal-content',
+                        '.baxia-dialog',
+                        '.next-dialog',
+                        '.next-overlay-wrapper',
+                        '.ant-modal',
+                        '.ant-popover',
+                        '[role="dialog"]',
+                        '[class*="modal"]',
+                        '[class*="dialog"]'
+                    ].join(',');
+                    const inPopupContext = (el) => {
+                        const overlay = el.closest(overlaySelector);
+                        if (overlay && visible(overlay)) return true;
+                        let node = el;
+                        while (node && node !== document.body) {
+                            const style = window.getComputedStyle(node);
+                            const zIndex = Number.parseInt(style.zIndex || '0', 10);
+                            if (style.position === 'fixed' && zIndex >= 900 && visible(node)) return true;
+                            node = node.parentElement;
+                        }
+                        return false;
+                    };
+                    let count = 0;
+                    for (const selector of selectors) {
+                        for (const el of Array.from(document.querySelectorAll(selector))) {
+                            if (!visible(el)) continue;
+                            if (!inPopupContext(el)) continue;
+                            try { el.click(); count += 1; } catch (_) {}
+                        }
+                    }
+                    const exactTexts = ['我知道了', '稍后再说', '关闭', '取消'];
+                    for (const target of exactTexts) {
+                        const el = Array.from(document.querySelectorAll('button, span, div, a'))
+                            .find((node) => visible(node) && inPopupContext(node) && (node.textContent || '').trim() === target);
+                        if (el) {
+                            try { el.click(); count += 1; } catch (_) {}
+                        }
+                    }
+                    return count > 0;
+                }""",
+                selectors,
+            )
+        )
+    except Exception:
+        closed = False
+    if closed:
+        print("已关闭弹窗/广告")
+        time.sleep(0.3)
+    return closed
+
+
+def is_logged_in(page) -> bool:
+    url = page.url.lower()
+    if "login" in url:
+        return False
+    try:
+        body = page.locator("body").inner_text(timeout=3000)
+        if "账号名" in body and "请输入登录密码" in body:
+            return False
+    except Exception:
+        pass
+    return True
+
+
+def fill_first(scope, selectors: list[str], value: str) -> bool:
+    for selector in selectors:
+        try:
+            loc = scope.locator(selector).first
+            loc.wait_for(state="visible", timeout=5000)
+            loc.fill(value)
+            return True
+        except Exception:
+            continue
+    return False
+
+
+def click_first(scope, selectors: list[str], timeout: int = 3000) -> bool:
+    for selector in selectors:
+        try:
+            loc = scope.locator(selector).first
+            loc.wait_for(state="visible", timeout=timeout)
+            loc.click()
+            return True
+        except Exception:
+            continue
+    return False
+
+
+def safe_goto(page, url: str, timeout: int = 60000, attempts: int = 3) -> None:
+    last_error: Exception | None = None
+    for attempt in range(1, attempts + 1):
+        try:
+            page.goto(url, wait_until="domcontentloaded", timeout=timeout)
+            return
+        except Exception as exc:
+            last_error = exc
+            print(f"  [WARN] 打开页面失败，第 {attempt}/{attempts} 次: {exc}")
+            time.sleep(3)
+    try:
+        page.evaluate("(url) => { window.location.href = url; }", url)
+        page.wait_for_load_state("domcontentloaded", timeout=timeout)
+        return
+    except Exception as exc:
+        last_error = exc
+    raise RuntimeError(f"打开页面失败: {url}; {last_error}")
+
+
+def goto_wanxiang_report_url(page) -> None:
+    try:
+        safe_goto(page, WANXIANG_REPORT_URL, attempts=2)
+        time.sleep(4)
+        if "one.alimama.com" in page.url:
+            return
+    except Exception as exc:
+        print(f"  [WARN] 直连创意报表失败，改用基础页加 hash: {exc}")
+
+    safe_goto(page, WANXIANG_BASE_URL, attempts=4)
+    time.sleep(5)
+    page.evaluate("(hash) => { window.location.hash = hash.slice(1); }", WANXIANG_REPORT_HASH)
+    time.sleep(8)
+    if "one.alimama.com" not in page.url:
+        raise RuntimeError(f"未能进入万相页面，当前 URL: {page.url}")
+
+
+def do_login(page, account: str, password: str, wait_seconds: int = 300) -> bool:
+    print(f"打开登录页: {LOGIN_URL}")
+    safe_goto(page, LOGIN_URL)
+    time.sleep(3)
+
+    try:
+        page.wait_for_selector("#alibaba-login-box", timeout=15000)
+        iframe_element = page.locator("#alibaba-login-box").element_handle()
+        login_frame = iframe_element.content_frame() if iframe_element else None
+    except Exception:
+        login_frame = None
+
+    frame_or_page = login_frame or page
+    account_selectors = [
+        'input[placeholder*="账号"]',
+        'input[name="fm-login-id"]',
+        'input[type="text"]',
+    ]
+    password_selectors = [
+        'input[placeholder*="密码"]',
+        'input[name="fm-login-password"]',
+        'input[type="password"]',
+    ]
+
+    print("输入淘宝账号...")
+    fill_first(frame_or_page, account_selectors, account)
+    time.sleep(0.5)
+    print("输入淘宝密码...")
+    fill_first(frame_or_page, password_selectors, password)
+    time.sleep(0.5)
+
+    print("点击登录...")
+    clicked = click_first(frame_or_page, ['button:has-text("登录")', 'button[type="submit"]', 'text=登录'])
+    if not clicked:
+        print("  [WARN] 未找到登录按钮，请在浏览器中手动点击登录")
+
+    print("=" * 50)
+    print("如出现滑块/验证码，请在浏览器中手动完成。")
+    print("=" * 50)
+    deadline = time.time() + wait_seconds
+    while time.time() < deadline:
+        close_known_popups(page)
+        if is_logged_in(page):
+            print(f"登录成功: {page.url}")
+            return True
+        time.sleep(1)
+    return False
+
+
+def ensure_logged_in(page, account: str, password: str) -> None:
+    print("检查淘宝卖家登录态...")
+    try:
+        safe_goto(page, SELLER_SUBWAY_URL)
+        time.sleep(4)
+        close_known_popups(page)
+        if is_logged_in(page):
+            print(f"[OK] 已登录: {page.url}")
+            return
+    except Exception as exc:
+        print(f"  [WARN] 登录态检查失败: {exc}")
+
+    if not do_login(page, account, password):
+        raise RuntimeError("淘宝登录失败或等待登录超时")
+
+
+def goto_wanxiang_from_seller(page) -> None:
+    print(f"进入卖家推广页: {SELLER_SUBWAY_URL}")
+    safe_goto(page, SELLER_SUBWAY_URL)
+    time.sleep(5)
+    close_known_popups(page)
+
+    print("滚动到底部并点击“前往万相AI无界”...")
+    clicked = False
+    for _ in range(8):
+        clicked = bool(
+            page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && rect.width > 0
+                            && rect.height > 0;
+                    };
+                    const target = Array.from(document.querySelectorAll('button, a, span, div'))
+                        .find((el) => visible(el) && (el.textContent || '').trim().includes('前往万相AI无界'));
+                    if (target) {
+                        target.scrollIntoView({block: 'center'});
+                        target.click();
+                        return true;
+                    }
+                    window.scrollTo(0, document.body.scrollHeight);
+                    return false;
+                }"""
+            )
+        )
+        if clicked:
+            break
+        time.sleep(1)
+
+    if clicked:
+        time.sleep(5)
+        print(f"点击后页面: {page.url}")
+    else:
+        print("  [WARN] 未找到“前往万相AI无界”，直接打开创意报表 URL")
+        goto_wanxiang_report_url(page)
+        time.sleep(5)
+
+
+def goto_creative_report(page) -> None:
+    if "one.alimama.com" not in page.url:
+        print(f"打开万相 AI 无界创意报表: {WANXIANG_REPORT_URL}")
+        goto_wanxiang_report_url(page)
+    time.sleep(6)
+    for _ in range(4):
+        if not close_known_popups(page):
+            break
+        time.sleep(1)
+
+    if "#!/report/creative" not in page.url:
+        print("跳转到创意报表路由...")
+        goto_wanxiang_report_url(page)
+        time.sleep(6)
+        close_known_popups(page)
+
+
+def _read_report_template_state(page) -> dict[str, Any]:
+    state = page.evaluate(
+        r"""(config) => {
+            const { labels, options } = config;
+            const normalize = (value) => String(value || '').replace(/\s+/g, '').trim();
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && Number(style.opacity || 1) !== 0
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const nodes = Array.from(document.querySelectorAll('label, button, div, span, p'));
+            const labelNodes = nodes
+                .filter(visible)
+                .filter((el) => config.labels.includes(normalize(el.textContent)))
+                .map((el) => {
+                    const rect = el.getBoundingClientRect();
+                    return { el, rect, area: rect.width * rect.height };
+                })
+                .sort((a, b) => a.area - b.area);
+            const label = labelNodes[0];
+            if (!label) return { found: false, current: '' };
+
+            const candidates = nodes
+                .filter(visible)
+                .map((el) => {
+                    const text = normalize(el.textContent);
+                    const rect = el.getBoundingClientRect();
+                    return {
+                        el,
+                        text,
+                        rect,
+                        area: rect.width * rect.height,
+                        vertical: Math.abs(rect.top - label.rect.bottom),
+                    };
+                })
+                .filter((item) => options.includes(item.text))
+                .filter((item) => item.rect.top >= label.rect.top - 10)
+                .filter((item) => item.rect.top <= label.rect.bottom + 180)
+                .filter((item) => item.rect.right >= label.rect.left - 80)
+                .filter((item) => item.rect.left <= label.rect.right + 420)
+                .sort((a, b) => (a.vertical - b.vertical) || (a.area - b.area));
+            const current = candidates[0];
+            return {
+                found: Boolean(current),
+                current: current ? current.text : '',
+            };
+        }""",
+        {
+            "labels": list(REPORT_TEMPLATE_LABELS),
+            "options": list(REPORT_TEMPLATE_OPTIONS),
+        },
+    )
+    return state if isinstance(state, dict) else {"found": False, "current": ""}
+
+
+def _click_report_template_control(page) -> bool:
+    return bool(
+        page.evaluate(
+            r"""(config) => {
+                const { options } = config;
+                const normalize = (value) => String(value || '').replace(/\s+/g, '').trim();
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && Number(style.opacity || 1) !== 0
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const nodes = Array.from(document.querySelectorAll('label, button, div, span, p'));
+                const label = nodes
+                    .filter(visible)
+                    .filter((el) => config.labels.includes(normalize(el.textContent)))
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        return { el, rect, area: rect.width * rect.height };
+                    })
+                    .sort((a, b) => a.area - b.area)[0];
+                if (!label) return false;
+
+                const current = nodes
+                    .filter(visible)
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        return {
+                            el,
+                            text: normalize(el.textContent),
+                            rect,
+                            area: rect.width * rect.height,
+                            vertical: Math.abs(rect.top - label.rect.bottom),
+                        };
+                    })
+                    .filter((item) => options.includes(item.text))
+                    .filter((item) => item.rect.top >= label.rect.top - 10)
+                    .filter((item) => item.rect.top <= label.rect.bottom + 180)
+                    .filter((item) => item.rect.right >= label.rect.left - 80)
+                    .filter((item) => item.rect.left <= label.rect.right + 420)
+                    .sort((a, b) => (a.vertical - b.vertical) || (a.area - b.area))[0];
+                if (!current) return false;
+
+                let clickable = current.el;
+                let ancestor = current.el;
+                for (let depth = 0; depth < 5 && ancestor.parentElement; depth += 1) {
+                    const parent = ancestor.parentElement;
+                    const rect = parent.getBoundingClientRect();
+                    const className = String(parent.className || '');
+                    if (
+                        parent.matches('button, [role="combobox"], [aria-haspopup="listbox"]')
+                        || /select|picker|trigger/i.test(className)
+                    ) {
+                        if (rect.width <= 600 && rect.height <= 140) clickable = parent;
+                        break;
+                    }
+                    if (
+                        rect.width <= 600
+                        && rect.height <= 140
+                        && window.getComputedStyle(parent).cursor === 'pointer'
+                    ) clickable = parent;
+                    ancestor = parent;
+                }
+                clickable.scrollIntoView({ block: 'center', inline: 'center' });
+                clickable.click();
+                return true;
+            }""",
+            {
+                "labels": list(REPORT_TEMPLATE_LABELS),
+                "options": list(REPORT_TEMPLATE_OPTIONS),
+            },
+        )
+    )
+
+
+def _click_report_template_option(page, target: str) -> bool:
+    return bool(
+        page.evaluate(
+            r"""(targetText) => {
+                const normalize = (value) => String(value || '').replace(/\s+/g, '').trim();
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && Number(style.opacity || 1) !== 0
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const candidates = Array.from(
+                    document.querySelectorAll('[role="option"], li, button, div, span')
+                )
+                    .filter(visible)
+                    .filter((el) => normalize(el.textContent) === normalize(targetText))
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        const semantic = Boolean(
+                            el.closest('[role="option"], li, [class*="option" i], [class*="menu" i]')
+                        );
+                        return { el, area: rect.width * rect.height, semantic };
+                    })
+                    .sort((a, b) => Number(b.semantic) - Number(a.semantic) || a.area - b.area);
+                const target = candidates[0]?.el;
+                if (!target) return false;
+                const clickable = target.closest(
+                    '[role="option"], li, button, [class*="option" i], [class*="menu-item" i]'
+                ) || target;
+                clickable.scrollIntoView({ block: 'nearest', inline: 'nearest' });
+                clickable.click();
+                return true;
+            }""",
+            target,
+        )
+    )
+
+
+def dump_report_template_debug(page, name: str) -> None:
+    debug_dir = TMP_ROOT / "debug" / "wanxiang_report_template"
+    debug_dir.mkdir(parents=True, exist_ok=True)
+    label = safe_filename(f"{datetime.now():%Y%m%d_%H%M%S}_{name}")
+    try:
+        page.screenshot(path=str(debug_dir / f"{label}.png"), full_page=False)
+        payload = page.evaluate(
+            r"""() => ({
+                url: window.location.href,
+                title: document.title,
+                readyState: document.readyState,
+                bodyText: String(document.body?.innerText || '').slice(0, 30000),
+            })"""
+        )
+        path = debug_dir / f"{label}.json"
+        path.write_text(
+            json.dumps(payload, ensure_ascii=False, indent=2),
+            encoding="utf-8",
+        )
+        print(f"  [DEBUG] 已保存报表模板页面信息: {path}")
+    except Exception as exc:
+        print(f"  [WARN] 保存报表模板调试信息失败: {exc}")
+
+
+def select_report_template(
+    page,
+    target: str = DEFAULT_REPORT_TEMPLATE,
+    *,
+    find_attempts: int = 30,
+    verify_attempts: int = 8,
+) -> None:
+    """选择并确认创意报表模板，防止误采全部计划数据。"""
+    if target not in REPORT_TEMPLATE_OPTIONS:
+        raise ValueError(f"不支持的报表模板: {target}")
+
+    state: dict[str, Any] = {"found": False, "current": ""}
+    for attempt in range(max(1, find_attempts)):
+        state = _read_report_template_state(page)
+        if state.get("found"):
+            break
+        if attempt < max(1, find_attempts) - 1:
+            page.wait_for_timeout(1000)
+
+    if state.get("current") == target:
+        print(f"[OK] 报表模板已是“{target}”")
+        return
+    if not state.get("found"):
+        dump_report_template_debug(page, "filter_not_found")
+        raise RuntimeError(
+            f"未找到“报表模板”筛选器，停止主图采集；当前 URL: {page.url}"
+        )
+
+    print(f"选择报表模板: {state.get('current') or '未知'} -> {target}")
+    if not _click_report_template_control(page):
+        raise RuntimeError("无法打开“报表模板”下拉框")
+    page.wait_for_timeout(500)
+    if not _click_report_template_option(page, target):
+        raise RuntimeError(f"未找到报表模板选项“{target}”")
+    page.wait_for_timeout(800)
+
+    for _ in range(max(1, verify_attempts)):
+        state = _read_report_template_state(page)
+        if state.get("current") == target:
+            print(f"[OK] 报表模板已切换为“{target}”")
+            wait_table_change(page, seconds=2.0)
+            return
+        page.wait_for_timeout(500)
+    dump_report_template_debug(page, "filter_verify_failed")
+    raise RuntimeError(f"报表模板未能确认切换为“{target}”，停止主图采集")
+
+
+def find_search_input_id(page) -> str:
+    input_id = page.evaluate(
+        r"""() => {
+            window.scrollTo(0, document.body.scrollHeight);
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const inputs = Array.from(document.querySelectorAll('input')).filter(visible);
+            const exact = inputs.find((el) => {
+                const ph = el.getAttribute('placeholder') || '';
+                return ph.includes('创意名称') || ph.includes('主体名称') || ph.includes('名称/ID');
+            });
+            const lowerExact = inputs.find((el) => {
+                const rect = el.getBoundingClientRect();
+                const ph = el.getAttribute('placeholder') || '';
+                return rect.top > window.innerHeight * 0.45
+                    && (ph.includes('创意名称') || ph.includes('主体名称') || ph.includes('名称/ID'));
+            });
+            const target = lowerExact || exact || inputs
+                .map((el) => ({ el, rect: el.getBoundingClientRect() }))
+                .filter((item) => item.rect.width >= 120 && item.rect.top > window.innerHeight * 0.35)
+                .sort((a, b) => (b.rect.top - a.rect.top) || (b.rect.left - a.rect.left))[0]?.el;
+            if (!target) return '';
+            if (!target.id) target.id = `codex-wanxiang-search-${Date.now()}`;
+            target.scrollIntoView({block: 'center', inline: 'center'});
+            return target.id;
+        }"""
+    )
+    if not input_id:
+        raise RuntimeError("未找到创意报表搜索输入框")
+    return str(input_id)
+
+
+def search_style_by_subject(page, style_name: str) -> None:
+    print(f"搜索款式主体名称: {style_name}")
+    input_id = find_search_input_id(page)
+    locator = page.locator(f"#{input_id}")
+    locator.click()
+    page.keyboard.press("Control+A")
+    page.keyboard.press("Backspace")
+    page.keyboard.type(style_name, delay=80)
+    time.sleep(1.5)
+
+    clicked = bool(
+        page.evaluate(
+            r"""(styleName) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const candidates = Array.from(document.querySelectorAll('div, li, span, a, button'))
+                    .filter((el) => visible(el))
+                    .filter((el) => {
+                        const text = (el.textContent || '').replace(/\s+/g, '');
+                        return text.includes(styleName)
+                            && text.includes('主体名称')
+                            && !text.includes('创意名称/ID搜索含有');
+                    })
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        return { el, rect, area: rect.width * rect.height };
+                    })
+                    .sort((a, b) => {
+                        const topDiff = b.rect.top - a.rect.top;
+                        if (Math.abs(topDiff) > 2) return topDiff;
+                        return a.area - b.area;
+                    });
+                const target = candidates[0];
+                if (!target) return false;
+                const x = target.rect.left + target.rect.width / 2;
+                const y = target.rect.top + target.rect.height / 2;
+                const hit = document.elementFromPoint(x, y);
+                if (hit) hit.click();
+                else target.el.click();
+                return true;
+            }""",
+            style_name,
+        )
+    )
+    if not clicked:
+        print(f"  [WARN] 未找到“{style_name}”的主体名称搜索选项，改用 URL 参数搜索")
+        safe_goto(page, build_style_search_url(style_name))
+    wait_table_change(page)
+
+
+def build_style_search_url(style_name: str) -> str:
+    return (
+        "https://one.alimama.com/index.html?spm=a21dvs.28495141.cc92828ac.dbf3edac0.778454dfoBlQB4"
+        f"#!/report/creative?rptType=creative&offset=0&searchKey=itemIdOrName&searchValue={quote(style_name)}"
+    )
+
+
+def wait_table_change(page, seconds: float = 1.2) -> None:
+    page.wait_for_timeout(int(seconds * 1000))
+    close_known_popups(page)
+
+
+def extract_current_page_rows(page) -> list[dict[str, Any]]:
+    rows = page.evaluate(
+        r"""() => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const normalize = (text) => String(text || '').replace(/\s+/g, ' ').trim();
+            const tables = Array.from(document.querySelectorAll('table')).filter(visible);
+            const table = tables
+                .map((el) => ({ el, text: normalize(el.innerText), rect: el.getBoundingClientRect() }))
+                .filter((item) => item.text.includes('创意ID') || item.text.includes('展现量'))
+                .sort((a, b) => b.rect.top - a.rect.top)[0]?.el || tables[tables.length - 1];
+            if (!table) return [];
+            const allRows = Array.from(table.querySelectorAll('tbody tr, tr')).filter(visible);
+            const normalizeUrl = (url) => {
+                const value = String(url || '').trim();
+                if (!value || value.startsWith('data:')) return '';
+                if (value.startsWith('//')) return `https:${value}`;
+                return value;
+            };
+            const imgUrl = (img) => {
+                if (!img) return '';
+                const attrs = [
+                    'currentSrc',
+                    'src',
+                    'data-src',
+                    'data-lazy-src',
+                    'data-original',
+                    'data-ks-lazyload',
+                    'data-img',
+                    'data-url',
+                ];
+                for (const attr of attrs) {
+                    const value = attr === 'currentSrc' ? img.currentSrc : img.getAttribute(attr);
+                    const normalized = normalizeUrl(value);
+                    if (normalized) return normalized;
+                }
+                return '';
+            };
+            const isCreativeImageUrl = (url) => /alicdn\.com|taobaocdn\.com|tbcdn\.cn/i.test(String(url || ''));
+            const visibleImages = Array.from(document.querySelectorAll('img'))
+                .filter(visible)
+                .map((img) => ({ img, rect: img.getBoundingClientRect(), url: imgUrl(img) }))
+                .filter((item) => item.url && isCreativeImageUrl(item.url) && item.rect.width >= 24 && item.rect.height >= 24);
+            const leftCreativeImages = visibleImages
+                .filter((item) => item.rect.left < 300)
+                .sort((a, b) => a.rect.top - b.rect.top);
+            const imageNearRow = (row, preferredLeft = true) => {
+                const rowRect = row.getBoundingClientRect();
+                const rowY = rowRect.top + rowRect.height / 2;
+                const candidates = visibleImages
+                    .filter((item) => Math.abs((item.rect.top + item.rect.height / 2) - rowY) <= Math.max(26, rowRect.height / 2 + 8))
+                    .filter((item) => !preferredLeft || item.rect.left <= rowRect.left + 520)
+                    .sort((a, b) => {
+                        const dyA = Math.abs((a.rect.top + a.rect.height / 2) - rowY);
+                        const dyB = Math.abs((b.rect.top + b.rect.height / 2) - rowY);
+                        if (Math.abs(dyA - dyB) > 1) return dyA - dyB;
+                        return a.rect.left - b.rect.left;
+                    });
+                return candidates[0]?.url || '';
+            };
+            const imageFromNode = (root) => {
+                if (!root) return '';
+                const attrs = [
+                    'src',
+                    'currentSrc',
+                    'data-src',
+                    'data-lazy-src',
+                    'data-original',
+                    'data-ks-lazyload',
+                    'data-img',
+                    'data-url',
+                ];
+                const imageFromHtml = (html) => {
+                    const text = String(html || '');
+                    const matches = text.match(/(?:https?:)?\/\/[^"')\s<>]+?\.(?:jpg|jpeg|png|webp|gif)(?:_[^"')\s<>]+)?/ig) || [];
+                    const ali = matches.find((url) => /alicdn|taobao|tbcdn/i.test(url));
+                    return normalizeUrl(ali || matches[0] || '');
+                };
+                const htmlHit = imageFromHtml(root.outerHTML);
+                if (htmlHit) return htmlHit;
+                const nodes = [root, ...Array.from(root.querySelectorAll('*'))];
+                for (const node of nodes) {
+                    if (node.tagName === 'IMG') {
+                        for (const attr of attrs) {
+                            const value = attr === 'currentSrc' ? node.currentSrc : node.getAttribute(attr);
+                            const normalized = normalizeUrl(value);
+                            if (normalized) return normalized;
+                        }
+                    }
+                    const inlineStyle = node.getAttribute('style') || '';
+                    const computedBg = window.getComputedStyle(node).backgroundImage || '';
+                    const bg = inlineStyle || computedBg;
+                    const match = bg.match(/url\(["']?([^"')]+)["']?\)/);
+                    if (match) {
+                        const normalized = normalizeUrl(match[1]);
+                        if (normalized) return normalized;
+                    }
+                }
+                return '';
+            };
+            return allRows.map((row, rowIndex) => {
+                const cells = Array.from(row.querySelectorAll('td, th')).filter(visible);
+                const cellTexts = cells.map((cell) => normalize(cell.innerText || cell.textContent));
+                const rowText = normalize(row.innerText || row.textContent);
+                const creativeMatch = rowText.match(/创意ID[:：]\s*([0-9A-Za-z_-]+)/);
+                const creativeCell = cells[0] || row;
+                const subjectCell = cells[1] || null;
+                return {
+                    rowText,
+                    rowHtml: row.outerHTML,
+                    cells: cellTexts,
+                    creativeId: creativeMatch ? creativeMatch[1] : '',
+                    creativeImageUrl: imageFromNode(creativeCell) || imageNearRow(row, true) || leftCreativeImages[rowIndex]?.url || '',
+                    subjectImageUrl: imageFromNode(subjectCell),
+                };
+            }).filter((row) => row.creativeId && row.cells.length >= 6);
+        }"""
+    )
+    if not isinstance(rows, list):
+        return []
+    parsed: list[dict[str, Any]] = []
+    for row in rows:
+        cells = row.get("cells") or []
+        impressions_text = cells[2] if len(cells) > 2 else ""
+        clicks_text = cells[3] if len(cells) > 3 else ""
+        cost_text = cells[4] if len(cells) > 4 else ""
+        ctr_text = cells[5] if len(cells) > 5 else ""
+        parsed.append(
+            {
+                "creative_id": str(row.get("creativeId") or ""),
+                "creative_image_url": row.get("creativeImageUrl") or "",
+                "creative_image_key": canonical_image_key(row.get("creativeImageUrl") or ""),
+                "subject_image_url": row.get("subjectImageUrl") or "",
+                "row_text": row.get("rowText") or "",
+                "row_html": row.get("rowHtml") or "",
+                "cells": cells,
+                "impressions": parse_number(impressions_text),
+                "clicks": parse_number(clicks_text),
+                "cost": parse_number(cost_text),
+                "source_ctr": parse_number(ctr_text),
+            }
+        )
+    return parsed
+
+
+def collect_current_page_image_map(page) -> dict[str, str]:
+    try:
+        image_map = page.evaluate(
+            r"""async () => {
+                const sleep = (ms) => new Promise((resolve) => setTimeout(resolve, ms));
+                const normalizeUrl = (url) => {
+                    const value = String(url || '').trim();
+                    if (!value || value.startsWith('data:')) return '';
+                    if (value.startsWith('//')) return `https:${value}`;
+                    return value;
+                };
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0
+                        && rect.bottom >= 0
+                        && rect.top <= window.innerHeight;
+                };
+                const imgUrl = (img) => {
+                    const attrs = ['currentSrc', 'src', 'data-src', 'data-lazy-src', 'data-original', 'data-ks-lazyload'];
+                    for (const attr of attrs) {
+                        const value = attr === 'currentSrc' ? img.currentSrc : img.getAttribute(attr);
+                        const normalized = normalizeUrl(value);
+                        if (normalized) return normalized;
+                    }
+                    return '';
+                };
+                const map = {};
+                const collect = () => {
+                    const nodes = Array.from(document.querySelectorAll('td, div, span, tr'))
+                        .filter((el) => visible(el) && (el.textContent || '').includes('创意ID'))
+                        .map((el) => {
+                            const rect = el.getBoundingClientRect();
+                            const text = (el.textContent || '').replace(/\s+/g, ' ').trim();
+                            const match = text.match(/创意ID[:：]\s*([0-9A-Za-z_-]+)/);
+                            return match ? { creativeId: match[1], rect } : null;
+                        })
+                        .filter(Boolean);
+                    const imgs = Array.from(document.querySelectorAll('img'))
+                        .filter(visible)
+                        .map((img) => ({ url: imgUrl(img), rect: img.getBoundingClientRect() }))
+                        .filter((item) => item.url && /alicdn\.com|taobaocdn\.com|tbcdn\.cn/i.test(item.url) && item.rect.left < 300);
+                    for (const node of nodes) {
+                        const nodeY = node.rect.top + node.rect.height / 2;
+                        const hit = imgs
+                            .map((img) => ({ ...img, dy: Math.abs((img.rect.top + img.rect.height / 2) - nodeY) }))
+                            .filter((img) => img.dy < 48)
+                            .sort((a, b) => a.dy - b.dy)[0];
+                        if (hit && !map[node.creativeId]) map[node.creativeId] = hit.url;
+                    }
+                };
+
+                const startY = window.scrollY;
+                const detail = Array.from(document.querySelectorAll('div, section'))
+                    .find((el) => (el.textContent || '').includes('创意数据明细'));
+                const detailTop = detail ? detail.getBoundingClientRect().top + window.scrollY : startY;
+                for (let i = 0; i < 7; i += 1) {
+                    window.scrollTo(0, detailTop + i * 430);
+                    await sleep(220);
+                    collect();
+                }
+                window.scrollTo(0, startY);
+                await sleep(120);
+                return map;
+            }"""
+        )
+        return {str(k): str(v) for k, v in (image_map or {}).items() if k and v}
+    except Exception as exc:
+        print(f"  [WARN] 页面图片映射采集失败: {exc}")
+        return {}
+
+
+def click_next_page(page, next_page_no: int) -> bool:
+    clicked = bool(
+        page.evaluate(
+            r"""(nextPageNo) => {
+                window.scrollTo(0, document.body.scrollHeight);
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const disabled = (el) => {
+                    const cls = el.className || '';
+                    const aria = el.getAttribute('aria-disabled') || '';
+                    return el.disabled || aria === 'true' || /disabled/.test(String(cls));
+                };
+                const pageText = String(nextPageNo);
+                const magixCandidates = Array.from(document.querySelectorAll('a[mx-click], button[mx-click], span[mx-click]'))
+                    .filter(visible)
+                    .filter((el) => {
+                        const rect = el.getBoundingClientRect();
+                        const mx = el.getAttribute('mx-click') || '';
+                        return mx.includes(`page:'${pageText}'`)
+                            && rect.top > window.innerHeight - 220
+                            && !disabled(el);
+                    })
+                    .sort((a, b) => b.getBoundingClientRect().left - a.getBoundingClientRect().left);
+                if (magixCandidates[0]) {
+                    magixCandidates[0].click();
+                    return true;
+                }
+
+                const pageCandidates = Array.from(document.querySelectorAll('button, li, a, span, div'))
+                    .filter(visible)
+                    .filter((el) => {
+                        const rect = el.getBoundingClientRect();
+                        const text = (el.textContent || '').trim();
+                        return text === pageText
+                            && rect.top > window.innerHeight - 140
+                            && rect.left > window.innerWidth * 0.45
+                            && rect.width <= 80
+                            && rect.height <= 80
+                            && !disabled(el);
+                    })
+                    .sort((a, b) => b.getBoundingClientRect().left - a.getBoundingClientRect().left);
+                if (pageCandidates[0]) {
+                    pageCandidates[0].click();
+                    return true;
+                }
+
+                const candidates = Array.from(document.querySelectorAll('button, li, a, span, div'))
+                    .filter(visible)
+                    .filter((el) => {
+                        const rect = el.getBoundingClientRect();
+                        const text = (el.textContent || '').trim();
+                        const title = el.getAttribute('title') || '';
+                        const aria = el.getAttribute('aria-label') || '';
+                        const cls = String(el.className || '');
+                        const isNext = text === '>' || text === '›' || title.includes('下一页') || aria.includes('下一页') || /next/.test(cls);
+                        return isNext
+                            && rect.top > window.innerHeight - 140
+                            && rect.left > window.innerWidth * 0.45
+                            && rect.width <= 90
+                            && rect.height <= 90
+                            && !disabled(el);
+                    })
+                    .sort((a, b) => b.getBoundingClientRect().left - a.getBoundingClientRect().left);
+                const target = candidates[0];
+                if (!target) return false;
+                target.click();
+                return true;
+            }""",
+            next_page_no,
+        )
+    )
+    if clicked:
+        wait_table_change(page, 1.5)
+    else:
+        if DEBUG:
+            dump_pagination_debug(page, f"next_{next_page_no}")
+    return clicked
+
+
+def dump_pagination_debug(page, name: str) -> None:
+    debug_dir = TMP_ROOT / "debug" / "wanxiang_pagination"
+    debug_dir.mkdir(parents=True, exist_ok=True)
+    label = f"{safe_filename(name)}_{int(time.time())}"
+    try:
+        page.screenshot(path=str(debug_dir / f"{label}.png"), full_page=False)
+    except Exception:
+        pass
+    try:
+        items = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                return Array.from(document.querySelectorAll('button, li, a, span, div'))
+                    .filter(visible)
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        return {
+                            tag: el.tagName,
+                            text: (el.textContent || '').replace(/\s+/g, ' ').trim().slice(0, 80),
+                            className: String(el.className || '').slice(0, 120),
+                            title: el.getAttribute('title') || '',
+                            aria: el.getAttribute('aria-label') || '',
+                            disabled: Boolean(el.disabled),
+                            rect: {
+                                left: Math.round(rect.left),
+                                top: Math.round(rect.top),
+                                width: Math.round(rect.width),
+                                height: Math.round(rect.height),
+                            },
+                        };
+                    })
+                    .filter((item) => item.rect.top > window.innerHeight - 220 && item.rect.left > window.innerWidth * 0.35);
+            }"""
+        )
+        (debug_dir / f"{label}.json").write_text(json.dumps(items, ensure_ascii=False, indent=2), encoding="utf-8")
+        print(f"  [DEBUG] 已保存分页调试信息: {debug_dir / f'{label}.json'}")
+    except Exception as exc:
+        print(f"  [DEBUG] 保存分页调试信息失败: {exc}")
+
+
+def dump_image_debug(page, name: str) -> None:
+    debug_dir = TMP_ROOT / "debug" / "wanxiang_images"
+    debug_dir.mkdir(parents=True, exist_ok=True)
+    label = f"{safe_filename(name)}_{int(time.time())}"
+    try:
+        data = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const attrs = ['src', 'currentSrc', 'data-src', 'data-lazy-src', 'data-original', 'data-ks-lazyload'];
+                const imageInfo = Array.from(document.querySelectorAll('img')).map((img) => {
+                    const rect = img.getBoundingClientRect();
+                    const values = {};
+                    for (const attr of attrs) values[attr] = attr === 'currentSrc' ? img.currentSrc : img.getAttribute(attr);
+                    return {
+                        visible: visible(img),
+                        className: String(img.className || ''),
+                        attrs: values,
+                        outerHTML: img.outerHTML.slice(0, 500),
+                        rect: {
+                            left: Math.round(rect.left),
+                            top: Math.round(rect.top),
+                            width: Math.round(rect.width),
+                            height: Math.round(rect.height),
+                        },
+                    };
+                });
+                const creativeNodes = Array.from(document.querySelectorAll('td, div, span, tr'))
+                    .filter((el) => (el.textContent || '').includes('创意ID'))
+                    .slice(0, 80)
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        return {
+                            visible: visible(el),
+                            tag: el.tagName,
+                            className: String(el.className || ''),
+                            text: (el.textContent || '').replace(/\s+/g, ' ').trim().slice(0, 180),
+                            outerHTML: el.outerHTML.slice(0, 1000),
+                            rect: {
+                                left: Math.round(rect.left),
+                                top: Math.round(rect.top),
+                                width: Math.round(rect.width),
+                                height: Math.round(rect.height),
+                            },
+                        };
+                    });
+                return { url: location.href, imageCount: imageInfo.length, images: imageInfo, creativeNodes };
+            }"""
+        )
+        path = debug_dir / f"{label}.json"
+        path.write_text(json.dumps(data, ensure_ascii=False, indent=2), encoding="utf-8")
+        print(f"  [DEBUG] 已保存图片调试信息: {path}")
+    except Exception as exc:
+        print(f"  [DEBUG] 保存图片调试信息失败: {exc}")
+
+
+def aggregate_rows(rows: list[dict[str, Any]]) -> dict[str, dict[str, Any]]:
+    aggregated: dict[str, dict[str, Any]] = {}
+    for row in rows:
+        creative_id = row["creative_id"]
+        image_key = row.get("creative_image_key") or f"creative_id:{creative_id}"
+        current = aggregated.setdefault(
+            image_key,
+            {
+                "image_key": image_key,
+                "creative_ids": [],
+                "creative_image_url": row.get("creative_image_url", ""),
+                "subject_image_url": row.get("subject_image_url", ""),
+                "impressions": 0.0,
+                "clicks": 0.0,
+                "cost": 0.0,
+                "click_rate": 0.0,
+                "row_count": 0,
+                "sample_rows": [],
+            },
+        )
+        if creative_id and creative_id not in current["creative_ids"]:
+            current["creative_ids"].append(creative_id)
+        current["impressions"] += row["impressions"]
+        current["clicks"] += row["clicks"]
+        current["cost"] += row["cost"]
+        current["row_count"] += 1
+        if not current.get("creative_image_url") and row.get("creative_image_url"):
+            current["creative_image_url"] = row["creative_image_url"]
+        if len(current["sample_rows"]) < 3:
+            current["sample_rows"].append(
+                {
+                    "creative_id": creative_id,
+                    "creative_image_url": row.get("creative_image_url", ""),
+                    "row_text": row.get("row_text", ""),
+                    "row_html": str(row.get("row_html", ""))[:2000],
+                    "cells": row.get("cells", []),
+                }
+            )
+    for item in aggregated.values():
+        impressions = item["impressions"]
+        clicks = item["clicks"]
+        item["click_rate"] = round(clicks / impressions * 100, 4) if impressions else 0.0
+        item["impressions"] = intish(item["impressions"])
+        item["clicks"] = intish(item["clicks"])
+        item["cost"] = round(float(item["cost"]), 4)
+    return aggregated
+
+
+def collect_style(page, style_name: str, max_pages: int, item_ids: list[str] | None = None) -> dict[str, Any]:
+    search_style_by_subject(page, style_name)
+    all_rows: list[dict[str, Any]] = []
+    visited_signatures: set[str] = set()
+
+    # 先检查款式名搜索是否有结果，无结果时用商品ID回退
+    first_check = extract_current_page_rows(page)
+    if not first_check and item_ids:
+        first_id = item_ids[0]
+        print(f"  [FALLBACK] 款式名搜不到结果，用商品ID {first_id} 重搜")
+        search_style_by_subject(page, first_id)
+
+    for page_no in range(1, max_pages + 1):
+        image_map = collect_current_page_image_map(page)
+        rows = extract_current_page_rows(page)
+        for row in rows:
+            image_url = image_map.get(row["creative_id"])
+            if image_url:
+                row["creative_image_url"] = image_url
+                row["creative_image_key"] = canonical_image_key(image_url)
+        if rows and not any(row.get("creative_image_url") for row in rows):
+            dump_image_debug(page, f"{style_name}_page_{page_no}")
+        signature = "|".join(row["creative_id"] for row in rows[:5])
+        print(f"  第 {page_no} 页，抓到 {len(rows)} 行")
+        if signature and signature in visited_signatures:
+            print("  [WARN] 检测到页面重复，停止翻页")
+            break
+        if signature:
+            visited_signatures.add(signature)
+        all_rows.extend(rows)
+        if not click_next_page(page, page_no + 1):
+            break
+
+    aggregated = aggregate_rows(all_rows)
+    return {
+        "style_name": style_name,
+        "collected_at": datetime.now().isoformat(timespec="seconds"),
+        "page_count": len(visited_signatures),
+        "raw_row_count": len(all_rows),
+        "creative_count": len(aggregated),
+        "creatives": sorted(aggregated.values(), key=lambda item: float(item["impressions"]), reverse=True),
+    }
+
+
+def write_style_outputs(result: dict[str, Any], output_dir: Path) -> tuple[Path, Path]:
+    style_name = str(result["style_name"])
+    date_dir = output_dir / now_label()
+    run_dir = date_dir / safe_filename(style_name)
+    run_dir.mkdir(parents=True, exist_ok=True)
+    images_dir = collect_images_dir(run_dir, "tm")
+
+    for item in result.get("creatives", []):
+        image_url = str(item.get("creative_image_url") or "")
+        image_key = safe_filename(str(item.get("image_key") or "unknown"))
+        image_path = images_dir / image_filename(image_key, image_url)
+        downloaded = download_file(image_url, image_path) if image_url else False
+        item["local_image_path"] = str(image_path) if downloaded else ""
+        item["image_downloaded"] = downloaded
+        item["image_fingerprint"] = image_fingerprint(image_path) if downloaded else {}
+
+    original_creative_count = len(result.get("creatives", []))
+    result["creatives"] = merge_creatives_by_visual_hash(result.get("creatives", []))
+    result["creative_count_before_visual_merge"] = original_creative_count
+    result["creative_count"] = len(result["creatives"])
+    result["visual_hash_distance"] = IMAGE_HASH_DISTANCE
+    cleanup_images_after_visual_merge(images_dir, result["creatives"])
+
+    json_path = run_dir / "wanxiang_creative_main_images.json"
+    md_path = run_dir / "wanxiang_creative_main_images.md"
+    json_path.write_text(json.dumps(result, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    lines = [
+        f"# {style_name} 万相创意主图汇总",
+        "",
+        f"- 采集时间: {result.get('collected_at', '')}",
+        f"- 翻页数: {result.get('page_count', 0)}",
+        f"- 原始行数: {result.get('raw_row_count', 0)}",
+        f"- 聚合主图数: {result.get('creative_count', 0)}",
+        "",
+        "| 主图Key | 创意ID列表 | 主图URL | 本地图片 | 展现量 | 点击量 | 点击率 | 花费 |",
+        "| --- | --- | --- | --- | ---: | ---: | ---: | ---: |",
+    ]
+    for item in result.get("creatives", []):
+        lines.append(
+            "| {image_key} | {creative_ids} | {creative_image_url} | {local_image_path} | {impressions} | {clicks} | {click_rate:.4f}% | {cost:.2f} |".format(
+                image_key=str(item.get("image_key", ""))[:120],
+                creative_ids=",".join(item.get("creative_ids", [])),
+                creative_image_url=item.get("creative_image_url", ""),
+                local_image_path=item.get("local_image_path", ""),
+                impressions=item.get("impressions", 0),
+                clicks=item.get("clicks", 0),
+                click_rate=float(item.get("click_rate", 0)),
+                cost=float(item.get("cost", 0)),
+            )
+        )
+    md_path.write_text("\n".join(lines), encoding="utf-8")
+    return json_path, md_path
+
+
+def write_overall_outputs(results: list[dict[str, Any]], output_dir: Path) -> Path:
+    date_dir = output_dir / now_label()
+    date_dir.mkdir(parents=True, exist_ok=True)
+    path = date_dir / f"wanxiang_creative_main_images_all_{now_label()}.json"
+    payload = {
+        "collected_at": datetime.now().isoformat(timespec="seconds"),
+        "style_count": len(results),
+        "results": results,
+    }
+    path.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+    return path
+
+
+def parse_styles(value: str | None) -> list[str]:
+    if not value:
+        return _get_style_names()
+    return [item.strip() for item in re.split(r"[,，\n]+", value) if item.strip()]
+
+
+def run(args: argparse.Namespace) -> None:
+    global _DATE_LABEL
+    output_dir = managed_data_path(args.output)
+    output_dir.mkdir(parents=True, exist_ok=True)
+    styles = parse_styles(args.styles)
+    _DATE_LABEL = args.date or ""
+
+    session = None
+    try:
+        session, page = start_stealthy_session(args.headless, args.user_data_dir)
+        ensure_logged_in(page, args.account, args.password)
+        goto_wanxiang_from_seller(page)
+        goto_creative_report(page)
+        select_report_template(page, args.report_template)
+
+        results: list[dict[str, Any]] = []
+        for index, style_name in enumerate(styles, 1):
+            print(f"\n[{index}/{len(styles)}] 开始采集: {style_name}")
+            item_ids = STYLE_ITEM_IDS.get(style_name, [])
+            result = collect_style(page, style_name, args.max_pages, item_ids=item_ids)
+            results.append(result)
+            json_path, md_path = write_style_outputs(result, output_dir)
+            print(f"  [OK] 聚合 JSON: {json_path}")
+            print(f"  [OK] 聚合 MD:   {md_path}")
+
+        overall_path = write_overall_outputs(results, output_dir)
+        print(f"\n[OK] 全部款式汇总: {overall_path}")
+
+        if args.keep_open:
+            print("[INFO] --keep-open 已开启，浏览器保留 60 秒")
+            time.sleep(60)
+    finally:
+        if session:
+            session.close()
+
+
+def build_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(description="采集万相 AI 无界创意报表主图聚合数据")
+    parser.add_argument("--account", default=ACCOUNT, help="淘宝卖家账号")
+    parser.add_argument("--password", default=PASSWORD, help="淘宝卖家密码")
+    parser.add_argument("--user-data-dir", default=USER_DATA_DIR, help="浏览器 profile 目录，用于复用登录态")
+    parser.add_argument("--output", default=str(OUTPUT_DIR), help="聚合结果输出目录")
+    parser.add_argument("--date", default=None,
+                        help="采集日期 YYYY-MM-DD,决定输出子目录名(默认 = today)。"
+                             "weekly pipeline 传「上周日」以与 insert 阶段对齐。")
+    parser.add_argument("--styles", default="", help="只采集指定款式，多个用逗号分隔；默认采集全部款式")
+    parser.add_argument("--max-pages", type=int, default=100, help="每个款式最多翻页数")
+    parser.add_argument(
+        "--report-template",
+        default=DEFAULT_REPORT_TEMPLATE,
+        choices=REPORT_TEMPLATE_OPTIONS,
+        help="创意报表模板，默认只采集关键词推广",
+    )
+    parser.add_argument("--headless", action="store_true", help="无头模式")
+    parser.add_argument("--keep-open", action="store_true", help="结束后保留浏览器 60 秒")
+    return parser
+
+
+if __name__ == "__main__":
+    run(build_parser().parse_args())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/upload_video_to_guanghe.py b/src/gyxx_flow/modules/product_commerce/runtime/upload_video_to_guanghe.py
new file mode 100644
index 0000000..282a3ea
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/upload_video_to_guanghe.py
@@ -0,0 +1,3048 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+读取飞书多维表格中「天猫上传」未勾选的短视频记录，
+下载视频附件，上传到淘宝光合平台，成功后回写表格为已勾选。
+
+表信息:
+  Base: https://bu0zgpibak.feishu.cn/base/TH1JbxhfUaxetis6F4pcoqvWncd
+  表 ID: tblGV2mO8t7aEO14
+
+运行前准备:
+  1. 在 .env 中配置 GUANGHE_USERNAME / GUANGHE_PASSWORD
+  2. 安装依赖: pip install python-dotenv playwright scrapling requests
+  3. 首次运行可能触发滑块验证，按提示处理；登录态会复用（默认保存到本地目录）
+
+用法:
+  python upload_video_to_guanghe.py --limit 1 --debug
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import shutil
+import subprocess
+import sys
+import time
+import traceback
+import urllib.request
+from pathlib import Path
+
+from runtime_paths import TMP_ROOT, profile_path
+from typing import Any, Dict, List, Optional, Tuple
+
+import io
+
+from dotenv import load_dotenv
+
+# 加载环境变量
+load_dotenv()
+
+# Windows GBK 环境下强制 stdout/stderr 为 UTF-8，并启用行缓冲
+# （裸 TextIOWrapper 默认 8KB 缓冲，redirect 到文件时不便观察进度；
+#  加 line_buffering 后每条 print 立刻刷盘）
+if sys.stdout.encoding != "utf-8":
+    sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8", errors="replace", line_buffering=True)
+if sys.stderr.encoding != "utf-8":
+    sys.stderr = io.TextIOWrapper(sys.stderr.buffer, encoding="utf-8", errors="replace", line_buffering=True)
+
+PROJECT_ROOT = Path(__file__).resolve().parent
+sys.path.insert(0, str(PROJECT_ROOT))
+
+from scrapling.fetchers import StealthySession
+
+USER_DATA_DIR = os.getenv("GUANGHE_USER_DATA_DIR", str(profile_path("guanghe")))
+LUGGAGE_USER_DATA_DIR = os.getenv(
+    "GUANGHE_LUGGAGE_USER_DATA_DIR",
+    str(profile_path("guanghe-luggage")),
+)
+
+# 复用 analyze_style_with_hermes 的 Hermes 调用（analyzer 端，端口 8642）
+try:
+    from analyze_style_with_hermes import call_hermes as _call_hermes, HERMES_API_KEY, HERMES_ANALYZER_URL
+    HAS_HERMES = True
+except Exception as _e:  # pragma: no cover
+    HAS_HERMES = False
+    _call_hermes = None
+    HERMES_API_KEY = ""
+    HERMES_ANALYZER_URL = ""
+    print(f"[WARN] 无法 import Hermes（不影响主流程）: {_e}", file=sys.stderr)
+
+# ─────────────────────────── 配置 ───────────────────────────
+BASE_TOKEN = "TH1JbxhfUaxetis6F4pcoqvWncd"
+TABLE_ID = "tblGV2mO8t7aEO14"
+LUGGAGE_STYLE_TABLE_ID = "tbl7sscXvLFM4tBe"
+LUGGAGE_STYLE_FIELD = "文本"
+TM_UPLOAD_FIELD = "天猫上传"
+TABLE_FIELDS = [
+    "代拍产品",
+    "代拍附件",
+    TM_UPLOAD_FIELD,
+    "日期",
+    "拍摄类型",
+    "负责人",
+    "品牌",
+    "性别",
+]
+
+STORE_FLAGSHIP = "flagship"
+STORE_LUGGAGE = "luggage"
+STORE_ALL = "all"
+
+VIDEO_EXTS = {".mp4", ".mov", ".m4v", ".avi", ".mkv", ".flv", ".wmv"}
+DOWNLOAD_DIR = TMP_ROOT / "videos"
+DEBUG_DIR = TMP_ROOT / "debug" / "guanghe_upload"
+
+GUANGHE_URL = "https://creator.guanghe.taobao.com/page/"
+
+HEADLESS = False  # 视频上传需要可视化，便于观察进度和调试
+
+
+# ─────────────────────────── lark-cli 工具 ───────────────────────────
+def _lark_cli() -> str:
+    """返回 lark-cli 可执行文件路径。"""
+    cli = shutil.which("lark-cli")
+    if cli:
+        return cli
+    # Windows 常见安装路径
+    candidates = [
+        Path.home() / "AppData" / "Roaming" / "npm" / "lark-cli.cmd",
+        Path.home() / "AppData" / "Roaming" / "npm" / "lark-cli",
+    ]
+    for c in candidates:
+        if c.exists():
+            return str(c)
+    raise RuntimeError("找不到 lark-cli，请先 npm install -g @larksuiteorg/lark-cli")
+
+
+def run_lark(args: List[str]) -> Tuple[int, str, str]:
+    """调用 lark-cli 并返回 (returncode, stdout, stderr)。"""
+    cmd = [_lark_cli()] + args
+    try:
+        proc = subprocess.run(cmd, capture_output=True, text=True, encoding="utf-8", timeout=120)
+    except subprocess.TimeoutExpired:
+        return -1, "", "lark-cli 调用超时"
+    return proc.returncode, proc.stdout, proc.stderr
+
+
+def _kill_guanghe_chrome(user_data_dir: str = USER_DATA_DIR):
+    """强制关闭所有由本脚本启动的 Chrome 进程（按 user_data_dir 路径精确匹配），
+    不会影响用户自己打开的普通 Chrome。
+
+    原因：scrapling 的 StealthySession 在 real_chrome/persistent 模式下，
+    self.browser 始终为 None，其 close() 不会调用 browser.close()，
+    导致 Chrome 进程残留、每次运行都新开一个窗口（“乱弹”）。这里兜底强杀。
+    """
+    marker = str(user_data_dir).replace("/", "\\")
+    try:
+        if sys.platform.startswith("win"):
+            ps_cmd = (
+                "Get-CimInstance Win32_Process -Filter \"Name='chrome.exe'\" | "
+                "Where-Object { $_.CommandLine -and $_.CommandLine -like '*" + marker + "*' } | "
+                "ForEach-Object { taskkill /PID $_.ProcessId /F /T }"
+            )
+            subprocess.run(
+                ["powershell", "-NoProfile", "-Command", ps_cmd],
+                capture_output=True, text=True, timeout=60,
+            )
+        else:
+            out = subprocess.run(["pgrep", "-f", marker], capture_output=True, text=True, timeout=30).stdout
+            for pid in out.split():
+                try:
+                    os.kill(int(pid), 9)
+                except Exception:
+                    pass
+    except Exception:
+        pass
+
+
+def fetch_records(limit: Optional[int] = None,
+                  product_name_contains: Optional[str] = None,
+                  max_total: int = 5000) -> List[Dict[str, Any]]:
+    """从飞书 Base 读取目标记录，返回包含 record_id 的字典列表。
+
+    - product_name_contains: 服务端 contains 过滤（代拍产品字段包含该子串）
+    - limit: 每页大小 1-200；不传则 200（lark-cli 单页上限）
+    - max_total: 全量上限（防意外超长表死循环），默认 5000
+    """
+    page_size = min(max(limit or 200, 1), 200)
+    cmd_base = [
+        "base", "+record-list",
+        "--base-token", BASE_TOKEN,
+        "--table-id", TABLE_ID,
+        "--format", "json",
+    ]
+    for field in TABLE_FIELDS:
+        cmd_base += ["--field-id", field]
+    if product_name_contains:
+        filter_json = (
+            '{"logic":"and","conditions":[["代拍产品","intersects","'
+            + product_name_contains.replace('"', '\\"')
+            + '"]]}'
+        )
+        cmd_base += ["--filter-json", filter_json]
+
+    all_results: List[Dict[str, Any]] = []
+    offset = 0
+    while len(all_results) < max_total:
+        cmd = cmd_base + ["--limit", str(page_size), "--offset", str(offset)]
+        rc, out, err = run_lark(cmd)
+        if rc != 0:
+            print(f"[lark-cli] 读取记录失败 (offset={offset}): rc={rc}\n{err}", file=sys.stderr)
+            break
+        try:
+            data = json.loads(out)
+        except json.JSONDecodeError:
+            print(f"[lark-cli] 返回非 JSON: {out[:500]}", file=sys.stderr)
+            break
+
+        inner = data.get("data", {}) or {}
+        fields = inner.get("fields", []) or []
+        rows = inner.get("data", []) or []
+        record_ids = inner.get("record_id_list", []) or []
+        has_more = inner.get("has_more", False)
+
+        added_this_page = 0
+        for i, row in enumerate(rows):
+            if len(all_results) >= max_total:
+                break
+            rec = {}
+            for j, field_name in enumerate(fields):
+                rec[field_name] = row[j] if j < len(row) else None
+            if i < len(record_ids):
+                rec["record_id"] = record_ids[i]
+            all_results.append(rec)
+            added_this_page += 1
+
+        if not has_more or added_this_page < page_size:
+            break
+        offset += page_size
+    return all_results
+
+
+def _normalize_style_name(value: Any) -> str:
+    return re.sub(r"\s+", "", str(value or "")).casefold()
+
+
+def _style_db_lookup_key(value: Any) -> str:
+    """Normalize a Feishu style label to the canonical dim_style lookup key."""
+    compact = re.sub(r"\s+", "", str(value or ""))
+    compact = re.sub(r"系列.*$", "", compact).strip("-—_·:：")
+    return compact.casefold()
+
+
+def _match_tm_item_ids_from_product_titles(
+    style_names: List[str],
+    catalog_rows: List[Tuple[Any, Any]],
+) -> Dict[str, List[str]]:
+    """Resolve style aliases by finding their exact text in database product titles."""
+    result: Dict[str, List[str]] = {}
+    normalized_rows = [
+        (
+            re.sub(r"\D", "", str(product_id or "")),
+            _style_db_lookup_key(product_name),
+        )
+        for product_id, product_name in catalog_rows
+    ]
+    for style_name in style_names:
+        style_key = _style_db_lookup_key(style_name)
+        if len(style_key) < 2:
+            continue
+        matched_ids: List[str] = []
+        for product_id, title_key in normalized_rows:
+            if product_id and style_key in title_key and product_id not in matched_ids:
+                matched_ids.append(product_id)
+        if matched_ids:
+            result[style_name] = matched_ids
+    return result
+
+
+def load_tm_item_ids_for_styles(style_names: List[str]) -> Dict[str, List[str]]:
+    """Read exact Tmall item IDs for the requested styles from dim_style.tm_spus."""
+    requested = [str(name).strip() for name in style_names if str(name).strip()]
+    if not requested:
+        return {}
+
+    requested_by_key: Dict[str, List[str]] = {}
+    for name in requested:
+        requested_by_key.setdefault(_style_db_lookup_key(name), []).append(name)
+
+    from db import get_conn
+
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT style_name, tm_spus
+                FROM dim_style
+                WHERE tm_spus IS NOT NULL AND cardinality(tm_spus) > 0
+                """
+            )
+            rows = cur.fetchall()
+
+    result: Dict[str, List[str]] = {}
+    for db_style_name, tm_spus in rows:
+        matching_names = requested_by_key.get(_style_db_lookup_key(db_style_name), [])
+        if not matching_names:
+            continue
+        clean_ids: List[str] = []
+        for raw_id in tm_spus or []:
+            item_id = re.sub(r"\D", "", str(raw_id or ""))
+            if item_id and item_id not in clean_ids:
+                clean_ids.append(item_id)
+        for requested_name in matching_names:
+            result[requested_name] = list(clean_ids)
+
+    unresolved = [name for name in requested if name not in result]
+    all_dim_item_ids = sorted({
+        re.sub(r"\D", "", str(raw_id or ""))
+        for _db_style_name, tm_spus in rows
+        for raw_id in (tm_spus or [])
+        if re.sub(r"\D", "", str(raw_id or ""))
+    })
+    if unresolved and all_dim_item_ids:
+        with get_conn() as conn:
+            with conn.cursor() as cur:
+                cur.execute(
+                    """
+                    SELECT DISTINCT product_id, product_name
+                    FROM product_daily_metrics
+                    WHERE platform = 'tm'
+                      AND product_id = ANY(%s)
+                      AND NULLIF(BTRIM(product_name), '') IS NOT NULL
+                    """,
+                    (all_dim_item_ids,),
+                )
+                catalog_rows = cur.fetchall()
+        result.update(
+            _match_tm_item_ids_from_product_titles(unresolved, catalog_rows)
+        )
+    return result
+
+
+def load_tm_product_titles_for_styles(
+    tm_item_ids_by_style: Dict[str, List[str]],
+) -> Dict[str, List[str]]:
+    """Read the latest known Tmall product titles for exact style item IDs."""
+    item_ids = sorted({
+        str(item_id).strip()
+        for ids in tm_item_ids_by_style.values()
+        for item_id in ids
+        if str(item_id).strip()
+    })
+    if not item_ids:
+        return {}
+
+    from db import get_conn
+
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT DISTINCT product_id, product_name
+                FROM product_daily_metrics
+                WHERE platform = 'tm'
+                  AND product_id = ANY(%s)
+                  AND NULLIF(BTRIM(product_name), '') IS NOT NULL
+                """,
+                (item_ids,),
+            )
+            rows = cur.fetchall()
+
+    titles_by_id: Dict[str, List[str]] = {}
+    for product_id, product_name in rows:
+        title = str(product_name or "").strip()
+        if title and title not in titles_by_id.setdefault(str(product_id), []):
+            titles_by_id[str(product_id)].append(title)
+
+    result: Dict[str, List[str]] = {}
+    for style_name, ids in tm_item_ids_by_style.items():
+        titles: List[str] = []
+        for item_id in ids:
+            for title in titles_by_id.get(str(item_id), []):
+                if title not in titles:
+                    titles.append(title)
+        if titles:
+            result[style_name] = titles
+    return result
+
+
+def fetch_luggage_style_whitelist() -> set[str]:
+    """Read the complete luggage-store style whitelist from its Base table."""
+    names: set[str] = set()
+    offset = 0
+    page_size = 200
+    while True:
+        rc, out, err = run_lark([
+            "base",
+            "+record-list",
+            "--base-token",
+            BASE_TOKEN,
+            "--table-id",
+            LUGGAGE_STYLE_TABLE_ID,
+            "--field-id",
+            LUGGAGE_STYLE_FIELD,
+            "--limit",
+            str(page_size),
+            "--offset",
+            str(offset),
+            "--format",
+            "json",
+        ])
+        if rc != 0:
+            raise RuntimeError(f"读取箱包店款式白名单失败: {err}")
+        try:
+            inner = (json.loads(out).get("data") or {})
+        except json.JSONDecodeError as exc:
+            raise RuntimeError("箱包店款式白名单返回非 JSON") from exc
+        rows = inner.get("data") or []
+        for row in rows:
+            value = row[0] if row else None
+            normalized = _normalize_style_name(value)
+            if normalized:
+                names.add(normalized)
+        if not inner.get("has_more"):
+            break
+        offset += page_size
+    if not names:
+        raise RuntimeError("箱包店款式白名单为空，停止上传")
+    return names
+
+
+def mark_tm_uploaded(record_id: str) -> bool:
+    """将「天猫上传」更新为 true。
+
+    2026-07-21 lark-cli 升级: subcommand 从 +record-update 改成 +record-upsert,
+    参数从 --payload (wrapped in fields) 改成 --json (top-level field map)。
+    """
+    payload = json.dumps({TM_UPLOAD_FIELD: True}, ensure_ascii=False)
+    for subcmd, flag in (("+record-upsert", "--json"), ("+record-update", "--payload")):
+        cmd = [
+            "base", subcmd,
+            "--base-token", BASE_TOKEN,
+            "--table-id", TABLE_ID,
+            "--record-id", record_id,
+            flag, payload,
+        ]
+        rc, out, err = run_lark(cmd)
+        if rc == 0:
+            print(f"[OK] 已回写{TM_UPLOAD_FIELD}: {record_id}")
+            return True
+        # 旧 CLI 报 unknown subcommand 时继续尝试下一个 subcmd
+        if "unknown subcommand" in err:
+            continue
+        # 旧 CLI 报 unknown flag 时也跳过
+        if "unknown flag" in err:
+            continue
+        print(f"[lark-cli] 更新记录失败 ({subcmd}): {record_id} {err}", file=sys.stderr)
+        return False
+    print(f"[ERR] 更新记录失败,新旧 lark-cli 都不可用", file=sys.stderr)
+    return False
+
+
+# ─────────────────────────── 附件下载 ───────────────────────────
+def download_videos(record_id: str, attachments: List[Dict[str, Any]]) -> List[Path]:
+    """下载记录里所有识别到的视频附件，按 attachments 顺序返回本地路径列表。
+    任一附件下载失败会被跳过；返回 [] 表示这条记录没有任何可上传的视频。"""
+    if not attachments:
+        return []
+    videos = [
+        a for a in attachments
+        if a.get("type") == "file" or Path(a.get("name", "")).suffix.lower() in VIDEO_EXTS
+    ]
+    if not videos:
+        videos = [a for a in attachments if Path(a.get("name", "")).suffix.lower() in VIDEO_EXTS]
+    if not videos:
+        print(f"[跳过] 记录 {record_id} 没有识别到视频附件")
+        return []
+
+    DOWNLOAD_DIR.mkdir(parents=True, exist_ok=True)
+    safe_record_id = re.sub(r"[^A-Za-z0-9_-]", "_", str(record_id))
+    record_download_dir = DOWNLOAD_DIR / safe_record_id
+    record_download_dir.mkdir(parents=True, exist_ok=True)
+    paths: List[Path] = []
+    for idx, att in enumerate(videos):
+        file_token = att.get("file_token")
+        file_name = att.get("name", f"video_{idx}.mp4")
+        if not file_token:
+            print(f"[跳过] 记录 {record_id} 第 {idx+1} 个附件缺少 file_token")
+            continue
+        output_path = record_download_dir / file_name
+
+        cmd = [
+            "base", "+record-download-attachment",
+            "--base-token", BASE_TOKEN,
+            "--table-id", TABLE_ID,
+            "--record-id", record_id,
+            "--file-token", file_token,
+            "--output", str(record_download_dir),
+            "--overwrite",
+        ]
+        rc, out, err = run_lark(cmd)
+        if rc != 0:
+            print(f"[下载失败] {record_id} {file_name}: {(err or '').strip()[:200]}", file=sys.stderr)
+            continue
+
+        if not output_path.exists():
+            # lark-cli 下载后会按 file_token/原名落盘；找不到时退回到「最新视频文件」兜底
+            candidates = sorted(
+                [
+                    p
+                    for p in record_download_dir.iterdir()
+                    if p.suffix.lower() in VIDEO_EXTS
+                ],
+                key=lambda p: p.stat().st_mtime,
+                reverse=True,
+            )
+            if candidates:
+                output_path = candidates[0]
+            else:
+                print(f"[下载失败] 未找到下载文件: {record_id} {file_name}", file=sys.stderr)
+                continue
+
+        size_mb = output_path.stat().st_size / 1024 / 1024
+        print(f"[OK] 已下载 [{idx+1}/{len(videos)}]: {output_path.name} ({size_mb:.2f} MB)")
+        paths.append(output_path)
+    return paths
+
+
+def download_video(record_id: str, attachments: List[Dict[str, Any]]) -> Optional[Path]:
+    """向后兼容：只返回第一个视频（脚本早期按 1 视频/记录的简化实现）。"""
+    paths = download_videos(record_id, attachments)
+    return paths[0] if paths else None
+
+
+def ensure_upload_compatible_video(video_path: Path) -> Path:
+    """将 HEVC 或宽/高不达标的视频转为光合兼容 H.264/AAC、1080P。"""
+    try:
+        import imageio_ffmpeg
+        ffmpeg_exe = imageio_ffmpeg.get_ffmpeg_exe()
+        reader = imageio_ffmpeg.read_frames(str(video_path))
+        metadata = next(reader)
+        reader.close()
+    except Exception as exc:
+        raise RuntimeError(
+            "无法读取视频编码/尺寸；请确认已安装 imageio-ffmpeg"
+        ) from exc
+
+    codec = str(metadata.get("codec", "")).lower()
+    width, height = metadata.get("source_size") or metadata.get("size") or (0, 0)
+    needs_codec = codec in {"hevc", "h265"}
+    needs_scale = bool(width and height and (width <= 720 or height <= 720))
+    if not needs_codec and not needs_scale:
+        return video_path
+
+    output_path = video_path.with_name(f"{video_path.stem}_compatible.mp4")
+    if (
+        output_path.exists()
+        and output_path.stat().st_size > 0
+        and output_path.stat().st_mtime >= video_path.stat().st_mtime
+    ):
+        print(f"[OK] 复用 H.264 兼容视频: {output_path.name}")
+        return output_path
+
+    reasons = []
+    if needs_codec:
+        reasons.append(f"编码={codec}")
+    if needs_scale:
+        reasons.append(f"尺寸={width}x{height}")
+    print(
+        f"[->] 检测到光合不兼容视频（{', '.join(reasons)}），"
+        f"转为 H.264/1080P: {video_path.name}"
+    )
+    cmd = [
+        ffmpeg_exe,
+        "-y",
+        "-hide_banner",
+        "-loglevel", "warning",
+        "-i", str(video_path),
+        "-map", "0:v:0",
+        "-map", "0:a?",
+        "-c:v", "libx264",
+        "-preset", "fast",
+        "-crf", "20",
+        "-pix_fmt", "yuv420p",
+        "-c:a", "aac",
+        "-b:a", "192k",
+        "-movflags", "+faststart",
+    ]
+    if needs_scale:
+        if width <= height:
+            cmd.extend(["-vf", "scale=1080:-2"])
+        else:
+            cmd.extend(["-vf", "scale=-2:1080"])
+    cmd.append(str(output_path))
+    proc = subprocess.run(cmd, text=True, capture_output=True, check=False)
+    if proc.returncode != 0 or not output_path.exists() or output_path.stat().st_size == 0:
+        raise RuntimeError(
+            f"视频兼容转码失败: {(proc.stderr or proc.stdout).strip()[-500:]}"
+        )
+    print(
+        f"[OK] 兼容转码完成: {output_path.name} "
+        f"({output_path.stat().st_size / 1024 / 1024:.2f} MB)"
+    )
+    return output_path
+
+
+# ─────────────────────────── 浏览器/光合平台 ───────────────────────────
+class GuangHeUploader:
+    FIXED_TOPIC = "在淘宝种草一夏"
+    FIXED_CONTENT_LABELS = ("商品展示",)
+    TITLE_SIMILARITY_LIMIT = 0.32
+    TITLE_VARIANTS = (
+        ("钥匙卡包各归其位,早高峰取物不用来回翻找", "快速取物", "随身小物各归其位,拿取不用翻找"),
+        ("水杯雨伞随手装好,短途出门少些手忙脚乱", "短途准备", "出门用品集中带齐,短途少些手忙"),
+        ("文件耳机分开放置,办公路上拿取更加顺手", "办公携带", "办公用品分别放好,途中拿取顺手"),
+        ("日常随身物有处可放,临时出门快速完成打包", "临时出门", "随身物件有处可放,出门快速打包"),
+        ("简洁轮廓适配日常,上班休闲搭配切换自然", "日常搭配", "简洁轮廓适配日常,搭配切换自然"),
+        ("肩背拿取节奏利落,赶车转场也不会显得拖沓", "赶车转场", "拿取节奏保持利落,赶车不拖沓"),
+        ("常用小物一眼找到,进出地铁不用来回翻找", "地铁出行", "常用小物一眼找到,不用反复翻找"),
+        ("衣物配件按需装入,周末短途说走就走更省心", "周末短途", "衣物配件按需装入,周末轻松启程"),
+        ("桌面物件集中带走,临时换个工位也不会忙乱", "工位切换", "桌面物件集中带走,换工位不忙乱"),
+        ("随身装备整理妥当,晨间出门准备更加有效率", "晨间准备", "随身装备整理妥当,出门有效率"),
+        ("工作用品稳妥归置,会议通勤之间从容切换", "会议通勤", "工作用品稳妥归置,行程从容切换"),
+        ("零碎小物减少散落,一整天行程都能保持利落", "防止散落", "零碎小物减少散落,行程保持利落"),
+        ("出门所需集中装下,临时安排也能够轻松应对", "临时安排", "出门所需集中装下,临时安排不慌"),
+        ("不同用途分别放置,拿取顺序清楚不会忙乱", "用途分类", "物品按用途放置,拿取顺序清楚"),
+        ("日常搭配保持简洁,城市出行轻松不会费劲", "城市搭配", "日常搭配保持简洁,城市出行轻松"),
+        ("随手物件安排到位,忙碌途中少些来回寻找", "忙碌行程", "随手物件安排到位,途中少找东西"),
+    )
+    # 每个主情境可与全部结果表达交叉组合；一次全量发布 48 条时仍有充足余量。
+    TITLE_VARIATION_COUNT = len(TITLE_VARIANTS) * len(TITLE_VARIANTS)
+    TITLE_LEAD_REFRAMES = (
+        ("门卡耳机放在手边", "高频小物随手可取"),
+        ("出门用品提前带齐", "日常所需一次备好"),
+        ("办公物件各自放好", "文件配件分开放置"),
+        ("随身用品安置妥当", "日用物件都有位置"),
+        ("清爽造型融入日常", "利落外形轻松搭配"),
+        ("肩背动作轻快干脆", "移动拿取保持利落"),
+        ("常带小件快速找到", "高频用品不用细找"),
+        ("短途用品按需带上", "周末物件一次备齐"),
+        ("工位物品随时带走", "桌面所需快速转移"),
+        ("每日装备提前理好", "随行用品准备妥当"),
+        ("工作所需有序带齐", "办公装备稳妥随行"),
+        ("细碎物件不再散落", "零散用品集中放好"),
+        ("外出所需一次带上", "当天用品集中备齐"),
+        ("随行物品分开放好", "不同物件各有位置"),
+        ("清爽搭配适合日常", "利落造型融入城市"),
+        ("常用物件提前放好", "随行小件准备到位"),
+    )
+    TITLE_TAIL_REFRAMES = (
+        ("高峰进站拿取不慌", "赶早出门不用翻包"),
+        ("短程安排从容一点", "说走就走不用忙乱"),
+        ("工作路上随拿随用", "办公移动更加顺手"),
+        ("突然出门也能快走", "临时启程不用久等"),
+        ("工作休闲自然过渡", "不同场合搭配省心"),
+        ("换乘赶路保持轻快", "连续转场依旧利落"),
+        ("刷卡换乘不耽时间", "上下地铁少些停顿"),
+        ("周末启程更加轻松", "两天行程准备省心"),
+        ("更换位置不再慌乱", "移动办公收拾更快"),
+        ("晨间收拾不再拖延", "每天出门更加高效"),
+        ("会议办公切换自然", "工作行程衔接从容"),
+        ("全天移动依旧整齐", "忙完一天仍然利落"),
+        ("行程变化也能应对", "突然加项心里不慌"),
+        ("使用时一眼看清楚", "需要什么快速拿到"),
+        ("城市移动轻松自然", "日常上身简洁耐看"),
+        ("赶路途中少翻几次", "忙碌时候拿取不乱"),
+    )
+    # 本地采集的官方商品标题/商品档案优先于联网摘要。联网结果里经常同时出现
+    # “电脑包、单肩包、斜挎包”等用途词，不能用这些词覆盖商品的主品类。
+    PRODUCT_CATEGORY_OVERRIDES = {
+        "波塞冬edge": "托特包",
+        "星迹2": "斜挎包",
+    }
+    CATALOG_CATEGORY_OVERRIDES: Dict[str, str] = {}
+    CATALOG_PRODUCT_INFO: Dict[str, str] = {}
+    VERIFIED_PRODUCT_INFO = {
+        "波塞冬edge": (
+            "主包型：托特包；尺寸：约39×30×14cm；容量/适配：约17L、适配16寸笔记本电脑；"
+            "材质：1260D考杜拉面料；卖点：防泼水拉链、多分区收纳、手提/单肩/斜挎多种背法"
+        ),
+        "星迹2": (
+            "主包型：斜挎包；尺寸：未查到；容量/适配：中号适配11寸平板；材质：未查到；"
+            "卖点：主袋/前袋/内袋科学分区、轻量便携、通勤出行"
+        ),
+    }
+
+    """基于 Scrapling/Playwright 的光合平台上传器。"""
+
+    def __init__(
+        self,
+        username: str,
+        password: str,
+        headless: bool = False,
+        debug: bool = False,
+        manual: bool = False,
+        user_data_dir: str = USER_DATA_DIR,
+    ):
+        self.username = username
+        self.login_secret = password
+        self.headless = headless
+        self.debug = debug
+        self.manual = manual
+        self.user_data_dir = user_data_dir
+        self.session = None
+        self.page = None
+        # 发布页表单实际嵌在 huodong.taobao.com 的 iframe 里（2026-07-21 实测），
+        # Playwright locator 不穿透 iframe，所有发布页交互都要路由到这个 frame
+        self.pub_frame = None
+        self._used_titles_by_products: Dict[Tuple[str, ...], set[str]] = {}
+        self._used_titles_global: List[str] = []
+        self._title_variant_usage: Dict[int, int] = {}
+
+    def _snapshot(self, name: str):
+        if not self.debug or not self.page:
+            return
+        DEBUG_DIR.mkdir(parents=True, exist_ok=True)
+        ts = time.strftime("%H%M%S")
+        try:
+            self.page.screenshot(path=str(DEBUG_DIR / f"{name}_{ts}.png"), full_page=False)
+        except Exception:
+            pass
+
+    def start(self):
+        """启动浏览器并打开光合平台。"""
+        # 先清掉上一次残留的脚本 Chrome 窗口，避免“乱弹”
+        print("[->] 清理上一次残留的浏览器窗口...")
+        _kill_guanghe_chrome(self.user_data_dir)
+        time.sleep(1)
+        print("[->] 正在启动 Chrome 浏览器（非 headless，你应该能看到窗口弹出）...")
+        self.session = StealthySession(
+            headless=self.headless,
+            real_chrome=True,
+            user_data_dir=self.user_data_dir,
+            locale="zh-CN",
+            timezone_id="Asia/Shanghai",
+            timeout=60000,
+            network_idle=False,
+            disable_resources=False,
+            block_ads=False,
+            # 关闭"Chrome 未正常关闭"恢复提示弹窗（_kill_guanghe_chrome 强杀后会出现）
+            extra_flags=["--disable-session-crashed-bubble", "--disable-features=InfiniteSessionRestore"],
+        )
+        self.session.start()
+        context = self.session.context
+        self.page = context.pages[-1] if context.pages else context.new_page()
+        try:
+            self.page.set_viewport_size({"width": 1920, "height": 1080})
+        except Exception:
+            pass
+        print(f"[OK] Chrome 已启动，当前页面: {self.page.url}")
+        self.page.goto(GUANGHE_URL, wait_until="domcontentloaded", timeout=60000)
+        # 淘宝可能跳到新 tab，需要重新同步 self.page 指向真正显示登录页/后台的 page
+        self._sync_page()
+        print(f"[OK] 已打开光合平台，当前 URL: {self.page.url}")
+        self._snapshot("01_open")
+
+    def _sync_page(self):
+        """淘宝登录可能跳到新 tab，遍历 context.pages 找到真正在 creator 后台的 page 切过去。"""
+        try:
+            context = self.session.context
+            for p in context.pages:
+                url = (p.url or "").lower()
+                if "creator.guanghe.taobao.com" in url and "login" not in url and "passport" not in url:
+                    if p is not self.page:
+                        print(f"    [同步] 切到后台 page: {p.url[:80]}")
+                        self.page = p
+                    return True
+        except Exception:
+            pass
+        return False
+
+    def _ensure_login_page(self):
+        """在所有 page 里找含登录表单（#fm-login-id）的 page 切过去，
+        避免 self.page 指向空白/中间 page 导致自动填表静默失败。"""
+        try:
+            context = self.session.context
+            for p in context.pages:
+                try:
+                    if p.locator("#fm-login-id").count() > 0:
+                        if p is not self.page:
+                            print(f"    [同步] 切到登录 page: {p.url[:80]}")
+                            self.page = p
+                        return True
+                except Exception:
+                    continue
+        except Exception:
+            pass
+        return False
+
+    def _wait_home_loaded(self, timeout: int = 30) -> bool:
+        """等光合平台首页 SPA 真正加载完（关键元素出现），避免点不到 +发布作品。"""
+        markers = ["平台爆品", "内容灵感", "创作者服务", "+ 发布作品", "发布作品"]
+        deadline = time.time() + timeout
+        while time.time() < deadline:
+            for m in markers:
+                try:
+                    if self.page.locator(f"text={m}").count() > 0:
+                        print(f"    [OK] 后台首页已加载（命中: {m}）")
+                        return True
+                except Exception:
+                    continue
+            time.sleep(0.5)
+        print(f"[WARN] {timeout} 秒未等到后台首页完全加载")
+        return False
+
+    def _is_logged_in(self) -> bool:
+        """判断是否已登录：URL 域名判定优先，元素判定作兜底。"""
+        try:
+            # 1) URL 判定（最稳）：在 creator 域且不在登录关键字
+            url = (self.page.url or "").lower()
+            # 诊断：每个新 URL 第一次出现时打印
+            if not hasattr(self, "_logged_urls"):
+                self._logged_urls = set()
+            if url not in self._logged_urls:
+                self._logged_urls.add(url)
+                print(f"    [诊断 _is_logged_in] URL: {url[:120]}")
+            if "creator.guanghe.taobao.com" in url:
+                # 把 "login" 加回，避免 creator 域内 login 路径误判
+                login_keywords = ["login", "passport", "havanalogin", "authorize", "qrcode"]
+                if not any(kw in url for kw in login_keywords):
+                    return True
+            # 2) 元素兜底（短超时逐个查，避免 locator 逗号语法报错吞掉 URL 判定）
+            for m in ["发布作品", "创作者服务", "退出登录", "内容创作", "光合平台"]:
+                try:
+                    if self.page.locator(f"text={m}").count() > 0:
+                        return True
+                except Exception:
+                    continue
+        except Exception:
+            pass
+        return False
+
+    def ensure_login(self):
+        """确保登录到光合平台。"""
+        if not self.page:
+            self.start()
+
+        # 先尝试切到真正含登录表单的 page（避免 self.page 指向空白/中间 page 导致自动填表静默失败）
+        self._ensure_login_page()
+        if self._is_logged_in():
+            print("[OK] 检测到已登录（复用登录态）")
+            self._wait_home_loaded()
+            return
+
+        # 先尝试同步到真正的后台 page
+        self._sync_page()
+        if self._is_logged_in():
+            print("[OK] 检测到已登录（复用登录态）")
+            self._wait_home_loaded()
+            return
+
+        if self.manual:
+            print("[->] 手动登录模式：请在浏览器中完成登录（点登录/滑块/扫码等）")
+            print("    登录成功后脚本会自动继续，无需关闭窗口")
+            for i in range(120):
+                time.sleep(1)
+                self._sync_page()
+                if i % 10 == 0 and i > 0:
+                    print(f"    [诊断] 当前 URL: {self.page.url[:100]}")
+                if self._is_logged_in():
+                    print("[OK] 手动登录成功")
+                    self._snapshot("04_logged_in")
+                    return
+            self._snapshot("04_login_timeout")
+            raise RuntimeError("手动登录超时（120秒），请检查窗口")
+            return  # unreachable
+
+        # === 自动登录 ===
+        print("[->] 需要登录，尝试自动填写...")
+
+        # 0. 再 sync 一次到含登录表单的 page
+        self._ensure_login_page()
+
+        # 1. 切换到「密码登录」标签
+        try:
+            self.page.locator("a.password-login-tab-item").first.click(timeout=3000)
+            print("[OK] 已切换到密码登录")
+        except Exception:
+            pass
+
+        self._snapshot("02_login_tab")
+
+        # 2. 填账号
+        try:
+            self.page.locator("#fm-login-id").first.fill(self.username)
+            time.sleep(0.3)
+        except Exception as e:
+            print(f"[WARN] 填账号失败: {e}，请手动填写")
+
+        # 3. 填密码
+        try:
+            self.page.locator("#fm-login-password").first.fill(self.login_secret)
+            time.sleep(0.3)
+        except Exception as e:
+            print(f"[WARN] 填密码失败: {e}，请手动填写")
+
+        # 4. 勾选协议（关键！不勾的话登录按钮是 button-low-light 灰色不可点）
+        try:
+            cb = self.page.locator("#fm-agreement-checkbox").first
+            if cb.is_visible() and not cb.is_checked():
+                cb.check()
+                time.sleep(0.3)
+                print("[OK] 已勾选协议")
+        except Exception as e:
+            print(f"[WARN] 勾选协议失败: {e}，请手动勾选")
+
+        self._snapshot("03_login_filled")
+
+        # 5. 点击登录按钮
+        clicked_login = False
+        try:
+            btn = self.page.locator("button.fm-submit, #login-form button[type='submit']").first
+            if btn.count() > 0:
+                btn.click()
+                clicked_login = True
+                print("[OK] 已点击登录按钮")
+        except Exception as e:
+            print(f"[WARN] 精准定位登录按钮失败: {e}")
+
+        if not clicked_login:
+            try:
+                self.page.locator("button:has-text('登录')").first.click()
+                clicked_login = True
+                print("[OK] 已用文本匹配点击登录")
+            except Exception as e:
+                print(f"[WARN] 文本点击登录也失败: {e}")
+
+        self._snapshot("03_login_clicked")
+        if not clicked_login:
+            print("[WARN] 未自动点击到登录按钮，请手动点击「登录」按钮")
+
+        # 6. 等待登录完成：后台元素命中即返回，最多 60 秒（带 URL 诊断日志）
+        print("[->] 等待登录完成（自动或手动均可）...")
+        for i in range(60):
+            time.sleep(1)
+            # 淘宝登录可能跳到新 tab，先同步 page
+            self._sync_page()
+            if i % 5 == 0 and i > 0:
+                print(f"    [诊断 i={i}s] URL: {self.page.url[:120]}")
+            if self._is_logged_in():
+                print("[OK] 登录成功")
+                self._snapshot("04_logged_in")
+                self._wait_home_loaded()
+                return
+
+        self._snapshot("04_login_timeout")
+        raise RuntimeError(f"登录超时（60秒），最后 URL: {self.page.url[:120]}")
+
+    def _loc(self, selector: str):
+        """在发布页 iframe 和主页面里找第一个有匹配的 locator。
+        发布页表单嵌在 iframe 里，直接 self.page.locator 永远匹配不到。"""
+        ctxs = []
+        if self.pub_frame is not None:
+            ctxs.append(self.pub_frame)
+        ctxs.append(self.page)
+        for ctx in ctxs:
+            try:
+                loc = ctx.locator(selector)
+                if loc.count() > 0:
+                    return loc
+            except Exception:
+                continue
+        # 都没匹配到时返回主页面的 locator，让调用方拿到标准超时错误
+        return self.page.locator(selector)
+
+    def _wait_publish_video_page(self, timeout: int = 120) -> bool:
+        """等发布视频页真正加载完（出现上传视频按钮/视频封面等标志元素）。
+        标志元素在 iframe 里，命中时顺带记录 self.pub_frame 供后续交互用。
+
+        优先用 SPA URL `/page/pubNew/video` 判定，因为下拉菜单被点中后路由会稳定
+        切到这个地址，比等具体元素出现更不容易被浮层误判。
+        """
+        url_marker = "/page/pubNew/video"
+        markers = ["上传视频", "视频封面", "点击上传视频", "视频描述",
+                   "获取流量攻略", "参与话题活动"]
+        deadline = time.time() + timeout
+        while time.time() < deadline:
+            # 「发视频」可能开新标签页，每次检测前先同步到真正的后台 page
+            self._sync_page()
+            url = (self.page.url or "").lower()
+            if url_marker in url:
+                # 顺手记 pub_frame，供后续 publish_video() 的 _loc 用
+                for ctx in list(self.page.frames):
+                    try:
+                        loc = ctx.locator("text=视频封面 >> visible=true").first
+                        if loc.count() > 0:
+                            self.pub_frame = ctx
+                            print(f"    [定位] 发布表单在 iframe: {ctx.url[:80]}")
+                            break
+                    except Exception:
+                        continue
+                print(f"    [OK] 发布视频页已加载（URL 命中 {url_marker}）")
+                return True
+            for m in markers:
+                for ctx in [self.page] + list(self.page.frames):
+                    try:
+                        loc = ctx.locator(f"text={m} >> visible=true").first
+                        if loc.count() > 0:
+                            if ctx is not self.page:
+                                self.pub_frame = ctx
+                                print(f"    [定位] 发布表单在 iframe: {ctx.url[:80]}")
+                            print(f"    [OK] 发布视频页已加载（命中: {m}）")
+                            return True
+                    except Exception:
+                        continue
+            time.sleep(1)
+        return False
+
+    def _click_publish_video_menu_item(self) -> bool:
+        """点下拉菜单里的「发视频」。必须只点可见项——DOM 里有隐藏的同名元素，
+        用 text=发视频 .first 会命中隐藏节点导致点空（菜单不收起、页面不跳转）。"""
+        try:
+            item = self.page.locator("text=发视频 >> visible=true").first
+            item.wait_for(state="visible", timeout=8_000)
+            item.click(timeout=8_000)
+            print("[OK] 已点「发视频」菜单项")
+            return True
+        except Exception as e:
+            print(f"[WARN] 点击「发视频」菜单项失败: {e}")
+            return False
+
+    def navigate_to_publish_video(self):
+        """导航到发布视频：点 +发布作品 -> 发视频，并确认真的跳到了发布页。"""
+        # 回到创作后台首页
+        self.page.goto("https://creator.guanghe.taobao.com/page/", wait_until="domcontentloaded", timeout=60000)
+        self._sync_page()
+        # 再次确认登录态（淘宝可能因 cookie 过期把页面踢回 login）
+        if not self._is_logged_in():
+            print("[WARN] 导航后被踢回登录页，重新登录...")
+            self.ensure_login()
+            self._sync_page()
+        # 等后台首页 SPA 真正加载完（否则 +发布作品 按钮还没渲染，点不到）
+        self._wait_home_loaded()
+        self._snapshot("05_home")
+
+        # 点「+ 发布作品」展开下拉菜单，再点「发视频」；没跳到发布页就重试一轮
+        # （注意：发布页直链 /page/pubNew/video 必须带完整 pub_url 参数，单独打开会报"URL不合法"，
+        #   所以只能走菜单点击路径；菜单展开有 ~1s 动画，sleep 不够长就会点不到「发视频」）
+        for attempt in range(3):
+            if attempt > 0:
+                print(f"[->] 未进入发布视频页，重试第 {attempt + 1} 次...")
+                # 回到 home 再走一遍菜单点击路径（防止 SPA 路由卡死）
+                self.page.goto("https://creator.guanghe.taobao.com/page/",
+                               wait_until="domcontentloaded", timeout=60000)
+                self._sync_page()
+                self._wait_home_loaded()
+            try:
+                self._click_by_text("发布作品", timeout=10_000)
+                print("[OK] 已点「+发布作品」")
+            except Exception as e:
+                print(f"[WARN] 点击「+发布作品」失败: {e}")
+                continue  # 跳到下一轮重新加载首页
+
+            # 菜单展开动画需要时间，0.8s 太短常点不到「发视频」
+            time.sleep(1.5)
+            self._snapshot("06_publish_menu")
+
+            if not self._click_publish_video_menu_item():
+                # 单独再点一次 +发布作品 让菜单重新展开（常见：菜单被 click 提前收起）
+                print("[->] 「发视频」未点中，再次展开下拉菜单...")
+                try:
+                    self._click_by_text("发布作品", timeout=5_000)
+                    time.sleep(1.5)
+                    if not self._click_publish_video_menu_item():
+                        print("[WARN] 二次点「发视频」仍失败，进入下一轮")
+                        continue
+                except Exception as e:
+                    print(f"[WARN] 二次展开菜单失败: {e}")
+                    continue
+
+            # 发布页是重 SPA 路由，加载慢，最多等 120 秒
+            if self._wait_publish_video_page(timeout=120):
+                self._snapshot("07_publish_video_tab")
+                return
+
+        self._snapshot("07_publish_video_tab")
+        raise RuntimeError("多次尝试后仍未进入发布视频页")
+
+    @staticmethod
+    def _parse_product_names(field_value) -> List[str]:
+        """把飞书「代拍产品」字段拆成单个款式名列表。
+
+        支持的分隔：`,` `，` `、` `/` `\\`（不在名字里加空格分割，
+        避免「宙斯Z air」被错拆成 2 段）。
+
+        字段中含有「未到货 / 先不发 / 不发该视频」等否决关键词时返回 ['__SKIP__']
+        哨兵值，由调用方识别后整条记录跳过。
+        """
+        if field_value is None:
+            return []
+        if isinstance(field_value, list):
+            return [str(x).strip() for x in field_value if str(x).strip()]
+        text = str(field_value)
+        if not text.strip():
+            return []
+        if any(k in text for k in ["未到货", "先不发", "不发该视频"]):
+            return ["__SKIP__"]
+        parts = re.split(r'[,，、/\\]+', text)
+        cleaned = [p.strip() for p in parts if p and p.strip()]
+        # 归一化款式名：
+        # 1) alias 表（精确匹配，像「星云 2」->「星云2」）
+        # 2) 把「中文/字母 + 一个或多个空白 + 数字」合并（星云 2 / 星云  2 / 星云\t2 → 星云2）
+        #    不动「宙斯Z air」这种「字母+空格+字母」命名（因为后面跟的不是数字）
+        _aliases = {"星云 2": "星云2"}
+        _ws_digit = re.compile(r'([一-鿿A-Za-z])\s+(\d)')
+        def _norm(n):
+            n = n.strip()
+            if n in _aliases:
+                return _aliases[n]
+            # lambda 取 group 而不用 \\1 \\2，避免 Python 字符串里的反斜杠被吞
+            prev = None
+            while prev != n:
+                prev = n
+                n = _ws_digit.sub(lambda m: m.group(1) + m.group(2), n)
+            return n
+        cleaned = [_norm(name) for name in cleaned]
+        return cleaned
+
+    _feature_cache: Dict[str, str] = {}  # 类级缓存，同款名不重复搜索
+
+    @classmethod
+    def configure_catalog_product_titles(
+        cls, titles_by_style: Dict[str, List[str]]
+    ) -> Dict[str, str]:
+        """Derive authoritative style bag types from database Tmall titles."""
+        cls.CATALOG_CATEGORY_OVERRIDES = {}
+        cls.CATALOG_PRODUCT_INFO = {}
+        cls._feature_cache.clear()
+        aliases = {
+            "双肩背包": "双肩包",
+            "相机/微单包": "相机包",
+            "托特手提包": "托特包",
+            "吐司腋下包": "腋下包",
+        }
+        for style_name, titles in titles_by_style.items():
+            explicit_style_type = cls._infer_bag_type(style_name)
+            category = aliases.get(explicit_style_type, explicit_style_type)
+            if category == "包袋":
+                for title in titles:
+                    inferred_type = cls._infer_bag_type(title)
+                    inferred = aliases.get(inferred_type, inferred_type)
+                    if inferred != "包袋":
+                        category = inferred
+                        break
+            if category == "包袋":
+                continue
+            cls.CATALOG_CATEGORY_OVERRIDES[style_name] = category
+            cls.CATALOG_PRODUCT_INFO[style_name] = (
+                f"主包型：{category}；尺寸：未查到；容量/适配：未查到；材质：未查到；"
+                f"卖点：数据库天猫商品标题已确认主包型为{category}"
+            )
+        return dict(cls.CATALOG_CATEGORY_OVERRIDES)
+
+    @classmethod
+    def _verified_product_info(cls, product_name: str) -> str:
+        normalized_name = re.sub(r"\s+", "", product_name or "").lower()
+        for model_name, info in cls.CATALOG_PRODUCT_INFO.items():
+            if re.sub(r"\s+", "", model_name).lower() in normalized_name:
+                return info
+        for model_name, info in cls.VERIFIED_PRODUCT_INFO.items():
+            if re.sub(r"\s+", "", model_name).lower() in normalized_name:
+                return info
+        return ""
+
+    @classmethod
+    def _validate_online_product_info(cls, product_name: str, info: str) -> str:
+        normalized = info or ""
+        for laptop_storage_term in ["笔记本电脑仓", "笔记本仓", "笔电仓"]:
+            normalized = normalized.replace(laptop_storage_term, "电脑包收纳")
+        normalized = normalized.replace("笔电", "电脑包").strip()
+        required_fields = ["主包型：", "尺寸：", "容量/适配：", "材质：", "卖点："]
+        if not normalized.startswith("主包型：") or not all(
+            field in normalized for field in required_fields
+        ):
+            return ""
+        main_category = normalized.split("；", 1)[0]
+        expected_category = cls._resolve_title_category(product_name)
+        if expected_category != "包袋" and expected_category not in main_category:
+            return ""
+        return normalized[:240]
+
+    def _search_product_features(self, product_name: str) -> str:
+        """调用 Hermes (MiMO) 联网搜索结构化产品资料。
+
+        使用 search_domain_filter 触发 MiMo 的联网搜索能力，
+        优先搜索电商商品页，并提取主包型、尺寸、适配设备、材质和核心卖点。
+        失败时返回空字符串（调用方会 fallback 到通用卖点）。
+        结果缓存到 _feature_cache，同款名不重复搜索。
+        """
+        if not HAS_HERMES:
+            return ""
+        # 缓存命中
+        if product_name in self._feature_cache:
+            cached = self._feature_cache[product_name]
+            if cached:
+                print(f"[OK] 联网搜索卖点(缓存): {product_name} -> {cached[:50]}")
+            return cached
+        verified = self._verified_product_info(product_name)
+        if verified:
+            self._feature_cache[product_name] = verified
+            print(f"[OK] 使用数据库/已核验商品资料: {product_name} -> {verified[:60]}")
+            return verified
+        try:
+            system = (
+                "你是电商产品资料核验助手。只根据搜索到的同型号商品页提取事实，优先采用京东、"
+                "天猫、淘宝商品标题和参数，多来源一致时才采用媒体实测信息。\n"
+                "必须区分主包型与背负方式：托特包可斜挎或手提，但主包型仍是托特包。"
+                "不得因为标题中出现多个用途词，就把后面的电脑包、斜挎包或双肩背负方式当主品类。\n"
+                "严格输出一行：主包型：...；尺寸：...；容量/适配：...；材质：...；卖点：...。"
+                "没有可靠依据的字段写未查到，严禁猜测或用同品牌其他型号补齐。"
+            )
+            user = f'联网核验「光影行星 {product_name}」的主包型、尺寸、容量或适配设备、材质和核心卖点'
+            payload = {
+                "model": "mimo-v2.5-pro",
+                "messages": [
+                    {"role": "system", "content": system},
+                    {"role": "user", "content": user},
+                ],
+                "temperature": 0.2,
+                "max_tokens": 256,
+                "search_domain_filter": [
+                    "jd.com", "taobao.com", "tmall.com", "zol.com.cn",
+                    "smzdm.com", "sohu.com",
+                ],
+            }
+            body = json.dumps(payload, ensure_ascii=False).encode("utf-8")
+            req = urllib.request.Request(
+                HERMES_ANALYZER_URL, data=body, method="POST",
+                headers={
+                    "Authorization": f"Bearer {HERMES_API_KEY}",
+                    "Content-Type": "application/json",
+                },
+            )
+            with urllib.request.urlopen(req, timeout=60) as resp:
+                raw = resp.read().decode("utf-8", errors="replace")
+            obj = json.loads(raw)
+            features = obj["choices"][0]["message"]["content"].strip()
+            # 清理 MiMo 可能加的"无需联网/家族速查表"等前缀噪音
+            for prefix in ["直接给答案", "根据搜索", "搜索结果", "技能触发", "家族速查表"]:
+                if prefix in features:
+                    for sep in ["：", ":", "\n"]:
+                        idx = features.find(sep)
+                        if idx >= 0:
+                            features = features[idx + 1:].strip()
+                            break
+            # 过滤 MiMo 的对话式回复（质疑款名/反问句）
+            if any(q in features for q in ["没有", "不存在", "你确认", "确认下", "可参", "在动手搜之前"]):
+                features = ""
+            # 结构化摘要至少应有字段分隔符；否则视为对话式噪音。
+            if features and not any(sep in features for sep in ["；", ";", "、", ",", "，"]):
+                # 单行长串没分隔符，可能是对话式回复
+                if len(features) > 50:
+                    features = ""
+            features = self._validate_online_product_info(product_name, features)
+            if not features:
+                features = self._verified_product_info(product_name)
+            self._feature_cache[product_name] = features
+            return features
+        except Exception as e:
+            print(f"[WARN] 联网搜索卖点失败({product_name}): {e}")
+            fallback = self._verified_product_info(product_name)
+            self._feature_cache[product_name] = fallback
+            return fallback
+
+    @staticmethod
+    def _infer_bag_type(product_name: str) -> str:
+        """从商品/款式名里推断包型，用于给标题生成提供卖点上下文。
+
+        数据库 dim_style.category 目前全空，且飞书里的商品名不一定能映射到 dim_style，
+        所以这里直接按商品名关键词推断（款名常自带包型，如「盖亚斜挎」「极星双肩」）。
+        推断不出来时回退到通用「包袋」。
+        """
+        name = product_name or ""
+        # 顺序有讲究：更具体的包型放前面
+        keyword_map = [
+            (("双肩", "背包"), "双肩背包"),
+            (("斜挎", "斜跨"), "斜挎包"),
+            (("微单", "相机"), "相机/微单包"),
+            (("托特",), "托特手提包"),
+            (("电脑",), "电脑包"),
+            (("手提",), "手提包"),
+            (("吐司",), "吐司腋下包"),
+            (("腋下",), "腋下包"),
+            (("单肩",), "单肩包"),
+            (("手拿",), "手拿包"),
+            (("腰包",), "腰包"),
+            (("胸包",), "胸包"),
+            (("水桶",), "水桶包"),
+        ]
+        for keywords, bag_type in keyword_map:
+            if any(k in name for k in keywords):
+                return bag_type
+        return "包袋"
+
+    @classmethod
+    def _resolve_title_category(cls, product_name: str, features: str = "") -> str:
+        """Resolve a searchable, explicit bag category from name then features."""
+        normalized_name = re.sub(r"\s+", "", product_name or "").lower()
+        for model_name, authoritative_category in cls.PRODUCT_CATEGORY_OVERRIDES.items():
+            if re.sub(r"\s+", "", model_name).lower() in normalized_name:
+                return authoritative_category
+
+        category = cls._infer_bag_type(product_name)
+        aliases = {
+            "双肩背包": "双肩包",
+            "相机/微单包": "相机包",
+            "托特手提包": "托特包",
+            "吐司腋下包": "腋下包",
+        }
+        category = aliases.get(category, category)
+        if category != "包袋":
+            return category
+
+        for model_name, authoritative_category in cls.CATALOG_CATEGORY_OVERRIDES.items():
+            if re.sub(r"\s+", "", model_name).lower() in normalized_name:
+                return authoritative_category
+
+        feature_categories = [
+            (("双肩包", "双肩背包", "背包"), "双肩包"),
+            (("斜挎包", "斜跨包"), "斜挎包"),
+            (("单肩包",), "单肩包"),
+            (("相机包", "微单包", "摄影包"), "相机包"),
+            (("笔记本电脑包", "电脑包"), "电脑包"),
+            (("托特包",), "托特包"),
+            (("手提包",), "手提包"),
+            (("腋下包",), "腋下包"),
+            (("胸包",), "胸包"),
+            (("腰包",), "腰包"),
+        ]
+        text = features or ""
+        for keywords, resolved in feature_categories:
+            if any(keyword in text for keyword in keywords):
+                return resolved
+        return "包袋"
+
+    @classmethod
+    def _title_prefix(cls, product_names: List[str], category: str) -> str:
+        name = product_names[0] if product_names else "精选款"
+        if category.endswith("包"):
+            stem = category[:-1]
+            if name.endswith(stem):
+                product_part = name + "包"
+            elif category in name:
+                product_part = name
+            else:
+                product_part = name + category
+        elif category in name:
+            product_part = name
+        else:
+            product_part = name + category
+        if len(product_names) > 1:
+            product_part += "等多款"
+        return product_part
+
+    @classmethod
+    def _title_body(cls, title: str) -> str:
+        """Return the reusable copy body while ignoring model/category prefixes."""
+        cleaned = re.sub(r"[！!。?？；;：:\s]+", "", title or "")
+        cleaned = cleaned.replace("，", ",").strip(",")
+        first, separator, remainder = cleaned.partition(",")
+        looks_like_product_prefix = (
+            first.endswith(("包", "包袋", "等多款"))
+            or any(
+                category in first
+                for category in (
+                    "双肩包", "斜挎包", "单肩包", "电脑包", "相机包",
+                    "托特包", "手提包", "腋下包", "胸包", "腰包", "包袋",
+                )
+            )
+        )
+        return remainder if separator and looks_like_product_prefix else cleaned
+
+    @staticmethod
+    def _title_ngrams(text: str, size: int) -> set[str]:
+        compact = re.sub(r"[,，！!。?？；;：:\s]+", "", text or "")
+        if not compact:
+            return set()
+        if len(compact) <= size:
+            return {compact}
+        return {compact[index:index + size] for index in range(len(compact) - size + 1)}
+
+    @classmethod
+    def _title_similarity(cls, first: str, second: str) -> float:
+        """Compare copy skeletons, ignoring product/model prefixes."""
+        first_body = cls._title_body(first)
+        second_body = cls._title_body(second)
+        if first_body == second_body:
+            return 1.0
+        scores: List[float] = []
+        for size in (2, 3):
+            left = cls._title_ngrams(first_body, size)
+            right = cls._title_ngrams(second_body, size)
+            union = left | right
+            scores.append(len(left & right) / len(union) if union else 0.0)
+        return max(scores, default=0.0)
+
+    @classmethod
+    def _normalize_generated_title(
+        cls,
+        title: str,
+        product_names: List[str],
+        category: str,
+    ) -> str:
+        """Enforce product/category prefix, readable clauses and searchable vocabulary."""
+        normalized = (title or "").replace("，", ",").replace("！", "!")
+        normalized = re.sub(r"[\s。.?？、；;：:]+", "", normalized)
+        normalized = re.sub(r",+", ",", normalized)
+        for laptop_storage_term in ["笔记本电脑仓", "笔记本仓", "笔电仓"]:
+            normalized = normalized.replace(laptop_storage_term, "电脑包收纳")
+        for laptop_term in ["笔记本电脑包", "笔记本电脑", "笔电包", "笔电"]:
+            normalized = normalized.replace(laptop_term, "电脑包")
+        normalized = normalized.replace("电脑包包", "电脑包")
+        phrase_replacements = {
+            "被同事追着要链接": "高颜值百搭",
+            "被问链接": "高颜值百搭",
+            "终于解放了": "轻便实用",
+            "终于解放": "轻便实用",
+            "真能装": "大容量",
+            "一只搞定": "多场景实用",
+            "通勤的秘密": "通勤轻便",
+        }
+        for source, target in phrase_replacements.items():
+            normalized = normalized.replace(source, target)
+
+        prefix = cls._title_prefix(product_names, category)
+        body = normalized
+        for token in ["光影行星", *(product_names[:1] or []), category]:
+            if token:
+                body = body.replace(token, "", 1)
+                compact_token = re.sub(r"\s+", "", token)
+                compact_body = re.sub(r"\s+", "", body)
+                if compact_token and compact_token.casefold() in compact_body.casefold():
+                    body = re.sub(
+                        re.escape(compact_token),
+                        "",
+                        body,
+                        count=1,
+                        flags=re.IGNORECASE,
+                    )
+        for other_category in [
+            "双肩背包", "双肩包", "斜挎包", "单肩包", "相机包",
+            "托特包", "手提包", "腋下包", "胸包", "腰包", "包袋",
+        ]:
+            if other_category != category:
+                body = body.replace(other_category, "")
+        if prefix.endswith("包"):
+            body = body.lstrip("包,!")
+        body = body.strip(",!")
+        body_parts = [part for part in body.split(",") if part]
+        if body_parts and (
+            len(body_parts[-1]) <= 3
+            or body_parts[-1].endswith(("从容应", "从容有", "通勤出行物品"))
+        ):
+            body_parts[-1] = "通勤出行更从容"
+            body = ",".join(body_parts)
+        if len(body) < 6:
+            body += "科学分区收纳,轻装通勤出行"
+        if "," not in body:
+            for result_phrase in [
+                "从容出行", "轻松出行", "轻装通勤", "收纳不乱",
+                "拿捏通勤出行", "通勤出行",
+            ]:
+                split_at = body.rfind(result_phrase)
+                if split_at > 3:
+                    body = body[:split_at] + "," + body[split_at:]
+                    break
+        if "," not in body and len(body) >= 12:
+            split_at = max(6, len(body) * 2 // 3)
+            body = body[:split_at] + "," + body[split_at:]
+        result = prefix + "," + body
+        padding_phrases = (
+            "拿取顺手", "出门不慌", "行程利落", "搭配自然",
+            "准备高效", "转场轻松", "随行省心", "使用从容",
+        )
+        padding_start = sum(ord(char) for char in body) % len(padding_phrases)
+        for offset in range(len(padding_phrases)):
+            if len(result) >= 25:
+                break
+            phrase = padding_phrases[(padding_start + offset) % len(padding_phrases)]
+            if (
+                phrase not in result
+                and phrase[:2] not in result
+                and len(result) + len(phrase) + 1 <= 29
+            ):
+                result += "," + phrase
+        if len(result) > 29:
+            compact_options = [
+                (("防泼", "雨天"), "雨天用品稳妥随行,通勤少些顾虑"),
+                (("电脑包",), "电脑包用品分开放好,办公拿取顺手"),
+                (("大容量", "容量"), "出门所需集中装下,临时安排不慌"),
+                (("轻量", "轻盈"), "肩背节奏保持轻松,赶车转场不拖沓"),
+                (("隔层", "分区", "分类"), "常用小物分别放好,拿取不用翻找"),
+            ]
+            compact_body = next(
+                (
+                    replacement
+                    for markers, replacement in compact_options
+                    if any(marker in body for marker in markers)
+                ),
+                "",
+            )
+            if not compact_body:
+                compact_body = cls.TITLE_VARIANTS[
+                    sum(ord(char) for char in body) % len(cls.TITLE_VARIANTS)
+                ][2]
+            result = f"{prefix},{compact_body}"
+            if len(result) > 29:
+                complete_short_bodies = (
+                    "随身物件安排到位,途中少找东西",
+                    "常用小物分别放好,拿取更顺手",
+                    "出门用品集中带齐,行程不忙乱",
+                    "日常所需稳妥带走,转场更利落",
+                )
+                fitting = [
+                    option
+                    for option in complete_short_bodies
+                    if len(prefix) + len(option) + 1 <= 29
+                ]
+                if fitting:
+                    compact_body = max(fitting, key=len)
+                    result = f"{prefix},{compact_body}"
+        for phrase in padding_phrases:
+            if len(result) >= 25:
+                break
+            if (
+                phrase not in result
+                and phrase[:2] not in result
+                and len(result) + len(phrase) + 1 <= 29
+            ):
+                result += "," + phrase
+        return result.rstrip(",!") + "!"
+
+    @classmethod
+    def _fallback_title(
+        cls,
+        product_names: List[str],
+        category: str,
+        variant_index: int = 1,
+    ) -> str:
+        zero_based = (max(1, variant_index) - 1) % cls.TITLE_VARIATION_COUNT
+        lead_index = zero_based % len(cls.TITLE_VARIANTS)
+        cycle = zero_based // len(cls.TITLE_VARIANTS)
+        tail_index = (lead_index + cycle * 5) % len(cls.TITLE_VARIANTS)
+        lead_variant = cls.TITLE_VARIANTS[lead_index]
+        tail_variant = cls.TITLE_VARIANTS[tail_index]
+        full_lead = lead_variant[0].split(",", 1)[0]
+        full_tail = tail_variant[0].split(",", 1)[1]
+        compact_lead = lead_variant[2].split(",", 1)[0]
+        compact_tail = tail_variant[2].split(",", 1)[1]
+        wording_round = cycle % 3
+        if wording_round:
+            full_lead = cls.TITLE_LEAD_REFRAMES[lead_index][wording_round - 1]
+            full_tail = cls.TITLE_TAIL_REFRAMES[tail_index][wording_round - 1]
+            compact_lead = full_lead
+            compact_tail = full_tail
+        prefix = cls._title_prefix(product_names, category)
+        body = f"{full_lead},{full_tail}"
+        if len(prefix) + len(body) + 2 > 30:
+            body = f"{compact_lead},{compact_tail}"
+        return cls._normalize_generated_title(
+            body,
+            product_names,
+            category,
+        )
+
+    @classmethod
+    def _variant_angle(cls, variant_index: int) -> str:
+        zero_based = (max(1, variant_index) - 1) % cls.TITLE_VARIATION_COUNT
+        lead_index = zero_based % len(cls.TITLE_VARIANTS)
+        cycle = zero_based // len(cls.TITLE_VARIANTS)
+        tail_index = (lead_index + cycle * 5) % len(cls.TITLE_VARIANTS)
+        lead_angle = cls.TITLE_VARIANTS[lead_index][1]
+        tail_angle = cls.TITLE_VARIANTS[tail_index][1]
+        return lead_angle if lead_index == tail_index else f"{lead_angle}+{tail_angle}"
+
+    def _claim_unique_title(
+        self,
+        candidate: str,
+        product_names: List[str],
+        category: str,
+        variant_index: int,
+    ) -> str:
+        """Reject exact and structurally similar copy across the whole upload run."""
+        title_map = getattr(self, "_used_titles_by_products", None)
+        if title_map is None:
+            title_map = {}
+            self._used_titles_by_products = title_map
+        global_used = getattr(self, "_used_titles_global", None)
+        if global_used is None:
+            global_used = []
+            self._used_titles_global = global_used
+        variant_usage = getattr(self, "_title_variant_usage", None)
+        if variant_usage is None:
+            variant_usage = {}
+            self._title_variant_usage = variant_usage
+        key = tuple(_style_db_lookup_key(name) for name in product_names)
+        used = title_map.setdefault(key, set())
+
+        def max_similarity(title: str) -> float:
+            return max(
+                (self._title_similarity(title, previous) for previous in global_used),
+                default=0.0,
+            )
+
+        candidate_similarity = max_similarity(candidate)
+        if candidate not in used and candidate_similarity < self.TITLE_SIMILARITY_LIMIT:
+            used.add(candidate)
+            global_used.append(candidate)
+            return candidate
+
+        replacements = []
+        for offset in range(self.TITLE_VARIATION_COUNT):
+            candidate_variant = (
+                (max(1, variant_index) - 1 + offset) % self.TITLE_VARIATION_COUNT
+            ) + 1
+            replacement = self._fallback_title(
+                product_names,
+                category,
+                variant_index=candidate_variant,
+            )
+            if replacement in used:
+                continue
+            similarity = max_similarity(replacement)
+            usage = variant_usage.get(candidate_variant, 0)
+            replacements.append((similarity, usage, offset, candidate_variant, replacement))
+
+        sufficiently_different = [
+            item for item in replacements if item[0] < self.TITLE_SIMILARITY_LIMIT
+        ]
+        pool = sufficiently_different or replacements
+        if not pool:
+            raise RuntimeError("标题角度池已耗尽，无法生成不重复标题")
+        similarity, _usage, _offset, chosen_variant, replacement = min(
+            pool,
+            key=lambda item: (item[1], item[0], item[2]),
+        )
+        reason = "同款标题重复" if candidate in used else f"正文相似度{candidate_similarity:.0%}过高"
+        angle = self._variant_angle(chosen_variant)
+        print(f"[INFO] {reason}，已改用“{angle}”角度")
+        used.add(replacement)
+        global_used.append(replacement)
+        variant_usage[chosen_variant] = variant_usage.get(chosen_variant, 0) + 1
+        return replacement
+
+    def _generate_title(
+        self,
+        product_names: List[str],
+        video_index: int = 1,
+        video_total: int = 1,
+    ) -> str:
+        """Generate a 26-30 char searchable title with a mandatory structure."""
+        video_index = max(1, int(video_index or 1))
+        video_total = max(video_index, int(video_total or 1))
+        primary_name = product_names[0] if product_names else "精选款"
+        product_features = self._search_product_features(primary_name) if product_names else ""
+        category = self._resolve_title_category(primary_name, product_features)
+        fallback = self._fallback_title(product_names, category, variant_index=video_index)
+        variant_angle = self._variant_angle(video_index)
+        recent_titles = getattr(self, "_used_titles_global", [])[-8:]
+        recent_bodies = [self._title_body(title) for title in recent_titles]
+        print(f"[->] 标题品类: {category}（多款式={len(product_names) > 1}）")
+        if product_features:
+            print(f"[OK] 联网搜索卖点: {product_features[:60]}")
+        else:
+            print("[INFO] 联网搜索卖点未获取，使用通用卖点")
+
+        if not HAS_HERMES:
+            print("[WARN] Hermes 不可用，使用结构化回退标题")
+            self._last_title_category = category
+            return self._claim_unique_title(
+                fallback, product_names, category, video_index
+            )
+
+        try:
+            system = (
+                "你是淘宝搜索型种草标题编辑，为光影行星（GYXX）包袋写短视频标题。\n"
+                "标题严格采用：产品昵称或型号+指定品类,核心痛点解决方案,场景或结果反馈!\n"
+                "产品和指定品类必须前置，让用户一眼看到重点；指定品类已经由商品档案确认，"
+                "必须原样使用，禁止改成其他包型。\n"
+                "场景要具体，如暴雨通勤、早高峰通勤、周末短途；解决方案要写防泼水、电脑包收纳、"
+                "多隔层、大容量、轻量等真实功能；结果反馈要简短，如从容出行、轻装通勤、收纳不乱。\n"
+                "禁止第一人称、反问、故事口吻及“被问链接、终于解放、秘密、一只搞定、真能装”"
+                "等口语钩子；不要文艺比喻。\n"
+                "涉及电脑收纳时必须写“电脑包”，禁止写“笔电”或“笔记本电脑”。"
+                "不用#话题、emoji、极限词。\n"
+                "必须用英文逗号分隔短句，并用英文感叹号结尾；只使用逗号和感叹号。\n"
+                "总长必须为26至30个汉字或字符，只输出标题本身。\n"
+                "示例：星迹2斜挎包,门卡耳机随手拿取,早高峰进站不耽搁!\n"
+                "示例：波塞冬edge托特包,文件水杯集中带齐,临时转场不忙乱!"
+            )
+            user = (
+                f"产品名：{primary_name}{'等多款' if len(product_names) > 1 else ''}\n"
+                f"品类词：{category}\n"
+                f"联网商品资料：{product_features or '未获得已验证功能；只写通用携带场景，不补充材质、尺寸或特殊功能'}\n"
+                f"这是同款第{video_index}/{video_total}条视频，本条创作角度：{variant_angle}。"
+                "必须围绕本条角度写具体使用情境，不要套用分区收纳、通勤出行这类固定骨架。\n"
+                f"本次任务已经使用的正文：{'；'.join(recent_bodies) if recent_bodies else '无'}。"
+                "不得复用这些正文的句式、连续四字短语或结果表达。\n"
+                "按固定结构写一个重点清晰的搜索型种草标题："
+            )
+            title, _usage = _call_hermes(system, user, timeout=90)
+            title = title.strip().strip('"').strip("'").strip("「」").strip("『』").strip("《》")
+            title = title.split("\n")[0].strip()
+            if not title:
+                raise RuntimeError("Hermes 返回空")
+            if category == "包袋":
+                category = self._resolve_title_category(primary_name, title)
+            normalized = self._normalize_generated_title(title, product_names, category)
+            self._last_title_category = category
+            normalized = self._claim_unique_title(
+                normalized, product_names, category, video_index
+            )
+            print(f"[OK] Hermes 生成并规范标题: {normalized}")
+            return normalized
+        except Exception as e:
+            print(f"[WARN] Hermes 生成标题失败: {e}，使用结构化回退标题")
+            self._last_title_category = category
+            return self._claim_unique_title(
+                fallback, product_names, category, video_index
+            )
+
+    @classmethod
+    def _content_label_for_products(cls, product_names: List[str]) -> str:
+        """Use the first product's primary bag type as the content label."""
+        if not product_names:
+            return "包袋"
+        authoritative = cls._resolve_title_category(product_names[0])
+        if authoritative != "包袋":
+            return authoritative
+        bag_type = cls._infer_bag_type(product_names[0])
+        # 光合标签使用简短包型名，不使用标题生成用的复合说明。
+        aliases = {
+            "相机/微单包": "相机包",
+            "托特手提包": "托特包",
+            "吐司腋下包": "腋下包",
+        }
+        return aliases.get(bag_type, bag_type)
+
+    def _active_overlay(self, ctx, marker: str):
+        """Return the visible dialog/dropdown containing ``marker`` when possible."""
+        candidates = [
+            "[role='dialog']",
+            ".next-dialog-inner",
+            "[class*='dialog-body']",
+            "[class*='popup']",
+            "[class*='overlay']",
+        ]
+        for selector in candidates:
+            try:
+                loc = ctx.locator(selector).filter(has_text=marker)
+                for idx in range(loc.count() - 1, -1, -1):
+                    item = loc.nth(idx)
+                    if item.is_visible():
+                        return item
+            except Exception:
+                continue
+        return ctx
+
+    @staticmethod
+    def _first_visible(locator):
+        for idx in range(locator.count()):
+            item = locator.nth(idx)
+            try:
+                if item.is_visible():
+                    return item
+            except Exception:
+                continue
+        return None
+
+    @staticmethod
+    def _content_label_text_matches(text: str, label: str) -> bool:
+        """Match one exact tag row while allowing a trailing browse-count column."""
+        compact = re.sub(r"\s+", "", text or "")
+        target = re.escape((label or "").strip())
+        return bool(
+            re.fullmatch(rf"#?{target}(?:\d+(?:\.\d+)?(?:万|亿)?次浏览)?", compact)
+        )
+
+    def _find_content_label_option(self, ctx, label: str):
+        """Find an exact selectable tag card, never an already-applied editor token."""
+        pattern = re.compile(rf"#\s*{re.escape(label)}")
+        candidates = []
+        for selector in [
+            ".hashTag-card",
+            ".hashTag-recommend-item",
+        ]:
+            try:
+                locator = ctx.locator(selector)
+                for idx in range(locator.count()):
+                    item = locator.nth(idx)
+                    if not item.is_visible():
+                        continue
+                    text = item.inner_text(timeout=1_000).strip()
+                    if self._content_label_text_matches(text, label):
+                        candidates.append((len(re.sub(r"\s+", "", text)), item))
+            except Exception:
+                continue
+        try:
+            locator = ctx.get_by_text(pattern)
+            for idx in range(locator.count()):
+                item = locator.nth(idx)
+                try:
+                    if not item.is_visible():
+                        continue
+                    if item.evaluate(
+                        "(el) => Boolean(el.closest('.richText-plugin-theme'))"
+                    ):
+                        continue
+                    text = item.inner_text(timeout=1_000).strip()
+                    if self._content_label_text_matches(text, label):
+                        candidates.append((len(re.sub(r"\s+", "", text)), item))
+                except Exception:
+                    continue
+        except Exception:
+            pass
+        if not candidates:
+            return None
+        candidates.sort(key=lambda pair: pair[0])
+        return candidates[0][1]
+
+    def _content_label_is_applied(self, ctx, label: str) -> bool:
+        """Return True only when the tag persists as a token inside the rich-text editor."""
+        for selector in [
+            ".richText-plugin-theme",
+            ".richText-container [data-cangjie-void='true']",
+        ]:
+            try:
+                locator = ctx.locator(selector)
+                for idx in range(locator.count()):
+                    item = locator.nth(idx)
+                    if not item.is_visible():
+                        continue
+                    if self._content_label_text_matches(
+                        item.inner_text(timeout=1_000), label
+                    ):
+                        return True
+            except Exception:
+                continue
+        return False
+
+    def _pick_content_label(self, label: str) -> None:
+        """Select one exact content tag from recommendations or the tag drawer."""
+        ctx = self.pub_frame if self.pub_frame is not None else self.page
+        if self._content_label_is_applied(ctx, label):
+            print(f"[OK] 内容标签已存在: #{label}")
+            return
+        option = self._find_content_label_option(ctx, label)
+        if option is None:
+            self._click_by_text("内容标签", timeout=10_000)
+            print(f"[OK] 已打开内容标签: {label}")
+            time.sleep(0.8)
+            option = self._find_content_label_option(ctx, label)
+        search = None
+        for selector in [
+            "input[placeholder*='输入文本检索']",
+            "input[placeholder*='检索更多']",
+            "input[placeholder*='搜索']",
+        ]:
+            try:
+                search = self._first_visible(ctx.locator(selector))
+                if search is not None:
+                    break
+            except Exception:
+                continue
+        if option is None and search is not None:
+            search.fill(label)
+            for _ in range(12):
+                time.sleep(0.25)
+                option = self._find_content_label_option(ctx, label)
+                if option is not None:
+                    break
+        if option is None and search is None:
+            # 光合当前版本没有独立搜索框：点击“内容标签”后会在正文光标处插入“#”，
+            # 继续键入标签名会直接生成正文标签 token；此时不能再次点击 token/候选卡，
+            # 否则平台会把刚生成的标签删除。
+            self.page.keyboard.type(label)
+            for _ in range(12):
+                time.sleep(0.25)
+                if self._content_label_is_applied(ctx, label):
+                    break
+            if self._content_label_is_applied(ctx, label):
+                try:
+                    self.page.keyboard.press("Enter")
+                except Exception:
+                    pass
+                time.sleep(0.3)
+                if not self._content_label_is_applied(ctx, label):
+                    raise RuntimeError(f"内容标签确认输入后丢失: #{label}")
+                print(f"[OK] 已输入并确认内容标签: #{label}")
+                return
+            option = self._find_content_label_option(ctx, label)
+        if option is None:
+            raise RuntimeError(f"内容标签没有精确匹配项: #{label}")
+        option.click(timeout=5_000)
+        for _ in range(12):
+            time.sleep(0.25)
+            if self._content_label_is_applied(ctx, label):
+                break
+        else:
+            raise RuntimeError(f"内容标签点击后未写入描述区: #{label}")
+        try:
+            self.page.keyboard.press("Escape")
+        except Exception:
+            pass
+        if not self._content_label_is_applied(ctx, label):
+            raise RuntimeError(f"内容标签关闭候选层后丢失: #{label}")
+        print(f"[OK] 已添加内容标签: #{label}")
+        time.sleep(0.5)
+
+    def _pick_content_labels(self, labels) -> None:
+        for label in labels:
+            self._pick_content_label(label)
+
+    @staticmethod
+    def _topic_result_card_text_matches(text: str, topic: str) -> bool:
+        compact = re.sub(r"\s+", "", text or "")
+        return topic in compact and "作品数" in compact and "参与数" in compact
+
+    @staticmethod
+    def _topic_card_box_is_clickable(width: float, height: float) -> bool:
+        return width >= 400 and 70 <= height <= 220
+
+    def _find_topic_result_cards(self, dialog, topic: str):
+        candidates = []
+        try:
+            locator = dialog.locator("div").filter(has_text=topic)
+            for idx in range(locator.count()):
+                item = locator.nth(idx)
+                try:
+                    if not item.is_visible():
+                        continue
+                    text = item.inner_text(timeout=1_000).strip()
+                    if self._topic_result_card_text_matches(text, topic):
+                        box = item.bounding_box()
+                        if box and self._topic_card_box_is_clickable(
+                            box.get("width", 0), box.get("height", 0)
+                        ):
+                            candidates.append((box["width"] * box["height"], item))
+                except Exception:
+                    continue
+        except Exception:
+            pass
+        candidates.sort(key=lambda pair: pair[0])
+        return [item for _area, item in candidates]
+
+    def _pick_topic(self, topic: str | None = None):
+        """Search and select the exact fixed topic instead of a recommended item."""
+        topic = topic or self.FIXED_TOPIC
+        self._click_by_text("点击添加话题", timeout=10_000)
+        print("[OK] 已点开话题弹窗")
+        time.sleep(1.5)
+
+        ctx = self.pub_frame if self.pub_frame is not None else self.page
+        dialog = self._active_overlay(ctx, "话题选择")
+        search = None
+        for selector in ["input[placeholder*='搜索']", "input"]:
+            search = self._first_visible(dialog.locator(selector))
+            if search is not None:
+                break
+        if search is None:
+            raise RuntimeError("话题弹窗没有搜索输入框")
+        search.fill(topic)
+
+        search_button = self._first_visible(dialog.locator("button:has-text('搜索')"))
+        if search_button is None:
+            search_button = self._first_visible(dialog.get_by_text("搜索", exact=True))
+        if search_button is None:
+            raise RuntimeError("话题弹窗没有搜索按钮")
+        search_button.click(timeout=5_000)
+        print(f"[OK] 已搜索固定话题: {topic}")
+
+        cards = []
+        for _ in range(20):
+            time.sleep(0.5)
+            cards = self._find_topic_result_cards(dialog, topic)
+            if cards:
+                break
+        if not cards:
+            raise RuntimeError(f"没有找到固定话题: #{topic}")
+
+        confirm = self._first_visible(dialog.locator("button:has-text('确认提交')"))
+        if confirm is None:
+            confirm = self._first_visible(dialog.get_by_text("确认提交", exact=True))
+        if confirm is None:
+            raise RuntimeError("话题弹窗没有确认提交按钮")
+        for attempt, card in enumerate(cards, 1):
+            card.click(timeout=5_000)
+            print(f"[OK] 已点击固定话题卡片外层({attempt}/{len(cards)}): #{topic}")
+            time.sleep(0.4)
+            confirm.click(timeout=5_000)
+            try:
+                dialog.wait_for(state="hidden", timeout=2_500)
+                print(f"[OK] 已确认固定话题: #{topic}")
+                return
+            except Exception:
+                continue
+        raise RuntimeError(f"点击结果卡和确认提交后话题弹窗仍未关闭: #{topic}")
+
+    def _prepare_publish_metadata(self, product_names: List[str]) -> None:
+        primary_label = self._content_label_for_products(product_names)
+        labels = [primary_label]
+        labels.extend(
+            label for label in self.FIXED_CONTENT_LABELS if label not in labels
+        )
+        self._pick_content_labels(labels)
+        self._snapshot("09b_content_label_selected")
+        self._pick_topic(self.FIXED_TOPIC)
+        self._snapshot("10_topic_selected")
+
+    @staticmethod
+    def _normalize_product_search_term(product_name: str) -> str:
+        """Convert a business style label into the shorter storefront search term."""
+        original = (product_name or "").strip()
+        compact = re.sub(r"\s+", "", original)
+        without_series = re.sub(r"系列.*$", "", compact).strip("-—_·:：")
+        return without_series or compact
+
+    @classmethod
+    def _build_product_search_specs(
+        cls,
+        product_names: List[str],
+        tm_item_ids_by_style: Optional[Dict[str, List[str]]] = None,
+    ) -> List[Dict[str, Any]]:
+        """Build exact-ID searches, falling back to one style keyword when IDs are absent."""
+        ids_by_key = {
+            _style_db_lookup_key(style_name): list(item_ids or [])
+            for style_name, item_ids in (tm_item_ids_by_style or {}).items()
+        }
+        primary_specs: List[Dict[str, Any]] = []
+        extra_specs: List[Dict[str, Any]] = []
+        seen_ids: set[str] = set()
+        for product_name in product_names:
+            item_ids = ids_by_key.get(_style_db_lookup_key(product_name), [])
+            clean_ids: List[str] = []
+            for raw_id in item_ids:
+                item_id = re.sub(r"\D", "", str(raw_id or ""))
+                if item_id and item_id not in seen_ids:
+                    clean_ids.append(item_id)
+                    seen_ids.add(item_id)
+            if clean_ids:
+                primary_specs.append(
+                    {
+                        "product_name": product_name,
+                        "search_term": clean_ids[0],
+                        "exact_id": True,
+                    }
+                )
+                extra_specs.extend(
+                    {
+                        "product_name": product_name,
+                        "search_term": item_id,
+                        "exact_id": True,
+                    }
+                    for item_id in clean_ids[1:]
+                )
+                continue
+            primary_specs.append(
+                {
+                    "product_name": product_name,
+                    "search_term": cls._normalize_product_search_term(product_name),
+                    "exact_id": False,
+                }
+            )
+        return primary_specs + extra_specs
+
+    @classmethod
+    def _build_exact_search_fallback_specs(
+        cls, search_specs: List[Dict[str, Any]]
+    ) -> List[Dict[str, Any]]:
+        """Add one keyword fallback for styles whose exact IDs may belong to another shop."""
+        fallback_specs: List[Dict[str, Any]] = []
+        seen_styles: set[str] = set()
+        for spec in search_specs:
+            if not spec.get("exact_id"):
+                continue
+            product_name = str(spec.get("product_name") or "").strip()
+            style_key = _style_db_lookup_key(product_name)
+            if not product_name or style_key in seen_styles:
+                continue
+            seen_styles.add(style_key)
+            fallback_specs.append(
+                {
+                    "product_name": product_name,
+                    "search_term": cls._normalize_product_search_term(product_name),
+                    "exact_id": False,
+                }
+            )
+        return fallback_specs
+
+    def _link_product(
+        self,
+        product_names: List[str],
+        tm_item_ids_by_style: Optional[Dict[str, List[str]]] = None,
+    ) -> bool:
+        """点「添加商品」后优先按数据库天猫商品 ID 精确搜索并勾选。
+
+        某款式没有数据库 ID，或数据库 ID 不属于当前登录店铺时，退回款式关键词搜索；
+        每个搜索项只勾选第一个结果，累计最多 6 个商品。
+
+        单款式时等同原行为。
+
+        Returns:
+            True  - 至少勾到 1 个商品并确认
+            False - 全部款式都无匹配 / 勾选失败 / 确定按钮失败（弹窗会尽力关掉）
+        Raises:
+            RuntimeError - 弹窗没打开这种前置失败
+        """
+        if not product_names:
+            return False
+        search_specs = self._build_product_search_specs(
+            product_names, tm_item_ids_by_style
+        )
+        search_specs.extend(self._build_exact_search_fallback_specs(search_specs))
+        is_multi = len(product_names) > 1
+        MAX_LINK = 6
+        exact_count = sum(1 for spec in search_specs if spec["exact_id"])
+        print(
+            f"[->] 关联商品：共 {len(product_names)} 个款式"
+            f"（{'多款式' if is_multi else '单款式'}），"
+            f"数据库精确 ID {exact_count} 个"
+        )
+
+        # 一次性打开弹窗，复用同一搜索框
+        self._click_by_text("添加商品", timeout=10_000)
+        print("[OK] 已点开关联商品弹窗")
+        time.sleep(1.5)
+
+        # 搜索框（占位符含「搜索」）
+        search = None
+        for sel in [
+            "input[placeholder*='搜索']",
+            "input[type='search']",
+            "input[placeholder*='商品']",
+        ]:
+            try:
+                loc = self._loc(sel).first
+                if loc.count() > 0:
+                    search = loc
+                    break
+            except Exception:
+                continue
+        if search is None:
+            print("[ERR] 关联商品弹窗内找不到搜索框 → 跳过本次发布")
+            for label in ["取消", "关 闭", "关闭"]:
+                try:
+                    self._click_by_text(label, timeout=2000)
+                    break
+                except Exception:
+                    continue
+            else:
+                try:
+                    self.page.keyboard.press("Escape")
+                except Exception:
+                    pass
+            time.sleep(0.5)
+            return False
+
+        ctxs = ([self.pub_frame] if self.pub_frame is not None else []) + [self.page]
+        total_selected = 0
+        linked_styles: set[str] = set()
+
+        for p_idx, spec in enumerate(search_specs):
+            if total_selected >= MAX_LINK:
+                print(
+                    f"[->] 已达关联上限 {MAX_LINK}/{MAX_LINK}，跳过剩余"
+                    f" {len(search_specs) - p_idx} 个搜索项"
+                )
+                break
+            product_name = str(spec["product_name"])
+            search_term = str(spec["search_term"])
+            exact_id = bool(spec["exact_id"])
+            if not exact_id and product_name in linked_styles:
+                print(f"[OK] 「{product_name}」精确 ID 已匹配，跳过关键词兜底")
+                continue
+            if exact_id:
+                print(
+                    f"\n[->] 精确搜索 [{p_idx+1}/{len(search_specs)}]: "
+                    f"「{product_name}」天猫商品 ID={search_term}"
+                )
+            elif search_term != re.sub(r"\s+", "", product_name):
+                print(
+                    f"\n[->] 关键词兜底 [{p_idx+1}/{len(search_specs)}]: "
+                    f"「{product_name}」→「{search_term}」"
+                )
+            else:
+                print(
+                    f"\n[->] 关键词兜底 [{p_idx+1}/{len(search_specs)}]: "
+                    f"「{search_term}」"
+                )
+
+            # 清空旧搜索词（多款式可能需要切换关键词）；优先用 fill('')
+            try:
+                search.fill("")
+            except Exception:
+                try:
+                    search.clear()
+                except Exception:
+                    pass
+            time.sleep(0.3)
+            try:
+                search.fill(search_term)
+            except Exception as e:
+                print(f"[WARN] 填入搜索词失败: {e}，跳过此款式")
+                continue
+            try:
+                self._loc("button:has-text('搜索')").first.click(timeout=2000)
+            except Exception:
+                try:
+                    search.press("Enter")
+                except Exception:
+                    pass
+            time.sleep(2.5)  # 等 AJAX 渲染
+
+            # 检测当前可勾选的 checkbox
+            checkbox_count = 0
+            for ctx in ctxs:
+                try:
+                    loc = ctx.locator(
+                        ".next-dialog input[type='checkbox']:not([disabled])"
+                    )
+                    n = loc.count()
+                    if n > 0:
+                        checkbox_count = n
+                        break
+                except Exception:
+                    continue
+            if checkbox_count == 0:
+                for ctx in ctxs:
+                    try:
+                        loc = ctx.locator("input[type='checkbox']:not([disabled])")
+                        n = loc.count()
+                        if n > 0:
+                            checkbox_count = n
+                            break
+                    except Exception:
+                        continue
+
+            # 检测「无匹配」提示文字
+            has_empty_text = False
+            for marker in ["未找到", "无搜索结果", "没有匹配", "没有找到",
+                           "暂无", "未搜到", "搜索为空", "暂无数据",
+                           "共 0 条", "共0条"]:
+                for ctx in ctxs:
+                    try:
+                        if ctx.locator(f"text={marker}").count() > 0:
+                            has_empty_text = True
+                            break
+                    except Exception:
+                        continue
+                if has_empty_text:
+                    break
+
+            if checkbox_count == 0 or has_empty_text:
+                print(
+                    f"[WARN] 「{product_name}」无匹配（checkbox={checkbox_count}），"
+                    "本款式不挂"
+                )
+                continue
+
+            # 精确 ID 和关键词兜底都只取首个结果，避免关键词误挂多个相似商品。
+            target_n = 1
+
+            # 收集前 N 个未选 checkbox
+            all_cbs = []
+            picked_ctx = None
+            for ctx in ctxs:
+                try:
+                    loc = ctx.locator("input[type='checkbox']:not([disabled])")
+                    n = loc.count()
+                    if n >= target_n:
+                        all_cbs = [loc.nth(i) for i in range(target_n)]
+                        picked_ctx = ctx
+                        break
+                    if n > 0 and picked_ctx is None:
+                        all_cbs = [loc.nth(i) for i in range(min(n, target_n))]
+                        picked_ctx = ctx
+                except Exception:
+                    continue
+            if not all_cbs:
+                print(
+                    f"[WARN] 「{product_name}」checkbox={checkbox_count}"
+                    " 但拿不到任何 locator，本款式不挂"
+                )
+                continue
+
+            selected_this = 0
+            for cb in all_cbs:
+                try:
+                    if cb.is_checked():
+                        selected_this += 1
+                        continue
+                except Exception:
+                    pass
+                ok = False
+                try:
+                    cb.check(timeout=3000, force=True)
+                    ok = True
+                except Exception:
+                    try:
+                        cb.click(timeout=2000, force=True)
+                        ok = True
+                    except Exception as e2:
+                        print(f"    [诊断] 单个勾选失败: {e2}")
+                if ok:
+                    selected_this += 1
+                time.sleep(0.15)
+
+            total_selected += selected_this
+            if selected_this:
+                linked_styles.add(product_name)
+            print(
+                f"[OK] 「{product_name}」{'ID' if exact_id else '关键词'}结果"
+                f"已勾 {selected_this}/{checkbox_count} "
+                f"（累计 {total_selected}/{MAX_LINK}）"
+            )
+
+        if total_selected == 0:
+            # 全部款式都无匹配 → 视同整体跳过
+            print(
+                f"[WARN] {len(product_names)} 个款式全部无匹配 → 跳过本次发布"
+            )
+            try:
+                self._snapshot("11b_no_product_match")
+            except Exception:
+                pass
+            for label in ["取消", "关 闭", "关闭"]:
+                try:
+                    self._click_by_text(label, timeout=2000)
+                    break
+                except Exception:
+                    continue
+            else:
+                try:
+                    self.page.keyboard.press("Escape")
+                except Exception:
+                    pass
+            time.sleep(0.5)
+            return False
+
+        # 全部款式搜过 → 点「确定」
+        try:
+            self._click_by_text("确定", timeout=5_000)
+            print(
+                f"[OK] 已确认关联商品（{total_selected} 个槽位，"
+                f"覆盖 {len(linked_styles)}/{len(product_names)} 个款式）"
+            )
+            return True
+        except Exception as e:
+            print(f"[WARN] 关联商品点击「确定」失败: {e} → 跳过本次发布")
+            try:
+                self._snapshot("11d_confirm_failed")
+            except Exception:
+                pass
+            for label in ["取消", "关 闭", "关闭"]:
+                try:
+                    self._click_by_text(label, timeout=2000)
+                    break
+                except Exception:
+                    continue
+            else:
+                try:
+                    self.page.keyboard.press("Escape")
+                except Exception:
+                    pass
+            time.sleep(0.5)
+            return False
+
+    def _click_upload_button(self):
+        """点发布页里的橙色「上传视频」按钮（等同手动点击，会弹系统文件选择框）。
+        依次尝试：button 元素 → 可见文本 → 上传拖拽区域。"""
+        for sel in [
+            "button:has-text('上传视频')",
+            "text=上传视频 >> visible=true",
+            "text=点击上传视频 >> visible=true",
+        ]:
+            try:
+                loc = self._loc(sel).first
+                if loc.count() > 0:
+                    loc.click(timeout=5_000)
+                    return
+            except Exception:
+                continue
+        raise TimeoutError("未找到「上传视频」按钮")
+
+    def _find_file_input(self):
+        """在主页面和所有 iframe 里找 file input，优先 accept 含 video/mp4 的那个。"""
+        for frame in self.page.frames:
+            try:
+                loc = frame.locator("input[type='file']")
+                n = loc.count()
+                if n == 0:
+                    continue
+                for i in range(n):
+                    accept = (loc.nth(i).get_attribute("accept") or "").lower()
+                    if "video" in accept or "mp4" in accept:
+                        return loc.nth(i)
+                return loc.first
+            except Exception:
+                continue
+        return None
+
+    def _upload_video_file(self, video_path: Path):
+        """上传视频文件。主路径：点「上传视频」按钮 + file chooser（等同手动操作）；
+        兜底：直接对 file input set_input_files。两条路都失败才抛错。"""
+        chooser_err = None
+        try:
+            with self.page.expect_file_chooser(timeout=15_000) as fc_info:
+                self._click_upload_button()
+            fc_info.value.set_files(str(video_path))
+            print("[OK] 已通过文件选择框选择视频文件（点按钮模式）")
+            return
+        except Exception as e:
+            chooser_err = e
+            print(f"[WARN] 点按钮+文件选择框失败: {e}，尝试直接 set_input_files")
+
+        try:
+            file_input = self._find_file_input()
+            if file_input is None:
+                raise RuntimeError("页面里没找到 input[type='file']")
+            file_input.set_input_files(str(video_path), timeout=15000)
+            print("[OK] 已通过 file input 选择视频文件（set_input_files 模式）")
+            return
+        except Exception as e:
+            raise RuntimeError(
+                f"上传视频失败: 文件选择框: {chooser_err}; set_input_files: {e}"
+            )
+
+    @staticmethod
+    def _cover_generation_is_pending(text: str) -> bool:
+        return "封面生成中" in (text or "")
+
+    def _wait_upload_done(self, timeout: int = 300) -> bool:
+        """等视频上传完成。带进度日志和定期截图，便于排障。
+
+        视频描述输入框在文件刚选中、进度仍为 0% 时就会出现，不能作为完成依据。
+        必须等到视频预览或智能封面区域出现；若页面明确显示上传失败则立即返回，
+        交给上层重新打开干净发布页重试。
+        所有检查都穿透 pub_frame，因为发布页表单嵌在 iframe 里
+        """
+        text_markers = ["智能封面", "上传成功", "上传完成", "更换视频", "编辑封面"]
+        failure_markers = ["视频上传失败", "上传视频失败"]
+        ctxs = ([self.pub_frame] if self.pub_frame is not None else []) + [self.page]
+        deadline = time.time() + timeout
+        last_log = 0
+        while time.time() < deadline:
+            for marker in failure_markers:
+                for ctx in ctxs:
+                    try:
+                        failed = ctx.get_by_text(marker, exact=False).first
+                        if failed.count() > 0 and failed.is_visible():
+                            print(f"    [WARN] 检测到上传失败标志: {marker}")
+                            return False
+                    except Exception:
+                        continue
+            cover_pending = False
+            for ctx in ctxs:
+                for marker in ["封面生成中", "封面图生成中"]:
+                    try:
+                        pending = ctx.get_by_text(marker, exact=False).first
+                        if pending.count() > 0 and pending.is_visible():
+                            cover_pending = True
+                            break
+                    except Exception:
+                        continue
+                if cover_pending:
+                    break
+            if cover_pending:
+                time.sleep(1)
+                continue
+            # 视频预览只会在文件真正上传并完成转码后出现。
+            for ctx in ctxs:
+                try:
+                    preview = ctx.locator("video").first
+                    if preview.count() > 0 and preview.is_visible():
+                        print("    [OK] 检测到上传完成标志: 视频预览")
+                        return True
+                except Exception:
+                    continue
+            # 兼容平台不使用 <video> 元素的版本。
+            for m in text_markers:
+                for ctx in ctxs:
+                    try:
+                        loc = ctx.locator(f"text={m} >> visible=true").first
+                        if loc.count() > 0:
+                            print(f"    [OK] 检测到上传完成标志: {m}")
+                            return True
+                    except Exception:
+                        continue
+            elapsed = int(timeout - (deadline - time.time()))
+            if elapsed - last_log >= 15:
+                last_log = elapsed
+                # 顺手抓一下页面上的百分比进度文本，方便看是不是卡在某个进度
+                pct = ""
+                for ctx in ctxs:
+                    try:
+                        pct_loc = ctx.locator("text=/%/ >> visible=true").first
+                        if pct_loc.count() > 0:
+                            pct = pct_loc.inner_text(timeout=500)
+                            break
+                    except Exception:
+                        continue
+                print(f"    [上传中] 已等待 {elapsed}s{f'，页面进度: {pct}' if pct else ''}")
+                self._snapshot(f"08_uploading_{elapsed}s")
+            time.sleep(1)
+        return False
+
+    def publish_video(
+        self,
+        video_path: Path,
+        product_names: List[str],
+        stop_before_publish: bool = False,
+        video_index: int = 1,
+        video_total: int = 1,
+        tm_item_ids_by_style: Optional[Dict[str, List[str]]] = None,
+    ) -> bool:
+        """上传视频 → AI标题 → 内容标签 → 固定话题 → 商品 → 声明 → 发布。"""
+        print(f"[->] 开始上传视频: {video_path.name}")
+
+        # 1) 上传视频：优先点「上传视频」按钮走系统文件选择框（等同手动操作），
+        #    兜底才是直接 set_input_files 到 hidden file input。
+        #    （2026-07-20 实测：直接 set_input_files 后上传不触发，页面一直停在"等待视频上传"）
+        upload_ok = False
+        for attempt in range(1, 3):
+            self._upload_video_file(video_path)
+            if self._wait_upload_done(timeout=300):
+                upload_ok = True
+                break
+            self._snapshot(f"08_upload_failed_attempt_{attempt}")
+            if attempt < 2:
+                print("[WARN] 视频上传失败，重新打开发布页后自动重试 1 次")
+                self.navigate_to_publish_video()
+        if not upload_ok:
+            raise RuntimeError("视频上传失败（已自动重试 1 次）")
+        print("[OK] 视频上传完成")
+        self._snapshot("08_upload_done")
+
+        # 2) 视频描述标题：Hermes 生成 30 字以内（多款式时走通用模式）
+        title = self._generate_title(
+            product_names,
+            video_index=video_index,
+            video_total=video_total,
+        )
+        try:
+            # 优先匹配 placeholder 含「标题」/「加个」的输入框（要穿透 pub_frame）
+            title_loc = None
+            for sel in [
+                "input[placeholder*='标题']",
+                "input[placeholder*='加个']",
+                "input[placeholder*='吸引']",
+            ]:
+                try:
+                    loc = self._loc(sel).first
+                    if loc.count() > 0:
+                        title_loc = loc
+                        break
+                except Exception:
+                    continue
+            if title_loc is None:
+                # 兜底：第一个 type=text 的 input（在 pub_frame 或主页面）
+                title_loc = self._loc("input[type='text']").first
+            title_loc.fill(title)
+            print(f"[OK] 已填写视频标题: {title}")
+        except Exception as e:
+            print(f"[WARN] 填标题失败: {e}，请手动填写")
+        self._snapshot("09_title_filled")
+
+        # 3) 内容标签 + 固定参与话题。任一步失败都停止本条，避免带错元数据发布。
+        try:
+            self._prepare_publish_metadata(product_names)
+        except Exception as e:
+            print(f"[WARN] 内容标签/固定话题设置失败: {e}，停止本条发布")
+            self._snapshot("10_metadata_failed")
+            return False
+
+        # 4) 关联商品（任何「无法成功关联」均跳过发布 — 不点「立即发布」、不回写飞书）
+        try:
+            linked = self._link_product(product_names, tm_item_ids_by_style)
+        except Exception as e:
+            print(f"[WARN] 关联商品步骤前置失败: {e}，跳过本次发布")
+            linked = False
+        if not linked:
+            print(f"[WARN] 本条记录跳过发布；下次循环会重新走发布页入口")
+            try:
+                self._snapshot("11_product_no_match_skip")
+            except Exception:
+                pass
+            # 弹窗可能还挡着，先用 ESC 兜底关掉，再回首页丢弃表单
+            try:
+                self.page.keyboard.press("Escape")
+            except Exception:
+                pass
+            time.sleep(0.5)
+            try:
+                self.page.goto("https://creator.guanghe.taobao.com/page/",
+                               wait_until="domcontentloaded", timeout=30000)
+                self._sync_page()
+            except Exception:
+                pass
+            self.pub_frame = None
+            return False
+        self._snapshot("11_product_linked")
+
+        # 5) 创作声明（穿透 pub_frame）
+        try:
+            self._loc("text=内容无需标注").first.click(timeout=5_000)
+            print("[OK] 已选创作声明: 内容无需标注")
+        except Exception as e:
+            print(f"[WARN] 选创作声明失败: {e}")
+        self._snapshot("12_declaration_set")
+
+        # 6) 立即发布
+        if stop_before_publish:
+            print("[OK] stop-before-publish 模式：所有信息已填好，停在点击「立即发布」之前")
+            self._snapshot("13_ready_to_publish")
+            return False
+        try:
+            self._loc("text=立即发布").first.click(timeout=5_000)
+            print("[OK] 已点击「立即发布」")
+        except Exception as e:
+            print(f"[ERR] 点击「立即发布」失败: {e}")
+            return False
+
+        # 等待发布完成（提示「发布成功」或「内容已发布」之类）
+        published = False
+        for _ in range(30):
+            time.sleep(1)
+            for marker in ["发布成功", "已发布", "发布完成", "内容已发布"]:
+                try:
+                    if self._loc(f"text={marker}").count() > 0:
+                        published = True
+                        break
+                except Exception:
+                    continue
+            if published:
+                break
+        time.sleep(2)
+        self._snapshot("13_published")
+        if published:
+            print("[OK] 发布成功")
+        else:
+            print("[WARN] 未明确检测到发布成功提示，请人工确认")
+        return published
+
+    def _click_by_text(self, text: str, timeout: int = 10_000):
+        """根据文本点击元素，支持逗号分隔多个候选文本。优先在 pub_frame(iframe)里找，fallback 到主页面。"""
+        candidates = [t.strip() for t in text.split(",")]
+        ctxs = ([self.pub_frame] if self.pub_frame is not None else []) + [self.page]
+        for t in candidates:
+            for ctx in ctxs:
+                try:
+                    loc = ctx.locator(f"text={t}").first
+                    if loc.count() > 0:
+                        loc.click(timeout=timeout)
+                        return
+                except Exception:
+                    continue
+        raise TimeoutError(f"无法点击文本: {text}")
+
+    def _click_button_by_text(self, text: str, timeout: int = 10_000):
+        """精准点 button 内含指定文本的元素（避免 text= 命中非按钮的提示文字）。
+        同样优先在 pub_frame(iframe) 里找。"""
+        candidates = [t.strip() for t in text.split(",")]
+        ctxs = ([self.pub_frame] if self.pub_frame is not None else []) + [self.page]
+        for ctx in ctxs:
+            for t in candidates:
+                try:
+                    loc = ctx.locator(f"button:has-text('{t}')").first
+                    if loc.count() > 0:
+                        loc.click(timeout=timeout)
+                        return
+                except Exception:
+                    continue
+            for t in candidates:
+                try:
+                    loc = ctx.locator(f"text={t}").first
+                    if loc.count() > 0:
+                        loc.click(timeout=timeout)
+                        return
+                except Exception:
+                    continue
+        raise TimeoutError(f"未找到 button: {text}")
+
+    def close(self):
+        if self.session:
+            try:
+                self.session.close()
+            except Exception:
+                pass
+        # 兜底：StealthySession 在 persistent 模式下不会真正关闭 Chrome 进程，这里强制清理
+        _kill_guanghe_chrome(self.user_data_dir)
+
+
+# ─────────────────────────── 主流程 ───────────────────────────
+def parse_args() -> argparse.Namespace:
+    p = argparse.ArgumentParser(description="将飞书 Base 视频上传到淘宝光合平台")
+    p.add_argument("--limit", type=int, default=None, help="最多处理多少条记录")
+    p.add_argument("--debug", action="store_true", help="启用调试截图")
+    p.add_argument("--dry-run", action="store_true", help="只下载和检查，不实际上传")
+    p.add_argument("--skip-upload", action="store_true", help="下载后不上传也不更新表格")
+    p.add_argument("--manual-login", action="store_true", help="只开浏览器，等待手动登录成功后再继续（推荐，最稳）")
+    p.add_argument("--stop-before-publish", action="store_true",
+                   help="测试用：填完所有信息后停在「立即发布」之前，不真正发布、不回写表格")
+    p.add_argument("--product-name", type=str, default=None,
+                   help="只处理「代拍产品」字段包含此子串的记录，如「觅光」「白鹭」")
+    p.add_argument("--record-id", type=str, default=None,
+                   help="只处理指定飞书记录 ID，用于安全重试单条记录")
+    p.add_argument("--video-name", type=str, default=None,
+                   help="只处理附件名完全匹配的单个视频，需配合 --record-id 使用")
+    p.add_argument(
+        "--exclude-video-name",
+        action="append",
+        default=[],
+        help="续跑单条记录时排除已发布的附件名，可重复传入；需配合 --record-id 使用",
+    )
+    p.add_argument(
+        "--store",
+        choices=[STORE_FLAGSHIP, STORE_LUGGAGE, STORE_ALL],
+        default=STORE_FLAGSHIP,
+        help="上传店铺：flagship=旗舰店，luggage=箱包店，all=顺序运行两家",
+    )
+    p.add_argument(
+        "--republish-from-date",
+        type=str,
+        default=None,
+        metavar="YYYY-MM-DD",
+        help=(
+            "旗舰店显式重发模式：选择日期不早于该日的记录，并忽略已勾选的天猫上传状态；"
+            "日常运行不要传此参数"
+        ),
+    )
+    return p.parse_args()
+
+
+def _select_values(value: Any) -> List[str]:
+    if value is None:
+        return []
+    if isinstance(value, list):
+        return [str(item).strip() for item in value if str(item).strip()]
+    text = str(value).strip()
+    return [text] if text else []
+
+
+def select_pending_records(
+    records: List[Dict[str, Any]],
+    store: str = STORE_FLAGSHIP,
+    luggage_style_whitelist: Optional[set[str]] = None,
+    republish_from_date: Optional[str] = None,
+) -> List[Dict[str, Any]]:
+    """Apply the explicit brand/gender/status route for one Guanghe store."""
+    if store not in {STORE_FLAGSHIP, STORE_LUGGAGE}:
+        raise ValueError(f"未知光合店铺: {store}")
+    if store == STORE_LUGGAGE and not luggage_style_whitelist:
+        raise ValueError("箱包店款式白名单未加载或为空")
+    if republish_from_date and store != STORE_FLAGSHIP:
+        raise ValueError("--republish-from-date 仅支持旗舰店")
+    normalized_whitelist = {
+        _normalize_style_name(name) for name in (luggage_style_whitelist or set())
+    }
+
+    result = []
+    for rec in records:
+        if "光影行星" not in _select_values(rec.get("品牌")):
+            continue
+        genders = _select_values(rec.get("性别"))
+        if store == STORE_FLAGSHIP and not any(
+            gender in {"男", "女"} for gender in genders
+        ):
+            continue
+        if store == STORE_LUGGAGE and "女" not in genders:
+            continue
+        if store == STORE_LUGGAGE:
+            product_names = GuangHeUploader._parse_product_names(
+                rec.get("代拍产品")
+            )
+            normalized_products = {
+                _normalize_style_name(name)
+                for name in product_names
+                if name and name != "__SKIP__"
+            }
+            if (
+                not normalized_products
+                or "__SKIP__" in product_names
+                or not normalized_products.issubset(normalized_whitelist)
+            ):
+                continue
+        if store == STORE_FLAGSHIP:
+            if republish_from_date:
+                record_date = str(rec.get("日期") or "")[:10]
+                if not record_date or record_date < republish_from_date:
+                    continue
+            elif rec.get(TM_UPLOAD_FIELD) is True:
+                continue
+        if not rec.get("代拍附件"):
+            continue
+        result.append(rec)
+    return result
+
+
+def _store_runtime(store: str) -> Tuple[str, str, str, str]:
+    if store == STORE_FLAGSHIP:
+        return (
+            "光影行星旗舰店",
+            os.getenv("GUANGHE_USERNAME", ""),
+            os.getenv("GUANGHE_PASSWORD", ""),
+            USER_DATA_DIR,
+        )
+    if store == STORE_LUGGAGE:
+        return (
+            "光影行星箱包旗舰店",
+            os.getenv("GUANGHE_LUGGAGE_USERNAME", ""),
+            os.getenv("GUANGHE_LUGGAGE_PASSWORD", ""),
+            LUGGAGE_USER_DATA_DIR,
+        )
+    raise ValueError(f"未知光合店铺: {store}")
+
+
+def _run_store(
+    args,
+    store: str,
+    records: List[Dict[str, Any]],
+    luggage_style_whitelist: Optional[set[str]] = None,
+    title_registry: Optional[Dict[str, Any]] = None,
+) -> None:
+    store_name, username, password, user_data_dir = _store_runtime(store)
+    pending = select_pending_records(
+        records,
+        store,
+        luggage_style_whitelist=luggage_style_whitelist,
+        republish_from_date=args.republish_from_date,
+    )
+
+    rule = (
+        (
+            "品牌=光影行星 + 性别=男/女 + "
+            f"日期>={args.republish_from_date} + 有附件（显式重发，忽略{TM_UPLOAD_FIELD}）"
+            if args.republish_from_date
+            else "品牌=光影行星 + 性别=男/女 + 天猫未勾选 + 有附件"
+        )
+        if store == STORE_FLAGSHIP
+        else f"品牌=光影行星 + 性别=女 + 款式白名单 + 有附件（忽略{TM_UPLOAD_FIELD}）"
+    )
+    print(f"\n[店铺] {store_name}")
+    print(f"[OK] 符合条件（{rule}）: {len(pending)} 条")
+
+    if args.dry_run:
+        for rec in pending:
+            print(
+                f"  - {rec.get('代拍产品')} | 性别={_select_values(rec.get('性别'))} "
+                f"| 天猫={rec.get(TM_UPLOAD_FIELD)} "
+                f"(record_id={rec.get('record_id')})"
+            )
+        return
+
+    if not username or not password:
+        prefix = "GUANGHE" if store == STORE_FLAGSHIP else "GUANGHE_LUGGAGE"
+        raise RuntimeError(f"请在 .env 中配置 {prefix}_USERNAME / {prefix}_PASSWORD")
+
+    if not pending:
+        print(f"[OK] {store_name} 没有待上传的视频")
+        return
+
+    pending_styles: List[str] = []
+    for rec in pending:
+        for style_name in GuangHeUploader._parse_product_names(
+            rec.get("代拍产品") or ""
+        ):
+            if style_name != "__SKIP__" and style_name not in pending_styles:
+                pending_styles.append(style_name)
+    try:
+        tm_item_ids_by_style = load_tm_item_ids_for_styles(pending_styles)
+        total_tm_ids = sum(len(ids) for ids in tm_item_ids_by_style.values())
+        print(
+            f"[OK] 数据库天猫商品 ID：覆盖 "
+            f"{len(tm_item_ids_by_style)}/{len(pending_styles)} 个待传款式，"
+            f"共 {total_tm_ids} 个 ID"
+        )
+        tm_product_titles_by_style = load_tm_product_titles_for_styles(
+            tm_item_ids_by_style
+        )
+        catalog_categories = GuangHeUploader.configure_catalog_product_titles(
+            tm_product_titles_by_style
+        )
+        if catalog_categories:
+            category_summary = "、".join(
+                f"{style}={category}" for style, category in catalog_categories.items()
+            )
+            print(f"[OK] 数据库商品标题包型: {category_summary}")
+    except Exception as exc:
+        tm_item_ids_by_style = {}
+        GuangHeUploader.configure_catalog_product_titles({})
+        print(f"[WARN] 读取 dim_style.tm_spus 失败，将退回款式关键词搜索: {exc}")
+
+    uploader = GuangHeUploader(
+        username,
+        password,
+        headless=HEADLESS,
+        debug=args.debug,
+        manual=args.manual_login,
+        user_data_dir=user_data_dir,
+    )
+    if title_registry is not None:
+        uploader._used_titles_by_products = title_registry.setdefault("by_products", {})
+        uploader._used_titles_global = title_registry.setdefault("global", [])
+        uploader._title_variant_usage = title_registry.setdefault("variant_usage", {})
+    try:
+        uploader.start()
+        uploader.ensure_login()
+
+        for rec in pending:
+            record_id = rec.get("record_id")
+            product_field = rec.get("代拍产品") or ""
+            product_names = GuangHeUploader._parse_product_names(product_field)
+
+            # 含「未到货 / 不发该视频」等否决关键词 → 整条跳过
+            if "__SKIP__" in product_names:
+                print(
+                    f"\n[跳过] {record_id} 「{product_field}」含「未到货 / 不发」标注,整条跳过"
+                )
+                continue
+
+            if not product_names:
+                product_names = ["光影行星短视频"]
+            multi_flag = " [多款式]" if len(product_names) > 1 else ""
+            display_name = " / ".join(product_names) if len(product_names) > 1 else product_names[0]
+            attachments = rec.get("代拍附件", [])
+            if args.video_name:
+                expected_name = args.video_name.strip().casefold()
+                attachments = [
+                    att for att in attachments
+                    if str(att.get("name", "")).strip().casefold() == expected_name
+                ]
+                if not attachments:
+                    print(
+                        f"\n[跳过] {record_id} 未找到精确匹配附件: {args.video_name}"
+                    )
+                    continue
+            if args.exclude_video_name:
+                excluded_names = {
+                    str(name).strip().casefold()
+                    for name in args.exclude_video_name
+                    if str(name).strip()
+                }
+                before_count = len(attachments)
+                attachments = [
+                    att for att in attachments
+                    if str(att.get("name", "")).strip().casefold()
+                    not in excluded_names
+                ]
+                excluded_count = before_count - len(attachments)
+                if excluded_count:
+                    print(
+                        f"[OK] 续跑排除已发布附件: {excluded_count} 个 "
+                        f"({', '.join(args.exclude_video_name)})"
+                    )
+                if not attachments:
+                    print(f"\n[跳过] {record_id} 排除后没有待上传附件")
+                    continue
+
+            print(f"\n[->] 处理: {display_name} ({record_id}){multi_flag}")
+            video_paths = download_videos(record_id, attachments)
+            if not video_paths:
+                continue
+
+            if args.skip_upload:
+                print("[OK] skip-upload 模式，不上传")
+                continue
+
+            total = len(video_paths)
+            n_published = 0
+            n_failed = 0
+
+            for idx, video_path in enumerate(video_paths, 1):
+                print(f"\n[->] [{idx}/{total}] 上传视频: {video_path.name}")
+                # 每条视频都重新走一遍发布页（视频之间状态不共享）
+                uploader.navigate_to_publish_video()
+                upload_path = ensure_upload_compatible_video(video_path)
+                published = uploader.publish_video(
+                    upload_path, product_names,
+                    stop_before_publish=args.stop_before_publish,
+                    video_index=idx,
+                    video_total=total,
+                    tm_item_ids_by_style=tm_item_ids_by_style,
+                )
+                if published:
+                    n_published += 1
+                else:
+                    n_failed += 1
+                    print(f"[WARN] 第 {idx}/{total} 个视频未确认发布成功")
+
+            if n_failed == 0 and n_published == total:
+                if store == STORE_LUGGAGE:
+                    print(
+                        f"[OK] 箱包店全部 {total} 个视频发布成功；"
+                        f"按规则不修改「{TM_UPLOAD_FIELD}」"
+                    )
+                elif mark_tm_uploaded(record_id):
+                    print(
+                        f"[OK] 全部 {total} 个视频发布成功，"
+                        f"已回写飞书「{TM_UPLOAD_FIELD}」= true"
+                    )
+                else:
+                    print(f"[ERR] 全部 {total} 个视频发布成功，但飞书回写失败: {record_id}")
+            else:
+                suffix = (
+                    f"不修改「{TM_UPLOAD_FIELD}」"
+                    if store == STORE_LUGGAGE
+                    else f"不勾选「{TM_UPLOAD_FIELD}」"
+                )
+                print(f"[WARN] 仅完成 {n_published}/{total}，{suffix}(record_id={record_id})")
+    finally:
+        uploader.close()
+
+
+def main():
+    args = parse_args()
+
+    if args.republish_from_date:
+        try:
+            time.strptime(args.republish_from_date, "%Y-%m-%d")
+        except ValueError as exc:
+            raise ValueError("--republish-from-date 必须是 YYYY-MM-DD") from exc
+        if args.store != STORE_FLAGSHIP:
+            raise ValueError("--republish-from-date 仅支持 --store flagship")
+
+    print("[->] 读取飞书 Base 记录...")
+    records = fetch_records(
+        limit=args.limit,
+        product_name_contains=args.product_name if args.product_name else None,
+    )
+    if args.video_name and not args.record_id:
+        raise ValueError("--video-name 必须配合 --record-id 使用，避免误传同名附件")
+    if args.exclude_video_name and not args.record_id:
+        raise ValueError("--exclude-video-name 必须配合 --record-id 使用，避免误排附件")
+    if args.record_id:
+        records = [
+            rec for rec in records
+            if str(rec.get("record_id", "")).strip() == args.record_id.strip()
+        ]
+    print(f"[OK] 读取到 {len(records)} 条记录")
+
+    stores = (
+        [STORE_FLAGSHIP, STORE_LUGGAGE]
+        if args.store == STORE_ALL
+        else [args.store]
+    )
+    luggage_style_whitelist = None
+    if STORE_LUGGAGE in stores:
+        luggage_style_whitelist = fetch_luggage_style_whitelist()
+        print(f"[OK] 箱包店款式白名单已加载: {len(luggage_style_whitelist)} 款")
+    title_registry: Dict[str, Any] = {
+        "by_products": {},
+        "global": [],
+        "variant_usage": {},
+    }
+    for store in stores:
+        _run_store(
+            args,
+            store,
+            records,
+            luggage_style_whitelist=luggage_style_whitelist,
+            title_registry=title_registry,
+        )
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/adaptive_selectors.py b/src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/adaptive_selectors.py
new file mode 100644
index 0000000..cfe944a
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/adaptive_selectors.py
@@ -0,0 +1,310 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Adaptive selector helpers for Scrapling response pages and Playwright pages.
+
+Scrapling's native adaptive selectors work on Scrapling Response objects. The
+current scraping flow also uses Playwright pages inside page_action, so this
+module adds a small fingerprint fallback for text clicks and inputs there.
+"""
+
+import json
+import os
+import re
+from typing import Any
+
+from scrapling import DynamicFetcher, Fetcher
+
+try:
+    from scrapling import StealthyFetcher
+except ModuleNotFoundError:
+    StealthyFetcher = None
+
+
+ADAPTIVE_SELECTOR_DIR = os.path.join(os.path.dirname(__file__), "data", "adaptive_selectors")
+PLAYWRIGHT_SELECTOR_STORE = os.path.join(ADAPTIVE_SELECTOR_DIR, "playwright_selectors.json")
+
+
+def enable_adaptive_fetchers() -> None:
+    Fetcher.adaptive = True
+    DynamicFetcher.adaptive = True
+    if StealthyFetcher is not None:
+        StealthyFetcher.adaptive = True
+
+
+def adaptive_css(page: Any, selector: str, identifier: str = "", percentage: int = 40):
+    """Use Scrapling native adaptive CSS and save successful element features."""
+    identifier = identifier or selector
+    try:
+        result = page.css(selector, identifier=identifier, adaptive=True, percentage=percentage)
+        if result:
+            return result
+    except Exception:
+        pass
+    return page.css(selector, identifier=identifier, auto_save=True, percentage=percentage)
+
+
+def adaptive_xpath(page: Any, selector: str, identifier: str = "", percentage: int = 40, **kwargs):
+    """Use Scrapling native adaptive XPath and save successful element features."""
+    identifier = identifier or selector
+    try:
+        result = page.xpath(selector, identifier=identifier, adaptive=True, percentage=percentage, **kwargs)
+        if result:
+            return result
+    except Exception:
+        pass
+    return page.xpath(selector, identifier=identifier, auto_save=True, percentage=percentage, **kwargs)
+
+
+def _load_store() -> dict:
+    try:
+        with open(PLAYWRIGHT_SELECTOR_STORE, "r", encoding="utf-8") as f:
+            return json.load(f)
+    except Exception:
+        return {}
+
+
+def _save_store(store: dict) -> None:
+    os.makedirs(ADAPTIVE_SELECTOR_DIR, exist_ok=True)
+    with open(PLAYWRIGHT_SELECTOR_STORE, "w", encoding="utf-8") as f:
+        json.dump(store, f, ensure_ascii=False, indent=2)
+
+
+def _key(namespace: str, identifier: str) -> str:
+    return f"{namespace}:{identifier}"
+
+
+def _safe_identifier(value: str) -> str:
+    return re.sub(r"\s+", " ", str(value or "")).strip()
+
+
+def _get_fingerprint(page, selector: str, text: str = "", exact: bool = True) -> dict | None:
+    try:
+        return page.evaluate(
+            """([selector, text, exact]) => {
+                const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const classTokens = (el) => String(el.className || '')
+                    .split(/\\s+/)
+                    .filter(Boolean)
+                    .filter(token => token.length > 2)
+                    .slice(0, 12);
+                const candidates = Array.from(document.querySelectorAll(selector))
+                    .filter(isVisible)
+                    .filter(el => {
+                        if (!text) return true;
+                        const value = clean(el.innerText || el.textContent || el.value || el.placeholder);
+                        return exact ? value === text : value.includes(text);
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return null;
+                const el = candidates[0];
+                const rect = el.getBoundingClientRect();
+                return {
+                    tag: el.tagName.toLowerCase(),
+                    text: clean(el.innerText || el.textContent || ''),
+                    value: clean(el.value || ''),
+                    placeholder: clean(el.getAttribute('placeholder') || ''),
+                    aria: clean(el.getAttribute('aria-label') || ''),
+                    title: clean(el.getAttribute('title') || ''),
+                    role: clean(el.getAttribute('role') || ''),
+                    type: clean(el.getAttribute('type') || ''),
+                    class_tokens: classTokens(el),
+                    width: Math.round(rect.width),
+                    height: Math.round(rect.height),
+                };
+            }""",
+            [selector, text, exact],
+        )
+    except Exception:
+        return None
+
+
+def _find_by_fingerprint(page, selector: str, fingerprint: dict) -> dict | None:
+    if not fingerprint:
+        return None
+    try:
+        return page.evaluate(
+            """([selector, fp]) => {
+                const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const classTokens = (el) => String(el.className || '').split(/\\s+/).filter(Boolean);
+                const scoreText = (a, b) => {
+                    a = clean(a); b = clean(b);
+                    if (!a || !b) return 0;
+                    if (a === b) return 40;
+                    if (a.includes(b) || b.includes(a)) return 24;
+                    return 0;
+                };
+                const candidates = Array.from(document.querySelectorAll(selector))
+                    .filter(isVisible)
+                    .map(el => {
+                        const rect = el.getBoundingClientRect();
+                        const tokens = classTokens(el);
+                        let score = 0;
+                        if (el.tagName.toLowerCase() === fp.tag) score += 12;
+                        if (clean(el.getAttribute('role') || '') === fp.role && fp.role) score += 10;
+                        if (clean(el.getAttribute('type') || '') === fp.type && fp.type) score += 8;
+                        score += scoreText(el.innerText || el.textContent || '', fp.text);
+                        score += scoreText(el.value || '', fp.value);
+                        score += scoreText(el.getAttribute('placeholder') || '', fp.placeholder);
+                        score += scoreText(el.getAttribute('aria-label') || '', fp.aria);
+                        score += scoreText(el.getAttribute('title') || '', fp.title);
+                        for (const token of (fp.class_tokens || [])) {
+                            if (tokens.includes(token)) score += 3;
+                        }
+                        const sizeGap = Math.abs(Math.round(rect.width) - (fp.width || 0))
+                            + Math.abs(Math.round(rect.height) - (fp.height || 0));
+                        score -= Math.min(20, sizeGap / 80);
+                        return { el, score };
+                    })
+                    .filter(item => item.score >= 25)
+                    .sort((a, b) => b.score - a.score);
+                if (!candidates.length) return null;
+                const rect = candidates[0].el.getBoundingClientRect();
+                return {
+                    x: rect.left + rect.width / 2,
+                    y: rect.top + rect.height / 2,
+                    score: candidates[0].score,
+                };
+            }""",
+            [selector, fingerprint],
+        )
+    except Exception:
+        return None
+
+
+def remember_playwright_selector(page, namespace: str, identifier: str, selector: str, text: str = "", exact: bool = True) -> None:
+    fingerprint = _get_fingerprint(page, selector, text=text, exact=exact)
+    if not fingerprint:
+        return
+    store = _load_store()
+    store[_key(namespace, identifier)] = {
+        "selector": selector,
+        "text": text,
+        "exact": exact,
+        "fingerprint": fingerprint,
+    }
+    _save_store(store)
+
+
+def remember_active_playwright_element(page, namespace: str, identifier: str, selector: str) -> None:
+    try:
+        fingerprint = page.evaluate(
+            """(selector) => {
+                const el = document.activeElement;
+                if (!el || !el.matches(selector)) return null;
+                const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const rect = el.getBoundingClientRect();
+                return {
+                    tag: el.tagName.toLowerCase(),
+                    text: clean(el.innerText || el.textContent || ''),
+                    value: clean(el.value || ''),
+                    placeholder: clean(el.getAttribute('placeholder') || ''),
+                    aria: clean(el.getAttribute('aria-label') || ''),
+                    title: clean(el.getAttribute('title') || ''),
+                    role: clean(el.getAttribute('role') || ''),
+                    type: clean(el.getAttribute('type') || ''),
+                    class_tokens: String(el.className || '').split(/\\s+/).filter(Boolean).slice(0, 12),
+                    width: Math.round(rect.width),
+                    height: Math.round(rect.height),
+                };
+            }""",
+            selector,
+        )
+        if not fingerprint:
+            return
+        store = _load_store()
+        store[_key(namespace, identifier)] = {
+            "selector": selector,
+            "text": "",
+            "exact": True,
+            "fingerprint": fingerprint,
+        }
+        _save_store(store)
+    except Exception:
+        return
+
+
+def adaptive_click_text(page, text: str, exact: bool = True, timeout: int = 10000) -> bool:
+    selector = "button, [role='button'], a, label, span, div, li"
+    identifier = _safe_identifier(f"{text}|exact={exact}")
+    try:
+        locator = page.get_by_text(text, exact=exact)
+        remember_playwright_selector(page, "text", identifier, selector, text=text, exact=exact)
+        locator.first.click(timeout=timeout)
+        return True
+    except Exception:
+        pass
+
+    store = _load_store()
+    entry = store.get(_key("text", identifier))
+    match = _find_by_fingerprint(page, entry.get("selector", selector), entry.get("fingerprint", {})) if entry else None
+    if match:
+        try:
+            page.mouse.click(match["x"], match["y"])
+            return True
+        except Exception:
+            pass
+    return False
+
+
+def adaptive_hover_text(page, text: str, exact: bool = True, timeout: int = 10000) -> bool:
+    selector = "button, [role='button'], a, label, span, div, li"
+    identifier = _safe_identifier(f"{text}|exact={exact}")
+    try:
+        locator = page.get_by_text(text, exact=exact)
+        remember_playwright_selector(page, "text", identifier, selector, text=text, exact=exact)
+        locator.first.hover(timeout=timeout)
+        return True
+    except Exception:
+        pass
+
+    store = _load_store()
+    entry = store.get(_key("text", identifier))
+    match = _find_by_fingerprint(page, entry.get("selector", selector), entry.get("fingerprint", {})) if entry else None
+    if match:
+        try:
+            page.mouse.move(match["x"], match["y"])
+            return True
+        except Exception:
+            pass
+    return False
+
+
+def adaptive_fill_input(page, identifier: str, selector: str, value: str, timeout: int = 8000) -> bool:
+    try:
+        locator = page.locator(selector).first
+        locator.click(timeout=timeout)
+        remember_active_playwright_element(page, "input", identifier, selector)
+        locator.fill(value, timeout=timeout)
+        return True
+    except Exception:
+        pass
+
+    store = _load_store()
+    entry = store.get(_key("input", identifier))
+    match = _find_by_fingerprint(page, entry.get("selector", selector), entry.get("fingerprint", {})) if entry else None
+    if not match:
+        return False
+    try:
+        page.mouse.click(match["x"], match["y"])
+        page.keyboard.press("Control+A")
+        page.keyboard.type(value, delay=20)
+        return True
+    except Exception:
+        return False
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/dy_store_competitor_store_scraping.py b/src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/dy_store_competitor_store_scraping.py
new file mode 100644
index 0000000..1ea3298
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/dy_store_competitor_store_scraping.py
@@ -0,0 +1,2307 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+抖音罗盘经营数据抓取脚本 (基于 scrapling.DynamicFetcher)
+"""
+
+import json
+import os
+import re
+import sys
+import time
+from datetime import datetime, timedelta
+from pathlib import Path
+
+from scrapling import DynamicFetcher
+from runtime_paths import RAW_DATA_ROOT, STATE_ROOT, managed_data_path, profile_path
+
+from adaptive_selectors import adaptive_click_text, adaptive_fill_input, adaptive_hover_text, enable_adaptive_fetchers
+
+try:
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")
+    sys.stderr.reconfigure(encoding="utf-8", errors="replace")
+except Exception:
+    pass
+
+OUTPUT_DIR = os.path.dirname(os.path.abspath(__file__))
+DATA_DIR = str(
+    managed_data_path(os.getenv("DY_DATA_DIR", str(RAW_DATA_ROOT / "dy-store")))
+)
+STORE_DATA_DIR = os.path.join(DATA_DIR, "店铺数据")
+COMPETITOR_DATA_DIR = os.path.join(DATA_DIR, "竞店数据")
+_AUTH_COOKIE_DOMAINS = (
+    "compass.jinritemai.com",
+    "doudian-sso.jinritemai.com",
+    "fxg.jinritemai.com",
+)
+
+
+def _cookie_bundle_score(root: Path) -> tuple[int, int, float]:
+    """Rank saved sessions by SSO coverage, cookie count, then freshness."""
+    cookie_file = root / "compass_cookies.json"
+    try:
+        cookies = json.loads(cookie_file.read_text(encoding="utf-8"))
+        domains = {str(item.get("domain", "")).lstrip(".").lower() for item in cookies}
+        coverage = sum(
+            any(domain == required or domain.endswith(f".{required}") for domain in domains)
+            for required in _AUTH_COOKIE_DOMAINS
+        )
+        return coverage, len(cookies), cookie_file.stat().st_mtime
+    except (OSError, ValueError, TypeError):
+        return 0, 0, 0.0
+
+
+def _choose_session_root(candidates, explicit=None) -> Path:
+    """Choose one coherent cookie/storage/profile bundle without mixing roots."""
+    if explicit:
+        return Path(explicit).expanduser().resolve()
+    roots = [Path(item).expanduser().resolve() for item in candidates]
+    if not roots:
+        raise ValueError("at least one Douyin session root is required")
+    return max(roots, key=_cookie_bundle_score)
+
+
+SESSION_ROOT = Path(os.getenv("DY_SESSION_ROOT", str(STATE_ROOT / "dy"))).resolve()
+COOKIES_FILE = os.getenv("DY_COOKIES_FILE", str(SESSION_ROOT / "compass_cookies.json"))
+STORAGE_STATE_FILE = os.getenv(
+    "DY_STORAGE_STATE_FILE", str(SESSION_ROOT / "compass_storage_state.json")
+)
+USER_DATA_DIR = os.getenv("DY_USER_DATA_DIR", str(profile_path("dy")))
+SHOP_NAME = "光影行星箱包旗舰店"
+LOGIN_TIMEOUT = int(os.getenv("DY_LOGIN_TIMEOUT", "240"))
+KEEP_BROWSER_OPEN = os.getenv("KEEP_BROWSER_OPEN", "0") != "0"
+ACTION_SETTLE_MS = int(os.getenv("ACTION_SETTLE_MS", "1200"))
+PAGE_SETTLE_MS = int(os.getenv("PAGE_SETTLE_MS", "2200"))
+_scraped_metrics = {}
+_market_rankings = {}
+_competitor_rankings = {}
+_qianchuan_metrics = {}
+_experience_score = {}
+_results_saved = False
+MARKET_CATEGORIES = ["单肩包/斜挎包", "双肩包", "腰包/胸包"]
+COMPETITOR_BRANDS = ["Bellroy", "tomtoc", "BAGSMART", "GROTTO", "NIID", "reeyee", "VAOPER", "探迹者"]
+
+
+def _previous_week_range():
+    today = datetime.now().date()
+    this_monday = today - timedelta(days=today.weekday())
+    start = this_monday - timedelta(days=7)
+    end = this_monday - timedelta(days=1)
+    return start, end
+
+
+def _date_range_label(start, end) -> str:
+    return f"{start:%Y/%m/%d} - {end:%Y/%m/%d}"
+
+
+def _load_cookies():
+    if os.path.exists(COOKIES_FILE):
+        try:
+            with open(COOKIES_FILE, "r", encoding="utf-8") as f:
+                cookies = json.load(f)
+            print(f"[OK] Loaded {len(cookies)} cookies")
+            return cookies
+        except Exception as e:
+            print(f"[WARN] Load cookies failed: {e}")
+    return None
+
+
+def _save_cookies(cookies):
+    try:
+        with open(COOKIES_FILE, "w", encoding="utf-8") as f:
+            json.dump(cookies, f, ensure_ascii=False, indent=2)
+        print(f"[OK] Saved {len(cookies)} cookies -> {COOKIES_FILE}")
+        return True
+    except Exception as e:
+        import traceback
+        print(f"[WARN] Save cookies failed: {e}")
+        traceback.print_exc()
+        return False
+
+
+def _save_browser_state(page, stage=""):
+    label = f" ({stage})" if stage else ""
+    try:
+        cookies = page.context.cookies()
+        if cookies:
+            _save_cookies(cookies)
+        else:
+            print(f"[WARN] No cookies found to save{label}")
+
+        page.context.storage_state(path=STORAGE_STATE_FILE)
+        print(f"[OK] Saved storage state{label} -> {STORAGE_STATE_FILE}")
+    except Exception as e:
+        print(f"[WARN] Failed to save browser state{label}: {e}")
+
+
+def _active_page(page):
+    try:
+        pages = [p for p in page.context.pages if not p.is_closed()]
+        if pages:
+            return pages[-1]
+    except Exception:
+        pass
+    return page
+
+
+def _looks_like_login_url(url: str) -> bool:
+    value = str(url or "").lower()
+    if not value:
+        return False
+    return any(
+        marker in value
+        for marker in (
+            "/login",
+            "passport.",
+            "/passport/",
+            "doudian-sso.jinritemai.com",
+        )
+    )
+
+
+def _maximize_page(page) -> None:
+    try:
+        page.set_viewport_size({"width": 1920, "height": 1080})
+    except Exception as e:
+        print(f"[WARN] Set viewport size failed: {e}")
+    try:
+        page.evaluate("() => { window.moveTo(0, 0); window.resizeTo(screen.availWidth, screen.availHeight); }")
+    except Exception:
+        pass
+
+
+def _check_login_status(page) -> bool:
+    """多种方式检测是否已登录"""
+    page = _active_page(page)
+    url = page.url
+    print(f"[DEBUG] check_login: url={url}")
+
+    if _looks_like_login_url(url):
+        print("[DEBUG] Login redirect detected by URL")
+        return False
+
+    # 身份选择页、手机登录页都不是登录完成状态，不能只靠 URL 判断。
+    try:
+        login_markers = [
+            "请选择您的身份",
+            "手机登录",
+            "邮箱登录",
+            "手机号",
+            "验证码",
+            "发送验证码",
+            "登录即代表同意",
+        ]
+        for marker in login_markers:
+            if page.locator(f"text={marker}").count() > 0:
+                print(f"[DEBUG] Still on login/identity page: {marker}")
+                return False
+    except Exception as e:
+        print(f"[DEBUG] Login marker check error: {e}")
+
+    # 方式2: 页面中存在已登录标志元素（店铺名、退出按钮等）
+    try:
+        # 检查是否出现了店铺名称或用户中心
+        if page.locator('text=光影行星箱包旗舰店').count() > 0:
+            print("[OK] Login detected by shop name")
+            return True
+        if page.locator('text=退出登录').count() > 0:
+            print("[OK] Login detected by logout button")
+            return True
+        if page.locator('text=个人中心').count() > 0:
+            print("[OK] Login detected by user center")
+            return True
+        if page.locator('text=成交金额').count() > 0:
+            print("[OK] Login detected by business metric")
+            return True
+        if page.locator('text=成交人数').count() > 0:
+            print("[OK] Login detected by business metric")
+            return True
+    except Exception as e:
+        print(f"[DEBUG] Element check error: {e}")
+
+    return False
+
+
+def _ensure_login(page):
+    print("\n" + "=" * 60)
+    print("【请登录】请在弹出的浏览器窗口中完成登录")
+    print("提示：扫码/短信/密码均可，登录成功后页面会自动跳转")
+    print("=" * 60 + "\n")
+
+    for i in range(LOGIN_TIMEOUT):
+        page = _active_page(page)
+        if _check_login_status(page):
+            print(f"[OK] Login success! URL: {page.url}")
+            page.wait_for_timeout(3000)
+            return page
+        if i % 5 == 0:
+            print(f"[WAIT] Waiting for login... {i}s / {LOGIN_TIMEOUT}s")
+        time.sleep(1)
+
+    print(f"[ERR] Login timeout ({LOGIN_TIMEOUT}s)")
+    return None
+
+
+def _select_menu_and_date(page) -> None:
+    week_start, week_end = _previous_week_range()
+    target_url = "https://compass.jinritemai.com/shop/business-part"
+    print(f"[INFO] Navigating to: {target_url}")
+    print(f"[INFO] Business date range: {_date_range_label(week_start, week_end)}")
+    page.goto(target_url, wait_until="domcontentloaded", timeout=60000)
+    try:
+        page.wait_for_load_state("load", timeout=8000)
+    except Exception as e:
+        print(f"[WARN] load wait skipped: {e}")
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+    print(f"[OK] Page loaded: {page.url}")
+
+    _select_business_previous_week(page, week_start, week_end)
+
+    print("[INFO] Waiting for data render...")
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+    try:
+        page.wait_for_load_state("load", timeout=5000)
+    except Exception:
+        pass
+    print("[OK] Data rendered")
+
+
+def _extract_metrics_via_js(page) -> dict:
+    print("[INFO] Extracting via JS...")
+    metrics = page.evaluate("""
+        () => {
+            const result = {};
+            const targets = [
+                {key: '成交金额', labels: ['成交金额']},
+                {key: '成交人数', labels: ['成交人数']},
+                {key: '客单价', labels: ['客单价']},
+                {key: '成交退款金额', labels: ['成交退款金额']},
+                {key: '商品点击人数', labels: ['商品点击人数']}
+            ];
+            const lines = (document.body.innerText || '')
+                .split('\\n')
+                .map(s => s.trim())
+                .filter(Boolean);
+            const valueRe = /^(¥\\s*)?[\\d,]+(\\.\\d+)?%?$/;
+            const unitRe = /^[万亿]$/;
+
+            function firstValueAfter(index) {
+                for (let i = index + 1; i < Math.min(lines.length, index + 8); i++) {
+                    if (valueRe.test(lines[i])) {
+                        const value = lines[i].replace(/\\s+/g, '');
+                        if (unitRe.test(lines[i + 1] || '')) return value + lines[i + 1];
+                        return value;
+                    }
+                }
+                return null;
+            }
+
+            for (const target of targets) {
+                for (let i = 0; i < lines.length; i++) {
+                    if (target.labels.some(label => lines[i].includes(label))) {
+                        const value = firstValueAfter(i);
+                        if (value) {
+                            result[target.key] = value;
+                            break;
+                        }
+                    }
+                }
+            }
+            return result;
+        }
+    """)
+    cleaned = {k: v for k, v in metrics.items() if v}
+    print(f"[OK] JS extract: {cleaned}")
+    return metrics
+
+
+def _extract_metrics_via_dom_fallback(page) -> dict:
+    print("[INFO] Fallback DOM extract...")
+    metrics = page.evaluate("""
+        () => {
+            const result = {};
+            const targets = ['成交金额', '成交人数', '客单价', '成交退款金额', '商品点击人数'];
+            const allEls = document.querySelectorAll('div, span, p, label');
+            allEls.forEach(el => {
+                const text = el.textContent.trim();
+                targets.forEach(keyword => {
+                    if (text.includes(keyword)) {
+                        const parent = el.closest('[class*="card"], [class*="item"], [class*="metric"], [class*="cell"], [class*="row"]') || el.parentElement;
+                        if (parent) {
+                            const nums = parent.querySelectorAll('span, div, p');
+                            for (const n of nums) {
+                                const nt = n.textContent.trim();
+                                if (/^(¥|￥)?[\\d,.]+(\\.\\d+)?%?$/.test(nt) && nt !== text) {
+                                    result[keyword] = nt;
+                                    break;
+                                }
+                            }
+                        }
+                    }
+                });
+            });
+            return result;
+        }
+    """)
+    cleaned = {k: v for k, v in metrics.items() if v}
+    print(f"[OK] Fallback extract: {cleaned}")
+    return metrics
+
+
+def _wait_after_action(page, label="", timeout=2500, settle_ms=None, wait_network=False) -> None:
+    if wait_network:
+        try:
+            page.wait_for_load_state("networkidle", timeout=timeout)
+        except Exception as e:
+            if label:
+                print(f"[WARN] networkidle wait skipped after {label}: {e}")
+    page.wait_for_timeout(ACTION_SETTLE_MS if settle_ms is None else settle_ms)
+
+
+def _click_calendar_date(page, target_date) -> bool:
+    date_dash = target_date.strftime("%Y-%m-%d")
+    date_slash = target_date.strftime("%Y/%m/%d")
+    day_text = str(target_date.day)
+    print(f"[INFO] Clicking calendar date: {date_dash}")
+    target_month_key = target_date.year * 12 + target_date.month
+
+    def try_click_visible_date() -> bool:
+        try:
+            clicked = page.evaluate(
+                """([dateDash, dateSlash, dayText, targetMonthKey]) => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const inDatePicker = (el) => {
+                    let cur = el;
+                    for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                        const cls = String(cur.className || '').toLowerCase();
+                        const role = (cur.getAttribute && cur.getAttribute('role') || '').toLowerCase();
+                        if (cls.includes('picker') || cls.includes('calendar') || cls.includes('date')
+                            || role.includes('dialog') || role.includes('grid')) {
+                            return true;
+                        }
+                    }
+                    return false;
+                };
+                const disabled = (el) => {
+                    const text = [el.className, el.getAttribute('aria-disabled'), el.getAttribute('disabled')]
+                        .join(' ').toLowerCase();
+                    return text.includes('disabled') || text.includes('true');
+                };
+                const attrs = ['title', 'aria-label', 'data-date', 'data-value', 'value'];
+                const dateHit = Array.from(document.querySelectorAll('td, button, div, span'))
+                    .filter(isVisible)
+                    .filter(inDatePicker)
+                    .filter(el => !disabled(el))
+                    .find(el => attrs.some(attr => {
+                        const value = el.getAttribute(attr) || '';
+                        return value.includes(dateDash) || value.includes(dateSlash);
+                    }));
+                if (dateHit) {
+                    (dateHit.closest('td, button') || dateHit).click();
+                    return true;
+                }
+
+                const visibleMonthKeys = Array.from(document.querySelectorAll('div, span, button, th'))
+                    .filter(isVisible)
+                    .map(el => (el.innerText || el.textContent || '').replace(/\\s+/g, ' ').trim())
+                    .map(text => text.match(/(20\\d{2})\\s*年\\s*(\\d{1,2})\\s*月/))
+                    .filter(Boolean)
+                    .map(match => Number(match[1]) * 12 + Number(match[2]));
+                if (visibleMonthKeys.length && !visibleMonthKeys.includes(targetMonthKey)) {
+                    const nearestGap = Math.min(...visibleMonthKeys.map(key => Math.abs(key - targetMonthKey)));
+                    // Calendars often show previous/next-month dates in gray in the visible month grid.
+                    // Allow adjacent months so a June panel can still select May 25-31 as last week.
+                    if (nearestGap > 1) {
+                        return false;
+                    }
+                }
+
+                const targetAdjacentDirection = visibleMonthKeys.includes(targetMonthKey)
+                    ? 0
+                    : (visibleMonthKeys.length && targetMonthKey < Math.min(...visibleMonthKeys) ? -1 : 1);
+                const dayCandidates = Array.from(document.querySelectorAll('td, button, div, span'))
+                    .filter(isVisible)
+                    .filter(inDatePicker)
+                    .filter(el => !disabled(el))
+                    .filter(el => (el.innerText || el.textContent || '').trim() === dayText)
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        const adjacentBiasA = targetAdjacentDirection < 0 ? ar.top : (targetAdjacentDirection > 0 ? -ar.top : 0);
+                        const adjacentBiasB = targetAdjacentDirection < 0 ? br.top : (targetAdjacentDirection > 0 ? -br.top : 0);
+                        const ascore = adjacentBiasA + (ar.width * ar.height) / 10000 + ar.left / 100000;
+                        const bscore = adjacentBiasB + (br.width * br.height) / 10000 + br.left / 100000;
+                        return ascore - bscore;
+                    });
+                if (!dayCandidates.length) return false;
+                (dayCandidates[0].closest('td, button') || dayCandidates[0]).click();
+                return true;
+            }""",
+                [date_dash, date_slash, day_text, target_month_key],
+            )
+            if clicked:
+                page.wait_for_timeout(1000)
+            return bool(clicked)
+        except Exception as e:
+            print(f"[WARN] Calendar date click failed ({date_dash}): {e}")
+            return False
+
+    def visible_month_keys() -> list[int]:
+        try:
+            return page.evaluate(
+                """() => {
+                    const isVisible = (el) => {
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style && style.visibility !== 'hidden' && style.display !== 'none'
+                            && rect.width > 0 && rect.height > 0;
+                    };
+                    const keys = [];
+                    for (const el of Array.from(document.querySelectorAll('div, span, button, th'))) {
+                        if (!isVisible(el)) continue;
+                        const text = (el.innerText || el.textContent || '').replace(/\\s+/g, ' ').trim();
+                        const match = text.match(/(20\\d{2})\\s*年\\s*(\\d{1,2})\\s*月/);
+                        if (match) keys.push(Number(match[1]) * 12 + Number(match[2]));
+                    }
+                    return Array.from(new Set(keys)).sort((a, b) => a - b);
+                }"""
+            ) or []
+        except Exception:
+            return []
+
+    def click_calendar_nav(direction: str) -> bool:
+        try:
+            clicked = page.evaluate(
+                """(direction) => {
+                    const isVisible = (el) => {
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style && style.visibility !== 'hidden' && style.display !== 'none'
+                            && rect.width > 0 && rect.height > 0;
+                    };
+                    const inDatePicker = (el) => {
+                        let cur = el;
+                        for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                            const cls = String(cur.className || '').toLowerCase();
+                            const role = (cur.getAttribute && cur.getAttribute('role') || '').toLowerCase();
+                            if (cls.includes('picker') || cls.includes('calendar') || cls.includes('date')
+                                || role.includes('dialog') || role.includes('grid')) return true;
+                        }
+                        return false;
+                    };
+                    const candidates = Array.from(document.querySelectorAll('button, span, div, a'))
+                        .filter(isVisible)
+                        .filter(inDatePicker)
+                        .map(el => {
+                            const rect = el.getBoundingClientRect();
+                            const text = (el.innerText || el.textContent || '').trim();
+                            const title = el.getAttribute('title') || '';
+                            const aria = el.getAttribute('aria-label') || '';
+                            const cls = String(el.className || '').toLowerCase();
+                            return { el, rect, text, title, aria, cls };
+                        })
+                        .filter(item => {
+                            const signal = `${item.text} ${item.title} ${item.aria} ${item.cls}`;
+                            if (direction === 'prev') {
+                                return signal.includes('上一月') || signal.includes('prev-month') || signal.includes('previous-month')
+                                    || signal.includes('prevmonth') || signal.includes('chevron_left') || signal.includes('arrowleft')
+                                    || item.text === '‹' || item.text === '<'
+                                    || item.text === '«' || item.text === '<<';
+                            }
+                            return signal.includes('下一月') || signal.includes('next-month') || signal.includes('nextmonth')
+                                || signal.includes('chevron_right') || signal.includes('arrowright')
+                                || item.text === '›' || item.text === '>' || item.text === '»' || item.text === '>>';
+                        })
+                        .map(item => {
+                            const signal = `${item.text} ${item.title} ${item.aria} ${item.cls}`;
+                            let score = 0;
+                            if (direction === 'prev') {
+                                if (signal.includes('上一月') || signal.includes('prev-month') || signal.includes('previous-month') || signal.includes('prevmonth')) score += 100;
+                                if (signal.includes('chevron_left') || signal.includes('arrowleft')) score += 85;
+                                if (item.text === '‹' || item.text === '<') score += 80;
+                                if (item.text === '«' || item.text === '<<') score -= 60;
+                            } else {
+                                if (signal.includes('下一月') || signal.includes('next-month') || signal.includes('nextmonth')) score += 100;
+                                if (signal.includes('chevron_right') || signal.includes('arrowright')) score += 85;
+                                if (item.text === '›' || item.text === '>') score += 80;
+                                if (item.text === '»' || item.text === '>>') score -= 60;
+                            }
+                            score -= Math.abs(item.rect.top - window.innerHeight * 0.18) / 100;
+                            return { ...item, score };
+                        })
+                        .sort((a, b) => b.score - a.score || (direction === 'prev' ? a.rect.left - b.rect.left : b.rect.left - a.rect.left));
+                    let target = candidates[0]?.el || null;
+                    if (!target) {
+                        const pickerItems = Array.from(document.querySelectorAll('button, span, div, a'))
+                            .filter(isVisible)
+                            .filter(inDatePicker)
+                            .map(el => ({ el, rect: el.getBoundingClientRect(), text: (el.innerText || el.textContent || '').trim() }))
+                            .filter(item => item.rect.width <= 44 && item.rect.height <= 44 && item.rect.top < window.innerHeight * 0.4);
+                        const sorted = pickerItems
+                            .sort((a, b) => {
+                                const at = Math.abs(a.rect.top - window.innerHeight * 0.18);
+                                const bt = Math.abs(b.rect.top - window.innerHeight * 0.18);
+                                return at - bt || a.rect.left - b.rect.left;
+                            })
+                            .slice(0, 8)
+                            .sort((a, b) => a.rect.left - b.rect.left);
+                        if (direction === 'prev') {
+                            target = (sorted[1] || sorted[0])?.el || null;
+                        } else {
+                            target = (sorted[sorted.length - 2] || sorted[sorted.length - 1])?.el || null;
+                        }
+                    }
+                    if (!target) return false;
+                    target.click();
+                    return true;
+                }""",
+                direction,
+            )
+            if clicked:
+                page.wait_for_timeout(800)
+            return bool(clicked)
+        except Exception as e:
+            print(f"[WARN] Calendar nav click failed ({direction}): {e}")
+            return False
+
+    if try_click_visible_date():
+        return True
+
+    for attempt in range(1, 5):
+        months = visible_month_keys()
+        direction = "prev"
+        if months:
+            if target_month_key < min(months):
+                direction = "prev"
+            elif target_month_key > max(months):
+                direction = "next"
+            else:
+                direction = "prev" if attempt % 2 == 1 else "next"
+        print(f"[INFO] Calendar date not visible; switching {direction} month for {date_dash} (attempt {attempt})")
+        if not click_calendar_nav(direction):
+            break
+        if try_click_visible_date():
+            return True
+
+    print(f"[WARN] Calendar date not selected after month switching: {date_dash}")
+    return False
+
+
+def _visible_calendar_month_keys(page) -> list[int]:
+    try:
+        return page.evaluate(
+            """() => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const keys = [];
+                for (const el of Array.from(document.querySelectorAll('div, span, button, th'))) {
+                    if (!isVisible(el)) continue;
+                    const text = (el.innerText || el.textContent || '').replace(/\\s+/g, ' ').trim();
+                    const match = text.match(/(20\\d{2})\\s*年\\s*(\\d{1,2})\\s*月/);
+                    if (match) keys.push(Number(match[1]) * 12 + Number(match[2]));
+                }
+                return Array.from(new Set(keys)).sort((a, b) => a - b);
+            }"""
+        ) or []
+    except Exception:
+        return []
+
+
+def _click_calendar_month_nav(page, direction: str) -> bool:
+    print(f"[INFO] Clicking calendar {direction} month arrow")
+    try:
+        clicked = page.evaluate(
+            """(direction) => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const inDatePicker = (el) => {
+                    let cur = el;
+                    for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                        const cls = String(cur.className || '').toLowerCase();
+                        const role = (cur.getAttribute && cur.getAttribute('role') || '').toLowerCase();
+                        if (cls.includes('picker') || cls.includes('calendar') || cls.includes('date')
+                            || role.includes('dialog') || role.includes('grid')) return true;
+                    }
+                    return false;
+                };
+                const headerCandidates = Array.from(document.querySelectorAll('div, span, button, th'))
+                    .filter(isVisible)
+                    .filter(inDatePicker)
+                    .map(el => ({ el, rect: el.getBoundingClientRect(), text: (el.innerText || el.textContent || '').replace(/\\s+/g, ' ').trim() }))
+                    .filter(item => /(20\\d{2})\\s*年\\s*(\\d{1,2})\\s*月/.test(item.text))
+                    .sort((a, b) => a.rect.top - b.rect.top || a.rect.left - b.rect.left);
+                const header = headerCandidates[0]?.rect || null;
+                const directSelector = direction === 'prev'
+                    ? '.ecom-picker-header-prev-btn:not(.ecom-picker-header-super-prev-btn), .ecom-picker-prev-icon'
+                    : '.ecom-picker-header-next-btn:not(.ecom-picker-header-super-next-btn), .ecom-picker-next-icon';
+                const direct = Array.from(document.querySelectorAll(directSelector))
+                    .filter(isVisible)
+                    .filter(inDatePicker)
+                    .map(el => {
+                        const btn = el.closest('button') || el;
+                        return { el: btn, rect: btn.getBoundingClientRect() };
+                    })
+                    .filter(item => item.rect.width > 0 && item.rect.height > 0)
+                    .sort((a, b) => {
+                        if (!header) return a.rect.left - b.rect.left;
+                        return Math.abs(a.rect.top - header.top) - Math.abs(b.rect.top - header.top);
+                    });
+                if (direct.length) {
+                    direct[0].el.click();
+                    return true;
+                }
+                const items = Array.from(document.querySelectorAll('button, span, div, a'))
+                    .filter(isVisible)
+                    .filter(inDatePicker)
+                    .map(el => {
+                        const rect = el.getBoundingClientRect();
+                        const text = (el.innerText || el.textContent || '').trim();
+                        const title = el.getAttribute('title') || '';
+                        const aria = el.getAttribute('aria-label') || '';
+                        const cls = String(el.className || '').toLowerCase();
+                        const signal = `${text} ${title} ${aria} ${cls}`;
+                        let score = 0;
+                        if (direction === 'prev') {
+                            if (signal.includes('上一月') || signal.includes('prev-month') || signal.includes('previous-month') || signal.includes('prevmonth')) score += 100;
+                            if (signal.includes('chevron_left') || signal.includes('arrowleft')) score += 80;
+                            if (text === '‹' || text === '<') score += 80;
+                            if (text === '«' || text === '<<') score -= 80;
+                            if (signal.includes('super-prev')) score -= 200;
+                        } else {
+                            if (signal.includes('下一月') || signal.includes('next-month') || signal.includes('nextmonth')) score += 100;
+                            if (signal.includes('chevron_right') || signal.includes('arrowright')) score += 80;
+                            if (text === '›' || text === '>') score += 80;
+                            if (text === '»' || text === '>>') score -= 80;
+                            if (signal.includes('super-next')) score -= 200;
+                        }
+                        if (header) {
+                            score -= Math.abs(rect.top - header.top) / 5;
+                            if (direction === 'prev' && rect.left < header.left) score += 40;
+                            if (direction === 'next' && rect.left > header.right) score += 40;
+                        }
+                        return { el, rect, score, text };
+                    })
+                    .filter(item => item.score > 0)
+                    .sort((a, b) => b.score - a.score || (direction === 'prev' ? b.rect.left - a.rect.left : a.rect.left - b.rect.left));
+                let target = items[0]?.el || null;
+                if (!target && header) {
+                    const smallNearHeader = Array.from(document.querySelectorAll('button, span, div, a'))
+                        .filter(isVisible)
+                        .filter(inDatePicker)
+                        .map(el => ({ el, rect: el.getBoundingClientRect() }))
+                        .filter(item => item.rect.width <= 42 && item.rect.height <= 42 && Math.abs(item.rect.top - header.top) < 36)
+                        .sort((a, b) => a.rect.left - b.rect.left);
+                    if (direction === 'prev') {
+                        target = smallNearHeader.filter(item => item.rect.left < header.left).slice(-1)[0]?.el || null;
+                    } else {
+                        target = smallNearHeader.find(item => item.rect.left > header.right)?.el || null;
+                    }
+                }
+                if (!target) return false;
+                target.click();
+                return true;
+            }""",
+            direction,
+        )
+        if clicked:
+            page.wait_for_timeout(800)
+        return bool(clicked)
+    except Exception as e:
+        print(f"[WARN] Calendar month nav failed ({direction}): {e}")
+        return False
+
+
+def _ensure_calendar_month(page, target_date) -> None:
+    target_month_key = target_date.year * 12 + target_date.month
+    for attempt in range(1, 4):
+        months = _visible_calendar_month_keys(page)
+        if not months or target_month_key in months:
+            return
+        direction = "prev" if target_month_key < min(months) else "next"
+        print(f"[INFO] Calendar month mismatch: visible={months}, target={target_month_key}; switching {direction} (attempt {attempt})")
+        if not _click_calendar_month_nav(page, direction):
+            return
+
+
+def _prepare_previous_week_calendar(page, week_start) -> None:
+    target_month_key = week_start.year * 12 + week_start.month
+    current_month_key = datetime.now().year * 12 + datetime.now().month
+    months = _visible_calendar_month_keys(page)
+    if months:
+        if target_month_key in months:
+            return
+        direction = "prev" if target_month_key < min(months) else "next"
+        _click_calendar_month_nav(page, direction)
+        return
+    if target_month_key < current_month_key:
+        _click_calendar_month_nav(page, "prev")
+    elif target_month_key > current_month_key:
+        _click_calendar_month_nav(page, "next")
+
+
+def _click_date_panel_quick_option(page, label: str) -> bool:
+    print(f"[INFO] Clicking date quick option: {label}")
+    try:
+        target = page.evaluate(
+            """(label) => {
+                const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const panelSignals = ['picker', 'calendar', 'date', 'popover', 'dropdown', 'overlay', 'tooltip'];
+                const inDatePanel = (el) => {
+                    let cur = el;
+                    for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                        const cls = String(cur.className || '').toLowerCase();
+                        const role = (cur.getAttribute && cur.getAttribute('role') || '').toLowerCase();
+                        if (panelSignals.some(signal => cls.includes(signal)) || role === 'dialog' || role === 'listbox') {
+                            return true;
+                        }
+                    }
+                    return false;
+                };
+                const candidates = Array.from(document.querySelectorAll('button, [role="button"], li, div, span'))
+                    .filter(isVisible)
+                    .filter(el => normalize(el.innerText || el.textContent) === label)
+                    .map(el => {
+                        let target = el;
+                        for (let cur = el; cur && cur !== document.body; cur = cur.parentElement) {
+                            const cls = String(cur.className || '').toLowerCase();
+                            const role = (cur.getAttribute && cur.getAttribute('role') || '').toLowerCase();
+                            const rect = cur.getBoundingClientRect();
+                            if ((role === 'menuitem' || cls.includes('menu') || cls.includes('item') || cls.includes('option'))
+                                && rect.width >= 40 && rect.height >= 24) {
+                                target = cur;
+                                break;
+                            }
+                        }
+                        const rect = target.getBoundingClientRect();
+                        return {
+                            el: target,
+                            rect,
+                            inPanel: inDatePanel(el),
+                            area: rect.width * rect.height,
+                        };
+                    })
+                    .sort((a, b) => Number(b.inPanel) - Number(a.inPanel) || a.rect.left - b.rect.left || a.rect.top - b.rect.top || a.area - b.area);
+                if (!candidates.length) return null;
+                const rect = candidates[0].rect;
+                return { x: rect.left + rect.width / 2, y: rect.top + rect.height / 2 };
+            }""",
+            label,
+        )
+        if target:
+            page.mouse.click(target["x"], target["y"])
+            page.wait_for_timeout(800)
+        return bool(target)
+    except Exception as e:
+        print(f"[WARN] Date quick option click failed ({label}): {e}")
+        return False
+
+
+def _open_business_date_picker(page) -> bool:
+    try:
+        clicked = page.evaluate(
+            """() => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const dateRangeRe = /\\d{4}\\/\\d{2}\\/\\d{2}\\s*-\\s*\\d{4}\\/\\d{2}\\/\\d{2}/;
+                const candidates = Array.from(document.querySelectorAll('button, [role="button"], div, span, input'))
+                    .filter(isVisible)
+                    .filter(el => {
+                        const text = normalize(el.innerText || el.textContent || el.value || el.getAttribute('placeholder'));
+                        return dateRangeRe.test(text) || text.includes('开始日期') || text.includes('结束日期');
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return false;
+                candidates[0].click();
+                return true;
+            }"""
+        )
+        if clicked:
+            return True
+    except Exception:
+        pass
+    try:
+        page.locator('input[placeholder*="开始日期"], input[placeholder*="结束日期"]').first.click(timeout=8000)
+        return True
+    except Exception:
+        return False
+
+
+def _select_business_previous_week(page, week_start, week_end) -> None:
+    print("[INFO] Selecting business natural week: " + _date_range_label(week_start, week_end))
+    if not _hover_visible_text(page, "更多", exact=True, timeout=8000):
+        _click_visible_text(page, "更多", exact=True, timeout=8000)
+    page.wait_for_timeout(800)
+
+    if not _click_date_panel_quick_option(page, "自然周") and not _click_visible_text(page, "自然周", exact=True, timeout=8000):
+        print("[WARN] Business natural week option not clicked")
+        return
+    page.wait_for_timeout(1000)
+
+    _prepare_previous_week_calendar(page, week_start)
+    if not _click_calendar_date(page, week_start) and not _click_calendar_date(page, week_end):
+        print(f"[WARN] Business natural week date not clicked: {week_start:%Y-%m-%d} - {week_end:%Y-%m-%d}")
+
+    _wait_after_action(page, "select business natural week", settle_ms=1800)
+
+
+def _hover_visible_text(page, text, exact=True, timeout=10000) -> bool:
+    print(f"[INFO] Hovering text: {text}")
+    try:
+        if adaptive_hover_text(page, text, exact=exact, timeout=timeout):
+            return True
+    except Exception as e:
+        print(f"[WARN] Adaptive hover text failed ({text}): {e}")
+    try:
+        page.get_by_text(text, exact=exact).first.hover(timeout=timeout)
+        return True
+    except Exception as e:
+        print(f"[WARN] Hover text failed ({text}): {e}")
+        return False
+
+
+def _click_visible_text(page, text, exact=True, timeout=10000) -> bool:
+    match_type = "exact text" if exact else "text"
+    print(f"[INFO] Clicking {match_type}: {text}")
+    try:
+        if adaptive_click_text(page, text, exact=exact, timeout=timeout):
+            return True
+    except Exception as e:
+        print(f"[WARN] Adaptive click text failed ({text}): {e}")
+
+    try:
+        locator = page.get_by_text(text, exact=exact)
+        locator.first.click(timeout=timeout)
+        return True
+    except Exception:
+        pass
+
+    try:
+        clicked = page.evaluate(
+            """([text, exact]) => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const candidates = Array.from(document.querySelectorAll('button, [role="button"], a, label, span, div, li'))
+                    .filter(isVisible)
+                    .filter(el => {
+                        const value = normalize(el.innerText || el.textContent);
+                        return exact ? value === text : value.includes(text);
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return false;
+                candidates[0].click();
+                return true;
+            }""",
+            [text, exact],
+        )
+        return bool(clicked)
+    except Exception as e:
+        print(f"[WARN] Click text failed ({text}): {e}")
+        return False
+
+
+def _click_market_category_dropdown(page) -> bool:
+    print("[INFO] Opening category dropdown...")
+    try:
+        clicked = page.evaluate(
+            """() => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const candidates = Array.from(document.querySelectorAll('button, [role="button"], div, span'))
+                    .filter(isVisible)
+                    .filter(el => {
+                        const text = (el.innerText || el.textContent || '').replace(/\\s+/g, '');
+                        return text.includes('鞋靴箱包') && text.includes('/');
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return false;
+                candidates[0].click();
+                return true;
+            }"""
+        )
+        return bool(clicked)
+    except Exception as e:
+        print(f"[WARN] Open category dropdown failed: {e}")
+        return False
+
+
+def _select_market_category(page, category_name: str) -> None:
+    try:
+        page.evaluate("window.scrollTo(0, 0)")
+    except Exception:
+        pass
+    page.wait_for_timeout(300)
+    if not _click_market_category_dropdown(page):
+        print("[WARN] Category dropdown not found")
+        return
+
+    page.wait_for_timeout(800)
+    for label in ["箱包", "时尚箱包", category_name]:
+        if not _click_visible_text(page, label, exact=True, timeout=5000):
+            print(f"[WARN] Category option not clicked: {label}")
+        page.wait_for_timeout(800)
+    _wait_after_action(page, f"select category {category_name}", settle_ms=1800)
+
+
+def _search_market_shop(page) -> None:
+    print(f"[INFO] Searching shop: {SHOP_NAME}")
+    try:
+        selector = 'input[placeholder*="店铺名"], input[placeholder*="搜索"], input'
+        if not adaptive_fill_input(page, "market_shop_search", selector, SHOP_NAME, timeout=10000):
+            locator = page.locator('input[placeholder*="店铺名"], input[placeholder*="搜索"]')
+            if locator.count() == 0:
+                locator = page.locator("input").first
+            else:
+                locator = locator.first
+            locator.click(timeout=10000)
+            locator.fill(SHOP_NAME)
+        page.keyboard.press("Enter")
+    except Exception as e:
+        print(f"[WARN] Shop search failed: {e}")
+        return
+    _wait_after_action(page, "shop search")
+    page.wait_for_timeout(2500)
+
+
+def _search_market_keyword(page, keyword: str) -> None:
+    print(f"[INFO] Searching market keyword: {keyword}")
+    try:
+        locator = page.locator(
+            'input[placeholder*="可搜索所选条件下的店铺名"], '
+            'input[placeholder*="所选条件"][placeholder*="店铺名"], '
+            'input[placeholder*="店铺名"][placeholder*="回车键搜索"]'
+        ).first
+        locator.wait_for(state="visible", timeout=5000)
+        locator.click(timeout=5000)
+        locator.fill(keyword, timeout=5000)
+        actual = locator.input_value(timeout=3000)
+        if keyword and keyword not in actual:
+            locator.click(timeout=3000)
+            page.keyboard.press("Control+A")
+            page.keyboard.type(keyword, delay=20)
+        elif not keyword and actual:
+            locator.fill("", timeout=3000)
+        page.keyboard.press("Enter")
+        print(f"[OK] Market search filled via locator: {keyword}")
+        _wait_after_action(page, f"keyword search {keyword}", settle_ms=1800)
+        return
+    except Exception as locator_error:
+        print(f"[WARN] Market search locator fill failed ({keyword}): {locator_error}")
+
+    try:
+        info = page.evaluate(
+            """() => {
+                const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const scoreInput = (input) => {
+                    const rect = input.getBoundingClientRect();
+                    const placeholder = normalize(input.placeholder || input.getAttribute('placeholder') || '');
+                    const text = `${placeholder} ${normalize(input.value || '')}`;
+                    let score = 0;
+                    if (placeholder.includes('可搜索所选条件下的店铺名')) score += 200;
+                    if (placeholder.includes('所选条件') && placeholder.includes('店铺')) score += 160;
+                    if (placeholder.includes('店铺名')) score += 120;
+                    if (placeholder.includes('按回车键搜索')) score += 80;
+                    if (placeholder.includes('搜索') && !placeholder.includes('智能')) score += 20;
+                    if (text.includes('AI') || text.includes('智能搜索')) score -= 220;
+                    if (rect.top < 130) score -= 180;
+                    if (rect.top >= 150 && rect.top <= 380) score += 80;
+                    if (rect.left < window.innerWidth * 0.45) score += 30;
+                    if (rect.width >= 220 && rect.width <= 620) score += 20;
+                    return { input, score, placeholder, rect };
+                };
+                const candidates = Array.from(document.querySelectorAll('input'))
+                    .filter(isVisible)
+                    .map(scoreInput)
+                    .filter(item => item.score > 0)
+                    .sort((a, b) => b.score - a.score || a.rect.top - b.rect.top);
+                const target = candidates[0];
+                if (!target) return { ok: false, reason: 'not_found' };
+                const input = target.input;
+                input.scrollIntoView({ block: 'center', inline: 'center' });
+                const freshRect = input.getBoundingClientRect();
+                return {
+                    ok: true,
+                    placeholder: target.placeholder,
+                    score: target.score,
+                    top: Math.round(freshRect.top),
+                    left: Math.round(freshRect.left),
+                    x: freshRect.left + freshRect.width / 2,
+                    y: freshRect.top + freshRect.height / 2,
+                };
+            }""",
+        )
+        if not info or not info.get("ok"):
+            print(f"[WARN] Keyword search input not found: {keyword} | {info}")
+        else:
+            print(
+                f"[OK] Market search input: placeholder={info.get('placeholder')} "
+                f"top={info.get('top')} left={info.get('left')} score={info.get('score')}"
+            )
+            page.mouse.click(info["x"], info["y"])
+            page.keyboard.press("Control+A")
+            if keyword:
+                page.keyboard.type(keyword, delay=20)
+            else:
+                page.keyboard.press("Backspace")
+            page.keyboard.press("Enter")
+    except Exception as e:
+        print(f"[WARN] Keyword search failed ({keyword}): {e}")
+    _wait_after_action(page, f"keyword search {keyword}", settle_ms=1800)
+
+
+def _extract_market_ranking(page, category_name: str) -> dict:
+    print(f"[INFO] Extracting market ranking: {category_name}")
+    result = page.evaluate(
+        """([shopName, categoryName]) => {
+            const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const headers = Array.from(document.querySelectorAll('table thead th'))
+                .map(th => clean(th.innerText || th.textContent))
+                .filter(Boolean);
+            const rows = Array.from(document.querySelectorAll('table tbody tr, table tr'));
+            const parseRank = (text, rankCell) => {
+                const normalized = clean(text);
+                const numbers = normalized.match(/\\d+/g) || [];
+                const rankMatch = numbers[0] || '';
+                const upMatch = normalized.match(/(?:↑|上升|升)\\s*(\\d+)/);
+                const downMatch = normalized.match(/(?:↓|下降|降)\\s*(\\d+)/);
+                let direction = '';
+                let change = '';
+                if (upMatch) {
+                    direction = '上升';
+                    change = upMatch[1];
+                } else if (downMatch) {
+                    direction = '下降';
+                    change = downMatch[1];
+                }
+                if (!change && numbers.length > 1) change = numbers[1];
+                if (!direction && rankCell) {
+                    const signals = [];
+                    const collect = (el) => {
+                        if (!el) return;
+                        const style = window.getComputedStyle(el);
+                        signals.push(String(el.className || '').toLowerCase());
+                        signals.push(String(el.getAttribute('style') || '').toLowerCase());
+                        signals.push(String(el.getAttribute('aria-label') || '').toLowerCase());
+                        signals.push(String(el.getAttribute('title') || '').toLowerCase());
+                        signals.push(style.color || '');
+                        Array.from(el.children || []).forEach(collect);
+                    };
+                    collect(rankCell);
+                    const signal = signals.join(' ');
+                    if (signal.includes('up') || signal.includes('rise') || signal.includes('increase')
+                        || signal.includes('red') || signal.includes('rgb(255') || signal.includes('rgb(245')
+                        || signal.includes('rgb(238') || signal.includes('rgb(239')) {
+                        direction = '上升';
+                    } else if (signal.includes('down') || signal.includes('fall') || signal.includes('decrease')
+                        || signal.includes('green') || signal.includes('rgb(0') || signal.includes('rgb(34')
+                        || signal.includes('rgb(52') || signal.includes('rgb(76') || signal.includes('rgb(82')) {
+                        direction = '下降';
+                    }
+                }
+                return {
+                    rank: rankMatch,
+                    rank_change_direction: direction,
+                    rank_change_value: change,
+                    rank_change: direction && change ? `${direction}${change}` : ''
+                };
+            };
+
+            for (const row of rows) {
+                const rowText = clean(row.innerText || row.textContent);
+                if (!rowText.includes('我的店铺') && !rowText.includes(shopName)) continue;
+
+                const cells = Array.from(row.querySelectorAll('td'))
+                    .map(td => clean(td.innerText || td.textContent));
+                const mapped = {};
+                if (headers.length && cells.length) {
+                    cells.forEach((cell, index) => {
+                        mapped[headers[index] || `列${index + 1}`] = cell;
+                    });
+                }
+                return {
+                    category: categoryName,
+                    row_text: rowText,
+                    columns: mapped,
+                    ...parseRank(cells[0] || rowText, row.querySelector('td')),
+                    captured_at: new Date().toISOString()
+                };
+            }
+
+            const lines = clean(document.body.innerText || '').split(' ').filter(Boolean);
+            const idx = lines.findIndex(line => line.includes('我的店铺') || line.includes(shopName));
+            const rowText = idx >= 0 ? lines.slice(Math.max(0, idx - 4), idx + 20).join(' ') : '';
+            return {
+                category: categoryName,
+                row_text: rowText,
+                columns: {},
+                ...parseRank(rowText, null),
+                captured_at: new Date().toISOString()
+            };
+        }""",
+        [SHOP_NAME, category_name],
+    )
+    print(f"[OK] Market ranking ({category_name}): {result.get('row_text', '')[:160]}")
+    return result
+
+
+def _extract_competitor_rows(page, brand: str, category_name: str) -> list:
+    print(f"[INFO] Extracting competitor rows: {brand} / {category_name}")
+    rows = page.evaluate(
+        """([brand, categoryName]) => {
+            const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const headers = Array.from(document.querySelectorAll('table thead th'))
+                .map(th => clean(th.innerText || th.textContent))
+                .filter(Boolean);
+            const trs = Array.from(document.querySelectorAll('table tbody tr, table tr'));
+            const result = [];
+
+            for (const row of trs) {
+                const rowText = clean(row.innerText || row.textContent);
+                if (!rowText || rowText.includes('我的店铺')) continue;
+
+                const cells = Array.from(row.querySelectorAll('td'))
+                    .map(td => clean(td.innerText || td.textContent));
+                if (cells.length < 4) continue;
+                const searchIndex = (cells[0].match(/\\d+/) || [''])[0];
+
+                const columns = {};
+                cells.forEach((cell, index) => {
+                    columns[headers[index] || `列${index + 1}`] = cell;
+                });
+                const shopName = cells[1] || '';
+                result.push({
+                    brand,
+                    category: categoryName,
+                    rank: '',
+                    search_index: searchIndex,
+                    shop_name: shopName,
+                    row_text: rowText,
+                    columns,
+                    captured_at: new Date().toISOString()
+                });
+            }
+            return result;
+        }""",
+        [brand, category_name],
+    )
+    print(f"[OK] Competitor rows ({brand} / {category_name}): {len(rows)}")
+    return rows
+
+
+def _extract_visible_rank_rows(page, category_name: str) -> list:
+    rows = page.evaluate(
+        """(categoryName) => {
+            const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const headers = Array.from(document.querySelectorAll('table thead th'))
+                .map(th => clean(th.innerText || th.textContent))
+                .filter(Boolean);
+            const result = [];
+            for (const row of Array.from(document.querySelectorAll('table tbody tr, table tr'))) {
+                const rowText = clean(row.innerText || row.textContent);
+                if (!rowText || rowText.includes('我的店铺')) continue;
+                const cells = Array.from(row.querySelectorAll('td')).map(td => clean(td.innerText || td.textContent));
+                if (cells.length < 4) continue;
+                const rank = (cells[0].match(/\\d+/) || [''])[0];
+                if (!rank) continue;
+                const columns = {};
+                cells.forEach((cell, index) => {
+                    columns[headers[index] || `列${index + 1}`] = cell;
+                });
+                result.push({
+                    category: categoryName,
+                    rank,
+                    shop_name: cells[1] || '',
+                    row_text: rowText,
+                    columns,
+                    captured_at: new Date().toISOString()
+                });
+            }
+            return result;
+        }""",
+        category_name,
+    )
+    return rows
+
+
+def _click_market_page_number(page, page_number: int) -> bool:
+    print(f"[INFO] Opening market ranking page number: {page_number}")
+    try:
+        clicked = page.evaluate(
+            """(pageNumber) => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const target = String(pageNumber);
+                const candidates = Array.from(document.querySelectorAll('button, li, a, span, div'))
+                    .filter(isVisible)
+                    .filter(el => (el.innerText || el.textContent || '').trim() === target)
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (br.top - ar.top) || ((ar.width * ar.height) - (br.width * br.height));
+                    });
+                if (!candidates.length) return false;
+                (candidates[0].closest('button, li, a') || candidates[0]).click();
+                return true;
+            }""",
+            page_number,
+        )
+        if clicked:
+            _wait_after_action(page, f"open page {page_number}", settle_ms=1000)
+        return bool(clicked)
+    except Exception as e:
+        print(f"[WARN] Page number click failed ({page_number}): {e}")
+        return False
+
+
+def _normalize_shop_name(name: str) -> str:
+    return re.sub(r"\s+", "", name or "").lower()
+
+
+def _find_competitor_ranks_in_category(
+    page,
+    category_name: str,
+    shop_names: list,
+    max_pages=10,
+    week_start=None,
+    week_end=None,
+) -> dict:
+    target_names = {_normalize_shop_name(name): name for name in shop_names if name}
+    found = {}
+    if not target_names:
+        return found
+
+    print(f"[INFO] Scanning ranking pages for {len(target_names)} shops / {category_name}")
+    _search_market_keyword(page, "")
+    if week_start and week_end:
+        _select_market_previous_week(page, week_start, week_end)
+    if not _click_market_page_number(page, 1):
+        _wait_after_action(page, "reset market search", settle_ms=1000)
+
+    for page_number in range(1, max_pages + 1):
+        if page_number > 1 and not _click_market_page_number(page, page_number):
+            print(f"[WARN] Market page {page_number} not available; stop scanning")
+            break
+
+        visible_rows = _extract_visible_rank_rows(page, category_name)
+        for row in visible_rows:
+            row_name = _normalize_shop_name(row.get("shop_name", ""))
+            for target_norm, original_name in target_names.items():
+                if target_norm in found:
+                    continue
+                if row_name == target_norm or target_norm in row_name or row_name in target_norm:
+                    found[target_norm] = {
+                        "shop_name": original_name,
+                        "rank": row.get("rank", ""),
+                        "rank_row": row,
+                        "page_number": page_number,
+                    }
+        if len(found) == len(target_names):
+            break
+
+    print(f"[OK] Ranking scan found {len(found)} / {len(target_names)} shops in {category_name}")
+    return found
+
+
+def _column_value(columns: dict, *needles: str) -> str:
+    for key, value in columns.items():
+        if all(needle in key for needle in needles):
+            return value
+    return ""
+
+
+def _parse_market_number_range(value: str) -> dict:
+    raw = (value or "").strip()
+    if not raw:
+        return {"raw": raw, "min": None, "max": None, "open_ended": False}
+
+    text = raw.replace(",", "").replace("￥", "¥").replace("－", "-").replace("—", "-").replace("–", "-")
+    open_ended = text.endswith("+")
+    text = text.rstrip("+")
+
+    def parse_part(part: str):
+        part = part.strip().replace("¥", "")
+        if not part:
+            return None
+        multiplier = 10000 if "万" in part else 1
+        part = part.replace("万", "")
+        match = re.search(r"-?\d+(?:\.\d+)?", part)
+        return float(match.group(0)) * multiplier if match else None
+
+    parts = [p for p in text.split("-") if p.strip()]
+    nums = [parse_part(part) for part in parts]
+    nums = [n for n in nums if n is not None]
+    if not nums:
+        return {"raw": raw, "min": None, "max": None, "open_ended": open_ended}
+    if len(nums) == 1:
+        return {"raw": raw, "min": nums[0], "max": None if open_ended else nums[0], "open_ended": open_ended}
+    return {"raw": raw, "min": nums[0], "max": nums[-1], "open_ended": open_ended}
+
+
+def _format_market_number(value, money=False) -> str:
+    if value is None:
+        return ""
+    prefix = "¥" if money else ""
+    if abs(value) >= 10000:
+        amount = value / 10000
+        if amount.is_integer():
+            return f"{prefix}{int(amount):,}万"
+        return f"{prefix}{amount:,.2f}".rstrip("0").rstrip(".") + "万"
+    if float(value).is_integer():
+        return f"{prefix}{int(value):,}"
+    return f"{prefix}{value:,.2f}".rstrip("0").rstrip(".")
+
+
+def _format_market_range(summary: dict, money=False) -> str:
+    min_value = summary.get("min")
+    max_value = summary.get("max")
+    if min_value is None:
+        return "未获取"
+    if summary.get("open_ended") or max_value is None:
+        return f"{_format_market_number(min_value, money=money)}+"
+    if min_value == max_value:
+        return _format_market_number(min_value, money=money)
+    return f"{_format_market_number(min_value, money=money)}-{_format_market_number(max_value, money=money)}"
+
+
+def _sum_market_ranges(values: list) -> dict:
+    parsed = [_parse_market_number_range(value) for value in values if value]
+    parsed = [item for item in parsed if item.get("min") is not None]
+    if not parsed:
+        return {"min": None, "max": None, "open_ended": False, "raw_values": values}
+    min_sum = sum(item["min"] for item in parsed)
+    open_ended = any(item.get("open_ended") or item.get("max") is None for item in parsed)
+    max_sum = None if open_ended else sum(item["max"] for item in parsed)
+    return {"min": min_sum, "max": max_sum, "open_ended": open_ended, "raw_values": values}
+
+
+def _summarize_competitor_rows(brand: str, category_name: str, rows: list) -> dict:
+    ranked_rows = [row for row in rows if str(row.get("rank", "")).isdigit()]
+    best_rank = min((int(row["rank"]) for row in ranked_rows), default=None)
+
+    payment_values = []
+    order_values = []
+    click_values = []
+    stores = []
+    for row in rows:
+        columns = row.get("columns", {})
+        payment_values.append(_column_value(columns, "用户", "支付", "金额"))
+        order_values.append(_column_value(columns, "成交", "订单"))
+        click_values.append(_column_value(columns, "商品", "点击", "人数"))
+        stores.append({
+            "shop_name": row.get("shop_name", ""),
+            "rank": row.get("rank", ""),
+            "search_index": row.get("search_index", ""),
+            "用户支付金额": payment_values[-1],
+            "成交订单数": order_values[-1],
+            "商品点击人数": click_values[-1],
+            "row_text": row.get("row_text", ""),
+        })
+
+    payment_summary = _sum_market_ranges(payment_values)
+    order_summary = _sum_market_ranges(order_values)
+    click_summary = _sum_market_ranges(click_values)
+    return {
+        "brand": brand,
+        "category": category_name,
+        "matched_shop_count": len(rows),
+        "rank": str(best_rank) if best_rank is not None else "",
+        "rank_display": str(best_rank) if best_rank is not None else "未进入前200名",
+        "用户支付金额": _format_market_range(payment_summary, money=True),
+        "成交订单数": _format_market_range(order_summary),
+        "商品点击人数": _format_market_range(click_summary),
+        "payment_amount": payment_summary,
+        "order_count": order_summary,
+        "product_click_users": click_summary,
+        "stores": stores,
+        "captured_at": datetime.now().isoformat(),
+    }
+
+
+def _select_market_previous_week(page, week_start, week_end) -> None:
+    print(f"[INFO] Selecting market natural week: {_date_range_label(week_start, week_end)}")
+    if not _hover_visible_text(page, "更多", exact=True, timeout=8000):
+        _click_visible_text(page, "更多", exact=True, timeout=8000)
+    page.wait_for_timeout(800)
+    if not _click_date_panel_quick_option(page, "自然周") and not _click_visible_text(page, "自然周", exact=True, timeout=8000):
+        print("[WARN] Natural week option not clicked")
+        return
+    page.wait_for_timeout(1000)
+
+    # 自然周模式下点击周内任意一天即可选中整周，使用上周周一。
+    _prepare_previous_week_calendar(page, week_start)
+    if not _click_calendar_date(page, week_start) and not _click_calendar_date(page, week_end):
+        print(f"[WARN] Market natural week date not clicked: {week_start:%Y-%m-%d}")
+    _wait_after_action(page, "select market natural week", settle_ms=1800)
+    _verify_market_previous_week_selected(page, week_start, week_end)
+    try:
+        page.keyboard.press("Escape")
+        page.mouse.click(120, 120)
+    except Exception:
+        pass
+    page.wait_for_timeout(500)
+
+
+def _verify_market_previous_week_selected(page, week_start, week_end) -> bool:
+    expected = _date_range_label(week_start, week_end)
+    try:
+        ranges = page.evaluate(
+            """() => Array.from(document.body.innerText.matchAll(/20\\d{2}\\/\\d{2}\\/\\d{2}\\s*-\\s*20\\d{2}\\/\\d{2}\\/\\d{2}/g))
+                .map(match => match[0].replace(/\\s+/g, ' '))
+                .slice(0, 10)"""
+        )
+    except Exception:
+        ranges = []
+    ok = expected in ranges
+    if ok:
+        print(f"[OK] Market date selected: {expected}")
+    else:
+        print(f"[WARN] Market date not confirmed as {expected}; visible ranges: {ranges}")
+    return ok
+
+
+def _open_market_rank_shop_page(page, label="market ranking") -> None:
+    target_url = "https://compass.jinritemai.com/shop/chance/rank-shop"
+    print(f"[INFO] Navigating to {label}: {target_url}")
+    page.goto(target_url, wait_until="domcontentloaded", timeout=60000)
+    try:
+        page.wait_for_load_state("load", timeout=8000)
+    except Exception as e:
+        print(f"[WARN] Market rank page load wait skipped: {e}")
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+    print(f"[OK] Market rank page loaded: {page.url}")
+
+
+def _collect_market_rankings(page) -> dict:
+    week_start, week_end = _previous_week_range()
+    print("[INFO] Opening market ranking page...")
+    _open_market_rank_shop_page(page, "market shop ranking")
+
+    _select_market_previous_week(page, week_start, week_end)
+
+    rankings = {}
+    for category_name in MARKET_CATEGORIES:
+        _select_market_category(page, category_name)
+        rankings[category_name] = _extract_market_ranking(page, category_name)
+
+    return rankings
+
+
+def _collect_competitor_rankings(page) -> dict:
+    print("[INFO] Collecting competitor rankings...")
+    week_start, week_end = _previous_week_range()
+    print("[INFO] Re-opening market ranking page for competitors...")
+    _open_market_rank_shop_page(page, "competitor market shop ranking")
+    _select_market_previous_week(page, week_start, week_end)
+
+    data = {
+        "date_range": _date_range_label(*_previous_week_range()),
+        "categories": MARKET_CATEGORIES,
+        "brands": COMPETITOR_BRANDS,
+        "captured_at": datetime.now().isoformat(),
+        "url": page.url,
+        "results": {},
+    }
+
+    for category_name in MARKET_CATEGORIES:
+        if not _verify_market_previous_week_selected(page, week_start, week_end):
+            _select_market_previous_week(page, week_start, week_end)
+        _select_market_category(page, category_name)
+        if not _verify_market_previous_week_selected(page, week_start, week_end):
+            _select_market_previous_week(page, week_start, week_end)
+        category_results = {}
+        for brand in COMPETITOR_BRANDS:
+            _search_market_keyword(page, brand)
+            rows = _extract_competitor_rows(page, brand, category_name)
+            category_results[brand] = _summarize_competitor_rows(brand, category_name, rows)
+
+        shop_names = []
+        for item in category_results.values():
+            for store in item.get("stores", []):
+                if store.get("shop_name"):
+                    shop_names.append(store["shop_name"])
+
+        rank_map = _find_competitor_ranks_in_category(
+            page,
+            category_name,
+            shop_names,
+            week_start=week_start,
+            week_end=week_end,
+        )
+        for item in category_results.values():
+            best_rank = None
+            for store in item.get("stores", []):
+                key = _normalize_shop_name(store.get("shop_name", ""))
+                rank_info = rank_map.get(key, {})
+                if rank_info.get("rank"):
+                    store["rank"] = rank_info["rank"]
+                    store["rank_page_number"] = rank_info.get("page_number")
+                    rank_value = int(rank_info["rank"])
+                    if best_rank is None or rank_value < best_rank:
+                        best_rank = rank_value
+            item["rank"] = str(best_rank) if best_rank is not None else ""
+            item["rank_display"] = item["rank"] or "未进入前200名"
+        data["results"][category_name] = category_results
+
+    return data
+
+
+def _select_qianchuan_previous_week(page) -> None:
+    target_url = "https://qianchuan.jinritemai.com/home?aavid=1813687011565580"
+    week_start, week_end = _previous_week_range()
+    print(f"[INFO] Navigating to Qianchuan home: {target_url}")
+    print(f"[INFO] Qianchuan date range: {_date_range_label(week_start, week_end)}")
+    page.goto(target_url, wait_until="domcontentloaded", timeout=60000)
+    try:
+        page.wait_for_load_state("load", timeout=8000)
+    except Exception as e:
+        print(f"[WARN] Qianchuan load wait skipped: {e}")
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+    print(f"[OK] Qianchuan page loaded: {page.url}")
+
+    qianchuan_week_ok = False
+    if _open_qianchuan_date_picker(page):
+        page.wait_for_timeout(800)
+        qianchuan_week_ok = _click_date_panel_quick_option(page, "上周")
+    if not qianchuan_week_ok:
+        print("[WARN] Qianchuan date panel quick option not clicked; trying direct text click")
+        qianchuan_week_ok = _click_visible_text(page, "上周", exact=True, timeout=8000)
+    if not qianchuan_week_ok:
+        print("[WARN] Qianchuan previous-week option still not clicked")
+
+    _wait_after_action(page, "select qianchuan previous week")
+    try:
+        page.wait_for_function(
+            "() => (document.body.innerText || '').includes('整体消耗')"
+            " && (document.body.innerText || '').includes('净成交ROI')",
+            timeout=8000,
+        )
+    except Exception as e:
+        print(f"[WARN] Qianchuan metric text wait skipped: {e}")
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+
+
+def _open_qianchuan_date_picker(page) -> bool:
+    print("[INFO] Opening Qianchuan date picker...")
+    try:
+        clicked = page.evaluate(
+            """() => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const rangeRe = /\\d{4}[-/]\\d{2}[-/]\\d{2}\\s*~\\s*\\d{4}[-/]\\d{2}[-/]\\d{2}/;
+                const candidates = Array.from(document.querySelectorAll('button, [role="button"], div, span, input'))
+                    .filter(isVisible)
+                    .filter(el => {
+                        const text = normalize(el.innerText || el.textContent || el.value || '');
+                        return text.includes('统计周期') || rangeRe.test(text);
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return false;
+                let target = candidates[0];
+                for (let cur = target; cur && cur !== document.body; cur = cur.parentElement) {
+                    const text = normalize(cur.innerText || cur.textContent || cur.value || '');
+                    const role = (cur.getAttribute && cur.getAttribute('role')) || '';
+                    if (cur.tagName === 'BUTTON' || role === 'button' || rangeRe.test(text)) {
+                        target = cur;
+                        break;
+                    }
+                }
+                target.click();
+                return true;
+            }"""
+        )
+        return bool(clicked)
+    except Exception as e:
+        print(f"[WARN] Open Qianchuan date picker failed: {e}")
+        return False
+
+
+def _extract_qianchuan_metrics(page) -> dict:
+    print("[INFO] Extracting Qianchuan metrics...")
+    metrics = page.evaluate(
+        """() => {
+            const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const isVisible = (el) => {
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style && style.visibility !== 'hidden' && style.display !== 'none'
+                    && rect.width > 0 && rect.height > 0;
+            };
+            const labels = [
+                { key: '整体消耗', candidates: ['整体消耗(元)', '整体消耗（元）'] },
+                { key: '净成交ROI', candidates: ['净成交ROI', '净成本ROI'] },
+            ];
+            const valueRe = /^[-+]?\\d{1,3}(,\\d{3})*(\\.\\d+)?$|^[-+]?\\d+(\\.\\d+)?$/;
+            const pctRe = /^[-+]?\\d+(\\.\\d+)?%$/;
+
+            function directionFromPercent(percent) {
+                if (!percent || percent === '--') return '';
+                const n = Number(String(percent).replace('%', ''));
+                if (Number.isNaN(n) || n === 0) return '持平';
+                return n > 0 ? '上升' : '下降';
+            }
+
+            function joinDecimalTail(lines, index) {
+                const value = lines[index] || '';
+                const tail = lines[index + 1] || '';
+                return /^\\.\\d+$/.test(tail) ? value + tail : value;
+            }
+
+            function parseFromText(text, labelText) {
+                const lines = normalize(text).split(' ').filter(Boolean);
+                const idx = lines.findIndex(line => line === labelText || line.includes(labelText));
+                if (idx < 0) return null;
+                let value = '';
+                let percent = '';
+                for (let i = idx + 1; i < Math.min(lines.length, idx + 8); i++) {
+                    if (!value && valueRe.test(lines[i])) {
+                        value = joinDecimalTail(lines, i);
+                        continue;
+                    }
+                    if (!percent && (pctRe.test(lines[i]) || lines[i] === '--')) {
+                        percent = lines[i];
+                    }
+                }
+                if (!value && !percent) return null;
+                return {
+                    value,
+                    change_percent: percent,
+                    change_direction: directionFromPercent(percent),
+                    raw_text: normalize(text)
+                };
+            }
+
+            function smallestUsefulContainer(labelEl, labelText) {
+                let cur = labelEl;
+                for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                    const text = normalize(cur.innerText || cur.textContent);
+                    if (text.includes(labelText) && text.length <= 180) {
+                        const parsed = parseFromText(text, labelText);
+                        if (parsed && parsed.value) return parsed;
+                    }
+                }
+                return null;
+            }
+
+            const result = {};
+            const visibleEls = Array.from(document.querySelectorAll('div, span, p, label'))
+                .filter(isVisible);
+
+            for (const target of labels) {
+                for (const labelText of target.candidates) {
+                    const labelEl = visibleEls.find(el => normalize(el.innerText || el.textContent) === labelText);
+                    if (labelEl) {
+                        const parsed = smallestUsefulContainer(labelEl, labelText);
+                        if (parsed) {
+                            result[target.key] = parsed;
+                            break;
+                        }
+                    }
+                }
+                if (result[target.key]) continue;
+
+                const bodyLines = (document.body.innerText || '').split('\\n').map(normalize).filter(Boolean);
+                for (const labelText of target.candidates) {
+                    const idx = bodyLines.findIndex(line => line === labelText);
+                    if (idx < 0) continue;
+                    let value = '';
+                    let percent = '';
+                    for (let i = idx + 1; i < Math.min(bodyLines.length, idx + 8); i++) {
+                        if (!value && valueRe.test(bodyLines[i])) value = joinDecimalTail(bodyLines, i);
+                        if (!percent && (pctRe.test(bodyLines[i]) || bodyLines[i] === '--')) percent = bodyLines[i];
+                    }
+                    if (value || percent) {
+                        result[target.key] = {
+                            value,
+                            change_percent: percent,
+                            change_direction: directionFromPercent(percent),
+                            raw_text: bodyLines.slice(idx, idx + 8).join(' ')
+                        };
+                        break;
+                    }
+                }
+            }
+
+            return result;
+        }"""
+    )
+    cleaned = {k: v for k, v in metrics.items() if v}
+    print(f"[OK] Qianchuan extract: {cleaned}")
+    return cleaned
+
+
+def _collect_qianchuan_metrics(page) -> dict:
+    _select_qianchuan_previous_week(page)
+    metrics = _extract_qianchuan_metrics(page)
+    metrics["captured_at"] = datetime.now().isoformat()
+    metrics["url"] = page.url
+    metrics["date_range"] = _date_range_label(*_previous_week_range())
+    return metrics
+
+
+def _extract_experience_score(page) -> dict:
+    print("[INFO] Extracting experience score...")
+    score = page.evaluate(
+        """() => {
+            const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const isVisible = (el) => {
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style && style.visibility !== 'hidden' && style.display !== 'none'
+                    && rect.width > 0 && rect.height > 0;
+            };
+            const scoreRe = /^\\d+(\\.\\d+)?分?$/;
+
+            function parseScoreFromText(text, labelText) {
+                const lines = normalize(text).split(' ').filter(Boolean);
+                const idx = lines.findIndex(line => line === labelText || line.includes(labelText));
+                if (idx < 0) return '';
+                for (let i = idx + 1; i < Math.min(lines.length, idx + 8); i++) {
+                    const value = lines[i].replace(/分$/, '');
+                    if (scoreRe.test(lines[i]) || /^\\d+(\\.\\d+)?$/.test(value)) return value;
+                }
+                const compact = normalize(text).replace(/\\s+/g, '');
+                const labelIndex = compact.indexOf(labelText);
+                if (labelIndex < 0) return '';
+                const afterLabel = compact.slice(labelIndex + labelText.length);
+                const match = afterLabel.match(/^(\\d+(?:\\.\\d+)?)分?/);
+                return match ? match[1] : '';
+            }
+
+            function findByLabel(labelText) {
+                const visibleEls = Array.from(document.querySelectorAll('div, span, p, label'))
+                    .filter(isVisible);
+                const labelEl = visibleEls.find(el => normalize(el.innerText || el.textContent).includes(labelText));
+                if (labelEl) {
+                    let cur = labelEl;
+                    for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                        const text = normalize(cur.innerText || cur.textContent);
+                        if (text.includes(labelText) && text.length <= 220) {
+                            const score = parseScoreFromText(text, labelText);
+                            if (score) return { score, raw_text: text };
+                        }
+                    }
+                }
+
+                const bodyLines = (document.body.innerText || '').split('\\n').map(normalize).filter(Boolean);
+                const idx = bodyLines.findIndex(line => line.includes(labelText));
+                if (idx >= 0) {
+                    const chunk = bodyLines.slice(idx, idx + 8).join(' ');
+                    const score = parseScoreFromText(chunk, labelText);
+                    if (score) return { score, raw_text: chunk };
+                }
+                return null;
+            }
+
+            const result = {};
+            const main = findByLabel('我的体验分') || findByLabel('体检分') || findByLabel('商家体验分');
+            if (main) {
+                result.score = main.score;
+                result.raw_text = main.raw_text;
+            }
+
+            const detailLabels = [
+                [['商品体验得分', '商品体验分'], 'product_experience_score'],
+                [['物流体验得分', '物流体验分'], 'logistics_experience_score'],
+                [['服务体验得分', '服务体验分'], 'service_experience_score'],
+                [['差行为扣分'], 'penalty_score']
+            ];
+            for (const [labels, key] of detailLabels) {
+                let parsed = null;
+                for (const label of labels) {
+                    parsed = findByLabel(label);
+                    if (parsed) break;
+                }
+                if (parsed) {
+                    result[key] = parsed.score;
+                    result[`${key}_raw_text`] = parsed.raw_text;
+                }
+            }
+            return result;
+        }"""
+    )
+    print(f"[OK] Experience score extract: {score}")
+    return score or {}
+
+
+def _collect_experience_score(page) -> dict:
+    target_url = (
+        "https://fxg.jinritemai.com/ffa/eco/experience-score"
+        "?btm_ppre=a0.b0.c0.d0&btm_pre=a2427.b76571.c902327.d871297"
+        "&btm_show_id=ed5a73dc-4e3f-4f2f-bd62-f31a0a8b496c"
+    )
+    print(f"[INFO] Navigating to experience score page: {target_url}")
+    page.goto(target_url, wait_until="domcontentloaded", timeout=60000)
+    try:
+        page.wait_for_load_state("load", timeout=8000)
+    except Exception as e:
+        print(f"[WARN] Experience score load wait skipped: {e}")
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+    result = _extract_experience_score(page)
+    result["captured_at"] = datetime.now().isoformat()
+    result["url"] = page.url
+    return result
+
+
+def _keep_browser_open(page) -> None:
+    if not KEEP_BROWSER_OPEN:
+        return
+
+    print("\n[INFO] Browser will stay open. Close the browser window or press Ctrl+C to exit.")
+    try:
+        while not page.is_closed():
+            page.wait_for_timeout(1000)
+    except KeyboardInterrupt:
+        print("\n[INFO] Exit requested by user.")
+    except Exception as e:
+        print(f"[INFO] Browser keep-open ended: {e}")
+
+
+def _page_automation(page) -> None:
+    global _scraped_metrics, _market_rankings, _competitor_rankings, _qianchuan_metrics, _experience_score, _results_saved
+
+    _maximize_page(page)
+    print(f"[DEBUG] Current URL: {page.url}")
+
+    # 1. Check and wait for login
+    if not _check_login_status(page):
+        logged_page = _ensure_login(page)
+        if not logged_page:
+            print("[ERR] Login failed, will try save cookies anyway...")
+        else:
+            page = logged_page
+            _maximize_page(page)
+    
+    # 2. Save state after login (best effort)
+    _save_browser_state(page, "after login")
+
+    # 3. Navigate
+    _select_menu_and_date(page)
+    _save_browser_state(page, "after business page")
+
+    # 4. Extract data
+    data = _extract_metrics_via_js(page)
+    expected_keys = {'成交金额', '成交人数', '客单价', '成交退款金额', '商品点击人数'}
+    missing = expected_keys - set(data.keys())
+    if missing:
+        print(f"[WARN] Missing: {missing}, trying fallback...")
+        data2 = _extract_metrics_via_dom_fallback(page)
+        data = {**data2, **data}
+
+    if '成交退款金额（支付时间）' in data and '成交退款金额' not in data:
+        data['成交退款金额'] = data.pop('成交退款金额（支付时间）')
+
+    _scraped_metrics = data
+
+    # 5. Screenshot
+    try:
+        os.makedirs(STORE_DATA_DIR, exist_ok=True)
+        screenshot_path = os.path.join(STORE_DATA_DIR, f"compass_screenshot_{datetime.now().strftime('%Y%m%d_%H%M%S')}.png")
+        page.screenshot(path=screenshot_path, full_page=True)
+        print(f"[OK] Screenshot: {screenshot_path}")
+    except Exception as e:
+        print(f"[WARN] Screenshot failed: {e}")
+
+    # 6. Market shop ranking data
+    try:
+        _market_rankings = _collect_market_rankings(page)
+    except Exception as e:
+        print(f"[WARN] Market rankings collection failed: {e}")
+        _market_rankings = {}
+
+    # 7. Qianchuan previous-week metrics
+    try:
+        _qianchuan_metrics = _collect_qianchuan_metrics(page)
+    except Exception as e:
+        print(f"[WARN] Qianchuan metrics collection failed: {e}")
+        _qianchuan_metrics = {}
+
+    # 8. Shop experience score
+    try:
+        _experience_score = _collect_experience_score(page)
+    except Exception as e:
+        print(f"[WARN] Experience score collection failed: {e}")
+        _experience_score = {}
+
+    # 9. Competitor shop ranking data
+    try:
+        _competitor_rankings = _collect_competitor_rankings(page)
+        save_competitor_results(_competitor_rankings)
+    except Exception as e:
+        print(f"[WARN] Competitor rankings collection failed: {e}")
+        _competitor_rankings = {}
+
+    final_data = {
+        "shop_name": SHOP_NAME,
+        "date_range": _date_range_label(*_previous_week_range()),
+        "scraped_at": datetime.now().isoformat(),
+        "url": page.url,
+        "metrics": _scraped_metrics,
+        "market_rankings": _market_rankings,
+        "qianchuan_metrics": _qianchuan_metrics,
+        "experience_score": _experience_score,
+    }
+    save_results(final_data)
+    _results_saved = True
+
+    _keep_browser_open(page)
+
+
+def save_competitor_results(data: dict) -> None:
+    date_stamp = datetime.now().strftime("%Y%m%d")
+    base_name = f"竞店数据汇总_{date_stamp}"
+
+    os.makedirs(COMPETITOR_DATA_DIR, exist_ok=True)
+
+    json_path = os.path.join(COMPETITOR_DATA_DIR, f"{base_name}.json")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(data, f, ensure_ascii=False, indent=2)
+    print(f"[OK] Competitor JSON: {json_path}")
+
+    md_path = os.path.join(COMPETITOR_DATA_DIR, f"{base_name}.md")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("# 竞店数据汇总\n\n")
+        f.write(f"- **周期**: {data.get('date_range', '')}\n")
+        f.write(f"- **采集时间**: {data.get('captured_at', '')}\n")
+        f.write(f"- **品牌**: {', '.join(data.get('brands', []))}\n\n")
+
+        results = data.get("results", {})
+        for category in data.get("categories", []):
+            f.write(f"## {category}\n\n")
+            f.write("| 品牌 | 排名 | 用户支付金额 | 成交订单数 | 商品点击人数 | 匹配店铺数 |\n")
+            f.write("|------|------|------|------|------|------|\n")
+            category_results = results.get(category, {})
+            for brand in data.get("brands", []):
+                item = category_results.get(brand, {})
+                rank = item.get("rank") or "未进入前200名"
+                f.write(
+                    f"| {brand} | {rank} | "
+                    f"{item.get('用户支付金额', '未获取')} | "
+                    f"{item.get('成交订单数', '未获取')} | "
+                    f"{item.get('商品点击人数', '未获取')} | "
+                    f"{item.get('matched_shop_count', 0)} |\n"
+                )
+            f.write("\n")
+
+        f.write("\n---\n*Generated by script*\n")
+    print(f"[OK] Competitor Markdown: {md_path}")
+
+
+def _parse_numeric_value(value) -> float | None:
+    text = str(value or "").strip()
+    if not text or text in {"未获取", "--", "-"}:
+        return None
+    multiplier = 1.0
+    if "万" in text:
+        multiplier = 10000.0
+    elif "亿" in text:
+        multiplier = 100000000.0
+    text = (
+        text.replace(",", "")
+        .replace("¥", "")
+        .replace("￥", "")
+        .replace("元", "")
+        .replace("万", "")
+        .replace("亿", "")
+        .replace("%", "")
+        .strip()
+    )
+    match = re.search(r"-?\d+(?:\.\d+)?", text)
+    if not match:
+        return None
+    return float(match.group(0)) * multiplier
+
+
+def _format_money_value(value: float | None) -> str:
+    if value is None:
+        return "未获取"
+    return f"¥{value:,.2f}"
+
+
+def _format_percent_value(value: float | None) -> str:
+    if value is None:
+        return "未获取"
+    return f"{value:.2f}%"
+
+
+def _format_ratio_value(value: float | None) -> str:
+    if value is None:
+        return "未获取"
+    return f"{value:.4f}"
+
+
+def _calculate_store_derived_metrics(data: dict) -> dict:
+    metrics = data.get("metrics", {}) or {}
+    qianchuan = data.get("qianchuan_metrics", {}) or {}
+
+    transaction_amount = _parse_numeric_value(metrics.get("成交金额"))
+    refund_amount = _parse_numeric_value(metrics.get("成交退款金额"))
+    ad_spend = _parse_numeric_value((qianchuan.get("整体消耗") or {}).get("value"))
+    net_deal_roi = _parse_numeric_value((qianchuan.get("净成交ROI") or {}).get("value"))
+
+    sales_amount = None
+    paid_ratio = None
+    adjusted_roi = None
+    if transaction_amount is not None and refund_amount is not None:
+        sales_amount = transaction_amount - refund_amount
+    if ad_spend is not None and sales_amount not in (None, 0):
+        paid_ratio = ad_spend / sales_amount * 100
+    if net_deal_roi is not None and transaction_amount not in (None, 0) and refund_amount is not None:
+        adjusted_roi = net_deal_roi * (refund_amount / transaction_amount)
+
+    return {
+        "销售额": _format_money_value(sales_amount),
+        "付费占比": _format_percent_value(paid_ratio),
+        "ROI": _format_ratio_value(adjusted_roi),
+        "raw": {
+            "成交金额": transaction_amount,
+            "成交退款金额": refund_amount,
+            "整体消耗": ad_spend,
+            "净成交ROI": net_deal_roi,
+            "销售额": sales_amount,
+            "付费占比": paid_ratio,
+            "ROI": adjusted_roi,
+        },
+        "formula": {
+            "销售额": "成交金额 - 成交退款金额",
+            "付费占比": "整体消耗 / 销售额 * 100%",
+            "ROI": "净成交ROI * (成交退款金额 / 成交金额)",
+        },
+    }
+
+
+def save_results(data: dict) -> None:
+    data["derived_metrics"] = _calculate_store_derived_metrics(data)
+    date_stamp = datetime.now().strftime("%Y%m%d")
+    base_name = f"光影行星店铺信息汇总_{date_stamp}"
+
+    os.makedirs(STORE_DATA_DIR, exist_ok=True)
+
+    json_path = os.path.join(STORE_DATA_DIR, f"{base_name}.json")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(data, f, ensure_ascii=False, indent=2)
+    print(f"[OK] JSON: {json_path}")
+
+    md_path = os.path.join(STORE_DATA_DIR, f"{base_name}.md")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("# 抖音罗盘经营数据报表\n\n")
+        f.write(f"- **店铺**: {data.get('shop_name', '未知')}\n")
+        f.write(f"- **周期**: {data.get('date_range', '近7天')}\n")
+        f.write(f"- **时间**: {data.get('scraped_at', '')}\n")
+        f.write(f"- **URL**: {data.get('url', '')}\n\n")
+        f.write("## 核心指标\n\n")
+        f.write("| 指标 | 数值 |\n")
+        f.write("|------|------|\n")
+        metrics = data.get("metrics", {})
+        for label, value in [
+            ("成交金额", metrics.get("成交金额", "未获取")),
+            ("成交人数", metrics.get("成交人数", "未获取")),
+            ("客单价", metrics.get("客单价", "未获取")),
+            ("成交退款金额（支付时间）", metrics.get("成交退款金额", "未获取")),
+            ("商品点击人数", metrics.get("商品点击人数", "未获取")),
+        ]:
+            f.write(f"| {label} | {value} |\n")
+        if not metrics:
+            f.write("| (无数据) | 请查看截图确认页面结构 |\n")
+        derived_metrics = data.get("derived_metrics", {})
+        if derived_metrics:
+            f.write("\n## 计算指标\n\n")
+            f.write("| 指标 | 数值 | 公式 |\n")
+            f.write("|------|------|------|\n")
+            formulas = derived_metrics.get("formula", {})
+            for label in ["销售额", "付费占比", "ROI"]:
+                f.write(f"| {label} | {derived_metrics.get(label, '未获取')} | {formulas.get(label, '')} |\n")
+        qianchuan_metrics = data.get("qianchuan_metrics", {})
+        if qianchuan_metrics:
+            f.write("\n## 千川投放数据\n\n")
+            f.write(f"- **周期**: {qianchuan_metrics.get('date_range', data.get('date_range', ''))}\n")
+            f.write(f"- **URL**: {qianchuan_metrics.get('url', '')}\n\n")
+            f.write("| 指标 | 数值 | 变化 | 百分比 |\n")
+            f.write("|------|------|------|------|\n")
+            for label in ["整体消耗", "净成交ROI"]:
+                item = qianchuan_metrics.get(label, {})
+                if item:
+                    f.write(
+                        f"| {label} | {item.get('value', '未获取')} | "
+                        f"{item.get('change_direction', '未获取')} | {item.get('change_percent', '未获取')} |\n"
+                    )
+            f.write(f"\n- **采集时间**: {qianchuan_metrics.get('captured_at', '')}\n")
+        experience_score = data.get("experience_score", {})
+        if experience_score:
+            f.write("\n## 店铺评分（体检分）\n\n")
+            f.write(f"- **体检分**: {experience_score.get('score', '未获取')}\n")
+            f.write(f"- **商品体验分**: {experience_score.get('product_experience_score', '未获取')}\n")
+            f.write(f"- **物流体验分**: {experience_score.get('logistics_experience_score', '未获取')}\n")
+            f.write(f"- **服务体验分**: {experience_score.get('service_experience_score', '未获取')}\n")
+            f.write(f"- **差行为扣分**: {experience_score.get('penalty_score', '未获取')}\n")
+            f.write(f"- **URL**: {experience_score.get('url', '')}\n")
+            f.write(f"- **采集时间**: {experience_score.get('captured_at', '')}\n")
+        market_rankings = data.get("market_rankings", {})
+        if market_rankings:
+            f.write("\n## 市场店铺榜单\n\n")
+            for category, ranking in market_rankings.items():
+                f.write(f"### {category}\n\n")
+                if ranking.get("rank") or ranking.get("rank_change"):
+                    f.write(f"- **排名**: {ranking.get('rank', '未获取')}\n")
+                    f.write(f"- **排名变化**: {ranking.get('rank_change', '无')}\n\n")
+                columns = ranking.get("columns", {})
+                if columns:
+                    f.write("| 字段 | 数值 |\n")
+                    f.write("|------|------|\n")
+                    for key, value in columns.items():
+                        f.write(f"| {key} | {value} |\n")
+                else:
+                    f.write(f"- **榜单行**: {ranking.get('row_text', '未获取')}\n")
+                f.write(f"- **采集时间**: {ranking.get('captured_at', '')}\n\n")
+        f.write("\n---\n*Generated by script*\n")
+    print(f"[OK] Markdown: {md_path}")
+
+
+def main() -> int:
+    print("=" * 60)
+    print("  抖音罗盘经营数据抓取")
+    print("=" * 60)
+    print(f"  Output: {OUTPUT_DIR}")
+    print(f"  Data: {DATA_DIR}")
+    print(f"  Store data: {STORE_DATA_DIR}")
+    print(f"  Competitor data: {COMPETITOR_DATA_DIR}")
+    print(f"  Cookies: {COOKIES_FILE}")
+    print(f"  Profile: {USER_DATA_DIR}")
+    print(f"  Keep browser open: {KEEP_BROWSER_OPEN}")
+    print("=" * 60 + "\n")
+
+    os.makedirs(OUTPUT_DIR, exist_ok=True)
+    os.makedirs(DATA_DIR, exist_ok=True)
+    os.makedirs(STORE_DATA_DIR, exist_ok=True)
+    os.makedirs(COMPETITOR_DATA_DIR, exist_ok=True)
+    os.makedirs(USER_DATA_DIR, exist_ok=True)
+    enable_adaptive_fetchers()
+    print("[OK] Adaptive enabled")
+
+    cookies = _load_cookies()
+
+    response_url = ""
+    response_status = 0
+
+    print("[INFO] Launching browser...")
+    try:
+        fetch_kwargs = dict(
+            headless=False,
+            real_chrome=True,
+            cookies=cookies,
+            timeout=120000,
+            page_action=_page_automation,
+            wait=2000,
+            extra_flags=["--start-maximized", "--window-size=1920,1080"],
+        )
+        try:
+            response = DynamicFetcher.fetch(
+                "https://compass.jinritemai.com/shop/business-part",
+                user_data_dir=USER_DATA_DIR,
+                **fetch_kwargs,
+            )
+        except Exception as profile_error:
+            print(f"[WARN] Launch with profile failed: {profile_error}")
+            print("[INFO] Retrying with saved cookies only...")
+            response = DynamicFetcher.fetch(
+                "https://compass.jinritemai.com/shop/business-part",
+                **fetch_kwargs,
+            )
+        response_url = response.url
+        response_status = response.status
+        print(f"\n[OK] Done: HTTP {response_status} | {response_url}")
+    except Exception as e:
+        print(f"\n[ERR] {e}")
+        import traceback
+        traceback.print_exc()
+        return 1
+
+    final_data = {
+        "shop_name": SHOP_NAME,
+        "date_range": _date_range_label(*_previous_week_range()),
+        "scraped_at": datetime.now().isoformat(),
+        "url": response_url,
+        "metrics": _scraped_metrics,
+        "market_rankings": _market_rankings,
+        "qianchuan_metrics": _qianchuan_metrics,
+        "experience_score": _experience_score,
+    }
+
+    print("\n" + "=" * 60)
+    print("  Results")
+    print("=" * 60)
+    for k, v in final_data["metrics"].items():
+        print(f"  {k}: {v}")
+    if not final_data["metrics"]:
+        print("  (No data)")
+    print("=" * 60 + "\n")
+
+    if not _results_saved:
+        save_results(final_data)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/dynamic_session_src.py b/src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/dynamic_session_src.py
new file mode 100644
index 0000000..ab3f709
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/vendors/dy-data-flow/dynamic_session_src.py
@@ -0,0 +1,191 @@
+class DynamicSession(SyncSession, DynamicSessionMixin):
+    """A Browser session manager with page pooling."""
+
+    __slots__ = (
+        "_config",
+        "_context_options",
+        "_browser_options",
+        "_user_data_dir",
+        "_headers_keys",
+        "max_pages",
+        "page_pool",
+        "_max_wait_for_page",
+        "playwright",
+        "context",
+    )
+
+    def __init__(self, **kwargs: Unpack[PlaywrightSession]):
+        """A Browser session manager with page pooling, it's using a persistent browser Context by default with a temporary user profile directory.
+
+        :param headless: Run the browser in headless/hidden (default), or headful/visible mode.
+        :param disable_resources: Drop requests for unnecessary resources for a speed boost.
+            Requests dropped are of type `font`, `image`, `media`, `beacon`, `object`, `imageset`, `texttrack`, `websocket`, `csp_report`, and `stylesheet`.
+        :param blocked_domains: A set of domain names to block requests to. Subdomains are also matched (e.g., ``"example.com"`` blocks ``"sub.example.com"`` too).
+        :param useragent: Pass a useragent string to be used. Otherwise the fetcher will generate a real Useragent of the same browser and use it.
+        :param cookies: Set cookies for the next request.
+        :param network_idle: Wait for the page until there are no network connections for at least 500 ms.
+        :param timeout: The timeout in milliseconds that is used in all operations and waits through the page. The default is 30,000
+        :param wait: The time (milliseconds) the fetcher will wait after everything finishes before closing the page and returning the ` Response ` object.
+        :param page_action: Added for automation. A function that takes the `page` object, runs after navigation, and does the automation you need.
+        :param page_setup: A function that takes the `page` object, runs before navigation. Use it to register event listeners or routes that must be set up before the page loads.
+        :param wait_selector: Wait for a specific CSS selector to be in a specific state.
+        :param init_script: An absolute path to a JavaScript file to be executed on page creation for all pages in this session.
+        :param locale: Specify user locale, for example, `en-GB`, `de-DE`, etc. Locale will affect navigator.language value, Accept-Language request header value as well as number and date formatting
+            rules. Defaults to the system default locale.
+        :param timezone_id: Changes the timezone of the browser. Defaults to the system timezone.
+        :param wait_selector_state: The state to wait for the selector given with `wait_selector`. The default state is `attached`.
+        :param real_chrome: If you have a Chrome browser installed on your device, enable this, and the Fetcher will launch an instance of your browser and use it.
+        :param load_dom: Enabled by default, wait for all JavaScript on page(s) to fully load and execute.
+        :param cdp_url: Instead of launching a new browser instance, connect to this CDP URL to control real browsers through CDP.
+        :param google_search: Enabled by default, Scrapling will set a Google referer header.
+        :param extra_headers: A dictionary of extra headers to add to the request. _The referer set by `google_search` takes priority over the referer set here if used together._
+        :param proxy: The proxy to be used with requests, it can be a string or a dictionary with the keys 'server', 'username', and 'password' only.
+        :param user_data_dir: Path to a User Data Directory, which stores browser session data like cookies and local storage. The default is to create a temporary directory.
+        :param extra_flags: A list of additional browser flags to pass to the browser on launch.
+        :param selector_config: The arguments that will be passed in the end while creating the final Selector's class.
+        :param additional_args: Additional arguments to be passed to Playwright's context as additional settings, and it takes higher priority than Scrapling's settings.
+        """
+        self.__validate__(**kwargs)
+        super().__init__()
+
+    def start(self):
+        """Create a browser for this instance and context."""
+        if not self.playwright:
+            self.playwright = sync_playwright().start()
+
+            try:
+                if self._config.cdp_url:  # pragma: no cover
+                    self.browser = self.playwright.chromium.connect_over_cdp(endpoint_url=self._config.cdp_url)
+                    if not self._config.proxy_rotator and self.browser:
+                        self.context = self.browser.new_context(**self._context_options)
+                elif self._config.proxy_rotator:
+                    self.browser = self.playwright.chromium.launch(**self._browser_options)
+                else:
+                    persistent_options = (
+                        self._browser_options | self._context_options | {"user_data_dir": self._user_data_dir}
+                    )
+                    self.context = self.playwright.chromium.launch_persistent_context(**persistent_options)
+
+                if self.context:
+                    self.context = self._initialize_context(self._config, self.context)
+
+                self._is_alive = True
+            except Exception:
+                # Clean up playwright if browser setup fails
+                self.playwright.stop()
+                self.playwright = None
+                raise
+        else:
+            raise RuntimeError("Session has been already started")
+
+    def fetch(self, url: str, **kwargs: Unpack[PlaywrightFetchParams]) -> Response:
+        """Opens up the browser and do your request based on your chosen options.
+
+        :param url: The Target url.
+        :param google_search: Enabled by default, Scrapling will set a Google referer header.
+        :param timeout: The timeout in milliseconds that is used in all operations and waits through the page. The default is 30,000
+        :param wait: The time (milliseconds) the fetcher will wait after everything finishes before closing the page and returning the ` Response ` object.
+        :param page_action: Added for automation. A function that takes the `page` object, runs after navigation, and does the automation you need.
+        :param page_setup: A function that takes the `page` object, runs before navigation. Use it to register event listeners or routes that must be set up before the page loads.
+        :param extra_headers: A dictionary of extra headers to add to the request. _The referer set by `google_search` takes priority over the referer set here if used together._
+        :param disable_resources: Drop requests for unnecessary resources for a speed boost.
+            Requests dropped are of type `font`, `image`, `media`, `beacon`, `object`, `imageset`, `texttrack`, `websocket`, `csp_report`, and `stylesheet`.
+        :param blocked_domains: A set of domain names to block requests to. Subdomains are also matched (e.g., ``"example.com"`` blocks ``"sub.example.com"`` too).
+        :param wait_selector: Wait for a specific CSS selector to be in a specific state.
+        :param wait_selector_state: The state to wait for the selector given with `wait_selector`. The default state is `attached`.
+        :param network_idle: Wait for the page until there are no network connections for at least 500 ms.
+        :param load_dom: Enabled by default, wait for all JavaScript on page(s) to fully load and execute.
+        :param selector_config: The arguments that will be passed in the end while creating the final Selector's class.
+        :param proxy: Static proxy to override rotator and session proxy. A new browser context will be created and used with it.
+        :return: A `Response` object.
+        """
+        static_proxy = kwargs.pop("proxy", None)
+
+        params = _validate(kwargs, self, PlaywrightConfig)
+        if not self._is_alive:  # pragma: no cover
+            raise RuntimeError("Context manager has been closed")
+
+        request_headers_keys = {h.lower() for h in params.extra_headers.keys()} if params.extra_headers else set()
+        referer = (
+            "https://www.google.com/" if (params.google_search and "referer" not in request_headers_keys) else None
+        )
+
+        for attempt in range(self._config.retries):
+            proxy: Optional[ProxyType] = None
+            if self._config.proxy_rotator and static_proxy is None:
+                proxy = self._config.proxy_rotator.get_proxy()
+            else:
+                proxy = static_proxy
+
+            with self._page_generator(
+                params.timeout, params.extra_headers, params.disable_resources, proxy, params.blocked_domains
+            ) as page_info:
+                final_response: List = [None]
+                xhr_captured: List = []
+                page = page_info.page
+                page.on(
+                    "response",
+                    self._create_response_handler(
+                        page_info,
+                        final_response,
+                        xhr_pattern=self._config.capture_xhr,
+                        xhr_container=xhr_captured,
+                    ),
+                )
+
+                if params.page_setup:
+                    try:
+                        params.page_setup(page)
+                    except Exception as e:  # pragma: no cover
+                        log.error(f"Error executing page_setup: {e}")
+
+                try:
+                    first_response = page.goto(url, referer=referer)
+                    self._wait_for_page_stability(page, params.load_dom, params.network_idle)
+
+                    if not first_response:
+                        raise RuntimeError(f"Failed to get response for {url}")
+
+                    if params.page_action:
+                        try:
+                            _ = params.page_action(page)
+                        except Exception as e:  # pragma: no cover
+                            log.error(f"Error executing page_action: {e}")
+
+                    if params.wait_selector:
+                        try:
+                            waiter: Locator = page.locator(params.wait_selector)
+                            waiter.first.wait_for(state=params.wait_selector_state)
+                            self._wait_for_page_stability(page, params.load_dom, params.network_idle)
+                        except Exception as e:  # pragma: no cover
+                            log.error(f"Error waiting for selector {params.wait_selector}: {e}")
+
+                    page.wait_for_timeout(params.wait)
+
+                    response = ResponseFactory.from_playwright_response(
+                        page,
+                        first_response,
+                        final_response[0],
+                        params.selector_config,
+                        meta={"proxy": proxy},
+                        xhr_captured=xhr_captured,
+                    )
+                    return response
+
+                except Exception as e:
+                    page_info.mark_error()
+                    if attempt < self._config.retries - 1:
+                        if is_proxy_error(e):
+                            log.warning(
+                                f"Proxy '{proxy}' failed (attempt {attempt + 1}) | Retrying in {self._config.retry_delay}s..."
+                            )
+                        else:
+                            log.warning(
+                                f"Attempt {attempt + 1} failed: {e}. Retrying in {self._config.retry_delay}s..."
+                            )
+                        time_sleep(self._config.retry_delay)
+                    else:
+                        log.error(f"Failed after {self._config.retries} attempts: {e}")
+                        raise
+
+        raise RuntimeError("Request failed")  # pragma: no cover
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/collection_progress.py b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/collection_progress.py
new file mode 100644
index 0000000..b4a9774
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/collection_progress.py
@@ -0,0 +1,55 @@
+import json
+import os
+import traceback
+from datetime import datetime
+
+
+class SpuProgress:
+    """Append-only progress log for long SPU collection jobs."""
+
+    def __init__(self, output_dir: str, job_name: str):
+        today = datetime.now().strftime("%Y-%m-%d")
+        progress_dir = os.path.join(output_dir, ".progress")
+        os.makedirs(progress_dir, exist_ok=True)
+        self.path = os.path.join(progress_dir, f"{job_name}_{today}.jsonl")
+
+    def append(self, status: str, item: dict, **extra):
+        record = {
+            "time": datetime.now().isoformat(timespec="seconds"),
+            "status": status,
+            "spu": str(item.get("spu", "")),
+            "product": item.get("product", ""),
+            "brand": item.get("brand", ""),
+            **extra,
+        }
+        with open(self.path, "a", encoding="utf-8") as f:
+            f.write(json.dumps(record, ensure_ascii=False) + "\n")
+
+    def mark_running(self, item: dict, index: int, total: int):
+        self.append("running", item, index=index, total=total)
+
+    def mark_success(self, item: dict, **extra):
+        self.append("success", item, **extra)
+
+    def mark_failed(self, item: dict, error: Exception):
+        self.append(
+            "failed",
+            item,
+            error=str(error),
+            traceback=traceback.format_exc(),
+        )
+
+    def successful_spus(self) -> set[str]:
+        if not os.path.exists(self.path):
+            return set()
+
+        completed = set()
+        with open(self.path, "r", encoding="utf-8") as f:
+            for line in f:
+                try:
+                    record = json.loads(line)
+                except json.JSONDecodeError:
+                    continue
+                if record.get("status") == "success" and record.get("spu"):
+                    completed.add(str(record["spu"]))
+        return completed
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/config.py b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/config.py
new file mode 100644
index 0000000..dbb2040
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/config.py
@@ -0,0 +1,125 @@
+import os
+from gyxx_flow.adapters import RuntimeServicePolicy
+
+os.environ.update(RuntimeServicePolicy().apply(os.environ))
+import sys
+from pathlib import Path
+
+RUNTIME_ROOT = Path(__file__).resolve().parents[2]
+if str(RUNTIME_ROOT) not in sys.path:
+    sys.path.insert(0, str(RUNTIME_ROOT))
+
+from runtime_paths import RAW_DATA_ROOT, managed_data_path, runtime_script
+from dotenv import load_dotenv
+
+load_dotenv()
+
+SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
+
+# ── Hermes API Server 配置（采集端/分析端本地服务，固定端口）──────────────────
+
+HERMES_API_KEY = os.getenv("HERMES_API_KEY", "")
+
+HERMES_API_SERVERS = {
+    "collector": {
+        "url": os.getenv("COLLECTOR_API_SERVER_URL", "http://localhost:8643/v1/chat/completions"),
+        "profile": "data-collector",
+        "model": "mimo-v2.5-pro",
+    },
+    "analyzer": {
+        "url": os.getenv("ANALYZER_API_SERVER_URL", "http://localhost:8642/v1/chat/completions"),
+        "profile": "data-analyzer",
+        "model": "mimo-v2.5-pro",
+    },
+}
+
+# ── Hermes Gateway 配置（端口随机，由环境变量传入）────────────────────────────
+
+HERMES_GATEWAYS = {
+    "collector": {
+        "url": os.getenv("COLLECTOR_HERMES_GATEWAY_URL", ""),
+        "profile": "data-collector",
+        "model": "mimo-v2.5-pro",
+    },
+    "analyzer": {
+        "url": os.getenv("ANALYZER_HERMES_GATEWAY_URL", ""),
+        "profile": "data-analyzer",
+        "model": "mimo-v2.5-pro",
+    },
+}
+
+# ── 采集默认配置 ─────────────────────────────────────────────────────────────
+
+DEFAULT_SHOP = os.getenv("SHOP_NAME", "光影行星研发")
+DEFAULT_PASSWORD = os.getenv("SHOP_PASSWORD", "")
+DEFAULT_INDUSTRY = os.getenv("INDUSTRY", "男包")
+DEFAULT_OUTPUT = str(
+    managed_data_path(
+        os.getenv("OUTPUT_DIR", str(RAW_DATA_ROOT / "jd-vendor" / "店铺数据"))
+    )
+)
+DEFAULT_CALIBER_DAYS = int(os.getenv("CALIBER_DAYS", "7"))
+
+# ── 数据目录 ─────────────────────────────────────────────────────────────────
+
+DATA_DIR = str(
+    managed_data_path(os.getenv("DATA_DIR", str(RAW_DATA_ROOT / "jd-vendor")))
+)
+PEER_PRODUCT_OUTPUT = str(
+    managed_data_path(
+        os.getenv("PEER_PRODUCT_OUTPUT", os.path.join(DATA_DIR, "竞品数据"))
+    )
+)
+PEER_STORE_OUTPUT = str(
+    managed_data_path(
+        os.getenv("PEER_STORE_OUTPUT", os.path.join(DATA_DIR, "竞店数据"))
+    )
+)
+PRODUCT_OUTPUT = str(
+    managed_data_path(
+        os.getenv("PRODUCT_OUTPUT", os.path.join(DATA_DIR, "单品数据"))
+    )
+)
+
+# ── 采集脚本路径 ─────────────────────────────────────────────────────────────
+
+COLLECTOR_SCRIPT = str(runtime_script("vendors/jd-data-flow/jd_data_collector.py"))
+PEER_PRODUCT_SCRIPT = str(
+    runtime_script("vendors/jd-data-flow/jd_peer_product_data_collector.py")
+)
+# The source project did not contain a peer-store collector implementation.
+PEER_STORE_SCRIPT = None
+PRODUCT_SCRIPT = str(runtime_script("vendors/jd-data-flow/jd_product_data_collector.py"))
+
+# ── 通知目标 ─────────────────────────────────────────────────────────────────
+# 飞书 open_id 列表，分析端会将报告发送给这些用户
+# 可通过环境变量 NOTIFY_OPEN_IDS 逗号分隔覆盖
+
+NOTIFY_OPEN_IDS = [
+    id.strip()
+    for id in os.getenv("NOTIFY_OPEN_IDS", "ou_7ad5fc8012e2f741afc5346e05ffd447").split(",")
+    if id.strip()
+]
+
+# 分析端绑定的飞书应用 ID（用于发送消息）
+ANALYZER_FEISHU_APP_ID = os.getenv("ANALYZER_FEISHU_APP_ID", "cli_aa8c4fc918b85cce")
+
+# ── 多维表格写入配置 ──────────────────────────────────────────────────────────
+# 所有 base_writer 节点都通过 lark-cli base +record-upsert/+record-list 操作
+
+BASE_APP_TOKEN = os.getenv("BASE_APP_TOKEN", "")
+
+# 店铺表（运营OKR（周）🚫🚫🚫）
+SHOP_BASE_TABLE_ID = os.getenv("SHOP_BASE_TABLE_ID", "tblsvXfxVkr3Zzwm")
+SHOP_BASE_PARENT_RECORD_ID = os.getenv("SHOP_BASE_PARENT_RECORD_ID", "recv7vJnwmVjan")  # 京东父锚
+SHOP_BASE_PLATFORM = os.getenv("SHOP_BASE_PLATFORM", "京东")
+SHOP_BASE_FIELDS = {
+    "父记录": "fldRCOjzP0",
+    "时间": "fldHQecrCX",
+    "月份": "fldt9txdb8",
+    "平台": "fldlejDqfr",
+    "上周销售额": "fldogTrPCM",
+    "店铺评分": "fldygNpysx",
+    "AI分析": "fldqiw6Ens",
+    "备注": "fld00lqdcf",
+}
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_data_collector.py b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_data_collector.py
new file mode 100644
index 0000000..ebc732d
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_data_collector.py
@@ -0,0 +1,2448 @@
+"""
+京东店铺数据采集脚本
+自动登录京东商家平台，采集商智和京准通数据，生成汇总报告。
+
+依赖:
+    pip install playwright
+    playwright install chromium
+
+用法:
+    python jd_data_collector.py
+    python jd_data_collector.py --shop "店铺名" --password "密码" --output "输出目录" --industry "行业" --caliber 7
+"""
+
+import argparse
+import json
+import os
+import re
+from datetime import datetime, timedelta
+from urllib.parse import urlparse
+from playwright.sync_api import sync_playwright, Page, BrowserContext, TimeoutError as PWTimeoutError
+from runtime_paths import RAW_DATA_ROOT, TMP_ROOT, managed_data_path
+
+SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
+
+# ── 配置 ─────────────────────────────────────────────────────────────────────
+
+DEFAULT_CONFIG = {
+    "shop": os.getenv("JD_SHOP", ""),
+    "password": os.getenv("JD_PASSWORD", ""),
+    "output": str(
+        managed_data_path(
+            os.getenv(
+                "JD_STORE_OUTPUT_DIR",
+                str(RAW_DATA_ROOT / "jd-vendor" / "店铺数据"),
+            )
+        )
+    ),
+    "industry": os.getenv("INDUSTRY", "箱包"),
+    "caliber_days": int(os.getenv("CALIBER_DAYS", "7")),
+}
+
+# ── 工具函数 ──────────────────────────────────────────────────────────────────
+
+def get_last_week_range():
+    """返回上周一和上周日的日期 (datetime, datetime)"""
+    today = datetime.now()
+    # 本周一 = today - (weekday) 天
+    this_monday = today - timedelta(days=today.weekday())
+    last_monday = this_monday - timedelta(days=7)
+    last_sunday = this_monday - timedelta(days=1)
+    return last_monday, last_sunday
+
+
+def extract(page: Page, js_expr: str, default=None):
+    """用 page.evaluate 提取数据，失败返回 default"""
+    try:
+        result = page.evaluate(js_expr)
+        if result is None:
+            return default
+        return str(result).strip()
+    except Exception:
+        return default
+
+
+def safe_click(page: Page, text: str, timeout=5000):
+    """用 text= 选择器点击元素，等待并忽略不存在的情况"""
+    try:
+        page.click(f"text={text}", timeout=timeout)
+        return True
+    except PWTimeoutError:
+        print(f"  [!] 未找到文本: {text}")
+        return False
+
+
+def wait_for_navigation(page: Page, timeout=15000):
+    """等待页面加载完成"""
+    try:
+        page.wait_for_load_state("networkidle", timeout=timeout)
+    except PWTimeoutError:
+        pass
+
+
+def _is_shangzhi_url(url: str) -> bool:
+    """Match the actual Shangzhi host, never a returnUrl/query substring."""
+    try:
+        hostname = (urlparse(str(url)).hostname or "").lower().rstrip(".")
+    except (TypeError, ValueError):
+        return False
+    allowed_hosts = ("sz.jd.com", "jdsz.jd.com")
+    return any(
+        hostname == allowed or hostname.endswith(f".{allowed}")
+        for allowed in allowed_hosts
+    )
+
+
+def _shop_home_overlay_visible(page: Page) -> bool:
+    """Return whether the JD shop-home promotion overlay is blocking the page."""
+    try:
+        return page.locator("div.all-dialog:visible").first.is_visible(timeout=500)
+    except Exception:
+        return False
+
+
+def _shop_home_overlay_blocks_shangzhi(page: Page) -> bool:
+    """Return whether an all-dialog overlay is the topmost element over 商智."""
+    try:
+        return bool(
+            page.evaluate(
+                """() => {
+                    const visible = (el) => {
+                        const style = getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style.display !== 'none' && style.visibility !== 'hidden' &&
+                            rect.width > 0 && rect.height > 0;
+                    };
+                    const entry = Array.from(document.querySelectorAll(
+                        '.shop-menu-navigate__menu-name, a, [role="menuitem"]'
+                    )).find((el) => visible(el) && (el.textContent || '').trim() === '商智');
+                    if (!entry) return false;
+                    const rect = entry.getBoundingClientRect();
+                    const top = document.elementFromPoint(
+                        rect.left + rect.width / 2,
+                        rect.top + rect.height / 2
+                    );
+                    return Boolean(top && top.closest('div.all-dialog'));
+                }"""
+            )
+        )
+    except Exception:
+        return False
+
+
+def dismiss_shop_home_overlays(page: Page) -> bool:
+    """Dismiss JD shop-home promotions, including variants without a close button."""
+    if not _shop_home_overlay_visible(page):
+        return True
+    if not _shop_home_overlay_blocks_shangzhi(page):
+        return True
+
+    close_selectors = (
+        "div.all-dialog .close-btn:visible",
+        "div.all-dialog button:has-text('关闭'):visible",
+        "div.all-dialog button:has-text('稍后再说'):visible",
+        "div.all-dialog [role='button'][aria-label*='关闭']:visible",
+        "div.all-dialog button[class*='close']:visible",
+        "div.all-dialog [role='button'][class*='close']:visible",
+        "div.all-dialog [class*='dialog-close']:visible",
+    )
+    for selector in close_selectors:
+        try:
+            close_button = page.locator(selector).first
+            if not close_button.is_visible(timeout=300):
+                continue
+            close_button.click(timeout=1500)
+            page.wait_for_timeout(300)
+            if not _shop_home_overlay_blocks_shangzhi(page):
+                print("  [OK] 已关闭商家后台推广弹窗")
+                return True
+        except Exception:
+            continue
+
+    try:
+        page.keyboard.press("Escape")
+        page.wait_for_timeout(300)
+    except Exception:
+        pass
+    if not _shop_home_overlay_blocks_shangzhi(page):
+        return True
+
+    # Some promotion variants expose only a pointer-blocking backdrop. Hide only
+    # the all-dialog that elementFromPoint confirms is covering the 商智 entry.
+    try:
+        hidden_count = page.evaluate(
+            """() => {
+                const visible = (el) => {
+                    const style = getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none' && style.visibility !== 'hidden' &&
+                        rect.width > 0 && rect.height > 0;
+                };
+                const entry = Array.from(document.querySelectorAll(
+                    '.shop-menu-navigate__menu-name, a, [role="menuitem"]'
+                )).find((el) => visible(el) && (el.textContent || '').trim() === '商智');
+                if (!entry) return 0;
+                const rect = entry.getBoundingClientRect();
+                const top = document.elementFromPoint(
+                    rect.left + rect.width / 2,
+                    rect.top + rect.height / 2
+                );
+                const overlay = top && top.closest('div.all-dialog');
+                if (!overlay || !visible(overlay)) return 0;
+                overlay.style.setProperty('pointer-events', 'none', 'important');
+                overlay.style.setProperty('display', 'none', 'important');
+                overlay.setAttribute('aria-hidden', 'true');
+                return 1;
+            }"""
+        )
+        if hidden_count:
+            print(f"  [OK] 已清理 {hidden_count} 个无关闭入口的推广遮罩")
+        page.wait_for_timeout(300)
+    except Exception as exc:
+        print(f"  [WARN] 推广遮罩兜底清理失败: {exc}")
+
+    return not _shop_home_overlay_blocks_shangzhi(page)
+
+
+def save_date_picker_debug(page: Page, reason: str):
+    """保存日期选择失败时的页面现场，便于定位页面结构变化。"""
+    debug_dir = str(TMP_ROOT / "debug" / "jd_date_picker")
+    os.makedirs(debug_dir, exist_ok=True)
+    ts = datetime.now().strftime("%Y%m%d_%H%M%S")
+    safe_reason = re.sub(r"[^0-9A-Za-z_-]+", "_", reason)[:40] or "date_picker"
+    screenshot_path = os.path.join(debug_dir, f"{ts}_{safe_reason}.png")
+    html_path = os.path.join(debug_dir, f"{ts}_{safe_reason}.html")
+
+    try:
+        page.screenshot(path=screenshot_path, full_page=True)
+    except Exception as e:
+        print(f"    [WARN] 保存日期调试截图失败: {e}")
+        screenshot_path = ""
+
+    try:
+        html = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    const before = window.getComputedStyle(el, '::before');
+                    const after = window.getComputedStyle(el, '::after');
+                    const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                        || (after.content && after.content !== 'none' && after.content !== '');
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && (rect.width > 0 || hasPseudo)
+                        && (rect.height > 0 || hasPseudo);
+                };
+                const roots = Array.from(document.querySelectorAll(
+                    '.jmt-date-picker-panels-wrapper, .date-picker-popup, .calendar-popup, ' +
+                    '.jmt-popover, .jmt-popover-content, .jmt-dropdown, .jmt-dropdown-menu, ' +
+                    '[class*="picker"], [class*="calendar"], [class*="popover"], [class*="dropdown"], ' +
+                    '.ant-picker-panel, .ant-picker-dropdown, .el-picker-panel'
+                )).filter(visible);
+                return roots.map((el, index) => {
+                    const rect = el.getBoundingClientRect();
+                    return `<!-- candidate ${index} ${Math.round(rect.width)}x${Math.round(rect.height)} -->\n${el.outerHTML}`;
+                }).join('\n\n');
+            }"""
+        )
+        if not html.strip():
+            html = page.content()
+        with open(html_path, "w", encoding="utf-8") as f:
+            f.write(html)
+    except Exception as e:
+        print(f"    [WARN] 保存日期调试 HTML 失败: {e}")
+        html_path = ""
+
+    if screenshot_path or html_path:
+        print(f"    [DEBUG] 日期选择现场已保存: {screenshot_path or '无截图'} {html_path or '无HTML'}")
+
+
+# ── Step 1: 登录 ──────────────────────────────────────────────────────────────
+
+def step_login(page: Page, shop: str, password: str):
+    """登录京东商家平台"""
+    print("[Step 1] 登录京东商家平台...")
+
+    page.goto("https://shop.jd.com/jdm/home", wait_until="networkidle")
+    page.wait_for_timeout(3000)
+
+    # 1. 定位账号输入框
+    account_input = page.locator('input[placeholder="请输入账号名/邮箱"]')
+    account_input.fill(shop)
+    page.wait_for_timeout(1000)
+
+    # 2. 定位密码输入框
+    password_input = page.locator('input[placeholder="请输入登录密码"]')
+    password_input.fill(password)
+    page.wait_for_timeout(1000)
+
+    # 3. 定位立即登录按钮并点击
+    login_button = page.locator('button:has-text("立即登录")')
+    login_button.click()
+
+    # 等待登录完成
+    page.wait_for_timeout(8000)
+    wait_for_navigation(page)
+
+    # 尝试关闭推广弹窗
+    print("  尝试关闭推广弹窗...")
+    dismiss_shop_home_overlays(page)
+    page.wait_for_timeout(1000)
+
+    print("  登录完成")
+
+
+def select_last_week_any_day(page):
+    """Hover the week tab, wait for the date panel, then select last week range."""
+    print("  选择上周时间范围...")
+
+    last_monday, last_sunday = get_last_week_range()
+    expected_start = last_monday.strftime("%Y-%m-%d")
+    expected_end = last_sunday.strftime("%Y-%m-%d")
+    target_date = last_monday + timedelta(days=2)
+    target_iso = target_date.strftime("%Y-%m-%d")
+    target_day = target_date.day
+    expected_start_day = last_monday.day
+    expected_end_day = last_sunday.day
+    print(f"    上周范围: {expected_start} ~ {expected_end}")
+
+    week_selectors = [
+        '.index-realtime-datepicker .jmt-date-picker >> nth=1',
+        '.index-realtime-datepicker .jmt-date-picker:has-text("周")',
+        '.index-realtime-datepicker li[data-event-content="当前时间_周"]',
+        '.jmt-combo-date-picker .jmt-date-picker >> nth=1',
+        '.jmt-combo-date-picker .jmt-date-picker:has-text("周")',
+        'li[data-event-content="当前时间_周"]',
+    ]
+    week_tab = None
+    for selector in week_selectors:
+        try:
+            week_tab = page.locator(selector).first
+            week_tab.wait_for(state="visible", timeout=3000)
+            week_tab.scroll_into_view_if_needed(timeout=2000)
+            week_tab.hover(timeout=3000)
+            print("    [OK] 已悬停周筛选")
+            break
+        except Exception:
+            week_tab = None
+            continue
+    if week_tab is None:
+        save_date_picker_debug(page, "week_tab_not_found")
+        raise RuntimeError("未找到顶部周筛选")
+
+    page.evaluate(
+        """(el) => {
+            for (const node of [el, el.querySelector('li'), el.parentElement].filter(Boolean)) {
+                node.dispatchEvent(new MouseEvent('mouseenter', { bubbles: true, cancelable: true, view: window }));
+                node.dispatchEvent(new MouseEvent('mouseover', { bubbles: true, cancelable: true, view: window }));
+            }
+        }""",
+        week_tab.element_handle(),
+    )
+
+    panel_ready = page.wait_for_function(
+        """() => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                const before = window.getComputedStyle(el, '::before');
+                const after = window.getComputedStyle(el, '::after');
+                const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                    || (after.content && after.content !== 'none' && after.content !== '');
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && (rect.width > 0 || hasPseudo)
+                    && (rect.height > 0 || hasPseudo);
+            };
+            return Array.from(document.querySelectorAll(
+                '.jmt-base-popover, .jmt-popover, .jmt-date-picker-dropdown-wrapper, .jmt-date-picker-panels-wrapper'
+            )).some(el => visible(el) && /202\\d年\\d{1,2}月/.test(el.textContent || '') && /一\\s*二\\s*三\\s*四\\s*五\\s*六\\s*日/.test(el.textContent || ''));
+        }""",
+        timeout=6000,
+    )
+    if not panel_ready:
+        save_date_picker_debug(page, "week_panel_not_ready")
+        raise RuntimeError("悬停周后日期弹层未出现")
+
+    target_year = target_date.year
+    target_month = target_date.month
+    for _ in range(12):
+        month_nav_result = page.evaluate(
+            r"""({targetYear, targetMonth}) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    const before = window.getComputedStyle(el, '::before');
+                    const after = window.getComputedStyle(el, '::after');
+                    const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                        || (after.content && after.content !== 'none' && after.content !== '');
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && (rect.width > 0 || hasPseudo)
+                        && (rect.height > 0 || hasPseudo);
+                };
+                const roots = Array.from(document.querySelectorAll(
+                    '.jmt-base-popover, .jmt-popover, .jmt-date-picker-dropdown-wrapper, .jmt-date-picker-dropdown-content, .jmt-date-picker-panels-wrapper'
+                )).filter((el) => visible(el) && /20\d{2}年\d{1,2}月/.test(el.textContent || ''));
+                const root = roots
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        return { el, area: rect.width * rect.height };
+                    })
+                    .sort((a, b) => a.area - b.area)[0]?.el || document;
+                const text = root.textContent || '';
+                const match = text.match(/(20\d{2})年(\d{1,2})月/);
+                if (!match) return 'no_month_title';
+                const currentYear = Number(match[1]);
+                const currentMonth = Number(match[2]);
+                if (currentYear === targetYear && currentMonth === targetMonth) {
+                    return `aligned_${currentYear}_${currentMonth}`;
+                }
+                const diff = (targetYear - currentYear) * 12 + (targetMonth - currentMonth);
+                const selector = diff < 0
+                    ? '.jmt-date-picker-header-prev-month-icon'
+                    : '.jmt-date-picker-header-next-month-icon';
+                const btn = Array.from(root.querySelectorAll(selector)).find(visible)
+                    || Array.from(document.querySelectorAll(selector)).find(visible);
+                if (!btn) return `no_nav_${currentYear}_${currentMonth}_${diff}`;
+                btn.dispatchEvent(new MouseEvent('mousedown', { bubbles: true, cancelable: true, view: window }));
+                btn.dispatchEvent(new MouseEvent('mouseup', { bubbles: true, cancelable: true, view: window }));
+                btn.click();
+                return `clicked_${diff < 0 ? 'prev' : 'next'}_${currentYear}_${currentMonth}`;
+            }""",
+            {"targetYear": target_year, "targetMonth": target_month},
+        )
+        print(f"    日期面板月份定位: {month_nav_result}")
+        if str(month_nav_result).startswith("aligned_"):
+            break
+        if not str(month_nav_result).startswith("clicked_"):
+            break
+        page.wait_for_timeout(800)
+
+    click_result = page.evaluate(
+        r"""({expectedStart, expectedEnd, targetIso, targetDay, expectedStartDay, expectedEndDay}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                const before = window.getComputedStyle(el, '::before');
+                const after = window.getComputedStyle(el, '::after');
+                const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                    || (after.content && after.content !== 'none' && after.content !== '');
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && (rect.width > 0 || hasPseudo)
+                    && (rect.height > 0 || hasPseudo);
+            };
+            const disabled = (el) => {
+                if (!el) return false;
+                const classText = `${el.className || ''} ${el.parentElement?.className || ''}`;
+                return el.disabled
+                    || el.getAttribute('aria-disabled') === 'true'
+                    || /disabled|unavailable/.test(classText);
+            };
+            const clickEl = (el) => {
+                const target = el.closest('td, [role="gridcell"], li, button, a, div') || el;
+                target.dispatchEvent(new MouseEvent('mousedown', { bubbles: true, cancelable: true, view: window }));
+                target.dispatchEvent(new MouseEvent('mouseup', { bubbles: true, cancelable: true, view: window }));
+                target.click();
+            };
+
+            const candidates = Array.from(document.querySelectorAll(
+                '.jmt-base-popover, .jmt-popover, .jmt-date-picker-dropdown-wrapper, .jmt-date-picker-dropdown-content, .jmt-date-picker-panels-wrapper'
+            )).filter(el => {
+                const text = el.textContent || '';
+                return visible(el)
+                    && /202\d年\d{1,2}月/.test(text)
+                    && /一\s*二\s*三\s*四\s*五\s*六\s*日/.test(text);
+            });
+            const root = candidates
+                .map(el => {
+                    const rect = el.getBoundingClientRect();
+                    const text = (el.textContent || '').replace(/\s+/g, '');
+                    const hasWeekHeader = /一二三四五六日/.test(text);
+                    const hasGrid = !!el.querySelector('tr, [role="row"], td, [role="gridcell"]');
+                    const area = rect.width * rect.height;
+                    const score = (hasWeekHeader ? 4 : 0) + (hasGrid ? 2 : 0) + (area < 300000 ? 1 : 0);
+                    return { el, area, score };
+                })
+                .sort((a, b) => b.score - a.score || a.area - b.area)[0]?.el || document;
+
+            const norm = (s) => String(s || '')
+                .replace(/\s+/g, '')
+                .replace(/[./年月]/g, '-')
+                .replace(/日/g, '')
+                .replace(/[~～至到—–]/g, '-')
+                .replace(/-+/g, '-')
+                .replace(/(^|-)(0+)(\d)/g, '$1$3')
+                .replace(/-$/g, '');
+            const sm = (s) => norm(s);
+            const startShort = expectedStart.slice(5);
+            const endShort = expectedEnd.slice(5);
+            const startNoZero = sm(startShort);
+            const endNoZero = sm(endShort);
+            const startCN = `${parseInt(expectedStart.slice(5,7),10)}月${parseInt(expectedStart.slice(8,10),10)}日`;
+            const endCN = `${parseInt(expectedEnd.slice(5,7),10)}月${parseInt(expectedEnd.slice(8,10),10)}日`;
+            const expectedCompact = norm(expectedStart + '-' + expectedEnd);
+            const shortCompact = norm(startShort + '-' + endShort);
+
+            const weekPatterns = [
+                expectedStart + ' ~ ' + expectedEnd,
+                expectedStart + ' - ' + expectedEnd,
+                expectedStart.replaceAll('-', '.') + '-' + expectedEnd.replaceAll('-', '.'),
+                expectedStart.replaceAll('-', '/') + '-' + expectedEnd.replaceAll('-', '/'),
+                startShort + ' ~ ' + endShort,
+                startShort + ' - ' + endShort,
+                startShort.replace('-', '.') + '-' + endShort.replace('-', '.'),
+                startShort.replace('-', '/') + '-' + endShort.replace('-', '/'),
+                startNoZero + ' ~ ' + endNoZero,
+                startNoZero + ' - ' + endNoZero,
+                startCN + ' ~ ' + endCN,
+                startCN + ' - ' + endCN,
+            ];
+
+            const allCells = Array.from(root.querySelectorAll(
+                'td, [role="gridcell"], .ant-picker-cell, .el-date-table td, ' +
+                'button, a, span, div, li'
+            )).filter(el => visible(el));
+
+            const attrsText = (cell) => [
+                cell.getAttribute('title'),
+                cell.getAttribute('aria-label'),
+                cell.getAttribute('data-date'),
+                cell.getAttribute('data-value'),
+                cell.getAttribute('data-day'),
+                cell.parentElement?.getAttribute('title'),
+                cell.parentElement?.getAttribute('aria-label'),
+                cell.parentElement?.getAttribute('data-date'),
+                cell.parentElement?.getAttribute('data-value'),
+                cell.parentElement?.getAttribute('data-day'),
+            ].filter(Boolean).join(' ');
+            const atomicCells = allCells
+                .map(el => {
+                    const rect = el.getBoundingClientRect();
+                    const tag = el.tagName.toLowerCase();
+                    const hasDateAttr = ['title', 'aria-label', 'data-date', 'data-value', 'data-day']
+                        .some(name => el.hasAttribute(name) || el.parentElement?.hasAttribute(name));
+                    const looksAtomic = hasDateAttr
+                        || ['td', 'button', 'a', 'span', 'li'].includes(tag)
+                        || el.getAttribute('role') === 'gridcell'
+                        || /cell|date|day|week/i.test(`${el.className || ''} ${el.parentElement?.className || ''}`);
+                    return { el, area: rect.width * rect.height, textLen: (el.textContent || '').trim().length, looksAtomic };
+                })
+                .filter(item => item.looksAtomic && item.area > 0 && item.area < 30000 && item.textLen < 120)
+                .sort((a, b) => a.area - b.area)
+                .map(item => item.el);
+
+            const findByIso = (iso) => {
+                const isoAltDot = iso.replaceAll('-', '.');
+                const isoAltSlash = iso.replaceAll('-', '/');
+                const md = iso.slice(5);
+                const mdDot = md.replace('-', '.');
+                const mdSlash = md.replace('-', '/');
+                const month = parseInt(iso.slice(5, 7), 10);
+                const day = parseInt(iso.slice(8, 10), 10);
+                const cn = `${month}月${day}日`;
+                for (const cell of atomicCells) {
+                    if (disabled(cell)) continue;
+                    const haystack = `${attrsText(cell)} ${(cell.textContent || '').trim()}`;
+                    const normalized = norm(haystack);
+                    if (
+                        haystack.includes(iso) ||
+                        haystack.includes(isoAltDot) ||
+                        haystack.includes(isoAltSlash) ||
+                        haystack.includes(md) ||
+                        haystack.includes(mdDot) ||
+                        haystack.includes(mdSlash) ||
+                        haystack.includes(cn) ||
+                        normalized.includes(norm(iso)) ||
+                        normalized.includes(norm(md))
+                    ) {
+                        return cell;
+                    }
+                }
+                return null;
+            };
+
+            const targetCell = findByIso(targetIso);
+            if (targetCell) {
+                clickEl(targetCell);
+                return `clicked_full_date_${targetIso}`;
+            }
+
+            for (const cell of allCells) {
+                if (disabled(cell)) continue;
+                const txt = (cell.textContent || '').trim().replace(/\s+/g, ' ');
+                if (!txt) continue;
+                const normalized = norm(txt);
+                if (normalized.includes(expectedCompact) || normalized.includes(shortCompact)) {
+                    (cell.querySelector('button, a, span, div') || cell).click();
+                    return `clicked_normalized_week_range_${txt.slice(0, 80)}`;
+                }
+                for (const pat of weekPatterns) {
+                    if (txt.includes(pat)) {
+                        (cell.querySelector('button, a, span, div') || cell).click();
+                        return `clicked_week_range_${pat}`;
+                    }
+                }
+            }
+
+            const numberCells = Array.from(root.querySelectorAll(
+                '.ant-calendar-date, .ant-picker-cell-inner, .el-date-table td, td[role="gridcell"], button, span, div'
+            ));
+
+            const normalizeDayText = (value) => String(value || '').trim().replace(/^0+(\d)$/, '$1');
+            const dayText = String(targetDay);
+            const dayText2 = dayText.padStart(2, '0');
+            const startText = String(expectedStartDay);
+            const endText = String(expectedEndDay);
+            const rows = Array.from(root.querySelectorAll('tr, [role="row"]')).filter(visible);
+            for (const row of rows) {
+                const cells = Array.from(row.querySelectorAll('td[role="gridcell"], [role="gridcell"], .ant-picker-cell, .el-date-table td'))
+                    .filter(el => visible(el) && !disabled(el))
+                    .filter(el => {
+                        const rect = el.getBoundingClientRect();
+                        return rect.width >= 18 && rect.width <= 60 && rect.height >= 18 && rect.height <= 60;
+                    })
+                    .map(el => ({ el, txt: normalizeDayText(el.textContent), x: el.getBoundingClientRect().left }))
+                    .filter(item => /^\d{1,2}$/.test(item.txt));
+                const hasStart = cells.some(item => item.txt === startText);
+                const hasEnd = cells.some(item => item.txt === endText);
+                if (hasStart && hasEnd) {
+                    const target = cells.find(item => item.txt === dayText || item.txt === dayText2);
+                    if (target) {
+                        clickEl(target.el);
+                        return `clicked_day_in_last_week_row_${dayText}`;
+                    }
+                }
+            }
+            const debugTexts = Array.from(root.querySelectorAll('td, [role="gridcell"], button, a, span, div, li'))
+                .filter(visible)
+                .map(el => (el.textContent || '').trim().replace(/\s+/g, ' '))
+                .filter(Boolean)
+                .filter((txt, index, arr) => arr.indexOf(txt) === index)
+                .slice(0, 30);
+            return `not_found|${debugTexts.join(' || ')}`;
+        }""",
+        {
+            "expectedStart": expected_start,
+            "expectedEnd": expected_end,
+            "targetIso": target_iso,
+            "targetDay": target_day,
+            "expectedStartDay": expected_start_day,
+            "expectedEndDay": expected_end_day,
+        },
+    )
+    print(f"    日期选择结果: {click_result}")
+
+    if str(click_result).startswith("clicked_"):
+        for text in ["确定", "应用", "查询", "完成"]:
+            try:
+                btn = page.locator(f'text="{text}"').last
+                btn.wait_for(state="visible", timeout=1000)
+                btn.click(timeout=1500)
+                print(f"    [OK] 已点击日期确认按钮: {text}")
+                break
+            except Exception:
+                pass
+        page.wait_for_timeout(2500)
+
+    if str(click_result).startswith("not_found"):
+        print("    [WARN] 未在日期弹层中找到上周日期，尝试页面级兜底点击...")
+        fallback = page.evaluate(
+            r"""({expectedStart, expectedEnd}) => {
+                const norm = (s) => String(s || '')
+                    .replace(/\s+/g, '')
+                    .replace(/[./年月]/g, '-')
+                    .replace(/日/g, '')
+                    .replace(/[~～至到—–]/g, '-')
+                    .replace(/-+/g, '-')
+                    .replace(/(^|-)(0+)(\d)/g, '$1$3')
+                    .replace(/-$/g, '');
+                const expected = norm(expectedStart + '-' + expectedEnd);
+                const shortExpected = norm(expectedStart.slice(5) + '-' + expectedEnd.slice(5));
+                const els = document.querySelectorAll('td, [role="gridcell"], button, span, div, li');
+                for (const el of els) {
+                    const txt = (el.textContent || '').trim();
+                    const normalized = norm(txt);
+                    if (txt.includes(expectedStart) || normalized.includes(expected) || normalized.includes(shortExpected)) {
+                        el.click();
+                        return 'fallback_' + txt;
+                    }
+                }
+                return 'not_found';
+            }""",
+            {"expectedStart": expected_start, "expectedEnd": expected_end},
+        )
+        print(f"    兜底结果: {fallback}")
+        if fallback == "not_found":
+            save_date_picker_debug(page, "last_week_not_found")
+            if "|" in str(click_result):
+                print(f"    [DEBUG] 日期弹层候选文本: {str(click_result).split('|', 1)[1]}")
+            raise RuntimeError(f"未在日期弹层中找到上周日期范围: {expected_start} ~ {expected_end}")
+
+    try:
+        page.wait_for_function(
+            """() => !Array.from(document.querySelectorAll('.jmt-base-popover, .jmt-popover')).some(el => {
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none' && style.visibility !== 'hidden' && rect.width > 0 && rect.height > 0
+                    && /202\\d年\\d{1,2}月/.test(el.textContent || '');
+            })""",
+            timeout=5000,
+        )
+    except Exception:
+        pass
+    page.wait_for_timeout(5000)
+    print("  已切换到上周数据")
+
+
+def close_profit_plan_popup_if_present(page):
+    """Close the profit-plan promotion popup and its abandon-confirm dialog."""
+    def click_abandon_confirm():
+        try:
+            confirm = page.locator("text=\u786e\u8ba4\u653e\u5f03").last
+            confirm.wait_for(state="visible", timeout=1200)
+            confirm.click()
+            page.wait_for_timeout(1500)
+            print("  [OK] profit plan popup closed and abandoned")
+            return True
+        except Exception:
+            pass
+
+        try:
+            clicked_confirm = page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        const before = window.getComputedStyle(el, '::before');
+                        const after = window.getComputedStyle(el, '::after');
+                        const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                            || (after.content && after.content !== 'none' && after.content !== '');
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && (rect.width > 0 || hasPseudo)
+                            && (rect.height > 0 || hasPseudo);
+                    };
+                    const target = Array.from(document.querySelectorAll('button, span, div'))
+                        .filter((el) => visible(el) && (el.textContent || '').trim().includes('\u786e\u8ba4\u653e\u5f03'))
+                        .pop();
+                    if (!target) return false;
+                    (target.closest('button') || target).click();
+                    return true;
+                }"""
+            )
+            if clicked_confirm:
+                page.wait_for_timeout(1500)
+                print("  [OK] profit plan popup closed and abandoned")
+                return True
+        except Exception:
+            pass
+
+        return False
+
+    try:
+        close_point = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    const before = window.getComputedStyle(el, '::before');
+                    const after = window.getComputedStyle(el, '::after');
+                    const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                        || (after.content && after.content !== 'none' && after.content !== '');
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && (rect.width > 0 || hasPseudo)
+                        && (rect.height > 0 || hasPseudo);
+                };
+                const keywords = [
+                    '\u5f00\u542f\u7a33\u8d5a\u8ba1\u5212',
+                    '\u7a33\u8d5a\u8ba1\u5212',
+                    '100%\u4fdd\u6536\u76ca',
+                    '\u6295\u4ea7\u6bd4\u4e0d\u8fbe\u6807'
+                ];
+                const exactCloseSvg = Array.from(document.querySelectorAll('svg[viewBox="0 0 1024 1024"], svg[viewbox="0 0 1024 1024"]'))
+                    .filter(visible)
+                    .find((svg) => {
+                        const path = svg.querySelector('path');
+                        const d = path?.getAttribute('d') || '';
+                        const r = svg.getBoundingClientRect();
+                        return d.includes('M817.01 282.4')
+                            && d.includes('587.34 512')
+                            && r.top >= 0
+                            && r.left >= 0
+                            && r.top < window.innerHeight * 0.8
+                            && r.left < window.innerWidth;
+                    });
+                if (exactCloseSvg) {
+                    const clickable = exactCloseSvg.closest('button, span, i, div') || exactCloseSvg;
+                    const r = clickable.getBoundingClientRect();
+                    return {
+                        x: Math.round(r.left + r.width / 2),
+                        y: Math.round(r.top + r.height / 2),
+                        method: 'exact-close-svg'
+                    };
+                }
+
+                const centeredPopup = Array.from(document.querySelectorAll('body *'))
+                    .filter((el) => {
+                        if (!visible(el)) return false;
+                        const r = el.getBoundingClientRect();
+                        const style = window.getComputedStyle(el);
+                        const zIndex = Number.parseInt(style.zIndex, 10);
+                        const centered = r.left > window.innerWidth * 0.2
+                            && r.right < window.innerWidth * 0.85
+                            && r.top > 80
+                            && r.top < window.innerHeight * 0.5;
+                        const modalSize = r.width >= 450
+                            && r.width <= window.innerWidth * 0.75
+                            && r.height >= 260
+                            && r.height <= window.innerHeight * 0.8;
+                        return centered && modalSize && (!Number.isNaN(zIndex) || style.position === 'fixed' || style.position === 'absolute');
+                    })
+                    .map((el) => {
+                        const r = el.getBoundingClientRect();
+                        const style = window.getComputedStyle(el);
+                        const zIndex = Number.parseInt(style.zIndex, 10);
+                        return { el, r, score: (Number.isNaN(zIndex) ? 0 : zIndex * 1000000) + r.width * r.height };
+                    })
+                    .sort((a, b) => b.score - a.score)[0];
+                if (centeredPopup) {
+                    const r = centeredPopup.r;
+                    return [
+                        { x: Math.round(Math.min(r.right - 14, window.innerWidth - 5)), y: Math.round(Math.max(r.top + 14, 5)), method: 'centered-popup-corner' },
+                        { x: Math.round(Math.min(r.right - 18, window.innerWidth - 5)), y: Math.round(Math.max(r.top + 18, 5)), method: 'centered-popup-corner' },
+                        { x: Math.round(Math.min(r.right - 10, window.innerWidth - 5)), y: Math.round(Math.max(r.top + 10, 5)), method: 'centered-popup-corner' },
+                        { x: Math.round(Math.min(r.right - 26, window.innerWidth - 5)), y: Math.round(Math.max(r.top + 18, 5)), method: 'centered-popup-corner' },
+                        { x: Math.round(Math.min(r.right - 18, window.innerWidth - 5)), y: Math.round(Math.max(r.top + 26, 5)), method: 'centered-popup-corner' }
+                    ];
+                }
+
+                const roots = Array.from(document.querySelectorAll(
+                    '.jmt-portal, .jmt-dialog, .jmt-modal, .jmt-base-modal, ' +
+                    '[role="dialog"], [class*="dialog"], [class*="modal"]'
+                )).filter((el) => visible(el) && keywords.some((kw) => (el.textContent || '').includes(kw)));
+                const scoredRoots = roots.map((el) => {
+                    const r = el.getBoundingClientRect();
+                    const area = r.width * r.height;
+                    const looksLikeDialog = r.width < window.innerWidth * 0.9
+                        && r.height < window.innerHeight * 0.9
+                        && r.width > 400
+                        && r.height > 250;
+                    return { el, area, score: (looksLikeDialog ? 100000000 : 0) + area };
+                }).sort((a, b) => b.score - a.score);
+                const popup = scoredRoots[0]?.el;
+                if (!popup) return false;
+
+                const pr = popup.getBoundingClientRect();
+                const closeSvg = Array.from(popup.querySelectorAll('svg[viewBox="0 0 1024 1024"], svg[viewbox="0 0 1024 1024"]'))
+                    .filter(visible)
+                    .find((svg) => {
+                        const path = svg.querySelector('path');
+                        const d = path?.getAttribute('d') || '';
+                        return d.includes('M817.01 282.4') && d.includes('587.34 512');
+                    });
+                if (closeSvg) {
+                    const clickable = closeSvg.closest('button, span, i, div') || closeSvg;
+                    const r = clickable.getBoundingClientRect();
+                    return {
+                        x: Math.round(r.left + r.width / 2),
+                        y: Math.round(r.top + r.height / 2),
+                        method: 'close-svg'
+                    };
+                }
+
+                const closeables = Array.from(popup.querySelectorAll(
+                    'button, span, i, svg, div, [class*="close"], [aria-label*="\u5173\u95ed"], [title*="\u5173\u95ed"]'
+                )).filter(visible);
+                const scored = closeables.map((el) => {
+                    const r = el.getBoundingClientRect();
+                    const label = [
+                        el.getAttribute('class') || '',
+                        el.getAttribute('aria-label') || '',
+                        el.getAttribute('title') || '',
+                        el.textContent || ''
+                    ].join(' ');
+                    let score = 0;
+                    if (r.left >= pr.right - 120) score += 20;
+                    if (r.top <= pr.top + 100) score += 20;
+                    if (/close|\u5173\u95ed|icon|svg/i.test(label)) score += 8;
+                    if (r.width <= 80 && r.height <= 80) score += 5;
+                    return { el, score };
+                }).sort((a, b) => b.score - a.score);
+                const target = scored[0]?.el;
+                if (target && scored[0].score >= 20) {
+                    const r = target.getBoundingClientRect();
+                    return {
+                        x: Math.round(r.left + r.width / 2),
+                        y: Math.round(r.top + r.height / 2),
+                        method: 'element'
+                    };
+                }
+                return {
+                    x: Math.round(Math.min(pr.right - 24, window.innerWidth - 5)),
+                    y: Math.round(Math.max(pr.top + 24, 5)),
+                    method: 'corner'
+                };
+            }"""
+        )
+    except Exception:
+        close_point = None
+
+    if not close_point:
+        return False
+
+    close_points = close_point if isinstance(close_point, list) else [close_point]
+    clicked_any = False
+    for point in close_points:
+        try:
+            page.mouse.click(point["x"], point["y"])
+            clicked_any = True
+        except Exception:
+            try:
+                page.evaluate(
+                    """({x, y}) => {
+                        const el = document.elementFromPoint(x, y);
+                        if (el) (el.closest('button, span, i, div') || el).click();
+                    }""",
+                    point,
+                )
+                clicked_any = True
+            except Exception:
+                continue
+
+        page.wait_for_timeout(600)
+        if click_abandon_confirm():
+            return True
+
+    if not clicked_any:
+        return False
+
+    try:
+        confirm = page.locator("text=\u786e\u8ba4\u653e\u5f03").last
+        confirm.wait_for(state="visible", timeout=4000)
+        confirm.click()
+        page.wait_for_timeout(1500)
+        print("  [OK] profit plan popup closed and abandoned")
+        return True
+    except Exception:
+        try:
+            clicked_confirm = page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        const before = window.getComputedStyle(el, '::before');
+                        const after = window.getComputedStyle(el, '::after');
+                        const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                            || (after.content && after.content !== 'none' && after.content !== '');
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && (rect.width > 0 || hasPseudo)
+                            && (rect.height > 0 || hasPseudo);
+                    };
+                    const target = Array.from(document.querySelectorAll('button, span, div'))
+                        .filter((el) => visible(el) && (el.textContent || '').trim().includes('\u786e\u8ba4\u653e\u5f03'))
+                        .pop();
+                    if (!target) return false;
+                    (target.closest('button') || target).click();
+                    return true;
+                }"""
+            )
+            if clicked_confirm:
+                page.wait_for_timeout(1500)
+                print("  [OK] profit plan popup closed and abandoned")
+                return True
+        except Exception:
+            pass
+
+        try:
+            fallback_point = page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        const before = window.getComputedStyle(el, '::before');
+                        const after = window.getComputedStyle(el, '::after');
+                        const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                            || (after.content && after.content !== 'none' && after.content !== '');
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && (rect.width > 0 || hasPseudo)
+                            && (rect.height > 0 || hasPseudo);
+                    };
+                    const keywords = [
+                        '\u5f00\u542f\u7a33\u8d5a\u8ba1\u5212',
+                        '\u7a33\u8d5a\u8ba1\u5212',
+                        '100%\u4fdd\u6536\u76ca',
+                        '\u6295\u4ea7\u6bd4\u4e0d\u8fbe\u6807'
+                    ];
+                    const roots = Array.from(document.querySelectorAll(
+                        '.jmt-portal, .jmt-dialog, .jmt-modal, .jmt-base-modal, ' +
+                        '[role="dialog"], [class*="dialog"], [class*="modal"]'
+                    )).filter((el) => visible(el) && keywords.some((kw) => (el.textContent || '').includes(kw)));
+                    const popup = roots.map((el) => {
+                        const r = el.getBoundingClientRect();
+                        const area = r.width * r.height;
+                        const looksLikeDialog = r.width < window.innerWidth * 0.9
+                            && r.height < window.innerHeight * 0.9
+                            && r.width > 400
+                            && r.height > 250;
+                        return { el, area, score: (looksLikeDialog ? 100000000 : 0) + area };
+                    }).sort((a, b) => b.score - a.score)[0]?.el;
+                    if (!popup) return null;
+                    const r = popup.getBoundingClientRect();
+                    return {
+                        x: Math.round(Math.min(r.right - 24, window.innerWidth - 5)),
+                        y: Math.round(Math.max(r.top + 24, 5))
+                    };
+                }"""
+            )
+            if fallback_point:
+                page.mouse.click(fallback_point["x"], fallback_point["y"])
+                page.wait_for_timeout(800)
+                confirm = page.locator("text=\u786e\u8ba4\u653e\u5f03").last
+                confirm.wait_for(state="visible", timeout=3000)
+                confirm.click()
+                page.wait_for_timeout(1500)
+                print("  [OK] profit plan popup closed and abandoned")
+                return True
+        except Exception:
+            pass
+
+    try:
+        popup_still_visible = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    const before = window.getComputedStyle(el, '::before');
+                    const after = window.getComputedStyle(el, '::after');
+                    const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                        || (after.content && after.content !== 'none' && after.content !== '');
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && (rect.width > 0 || hasPseudo)
+                        && (rect.height > 0 || hasPseudo);
+                };
+                const keywords = [
+                    '\u5f00\u542f\u7a33\u8d5a\u8ba1\u5212',
+                    '\u7a33\u8d5a\u8ba1\u5212',
+                    '100%\u4fdd\u6536\u76ca',
+                    '\u6295\u4ea7\u6bd4\u4e0d\u8fbe\u6807'
+                ];
+                const hasTextPopup = Array.from(document.querySelectorAll('body *'))
+                    .some((el) => visible(el) && keywords.some((kw) => (el.textContent || '').includes(kw)));
+                const hasCloseSvg = Array.from(document.querySelectorAll('svg[viewBox="0 0 1024 1024"], svg[viewbox="0 0 1024 1024"]'))
+                    .filter(visible)
+                    .some((svg) => {
+                        const d = svg.querySelector('path')?.getAttribute('d') || '';
+                        return d.includes('M817.01 282.4') && d.includes('587.34 512');
+                    });
+                return hasTextPopup || hasCloseSvg;
+            }"""
+        )
+        if not popup_still_visible:
+            print("  [OK] profit plan popup dismissed")
+            return True
+    except Exception:
+        pass
+
+    print("  [WARN] profit plan popup close clicked, but abandon confirm was not found")
+    return True
+
+
+def close_popup_if_present(page):
+    """
+    检测并关闭广告弹窗（如果存在）
+    使用 try/except 避免弹窗未出现时报错
+    """
+    closed_profit_plan = False
+    for _ in range(4):
+        if close_profit_plan_popup_if_present(page):
+            closed_profit_plan = True
+            page.wait_for_timeout(800)
+            continue
+        break
+
+    close_btn = page.locator('body > div.jmt-portal.jmt-portal-css-1nrmhc.jmtv2.jmt-dialog-portal.jmt-dialog-hidden.jmt-dialog-hidden-css-grlj9t.jmtv2.win-notice-modal > div:nth-child(2) > div > div > div > div > span > span > svg')
+    try:
+        close_btn.wait_for(state="visible", timeout=2000 if closed_profit_plan else 5000)
+        close_btn.click()
+        print("  [OK] 广告弹窗已关闭")
+        page.wait_for_timeout(3000)
+    except Exception:
+        if closed_profit_plan:
+            return
+        print("  [INFO] 无广告弹窗，继续执行")
+
+
+# ── Step 2: 进入商智 ─────────────────────────────────────────────────────────
+
+def select_last_week_in_old_datepicker(page, label="old_datepicker"):
+    """Select one day from last week in the legacy JD date picker."""
+    last_monday, last_sunday = get_last_week_range()
+    target_date = last_monday + timedelta(days=2)
+    target_year = target_date.year
+    target_month = target_date.month
+    target_day = target_date.day
+
+    print(f"  旧版日期控件选择上周: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+
+    already_set = page.evaluate(
+        r"""({start, end}) => {
+            const text = document.body?.textContent || '';
+            const norm = (s) => s.replace(/[\s ]/g, '');
+            const target = `${norm(start)}至${norm(end)}`;
+            const target2 = `${start} ~ ${end}`;
+            const target3 = `${start}至${end}`;
+            return norm(text).includes(norm(`${start}至${end}`))
+                || norm(text).includes(norm(`${start} ~ ${end}`))
+                || norm(text).includes(norm(`${start} - ${end}`))
+                || norm(text).includes(norm(`${start}-${end}`));
+        }""",
+        {"start": last_monday.strftime('%Y-%m-%d'), "end": last_sunday.strftime('%Y-%m-%d')},
+    )
+    if already_set:
+        print(f"  [INFO] 日期筛选已为目标范围，跳过日期控件操作")
+        return True
+
+    for _ in range(12):
+        nav_info = page.evaluate(
+            r"""({targetYear, targetMonth}) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    const before = window.getComputedStyle(el, '::before');
+                    const after = window.getComputedStyle(el, '::after');
+                    const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                        || (after.content && after.content !== 'none' && after.content !== '');
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && (rect.width > 0 || hasPseudo)
+                        && (rect.height > 0 || hasPseudo);
+                };
+                const roots = Array.from(document.querySelectorAll(
+                    '.jmtd-popover-open, .jmtd-date-picker-panel, .jmtd-date-picker-week-panel, ' +
+                    '.dropdown-menu, .grace-datepicker, .grace-sz-datepicker, .uib-datepicker-popup, ' +
+                    '.ant-calendar-picker-container, .ant-picker-dropdown, .el-picker-panel, ' +
+                    '[class*="datepicker"], [class*="calendar"], [class*="date-picker"]'
+                )).filter((el) => visible(el) && /(20\d{2})[-/.年](\d{1,2})/.test(el.textContent || ''));
+                const root = roots
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        const hasCells = !!el.querySelector('td, [role="gridcell"], button, .ant-calendar-date');
+                        const hasHeader = !!el.querySelector(
+                            '.jmtd-date-picker-header-content, .jmt-date-picker-header, .ant-calendar-header, .ant-picker-header, .el-date-picker__header'
+                        );
+                        const isTrigger = el.classList.contains('jmtd-date-picker-combo')
+                            || el.classList.contains('jmtd-base-input')
+                            || el.classList.contains('jmt-combo-date-picker')
+                            || el.classList.contains('index-realtime-datepicker');
+                        return {
+                            el,
+                            area: rect.width * rect.height,
+                            score: (hasHeader ? 1000000000 : 0) + (hasCells ? 100000000 : 0) - (isTrigger ? 10000000000 : 0) - rect.width * rect.height
+                        };
+                    })
+                    .sort((a, b) => b.score - a.score)[0]?.el || document;
+                const text = root.textContent || '';
+                const match = text.match(/(20\d{2})[-/.年]\s*(\d{1,2})/);
+                if (!match) return { status: 'no_month_title', rootTag: root.tagName, rootClass: root.className?.substring?.(0,100) };
+                const currentYear = Number(match[1]);
+                const currentMonth = Number(match[2]);
+                if (currentYear === targetYear && currentMonth === targetMonth) {
+                    return { status: 'aligned', currentYear, currentMonth };
+                }
+                const diff = (targetYear - currentYear) * 12 + (targetMonth - currentMonth);
+                const preferred = diff < 0
+                    ? [
+                        '.jmtd-date-picker-header-btn-prev-month',
+                        '[data-event-content="prevMonth"]',
+                        '.jmt-date-picker-header-prev-month-icon',
+                        '.ant-calendar-prev-month-btn',
+                        '.ant-picker-header-prev-btn',
+                        '.el-picker-panel__icon-btn.prev-month',
+                        '.uib-left',
+                        '[ng-click*="move(-1"]',
+                        '[ng-click*="move(-1)"]',
+                        '[aria-label*="previous month" i]',
+                        '[aria-label*="上一月"]',
+                        '[title*="上一月"]'
+                    ]
+                    : [
+                        '.jmtd-date-picker-header-btn-next-month',
+                        '[data-event-content="nextMonth"]',
+                        '.jmt-date-picker-header-next-month-icon',
+                        '.ant-calendar-next-month-btn',
+                        '.ant-picker-header-next-btn',
+                        '.el-picker-panel__icon-btn.next-month',
+                        '.uib-right',
+                        '[ng-click*="move(1"]',
+                        '[ng-click*="move(1)"]',
+                        '[aria-label*="next month" i]',
+                        '[aria-label*="下一月"]',
+                        '[title*="下一月"]'
+                    ];
+                for (const selector of preferred) {
+                    let btn = Array.from(root.querySelectorAll(selector)).find(visible)
+                        || Array.from(document.querySelectorAll(selector)).find(visible);
+                    if (!btn) {
+                        // fallback: icon-only buttons may have zero rect height
+                        btn = Array.from(root.querySelectorAll(selector))[0]
+                            || Array.from(document.querySelectorAll(selector))[0];
+                    }
+                    if (btn) {
+                        return { status: 'click', selector, diff, currentYear, currentMonth };
+                    }
+                }
+                const rootRect = root.getBoundingClientRect();
+                const buttons = Array.from(root.querySelectorAll('button, a, span, i'))
+                    .filter(visible)
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        const label = [
+                            el.getAttribute('class') || '',
+                            el.getAttribute('title') || '',
+                            el.getAttribute('aria-label') || '',
+                            el.textContent || ''
+                        ].join(' ');
+                        return { el, rect, label };
+                    });
+                const candidates = buttons.filter(({rect, label}) => {
+                    const isLeft = rect.left < rootRect.left + 120;
+                    const isRight = rect.right > rootRect.right - 120;
+                    const isTop = rect.top < rootRect.top + 80;
+                    const looksMonthly = !/year|年/.test(label) || /month|月/.test(label);
+                    return isTop && looksMonthly && (diff < 0 ? isLeft : isRight);
+                }).sort((a, b) => diff < 0 ? b.rect.left - a.rect.left : a.rect.left - b.rect.left);
+                const btn = candidates[0];
+                if (!btn) return { status: 'no_nav', currentYear, currentMonth, diff };
+                return {
+                    status: 'corner_click',
+                    x: btn.rect.left + btn.rect.width / 2,
+                    y: btn.rect.top + btn.rect.height / 2,
+                    diff, currentYear, currentMonth
+                };
+            }""",
+            {"targetYear": target_year, "targetMonth": target_month},
+        )
+        print(f"    旧版日期面板月份定位: {nav_info}")
+        if not isinstance(nav_info, dict):
+            nav_info = {"status": str(nav_info)}
+        status = nav_info.get("status")
+        if status == "aligned":
+            break
+        if status == "click":
+            try:
+                sel = nav_info["selector"]
+                page.locator(
+                    f'.jmtd-date-picker-panel {sel}, .jmtd-date-picker-week-panel {sel}, .ant-picker-dropdown {sel}, .el-picker-panel {sel}, {sel}'
+                ).first.click(timeout=3000)
+            except Exception as e:
+                print(f"    [WARN] 月份导航按钮点击失败: {e}")
+                break
+        elif status == "corner_click":
+            try:
+                page.mouse.click(nav_info["x"], nav_info["y"])
+            except Exception as e:
+                print(f"    [WARN] 月份导航角落点击失败: {e}")
+                break
+        else:
+            page.evaluate(
+                r"""({targetMonth}) => {
+                    const monthCell = document.querySelector(
+                        `.jmtd-date-picker-month-cell:nth-child(${targetMonth}) .jmtd-date-picker-date-cell-content-inner, ` +
+                        `td.jmtd-date-picker-month-cell:nth-of-type(${targetMonth})`
+                    );
+                    if (monthCell) {
+                        monthCell.dispatchEvent(new MouseEvent('mousedown', { bubbles: true, cancelable: true, view: window }));
+                        monthCell.dispatchEvent(new MouseEvent('mouseup', { bubbles: true, cancelable: true, view: window }));
+                        monthCell.click();
+                        return true;
+                    }
+                    return false;
+                }""",
+                {"targetMonth": target_month},
+            )
+            page.wait_for_timeout(800)
+            break
+        page.wait_for_timeout(1000)
+
+    click_info = page.evaluate(
+        r"""({targetDay}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                const before = window.getComputedStyle(el, '::before');
+                const after = window.getComputedStyle(el, '::after');
+                const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                    || (after.content && after.content !== 'none' && after.content !== '');
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && (rect.width > 0 || hasPseudo)
+                    && (rect.height > 0 || hasPseudo);
+            };
+            const disabled = (el) => {
+                const classText = `${el.className || ''} ${el.parentElement?.className || ''}`;
+                return el.disabled
+                    || el.getAttribute('aria-disabled') === 'true'
+                    || /disabled|off|unavailable|diff-month|other-month|prev-month|next-month/.test(classText);
+            };
+            const roots = Array.from(document.querySelectorAll(
+                '.jmtd-popover-open, .jmtd-date-picker-panel, .jmtd-date-picker-week-panel, ' +
+                '.dropdown-menu, .grace-datepicker, .grace-sz-datepicker, .uib-datepicker-popup, ' +
+                '.ant-calendar-picker-container, .ant-picker-dropdown, .el-picker-panel, ' +
+                '[class*="datepicker"], [class*="calendar"], [class*="date-picker"]'
+            )).filter((el) => visible(el));
+            const root = roots
+                .map((el) => {
+                    const rect = el.getBoundingClientRect();
+                    const hasCells = !!el.querySelector('td, [role="gridcell"], button, .ant-calendar-date');
+                    const hasHeader = !!el.querySelector(
+                        '.jmtd-date-picker-header-content, .jmt-date-picker-header, .ant-calendar-header, .ant-picker-header, .el-date-picker__header'
+                    );
+                    return {
+                        el,
+                        area: rect.width * rect.height,
+                        score: (hasHeader ? 1000000000 : 0) + (hasCells ? 100000000 : 0) - rect.width * rect.height
+                    };
+                })
+                .sort((a, b) => b.score - a.score)[0]?.el || document;
+            const dayText = String(targetDay);
+            const dayText2 = dayText.padStart(2, '0');
+            const startText = String(targetDay - 2);
+            const endText = String(targetDay + 4);
+            const normalizeDay = (value) => String(value || '').trim().replace(/^0+(\d)$/, '$1');
+            const rows = Array.from(root.querySelectorAll('tr, [role="row"], .jmtd-date-picker-week-row, .jmt-date-picker-calendar-week-row'))
+                .filter(visible);
+            for (const row of rows) {
+                const rowCells = Array.from(row.querySelectorAll(
+                    'td[role="gridcell"], [role="gridcell"], .ant-calendar-date, .ant-picker-cell, .el-date-table td, td'
+                ))
+                    .filter(visible)
+                    .filter((cell) => !disabled(cell))
+                    .map((cell) => ({ cell, text: normalizeDay(cell.textContent) }))
+                    .filter((item) => /^\d{1,2}$/.test(item.text));
+                if (!rowCells.length) continue;
+                const hasStart = rowCells.some((item) => item.text === startText);
+                const hasEnd = rowCells.some((item) => item.text === endText);
+                if (hasStart && hasEnd) {
+                    const target = rowCells.find((item) => item.text === dayText || item.text === dayText2);
+                    if (target) {
+                        const tempId = '__jd_date_target_' + Date.now() + '_' + Math.random().toString(36).slice(2);
+                        target.cell.setAttribute('id', tempId);
+                        return { status: 'found_week_row', id: tempId, text: target.text };
+                    }
+                }
+            }
+
+            const cells = Array.from(root.querySelectorAll(
+                'td[role="gridcell"], [role="gridcell"], .ant-calendar-date, .ant-picker-cell, ' +
+                '.el-date-table td, td'
+            )).filter(visible);
+            for (const cell of cells) {
+                if (disabled(cell)) continue;
+                const text = (cell.textContent || '').trim();
+                if (text === dayText || text === dayText2) {
+                    const tempId = '__jd_date_target_' + Date.now() + '_' + Math.random().toString(36).slice(2);
+                    cell.setAttribute('id', tempId);
+                    return { status: 'found_cell', id: tempId, text };
+                }
+            }
+            return { status: 'not_found' };
+        }""",
+        {"targetDay": target_day},
+    )
+    print(f"    旧版日期选择结果: {click_info}")
+    if not isinstance(click_info, dict):
+        click_info = {"status": str(click_info)}
+    click_status = click_info.get("status")
+    if click_status in ("found_week_row", "found_cell"):
+        try:
+            page.locator(f"#{click_info['id']}").click(timeout=3000)
+        except Exception as e:
+            print(f"    [WARN] 日期单元格点击失败: {e}")
+            save_date_picker_debug(page, f"{label}_old_week_not_found")
+            raise RuntimeError(f"旧版日期控件未能选择上周日期: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+    else:
+        save_date_picker_debug(page, f"{label}_old_week_not_found")
+        raise RuntimeError(f"旧版日期控件未能选择上周日期: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+    page.wait_for_timeout(3000)
+    return True
+
+
+def step_enter_shangzhi(context: BrowserContext, page: Page):
+    """点击商智链接，切换到新标签页，返回旧版"""
+    print("[Step 2] 进入商智...")
+
+    if page.is_closed():
+        print("  [INFO] 原页面已关闭，重新打开商家后台首页...")
+        page = context.new_page()
+        page.goto("https://shop.jd.com/jdm/home", wait_until="domcontentloaded", timeout=60000)
+        page.wait_for_timeout(3000)
+    elif "shop.jd.com" not in page.url:
+        page.goto("https://shop.jd.com/jdm/home", wait_until="domcontentloaded", timeout=60000)
+        page.wait_for_timeout(3000)
+
+    last_error = None
+    sz_page = None
+    entry_selectors = (
+        '.shop-menu-navigate__menu-name:has-text("商智"):visible',
+        'a:has-text("商智"):visible',
+        '[role="menuitem"]:has-text("商智"):visible',
+    )
+
+    for attempt in range(1, 4):
+        dismiss_shop_home_overlays(page)
+
+        entry = None
+        for selector in entry_selectors:
+            try:
+                candidate = page.locator(selector).first
+                if candidate.is_visible(timeout=1500):
+                    entry = candidate
+                    break
+            except Exception:
+                continue
+        if entry is None:
+            last_error = RuntimeError("未找到可见的商智菜单入口")
+            page.wait_for_timeout(500)
+            continue
+
+        pages_before = list(context.pages)
+        try:
+            entry.click(timeout=6000)
+        except Exception as exc:
+            last_error = exc
+            print(f"  [WARN] 点击商智失败 ({attempt}/3): {exc}")
+            page.wait_for_timeout(500)
+            continue
+
+        observed_new_urls = set()
+        for _ in range(40):
+            try:
+                current_url = "" if page.is_closed() else page.url.lower()
+            except Exception:
+                current_url = ""
+            if _is_shangzhi_url(current_url):
+                sz_page = page
+                break
+
+            new_pages = [
+                candidate
+                for candidate in context.pages
+                if all(candidate is not old_page for old_page in pages_before)
+            ]
+            for candidate in reversed(new_pages):
+                try:
+                    candidate_url = candidate.url.lower()
+                except Exception:
+                    candidate_url = ""
+                if candidate_url:
+                    observed_new_urls.add(candidate_url)
+                if _is_shangzhi_url(candidate_url):
+                    sz_page = candidate
+                    break
+            if sz_page is not None:
+                break
+            try:
+                page.wait_for_timeout(250)
+            except Exception:
+                pass
+
+        if sz_page is not None:
+            break
+        observed = ", ".join(sorted(observed_new_urls)) or "无新页面"
+        last_error = RuntimeError(
+            f"点击商智后未检测到目标页面，观察到: {observed}"
+        )
+        print(f"  [WARN] 商智页面未打开 ({attempt}/3)，观察到: {observed}，准备重试")
+
+    if sz_page is None:
+        raise RuntimeError(f"进入商智失败，已重试 3 次: {last_error}") from last_error
+
+    try:
+        sz_page.wait_for_load_state("domcontentloaded", timeout=60000)
+    except Exception:
+        pass
+
+    # 先关闭广告弹窗（如果有）
+    close_popup_if_present(sz_page)
+
+    # 不返回旧版，直接在新页面采集
+    sz_page.wait_for_timeout(3000)
+    print("  已在商智页面")
+    return sz_page
+
+
+# ── Step 3: 采集店铺星级 ─────────────────────────────────────────────────────
+
+def step_shop_star_and_trade(page: Page):
+    """在商智新页面先选上周时间，然后采集店铺星级和交易数据"""
+    print("[Step 3] 采集店铺星级与交易数据...")
+
+    page.wait_for_timeout(5000)
+
+    # 先选择上周时间
+    # Close promotion overlays before locating the top week filter.
+    close_popup_if_present(page)
+    select_last_week_any_day(page)
+
+    # 抓取店铺星级（只提取星级数字，避免捕获多余文案）
+    star = extract(page, "document.querySelector('#jdsz-container .shop-info-left2 > div')?.textContent", "未获取到")
+    if star == "未获取到" or not star:
+        # 兜底：用更宽泛的选择器
+        star = extract(page, """() => {
+            const el = document.querySelector('.shop-info-left2 > div, [class*="shop-info"] [class*="left2"] > div');
+            return el?.textContent?.trim();
+        }""", "未获取到")
+    # 过滤：只保留星级数字部分（如 4.5星、5星）
+    star_match = re.search(r'\d+(?:\.\d+)?星', star)
+    if star_match:
+        star = star_match.group(0)
+    print(f"  店铺星级: {star}")
+
+    # 抓取交易数据（新版商智首页卡片式布局，含环比和同行均值）
+    transaction_raw = extract(page, """() => {
+        const cards = document.querySelectorAll('.jmt-summary-card-main');
+        const result = {};
+        for (const card of cards) {
+            const title = card.querySelector('.jmt-summary-card-title-left')?.textContent?.trim();
+            const value = card.querySelector('.jmt-summary-card-index-main-info')?.textContent?.trim();
+            if (!title || !value) continue;
+
+            const item = { value: value };
+
+            // 抓取环比、同行同级均值、同行上级均值
+            const descRows = card.querySelectorAll('.jmt-summary-card-description');
+            for (const row of descRows) {
+                const label = row.querySelector('.jmt-summary-card-description-left')?.textContent?.trim();
+                const val = row.querySelector('.jmt-summary-card-description-right')?.textContent?.trim();
+                if (label && val) {
+                    item[label] = val;
+                }
+            }
+
+            result[title] = item;
+        }
+        return JSON.stringify(result);
+    }""", "{}")
+
+    transaction = {}
+    if isinstance(transaction_raw, str):
+        try:
+            transaction = json.loads(transaction_raw)
+        except json.JSONDecodeError:
+            transaction = {}
+
+    # 确保关键字段存在
+    for key in ["成交金额", "商品访客数", "成交客户数", "成交转化率", "成交单量"]:
+        if key not in transaction:
+            transaction[key] = {"value": "N/A", "环比": "N/A", "同行同级均值": "N/A", "同行上级均值": "N/A"}
+
+    print(f"  交易数据: {json.dumps(transaction, ensure_ascii=False)}")
+
+    # 抓取完后点击返回旧版（供后续步骤使用）
+    print("  返回旧版...")
+    try:
+        back_button = page.locator('#header > div > div.page-top-header-tools > img')
+        back_button.wait_for(state="visible", timeout=5000)
+        back_button.click()
+        page.wait_for_timeout(5000)
+        page.wait_for_load_state("networkidle")
+        print("  已进入商智旧版")
+    except Exception:
+        print("  [INFO] 未找到返回旧版按钮，继续当前页面")
+
+    return star, transaction
+
+
+# ── Step 4: 交易概况-客单价（旧版商智）───────────────────────────────────────
+
+def step_deal_summary(page: Page):
+    """跳转交易概况页面，选择上周，采集客单价（数值+环比）"""
+    print("[Step 4] 采集交易概况-客单价...")
+
+    page.goto("https://sz.jd.com/sz/view/dealAnalysis/dealSummarys.html",
+              wait_until="networkidle")
+    page.wait_for_timeout(5000)
+
+    # 选择日期为上周
+    last_monday, last_sunday = get_last_week_range()
+    print(f"  选择上周: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+
+    try:
+        # 1. 点击日期选择器打开面板
+        page.locator('input[grace-sz-datepicker-popup]').click()
+        page.wait_for_timeout(2000)
+
+        # 2. 点击"按周查询"
+        page.locator('.quick-type-list li:has-text("按周查询")').click()
+        page.wait_for_timeout(3000)
+
+        select_last_week_in_old_datepicker(page, "deal_summary")
+    except Exception as e:
+        print(f"  [WARN] 日期选择失败: {e}")
+
+    # 提取客单价（数值 + 环比）
+    customer_price_raw = extract(page, """() => {
+        // 在 grace-summary 列表中找到"客单价"对应的 summary-item
+        const items = document.querySelectorAll('.grace-summary .summary-item, .summary-list .summary-item');
+        for (const item of items) {
+            const titleEl = item.querySelector('.summary-title label, h6 label, .ng-binding');
+            if (titleEl && titleEl.textContent.trim() === '客单价') {
+                // 数值：.summary-body 里的 .odometer
+                const unitEl = item.querySelector('.summary-body .unit');
+                const numEl = item.querySelector('.summary-body .odometer');
+                const value = (unitEl ? unitEl.textContent.trim() : '') + (numEl ? numEl.textContent.trim() : '');
+                // 环比：.summary-detail 里的 .detail-value
+                const changeEl = item.querySelector('.summary-detail .detail-value');
+                const change = changeEl ? changeEl.textContent.trim() : '';
+                return JSON.stringify({value: value, change: change});
+            }
+        }
+        return JSON.stringify({value: '', change: ''});
+    }""", '{"value": "", "change": ""}')
+
+    cp_data = {"value": "N/A", "change": ""}
+    if isinstance(customer_price_raw, str):
+        try:
+            cp_data = json.loads(customer_price_raw)
+        except json.JSONDecodeError:
+            cp_data = {"value": "N/A", "change": ""}
+
+    cp_val = cp_data.get('value', 'N/A')
+    cp_change = cp_data.get('change', '')
+    try:
+        print(f"  客单价: {cp_val}, 环比: {cp_change}")
+    except UnicodeEncodeError:
+        print(f"  客单价: {json.dumps(cp_val, ensure_ascii=True)}, 环比: {json.dumps(cp_change, ensure_ascii=True)}")
+    return {"客单价": {"value": cp_data.get("value", "N/A"), "change": cp_data.get("change", "")}}
+
+
+# ── Step 4-old: 交易概况（按周查询）────────────────────────────────────────────
+
+def step_transaction_overview(page: Page):
+    """切换按周查询，选择上周，采集交易数据"""
+    print("[Step 4-old] 采集交易概况（上周）...")
+
+    last_monday, last_sunday = get_last_week_range()
+    print(f"  上周范围: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+
+    # 点击日期选择器
+    page.click('input[placeholder*="时间"]')
+    page.wait_for_timeout(3000)
+
+    # 切换到按周查询
+    safe_click(page, "按周查询")
+    page.wait_for_timeout(3000)
+
+    # 在日期选择器中找到并点击上周
+    # 计算上周的日期文本用于匹配
+    last_week_day = last_monday.day
+    # 尝试点击上周一对应的日期数字
+    clicked = page.evaluate(f"""() => {{
+        // 查找日期选择器中所有可点击的日期单元格
+        const cells = document.querySelectorAll('.ant-calendar-date, .el-date-table td, td[role="gridcell"]');
+        for (const cell of cells) {{
+            const text = cell.textContent?.trim();
+            if (text === '{last_week_day}' || text === '{last_week_day:02d}') {{
+                cell.click();
+                return true;
+            }}
+        }}
+        return false;
+    }}""")
+
+    if not clicked:
+        print("  [!] 自动选择上周失败，尝试 snapshot 辅助定位...")
+        # 如果自动选择失败，需要人工介入或使用更复杂的逻辑
+        page.wait_for_timeout(3000)
+
+    page.wait_for_timeout(5000)
+    wait_for_navigation(page)
+
+    # 提取交易数据
+    data = {}
+    metrics = {
+        "成交金额": ".deal-amount",
+        "访客数": ".visitor-count",
+        "成交转化率": ".conversion-rate",
+        "成交单量": ".order-count",
+        "客单价": ".customer-price",
+    }
+
+    for name, selector in metrics.items():
+        val = extract(page, f"document.querySelector('{selector}')?.textContent")
+        data[name] = val
+
+    # 客单价兜底：成交金额 / 成交单量
+    if not data.get("客单价"):
+        data["客单价"] = extract(page, """() => {
+            const amount = parseFloat(document.querySelector('.deal-amount')?.textContent?.replace(/,/g, '') || '0');
+            const orders = parseFloat(document.querySelector('.order-count')?.textContent?.replace(/,/g, '') || '0');
+            return orders > 0 ? (amount / orders).toFixed(2) : 'N/A';
+        }""")
+
+    print(f"  交易数据: {json.dumps(data, ensure_ascii=False)}")
+    return data
+
+
+# ── Step 5: 商家榜单 ─────────────────────────────────────────────────────────
+
+def switch_category(page: Page, category_name: str, parent_category: str = None):
+    """切换商家榜单类目，支持级联下拉（hover父类目展开子类目后再点击）"""
+    label = f"{parent_category} → {category_name}" if parent_category else category_name
+    print(f"  切换类目到: {label}")
+    try:
+        # 1. 点击当前类目下拉框，打开下拉列表
+        page.evaluate("""() => {
+            const el = document.querySelector('.jmtd-base-input-content, .jmtd-select-input');
+            if (el) el.click();
+        }""")
+        page.wait_for_timeout(2000)
+
+        # 2. 悬停父类目以展开子类目列表
+        if parent_category:
+            try:
+                parent_locator = page.locator('.jmtd-dropdown-content .jmtd-label', has_text=parent_category).first
+                parent_locator.hover(timeout=3000)
+                page.wait_for_timeout(2000)  # 等待子类目加载
+            except Exception as e:
+                print(f"    [WARN] 悬停父类目失败: {e}")
+
+        # 3. 点击目标子类目
+        try:
+            child_locator = page.locator('.jmtd-dropdown-content .jmtd-label', has_text=category_name).first
+            child_locator.click(timeout=3000)
+            page.wait_for_timeout(5000)
+            print(f"    [OK] 已切换到 {label}")
+        except Exception as e:
+            print(f"    [WARN] 点击子类目失败: {e}")
+    except Exception as e:
+        print(f"    [WARN] 切换类目失败: {e}")
+
+
+def extract_ranking_table(page):
+    """提取商家榜单表格数据"""
+    ranking_data = extract(page, """() => {
+        const rows = document.querySelectorAll('.jmtd-table-main tbody tr');
+        const result = [];
+        let myRank = null;
+        for (const row of rows) {
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 7) continue;
+
+            const rankEl = cells[0].querySelector('.jmtd-table-formatter-rank-number');
+            const rank = rankEl ? rankEl.textContent.trim() : cells[0].textContent.trim();
+
+            const shopLink = cells[1].querySelector('a');
+            const shop = shopLink ? shopLink.textContent.trim() : cells[1].textContent.trim();
+
+            const isMyShop = cells[1].querySelector('.jmti-star-f') !== null;
+
+            const item = {
+                rank: rank,
+                shop: shop,
+                amount: cells[2]?.textContent?.trim() || '',
+                orders: cells[3]?.textContent?.trim() || '',
+                followers: cells[4]?.textContent?.trim() || '',
+                visitors: cells[5]?.textContent?.trim() || '',
+                searchClicks: cells[6]?.textContent?.trim() || '',
+                isMyShop: isMyShop
+            };
+
+            result.push(item);
+            if (isMyShop) {
+                myRank = item;
+            }
+        }
+
+        // 提取排名环比
+        const rankChangeEl = document.querySelector('.rank-change');
+        let rankChange = '';
+        if (rankChangeEl) {
+            const arrow = rankChangeEl.querySelector('.jmti-arrow-down, .jmti-arrow-up');
+            const arrowClass = arrow ? arrow.className : '';
+            const direction = arrowClass.includes('arrow-down') ? '下降' : (arrowClass.includes('arrow-up') ? '上升' : '');
+            const match = rankChangeEl.textContent.match(/排名环比[：:]?\\D*(\\d+)/);
+            const num = match ? match[1] : '';
+            if (direction && num) {
+                rankChange = direction + num + '位';
+            }
+        }
+
+        return JSON.stringify({ rows: result, myRank: myRank, rankChange: rankChange });
+    }""", "{}")
+
+    ranking = {"rows": [], "myRank": None}
+    if isinstance(ranking_data, str):
+        try:
+            ranking = json.loads(ranking_data)
+        except json.JSONDecodeError:
+            ranking = {"rows": [], "myRank": None}
+
+    myRank = ranking.get("myRank") or {}
+    allRows = ranking.get("rows", [])
+
+    print(f"  本店排名: {myRank.get('rank', 'N/A')}, 商家: {myRank.get('shop', 'N/A')}")
+    print(f"  共抓取商家数: {len(allRows)}")
+
+    return {
+        "rank": myRank.get("rank", "N/A") if myRank else "N/A",
+        "shop": myRank.get("shop", "N/A") if myRank else "N/A",
+        "amount": myRank.get("amount", "N/A") if myRank else "N/A",
+        "orders": myRank.get("orders", "N/A") if myRank else "N/A",
+        "followers": myRank.get("followers", "N/A") if myRank else "N/A",
+        "visitors": myRank.get("visitors", "N/A") if myRank else "N/A",
+        "searchClicks": myRank.get("searchClicks", "N/A") if myRank else "N/A",
+        "rankChange": ranking.get("rankChange", ""),
+        "all_rows": allRows,
+    }
+
+
+def step_merchant_ranking(page: Page, industry: str):
+    """采集行业商家榜单（新版商智表格），支持多类目"""
+    print("[Step 5] 采集商家榜单...")
+
+    page.goto("https://sz.jd.com/szweb/sz/view/industryMarket/venderRanks.html",
+              wait_until="networkidle")
+    page.wait_for_timeout(8000)
+
+    # 确认行业分类
+    current_industry = extract(page, """() => {
+        const el = document.querySelector('.industry-name, .category-current, [class*="industry"]');
+        return el?.textContent?.trim();
+    }""")
+
+    if current_industry and industry not in current_industry:
+        safe_click(page, industry)
+        page.wait_for_timeout(3000)
+
+    # 切换到按周查询
+    safe_click(page, "按天查询")
+    page.wait_for_timeout(3000)
+    safe_click(page, "按周查询")
+    page.wait_for_timeout(5000)
+
+    select_last_week_in_old_datepicker(page, "merchant_ranking")
+    page.wait_for_timeout(8000)
+
+    # 采集默认类目（男包）
+    results = {}
+    results["男包"] = extract_ranking_table(page)
+
+    # 采集其他子类目（级联下拉：先点父类目展开，再点子类目）
+    extra_categories = [
+        ("男士双肩包", "男包"),
+        ("男士单肩/斜挎包", "男包"),
+        ("电脑包", "功能箱包"),
+    ]
+    for cat, parent in extra_categories:
+        switch_category(page, cat, parent)
+        # 切换类目后可能需要重新选择上周
+        page.wait_for_timeout(3000)
+        safe_click(page, "按周查询")
+        page.wait_for_timeout(2500)
+        select_last_week_in_old_datepicker(page, f"merchant_ranking_{cat}")
+        page.wait_for_timeout(8000)
+        results[cat] = extract_ranking_table(page)
+
+    return results
+
+
+# ── Step 6: 京准通-非全站营销 ─────────────────────────────────────────────────
+
+def switch_account_filter(page):
+    """取消勾选'自有账户'，勾选'授权账户'，确定"""
+    print("  配置账号筛选...")
+
+    # 1. 点击已选账号区域，打开筛选弹窗
+    try:
+        page.evaluate("""() => {
+            const el = document.querySelector('.checked-filters, .filter-index-item');
+            if (el) el.click();
+        }""")
+        page.wait_for_timeout(2000)
+    except Exception as e:
+        print(f"    [WARN] 点击账号筛选失败: {e}")
+        return
+
+    # 2. 取消勾选"自有账户"
+    try:
+        page.evaluate("""() => {
+            const wrappers = document.querySelectorAll('.jad-checkbox-wrapper, label');
+            for (const w of wrappers) {
+                if (w.textContent.includes('\u81ea\u6709\u8d26\u6237')) {
+                    const cb = w.querySelector('input[type="checkbox"]');
+                    if (cb && cb.checked) cb.click();
+                }
+            }
+        }""")
+        print("    [OK] 已取消勾选'自有账户'")
+        page.wait_for_timeout(500)
+    except Exception as e:
+        print(f"    [WARN] 取消自有账户失败: {e}")
+
+    # 3. 勾选"授权账户"
+    try:
+        page.evaluate("""() => {
+            const wrappers = document.querySelectorAll('.jad-checkbox-wrapper, label');
+            for (const w of wrappers) {
+                if (w.textContent.includes('\u6388\u6743\u8d26\u6237')) {
+                    const cb = w.querySelector('input[type="checkbox"]');
+                    if (cb && !cb.checked) cb.click();
+                }
+            }
+        }""")
+        print("    [OK] 已勾选'授权账户'")
+        page.wait_for_timeout(500)
+    except Exception as e:
+        print(f"    [WARN] 勾选授权账户失败: {e}")
+
+    # 4. 点击弹窗内的确定按钮（.filter-bottom 里的 jad-btn-primary）
+    try:
+        page.evaluate("""() => {
+            const btn = document.querySelector('.filter-bottom button.jad-btn-primary');
+            if (btn) btn.click();
+        }""")
+        print("    [OK] 已点击弹窗确定按钮")
+        page.wait_for_timeout(3000)
+    except Exception as e:
+        print(f"    [WARN] 点击弹窗确定失败: {e}")
+        page.keyboard.press("Escape")
+        page.wait_for_timeout(1000)
+
+
+def step_jzt_non_full_site(page: Page, caliber_days: int):
+    """采集京准通非全站营销数据"""
+    print("[Step 6] 采集京准通-非全站营销...")
+
+    page.goto("https://jzt.jd.com/custom-report/#/overview", wait_until="domcontentloaded")
+    page.wait_for_timeout(8000)
+
+    # 点击非全站营销标签
+    safe_click(page, "非全站营销")
+    page.wait_for_timeout(3000)
+
+    # 配置账号筛选
+    switch_account_filter(page)
+
+    # 配置口径：设置为7天
+    cal_btn = page.query_selector('button:has-text("天")')
+    if cal_btn:
+        cal_btn.click()
+        page.wait_for_timeout(2000)
+
+    page.evaluate(f"""() => {{
+        const radios = document.querySelectorAll('.jad-radio-wrapper');
+        for (const r of radios) {{
+            if (r.textContent.trim() === '{caliber_days}天') {{
+                r.click();
+                return;
+            }}
+        }}
+    }}""")
+    page.wait_for_timeout(1000)
+    page.keyboard.press("Escape")
+    page.wait_for_timeout(1000)
+
+    # 点击页面上的"查询"按钮
+    try:
+        page.evaluate("""() => {
+            const btns = document.querySelectorAll('button.jad-btn-primary');
+            for (const b of btns) {
+                if (b.textContent.trim() === '\u67e5\u8be2') {
+                    b.click();
+                    return true;
+                }
+            }
+            return false;
+        }""")
+        print("  [OK] 已点击查询按钮")
+        page.wait_for_timeout(10000)
+    except Exception as e:
+        print(f"  [WARN] 点击查询失败: {e}")
+
+    # 提取数据：花费(cost)、投产比(ROI)、点击率(CTR) + 涨跌对比
+    data = extract(page, """() => {
+        const result = {};
+        const cards = document.querySelectorAll('.cards li, .cards .checked');
+        for (const card of cards) {
+            const label = card.querySelector('.jad-checkbox-label span')?.textContent?.trim();
+            const value = card.querySelector('.current-number')?.textContent?.trim();
+            const siblings = card.querySelectorAll('.current-number ~ span');
+            let change = '';
+            for (const s of siblings) {
+                const txt = s.textContent.trim();
+                if (txt && txt !== '') { change = txt; break; }
+            }
+            const compare = card.querySelector('.compare-number')?.textContent?.trim();
+            if (label && value) {
+                result[label] = { value: value, change: change, compare: compare };
+            }
+        }
+        return JSON.stringify(result);
+    }""", "{}")
+
+    if isinstance(data, str):
+        try:
+            data = json.loads(data)
+        except json.JSONDecodeError:
+            data = {}
+
+    # 兜底：根据 checkbox value 直接找卡片
+    for key, val_attr in [('cost', 'cost'), ('ROI', 'ROI'), ('CTR', 'CTR')]:
+        if not data.get(key):
+            raw = extract(page, f"""() => {{
+                const cards = document.querySelectorAll('.cards li');
+                for (const card of cards) {{
+                    const input = card.querySelector('input[value="{val_attr}"]');
+                    if (input) {{
+                        const value = card.querySelector('.current-number')?.textContent?.trim() || 'N/A';
+                        const siblings = card.querySelectorAll('.current-number ~ span');
+                        let change = '';
+                        for (const s of siblings) {{
+                            const txt = s.textContent.trim();
+                            if (txt && txt !== '') {{ change = txt; break; }}
+                        }}
+                        const compare = card.querySelector('.compare-number')?.textContent?.trim() || '';
+                        return JSON.stringify({{value: value, change: change, compare: compare}});
+                    }}
+                }}
+                return JSON.stringify({{value: 'N/A', change: '', compare: ''}});
+            }}""")
+            try:
+                data[key] = json.loads(raw) if isinstance(raw, str) else {"value": "N/A", "change": "", "compare": ""}
+            except:
+                data[key] = {"value": "N/A", "change": "", "compare": ""}
+
+    try:
+        print(f"  京准通数据: {json.dumps(data, ensure_ascii=False)}")
+    except UnicodeEncodeError:
+        print(f"  京准通数据: {json.dumps(data, ensure_ascii=True)}")
+    return data
+
+
+# ── Step 7: 京准通-全站营销 ──────────────────────────────────────────────────
+
+def step_jzt_full_site(page: Page, caliber_days: int):
+    """采集京准通全站营销数据（流程与非全站相同）"""
+    print("[Step 7] 采集京准通-全站营销...")
+
+    # 点击全站营销标签（精确选择器）
+    try:
+        page.evaluate("""() => {
+            const tabs = document.querySelectorAll('.jad-tabs-nav-tab-no-icon');
+            for (const tab of tabs) {
+                if (tab.textContent.trim() === '\u5168\u7ad9\u8425\u9500') {
+                    tab.click();
+                    return true;
+                }
+            }
+            return false;
+        }""")
+        print("  [OK] 已点击全站营销标签")
+    except Exception as e:
+        print(f"  [WARN] 点击全站营销标签失败: {e}")
+    page.wait_for_timeout(8000)
+
+    # 配置账号筛选（确保操作当前可见的弹窗）
+    switch_account_filter(page)
+
+    # 配置口径：点击口径下拉，选7天
+    try:
+        page.evaluate("""() => {
+            const btn = document.querySelector('button.jad-btn-default');
+            if (btn && btn.textContent.includes('\u5929')) btn.click();
+        }""")
+        page.wait_for_timeout(2000)
+        page.evaluate(f"""() => {{
+            const radios = document.querySelectorAll('.jad-radio-wrapper');
+            for (const r of radios) {{
+                if (r.textContent.trim() === '{caliber_days}\u5929') {{
+                    r.click();
+                    return;
+                }}
+            }}
+        }}""")
+        page.wait_for_timeout(1000)
+        page.keyboard.press("Escape")
+        page.wait_for_timeout(1000)
+    except Exception:
+        pass
+
+    # 点击页面上的"查询"按钮
+    try:
+        page.evaluate("""() => {
+            const btns = document.querySelectorAll('button.jad-btn-primary');
+            for (const b of btns) {
+                if (b.textContent.trim() === '\u67e5\u8be2') {
+                    b.click();
+                    return true;
+                }
+            }
+            return false;
+        }""")
+        print("  [OK] 已点击查询按钮")
+        page.wait_for_timeout(10000)
+    except Exception as e:
+        print(f"  [WARN] 点击查询失败: {e}")
+
+    # 提取数据：花费(cost)、投产比(ROI)、点击率(CTR) + 对比涨跌
+    data = extract(page, """() => {
+        const result = {};
+        const cards = document.querySelectorAll('.cards li, .cards .checked');
+        for (const card of cards) {
+            const label = card.querySelector('.jad-checkbox-label span')?.textContent?.trim();
+            const value = card.querySelector('.current-number')?.textContent?.trim();
+            // 涨跌：current-number 后面的兄弟 span
+            const siblings = card.querySelectorAll('.current-number ~ span');
+            let change = '';
+            for (const s of siblings) {
+                const txt = s.textContent.trim();
+                if (txt && txt !== '') { change = txt; break; }
+            }
+            // 对比数值
+            const compare = card.querySelector('.compare-number')?.textContent?.trim();
+            if (label && value) {
+                result[label] = { value: value, change: change, compare: compare };
+            }
+        }
+        return JSON.stringify(result);
+    }""", "{}")
+
+    if isinstance(data, str):
+        try:
+            data = json.loads(data)
+        except json.JSONDecodeError:
+            data = {}
+
+    # 兜底：根据 checkbox value 直接找卡片（全站指标：cost, swaClicks, swaOrderROI, swaOrderSum）
+    for key, val_attr in [('cost', 'cost'), ('swaClicks', 'swaClicks'), ('swaOrderROI', 'swaOrderROI'), ('swaOrderSum', 'swaOrderSum')]:
+        if not data.get(key):
+            raw = extract(page, f"""() => {{
+                const cards = document.querySelectorAll('.cards li');
+                for (const card of cards) {{
+                    const input = card.querySelector('input[value="{val_attr}"]');
+                    if (input) {{
+                        const value = card.querySelector('.current-number')?.textContent?.trim() || 'N/A';
+                        const siblings = card.querySelectorAll('.current-number ~ span');
+                        let change = '';
+                        for (const s of siblings) {{
+                            const txt = s.textContent.trim();
+                            if (txt && txt !== '') {{ change = txt; break; }}
+                        }}
+                        const compare = card.querySelector('.compare-number')?.textContent?.trim() || '';
+                        return JSON.stringify({{value: value, change: change, compare: compare}});
+                    }}
+                }}
+                return JSON.stringify({{value: 'N/A', change: '', compare: ''}});
+            }}""")
+            try:
+                data[key] = json.loads(raw) if isinstance(raw, str) else {"value": "N/A", "change": "", "compare": ""}
+            except:
+                data[key] = {"value": "N/A", "change": "", "compare": ""}
+
+    # 字段映射：swaOrderROI -> ROI，swaClicks -> 全站点击量，swaOrderSum -> 全站交易额
+    if data.get('swaOrderROI'):
+        data['ROI'] = data['swaOrderROI']
+    if data.get('swaClicks'):
+        data['全站点击量'] = data['swaClicks']
+    if data.get('swaOrderSum'):
+        data['全站交易额'] = data['swaOrderSum']
+
+    try:
+        print(f"  全站营销数据: {json.dumps(data, ensure_ascii=False)}")
+    except UnicodeEncodeError:
+        print(f"  全站营销数据: {json.dumps(data, ensure_ascii=True)}")
+    return data
+
+
+# ── Step 8: 生成报告 ─────────────────────────────────────────────────────────
+
+def generate_report(config, star, transaction, ranking, jzt_non_full, jzt_full):
+    """生成 Markdown 汇总报告"""
+    print("[Step 8] 生成汇总报告...")
+
+    shop = config["shop"]
+    industry = config["industry"]
+    caliber_days = config["caliber_days"]
+    output_dir = config["output"]
+    today = datetime.now().strftime("%Y-%m-%d")
+    last_monday, last_sunday = get_last_week_range()
+    week_range = f"{last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}"
+
+    # 构建交易表格行（含环比、同行均值）
+    trans_rows = ""
+    for name in ["成交金额", "商品访客数", "成交客户数", "成交转化率", "成交单量", "客单价"]:
+        item = transaction.get(name, {})
+        val = item.get("value", "N/A") if isinstance(item, dict) else item
+        hb = item.get("环比", "N/A") if isinstance(item, dict) else "N/A"
+        same = item.get("同行同级均值", "N/A") if isinstance(item, dict) else "N/A"
+        upper = item.get("同行上级均值", "N/A") if isinstance(item, dict) else "N/A"
+        trans_rows += f"| {name} | {val} | {hb} | {same} | {upper} |\n"
+
+    # 构建商家榜单对比表格（支持多类目）
+    def build_ranking_section(cat_name, cat_data):
+        if not cat_data:
+            return f"### 类目：{cat_name}\n\n暂无数据。\n\n"
+        rows = ""
+        for item in cat_data.get("all_rows", [])[:20]:  # 取前20名
+            is_my = item.get("isMyShop", False)
+            shop_name = item.get("shop", "")
+            marker = " ⭐" if is_my else ""
+            rows += (f"| {item.get('rank', '')} | {shop_name}{marker} "
+                     f"| {item.get('amount', '')} | {item.get('orders', '')} "
+                     f"| {item.get('followers', '')} | {item.get('visitors', '')} "
+                     f"| {item.get('searchClicks', '')} |\n")
+        rank_change = cat_data.get('rankChange', '')
+        rank_change_str = f" | 排名环比：{rank_change}" if rank_change else ""
+        return (
+            f"### 类目：{cat_name}\n\n"
+            f"**本店排名：** 第{cat_data.get('rank', 'N/A')}名 — {cat_data.get('shop', 'N/A')}{rank_change_str}\n"
+            f"- 成交金额：{cat_data.get('amount', 'N/A')} | 成交单量：{cat_data.get('orders', 'N/A')} | 关注人数：{cat_data.get('followers', 'N/A')} | 访客数：{cat_data.get('visitors', 'N/A')} | 搜索点击：{cat_data.get('searchClicks', 'N/A')}\n\n"
+            f"| 排名 | 商家 | 成交金额 | 成交单量 | 关注人数 | 访客数 | 搜索点击 |\n"
+            f"|------|------|----------|----------|----------|--------|----------|\n"
+            f"{rows}\n"
+        )
+
+    # 非全站营销表格（cost=付费占比, ROI=ROI, CTR=点击率，含涨跌对比）
+    non_full_rows = ""
+    non_full_map = {
+        "付费占比 (cost)": jzt_non_full.get("cost", {}),
+        "ROI": jzt_non_full.get("ROI", {}),
+        "点击率 (CTR)": jzt_non_full.get("CTR", {}),
+    }
+    for label, item in non_full_map.items():
+        if isinstance(item, dict):
+            val = item.get("value", "N/A")
+            change = item.get("change", "")
+            compare = item.get("compare", "")
+            non_full_rows += f"| {label} | {val} | {change} ({compare}) |\n"
+        else:
+            non_full_rows += f"| {label} | {item} | |\n"
+
+    # 全站营销概览表格（含付费占比、ROI、点击量、交易额、涨跌对比）
+    full_rows = ""
+    full_map = {
+        "付费占比 (cost)": jzt_full.get("cost", {}),
+        "ROI": jzt_full.get("ROI", {}),
+        "全站点击量": jzt_full.get("全站点击量", {}),
+        "全站交易额": jzt_full.get("全站交易额", {}),
+    }
+    for label, item in full_map.items():
+        if isinstance(item, dict):
+            val = item.get("value", "N/A")
+            change = item.get("change", "")
+            compare = item.get("compare", "")
+            full_rows += f"| {label} | {val} | {change} ({compare}) |\n"
+        else:
+            full_rows += f"| {label} | {item} | |\n"
+
+    # 汇总表
+    def get_val(obj):
+        return obj.get("value", "N/A") if isinstance(obj, dict) else obj
+
+    non_full_cost = get_val(jzt_non_full.get("cost", "N/A"))
+    non_full_roi = get_val(jzt_non_full.get("ROI", "N/A"))
+    non_full_ctr = get_val(jzt_non_full.get("CTR", "N/A"))
+
+    full_cost_val = get_val(jzt_full.get("cost", "N/A"))
+    full_roi_val = get_val(jzt_full.get("ROI", "N/A"))
+    full_clicks_val = get_val(jzt_full.get("全站点击量", "N/A"))
+    full_sum_val = get_val(jzt_full.get("全站交易额", "N/A"))
+
+    report = f"""# {shop} 数据汇总
+
+**数据采集日期：** {today}
+**店铺名称：** {shop}
+**数据周期：** {week_range}
+
+## 一、店铺星级
+
+| 项目 | 详情 |
+|------|------|
+| 当前星级 | {star} |
+
+## 二、交易概况（上周 {week_range}）
+
+| 指标 | 数值 | 环比 | 同行同级均值 | 同行上级均值 |
+|------|------|------|-------------|-------------|
+{trans_rows}
+## 三、商家榜单（行业：{industry}，上周）
+
+{''.join(build_ranking_section(cat, data) for cat, data in ranking.items())}
+## 四、非全站营销（口径：{caliber_days}天）
+
+| 指标 | 数值 | 环比 |
+|------|------|------|
+{non_full_rows}
+## 五、全站营销（口径：{caliber_days}天）
+
+| 指标 | 数值 | 涨跌对比 |
+|------|------|----------|
+{full_rows}
+## 六、汇总
+
+| 渠道 | 付费占比 | ROI | 点击率 | 点击量 | 交易额 |
+|------|----------|-----|--------|--------|--------|
+| 非全站 | {non_full_cost} | {non_full_roi} | {non_full_ctr} | -- | -- |
+| 全站 | {full_cost_val} | {full_roi_val} | -- | {full_clicks_val} | {full_sum_val} |
+| 合计 | -- | -- | -- | -- | -- |
+"""
+
+    # 写入 MD 报告
+    os.makedirs(output_dir, exist_ok=True)
+    filename = f"{shop}_店铺数据汇总_{today}.md"
+    filepath = os.path.join(output_dir, filename)
+
+    with open(filepath, "w", encoding="utf-8") as f:
+        f.write(report)
+
+    # 结构化指标 JSON（供 base_writer 节点写入多维表格使用）
+    metrics = {
+        "shop": shop,
+        "industry": industry,
+        "caliber_days": caliber_days,
+        "data_date": today,
+        "week_start": last_monday.strftime("%Y-%m-%d"),
+        "week_end": last_sunday.strftime("%Y-%m-%d"),
+        "week_range": week_range,
+        "star_rating_raw": star,
+        "transaction": transaction,
+        "ranking": ranking,
+        "jzt_non_full": jzt_non_full,
+        "jzt_full": jzt_full,
+    }
+    metrics_filename = f"{shop}_指标数据_{today}.json"
+    metrics_filepath = os.path.join(output_dir, metrics_filename)
+    try:
+        with open(metrics_filepath, "w", encoding="utf-8") as f:
+            json.dump(metrics, f, ensure_ascii=False, indent=2)
+        print(f"  指标数据已保存: {metrics_filepath}")
+    except Exception as e:
+        print(f"  [WARN] 保存指标 JSON 失败: {e}")
+
+    # 多类目原始数据保存为 JSON（供后续使用）
+    if isinstance(ranking, dict):
+        json_filename = f"{shop}_榜单数据_{today}.json"
+        json_filepath = os.path.join(output_dir, json_filename)
+        try:
+            with open(json_filepath, "w", encoding="utf-8") as f:
+                json.dump(ranking, f, ensure_ascii=False, indent=2)
+            print(f"  榜单原始数据已保存: {json_filepath}")
+        except Exception as e:
+            print(f"  [WARN] 保存 JSON 失败: {e}")
+
+    print(f"  报告已保存: {filepath}")
+    return filepath
+
+
+# ── 主流程 ────────────────────────────────────────────────────────────────────
+
+def main():
+    parser = argparse.ArgumentParser(description="京东店铺数据采集脚本")
+    parser.add_argument("--shop", default=DEFAULT_CONFIG["shop"], help="店铺账号名")
+    parser.add_argument("--password", default=DEFAULT_CONFIG["password"], help="登录密码")
+    parser.add_argument("--output", default=DEFAULT_CONFIG["output"], help="报告输出目录")
+    parser.add_argument("--industry", default=DEFAULT_CONFIG["industry"], help="行业分类")
+    parser.add_argument("--caliber", type=int, default=DEFAULT_CONFIG["caliber_days"], help="口径天数")
+    args = parser.parse_args()
+    args.output = str(managed_data_path(args.output))
+
+    config = {
+        "shop": args.shop,
+        "password": args.password,
+        "output": args.output,
+        "industry": args.industry,
+        "caliber_days": args.caliber,
+    }
+
+    print("=" * 60)
+    print("京东店铺数据采集")
+    print(f"  店铺: {config['shop']}")
+    print(f"  行业: {config['industry']}")
+    print(f"  输出: {config['output']}")
+    print("=" * 60)
+
+    last_monday, last_sunday = get_last_week_range()
+    print(f"  上周: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+    print()
+
+    with sync_playwright() as p:
+        # 启动浏览器（headed模式）
+        browser = p.chromium.launch(headless=False)
+        context = browser.new_context(
+            viewport={"width": 1920, "height": 1080},
+            locale="zh-CN",
+        )
+        page = context.new_page()
+
+        try:
+            # Step 1: 登录
+            step_login(page, config["shop"], config["password"])
+
+            # Step 2: 进入商智
+            sz_page = step_enter_shangzhi(context, page)
+
+            # Step 3: 采集店铺星级与交易数据（新版商智先选时间再抓取）
+            star, transaction = step_shop_star_and_trade(sz_page)
+
+            # Step 4: 采集旧版交易概况-客单价
+            deal_summary = step_deal_summary(sz_page)
+            # 合并到 transaction 中
+            if deal_summary.get("客单价") and deal_summary["客单价"].get("value") != "N/A":
+                cp = deal_summary["客单价"]
+                transaction["客单价"] = {
+                    "value": cp["value"],
+                    "环比": cp.get("change", "N/A"),
+                    "同行同级均值": "N/A",
+                    "同行上级均值": "N/A"
+                }
+
+            # Step 5: 商家榜单
+            ranking = step_merchant_ranking(sz_page, config["industry"])
+
+            # Step 6: 京准通-非全站营销
+            jzt_non_full = step_jzt_non_full_site(sz_page, config["caliber_days"])
+
+            # Step 7: 京准通-全站营销
+            jzt_full = step_jzt_full_site(sz_page, config["caliber_days"])
+
+            # Step 8: 生成报告
+            report_path = generate_report(
+                config, star, transaction, ranking, jzt_non_full, jzt_full
+            )
+
+            print()
+            print("=" * 60)
+            print(f"采集完成！报告: {report_path}")
+            print("=" * 60)
+
+        except Exception as e:
+            print(f"\n[ERROR] 采集过程中出错: {e}")
+            import traceback
+            traceback.print_exc()
+
+        finally:
+            context.close()
+            browser.close()
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_peer_product_data_collector.py b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_peer_product_data_collector.py
new file mode 100644
index 0000000..646a6d6
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_peer_product_data_collector.py
@@ -0,0 +1,2401 @@
+"""
+京东商品榜单竞品数据采集脚本
+复用 jd_data_collector 的登录和商智跳转逻辑，
+在商品榜单页面搜索品牌、选择类目、采集数据（含SPU）。
+
+依赖:
+    pip install playwright
+    playwright install chromium
+
+用法:
+    python jd_peer_product_data_collector.py
+    python jd_peer_product_data_collector.py --shop "店铺名" --password "密码" --output "输出目录"
+"""
+
+import argparse
+import json
+import os
+import re
+from datetime import datetime, timedelta
+
+from playwright.sync_api import sync_playwright, BrowserContext
+from runtime_paths import RAW_DATA_ROOT, managed_data_path
+
+from collection_progress import SpuProgress
+
+# 复用 jd_data_collector 的配置和函数
+from jd_data_collector import (
+    DEFAULT_CONFIG,
+    step_login,
+    step_enter_shangzhi,
+    extract,
+    get_last_week_range,
+    select_last_week_in_old_datepicker,
+    close_popup_if_present,
+)
+
+BRANDS = ["Bellroy", "tomtoc", "探迹者", "BAGSMART", "GROTTO", "NIID"]
+CATEGORIES = [
+    ("男士双肩包", "男包"),
+    ("男士单肩/斜挎包", "男包"),
+]
+STORE_SPU = "10027762032411"  # 本店固定SPU，用于竞品对比
+
+
+# ── 页面导航 ──────────────────────────────────────────────────────────────────
+
+def step_return_old_version(page):
+    """返回商智旧版"""
+    print("[Step 3] 返回商智旧版...")
+    close_popup_if_present(page)
+    try:
+        back_button = page.locator('#header > div > div.page-top-header-tools > img')
+        back_button.wait_for(state="visible", timeout=5000)
+        back_button.click()
+        page.wait_for_timeout(5000)
+        try:
+            page.wait_for_load_state("domcontentloaded", timeout=60000)
+        except Exception:
+            pass
+        print("  [OK] 已进入商智旧版")
+    except Exception:
+        print("  [INFO] 未找到返回旧版按钮，继续当前页面")
+
+
+def step_goto_product_ranks(page):
+    """跳转到商品榜单页面"""
+    print("[Step 4] 跳转到商品榜单页面...")
+    page.goto(
+        "https://sz.jd.com/szweb/sz/view/industryMarket/productRanks.html",
+        wait_until="networkidle"
+    )
+    page.wait_for_timeout(5000)
+    print("  [OK] 已到达商品榜单页面")
+
+
+# ── 日期选择（复用 jd_data_collector.get_last_week_range 计算日期）────────────
+
+def select_last_week_any_day(page):
+    """选择上周时间范围"""
+    print("  选择上周时间范围...")
+
+    # 1. 点击日期选择器
+    try:
+        date_picker = page.locator('.jmtd-date-picker .jmtd-base-input-top').first
+        date_picker.wait_for(state="visible", timeout=5000)
+        date_picker.click()
+        print("    [OK] 点击日期选择器")
+    except Exception as e:
+        print(f"    [WARN] 点击日期选择器失败: {e}")
+        try:
+            page.locator('text=按天查询').first.click()
+            print("    [OK] 使用文本定位点击日期选择器")
+        except Exception:
+            pass
+    page.wait_for_timeout(2000)
+
+    # 2. 点击"按周查询"
+    try:
+        week_switch = page.locator('.jmtd-date-picker-switch-panel-item', has_text="按周查询").first
+        week_switch.wait_for(state="visible", timeout=5000)
+        week_switch.click()
+        print("    [OK] 点击按周查询")
+    except Exception as e:
+        print(f"    [WARN] 点击按周查询失败: {e}")
+        page.evaluate("""() => {
+            const items = document.querySelectorAll('.jmtd-date-picker-switch-panel-item');
+            for (const item of items) {
+                if (item.textContent.includes('按周查询')) {
+                    item.click();
+                    return true;
+                }
+            }
+            return false;
+        }""")
+    page.wait_for_timeout(2000)
+
+    # 3. 等待日历弹窗并选择上周某一天
+    calendar = page.locator('.jmtd-date-picker-popup, .jmtd-date-picker-panel, [class*="picker-panel"], [class*="calendar-panel"]')
+    try:
+        calendar.wait_for(state="visible", timeout=5000)
+    except Exception:
+        print("  [WARN] 未检测到日历弹窗，尝试继续...")
+        page.wait_for_timeout(3000)
+
+    select_last_week_in_old_datepicker(page, "peer_product_rank")
+
+    try:
+        calendar.wait_for(state="hidden", timeout=5000)
+    except Exception:
+        pass
+    page.wait_for_timeout(5000)
+    print("  已切换到上周数据")
+
+
+# ── 类目切换（针对商品榜单页面的结构）──────────────────────────────────────────
+
+def switch_product_category(page, category_name, parent_category="男包"):
+    """在商品榜单页面切换类目：先点击'男包'选择器，再悬停父类目，最后点击子类目"""
+    label = f"{parent_category} → {category_name}"
+    print(f"  切换类目到: {label}")
+
+    # 1. 点击包含父类目名称的类目选择器（如显示"男包"的那个下拉框）
+    try:
+        category_trigger = page.locator(
+            f'.jmtd-base-input:has(.jmtd-label:has-text("{parent_category}")) .jmtd-base-input-top'
+        ).first
+        category_trigger.wait_for(state="visible", timeout=5000)
+        category_trigger.click()
+        print(f"    [OK] 点击类目选择器: {parent_category}")
+    except Exception as e:
+        print(f"    [WARN] 精确定位类目选择器失败: {e}")
+        # 兜底：尝试点击第2个 jmtd-base-input-top（0=SPU, 1=类目, 2=渠道, 3=日期）
+        try:
+            page.locator('.jmtd-base-input-top').nth(1).click()
+            print("    [OK] 兜底点击第2个选择器")
+        except Exception as e2:
+            print(f"    [WARN] 兜底点击也失败: {e2}")
+    page.wait_for_timeout(2000)
+
+    # 2. 悬停父类目以展开子类目列表
+    try:
+        parent_locator = page.locator('.jmtd-dropdown-content .jmtd-label', has_text=parent_category).first
+        parent_locator.hover(timeout=3000)
+        page.wait_for_timeout(2000)
+        print(f"    [OK] 悬停父类目: {parent_category}")
+    except Exception as e:
+        print(f"    [WARN] 悬停父类目失败: {e}")
+
+    # 3. 点击目标子类目
+    try:
+        child_locator = page.locator('.jmtd-dropdown-content .jmtd-label', has_text=category_name).first
+        child_locator.click(timeout=3000)
+        page.wait_for_timeout(5000)
+        print(f"    [OK] 已切换到 {category_name}")
+    except Exception as e:
+        print(f"    [WARN] 点击子类目失败: {e}")
+
+
+# ── 搜索与采集 ─────────────────────────────────────────────────────────────────
+
+def search_brand(page, brand_name):
+    """在商品榜单搜索框中搜索品牌（定位带搜索图标的输入框）"""
+    print(f"  搜索品牌: {brand_name}")
+
+    try:
+        # 定位带搜索图标（.jmtd-base-input-start-icon）的输入框
+        search_input = page.locator('.jmtd-base-input:has(.jmtd-base-input-start-icon) .jmtd-input-input').first
+        search_input.wait_for(state="visible", timeout=5000)
+        search_input.click()
+        search_input.fill("")
+        page.wait_for_timeout(500)
+        search_input.fill(brand_name)
+        page.wait_for_timeout(1000)
+        search_input.press("Enter")
+        print(f"    [OK] 已搜索: {brand_name}")
+    except Exception as e:
+        print(f"    [WARN] 精确搜索失败: {e}")
+        # 兜底：用 JS 查找 placeholder 为"请输入商品名称"的输入框
+        page.evaluate(f"""() => {{
+            const inputs = document.querySelectorAll('input');
+            for (const input of inputs) {{
+                const ph = input.placeholder || '';
+                if (ph.includes('商品名称') || ph.includes('搜索')) {{
+                    input.value = '{brand_name}';
+                    input.dispatchEvent(new Event('input', {{ bubbles: true }}));
+                    input.dispatchEvent(new KeyboardEvent('keydown', {{ key: 'Enter', bubbles: true }}));
+                    return true;
+                }}
+            }}
+            return false;
+        }}""")
+        print(f"    [OK] JS兜底搜索: {brand_name}")
+    page.wait_for_timeout(8000)
+
+
+def extract_product_ranking_table(page):
+    """提取商品榜单表格数据（同时提取商品链接中的SKU）"""
+    ranking_data = extract(page, """() => {
+        const rows = document.querySelectorAll('.jmtd-table-main tbody tr');
+        const result = [];
+        for (const row of rows) {
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 7) continue;
+
+            let rank = '';
+            const rankNumEl = cells[0].querySelector('.jmtd-table-formatter-rank-number');
+            if (rankNumEl) {
+                rank = rankNumEl.textContent.trim();
+            }
+            if (!rank) {
+                const rankIconEl = cells[0].querySelector('.jmtd-table-formatter-rank');
+                if (rankIconEl) {
+                    const iconMatch = rankIconEl.className.match(/jmtd-table-formatter-rank-icon-(\\d)/);
+                    if (iconMatch) {
+                        rank = iconMatch[1];
+                    }
+                }
+            }
+            if (!rank) {
+                rank = cells[0].textContent.trim();
+            }
+
+            const productLink = cells[1].querySelector('a');
+            const product = productLink ? productLink.textContent.trim() : cells[1].textContent.trim();
+            const href = productLink ? productLink.getAttribute('href') || '' : '';
+            const productCell = cells[1];
+
+            function normalizeUrl(url) {
+                if (!url) return '';
+                url = String(url).trim();
+                if (!url) return '';
+                if (url.startsWith('//')) return window.location.protocol + url;
+                try {
+                    return new URL(url, window.location.href).href;
+                } catch (e) {
+                    return url;
+                }
+            }
+
+            let fullHref = '';
+            if (href) {
+                try {
+                    fullHref = new URL(href, window.location.href).href;
+                } catch (e) {
+                    fullHref = href.startsWith('//') ? window.location.protocol + href : href;
+                }
+            }
+
+            let imageUrl = '';
+            const img = productCell.querySelector('img');
+            if (img) {
+                imageUrl = img.getAttribute('src')
+                    || img.getAttribute('data-src')
+                    || img.getAttribute('data-lazy-img')
+                    || img.getAttribute('data-original')
+                    || img.getAttribute('data-url')
+                    || '';
+            }
+            if (!imageUrl) {
+                const imageNode = productCell.querySelector('[style*="background-image"]');
+                if (imageNode) {
+                    const style = imageNode.getAttribute('style') || '';
+                    const match = style.match(/url\\((['"]?)(.*?)\\1\\)/);
+                    if (match) imageUrl = match[2];
+                }
+            }
+            imageUrl = normalizeUrl(imageUrl);
+
+            // 尝试从 href 中直接提取 SKU
+            let sku = '';
+            let itemUrl = '';
+            if (fullHref) {
+                const skuMatch = fullHref.match(/item\\.jd\\.com\\/(\\d{14})\\.html/);
+                if (skuMatch) {
+                    sku = skuMatch[1];
+                    itemUrl = fullHref;
+                } else if (fullHref.startsWith('http')) {
+                    itemUrl = fullHref;
+                    const fallbackMatch = fullHref.match(/(\\d{14})/);
+                    if (fallbackMatch) sku = fallbackMatch[1];
+                }
+            }
+
+            const item = {
+                rank: rank,
+                product: product,
+                amount: cells[2]?.textContent?.trim() || '',
+                orders: cells[3]?.textContent?.trim() || '',
+                visitors: cells[4]?.textContent?.trim() || '',
+                searchClicks: cells[5]?.textContent?.trim() || '',
+                followers: cells[6]?.textContent?.trim() || '',
+                sku: sku,
+                item_url: itemUrl,
+                href: href,
+                product_url: fullHref,
+                image_url: imageUrl,
+            };
+            result.push(item);
+        }
+        return JSON.stringify(result);
+    }""", "[]")
+
+    rows = []
+    if isinstance(ranking_data, str):
+        try:
+            rows = json.loads(ranking_data)
+        except json.JSONDecodeError:
+            rows = []
+    return rows
+
+
+def normalize_product_url(url):
+    """把商品链接规范成完整 URL。"""
+    if not url:
+        return ''
+    url = str(url).strip()
+    if not url:
+        return ''
+    if url.startswith('//'):
+        return 'https:' + url
+    if url.startswith('/'):
+        return 'https://item.jd.com' + url
+    return url
+
+
+def normalize_asset_url(url):
+    """把图片等资源链接规范成完整 URL。"""
+    if not url:
+        return ''
+    url = str(url).strip()
+    if not url:
+        return ''
+    if url.startswith('//'):
+        url = 'https:' + url
+    if url.startswith('/'):
+        url = 'https://img.jd.com' + url
+    url = re.sub(r'/n\d+/jfs/', '/n1/jfs/', url)
+    url = re.sub(r'/s\d+x\d+_jfs/', '/n1/jfs/', url)
+    return url
+
+
+def get_product_url(row):
+    """优先取已解析的商品完整链接，兼容旧字段。"""
+    return normalize_product_url(row.get('item_url') or row.get('product_url') or row.get('href') or row.get('itemUrl') or '')
+
+
+def get_product_image_url(row):
+    """获取商品图片链接，兼容可能的字段名。"""
+    return normalize_asset_url(
+        row.get('image_url')
+        or row.get('product_image')
+        or row.get('imageUrl')
+        or row.get('img_url')
+        or row.get('imgUrl')
+        or ''
+    )
+
+
+def md_cell(value):
+    """Markdown 表格单元格转义。"""
+    return str(value or '').replace('|', '\\|').replace('\n', ' ').replace('\r', ' ')
+
+
+def make_link_ref(prefix, index, url, refs):
+    """生成短引用链接，避免 Markdown 表格被长 URL 撑开。"""
+    if not url:
+        return ''
+    ref = f"{prefix}-{index}"
+    refs.append(f"[{ref}]: {url}")
+    return f"[查看商品][{ref}]"
+
+
+def make_image_ref(prefix, index, url, refs):
+    """生成短图片引用，避免 Markdown 表格被图片 URL 撑开。"""
+    if not url:
+        return ''
+    ref = f"{prefix}-{index}"
+    refs.append(f"[{ref}]: {url}")
+    return f"![图][{ref}]"
+
+
+def extract_sku_from_url(url):
+    """从京东商品URL中提取14位SKU"""
+    match = re.search(r'item\.jd\.com/(\d{14})\.html', url)
+    if match:
+        return match.group(1)
+    return ''
+
+
+def collect_product_skus(page, context, rows):
+    """对表格中未提取到SKU的商品，兜底点击跳转获取"""
+    # 先过滤出需要点击兜底的商品
+    need_click = [(i, row) for i, row in enumerate(rows) if not row.get('sku')]
+    if not need_click:
+        print(f"    全部 {len(rows)} 个商品已从表格链接中提取到SKU，无需跳转")
+        return rows
+
+    print(f"    {len(need_click)} 个商品需要点击跳转获取SKU（已间隔2秒降低风控风险）...")
+
+    for idx, (i, row) in enumerate(need_click):
+        product_name = row.get('product', '')[:40]
+        print(f"      [{idx+1}/{len(need_click)}] {product_name}...")
+
+        try:
+            with context.expect_page() as new_page_info:
+                page.locator('.jmtd-table-main tbody tr').nth(i).locator('td').nth(1).locator('a').first.click()
+
+            new_page = new_page_info.value
+            new_page.wait_for_load_state("networkidle", timeout=15000)
+
+            url = new_page.url
+            sku = extract_sku_from_url(url)
+
+            row['sku'] = sku
+            row['item_url'] = url
+
+            if sku:
+                print(f"        [OK] SKU: {sku}")
+            else:
+                print(f"        [WARN] URL中未找到14位SKU")
+
+            new_page.close()
+            # 间隔2秒再点下一个，降低风控风险
+            page.wait_for_timeout(2000)
+        except Exception as e:
+            print(f"        [FAIL] 获取失败: {e}")
+            row['sku'] = ''
+            row['item_url'] = ''
+
+    return rows
+
+
+def collect_data(page, context):
+    """主采集逻辑：遍历品牌和类目，采集商品榜单数据（含SPU）"""
+    print("[Step 5] 采集商品榜单数据...")
+
+    # 先选择上周
+    select_last_week_any_day(page)
+
+    results = {}
+
+    for brand_name in BRANDS:
+        print(f"\n{'='*50}")
+        print(f"  品牌: {brand_name}")
+        print(f"{'='*50}")
+
+        brand_data = {}
+
+        # 搜索品牌
+        search_brand(page, brand_name)
+
+        # 遍历类目
+        for i, (cat_name, parent_cat) in enumerate(CATEGORIES):
+            print(f"\n  [{i+1}/{len(CATEGORIES)}] 类目: {parent_cat} → {cat_name}")
+
+            switch_product_category(page, cat_name, parent_cat)
+            page.wait_for_timeout(5000)
+
+            # 提取表格
+            rows = extract_product_ranking_table(page)
+            # 采集每个商品的 SKU
+            rows = collect_product_skus(page, context, rows)
+            brand_data[cat_name] = rows
+
+            print(f"    抓取到 {len(rows)} 条商品数据（含SKU）")
+
+        results[brand_name] = brand_data
+
+    return results
+
+
+# ── 竞品配置页面操作 ──────────────────────────────────────────────────────────
+
+def step_goto_competition_config(page):
+    """跳转到竞品分析配置页面"""
+    print("[Step 6] 跳转到竞品分析配置页面...")
+    page.goto(
+        "https://sz.jd.com/szweb/sz/view/competitionAnalysis/competitionConfig.html",
+        wait_until="networkidle"
+    )
+    page.wait_for_timeout(5000)
+    print("  [OK] 已到达竞品分析配置页面")
+
+
+def click_competitor_config_tab(page):
+    """点击竞品配置 tab"""
+    print("[Step 7] 切换到竞品配置...")
+    try:
+        tab = page.locator('.jmtd-tabs-tab', has_text="竞品配置").first
+        tab.wait_for(state="visible", timeout=5000)
+        tab.click()
+        print("  [OK] 已点击竞品配置")
+    except Exception as e:
+        print(f"  [WARN] 精确点击失败: {e}")
+        page.evaluate("""() => {
+            const tabs = document.querySelectorAll('.jmtd-tabs-tab');
+            for (const tab of tabs) {
+                if (tab.textContent.includes('竞品配置')) {
+                    tab.click();
+                    return true;
+                }
+            }
+            return false;
+        }""")
+        print("  [OK] JS兜底点击竞品配置")
+    page.wait_for_timeout(5000)
+
+
+def get_monitored_count(page):
+    """获取已监控数量（统计商品卡片数量）"""
+    print("  获取已监控数量...")
+    try:
+        count = page.locator('.card-wrapper').count()
+        print(f"    [OK] 已监控数量: {count}")
+        return count
+    except Exception as e:
+        print(f"    [WARN] 获取数量失败: {e}")
+        return 0
+
+
+def remove_all_monitored_products(page):
+    """逐个删除第一个位置的已监控竞品"""
+    count = get_monitored_count(page)
+    if count == 0:
+        print("  没有已监控的竞品，跳过删除")
+        return
+
+    print(f"[Step 8] 开始删除 {count} 个已监控竞品...")
+
+    for i in range(count):
+        print(f"  [{i+1}/{count}] 删除第 1 个位置的竞品...")
+
+        try:
+            # 1. hover 第一个 card-wrapper，等待删除按钮出现
+            first_card = page.locator('.card-wrapper').first
+            first_card.hover(timeout=5000)
+            page.wait_for_timeout(1000)
+
+            # 2. 点击删除按钮（在 .del-button 下）
+            delete_btn = first_card.locator('.del-button .jmti-close').first
+            delete_btn.wait_for(state="visible", timeout=3000)
+            delete_btn.click()
+            print("    [OK] 点击删除按钮")
+
+            # 3. 等待弹窗出现并点击确定
+            page.wait_for_timeout(1000)
+            confirm_btn = page.locator('.jmtd-modal-footer .jmtd-button-primary').last
+            confirm_btn.wait_for(state="visible", timeout=5000)
+            confirm_btn.click()
+            print("    [OK] 点击确定删除")
+
+            # 4. 等待 2 秒再继续下一个
+            page.wait_for_timeout(2000)
+
+        except Exception as e:
+            print(f"    [FAIL] 删除失败: {e}")
+            # 兜底：直接用 JS 触发删除
+            try:
+                page.evaluate("""() => {
+                    const cards = document.querySelectorAll('.card-wrapper');
+                    if (cards.length > 0) {
+                        const card = cards[0];
+                        card.dispatchEvent(new MouseEvent('mouseenter', { bubbles: true }));
+                        const closeBtn = card.querySelector('.del-button .jmti-close');
+                        if (closeBtn) {
+                            closeBtn.click();
+                            return 'clicked_close';
+                        }
+                    }
+                    return 'no_card';
+                }""")
+                page.wait_for_timeout(1500)
+                page.evaluate("""() => {
+                    const confirmBtn = document.querySelector('.jmtd-modal-footer .jmtd-button-primary');
+                    if (confirmBtn) {
+                        confirmBtn.click();
+                        return 'clicked_confirm';
+                    }
+                    return 'no_confirm';
+                }""")
+                page.wait_for_timeout(2000)
+            except Exception as e2:
+                print(f"    [FAIL] 兜底删除也失败: {e2}")
+
+    print("  [OK] 全部已监控竞品删除完成")
+
+
+def add_spus_to_monitor(page, results):
+    """将采集到的 SKU 逐个添加到竞品监控列表，同时收集 SPU"""
+    print("[Step 9] 开始添加采集到的 SKU 到监控列表...")
+
+    # 收集所有不重复的 SKU，同时建立 SKU -> rows 映射用于回写 SPU
+    all_skus = []
+    sku_to_rows = {}  # sku -> list of (brand_name, cat_name, row)
+    seen = set()
+    for brand_name in BRANDS:
+        for cat_name, _ in CATEGORIES:
+            rows = results.get(brand_name, {}).get(cat_name, [])
+            for row in rows:
+                # 兼容新旧数据：优先取 sku 字段，没有则取 spu 字段（旧数据实际存的是 SKU）
+                sku = row.get('sku', '') or row.get('spu', '')
+                if sku and sku not in seen:
+                    seen.add(sku)
+                    all_skus.append({
+                        'sku': sku,
+                        'product': row.get('product', ''),
+                        'brand': brand_name
+                    })
+                    sku_to_rows[sku] = []
+                if sku:
+                    sku_to_rows[sku].append((brand_name, cat_name, row))
+
+    print(f"  共 {len(all_skus)} 个 SKU 需要添加")
+    if not all_skus:
+        print("  没有需要添加的 SKU，跳过")
+        return
+
+    total = len(all_skus)
+    for idx, item in enumerate(all_skus):
+        sku = item['sku']
+        product = item['product'][:40]
+        print(f"\n  [{idx+1}/{total}] 添加 SKU: {sku} ({product}...)")
+
+        try:
+            # 1. 清空输入框并输入 SKU
+            input_box = page.locator('input.jmtd-input-input').first
+            input_box.wait_for(state="visible", timeout=5000)
+            input_box.click()
+            input_box.fill("")
+            page.wait_for_timeout(500)
+            input_box.fill(sku)
+            print(f"    [OK] 已输入 SKU: {sku}")
+            page.wait_for_timeout(500)
+
+            # 2. 点击"添加商品"按钮
+            add_btn = page.locator('button.jmtd-button-primary', has_text="添加商品").first
+            add_btn.wait_for(state="visible", timeout=5000)
+            add_btn.click()
+            print("    [OK] 点击添加商品")
+            page.wait_for_timeout(1000)
+
+            # 3. 等待商品预览弹窗出现
+            preview = page.locator('.product-pre-select').first
+            preview.wait_for(state="visible", timeout=10000)
+            print("    [OK] 商品预览弹窗已出现")
+
+            # 4. 从弹窗中提取 SPU
+            spu = ''
+            try:
+                spu_el = preview.locator('.jmtd-typography-text.product-id', has_text='SPU').locator('span').first
+                if spu_el.count() > 0:
+                    spu = spu_el.inner_text(timeout=3000).strip()
+            except Exception:
+                pass
+
+            if not spu:
+                # JS 兜底提取 SPU
+                spu = preview.evaluate("""() => {
+                    const spans = document.querySelectorAll('.product-pre-select .jmtd-typography-text.product-id');
+                    for (const span of spans) {
+                        if (span.textContent.includes('SPU')) {
+                            const numSpan = span.querySelector('span');
+                            return numSpan ? numSpan.textContent.trim() : '';
+                        }
+                    }
+                    return '';
+                }""")
+
+            if spu:
+                # 回写 SPU 到 results 中对应的所有 row
+                for brand_name, cat_name, row in sku_to_rows[sku]:
+                    row['spu'] = spu
+                print(f"    [OK] 提取并记录 SPU: {spu}")
+            else:
+                print(f"    [WARN] 未能从弹窗提取 SPU")
+
+            # 5. 检查 SPU 是否已监控（按钮 disabled 且显示"已监控"）
+            # 弹窗中第一个 .cell 是 SPU
+            spu_add_btn = preview.locator('.cell').first.locator('button').first
+            btn_text = spu_add_btn.inner_text(timeout=3000)
+            is_disabled = spu_add_btn.is_disabled()
+
+            if is_disabled and '已监控' in btn_text:
+                print("    [SKIP] 该 SPU 已在监控列表中")
+                # 关闭弹窗：先尝试 × 按钮，再试 ESC
+                try:
+                    close_btn = page.locator('.jmtd-modal-close, .jmti-close').last
+                    close_btn.wait_for(state="visible", timeout=3000)
+                    close_btn.click()
+                except Exception:
+                    page.keyboard.press("Escape")
+                page.wait_for_timeout(500)
+                continue
+
+            # 6. 点击 SPU 旁边的"添加"按钮
+            spu_add_btn.click()
+            print("    [OK] 点击 SPU 添加按钮")
+            page.wait_for_timeout(800)
+
+            # 7. 等待确认弹窗出现并点击确定
+            confirm_modal = page.locator('.jmtd-modal-tips-content', has_text="确定要添加此商品吗")
+            confirm_modal.wait_for(state="visible", timeout=5000)
+            confirm_btn = confirm_modal.locator('.jmtd-button-primary').last
+            confirm_btn.click()
+            print("    [OK] 确认添加")
+            page.wait_for_timeout(800)
+
+            # 8. 等待成功弹窗出现并点击确定
+            success_modal = page.locator('.jmtd-modal-tips-content', has_text="添加成功")
+            success_modal.wait_for(state="visible", timeout=10000)
+            success_ok_btn = success_modal.locator('.jmtd-button-primary').first
+            success_ok_btn.click()
+            print("    [OK] 添加成功")
+            page.wait_for_timeout(500)
+
+        except Exception as e:
+            print(f"    [FAIL] 添加失败: {e}")
+            # 兜底：尝试关闭所有弹窗，继续下一个
+            try:
+                page.keyboard.press("Escape")
+                page.evaluate("""() => {
+                    const modals = document.querySelectorAll('.jmtd-modal-mask');
+                    modals.forEach(m => { if(m.click) m.click(); });
+                    return 'closed';
+                }""")
+                page.wait_for_timeout(1000)
+            except Exception:
+                pass
+
+    print(f"\n  [OK] 全部 {total} 个 SKU 处理完成")
+
+
+# ── 保存结果 ───────────────────────────────────────────────────────────────────
+
+def save_results(results, output_dir):
+    """保存采集结果到 JSON 和 Markdown（含SKU汇总表）"""
+    today = datetime.now().strftime("%Y-%m-%d")
+    last_monday, last_sunday = get_last_week_range()
+    week_range = f"{last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}"
+
+    os.makedirs(output_dir, exist_ok=True)
+
+    # ── JSON ──
+    output = {
+        "采集日期": today,
+        "数据周期": week_range,
+        "数据": results,
+    }
+    json_path = os.path.join(output_dir, f"商品榜单竞品数据_{today}.json")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(output, f, ensure_ascii=False, indent=2)
+    print(f"\nJSON 已保存: {json_path}")
+
+    # ── 详细数据 Markdown（含SKU列）──
+    md = []
+    md.append("# 商品榜单竞品数据汇总")
+    md.append("")
+    md.append(f"**采集日期：** {today}")
+    md.append(f"**数据周期：** 上周 {week_range}")
+    md.append("")
+    md_refs = []
+    md_link_index = 1
+    md_image_refs = []
+    md_image_index = 1
+
+    for brand_name in BRANDS:
+        md.append(f"## {brand_name}")
+        md.append("")
+        for cat_name, _ in CATEGORIES:
+            md.append(f"### {cat_name}")
+            md.append("")
+            md.append("| 排名 | 商品图片 | 商品信息 | 商品链接 | SKU | SPU | 成交金额 | 成交单量 | 访客数 | 搜索点击 | 关注人数 |")
+            md.append("|------|----------|----------|----------|-----|-----|----------|----------|--------|----------|----------|")
+            rows = results.get(brand_name, {}).get(cat_name, [])
+            if not rows:
+                md.append("| -- | -- | 暂无数据 | -- | -- | -- | -- | -- | -- | -- | -- |")
+            for row in rows:
+                product_url = get_product_url(row)
+                product_link = make_link_ref("item", md_link_index, product_url, md_refs)
+                if product_url:
+                    md_link_index += 1
+                image_url = get_product_image_url(row)
+                product_image = make_image_ref("img", md_image_index, image_url, md_image_refs)
+                if image_url:
+                    md_image_index += 1
+                md.append(f"| {md_cell(row.get('rank', ''))} | {product_image} | {md_cell(row.get('product', ''))} | {product_link} "
+                          f"| {md_cell(row.get('sku', ''))} | {md_cell(row.get('spu', ''))} "
+                          f"| {md_cell(row.get('amount', ''))} | {md_cell(row.get('orders', ''))} "
+                          f"| {md_cell(row.get('visitors', ''))} | {md_cell(row.get('searchClicks', ''))} "
+                          f"| {md_cell(row.get('followers', ''))} |")
+            md.append("")
+
+    if md_refs:
+        md.append("## 商品链接引用")
+        md.append("")
+        md.extend(md_refs)
+        md.append("")
+
+    if md_image_refs:
+        md.append("## 商品图片引用")
+        md.append("")
+        md.extend(md_image_refs)
+        md.append("")
+
+    md_path = os.path.join(output_dir, f"商品榜单竞品数据汇总_{today}.md")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md))
+    print(f"详细数据 MD 已保存: {md_path}")
+
+    # ── SKU 汇总 Markdown（店铺 | 商品名称 | 商品图片 | 商品链接 | SKU | SPU）──
+    md_sku = []
+    md_sku.append("# 商品 SPU 汇总表")
+    md_sku.append("")
+    md_sku.append(f"**采集日期：** {today}")
+    md_sku.append(f"**数据周期：** 上周 {week_range}")
+    md_sku.append("")
+    md_sku.append("| 店铺（品牌） | 商品图片 | 商品名称 | 商品链接 | SKU | SPU |")
+    md_sku.append("|-------------|----------|----------|----------|-----|-----|")
+    md_sku_refs = []
+    md_sku_link_index = 1
+    md_sku_image_refs = []
+    md_sku_image_index = 1
+
+    for brand_name in BRANDS:
+        for cat_name, _ in CATEGORIES:
+            rows = results.get(brand_name, {}).get(cat_name, [])
+            for row in rows:
+                sku = row.get('sku', '')
+                spu = row.get('spu', '')
+                product = row.get('product', '')
+                product_url = get_product_url(row)
+                product_link = make_link_ref("sku-item", md_sku_link_index, product_url, md_sku_refs)
+                if product_url:
+                    md_sku_link_index += 1
+                image_url = get_product_image_url(row)
+                product_image = make_image_ref("sku-img", md_sku_image_index, image_url, md_sku_image_refs)
+                if image_url:
+                    md_sku_image_index += 1
+                md_sku.append(f"| {md_cell(brand_name)} | {product_image} | {md_cell(product)} | {product_link} | {md_cell(sku)} | {md_cell(spu)} |")
+
+    if md_sku_refs:
+        md_sku.append("")
+        md_sku.append("## 商品链接引用")
+        md_sku.append("")
+        md_sku.extend(md_sku_refs)
+
+    if md_sku_image_refs:
+        md_sku.append("")
+        md_sku.append("## 商品图片引用")
+        md_sku.append("")
+        md_sku.extend(md_sku_image_refs)
+
+    sku_path = os.path.join(output_dir, f"商品SPU汇总_{today}.md")
+    with open(sku_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md_sku))
+    print(f"SPU 汇总 MD 已保存: {sku_path}")
+
+    return md_path, sku_path
+
+
+# ── 竞品对比页面数据采集 ───────────────────────────────────────────────────────
+
+def goto_product_contrasts(page):
+    """跳转到竞品对比页面"""
+    if "productContrasts.html" in page.url:
+        print("[Step 10] 已在竞品对比页面，跳过重复跳转")
+        return
+
+    print("[Step 10] 跳转到竞品对比页面...")
+    page.goto(
+        "https://sz.jd.com/szweb/sz/view/competitionAnalysis/productContrasts.html",
+        wait_until="networkidle"
+    )
+    page.wait_for_timeout(5000)
+    print("  [OK] 已到达竞品对比页面")
+
+
+def ensure_spu_tab(page):
+    """确保SPU标签被选中"""
+    print("[Step 11] 确保SPU标签被选中...")
+    try:
+        spu_tab = page.locator('.jmtd-tabs-tab', has_text="SPU").first
+        is_active = spu_tab.evaluate("el => el.classList.contains('jmtd-tabs-tab-active')")
+        if not is_active:
+            spu_tab.click()
+            page.wait_for_timeout(2000)
+            print("  [OK] 已点击SPU标签")
+        else:
+            print("  [OK] SPU标签已激活")
+    except Exception as e:
+        print(f"  [WARN] 确保SPU标签失败: {e}")
+        page.evaluate("""() => {
+            const tabs = document.querySelectorAll('.jmtd-tabs-tab');
+            for (const tab of tabs) {
+                if (tab.textContent.trim() === 'SPU' && !tab.classList.contains('jmtd-tabs-tab-active')) {
+                    tab.click();
+                    return true;
+                }
+            }
+            return false;
+        }""")
+        page.wait_for_timeout(2000)
+
+
+def set_date_to_last_week(page):
+    """设置核心指标模块的日期为上周（按周查询模式）"""
+    return set_module_date(page, '.core-container')
+
+
+def is_module_date_last_week(page, module_selector):
+    """检查模块日期是否已经是上周，避免重复打开日期控件。"""
+    last_monday, last_sunday = get_last_week_range()
+    expected_start = last_monday.strftime("%Y-%m-%d")
+    expected_end = last_sunday.strftime("%Y-%m-%d")
+    label_text = get_module_date_label(page, module_selector)
+    return '按周查询' in label_text and expected_start in label_text and expected_end in label_text
+
+
+def set_module_date_by_title(page, title):
+    """按模块标题定位并设置日期为上周。"""
+    marker = "data-date-module-target"
+    result = page.evaluate("""({title, marker}) => {
+        document.querySelectorAll('[' + marker + ']').forEach(el => el.removeAttribute(marker));
+        const modules = document.querySelectorAll('.container-right-modules, .core-container, .keyword-container, .viewsource-container');
+        for (const module of modules) {
+            const h2 = module.querySelector('h2');
+            if (h2 && h2.textContent.includes(title)) {
+                module.setAttribute(marker, '1');
+                return 'marked';
+            }
+        }
+        for (const module of modules) {
+            if (module.textContent.includes(title) && module.querySelector('.jmtd-date-picker')) {
+                module.setAttribute(marker, '1');
+                return 'marked_by_text';
+            }
+        }
+        return 'not_found';
+    }""", {"title": title, "marker": marker})
+    if not str(result).startswith('marked'):
+        print(f"    [WARN] 未找到模块标题: {title}")
+        return False
+    return set_module_date(page, f'[{marker}="1"]')
+
+
+def set_module_date(page, module_selector):
+    """在指定模块内设置日期为上周（按周查询模式），支持跨月导航"""
+    last_monday, last_sunday = get_last_week_range()
+    target_year = last_monday.year
+    target_month = last_monday.month
+    target_day = last_monday.day
+    expected_start = last_monday.strftime("%Y-%m-%d")
+    expected_end = last_sunday.strftime("%Y-%m-%d")
+
+    try:
+        if is_module_date_last_week(page, module_selector):
+            print(f"    [OK] 日期已是上周: {expected_start} 至 {expected_end}")
+            return True
+
+        module = page.locator(module_selector).first
+        if module.count() == 0:
+            print(f"    [WARN] 未找到模块: {module_selector}")
+            return False
+
+        # 1. 用 Playwright 真实点击模块内的日期选择器。
+        picker = module.locator('.jmtd-date-picker .jmtd-base-input-top, .jmtd-date-picker').first
+        if picker.count() == 0:
+            print("    [WARN] no_picker")
+            return False
+        picker.click(timeout=5000)
+
+        dropdown = wait_for_visible_date_dropdown(page)
+        if not dropdown:
+            print("    [WARN] 日期弹层未出现")
+            return False
+
+        # 2. 切换到按周查询。这里必须用真实点击，JS click 会导致组件仍停留在按天模式。
+        week_result = click_week_switch(page)
+
+        print(f"    周模式切换: {week_result}")
+        if week_result == 'no_week_switch':
+            page.keyboard.press("Escape")
+            return False
+
+        if not wait_for_week_mode(page):
+            print("    [WARN] 未确认日期弹层切到按周模式，继续点击上周日期")
+
+        # 3. 导航到正确年月 + 真实鼠标点击上周任意一天。
+        # 个别模块会忽略第一次点击，所以尝试上周 7 天，直到 label 确认命中。
+        click_result = ""
+        for offset in range(7):
+            target_date = last_monday + timedelta(days=offset)
+            click_result = click_week_day(page, target_date.year, target_date.month, target_date.day)
+            print(f"    日期选择结果: {click_result}")
+
+            if str(click_result).startswith('clicked'):
+                selected = wait_for_module_week_label(page, module_selector, expected_start, expected_end)
+                if selected:
+                    print(f"    [OK] 已切换上周: {expected_start} 至 {expected_end}")
+                    return True
+
+            actual_label = get_module_date_label(page, module_selector)
+            print(f"    [WARN] 第 {offset + 1} 次未命中上周，当前: {actual_label}")
+            if offset < 6:
+                picker.click(timeout=5000)
+                dropdown = wait_for_visible_date_dropdown(page)
+                if not dropdown:
+                    print("    [WARN] 重试时日期弹层未出现")
+                    break
+                week_result = click_week_switch(page)
+                if week_result == 'no_week_switch':
+                    print("    [WARN] 重试时未找到按周查询")
+                    break
+                page.wait_for_timeout(300)
+
+        page.keyboard.press("Escape")
+        page.wait_for_timeout(500)
+        actual_label = get_module_date_label(page, module_selector)
+        print(f"    [WARN] 未确认日期标签切到上周: {expected_start} 至 {expected_end}，当前: {actual_label}")
+        return False
+
+    except Exception as e:
+        print(f"    [WARN] 设置日期失败: {e}")
+        return False
+
+
+def wait_for_visible_date_dropdown(page, timeout_ms=8000):
+    """等待当前可见的日期选择外层弹层，切换面板和日历都在这个容器里。"""
+    deadline = datetime.now().timestamp() + timeout_ms / 1000
+    while datetime.now().timestamp() < deadline:
+        try:
+            dropdowns = page.locator('.jmtd-dropdown-panel:visible').filter(
+                has=page.locator('.jmtd-date-picker-switch-panel')
+            )
+            count = dropdowns.count()
+            if count > 0:
+                return dropdowns.nth(count - 1)
+        except Exception:
+            pass
+        page.wait_for_timeout(300)
+    return None
+
+
+def click_week_switch(page):
+    """真实鼠标点击日期弹层里的“按周查询”。"""
+    try:
+        dropdown = wait_for_visible_date_dropdown(page)
+        if not dropdown:
+            return "no_popup"
+        week_item = dropdown.locator('.jmtd-date-picker-switch-panel-item', has_text="按周查询").first
+        if week_item.count() == 0:
+            return "no_week_switch"
+        current = week_item.evaluate(
+            "el => el.classList.contains('jmtd-date-picker-switch-panel-current') || el.classList.contains('jmtd-date-picker-switch-panel-selected')"
+        )
+        if current:
+            return "already_week"
+        week_item.scroll_into_view_if_needed(timeout=3000)
+        box = week_item.bounding_box(timeout=3000)
+        if not box:
+            return "no_week_box"
+        page.mouse.move(box["x"] + box["width"] / 2, box["y"] + box["height"] / 2)
+        page.wait_for_timeout(100)
+        page.mouse.down()
+        page.wait_for_timeout(80)
+        page.mouse.up()
+        page.wait_for_timeout(800)
+        return "switched_week"
+    except Exception as e:
+        return f"switch_error: {e}"
+
+
+def click_week_day(page, target_year, target_month, target_day):
+    """用 Playwright 真实鼠标点击周模式下的目标日期格。"""
+    try:
+        for _ in range(24):
+            info = page.evaluate("""({targetYear, targetMonth, targetDay}) => {
+                const popups = Array.from(document.querySelectorAll('.jmtd-dropdown-panel'));
+                const popup = popups.reverse().find(el => {
+                    const style = getComputedStyle(el);
+                    return style.display !== 'none' && style.visibility !== 'hidden' && el.querySelector('.jmtd-date-picker-calendar');
+                });
+                if (!popup) return {status: 'no_popup'};
+
+                const header = popup.querySelector('.jmtd-date-picker-header-content, .jmtd-date-picker-header');
+                const text = header ? header.textContent.trim() : '';
+                const m = text.match(/(\\d{4})[^\\d]*(\\d{1,2})/);
+                if (!m) return {status: 'no_header'};
+                const year = parseInt(m[1], 10);
+                const month = parseInt(m[2], 10);
+                const currentTotal = year * 12 + month;
+                const targetTotal = targetYear * 12 + targetMonth;
+
+                if (currentTotal !== targetTotal) {
+                    const selector = currentTotal > targetTotal
+                        ? '.jmtd-date-picker-header-btn-prev-month, [data-event-content="prevMonth"]'
+                        : '.jmtd-date-picker-header-btn-next-month, [data-event-content="nextMonth"]';
+                    const btn = popup.querySelector(selector);
+                    if (!btn) return {status: 'no_nav_btn', year, month};
+                    const r = btn.getBoundingClientRect();
+                    return {status: 'nav', x: r.left + r.width / 2, y: r.top + r.height / 2, year, month};
+                }
+
+                const cells = Array.from(popup.querySelectorAll('td[role="gridcell"], .jmtd-date-picker-cell'));
+                for (const cell of cells) {
+                    const inner = cell.querySelector('.jmtd-date-picker-cell-content-inner');
+                    const txt = (inner ? inner.textContent : cell.textContent || '').trim();
+                    const num = parseInt(txt, 10);
+                    if (num !== targetDay) continue;
+                    const cls = cell.className || '';
+                    if (cls.includes('diff-month') || cls.includes('disabled')) continue;
+                    const r = cell.getBoundingClientRect();
+                    return {status: 'day', x: r.left + r.width / 2, y: r.top + r.height / 2, day: targetDay};
+                }
+                return {status: 'day_not_found', year, month};
+            }""", {"targetYear": target_year, "targetMonth": target_month, "targetDay": target_day})
+
+            status = info.get("status")
+            if status == "nav":
+                page.mouse.click(info["x"], info["y"])
+                page.wait_for_timeout(500)
+                continue
+            if status == "day":
+                page.mouse.click(info["x"], info["y"])
+                page.wait_for_timeout(1000)
+                return f"clicked_week_day_{target_day}"
+            return status
+
+        return "navigate_timeout"
+    except Exception as e:
+        return f"day_click_error: {e}"
+
+
+def wait_for_week_mode(page, timeout_ms=5000):
+    """等待日期弹层左侧切换项确认进入按周查询。"""
+    deadline = datetime.now().timestamp() + timeout_ms / 1000
+    while datetime.now().timestamp() < deadline:
+        try:
+            ok = page.evaluate("""() => {
+                const popups = Array.from(document.querySelectorAll('.jmtd-dropdown-panel'));
+                const popup = popups.reverse().find(el => {
+                    const style = getComputedStyle(el);
+                    return style.display !== 'none' && style.visibility !== 'hidden' && el.querySelector('.jmtd-date-picker-switch-panel');
+                });
+                if (!popup) return false;
+                const current = popup.querySelector('.jmtd-date-picker-switch-panel-current, .jmtd-date-picker-switch-panel-selected');
+                return !!current && current.textContent.includes('按周查询');
+            }""")
+            if ok:
+                return True
+        except Exception:
+            pass
+        page.wait_for_timeout(300)
+    return False
+
+
+def wait_for_module_week_label(page, module_selector, expected_start, expected_end, timeout_ms=10000):
+    """确认模块日期 label 已变成按周查询的上周区间。"""
+    deadline = datetime.now().timestamp() + timeout_ms / 1000
+    while datetime.now().timestamp() < deadline:
+        try:
+            label_text = get_module_date_label(page, module_selector)
+            if '按周查询' in label_text and expected_start in label_text and expected_end in label_text:
+                return True
+        except Exception:
+            pass
+        page.wait_for_timeout(500)
+    return False
+
+
+def get_module_date_label(page, module_selector):
+    """读取模块日期选择器当前显示文案。"""
+    return page.evaluate("""(moduleSelector) => {
+        const module = document.querySelector(moduleSelector);
+        if (!module) return '';
+        const picker = module.querySelector('.jmtd-date-picker');
+        const label = picker ? picker.querySelector('.jmtd-label') : null;
+        return label ? label.textContent.trim() : '';
+    }""", module_selector)
+
+
+def add_spu_to_compare(page, spu):
+    """在搜索面板中搜索并选择指定SPU添加到对比（点击原来竞品位置的空槽位）"""
+    print(f"  添加SPU到对比: {spu}")
+
+    try:
+        close_compare_search_panel(page)
+        scroll_to_compare_top(page)
+        wait_for_competitor_slot_ready(page, timeout_ms=5000)
+
+        # 优先点击竞品侧的空槽位，避免误点第一个本店商品槽位。
+        panel = None
+        for attempt in range(4):
+            try:
+                js_result = page.evaluate("""() => {
+                    const wrapers = document.querySelectorAll('.compare-wraper');
+
+                    function clickSlot(slot) {
+                        slot.scrollIntoView({block: 'center', inline: 'center'});
+                        const rect = slot.getBoundingClientRect();
+                        const x = rect.left + rect.width / 2;
+                        const y = rect.top + rect.height / 2;
+                        for (const type of ['mouseover', 'mousemove', 'mousedown', 'mouseup', 'click']) {
+                            slot.dispatchEvent(new MouseEvent(type, {
+                                bubbles: true,
+                                cancelable: true,
+                                view: window,
+                                clientX: x,
+                                clientY: y
+                            }));
+                        }
+                    }
+
+                    for (let i = 1; i < wrapers.length; i++) {
+                        const wraper = wrapers[i];
+                        const hasContent = !!wraper.querySelector('.compare-content');
+                        const slot = wraper.querySelector('.add.type-add, .type-add, .add');
+                        if (!hasContent && slot) {
+                            clickSlot(slot);
+                            return 'clicked_competitor_empty_slot_' + i;
+                        }
+                    }
+
+                    const greenSlots = Array.from(document.querySelectorAll('.type-add[style*="244, 252, 244"], .add.type-add'));
+                    const slot = greenSlots.find((el) => {
+                        const wraper = el.closest('.compare-wraper');
+                        if (!wraper) return false;
+                        return Array.from(wrapers).indexOf(wraper) > 0;
+                    });
+                    if (slot) {
+                        clickSlot(slot);
+                        return 'clicked_competitor_type_add';
+                    }
+                    return 'not_found';
+                }""")
+                print(f"    [OK] JS点击添加按钮: {js_result}")
+                if not js_result.startswith('clicked'):
+                    page.wait_for_timeout(1000)
+                    continue
+
+                panel = wait_for_visible_search_panel(page, timeout_ms=4000)
+                if panel:
+                    break
+                if click_competitor_empty_slot_with_mouse(page):
+                    panel = wait_for_visible_search_panel(page, timeout_ms=4000)
+                    if panel:
+                        break
+                print(f"    [WARN] 点击后搜索面板未出现，重试({attempt+1}/4)")
+            except Exception as e:
+                print(f"    [WARN] JS点击失败(尝试{attempt+1}): {e}")
+            close_compare_search_panel(page)
+            page.wait_for_timeout(1000)
+
+        if not panel:
+            print("    [FAIL] 添加失败: 搜索面板始终未出现")
+            close_compare_search_panel(page)
+            return False
+
+        search_input = panel.locator('input.jmtd-input-input, .jmtd-base-input input, input[placeholder*="搜索"], input[placeholder*="商品"]').first
+        search_input.wait_for(state="visible", timeout=5000)
+        print("    [OK] 搜索输入框已出现")
+
+        # 在搜索框中输入 SPU 并搜索
+        try:
+            search_input.click()
+            search_input.fill("")
+            page.wait_for_timeout(300)
+            search_input.fill(str(spu))
+            page.wait_for_timeout(500)
+            search_input.press("Enter")
+            print(f"    [OK] 已搜索SPU: {spu}")
+            page.wait_for_timeout(2000)
+        except Exception as e:
+            print(f"    [WARN] 搜索框输入失败: {e}")
+
+        # 点击目标SPU，只在当前可见搜索面板内查找，避免命中过期面板中的旧结果。
+        spu_li = panel.locator(f'li[data-compare-id="{spu}"]').first
+        if spu_li.count() == 0:
+            spu_li = panel.locator('li.li-wrapper', has_text=str(spu)).first
+
+        if spu_li.count() == 0:
+            print(f"    [WARN] 未找到SPU: {spu}")
+            close_compare_search_panel(page)
+            return False
+
+        if not click_search_result_item(page, spu_li):
+            print(f"    [WARN] 点击SPU搜索结果失败: {spu}")
+            close_compare_search_panel(page)
+            return False
+        print(f"    [OK] 已选择SPU: {spu}")
+
+        try:
+            panel.wait_for(state="hidden", timeout=5000)
+        except Exception:
+            close_compare_search_panel(page)
+
+        # 额外等待页面加载新SPU的对比数据
+        page.wait_for_timeout(3000)
+        return True
+
+    except Exception as e:
+        print(f"    [FAIL] 添加失败: {e}")
+        close_compare_search_panel(page)
+        return False
+
+
+def close_compare_search_panel(page):
+    """关闭可能残留的搜索面板，避免下一次添加命中过期输入框。"""
+    try:
+        page.keyboard.press("Escape")
+        page.wait_for_timeout(500)
+    except Exception:
+        pass
+
+
+def scroll_to_compare_top(page):
+    """滚回竞品对比页顶部，确保添加/移除竞品槽位在可操作区域。"""
+    try:
+        page.evaluate("""() => {
+            window.scrollTo(0, 0);
+            document.documentElement.scrollTop = 0;
+            document.body.scrollTop = 0;
+
+            const selectors = [
+                '.container-right',
+                '.container-right-content',
+                '.content-wrapper',
+                '.page-container',
+                '.jmtd-layout-content',
+                '[class*="scroll"]'
+            ];
+            for (const selector of selectors) {
+                document.querySelectorAll(selector).forEach((el) => {
+                    if (el && el.scrollTop > 0) el.scrollTop = 0;
+                });
+            }
+
+            let node = document.querySelector('.compare-container, .compare-wraper, .core-container');
+            while (node) {
+                if (node.scrollTop > 0) node.scrollTop = 0;
+                node = node.parentElement;
+            }
+        }""")
+        page.wait_for_timeout(500)
+    except Exception as e:
+        print(f"    [WARN] 回到页面顶部失败: {e}")
+
+
+def click_competitor_empty_slot_with_mouse(page):
+    """用真实鼠标坐标点击竞品空槽，弥补 JS click 偶尔不触发的问题。"""
+    try:
+        rect = page.evaluate("""() => {
+            const wrapers = Array.from(document.querySelectorAll('.compare-wraper'));
+            for (let i = 1; i < wrapers.length; i++) {
+                const wraper = wrapers[i];
+                if (wraper.querySelector('.compare-content')) continue;
+                const slot = wraper.querySelector('.add.type-add, .type-add, .add');
+                if (!slot) continue;
+                slot.scrollIntoView({block: 'center', inline: 'center'});
+                const r = slot.getBoundingClientRect();
+                return {x: r.left + r.width / 2, y: r.top + r.height / 2};
+            }
+            return null;
+        }""")
+        if not rect:
+            return False
+        page.mouse.move(rect["x"], rect["y"])
+        page.mouse.down()
+        page.mouse.up()
+        page.wait_for_timeout(500)
+        return True
+    except Exception:
+        return False
+
+
+def click_search_result_item(page, item_locator):
+    """用真实鼠标点击搜索结果，避免列表刷新时 Locator.click 等待稳定超时。"""
+    try:
+        item_locator.scroll_into_view_if_needed(timeout=3000)
+        box = item_locator.bounding_box(timeout=3000)
+        if box:
+            page.mouse.move(box["x"] + box["width"] / 2, box["y"] + box["height"] / 2)
+            page.wait_for_timeout(100)
+            page.mouse.down()
+            page.wait_for_timeout(80)
+            page.mouse.up()
+            page.wait_for_timeout(500)
+            return True
+    except Exception:
+        pass
+
+    try:
+        item_locator.click(timeout=5000, force=True)
+        return True
+    except Exception:
+        return False
+
+
+def wait_for_visible_search_panel(page, timeout_ms=10000):
+    """等待并返回当前可见的竞品搜索面板。"""
+    deadline = datetime.now().timestamp() + timeout_ms / 1000
+    while datetime.now().timestamp() < deadline:
+        try:
+            panels = page.locator('.search-panel-list:visible')
+            count = panels.count()
+            if count > 0:
+                return panels.nth(count - 1)
+        except Exception:
+            pass
+        page.wait_for_timeout(300)
+    return None
+
+
+def wait_for_competitor_slot_ready(page, timeout_ms=15000):
+    """等待竞品已被移除，并且竞品侧出现可添加空槽位。"""
+    deadline = datetime.now().timestamp() + timeout_ms / 1000
+    while datetime.now().timestamp() < deadline:
+        try:
+            status = page.evaluate("""() => {
+                const wrapers = Array.from(document.querySelectorAll('.compare-wraper'));
+                const hasCompetitor = wrapers.some((wraper) => {
+                    const tail = wraper.querySelector('.compare-tail');
+                    return tail && tail.textContent.includes('竞品');
+                });
+                const hasCompetitorEmptySlot = wrapers.slice(1).some((wraper) => {
+                    return !wraper.querySelector('.compare-content') && !!wraper.querySelector('.add, .type-add');
+                });
+                return {hasCompetitor, hasCompetitorEmptySlot};
+            }""")
+            if not status.get('hasCompetitor') and status.get('hasCompetitorEmptySlot'):
+                return True
+        except Exception:
+            pass
+        page.wait_for_timeout(500)
+    return False
+
+
+def add_store_product_to_compare(page, store_spu):
+    """点击左侧本店商品的添加按钮，搜索并选中本店SPU"""
+    print(f"  添加本店商品到对比: {store_spu}")
+
+    try:
+        # 点击第一个 .add.type-add（蓝色背景，本店侧）
+        add_btn = page.locator('.add.type-add').first
+        if add_btn.count() == 0:
+            print("    [WARN] 未找到本店添加按钮")
+            return False
+
+        add_btn.click()
+        print("    [OK] 点击本店添加按钮")
+        page.wait_for_timeout(2000)
+
+        # 等待搜索面板
+        panel = page.locator('.search-panel-list').first
+        panel.wait_for(state="visible", timeout=10000)
+        print("    [OK] 搜索面板已出现")
+
+        # 搜索本店SPU
+        try:
+            search_input = panel.locator('input.jmtd-input-input, .jmtd-base-input input, input[placeholder*="搜索"], input[placeholder*="商品"]').first
+            if search_input.count() > 0:
+                search_input.click()
+                search_input.fill("")
+                page.wait_for_timeout(300)
+                search_input.fill(str(store_spu))
+                page.wait_for_timeout(500)
+                search_input.press("Enter")
+                print(f"    [OK] 已搜索本店SPU: {store_spu}")
+                page.wait_for_timeout(2000)
+        except Exception as e:
+            print(f"    [WARN] 搜索框输入失败: {e}")
+
+        # 选中目标SPU
+        spu_li = panel.locator(f'li[data-compare-id="{store_spu}"]').first
+        if spu_li.count() == 0:
+            spu_li = panel.locator('li.li-wrapper', has_text=str(store_spu)).first
+
+        if spu_li.count() == 0:
+            print(f"    [WARN] 搜索面板中未找到本店SPU: {store_spu}")
+            page.keyboard.press("Escape")
+            return False
+
+        if not click_search_result_item(page, spu_li):
+            print(f"    [WARN] 点击本店SPU搜索结果失败: {store_spu}")
+            page.keyboard.press("Escape")
+            return False
+        print(f"    [OK] 已选择本店SPU: {store_spu}")
+        page.wait_for_timeout(3000)
+
+        # 关闭搜索面板
+        try:
+            panel.wait_for(state="hidden", timeout=5000)
+        except Exception:
+            page.keyboard.press("Escape")
+            page.wait_for_timeout(1000)
+
+        return True
+
+    except Exception as e:
+        print(f"    [FAIL] 添加本店商品失败: {e}")
+        page.keyboard.press("Escape")
+        return False
+
+
+def has_store_product(page):
+    """本店商品槽位是否已添加。"""
+    try:
+        return page.locator('.compare-tail', has_text="本店商品").count() > 0
+    except Exception:
+        return False
+
+
+def prepare_product_contrast_once(page):
+    """初始化竞品对比页：只做一次本店商品、竞品空槽和日期。"""
+    print("[Step 12-Init] 初始化竞品对比页...")
+    goto_product_contrasts(page)
+    ensure_spu_tab(page)
+
+    if not has_store_product(page):
+        if not add_store_product_to_compare(page, STORE_SPU):
+            print("  [WARN] 本店商品添加失败，后续数据可能缺少本店对比")
+    else:
+        print("  [OK] 本店SPU已存在，跳过重复添加")
+
+    remove_compared_spu(page)
+
+    print("  设置各模块日期为上周（仅初始化一次）...")
+    set_date_to_last_week(page)
+    page.wait_for_timeout(1000)
+
+    set_module_date(page, '.keyword-container')
+    page.wait_for_timeout(1000)
+
+    # 切一次交易关键词，确保同一关键词模块的日期状态也被组件记录。
+    try:
+        tab = page.locator('.keyword-tabs .jmtd-tabs-tab', has_text="交易关键词").first
+        if tab.count() > 0:
+            tab.click()
+            page.wait_for_timeout(1500)
+            set_module_date(page, '.keyword-container')
+            page.wait_for_timeout(1000)
+        drain_tab = page.locator('.keyword-tabs .jmtd-tabs-tab', has_text="引流关键词").first
+        if drain_tab.count() > 0:
+            drain_tab.click()
+            page.wait_for_timeout(1000)
+    except Exception as e:
+        print(f"  [WARN] 初始化关键词日期失败: {e}")
+
+    set_module_date(page, '.viewsource-container')
+    page.wait_for_timeout(1000)
+
+    set_module_date_by_title(page, "成交客户")
+    page.wait_for_timeout(1000)
+    print("  [OK] 竞品对比页初始化完成")
+
+
+def remove_compared_spu(page):
+    """移除竞品（保留本店商品），通过 .compare-tail 文本识别竞品"""
+    print("  移除已添加的竞品...")
+    try:
+        close_compare_search_panel(page)
+        scroll_to_compare_top(page)
+
+        # 检查是否有竞品（.compare-tail 包含"竞品"文本）
+        has_competitor = page.evaluate("""() => {
+            const tails = document.querySelectorAll('.compare-tail');
+            for (const tail of tails) {
+                if (tail.textContent.includes('竞品')) return true;
+            }
+            return false;
+        }""")
+        if not has_competitor:
+            print("    [INFO] 没有已添加的竞品，跳过移除")
+            return wait_for_competitor_slot_ready(page, timeout_ms=5000)
+
+        # 找到包含"竞品"的 .compare-wraper，点击其关闭按钮
+        result = page.evaluate("""() => {
+            const wrapers = document.querySelectorAll('.compare-wraper');
+            for (const wraper of wrapers) {
+                const tail = wraper.querySelector('.compare-tail');
+                if (tail && tail.textContent.includes('竞品')) {
+                    const closeBtn = wraper.querySelector('.compare-header .jmti-close');
+                    if (closeBtn) {
+                        closeBtn.click();
+                        return 'clicked';
+                    }
+                    return 'no_close_btn_in_wraper';
+                }
+            }
+            return 'no_competitor_wraper';
+        }""")
+        print(f"    JS移除结果: {result}")
+
+        if result.startswith('no_'):
+            print(f"    [WARN] 未找到竞品关闭按钮: {result}")
+            return False
+
+        # 等待弹窗出现并处理确认（多次重试）
+        for attempt in range(5):
+            page.wait_for_timeout(1000)
+            try:
+                confirm = page.locator('.jmtd-modal-footer .jmtd-button-primary').last
+                if confirm.count() > 0 and confirm.is_visible():
+                    confirm.click(timeout=3000)
+                    page.wait_for_timeout(1500)
+                    print("    [OK] 已确认移除")
+                    break
+            except Exception:
+                pass
+        else:
+            print("    [INFO] 未检测到确认弹窗，继续")
+
+        if wait_for_competitor_slot_ready(page):
+            print("    [OK] 竞品槽位已清空")
+            return True
+
+        print("    [WARN] 竞品槽位未在预期时间内清空")
+        return False
+
+    except Exception as e:
+        print(f"    [WARN] 移除失败: {e}")
+        return False
+
+
+def collect_core_metrics(page, ensure_date=False):
+    """收集核心指标数据（含本店+竞品多组卡片和表格）"""
+    print("  收集核心指标...")
+    if ensure_date:
+        set_date_to_last_week(page)
+        page.wait_for_timeout(3000)
+
+    data = extract(page, """() => {
+        const result = {};
+        const container = document.querySelector('.core-container, .container-right-modules');
+        if (!container) return JSON.stringify(result);
+
+        // 提取列头（.summary-label 里的 .summary-card-description 文本）
+        const columnNames = [];
+        const labelCard = container.querySelector('.summary-label .summary-card');
+        if (labelCard) {
+            labelCard.querySelectorAll('.summary-card-description').forEach(desc => {
+                columnNames.push(desc.textContent.trim());
+            });
+        }
+
+        // 提取数据卡片（.summary-customer 里，每张卡有 title + 多个 description）
+        const cards = [];
+        container.querySelectorAll('.summary-customer .summary-card').forEach(card => {
+            const titleEl = card.querySelector('.summary-card-title');
+            const descEls = card.querySelectorAll('.summary-card-description');
+            if (titleEl && descEls.length > 0) {
+                const item = {name: titleEl.textContent.trim()};
+                descEls.forEach((desc, i) => {
+                    const colName = columnNames[i] || ('列' + (i + 1));
+                    item[colName] = desc.textContent.trim();
+                });
+                cards.push(item);
+            }
+        });
+        if (cards.length > 0) result.cards = cards;
+
+        // 提取所有表格（包含多列：指标 | 本店 | 竞品）
+        const tableData = [];
+        container.querySelectorAll('.jmtd-table-main').forEach(table => {
+            const headers = [];
+            table.querySelectorAll('thead th').forEach(th => headers.push(th.textContent.trim()));
+            const rows = [];
+            table.querySelectorAll('tbody tr').forEach(row => {
+                const cells = row.querySelectorAll('td');
+                const obj = {};
+                cells.forEach((cell, i) => { obj[headers[i] || ('col_' + i)] = cell.textContent.trim(); });
+                if (Object.keys(obj).length > 0) rows.push(obj);
+            });
+            if (rows.length > 0) tableData.push({headers, rows});
+        });
+        if (tableData.length > 0) result.tables = tableData;
+
+        return JSON.stringify(result);
+    }""", "{}")
+
+    try:
+        result = json.loads(data) if isinstance(data, str) else {}
+    except:
+        result = {}
+
+    print(f"    核心指标: {len(result.get('tables', []))} 表格, {len(result.get('cards', []))} 张卡片")
+    return result
+
+
+def collect_keyword_section(page, tab_name, ensure_date=False):
+    """收集关键词数据"""
+    print(f"  收集{tab_name}...")
+
+    try:
+        # 点击tab
+        tab = page.locator('.keyword-tabs .jmtd-tabs-tab', has_text=tab_name).first
+        if tab.count() > 0:
+            is_active = tab.evaluate("el => el.classList.contains('jmtd-tabs-tab-active')")
+            if not is_active:
+                tab.click()
+                print(f"    [OK] 已切换{tab_name}")
+                page.wait_for_timeout(3000)
+
+        if ensure_date:
+            set_module_date(page, '.keyword-container')
+            page.wait_for_timeout(3000)
+
+        # 提取表格（含本店+竞品多列）
+        data = extract(page, f"""() => {{
+            const result = [];
+            const section = document.querySelector('.keyword-container');
+            if (!section) return JSON.stringify(result);
+            const tables = section.querySelectorAll('.jmtd-table-main');
+            for (const table of tables) {{
+                const headers = [];
+                table.querySelectorAll('thead th').forEach(th => headers.push(th.textContent.trim()));
+                const rows = [];
+                table.querySelectorAll('tbody tr').forEach(row => {{
+                    const cells = row.querySelectorAll('td');
+                    const obj = {{}};
+                    cells.forEach((cell, i) => {{
+                        obj[headers[i] || ('col_' + i)] = cell.textContent.trim();
+                    }});
+                    if (Object.keys(obj).length > 0) rows.push(obj);
+                }});
+                if (rows.length > 0) result.push({{headers, rows}});
+            }}
+            return JSON.stringify(result);
+        }}""", "[]")
+
+        try:
+            result = json.loads(data) if isinstance(data, str) else []
+        except:
+            result = []
+
+        totalRows = sum(len(t.get('rows', [])) for t in result)
+        print(f"    {tab_name}: {len(result)} 个表格, {totalRows} 条")
+        return result
+
+    except Exception as e:
+        print(f"    [FAIL] 收集{tab_name}失败: {e}")
+        return []
+
+
+def collect_traffic_source(page, ensure_date=False):
+    """收集流量来源对比"""
+    print("  收集流量来源对比...")
+
+    try:
+        if ensure_date:
+            set_module_date(page, '.viewsource-container')
+            page.wait_for_timeout(3000)
+
+        data = extract(page, """() => {
+            const result = [];
+            const sections = document.querySelectorAll('.viewsource-container');
+            for (const section of sections) {
+                const tables = section.querySelectorAll('.jmtd-table-main');
+                for (const table of tables) {
+                    const headers = [];
+                    table.querySelectorAll('thead th').forEach(th => headers.push(th.textContent.trim()));
+                    const rows = [];
+                    table.querySelectorAll('tbody tr').forEach(row => {
+                        const cells = row.querySelectorAll('td');
+                        const obj = {};
+                        cells.forEach((cell, i) => {
+                            obj[headers[i] || ('col_' + i)] = cell.textContent.trim();
+                        });
+                        if (Object.keys(obj).length > 0) rows.push(obj);
+                    });
+                    if (rows.length > 0) result.push({headers, rows});
+                }
+            }
+            return JSON.stringify(result);
+        }""", "[]")
+
+        try:
+            result = json.loads(data) if isinstance(data, str) else []
+        except:
+            result = []
+
+        totalRows = sum(len(t.get('rows', [])) for t in result)
+        print(f"    流量来源: {len(result)} 个表格, {totalRows} 条")
+        return result
+
+    except Exception as e:
+        print(f"    [FAIL] 收集流量来源失败: {e}")
+        return []
+
+
+def collect_customer_profile(page, ensure_date=False):
+    """收集成交客户对比"""
+    print("  收集成交客户对比...")
+
+    try:
+        if ensure_date:
+            set_module_date_by_title(page, "成交客户")
+            page.wait_for_timeout(3000)
+
+        # 检查提示
+        tip = extract(page, """() => {
+            const emptyEls = document.querySelectorAll('.jmtd-empty, .no-data, [class*="empty"]');
+            for (const el of emptyEls) {
+                const text = el.textContent.trim();
+                if (text.includes('人群规模较小') || text.includes('不具有业务统计意义')) {
+                    return text;
+                }
+            }
+            if (document.body.innerText.includes('人群规模较小') && document.body.innerText.includes('不具有业务统计意义')) {
+                return '人群规模较小，不具有业务统计意义，暂无法生成画像';
+            }
+            return '';
+        }""", "")
+
+        if tip and ('人群规模较小' in tip or '不具有业务统计意义' in tip):
+            print(f"    [SKIP] {tip}")
+            return {"data": {}, "tip": tip}
+
+        # 提取数据
+        data = extract(page, """() => {
+            const result = {};
+            const modules = document.querySelectorAll('.container-right-modules');
+            for (const mod of modules) {
+                const h2 = mod.querySelector('h2');
+                if (h2 && h2.textContent.includes('成交客户')) {
+                    const charts = [];
+                    mod.querySelectorAll('.chart, [class*="chart"]').forEach(c => {
+                        const label = c.querySelector('[class*="title"], .legend')?.textContent?.trim();
+                        const value = c.querySelector('[class*="value"], .percent')?.textContent?.trim();
+                        if (label || value) charts.push({label: label || '', value: value || ''});
+                    });
+                    if (charts.length > 0) result.charts = charts;
+
+                    const tables = [];
+                    mod.querySelectorAll('.jmtd-table-main').forEach(table => {
+                        const headers = [];
+                        table.querySelectorAll('thead th').forEach(th => headers.push(th.textContent.trim()));
+                        const rows = [];
+                        table.querySelectorAll('tbody tr').forEach(row => {
+                            const cells = row.querySelectorAll('td');
+                            const obj = {};
+                            cells.forEach((cell, i) => {
+                                obj[headers[i] || ('col_' + i)] = cell.textContent.trim();
+                            });
+                            if (Object.keys(obj).length > 0) rows.push(obj);
+                        });
+                        if (rows.length > 0) tables.push({headers, rows});
+                    });
+                    if (tables.length > 0) result.tables = tables;
+                    break;
+                }
+            }
+            return JSON.stringify(result);
+        }""", "{}")
+
+        try:
+            result = json.loads(data) if isinstance(data, str) else {}
+        except:
+            result = {}
+
+        print(f"    成交客户: {len(result.get('charts', []))} 图表, {len(result.get('tables', []))} 表格")
+        return {"data": result, "tip": ""}
+
+    except Exception as e:
+        print(f"    [FAIL] 收集成交客户失败: {e}")
+        return {"data": {}, "tip": f"失败: {str(e)}"}
+
+
+def screenshot_module(page, module_selector, save_path):
+    """对指定模块截图"""
+    try:
+        module = page.locator(module_selector).first
+        if module.count() > 0:
+            module.screenshot(path=save_path, timeout=10000)
+            print(f"    [OK] 截图已保存: {os.path.basename(save_path)}")
+            return True
+    except Exception as e:
+        print(f"    [WARN] 截图失败 {module_selector}: {e}")
+    return False
+
+
+def generate_spu_markdown(spu_data, screenshot_dir, md_path):
+    """生成SPU详情的Markdown报告"""
+    spu = spu_data.get('spu', '')
+    product = spu_data.get('product', '')
+    brand = spu_data.get('brand', '')
+    store_spu = spu_data.get('本店SPU', '')
+    collect_date = spu_data.get('采集日期', '')
+    week_range = spu_data.get('数据周期', '')
+
+    def img_tag(name):
+        p = os.path.join(screenshot_dir, f"{name}.png")
+        return f"![{name}](./screenshots/{name}.png)" if os.path.exists(p) else ""
+
+    md = []
+    md.append(f"# {brand} {product} - SPU详情")
+    md.append("")
+    md.append(f"**品牌：** {brand}")
+    md.append(f"**商品：** {product}")
+    md.append(f"**SPU：** {spu}")
+    md.append(f"**本店SPU：** {store_spu}")
+    md.append(f"**采集日期：** {collect_date}")
+    md.append(f"**数据周期：** {week_range}")
+    md.append("")
+
+    # 核心指标
+    md.append("## 核心指标")
+    md.append("")
+    img = img_tag("核心指标")
+    if img:
+        md.append(img)
+        md.append("")
+    core = spu_data.get("核心指标", {})
+    # 卡片（每张卡有 name + 多列数值：本店商品、竞品1...）
+    cards = core.get("cards", [])
+    if cards:
+        # 动态获取所有列头
+        sample = cards[0]
+        headers = [k for k in sample.keys() if k != 'name']
+        md.append("| 指标 | " + " | ".join(headers) + " |")
+        md.append("|------|" + "|".join(["------"] * len(headers)) + "|")
+        for c in cards:
+            vals = [c.get(h, "") for h in headers]
+            md.append(f"| {c.get('name', '')} | " + " | ".join(vals) + " |")
+        md.append("")
+    # 表格（含多列）
+    tables = core.get("tables", [])
+    for t in tables:
+        headers = t.get("headers", [])
+        rows = t.get("rows", [])
+        if headers:
+            md.append("### 表格数据")
+            md.append("| " + " | ".join(headers) + " |")
+            md.append("| " + " | ".join(["---"] * len(headers)) + " |")
+        for row in rows:
+            vals = [str(row.get(h, "")) for h in headers]
+            md.append("| " + " | ".join(vals) + " |")
+        md.append("")
+    if not cards and not tables:
+        md.append("*暂无数据*")
+        md.append("")
+    md.append("---")
+    md.append("")
+
+    def render_tables(section_data, section_name):
+        """渲染多列表格数据（本店+竞品）"""
+        md.append(f"## {section_name}")
+        md.append("")
+        img = img_tag(section_name)
+        if img:
+            md.append(img)
+            md.append("")
+        tables = section_data if isinstance(section_data, list) else []
+        if tables:
+            for t in tables:
+                headers = t.get("headers", [])
+                rows = t.get("rows", [])
+                if headers:
+                    md.append("| " + " | ".join(headers) + " |")
+                    md.append("| " + " | ".join(["---"] * len(headers)) + " |")
+                for row in rows:
+                    vals = [str(row.get(h, "")) for h in headers]
+                    md.append("| " + " | ".join(vals) + " |")
+                md.append("")
+        else:
+            md.append("*暂无数据*")
+            md.append("")
+        md.append("---")
+        md.append("")
+
+    render_tables(spu_data.get("引流关键词", []), "引流关键词")
+    render_tables(spu_data.get("交易关键词", []), "交易关键词")
+    render_tables(spu_data.get("流量来源对比", []), "流量来源对比")
+
+    # 成交客户对比
+    md.append("## 成交客户对比")
+    md.append("")
+    img = img_tag("成交客户对比")
+    if img:
+        md.append(img)
+        md.append("")
+    customer = spu_data.get("成交客户对比", {})
+    tip = customer.get("tip", "") if isinstance(customer, dict) else ""
+    if tip:
+        md.append(f"*{tip}*")
+        md.append("")
+    else:
+        cdata = customer.get("data", {}) if isinstance(customer, dict) else {}
+        charts = cdata.get("charts", []) if isinstance(cdata, dict) else []
+        if charts:
+            md.append("### 图表")
+            for c in charts:
+                md.append(f"- {c.get('label', '')}: {c.get('value', '')}")
+            md.append("")
+        tables = cdata.get("tables", []) if isinstance(cdata, dict) else []
+        if tables:
+            for t in tables:
+                headers = t.get("headers", [])
+                rows = t.get("rows", [])
+                if headers:
+                    md.append("### 表格")
+                    md.append("| " + " | ".join(headers) + " |")
+                    md.append("| " + " | ".join(["---"] * len(headers)) + " |")
+                for row in rows:
+                    vals = [str(row.get(h, "")) for h in headers]
+                    md.append("| " + " | ".join(vals) + " |")
+                md.append("")
+        if not charts and not tables:
+            md.append("*暂无数据*")
+            md.append("")
+    md.append("---")
+
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md))
+    print(f"    [OK] Markdown已保存: {md_path}")
+
+
+def collect_single_spu_data(page, spu_item, output_dir):
+    """收集单个SPU的详细数据（含本店商品对比），截图并保存到独立目录"""
+    spu = spu_item['spu']
+    product = spu_item['product']
+    brand = spu_item['brand']
+
+    print(f"\n{'='*50}")
+    print(f"  SPU: {spu}")
+    print(f"  商品: {product[:50]}...")
+    print(f"{'='*50}")
+
+    scroll_to_compare_top(page)
+
+    # 创建目录结构
+    safe_product = re.sub(r'[\\/:*?"<>|]', '', product)[:30]
+    safe_brand = re.sub(r'[\\/:*?"<>|]', '', brand)
+    today = datetime.now().strftime("%Y-%m-%d")
+    last_monday, last_sunday = get_last_week_range()
+    week_range = f"{last_monday.strftime('%Y%m%d')}_{last_sunday.strftime('%Y%m%d')}"
+
+    dir_name = f"{safe_brand}_{safe_product}_{spu}"
+    detail_dir = os.path.join(output_dir, f"SPU详情数据_{today}", dir_name)
+    os.makedirs(detail_dir, exist_ok=True)
+    screenshot_dir = os.path.join(detail_dir, "screenshots")
+    os.makedirs(screenshot_dir, exist_ok=True)
+
+    # 移除已有竞品（通过 .compare-tail 文本识别），为新SPU腾位置
+    existing_competitor = page.locator('.compare-tail', has_text=re.compile(r"竞品\d*")).count() > 0
+    if existing_competitor:
+        if not remove_compared_spu(page):
+            print("    [WARN] 旧竞品未清理干净，交给外层刷新后重试")
+            return None, False
+        page.wait_for_timeout(1000)
+
+    # 添加竞品SPU到空槽位(.type-add)
+    if not add_spu_to_compare(page, spu):
+        return None, False
+
+    # 等待新SPU数据加载到页面
+    print("    等待页面加载新SPU数据...")
+    for attempt in range(30):
+        found = page.evaluate(f"""() => {{
+            const els = document.querySelectorAll('.sxu-id, .content-title');
+            for (const el of els) {{
+                if (el.textContent.includes('{spu}')) return true;
+            }}
+            return false;
+        }}""")
+        if found:
+            print("    [OK] 页面已显示新SPU")
+            break
+        page.wait_for_timeout(1000)
+    else:
+        print("    [WARN] 页面可能未刷新到新SPU，继续采集")
+
+    # 再额外等待数据完全加载
+    page.wait_for_timeout(3000)
+
+    # 收集核心指标
+    core_metrics = collect_core_metrics(page, ensure_date=False)
+    screenshot_module(page, '.core-container', os.path.join(screenshot_dir, "核心指标.png"))
+    page.wait_for_timeout(2000)
+
+    # 收集引流关键词
+    drain_keywords = collect_keyword_section(page, "引流关键词", ensure_date=False)
+    screenshot_module(page, '.keyword-container', os.path.join(screenshot_dir, "引流关键词.png"))
+    page.wait_for_timeout(2000)
+
+    # 收集交易关键词
+    trade_keywords = collect_keyword_section(page, "交易关键词", ensure_date=False)
+    screenshot_module(page, '.keyword-container', os.path.join(screenshot_dir, "交易关键词.png"))
+    page.wait_for_timeout(2000)
+
+    # 收集流量来源
+    traffic_source = collect_traffic_source(page, ensure_date=False)
+    screenshot_module(page, '.viewsource-container', os.path.join(screenshot_dir, "流量来源对比.png"))
+    page.wait_for_timeout(2000)
+
+    # 收集成交客户
+    customer_profile = collect_customer_profile(page, ensure_date=False)
+    # 成交客户模块没有固定选择器，尝试找包含"成交客户"的模块
+    try:
+        customer_module = page.locator('.container-right-modules', has_text="成交客户").first
+        if customer_module.count() > 0:
+            customer_module.screenshot(path=os.path.join(screenshot_dir, "成交客户对比.png"), timeout=10000)
+            print(f"    [OK] 截图已保存: 成交客户对比.png")
+    except Exception as e:
+        print(f"    [WARN] 截图失败 成交客户对比: {e}")
+
+    # 移除竞品SPU（为下一个SPU腾位置）
+    removed = remove_compared_spu(page)
+    scroll_to_compare_top(page)
+
+    # 构建数据
+    data = {
+        "spu": spu,
+        "product": product,
+        "brand": brand,
+        "本店SPU": STORE_SPU,
+        "采集日期": today,
+        "数据周期": f"{last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}",
+        "核心指标": core_metrics,
+        "引流关键词": drain_keywords,
+        "交易关键词": trade_keywords,
+        "流量来源对比": traffic_source,
+        "成交客户对比": customer_profile,
+    }
+
+    # 保存JSON
+    json_name = f"{safe_brand}_{safe_product}_{spu}_{week_range}.json"
+    json_path = os.path.join(detail_dir, json_name)
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(data, f, ensure_ascii=False, indent=2)
+    print(f"    [OK] JSON已保存: {json_path}")
+
+    # 生成Markdown
+    md_name = f"{safe_brand}_{safe_product}_{spu}.md"
+    md_path = os.path.join(detail_dir, md_name)
+    generate_spu_markdown(data, screenshot_dir, md_path)
+
+    print(f"    [OK] SPU详情已保存到目录: {detail_dir}")
+    return data, removed
+
+
+def collect_each_spu_detail(page, spus, output_dir, resume=False):
+    """逐个收集每个SPU的详细数据"""
+    print(f"\n[Step 12] 开始收集 {len(spus)} 个SPU的详细数据...")
+    progress = SpuProgress(output_dir, "peer_product_spu_detail")
+
+    # resume 模式：跳过已有 JSON 的 SPU
+    if resume:
+        completed_spus = progress.successful_spus()
+        today = datetime.now().strftime("%Y-%m-%d")
+        detail_base = os.path.join(output_dir, f"SPU详情数据_{today}")
+        skipped = 0
+        filtered = []
+        for item in spus:
+            spu = item['spu']
+            product = item['product']
+            brand = item['brand']
+            safe_product = re.sub(r'[\\/:*?"<>|]', '', product)[:30]
+            safe_brand = re.sub(r'[\\/:*?"<>|]', '', brand)
+            dir_name = f"{safe_brand}_{safe_product}_{spu}"
+            spu_dir = os.path.join(detail_base, dir_name)
+            # 检查目录下是否有 JSON 文件
+            has_output = os.path.isdir(spu_dir) and any(f.endswith('.json') for f in os.listdir(spu_dir))
+            if spu in completed_spus or has_output:
+                skipped += 1
+            else:
+                filtered.append(item)
+        if skipped:
+            print(f"  [Resume] 跳过已成功采集的 {skipped} 个 SPU，剩余 {len(filtered)} 个待采集")
+        spus = filtered
+
+    if not spus:
+        print("  [OK] 所有 SPU 均已采集，无需继续")
+        return 0, 0
+
+    prepare_product_contrast_once(page)
+
+    success = 0
+    failed = 0
+
+    for idx, item in enumerate(spus):
+        print(f"\n  进度: [{idx+1}/{len(spus)}]")
+        progress.mark_running(item, idx + 1, len(spus))
+
+        try:
+            data, removed = collect_single_spu_data(page, item, output_dir)
+            if data:
+                success += 1
+                progress.mark_success(item, removed=removed)
+                if not removed:
+                    print("  移除失败，刷新页面...")
+                    prepare_product_contrast_once(page)
+            else:
+                failed += 1
+                progress.append("failed", item, error="collect_single_spu_data 返回空数据")
+                prepare_product_contrast_once(page)
+        except Exception as e:
+            print(f"    [FAIL] 异常: {e}")
+            failed += 1
+            progress.mark_failed(item, e)
+            try:
+                prepare_product_contrast_once(page)
+            except Exception:
+                pass
+
+        page.wait_for_timeout(2000)
+
+    print(f"\n  [OK] 完成: 成功 {success}, 失败 {failed}")
+    print(f"  [Progress] 进度日志: {progress.path}")
+    return success, failed
+
+
+# ── 断点续爬辅助 ─────────────────────────────────────────────────────────────
+
+def _load_spus_from_existing(output_dir):
+    """从已有的 SPU 详情目录中重建 SPU 列表（用于 resume 模式跳过步骤8-9后）"""
+    today = datetime.now().strftime("%Y-%m-%d")
+    detail_base = os.path.join(output_dir, f"SPU详情数据_{today}")
+    if not os.path.isdir(detail_base):
+        return []
+
+    spus = []
+    for dir_name in os.listdir(detail_base):
+        dir_path = os.path.join(detail_base, dir_name)
+        if not os.path.isdir(dir_path):
+            continue
+        # 格式: {brand}_{product}_{spu}
+        parts = dir_name.rsplit("_", 1)
+        if len(parts) < 2:
+            continue
+        spu = parts[-1]
+        brand_product = parts[0]
+        # 从 JSON 文件读取 brand 信息
+        json_files = [f for f in os.listdir(dir_path) if f.endswith(".json")]
+        brand = ""
+        product = ""
+        if json_files:
+            try:
+                with open(os.path.join(dir_path, json_files[0]), "r", encoding="utf-8") as f:
+                    data = json.load(f)
+                brand = data.get("brand", "")
+                product = data.get("product", "")
+            except Exception:
+                pass
+        if not brand:
+            # 从目录名推断 brand（第一个 _ 前的部分）
+            brand = brand_product.split("_")[0] if "_" in brand_product else ""
+        if not product:
+            product = brand_product
+        spus.append({"spu": spu, "product": product, "brand": brand})
+    print(f"  [Resume] 从已有目录恢复 {len(spus)} 个 SPU")
+    return spus
+
+
+# ── 主流程 ─────────────────────────────────────────────────────────────────────
+
+def main():
+    parser = argparse.ArgumentParser(description="京东商品榜单竞品数据采集脚本")
+    parser.add_argument("--shop", default=DEFAULT_CONFIG["shop"], help="店铺账号名")
+    parser.add_argument("--password", default=DEFAULT_CONFIG["password"], help="登录密码")
+    parser.add_argument(
+        "--output",
+        default=os.getenv(
+            "JD_PEER_PRODUCT_OUTPUT_DIR",
+            str(RAW_DATA_ROOT / "jd-vendor" / "竞品数据"),
+        ),
+        help="输出目录",
+    )
+    parser.add_argument("--resume", action="store_true", help="跳过已成功采集的SPU，从断点继续")
+    args = parser.parse_args()
+    args.output = str(managed_data_path(args.output))
+
+    print("=" * 60)
+    print("京东商品榜单竞品数据采集（含SKU/SPU）")
+    print(f"  店铺: {args.shop}")
+    print(f"  品牌: {', '.join(BRANDS)}")
+    print(f"  类目: {', '.join(c[0] for c in CATEGORIES)}")
+    print("=" * 60)
+
+    last_monday, last_sunday = get_last_week_range()
+    print(f"  上周: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+    print()
+
+    with sync_playwright() as p:
+        browser = p.chromium.launch(headless=False)
+        context = browser.new_context(
+            viewport={"width": 1920, "height": 1080},
+            locale="zh-CN",
+        )
+        page = context.new_page()
+
+        try:
+            # Step 1: 登录（复用 jd_data_collector）
+            step_login(page, args.shop, args.password)
+
+            # Step 2: 进入商智（复用 jd_data_collector，会打开新标签页）
+            sz_page = step_enter_shangzhi(context, page)
+
+            # Step 3: 返回旧版
+            step_return_old_version(sz_page)
+
+            # Step 4: 跳转到商品榜单页面
+            step_goto_product_ranks(sz_page)
+
+            # Step 5: 采集数据（传入 context 用于打开新标签页获取SPU）
+            results = collect_data(sz_page, context)
+
+            # 保存初始采集结果（仅含 SKU，后续步骤会补充 SPU）
+            save_results(results, args.output)
+
+            # Step 6: 跳转到竞品分析配置页面
+            step_goto_competition_config(sz_page)
+
+            # Step 7: 切换到竞品配置
+            click_competitor_config_tab(sz_page)
+
+            # Step 8-9: 删除旧竞品 & 添加新 SKU（今天执行过则跳过）
+            today = datetime.now().strftime("%Y-%m-%d")
+            marker = os.path.join(args.output, f".sku_config_done_{today}")
+            spus = []
+            if os.path.exists(marker):
+                print(f"\n[Step 8-9] 今天的 SKU 配置已执行过，跳过删除/添加步骤")
+                # 从已有目录恢复 SPU 列表
+                spus = _load_spus_from_existing(args.output)
+            else:
+                # Step 8: 删除所有已监控竞品
+                remove_all_monitored_products(sz_page)
+
+                # Step 9: 添加采集到的 SKU 到监控列表（同时收集 SPU）
+                add_spus_to_monitor(sz_page, results)
+
+                # 标记今天已完成
+                with open(marker, "w") as f:
+                    f.write(today)
+                print(f"  [OK] SKU 配置标记已写入: {marker}")
+
+                # 从 results 构建 SPU 列表
+                seen = set()
+                for brand_name in BRANDS:
+                    for cat_name, _ in CATEGORIES:
+                        rows = results.get(brand_name, {}).get(cat_name, [])
+                        for row in rows:
+                            spu = row.get('spu', '')
+                            if spu and spu not in seen:
+                                seen.add(spu)
+                                spus.append({
+                                    'spu': spu,
+                                    'product': row.get('product', ''),
+                                    'brand': brand_name
+                                })
+
+            if not spus:
+                print("\n[WARN] 未收集到任何 SPU，跳过竞品对比详情采集")
+            else:
+                print(f"\n[INFO] 共 {len(spus)} 个 SPU 需要采集对比详情")
+                collect_each_spu_detail(sz_page, spus, args.output, resume=args.resume)
+
+            # 保存结果
+            md_path, sku_path = save_results(results, args.output)
+
+            print()
+            print("=" * 60)
+            print("采集完成！")
+            print(f"  详细数据: {md_path}")
+            print(f"  SPU 汇总: {sku_path}")
+            print("=" * 60)
+
+        except KeyboardInterrupt:
+            print("\n用户中断，退出。")
+        except Exception as e:
+            print(f"\n[ERROR] 出错: {e}")
+            import traceback
+            traceback.print_exc()
+
+        finally:
+            context.close()
+            browser.close()
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_product_data_collector.py b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_product_data_collector.py
new file mode 100644
index 0000000..803cef2
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/jd_product_data_collector.py
@@ -0,0 +1,3604 @@
+"""
+京东单品（商品明细）数据采集脚本
+1. 在商品明细页面翻页采集所有 SPU
+2. 逐个 SPU 搜索 → 点击详情 → 新页面
+3. 在新页面切到"核心数据"tab，采集指标（含比较数据）
+4. 结果保存到 单品详情数据_日期/ 目录
+
+依赖:
+    pip install playwright
+    playwright install chromium
+
+用法:
+    python jd_product_data_collector.py
+    python jd_product_data_collector.py --shop "店铺名" --password "密码"
+"""
+
+import argparse
+import json
+import math
+import os
+import re
+from datetime import datetime, timedelta
+
+from playwright.sync_api import sync_playwright, BrowserContext
+
+from collection_progress import SpuProgress
+
+from jd_data_collector import (
+    DEFAULT_CONFIG,
+    step_login,
+    step_enter_shangzhi,
+    extract,
+)
+from jd_peer_product_data_collector import step_return_old_version
+from config import PRODUCT_OUTPUT
+from runtime_paths import managed_data_path
+
+
+# ── 常量 ──────────────────────────────────────────────────────────────────────
+
+OUTPUT_BASE = PRODUCT_OUTPUT
+
+# 阿波罗2 两个 SPU，合并统计
+APOLLO2_SPU_LIST = ["10028594796004", "10028595026975"]
+APOLLO2_FOLDER = "阿波罗2"
+
+# 宙斯3 两个 SPU，合并统计
+ZEUS3_SPU_LIST = ["10033333485894", "10025345724491"]
+ZEUS3_FOLDER = "宙斯3"
+
+# 星迹 三个 SPU，合并统计
+XINGJI_SPU_LIST = ["10033913254016", "10025329514925", "10025329599430"]
+XINGJI_FOLDER = "星迹"
+
+STYLE_SPU_GROUPS = [
+    {"style": "盖亚斜挎", "spus": ["10022621048112", "10023155725944", "10030097050220"]},
+    {"style": "极星pro", "spus": ["10025628785945", "10026601031834", "10026600460605"]},
+    {"style": "盖亚微单pro", "spus": ["10025310117864", "10023909449487"]},
+    {"style": "盖世m1", "spus": ["10026326257225", "10029721108947"]},
+    {"style": "星迹2", "spus": ["10033913254016", "10025329514925", "10025329599430"]},
+    {"style": "星云2", "spus": ["10027762032411"]},
+    {"style": "宙斯", "spus": ["10033333485894", "10025292822415", "10028005218343", "10025020862662", "10032720707692", "10028005207681"]},
+    {"style": "阿波罗2", "spus": ["10028594796004", "10028595026975"]},
+    {"style": "阿波罗x1", "spus": ["10024429591100", "10024753441908"]},
+    {"style": "波塞冬edge", "spus": ["10032366483452"]},
+    {"style": "逐星", "spus": ["10026025497958", "10031326808929"]},
+    {"style": "逐星GT", "spus": ["10033470328729"]},
+    {"style": "星云mini", "spus": ["10032211469888"]},
+    {"style": "晨星2", "spus": ["10031295279033"]},
+    {"style": "极星双肩", "spus": ["10029986282219", "10029920288281", "10033941202259"]},
+    {"style": "凌云air", "spus": ["10029593796467"]},
+    {"style": "觅光", "spus": ["10033787463147", "10033860225085"]},
+    {"style": "云卷2", "spus": ["10032963233896", "10034088808495"]},
+    {"style": "凌云2", "spus": ["10028043067231"]},
+    {"style": "瑞白", "spus": ["10033366316587"]},
+    {"style": "黑曜石山司包", "spus": ["10033995817475"]},
+    {"style": "黑曜石托特包", "spus": ["10033995834580"]},
+    {"style": "曜影", "spus": ["10034534903226"]},
+    {"style": "黑曜石电脑包", "spus": ["10034651877353"]},
+    {"style": "凌云3", "spus": ["10034906180847"]},
+    {"style": "拓界2", "spus": ["10034905606265"]},
+]
+
+
+# ── 页面导航 ──────────────────────────────────────────────────────────────────
+
+def step_goto_product_detail(page):
+    """跳转到商品明细页面"""
+    print("[Step 4] 跳转到商品明细页面...")
+    page.goto(
+        "https://sz.jd.com/szweb/sz/view/productAnalysis/productDetail.html",
+        wait_until="networkidle"
+    )
+    page.wait_for_timeout(5000)
+    print("  [OK] 已到达商品明细页面")
+
+
+def close_guide_if_present(page):
+    """检测并关闭新功能引导弹窗"""
+    print("  检测引导弹窗...")
+    try:
+        page.locator('.jmtp-guide-content, .jmtp-guide-con').first.wait_for(state="visible", timeout=3000)
+        page.locator('.jmtp-guide-footer button', has_text="完成").first.click()
+        print("    [OK] 引导弹窗已关闭")
+        page.wait_for_timeout(2000)
+    except Exception:
+        print("    [INFO] 无引导弹窗")
+
+
+# ── 日期选择 ───────────────────────────────────────────────────────────────────
+
+def get_yesterday_date():
+    return datetime.now() - timedelta(days=1)
+
+
+def get_yesterday_range_text():
+    day = get_yesterday_date().strftime("%Y-%m-%d")
+    return day, day
+
+
+def _date_text_matches_day(text, day):
+    if not text:
+        return False
+    normalized = re.sub(r"\s+", "", text)
+    return (
+        day in text
+        and (
+            f"{day}~{day}" in normalized
+            or f"{day}-{day}" in normalized
+            or f"{day}至{day}" in normalized
+            or normalized.count(day) >= 1
+        )
+    )
+
+
+def _combo_date_label(page):
+    try:
+        return page.locator('.jmtd-date-picker-combo-tag-echo .jmtd-label').first.inner_text(timeout=3000).strip()
+    except Exception:
+        return ""
+
+
+def _click_combo_mode(page, mode):
+    try:
+        tag = page.locator(f'.jmtd-date-picker-combo-tag[data-event-content="{mode}"]').first
+        tag.wait_for(state="visible", timeout=5000)
+        tag.click()
+        return True
+    except Exception:
+        return page.evaluate(
+            """(modeText) => {
+                for (const tag of document.querySelectorAll('.jmtd-date-picker-combo-tag')) {
+                    if ((tag.textContent || '').trim() === modeText) {
+                        tag.click();
+                        return true;
+                    }
+                }
+                return false;
+            }""",
+            "天" if mode == "day" else mode,
+        )
+
+
+def _click_combo_day_flip(page, direction):
+    event_name = "dayPrevClick" if direction == "prev" else "dayNextClick"
+    event_content = "dayPrev" if direction == "prev" else "dayNext"
+    click_point = page.evaluate(
+        r"""({eventName, eventContent}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const buttons = Array.from(document.querySelectorAll(
+                `button[data-event-name="${eventName}"], button[data-event-content="${eventContent}"]`
+            )).filter((btn) => visible(btn) && !btn.disabled && btn.getAttribute('aria-disabled') !== 'true');
+            const mainButtons = buttons.filter((btn) => !btn.closest('.jmtd-popover-portal, .jmtd-tooltip-portal'));
+            const btn = mainButtons[0] || buttons[0];
+            if (!btn) return null;
+            const rect = btn.getBoundingClientRect();
+            return { x: rect.left + rect.width / 2, y: rect.top + rect.height / 2 };
+        }""",
+        {"eventName": event_name, "eventContent": event_content},
+    )
+    if not click_point:
+        return False
+    page.mouse.click(click_point["x"], click_point["y"])
+    return True
+
+
+def _select_yesterday_by_combo_flip(page, target_day):
+    for _ in range(8):
+        label = _combo_date_label(page)
+        if _date_text_matches_day(label, target_day):
+            print(f"    [OK] 日期已是昨天: {label}")
+            return True
+
+        dates = re.findall(r"\d{4}-\d{2}-\d{2}", label)
+        if len(dates) >= 1 and dates[0] < target_day:
+            direction = "next"
+        else:
+            direction = "prev"
+
+        if not _click_combo_day_flip(page, direction):
+            return False
+        page.wait_for_timeout(1500)
+
+    return _date_text_matches_day(_combo_date_label(page), target_day)
+
+
+def _open_combo_date_panel(page):
+    for action in ("hover", "click"):
+        try:
+            echo = page.locator('.jmtd-date-picker-combo-tag-echo').first
+            if action == "hover":
+                echo.hover(timeout=5000)
+            else:
+                echo.click(timeout=5000)
+            page.wait_for_timeout(1200)
+            if page.locator('.jmtd-date-picker-panel, .jmtd-dropdown-list, .jmtd-popover-open').first.count() > 0:
+                return True
+        except Exception:
+            pass
+    return False
+
+
+def _click_day_in_open_datepicker(page, target_date, label):
+    target_year = target_date.year
+    target_month = target_date.month
+    target_day = target_date.day
+
+    for _ in range(12):
+        nav_info = page.evaluate(
+            r"""({targetYear, targetMonth}) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && rect.width > 0
+                        && rect.height > 0;
+                };
+                const roots = Array.from(document.querySelectorAll(
+                    '.jmtd-popover-open, .jmtd-date-picker-panel, .jmtd-date-picker-week-panel, ' +
+                    '.dropdown-menu, .ant-picker-dropdown, .el-picker-panel, [class*="datepicker"], [class*="calendar"], [class*="date-picker"]'
+                )).filter((el) => visible(el) && /(20\d{2})[-/.年]\s*(\d{1,2})/.test(el.textContent || ''));
+                const root = roots
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        const hasHeader = !!el.querySelector('.jmtd-date-picker-header-content, .jmt-date-picker-header, .ant-calendar-header, .ant-picker-header, .el-date-picker__header');
+                        const hasCells = !!el.querySelector('td, [role="gridcell"], button, .ant-calendar-date');
+                        return { el, score: (hasHeader ? 1000000000 : 0) + (hasCells ? 100000000 : 0) - rect.width * rect.height };
+                    })
+                    .sort((a, b) => b.score - a.score)[0]?.el || document;
+                const match = (root.textContent || '').match(/(20\d{2})[-/.年]\s*(\d{1,2})/);
+                if (!match) return { status: 'no_month_title' };
+                const currentYear = Number(match[1]);
+                const currentMonth = Number(match[2]);
+                if (currentYear === targetYear && currentMonth === targetMonth) {
+                    return { status: 'aligned', currentYear, currentMonth };
+                }
+                const diff = (targetYear - currentYear) * 12 + (targetMonth - currentMonth);
+                const selectors = diff < 0
+                    ? ['.jmtd-date-picker-header-btn-prev-month', '[data-event-content="prevMonth"]', '.ant-picker-header-prev-btn', '.el-picker-panel__icon-btn.prev-month']
+                    : ['.jmtd-date-picker-header-btn-next-month', '[data-event-content="nextMonth"]', '.ant-picker-header-next-btn', '.el-picker-panel__icon-btn.next-month'];
+                for (const selector of selectors) {
+                    const btn = Array.from(root.querySelectorAll(selector)).find(visible)
+                        || Array.from(document.querySelectorAll(selector)).find(visible);
+                    if (btn) {
+                        const rect = btn.getBoundingClientRect();
+                        return { status: 'click', x: rect.left + rect.width / 2, y: rect.top + rect.height / 2, currentYear, currentMonth };
+                    }
+                }
+                return { status: 'no_nav', currentYear, currentMonth, diff };
+            }""",
+            {"targetYear": target_year, "targetMonth": target_month},
+        )
+        if not isinstance(nav_info, dict):
+            nav_info = {"status": str(nav_info)}
+        if nav_info.get("status") == "aligned":
+            break
+        if nav_info.get("status") == "click":
+            page.mouse.click(nav_info["x"], nav_info["y"])
+            page.wait_for_timeout(800)
+            continue
+        break
+
+    click_info = page.evaluate(
+        r"""({targetDay}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && rect.width > 0
+                    && rect.height > 0;
+            };
+            const disabled = (el) => {
+                const classText = `${el.className || ''} ${el.parentElement?.className || ''}`;
+                return el.disabled
+                    || el.getAttribute('aria-disabled') === 'true'
+                    || /disabled|off|unavailable|diff-month|other-month|prev-month|next-month/.test(classText);
+            };
+            const roots = Array.from(document.querySelectorAll(
+                '.jmtd-popover-open, .jmtd-date-picker-panel, .jmtd-date-picker-week-panel, .dropdown-menu, ' +
+                '.ant-picker-dropdown, .el-picker-panel, [class*="datepicker"], [class*="calendar"], [class*="date-picker"]'
+            )).filter(visible);
+            const root = roots
+                .map((el) => {
+                    const rect = el.getBoundingClientRect();
+                    const hasHeader = !!el.querySelector('.jmtd-date-picker-header-content, .jmt-date-picker-header, .ant-calendar-header, .ant-picker-header, .el-date-picker__header');
+                    const hasCells = !!el.querySelector('td, [role="gridcell"], button, .ant-calendar-date');
+                    return { el, score: (hasHeader ? 1000000000 : 0) + (hasCells ? 100000000 : 0) - rect.width * rect.height };
+                })
+                .sort((a, b) => b.score - a.score)[0]?.el || document;
+            const wanted = String(targetDay);
+            const cells = Array.from(root.querySelectorAll('td[role="gridcell"], [role="gridcell"], .ant-calendar-date, .ant-picker-cell, .el-date-table td, td'))
+                .filter(visible)
+                .filter((cell) => !disabled(cell));
+            for (const cell of cells) {
+                const text = (cell.textContent || '').trim().replace(/^0+(\d)$/, '$1');
+                if (text === wanted) {
+                    const rect = cell.getBoundingClientRect();
+                    return { status: 'found', x: rect.left + rect.width / 2, y: rect.top + rect.height / 2, text };
+                }
+            }
+            return { status: 'not_found' };
+        }""",
+        {"targetDay": target_day},
+    )
+    print(f"    日期选择结果: {click_info}")
+    if isinstance(click_info, dict) and click_info.get("status") == "found":
+        page.mouse.click(click_info["x"], click_info["y"])
+        page.wait_for_timeout(2000)
+        return True
+
+    try:
+        from jd_data_collector import save_date_picker_debug
+        save_date_picker_debug(page, f"{label}_yesterday_not_found")
+    except Exception:
+        pass
+    return False
+
+
+def select_yesterday_in_combo_datepicker(page, label="product_date"):
+    target = get_yesterday_date()
+    target_day = target.strftime("%Y-%m-%d")
+    print(f"  设置日期为昨天: {target_day}")
+
+    if _date_text_matches_day(_combo_date_label(page), target_day):
+        print(f"    [OK] 日期已是昨天: {_combo_date_label(page)}")
+        return True
+
+    _click_combo_mode(page, "day")
+    page.wait_for_timeout(1200)
+
+    if _select_yesterday_by_combo_flip(page, target_day):
+        print("  [OK] 已切换到昨天数据")
+        return True
+
+    _open_combo_date_panel(page)
+    if _click_day_in_open_datepicker(page, target, label):
+        if _date_text_matches_day(_combo_date_label(page), target_day):
+            print("  [OK] 已切换到昨天数据")
+            return True
+
+    raise RuntimeError(f"未能选择昨天日期: {target_day}")
+
+
+def select_week_and_last_week(page):
+    """选择天模式并切换到昨天。函数名保留，避免改动调用链。"""
+    print("  选择时间范围为'天'并切换到昨天...")
+    select_yesterday_in_combo_datepicker(page, "product_detail_list")
+    page.wait_for_timeout(5000)
+
+
+# ── 商品明细页：翻页采集所有 SPU ─────────────────────────────────────────────
+
+def extract_product_detail_rows(page):
+    """提取当前页商品明细表格数据（排除合计行）"""
+    data = extract(page, """() => {
+        const rows = document.querySelectorAll('.jmtd-table-main tbody tr');
+        const result = [];
+        for (const row of rows) {
+            if (row.classList.contains('jmtd-table-plugin-total-row')) continue;
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 2) continue;
+
+            const productCell = cells[0];
+            const nameEl = productCell.querySelector('.jmtd-table-formatter-name-name .jmtd-label');
+            const productName = nameEl ? nameEl.textContent.trim() : '';
+
+            const descEl = productCell.querySelector('.jmtd-table-formatter-name-desc');
+            const descText = descEl ? descEl.textContent.trim() : '';
+            const spuMatch = descText.match(/SPU[\\s:：]*(\\d+)/);
+            const spu = spuMatch ? spuMatch[1] : '';
+
+            const imgEl = productCell.querySelector('.jmtd-table-formatter-name-img');
+            let imgSrc = imgEl ? (imgEl.getAttribute('src') || imgEl.getAttribute('data-src') || '') : '';
+            if (imgSrc.startsWith('//')) imgSrc = 'https:' + imgSrc;
+
+            const linkEl = productCell.querySelector('.jmtd-table-formatter-name-link');
+            let productUrl = linkEl ? (linkEl.getAttribute('href') || '') : '';
+            if (productUrl.startsWith('//')) productUrl = 'https:' + productUrl;
+
+            function extractMetric(cell) {
+                const wrap = cell.querySelector('.rate-target-warp');
+                if (!wrap) return { value: cell.textContent.trim(), compare: '', rate: '' };
+                return {
+                    value: wrap.querySelector('.first-value')?.textContent?.trim() || '',
+                    compare: wrap.querySelector('.second-value')?.textContent?.trim() || '',
+                    rate: wrap.querySelector('.target-value')?.textContent?.trim() || '',
+                };
+            }
+            const headers = ['访客数','浏览量','成交客户数','成交商品件数','成交金额'];
+            const metrics = {};
+            for (let i = 1; i < cells.length; i++) {
+                metrics[headers[i-1] || 'col_'+i] = extractMetric(cells[i]);
+            }
+            result.push({ spu, product: productName, image: imgSrc, url: productUrl, metrics });
+        }
+        return JSON.stringify(result);
+    }""", "[]")
+    try:
+        return json.loads(data) if isinstance(data, str) else []
+    except:
+        return []
+
+
+def has_next_page(page):
+    try:
+        btn = page.locator('.jmtd-pagination-item-next').first
+        return btn.count() > 0 and not btn.evaluate("el => el.classList.contains('jmtd-pagination-item-disabled')")
+    except:
+        return False
+
+
+def click_next_page(page):
+    try:
+        # 记录当前第一条 SPU，用于判断翻页后是否刷新
+        first_spu_before = page.evaluate("""() => {
+            const desc = document.querySelector('.jmtd-table-main tbody tr:not(.jmtd-table-plugin-total-row) .jmtd-table-formatter-name-desc');
+            if (!desc) return '';
+            const m = desc.textContent.match(/SPU[\\s:：]*(\\d+)/);
+            return m ? m[1] : '';
+        }""")
+
+        page.locator('.jmtd-pagination-item-next').first.click()
+        print("    [OK] 点击下一页")
+
+        # 智能等待：等表格第一条 SPU 变化 或 loading 消失，最多 4 秒
+        for _ in range(20):
+            page.wait_for_timeout(200)
+            # 检查是否还有 loading
+            loading = page.evaluate("""() => {
+                const el = document.querySelector('.jmtd-loading');
+                return el ? window.getComputedStyle(el).display !== 'none' : false;
+            }""")
+            if loading:
+                continue
+            # 检查第一条 SPU 是否已变化
+            first_spu_after = page.evaluate("""() => {
+                const desc = document.querySelector('.jmtd-table-main tbody tr:not(.jmtd-table-plugin-total-row) .jmtd-table-formatter-name-desc');
+                if (!desc) return '';
+                const m = desc.textContent.match(/SPU[\\s:：]*(\\d+)/);
+                return m ? m[1] : '';
+            }""")
+            if first_spu_after and first_spu_after != first_spu_before:
+                break
+        page.wait_for_timeout(1000)
+        return True
+    except Exception as e:
+        print(f"    [WARN] 点击下一页异常: {e}")
+        return False
+
+
+def collect_all_spus(page):
+    """翻页采集商品明细所有 SPU"""
+    print("[Step 5] 翻页采集商品明细所有 SPU...")
+    select_week_and_last_week(page)
+
+    all_rows = []
+    page_num = 1
+    while True:
+        print(f"\n  采集第 {page_num} 页...")
+        rows = extract_product_detail_rows(page)
+        print(f"    提取到 {len(rows)} 条")
+        for r in rows:
+            all_rows.append(r)
+            print(f"      SPU: {r.get('spu','')} | {r.get('product','')[:40]}...")
+        if not has_next_page(page):
+            print(f"\n  [OK] 共采集 {len(all_rows)} 个 SPU")
+            break
+        click_next_page(page)
+        page_num += 1
+    return all_rows
+
+
+# ── 逐个 SPU 搜索 → 点击详情 → 新页面 ─────────────────────────────────────────
+
+def clear_search_tags(page):
+    """清除搜索框中已存在的 SPU tag"""
+    print("    清除已有搜索标签...")
+    try:
+        # 优先点击"清空全部"按钮
+        clear_all = page.locator('.jmtd-base-input-icon-clear button').first
+        if clear_all.count() > 0 and clear_all.is_visible():
+            clear_all.click()
+            page.wait_for_timeout(800)
+            print("      [OK] 已点击清空全部")
+            return
+    except Exception:
+        pass
+
+    # 兜底：逐个点击 tag 的关闭按钮
+    try:
+        tags = page.locator('.search-multiple-input .jmtd-tag .jmtd-tag-close-icon').all()
+        if tags:
+            for tag in tags:
+                tag.click()
+                page.wait_for_timeout(300)
+            print(f"      [OK] 已关闭 {len(tags)} 个标签")
+        else:
+            print("      [INFO] 没有已有标签")
+    except Exception as e:
+        print(f"      [WARN] 关闭标签失败: {e}")
+
+
+def search_spu_in_detail_page(page, spu):
+    """在商品明细页搜索框输入 SPU 并搜索"""
+    print(f"  搜索 SPU: {spu}")
+    try:
+        # 1. 先清除已有标签
+        clear_search_tags(page)
+
+        # 2. 定位输入框并聚焦
+        input_box = page.locator('.search-multiple-input .jmtd-multiple-input-input').first
+        input_box.wait_for(state="visible", timeout=5000)
+        input_box.click()
+        page.wait_for_timeout(300)
+
+        # 3. 逐字输入（某些组件对 fill 不敏感，type 更可靠）
+        input_box.type(spu, delay=50)
+        page.wait_for_timeout(500)
+
+        # 4. 按 Enter 把输入转成 tag（multi-input 通常需要 Enter 确认）
+        input_box.press("Enter")
+        page.wait_for_timeout(500)
+
+        # 5. 额外触发 input 事件确保组件收到值
+        page.evaluate(f"""() => {{
+            const input = document.querySelector('.search-multiple-input .jmtd-multiple-input-input');
+            if (input) {{
+                input.value = '{spu}';
+                input.dispatchEvent(new Event('input', {{ bubbles: true }}));
+                input.dispatchEvent(new Event('change', {{ bubbles: true }}));
+            }}
+        }}""")
+        page.wait_for_timeout(300)
+
+        # 6. 点击搜索按钮
+        search_btn = page.locator('.search-multiple-input .jmtd-input-search-button').first
+        search_btn.wait_for(state="visible", timeout=3000)
+        search_btn.click()
+        print("    [OK] 已点击搜索")
+    except Exception as e:
+        print(f"    [WARN] 精确搜索失败: {e}")
+        # 兜底：纯 JS 操作
+        page.evaluate(f"""() => {{
+            // 先清除已有 tag
+            document.querySelectorAll('.search-multiple-input .jmtd-tag .jmtd-tag-close-icon').forEach(el => el.click());
+            const input = document.querySelector('.search-multiple-input .jmtd-multiple-input-input');
+            const btn = document.querySelector('.jmtd-input-search-button.jmtd-button-small');
+            if (input) {{
+                input.focus();
+                input.value = '{spu}';
+                input.dispatchEvent(new Event('input', {{ bubbles: true }}));
+                input.dispatchEvent(new Event('change', {{ bubbles: true }}));
+                // 模拟回车生成 tag
+                input.dispatchEvent(new KeyboardEvent('keydown', {{ key: 'Enter', bubbles: true }}));
+            }}
+            if (btn) {{ btn.click(); return 'js_searched'; }}
+            return 'not_found';
+        }}""")
+
+    # 7. 等待搜索结果刷新
+    for _ in range(25):
+        page.wait_for_timeout(200)
+        loading = page.evaluate("""() => {
+            const el = document.querySelector('.jmtd-loading');
+            return el ? window.getComputedStyle(el).display !== 'none' : false;
+        }""")
+        if not loading:
+            break
+    page.wait_for_timeout(1000)
+
+
+def click_detail_and_wait_new_page(page, context, spu):
+    """点击搜索结果中的'详情'按钮，等待并返回新页面"""
+    print(f"  点击'详情'按钮...")
+    try:
+        # 等待表格刷新（搜索结果应该只有1条）
+        page.wait_for_selector('.jmtd-table-main tbody tr:not(.jmtd-table-plugin-total-row)', timeout=10000)
+
+        # 点击详情按钮（不是取消/关注按钮）
+        with context.expect_page(timeout=15000) as new_page_info:
+            detail_btn = page.locator(
+                '.jmtd-table-main tbody tr:not(.jmtd-table-plugin-total-row) td:last-child .detail-btn',
+                has_text="详情"
+            ).first
+            detail_btn.wait_for(state="visible", timeout=5000)
+            detail_btn.click()
+            print("    [OK] 已点击详情")
+
+        new_page = new_page_info.value
+        new_page.wait_for_load_state("networkidle", timeout=15000)
+        page.wait_for_timeout(3000)
+        print(f"    [OK] 新页面已打开: {new_page.url[:80]}...")
+        return new_page
+    except Exception as e:
+        print(f"    [FAIL] 点击详情或打开新页面失败: {e}")
+        return None
+
+
+# ── 新页面：核心数据 tab ──────────────────────────────────────────────────────
+
+def click_core_data_tab(page):
+    """在新页面点击'核心数据'tab"""
+    print("  切换到'核心数据'...")
+    try:
+        tab = page.locator('.single-header-tabs .jmtd-tabs-tab', has_text="核心数据").first
+        tab.wait_for(state="visible", timeout=5000)
+        is_active = tab.evaluate("el => el.classList.contains('jmtd-tabs-tab-active')")
+        if not is_active:
+            tab.click()
+            page.wait_for_timeout(3000)
+            print("    [OK] 已点击核心数据")
+        else:
+            print("    [OK] 核心数据已激活")
+    except Exception as e:
+        print(f"    [WARN] 点击核心数据失败: {e}")
+        page.evaluate("""() => {
+            for (const tab of document.querySelectorAll('.single-header-tabs .jmtd-tabs-tab')) {
+                if (tab.textContent.includes('核心数据') && !tab.classList.contains('jmtd-tabs-tab-active')) {
+                    tab.click(); return 'clicked';
+                }
+            }
+            return 'already_active_or_not_found';
+        }""")
+        page.wait_for_timeout(3000)
+
+
+# ── 新页面：流量来源 tab ───────────────────────────────────────────────────────
+
+def click_traffic_source_tab(page):
+    """在新页面点击'流量来源'tab"""
+    print("  切换到'流量来源'...")
+    try:
+        tab = page.locator('.single-header-tabs .jmtd-tabs-tab', has_text="流量来源").first
+        tab.wait_for(state="visible", timeout=5000)
+        is_active = tab.evaluate("el => el.classList.contains('jmtd-tabs-tab-active')")
+        if not is_active:
+            tab.click()
+            page.wait_for_timeout(3000)
+            print("    [OK] 已点击流量来源")
+        else:
+            print("    [OK] 流量来源已激活")
+    except Exception as e:
+        print(f"    [WARN] 点击流量来源失败: {e}")
+        page.evaluate("""() => {
+            for (const tab of document.querySelectorAll('.single-header-tabs .jmtd-tabs-tab')) {
+                if (tab.textContent.includes('流量来源') && !tab.classList.contains('jmtd-tabs-tab-active')) {
+                    tab.click(); return 'clicked';
+                }
+            }
+            return 'already_active_or_not_found';
+        }""")
+        page.wait_for_timeout(3000)
+
+
+def expand_all_traffic_rows(page):
+    """展开流量来源表格的所有行"""
+    print("  展开所有行...")
+    page.evaluate("""() => {
+        const expandIcons = document.querySelectorAll('.jmti-plus-square-f');
+        for (const icon of expandIcons) { icon.click(); }
+        return expandIcons.length;
+    }""")
+    page.wait_for_timeout(2000)
+    print("    [OK] 已展开所有行")
+
+
+def extract_traffic_data(page):
+    """提取流量来源表格数据（含数值和涨跌百分比）"""
+    print("  采集流量来源数据...")
+    data = extract(page, """() => {
+        // 定位流量来源卡片内的表格（避免和关键词表格混淆）
+        const cards = document.querySelectorAll('.jmtd-card');
+        let targetTable = null;
+        for (const card of cards) {
+            const header = card.querySelector('.jmtd-card-header');
+            if (header && header.textContent.includes('流量来源')) {
+                targetTable = card.querySelector('.jmtd-table-main');
+                break;
+            }
+        }
+        if (!targetTable) {
+            // 兜底：直接取页面第一个 jmtd-table-main
+            targetTable = document.querySelector('.jmtd-table-main');
+        }
+        if (!targetTable) return '[]';
+
+        const rows = targetTable.querySelectorAll('tbody tr');
+        const headers = Array.from(targetTable.querySelectorAll('thead th')).map(th => th.textContent.trim());
+        const result = [];
+
+        for (const row of rows) {
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 2) continue;
+            const obj = {};
+            cells.forEach((cell, i) => {
+                if (headers[i] === '操作') return;
+                if (i === 0) {
+                    // 流量来源名称
+                    const sourceName = cell.querySelector('.source-name span')?.textContent?.trim()
+                        || cell.textContent.trim();
+                    obj[headers[i]] = sourceName;
+                } else {
+                    // 数值 + 环比：提取所有 <p> 子元素
+                    const ps = cell.querySelectorAll(':scope > p');
+                    if (ps.length >= 2) {
+                        const value = ps[0].textContent.trim();
+                        const rateEl = ps[1];
+                        const rate = rateEl.textContent.trim();
+                        // 判断涨跌趋势
+                        let trend = '';
+                        const cls = rateEl.className || '';
+                        if (cls.includes('sz-frontend-ft')) trend = 'up';
+                        else if (cls.includes('sz-frontend-fu')) trend = 'down';
+                        else if (rate.startsWith('+')) trend = 'up';
+                        else if (rate.startsWith('-')) trend = 'down';
+                        obj[headers[i]] = { value, rate, trend };
+                    } else if (ps.length === 1) {
+                        // 只有数值没有环比
+                        obj[headers[i]] = { value: ps[0].textContent.trim(), rate: '', trend: '' };
+                    } else {
+                        // 兜底：纯文本，尝试拆分
+                        const txt = cell.textContent.trim();
+                        const numMatch = txt.match(/^([^+\x2d±]*[0-9]*)([±+\x2d]?[0-9,.%￥]+)(.*)$/);
+                        if (numMatch) {
+                            obj[headers[i]] = { value: numMatch[1].trim() || numMatch[2].trim(), rate: numMatch[3].trim() || '', trend: '' };
+                        } else {
+                            obj[headers[i]] = { value: txt, rate: '', trend: '' };
+                        }
+                    }
+                }
+            });
+            if (Object.keys(obj).length > 0) result.push(obj);
+        }
+        return JSON.stringify(result);
+    }""", "[]")
+    try:
+        result = json.loads(data) if isinstance(data, str) else []
+    except:
+        result = []
+    print(f"    [OK] 提取到 {len(result)} 条流量来源数据")
+    return result
+
+
+def extract_keyword_data(page):
+    """提取关键词来源成交表格数据（翻页收集最多20个）"""
+    print("  采集关键词来源成交数据...")
+    all_keywords = []
+    max_keywords = 20
+
+    for page_num in range(1, 5):  # 最多4页
+        print(f"    采集第 {page_num} 页...")
+        # 等待表格刷新
+        for _ in range(20):
+            page.wait_for_timeout(200)
+            loading = page.evaluate("""() => {
+                const el = document.querySelector('.jmtd-loading');
+                return el ? window.getComputedStyle(el).display !== 'none' : false;
+            }""")
+            if not loading:
+                break
+
+        data = extract(page, """() => {
+            const cards = document.querySelectorAll('.jmtd-card');
+            let targetTable = null;
+            for (const card of cards) {
+                const header = card.querySelector('.jmtd-card-header');
+                if (header && header.textContent.includes('关键词来源成交')) {
+                    targetTable = card.querySelector('.jmtd-table-main');
+                    break;
+                }
+            }
+            if (!targetTable) return '[]';
+            const rows = targetTable.querySelectorAll('tbody tr');
+            const headers = Array.from(targetTable.querySelectorAll('thead th')).map(th => th.textContent.trim());
+            const result = [];
+            for (const row of rows) {
+                const cells = row.querySelectorAll('td');
+                if (cells.length < 2) continue;
+                const obj = {};
+                cells.forEach((cell, i) => {
+                    if (headers[i] === '操作') return;
+                    obj[headers[i]] = cell.textContent.trim();
+                });
+                if (Object.keys(obj).length > 0) result.push(obj);
+            }
+            return JSON.stringify(result);
+        }""", "[]")
+
+        try:
+            rows = json.loads(data) if isinstance(data, str) else []
+        except:
+            rows = []
+        print(f"      提取到 {len(rows)} 条")
+        for r in rows:
+            all_keywords.append(r)
+
+        if len(all_keywords) >= max_keywords:
+            break
+
+        # 点击下一页（限定在关键词卡片内）
+        has_next = page.evaluate("""() => {
+            const cards = document.querySelectorAll('.jmtd-card');
+            let nextBtn = null;
+            for (const card of cards) {
+                const header = card.querySelector('.jmtd-card-header');
+                if (header && header.textContent.includes('关键词来源成交')) {
+                    nextBtn = card.querySelector('.jmtd-pagination-simple .jmtd-pagination-item-next');
+                    break;
+                }
+            }
+            if (!nextBtn) return false;
+            if (nextBtn.classList.contains('jmtd-pagination-item-disabled')) return false;
+            nextBtn.click();
+            return true;
+        }""")
+
+        if not has_next:
+            print("    [INFO] 无更多分页")
+            break
+        page.wait_for_timeout(2000)
+
+    print(f"    [OK] 共提取 {len(all_keywords)} 条关键词数据")
+    return all_keywords[:max_keywords]
+
+
+# ── 新页面：人群画像 tab ──────────────────────────────────────────────────────
+
+def click_crowd_portrait_tab(page):
+    """在新页面点击'人群画像'tab"""
+    print("  切换到'人群画像'...")
+    try:
+        tab = page.locator('.single-header-tabs .jmtd-tabs-tab', has_text="人群画像").first
+        tab.wait_for(state="visible", timeout=5000)
+        is_active = tab.evaluate("el => el.classList.contains('jmtd-tabs-tab-active')")
+        if not is_active:
+            tab.click()
+            page.wait_for_timeout(3000)
+            print("    [OK] 已点击人群画像")
+        else:
+            print("    [OK] 人群画像已激活")
+    except Exception as e:
+        print(f"    [WARN] 点击人群画像失败: {e}")
+        page.evaluate("""() => {
+            for (const tab of document.querySelectorAll('.single-header-tabs .jmtd-tabs-tab')) {
+                if (tab.textContent.includes('人群画像') && !tab.classList.contains('jmtd-tabs-tab-active')) {
+                    tab.click(); return 'clicked';
+                }
+            }
+            return 'already_active_or_not_found';
+        }""")
+        page.wait_for_timeout(3000)
+
+
+def extract_crowd_texts(page, section_title):
+    """从人群画像指定区域提取所有可见文本"""
+    data = page.evaluate(f"""() => {{
+        const result = {{title: '{section_title}', texts: []}};
+        // 提取所有 .jmtd-typography-text 的文本
+        document.querySelectorAll('.jmtd-typography-text').forEach(el => {{
+            const txt = el.textContent.trim();
+            if (txt && txt.length > 2 && txt.length < 200) result.texts.push(txt);
+        }});
+        // 去重
+        result.texts = [...new Set(result.texts)];
+        return JSON.stringify(result);
+    }}""")
+    try:
+        return json.loads(data) if isinstance(data, str) else {"texts": []}
+    except:
+        return {"texts": []}
+
+
+def extract_crowd_text_fallback(page, keywords):
+    """从页面文本中提取匹配关键词的信息"""
+    result = page.evaluate("""(keywords) => {
+        const texts = document.querySelectorAll('.jmtd-typography-text');
+        for (const el of texts) {
+            const txt = el.textContent.trim();
+            for (const kw of keywords) {
+                if (txt.includes(kw)) return txt;
+            }
+        }
+        return '';
+    }""", keywords)
+    return result if result else ''
+
+
+def extract_region_table_data(page):
+    """提取地区分布表格"""
+    data = page.evaluate("""() => {
+        const rows = document.querySelectorAll('.jmtd-table-main tbody tr');
+        const result = [];
+        for (const row of rows) {
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 3) continue;
+            const rank = cells[0].textContent.trim();
+            const region = cells[1].textContent.trim();
+            const ratio = cells[2].querySelector('.jmtd-table-formatter-ratio-bar-number')?.textContent?.trim() || cells[2].textContent.trim();
+            result.push({rank, region, ratio});
+        }
+        return JSON.stringify(result);
+    }""")
+    try:
+        return json.loads(data) if isinstance(data, str) else []
+    except:
+        return []
+
+
+def click_city_radio(page):
+    """点击城市radio按钮"""
+    clicked = page.evaluate("""() => {
+        const labels = document.querySelectorAll('.jmtd-form-control-button-label');
+        for (const label of labels) {
+            if (label.textContent.includes('城市')) {
+                label.click();
+                return true;
+            }
+        }
+        return false;
+    }""")
+    if clicked:
+        print("    [OK] 已点击城市radio")
+    else:
+        print("    [WARN] 未找到城市radio")
+    page.wait_for_timeout(2000)
+
+
+def extract_echarts_instance_data(page):
+    """尝试通过ECharts实例直接获取图表数据（最优先策略）"""
+    try:
+        data = page.evaluate("""() => {
+            const results = [];
+
+        function visible(el) {
+            if (!el) return false;
+            const rect = el.getBoundingClientRect();
+            const style = window.getComputedStyle(el);
+            return rect.width > 30 && rect.height > 30 && style.display !== 'none' && style.visibility !== 'hidden';
+        }
+
+        function closestTitle(dom) {
+            const card = dom.closest('.jmtd-card, .jmtd-col, .jmtd-row-item, [class*="card"], [class*="panel"]');
+            if (card) {
+                const titleEl = card.querySelector('.jmtd-card-header, .jmtd-typography-title, .jmtd-label, [class*="title"]');
+                const title = titleEl ? titleEl.textContent.trim() : '';
+                if (title) return title;
+            }
+            let p = dom.parentElement;
+            for (let i = 0; p && i < 5; i++, p = p.parentElement) {
+                const text = p.textContent.trim();
+                if (/性别|年龄|职业/.test(text)) {
+                    if (text.includes('性别')) return '性别';
+                    if (text.includes('年龄')) return '年龄';
+                    if (text.includes('职业')) return '职业';
+                }
+            }
+            return '';
+        }
+
+        function findInstance(dom) {
+            let instance = null;
+            // 方式1: 常见内部属性
+            const keys = ['_echartsInstance', '__echartsInstance', '_echarts', '__echarts__', '__EC_CHARTS__', '_echarts_instance_'];
+            for (const k of keys) {
+                if (dom[k] && typeof dom[k].getOption === 'function') {
+                    instance = dom[k];
+                    break;
+                }
+            }
+            // 方式2: ECharts常见实例ID属性
+            if (!instance && window.echarts && typeof window.echarts.getInstanceByDom === 'function') {
+                try { instance = window.echarts.getInstanceByDom(dom); } catch(e) {}
+            }
+            // 方式3: 遍历所有可枚举属性
+            if (!instance) {
+                for (const key of Object.keys(dom)) {
+                    try {
+                        const val = dom[key];
+                        if (val && typeof val.getOption === 'function') {
+                            instance = val;
+                            break;
+                        }
+                    } catch(e) {}
+                }
+            }
+            return instance;
+        }
+
+        function findChartDomCandidates() {
+            const set = new Set();
+            document.querySelectorAll('.jmtc-container, [_echarts_instance_]').forEach(el => {
+                if (visible(el)) set.add(el);
+            });
+            document.querySelectorAll('canvas[data-zr-dom-id], canvas').forEach(canvas => {
+                let p = canvas.parentElement;
+                for (let i = 0; p && i < 6; i++, p = p.parentElement) {
+                    if (visible(p)) set.add(p);
+                    if (findInstance(p)) break;
+                }
+            });
+            return [...set];
+        }
+
+        function toNumber(v) {
+            if (typeof v === 'number') return v;
+            if (Array.isArray(v)) return Number(v.find(x => !Number.isNaN(Number(x))) || 0);
+            if (v && typeof v === 'object') return Number(v.value || v[1] || 0);
+            return Number(String(v || '').replace(/,/g, '').replace('%', ''));
+        }
+
+        function formatPercent(num) {
+            if (!Number.isFinite(num)) return '';
+            const fixed = num.toFixed(2).replace(/\\.00$/, '').replace(/(\\.\\d)0$/, '$1');
+            return fixed + '%';
+        }
+
+        const charts = findChartDomCandidates();
+        const seenCharts = new Set();
+        for (const dom of charts) {
+            const instance = findInstance(dom);
+            const chartKey = instance && (instance.id || dom.getAttribute('_echarts_instance_'));
+            if (!instance || (chartKey && seenCharts.has(chartKey))) continue;
+            if (chartKey) seenCharts.add(chartKey);
+
+            try {
+                const opt = instance.getOption();
+                const seriesList = Array.isArray(opt.series) ? opt.series : [];
+                for (const series of seriesList) {
+                    const sourceData = Array.isArray(series.data) ? series.data : [];
+                    if (!sourceData.length) continue;
+
+                    const numericValues = sourceData.map(d => toNumber(d && d.value !== undefined ? d.value : d));
+                    const total = numericValues.reduce((sum, n) => sum + (Number.isFinite(n) ? n : 0), 0);
+                    const isPie = series.type === 'pie' || (Array.isArray(series.radius) && series.radius.length > 0);
+
+                    const chartData = sourceData.map((d, idx) => {
+                        const raw = d && d.value !== undefined ? d.value : d;
+                        let rawValue = raw;
+                        if (raw && typeof raw === 'object' && !Array.isArray(raw)) rawValue = raw.value;
+                        const num = numericValues[idx];
+                        let value = String(rawValue ?? '').trim();
+                        if (isPie && total > 0 && Number.isFinite(num)) {
+                            value = formatPercent(num / total * 100);
+                        }
+                        return {
+                            name: String((d && (d.name || d[0])) || '').trim(),
+                            value,
+                            raw_value: String(rawValue ?? '').trim()
+                        };
+                    }).filter(item => item.name && item.value);
+
+                    if (chartData.length > 0) {
+                        results.push({
+                            title: closestTitle(dom),
+                            seriesType: series.type || '',
+                            data: chartData
+                        });
+                    }
+                }
+            } catch(e) {}
+        }
+
+            return JSON.stringify(results);
+        }""")
+        return json.loads(data) if isinstance(data, str) else []
+    except Exception as e:
+        print(f"    [WARN] ECharts实例提取失败: {e}")
+        return []
+
+
+def extract_echarts_canvas_data(page):
+    """从canvas对应的ECharts/ZRender容器提取环形图数据，补充实例扫描失败的场景"""
+    try:
+        data = page.evaluate("""() => {
+            const results = [];
+            if (!window.echarts || typeof window.echarts.getInstanceByDom !== 'function') {
+                return JSON.stringify(results);
+            }
+
+        function visible(el) {
+            if (!el) return false;
+            const rect = el.getBoundingClientRect();
+            const style = window.getComputedStyle(el);
+            return rect.width > 30 && rect.height > 30 && style.display !== 'none' && style.visibility !== 'hidden';
+        }
+
+        function toNumber(v) {
+            if (typeof v === 'number') return v;
+            if (Array.isArray(v)) return Number(v.find(x => !Number.isNaN(Number(x))) || 0);
+            if (v && typeof v === 'object') return Number(v.value || v[1] || 0);
+            return Number(String(v || '').replace(/,/g, '').replace('%', ''));
+        }
+
+        function formatPercent(num) {
+            if (!Number.isFinite(num)) return '';
+            return num.toFixed(2).replace(/\\.00$/, '').replace(/(\\.\\d)0$/, '$1') + '%';
+        }
+
+        const candidates = new Set();
+        document.querySelectorAll('canvas[data-zr-dom-id]').forEach(canvas => {
+            let p = canvas.parentElement;
+            for (let i = 0; p && i < 8; i++, p = p.parentElement) {
+                if (visible(p)) candidates.add(p);
+                if (p.getAttribute && p.getAttribute('_echarts_instance_')) break;
+            }
+        });
+
+        const seen = new Set();
+        for (const dom of candidates) {
+            let instance = null;
+            try { instance = window.echarts.getInstanceByDom(dom); } catch(e) {}
+            if (!instance || seen.has(instance.id)) continue;
+            seen.add(instance.id);
+
+            try {
+                const opt = instance.getOption();
+                const seriesList = Array.isArray(opt.series) ? opt.series : [];
+                for (const series of seriesList) {
+                    const rows = Array.isArray(series.data) ? series.data : [];
+                    if (!rows.length) continue;
+                    const nums = rows.map(d => toNumber(d && d.value !== undefined ? d.value : d));
+                    const total = nums.reduce((sum, n) => sum + (Number.isFinite(n) ? n : 0), 0);
+                    const isPie = series.type === 'pie' || (Array.isArray(series.radius) && series.radius.length > 0);
+                    const items = rows.map((d, idx) => {
+                        const raw = d && d.value !== undefined ? d.value : d;
+                        const num = nums[idx];
+                        return {
+                            name: String((d && (d.name || d[0])) || '').trim(),
+                            value: isPie && total > 0 && Number.isFinite(num) ? formatPercent(num / total * 100) : String(raw ?? '').trim(),
+                            raw_value: String(raw ?? '').trim()
+                        };
+                    }).filter(item => item.name && item.value);
+                    if (items.length) results.push({seriesType: series.type || '', data: items});
+                }
+            } catch(e) {}
+        }
+
+            return JSON.stringify(results);
+        }""")
+        return json.loads(data) if isinstance(data, str) else []
+    except Exception as e:
+        print(f"    [WARN] Canvas环形图提取失败: {e}")
+        return []
+
+
+def extract_echarts_by_hover(page):
+    """通过模拟鼠标悬停环形图圆环，读取tooltip数据（备选策略）"""
+    charts = page.locator('canvas[data-zr-dom-id], .jmtc-container').all()[:6]
+    all_data = {"性别": [], "年龄": [], "职业": []}
+    for idx, chart in enumerate(charts):
+        try:
+            box = chart.bounding_box()
+            if not box or box['width'] < 50 or box['height'] < 50:
+                continue
+            seen = set()
+            chart_items = []
+            cx = box['x'] + box['width'] / 2
+            cy = box['y'] + box['height'] / 2
+            base_r = min(box['width'], box['height']) * 0.36
+            # 空心环不能悬停中心点，沿圆环两圈采样以触发 tooltip
+            hover_points = []
+            for radius_scale in [1.0, 0.78]:
+                radius = base_r * radius_scale
+                for angle in [0, 30, 60, 90, 120, 150, 180, 210, 240, 270, 300, 330]:
+                    rad = angle * 3.1415926 / 180
+                    hover_points.append((cx + radius * math.cos(rad), cy + radius * math.sin(rad)))
+
+            for x, y in hover_points:
+                page.mouse.move(x, y)
+                page.wait_for_timeout(180)
+                tooltip_result = page.evaluate("""() => {
+                    const items = [];
+                    function addPair(name, value) {
+                        name = String(name || '').replace(/[：:\\s]+$/g, '').trim();
+                        value = String(value || '').trim();
+                        if (!name || !value) return;
+                        if (!/%|\\d/.test(value)) return;
+                        items.push({name, value});
+                    }
+                    document.querySelectorAll('.sz-frontend-bc, .echarts-tooltip, [class*="tooltip"]').forEach(el => {
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        if (style.visibility === 'hidden' || style.display === 'none' || rect.width === 0 || rect.height === 0) return;
+                        const nameEl = el.querySelector('.name');
+                        const valueEl = el.querySelector('.value');
+                        if (nameEl && valueEl) {
+                            addPair(nameEl.textContent, valueEl.textContent);
+                            return;
+                        }
+                        const text = el.textContent.replace(/\\u00a0/g, ' ').trim();
+                        const lines = text.split(/\\n/).map(s => s.trim()).filter(Boolean);
+                        for (const line of lines) {
+                            let m = line.match(/^(.+?)[：:]\\s*(\\d+(?:\\.\\d+)?%?.*)$/);
+                            if (m) {
+                                addPair(m[1], m[2]);
+                                continue;
+                            }
+                            m = line.match(/^([^\\d%：:]{1,30})\\s+(\\d+(?:\\.\\d+)?%.*)$/);
+                            if (m) addPair(m[1], m[2]);
+                        }
+                    });
+                    return JSON.stringify(items);
+                }""")
+                items = json.loads(tooltip_result) if isinstance(tooltip_result, str) else []
+                for item in items:
+                    key = f"{item['name']}:{item['value']}"
+                    if key not in seen:
+                        seen.add(key)
+                        chart_items.append(item)
+            # 判断图表类型
+            if chart_items:
+                names_str = ' '.join([d['name'] for d in chart_items])
+                if any(n in names_str for n in ['男', '女']):
+                    all_data["性别"] = chart_items
+                elif '岁' in names_str:
+                    all_data["年龄"] = chart_items
+                else:
+                    all_data["职业"] = chart_items
+        except Exception as e:
+            print(f"    [WARN] 图表{idx}悬停提取失败: {e}")
+    # 鼠标移回安全区域
+    page.mouse.move(0, 0)
+    return all_data
+
+
+def extract_crowd_charts_by_card(page):
+    """按卡片标题定向读取人群画像图表，避免性别/年龄/职业串数据"""
+    try:
+        data = page.evaluate("""() => {
+            const result = { "性别": [], "年龄": [], "职业": [] };
+
+            function visible(el) {
+                if (!el) return false;
+                const rect = el.getBoundingClientRect();
+                const style = window.getComputedStyle(el);
+                return rect.width > 30 && rect.height > 30 && style.display !== 'none' && style.visibility !== 'hidden';
+            }
+
+            function toNumber(v) {
+                if (typeof v === 'number') return v;
+                if (Array.isArray(v)) return Number(v.find(x => !Number.isNaN(Number(x))) || 0);
+                if (v && typeof v === 'object') return Number(v.value || v[1] || 0);
+                return Number(String(v || '').replace(/,/g, '').replace('%', ''));
+            }
+
+            function formatPercent(num) {
+                if (!Number.isFinite(num)) return '';
+                return num.toFixed(2).replace(/\\.00$/, '').replace(/(\\.\\d)0$/, '$1') + '%';
+            }
+
+            function findInstance(dom) {
+                if (!dom) return null;
+                if (window.echarts && typeof window.echarts.getInstanceByDom === 'function') {
+                    try {
+                        const instance = window.echarts.getInstanceByDom(dom);
+                        if (instance) return instance;
+                    } catch(e) {}
+                }
+                for (const key of Object.keys(dom)) {
+                    try {
+                        const val = dom[key];
+                        if (val && typeof val.getOption === 'function') return val;
+                    } catch(e) {}
+                }
+                return null;
+            }
+
+            function findChartBlock(title) {
+                const aliases = title === '职业' ? ['职业分布', '职业'] : [title];
+                const titleEls = Array.from(document.querySelectorAll('.jmtd-typography-text, .jmtd-typography-title, span, h4'))
+                    .filter(el => aliases.some(alias => el.textContent.trim() === alias));
+                for (const titleEl of titleEls) {
+                    const block = titleEl.closest('.jmtd-grid-col-12, .jmtd-grid-col, .jmtd-card-section, .jmtd-card');
+                    if (block && block.querySelector('canvas, .jmtc-container, [_echarts_instance_]')) return block;
+                }
+                return null;
+            }
+
+            function chartDomsInCard(card) {
+                const set = new Set();
+                if (!card) return [];
+                card.querySelectorAll('.jmtc-container, [_echarts_instance_]').forEach(el => {
+                    if (visible(el)) set.add(el);
+                });
+                card.querySelectorAll('canvas[data-zr-dom-id], canvas').forEach(canvas => {
+                    let p = canvas.parentElement;
+                    for (let i = 0; p && i < 6 && card.contains(p); i++, p = p.parentElement) {
+                        if (visible(p)) set.add(p);
+                        if (findInstance(p)) break;
+                    }
+                });
+                return [...set];
+            }
+
+            function normalizeSeries(opt, title) {
+                const seriesList = Array.isArray(opt.series) ? opt.series : [];
+                if (!seriesList.length) return [];
+                const wantPie = title === '性别' || title === '职业';
+                const wantBar = title === '年龄';
+                let series = seriesList.find(s => wantPie && (s.type === 'pie' || Array.isArray(s.radius)));
+                if (!series) series = seriesList.find(s => wantBar && s.type === 'bar');
+                if (!series) series = seriesList.find(s => Array.isArray(s.data) && s.data.length);
+                if (!series || !Array.isArray(series.data)) return [];
+
+                const xData = Array.isArray(opt.xAxis) && opt.xAxis[0] && Array.isArray(opt.xAxis[0].data) ? opt.xAxis[0].data : [];
+                const nums = series.data.map(d => toNumber(d && d.value !== undefined ? d.value : d));
+                const total = nums.reduce((sum, n) => sum + (Number.isFinite(n) ? n : 0), 0);
+                const isPie = series.type === 'pie' || Array.isArray(series.radius);
+                const smallRatio = !isPie && nums.length && nums.every(n => Number.isFinite(n) && n >= 0 && n <= 1);
+
+                return series.data.map((d, idx) => {
+                    const raw = d && d.value !== undefined ? d.value : d;
+                    const name = String((d && (d.name || d[0])) || xData[idx] || '').trim();
+                    const num = nums[idx];
+                    let value = String(raw ?? '').trim();
+                    if (isPie && total > 0 && Number.isFinite(num)) {
+                        value = formatPercent(num / total * 100);
+                    } else if (smallRatio && Number.isFinite(num)) {
+                        value = formatPercent(num * 100);
+                    }
+                    return { name, value, raw_value: String(raw ?? '').trim() };
+                }).filter(item => item.name && item.value);
+            }
+
+            for (const title of ['性别', '年龄', '职业']) {
+                const card = findChartBlock(title);
+                if (!card) continue;
+                const seen = new Set();
+                for (const dom of chartDomsInCard(card)) {
+                    const instance = findInstance(dom);
+                    if (!instance) continue;
+                    const key = instance.id || dom.getAttribute('_echarts_instance_') || '';
+                    if (key && seen.has(key)) continue;
+                    if (key) seen.add(key);
+                    try {
+                        const items = normalizeSeries(instance.getOption(), title);
+                        if (items.length) {
+                            result[title] = items;
+                            break;
+                        }
+                    } catch(e) {}
+                }
+            }
+
+            return JSON.stringify(result);
+        }""")
+        return json.loads(data) if isinstance(data, str) else {"性别": [], "年龄": [], "职业": []}
+    except Exception as e:
+        print(f"    [WARN] 按卡片读取图表失败: {e}")
+        return {"性别": [], "年龄": [], "职业": []}
+
+
+def extract_crowd_chart_by_hover(page, title, chart_type="ring"):
+    """只在指定标题卡片内悬停采集，避免不同图表数据混淆"""
+    try:
+        box = page.evaluate("""(title) => {
+            function visible(el) {
+                if (!el) return false;
+                const rect = el.getBoundingClientRect();
+                const style = window.getComputedStyle(el);
+                return rect.width > 30 && rect.height > 30 && style.display !== 'none' && style.visibility !== 'hidden';
+            }
+            const aliases = title === '职业' ? ['职业分布', '职业'] : [title];
+            const titleEl = Array.from(document.querySelectorAll('.jmtd-typography-text, .jmtd-typography-title, span, h4'))
+                .find(el => aliases.some(alias => el.textContent.trim() === alias));
+            if (!titleEl) return null;
+            const block = titleEl.closest('.jmtd-grid-col-12, .jmtd-grid-col, .jmtd-card-section, .jmtd-card');
+            if (!block) return null;
+            const chart = Array.from(block.querySelectorAll('canvas[data-zr-dom-id], canvas, .jmtc-container')).find(visible);
+            if (!chart) return null;
+            const rect = chart.getBoundingClientRect();
+            return {x: rect.x, y: rect.y, width: rect.width, height: rect.height};
+        }""", title)
+        if not box or box['width'] < 50 or box['height'] < 50:
+            return []
+
+        points = []
+        if chart_type == "ring":
+            cx = box['x'] + box['width'] / 2
+            cy = box['y'] + box['height'] / 2
+            base_r = min(box['width'], box['height']) * 0.36
+            for radius_scale in [1.0, 0.78]:
+                radius = base_r * radius_scale
+                for angle in [0, 30, 60, 90, 120, 150, 180, 210, 240, 270, 300, 330]:
+                    rad = angle * math.pi / 180
+                    points.append((cx + radius * math.cos(rad), cy + radius * math.sin(rad)))
+        else:
+            for y_scale in [0.2, 0.35, 0.5, 0.65, 0.8]:
+                y = box['y'] + box['height'] * y_scale
+                for i in range(10):
+                    points.append((box['x'] + box['width'] * (i + 0.5) / 10, y))
+
+        seen = set()
+        items = []
+        def accept_tooltip_item(item):
+            name = (item.get('name') or '').strip()
+            if title == "性别":
+                return name in ("男", "女", "男性", "女性")
+            if title == "年龄":
+                return "岁" in name
+            if title == "职业":
+                return name not in ("男", "女", "男性", "女性") and "岁" not in name
+            return True
+
+        for x, y in points:
+            page.mouse.move(x, y)
+            page.wait_for_timeout(160)
+            tooltip_result = page.evaluate("""() => {
+                const items = [];
+                function addPair(name, value) {
+                    name = String(name || '').replace(/[：:\\s]+$/g, '').trim();
+                    value = String(value || '').trim();
+                    if (!name || !value || !/%|\\d/.test(value)) return;
+                    items.push({name, value});
+                }
+                document.querySelectorAll('.sz-frontend-bc, .echarts-tooltip, [class*="tooltip"]').forEach(el => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    if (style.visibility === 'hidden' || style.display === 'none' || rect.width === 0 || rect.height === 0) return;
+                    const text = el.textContent.replace(/\\u00a0/g, ' ').trim();
+                    const lines = text.split(/\\n/).map(s => s.trim()).filter(Boolean);
+                    for (const line of lines) {
+                        let m = line.match(/^(.+?)[：:]\\s*(\\d+(?:\\.\\d+)?%?.*)$/);
+                        if (!m) m = line.match(/^([^\\d%：:]{1,30})\\s+(\\d+(?:\\.\\d+)?%.*)$/);
+                        if (m) addPair(m[1], m[2]);
+                    }
+                });
+                return JSON.stringify(items);
+            }""")
+            tooltip_items = json.loads(tooltip_result) if isinstance(tooltip_result, str) else []
+            for item in tooltip_items:
+                if not accept_tooltip_item(item):
+                    continue
+                key = f"{item.get('name', '')}:{item.get('value', '')}"
+                if key not in seen:
+                    seen.add(key)
+                    items.append(item)
+        page.mouse.move(0, 0)
+        return items
+    except Exception as e:
+        print(f"    [WARN] {title}图表悬停提取失败: {e}")
+        return []
+
+
+def extract_crowd_tooltip_text_by_block(page, title):
+    """读取指定画像块内已渲染的tooltip文本（即使tooltip处于隐藏状态）"""
+    try:
+        data = page.evaluate("""(title) => {
+            const aliases = title === '职业' ? ['职业分布', '职业'] : [title];
+            const titleEl = Array.from(document.querySelectorAll('.jmtd-typography-text, .jmtd-typography-title, span, h4'))
+                .find(el => aliases.some(alias => el.textContent.trim() === alias));
+            if (!titleEl) return JSON.stringify([]);
+            const block = titleEl.closest('.jmtd-grid-col-12, .jmtd-grid-col, .jmtd-card-section, .jmtd-card');
+            if (!block) return JSON.stringify([]);
+
+            const result = [];
+            function cleanName(text) {
+                return String(text || '').replace(/\\u00a0/g, ' ').replace(/[：:\\s]+$/g, '').trim();
+            }
+            function cleanValue(text) {
+                return String(text || '').replace(/\\u00a0/g, ' ').trim();
+            }
+            block.querySelectorAll('.sz-frontend-bc').forEach(el => {
+                const name = cleanName(el.querySelector('.name')?.textContent || '');
+                const value = cleanValue(el.querySelector('.value')?.textContent || '');
+                if (name && value && /%|\\d/.test(value)) {
+                    result.push({name, value});
+                }
+            });
+
+            return JSON.stringify([...new Map(result.map(r => [r.name + ':' + r.value, r])).values()]);
+        }""", title)
+        return json.loads(data) if isinstance(data, str) else []
+    except Exception as e:
+        print(f"    [WARN] {title}隐藏tooltip提取失败: {e}")
+        return []
+
+
+def extract_age_distribution_data(page):
+    """单独提取年龄柱状图：优先按x轴/数据名包含'岁'定位，保持原始年龄段和值"""
+    try:
+        data = page.evaluate("""() => {
+            const results = [];
+
+            function visible(el) {
+                if (!el) return false;
+                const rect = el.getBoundingClientRect();
+                const style = window.getComputedStyle(el);
+                return rect.width > 30 && rect.height > 30 && style.display !== 'none' && style.visibility !== 'hidden';
+            }
+
+            function findInstance(dom) {
+                if (!dom) return null;
+                if (window.echarts && typeof window.echarts.getInstanceByDom === 'function') {
+                    try {
+                        const instance = window.echarts.getInstanceByDom(dom);
+                        if (instance) return instance;
+                    } catch(e) {}
+                }
+                for (const key of Object.keys(dom)) {
+                    try {
+                        const val = dom[key];
+                        if (val && typeof val.getOption === 'function') return val;
+                    } catch(e) {}
+                }
+                return null;
+            }
+
+            function xAxisData(opt) {
+                const xAxis = Array.isArray(opt.xAxis) ? opt.xAxis[0] : opt.xAxis;
+                if (xAxis && Array.isArray(xAxis.data)) return xAxis.data;
+                return [];
+            }
+
+            function rawValue(d) {
+                if (d && typeof d === 'object' && !Array.isArray(d) && d.value !== undefined) return d.value;
+                if (Array.isArray(d)) return d.find(x => !Number.isNaN(Number(x))) ?? d[1] ?? d[0];
+                return d;
+            }
+
+            function formatValue(v) {
+                const s = String(v ?? '').trim();
+                if (!s) return '';
+                if (s.includes('%')) return s;
+                const n = Number(s.replace(/,/g, ''));
+                if (Number.isFinite(n) && n >= 0 && n <= 1) {
+                    return (n * 100).toFixed(2).replace(/\\.00$/, '').replace(/(\\.\\d)0$/, '$1') + '%';
+                }
+                if (Number.isFinite(n)) {
+                    return n.toFixed(2).replace(/\\.00$/, '').replace(/(\\.\\d)0$/, '$1') + '%';
+                }
+                return s;
+            }
+
+            const candidates = new Set();
+            const titleEl = Array.from(document.querySelectorAll('.jmtd-typography-text, .jmtd-typography-title, span, h4'))
+                .find(el => el.textContent.trim() === '年龄');
+            const ageBlock = titleEl ? titleEl.closest('.jmtd-grid-col-12, .jmtd-grid-col, .jmtd-card-section, .jmtd-card') : null;
+            const root = ageBlock || document;
+            root.querySelectorAll('.jmtc-container, [_echarts_instance_], canvas[data-zr-dom-id], canvas').forEach(el => {
+                let p = el;
+                for (let i = 0; p && i < 6 && root.contains(p); i++, p = p.parentElement) {
+                    if (visible(p)) candidates.add(p);
+                    if (findInstance(p)) break;
+                }
+            });
+
+            if (ageBlock) {
+                ageBlock.querySelectorAll('.sz-frontend-bc').forEach(el => {
+                    const name = el.querySelector('.name')?.textContent?.replace(/[：:\\s]+/g, '').trim() || '';
+                    const value = el.querySelector('.value')?.textContent?.trim() || '';
+                    if (name.includes('岁') && value) {
+                        results.push([{name, value: formatValue(value), raw_value: value}]);
+                    }
+                });
+            }
+
+            const seen = new Set();
+            for (const dom of candidates) {
+                const instance = findInstance(dom);
+                const key = instance && (instance.id || dom.getAttribute('_echarts_instance_'));
+                if (!instance || (key && seen.has(key))) continue;
+                if (key) seen.add(key);
+
+                try {
+                    const opt = instance.getOption();
+                    const xData = xAxisData(opt).map(x => String(x ?? '').trim());
+                    const seriesList = Array.isArray(opt.series) ? opt.series : [];
+                    for (const series of seriesList) {
+                        if (!Array.isArray(series.data) || !series.data.length) continue;
+                        const labels = series.data.map((d, idx) => String((d && (d.name || d[0])) || xData[idx] || '').trim());
+                        const labelText = labels.join(' ');
+                        if (!labelText.includes('岁')) continue;
+                        const items = series.data.map((d, idx) => {
+                            const name = labels[idx];
+                            const rv = rawValue(d);
+                            return {name, value: formatValue(rv), raw_value: String(rv ?? '').trim()};
+                        }).filter(item => item.name && item.value);
+                        if (items.length) results.push(items);
+                    }
+                } catch(e) {}
+            }
+
+            results.sort((a, b) => b.length - a.length);
+            return JSON.stringify(results[0] || []);
+        }""")
+        return json.loads(data) if isinstance(data, str) else []
+    except Exception as e:
+        print(f"    [WARN] 年龄分布柱状图提取失败: {e}")
+        return []
+
+
+def extract_career_text_data(page):
+    """从职业分布区域直接提取文本数据（职业卡片内的文本列表）"""
+    data = page.evaluate("""() => {
+        const result = [];
+        // 1. 先找到包含"职业"的卡片
+        let careerCard = null;
+        const cards = document.querySelectorAll('.jmtd-card');
+        for (const card of cards) {
+            const header = card.querySelector('.jmtd-card-header');
+            if (header && header.textContent.includes('职业')) {
+                careerCard = card;
+                break;
+            }
+        }
+        if (!careerCard) {
+            // 兜底：找包含"职业"标题的最近父容器
+            const titles = document.querySelectorAll('.jmtd-typography-title, .jmtd-label, .jmtp-basic-summary-label');
+            for (const t of titles) {
+                if (t.textContent.includes('职业')) {
+                    careerCard = t.closest('.jmtd-card, .jmtd-col, .jmtd-row-item') || t.parentElement;
+                    break;
+                }
+            }
+        }
+        if (careerCard) {
+            // 在职业卡片内搜索 "名称 数值%" 或 "名称：数值%" 格式的文本
+            careerCard.querySelectorAll('.jmtd-typography-text, span, .jmtd-label, .jmtp-basic-summary-label, p').forEach(el => {
+                const text = el.textContent.trim();
+                // 匹配格式：职业名 + 空格/冒号 + 百分比数字
+                const m = text.match(/^([^\\d%:：\\s]{2,20})[:：\\s]+(\\d+(?:\\.\\d+)?%)$/);
+                if (m) {
+                    result.push({name: m[1].trim(), value: m[2].trim()});
+                }
+            });
+        }
+        // 2. 如果卡片内没找到，全局搜索职业相关文本（兜底）
+        if (result.length === 0) {
+            document.querySelectorAll('.jmtd-typography-text, .jmtd-label').forEach(el => {
+                const text = el.textContent.trim();
+                // 更宽泛的匹配：允许前面有一些前缀词
+                const m = text.match(/(?:职业|行业|身份)[:：\\s]*([^\\d%:：\\s]{2,20})[:：\\s]*(\\d+(?:\\.\\d+)?%)/);
+                if (m) {
+                    result.push({name: m[1].trim(), value: m[2].trim()});
+                }
+            });
+        }
+        // 去重
+        return JSON.stringify([...new Map(result.map(r => [r.name + ':' + r.value, r])).values()]);
+    }""")
+    try:
+        return json.loads(data) if isinstance(data, str) else []
+    except:
+        return []
+
+
+def extract_crowd_portrait_data(page):
+    """采集人群画像数据（性别/年龄/职业/地区）"""
+    print("  采集人群画像数据...")
+    crowd_data = {
+        "性别": [],
+        "年龄": [],
+        "职业": [],
+        "地区_省份": [],
+        "地区_城市": [],
+    }
+    page.wait_for_timeout(3000)
+
+    def filter_crowd_items(kind, items):
+        filtered = []
+        for item in items or []:
+            name = str(item.get("name", "")).strip()
+            value = str(item.get("value", "")).strip()
+            if not name or not value:
+                continue
+            if kind == "性别" and name not in ("男", "女", "男性", "女性"):
+                continue
+            if kind == "年龄" and "岁" not in name:
+                continue
+            if kind == "职业" and (name in ("男", "女", "男性", "女性") or "岁" in name):
+                continue
+            filtered.append(item)
+        return filtered
+
+    # ── 策略1: 按卡片标题精确读取图表，避免性别/年龄/职业数据混在一起 ──
+    card_chart_data = extract_crowd_charts_by_card(page)
+    crowd_data["性别"] = filter_crowd_items("性别", card_chart_data.get("性别", []))
+    crowd_data["职业"] = filter_crowd_items("职业", card_chart_data.get("职业", []))
+
+    # 年龄柱状图单独按“岁”标签提取，恢复原先稳定的年龄段逻辑
+    crowd_data["年龄"] = filter_crowd_items("年龄", extract_age_distribution_data(page))
+    if not crowd_data["年龄"]:
+        crowd_data["年龄"] = filter_crowd_items("年龄", card_chart_data.get("年龄", []))
+
+    print(
+        f"    [INFO] 卡片图表读取: 性别{len(crowd_data['性别'])}条, "
+        f"年龄{len(crowd_data['年龄'])}条, 职业{len(crowd_data['职业'])}条"
+    )
+
+    # ── 策略2: 指定卡片内悬停补采，性别/职业是环形图，年龄是柱状图 ──
+    if not crowd_data["性别"]:
+        print("    [INFO] 性别分布环形图悬停补采...")
+        crowd_data["性别"] = filter_crowd_items("性别", extract_crowd_chart_by_hover(page, "性别", "ring"))
+    if not crowd_data["年龄"]:
+        print("    [INFO] 年龄分布柱状图悬停补采...")
+        crowd_data["年龄"] = filter_crowd_items("年龄", extract_crowd_chart_by_hover(page, "年龄", "bar"))
+    if not crowd_data["职业"]:
+        print("    [INFO] 职业分布隐藏tooltip补采...")
+        crowd_data["职业"] = filter_crowd_items("职业", extract_crowd_tooltip_text_by_block(page, "职业"))
+    if len(crowd_data["职业"]) < 2:
+        print("    [INFO] 职业分布环形图悬停补采...")
+        hover_career = filter_crowd_items("职业", extract_crowd_chart_by_hover(page, "职业", "ring"))
+        merged = {}
+        for item in crowd_data["职业"] + hover_career:
+            merged[f"{item.get('name', '')}:{item.get('value', '')}"] = item
+        crowd_data["职业"] = list(merged.values())
+
+    # ── 策略3: 职业分布文本直接提取（最后兜底）──
+    if not crowd_data["职业"]:
+        career_text_data = extract_career_text_data(page)
+        if career_text_data:
+            print(f"    [INFO] 职业文本直接提取到 {len(career_text_data)} 条")
+            crowd_data["职业"] = filter_crowd_items("职业", career_text_data)
+
+    # ── 策略4: 文本结论兜底 ──
+    if not crowd_data["性别"]:
+        gender_text = extract_crowd_text_fallback(page, ["男性客户偏多", "女性客户偏多", "男客户偏多", "女客户偏多"])
+        if gender_text:
+            crowd_data["性别"] = [{"name": "结论", "value": gender_text}]
+    if not crowd_data["年龄"]:
+        age_text = extract_crowd_text_fallback(page, ["岁", "年龄"])
+        if age_text and "岁" in age_text and len(age_text) < 80:
+            crowd_data["年龄"] = [{"name": "结论", "value": age_text}]
+    if not crowd_data["职业"]:
+        occ_text = extract_crowd_text_fallback(page, ["主力军是", "购买的主力军", "职业"])
+        if occ_text and len(occ_text) < 80:
+            crowd_data["职业"] = [{"name": "结论", "value": occ_text}]
+
+    # ── 地区-省份 ──
+    crowd_data["地区_省份"] = extract_region_table_data(page)
+
+    # ── 地区-城市 ──
+    click_city_radio(page)
+    crowd_data["地区_城市"] = extract_region_table_data(page)
+
+    print(f"    [OK] 人群画像采集完成: 性别{len(crowd_data['性别'])}条, 年龄{len(crowd_data['年龄'])}条, 职业{len(crowd_data['职业'])}条, 省份{len(crowd_data['地区_省份'])}条, 城市{len(crowd_data['地区_城市'])}条")
+    return crowd_data
+
+
+# ── 新页面：选择上周日期 ───────────────────────────────────────────────────────
+
+def select_last_week_on_detail_page(page):
+    """在单品详情页面选择昨天（天模式）。函数名保留，避免改动调用链。"""
+    print("  选择昨天时间范围...")
+    select_yesterday_in_combo_datepicker(page, "product_detail")
+    page.wait_for_timeout(3000)
+
+
+# ── 新页面：采集核心数据 ───────────────────────────────────────────────────────
+
+def extract_core_data(page):
+    """在核心数据 tab 采集指标（含比较数据）"""
+    print("  采集核心数据...")
+    page.wait_for_timeout(3000)
+
+    data = extract(page, """() => {
+        const result = {};
+        const groups = document.querySelectorAll('.jmtp-basic-summary-group');
+
+        for (const group of groups) {
+            // 分组名称
+            const labelEl = group.querySelector('.jmtp-basic-summary-label');
+            const groupName = labelEl ? labelEl.textContent.trim() : '其他';
+            const groupData = {};
+
+            // 提取该分组下的所有卡片
+            const cards = group.querySelectorAll('.jmtp-basic-summary-card');
+            for (const card of cards) {
+                const titleEl = card.querySelector('.jmtp-summary-card-title .jmtd-label');
+                const title = titleEl ? titleEl.textContent.trim() : '';
+                if (!title) continue;
+
+                const valueEl = card.querySelector('.summary-index');
+                const value = valueEl ? valueEl.textContent.trim() : '';
+
+                const descLeft = card.querySelector('.jmtp-summary-card-description-left');
+                const compareLabel = descLeft ? descLeft.textContent.trim() : '';
+
+                const descRight = card.querySelector('.jmtp-summary-card-description-right');
+                const compareValue = descRight ? descRight.textContent.trim() : '';
+
+                // 判断涨跌趋势
+                let trend = '';
+                const arrowGreen = card.querySelector('.jmtp-summary-card-arrow-green');
+                const arrowRed = card.querySelector('.jmtp-summary-card-arrow-red');
+                if (arrowGreen) trend = 'up';
+                else if (arrowRed) trend = 'down';
+
+                groupData[title] = {
+                    value: value,
+                    compare_label: compareLabel,
+                    compare_value: compareValue,
+                    trend: trend,
+                };
+            }
+
+            if (Object.keys(groupData).length > 0) {
+                result[groupName] = groupData;
+            }
+        }
+
+        return JSON.stringify(result);
+    }""", "{}")
+
+    try:
+        result = json.loads(data) if isinstance(data, str) else {}
+    except:
+        result = {}
+
+    # 打印关键指标
+    total_cards = sum(len(v) for v in result.values() if isinstance(v, dict))
+    print(f"    共提取 {len(result)} 个分组, {total_cards} 个指标")
+    for group_name, items in result.items():
+        print(f"    [{group_name}]")
+        for title, v in items.items():
+            print(f"      {title}: {v.get('value','')} ({v.get('compare_label','')} {v.get('compare_value','')})")
+    return result
+
+
+# ── 商品评价差评采集 ───────────────────────────────────────────────────────────
+
+def navigate_to_goods_appraise(page):
+    """从商家后台首页导航到商品评价页面"""
+    print("[Step] 导航到商品评价页面...")
+    if 'shop.jd.com' not in page.url:
+        page.goto("https://shop.jd.com/jdm/home", wait_until="networkidle")
+        page.wait_for_timeout(5000)
+    try:
+        goods_menu = page.locator('.shop-menu-sidebar__menu-first-item:has-text("商品")').first
+        goods_menu.wait_for(state="visible", timeout=10000)
+        goods_menu.hover()
+        print("  [OK] 已悬停'商品'菜单")
+    except Exception as e:
+        print(f"  [WARN] 悬停商品菜单失败: {e}")
+        page.goto("https://wares-jdm.jd.com/goodsAppraise", wait_until="networkidle")
+        page.wait_for_timeout(5000)
+        return page
+    page.wait_for_timeout(2000)
+    try:
+        appraise_link = page.locator('.shop-menu-sidebar__hover-item-menu:has-text("商品评价")').first
+        appraise_link.wait_for(state="visible", timeout=5000)
+        appraise_link.click()
+        print("  [OK] 已点击'商品评价'")
+    except Exception as e:
+        print(f"  [WARN] 点击商品评价失败: {e}")
+        page.goto("https://wares-jdm.jd.com/goodsAppraise", wait_until="networkidle")
+        page.wait_for_timeout(5000)
+        return page
+    page.wait_for_timeout(5000)
+    pages = page.context.pages
+    for p in pages:
+        if 'goodsAppraise' in p.url or 'wares-jdm' in p.url:
+            print(f"  [OK] 已到达商品评价页面")
+            return p
+    print(f"  [OK] 当前页面: {page.url}")
+    return page
+
+
+def click_negative_reviews_filter(page):
+    """点击差评筛选"""
+    print("  选择差评筛选...")
+    try:
+        negative_tag = page.locator('.filter-button-group__options .tag:has-text("差评")').first
+        negative_tag.wait_for(state="visible", timeout=5000)
+        negative_tag.click()
+        print("  [OK] 已点击差评")
+    except Exception as e:
+        print(f"  [WARN] 点击差评失败: {e}")
+        page.evaluate("""() => {
+            const tags = document.querySelectorAll('.filter-button-group__options .tag');
+            for (const tag of tags) {
+                if (tag.textContent.includes('差评')) { tag.click(); return 'clicked'; }
+            }
+            return 'not_found';
+        }""")
+    page.wait_for_timeout(1500)
+
+
+def select_last_week_date_range_appraise(page):
+    """在商品评价页面选择昨天日期范围（直接填写输入框，避免面板月份错位）"""
+    print("  设置评价时间为昨天...")
+    last_monday = get_yesterday_date()
+    last_sunday = last_monday
+    expected_start = last_monday.strftime('%Y-%m-%d')
+    expected_end = last_sunday.strftime('%Y-%m-%d')
+    input_selector = '.rcd-date-editor--datetimerange .rcd-range-input, .rcd-date-editor--daterange .rcd-range-input'
+    editor_selector = '.rcd-date-editor--datetimerange, .rcd-date-editor--daterange'
+    expected_start_value = f'{expected_start} 00:00:00'
+    expected_end_value = f'{expected_end} 23:59:59'
+
+    try:
+        direct_set = page.evaluate(
+            r"""({start, end, inputSelector}) => {
+                const inputs = Array.from(document.querySelectorAll(inputSelector));
+                if (inputs.length < 2) return { ok: false, reason: 'inputs_not_found' };
+                const setValue = (input, value) => {
+                    input.focus();
+                    const setter = Object.getOwnPropertyDescriptor(HTMLInputElement.prototype, 'value')?.set;
+                    if (setter) setter.call(input, value);
+                    else input.value = value;
+                    input.dispatchEvent(new Event('input', { bubbles: true }));
+                    input.dispatchEvent(new Event('change', { bubbles: true }));
+                    input.dispatchEvent(new KeyboardEvent('keydown', { key: 'Enter', bubbles: true, cancelable: true }));
+                    input.dispatchEvent(new KeyboardEvent('keyup', { key: 'Enter', bubbles: true, cancelable: true }));
+                    input.blur();
+                };
+                setValue(inputs[0], start);
+                setValue(inputs[1], end);
+                return { ok: inputs[0].value === start && inputs[1].value === end, start: inputs[0].value, end: inputs[1].value };
+            }""",
+            {"start": expected_start_value, "end": expected_end_value, "inputSelector": input_selector},
+        )
+        page.keyboard.press("Enter")
+        page.wait_for_timeout(1200)
+        inputs = page.locator(input_selector).all()
+        if len(inputs) >= 2:
+            actual_start = inputs[0].get_attribute('value') or ''
+            actual_end = inputs[1].get_attribute('value') or ''
+            if actual_start.startswith(expected_start) and actual_end.startswith(expected_end):
+                print(f"    [OK] appraise date set by native events: {actual_start} ~ {actual_end}")
+                return
+        print(f"    [WARN] native event date set did not stick: {direct_set}")
+    except Exception as e:
+        print(f"    [WARN] native event date set failed: {e}")
+
+    # 检查当前是否已是目标日期
+    try:
+        inputs = page.locator(input_selector).all()
+        if len(inputs) >= 2:
+            current_start = inputs[0].get_attribute('value') or ''
+            current_end = inputs[1].get_attribute('value') or ''
+            if current_start.startswith(expected_start) and current_end.startswith(expected_end):
+                print(f"    [OK] 日期已是昨天: {current_start} ~ {current_end}")
+                return
+    except Exception:
+        pass
+
+    # 策略1: 直接填写输入框（最可靠）
+    try:
+        inputs = page.locator(input_selector).all()
+        if len(inputs) >= 2:
+            # 先点击一下输入框确保焦点
+            inputs[0].click()
+            page.wait_for_timeout(300)
+            # 清空并填写开始日期
+            inputs[0].fill("")
+            page.wait_for_timeout(200)
+            inputs[0].fill(expected_start_value)
+            page.wait_for_timeout(300)
+            # 填写结束日期
+            inputs[1].fill("")
+            page.wait_for_timeout(200)
+            inputs[1].fill(expected_end_value)
+            page.wait_for_timeout(300)
+            # 按 Enter 确认
+            page.keyboard.press("Enter")
+            page.wait_for_timeout(1000)
+
+            # 验证
+            inputs = page.locator(input_selector).all()
+            if len(inputs) >= 2:
+                actual_start = inputs[0].get_attribute('value') or ''
+                actual_end = inputs[1].get_attribute('value') or ''
+                if actual_start.startswith(expected_start) and actual_end.startswith(expected_end):
+                    print(f"    [OK] 日期已设置: {actual_start} ~ {actual_end}")
+                    return
+                else:
+                    print(f"    [WARN] 直接填写结果不符: {actual_start} ~ {actual_end}，尝试面板点击...")
+        else:
+            print("    [WARN] 未找到日期输入框，尝试面板点击...")
+    except Exception as e:
+        print(f"    [WARN] 直接填写失败: {e}，尝试面板点击...")
+
+    # 策略2: 面板点击（兜底）
+    date_editor = page.locator(editor_selector).first
+    date_editor.click()
+    page.wait_for_timeout(1500)
+    popper = page.locator('.rcd-picker__popper').first
+    try:
+        popper.wait_for(state="visible", timeout=5000)
+    except:
+        print("    [WARN] 日历弹窗未出现")
+        return
+
+    # 获取左面板当前月份
+    try:
+        left_header = popper.locator('.rcd-date-range-picker__content.is-left .rcd-date-range-picker__header div').first
+        left_month_text = left_header.inner_text()
+        m = re.search(r'(\d+)\s*年\s*(\d+)\s*月', left_month_text)
+        if m:
+            current_left_year, current_left_month = int(m.group(1)), int(m.group(2))
+        else:
+            current_left_year, current_left_month = datetime.now().year, datetime.now().month
+    except Exception:
+        current_left_year, current_left_month = datetime.now().year, datetime.now().month
+
+    # 调整到目标开始月份
+    target_start_year, target_start_month = last_monday.year, last_monday.month
+    month_diff = (target_start_year - current_left_year) * 12 + (target_start_month - current_left_month)
+    if month_diff < 0:
+        left_arrow = popper.locator('.rcd-date-range-picker__content.is-left .arrow-left').first
+        for _ in range(abs(month_diff)):
+            if left_arrow.count() > 0:
+                left_arrow.click()
+                page.wait_for_timeout(300)
+    elif month_diff > 0:
+        right_arrow = popper.locator('.rcd-date-range-picker__content.is-left .arrow-right').first
+        for _ in range(month_diff):
+            if right_arrow.count() > 0:
+                right_arrow.click()
+                page.wait_for_timeout(300)
+    page.wait_for_timeout(800)
+
+    # 判断是否需要跨月选择
+    same_month = (last_monday.year == last_sunday.year and last_monday.month == last_sunday.month)
+
+    if same_month:
+        # 同月：两个日期都在左面板选择
+        # 先点击开始日期
+        start_clicked = page.evaluate(f"""() => {{
+            const panel = document.querySelector('.rcd-date-range-picker__content.is-left');
+            if (!panel) return false;
+            const cells = panel.querySelectorAll('td');
+            for (const cell of cells) {{
+                if (cell.classList.contains('prev-month') || cell.classList.contains('next-month')) continue;
+                const text = cell.textContent.trim();
+                if (text === '{last_monday.day}') {{ cell.click(); return true; }}
+            }}
+            return false;
+        }}""")
+        page.wait_for_timeout(500)
+        # 再点击结束日期（仍在左面板）
+        end_clicked = page.evaluate(f"""() => {{
+            const panel = document.querySelector('.rcd-date-range-picker__content.is-left');
+            if (!panel) return false;
+            const cells = panel.querySelectorAll('td');
+            for (const cell of cells) {{
+                if (cell.classList.contains('prev-month') || cell.classList.contains('next-month')) continue;
+                const text = cell.textContent.trim();
+                if (text === '{last_sunday.day}') {{ cell.click(); return true; }}
+            }}
+            return false;
+        }}""")
+    else:
+        # 跨月：开始日期在左面板，结束日期在右面板
+        start_clicked = page.evaluate(f"""() => {{
+            const panel = document.querySelector('.rcd-date-range-picker__content.is-left');
+            if (!panel) return false;
+            const cells = panel.querySelectorAll('td');
+            for (const cell of cells) {{
+                if (cell.classList.contains('prev-month') || cell.classList.contains('next-month')) continue;
+                const text = cell.textContent.trim();
+                if (text === '{last_monday.day}') {{ cell.click(); return true; }}
+            }}
+            return false;
+        }}""")
+        page.wait_for_timeout(500)
+        end_clicked = page.evaluate(f"""() => {{
+            const panel = document.querySelector('.rcd-date-range-picker__content.is-right');
+            if (!panel) return false;
+            const cells = panel.querySelectorAll('td');
+            for (const cell of cells) {{
+                if (cell.classList.contains('prev-month') || cell.classList.contains('next-month')) continue;
+                const text = cell.textContent.trim();
+                if (text === '{last_sunday.day}') {{ cell.click(); return true; }}
+            }}
+            return false;
+        }}""")
+
+    if start_clicked:
+        print(f"    [OK] 点击开始日期: {expected_start}")
+    if end_clicked:
+        print(f"    [OK] 点击结束日期: {expected_end}")
+
+    page.wait_for_timeout(1000)
+    try:
+        popper.wait_for(state="hidden", timeout=5000)
+    except:
+        pass
+
+    # 最终验证
+    try:
+        inputs = page.locator(input_selector).all()
+        if len(inputs) >= 2:
+            actual_start = inputs[0].get_attribute('value') or ''
+            actual_end = inputs[1].get_attribute('value') or ''
+            print(f"    [INFO] 当前日期范围: {actual_start} ~ {actual_end}")
+    except:
+        pass
+    print("  [OK] 日期范围设置完成")
+
+
+def click_query_button_appraise(page):
+    """点击查询按钮"""
+    print("  点击查询...")
+    try:
+        query_btn = page.locator('.rcd-button--primary:has-text("查询")').first
+        query_btn.click()
+    except Exception as e:
+        print(f"  [WARN] 点击查询失败: {e}")
+        page.evaluate("""() => {
+            const btns = document.querySelectorAll('.rcd-button');
+            for (const btn of btns) {
+                if (btn.textContent.includes('查询')) { btn.click(); return 'clicked'; }
+            }
+            return 'not_found';
+        }""")
+    page.wait_for_timeout(3000)
+    print("  [OK] 查询完成")
+
+
+def extract_negative_reviews(page):
+    """提取当前页差评表格数据"""
+    data = page.evaluate("""() => {
+        const rows = document.querySelectorAll('.rcd-table__body tbody tr');
+        const result = [];
+        for (const row of rows) {
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 5) continue;
+            const ratingCell = cells[1];
+            const userName = ratingCell.querySelector('.appraise-manage_user')?.textContent?.trim() || '';
+            const scoreEl = ratingCell.querySelector('.rcd-rate');
+            const score = scoreEl ? (scoreEl.getAttribute('aria-valuenow') || '') : '';
+            const tags = Array.from(ratingCell.querySelectorAll('.rcd-tag__content')).map(t => t.textContent.trim()).filter(Boolean);
+            const contentCell = cells[2];
+            const contentMain = contentCell.querySelector('.appraise-content_main')?.textContent?.trim() || '';
+            const contentTime = contentCell.querySelector('.appraise-content_time')?.textContent?.trim() || '';
+            const orderCell = cells[3];
+            const orderIdSpan = orderCell.querySelector('.order-id span');
+            const orderId = orderIdSpan ? orderIdSpan.textContent.replace('订单编号：', '').trim() : '';
+            const skuName = orderCell.querySelector('.sku-name-text')?.textContent?.trim() || '';
+            const skuIdText = orderCell.querySelector('.sku-id')?.textContent?.trim() || '';
+            const skuId = skuIdText.replace('SKU编码：', '').trim();
+            const skuSize = orderCell.querySelector('.sku-size')?.textContent?.trim() || '';
+            result.push({
+                user_name: userName,
+                score: score,
+                tags: tags,
+                content: contentMain,
+                time: contentTime,
+                order_id: orderId,
+                sku_name: skuName,
+                sku_id: skuId,
+                sku_size: skuSize,
+            });
+        }
+        return JSON.stringify(result);
+    }""")
+    try:
+        reviews = json.loads(data) if isinstance(data, str) else []
+    except:
+        reviews = []
+    return reviews
+
+
+def collect_all_negative_reviews(page):
+    """翻页采集所有差评"""
+    all_reviews = []
+    page_num = 1
+    max_pages = 50
+    while page_num <= max_pages:
+        print(f"    采集第 {page_num} 页...")
+        reviews = extract_negative_reviews(page)
+        all_reviews.extend(reviews)
+        try:
+            next_btn = page.locator('.rcd-pagination .btn-next').first
+            if next_btn.count() == 0:
+                print("    无分页组件")
+                break
+            if next_btn.is_disabled():
+                print("    已到达最后一页")
+                break
+            next_btn.click()
+            page.wait_for_timeout(3000)
+            page_num += 1
+        except Exception as e:
+            print(f"    [WARN] 分页处理异常: {e}")
+            break
+    print(f"  [OK] 共采集 {len(all_reviews)} 条差评")
+    return all_reviews
+
+
+def save_negative_reviews_md(reviews, output_base):
+    """保存差评数据为 Markdown"""
+    today = datetime.now().strftime("%Y-%m-%d")
+    start_day, end_day = get_yesterday_range_text()
+    week_range = f"{start_day} ~ {end_day}"
+    md_path = os.path.join(output_base, f"差评数据_{today}.md")
+    md = []
+    md.append(f"# 商品差评数据")
+    md.append("")
+    md.append(f"**采集日期：** {today}")
+    md.append(f"**评价时间范围：** {week_range}")
+    md.append(f"**差评总数：** {len(reviews)} 条")
+    md.append("")
+    if not reviews:
+        md.append("> 昨天暂无差评数据")
+    else:
+        for i, review in enumerate(reviews, 1):
+            md.append(f"## 差评 {i}")
+            md.append("")
+            md.append(f"- **用户昵称：** {review.get('user_name', '')}")
+            score = review.get('score', '')
+            md.append(f"- **评分：** {score} 星" if score else "- **评分：** 差评")
+            tags = review.get('tags', [])
+            if tags:
+                md.append(f"- **标签：** {', '.join(tags)}")
+            md.append(f"- **评价时间：** {review.get('time', '')}")
+            md.append(f"- **评价内容：** {review.get('content', '')}")
+            md.append(f"- **订单编号：** {review.get('order_id', '')}")
+            md.append(f"- **商品名称：** {review.get('sku_name', '')}")
+            md.append(f"- **SKU编码：** {review.get('sku_id', '')}")
+            md.append(f"- **规格：** {review.get('sku_size', '')}")
+            md.append("")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md))
+    print(f"  [OK] 差评数据已保存: {md_path}")
+    return md_path
+
+
+# ── 保存单个 SPU 详情 ─────────────────────────────────────────────────────────
+
+def save_single_product_detail(spu_info, core_data, traffic_data, keyword_data, crowd_data, output_base,
+                          is_apollo2=False, is_zeus3=False, is_xingji=False,
+                          apollo2_summary=None, zeus3_summary=None, xingji_summary=None):
+    """保存单个 SPU 的核心数据、流量来源、关键词来源成交、人群画像到 单品详情数据_日期/SPU子文件夹/（JSON + Markdown）"""
+    today = datetime.now().strftime("%Y-%m-%d")
+    detail_dir = os.path.join(output_base, f"单品详情数据_{today}")
+
+    spu = spu_info.get('spu', '')
+    product = spu_info.get('product', '')
+    safe_name = re.sub(r'[\\/:*?"<>|]', '', product)[:30]
+    file_name = f"{spu}_{safe_name}" if safe_name else spu
+
+    # 阿波罗2或宙斯3或星迹使用共享文件夹
+    if is_apollo2:
+        spu_dir = os.path.join(detail_dir, APOLLO2_FOLDER)
+    elif is_zeus3:
+        spu_dir = os.path.join(detail_dir, ZEUS3_FOLDER)
+    elif is_xingji:
+        spu_dir = os.path.join(detail_dir, XINGJI_FOLDER)
+    else:
+        spu_dir = os.path.join(detail_dir, file_name)
+    os.makedirs(spu_dir, exist_ok=True)
+
+    start_day, end_day = get_yesterday_range_text()
+    week_range = f"{start_day} ~ {end_day}"
+
+    record = {
+        "spu": spu,
+        "product": product,
+        "image": spu_info.get('image', ''),
+        "url": spu_info.get('url', ''),
+        "采集日期": today,
+        "数据周期": week_range,
+        "核心数据": core_data,
+        "流量来源": traffic_data,
+        "关键词来源成交": keyword_data,
+        "人群画像": crowd_data,
+    }
+
+    # ── 计算转化率和加购率 ──────────────────────────────────────────────────────
+    # 转化率 = 加购客户数 / 访客数
+    # 加购率 = 加购商品件数 / 访客数
+    try:
+        visitors = 0
+        add_cart_customers = 0
+        add_cart_items = 0
+        # 从核心数据的"交易"分组中提取访客数、加购客户数、加购商品件数
+        for group_name, items in core_data.items():
+            for title, v in items.items():
+                val = v.get('value', '')
+                if not val or val == 'N/A':
+                    continue
+                # 尝试解析数值（去除逗号）
+                try:
+                    num = float(str(val).replace(',', '').replace('%', ''))
+                except:
+                    continue
+                if '访客数' in title or '访客' in title:
+                    visitors = num
+                if '加购客户' in title or '加购人' in title:
+                    add_cart_customers = num
+                if '加购商品件' in title or '加购商品' in title:
+                    add_cart_items = num
+
+        if visitors > 0:
+            record['转化率'] = f"{(add_cart_customers / visitors * 100):.2f}%"
+            record['加购率'] = f"{(add_cart_items / visitors * 100):.2f}%"
+        else:
+            record['转化率'] = 'N/A'
+            record['加购率'] = 'N/A'
+    except Exception as e:
+        print(f"    [WARN] 计算转化率/加购率异常: {e}")
+        record['转化率'] = 'N/A'
+        record['加购率'] = 'N/A'
+
+    # ── 提取待插入数据指标 ────────────────────────────────────────────────────
+    # 销量=成交金额, 访客数, 转化率, 加购数(=加购客户数), 退款率(=退货率)
+    # 注：访客数、加购客户数、加购商品件数已在上面计算转化率时提取
+    sales_amount = 0
+    refund_rate = 'N/A'
+    for group_name, items in core_data.items():
+        for title, v in items.items():
+            val = v.get('value', '')
+            if not val or val == 'N/A':
+                continue
+            if '成交金额' in title:
+                sales_amount = val  # 保留原始格式
+            if '退货率' in title:
+                refund_rate = val
+
+    # 阿波罗2或宙斯3或星迹：暂存汇总数据，不在此处保存待插入数据
+    target_summary = None
+    if is_apollo2:
+        target_summary = apollo2_summary
+    elif is_zeus3:
+        target_summary = zeus3_summary
+    elif is_xingji:
+        target_summary = xingji_summary
+
+    if target_summary is not None:
+        try:
+            target_summary['sales_amount'] += float(str(sales_amount).replace(',', '') if sales_amount else 0)
+        except:
+            pass
+        try:
+            target_summary['visitors'] += float(visitors) if visitors else 0
+        except:
+            pass
+        try:
+            target_summary['add_cart_customers'] += float(add_cart_customers) if add_cart_customers else 0
+        except:
+            pass
+        try:
+            target_summary['add_cart_items'] += float(add_cart_items) if add_cart_items else 0
+        except:
+            pass
+        if refund_rate != 'N/A':
+            try:
+                target_summary['refund_rate_sum'] += float(str(refund_rate).replace('%', ''))
+                target_summary['refund_rate_count'] += 1
+            except:
+                pass
+        # 多 SPU 产品：每个 SPU 仍然保存各自的待插入数据（文件名含 SPU 区分），不 return
+
+    summary_record = {
+        "spu": spu,
+        "product": product,
+        "采集日期": today,
+        "数据周期": week_range,
+        "销量（成交金额）": sales_amount if sales_amount else 'N/A',
+        "访客数": visitors if visitors else 'N/A',
+        "转化率": record.get('转化率', 'N/A'),
+        "加购数（加购客户数）": add_cart_customers if add_cart_customers else 'N/A',
+        "退款率（退货率）": refund_rate,
+    }
+
+    # ── 保存待插入数据 JSON ─────────────────────────────────────────────────
+    # 多 SPU 产品用 SPU 编号区分文件名，单 SPU 产品直接用"待插入数据"
+    if target_summary is not None:
+        summary_json_path = os.path.join(spu_dir, f"待插入数据_{spu}.json")
+    else:
+        summary_json_path = os.path.join(spu_dir, "待插入数据.json")
+    with open(summary_json_path, "w", encoding="utf-8") as f:
+        json.dump(summary_record, f, ensure_ascii=False, indent=2)
+    print(f"    [OK] 待插入数据 JSON 已保存: {summary_json_path}")
+
+    # ── 保存待插入数据 Markdown ─────────────────────────────────────────────
+    summary_md = []
+    summary_md.append(f"# {product} - 待插入数据")
+    summary_md.append("")
+    summary_md.append(f"**SPU：** {spu}")
+    summary_md.append(f"**采集日期：** {today}")
+    summary_md.append(f"**数据周期：** {week_range}")
+    summary_md.append("")
+    summary_md.append("| 指标 | 数值 |")
+    summary_md.append("|------|------|")
+    summary_md.append(f"| 销量（成交金额） | {summary_record['销量（成交金额）']} |")
+    summary_md.append(f"| 访客数 | {summary_record['访客数']} |")
+    summary_md.append(f"| 转化率 | {summary_record['转化率']} |")
+    summary_md.append(f"| 加购数（加购客户数） | {summary_record['加购数（加购客户数）']} |")
+    summary_md.append(f"| 退款率（退货率） | {summary_record['退款率（退货率）']} |")
+    if target_summary is not None:
+        summary_md_path = os.path.join(spu_dir, f"待插入数据_{spu}.md")
+    else:
+        summary_md_path = os.path.join(spu_dir, "待插入数据.md")
+    with open(summary_md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(summary_md))
+    print(f"    [OK] 待插入数据 Markdown 已保存: {summary_md_path}")
+
+    # ── JSON ──
+    json_path = os.path.join(spu_dir, f"{file_name}.json")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(record, f, ensure_ascii=False, indent=2)
+    print(f"    [OK] JSON 已保存: {json_path}")
+
+    # ── Markdown ──
+    md = []
+    md.append(f"# {product}")
+    md.append("")
+    md.append(f"**SPU：** {spu}")
+    md.append(f"**采集日期：** {today}")
+    md.append(f"**数据周期：** {week_range}")
+    if spu_info.get('url'):
+        md.append(f"**商品链接：** [{spu_info['url']}]({spu_info['url']})")
+    md.append("")
+
+    # 核心数据
+    for group_name, items in core_data.items():
+        md.append(f"## {group_name}")
+        md.append("")
+        md.append("| 指标 | 数值 | 对比标签 | 对比值 | 趋势 |")
+        md.append("|------|------|----------|--------|------|")
+        for title, v in items.items():
+            value = v.get('value', '')
+            compare_label = v.get('compare_label', '')
+            compare_value = v.get('compare_value', '')
+            trend = v.get('trend', '')
+            trend_icon = '📈' if trend == 'up' else ('📉' if trend == 'down' else '')
+            md.append(f"| {title} | {value} | {compare_label} | {compare_value} | {trend_icon} |")
+        md.append("")
+
+    # 流量来源
+    if traffic_data:
+        md.append("## 流量来源")
+        md.append("")
+        headers = list(traffic_data[0].keys())
+        md.append("| " + " | ".join(headers) + " |")
+        md.append("| " + " | ".join(["------"] * len(headers)) + " |")
+        for row in traffic_data:
+            cells = []
+            for h in headers:
+                val = row.get(h, '')
+                if isinstance(val, dict):
+                    v = val.get('value', '')
+                    r = val.get('rate', '')
+                    t = val.get('trend', '')
+                    icon = '📈' if t == 'up' else ('📉' if t == 'down' else '')
+                    cells.append(f"{v} {r}{icon}" if r else v)
+                else:
+                    cells.append(str(val))
+            md.append("| " + " | ".join(cells) + " |")
+        md.append("")
+
+    # 关键词来源成交
+    if keyword_data:
+        md.append("## 关键词来源成交")
+        md.append("")
+        headers = list(keyword_data[0].keys())
+        md.append("| " + " | ".join(headers) + " |")
+        md.append("| " + " | ".join(["------"] * len(headers)) + " |")
+        for row in keyword_data:
+            cells = [str(row.get(h, '')) for h in headers]
+            md.append("| " + " | ".join(cells) + " |")
+        md.append("")
+
+    # 人群画像
+    if crowd_data:
+        md.append("## 人群画像")
+        md.append("")
+        if crowd_data.get("性别"):
+            md.append("### 性别分布")
+            md.append("")
+            md.append("| 性别 | 占比 |")
+            md.append("|------|------|")
+            for item in crowd_data["性别"]:
+                md.append(f"| {item.get('name', '')} | {item.get('value', '')} |")
+            md.append("")
+        if crowd_data.get("年龄"):
+            md.append("### 年龄分布")
+            md.append("")
+            md.append("| 年龄段 | 占比 |")
+            md.append("|--------|------|")
+            for item in crowd_data["年龄"]:
+                md.append(f"| {item.get('name', '')} | {item.get('value', '')} |")
+            md.append("")
+        if crowd_data.get("地区_省份"):
+            md.append("### 地区分布（省份）")
+            md.append("")
+            md.append("| 排名 | 地区 | 占比 |")
+            md.append("|------|------|------|")
+            for item in crowd_data["地区_省份"]:
+                md.append(f"| {item.get('rank', '')} | {item.get('region', '')} | {item.get('ratio', '')} |")
+            md.append("")
+        if crowd_data.get("地区_城市"):
+            md.append("### 地区分布（城市）")
+            md.append("")
+            md.append("| 排名 | 城市 | 占比 |")
+            md.append("|------|------|------|")
+            for item in crowd_data["地区_城市"]:
+                md.append(f"| {item.get('rank', '')} | {item.get('region', '')} | {item.get('ratio', '')} |")
+            md.append("")
+        if crowd_data.get("职业"):
+            md.append("### 职业分布")
+            md.append("")
+            md.append("| 职业 | 占比 |")
+            md.append("|------|------|")
+            for item in crowd_data["职业"]:
+                md.append(f"| {item.get('name', '')} | {item.get('value', '')} |")
+            md.append("")
+
+    md_path = os.path.join(spu_dir, f"{file_name}.md")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md))
+    print(f"    [OK] Markdown 已保存: {md_path}")
+
+    return json_path, md_path, spu_dir
+
+
+# ── 逐个 SPU 采集详情主流程 ───────────────────────────────────────────────────
+
+def _single_product_output_exists(spu_info, output_base):
+    today = datetime.now().strftime("%Y-%m-%d")
+    spu = spu_info.get("spu", "")
+    product = spu_info.get("product", "")[:40]
+    safe_name = re.sub(r'[\\/:*?"<>|]', '', product)[:30]
+    file_name = f"{spu}_{safe_name}" if safe_name else spu
+
+    if spu in APOLLO2_SPU_LIST:
+        spu_dir = os.path.join(output_base, f"单品详情数据_{today}", APOLLO2_FOLDER)
+        json_path = os.path.join(spu_dir, f"{file_name}.json")
+    elif spu in ZEUS3_SPU_LIST:
+        spu_dir = os.path.join(output_base, f"单品详情数据_{today}", ZEUS3_FOLDER)
+        json_path = os.path.join(spu_dir, f"{file_name}.json")
+    elif spu in XINGJI_SPU_LIST:
+        spu_dir = os.path.join(output_base, f"单品详情数据_{today}", XINGJI_FOLDER)
+        json_path = os.path.join(spu_dir, f"{file_name}.json")
+    else:
+        spu_dir = os.path.join(output_base, f"单品详情数据_{today}", file_name)
+        json_path = os.path.join(spu_dir, f"{file_name}.json")
+
+    return os.path.exists(json_path)
+
+
+def collect_each_spu_detail(page, context, all_spus, output_base, resume=False):
+    """逐个搜索 SPU → 点击详情 → 新页面 → 核心数据 → 保存"""
+    start_day, end_day = get_yesterday_range_text()
+    week_range = f"{start_day} ~ {end_day}"
+    total = len(all_spus)
+    print(f"\n[Step 6] 开始逐个采集 {total} 个 SPU 的详情...")
+    progress = SpuProgress(output_base, "product_spu_detail")
+
+    if resume:
+        completed_spus = progress.successful_spus()
+        filtered = []
+        skipped = 0
+        for spu_info in all_spus:
+            spu = str(spu_info.get("spu", ""))
+            if spu in completed_spus or _single_product_output_exists(spu_info, output_base):
+                skipped += 1
+            else:
+                filtered.append(spu_info)
+        if skipped:
+            print(f"  [Resume] 跳过已成功采集的 {skipped} 个 SPU，剩余 {len(filtered)} 个待采集")
+        all_spus = filtered
+        total = len(all_spus)
+
+    if not all_spus:
+        print("  [OK] 所有 SPU 均已采集，无需继续")
+        return 0, 0
+
+    success = 0
+    failed = 0
+
+    def make_summary():
+        return {"sales_amount": 0, "visitors": 0, "add_cart_customers": 0,
+                "add_cart_items": 0, "refund_rate_sum": 0, "refund_rate_count": 0}
+
+    apollo2_summary = make_summary()
+    zeus3_summary = make_summary()
+    xingji_summary = make_summary()
+
+    for idx, spu_info in enumerate(all_spus):
+        spu = spu_info.get('spu', '')
+        product = spu_info.get('product', '')[:40]
+
+        # 判断所属分组
+        if spu in APOLLO2_SPU_LIST:
+            group_name, group_folder, group_summary = "阿波罗2", APOLLO2_FOLDER, apollo2_summary
+        elif spu in ZEUS3_SPU_LIST:
+            group_name, group_folder, group_summary = "宙斯3", ZEUS3_FOLDER, zeus3_summary
+        elif spu in XINGJI_SPU_LIST:
+            group_name, group_folder, group_summary = "星迹", XINGJI_FOLDER, xingji_summary
+        else:
+            group_name, group_folder, group_summary = None, None, None
+
+        print(f"\n{'='*50}")
+        print(f"  [{idx+1}/{total}] SPU: {spu}")
+        print(f"  商品: {product}...")
+        if group_name:
+            print(f"  [{group_name}] 共享文件夹模式")
+        print(f"{'='*50}")
+
+        if not spu:
+            print("  [SKIP] SPU 为空，跳过")
+            failed += 1
+            progress.append("failed", spu_info, error="SPU 为空")
+            continue
+
+        progress.mark_running(spu_info, idx + 1, total)
+
+        try:
+            # 1. 搜索 SPU
+            search_spu_in_detail_page(page, spu)
+
+            # 2. 点击详情，打开新页面
+            detail_page = click_detail_and_wait_new_page(page, context, spu)
+            if not detail_page:
+                failed += 1
+                progress.append("failed", spu_info, error="详情页未打开")
+                continue
+
+            # 3. 点击核心数据 tab
+            click_core_data_tab(detail_page)
+
+            # 4. 选择昨天日期
+            select_last_week_on_detail_page(detail_page)
+
+            # 5. 采集核心数据
+            core_data = extract_core_data(detail_page)
+
+            # 6. 点击流量来源 tab
+            click_traffic_source_tab(detail_page)
+
+            # 7. 展开所有行
+            expand_all_traffic_rows(detail_page)
+
+            # 8. 采集流量来源数据
+            traffic_data = extract_traffic_data(detail_page)
+
+            # 10. 采集关键词来源成交数据
+            keyword_data = extract_keyword_data(detail_page)
+
+            # 11. 点击人群画像tab
+            click_crowd_portrait_tab(detail_page)
+
+            # 12. 采集人群画像数据
+            crowd_data = extract_crowd_portrait_data(detail_page)
+
+            # 13. 截图（保存到 SPU 子文件夹）
+            today = datetime.now().strftime("%Y-%m-%d")
+            safe_name = re.sub(r'[\\/:*?"<>|]', '', product)[:30]
+            file_name = f"{spu}_{safe_name}" if safe_name else spu
+
+            if group_name:
+                spu_dir = os.path.join(output_base, f"单品详情数据_{today}", group_folder)
+                os.makedirs(spu_dir, exist_ok=True)
+            else:
+                spu_dir = os.path.join(output_base, f"单品详情数据_{today}", file_name)
+                os.makedirs(spu_dir, exist_ok=True)
+
+            if group_name:
+                screenshot_path = os.path.join(spu_dir, f"screenshot_{spu}.png")
+            else:
+                screenshot_path = os.path.join(spu_dir, "screenshot.png")
+            detail_page.screenshot(path=screenshot_path, full_page=True)
+            print(f"    [OK] 截图已保存: {screenshot_path}")
+
+            # 14. 保存（传入 crowd_data）
+            save_single_product_detail(spu_info, core_data, traffic_data, keyword_data, crowd_data, output_base,
+                                    is_apollo2=(group_name == "阿波罗2"), is_zeus3=(group_name == "宙斯3"), is_xingji=(group_name == "星迹"),
+                                    apollo2_summary=apollo2_summary, zeus3_summary=zeus3_summary, xingji_summary=xingji_summary)
+            success += 1
+            progress.mark_success(spu_info, group=group_name or "")
+
+            # 15. 关闭详情页后，立即采集当前 SPU 的商品榜单排名并追加到 JSON/MD
+            detail_page.close()
+            page.wait_for_timeout(2000)
+            try:
+                collect_rank_for_single_spu(context, spu_info, output_base)
+            except Exception as e:
+                print(f"  [WARN] 当前 SPU 榜单排名采集失败: {e}")
+
+        except Exception as e:
+            print(f"  [FAIL] 异常: {e}")
+            failed += 1
+            progress.mark_failed(spu_info, e)
+            try:
+                # 只关闭当前 SPU 的详情页，避免误关主页面
+                detail_page_local = locals().get('detail_page')
+                if detail_page_local and detail_page_local != page:
+                    detail_page_local.close()
+            except:
+                pass
+
+    print(f"\n[OK] SPU 详情采集完成: 成功 {success}, 失败 {failed}")
+    print(f"  [Progress] 进度日志: {progress.path}")
+
+    # ── 汇总函数：阿波罗2/宙斯3 通用 ─────────────────────────────────────────
+    def save_group_summary(group_name, folder, spu_list, summary, week_range):
+        if summary['visitors'] > 0:
+            summary['转化率'] = f"{(summary['add_cart_customers'] / summary['visitors'] * 100):.2f}%"
+        else:
+            summary['转化率'] = 'N/A'
+        if summary['refund_rate_count'] > 0:
+            summary['退款率'] = f"{(summary['refund_rate_sum'] / summary['refund_rate_count']):.2f}%"
+        else:
+            summary['退款率'] = 'N/A'
+
+        today = datetime.now().strftime("%Y-%m-%d")
+        group_dir = os.path.join(output_base, f"单品详情数据_{today}", folder)
+        os.makedirs(group_dir, exist_ok=True)
+
+        record = {
+            "product": group_name,
+            "spu_list": spu_list,
+            "采集日期": today,
+            "数据周期": week_range,
+            "销量（成交金额）": f"{summary['sales_amount']:.2f}" if summary['sales_amount'] else 'N/A',
+            "访客数": int(summary['visitors']) if summary['visitors'] else 'N/A',
+            "转化率": summary['转化率'],
+            "加购数（加购客户数）": int(summary['add_cart_customers']) if summary['add_cart_customers'] else 'N/A',
+            "退款率（退货率）": summary['退款率'],
+        }
+
+        json_path = os.path.join(group_dir, "待插入数据.json")
+        with open(json_path, "w", encoding="utf-8") as f:
+            json.dump(record, f, ensure_ascii=False, indent=2)
+        print(f"    [OK] {group_name} 待插入数据 JSON 已保存: {json_path}")
+
+        md = []
+        md.append(f"# {group_name} - 待插入数据")
+        md.append("")
+        md.append(f"**SPU列表：** {', '.join(spu_list)}")
+        md.append(f"**采集日期：** {today}")
+        md.append(f"**数据周期：** {week_range}")
+        md.append("")
+        md.append("| 指标 | 数值 |")
+        md.append("|------|------|")
+        md.append(f"| 销量（成交金额） | {record['销量（成交金额）']} |")
+        md.append(f"| 访客数 | {record['访客数']} |")
+        md.append(f"| 转化率 | {record['转化率']} |")
+        md.append(f"| 加购数（加购客户数） | {record['加购数（加购客户数）']} |")
+        md.append(f"| 退款率（退货率） | {record['退款率（退货率）']} |")
+        md_path = os.path.join(group_dir, "待插入数据.md")
+        with open(md_path, "w", encoding="utf-8") as f:
+            f.write("\n".join(md))
+        print(f"    [OK] {group_name} 待插入数据 Markdown 已保存: {md_path}")
+
+    # 断点续采时本次只包含剩余 SPU，避免用不完整累加值覆盖已有分组汇总。
+    if resume:
+        print("  [Resume] 跳过分组总汇总覆盖，保留已有汇总文件")
+    else:
+        save_group_summary("阿波罗2", APOLLO2_FOLDER, APOLLO2_SPU_LIST, apollo2_summary, week_range)
+        save_group_summary("宙斯3", ZEUS3_FOLDER, ZEUS3_SPU_LIST, zeus3_summary, week_range)
+        save_group_summary("星迹", XINGJI_FOLDER, XINGJI_SPU_LIST, xingji_summary, week_range)
+
+    return success, failed
+
+
+# ── 商品榜单排名采集 ───────────────────────────────────────────────────────────
+
+def select_week_and_last_week_on_rank_page(page):
+    """在榜单页面设置日期为昨天、按天查询。函数名保留，避免改动调用链。"""
+    print("  设置榜单页面日期为昨天...")
+    target_date = get_yesterday_date()
+    target_day = target_date.strftime("%Y-%m-%d")
+
+    # 检查当前是否已是目标日期
+    try:
+        current_label = page.locator('.jmtd-card-header .jmtd-base-input[data-component-name="DatePicker"] .jmtd-label').first.inner_text(timeout=3000)
+        if "按天查询" in current_label and _date_text_matches_day(current_label, target_day):
+            print(f"    [OK] 日期已是昨天: {current_label}")
+            return
+    except Exception:
+        pass
+
+    # 1. 点击日期选择器
+    date_picker = page.locator('.jmtd-card-header .jmtd-base-input[data-component-name="DatePicker"]').first
+    date_picker.click()
+    page.wait_for_timeout(1500)
+
+    # 2. 点击"按天查询"
+    try:
+        day_option = page.locator('.jmtd-date-picker-switch-panel-item:has-text("按天查询")').first
+        day_option.wait_for(state="visible", timeout=5000)
+        day_option.click()
+        print("    [OK] 选择按天查询")
+    except Exception as e:
+        print(f"    [WARN] 点击按天查询失败: {e}")
+        page.evaluate("""() => {
+            const items = document.querySelectorAll('.jmtd-date-picker-switch-panel-item');
+            for (const item of items) {
+                if (item.textContent.includes('按天查询')) { item.click(); return 'clicked'; }
+            }
+            return 'not_found';
+        }""")
+    page.wait_for_timeout(1500)
+
+    if not _click_day_in_open_datepicker(page, target_date, "product_rank"):
+        raise RuntimeError(f"榜单页面未能选择昨天日期: {target_day}")
+
+    page.wait_for_timeout(3000)
+
+    # 4. 等待弹窗关闭
+    try:
+        page.locator('.jmtd-dropdown-lists').first.wait_for(state="hidden", timeout=5000)
+    except:
+        pass
+
+    # 5. 验证
+    new_label = ""
+    try:
+        new_label = page.locator('.jmtd-card-header .jmtd-base-input[data-component-name="DatePicker"] .jmtd-label').first.inner_text(timeout=3000)
+        print(f"    [INFO] 当前日期: {new_label}")
+    except Exception:
+        pass
+    if new_label and not _date_text_matches_day(new_label, target_day):
+        raise RuntimeError(f"榜单日期未切换到昨天: {new_label}")
+    print("  [OK] 榜单日期设置完成")
+
+
+def select_category_on_rank_page(page, category_path):
+    """在榜单页面选择类目（支持两级类目）"""
+    cat_label = " > ".join(category_path)
+    print(f"  选择类目: {cat_label}")
+
+    # 点击类目 Select（在卡片头部右侧）
+    try:
+        cat_select = page.locator('.jmtd-card-header-right .jmtd-base-input:has(.jmtd-label)').nth(1)
+        cat_select.click()
+    except Exception:
+        # 兜底：点击包含"男包"或"功能箱包"的 select
+        cat_select = page.locator('.jmtd-card-header .jmtd-base-input:has(.jmtd-label:has-text("男包")), .jmtd-card-header .jmtd-base-input:has(.jmtd-label:has-text("功能箱包"))').first
+        cat_select.click()
+    page.wait_for_timeout(1500)
+
+    # 第一列：选择大类目
+    big_cat = category_path[0]
+    try:
+        big_option = page.locator(f'.jmtd-dropdown-option:has-text("{big_cat}")').first
+        big_option.wait_for(state="visible", timeout=5000)
+        big_option.click()
+        print(f"    [OK] 选择大类目: {big_cat}")
+    except Exception as e:
+        print(f"    [WARN] 选择大类目失败: {e}")
+        page.evaluate(f"""() => {{
+            const opts = document.querySelectorAll('.jmtd-dropdown-option');
+            for (const opt of opts) {{
+                if (opt.textContent.trim() === '{big_cat}') {{ opt.click(); return 'clicked'; }}
+            }}
+            return 'not_found';
+        }}""")
+    page.wait_for_timeout(800)
+
+    # 如果有子类目，选择子类目
+    if len(category_path) > 1:
+        sub_cat = category_path[1]
+        try:
+            sub_option = page.locator(f'.jmtd-dropdown-option:has-text("{sub_cat}")').first
+            sub_option.wait_for(state="visible", timeout=5000)
+            sub_option.click()
+            print(f"    [OK] 选择子类目: {sub_cat}")
+        except Exception as e:
+            print(f"    [WARN] 选择子类目失败: {e}")
+            page.evaluate(f"""() => {{
+                const opts = document.querySelectorAll('.jmtd-dropdown-option');
+                for (const opt of opts) {{
+                    if (opt.textContent.trim() === '{sub_cat}') {{ opt.click(); return 'clicked'; }}
+                }}
+                return 'not_found';
+            }}""")
+        page.wait_for_timeout(800)
+    else:
+        # 只选大类目：按 Escape 关闭下拉
+        page.keyboard.press("Escape")
+        page.wait_for_timeout(500)
+
+    # 等待表格刷新
+    page.wait_for_timeout(2000)
+
+
+def search_product_on_rank_page(page, product_name):
+    """在榜单页面搜索商品"""
+    try:
+        # 优先使用工具栏中的搜索框（更靠近表格）
+        search_input = page.locator('.tool-bar input.jmtd-input-input').first
+        if search_input.count() == 0:
+            # 兜底：页面顶部的搜索框
+            search_input = page.locator('input.jmtd-input-input[placeholder*="商品名称"]').first
+        search_input.fill("")
+        page.wait_for_timeout(300)
+        search_input.fill(product_name)
+        page.wait_for_timeout(3000)
+        # 按 Enter 确保搜索生效
+        search_input.press("Enter")
+        page.wait_for_timeout(2000)
+    except Exception as e:
+        print(f"    [WARN] 搜索商品失败: {e}")
+
+
+def extract_rank_table_data(page, product_name):
+    """提取榜单表格中匹配商品的数据"""
+    data = page.evaluate("""(productName) => {
+        const rows = document.querySelectorAll('.jmtd-table-main tbody tr');
+        for (const row of rows) {
+            const nameEl = row.querySelector('.prod-name');
+            const name = nameEl ? (nameEl.getAttribute('title') || nameEl.textContent.trim()) : '';
+            if (!name) continue;
+            // 模糊匹配：互相包含或去除空格后匹配
+            const n1 = name.toLowerCase().replace(/\\s+/g, '');
+            const n2 = productName.toLowerCase().replace(/\\s+/g, '');
+            if (n1.includes(n2) || n2.includes(n1) || name.includes(productName) || productName.includes(name)) {
+                const rankEl = row.querySelector('.jmtd-table-formatter-rank-number');
+                const rank = rankEl ? rankEl.textContent.trim() : '';
+                const cells = row.querySelectorAll('td');
+                return {
+                    rank: rank,
+                    product_name: name,
+                    成交金额: cells[2] ? cells[2].textContent.trim() : '',
+                    成交单量: cells[3] ? cells[3].textContent.trim() : '',
+                    访客数: cells[4] ? cells[4].textContent.trim() : '',
+                    搜索点击次数: cells[5] ? cells[5].textContent.trim() : '',
+                    关注人数: cells[6] ? cells[6].textContent.trim() : '',
+                };
+            }
+        }
+        return null;
+    }""", product_name)
+    return data
+
+
+def append_rank_data_to_spu(spu, file_name, rank_records, detail_dir):
+    """将榜单数据追加到 SPU 的 JSON 和 MD 文件"""
+    spu_dir = os.path.join(detail_dir, file_name)
+    if not os.path.exists(spu_dir):
+        print(f"    [WARN] SPU目录不存在: {spu_dir}")
+        return
+
+    # 更新 JSON
+    json_path = os.path.join(spu_dir, f"{file_name}.json")
+    if os.path.exists(json_path):
+        try:
+            with open(json_path, "r", encoding="utf-8") as f:
+                record = json.load(f)
+            record["商品榜单排名"] = rank_records
+            with open(json_path, "w", encoding="utf-8") as f:
+                json.dump(record, f, ensure_ascii=False, indent=2)
+            print(f"    [OK] JSON 已追加榜单数据")
+        except Exception as e:
+            print(f"    [WARN] JSON更新失败: {e}")
+
+    # 追加 MD
+    md_path = os.path.join(spu_dir, f"{file_name}.md")
+    if os.path.exists(md_path) and rank_records:
+        try:
+            with open(md_path, "a", encoding="utf-8") as f:
+                f.write("\n\n## 商品榜单排名\n\n")
+                f.write("| 类目 | 排名 | 成交金额 | 成交单量 | 访客数 | 搜索点击次数 | 关注人数 |\n")
+                f.write("|------|------|----------|----------|--------|--------------|----------|\n")
+                for r in rank_records:
+                    f.write(f"| {r.get('类目', '')} | {r.get('rank', '')} | {r.get('成交金额', '')} | {r.get('成交单量', '')} | {r.get('访客数', '')} | {r.get('搜索点击次数', '')} | {r.get('关注人数', '')} |\n")
+            print(f"    [OK] Markdown 已追加榜单数据")
+        except Exception as e:
+            print(f"    [WARN] MD追加失败: {e}")
+
+
+def collect_rank_for_single_spu(context, spu_info, output_base):
+    """单个SPU采集完详情后，打开新标签页去榜单查排名，追加到其JSON/MD"""
+    spu = spu_info.get('spu', '')
+    product = spu_info.get('product', '')
+    if not product:
+        return
+
+    print(f"  采集商品榜单排名...")
+
+    # 打开新标签页去榜单页面
+    rank_page = context.new_page()
+
+    try:
+        # 1. 跳转榜单页面
+        rank_page.goto("https://sz.jd.com/szweb/sz/view/industryMarket/productRanks.html", wait_until="domcontentloaded", timeout=30000)
+        rank_page.wait_for_timeout(5000)
+
+        # 2. 设置日期
+        select_week_and_last_week_on_rank_page(rank_page)
+
+        # 3. 先输入一次商品名称，后续只切换类目
+        search_product_on_rank_page(rank_page, product)
+
+        # 4. 类目组合
+        categories = [
+            {"path": ["男包"]},
+            {"path": ["男包", "男士双肩包"]},
+            {"path": ["男包", "男士单肩/斜挎包"]},
+            {"path": ["功能箱包"]},
+            {"path": ["功能箱包", "电脑包"]},
+        ]
+
+        rank_records = []
+
+        for cat in categories:
+            cat_label = " > ".join(cat["path"])
+            print(f"    查询类目: {cat_label}")
+
+            try:
+                select_category_on_rank_page(rank_page, cat["path"])
+
+                rank_data = extract_rank_table_data(rank_page, product)
+                if rank_data:
+                    print(f"      [OK] 排名: {rank_data.get('rank', '')}")
+                    rank_records.append({"类目": cat_label, **rank_data})
+                else:
+                    print(f"      [INFO] 该类目下无此商品")
+            except Exception as e:
+                print(f"      [ERROR] 查询类目异常: {e}")
+
+        # 4. 追加到当前SPU的文件；没有命中时 JSON 也写入空数组，便于后续判断已采集
+        today = datetime.now().strftime("%Y-%m-%d")
+        detail_dir = os.path.join(output_base, f"单品详情数据_{today}")
+        safe_name = re.sub(r'[\\/:*?"<>|]', '', product)[:30]
+        file_name = f"{spu}_{safe_name}" if safe_name else spu
+        append_rank_data_to_spu(spu, file_name, rank_records, detail_dir)
+        if not rank_records:
+            print(f"    [INFO] 所有类目均无此商品排名数据")
+
+    except Exception as e:
+        print(f"    [ERROR] 榜单采集异常: {e}")
+    finally:
+        rank_page.close()
+
+
+# ── 主流程 ─────────────────────────────────────────────────────────────────────
+
+def safe_filename(value, limit=80):
+    text = str(value or "").strip()
+    text = re.sub(r'[\\/:*?"<>|]+', "_", text)
+    text = re.sub(r"\s+", " ", text).strip(" ._")
+    return (text[:limit] if text else "未命名")
+
+
+def _collect_style_spu_tasks():
+    tasks = []
+    for group in STYLE_SPU_GROUPS:
+        style = group["style"]
+        for spu in group["spus"]:
+            tasks.append({"style": style, "spu": str(spu), "product": "", "image": "", "url": ""})
+    return tasks
+
+
+def collect_all_spus(page):
+    """使用固定款式 SPU 清单，不再翻页全量采集商品列表。"""
+    print("\n[Step 5] 使用固定款式 SPU 清单，不再全量翻页采集...")
+    try:
+        select_week_and_last_week(page)
+    except Exception as e:
+        print(f"  [WARN] 商品明细列表日期设置失败，继续按固定 SPU 采集详情: {e}")
+    tasks = _collect_style_spu_tasks()
+    print(f"  [OK] 款式 {len(STYLE_SPU_GROUPS)} 个，SPU {len(tasks)} 个")
+    return tasks
+
+
+def _parse_number(value):
+    if value is None:
+        return 0.0
+    text = str(value).strip()
+    if not text or text in {"-", "--", "N/A", "nan"}:
+        return 0.0
+    negative = text.startswith("-")
+    text = text.replace(",", "").replace("，", "")
+    match = re.search(r"\d+(?:\.\d+)?", text)
+    if not match:
+        return 0.0
+    num = float(match.group(0))
+    if "万" in text:
+        num *= 10000
+    if "亿" in text:
+        num *= 100000000
+    return -num if negative else num
+
+
+def _metric_from_core(core_data, keywords):
+    for group_items in (core_data or {}).values():
+        if not isinstance(group_items, dict):
+            continue
+        for title, item in group_items.items():
+            if any(keyword in str(title) for keyword in keywords):
+                if isinstance(item, dict):
+                    return _parse_number(item.get("value"))
+                return _parse_number(item)
+    return 0.0
+
+
+def _rate(numerator, denominator):
+    return (float(numerator) / float(denominator)) if denominator else 0.0
+
+
+def _percent(value):
+    return f"{value * 100:.2f}%"
+
+
+def _build_spu_metrics(core_data):
+    sales = _metric_from_core(core_data, ["成交单量", "成交订单", "下单单量"])
+    visitors = _metric_from_core(core_data, ["访客数"])
+    add_cart_customers = _metric_from_core(core_data, ["加购客户数"])
+    refund_items = _metric_from_core(core_data, ["申请退货件数", "退货件数"])
+    return {
+        "销量": int(sales),
+        "访客数": int(visitors),
+        "加购客户数": int(add_cart_customers),
+        "申请退货件数": int(refund_items),
+        "转化率": _percent(_rate(sales, visitors)),
+        "加购率": _percent(_rate(add_cart_customers, visitors)),
+        "退款率": _percent(_rate(refund_items, visitors)),
+    }
+
+
+def _aggregate_style_metrics(spu_results):
+    sales = sum(_parse_number(r.get("指标汇总", {}).get("销量")) for r in spu_results)
+    visitors = sum(_parse_number(r.get("指标汇总", {}).get("访客数")) for r in spu_results)
+    add_cart_customers = sum(_parse_number(r.get("指标汇总", {}).get("加购客户数")) for r in spu_results)
+    refund_items = sum(_parse_number(r.get("指标汇总", {}).get("申请退货件数")) for r in spu_results)
+    return {
+        "销量": int(sales),
+        "访客数": int(visitors),
+        "加购客户数": int(add_cart_customers),
+        "申请退货件数": int(refund_items),
+        "转化率": _percent(_rate(sales, visitors)),
+        "加购率": _percent(_rate(add_cart_customers, visitors)),
+        "退款率": _percent(_rate(refund_items, visitors)),
+    }
+
+
+def _today():
+    return datetime.now().strftime("%Y-%m-%d")
+
+
+def _style_dir(output_base, style):
+    return os.path.join(output_base, f"{safe_filename(style)}_{_today()}")
+
+
+def _spu_dir(style_dir, product, spu):
+    name = safe_filename(product or spu, 60)
+    return os.path.join(style_dir, f"{name}_{spu}_{_today()}")
+
+
+def _find_existing_spu_result(style_dir, spu):
+    if not os.path.isdir(style_dir):
+        return None
+    pattern = f"_{spu}_{_today()}"
+    for name in os.listdir(style_dir):
+        subdir = os.path.join(style_dir, name)
+        if not os.path.isdir(subdir) or pattern not in name:
+            continue
+        for file_name in os.listdir(subdir):
+            if file_name.endswith(".json"):
+                path = os.path.join(subdir, file_name)
+                try:
+                    with open(path, "r", encoding="utf-8") as f:
+                        return json.load(f)
+                except Exception:
+                    return None
+    return None
+
+
+def _write_spu_files(record, spu_dir_path):
+    os.makedirs(spu_dir_path, exist_ok=True)
+    file_base = safe_filename(f"{record.get('商品名')}_{record.get('spu')}_{_today()}", 120)
+    json_path = os.path.join(spu_dir_path, f"{file_base}.json")
+    md_path = os.path.join(spu_dir_path, f"{file_base}.md")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(record, f, ensure_ascii=False, indent=2)
+
+    metrics = record.get("指标汇总", {})
+    screenshot = record.get("人物画像截图", "")
+    md = [
+        f"# {record.get('商品名') or record.get('spu')} - 单品数据",
+        "",
+        f"**款式：** {record.get('款式', '')}",
+        f"**SPU：** {record.get('spu', '')}",
+        f"**采集日期：** {record.get('采集日期', '')}",
+        f"**数据周期：** {record.get('数据周期', '')}",
+        "",
+        "## 指标汇总",
+        "",
+        "| 指标 | 数值 |",
+        "|------|------|",
+        f"| 销量 | {metrics.get('销量', 0)} |",
+        f"| 访客数 | {metrics.get('访客数', 0)} |",
+        f"| 转化率 | {metrics.get('转化率', '0.00%')} |",
+        f"| 加购率 | {metrics.get('加购率', '0.00%')} |",
+        f"| 退款率 | {metrics.get('退款率', '0.00%')} |",
+        "",
+    ]
+    if screenshot:
+        md.extend(["## 人物画像截图", "", f"![人物画像]({os.path.basename(screenshot)})", ""])
+    rank_records = record.get("商品榜单排名") or []
+    if rank_records:
+        md.extend([
+            "## 商品榜单排名",
+            "",
+            "| 类目 | 排名 | 成交金额 | 成交单量 | 访客数 |",
+            "|------|------|----------|----------|--------|",
+        ])
+        for item in rank_records:
+            md.append(
+                f"| {item.get('类目', '')} | {item.get('rank', '')} | "
+                f"{item.get('成交金额', '')} | {item.get('成交单量', '')} | {item.get('访客数', '')} |"
+            )
+        md.append("")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md))
+    print(f"    [OK] SPU 数据已保存: {spu_dir_path}")
+    return json_path, md_path
+
+
+def _write_style_summary(style, style_dir_path, spu_results):
+    os.makedirs(style_dir_path, exist_ok=True)
+    start_day, end_day = get_yesterday_range_text()
+    metrics = _aggregate_style_metrics(spu_results)
+    file_base = safe_filename(f"{style}_{_today()}", 120)
+    record = {
+        "款式": style,
+        "采集日期": _today(),
+        "数据周期": f"{start_day} ~ {end_day}",
+        "spu列表": [r.get("spu", "") for r in spu_results],
+        "指标汇总": metrics,
+        "人物画像截图": [
+            os.path.relpath(r.get("人物画像截图", ""), style_dir_path).replace("\\", "/")
+            for r in spu_results if r.get("人物画像截图")
+        ],
+        "spu明细": [
+            {
+                "spu": r.get("spu", ""),
+                "商品名": r.get("商品名", ""),
+                "指标汇总": r.get("指标汇总", {}),
+                "目录": os.path.relpath(r.get("输出目录", ""), style_dir_path).replace("\\", "/")
+                if r.get("输出目录") else "",
+            }
+            for r in spu_results
+        ],
+    }
+    json_path = os.path.join(style_dir_path, f"{file_base}.json")
+    md_path = os.path.join(style_dir_path, f"{file_base}.md")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(record, f, ensure_ascii=False, indent=2)
+
+    md = [
+        f"# {style} - 款式汇总",
+        "",
+        f"**采集日期：** {_today()}",
+        f"**数据周期：** {start_day} ~ {end_day}",
+        "",
+        "## 指标汇总",
+        "",
+        "| 指标 | 数值 |",
+        "|------|------|",
+        f"| 销量 | {metrics.get('销量', 0)} |",
+        f"| 访客数 | {metrics.get('访客数', 0)} |",
+        f"| 转化率 | {metrics.get('转化率', '0.00%')} |",
+        f"| 加购率 | {metrics.get('加购率', '0.00%')} |",
+        f"| 退款率 | {metrics.get('退款率', '0.00%')} |",
+        "",
+        "## SPU 明细",
+        "",
+        "| SPU | 商品名 | 销量 | 访客数 | 转化率 | 加购率 | 退款率 |",
+        "|-----|--------|------|--------|--------|--------|--------|",
+    ]
+    for item in record["spu明细"]:
+        m = item.get("指标汇总", {})
+        md.append(
+            f"| {item.get('spu', '')} | {item.get('商品名', '')} | {m.get('销量', 0)} | "
+            f"{m.get('访客数', 0)} | {m.get('转化率', '0.00%')} | "
+            f"{m.get('加购率', '0.00%')} | {m.get('退款率', '0.00%')} |"
+        )
+    screenshots = record["人物画像截图"]
+    if screenshots:
+        md.extend(["", "## 人物画像截图", ""])
+        for path in screenshots:
+            md.append(f"![人物画像]({path})")
+            md.append("")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md))
+    print(f"  [OK] 款式汇总已保存: {style_dir_path}")
+    return json_path, md_path
+
+
+def _normalize_rank_record(record):
+    if not record:
+        return {}
+    normalized = {}
+    for key, value in record.items():
+        key_text = str(key)
+        if "金额" in key_text:
+            normalized["成交金额"] = value
+        elif "单量" in key_text:
+            normalized["成交单量"] = value
+        elif "访客" in key_text:
+            normalized["访客数"] = value
+        elif "搜索" in key_text:
+            normalized["搜索点击次数"] = value
+        elif "关注" in key_text:
+            normalized["关注人数"] = value
+        else:
+            normalized[key_text] = value
+    return normalized
+
+
+def collect_rank_records_for_single_spu(context, product):
+    if not product:
+        return []
+    print("  采集商品榜单排名...")
+    rank_page = context.new_page()
+    rank_records = []
+    try:
+        rank_page.goto("https://sz.jd.com/szweb/sz/view/industryMarket/productRanks.html", wait_until="domcontentloaded", timeout=30000)
+        rank_page.wait_for_timeout(5000)
+        select_week_and_last_week_on_rank_page(rank_page)
+        search_product_on_rank_page(rank_page, product)
+        categories = [
+            {"path": ["男包"]},
+            {"path": ["男包", "男士双肩包"]},
+            {"path": ["男包", "男士单肩/斜挎包"]},
+            {"path": ["功能箱包"]},
+            {"path": ["功能箱包", "电脑包"]},
+        ]
+        for cat in categories:
+            cat_label = " > ".join(cat["path"])
+            try:
+                select_category_on_rank_page(rank_page, cat["path"])
+                rank_data = extract_rank_table_data(rank_page, product)
+                if rank_data:
+                    rank_records.append({"类目": cat_label, **_normalize_rank_record(rank_data)})
+                    print(f"    [OK] {cat_label} 排名: {rank_data.get('rank', '')}")
+            except Exception as e:
+                print(f"    [WARN] 榜单类目采集失败 {cat_label}: {e}")
+    except Exception as e:
+        print(f"    [WARN] 榜单排名采集失败: {e}")
+    finally:
+        try:
+            rank_page.close()
+        except Exception:
+            pass
+    return rank_records
+
+
+def _spu_info_from_current_rows(page, fallback):
+    rows = extract_product_detail_rows(page)
+    target_spu = str(fallback.get("spu", ""))
+    for row in rows:
+        if str(row.get("spu", "")) == target_spu:
+            return {**fallback, **row}
+    if rows:
+        return {**fallback, **rows[0]}
+    return fallback
+
+
+def collect_each_spu_detail(page, context, all_spus, output_base, resume=False):
+    """按款式固定清单搜索 SPU，保存单 SPU 数据并生成款式汇总。"""
+    start_day, end_day = get_yesterday_range_text()
+    period = f"{start_day} ~ {end_day}"
+    print(f"\n[Step 6] 按固定款式采集 SPU 详情，数据周期: {period}")
+    progress = SpuProgress(output_base, "product_style_spu_detail")
+    task_map = {(task["style"], task["spu"]): task for task in all_spus}
+    success = 0
+    failed = 0
+
+    for style_group in STYLE_SPU_GROUPS:
+        style = style_group["style"]
+        style_dir_path = _style_dir(output_base, style)
+        os.makedirs(style_dir_path, exist_ok=True)
+        style_results = []
+        print(f"\n{'=' * 60}")
+        print(f"款式: {style} | SPU {len(style_group['spus'])} 个")
+        print(f"{'=' * 60}")
+
+        for index, spu in enumerate(style_group["spus"], 1):
+            spu = str(spu)
+            spu_info = task_map.get((style, spu), {"style": style, "spu": spu})
+            if resume:
+                existing = _find_existing_spu_result(style_dir_path, spu)
+                if existing:
+                    print(f"  [Resume] 跳过已采集 SPU: {spu}")
+                    style_results.append(existing)
+                    continue
+
+            print(f"\n  [{index}/{len(style_group['spus'])}] SPU: {spu}")
+            progress.mark_running(spu_info, index, len(style_group["spus"]))
+            detail_page = None
+            try:
+                search_spu_in_detail_page(page, spu)
+                spu_info = _spu_info_from_current_rows(page, spu_info)
+                product = spu_info.get("product") or spu
+                detail_page = click_detail_and_wait_new_page(page, context, spu)
+                if not detail_page:
+                    raise RuntimeError("详情页未打开")
+
+                click_core_data_tab(detail_page)
+                select_last_week_on_detail_page(detail_page)
+                core_data = extract_core_data(detail_page)
+
+                click_traffic_source_tab(detail_page)
+                expand_all_traffic_rows(detail_page)
+                traffic_data = extract_traffic_data(detail_page)
+                keyword_data = extract_keyword_data(detail_page)
+
+                click_crowd_portrait_tab(detail_page)
+                crowd_data = extract_crowd_portrait_data(detail_page)
+
+                metrics = _build_spu_metrics(core_data)
+                spu_dir_path = _spu_dir(style_dir_path, product, spu)
+                os.makedirs(spu_dir_path, exist_ok=True)
+                crowd_screenshot = os.path.join(spu_dir_path, "人物画像截图.png")
+                detail_page.screenshot(path=crowd_screenshot, full_page=True)
+                print(f"    [OK] 人物画像截图已保存: {crowd_screenshot}")
+
+                try:
+                    detail_page.close()
+                except Exception:
+                    pass
+                detail_page = None
+                page.wait_for_timeout(1500)
+
+                rank_records = collect_rank_records_for_single_spu(context, product)
+                record = {
+                    "款式": style,
+                    "spu": spu,
+                    "商品名": product,
+                    "商品图片": spu_info.get("image", ""),
+                    "商品链接": spu_info.get("url", ""),
+                    "采集日期": _today(),
+                    "数据周期": period,
+                    "指标汇总": metrics,
+                    "核心数据": core_data,
+                    "流量来源": traffic_data,
+                    "关键词来源成交": keyword_data,
+                    "人群画像": crowd_data,
+                    "人物画像截图": crowd_screenshot,
+                    "商品榜单排名": rank_records,
+                    "输出目录": spu_dir_path,
+                }
+                _write_spu_files(record, spu_dir_path)
+                style_results.append(record)
+                success += 1
+                progress.mark_success(spu_info, group=style)
+            except Exception as e:
+                failed += 1
+                print(f"  [FAIL] SPU {spu} 采集失败: {e}")
+                progress.mark_failed(spu_info, e)
+                try:
+                    if detail_page:
+                        detail_page.close()
+                except Exception:
+                    pass
+
+        if style_results:
+            _write_style_summary(style, style_dir_path, style_results)
+        else:
+            print(f"  [WARN] 款式 {style} 没有成功采集的 SPU，跳过汇总")
+
+    print(f"\n[OK] 固定款式 SPU 采集完成: 成功 {success}, 失败 {failed}")
+    print(f"  [Progress] 进度日志: {progress.path}")
+    return success, failed
+
+
+def main():
+    parser = argparse.ArgumentParser(description="京东商品明细+单品详情数据采集脚本")
+    parser.add_argument("--shop", default=DEFAULT_CONFIG["shop"], help="店铺账号名")
+    parser.add_argument("--password", default=DEFAULT_CONFIG["password"], help="登录密码")
+    parser.add_argument("--output", default=OUTPUT_BASE, help="输出目录")
+    parser.add_argument("--resume", action="store_true", help="跳过已成功采集的SPU，从断点继续")
+    args = parser.parse_args()
+    args.output = str(managed_data_path(args.output))
+
+    print("=" * 60)
+    print("京东商品明细 + 单品详情数据采集")
+    print(f"  店铺: {args.shop}")
+    print(f"  输出: {args.output}")
+    print("=" * 60)
+
+    start_day, end_day = get_yesterday_range_text()
+    print(f"  昨天: {start_day} ~ {end_day}")
+    print()
+
+    with sync_playwright() as p:
+        browser = p.chromium.launch(
+            headless=False,
+            args=["--disable-gpu"]
+        )
+        context = browser.new_context(
+            viewport={"width": 1920, "height": 1080},
+            locale="zh-CN",
+        )
+        page = context.new_page()
+
+        try:
+            # Step 1: 登录
+            step_login(page, args.shop, args.password)
+
+            # Step 2: 商品评价差评采集（新增）
+            print("\n" + "=" * 60)
+            print("[新增] 采集商品评价差评数据")
+            print("=" * 60)
+            appraise_page = None
+            try:
+                appraise_page = navigate_to_goods_appraise(page)
+                click_negative_reviews_filter(appraise_page)
+                select_last_week_date_range_appraise(appraise_page)
+                click_query_button_appraise(appraise_page)
+                reviews = collect_all_negative_reviews(appraise_page)
+                save_negative_reviews_md(reviews, args.output)
+            except Exception as e:
+                print(f"  [WARN] 商品评价差评采集失败，跳过继续: {e}")
+            finally:
+                if appraise_page and appraise_page != page and not appraise_page.is_closed():
+                    try:
+                        appraise_page.close()
+                    except Exception:
+                        pass
+
+            alive_pages = [p for p in context.pages if not p.is_closed()]
+            page = next((p for p in alive_pages if "shop.jd.com" in p.url), None) or (alive_pages[0] if alive_pages else context.new_page())
+
+            # Step 3~5: 商智 → 返回旧版 → 商品明细页
+            print("\n" + "=" * 60)
+            print("继续进入商智采集单品详情数据")
+            print("=" * 60)
+            if page.is_closed():
+                print("  [INFO] 原商家后台页面已关闭，重新打开页面...")
+                page = context.new_page()
+            if "shop.jd.com" not in page.url:
+                try:
+                    page.goto("https://shop.jd.com/jdm/home", wait_until="domcontentloaded", timeout=60000)
+                except Exception:
+                    page = context.new_page()
+                    page.goto("https://shop.jd.com/jdm/home", wait_until="domcontentloaded", timeout=60000)
+                page.wait_for_timeout(3000)
+            sz_page = step_enter_shangzhi(context, page)
+            step_return_old_version(sz_page)
+
+            # Step 6: 商品明细页
+            step_goto_product_detail(sz_page)
+            close_guide_if_present(sz_page)
+
+            # Step 7: 使用固定款式 SPU 清单
+            all_spus = collect_all_spus(sz_page)
+
+            if not all_spus:
+                print("\n[WARN] 未采集到任何 SPU，退出")
+                return
+
+            # Step 8: 按款式逐个搜索 SPU → 详情 → 单品保存 → 款式汇总
+            try:
+                collect_each_spu_detail(sz_page, context, all_spus, args.output, resume=args.resume)
+            except Exception as e:
+                print(f"\n[ERROR] 单品详情采集异常: {e}")
+                import traceback
+                traceback.print_exc()
+
+            print()
+            print("=" * 60)
+            print("全部采集完成！")
+            detail_output = args.output
+            print(f"  差评数据: {os.path.join(args.output, '差评数据_' + datetime.now().strftime('%Y-%m-%d') + '.md')}")
+            print(f"  单品数据: {detail_output}")
+            print("=" * 60)
+
+        except KeyboardInterrupt:
+            print("\n用户中断，退出。")
+        except Exception as e:
+            print(f"\n[ERROR] 出错: {e}")
+            import traceback
+            traceback.print_exc()
+        finally:
+            context.close()
+            browser.close()
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/state.py b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/state.py
new file mode 100644
index 0000000..191f6ca
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/vendors/jd-data-flow/state.py
@@ -0,0 +1,15 @@
+from typing import TypedDict
+
+
+class PipelineState(TypedDict):
+    config: dict
+    report_path: str
+    report_paths: dict
+    raw_data_paths: dict
+    metrics_paths: dict
+    analysis_result: str
+    collector_response: str
+    notification_sent: bool
+    base_written: bool
+    task_errors: list[str]
+    error: str | None
diff --git a/src/gyxx_flow/modules/product_commerce/runtime/weekly_aggregate.py b/src/gyxx_flow/modules/product_commerce/runtime/weekly_aggregate.py
new file mode 100644
index 0000000..7309835
--- /dev/null
+++ b/src/gyxx_flow/modules/product_commerce/runtime/weekly_aggregate.py
@@ -0,0 +1,362 @@
+"""周汇总：从 PG daily_style_metrics 拉 7 天 (款式, 平台) 维度的销量/退款/访客/加购，求和后
+计算转化率/加购率/退款率，写入 data/<几月第N周>/<款式>/weekly_summary.{json,md}。
+
+京东自营供应链库存出库件数仍由 jd_self_inventory_sales_collector 提供，合并到 jd 行的销量里。
+
+默认 end-date 是「上周日」(周一跑周报时，统计完整的上周一~周日)；
+可用 --end-date 指定任意一天，会回溯 6 天共 7 天。
+
+用法:
+    .venv/Scripts/python.exe weekly_aggregate.py
+    .venv/Scripts/python.exe weekly_aggregate.py --end-date 2026-06-07
+    .venv/Scripts/python.exe weekly_aggregate.py --end-date 2026-06-07 --dry-run
+"""
+
+import argparse
+import json
+from collections import defaultdict
+from datetime import date, datetime, timedelta
+from pathlib import Path
+
+from erp_metric_overrides import DATA_ROOT, format_percent, parse_number
+from runtime_paths import CURATED_DATA_ROOT, EXPORTS_DATA_ROOT
+
+
+PLATFORM_DIR_TO_NAME = {
+    "jd单品数据": "京东", "dy单品数据": "抖音", "tm单品数据": "天猫"
+}
+PLATFORM_KEYS = ["jd", "dy", "tm"]
+
+# 京东自营供应链库存明细的销量数据 (jd_self_inventory_sales_collector 产物)
+JD_SELF_DIR = DATA_ROOT / "jd自营库存销量数据"
+JD_SELF_PLATFORM = "jd"  # 合并时归属到京东 POP 这一行
+
+
+def read_json(path: Path) -> dict:
+    return json.loads(path.read_text(encoding="utf-8-sig"))
+
+
+def safe_name(value: str) -> str:
+    return "".join("_" if ch in '\\/:*?"<>|' else ch for ch in str(value)).strip(" ._") or "未命名"
+
+
+def parse_date(value: str) -> date:
+    return datetime.strptime(value, "%Y-%m-%d").date()
+
+
+def week_dates(end_date: date) -> list[date]:
+    """返回 [周一, ..., 周日] 共 7 天（end_date 是周日）。"""
+    return [end_date - timedelta(days=offset) for offset in range(6, -1, -1)]
+
+
+def last_week_sunday(today: date) -> date:
+    """返回上周日的日期。周一 on_week_start=True 时使用。"""
+    return today - timedelta(days=today.weekday() + 1)
+
+
+def month_week_label(target: date) -> str:
+    """把日期转成 'M月第N周' 标签（N=1..4，按当月 1~7 / 8~14 / 15~21 / 22+ 分）。"""
+    week = min(((target.day - 1) // 7) + 1, 4)
+    return f"{target.month}月第{week}周"
+
+
+def month_week_label_anchored(target: date) -> str:
+    """Mon-anchored 周标签: W1 = 包含该月第 1 个周一的 Mon-Sun 7 天。
+
+    与 month_week_label (按 day-of-month 切) 不同: 本函数按 ISO 周对齐,
+    5.4 落在 5 月第 1 周、5.11 第 2 周、5.18 第 3 周、5.25 第 4 周 (假设 5.4 是周一)。
+
+    例 2026 年 5 月: 第1周=5.4~5.10, 第2周=5.11~5.17, 第3周=5.18~5.24, 第4周=5.25~5.31
+
+    跨月周 (例如 6.29-7.5, target_monday=6.29 在 6 月, end_date=7.5 在 7 月):
+    按 target_monday 所在月份算 label, 即 "6月第5周"。这样 6.29-7.5 (6月第5周) 不会
+    跟 7.6-7.12 (7月第1周) 撞同 label。
+    """
+    target_monday = target - timedelta(days=target.weekday())
+    first_day = date(target_monday.year, target_monday.month, 1)
+    first_monday = first_day
+    while first_monday.weekday() != 0:
+        first_monday += timedelta(days=1)
+    week_index = ((target_monday - first_monday).days // 7) + 1
+    return f"{target_monday.month}月第{week_index}周"
+
+
+def load_week_from_pg(end_date: date) -> tuple[dict, list[str]]:
+    """从 PG daily_style_metrics 拉 [start, end] 7 天的 (platform, style) 汇总。
+
+    返回 (buckets, missing_dates):
+        buckets: {(plat, style): {sales, refund_orders, visitors, cart_users, erp_style_codes, present_days}}
+        missing_dates: 7 天里有哪几天 PG 完全没数据
+    """
+    from db import fetch_daily_for_range, get_conn
+    start_date = end_date - timedelta(days=6)
+    days = week_dates(end_date)
+    with get_conn() as conn:
+        agg = fetch_daily_for_range(conn, start_date.isoformat(), end_date.isoformat())
+    buckets: dict[tuple[str, str], dict] = {}
+    for (plat, style), data in agg.items():
+        buckets[(style, plat)] = {
+            "style_name": style,
+            "platform": plat,
+            "sales": data["sales"],
+            "refund_orders": data["refund_orders"],
+            "visitors": data["visitors"],
+            "cart_users": data["cart_users"],
+            "erp_style_codes": data["erp_style_codes"],
+            "present_days": data["present_days"],
+        }
+    # 找出 7 天里 PG 完全没数据的日期 (按 (plat, style) 全检, 但实用上只看 metric_date 维度即可)
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                SELECT DISTINCT metric_date FROM daily_style_metrics
+                WHERE metric_date BETWEEN %s AND %s
+                """,
+                (start_date.isoformat(), end_date.isoformat()),
+            )
+            present_dates = {r[0].isoformat() for r in cur.fetchall()}
+    missing = [d.isoformat() for d in days if d.isoformat() not in present_dates]
+    return buckets, missing
+
+
+def load_jd_self_sales_data(end_date: date) -> dict[str, int]:
+    """读 jd_self_inventory_sales_collector 产出的「上周京东自营出库商品件数」。
+
+    落点: data/jd自营库存销量数据/<start>_<end>/jd_self_inventory_sales_summary.json
+    返回 {style_name: self_sales}；缺失则返回 {}。
+    """
+    start_date = end_date - timedelta(days=6)
+    label = f"{start_date.isoformat()}_{end_date.isoformat()}"
+    summary_path = JD_SELF_DIR / label / "jd_self_inventory_sales_summary.json"
+    if not summary_path.exists():
+        return {}
+    try:
+        payload = read_json(summary_path)
+    except Exception as e:
+        print(f"[WARN] 解析 {summary_path} 失败: {e}")
+        return {}
+    result: dict[str, int] = {}
+    for item in payload.get("results") or []:
+        name = item.get("style_name")
+        sales = int(parse_number(item.get("sales")))
+        if name:
+            result[name] = result.get(name, 0) + sales
+    return result
+
+
+def aggregate_week(end_date: date) -> tuple[dict, list[dict]]:
+    """从 PG daily_style_metrics 拉最近 7 天数据，合并京东自营出库件数，生成 weekly_records。"""
+    start_date = end_date - timedelta(days=6)
+
+    buckets, missing_dates = load_week_from_pg(end_date)
+
+    # 京东自营出库商品件数 (jd_self_inventory_sales_collector 产物)，仅京东款式参与追加销量
+    jd_self_sales_by_style = load_jd_self_sales_data(end_date)
+
+    weekly_records = []
+    for (style_name, platform), b in sorted(buckets.items()):
+        pop_sales = b["sales"]
+        refund_orders = b["refund_orders"]
+        erp_codes = b["erp_style_codes"]
+
+        # jd_self 是京东自营供应链出库件数 (出 POP 店, 但计入"京东渠道总销量")
+        # total_sales 写飞书「销量」列, 让业务方看到 POP + 自营的真实规模
+        # 但转化率/退款率仍以 pop_sales 为分母, 因为自营出库没有访客/加购数据,
+        # 强行加进去会让比率被稀释失真
+        jd_self = jd_self_sales_by_style.get(style_name, 0) if platform == "jd" else 0
+        total_sales = pop_sales + jd_self
+
+        weekly_records.append({
+            "style_name": style_name,
+            "platform": platform,
+            "platform_name": PLATFORM_DIR_TO_NAME.get(f"{platform}单品数据", platform),
+            "sales": total_sales,
+            "jd_self_sales": jd_self if platform == "jd" else 0,
+            "refund_orders": refund_orders,
+            "visitors": b["visitors"],
+            "cart_users": b["cart_users"],
+            # 注意：转化率 / 退款率 用 pop_sales，不加自营
+            "conversion_rate": format_percent(pop_sales, b["visitors"]),
+            "cart_rate": format_percent(b["cart_users"], b["visitors"]),
+            "refund_rate": format_percent(refund_orders, pop_sales),
+            "erp_style_codes": erp_codes,
+            "erp_source": "daily_pg",
+            "present_days": b["present_days"],
+        })
+
+    meta = {
+        "period": "weekly",
+        "start_date": start_date.isoformat(),
+        "end_date": end_date.isoformat(),
+        "time_period": month_week_label_anchored(end_date),
+        "date_count_expected": 7,
+        "missing_dates": missing_dates,
+        "erp_source": "daily_pg",
+        "formula": {
+            "转化率": "7天POP店销量 / 7天总访客数 * 100%（京东款式的销量含自营店出库，但分母用POP销量）",
+            "加购率": "7天总加购人数 / 7天总访客数 * 100%",
+            "退款率": "7天总退款订单数 / 7天POP店销量 * 100%（京东款式的销量含自营店出库，但分母用POP销量）",
+        },
+        "generated_at": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+    }
+    return meta, weekly_records
+
+
+def write_weekly_outputs(meta: dict, weekly_records: list[dict]) -> list[Path]:
+    """把 weekly_records 按款式拆目录写到 data/<M月第N周>/<款式>/weekly_summary.{json,md}。
+
+    同时把汇总的 bitable_records.json 写到 data/ready_for_bitable/<start>_<end>/bitable_records.json
+    让 insert_bitable_records.py --record-file 复用。
+    """
+    time_period = meta["time_period"]
+    period_root = CURATED_DATA_ROOT / time_period
+    period_root.mkdir(parents=True, exist_ok=True)
+
+    written = []
+
+    # 按款式聚合
+    by_style: dict[str, list[dict]] = defaultdict(list)
+    for r in weekly_records:
+        by_style[r["style_name"]].append(r)
+
+    for style_name, records in by_style.items():
+        style_dir = period_root / safe_name(style_name)
+        style_dir.mkdir(parents=True, exist_ok=True)
+
+        # 按平台排序
+        records.sort(key=lambda r: r["platform"])
+        json_payload = {
+            **meta,
+            "style_name": style_name,
+            "platforms": records,
+        }
+        json_path = style_dir / "weekly_summary.json"
+        json_path.write_text(
+            json.dumps(json_payload, ensure_ascii=False, indent=2),
+            encoding="utf-8-sig",
+        )
+        written.append(json_path)
+
+        # markdown
+        md_lines = [
+            f"# {style_name} - 周汇总 ({meta['start_date']} ~ {meta['end_date']})",
+            "",
+            f"时间周期: {time_period}",
+            f"生成时间: {meta['generated_at']}",
+            "",
+            "| 平台 | 7天销量 | 7天访客 | 7天加购 | 7天退款 | 转化率 | 加购率 | 退款率 | 出现天数 |",
+            "|---|---:|---:|---:|---:|---:|---:|---:|---:|",
+        ]
+        for r in records:
+            sales_display = r["sales"]
+            if r.get("jd_self_sales"):
+                sales_display = f"{r['sales']} (含自营{r['jd_self_sales']})"
+            md_lines.append(
+                f"| {r['platform_name']} | {sales_display} | {r['visitors']} | "
+                f"{r['cart_users']} | {r['refund_orders']} | {r['conversion_rate']} | "
+                f"{r['cart_rate']} | {r['refund_rate']} | {r['present_days']}/7 |"
+            )
+        md_path = style_dir / "weekly_summary.md"
+        md_path.write_text("\n".join(md_lines), encoding="utf-8-sig")
+        written.append(md_path)
+
+    # bitable_records.json（给 insert_bitable_records.py --record-file 复用）
+    bitable_dir = EXPORTS_DATA_ROOT / f"{meta['start_date']}_{meta['end_date']}"
+    bitable_dir.mkdir(parents=True, exist_ok=True)
+    bitable_records = []
+    for r in weekly_records:
+        bitable_records.append({
+            "数据日期": meta["end_date"],
+            "周开始": meta["start_date"],
+            "周结束": meta["end_date"],
+            "时间周期": time_period,
+            "平台": r["platform_name"],
+            "平台代码": r["platform"],
+            "款式": r["style_name"],
+            "ERP款式编码": ",".join(r["erp_style_codes"]),
+            "销量": r["sales"],
+            "访客数": r["visitors"],
+            "加购人数": r["cart_users"],
+            "退款订单数": r["refund_orders"],
+            "转化率": r["conversion_rate"],
+            "加购率": r["cart_rate"],
+            "退款率": r["refund_rate"],
+            "出现天数": r["present_days"],
+        })
+    bitable_path = bitable_dir / "bitable_records.json"
+    bitable_path.write_text(
+        json.dumps(bitable_records, ensure_ascii=False, indent=2),
+        encoding="utf-8-sig",
+    )
+    written.append(bitable_path)
+    return written, bitable_path
+
+
+def main():
+    parser = argparse.ArgumentParser(description="把最近 7 天的款式汇总累加成周报")
+    parser.add_argument("--end-date", type=parse_date, default=None,
+                        help="周结束日（周日）。默认上周日")
+    parser.add_argument("--last-week", action="store_true", default=True,
+                        help="默认用上周（周一~周日）。--this-week 取本周日（可能未来）")
+    parser.add_argument("--this-week", dest="last_week", action="store_false")
+    parser.add_argument("--dry-run", action="store_true")
+    args = parser.parse_args()
+
+    if args.end_date is None:
+        today = date.today()
+        ref = today - timedelta(days=7) if args.last_week else today
+        end_date = last_week_sunday(ref) if args.last_week else ref
+    else:
+        end_date = args.end_date
+
+    print(f"[INFO] 周区间: {end_date - timedelta(days=6)} ~ {end_date}  ({end_date.strftime('%A')})")
+    meta, weekly_records = aggregate_week(end_date)
+    print(f"[INFO] 缺数据日期: {meta['missing_dates'] or '无'}")
+    print(f"[INFO] 款式×平台: {len(weekly_records)}")
+
+    if args.dry_run:
+        print("[INFO] DRY-RUN 模式，不写文件")
+        return
+
+    written, bitable_path = write_weekly_outputs(meta, weekly_records)
+    print(f"[OK] 时间周期: {meta['time_period']}")
+    print(f"[OK] 周报目录: {CURATED_DATA_ROOT / meta['time_period']}")
+    print(f"[OK] bitable_records.json: {bitable_path}")
+    for p in written:
+        print(f"[OK] {p}")
+
+    # 写库（库不可用时 WARN，不阻塞主流程）
+    try:
+        from db import get_conn, upsert_weekly_metrics
+        with get_conn() as conn:
+            rows = [
+                {
+                    "week_end": meta["end_date"],
+                    "week_start": meta["start_date"],
+                    "time_period": meta["time_period"],
+                    "style_name": r["style_name"],
+                    "platform": r["platform"],
+                    "sales": r["sales"],
+                    "pop_sales": r["sales"] - r.get("jd_self_sales", 0),
+                    "jd_self_sales": r.get("jd_self_sales", 0),
+                    "refund_orders": r["refund_orders"],
+                    "visitors": r["visitors"],
+                    "cart_users": r["cart_users"],
+                    "conversion_rate": r["conversion_rate"],
+                    "cart_rate": r["cart_rate"],
+                    "refund_rate": r["refund_rate"],
+                    "present_days": r["present_days"],
+                    "erp_style_codes": r["erp_style_codes"],
+                    "erp_source": r["erp_source"],
+                }
+                for r in weekly_records
+            ]
+            n = upsert_weekly_metrics(conn, rows)
+            print(f"[PG] weekly_style_metrics upsert: {n} 行")
+    except Exception as exc:
+        print(f"[WARN] PG upsert 失败 (库未配/不可达)，主流程不受影响: {exc}")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/registry.py b/src/gyxx_flow/modules/registry.py
new file mode 100644
index 0000000..21ebecd
--- /dev/null
+++ b/src/gyxx_flow/modules/registry.py
@@ -0,0 +1,69 @@
+"""Validated registry for composing business modules at the application edge."""
+
+from __future__ import annotations
+
+import re
+from types import MappingProxyType
+from typing import Iterable, Mapping
+
+from gyxx_flow.workflow import WorkflowDefinition
+
+from .contracts import BusinessModule
+
+_MODULE_ID = re.compile(r"^[a-z][a-z0-9]*(?:_[a-z0-9]+)*$")
+
+
+class ModuleRegistry:
+    """Immutable index of modules and their workflow definitions."""
+
+    def __init__(self, modules: Iterable[BusinessModule]) -> None:
+        module_index: dict[str, BusinessModule] = {}
+        workflow_index: dict[str, WorkflowDefinition] = {}
+
+        for module in modules:
+            if not isinstance(module, BusinessModule):
+                raise TypeError("module does not implement BusinessModule")
+            module_id = module.module_id
+            if not isinstance(module_id, str) or not _MODULE_ID.fullmatch(module_id):
+                raise ValueError(f"invalid module id: {module_id!r}")
+            if module_id in module_index:
+                raise ValueError(f"duplicate module id: {module_id}")
+            module_index[module_id] = module
+
+            definitions = module.workflow_definitions()
+            if not isinstance(definitions, tuple):
+                raise TypeError(
+                    f"module {module_id!r} must return a tuple of workflow definitions"
+                )
+            for definition in definitions:
+                if not isinstance(definition, WorkflowDefinition):
+                    raise TypeError(
+                        f"module {module_id!r} returned an invalid workflow definition"
+                    )
+                if definition.workflow_id in workflow_index:
+                    raise ValueError(
+                        f"duplicate workflow id: {definition.workflow_id}"
+                    )
+                workflow_index[definition.workflow_id] = definition
+
+        self._modules: Mapping[str, BusinessModule] = MappingProxyType(module_index)
+        self._workflows: Mapping[str, WorkflowDefinition] = MappingProxyType(
+            workflow_index
+        )
+
+    @property
+    def module_ids(self) -> tuple[str, ...]:
+        return tuple(sorted(self._modules))
+
+    @property
+    def workflow_ids(self) -> tuple[str, ...]:
+        return tuple(sorted(self._workflows))
+
+    def get(self, module_id: str) -> BusinessModule:
+        return self._modules[module_id]
+
+    def workflow(self, workflow_id: str) -> WorkflowDefinition:
+        return self._workflows[workflow_id]
+
+
+__all__ = ["ModuleRegistry"]
diff --git a/src/gyxx_flow/modules/shop_intelligence/__init__.py b/src/gyxx_flow/modules/shop_intelligence/__init__.py
new file mode 100644
index 0000000..ab52ec8
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/__init__.py
@@ -0,0 +1,82 @@
+"""Shop and competitor intelligence workflows."""
+
+from __future__ import annotations
+
+from collections.abc import Iterable
+from typing import Any
+
+from gyxx_flow.adapters.native import DeferredModuleCommandStep, ModuleCommandFactory
+from gyxx_flow.workflow import StepDefinition, WorkflowDefinition
+
+SHOP_WORKFLOW_IDS = (
+    "shop.metrics.weekly",
+    "shop.competitor.weekly",
+)
+COLLECTION_TIMEOUT_SECONDS = 4 * 60 * 60
+SHOP_RESOURCE = "module:shop_intelligence"
+
+
+class ShopIntelligenceModule:
+    """Shop-intelligence business module composition boundary."""
+
+    module_id = "shop_intelligence"
+
+    def __init__(self, definitions: Iterable[WorkflowDefinition] = ()) -> None:
+        self._definitions = tuple(definitions)
+
+    @classmethod
+    def from_catalog(
+        cls,
+        catalog: Any,
+        *,
+        command_factory: ModuleCommandFactory | None = None,
+    ) -> "ShopIntelligenceModule":
+        """Build only the two scheduled shop definitions from validated metadata."""
+
+        entries = {
+            entry.workflow_id: entry
+            for entry in catalog.workflows
+            if entry.module == cls.module_id and entry.trigger == "scheduled"
+        }
+        expected = set(SHOP_WORKFLOW_IDS)
+        if set(entries) != expected:
+            raise ValueError(
+                "shop catalog workflows do not match the migration contract: "
+                f"expected={sorted(expected)}, actual={sorted(entries)}"
+            )
+
+        definitions: list[WorkflowDefinition] = []
+        for workflow_id in SHOP_WORKFLOW_IDS:
+            entry = entries[workflow_id]
+            action = (
+                DeferredModuleCommandStep(entry)
+                if command_factory is None
+                else DeferredModuleCommandStep(entry, command_factory=command_factory)
+            )
+            definitions.append(
+                WorkflowDefinition(
+                    workflow_id,
+                    (
+                        StepDefinition(
+                            "module_collect",
+                            action,
+                            timeout_seconds=COLLECTION_TIMEOUT_SECONDS,
+                            max_attempts=1,
+                            resources=(SHOP_RESOURCE,),
+                            production_sink=True,
+                        ),
+                    ),
+                )
+            )
+        return cls(definitions)
+
+    def workflow_definitions(self) -> tuple[WorkflowDefinition, ...]:
+        return self._definitions
+
+
+__all__ = [
+    "COLLECTION_TIMEOUT_SECONDS",
+    "SHOP_RESOURCE",
+    "SHOP_WORKFLOW_IDS",
+    "ShopIntelligenceModule",
+]
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/__init__.py b/src/gyxx_flow/modules/shop_intelligence/runtime/__init__.py
new file mode 100644
index 0000000..39fd8ee
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/__init__.py
@@ -0,0 +1,2 @@
+"""Native shop-intelligence runtime migrated from the former standalone project."""
+
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/adaptive_selectors.py b/src/gyxx_flow/modules/shop_intelligence/runtime/adaptive_selectors.py
new file mode 100644
index 0000000..70a8bb1
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/adaptive_selectors.py
@@ -0,0 +1,312 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Adaptive selector helpers for Scrapling response pages and Playwright pages.
+
+Scrapling's native adaptive selectors work on Scrapling Response objects. The
+current scraping flow also uses Playwright pages inside page_action, so this
+module adds a small fingerprint fallback for text clicks and inputs there.
+"""
+
+import json
+import os
+import re
+from typing import Any
+
+from scrapling import DynamicFetcher, Fetcher
+
+from gyxx_flow.modules.shop_intelligence.runtime.paths import state_path
+
+try:
+    from scrapling import StealthyFetcher
+except ModuleNotFoundError:
+    StealthyFetcher = None
+
+
+ADAPTIVE_SELECTOR_DIR = str(state_path("adaptive_selectors"))
+PLAYWRIGHT_SELECTOR_STORE = os.path.join(ADAPTIVE_SELECTOR_DIR, "playwright_selectors.json")
+
+
+def enable_adaptive_fetchers() -> None:
+    Fetcher.adaptive = True
+    DynamicFetcher.adaptive = True
+    if StealthyFetcher is not None:
+        StealthyFetcher.adaptive = True
+
+
+def adaptive_css(page: Any, selector: str, identifier: str = "", percentage: int = 40):
+    """Use Scrapling native adaptive CSS and save successful element features."""
+    identifier = identifier or selector
+    try:
+        result = page.css(selector, identifier=identifier, adaptive=True, percentage=percentage)
+        if result:
+            return result
+    except Exception:
+        pass
+    return page.css(selector, identifier=identifier, auto_save=True, percentage=percentage)
+
+
+def adaptive_xpath(page: Any, selector: str, identifier: str = "", percentage: int = 40, **kwargs):
+    """Use Scrapling native adaptive XPath and save successful element features."""
+    identifier = identifier or selector
+    try:
+        result = page.xpath(selector, identifier=identifier, adaptive=True, percentage=percentage, **kwargs)
+        if result:
+            return result
+    except Exception:
+        pass
+    return page.xpath(selector, identifier=identifier, auto_save=True, percentage=percentage, **kwargs)
+
+
+def _load_store() -> dict:
+    try:
+        with open(PLAYWRIGHT_SELECTOR_STORE, "r", encoding="utf-8") as f:
+            return json.load(f)
+    except Exception:
+        return {}
+
+
+def _save_store(store: dict) -> None:
+    os.makedirs(ADAPTIVE_SELECTOR_DIR, exist_ok=True)
+    with open(PLAYWRIGHT_SELECTOR_STORE, "w", encoding="utf-8") as f:
+        json.dump(store, f, ensure_ascii=False, indent=2)
+
+
+def _key(namespace: str, identifier: str) -> str:
+    return f"{namespace}:{identifier}"
+
+
+def _safe_identifier(value: str) -> str:
+    return re.sub(r"\s+", " ", str(value or "")).strip()
+
+
+def _get_fingerprint(page, selector: str, text: str = "", exact: bool = True) -> dict | None:
+    try:
+        return page.evaluate(
+            """([selector, text, exact]) => {
+                const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const classTokens = (el) => String(el.className || '')
+                    .split(/\\s+/)
+                    .filter(Boolean)
+                    .filter(token => token.length > 2)
+                    .slice(0, 12);
+                const candidates = Array.from(document.querySelectorAll(selector))
+                    .filter(isVisible)
+                    .filter(el => {
+                        if (!text) return true;
+                        const value = clean(el.innerText || el.textContent || el.value || el.placeholder);
+                        return exact ? value === text : value.includes(text);
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return null;
+                const el = candidates[0];
+                const rect = el.getBoundingClientRect();
+                return {
+                    tag: el.tagName.toLowerCase(),
+                    text: clean(el.innerText || el.textContent || ''),
+                    value: clean(el.value || ''),
+                    placeholder: clean(el.getAttribute('placeholder') || ''),
+                    aria: clean(el.getAttribute('aria-label') || ''),
+                    title: clean(el.getAttribute('title') || ''),
+                    role: clean(el.getAttribute('role') || ''),
+                    type: clean(el.getAttribute('type') || ''),
+                    class_tokens: classTokens(el),
+                    width: Math.round(rect.width),
+                    height: Math.round(rect.height),
+                };
+            }""",
+            [selector, text, exact],
+        )
+    except Exception:
+        return None
+
+
+def _find_by_fingerprint(page, selector: str, fingerprint: dict) -> dict | None:
+    if not fingerprint:
+        return None
+    try:
+        return page.evaluate(
+            """([selector, fp]) => {
+                const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const classTokens = (el) => String(el.className || '').split(/\\s+/).filter(Boolean);
+                const scoreText = (a, b) => {
+                    a = clean(a); b = clean(b);
+                    if (!a || !b) return 0;
+                    if (a === b) return 40;
+                    if (a.includes(b) || b.includes(a)) return 24;
+                    return 0;
+                };
+                const candidates = Array.from(document.querySelectorAll(selector))
+                    .filter(isVisible)
+                    .map(el => {
+                        const rect = el.getBoundingClientRect();
+                        const tokens = classTokens(el);
+                        let score = 0;
+                        if (el.tagName.toLowerCase() === fp.tag) score += 12;
+                        if (clean(el.getAttribute('role') || '') === fp.role && fp.role) score += 10;
+                        if (clean(el.getAttribute('type') || '') === fp.type && fp.type) score += 8;
+                        score += scoreText(el.innerText || el.textContent || '', fp.text);
+                        score += scoreText(el.value || '', fp.value);
+                        score += scoreText(el.getAttribute('placeholder') || '', fp.placeholder);
+                        score += scoreText(el.getAttribute('aria-label') || '', fp.aria);
+                        score += scoreText(el.getAttribute('title') || '', fp.title);
+                        for (const token of (fp.class_tokens || [])) {
+                            if (tokens.includes(token)) score += 3;
+                        }
+                        const sizeGap = Math.abs(Math.round(rect.width) - (fp.width || 0))
+                            + Math.abs(Math.round(rect.height) - (fp.height || 0));
+                        score -= Math.min(20, sizeGap / 80);
+                        return { el, score };
+                    })
+                    .filter(item => item.score >= 25)
+                    .sort((a, b) => b.score - a.score);
+                if (!candidates.length) return null;
+                const rect = candidates[0].el.getBoundingClientRect();
+                return {
+                    x: rect.left + rect.width / 2,
+                    y: rect.top + rect.height / 2,
+                    score: candidates[0].score,
+                };
+            }""",
+            [selector, fingerprint],
+        )
+    except Exception:
+        return None
+
+
+def remember_playwright_selector(page, namespace: str, identifier: str, selector: str, text: str = "", exact: bool = True) -> None:
+    fingerprint = _get_fingerprint(page, selector, text=text, exact=exact)
+    if not fingerprint:
+        return
+    store = _load_store()
+    store[_key(namespace, identifier)] = {
+        "selector": selector,
+        "text": text,
+        "exact": exact,
+        "fingerprint": fingerprint,
+    }
+    _save_store(store)
+
+
+def remember_active_playwright_element(page, namespace: str, identifier: str, selector: str) -> None:
+    try:
+        fingerprint = page.evaluate(
+            """(selector) => {
+                const el = document.activeElement;
+                if (!el || !el.matches(selector)) return null;
+                const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const rect = el.getBoundingClientRect();
+                return {
+                    tag: el.tagName.toLowerCase(),
+                    text: clean(el.innerText || el.textContent || ''),
+                    value: clean(el.value || ''),
+                    placeholder: clean(el.getAttribute('placeholder') || ''),
+                    aria: clean(el.getAttribute('aria-label') || ''),
+                    title: clean(el.getAttribute('title') || ''),
+                    role: clean(el.getAttribute('role') || ''),
+                    type: clean(el.getAttribute('type') || ''),
+                    class_tokens: String(el.className || '').split(/\\s+/).filter(Boolean).slice(0, 12),
+                    width: Math.round(rect.width),
+                    height: Math.round(rect.height),
+                };
+            }""",
+            selector,
+        )
+        if not fingerprint:
+            return
+        store = _load_store()
+        store[_key(namespace, identifier)] = {
+            "selector": selector,
+            "text": "",
+            "exact": True,
+            "fingerprint": fingerprint,
+        }
+        _save_store(store)
+    except Exception:
+        return
+
+
+def adaptive_click_text(page, text: str, exact: bool = True, timeout: int = 10000) -> bool:
+    selector = "button, [role='button'], a, label, span, div, li"
+    identifier = _safe_identifier(f"{text}|exact={exact}")
+    try:
+        locator = page.get_by_text(text, exact=exact)
+        remember_playwright_selector(page, "text", identifier, selector, text=text, exact=exact)
+        locator.first.click(timeout=timeout)
+        return True
+    except Exception:
+        pass
+
+    store = _load_store()
+    entry = store.get(_key("text", identifier))
+    match = _find_by_fingerprint(page, entry.get("selector", selector), entry.get("fingerprint", {})) if entry else None
+    if match:
+        try:
+            page.mouse.click(match["x"], match["y"])
+            return True
+        except Exception:
+            pass
+    return False
+
+
+def adaptive_hover_text(page, text: str, exact: bool = True, timeout: int = 10000) -> bool:
+    selector = "button, [role='button'], a, label, span, div, li"
+    identifier = _safe_identifier(f"{text}|exact={exact}")
+    try:
+        locator = page.get_by_text(text, exact=exact)
+        remember_playwright_selector(page, "text", identifier, selector, text=text, exact=exact)
+        locator.first.hover(timeout=timeout)
+        return True
+    except Exception:
+        pass
+
+    store = _load_store()
+    entry = store.get(_key("text", identifier))
+    match = _find_by_fingerprint(page, entry.get("selector", selector), entry.get("fingerprint", {})) if entry else None
+    if match:
+        try:
+            page.mouse.move(match["x"], match["y"])
+            return True
+        except Exception:
+            pass
+    return False
+
+
+def adaptive_fill_input(page, identifier: str, selector: str, value: str, timeout: int = 8000) -> bool:
+    try:
+        locator = page.locator(selector).first
+        locator.click(timeout=timeout)
+        remember_active_playwright_element(page, "input", identifier, selector)
+        locator.fill(value, timeout=timeout)
+        return True
+    except Exception:
+        pass
+
+    store = _load_store()
+    entry = store.get(_key("input", identifier))
+    match = _find_by_fingerprint(page, entry.get("selector", selector), entry.get("fingerprint", {})) if entry else None
+    if not match:
+        return False
+    try:
+        page.mouse.click(match["x"], match["y"])
+        page.keyboard.press("Control+A")
+        page.keyboard.type(value, delay=20)
+        return True
+    except Exception:
+        return False
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/__init__.py b/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/dy_store_competitor_store_scraping.py b/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/dy_store_competitor_store_scraping.py
new file mode 100644
index 0000000..ba49918
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/dy_store_competitor_store_scraping.py
@@ -0,0 +1,2501 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+抖音罗盘经营数据抓取脚本 (基于 scrapling.DynamicFetcher)
+"""
+
+import json
+import os
+import re
+import sys
+import time
+from datetime import datetime, timedelta
+
+from scrapling import DynamicFetcher
+
+from gyxx_flow.modules.shop_intelligence.runtime.adaptive_selectors import (
+    adaptive_click_text,
+    adaptive_fill_input,
+    adaptive_hover_text,
+    enable_adaptive_fetchers,
+)
+from gyxx_flow.modules.shop_intelligence.runtime.paths import raw_path, state_path
+
+try:
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")
+    sys.stderr.reconfigure(encoding="utf-8", errors="replace")
+except Exception:
+    pass
+
+OUTPUT_DIR = os.path.dirname(os.path.abspath(__file__))
+DATA_DIR = str(raw_path("抖音"))
+STORE_DATA_DIR = os.path.join(DATA_DIR, "店铺数据")
+COMPETITOR_DATA_DIR = os.path.join(DATA_DIR, "竞店数据")
+COOKIES_FILE = os.getenv(
+    "GYXX_BROWSER_COOKIE_FILE", str(state_path("douyin", "compass_cookies.json"))
+)
+STORAGE_STATE_FILE = os.getenv(
+    "GYXX_BROWSER_STORAGE_STATE_FILE",
+    str(state_path("douyin", "compass_storage_state.json")),
+)
+USER_DATA_DIR = os.getenv(
+    "GYXX_BROWSER_PROFILE_DIR", str(state_path("douyin", "compass_profile"))
+)
+SHOP_NAME = "光影行星箱包旗舰店"
+LOGIN_TIMEOUT = 240
+KEEP_BROWSER_OPEN = os.getenv("KEEP_BROWSER_OPEN", "1") != "0"
+SKIP_EXPERIENCE_SCORE = os.getenv("SKIP_EXPERIENCE_SCORE", "0") == "1"
+SKIP_SHOP = os.getenv("SKIP_SHOP", "0") == "1"
+ACTION_SETTLE_MS = int(os.getenv("ACTION_SETTLE_MS", "1200"))
+PAGE_SETTLE_MS = int(os.getenv("PAGE_SETTLE_MS", "2200"))
+_scraped_metrics = {}
+_market_rankings = {}
+_competitor_rankings = {}
+_qianchuan_metrics = {}
+_experience_score = {}
+_results_saved = False
+MARKET_CATEGORIES = ["单肩包/斜挎包", "双肩包", "腰包/胸包"]
+COMPETITOR_BRANDS = ["Bellroy", "tomtoc", "BAGSMART", "GROTTO", "NIID", "reeyee", "VAOPER", "探迹者"]
+
+
+def _previous_week_range():
+    today = datetime.now().date()
+    this_monday = today - timedelta(days=today.weekday())
+    start = this_monday - timedelta(days=7)
+    end = this_monday - timedelta(days=1)
+    return start, end
+
+
+def _date_range_label(start, end) -> str:
+    return f"{start:%Y/%m/%d} - {end:%Y/%m/%d}"
+
+
+def _load_cookies():
+    if os.path.exists(COOKIES_FILE):
+        try:
+            with open(COOKIES_FILE, "r", encoding="utf-8") as f:
+                cookies = json.load(f)
+            print(f"[OK] Loaded {len(cookies)} cookies")
+            return cookies
+        except Exception as e:
+            print(f"[WARN] Load cookies failed: {e}")
+    return None
+
+
+def _save_cookies(cookies):
+    try:
+        with open(COOKIES_FILE, "w", encoding="utf-8") as f:
+            json.dump(cookies, f, ensure_ascii=False, indent=2)
+        print(f"[OK] Saved {len(cookies)} cookies -> {COOKIES_FILE}")
+        return True
+    except Exception as e:
+        import traceback
+        print(f"[WARN] Save cookies failed: {e}")
+        traceback.print_exc()
+        return False
+
+
+def _save_browser_state(page, stage=""):
+    label = f" ({stage})" if stage else ""
+    try:
+        cookies = page.context.cookies()
+        if cookies:
+            _save_cookies(cookies)
+        else:
+            print(f"[WARN] No cookies found to save{label}")
+
+        page.context.storage_state(path=STORAGE_STATE_FILE)
+        print(f"[OK] Saved storage state{label} -> {STORAGE_STATE_FILE}")
+    except Exception as e:
+        print(f"[WARN] Failed to save browser state{label}: {e}")
+
+
+def _active_page(page):
+    try:
+        pages = [p for p in page.context.pages if not p.is_closed()]
+        if pages:
+            return pages[-1]
+    except Exception:
+        pass
+    return page
+
+
+def _maximize_page(page) -> None:
+    try:
+        page.set_viewport_size({"width": 1920, "height": 1080})
+    except Exception as e:
+        print(f"[WARN] Set viewport size failed: {e}")
+    try:
+        page.evaluate("() => { window.moveTo(0, 0); window.resizeTo(screen.availWidth, screen.availHeight); }")
+    except Exception:
+        pass
+
+
+def _check_login_status(page) -> bool:
+    """多种方式检测是否已登录"""
+    page = _active_page(page)
+    url = page.url
+    print(f"[DEBUG] check_login: url={url}")
+
+    # 身份选择页、手机登录页都不是登录完成状态，不能只靠 URL 判断。
+    try:
+        login_markers = [
+            "请选择您的身份",
+            "手机登录",
+            "邮箱登录",
+            "手机号",
+            "验证码",
+            "发送验证码",
+            "登录即代表同意",
+        ]
+        for marker in login_markers:
+            if page.locator(f"text={marker}").count() > 0:
+                print(f"[DEBUG] Still on login/identity page: {marker}")
+                return False
+    except Exception as e:
+        print(f"[DEBUG] Login marker check error: {e}")
+
+    # 方式2: 页面中存在已登录标志元素（店铺名、退出按钮等）
+    try:
+        # 检查是否出现了店铺名称或用户中心
+        if page.locator('text=光影行星箱包旗舰店').count() > 0:
+            print("[OK] Login detected by shop name")
+            return True
+        if page.locator('text=退出登录').count() > 0:
+            print("[OK] Login detected by logout button")
+            return True
+        if page.locator('text=个人中心').count() > 0:
+            print("[OK] Login detected by user center")
+            return True
+        if page.locator('text=成交金额').count() > 0:
+            print("[OK] Login detected by business metric")
+            return True
+        if page.locator('text=成交人数').count() > 0:
+            print("[OK] Login detected by business metric")
+            return True
+    except Exception as e:
+        print(f"[DEBUG] Element check error: {e}")
+
+    if "/shop/" in url and "/login" not in url:
+        print("[OK] Login detected by shop URL")
+        return True
+
+    return False
+
+
+def _ensure_login(page):
+    print("\n" + "=" * 60)
+    print("【请登录】请在弹出的浏览器窗口中完成登录")
+    print("提示：扫码/短信/密码均可，登录成功后页面会自动跳转")
+    print("=" * 60 + "\n")
+
+    for i in range(LOGIN_TIMEOUT):
+        page = _active_page(page)
+        if _check_login_status(page):
+            print(f"[OK] Login success! URL: {page.url}")
+            page.wait_for_timeout(3000)
+            return page
+        if i % 5 == 0:
+            print(f"[WAIT] Waiting for login... {i}s / {LOGIN_TIMEOUT}s")
+        time.sleep(1)
+
+    print(f"[ERR] Login timeout ({LOGIN_TIMEOUT}s)")
+    return None
+
+
+def _select_menu_and_date(page) -> None:
+    week_start, week_end = _previous_week_range()
+    target_url = "https://compass.jinritemai.com/shop/business-part"
+    print(f"[INFO] Navigating to: {target_url}")
+    print(f"[INFO] Business date range: {_date_range_label(week_start, week_end)}")
+    page.goto(target_url, wait_until="domcontentloaded", timeout=60000)
+    try:
+        page.wait_for_load_state("load", timeout=8000)
+    except Exception as e:
+        print(f"[WARN] load wait skipped: {e}")
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+    print(f"[OK] Page loaded: {page.url}")
+
+    _select_business_previous_week(page, week_start, week_end)
+
+    print("[INFO] Waiting for data render...")
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+    try:
+        page.wait_for_load_state("load", timeout=5000)
+    except Exception:
+        pass
+    print("[OK] Data rendered")
+
+
+def _extract_metrics_via_js(page) -> dict:
+    print("[INFO] Extracting via JS...")
+    metrics = page.evaluate("""
+        () => {
+            const result = {};
+            const targets = [
+                {key: '成交金额', labels: ['成交金额']},
+                {key: '成交人数', labels: ['成交人数']},
+                {key: '成交订单数', labels: ['成交订单数']},
+                {key: '客单价', labels: ['客单价']},
+                {key: '成交退款金额', labels: ['成交退款金额']},
+                {key: '商品点击人数', labels: ['商品点击人数']},
+                {key: '访客数', labels: ['访客数', '店铺访客数']}
+            ];
+            const lines = (document.body.innerText || '')
+                .split('\\n')
+                .map(s => s.trim())
+                .filter(Boolean);
+            const valueRe = /^(¥\\s*)?[\\d,]+(\\.\\d+)?%?$/;
+            const unitRe = /^[万亿]$/;
+
+            function firstValueAfter(index) {
+                for (let i = index + 1; i < Math.min(lines.length, index + 8); i++) {
+                    if (valueRe.test(lines[i])) {
+                        const value = lines[i].replace(/\\s+/g, '');
+                        if (unitRe.test(lines[i + 1] || '')) return value + lines[i + 1];
+                        return value;
+                    }
+                }
+                return null;
+            }
+
+            for (const target of targets) {
+                for (let i = 0; i < lines.length; i++) {
+                    if (target.labels.some(label => lines[i].includes(label))) {
+                        const value = firstValueAfter(i);
+                        if (value) {
+                            result[target.key] = value;
+                            break;
+                        }
+                    }
+                }
+            }
+            return result;
+        }
+    """)
+    cleaned = {k: v for k, v in metrics.items() if v}
+    print(f"[OK] JS extract: {cleaned}")
+    return metrics
+
+
+def _extract_metrics_via_dom_fallback(page) -> dict:
+    print("[INFO] Fallback DOM extract...")
+    metrics = page.evaluate("""
+        () => {
+            const result = {};
+            const targets = ['成交金额', '成交人数', '成交订单数', '客单价', '成交退款金额', '商品点击人数', '访客数', '店铺访客数'];
+            const keyMap = {'店铺访客数': '访客数'};
+            const allEls = document.querySelectorAll('div, span, p, label');
+            allEls.forEach(el => {
+                const text = el.textContent.trim();
+                targets.forEach(keyword => {
+                    if (text.includes(keyword)) {
+                        const parent = el.closest('[class*="card"], [class*="item"], [class*="metric"], [class*="cell"], [class*="row"]') || el.parentElement;
+                        if (parent) {
+                            const nums = parent.querySelectorAll('span, div, p');
+                            for (const n of nums) {
+                                const nt = n.textContent.trim();
+                                if (/^(¥|￥)?[\\d,.]+(\\.\\d+)?%?$/.test(nt) && nt !== text) {
+                                    const key = keyMap[keyword] || keyword;
+                                    if (!result[key]) result[key] = nt;
+                                    break;
+                                }
+                            }
+                        }
+                    }
+                });
+            });
+            return result;
+        }
+    """)
+    cleaned = {k: v for k, v in metrics.items() if v}
+    print(f"[OK] Fallback extract: {cleaned}")
+    return metrics
+
+
+def _wait_after_action(page, label="", timeout=2500, settle_ms=None, wait_network=False) -> None:
+    if wait_network:
+        try:
+            page.wait_for_load_state("networkidle", timeout=timeout)
+        except Exception as e:
+            if label:
+                print(f"[WARN] networkidle wait skipped after {label}: {e}")
+    page.wait_for_timeout(ACTION_SETTLE_MS if settle_ms is None else settle_ms)
+
+
+def _click_calendar_date(page, target_date) -> bool:
+    date_dash = target_date.strftime("%Y-%m-%d")
+    date_slash = target_date.strftime("%Y/%m/%d")
+    day_text = str(target_date.day)
+    print(f"[INFO] Clicking calendar date: {date_dash}")
+    target_month_key = target_date.year * 12 + target_date.month
+
+    def try_click_visible_date() -> bool:
+        try:
+            clicked = page.evaluate(
+                """([dateDash, dateSlash, dayText, targetMonthKey]) => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const inDatePicker = (el) => {
+                    let cur = el;
+                    for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                        const cls = String(cur.className || '').toLowerCase();
+                        const role = (cur.getAttribute && cur.getAttribute('role') || '').toLowerCase();
+                        if (cls.includes('picker') || cls.includes('calendar') || cls.includes('date')
+                            || role.includes('dialog') || role.includes('grid')) {
+                            return true;
+                        }
+                    }
+                    return false;
+                };
+                const disabled = (el) => {
+                    const text = [el.className, el.getAttribute('aria-disabled'), el.getAttribute('disabled')]
+                        .join(' ').toLowerCase();
+                    return text.includes('disabled') || text.includes('true');
+                };
+                const attrs = ['title', 'aria-label', 'data-date', 'data-value', 'value'];
+                const dateHit = Array.from(document.querySelectorAll('td, button, div, span'))
+                    .filter(isVisible)
+                    .filter(inDatePicker)
+                    .filter(el => !disabled(el))
+                    .find(el => attrs.some(attr => {
+                        const value = el.getAttribute(attr) || '';
+                        return value.includes(dateDash) || value.includes(dateSlash);
+                    }));
+                if (dateHit) {
+                    (dateHit.closest('td, button') || dateHit).click();
+                    return true;
+                }
+
+                const visibleMonthKeys = Array.from(document.querySelectorAll('div, span, button, th'))
+                    .filter(isVisible)
+                    .map(el => (el.innerText || el.textContent || '').replace(/\\s+/g, ' ').trim())
+                    .map(text => text.match(/(20\\d{2})\\s*年\\s*(\\d{1,2})\\s*月/))
+                    .filter(Boolean)
+                    .map(match => Number(match[1]) * 12 + Number(match[2]));
+                if (visibleMonthKeys.length && !visibleMonthKeys.includes(targetMonthKey)) {
+                    const nearestGap = Math.min(...visibleMonthKeys.map(key => Math.abs(key - targetMonthKey)));
+                    // Calendars often show previous/next-month dates in gray in the visible month grid.
+                    // Allow adjacent months so a June panel can still select May 25-31 as last week.
+                    if (nearestGap > 1) {
+                        return false;
+                    }
+                }
+
+                const targetAdjacentDirection = visibleMonthKeys.includes(targetMonthKey)
+                    ? 0
+                    : (visibleMonthKeys.length && targetMonthKey < Math.min(...visibleMonthKeys) ? -1 : 1);
+                const dayCandidates = Array.from(document.querySelectorAll('td, button, div, span'))
+                    .filter(isVisible)
+                    .filter(inDatePicker)
+                    .filter(el => !disabled(el))
+                    .filter(el => (el.innerText || el.textContent || '').trim() === dayText)
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        const adjacentBiasA = targetAdjacentDirection < 0 ? ar.top : (targetAdjacentDirection > 0 ? -ar.top : 0);
+                        const adjacentBiasB = targetAdjacentDirection < 0 ? br.top : (targetAdjacentDirection > 0 ? -br.top : 0);
+                        const ascore = adjacentBiasA + (ar.width * ar.height) / 10000 + ar.left / 100000;
+                        const bscore = adjacentBiasB + (br.width * br.height) / 10000 + br.left / 100000;
+                        return ascore - bscore;
+                    });
+                if (!dayCandidates.length) return false;
+                (dayCandidates[0].closest('td, button') || dayCandidates[0]).click();
+                return true;
+            }""",
+                [date_dash, date_slash, day_text, target_month_key],
+            )
+            if clicked:
+                page.wait_for_timeout(1000)
+            return bool(clicked)
+        except Exception as e:
+            print(f"[WARN] Calendar date click failed ({date_dash}): {e}")
+            return False
+
+    def visible_month_keys() -> list[int]:
+        try:
+            return page.evaluate(
+                """() => {
+                    const isVisible = (el) => {
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style && style.visibility !== 'hidden' && style.display !== 'none'
+                            && rect.width > 0 && rect.height > 0;
+                    };
+                    const keys = [];
+                    for (const el of Array.from(document.querySelectorAll('div, span, button, th'))) {
+                        if (!isVisible(el)) continue;
+                        const text = (el.innerText || el.textContent || '').replace(/\\s+/g, ' ').trim();
+                        const match = text.match(/(20\\d{2})\\s*年\\s*(\\d{1,2})\\s*月/);
+                        if (match) keys.push(Number(match[1]) * 12 + Number(match[2]));
+                    }
+                    return Array.from(new Set(keys)).sort((a, b) => a - b);
+                }"""
+            ) or []
+        except Exception:
+            return []
+
+    def click_calendar_nav(direction: str) -> bool:
+        try:
+            clicked = page.evaluate(
+                """(direction) => {
+                    const isVisible = (el) => {
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        return style && style.visibility !== 'hidden' && style.display !== 'none'
+                            && rect.width > 0 && rect.height > 0;
+                    };
+                    const inDatePicker = (el) => {
+                        let cur = el;
+                        for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                            const cls = String(cur.className || '').toLowerCase();
+                            const role = (cur.getAttribute && cur.getAttribute('role') || '').toLowerCase();
+                            if (cls.includes('picker') || cls.includes('calendar') || cls.includes('date')
+                                || role.includes('dialog') || role.includes('grid')) return true;
+                        }
+                        return false;
+                    };
+                    const candidates = Array.from(document.querySelectorAll('button, span, div, a'))
+                        .filter(isVisible)
+                        .filter(inDatePicker)
+                        .map(el => {
+                            const rect = el.getBoundingClientRect();
+                            const text = (el.innerText || el.textContent || '').trim();
+                            const title = el.getAttribute('title') || '';
+                            const aria = el.getAttribute('aria-label') || '';
+                            const cls = String(el.className || '').toLowerCase();
+                            return { el, rect, text, title, aria, cls };
+                        })
+                        .filter(item => {
+                            const signal = `${item.text} ${item.title} ${item.aria} ${item.cls}`;
+                            if (direction === 'prev') {
+                                return signal.includes('上一月') || signal.includes('prev-month') || signal.includes('previous-month')
+                                    || signal.includes('prevmonth') || signal.includes('chevron_left') || signal.includes('arrowleft')
+                                    || item.text === '‹' || item.text === '<'
+                                    || item.text === '«' || item.text === '<<';
+                            }
+                            return signal.includes('下一月') || signal.includes('next-month') || signal.includes('nextmonth')
+                                || signal.includes('chevron_right') || signal.includes('arrowright')
+                                || item.text === '›' || item.text === '>' || item.text === '»' || item.text === '>>';
+                        })
+                        .map(item => {
+                            const signal = `${item.text} ${item.title} ${item.aria} ${item.cls}`;
+                            let score = 0;
+                            if (direction === 'prev') {
+                                if (signal.includes('上一月') || signal.includes('prev-month') || signal.includes('previous-month') || signal.includes('prevmonth')) score += 100;
+                                if (signal.includes('chevron_left') || signal.includes('arrowleft')) score += 85;
+                                if (item.text === '‹' || item.text === '<') score += 80;
+                                if (item.text === '«' || item.text === '<<') score -= 60;
+                            } else {
+                                if (signal.includes('下一月') || signal.includes('next-month') || signal.includes('nextmonth')) score += 100;
+                                if (signal.includes('chevron_right') || signal.includes('arrowright')) score += 85;
+                                if (item.text === '›' || item.text === '>') score += 80;
+                                if (item.text === '»' || item.text === '>>') score -= 60;
+                            }
+                            score -= Math.abs(item.rect.top - window.innerHeight * 0.18) / 100;
+                            return { ...item, score };
+                        })
+                        .sort((a, b) => b.score - a.score || (direction === 'prev' ? a.rect.left - b.rect.left : b.rect.left - a.rect.left));
+                    let target = candidates[0]?.el || null;
+                    if (!target) {
+                        const pickerItems = Array.from(document.querySelectorAll('button, span, div, a'))
+                            .filter(isVisible)
+                            .filter(inDatePicker)
+                            .map(el => ({ el, rect: el.getBoundingClientRect(), text: (el.innerText || el.textContent || '').trim() }))
+                            .filter(item => item.rect.width <= 44 && item.rect.height <= 44 && item.rect.top < window.innerHeight * 0.4);
+                        const sorted = pickerItems
+                            .sort((a, b) => {
+                                const at = Math.abs(a.rect.top - window.innerHeight * 0.18);
+                                const bt = Math.abs(b.rect.top - window.innerHeight * 0.18);
+                                return at - bt || a.rect.left - b.rect.left;
+                            })
+                            .slice(0, 8)
+                            .sort((a, b) => a.rect.left - b.rect.left);
+                        if (direction === 'prev') {
+                            target = (sorted[1] || sorted[0])?.el || null;
+                        } else {
+                            target = (sorted[sorted.length - 2] || sorted[sorted.length - 1])?.el || null;
+                        }
+                    }
+                    if (!target) return false;
+                    target.click();
+                    return true;
+                }""",
+                direction,
+            )
+            if clicked:
+                page.wait_for_timeout(800)
+            return bool(clicked)
+        except Exception as e:
+            print(f"[WARN] Calendar nav click failed ({direction}): {e}")
+            return False
+
+    if try_click_visible_date():
+        return True
+
+    for attempt in range(1, 5):
+        months = visible_month_keys()
+        direction = "prev"
+        if months:
+            if target_month_key < min(months):
+                direction = "prev"
+            elif target_month_key > max(months):
+                direction = "next"
+            else:
+                direction = "prev" if attempt % 2 == 1 else "next"
+        print(f"[INFO] Calendar date not visible; switching {direction} month for {date_dash} (attempt {attempt})")
+        if not click_calendar_nav(direction):
+            break
+        if try_click_visible_date():
+            return True
+
+    print(f"[WARN] Calendar date not selected after month switching: {date_dash}")
+    return False
+
+
+def _visible_calendar_month_keys(page) -> list[int]:
+    try:
+        return page.evaluate(
+            """() => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const keys = [];
+                for (const el of Array.from(document.querySelectorAll('div, span, button, th'))) {
+                    if (!isVisible(el)) continue;
+                    const text = (el.innerText || el.textContent || '').replace(/\\s+/g, ' ').trim();
+                    const match = text.match(/(20\\d{2})\\s*年\\s*(\\d{1,2})\\s*月/);
+                    if (match) keys.push(Number(match[1]) * 12 + Number(match[2]));
+                }
+                return Array.from(new Set(keys)).sort((a, b) => a - b);
+            }"""
+        ) or []
+    except Exception:
+        return []
+
+
+def _click_calendar_month_nav(page, direction: str) -> bool:
+    print(f"[INFO] Clicking calendar {direction} month arrow")
+    try:
+        clicked = page.evaluate(
+            """(direction) => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const inDatePicker = (el) => {
+                    let cur = el;
+                    for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                        const cls = String(cur.className || '').toLowerCase();
+                        const role = (cur.getAttribute && cur.getAttribute('role') || '').toLowerCase();
+                        if (cls.includes('picker') || cls.includes('calendar') || cls.includes('date')
+                            || role.includes('dialog') || role.includes('grid')) return true;
+                    }
+                    return false;
+                };
+                const headerCandidates = Array.from(document.querySelectorAll('div, span, button, th'))
+                    .filter(isVisible)
+                    .filter(inDatePicker)
+                    .map(el => ({ el, rect: el.getBoundingClientRect(), text: (el.innerText || el.textContent || '').replace(/\\s+/g, ' ').trim() }))
+                    .filter(item => /(20\\d{2})\\s*年\\s*(\\d{1,2})\\s*月/.test(item.text))
+                    .sort((a, b) => a.rect.top - b.rect.top || a.rect.left - b.rect.left);
+                const header = headerCandidates[0]?.rect || null;
+                const directSelector = direction === 'prev'
+                    ? '.ecom-picker-header-prev-btn:not(.ecom-picker-header-super-prev-btn), .ecom-picker-prev-icon'
+                    : '.ecom-picker-header-next-btn:not(.ecom-picker-header-super-next-btn), .ecom-picker-next-icon';
+                const direct = Array.from(document.querySelectorAll(directSelector))
+                    .filter(isVisible)
+                    .filter(inDatePicker)
+                    .map(el => {
+                        const btn = el.closest('button') || el;
+                        return { el: btn, rect: btn.getBoundingClientRect() };
+                    })
+                    .filter(item => item.rect.width > 0 && item.rect.height > 0)
+                    .sort((a, b) => {
+                        if (!header) return a.rect.left - b.rect.left;
+                        return Math.abs(a.rect.top - header.top) - Math.abs(b.rect.top - header.top);
+                    });
+                if (direct.length) {
+                    direct[0].el.click();
+                    return true;
+                }
+                const items = Array.from(document.querySelectorAll('button, span, div, a'))
+                    .filter(isVisible)
+                    .filter(inDatePicker)
+                    .map(el => {
+                        const rect = el.getBoundingClientRect();
+                        const text = (el.innerText || el.textContent || '').trim();
+                        const title = el.getAttribute('title') || '';
+                        const aria = el.getAttribute('aria-label') || '';
+                        const cls = String(el.className || '').toLowerCase();
+                        const signal = `${text} ${title} ${aria} ${cls}`;
+                        let score = 0;
+                        if (direction === 'prev') {
+                            if (signal.includes('上一月') || signal.includes('prev-month') || signal.includes('previous-month') || signal.includes('prevmonth')) score += 100;
+                            if (signal.includes('chevron_left') || signal.includes('arrowleft')) score += 80;
+                            if (text === '‹' || text === '<') score += 80;
+                            if (text === '«' || text === '<<') score -= 80;
+                            if (signal.includes('super-prev')) score -= 200;
+                        } else {
+                            if (signal.includes('下一月') || signal.includes('next-month') || signal.includes('nextmonth')) score += 100;
+                            if (signal.includes('chevron_right') || signal.includes('arrowright')) score += 80;
+                            if (text === '›' || text === '>') score += 80;
+                            if (text === '»' || text === '>>') score -= 80;
+                            if (signal.includes('super-next')) score -= 200;
+                        }
+                        if (header) {
+                            score -= Math.abs(rect.top - header.top) / 5;
+                            if (direction === 'prev' && rect.left < header.left) score += 40;
+                            if (direction === 'next' && rect.left > header.right) score += 40;
+                        }
+                        return { el, rect, score, text };
+                    })
+                    .filter(item => item.score > 0)
+                    .sort((a, b) => b.score - a.score || (direction === 'prev' ? b.rect.left - a.rect.left : a.rect.left - b.rect.left));
+                let target = items[0]?.el || null;
+                if (!target && header) {
+                    const smallNearHeader = Array.from(document.querySelectorAll('button, span, div, a'))
+                        .filter(isVisible)
+                        .filter(inDatePicker)
+                        .map(el => ({ el, rect: el.getBoundingClientRect() }))
+                        .filter(item => item.rect.width <= 42 && item.rect.height <= 42 && Math.abs(item.rect.top - header.top) < 36)
+                        .sort((a, b) => a.rect.left - b.rect.left);
+                    if (direction === 'prev') {
+                        target = smallNearHeader.filter(item => item.rect.left < header.left).slice(-1)[0]?.el || null;
+                    } else {
+                        target = smallNearHeader.find(item => item.rect.left > header.right)?.el || null;
+                    }
+                }
+                if (!target) return false;
+                target.click();
+                return true;
+            }""",
+            direction,
+        )
+        if clicked:
+            page.wait_for_timeout(800)
+        return bool(clicked)
+    except Exception as e:
+        print(f"[WARN] Calendar month nav failed ({direction}): {e}")
+        return False
+
+
+def _ensure_calendar_month(page, target_date) -> None:
+    target_month_key = target_date.year * 12 + target_date.month
+    for attempt in range(1, 4):
+        months = _visible_calendar_month_keys(page)
+        if not months or target_month_key in months:
+            return
+        direction = "prev" if target_month_key < min(months) else "next"
+        print(f"[INFO] Calendar month mismatch: visible={months}, target={target_month_key}; switching {direction} (attempt {attempt})")
+        if not _click_calendar_month_nav(page, direction):
+            return
+
+
+def _prepare_previous_week_calendar(page, week_start) -> None:
+    target_month_key = week_start.year * 12 + week_start.month
+    current_month_key = datetime.now().year * 12 + datetime.now().month
+    months = _visible_calendar_month_keys(page)
+    if months:
+        if target_month_key in months:
+            return
+        direction = "prev" if target_month_key < min(months) else "next"
+        _click_calendar_month_nav(page, direction)
+        return
+    if target_month_key < current_month_key:
+        _click_calendar_month_nav(page, "prev")
+    elif target_month_key > current_month_key:
+        _click_calendar_month_nav(page, "next")
+
+
+def _click_date_panel_quick_option(page, label: str) -> bool:
+    print(f"[INFO] Clicking date quick option: {label}")
+    try:
+        target = page.evaluate(
+            """(label) => {
+                const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const panelSignals = ['picker', 'calendar', 'date', 'popover', 'dropdown', 'overlay', 'tooltip'];
+                const inDatePanel = (el) => {
+                    let cur = el;
+                    for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                        const cls = String(cur.className || '').toLowerCase();
+                        const role = (cur.getAttribute && cur.getAttribute('role') || '').toLowerCase();
+                        if (panelSignals.some(signal => cls.includes(signal)) || role === 'dialog' || role === 'listbox') {
+                            return true;
+                        }
+                    }
+                    return false;
+                };
+                const candidates = Array.from(document.querySelectorAll('button, [role="button"], li, div, span'))
+                    .filter(isVisible)
+                    .filter(el => normalize(el.innerText || el.textContent) === label)
+                    .map(el => {
+                        let target = el;
+                        for (let cur = el; cur && cur !== document.body; cur = cur.parentElement) {
+                            const cls = String(cur.className || '').toLowerCase();
+                            const role = (cur.getAttribute && cur.getAttribute('role') || '').toLowerCase();
+                            const rect = cur.getBoundingClientRect();
+                            if ((role === 'menuitem' || cls.includes('menu') || cls.includes('item') || cls.includes('option'))
+                                && rect.width >= 40 && rect.height >= 24) {
+                                target = cur;
+                                break;
+                            }
+                        }
+                        const rect = target.getBoundingClientRect();
+                        return {
+                            el: target,
+                            rect,
+                            inPanel: inDatePanel(el),
+                            area: rect.width * rect.height,
+                        };
+                    })
+                    .sort((a, b) => Number(b.inPanel) - Number(a.inPanel) || a.rect.left - b.rect.left || a.rect.top - b.rect.top || a.area - b.area);
+                if (!candidates.length) return null;
+                const rect = candidates[0].rect;
+                return { x: rect.left + rect.width / 2, y: rect.top + rect.height / 2 };
+            }""",
+            label,
+        )
+        if target:
+            page.mouse.click(target["x"], target["y"])
+            page.wait_for_timeout(800)
+        return bool(target)
+    except Exception as e:
+        print(f"[WARN] Date quick option click failed ({label}): {e}")
+        return False
+
+
+def _open_business_date_picker(page) -> bool:
+    try:
+        clicked = page.evaluate(
+            """() => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const dateRangeRe = /\\d{4}\\/\\d{2}\\/\\d{2}\\s*-\\s*\\d{4}\\/\\d{2}\\/\\d{2}/;
+                const candidates = Array.from(document.querySelectorAll('button, [role="button"], div, span, input'))
+                    .filter(isVisible)
+                    .filter(el => {
+                        const text = normalize(el.innerText || el.textContent || el.value || el.getAttribute('placeholder'));
+                        return dateRangeRe.test(text) || text.includes('开始日期') || text.includes('结束日期');
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return false;
+                candidates[0].click();
+                return true;
+            }"""
+        )
+        if clicked:
+            return True
+    except Exception:
+        pass
+    try:
+        page.locator('input[placeholder*="开始日期"], input[placeholder*="结束日期"]').first.click(timeout=8000)
+        return True
+    except Exception:
+        return False
+
+
+def _select_business_previous_week(page, week_start, week_end) -> None:
+    print("[INFO] Selecting business natural week: " + _date_range_label(week_start, week_end))
+    if not _hover_visible_text(page, "更多", exact=True, timeout=8000):
+        _click_visible_text(page, "更多", exact=True, timeout=8000)
+    page.wait_for_timeout(800)
+
+    natural_week_clicked = (
+        _click_date_panel_quick_option(page, "自然周")
+        or _click_visible_text(page, "自然周", exact=True, timeout=8000)
+    )
+    if not natural_week_clicked:
+        print("[WARN] Business natural week option not clicked, falling back to direct calendar date selection")
+    page.wait_for_timeout(1000)
+
+    # 自然周模式下点击周内任意一天即可选中整周；如果未切换到自然周模式，
+    # 点 week_start 至少能选中上周起始日（business 页通常默认就是自然周模式）。
+    _prepare_previous_week_calendar(page, week_start)
+    if not _click_calendar_date(page, week_start) and not _click_calendar_date(page, week_end):
+        print(f"[WARN] Business natural week date not clicked: {week_start:%Y-%m-%d} - {week_end:%Y-%m-%d}")
+
+    _wait_after_action(page, "select business natural week", settle_ms=1800)
+
+
+def _hover_visible_text(page, text, exact=True, timeout=10000) -> bool:
+    print(f"[INFO] Hovering text: {text}")
+    try:
+        if adaptive_hover_text(page, text, exact=exact, timeout=timeout):
+            return True
+    except Exception as e:
+        print(f"[WARN] Adaptive hover text failed ({text}): {e}")
+    try:
+        page.get_by_text(text, exact=exact).first.hover(timeout=timeout)
+        return True
+    except Exception as e:
+        print(f"[WARN] Hover text failed ({text}): {e}")
+        return False
+
+
+def _click_visible_text(page, text, exact=True, timeout=10000) -> bool:
+    match_type = "exact text" if exact else "text"
+    print(f"[INFO] Clicking {match_type}: {text}")
+    try:
+        if adaptive_click_text(page, text, exact=exact, timeout=timeout):
+            return True
+    except Exception as e:
+        print(f"[WARN] Adaptive click text failed ({text}): {e}")
+
+    try:
+        locator = page.get_by_text(text, exact=exact)
+        locator.first.click(timeout=timeout)
+        return True
+    except Exception:
+        pass
+
+    try:
+        clicked = page.evaluate(
+            """([text, exact]) => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const candidates = Array.from(document.querySelectorAll('button, [role="button"], a, label, span, div, li'))
+                    .filter(isVisible)
+                    .filter(el => {
+                        const value = normalize(el.innerText || el.textContent);
+                        return exact ? value === text : value.includes(text);
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return false;
+                candidates[0].click();
+                return true;
+            }""",
+            [text, exact],
+        )
+        return bool(clicked)
+    except Exception as e:
+        print(f"[WARN] Click text failed ({text}): {e}")
+        return False
+
+
+def _click_market_category_dropdown(page) -> bool:
+    print("[INFO] Opening category dropdown...")
+    try:
+        clicked = page.evaluate(
+            """() => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const candidates = Array.from(document.querySelectorAll('button, [role="button"], div, span'))
+                    .filter(isVisible)
+                    .filter(el => {
+                        const text = (el.innerText || el.textContent || '').replace(/\\s+/g, '');
+                        return text.includes('鞋靴箱包') && text.includes('/');
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return false;
+                candidates[0].click();
+                return true;
+            }"""
+        )
+        return bool(clicked)
+    except Exception as e:
+        print(f"[WARN] Open category dropdown failed: {e}")
+        return False
+
+
+def _select_market_category(page, category_name: str) -> None:
+    try:
+        page.evaluate("window.scrollTo(0, 0)")
+    except Exception:
+        pass
+    page.wait_for_timeout(300)
+    if not _click_market_category_dropdown(page):
+        print("[WARN] Category dropdown not found")
+        return
+
+    page.wait_for_timeout(800)
+    for label in ["箱包", "时尚箱包", category_name]:
+        if not _click_visible_text(page, label, exact=True, timeout=5000):
+            print(f"[WARN] Category option not clicked: {label}")
+        page.wait_for_timeout(800)
+    _wait_after_action(page, f"select category {category_name}", settle_ms=1800)
+
+
+def _search_market_shop(page) -> None:
+    print(f"[INFO] Searching shop: {SHOP_NAME}")
+    try:
+        selector = 'input[placeholder*="店铺名"], input[placeholder*="搜索"], input'
+        if not adaptive_fill_input(page, "market_shop_search", selector, SHOP_NAME, timeout=10000):
+            locator = page.locator('input[placeholder*="店铺名"], input[placeholder*="搜索"]')
+            if locator.count() == 0:
+                locator = page.locator("input").first
+            else:
+                locator = locator.first
+            locator.click(timeout=10000)
+            locator.fill(SHOP_NAME)
+        page.keyboard.press("Enter")
+    except Exception as e:
+        print(f"[WARN] Shop search failed: {e}")
+        return
+    _wait_after_action(page, "shop search")
+    page.wait_for_timeout(2500)
+
+
+def _search_market_keyword(page, keyword: str) -> None:
+    print(f"[INFO] Searching market keyword: {keyword}")
+    try:
+        locator = page.locator(
+            'input[placeholder*="可搜索所选条件下的店铺名"], '
+            'input[placeholder*="所选条件"][placeholder*="店铺名"], '
+            'input[placeholder*="店铺名"][placeholder*="回车键搜索"]'
+        ).first
+        locator.wait_for(state="visible", timeout=5000)
+        locator.click(timeout=5000)
+        locator.fill(keyword, timeout=5000)
+        actual = locator.input_value(timeout=3000)
+        if keyword and keyword not in actual:
+            locator.click(timeout=3000)
+            page.keyboard.press("Control+A")
+            page.keyboard.type(keyword, delay=20)
+        elif not keyword and actual:
+            locator.fill("", timeout=3000)
+        page.keyboard.press("Enter")
+        print(f"[OK] Market search filled via locator: {keyword}")
+        _wait_after_action(page, f"keyword search {keyword}", settle_ms=1800)
+        return
+    except Exception as locator_error:
+        print(f"[WARN] Market search locator fill failed ({keyword}): {locator_error}")
+
+    try:
+        info = page.evaluate(
+            """() => {
+                const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const scoreInput = (input) => {
+                    const rect = input.getBoundingClientRect();
+                    const placeholder = normalize(input.placeholder || input.getAttribute('placeholder') || '');
+                    const text = `${placeholder} ${normalize(input.value || '')}`;
+                    let score = 0;
+                    if (placeholder.includes('可搜索所选条件下的店铺名')) score += 200;
+                    if (placeholder.includes('所选条件') && placeholder.includes('店铺')) score += 160;
+                    if (placeholder.includes('店铺名')) score += 120;
+                    if (placeholder.includes('按回车键搜索')) score += 80;
+                    if (placeholder.includes('搜索') && !placeholder.includes('智能')) score += 20;
+                    if (text.includes('AI') || text.includes('智能搜索')) score -= 220;
+                    if (rect.top < 130) score -= 180;
+                    if (rect.top >= 150 && rect.top <= 380) score += 80;
+                    if (rect.left < window.innerWidth * 0.45) score += 30;
+                    if (rect.width >= 220 && rect.width <= 620) score += 20;
+                    return { input, score, placeholder, rect };
+                };
+                const candidates = Array.from(document.querySelectorAll('input'))
+                    .filter(isVisible)
+                    .map(scoreInput)
+                    .filter(item => item.score > 0)
+                    .sort((a, b) => b.score - a.score || a.rect.top - b.rect.top);
+                const target = candidates[0];
+                if (!target) return { ok: false, reason: 'not_found' };
+                const input = target.input;
+                input.scrollIntoView({ block: 'center', inline: 'center' });
+                const freshRect = input.getBoundingClientRect();
+                return {
+                    ok: true,
+                    placeholder: target.placeholder,
+                    score: target.score,
+                    top: Math.round(freshRect.top),
+                    left: Math.round(freshRect.left),
+                    x: freshRect.left + freshRect.width / 2,
+                    y: freshRect.top + freshRect.height / 2,
+                };
+            }""",
+        )
+        if not info or not info.get("ok"):
+            print(f"[WARN] Keyword search input not found: {keyword} | {info}")
+        else:
+            print(
+                f"[OK] Market search input: placeholder={info.get('placeholder')} "
+                f"top={info.get('top')} left={info.get('left')} score={info.get('score')}"
+            )
+            page.mouse.click(info["x"], info["y"])
+            page.keyboard.press("Control+A")
+            if keyword:
+                page.keyboard.type(keyword, delay=20)
+            else:
+                page.keyboard.press("Backspace")
+            page.keyboard.press("Enter")
+    except Exception as e:
+        print(f"[WARN] Keyword search failed ({keyword}): {e}")
+    _wait_after_action(page, f"keyword search {keyword}", settle_ms=1800)
+
+
+def _extract_market_ranking(page, category_name: str) -> dict:
+    print(f"[INFO] Extracting market ranking: {category_name}")
+    result = page.evaluate(
+        """([shopName, categoryName]) => {
+            const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const headers = Array.from(document.querySelectorAll('table thead th'))
+                .map(th => clean(th.innerText || th.textContent))
+                .filter(Boolean);
+            const rows = Array.from(document.querySelectorAll('table tbody tr, table tr'));
+            const parseRank = (text, rankCell) => {
+                const normalized = clean(text);
+                const numbers = normalized.match(/\\d+/g) || [];
+                const rankMatch = numbers[0] || '';
+                const upMatch = normalized.match(/(?:↑|上升|升)\\s*(\\d+)/);
+                const downMatch = normalized.match(/(?:↓|下降|降)\\s*(\\d+)/);
+                let direction = '';
+                let change = '';
+                if (upMatch) {
+                    direction = '上升';
+                    change = upMatch[1];
+                } else if (downMatch) {
+                    direction = '下降';
+                    change = downMatch[1];
+                }
+                if (!change && numbers.length > 1) change = numbers[1];
+                if (!direction && rankCell) {
+                    const signals = [];
+                    const collect = (el) => {
+                        if (!el) return;
+                        const style = window.getComputedStyle(el);
+                        signals.push(String(el.className || '').toLowerCase());
+                        signals.push(String(el.getAttribute('style') || '').toLowerCase());
+                        signals.push(String(el.getAttribute('aria-label') || '').toLowerCase());
+                        signals.push(String(el.getAttribute('title') || '').toLowerCase());
+                        signals.push(style.color || '');
+                        Array.from(el.children || []).forEach(collect);
+                    };
+                    collect(rankCell);
+                    const signal = signals.join(' ');
+                    if (signal.includes('up') || signal.includes('rise') || signal.includes('increase')
+                        || signal.includes('red') || signal.includes('rgb(255') || signal.includes('rgb(245')
+                        || signal.includes('rgb(238') || signal.includes('rgb(239')) {
+                        direction = '上升';
+                    } else if (signal.includes('down') || signal.includes('fall') || signal.includes('decrease')
+                        || signal.includes('green') || signal.includes('rgb(0') || signal.includes('rgb(34')
+                        || signal.includes('rgb(52') || signal.includes('rgb(76') || signal.includes('rgb(82')) {
+                        direction = '下降';
+                    }
+                }
+                return {
+                    rank: rankMatch,
+                    rank_change_direction: direction,
+                    rank_change_value: change,
+                    rank_change: direction && change ? `${direction}${change}` : ''
+                };
+            };
+
+            for (const row of rows) {
+                const rowText = clean(row.innerText || row.textContent);
+                if (!rowText.includes('我的店铺') && !rowText.includes(shopName)) continue;
+
+                const cells = Array.from(row.querySelectorAll('td'))
+                    .map(td => clean(td.innerText || td.textContent));
+                const mapped = {};
+                if (headers.length && cells.length) {
+                    cells.forEach((cell, index) => {
+                        mapped[headers[index] || `列${index + 1}`] = cell;
+                    });
+                }
+                return {
+                    category: categoryName,
+                    row_text: rowText,
+                    columns: mapped,
+                    ...parseRank(cells[0] || rowText, row.querySelector('td')),
+                    captured_at: new Date().toISOString()
+                };
+            }
+
+            const lines = clean(document.body.innerText || '').split(' ').filter(Boolean);
+            const idx = lines.findIndex(line => line.includes('我的店铺') || line.includes(shopName));
+            const rowText = idx >= 0 ? lines.slice(Math.max(0, idx - 4), idx + 20).join(' ') : '';
+            return {
+                category: categoryName,
+                row_text: rowText,
+                columns: {},
+                ...parseRank(rowText, null),
+                captured_at: new Date().toISOString()
+            };
+        }""",
+        [SHOP_NAME, category_name],
+    )
+    print(f"[OK] Market ranking ({category_name}): {result.get('row_text', '')[:160]}")
+    return result
+
+
+def _extract_competitor_rows(page, brand: str, category_name: str) -> list:
+    print(f"[INFO] Extracting competitor rows: {brand} / {category_name}")
+    rows = page.evaluate(
+        """([brand, categoryName]) => {
+            const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const headers = Array.from(document.querySelectorAll('table thead th'))
+                .map(th => clean(th.innerText || th.textContent))
+                .filter(Boolean);
+            const trs = Array.from(document.querySelectorAll('table tbody tr, table tr'));
+            const result = [];
+
+            for (const row of trs) {
+                const rowText = clean(row.innerText || row.textContent);
+                if (!rowText || rowText.includes('我的店铺')) continue;
+
+                const cells = Array.from(row.querySelectorAll('td'))
+                    .map(td => clean(td.innerText || td.textContent));
+                if (cells.length < 4) continue;
+                const searchIndex = (cells[0].match(/\\d+/) || [''])[0];
+
+                const columns = {};
+                cells.forEach((cell, index) => {
+                    columns[headers[index] || `列${index + 1}`] = cell;
+                });
+                const shopName = cells[1] || '';
+                result.push({
+                    brand,
+                    category: categoryName,
+                    rank: '',
+                    search_index: searchIndex,
+                    shop_name: shopName,
+                    row_text: rowText,
+                    columns,
+                    captured_at: new Date().toISOString()
+                });
+            }
+            return result;
+        }""",
+        [brand, category_name],
+    )
+    print(f"[OK] Competitor rows ({brand} / {category_name}): {len(rows)}")
+    return rows
+
+
+def _extract_visible_rank_rows(page, category_name: str) -> list:
+    rows = page.evaluate(
+        """(categoryName) => {
+            const clean = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const headers = Array.from(document.querySelectorAll('table thead th'))
+                .map(th => clean(th.innerText || th.textContent))
+                .filter(Boolean);
+            const result = [];
+            for (const row of Array.from(document.querySelectorAll('table tbody tr, table tr'))) {
+                const rowText = clean(row.innerText || row.textContent);
+                if (!rowText || rowText.includes('我的店铺')) continue;
+                const cells = Array.from(row.querySelectorAll('td')).map(td => clean(td.innerText || td.textContent));
+                if (cells.length < 4) continue;
+                const rank = (cells[0].match(/\\d+/) || [''])[0];
+                if (!rank) continue;
+                const columns = {};
+                cells.forEach((cell, index) => {
+                    columns[headers[index] || `列${index + 1}`] = cell;
+                });
+                result.push({
+                    category: categoryName,
+                    rank,
+                    shop_name: cells[1] || '',
+                    row_text: rowText,
+                    columns,
+                    captured_at: new Date().toISOString()
+                });
+            }
+            return result;
+        }""",
+        category_name,
+    )
+    return rows
+
+
+def _click_market_page_number(page, page_number: int) -> bool:
+    print(f"[INFO] Opening market ranking page number: {page_number}")
+    try:
+        clicked = page.evaluate(
+            """(pageNumber) => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const target = String(pageNumber);
+                const candidates = Array.from(document.querySelectorAll('button, li, a, span, div'))
+                    .filter(isVisible)
+                    .filter(el => (el.innerText || el.textContent || '').trim() === target)
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (br.top - ar.top) || ((ar.width * ar.height) - (br.width * br.height));
+                    });
+                if (!candidates.length) return false;
+                (candidates[0].closest('button, li, a') || candidates[0]).click();
+                return true;
+            }""",
+            page_number,
+        )
+        if clicked:
+            _wait_after_action(page, f"open page {page_number}", settle_ms=1000)
+        return bool(clicked)
+    except Exception as e:
+        print(f"[WARN] Page number click failed ({page_number}): {e}")
+        return False
+
+
+def _normalize_shop_name(name: str) -> str:
+    return re.sub(r"\s+", "", name or "").lower()
+
+
+def _find_competitor_ranks_in_category(
+    page,
+    category_name: str,
+    shop_names: list,
+    max_pages=10,
+    week_start=None,
+    week_end=None,
+) -> dict:
+    target_names = {_normalize_shop_name(name): name for name in shop_names if name}
+    found = {}
+    if not target_names:
+        return found
+
+    print(f"[INFO] Scanning ranking pages for {len(target_names)} shops / {category_name}")
+    _search_market_keyword(page, "")
+    if week_start and week_end:
+        _select_market_previous_week(page, week_start, week_end)
+    if not _click_market_page_number(page, 1):
+        _wait_after_action(page, "reset market search", settle_ms=1000)
+
+    for page_number in range(1, max_pages + 1):
+        if page_number > 1 and not _click_market_page_number(page, page_number):
+            print(f"[WARN] Market page {page_number} not available; stop scanning")
+            break
+
+        visible_rows = _extract_visible_rank_rows(page, category_name)
+        for row in visible_rows:
+            row_name = _normalize_shop_name(row.get("shop_name", ""))
+            for target_norm, original_name in target_names.items():
+                if target_norm in found:
+                    continue
+                if row_name == target_norm or target_norm in row_name or row_name in target_norm:
+                    found[target_norm] = {
+                        "shop_name": original_name,
+                        "rank": row.get("rank", ""),
+                        "rank_row": row,
+                        "page_number": page_number,
+                    }
+        if len(found) == len(target_names):
+            break
+
+    print(f"[OK] Ranking scan found {len(found)} / {len(target_names)} shops in {category_name}")
+    return found
+
+
+def _column_value(columns: dict, *needles: str) -> str:
+    for key, value in columns.items():
+        if all(needle in key for needle in needles):
+            return value
+    return ""
+
+
+def _parse_market_number_range(value: str) -> dict:
+    raw = (value or "").strip()
+    if not raw:
+        return {"raw": raw, "min": None, "max": None, "open_ended": False}
+
+    text = raw.replace(",", "").replace("￥", "¥").replace("－", "-").replace("—", "-").replace("–", "-")
+    open_ended = text.endswith("+")
+    text = text.rstrip("+")
+
+    def parse_part(part: str):
+        part = part.strip().replace("¥", "")
+        if not part:
+            return None
+        multiplier = 10000 if "万" in part else 1
+        part = part.replace("万", "")
+        match = re.search(r"-?\d+(?:\.\d+)?", part)
+        return float(match.group(0)) * multiplier if match else None
+
+    parts = [p for p in text.split("-") if p.strip()]
+    nums = [parse_part(part) for part in parts]
+    nums = [n for n in nums if n is not None]
+    if not nums:
+        return {"raw": raw, "min": None, "max": None, "open_ended": open_ended}
+    if len(nums) == 1:
+        return {"raw": raw, "min": nums[0], "max": None if open_ended else nums[0], "open_ended": open_ended}
+    return {"raw": raw, "min": nums[0], "max": nums[-1], "open_ended": open_ended}
+
+
+def _format_market_number(value, money=False) -> str:
+    if value is None:
+        return ""
+    prefix = "¥" if money else ""
+    if abs(value) >= 10000:
+        amount = value / 10000
+        if amount.is_integer():
+            return f"{prefix}{int(amount):,}万"
+        return f"{prefix}{amount:,.2f}".rstrip("0").rstrip(".") + "万"
+    if float(value).is_integer():
+        return f"{prefix}{int(value):,}"
+    return f"{prefix}{value:,.2f}".rstrip("0").rstrip(".")
+
+
+def _format_market_range(summary: dict, money=False) -> str:
+    min_value = summary.get("min")
+    max_value = summary.get("max")
+    if min_value is None:
+        return "未获取"
+    if summary.get("open_ended") or max_value is None:
+        return f"{_format_market_number(min_value, money=money)}+"
+    if min_value == max_value:
+        return _format_market_number(min_value, money=money)
+    return f"{_format_market_number(min_value, money=money)}-{_format_market_number(max_value, money=money)}"
+
+
+def _sum_market_ranges(values: list) -> dict:
+    parsed = [_parse_market_number_range(value) for value in values if value]
+    parsed = [item for item in parsed if item.get("min") is not None]
+    if not parsed:
+        return {"min": None, "max": None, "open_ended": False, "raw_values": values}
+    min_sum = sum(item["min"] for item in parsed)
+    open_ended = any(item.get("open_ended") or item.get("max") is None for item in parsed)
+    max_sum = None if open_ended else sum(item["max"] for item in parsed)
+    return {"min": min_sum, "max": max_sum, "open_ended": open_ended, "raw_values": values}
+
+
+def _summarize_competitor_rows(brand: str, category_name: str, rows: list) -> dict:
+    ranked_rows = [row for row in rows if str(row.get("rank", "")).isdigit()]
+    best_rank = min((int(row["rank"]) for row in ranked_rows), default=None)
+
+    payment_values = []
+    order_values = []
+    click_values = []
+    stores = []
+    for row in rows:
+        columns = row.get("columns", {})
+        payment_values.append(_column_value(columns, "用户", "支付", "金额"))
+        order_values.append(_column_value(columns, "成交", "订单"))
+        click_values.append(_column_value(columns, "商品", "点击", "人数"))
+        stores.append({
+            "shop_name": row.get("shop_name", ""),
+            "rank": row.get("rank", ""),
+            "search_index": row.get("search_index", ""),
+            "用户支付金额": payment_values[-1],
+            "成交订单数": order_values[-1],
+            "商品点击人数": click_values[-1],
+            "row_text": row.get("row_text", ""),
+        })
+
+    payment_summary = _sum_market_ranges(payment_values)
+    order_summary = _sum_market_ranges(order_values)
+    click_summary = _sum_market_ranges(click_values)
+    return {
+        "brand": brand,
+        "category": category_name,
+        "matched_shop_count": len(rows),
+        "rank": str(best_rank) if best_rank is not None else "",
+        "rank_display": str(best_rank) if best_rank is not None else "未进入前200名",
+        "用户支付金额": _format_market_range(payment_summary, money=True),
+        "成交订单数": _format_market_range(order_summary),
+        "商品点击人数": _format_market_range(click_summary),
+        "payment_amount": payment_summary,
+        "order_count": order_summary,
+        "product_click_users": click_summary,
+        "stores": stores,
+        "captured_at": datetime.now().isoformat(),
+    }
+
+
+def _select_market_previous_week(page, week_start, week_end) -> None:
+    print(f"[INFO] Selecting market natural week: {_date_range_label(week_start, week_end)}")
+    if not _hover_visible_text(page, "更多", exact=True, timeout=8000):
+        _click_visible_text(page, "更多", exact=True, timeout=8000)
+    page.wait_for_timeout(800)
+    natural_week_clicked = (
+        _click_date_panel_quick_option(page, "自然周")
+        or _click_visible_text(page, "自然周", exact=True, timeout=8000)
+    )
+    if not natural_week_clicked:
+        print("[WARN] Natural week option not clicked, falling back to direct calendar date selection")
+    page.wait_for_timeout(1000)
+
+    # 自然周模式下点击周内任意一天即可选中整周，使用上周周一。
+    _prepare_previous_week_calendar(page, week_start)
+    if not _click_calendar_date(page, week_start) and not _click_calendar_date(page, week_end):
+        print(f"[WARN] Market natural week date not clicked: {week_start:%Y-%m-%d}")
+    _wait_after_action(page, "select market natural week", settle_ms=1800)
+    _verify_market_previous_week_selected(page, week_start, week_end)
+    try:
+        page.keyboard.press("Escape")
+        page.mouse.click(120, 120)
+    except Exception:
+        pass
+    page.wait_for_timeout(500)
+
+
+def _verify_market_previous_week_selected(page, week_start, week_end) -> bool:
+    expected = _date_range_label(week_start, week_end)
+    try:
+        ranges = page.evaluate(
+            """() => Array.from(document.body.innerText.matchAll(/20\\d{2}\\/\\d{2}\\/\\d{2}\\s*-\\s*20\\d{2}\\/\\d{2}\\/\\d{2}/g))
+                .map(match => match[0].replace(/\\s+/g, ' '))
+                .slice(0, 10)"""
+        )
+    except Exception:
+        ranges = []
+    ok = expected in ranges
+    if ok:
+        print(f"[OK] Market date selected: {expected}")
+    else:
+        print(f"[WARN] Market date not confirmed as {expected}; visible ranges: {ranges}")
+    return ok
+
+
+def _open_market_rank_shop_page(page, label="market ranking") -> None:
+    target_url = "https://compass.jinritemai.com/shop/chance/rank-shop"
+    print(f"[INFO] Navigating to {label}: {target_url}")
+    page.goto(target_url, wait_until="domcontentloaded", timeout=60000)
+    try:
+        page.wait_for_load_state("load", timeout=8000)
+    except Exception as e:
+        print(f"[WARN] Market rank page load wait skipped: {e}")
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+    print(f"[OK] Market rank page loaded: {page.url}")
+
+
+def _collect_market_rankings(page) -> dict:
+    week_start, week_end = _previous_week_range()
+    print("[INFO] Opening market ranking page...")
+    _open_market_rank_shop_page(page, "market shop ranking")
+
+    _select_market_previous_week(page, week_start, week_end)
+
+    rankings = {}
+    for category_name in MARKET_CATEGORIES:
+        _select_market_category(page, category_name)
+        rankings[category_name] = _extract_market_ranking(page, category_name)
+
+    return rankings
+
+
+def _collect_competitor_rankings(page) -> dict:
+    print("[INFO] Collecting competitor rankings...")
+    week_start, week_end = _previous_week_range()
+    print("[INFO] Re-opening market ranking page for competitors...")
+    _open_market_rank_shop_page(page, "competitor market shop ranking")
+    _select_market_previous_week(page, week_start, week_end)
+
+    data = {
+        "date_range": _date_range_label(*_previous_week_range()),
+        "categories": MARKET_CATEGORIES,
+        "brands": COMPETITOR_BRANDS,
+        "captured_at": datetime.now().isoformat(),
+        "url": page.url,
+        "results": {},
+    }
+
+    for category_name in MARKET_CATEGORIES:
+        if not _verify_market_previous_week_selected(page, week_start, week_end):
+            _select_market_previous_week(page, week_start, week_end)
+        _select_market_category(page, category_name)
+        if not _verify_market_previous_week_selected(page, week_start, week_end):
+            _select_market_previous_week(page, week_start, week_end)
+        category_results = {}
+        for brand in COMPETITOR_BRANDS:
+            _search_market_keyword(page, brand)
+            rows = _extract_competitor_rows(page, brand, category_name)
+            category_results[brand] = _summarize_competitor_rows(brand, category_name, rows)
+
+        shop_names = []
+        for item in category_results.values():
+            for store in item.get("stores", []):
+                if store.get("shop_name"):
+                    shop_names.append(store["shop_name"])
+
+        rank_map = _find_competitor_ranks_in_category(
+            page,
+            category_name,
+            shop_names,
+            week_start=week_start,
+            week_end=week_end,
+        )
+        for item in category_results.values():
+            best_rank = None
+            for store in item.get("stores", []):
+                key = _normalize_shop_name(store.get("shop_name", ""))
+                rank_info = rank_map.get(key, {})
+                if rank_info.get("rank"):
+                    store["rank"] = rank_info["rank"]
+                    store["rank_page_number"] = rank_info.get("page_number")
+                    rank_value = int(rank_info["rank"])
+                    if best_rank is None or rank_value < best_rank:
+                        best_rank = rank_value
+            item["rank"] = str(best_rank) if best_rank is not None else ""
+            item["rank_display"] = item["rank"] or "未进入前200名"
+        data["results"][category_name] = category_results
+
+    return data
+
+
+def _select_qianchuan_previous_week(page) -> None:
+    target_url = "https://qianchuan.jinritemai.com/home?aavid=1813687011565580"
+    week_start, week_end = _previous_week_range()
+    print(f"[INFO] Navigating to Qianchuan home: {target_url}")
+    print(f"[INFO] Qianchuan date range: {_date_range_label(week_start, week_end)}")
+    page.goto(target_url, wait_until="domcontentloaded", timeout=60000)
+    try:
+        page.wait_for_load_state("load", timeout=8000)
+    except Exception as e:
+        print(f"[WARN] Qianchuan load wait skipped: {e}")
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+
+    # 检测是否被重定向到登录页，如果是则等待用户登录
+    if _check_qianchuan_login(page):
+        print("[INFO] Qianchuan needs login, waiting...")
+        logged_page = _ensure_qianchuan_login(page)
+        if not logged_page:
+            print("[WARN] Qianchuan login failed, skipping date selection")
+            return
+        page = logged_page
+
+    qianchuan_week_ok = False
+    if _open_qianchuan_date_picker(page):
+        page.wait_for_timeout(800)
+        qianchuan_week_ok = _click_date_panel_quick_option(page, "上周")
+    if not qianchuan_week_ok:
+        print("[WARN] Qianchuan date panel quick option not clicked; trying direct text click")
+        qianchuan_week_ok = _click_visible_text(page, "上周", exact=True, timeout=8000)
+    if not qianchuan_week_ok:
+        print("[WARN] Qianchuan previous-week option still not clicked")
+
+    _wait_after_action(page, "select qianchuan previous week")
+    # 多次等待数据渲染，千川首页数据加载较慢
+    for attempt in range(3):
+        try:
+            page.wait_for_function(
+                "() => (document.body.innerText || '').includes('整体消耗')"
+                " || (document.body.innerText || '').includes('消耗')",
+                timeout=10000,
+            )
+            break
+        except Exception as e:
+            print(f"[WARN] Qianchuan metric text wait attempt {attempt + 1} skipped: {e}")
+            page.wait_for_timeout(PAGE_SETTLE_MS)
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+
+
+def _check_qianchuan_login(page) -> bool:
+    """检测千川页面是否需要登录"""
+    page = _active_page(page)
+    url = page.url
+    if "/login" in url:
+        print(f"[DEBUG] Qianchuan on login page: {url}")
+        return True
+    try:
+        login_markers = ["手机登录", "邮箱登录", "验证码", "扫码登录", "请登录"]
+        for marker in login_markers:
+            if page.locator(f"text={marker}").count() > 0:
+                print(f"[DEBUG] Qianchuan login marker found: {marker}")
+                return True
+    except Exception:
+        pass
+    return False
+
+
+def _ensure_qianchuan_login(page):
+    """等待千川页面登录完成"""
+    print("\n" + "=" * 60)
+    print("【千川登录】请在浏览器窗口中完成千川登录")
+    print("=" * 60 + "\n")
+    for i in range(LOGIN_TIMEOUT):
+        page = _active_page(page)
+        if not _check_qianchuan_login(page):
+            print(f"[OK] Qianchuan login success! URL: {page.url}")
+            page.wait_for_timeout(3000)
+            return page
+        if i % 5 == 0:
+            print(f"[WAIT] Waiting for Qianchuan login... {i}s / {LOGIN_TIMEOUT}s")
+        time.sleep(1)
+    print(f"[ERR] Qianchuan login timeout ({LOGIN_TIMEOUT}s)")
+    return None
+
+
+def _open_qianchuan_date_picker(page) -> bool:
+    print("[INFO] Opening Qianchuan date picker...")
+    try:
+        clicked = page.evaluate(
+            """() => {
+                const isVisible = (el) => {
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    return style && style.visibility !== 'hidden' && style.display !== 'none'
+                        && rect.width > 0 && rect.height > 0;
+                };
+                const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+                const rangeRe = /\\d{4}[-/]\\d{2}[-/]\\d{2}\\s*~\\s*\\d{4}[-/]\\d{2}[-/]\\d{2}/;
+                const candidates = Array.from(document.querySelectorAll('button, [role="button"], div, span, input'))
+                    .filter(isVisible)
+                    .filter(el => {
+                        const text = normalize(el.innerText || el.textContent || el.value || '');
+                        return text.includes('统计周期') || rangeRe.test(text);
+                    })
+                    .sort((a, b) => {
+                        const ar = a.getBoundingClientRect();
+                        const br = b.getBoundingClientRect();
+                        return (ar.width * ar.height) - (br.width * br.height);
+                    });
+                if (!candidates.length) return false;
+                let target = candidates[0];
+                for (let cur = target; cur && cur !== document.body; cur = cur.parentElement) {
+                    const text = normalize(cur.innerText || cur.textContent || cur.value || '');
+                    const role = (cur.getAttribute && cur.getAttribute('role')) || '';
+                    if (cur.tagName === 'BUTTON' || role === 'button' || rangeRe.test(text)) {
+                        target = cur;
+                        break;
+                    }
+                }
+                target.click();
+                return true;
+            }"""
+        )
+        return bool(clicked)
+    except Exception as e:
+        print(f"[WARN] Open Qianchuan date picker failed: {e}")
+        return False
+
+
+def _extract_qianchuan_metrics(page) -> dict:
+    print("[INFO] Extracting Qianchuan metrics...")
+    metrics = page.evaluate(
+        """() => {
+            const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const isVisible = (el) => {
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style && style.visibility !== 'hidden' && style.display !== 'none'
+                    && rect.width > 0 && rect.height > 0;
+            };
+            const labels = [
+                { key: '整体消耗', candidates: ['整体消耗(元)', '整体消耗（元）', '整体消耗'] },
+                { key: '净成交ROI', candidates: ['净成交ROI', '净成本ROI', 'ROI'] },
+            ];
+            const valueRe = /^[-+]?\\d{1,3}(,\\d{3})*(\\.\\d+)?$|^[-+]?\\d+(\\.\\d+)?$/;
+            const pctRe = /^[-+]?\\d+(\\.\\d+)?%$/;
+
+            function directionFromPercent(percent) {
+                if (!percent || percent === '--') return '';
+                const n = Number(String(percent).replace('%', ''));
+                if (Number.isNaN(n) || n === 0) return '持平';
+                return n > 0 ? '上升' : '下降';
+            }
+
+            function joinDecimalTail(lines, index) {
+                const value = lines[index] || '';
+                const tail = lines[index + 1] || '';
+                return /^\\.\\d+$/.test(tail) ? value + tail : value;
+            }
+
+            function parseFromText(text, labelText) {
+                const lines = normalize(text).split(' ').filter(Boolean);
+                const idx = lines.findIndex(line => line === labelText || line.includes(labelText));
+                if (idx < 0) return null;
+                let value = '';
+                let percent = '';
+                for (let i = idx + 1; i < Math.min(lines.length, idx + 8); i++) {
+                    if (!value && valueRe.test(lines[i])) {
+                        value = joinDecimalTail(lines, i);
+                        continue;
+                    }
+                    if (!percent && (pctRe.test(lines[i]) || lines[i] === '--')) {
+                        percent = lines[i];
+                    }
+                }
+                if (!value && !percent) return null;
+                return {
+                    value,
+                    change_percent: percent,
+                    change_direction: directionFromPercent(percent),
+                    raw_text: normalize(text)
+                };
+            }
+
+            function smallestUsefulContainer(labelEl, labelText) {
+                let cur = labelEl;
+                for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                    const text = normalize(cur.innerText || cur.textContent);
+                    if (text.includes(labelText) && text.length <= 180) {
+                        const parsed = parseFromText(text, labelText);
+                        if (parsed && parsed.value) return parsed;
+                    }
+                }
+                return null;
+            }
+
+            const result = {};
+            const visibleEls = Array.from(document.querySelectorAll('div, span, p, label, td, th, li'))
+                .filter(isVisible);
+
+            for (const target of labels) {
+                // 策略1: 精确标签匹配 + 向上查找容器
+                for (const labelText of target.candidates) {
+                    const labelEl = visibleEls.find(el => normalize(el.innerText || el.textContent) === labelText);
+                    if (labelEl) {
+                        const parsed = smallestUsefulContainer(labelEl, labelText);
+                        if (parsed) {
+                            result[target.key] = parsed;
+                            break;
+                        }
+                    }
+                }
+                if (result[target.key]) continue;
+
+                // 策略2: 模糊匹配包含标签的元素
+                for (const labelText of target.candidates) {
+                    const labelEls = visibleEls.filter(el => {
+                        const t = normalize(el.innerText || el.textContent);
+                        return t.includes(labelText) && t.length <= 120;
+                    });
+                    // 优先选更小（更精确）的元素
+                    labelEls.sort((a, b) => {
+                        const at = normalize(a.innerText || a.textContent).length;
+                        const bt = normalize(b.innerText || b.textContent).length;
+                        return at - bt;
+                    });
+                    for (const labelEl of labelEls) {
+                        const parsed = smallestUsefulContainer(labelEl, labelText);
+                        if (parsed) {
+                            result[target.key] = parsed;
+                            break;
+                        }
+                    }
+                    if (result[target.key]) break;
+                }
+                if (result[target.key]) continue;
+
+                // 策略3: body 文本逐行扫描
+                const bodyLines = (document.body.innerText || '').split('\\n').map(normalize).filter(Boolean);
+                for (const labelText of target.candidates) {
+                    const idx = bodyLines.findIndex(line => line === labelText || line.includes(labelText));
+                    if (idx < 0) continue;
+                    let value = '';
+                    let percent = '';
+                    for (let i = idx + 1; i < Math.min(bodyLines.length, idx + 8); i++) {
+                        if (!value && valueRe.test(bodyLines[i])) value = joinDecimalTail(bodyLines, i);
+                        if (!percent && (pctRe.test(bodyLines[i]) || bodyLines[i] === '--')) percent = bodyLines[i];
+                    }
+                    if (value || percent) {
+                        result[target.key] = {
+                            value,
+                            change_percent: percent,
+                            change_direction: directionFromPercent(percent),
+                            raw_text: bodyLines.slice(idx, idx + 8).join(' ')
+                        };
+                        break;
+                    }
+                }
+            }
+
+            return result;
+        }"""
+    )
+    cleaned = {k: v for k, v in metrics.items() if v}
+    print(f"[OK] Qianchuan extract: {cleaned}")
+    return cleaned
+
+
+def _collect_qianchuan_metrics(page) -> dict:
+    _select_qianchuan_previous_week(page)
+    metrics = _extract_qianchuan_metrics(page)
+
+    # 如果首次提取不到核心指标，等待后重试
+    for attempt in range(2):
+        has_spend = (metrics.get("整体消耗") or {}).get("value")
+        has_roi = (metrics.get("净成交ROI") or {}).get("value")
+        if has_spend or has_roi:
+            break
+        print(f"[WARN] Qianchuan metrics empty on attempt {attempt + 1}, waiting and retrying...")
+        page.wait_for_timeout(PAGE_SETTLE_MS * 2)
+        # 尝试滚动页面触发懒加载
+        try:
+            page.evaluate("window.scrollTo(0, 300)")
+            page.wait_for_timeout(1000)
+            page.evaluate("window.scrollTo(0, 0)")
+            page.wait_for_timeout(1000)
+        except Exception:
+            pass
+        metrics = _extract_qianchuan_metrics(page)
+
+    metrics["captured_at"] = datetime.now().isoformat()
+    metrics["url"] = page.url
+    metrics["date_range"] = _date_range_label(*_previous_week_range())
+    return metrics
+
+
+def _check_fxg_login(page) -> bool:
+    """检测抖店/FXG页面是否需要登录。先检查正向已登录标志，再检查登录标志。"""
+    page = _active_page(page)
+    url = page.url
+    if "/login" in url:
+        print(f"[DEBUG] FXG on login page: {url}")
+        return True
+    try:
+        # 体验分页面已登录的强标志
+        positive_markers = [
+            "我的体验分",
+            "商家体验分",
+            "综合体验分",
+            "体检分",
+            "商品体验",
+            "物流体验",
+            "服务体验",
+        ]
+        for marker in positive_markers:
+            if page.locator(f"text={marker}").count() > 0:
+                print(f"[DEBUG] FXG logged-in marker found: {marker}")
+                return False
+
+        login_markers = ["手机登录", "邮箱登录", "验证码", "扫码登录", "请登录", "密码登录", "登录即代表同意"]
+        for marker in login_markers:
+            if page.locator(f"text={marker}").count() > 0:
+                print(f"[DEBUG] FXG login marker found: {marker}")
+                return True
+    except Exception:
+        pass
+    return False
+
+
+def _ensure_fxg_login(page):
+    """等待抖店/FXG页面登录完成"""
+    print("\n" + "=" * 60)
+    print("【抖店登录】请在浏览器窗口中完成抖店登录（体验分页面）")
+    print("=" * 60 + "\n")
+    for i in range(LOGIN_TIMEOUT):
+        page = _active_page(page)
+        if not _check_fxg_login(page):
+            print(f"[OK] FXG login success! URL: {page.url}")
+            page.wait_for_timeout(3000)
+            return page
+        if i % 5 == 0:
+            print(f"[WAIT] Waiting for FXG login... {i}s / {LOGIN_TIMEOUT}s")
+        time.sleep(1)
+    print(f"[ERR] FXG login timeout ({LOGIN_TIMEOUT}s)")
+    return None
+
+
+def _extract_experience_score(page) -> dict:
+    print("[INFO] Extracting experience score...")
+    score = page.evaluate(
+        """() => {
+            const normalize = (s) => (s || '').replace(/\\s+/g, ' ').trim();
+            const isVisible = (el) => {
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style && style.visibility !== 'hidden' && style.display !== 'none'
+                    && rect.width > 0 && rect.height > 0;
+            };
+            const scoreRe = /^\\d+(\\.\\d+)?分?$/;
+
+            function parseScoreFromText(text, labelText) {
+                const lines = normalize(text).split(' ').filter(Boolean);
+                const idx = lines.findIndex(line => line === labelText || line.includes(labelText));
+                if (idx < 0) return '';
+                for (let i = idx + 1; i < Math.min(lines.length, idx + 8); i++) {
+                    const value = lines[i].replace(/分$/, '');
+                    if (scoreRe.test(lines[i]) || /^\\d+(\\.\\d+)?$/.test(value)) return value;
+                }
+                const compact = normalize(text).replace(/\\s+/g, '');
+                const labelIndex = compact.indexOf(labelText);
+                if (labelIndex < 0) return '';
+                const afterLabel = compact.slice(labelIndex + labelText.length);
+                const match = afterLabel.match(/^(\\d+(?:\\.\\d+)?)分?/);
+                return match ? match[1] : '';
+            }
+
+            function findByLabel(labelText) {
+                const visibleEls = Array.from(document.querySelectorAll('div, span, p, label'))
+                    .filter(isVisible);
+                const labelEl = visibleEls.find(el => normalize(el.innerText || el.textContent).includes(labelText));
+                if (labelEl) {
+                    let cur = labelEl;
+                    for (let depth = 0; cur && depth < 8; depth++, cur = cur.parentElement) {
+                        const text = normalize(cur.innerText || cur.textContent);
+                        if (text.includes(labelText) && text.length <= 220) {
+                            const score = parseScoreFromText(text, labelText);
+                            if (score) return { score, raw_text: text };
+                        }
+                    }
+                }
+
+                const bodyLines = (document.body.innerText || '').split('\\n').map(normalize).filter(Boolean);
+                const idx = bodyLines.findIndex(line => line.includes(labelText));
+                if (idx >= 0) {
+                    const chunk = bodyLines.slice(idx, idx + 8).join(' ');
+                    const score = parseScoreFromText(chunk, labelText);
+                    if (score) return { score, raw_text: chunk };
+                }
+                return null;
+            }
+
+            const result = {};
+            const main = findByLabel('我的体验分') || findByLabel('体检分') || findByLabel('商家体验分');
+            if (main) {
+                result.score = main.score;
+                result.raw_text = main.raw_text;
+            }
+
+            const detailLabels = [
+                [['商品体验得分', '商品体验分'], 'product_experience_score'],
+                [['物流体验得分', '物流体验分'], 'logistics_experience_score'],
+                [['服务体验得分', '服务体验分'], 'service_experience_score'],
+                [['差行为扣分'], 'penalty_score']
+            ];
+            for (const [labels, key] of detailLabels) {
+                let parsed = null;
+                for (const label of labels) {
+                    parsed = findByLabel(label);
+                    if (parsed) break;
+                }
+                if (parsed) {
+                    result[key] = parsed.score;
+                    result[`${key}_raw_text`] = parsed.raw_text;
+                }
+            }
+            return result;
+        }"""
+    )
+    print(f"[OK] Experience score extract: {score}")
+    return score or {}
+
+
+def _collect_experience_score(page) -> dict:
+    target_url = (
+        "https://fxg.jinritemai.com/ffa/eco/experience-score"
+        "?btm_ppre=a0.b0.c0.d0&btm_pre=a2427.b76571.c902327.d871297"
+        "&btm_show_id=ed5a73dc-4e3f-4f2f-bd62-f31a0a8b496c"
+    )
+    print(f"[INFO] Navigating to experience score page: {target_url}")
+    page.goto(target_url, wait_until="domcontentloaded", timeout=60000)
+    try:
+        page.wait_for_load_state("load", timeout=8000)
+    except Exception as e:
+        print(f"[WARN] Experience score load wait skipped: {e}")
+    page.wait_for_timeout(PAGE_SETTLE_MS)
+
+    # 等待体验分数据渲染
+    for attempt in range(3):
+        try:
+            page.wait_for_function(
+                "() => (document.body.innerText || '').includes('体验分')"
+                " || (document.body.innerText || '').includes('体检分')",
+                timeout=8000,
+            )
+            break
+        except Exception as e:
+            print(f"[WARN] Experience score text wait attempt {attempt + 1} skipped: {e}")
+            page.wait_for_timeout(PAGE_SETTLE_MS)
+
+    result = _extract_experience_score(page)
+
+    # 如果首次提取不到，等待后重试
+    if not result.get("score"):
+        print("[WARN] Experience score empty on first attempt, waiting and retrying...")
+        page.wait_for_timeout(PAGE_SETTLE_MS * 2)
+        try:
+            page.evaluate("window.scrollTo(0, 300)")
+            page.wait_for_timeout(1000)
+            page.evaluate("window.scrollTo(0, 0)")
+            page.wait_for_timeout(1000)
+        except Exception:
+            pass
+        result = _extract_experience_score(page)
+
+    result["captured_at"] = datetime.now().isoformat()
+    result["url"] = page.url
+    return result
+
+
+def _keep_browser_open(page) -> None:
+    if not KEEP_BROWSER_OPEN:
+        return
+
+    print("\n[INFO] Browser will stay open. Close the browser window or press Ctrl+C to exit.")
+    try:
+        while not page.is_closed():
+            page.wait_for_timeout(1000)
+    except KeyboardInterrupt:
+        print("\n[INFO] Exit requested by user.")
+    except Exception as e:
+        print(f"[INFO] Browser keep-open ended: {e}")
+
+
+def _page_automation(page) -> None:
+    global _scraped_metrics, _market_rankings, _competitor_rankings, _qianchuan_metrics, _experience_score, _results_saved
+
+    _maximize_page(page)
+    print(f"[DEBUG] Current URL: {page.url}")
+
+    # 1. Check and wait for login
+    if not _check_login_status(page):
+        logged_page = _ensure_login(page)
+        if not logged_page:
+            print("[ERR] Login failed, will try save cookies anyway...")
+        else:
+            page = logged_page
+            _maximize_page(page)
+
+    # 2. Save state after login (best effort)
+    _save_browser_state(page, "after login")
+
+    if SKIP_SHOP:
+        print("[INFO] SKIP_SHOP=1, skipping shop metrics/market/qianchuan/experience; will reuse existing shop JSON")
+        _load_existing_shop_data()
+    else:
+        # 3. Navigate
+        _select_menu_and_date(page)
+        _save_browser_state(page, "after business page")
+
+        # 4. Extract data
+        data = _extract_metrics_via_js(page)
+        expected_keys = {'成交金额', '成交人数', '成交订单数', '客单价', '成交退款金额', '商品点击人数', '访客数'}
+        missing = expected_keys - set(data.keys())
+        if missing:
+            print(f"[WARN] Missing: {missing}, trying fallback...")
+            data2 = _extract_metrics_via_dom_fallback(page)
+            data = {**data2, **data}
+
+        if '成交退款金额（支付时间）' in data and '成交退款金额' not in data:
+            data['成交退款金额'] = data.pop('成交退款金额（支付时间）')
+        if '店铺访客数' in data and '访客数' not in data:
+            data['访客数'] = data.pop('店铺访客数')
+
+        _scraped_metrics = data
+
+        # 5. Screenshot
+        try:
+            os.makedirs(STORE_DATA_DIR, exist_ok=True)
+            screenshot_path = os.path.join(STORE_DATA_DIR, f"compass_screenshot_{datetime.now().strftime('%Y%m%d_%H%M%S')}.png")
+            page.screenshot(path=screenshot_path, full_page=True)
+            print(f"[OK] Screenshot: {screenshot_path}")
+        except Exception as e:
+            print(f"[WARN] Screenshot failed: {e}")
+
+        # 6. Market shop ranking data
+        try:
+            _market_rankings = _collect_market_rankings(page)
+        except Exception as e:
+            print(f"[WARN] Market rankings collection failed: {e}")
+            _market_rankings = {}
+
+        # 7. Qianchuan previous-week metrics
+        try:
+            _qianchuan_metrics = _collect_qianchuan_metrics(page)
+        except Exception as e:
+            print(f"[WARN] Qianchuan metrics collection failed: {e}")
+            _qianchuan_metrics = {}
+
+        # 8. Shop experience score
+        if SKIP_EXPERIENCE_SCORE:
+            print("[INFO] SKIP_EXPERIENCE_SCORE=1, skip experience score")
+            _experience_score = {}
+        else:
+            try:
+                _experience_score = _collect_experience_score(page)
+            except Exception as e:
+                print(f"[WARN] Experience score collection failed: {e}")
+                _experience_score = {}
+
+    # 9. Competitor shop ranking data (skipped when SKIP_COMPETITOR=1)
+    if os.getenv("SKIP_COMPETITOR", "0") == "1":
+        print("[INFO] SKIP_COMPETITOR=1, skip competitor rankings")
+    else:
+        try:
+            _competitor_rankings = _collect_competitor_rankings(page)
+            save_competitor_results(_competitor_rankings)
+        except Exception as e:
+            print(f"[WARN] Competitor rankings collection failed: {e}")
+            _competitor_rankings = {}
+
+    final_data = {
+        "shop_name": SHOP_NAME,
+        "date_range": _date_range_label(*_previous_week_range()),
+        "scraped_at": datetime.now().isoformat(),
+        "url": page.url,
+        "metrics": _scraped_metrics,
+        "market_rankings": _market_rankings,
+        "qianchuan_metrics": _qianchuan_metrics,
+        "experience_score": _experience_score,
+    }
+    save_results(final_data)
+    _results_saved = True
+
+    _keep_browser_open(page)
+
+
+def save_competitor_results(data: dict) -> None:
+    date_stamp = datetime.now().strftime("%Y%m%d")
+    base_name = f"竞店数据汇总_{date_stamp}"
+
+    os.makedirs(COMPETITOR_DATA_DIR, exist_ok=True)
+
+    json_path = os.path.join(COMPETITOR_DATA_DIR, f"{base_name}.json")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(data, f, ensure_ascii=False, indent=2)
+    print(f"[OK] Competitor JSON: {json_path}")
+
+    md_path = os.path.join(COMPETITOR_DATA_DIR, f"{base_name}.md")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("# 竞店数据汇总\n\n")
+        f.write(f"- **周期**: {data.get('date_range', '')}\n")
+        f.write(f"- **采集时间**: {data.get('captured_at', '')}\n")
+        f.write(f"- **品牌**: {', '.join(data.get('brands', []))}\n\n")
+
+        results = data.get("results", {})
+        for category in data.get("categories", []):
+            f.write(f"## {category}\n\n")
+            f.write("| 品牌 | 排名 | 用户支付金额 | 成交订单数 | 商品点击人数 | 匹配店铺数 |\n")
+            f.write("|------|------|------|------|------|------|\n")
+            category_results = results.get(category, {})
+            for brand in data.get("brands", []):
+                item = category_results.get(brand, {})
+                rank = item.get("rank") or "未进入前200名"
+                f.write(
+                    f"| {brand} | {rank} | "
+                    f"{item.get('用户支付金额', '未获取')} | "
+                    f"{item.get('成交订单数', '未获取')} | "
+                    f"{item.get('商品点击人数', '未获取')} | "
+                    f"{item.get('matched_shop_count', 0)} |\n"
+                )
+            f.write("\n")
+
+        f.write("\n---\n*Generated by script*\n")
+    print(f"[OK] Competitor Markdown: {md_path}")
+
+
+def _parse_numeric_value(value) -> float | None:
+    text = str(value or "").strip()
+    if not text or text in {"未获取", "--", "-"}:
+        return None
+    multiplier = 1.0
+    if "万" in text:
+        multiplier = 10000.0
+    elif "亿" in text:
+        multiplier = 100000000.0
+    text = (
+        text.replace(",", "")
+        .replace("¥", "")
+        .replace("￥", "")
+        .replace("元", "")
+        .replace("万", "")
+        .replace("亿", "")
+        .replace("%", "")
+        .strip()
+    )
+    match = re.search(r"-?\d+(?:\.\d+)?", text)
+    if not match:
+        return None
+    return float(match.group(0)) * multiplier
+
+
+def _format_money_value(value: float | None) -> str:
+    if value is None:
+        return "未获取"
+    return f"¥{value:,.2f}"
+
+
+def _format_percent_value(value: float | None) -> str:
+    if value is None:
+        return "未获取"
+    return f"{value:.2f}%"
+
+
+def _format_ratio_value(value: float | None) -> str:
+    if value is None:
+        return "未获取"
+    return f"{value:.4f}"
+
+
+def _calculate_store_derived_metrics(data: dict) -> dict:
+    metrics = data.get("metrics", {}) or {}
+    qianchuan = data.get("qianchuan_metrics", {}) or {}
+
+    transaction_amount = _parse_numeric_value(metrics.get("成交金额"))
+    refund_amount = _parse_numeric_value(metrics.get("成交退款金额"))
+    ad_spend = _parse_numeric_value((qianchuan.get("整体消耗") or {}).get("value"))
+    net_deal_roi = _parse_numeric_value((qianchuan.get("净成交ROI") or {}).get("value"))
+
+    sales_amount = None
+    paid_ratio = None
+    adjusted_roi = None
+    if transaction_amount is not None and refund_amount is not None:
+        sales_amount = transaction_amount - refund_amount
+    if ad_spend is not None and sales_amount not in (None, 0):
+        paid_ratio = ad_spend / sales_amount * 100
+    if net_deal_roi is not None and transaction_amount not in (None, 0) and refund_amount is not None:
+        adjusted_roi = net_deal_roi * (refund_amount / transaction_amount)
+
+    return {
+        "销售额": _format_money_value(sales_amount),
+        "付费占比": _format_percent_value(paid_ratio),
+        "ROI": _format_ratio_value(adjusted_roi),
+        "raw": {
+            "成交金额": transaction_amount,
+            "成交退款金额": refund_amount,
+            "整体消耗": ad_spend,
+            "净成交ROI": net_deal_roi,
+            "销售额": sales_amount,
+            "付费占比": paid_ratio,
+            "ROI": adjusted_roi,
+        },
+        "formula": {
+            "销售额": "成交金额 - 成交退款金额",
+            "付费占比": "整体消耗 / 销售额 * 100%",
+            "ROI": "净成交ROI * (成交退款金额 / 成交金额)",
+        },
+    }
+
+
+def _load_existing_shop_data() -> None:
+    """SKIP_SHOP=1 时从今天已有的店铺 JSON 加载 metrics/market_rankings 等字段。
+
+    避免竞店任务覆盖店铺 JSON 为空数据：竞店任务只需写竞店 JSON，
+    但本店行的 market_rankings 仍需要从店铺数据中取。
+    """
+    global _scraped_metrics, _market_rankings, _qianchuan_metrics, _experience_score
+    import glob as _glob
+    today_stamp = datetime.now().strftime("%Y%m%d")
+    pattern = os.path.join(STORE_DATA_DIR, f"光影行星店铺信息汇总_{today_stamp}.json")
+    candidates = sorted(_glob.glob(pattern), key=os.path.getmtime, reverse=True)
+    if not candidates:
+        print(f"[WARN] No existing shop JSON for today under {STORE_DATA_DIR}; shop fields will be empty")
+        return
+    path = candidates[0]
+    try:
+        with open(path, encoding="utf-8") as f:
+            prev = json.load(f)
+    except Exception as e:
+        print(f"[WARN] Failed to load existing shop JSON {path}: {e}")
+        return
+    _scraped_metrics = prev.get("metrics", {}) or {}
+    _market_rankings = prev.get("market_rankings", {}) or {}
+    _qianchuan_metrics = prev.get("qianchuan_metrics", {}) or {}
+    _experience_score = prev.get("experience_score", {}) or {}
+    print(
+        f"[OK] Loaded shop data from {path}: "
+        f"metrics={len(_scraped_metrics)} keys, market_rankings={len(_market_rankings)} cats"
+    )
+
+
+def save_results(data: dict) -> None:
+    data["derived_metrics"] = _calculate_store_derived_metrics(data)
+    date_stamp = datetime.now().strftime("%Y%m%d")
+    base_name = f"光影行星店铺信息汇总_{date_stamp}"
+
+    os.makedirs(STORE_DATA_DIR, exist_ok=True)
+
+    json_path = os.path.join(STORE_DATA_DIR, f"{base_name}.json")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(data, f, ensure_ascii=False, indent=2)
+    print(f"[OK] JSON: {json_path}")
+
+    md_path = os.path.join(STORE_DATA_DIR, f"{base_name}.md")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("# 抖音罗盘经营数据报表\n\n")
+        f.write(f"- **店铺**: {data.get('shop_name', '未知')}\n")
+        f.write(f"- **周期**: {data.get('date_range', '近7天')}\n")
+        f.write(f"- **时间**: {data.get('scraped_at', '')}\n")
+        f.write(f"- **URL**: {data.get('url', '')}\n\n")
+        f.write("## 核心指标\n\n")
+        f.write("| 指标 | 数值 |\n")
+        f.write("|------|------|\n")
+        metrics = data.get("metrics", {})
+        for label, value in [
+            ("成交金额", metrics.get("成交金额", "未获取")),
+            ("成交人数", metrics.get("成交人数", "未获取")),
+            ("成交订单数", metrics.get("成交订单数", "未获取")),
+            ("客单价", metrics.get("客单价", "未获取")),
+            ("成交退款金额（支付时间）", metrics.get("成交退款金额", "未获取")),
+            ("商品点击人数", metrics.get("商品点击人数", "未获取")),
+            ("访客数", metrics.get("访客数", "未获取")),
+        ]:
+            f.write(f"| {label} | {value} |\n")
+        if not metrics:
+            f.write("| (无数据) | 请查看截图确认页面结构 |\n")
+        derived_metrics = data.get("derived_metrics", {})
+        if derived_metrics:
+            f.write("\n## 计算指标\n\n")
+            f.write("| 指标 | 数值 | 公式 |\n")
+            f.write("|------|------|------|\n")
+            formulas = derived_metrics.get("formula", {})
+            for label in ["销售额", "付费占比", "ROI"]:
+                f.write(f"| {label} | {derived_metrics.get(label, '未获取')} | {formulas.get(label, '')} |\n")
+        qianchuan_metrics = data.get("qianchuan_metrics", {})
+        if qianchuan_metrics:
+            f.write("\n## 千川投放数据\n\n")
+            f.write(f"- **周期**: {qianchuan_metrics.get('date_range', data.get('date_range', ''))}\n")
+            f.write(f"- **URL**: {qianchuan_metrics.get('url', '')}\n\n")
+            f.write("| 指标 | 数值 | 变化 | 百分比 |\n")
+            f.write("|------|------|------|------|\n")
+            for label in ["整体消耗", "净成交ROI"]:
+                item = qianchuan_metrics.get(label, {})
+                if item:
+                    f.write(
+                        f"| {label} | {item.get('value', '未获取')} | "
+                        f"{item.get('change_direction', '未获取')} | {item.get('change_percent', '未获取')} |\n"
+                    )
+            f.write(f"\n- **采集时间**: {qianchuan_metrics.get('captured_at', '')}\n")
+        experience_score = data.get("experience_score", {})
+        if experience_score:
+            f.write("\n## 店铺评分（体检分）\n\n")
+            f.write(f"- **体检分**: {experience_score.get('score', '未获取')}\n")
+            f.write(f"- **商品体验分**: {experience_score.get('product_experience_score', '未获取')}\n")
+            f.write(f"- **物流体验分**: {experience_score.get('logistics_experience_score', '未获取')}\n")
+            f.write(f"- **服务体验分**: {experience_score.get('service_experience_score', '未获取')}\n")
+            f.write(f"- **差行为扣分**: {experience_score.get('penalty_score', '未获取')}\n")
+            f.write(f"- **URL**: {experience_score.get('url', '')}\n")
+            f.write(f"- **采集时间**: {experience_score.get('captured_at', '')}\n")
+        market_rankings = data.get("market_rankings", {})
+        if market_rankings:
+            f.write("\n## 市场店铺榜单\n\n")
+            for category, ranking in market_rankings.items():
+                f.write(f"### {category}\n\n")
+                if ranking.get("rank") or ranking.get("rank_change"):
+                    f.write(f"- **排名**: {ranking.get('rank', '未获取')}\n")
+                    f.write(f"- **排名变化**: {ranking.get('rank_change', '无')}\n\n")
+                columns = ranking.get("columns", {})
+                if columns:
+                    f.write("| 字段 | 数值 |\n")
+                    f.write("|------|------|\n")
+                    for key, value in columns.items():
+                        f.write(f"| {key} | {value} |\n")
+                else:
+                    f.write(f"- **榜单行**: {ranking.get('row_text', '未获取')}\n")
+                f.write(f"- **采集时间**: {ranking.get('captured_at', '')}\n\n")
+        f.write("\n---\n*Generated by script*\n")
+    print(f"[OK] Markdown: {md_path}")
+
+
+def main() -> int:
+    print("=" * 60)
+    print("  抖音罗盘经营数据抓取")
+    print("=" * 60)
+    print(f"  Output: {OUTPUT_DIR}")
+    print(f"  Data: {DATA_DIR}")
+    print(f"  Store data: {STORE_DATA_DIR}")
+    print(f"  Competitor data: {COMPETITOR_DATA_DIR}")
+    print(f"  Cookies: {COOKIES_FILE}")
+    print(f"  Profile: {USER_DATA_DIR}")
+    print(f"  Keep browser open: {KEEP_BROWSER_OPEN}")
+    print("=" * 60 + "\n")
+
+    os.makedirs(OUTPUT_DIR, exist_ok=True)
+    os.makedirs(DATA_DIR, exist_ok=True)
+    os.makedirs(STORE_DATA_DIR, exist_ok=True)
+    os.makedirs(COMPETITOR_DATA_DIR, exist_ok=True)
+    os.makedirs(USER_DATA_DIR, exist_ok=True)
+    enable_adaptive_fetchers()
+    print("[OK] Adaptive enabled")
+
+    cookies = _load_cookies()
+
+    response_url = ""
+    response_status = 0
+
+    print("[INFO] Launching browser...")
+    try:
+        fetch_kwargs = dict(
+            headless=False,
+            real_chrome=True,
+            cookies=cookies,
+            timeout=120000,
+            page_action=_page_automation,
+            wait=2000,
+            extra_flags=["--start-maximized", "--window-size=1920,1080"],
+        )
+        chrome_path = os.getenv("CHROME_EXECUTABLE_PATH")
+        if chrome_path:
+            fetch_kwargs["executable_path"] = chrome_path
+            print(f"[INFO] Using CHROME_EXECUTABLE_PATH={chrome_path}")
+        try:
+            response = DynamicFetcher.fetch(
+                "https://compass.jinritemai.com/shop/business-part",
+                user_data_dir=USER_DATA_DIR,
+                **fetch_kwargs,
+            )
+        except Exception as profile_error:
+            print(f"[WARN] Launch with profile failed: {profile_error}")
+            print("[INFO] Retrying with saved cookies only...")
+            response = DynamicFetcher.fetch(
+                "https://compass.jinritemai.com/shop/business-part",
+                **fetch_kwargs,
+            )
+        response_url = response.url
+        response_status = response.status
+        print(f"\n[OK] Done: HTTP {response_status} | {response_url}")
+    except Exception as e:
+        print(f"\n[ERR] {e}")
+        import traceback
+        traceback.print_exc()
+        return 1
+
+    final_data = {
+        "shop_name": SHOP_NAME,
+        "date_range": _date_range_label(*_previous_week_range()),
+        "scraped_at": datetime.now().isoformat(),
+        "url": response_url,
+        "metrics": _scraped_metrics,
+        "market_rankings": _market_rankings,
+        "qianchuan_metrics": _qianchuan_metrics,
+        "experience_score": _experience_score,
+    }
+
+    print("\n" + "=" * 60)
+    print("  Results")
+    print("=" * 60)
+    for k, v in final_data["metrics"].items():
+        print(f"  {k}: {v}")
+    if not final_data["metrics"]:
+        print("  (No data)")
+    print("=" * 60 + "\n")
+
+    if not _results_saved:
+        save_results(final_data)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/jd_data_collector.py b/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/jd_data_collector.py
new file mode 100644
index 0000000..c636c2d
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/jd_data_collector.py
@@ -0,0 +1,2257 @@
+"""
+京东店铺数据采集脚本
+自动登录京东商家平台，采集商智和京准通数据，生成汇总报告。
+
+依赖:
+    pip install playwright
+    playwright install chromium
+
+用法:
+    python jd_data_collector.py
+    python jd_data_collector.py --shop "店铺名" --password "密码" --output "输出目录" --industry "行业" --caliber 7
+"""
+
+import argparse
+import json
+import os
+import re
+from datetime import datetime, timedelta
+
+from playwright.sync_api import BrowserContext, Page, sync_playwright
+from playwright.sync_api import TimeoutError as PWTimeoutError
+
+from gyxx_flow.modules.shop_intelligence.runtime.config import JD_SHOP_OUTPUT
+from gyxx_flow.modules.shop_intelligence.runtime.paths import (
+    TMP_ROOT,
+    resolve_raw_output,
+)
+
+# ── 配置 ─────────────────────────────────────────────────────────────────────
+
+DEFAULT_CONFIG = {
+    "shop": "光影行星研发",
+    "password": os.getenv("JD_PASSWORD", ""),
+    "output": JD_SHOP_OUTPUT,
+    "industry": "男包",
+    "caliber_days": 7,
+}
+
+
+# ── 工具函数 ──────────────────────────────────────────────────────────────────
+
+def get_last_week_range():
+    """返回上周一和上周日的日期 (datetime, datetime)"""
+    today = datetime.now()
+    # 本周一 = today - (weekday) 天
+    this_monday = today - timedelta(days=today.weekday())
+    last_monday = this_monday - timedelta(days=7)
+    last_sunday = this_monday - timedelta(days=1)
+    return last_monday, last_sunday
+
+
+def extract(page: Page, js_expr: str, default=None):
+    """用 page.evaluate 提取数据，失败返回 default"""
+    try:
+        result = page.evaluate(js_expr)
+        if result is None:
+            return default
+        return str(result).strip()
+    except Exception:
+        return default
+
+
+def safe_click(page: Page, text: str, timeout=5000):
+    """用 text= 选择器点击元素，等待并忽略不存在的情况"""
+    try:
+        page.click(f"text={text}", timeout=timeout)
+        return True
+    except PWTimeoutError:
+        print(f"  [!] 未找到文本: {text}")
+        return False
+
+
+def wait_for_navigation(page: Page, timeout=15000):
+    """等待页面加载完成"""
+    try:
+        page.wait_for_load_state("networkidle", timeout=timeout)
+    except PWTimeoutError:
+        pass
+
+
+def save_date_picker_debug(page: Page, reason: str):
+    """保存日期选择失败时的页面现场，便于定位页面结构变化。"""
+    debug_dir = TMP_ROOT / "debug" / "jd_date_picker"
+    os.makedirs(debug_dir, exist_ok=True)
+    ts = datetime.now().strftime("%Y%m%d_%H%M%S")
+    safe_reason = re.sub(r"[^0-9A-Za-z_-]+", "_", reason)[:40] or "date_picker"
+    screenshot_path = os.path.join(debug_dir, f"{ts}_{safe_reason}.png")
+    html_path = os.path.join(debug_dir, f"{ts}_{safe_reason}.html")
+
+    try:
+        page.screenshot(path=screenshot_path, full_page=True)
+    except Exception as e:
+        print(f"    [WARN] 保存日期调试截图失败: {e}")
+        screenshot_path = ""
+
+    try:
+        html = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    const before = window.getComputedStyle(el, '::before');
+                    const after = window.getComputedStyle(el, '::after');
+                    const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                        || (after.content && after.content !== 'none' && after.content !== '');
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && (rect.width > 0 || hasPseudo)
+                        && (rect.height > 0 || hasPseudo);
+                };
+                const roots = Array.from(document.querySelectorAll(
+                    '.jmt-date-picker-panels-wrapper, .date-picker-popup, .calendar-popup, ' +
+                    '.jmt-popover, .jmt-popover-content, .jmt-dropdown, .jmt-dropdown-menu, ' +
+                    '[class*="picker"], [class*="calendar"], [class*="popover"], [class*="dropdown"], ' +
+                    '.ant-picker-panel, .ant-picker-dropdown, .el-picker-panel'
+                )).filter(visible);
+                return roots.map((el, index) => {
+                    const rect = el.getBoundingClientRect();
+                    return `<!-- candidate ${index} ${Math.round(rect.width)}x${Math.round(rect.height)} -->\n${el.outerHTML}`;
+                }).join('\n\n');
+            }"""
+        )
+        if not html.strip():
+            html = page.content()
+        with open(html_path, "w", encoding="utf-8") as f:
+            f.write(html)
+    except Exception as e:
+        print(f"    [WARN] 保存日期调试 HTML 失败: {e}")
+        html_path = ""
+
+    if screenshot_path or html_path:
+        print(f"    [DEBUG] 日期选择现场已保存: {screenshot_path or '无截图'} {html_path or '无HTML'}")
+
+
+# ── Step 1: 登录 ──────────────────────────────────────────────────────────────
+
+def step_login(page: Page, shop: str, password: str):
+    """登录京东商家平台"""
+    print("[Step 1] 登录京东商家平台...")
+
+    page.goto("https://shop.jd.com/jdm/home", wait_until="networkidle")
+    page.wait_for_timeout(3000)
+
+    # 1. 定位账号输入框
+    account_input = page.locator('input[placeholder="请输入账号名/邮箱"]')
+    account_input.fill(shop)
+    page.wait_for_timeout(1000)
+
+    # 2. 定位密码输入框
+    password_input = page.locator('input[placeholder="请输入登录密码"]')
+    password_input.fill(password)
+    page.wait_for_timeout(1000)
+
+    # 3. 定位立即登录按钮并点击
+    login_button = page.locator('button:has-text("立即登录")')
+    login_button.click()
+
+    # 等待登录完成
+    page.wait_for_timeout(8000)
+    wait_for_navigation(page)
+
+    # 尝试关闭推广弹窗
+    print("  尝试关闭推广弹窗...")
+    page.evaluate("document.querySelector('.close-btn')?.click()")
+    page.wait_for_timeout(2000)
+    page.keyboard.press("Escape")
+    page.wait_for_timeout(2000)
+
+    print("  登录完成")
+
+
+def select_last_week_any_day(page):
+    """Hover the week tab, wait for the date panel, then select last week range."""
+    print("  选择上周时间范围...")
+
+    last_monday, last_sunday = get_last_week_range()
+    expected_start = last_monday.strftime("%Y-%m-%d")
+    expected_end = last_sunday.strftime("%Y-%m-%d")
+    target_date = last_monday + timedelta(days=2)
+    target_iso = target_date.strftime("%Y-%m-%d")
+    target_day = target_date.day
+    expected_start_day = last_monday.day
+    expected_end_day = last_sunday.day
+    print(f"    上周范围: {expected_start} ~ {expected_end}")
+
+    week_selectors = [
+        '.index-realtime-datepicker .jmt-date-picker >> nth=1',
+        '.index-realtime-datepicker .jmt-date-picker:has-text("周")',
+        '.index-realtime-datepicker li[data-event-content="当前时间_周"]',
+        '.jmt-combo-date-picker .jmt-date-picker >> nth=1',
+        '.jmt-combo-date-picker .jmt-date-picker:has-text("周")',
+        'li[data-event-content="当前时间_周"]',
+    ]
+    week_tab = None
+    for selector in week_selectors:
+        try:
+            week_tab = page.locator(selector).first
+            week_tab.wait_for(state="visible", timeout=3000)
+            week_tab.scroll_into_view_if_needed(timeout=2000)
+            week_tab.hover(timeout=3000)
+            print("    [OK] 已悬停周筛选")
+            break
+        except Exception:
+            week_tab = None
+            continue
+    if week_tab is None:
+        save_date_picker_debug(page, "week_tab_not_found")
+        raise RuntimeError("未找到顶部周筛选")
+
+    page.evaluate(
+        """(el) => {
+            for (const node of [el, el.querySelector('li'), el.parentElement].filter(Boolean)) {
+                node.dispatchEvent(new MouseEvent('mouseenter', { bubbles: true, cancelable: true, view: window }));
+                node.dispatchEvent(new MouseEvent('mouseover', { bubbles: true, cancelable: true, view: window }));
+            }
+        }""",
+        week_tab.element_handle(),
+    )
+
+    panel_ready = page.wait_for_function(
+        """() => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                const before = window.getComputedStyle(el, '::before');
+                const after = window.getComputedStyle(el, '::after');
+                const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                    || (after.content && after.content !== 'none' && after.content !== '');
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && (rect.width > 0 || hasPseudo)
+                    && (rect.height > 0 || hasPseudo);
+            };
+            return Array.from(document.querySelectorAll(
+                '.jmt-base-popover, .jmt-popover, .jmt-date-picker-dropdown-wrapper, .jmt-date-picker-panels-wrapper'
+            )).some(el => visible(el) && /202\\d年\\d{1,2}月/.test(el.textContent || '') && /一\\s*二\\s*三\\s*四\\s*五\\s*六\\s*日/.test(el.textContent || ''));
+        }""",
+        timeout=6000,
+    )
+    if not panel_ready:
+        save_date_picker_debug(page, "week_panel_not_ready")
+        raise RuntimeError("悬停周后日期弹层未出现")
+
+    target_year = target_date.year
+    target_month = target_date.month
+    for _ in range(12):
+        month_nav_result = page.evaluate(
+            r"""({targetYear, targetMonth}) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    const before = window.getComputedStyle(el, '::before');
+                    const after = window.getComputedStyle(el, '::after');
+                    const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                        || (after.content && after.content !== 'none' && after.content !== '');
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && (rect.width > 0 || hasPseudo)
+                        && (rect.height > 0 || hasPseudo);
+                };
+                const roots = Array.from(document.querySelectorAll(
+                    '.jmt-base-popover, .jmt-popover, .jmt-date-picker-dropdown-wrapper, .jmt-date-picker-dropdown-content, .jmt-date-picker-panels-wrapper'
+                )).filter((el) => visible(el) && /20\d{2}年\d{1,2}月/.test(el.textContent || ''));
+                const root = roots
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        return { el, area: rect.width * rect.height };
+                    })
+                    .sort((a, b) => a.area - b.area)[0]?.el || document;
+                const text = root.textContent || '';
+                const match = text.match(/(20\d{2})年(\d{1,2})月/);
+                if (!match) return 'no_month_title';
+                const currentYear = Number(match[1]);
+                const currentMonth = Number(match[2]);
+                if (currentYear === targetYear && currentMonth === targetMonth) {
+                    return `aligned_${currentYear}_${currentMonth}`;
+                }
+                const diff = (targetYear - currentYear) * 12 + (targetMonth - currentMonth);
+                const selector = diff < 0
+                    ? '.jmt-date-picker-header-prev-month-icon'
+                    : '.jmt-date-picker-header-next-month-icon';
+                const btn = Array.from(root.querySelectorAll(selector)).find(visible)
+                    || Array.from(document.querySelectorAll(selector)).find(visible);
+                if (!btn) return `no_nav_${currentYear}_${currentMonth}_${diff}`;
+                btn.dispatchEvent(new MouseEvent('mousedown', { bubbles: true, cancelable: true, view: window }));
+                btn.dispatchEvent(new MouseEvent('mouseup', { bubbles: true, cancelable: true, view: window }));
+                btn.click();
+                return `clicked_${diff < 0 ? 'prev' : 'next'}_${currentYear}_${currentMonth}`;
+            }""",
+            {"targetYear": target_year, "targetMonth": target_month},
+        )
+        print(f"    日期面板月份定位: {month_nav_result}")
+        if str(month_nav_result).startswith("aligned_"):
+            break
+        if not str(month_nav_result).startswith("clicked_"):
+            break
+        page.wait_for_timeout(800)
+
+    click_result = page.evaluate(
+        r"""({expectedStart, expectedEnd, targetIso, targetDay, expectedStartDay, expectedEndDay}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                const before = window.getComputedStyle(el, '::before');
+                const after = window.getComputedStyle(el, '::after');
+                const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                    || (after.content && after.content !== 'none' && after.content !== '');
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && (rect.width > 0 || hasPseudo)
+                    && (rect.height > 0 || hasPseudo);
+            };
+            const disabled = (el) => {
+                if (!el) return false;
+                const classText = `${el.className || ''} ${el.parentElement?.className || ''}`;
+                return el.disabled
+                    || el.getAttribute('aria-disabled') === 'true'
+                    || /disabled|unavailable/.test(classText);
+            };
+            const clickEl = (el) => {
+                const target = el.closest('td, [role="gridcell"], li, button, a, div') || el;
+                target.dispatchEvent(new MouseEvent('mousedown', { bubbles: true, cancelable: true, view: window }));
+                target.dispatchEvent(new MouseEvent('mouseup', { bubbles: true, cancelable: true, view: window }));
+                target.click();
+            };
+
+            const candidates = Array.from(document.querySelectorAll(
+                '.jmt-base-popover, .jmt-popover, .jmt-date-picker-dropdown-wrapper, .jmt-date-picker-dropdown-content, .jmt-date-picker-panels-wrapper'
+            )).filter(el => {
+                const text = el.textContent || '';
+                return visible(el)
+                    && /202\d年\d{1,2}月/.test(text)
+                    && /一\s*二\s*三\s*四\s*五\s*六\s*日/.test(text);
+            });
+            const root = candidates
+                .map(el => {
+                    const rect = el.getBoundingClientRect();
+                    const text = (el.textContent || '').replace(/\s+/g, '');
+                    const hasWeekHeader = /一二三四五六日/.test(text);
+                    const hasGrid = !!el.querySelector('tr, [role="row"], td, [role="gridcell"]');
+                    const area = rect.width * rect.height;
+                    const score = (hasWeekHeader ? 4 : 0) + (hasGrid ? 2 : 0) + (area < 300000 ? 1 : 0);
+                    return { el, area, score };
+                })
+                .sort((a, b) => b.score - a.score || a.area - b.area)[0]?.el || document;
+
+            const norm = (s) => String(s || '')
+                .replace(/\s+/g, '')
+                .replace(/[./年月]/g, '-')
+                .replace(/日/g, '')
+                .replace(/[~～至到—–]/g, '-')
+                .replace(/-+/g, '-')
+                .replace(/(^|-)(0+)(\d)/g, '$1$3')
+                .replace(/-$/g, '');
+            const sm = (s) => norm(s);
+            const startShort = expectedStart.slice(5);
+            const endShort = expectedEnd.slice(5);
+            const startNoZero = sm(startShort);
+            const endNoZero = sm(endShort);
+            const startCN = `${parseInt(expectedStart.slice(5,7),10)}月${parseInt(expectedStart.slice(8,10),10)}日`;
+            const endCN = `${parseInt(expectedEnd.slice(5,7),10)}月${parseInt(expectedEnd.slice(8,10),10)}日`;
+            const expectedCompact = norm(expectedStart + '-' + expectedEnd);
+            const shortCompact = norm(startShort + '-' + endShort);
+
+            const weekPatterns = [
+                expectedStart + ' ~ ' + expectedEnd,
+                expectedStart + ' - ' + expectedEnd,
+                expectedStart.replaceAll('-', '.') + '-' + expectedEnd.replaceAll('-', '.'),
+                expectedStart.replaceAll('-', '/') + '-' + expectedEnd.replaceAll('-', '/'),
+                startShort + ' ~ ' + endShort,
+                startShort + ' - ' + endShort,
+                startShort.replace('-', '.') + '-' + endShort.replace('-', '.'),
+                startShort.replace('-', '/') + '-' + endShort.replace('-', '/'),
+                startNoZero + ' ~ ' + endNoZero,
+                startNoZero + ' - ' + endNoZero,
+                startCN + ' ~ ' + endCN,
+                startCN + ' - ' + endCN,
+            ];
+
+            const allCells = Array.from(root.querySelectorAll(
+                'td, [role="gridcell"], .ant-picker-cell, .el-date-table td, ' +
+                'button, a, span, div, li'
+            )).filter(el => visible(el));
+
+            const attrsText = (cell) => [
+                cell.getAttribute('title'),
+                cell.getAttribute('aria-label'),
+                cell.getAttribute('data-date'),
+                cell.getAttribute('data-value'),
+                cell.getAttribute('data-day'),
+                cell.parentElement?.getAttribute('title'),
+                cell.parentElement?.getAttribute('aria-label'),
+                cell.parentElement?.getAttribute('data-date'),
+                cell.parentElement?.getAttribute('data-value'),
+                cell.parentElement?.getAttribute('data-day'),
+            ].filter(Boolean).join(' ');
+            const atomicCells = allCells
+                .map(el => {
+                    const rect = el.getBoundingClientRect();
+                    const tag = el.tagName.toLowerCase();
+                    const hasDateAttr = ['title', 'aria-label', 'data-date', 'data-value', 'data-day']
+                        .some(name => el.hasAttribute(name) || el.parentElement?.hasAttribute(name));
+                    const looksAtomic = hasDateAttr
+                        || ['td', 'button', 'a', 'span', 'li'].includes(tag)
+                        || el.getAttribute('role') === 'gridcell'
+                        || /cell|date|day|week/i.test(`${el.className || ''} ${el.parentElement?.className || ''}`);
+                    return { el, area: rect.width * rect.height, textLen: (el.textContent || '').trim().length, looksAtomic };
+                })
+                .filter(item => item.looksAtomic && item.area > 0 && item.area < 30000 && item.textLen < 120)
+                .sort((a, b) => a.area - b.area)
+                .map(item => item.el);
+
+            const findByIso = (iso) => {
+                const isoAltDot = iso.replaceAll('-', '.');
+                const isoAltSlash = iso.replaceAll('-', '/');
+                const md = iso.slice(5);
+                const mdDot = md.replace('-', '.');
+                const mdSlash = md.replace('-', '/');
+                const month = parseInt(iso.slice(5, 7), 10);
+                const day = parseInt(iso.slice(8, 10), 10);
+                const cn = `${month}月${day}日`;
+                for (const cell of atomicCells) {
+                    if (disabled(cell)) continue;
+                    const haystack = `${attrsText(cell)} ${(cell.textContent || '').trim()}`;
+                    const normalized = norm(haystack);
+                    if (
+                        haystack.includes(iso) ||
+                        haystack.includes(isoAltDot) ||
+                        haystack.includes(isoAltSlash) ||
+                        haystack.includes(md) ||
+                        haystack.includes(mdDot) ||
+                        haystack.includes(mdSlash) ||
+                        haystack.includes(cn) ||
+                        normalized.includes(norm(iso)) ||
+                        normalized.includes(norm(md))
+                    ) {
+                        return cell;
+                    }
+                }
+                return null;
+            };
+
+            const targetCell = findByIso(targetIso);
+            if (targetCell) {
+                clickEl(targetCell);
+                return `clicked_full_date_${targetIso}`;
+            }
+
+            for (const cell of allCells) {
+                if (disabled(cell)) continue;
+                const txt = (cell.textContent || '').trim().replace(/\s+/g, ' ');
+                if (!txt) continue;
+                const normalized = norm(txt);
+                if (normalized.includes(expectedCompact) || normalized.includes(shortCompact)) {
+                    (cell.querySelector('button, a, span, div') || cell).click();
+                    return `clicked_normalized_week_range_${txt.slice(0, 80)}`;
+                }
+                for (const pat of weekPatterns) {
+                    if (txt.includes(pat)) {
+                        (cell.querySelector('button, a, span, div') || cell).click();
+                        return `clicked_week_range_${pat}`;
+                    }
+                }
+            }
+
+            const numberCells = Array.from(root.querySelectorAll(
+                '.ant-calendar-date, .ant-picker-cell-inner, .el-date-table td, td[role="gridcell"], button, span, div'
+            ));
+
+            const normalizeDayText = (value) => String(value || '').trim().replace(/^0+(\d)$/, '$1');
+            const dayText = String(targetDay);
+            const dayText2 = dayText.padStart(2, '0');
+            const startText = String(expectedStartDay);
+            const endText = String(expectedEndDay);
+            const rows = Array.from(root.querySelectorAll('tr, [role="row"]')).filter(visible);
+            for (const row of rows) {
+                const cells = Array.from(row.querySelectorAll('td[role="gridcell"], [role="gridcell"], .ant-picker-cell, .el-date-table td'))
+                    .filter(el => visible(el) && !disabled(el))
+                    .filter(el => {
+                        const rect = el.getBoundingClientRect();
+                        return rect.width >= 18 && rect.width <= 60 && rect.height >= 18 && rect.height <= 60;
+                    })
+                    .map(el => ({ el, txt: normalizeDayText(el.textContent), x: el.getBoundingClientRect().left }))
+                    .filter(item => /^\d{1,2}$/.test(item.txt));
+                const hasStart = cells.some(item => item.txt === startText);
+                const hasEnd = cells.some(item => item.txt === endText);
+                if (hasStart && hasEnd) {
+                    const target = cells.find(item => item.txt === dayText || item.txt === dayText2);
+                    if (target) {
+                        clickEl(target.el);
+                        return `clicked_day_in_last_week_row_${dayText}`;
+                    }
+                }
+            }
+            const debugTexts = Array.from(root.querySelectorAll('td, [role="gridcell"], button, a, span, div, li'))
+                .filter(visible)
+                .map(el => (el.textContent || '').trim().replace(/\s+/g, ' '))
+                .filter(Boolean)
+                .filter((txt, index, arr) => arr.indexOf(txt) === index)
+                .slice(0, 30);
+            return `not_found|${debugTexts.join(' || ')}`;
+        }""",
+        {
+            "expectedStart": expected_start,
+            "expectedEnd": expected_end,
+            "targetIso": target_iso,
+            "targetDay": target_day,
+            "expectedStartDay": expected_start_day,
+            "expectedEndDay": expected_end_day,
+        },
+    )
+    print(f"    日期选择结果: {click_result}")
+
+    if str(click_result).startswith("clicked_"):
+        for text in ["确定", "应用", "查询", "完成"]:
+            try:
+                btn = page.locator(f'text="{text}"').last
+                btn.wait_for(state="visible", timeout=1000)
+                btn.click(timeout=1500)
+                print(f"    [OK] 已点击日期确认按钮: {text}")
+                break
+            except Exception:
+                pass
+        page.wait_for_timeout(2500)
+
+    if str(click_result).startswith("not_found"):
+        print("    [WARN] 未在日期弹层中找到上周日期，尝试页面级兜底点击...")
+        fallback = page.evaluate(
+            r"""({expectedStart, expectedEnd}) => {
+                const norm = (s) => String(s || '')
+                    .replace(/\s+/g, '')
+                    .replace(/[./年月]/g, '-')
+                    .replace(/日/g, '')
+                    .replace(/[~～至到—–]/g, '-')
+                    .replace(/-+/g, '-')
+                    .replace(/(^|-)(0+)(\d)/g, '$1$3')
+                    .replace(/-$/g, '');
+                const expected = norm(expectedStart + '-' + expectedEnd);
+                const shortExpected = norm(expectedStart.slice(5) + '-' + expectedEnd.slice(5));
+                const els = document.querySelectorAll('td, [role="gridcell"], button, span, div, li');
+                for (const el of els) {
+                    const txt = (el.textContent || '').trim();
+                    const normalized = norm(txt);
+                    if (txt.includes(expectedStart) || normalized.includes(expected) || normalized.includes(shortExpected)) {
+                        el.click();
+                        return 'fallback_' + txt;
+                    }
+                }
+                return 'not_found';
+            }""",
+            {"expectedStart": expected_start, "expectedEnd": expected_end},
+        )
+        print(f"    兜底结果: {fallback}")
+        if fallback == "not_found":
+            save_date_picker_debug(page, "last_week_not_found")
+            if "|" in str(click_result):
+                print(f"    [DEBUG] 日期弹层候选文本: {str(click_result).split('|', 1)[1]}")
+            raise RuntimeError(f"未在日期弹层中找到上周日期范围: {expected_start} ~ {expected_end}")
+
+    try:
+        page.wait_for_function(
+            """() => !Array.from(document.querySelectorAll('.jmt-base-popover, .jmt-popover')).some(el => {
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                return style.display !== 'none' && style.visibility !== 'hidden' && rect.width > 0 && rect.height > 0
+                    && /202\\d年\\d{1,2}月/.test(el.textContent || '');
+            })""",
+            timeout=5000,
+        )
+    except Exception:
+        pass
+    page.wait_for_timeout(5000)
+    print("  已切换到上周数据")
+
+
+def close_profit_plan_popup_if_present(page):
+    """Close the profit-plan promotion popup and its abandon-confirm dialog."""
+    def click_abandon_confirm():
+        try:
+            confirm = page.locator("text=\u786e\u8ba4\u653e\u5f03").last
+            confirm.wait_for(state="visible", timeout=1200)
+            confirm.click()
+            page.wait_for_timeout(1500)
+            print("  [OK] profit plan popup closed and abandoned")
+            return True
+        except Exception:
+            pass
+
+        try:
+            clicked_confirm = page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        const before = window.getComputedStyle(el, '::before');
+                        const after = window.getComputedStyle(el, '::after');
+                        const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                            || (after.content && after.content !== 'none' && after.content !== '');
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && (rect.width > 0 || hasPseudo)
+                            && (rect.height > 0 || hasPseudo);
+                    };
+                    const target = Array.from(document.querySelectorAll('button, span, div'))
+                        .filter((el) => visible(el) && (el.textContent || '').trim().includes('\u786e\u8ba4\u653e\u5f03'))
+                        .pop();
+                    if (!target) return false;
+                    (target.closest('button') || target).click();
+                    return true;
+                }"""
+            )
+            if clicked_confirm:
+                page.wait_for_timeout(1500)
+                print("  [OK] profit plan popup closed and abandoned")
+                return True
+        except Exception:
+            pass
+
+        return False
+
+    try:
+        close_point = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    const before = window.getComputedStyle(el, '::before');
+                    const after = window.getComputedStyle(el, '::after');
+                    const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                        || (after.content && after.content !== 'none' && after.content !== '');
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && (rect.width > 0 || hasPseudo)
+                        && (rect.height > 0 || hasPseudo);
+                };
+                const keywords = [
+                    '\u5f00\u542f\u7a33\u8d5a\u8ba1\u5212',
+                    '\u7a33\u8d5a\u8ba1\u5212',
+                    '100%\u4fdd\u6536\u76ca',
+                    '\u6295\u4ea7\u6bd4\u4e0d\u8fbe\u6807'
+                ];
+                const exactCloseSvg = Array.from(document.querySelectorAll('svg[viewBox="0 0 1024 1024"], svg[viewbox="0 0 1024 1024"]'))
+                    .filter(visible)
+                    .find((svg) => {
+                        const path = svg.querySelector('path');
+                        const d = path?.getAttribute('d') || '';
+                        const r = svg.getBoundingClientRect();
+                        return d.includes('M817.01 282.4')
+                            && d.includes('587.34 512')
+                            && r.top >= 0
+                            && r.left >= 0
+                            && r.top < window.innerHeight * 0.8
+                            && r.left < window.innerWidth;
+                    });
+                if (exactCloseSvg) {
+                    const clickable = exactCloseSvg.closest('button, span, i, div') || exactCloseSvg;
+                    const r = clickable.getBoundingClientRect();
+                    return {
+                        x: Math.round(r.left + r.width / 2),
+                        y: Math.round(r.top + r.height / 2),
+                        method: 'exact-close-svg'
+                    };
+                }
+
+                const centeredPopup = Array.from(document.querySelectorAll('body *'))
+                    .filter((el) => {
+                        if (!visible(el)) return false;
+                        const r = el.getBoundingClientRect();
+                        const style = window.getComputedStyle(el);
+                        const zIndex = Number.parseInt(style.zIndex, 10);
+                        const centered = r.left > window.innerWidth * 0.2
+                            && r.right < window.innerWidth * 0.85
+                            && r.top > 80
+                            && r.top < window.innerHeight * 0.5;
+                        const modalSize = r.width >= 450
+                            && r.width <= window.innerWidth * 0.75
+                            && r.height >= 260
+                            && r.height <= window.innerHeight * 0.8;
+                        return centered && modalSize && (!Number.isNaN(zIndex) || style.position === 'fixed' || style.position === 'absolute');
+                    })
+                    .map((el) => {
+                        const r = el.getBoundingClientRect();
+                        const style = window.getComputedStyle(el);
+                        const zIndex = Number.parseInt(style.zIndex, 10);
+                        return { el, r, score: (Number.isNaN(zIndex) ? 0 : zIndex * 1000000) + r.width * r.height };
+                    })
+                    .sort((a, b) => b.score - a.score)[0];
+                if (centeredPopup) {
+                    const r = centeredPopup.r;
+                    return [
+                        { x: Math.round(Math.min(r.right - 14, window.innerWidth - 5)), y: Math.round(Math.max(r.top + 14, 5)), method: 'centered-popup-corner' },
+                        { x: Math.round(Math.min(r.right - 18, window.innerWidth - 5)), y: Math.round(Math.max(r.top + 18, 5)), method: 'centered-popup-corner' },
+                        { x: Math.round(Math.min(r.right - 10, window.innerWidth - 5)), y: Math.round(Math.max(r.top + 10, 5)), method: 'centered-popup-corner' },
+                        { x: Math.round(Math.min(r.right - 26, window.innerWidth - 5)), y: Math.round(Math.max(r.top + 18, 5)), method: 'centered-popup-corner' },
+                        { x: Math.round(Math.min(r.right - 18, window.innerWidth - 5)), y: Math.round(Math.max(r.top + 26, 5)), method: 'centered-popup-corner' }
+                    ];
+                }
+
+                const roots = Array.from(document.querySelectorAll(
+                    '.jmt-portal, .jmt-dialog, .jmt-modal, .jmt-base-modal, ' +
+                    '[role="dialog"], [class*="dialog"], [class*="modal"]'
+                )).filter((el) => visible(el) && keywords.some((kw) => (el.textContent || '').includes(kw)));
+                const scoredRoots = roots.map((el) => {
+                    const r = el.getBoundingClientRect();
+                    const area = r.width * r.height;
+                    const looksLikeDialog = r.width < window.innerWidth * 0.9
+                        && r.height < window.innerHeight * 0.9
+                        && r.width > 400
+                        && r.height > 250;
+                    return { el, area, score: (looksLikeDialog ? 100000000 : 0) + area };
+                }).sort((a, b) => b.score - a.score);
+                const popup = scoredRoots[0]?.el;
+                if (!popup) return false;
+
+                const pr = popup.getBoundingClientRect();
+                const closeSvg = Array.from(popup.querySelectorAll('svg[viewBox="0 0 1024 1024"], svg[viewbox="0 0 1024 1024"]'))
+                    .filter(visible)
+                    .find((svg) => {
+                        const path = svg.querySelector('path');
+                        const d = path?.getAttribute('d') || '';
+                        return d.includes('M817.01 282.4') && d.includes('587.34 512');
+                    });
+                if (closeSvg) {
+                    const clickable = closeSvg.closest('button, span, i, div') || closeSvg;
+                    const r = clickable.getBoundingClientRect();
+                    return {
+                        x: Math.round(r.left + r.width / 2),
+                        y: Math.round(r.top + r.height / 2),
+                        method: 'close-svg'
+                    };
+                }
+
+                const closeables = Array.from(popup.querySelectorAll(
+                    'button, span, i, svg, div, [class*="close"], [aria-label*="\u5173\u95ed"], [title*="\u5173\u95ed"]'
+                )).filter(visible);
+                const scored = closeables.map((el) => {
+                    const r = el.getBoundingClientRect();
+                    const label = [
+                        el.getAttribute('class') || '',
+                        el.getAttribute('aria-label') || '',
+                        el.getAttribute('title') || '',
+                        el.textContent || ''
+                    ].join(' ');
+                    let score = 0;
+                    if (r.left >= pr.right - 120) score += 20;
+                    if (r.top <= pr.top + 100) score += 20;
+                    if (/close|\u5173\u95ed|icon|svg/i.test(label)) score += 8;
+                    if (r.width <= 80 && r.height <= 80) score += 5;
+                    return { el, score };
+                }).sort((a, b) => b.score - a.score);
+                const target = scored[0]?.el;
+                if (target && scored[0].score >= 20) {
+                    const r = target.getBoundingClientRect();
+                    return {
+                        x: Math.round(r.left + r.width / 2),
+                        y: Math.round(r.top + r.height / 2),
+                        method: 'element'
+                    };
+                }
+                return {
+                    x: Math.round(Math.min(pr.right - 24, window.innerWidth - 5)),
+                    y: Math.round(Math.max(pr.top + 24, 5)),
+                    method: 'corner'
+                };
+            }"""
+        )
+    except Exception:
+        close_point = None
+
+    if not close_point:
+        return False
+
+    close_points = close_point if isinstance(close_point, list) else [close_point]
+    clicked_any = False
+    for point in close_points:
+        try:
+            page.mouse.click(point["x"], point["y"])
+            clicked_any = True
+        except Exception:
+            try:
+                page.evaluate(
+                    """({x, y}) => {
+                        const el = document.elementFromPoint(x, y);
+                        if (el) (el.closest('button, span, i, div') || el).click();
+                    }""",
+                    point,
+                )
+                clicked_any = True
+            except Exception:
+                continue
+
+        page.wait_for_timeout(600)
+        if click_abandon_confirm():
+            return True
+
+    if not clicked_any:
+        return False
+
+    try:
+        confirm = page.locator("text=\u786e\u8ba4\u653e\u5f03").last
+        confirm.wait_for(state="visible", timeout=4000)
+        confirm.click()
+        page.wait_for_timeout(1500)
+        print("  [OK] profit plan popup closed and abandoned")
+        return True
+    except Exception:
+        try:
+            clicked_confirm = page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        const before = window.getComputedStyle(el, '::before');
+                        const after = window.getComputedStyle(el, '::after');
+                        const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                            || (after.content && after.content !== 'none' && after.content !== '');
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && (rect.width > 0 || hasPseudo)
+                            && (rect.height > 0 || hasPseudo);
+                    };
+                    const target = Array.from(document.querySelectorAll('button, span, div'))
+                        .filter((el) => visible(el) && (el.textContent || '').trim().includes('\u786e\u8ba4\u653e\u5f03'))
+                        .pop();
+                    if (!target) return false;
+                    (target.closest('button') || target).click();
+                    return true;
+                }"""
+            )
+            if clicked_confirm:
+                page.wait_for_timeout(1500)
+                print("  [OK] profit plan popup closed and abandoned")
+                return True
+        except Exception:
+            pass
+
+        try:
+            fallback_point = page.evaluate(
+                r"""() => {
+                    const visible = (el) => {
+                        if (!el) return false;
+                        const style = window.getComputedStyle(el);
+                        const rect = el.getBoundingClientRect();
+                        const before = window.getComputedStyle(el, '::before');
+                        const after = window.getComputedStyle(el, '::after');
+                        const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                            || (after.content && after.content !== 'none' && after.content !== '');
+                        return style.display !== 'none'
+                            && style.visibility !== 'hidden'
+                            && (rect.width > 0 || hasPseudo)
+                            && (rect.height > 0 || hasPseudo);
+                    };
+                    const keywords = [
+                        '\u5f00\u542f\u7a33\u8d5a\u8ba1\u5212',
+                        '\u7a33\u8d5a\u8ba1\u5212',
+                        '100%\u4fdd\u6536\u76ca',
+                        '\u6295\u4ea7\u6bd4\u4e0d\u8fbe\u6807'
+                    ];
+                    const roots = Array.from(document.querySelectorAll(
+                        '.jmt-portal, .jmt-dialog, .jmt-modal, .jmt-base-modal, ' +
+                        '[role="dialog"], [class*="dialog"], [class*="modal"]'
+                    )).filter((el) => visible(el) && keywords.some((kw) => (el.textContent || '').includes(kw)));
+                    const popup = roots.map((el) => {
+                        const r = el.getBoundingClientRect();
+                        const area = r.width * r.height;
+                        const looksLikeDialog = r.width < window.innerWidth * 0.9
+                            && r.height < window.innerHeight * 0.9
+                            && r.width > 400
+                            && r.height > 250;
+                        return { el, area, score: (looksLikeDialog ? 100000000 : 0) + area };
+                    }).sort((a, b) => b.score - a.score)[0]?.el;
+                    if (!popup) return null;
+                    const r = popup.getBoundingClientRect();
+                    return {
+                        x: Math.round(Math.min(r.right - 24, window.innerWidth - 5)),
+                        y: Math.round(Math.max(r.top + 24, 5))
+                    };
+                }"""
+            )
+            if fallback_point:
+                page.mouse.click(fallback_point["x"], fallback_point["y"])
+                page.wait_for_timeout(800)
+                confirm = page.locator("text=\u786e\u8ba4\u653e\u5f03").last
+                confirm.wait_for(state="visible", timeout=3000)
+                confirm.click()
+                page.wait_for_timeout(1500)
+                print("  [OK] profit plan popup closed and abandoned")
+                return True
+        except Exception:
+            pass
+
+    try:
+        popup_still_visible = page.evaluate(
+            r"""() => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    const before = window.getComputedStyle(el, '::before');
+                    const after = window.getComputedStyle(el, '::after');
+                    const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                        || (after.content && after.content !== 'none' && after.content !== '');
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && (rect.width > 0 || hasPseudo)
+                        && (rect.height > 0 || hasPseudo);
+                };
+                const keywords = [
+                    '\u5f00\u542f\u7a33\u8d5a\u8ba1\u5212',
+                    '\u7a33\u8d5a\u8ba1\u5212',
+                    '100%\u4fdd\u6536\u76ca',
+                    '\u6295\u4ea7\u6bd4\u4e0d\u8fbe\u6807'
+                ];
+                const hasTextPopup = Array.from(document.querySelectorAll('body *'))
+                    .some((el) => visible(el) && keywords.some((kw) => (el.textContent || '').includes(kw)));
+                const hasCloseSvg = Array.from(document.querySelectorAll('svg[viewBox="0 0 1024 1024"], svg[viewbox="0 0 1024 1024"]'))
+                    .filter(visible)
+                    .some((svg) => {
+                        const d = svg.querySelector('path')?.getAttribute('d') || '';
+                        return d.includes('M817.01 282.4') && d.includes('587.34 512');
+                    });
+                return hasTextPopup || hasCloseSvg;
+            }"""
+        )
+        if not popup_still_visible:
+            print("  [OK] profit plan popup dismissed")
+            return True
+    except Exception:
+        pass
+
+    print("  [WARN] profit plan popup close clicked, but abandon confirm was not found")
+    return True
+
+
+def close_popup_if_present(page):
+    """
+    检测并关闭广告弹窗（如果存在）
+    使用 try/except 避免弹窗未出现时报错
+    """
+    closed_profit_plan = False
+    for _ in range(4):
+        if close_profit_plan_popup_if_present(page):
+            closed_profit_plan = True
+            page.wait_for_timeout(800)
+            continue
+        break
+
+    close_btn = page.locator('body > div.jmt-portal.jmt-portal-css-1nrmhc.jmtv2.jmt-dialog-portal.jmt-dialog-hidden.jmt-dialog-hidden-css-grlj9t.jmtv2.win-notice-modal > div:nth-child(2) > div > div > div > div > span > span > svg')
+    try:
+        close_btn.wait_for(state="visible", timeout=2000 if closed_profit_plan else 5000)
+        close_btn.click()
+        print("  [OK] 广告弹窗已关闭")
+        page.wait_for_timeout(3000)
+    except Exception:
+        if closed_profit_plan:
+            return
+        print("  [INFO] 无广告弹窗，继续执行")
+
+
+# ── Step 2: 进入商智 ─────────────────────────────────────────────────────────
+
+def select_last_week_in_old_datepicker(page, label="old_datepicker"):
+    """Select one day from last week in the legacy JD date picker."""
+    last_monday, last_sunday = get_last_week_range()
+    target_date = last_monday + timedelta(days=2)
+    target_year = target_date.year
+    target_month = target_date.month
+    target_day = target_date.day
+
+    print(f"  旧版日期控件选择上周: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+
+    already_set = page.evaluate(
+        r"""({start, end}) => {
+            const text = document.body?.textContent || '';
+            const norm = (s) => s.replace(/[\s ]/g, '');
+            const target = `${norm(start)}至${norm(end)}`;
+            const target2 = `${start} ~ ${end}`;
+            const target3 = `${start}至${end}`;
+            return norm(text).includes(norm(`${start}至${end}`))
+                || norm(text).includes(norm(`${start} ~ ${end}`))
+                || norm(text).includes(norm(`${start} - ${end}`))
+                || norm(text).includes(norm(`${start}-${end}`));
+        }""",
+        {"start": last_monday.strftime('%Y-%m-%d'), "end": last_sunday.strftime('%Y-%m-%d')},
+    )
+    if already_set:
+        print("  [INFO] 日期筛选已为目标范围，跳过日期控件操作")
+        return True
+
+    for _ in range(12):
+        nav_info = page.evaluate(
+            r"""({targetYear, targetMonth}) => {
+                const visible = (el) => {
+                    if (!el) return false;
+                    const style = window.getComputedStyle(el);
+                    const rect = el.getBoundingClientRect();
+                    const before = window.getComputedStyle(el, '::before');
+                    const after = window.getComputedStyle(el, '::after');
+                    const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                        || (after.content && after.content !== 'none' && after.content !== '');
+                    return style.display !== 'none'
+                        && style.visibility !== 'hidden'
+                        && (rect.width > 0 || hasPseudo)
+                        && (rect.height > 0 || hasPseudo);
+                };
+                const roots = Array.from(document.querySelectorAll(
+                    '.jmtd-popover-open, .jmtd-date-picker-panel, .jmtd-date-picker-week-panel, ' +
+                    '.dropdown-menu, .grace-datepicker, .grace-sz-datepicker, .uib-datepicker-popup, ' +
+                    '.ant-calendar-picker-container, .ant-picker-dropdown, .el-picker-panel, ' +
+                    '[class*="datepicker"], [class*="calendar"], [class*="date-picker"]'
+                )).filter((el) => visible(el) && /(20\d{2})[-/.年](\d{1,2})/.test(el.textContent || ''));
+                const root = roots
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        const hasCells = !!el.querySelector('td, [role="gridcell"], button, .ant-calendar-date');
+                        const hasHeader = !!el.querySelector(
+                            '.jmtd-date-picker-header-content, .jmt-date-picker-header, .ant-calendar-header, .ant-picker-header, .el-date-picker__header'
+                        );
+                        const isTrigger = el.classList.contains('jmtd-date-picker-combo')
+                            || el.classList.contains('jmtd-base-input')
+                            || el.classList.contains('jmt-combo-date-picker')
+                            || el.classList.contains('index-realtime-datepicker');
+                        return {
+                            el,
+                            area: rect.width * rect.height,
+                            score: (hasHeader ? 1000000000 : 0) + (hasCells ? 100000000 : 0) - (isTrigger ? 10000000000 : 0) - rect.width * rect.height
+                        };
+                    })
+                    .sort((a, b) => b.score - a.score)[0]?.el || document;
+                const text = root.textContent || '';
+                const match = text.match(/(20\d{2})[-/.年]\s*(\d{1,2})/);
+                if (!match) return { status: 'no_month_title', rootTag: root.tagName, rootClass: root.className?.substring?.(0,100) };
+                const currentYear = Number(match[1]);
+                const currentMonth = Number(match[2]);
+                if (currentYear === targetYear && currentMonth === targetMonth) {
+                    return { status: 'aligned', currentYear, currentMonth };
+                }
+                const diff = (targetYear - currentYear) * 12 + (targetMonth - currentMonth);
+                const preferred = diff < 0
+                    ? [
+                        '.jmtd-date-picker-header-btn-prev-month',
+                        '[data-event-content="prevMonth"]',
+                        '.jmt-date-picker-header-prev-month-icon',
+                        '.ant-calendar-prev-month-btn',
+                        '.ant-picker-header-prev-btn',
+                        '.el-picker-panel__icon-btn.prev-month',
+                        '.uib-left',
+                        '[ng-click*="move(-1"]',
+                        '[ng-click*="move(-1)"]',
+                        '[aria-label*="previous month" i]',
+                        '[aria-label*="上一月"]',
+                        '[title*="上一月"]'
+                    ]
+                    : [
+                        '.jmtd-date-picker-header-btn-next-month',
+                        '[data-event-content="nextMonth"]',
+                        '.jmt-date-picker-header-next-month-icon',
+                        '.ant-calendar-next-month-btn',
+                        '.ant-picker-header-next-btn',
+                        '.el-picker-panel__icon-btn.next-month',
+                        '.uib-right',
+                        '[ng-click*="move(1"]',
+                        '[ng-click*="move(1)"]',
+                        '[aria-label*="next month" i]',
+                        '[aria-label*="下一月"]',
+                        '[title*="下一月"]'
+                    ];
+                for (const selector of preferred) {
+                    let btn = Array.from(root.querySelectorAll(selector)).find(visible)
+                        || Array.from(document.querySelectorAll(selector)).find(visible);
+                    if (!btn) {
+                        // fallback: icon-only buttons may have zero rect height
+                        btn = Array.from(root.querySelectorAll(selector))[0]
+                            || Array.from(document.querySelectorAll(selector))[0];
+                    }
+                    if (btn) {
+                        return { status: 'click', selector, diff, currentYear, currentMonth };
+                    }
+                }
+                const rootRect = root.getBoundingClientRect();
+                const buttons = Array.from(root.querySelectorAll('button, a, span, i'))
+                    .filter(visible)
+                    .map((el) => {
+                        const rect = el.getBoundingClientRect();
+                        const label = [
+                            el.getAttribute('class') || '',
+                            el.getAttribute('title') || '',
+                            el.getAttribute('aria-label') || '',
+                            el.textContent || ''
+                        ].join(' ');
+                        return { el, rect, label };
+                    });
+                const candidates = buttons.filter(({rect, label}) => {
+                    const isLeft = rect.left < rootRect.left + 120;
+                    const isRight = rect.right > rootRect.right - 120;
+                    const isTop = rect.top < rootRect.top + 80;
+                    const looksMonthly = !/year|年/.test(label) || /month|月/.test(label);
+                    return isTop && looksMonthly && (diff < 0 ? isLeft : isRight);
+                }).sort((a, b) => diff < 0 ? b.rect.left - a.rect.left : a.rect.left - b.rect.left);
+                const btn = candidates[0];
+                if (!btn) return { status: 'no_nav', currentYear, currentMonth, diff };
+                return {
+                    status: 'corner_click',
+                    x: btn.rect.left + btn.rect.width / 2,
+                    y: btn.rect.top + btn.rect.height / 2,
+                    diff, currentYear, currentMonth
+                };
+            }""",
+            {"targetYear": target_year, "targetMonth": target_month},
+        )
+        print(f"    旧版日期面板月份定位: {nav_info}")
+        if not isinstance(nav_info, dict):
+            nav_info = {"status": str(nav_info)}
+        status = nav_info.get("status")
+        if status == "aligned":
+            break
+        if status == "click":
+            try:
+                sel = nav_info["selector"]
+                page.locator(
+                    f'.jmtd-date-picker-panel {sel}, .jmtd-date-picker-week-panel {sel}, .ant-picker-dropdown {sel}, .el-picker-panel {sel}, {sel}'
+                ).first.click(timeout=3000)
+            except Exception as e:
+                print(f"    [WARN] 月份导航按钮点击失败: {e}")
+                break
+        elif status == "corner_click":
+            try:
+                page.mouse.click(nav_info["x"], nav_info["y"])
+            except Exception as e:
+                print(f"    [WARN] 月份导航角落点击失败: {e}")
+                break
+        else:
+            page.evaluate(
+                r"""({targetMonth}) => {
+                    const monthCell = document.querySelector(
+                        `.jmtd-date-picker-month-cell:nth-child(${targetMonth}) .jmtd-date-picker-date-cell-content-inner, ` +
+                        `td.jmtd-date-picker-month-cell:nth-of-type(${targetMonth})`
+                    );
+                    if (monthCell) {
+                        monthCell.dispatchEvent(new MouseEvent('mousedown', { bubbles: true, cancelable: true, view: window }));
+                        monthCell.dispatchEvent(new MouseEvent('mouseup', { bubbles: true, cancelable: true, view: window }));
+                        monthCell.click();
+                        return true;
+                    }
+                    return false;
+                }""",
+                {"targetMonth": target_month},
+            )
+            page.wait_for_timeout(800)
+            break
+        page.wait_for_timeout(1000)
+
+    click_info = page.evaluate(
+        r"""({targetDay}) => {
+            const visible = (el) => {
+                if (!el) return false;
+                const style = window.getComputedStyle(el);
+                const rect = el.getBoundingClientRect();
+                const before = window.getComputedStyle(el, '::before');
+                const after = window.getComputedStyle(el, '::after');
+                const hasPseudo = (before.content && before.content !== 'none' && before.content !== '')
+                    || (after.content && after.content !== 'none' && after.content !== '');
+                return style.display !== 'none'
+                    && style.visibility !== 'hidden'
+                    && (rect.width > 0 || hasPseudo)
+                    && (rect.height > 0 || hasPseudo);
+            };
+            const disabled = (el) => {
+                const classText = `${el.className || ''} ${el.parentElement?.className || ''}`;
+                return el.disabled
+                    || el.getAttribute('aria-disabled') === 'true'
+                    || /disabled|off|unavailable|diff-month|other-month|prev-month|next-month/.test(classText);
+            };
+            const roots = Array.from(document.querySelectorAll(
+                '.jmtd-popover-open, .jmtd-date-picker-panel, .jmtd-date-picker-week-panel, ' +
+                '.dropdown-menu, .grace-datepicker, .grace-sz-datepicker, .uib-datepicker-popup, ' +
+                '.ant-calendar-picker-container, .ant-picker-dropdown, .el-picker-panel, ' +
+                '[class*="datepicker"], [class*="calendar"], [class*="date-picker"]'
+            )).filter((el) => visible(el));
+            const root = roots
+                .map((el) => {
+                    const rect = el.getBoundingClientRect();
+                    const hasCells = !!el.querySelector('td, [role="gridcell"], button, .ant-calendar-date');
+                    const hasHeader = !!el.querySelector(
+                        '.jmtd-date-picker-header-content, .jmt-date-picker-header, .ant-calendar-header, .ant-picker-header, .el-date-picker__header'
+                    );
+                    return {
+                        el,
+                        area: rect.width * rect.height,
+                        score: (hasHeader ? 1000000000 : 0) + (hasCells ? 100000000 : 0) - rect.width * rect.height
+                    };
+                })
+                .sort((a, b) => b.score - a.score)[0]?.el || document;
+            const dayText = String(targetDay);
+            const dayText2 = dayText.padStart(2, '0');
+            const startText = String(targetDay - 2);
+            const endText = String(targetDay + 4);
+            const normalizeDay = (value) => String(value || '').trim().replace(/^0+(\d)$/, '$1');
+            const rows = Array.from(root.querySelectorAll('tr, [role="row"], .jmtd-date-picker-week-row, .jmt-date-picker-calendar-week-row'))
+                .filter(visible);
+            for (const row of rows) {
+                const rowCells = Array.from(row.querySelectorAll(
+                    'td[role="gridcell"], [role="gridcell"], .ant-calendar-date, .ant-picker-cell, .el-date-table td, td'
+                ))
+                    .filter(visible)
+                    .filter((cell) => !disabled(cell))
+                    .map((cell) => ({ cell, text: normalizeDay(cell.textContent) }))
+                    .filter((item) => /^\d{1,2}$/.test(item.text));
+                if (!rowCells.length) continue;
+                const hasStart = rowCells.some((item) => item.text === startText);
+                const hasEnd = rowCells.some((item) => item.text === endText);
+                if (hasStart && hasEnd) {
+                    const target = rowCells.find((item) => item.text === dayText || item.text === dayText2);
+                    if (target) {
+                        const tempId = '__jd_date_target_' + Date.now() + '_' + Math.random().toString(36).slice(2);
+                        target.cell.setAttribute('id', tempId);
+                        return { status: 'found_week_row', id: tempId, text: target.text };
+                    }
+                }
+            }
+
+            const cells = Array.from(root.querySelectorAll(
+                'td[role="gridcell"], [role="gridcell"], .ant-calendar-date, .ant-picker-cell, ' +
+                '.el-date-table td, td'
+            )).filter(visible);
+            for (const cell of cells) {
+                if (disabled(cell)) continue;
+                const text = (cell.textContent || '').trim();
+                if (text === dayText || text === dayText2) {
+                    const tempId = '__jd_date_target_' + Date.now() + '_' + Math.random().toString(36).slice(2);
+                    cell.setAttribute('id', tempId);
+                    return { status: 'found_cell', id: tempId, text };
+                }
+            }
+            return { status: 'not_found' };
+        }""",
+        {"targetDay": target_day},
+    )
+    print(f"    旧版日期选择结果: {click_info}")
+    if not isinstance(click_info, dict):
+        click_info = {"status": str(click_info)}
+    click_status = click_info.get("status")
+    if click_status in ("found_week_row", "found_cell"):
+        try:
+            page.locator(f"#{click_info['id']}").click(timeout=3000)
+        except Exception as e:
+            print(f"    [WARN] 日期单元格点击失败: {e}")
+            save_date_picker_debug(page, f"{label}_old_week_not_found")
+            raise RuntimeError(f"旧版日期控件未能选择上周日期: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+    else:
+        save_date_picker_debug(page, f"{label}_old_week_not_found")
+        raise RuntimeError(f"旧版日期控件未能选择上周日期: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+    page.wait_for_timeout(3000)
+    return True
+
+
+def step_enter_shangzhi(context: BrowserContext, page: Page):
+    """点击商智链接，切换到新标签页，返回旧版"""
+    print("[Step 2] 进入商智...")
+
+    if page.is_closed():
+        print("  [INFO] 原页面已关闭，重新打开商家后台首页...")
+        page = context.new_page()
+        page.goto("https://shop.jd.com/jdm/home", wait_until="domcontentloaded", timeout=60000)
+        page.wait_for_timeout(3000)
+    elif "shop.jd.com" not in page.url:
+        page.goto("https://shop.jd.com/jdm/home", wait_until="domcontentloaded", timeout=60000)
+        page.wait_for_timeout(3000)
+
+    # 点击商智链接，会新开一个标签页
+    with context.expect_page() as new_page_info:
+        page.click("text=商智")
+    sz_page = new_page_info.value
+
+    try:
+        sz_page.wait_for_load_state("domcontentloaded", timeout=60000)
+    except Exception:
+        pass
+
+    # 先关闭广告弹窗（如果有）
+    close_popup_if_present(sz_page)
+
+    # 不返回旧版，直接在新页面采集
+    sz_page.wait_for_timeout(3000)
+    print("  已在商智页面")
+    return sz_page
+
+
+# ── Step 3: 采集店铺星级 ─────────────────────────────────────────────────────
+
+def step_shop_star_and_trade(page: Page):
+    """在商智新页面先选上周时间，然后采集店铺星级和交易数据"""
+    print("[Step 3] 采集店铺星级与交易数据...")
+
+    page.wait_for_timeout(5000)
+
+    # 先选择上周时间
+    # Close promotion overlays before locating the top week filter.
+    close_popup_if_present(page)
+    select_last_week_any_day(page)
+
+    # 抓取店铺星级（只提取星级数字，避免捕获多余文案）
+    star = extract(page, "document.querySelector('#jdsz-container .shop-info-left2 > div')?.textContent", "未获取到")
+    if star == "未获取到" or not star:
+        # 兜底：用更宽泛的选择器
+        star = extract(page, """() => {
+            const el = document.querySelector('.shop-info-left2 > div, [class*="shop-info"] [class*="left2"] > div');
+            return el?.textContent?.trim();
+        }""", "未获取到")
+    # 过滤：只保留星级数字部分（如 4.5星、5星）
+    star_match = re.search(r'\d+(?:\.\d+)?星', star)
+    if star_match:
+        star = star_match.group(0)
+    print(f"  店铺星级: {star}")
+
+    # 抓取交易数据（新版商智首页卡片式布局，含环比和同行均值）
+    transaction_raw = extract(page, """() => {
+        const cards = document.querySelectorAll('.jmt-summary-card-main');
+        const result = {};
+        for (const card of cards) {
+            const title = card.querySelector('.jmt-summary-card-title-left')?.textContent?.trim();
+            const value = card.querySelector('.jmt-summary-card-index-main-info')?.textContent?.trim();
+            if (!title || !value) continue;
+
+            const item = { value: value };
+
+            // 抓取环比、同行同级均值、同行上级均值
+            const descRows = card.querySelectorAll('.jmt-summary-card-description');
+            for (const row of descRows) {
+                const label = row.querySelector('.jmt-summary-card-description-left')?.textContent?.trim();
+                const val = row.querySelector('.jmt-summary-card-description-right')?.textContent?.trim();
+                if (label && val) {
+                    item[label] = val;
+                }
+            }
+
+            result[title] = item;
+        }
+        return JSON.stringify(result);
+    }""", "{}")
+
+    transaction = {}
+    if isinstance(transaction_raw, str):
+        try:
+            transaction = json.loads(transaction_raw)
+        except json.JSONDecodeError:
+            transaction = {}
+
+    # 确保关键字段存在
+    for key in ["成交金额", "商品访客数", "成交客户数", "成交转化率", "成交单量"]:
+        if key not in transaction:
+            transaction[key] = {"value": "N/A", "环比": "N/A", "同行同级均值": "N/A", "同行上级均值": "N/A"}
+
+    print(f"  交易数据: {json.dumps(transaction, ensure_ascii=False)}")
+
+    # 抓取完后点击返回旧版（供后续步骤使用）
+    print("  返回旧版...")
+    try:
+        back_button = page.locator('#header > div > div.page-top-header-tools > img')
+        back_button.wait_for(state="visible", timeout=5000)
+        back_button.click()
+        page.wait_for_timeout(5000)
+        page.wait_for_load_state("networkidle")
+        print("  已进入商智旧版")
+    except Exception:
+        print("  [INFO] 未找到返回旧版按钮，继续当前页面")
+
+    return star, transaction
+
+
+# ── Step 4: 交易概况-客单价（旧版商智）───────────────────────────────────────
+
+def step_deal_summary(page: Page):
+    """跳转交易概况页面，选择上周，采集客单价（数值+环比）"""
+    print("[Step 4] 采集交易概况-客单价...")
+
+    page.goto("https://sz.jd.com/sz/view/dealAnalysis/dealSummarys.html",
+              wait_until="networkidle")
+    page.wait_for_timeout(5000)
+
+    # 选择日期为上周
+    last_monday, last_sunday = get_last_week_range()
+    print(f"  选择上周: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+
+    try:
+        # 1. 点击日期选择器打开面板
+        page.locator('input[grace-sz-datepicker-popup]').click()
+        page.wait_for_timeout(2000)
+
+        # 2. 点击"按周查询"
+        page.locator('.quick-type-list li:has-text("按周查询")').click()
+        page.wait_for_timeout(3000)
+
+        select_last_week_in_old_datepicker(page, "deal_summary")
+    except Exception as e:
+        print(f"  [WARN] 日期选择失败: {e}")
+
+    # 提取客单价（数值 + 环比）
+    customer_price_raw = extract(page, """() => {
+        // 在 grace-summary 列表中找到"客单价"对应的 summary-item
+        const items = document.querySelectorAll('.grace-summary .summary-item, .summary-list .summary-item');
+        for (const item of items) {
+            const titleEl = item.querySelector('.summary-title label, h6 label, .ng-binding');
+            if (titleEl && titleEl.textContent.trim() === '客单价') {
+                // 数值：.summary-body 里的 .odometer
+                const unitEl = item.querySelector('.summary-body .unit');
+                const numEl = item.querySelector('.summary-body .odometer');
+                const value = (unitEl ? unitEl.textContent.trim() : '') + (numEl ? numEl.textContent.trim() : '');
+                // 环比：.summary-detail 里的 .detail-value
+                const changeEl = item.querySelector('.summary-detail .detail-value');
+                const change = changeEl ? changeEl.textContent.trim() : '';
+                return JSON.stringify({value: value, change: change});
+            }
+        }
+        return JSON.stringify({value: '', change: ''});
+    }""", '{"value": "", "change": ""}')
+
+    cp_data = {"value": "N/A", "change": ""}
+    if isinstance(customer_price_raw, str):
+        try:
+            cp_data = json.loads(customer_price_raw)
+        except json.JSONDecodeError:
+            cp_data = {"value": "N/A", "change": ""}
+
+    cp_val = cp_data.get('value', 'N/A')
+    cp_change = cp_data.get('change', '')
+    try:
+        print(f"  客单价: {cp_val}, 环比: {cp_change}")
+    except UnicodeEncodeError:
+        print(f"  客单价: {json.dumps(cp_val, ensure_ascii=True)}, 环比: {json.dumps(cp_change, ensure_ascii=True)}")
+    return {"客单价": {"value": cp_data.get("value", "N/A"), "change": cp_data.get("change", "")}}
+
+
+# ── Step 4-old: 交易概况（按周查询）────────────────────────────────────────────
+
+def step_transaction_overview(page: Page):
+    """切换按周查询，选择上周，采集交易数据"""
+    print("[Step 4-old] 采集交易概况（上周）...")
+
+    last_monday, last_sunday = get_last_week_range()
+    print(f"  上周范围: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+
+    # 点击日期选择器
+    page.click('input[placeholder*="时间"]')
+    page.wait_for_timeout(3000)
+
+    # 切换到按周查询
+    safe_click(page, "按周查询")
+    page.wait_for_timeout(3000)
+
+    # 在日期选择器中找到并点击上周
+    # 计算上周的日期文本用于匹配
+    last_week_day = last_monday.day
+    # 尝试点击上周一对应的日期数字
+    clicked = page.evaluate(f"""() => {{
+        // 查找日期选择器中所有可点击的日期单元格
+        const cells = document.querySelectorAll('.ant-calendar-date, .el-date-table td, td[role="gridcell"]');
+        for (const cell of cells) {{
+            const text = cell.textContent?.trim();
+            if (text === '{last_week_day}' || text === '{last_week_day:02d}') {{
+                cell.click();
+                return true;
+            }}
+        }}
+        return false;
+    }}""")
+
+    if not clicked:
+        print("  [!] 自动选择上周失败，尝试 snapshot 辅助定位...")
+        # 如果自动选择失败，需要人工介入或使用更复杂的逻辑
+        page.wait_for_timeout(3000)
+
+    page.wait_for_timeout(5000)
+    wait_for_navigation(page)
+
+    # 提取交易数据
+    data = {}
+    metrics = {
+        "成交金额": ".deal-amount",
+        "访客数": ".visitor-count",
+        "成交转化率": ".conversion-rate",
+        "成交单量": ".order-count",
+        "客单价": ".customer-price",
+    }
+
+    for name, selector in metrics.items():
+        val = extract(page, f"document.querySelector('{selector}')?.textContent")
+        data[name] = val
+
+    # 客单价兜底：成交金额 / 成交单量
+    if not data.get("客单价"):
+        data["客单价"] = extract(page, """() => {
+            const amount = parseFloat(document.querySelector('.deal-amount')?.textContent?.replace(/,/g, '') || '0');
+            const orders = parseFloat(document.querySelector('.order-count')?.textContent?.replace(/,/g, '') || '0');
+            return orders > 0 ? (amount / orders).toFixed(2) : 'N/A';
+        }""")
+
+    print(f"  交易数据: {json.dumps(data, ensure_ascii=False)}")
+    return data
+
+
+# ── Step 5: 商家榜单 ─────────────────────────────────────────────────────────
+
+def switch_category(page: Page, category_name: str, parent_category: str = None):
+    """切换商家榜单类目，支持级联下拉（hover父类目展开子类目后再点击）"""
+    label = f"{parent_category} → {category_name}" if parent_category else category_name
+    print(f"  切换类目到: {label}")
+    try:
+        # 1. 点击当前类目下拉框，打开下拉列表
+        page.evaluate("""() => {
+            const el = document.querySelector('.jmtd-base-input-content, .jmtd-select-input');
+            if (el) el.click();
+        }""")
+        page.wait_for_timeout(2000)
+
+        # 2. 悬停父类目以展开子类目列表
+        if parent_category:
+            try:
+                parent_locator = page.locator('.jmtd-dropdown-content .jmtd-label', has_text=parent_category).first
+                parent_locator.hover(timeout=3000)
+                page.wait_for_timeout(2000)  # 等待子类目加载
+            except Exception as e:
+                print(f"    [WARN] 悬停父类目失败: {e}")
+
+        # 3. 点击目标子类目
+        try:
+            child_locator = page.locator('.jmtd-dropdown-content .jmtd-label', has_text=category_name).first
+            child_locator.click(timeout=3000)
+            page.wait_for_timeout(5000)
+            print(f"    [OK] 已切换到 {label}")
+        except Exception as e:
+            print(f"    [WARN] 点击子类目失败: {e}")
+    except Exception as e:
+        print(f"    [WARN] 切换类目失败: {e}")
+
+
+def extract_ranking_table(page):
+    """提取商家榜单表格数据"""
+    ranking_data = extract(page, """() => {
+        const rows = document.querySelectorAll('.jmtd-table-main tbody tr');
+        const result = [];
+        let myRank = null;
+        for (const row of rows) {
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 7) continue;
+
+            const rankEl = cells[0].querySelector('.jmtd-table-formatter-rank-number');
+            const rank = rankEl ? rankEl.textContent.trim() : cells[0].textContent.trim();
+
+            const shopLink = cells[1].querySelector('a');
+            const shop = shopLink ? shopLink.textContent.trim() : cells[1].textContent.trim();
+
+            const isMyShop = cells[1].querySelector('.jmti-star-f') !== null;
+
+            const item = {
+                rank: rank,
+                shop: shop,
+                amount: cells[2]?.textContent?.trim() || '',
+                orders: cells[3]?.textContent?.trim() || '',
+                followers: cells[4]?.textContent?.trim() || '',
+                visitors: cells[5]?.textContent?.trim() || '',
+                searchClicks: cells[6]?.textContent?.trim() || '',
+                isMyShop: isMyShop
+            };
+
+            result.push(item);
+            if (isMyShop) {
+                myRank = item;
+            }
+        }
+
+        // 提取排名环比
+        const rankChangeEl = document.querySelector('.rank-change');
+        let rankChange = '';
+        if (rankChangeEl) {
+            const arrow = rankChangeEl.querySelector('.jmti-arrow-down, .jmti-arrow-up');
+            const arrowClass = arrow ? arrow.className : '';
+            const direction = arrowClass.includes('arrow-down') ? '下降' : (arrowClass.includes('arrow-up') ? '上升' : '');
+            const match = rankChangeEl.textContent.match(/排名环比[：:]?\\D*(\\d+)/);
+            const num = match ? match[1] : '';
+            if (direction && num) {
+                rankChange = direction + num + '位';
+            }
+        }
+
+        return JSON.stringify({ rows: result, myRank: myRank, rankChange: rankChange });
+    }""", "{}")
+
+    ranking = {"rows": [], "myRank": None}
+    if isinstance(ranking_data, str):
+        try:
+            ranking = json.loads(ranking_data)
+        except json.JSONDecodeError:
+            ranking = {"rows": [], "myRank": None}
+
+    myRank = ranking.get("myRank") or {}
+    allRows = ranking.get("rows", [])
+
+    print(f"  本店排名: {myRank.get('rank', 'N/A')}, 商家: {myRank.get('shop', 'N/A')}")
+    print(f"  共抓取商家数: {len(allRows)}")
+
+    return {
+        "rank": myRank.get("rank", "N/A") if myRank else "N/A",
+        "shop": myRank.get("shop", "N/A") if myRank else "N/A",
+        "amount": myRank.get("amount", "N/A") if myRank else "N/A",
+        "orders": myRank.get("orders", "N/A") if myRank else "N/A",
+        "followers": myRank.get("followers", "N/A") if myRank else "N/A",
+        "visitors": myRank.get("visitors", "N/A") if myRank else "N/A",
+        "searchClicks": myRank.get("searchClicks", "N/A") if myRank else "N/A",
+        "rankChange": ranking.get("rankChange", ""),
+        "all_rows": allRows,
+    }
+
+
+def step_merchant_ranking(page: Page, industry: str):
+    """采集行业商家榜单（新版商智表格），支持多类目"""
+    print("[Step 5] 采集商家榜单...")
+
+    page.goto("https://sz.jd.com/szweb/sz/view/industryMarket/venderRanks.html",
+              wait_until="networkidle")
+    page.wait_for_timeout(8000)
+
+    # 确认行业分类
+    current_industry = extract(page, """() => {
+        const el = document.querySelector('.industry-name, .category-current, [class*="industry"]');
+        return el?.textContent?.trim();
+    }""")
+
+    if current_industry and industry not in current_industry:
+        safe_click(page, industry)
+        page.wait_for_timeout(3000)
+
+    # 切换到按周查询
+    safe_click(page, "按天查询")
+    page.wait_for_timeout(3000)
+    safe_click(page, "按周查询")
+    page.wait_for_timeout(5000)
+
+    select_last_week_in_old_datepicker(page, "merchant_ranking")
+    page.wait_for_timeout(8000)
+
+    # 采集默认类目（男包）
+    results = {}
+    results["男包"] = extract_ranking_table(page)
+
+    # 采集其他子类目（级联下拉：先点父类目展开，再点子类目）
+    extra_categories = [
+        ("男士双肩包", "男包"),
+        ("男士单肩/斜挎包", "男包"),
+        ("电脑包", "功能箱包"),
+    ]
+    for cat, parent in extra_categories:
+        switch_category(page, cat, parent)
+        # 切换类目后可能需要重新选择上周
+        page.wait_for_timeout(3000)
+        safe_click(page, "按周查询")
+        page.wait_for_timeout(2500)
+        select_last_week_in_old_datepicker(page, f"merchant_ranking_{cat}")
+        page.wait_for_timeout(8000)
+        results[cat] = extract_ranking_table(page)
+
+    return results
+
+
+# ── Step 6: 京准通-非全站营销 ─────────────────────────────────────────────────
+
+def switch_account_filter(page):
+    """取消勾选'自有账户'，勾选'授权账户'，确定"""
+    print("  配置账号筛选...")
+
+    # 1. 点击已选账号区域，打开筛选弹窗
+    try:
+        page.evaluate("""() => {
+            const el = document.querySelector('.checked-filters, .filter-index-item');
+            if (el) el.click();
+        }""")
+        page.wait_for_timeout(2000)
+    except Exception as e:
+        print(f"    [WARN] 点击账号筛选失败: {e}")
+        return
+
+    # 2. 取消勾选"自有账户"
+    try:
+        page.evaluate("""() => {
+            const wrappers = document.querySelectorAll('.jad-checkbox-wrapper, label');
+            for (const w of wrappers) {
+                if (w.textContent.includes('\u81ea\u6709\u8d26\u6237')) {
+                    const cb = w.querySelector('input[type="checkbox"]');
+                    if (cb && cb.checked) cb.click();
+                }
+            }
+        }""")
+        print("    [OK] 已取消勾选'自有账户'")
+        page.wait_for_timeout(500)
+    except Exception as e:
+        print(f"    [WARN] 取消自有账户失败: {e}")
+
+    # 3. 勾选"授权账户"
+    try:
+        page.evaluate("""() => {
+            const wrappers = document.querySelectorAll('.jad-checkbox-wrapper, label');
+            for (const w of wrappers) {
+                if (w.textContent.includes('\u6388\u6743\u8d26\u6237')) {
+                    const cb = w.querySelector('input[type="checkbox"]');
+                    if (cb && !cb.checked) cb.click();
+                }
+            }
+        }""")
+        print("    [OK] 已勾选'授权账户'")
+        page.wait_for_timeout(500)
+    except Exception as e:
+        print(f"    [WARN] 勾选授权账户失败: {e}")
+
+    # 4. 点击弹窗内的确定按钮（.filter-bottom 里的 jad-btn-primary）
+    try:
+        page.evaluate("""() => {
+            const btn = document.querySelector('.filter-bottom button.jad-btn-primary');
+            if (btn) btn.click();
+        }""")
+        print("    [OK] 已点击弹窗确定按钮")
+        page.wait_for_timeout(3000)
+    except Exception as e:
+        print(f"    [WARN] 点击弹窗确定失败: {e}")
+        page.keyboard.press("Escape")
+        page.wait_for_timeout(1000)
+
+
+def step_jzt_non_full_site(page: Page, caliber_days: int):
+    """采集京准通非全站营销数据"""
+    print("[Step 6] 采集京准通-非全站营销...")
+
+    page.goto("https://jzt.jd.com/custom-report/#/overview", wait_until="domcontentloaded")
+    page.wait_for_timeout(8000)
+
+    # 点击非全站营销标签
+    safe_click(page, "非全站营销")
+    page.wait_for_timeout(3000)
+
+    # 配置账号筛选
+    switch_account_filter(page)
+
+    # 配置口径：设置为7天
+    cal_btn = page.query_selector('button:has-text("天")')
+    if cal_btn:
+        cal_btn.click()
+        page.wait_for_timeout(2000)
+
+    page.evaluate(f"""() => {{
+        const radios = document.querySelectorAll('.jad-radio-wrapper');
+        for (const r of radios) {{
+            if (r.textContent.trim() === '{caliber_days}天') {{
+                r.click();
+                return;
+            }}
+        }}
+    }}""")
+    page.wait_for_timeout(1000)
+    page.keyboard.press("Escape")
+    page.wait_for_timeout(1000)
+
+    # 点击页面上的"查询"按钮
+    try:
+        page.evaluate("""() => {
+            const btns = document.querySelectorAll('button.jad-btn-primary');
+            for (const b of btns) {
+                if (b.textContent.trim() === '\u67e5\u8be2') {
+                    b.click();
+                    return true;
+                }
+            }
+            return false;
+        }""")
+        print("  [OK] 已点击查询按钮")
+        page.wait_for_timeout(10000)
+    except Exception as e:
+        print(f"  [WARN] 点击查询失败: {e}")
+
+    # 提取数据：花费(cost)、投产比(ROI)、点击率(CTR) + 涨跌对比
+    data = extract(page, """() => {
+        const result = {};
+        const cards = document.querySelectorAll('.cards li, .cards .checked');
+        for (const card of cards) {
+            const label = card.querySelector('.jad-checkbox-label span')?.textContent?.trim();
+            const value = card.querySelector('.current-number')?.textContent?.trim();
+            const siblings = card.querySelectorAll('.current-number ~ span');
+            let change = '';
+            for (const s of siblings) {
+                const txt = s.textContent.trim();
+                if (txt && txt !== '') { change = txt; break; }
+            }
+            const compare = card.querySelector('.compare-number')?.textContent?.trim();
+            if (label && value) {
+                result[label] = { value: value, change: change, compare: compare };
+            }
+        }
+        return JSON.stringify(result);
+    }""", "{}")
+
+    if isinstance(data, str):
+        try:
+            data = json.loads(data)
+        except json.JSONDecodeError:
+            data = {}
+
+    # 兜底：根据 checkbox value 直接找卡片
+    for key, val_attr in [('cost', 'cost'), ('ROI', 'ROI'), ('CTR', 'CTR')]:
+        if not data.get(key):
+            raw = extract(page, f"""() => {{
+                const cards = document.querySelectorAll('.cards li');
+                for (const card of cards) {{
+                    const input = card.querySelector('input[value="{val_attr}"]');
+                    if (input) {{
+                        const value = card.querySelector('.current-number')?.textContent?.trim() || 'N/A';
+                        const siblings = card.querySelectorAll('.current-number ~ span');
+                        let change = '';
+                        for (const s of siblings) {{
+                            const txt = s.textContent.trim();
+                            if (txt && txt !== '') {{ change = txt; break; }}
+                        }}
+                        const compare = card.querySelector('.compare-number')?.textContent?.trim() || '';
+                        return JSON.stringify({{value: value, change: change, compare: compare}});
+                    }}
+                }}
+                return JSON.stringify({{value: 'N/A', change: '', compare: ''}});
+            }}""")
+            try:
+                data[key] = json.loads(raw) if isinstance(raw, str) else {"value": "N/A", "change": "", "compare": ""}
+            except Exception:
+                data[key] = {"value": "N/A", "change": "", "compare": ""}
+
+    try:
+        print(f"  京准通数据: {json.dumps(data, ensure_ascii=False)}")
+    except UnicodeEncodeError:
+        print(f"  京准通数据: {json.dumps(data, ensure_ascii=True)}")
+    return data
+
+
+# ── Step 7: 京准通-全站营销 ──────────────────────────────────────────────────
+
+def step_jzt_full_site(page: Page, caliber_days: int):
+    """采集京准通全站营销数据（流程与非全站相同）"""
+    print("[Step 7] 采集京准通-全站营销...")
+
+    # 点击全站营销标签（精确选择器）
+    try:
+        page.evaluate("""() => {
+            const tabs = document.querySelectorAll('.jad-tabs-nav-tab-no-icon');
+            for (const tab of tabs) {
+                if (tab.textContent.trim() === '\u5168\u7ad9\u8425\u9500') {
+                    tab.click();
+                    return true;
+                }
+            }
+            return false;
+        }""")
+        print("  [OK] 已点击全站营销标签")
+    except Exception as e:
+        print(f"  [WARN] 点击全站营销标签失败: {e}")
+    page.wait_for_timeout(8000)
+
+    # 配置账号筛选（确保操作当前可见的弹窗）
+    switch_account_filter(page)
+
+    # 配置口径：点击口径下拉，选7天
+    try:
+        page.evaluate("""() => {
+            const btn = document.querySelector('button.jad-btn-default');
+            if (btn && btn.textContent.includes('\u5929')) btn.click();
+        }""")
+        page.wait_for_timeout(2000)
+        page.evaluate(f"""() => {{
+            const radios = document.querySelectorAll('.jad-radio-wrapper');
+            for (const r of radios) {{
+                if (r.textContent.trim() === '{caliber_days}\u5929') {{
+                    r.click();
+                    return;
+                }}
+            }}
+        }}""")
+        page.wait_for_timeout(1000)
+        page.keyboard.press("Escape")
+        page.wait_for_timeout(1000)
+    except Exception:
+        pass
+
+    # 点击页面上的"查询"按钮
+    try:
+        page.evaluate("""() => {
+            const btns = document.querySelectorAll('button.jad-btn-primary');
+            for (const b of btns) {
+                if (b.textContent.trim() === '\u67e5\u8be2') {
+                    b.click();
+                    return true;
+                }
+            }
+            return false;
+        }""")
+        print("  [OK] 已点击查询按钮")
+        page.wait_for_timeout(10000)
+    except Exception as e:
+        print(f"  [WARN] 点击查询失败: {e}")
+
+    # 提取数据：花费(cost)、投产比(ROI)、点击率(CTR) + 对比涨跌
+    data = extract(page, """() => {
+        const result = {};
+        const cards = document.querySelectorAll('.cards li, .cards .checked');
+        for (const card of cards) {
+            const label = card.querySelector('.jad-checkbox-label span')?.textContent?.trim();
+            const value = card.querySelector('.current-number')?.textContent?.trim();
+            // 涨跌：current-number 后面的兄弟 span
+            const siblings = card.querySelectorAll('.current-number ~ span');
+            let change = '';
+            for (const s of siblings) {
+                const txt = s.textContent.trim();
+                if (txt && txt !== '') { change = txt; break; }
+            }
+            // 对比数值
+            const compare = card.querySelector('.compare-number')?.textContent?.trim();
+            if (label && value) {
+                result[label] = { value: value, change: change, compare: compare };
+            }
+        }
+        return JSON.stringify(result);
+    }""", "{}")
+
+    if isinstance(data, str):
+        try:
+            data = json.loads(data)
+        except json.JSONDecodeError:
+            data = {}
+
+    # 兜底：根据 checkbox value 直接找卡片（全站指标：cost, swaClicks, swaOrderROI, swaOrderSum）
+    for key, val_attr in [('cost', 'cost'), ('swaClicks', 'swaClicks'), ('swaOrderROI', 'swaOrderROI'), ('swaOrderSum', 'swaOrderSum')]:
+        if not data.get(key):
+            raw = extract(page, f"""() => {{
+                const cards = document.querySelectorAll('.cards li');
+                for (const card of cards) {{
+                    const input = card.querySelector('input[value="{val_attr}"]');
+                    if (input) {{
+                        const value = card.querySelector('.current-number')?.textContent?.trim() || 'N/A';
+                        const siblings = card.querySelectorAll('.current-number ~ span');
+                        let change = '';
+                        for (const s of siblings) {{
+                            const txt = s.textContent.trim();
+                            if (txt && txt !== '') {{ change = txt; break; }}
+                        }}
+                        const compare = card.querySelector('.compare-number')?.textContent?.trim() || '';
+                        return JSON.stringify({{value: value, change: change, compare: compare}});
+                    }}
+                }}
+                return JSON.stringify({{value: 'N/A', change: '', compare: ''}});
+            }}""")
+            try:
+                data[key] = json.loads(raw) if isinstance(raw, str) else {"value": "N/A", "change": "", "compare": ""}
+            except Exception:
+                data[key] = {"value": "N/A", "change": "", "compare": ""}
+
+    # 字段映射：swaOrderROI -> ROI，swaClicks -> 全站点击量，swaOrderSum -> 全站交易额
+    if data.get('swaOrderROI'):
+        data['ROI'] = data['swaOrderROI']
+    if data.get('swaClicks'):
+        data['全站点击量'] = data['swaClicks']
+    if data.get('swaOrderSum'):
+        data['全站交易额'] = data['swaOrderSum']
+
+    try:
+        print(f"  全站营销数据: {json.dumps(data, ensure_ascii=False)}")
+    except UnicodeEncodeError:
+        print(f"  全站营销数据: {json.dumps(data, ensure_ascii=True)}")
+    return data
+
+
+# ── Step 8: 生成报告 ─────────────────────────────────────────────────────────
+
+def generate_report(config, star, transaction, ranking, jzt_non_full, jzt_full):
+    """生成 Markdown 汇总报告"""
+    print("[Step 8] 生成汇总报告...")
+
+    shop = config["shop"]
+    industry = config["industry"]
+    caliber_days = config["caliber_days"]
+    output_dir = config["output"]
+    today = datetime.now().strftime("%Y-%m-%d")
+    last_monday, last_sunday = get_last_week_range()
+    week_range = f"{last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}"
+
+    # 构建交易表格行（含环比、同行均值）
+    trans_rows = ""
+    for name in ["成交金额", "商品访客数", "成交客户数", "成交转化率", "成交单量", "客单价"]:
+        item = transaction.get(name, {})
+        val = item.get("value", "N/A") if isinstance(item, dict) else item
+        hb = item.get("环比", "N/A") if isinstance(item, dict) else "N/A"
+        same = item.get("同行同级均值", "N/A") if isinstance(item, dict) else "N/A"
+        upper = item.get("同行上级均值", "N/A") if isinstance(item, dict) else "N/A"
+        trans_rows += f"| {name} | {val} | {hb} | {same} | {upper} |\n"
+
+    # 构建商家榜单对比表格（支持多类目）
+    def build_ranking_section(cat_name, cat_data):
+        if not cat_data:
+            return f"### 类目：{cat_name}\n\n暂无数据。\n\n"
+        rows = ""
+        for item in cat_data.get("all_rows", [])[:20]:  # 取前20名
+            is_my = item.get("isMyShop", False)
+            shop_name = item.get("shop", "")
+            marker = " ⭐" if is_my else ""
+            rows += (f"| {item.get('rank', '')} | {shop_name}{marker} "
+                     f"| {item.get('amount', '')} | {item.get('orders', '')} "
+                     f"| {item.get('followers', '')} | {item.get('visitors', '')} "
+                     f"| {item.get('searchClicks', '')} |\n")
+        rank_change = cat_data.get('rankChange', '')
+        rank_change_str = f" | 排名环比：{rank_change}" if rank_change else ""
+        return (
+            f"### 类目：{cat_name}\n\n"
+            f"**本店排名：** 第{cat_data.get('rank', 'N/A')}名 — {cat_data.get('shop', 'N/A')}{rank_change_str}\n"
+            f"- 成交金额：{cat_data.get('amount', 'N/A')} | 成交单量：{cat_data.get('orders', 'N/A')} | 关注人数：{cat_data.get('followers', 'N/A')} | 访客数：{cat_data.get('visitors', 'N/A')} | 搜索点击：{cat_data.get('searchClicks', 'N/A')}\n\n"
+            f"| 排名 | 商家 | 成交金额 | 成交单量 | 关注人数 | 访客数 | 搜索点击 |\n"
+            f"|------|------|----------|----------|----------|--------|----------|\n"
+            f"{rows}\n"
+        )
+
+    # 非全站营销表格（cost=付费占比, ROI=ROI, CTR=点击率，含涨跌对比）
+    non_full_rows = ""
+    non_full_map = {
+        "付费占比 (cost)": jzt_non_full.get("cost", {}),
+        "ROI": jzt_non_full.get("ROI", {}),
+        "点击率 (CTR)": jzt_non_full.get("CTR", {}),
+    }
+    for label, item in non_full_map.items():
+        if isinstance(item, dict):
+            val = item.get("value", "N/A")
+            change = item.get("change", "")
+            compare = item.get("compare", "")
+            non_full_rows += f"| {label} | {val} | {change} ({compare}) |\n"
+        else:
+            non_full_rows += f"| {label} | {item} | |\n"
+
+    # 全站营销概览表格（含付费占比、ROI、点击量、交易额、涨跌对比）
+    full_rows = ""
+    full_map = {
+        "付费占比 (cost)": jzt_full.get("cost", {}),
+        "ROI": jzt_full.get("ROI", {}),
+        "全站点击量": jzt_full.get("全站点击量", {}),
+        "全站交易额": jzt_full.get("全站交易额", {}),
+    }
+    for label, item in full_map.items():
+        if isinstance(item, dict):
+            val = item.get("value", "N/A")
+            change = item.get("change", "")
+            compare = item.get("compare", "")
+            full_rows += f"| {label} | {val} | {change} ({compare}) |\n"
+        else:
+            full_rows += f"| {label} | {item} | |\n"
+
+    # 汇总表
+    def get_val(obj):
+        return obj.get("value", "N/A") if isinstance(obj, dict) else obj
+
+    non_full_cost = get_val(jzt_non_full.get("cost", "N/A"))
+    non_full_roi = get_val(jzt_non_full.get("ROI", "N/A"))
+    non_full_ctr = get_val(jzt_non_full.get("CTR", "N/A"))
+
+    full_cost_val = get_val(jzt_full.get("cost", "N/A"))
+    full_roi_val = get_val(jzt_full.get("ROI", "N/A"))
+    full_clicks_val = get_val(jzt_full.get("全站点击量", "N/A"))
+    full_sum_val = get_val(jzt_full.get("全站交易额", "N/A"))
+
+    report = f"""# {shop} 数据汇总
+
+**数据采集日期：** {today}
+**店铺名称：** {shop}
+**数据周期：** {week_range}
+
+## 一、店铺星级
+
+| 项目 | 详情 |
+|------|------|
+| 当前星级 | {star} |
+
+## 二、交易概况（上周 {week_range}）
+
+| 指标 | 数值 | 环比 | 同行同级均值 | 同行上级均值 |
+|------|------|------|-------------|-------------|
+{trans_rows}
+## 三、商家榜单（行业：{industry}，上周）
+
+{''.join(build_ranking_section(cat, data) for cat, data in ranking.items())}
+## 四、非全站营销（口径：{caliber_days}天）
+
+| 指标 | 数值 | 环比 |
+|------|------|------|
+{non_full_rows}
+## 五、全站营销（口径：{caliber_days}天）
+
+| 指标 | 数值 | 涨跌对比 |
+|------|------|----------|
+{full_rows}
+## 六、汇总
+
+| 渠道 | 付费占比 | ROI | 点击率 | 点击量 | 交易额 |
+|------|----------|-----|--------|--------|--------|
+| 非全站 | {non_full_cost} | {non_full_roi} | {non_full_ctr} | -- | -- |
+| 全站 | {full_cost_val} | {full_roi_val} | -- | {full_clicks_val} | {full_sum_val} |
+| 合计 | -- | -- | -- | -- | -- |
+"""
+
+    # 写入 MD 报告
+    os.makedirs(output_dir, exist_ok=True)
+    filename = f"{shop}_店铺数据汇总_{today}.md"
+    filepath = os.path.join(output_dir, filename)
+
+    with open(filepath, "w", encoding="utf-8") as f:
+        f.write(report)
+
+    # 结构化指标 JSON（供 base_writer 节点写入多维表格使用）
+    metrics = {
+        "shop": shop,
+        "industry": industry,
+        "caliber_days": caliber_days,
+        "data_date": today,
+        "week_start": last_monday.strftime("%Y-%m-%d"),
+        "week_end": last_sunday.strftime("%Y-%m-%d"),
+        "week_range": week_range,
+        "star_rating_raw": star,
+        "transaction": transaction,
+        "ranking": ranking,
+        "jzt_non_full": jzt_non_full,
+        "jzt_full": jzt_full,
+    }
+    metrics_filename = f"{shop}_指标数据_{today}.json"
+    metrics_filepath = os.path.join(output_dir, metrics_filename)
+    try:
+        with open(metrics_filepath, "w", encoding="utf-8") as f:
+            json.dump(metrics, f, ensure_ascii=False, indent=2)
+        print(f"  指标数据已保存: {metrics_filepath}")
+    except Exception as e:
+        print(f"  [WARN] 保存指标 JSON 失败: {e}")
+
+    # 多类目原始数据保存为 JSON（供后续使用）
+    if isinstance(ranking, dict):
+        json_filename = f"{shop}_榜单数据_{today}.json"
+        json_filepath = os.path.join(output_dir, json_filename)
+        try:
+            with open(json_filepath, "w", encoding="utf-8") as f:
+                json.dump(ranking, f, ensure_ascii=False, indent=2)
+            print(f"  榜单原始数据已保存: {json_filepath}")
+        except Exception as e:
+            print(f"  [WARN] 保存 JSON 失败: {e}")
+
+    print(f"  报告已保存: {filepath}")
+    return filepath
+
+
+# ── 主流程 ────────────────────────────────────────────────────────────────────
+
+def main():
+    parser = argparse.ArgumentParser(description="京东店铺数据采集脚本")
+    parser.add_argument("--shop", default=DEFAULT_CONFIG["shop"], help="店铺账号名")
+    parser.add_argument("--password", default=DEFAULT_CONFIG["password"], help="登录密码")
+    parser.add_argument("--output", default=DEFAULT_CONFIG["output"], help="报告输出目录")
+    parser.add_argument("--industry", default=DEFAULT_CONFIG["industry"], help="行业分类")
+    parser.add_argument("--caliber", type=int, default=DEFAULT_CONFIG["caliber_days"], help="口径天数")
+    args = parser.parse_args()
+    try:
+        args.output = str(
+            resolve_raw_output(args.output, default=DEFAULT_CONFIG["output"])
+        )
+    except ValueError as exc:
+        parser.error(str(exc))
+
+    config = {
+        "shop": args.shop,
+        "password": args.password,
+        "output": args.output,
+        "industry": args.industry,
+        "caliber_days": args.caliber,
+    }
+
+    print("=" * 60)
+    print("京东店铺数据采集")
+    print(f"  店铺: {config['shop']}")
+    print(f"  行业: {config['industry']}")
+    print(f"  输出: {config['output']}")
+    print("=" * 60)
+
+    last_monday, last_sunday = get_last_week_range()
+    print(f"  上周: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+    print()
+
+    with sync_playwright() as p:
+        # 启动浏览器（headed模式）
+        browser = p.chromium.launch(headless=False)
+        context = browser.new_context(
+            viewport={"width": 1920, "height": 1080},
+            locale="zh-CN",
+        )
+        page = context.new_page()
+
+        try:
+            # Step 1: 登录
+            step_login(page, config["shop"], config["password"])
+
+            # Step 2: 进入商智
+            sz_page = step_enter_shangzhi(context, page)
+
+            # Step 3: 采集店铺星级与交易数据（新版商智先选时间再抓取）
+            star, transaction = step_shop_star_and_trade(sz_page)
+
+            # Step 4: 采集旧版交易概况-客单价
+            deal_summary = step_deal_summary(sz_page)
+            # 合并到 transaction 中
+            if deal_summary.get("客单价") and deal_summary["客单价"].get("value") != "N/A":
+                cp = deal_summary["客单价"]
+                transaction["客单价"] = {
+                    "value": cp["value"],
+                    "环比": cp.get("change", "N/A"),
+                    "同行同级均值": "N/A",
+                    "同行上级均值": "N/A"
+                }
+
+            # Step 5: 商家榜单
+            ranking = step_merchant_ranking(sz_page, config["industry"])
+
+            # Step 6: 京准通-非全站营销
+            jzt_non_full = step_jzt_non_full_site(sz_page, config["caliber_days"])
+
+            # Step 7: 京准通-全站营销
+            jzt_full = step_jzt_full_site(sz_page, config["caliber_days"])
+
+            # Step 8: 生成报告
+            report_path = generate_report(
+                config, star, transaction, ranking, jzt_non_full, jzt_full
+            )
+
+            print()
+            print("=" * 60)
+            print(f"采集完成！报告: {report_path}")
+            print("=" * 60)
+
+        except Exception as e:
+            print(f"\n[ERROR] 采集过程中出错: {e}")
+            import traceback
+            traceback.print_exc()
+
+        finally:
+            context.close()
+            browser.close()
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/jd_peer_store_data_collector.py b/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/jd_peer_store_data_collector.py
new file mode 100644
index 0000000..ccab579
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/jd_peer_store_data_collector.py
@@ -0,0 +1,1026 @@
+# ruff: noqa: F841
+"""
+京东商家榜单竞品数据采集脚本
+登录京东商家平台,进入商智,跳转商家榜单,搜索竞品店铺,采集各类目数据。
+
+依赖:
+    pip install playwright
+    playwright install chromium
+
+用法:
+    python jd_login.py
+    python jd_login.py --shop "店铺名" --password "密码" --output "输出目录"
+"""
+
+import argparse
+import json
+import os
+from datetime import datetime, timedelta
+
+from playwright.sync_api import BrowserContext, Page, sync_playwright
+from playwright.sync_api import TimeoutError as PWTimeoutError
+
+from gyxx_flow.modules.shop_intelligence.runtime.collectors.jd_data_collector import (
+    close_profit_plan_popup_if_present,
+    select_last_week_in_old_datepicker,
+)
+from gyxx_flow.modules.shop_intelligence.runtime.config import JD_PEER_STORE_OUTPUT
+from gyxx_flow.modules.shop_intelligence.runtime.paths import resolve_raw_output
+
+# ── 配置 ─────────────────────────────────────────────────────────────────────
+
+DEFAULT_SHOP = "光影行星研发"
+DEFAULT_PASSWORD = os.getenv("JD_PASSWORD", "")
+DEFAULT_OUTPUT = JD_PEER_STORE_OUTPUT
+
+SHOPS = [
+    "Bellroy官方旗舰店",
+    "tomtoc官方旗舰店",
+    "探迹者TAJEZZO官方旗舰店",
+    "BAGSMART旗舰店",
+    "GROTTO个乐官方旗舰店",
+    "NIID箱包旗舰店",
+    "光影行星 GYXX官方旗舰店",
+]
+
+CATEGORIES = [
+    ("男包", None),
+    ("男士双肩包", "男包"),
+    ("男士单肩/斜挎包", "男包"),
+    ("电脑包", "功能箱包"),
+    ("男士腰包/胸包", "男包"),
+    ("旅行包", "功能箱包"),
+    ("休闲运动包", "功能箱包"),
+]
+
+CATEGORY_OUTPUT_NAMES = {
+    "男包": "男士包袋",
+}
+
+# 新品采集配置
+NEW_PRODUCT_SHOPS = [
+    "Bellroy",
+    "tomtoc",
+    "探迹者",
+    "BAGSMART",
+    "GROTTO",
+    "NIID",
+]
+
+NEW_PRODUCT_CATEGORIES = [
+    ("男士双肩包", "男包"),
+    ("男士单肩/斜挎包", "男包"),
+    ("电脑包", "功能箱包"),
+]
+
+
+# ── 工具函数 ──────────────────────────────────────────────────────────────────
+
+def category_output_name(category_name: str) -> str:
+    return CATEGORY_OUTPUT_NAMES.get(category_name, category_name)
+
+
+def wait_for_navigation(page: Page, timeout=15000):
+    try:
+        page.wait_for_load_state("networkidle", timeout=timeout)
+    except PWTimeoutError:
+        pass
+
+
+def extract(page: Page, js_expr: str, default=None):
+    try:
+        result = page.evaluate(js_expr)
+        if result is None:
+            return default
+        return str(result).strip()
+    except Exception:
+        return default
+
+
+def safe_click(page: Page, text: str, timeout=5000):
+    try:
+        page.click(f"text={text}", timeout=timeout)
+        return True
+    except PWTimeoutError:
+        print(f"  [!] 未找到文本: {text}")
+        return False
+
+
+def get_last_week_range():
+    today = datetime.now()
+    this_monday = today - timedelta(days=today.weekday())
+    last_monday = this_monday - timedelta(days=7)
+    last_sunday = this_monday - timedelta(days=1)
+    return last_monday, last_sunday
+
+
+def select_last_week_any_day(page):
+    """点击日期选择器 → 按周查询 → 点上周任意一天（自动选中整周）"""
+    print("  选择上周时间范围...")
+
+    # 1. 点击日期选择器（显示"按天查询..."的元素）
+    try:
+        date_picker = page.locator('.jmtd-date-picker .jmtd-base-input-top').first
+        date_picker.wait_for(state="visible", timeout=5000)
+        date_picker.click()
+        print("    [OK] 点击日期选择器")
+    except Exception as e:
+        print(f"    [WARN] 点击日期选择器失败: {e}")
+        # 兜底：点击包含"按天查询"文本的元素
+        try:
+            page.locator('text=按天查询').first.click()
+            print("    [OK] 使用文本定位点击日期选择器")
+        except Exception:
+            pass
+    page.wait_for_timeout(2000)
+
+    # 2. 点击"按周查询"
+    try:
+        week_switch = page.locator('.jmtd-date-picker-switch-panel-item', has_text="按周查询").first
+        week_switch.wait_for(state="visible", timeout=5000)
+        week_switch.click()
+        print("    [OK] 点击按周查询")
+    except Exception as e:
+        print(f"    [WARN] 点击按周查询失败: {e}")
+        # 兜底：用 JS 点击
+        page.evaluate("""() => {
+            const items = document.querySelectorAll('.jmtd-date-picker-switch-panel-item');
+            for (const item of items) {
+                if (item.textContent.includes('按周查询')) {
+                    item.click();
+                    return true;
+                }
+            }
+            return false;
+        }""")
+    page.wait_for_timeout(2000)
+
+    # 3. 等待日历弹窗并选择上周某一天
+    calendar = page.locator('.jmtd-date-picker-popup, .jmtd-date-picker-panel, [class*="picker-panel"], [class*="calendar-panel"]')
+    try:
+        calendar.wait_for(state="visible", timeout=5000)
+    except Exception:
+        print("  [WARN] 未检测到日历弹窗,尝试继续...")
+        page.wait_for_timeout(3000)
+
+    select_last_week_in_old_datepicker(page, "peer_store")
+
+    try:
+        calendar.wait_for(state="hidden", timeout=5000)
+    except Exception:
+        pass
+    page.wait_for_timeout(5000)
+    print("  已切换到上周数据")
+
+
+def reselect_last_week(page):
+    """切换类目后重新选择上周日期（点击日期选择器 → 按周查询 → 点日期）"""
+    # 重新点击日期选择器
+    try:
+        date_picker = page.locator('.jmtd-date-picker .jmtd-base-input-top').first
+        date_picker.click()
+    except Exception:
+        page.locator('text=按天查询').first.click()
+    page.wait_for_timeout(1500)
+
+    # 点击按周查询
+    try:
+        week_switch = page.locator('.jmtd-date-picker-switch-panel-item', has_text="按周查询").first
+        week_switch.click()
+    except Exception:
+        page.evaluate("""() => {
+            const items = document.querySelectorAll('.jmtd-date-picker-switch-panel-item');
+            for (const item of items) {
+                if (item.textContent.includes('按周查询')) {
+                    item.click();
+                    return true;
+                }
+            }
+        }""")
+    page.wait_for_timeout(1500)
+
+    select_last_week_in_old_datepicker(page, "peer_store_reselect")
+    page.wait_for_timeout(8000)
+
+
+def switch_category(page: Page, category_name: str, parent_category: str = None):
+    """切换商家榜单类目,支持级联下拉"""
+    label = f"{parent_category} → {category_name}" if parent_category else category_name
+    print(f"  切换类目到: {label}")
+    try:
+        page.evaluate("""() => {
+            const el = document.querySelector('.jmtd-base-input-content, .jmtd-select-input');
+            if (el) el.click();
+        }""")
+        page.wait_for_timeout(2000)
+
+        if parent_category:
+            try:
+                parent_locator = page.locator('.jmtd-dropdown-content .jmtd-label', has_text=parent_category).first
+                parent_locator.hover(timeout=3000)
+                page.wait_for_timeout(2000)
+            except Exception as e:
+                print(f"    [WARN] 悬停父类目失败: {e}")
+
+        try:
+            child_locator = page.locator('.jmtd-dropdown-content .jmtd-label', has_text=category_name).first
+            child_locator.click(timeout=3000)
+            page.wait_for_timeout(5000)
+            print(f"    [OK] 已切换到 {label}")
+        except Exception as e:
+            print(f"    [WARN] 点击子类目失败: {e}")
+    except Exception as e:
+        print(f"    [WARN] 切换类目失败: {e}")
+
+
+def extract_ranking_table(page):
+    """提取商家榜单表格数据"""
+    ranking_data = extract(page, """() => {
+        const rows = document.querySelectorAll('.jmtd-table-main tbody tr');
+        const result = [];
+        for (const row of rows) {
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 7) continue;
+
+            // 排名提取：优先数字文本，其次检查 1/2/3 名图标 class
+            let rank = '';
+            const rankNumEl = cells[0].querySelector('.jmtd-table-formatter-rank-number');
+            if (rankNumEl) {
+                rank = rankNumEl.textContent.trim();
+            }
+            if (!rank) {
+                const rankIconEl = cells[0].querySelector('.jmtd-table-formatter-rank');
+                if (rankIconEl) {
+                    const iconMatch = rankIconEl.className.match(/jmtd-table-formatter-rank-icon-(\\d)/);
+                    if (iconMatch) {
+                        rank = iconMatch[1];
+                    }
+                }
+            }
+            if (!rank) {
+                rank = cells[0].textContent.trim();
+            }
+
+            const shopLink = cells[1].querySelector('a');
+            const shop = shopLink ? shopLink.textContent.trim() : cells[1].textContent.trim();
+
+            const item = {
+                rank: rank,
+                shop: shop,
+                amount: cells[2]?.textContent?.trim() || '',
+                orders: cells[3]?.textContent?.trim() || '',
+                followers: cells[4]?.textContent?.trim() || '',
+                visitors: cells[5]?.textContent?.trim() || '',
+                searchClicks: cells[6]?.textContent?.trim() || '',
+            };
+            result.push(item);
+        }
+        return JSON.stringify(result);
+    }""", "[]")
+
+    rows = []
+    if isinstance(ranking_data, str):
+        try:
+            rows = json.loads(ranking_data)
+        except json.JSONDecodeError:
+            rows = []
+    return rows
+
+
+def find_shop_in_rows(rows, shop_name):
+    """从表格行中按店铺名匹配"""
+    for row in rows:
+        if shop_name in row.get("shop", ""):
+            return row
+    return None
+
+
+# ── 新品数据采集 ─────────────────────────────────────────────────────────────
+
+def step_goto_product_ranks(page: Page):
+    """跳转到商品榜单页面"""
+    print("[Step 6] 跳转到商品榜单页面...")
+    page.goto("https://sz.jd.com/szweb/sz/view/industryMarket/productRanks.html",
+              wait_until="networkidle")
+    page.wait_for_timeout(8000)
+    print("  已进入商品榜单页面")
+
+
+def switch_to_new_product_popularity(page: Page):
+    """切换到新品人气榜单标签"""
+    print("  切换到新品人气榜单...")
+    try:
+        # 点击"新品人气榜单"标签
+        tab = page.locator('.jmtd-tabs-tab', has_text="新品人气榜单").first
+        tab.wait_for(state="visible", timeout=5000)
+        tab.click()
+        page.wait_for_timeout(5000)
+        print("    [OK] 已切换到新品人气榜单")
+        return True
+    except Exception as e:
+        print(f"    [WARN] 切换标签失败: {e}")
+        # 兜底：JS点击
+        try:
+            page.evaluate("""() => {
+                const tabs = document.querySelectorAll('.jmtd-tabs-tab');
+                for (const tab of tabs) {
+                    if (tab.textContent.includes('新品人气榜单')) {
+                        tab.click();
+                        return true;
+                    }
+                }
+                return false;
+            }""")
+            page.wait_for_timeout(5000)
+            print("    [OK] JS点击切换到新品人气榜单")
+            return True
+        except Exception:
+            return False
+
+
+def select_product_category(page: Page, category_name: str, parent_category: str):
+    """在商品榜单页面选择类目（级联下拉）"""
+    print(f"    选择类目: {parent_category} → {category_name}")
+
+    # 1. 点击类目选择器（第二个 Select，宽度200px的那个）
+    try:
+        category_selectors = page.locator('.jmtd-select[data-component-name="Select"]')
+        # 找到类目选择器（通常包含父类目名的label）
+        category_selector = None
+        for i in range(category_selectors.count()):
+            selector = category_selectors.nth(i)
+            label_text = selector.locator('.jmtd-label').text_content()
+            if parent_category in label_text or label_text in ["男包", "女包", "功能箱包", "配件"]:
+                category_selector = selector
+                break
+
+        if not category_selector:
+            # 兜底：使用第二个select
+            category_selector = category_selectors.nth(1)
+
+        category_selector.click()
+        page.wait_for_timeout(1500)
+    except Exception as e:
+        print(f"    [WARN] 点击类目选择器失败: {e}")
+        return False
+
+    # 2. 悬停父类目
+    try:
+        parent_item = page.locator('.jmtd-dropdown-content .jmtd-label', has_text=parent_category).first
+        parent_item.hover(timeout=3000)
+        page.wait_for_timeout(1500)
+    except Exception as e:
+        print(f"    [WARN] 悬停父类目失败: {e}")
+
+    # 3. 点击子类目
+    try:
+        child_item = page.locator('.jmtd-dropdown-content .jmtd-label', has_text=category_name).first
+        child_item.click(timeout=3000)
+        page.wait_for_timeout(5000)
+        print("    [OK] 类目已切换")
+        return True
+    except Exception as e:
+        print(f"    [WARN] 点击子类目失败: {e}")
+        return False
+
+
+def search_product_by_name(page: Page, brand_name: str):
+    """在商品榜单页面搜索商品名称"""
+    print(f"    搜索品牌: {brand_name}")
+
+    # 找到搜索框 - 尝试多种选择器
+    search_input = None
+    selectors = [
+        '.jmtd-input-input',  # 商品名称输入框的class
+        'input[data-event-name="input"]',
+        '.tool-bar input.jmtd-input-input',
+        '.jmtd-base-input-content input',
+    ]
+
+    for selector in selectors:
+        try:
+            el = page.locator(selector).first
+            if el.is_visible(timeout=2000):
+                search_input = el
+                print(f"    [DEBUG] 使用选择器: {selector}")
+                break
+        except Exception:
+            continue
+
+    if not search_input:
+        # 兜底：用JS查找placeholder包含"商品名称"的input
+        try:
+            search_input = page.evaluate("""() => {
+                const inputs = document.querySelectorAll('input');
+                for (const input of inputs) {
+                    const placeholder = input.closest('.jmtd-base-input')?.querySelector('.jmtd-base-input-placeholder');
+                    if (placeholder && placeholder.textContent.includes('商品名称')) {
+                        return input;
+                    }
+                }
+                return null;
+            }""")
+            if search_input:
+                search_input = page.locator('input').first  # 重新定位
+                print("    [DEBUG] 使用JS找到搜索框")
+        except Exception:
+            pass
+
+    if not search_input:
+        print("    [WARN] 未找到搜索框")
+        return False
+
+    try:
+        search_input.click()
+        page.wait_for_timeout(500)
+        # 清空输入框
+        search_input.fill("")
+        page.wait_for_timeout(300)
+        search_input.fill(brand_name)
+        page.wait_for_timeout(1000)
+        # 按回车搜索
+        search_input.press("Enter")
+        page.wait_for_timeout(8000)
+        print("    [OK] 搜索完成")
+        return True
+    except Exception as e:
+        print(f"    [WARN] 搜索失败: {e}")
+        return False
+
+
+def extract_product_table(page: Page) -> list:
+    """提取商品榜单表格数据"""
+    product_data = page.evaluate("""() => {
+        const rows = document.querySelectorAll('.jmtd-table-main tbody tr');
+        const result = [];
+        for (const row of rows) {
+            const cells = row.querySelectorAll('td');
+            if (cells.length < 7) continue;
+
+            // 排名
+            let rank = '';
+            const rankNumEl = cells[0].querySelector('.jmtd-table-formatter-rank-number');
+            if (rankNumEl) {
+                rank = rankNumEl.textContent.trim();
+            }
+            if (!rank) {
+                const rankIconEl = cells[0].querySelector('.jmtd-table-formatter-rank');
+                if (rankIconEl) {
+                    const iconMatch = rankIconEl.className.match(/jmtd-table-formatter-rank-icon-(\\d)/);
+                    if (iconMatch) {
+                        rank = iconMatch[1];
+                    }
+                }
+            }
+            if (!rank) {
+                rank = cells[0].textContent.trim();
+            }
+
+            // 商品信息
+            const prodInfo = cells[1].querySelector('.prod-info');
+            let productName = '';
+            let shopName = '';
+            if (prodInfo) {
+                const nameLink = prodInfo.querySelector('.prod-name');
+                const shopLink = prodInfo.querySelector('.shop-name');
+                productName = nameLink ? nameLink.textContent.trim() : '';
+                shopName = shopLink ? shopLink.textContent.trim() : '';
+            }
+
+            const item = {
+                rank: rank,
+                productName: productName,
+                shopName: shopName,
+                visitors: cells[2]?.textContent?.trim() || '',
+                amount: cells[3]?.textContent?.trim() || '',
+                orders: cells[4]?.textContent?.trim() || '',
+                searchClicks: cells[5]?.textContent?.trim() || '',
+                followers: cells[6]?.textContent?.trim() || '',
+            };
+            result.push(item);
+        }
+        return result;
+    }""")
+
+    return product_data or []
+
+
+def collect_new_products(page: Page) -> dict:
+    """采集各店铺新品数据"""
+    print("\n[Step 7] 采集新品数据...")
+
+    # 1. 切换到新品人气榜单
+    switch_to_new_product_popularity(page)
+    page.wait_for_timeout(3000)
+
+    # 2. 选择上周日期
+    select_last_week_any_day(page)
+    page.wait_for_timeout(3000)
+
+    results = {}
+
+    # 3. 循环品牌
+    for brand_name in NEW_PRODUCT_SHOPS:
+        print(f"\n{'='*50}")
+        print(f"  品牌: {brand_name}")
+        print(f"{'='*50}")
+
+        # 清空搜索框并输入品牌名
+        clear_and_search_brand(page, brand_name)
+        page.wait_for_timeout(3000)
+
+        brand_data = {}
+
+        # 4. 遍历三个类目
+        for cat_name, parent_cat in NEW_PRODUCT_CATEGORIES:
+            print(f"\n  类目: {parent_cat} → {cat_name}")
+
+            # 选择类目
+            select_product_category(page, cat_name, parent_cat)
+            page.wait_for_timeout(3000)
+
+            # 提取表格数据
+            products = extract_product_table(page)
+
+            if products:
+                # 过滤出属于当前品牌店铺的商品
+                brand_products = []
+                for prod in products:
+                    # 匹配店铺名中包含品牌关键词的
+                    if (brand_name.lower() in prod.get('shopName', '').lower() or
+                        brand_name.lower() in prod.get('productName', '').lower()):
+                        brand_products.append(prod)
+
+                if brand_products:
+                    brand_data[cat_name] = brand_products
+                    print(f"    [OK] 找到 {len(brand_products)} 个新品")
+                else:
+                    brand_data[cat_name] = products  # 保存所有结果供参考
+                    print(f"    [INFO] 未精确匹配品牌，保存全部 {len(products)} 条数据")
+            else:
+                brand_data[cat_name] = []
+                print("    [WARN] 未找到数据")
+
+        results[brand_name] = brand_data
+
+    return results
+
+
+def clear_and_search_brand(page: Page, brand_name: str):
+    """清空搜索框并输入新品牌名"""
+    print(f"    搜索品牌: {brand_name}")
+
+    try:
+        # 找到搜索框
+        search_input = page.locator('.jmtd-input-input').first
+        search_input.wait_for(state="visible", timeout=5000)
+
+        # 清空搜索框
+        search_input.click()
+        page.wait_for_timeout(300)
+        # 全选并删除
+        search_input.press("Control+a")
+        page.wait_for_timeout(200)
+        search_input.press("Backspace")
+        page.wait_for_timeout(500)
+
+        # 输入品牌名
+        search_input.fill(brand_name)
+        page.wait_for_timeout(1000)
+
+        # 按回车搜索
+        search_input.press("Enter")
+        page.wait_for_timeout(8000)
+        print("    [OK] 搜索完成")
+    except Exception as e:
+        print(f"    [WARN] 搜索失败: {e}")
+
+
+def save_new_product_results(results: dict, output_dir: str):
+    """保存新品数据到 JSON 和 Markdown 文件"""
+    today = datetime.now().strftime("%Y-%m-%d")
+    last_monday, last_sunday = get_last_week_range()
+    week_range = f"{last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}"
+
+    for brand_name, brand_data in results.items():
+        # 构建店铺名（用于文件名）
+        shop_name = brand_name.replace("/", "_").replace("\\", "_")
+
+        # 创建店铺新品监控子目录
+        shop_dir = os.path.join(output_dir, f"{shop_name}店铺新品监控")
+        os.makedirs(shop_dir, exist_ok=True)
+
+        # ── 保存 JSON ──
+        output = {
+            "店铺名": brand_name,
+            "采集日期": today,
+            "数据周期": week_range,
+            "新品数据": brand_data,
+        }
+        json_path = os.path.join(shop_dir, f"{shop_name}_新品_{today}.json")
+        with open(json_path, "w", encoding="utf-8") as f:
+            json.dump(output, f, ensure_ascii=False, indent=2)
+        print(f"JSON 已保存: {json_path}")
+
+        # ── 生成 Markdown ──
+        md = []
+        md.append(f"# {brand_name} - 新品数据")
+        md.append("")
+        md.append(f"**采集日期：** {today}")
+        md.append(f"**数据周期：** 上周 {week_range}")
+        md.append("")
+
+        for cat_name, products in brand_data.items():
+            md.append(f"## {cat_name}")
+            md.append("")
+
+            if not products:
+                md.append("暂无数据")
+                md.append("")
+                continue
+
+            md.append("| 排名 | 商品名称 | 店铺 | 访客数 | 成交金额 | 成交单量 | 搜索点击 | 关注人数 |")
+            md.append("|------|----------|------|--------|----------|----------|----------|----------|")
+
+            for prod in products:
+                md.append(f"| {prod.get('rank', 'N/A')} "
+                          f"| {prod.get('productName', 'N/A')} "
+                          f"| {prod.get('shopName', 'N/A')} "
+                          f"| {prod.get('visitors', 'N/A')} "
+                          f"| {prod.get('amount', 'N/A')} "
+                          f"| {prod.get('orders', 'N/A')} "
+                          f"| {prod.get('searchClicks', 'N/A')} "
+                          f"| {prod.get('followers', 'N/A')} |")
+            md.append("")
+
+        md_path = os.path.join(shop_dir, f"{shop_name}_新品_{today}.md")
+        with open(md_path, "w", encoding="utf-8") as f:
+            f.write("\n".join(md))
+        print(f"MD 已保存: {md_path}")
+
+
+# ── 登录流程 ──────────────────────────────────────────────────────────────────
+
+def close_popup_if_present(page):
+    """检测并关闭推广弹窗（profit plan）+ 广告弹窗（SVG close btn）。"""
+    closed_profit_plan = close_profit_plan_popup_if_present(page)
+
+    close_btn = page.locator(
+        'body > div.jmt-portal.jmt-portal-css-1nrmhc.jmtv2.jmt-dialog-portal.jmt-dialog-hidden.jmt-dialog-hidden-css-grlj9t.jmtv2.win-notice-modal > div:nth-child(2) > div > div > div > div > span > span > svg'
+    )
+    try:
+        close_btn.wait_for(state="visible", timeout=2000 if closed_profit_plan else 5000)
+        close_btn.click()
+        print("  [OK] 广告弹窗已关闭")
+        page.wait_for_timeout(3000)
+    except Exception:
+        if closed_profit_plan:
+            return
+        print("  [INFO] 无广告弹窗,继续执行")
+
+
+def step_login(page: Page, shop: str, password: str):
+    print("[Step 1] 登录京东商家平台...")
+    page.goto("https://shop.jd.com/jdm/home", wait_until="networkidle")
+    page.wait_for_timeout(3000)
+
+    account_input = page.locator('input[placeholder="请输入账号名/邮箱"]')
+    account_input.fill(shop)
+    page.wait_for_timeout(1000)
+
+    password_input = page.locator('input[placeholder="请输入登录密码"]')
+    password_input.fill(password)
+    page.wait_for_timeout(1000)
+
+    login_button = page.locator('button:has-text("立即登录")')
+    login_button.click()
+
+    page.wait_for_timeout(8000)
+    wait_for_navigation(page)
+
+    print("  尝试关闭推广弹窗...")
+    page.evaluate("document.querySelector('.close-btn')?.click()")
+    page.wait_for_timeout(2000)
+    page.keyboard.press("Escape")
+    page.wait_for_timeout(2000)
+    print("  登录完成")
+
+
+def _dismiss_all_dialog(page: Page) -> None:
+    """关闭 .all-dialog 全屏弹窗(6/15 起曾拦截 '商智' click)。"""
+    try:
+        dlg = page.locator("div.all-dialog")
+        if dlg.count() > 0 and dlg.first.is_visible():
+            for sel in [
+                "div.all-dialog .el-dialog__headerbtn",
+                "div.all-dialog .close",
+                "div.all-dialog [aria-label='close']",
+                "div.all-dialog [class*='close']",
+            ]:
+                btn = page.locator(sel).first
+                if btn.count() > 0 and btn.is_visible():
+                    btn.click()
+                    page.wait_for_timeout(800)
+                    print("  [OK] 已关闭 all-dialog")
+                    return
+            page.keyboard.press("Escape")
+            page.wait_for_timeout(800)
+    except Exception:
+        pass
+
+
+def step_enter_shangzhi(context: BrowserContext, page: Page) -> Page:
+    print("[Step 2] 进入商智...")
+    # 先关闭 .all-dialog 拦截弹窗,再点商智
+    _dismiss_all_dialog(page)
+    try:
+        with context.expect_page(timeout=10000) as new_page_info:
+            page.click("text=商智", timeout=10000)
+        sz_page = new_page_info.value
+    except Exception:
+        # 兜底: 强制点击,有可能没开新标签页
+        page.click("text=商智", force=True, timeout=10000)
+        page.wait_for_timeout(3000)
+        pages = context.pages
+        sz_page = pages[-1] if pages else page
+    try:
+        sz_page.wait_for_load_state("domcontentloaded", timeout=60000)
+    except Exception:
+        pass
+    _dismiss_all_dialog(sz_page)
+    close_popup_if_present(sz_page)
+    sz_page.wait_for_timeout(3000)
+    if "sz.jd.com" not in (sz_page.url or ""):
+        raise RuntimeError(f"点击商智后未进入 sz.jd.com, 当前 url={sz_page.url}")
+    print("  已在商智页面")
+    return sz_page
+
+
+def step_return_old_version(page: Page):
+    print("[Step 3] 返回旧版...")
+    try:
+        back_button = page.locator('#header > div > div.page-top-header-tools > img')
+        back_button.wait_for(state="visible", timeout=5000)
+        back_button.click()
+        page.wait_for_timeout(5000)
+        page.wait_for_load_state("networkidle")
+        print("  已进入商智旧版")
+    except Exception:
+        print("  [INFO] 未找到返回旧版按钮,继续当前页面")
+
+
+def step_goto_vender_ranks(page: Page):
+    print("[Step 4] 跳转到商家榜单...")
+    page.goto("https://sz.jd.com/szweb/sz/view/industryMarket/venderRanks.html",
+              wait_until="domcontentloaded", timeout=60000)
+    page.wait_for_timeout(8000)
+    print("  已进入商家榜单页面")
+
+
+# ── 搜索店铺 ──────────────────────────────────────────────────────────────────
+
+def search_shop(page: Page, shop_name: str):
+    """在商家榜单页面搜索指定店铺"""
+    print(f"  搜索店铺: {shop_name}")
+
+    # 尝试多种搜索框选择器
+    search_input = None
+    for selector in [
+        'input[placeholder*="搜索"]',
+        'input[placeholder*="店铺"]',
+        'input[placeholder*="商家"]',
+        'input[type="search"]',
+        '.search-input input',
+        '.jmtd-search input',
+    ]:
+        try:
+            el = page.locator(selector).first
+            if el.is_visible(timeout=2000):
+                search_input = el
+                break
+        except Exception:
+            continue
+
+    if not search_input:
+        # 兜底：用 JS 查找所有 input
+        search_input = page.locator('input').first
+        print("    [WARN] 未找到明确的搜索框,使用第一个 input")
+
+    # 清空并输入店铺名
+    search_input.click()
+    search_input.fill("")
+    page.wait_for_timeout(500)
+    search_input.fill(shop_name)
+    page.wait_for_timeout(1000)
+
+    # 尝试点击搜索按钮或按回车
+    try:
+        search_btn = page.locator('button:has-text("搜索"), button:has-text("查询"), .search-btn, .jmtd-search-btn').first
+        search_btn.click(timeout=3000)
+    except Exception:
+        search_input.press("Enter")
+
+    page.wait_for_timeout(8000)
+    print(f"    [OK] 搜索完成: {shop_name}")
+
+
+# ── 采集竞品数据 ──────────────────────────────────────────────────────────────
+
+def step_collect_competitor_data(page: Page):
+    """搜索竞品店铺,采集各类目数据"""
+    print("[Step 5] 采集竞品数据...")
+
+    # 先选择上周
+    select_last_week_any_day(page)
+
+    results = {}
+
+    for shop_name in SHOPS:
+        print(f"\n{'='*50}")
+        print(f"  店铺: {shop_name}")
+        print(f"{'='*50}")
+
+        shop_data = {}
+
+        # 搜索店铺
+        search_shop(page, shop_name)
+
+        # 遍历类目
+        for i, (cat_name, parent_cat) in enumerate(CATEGORIES):
+            output_cat_name = category_output_name(cat_name)
+            print(f"\n  [{i+1}/{len(CATEGORIES)}] 类目: {parent_cat or '一级类目'} → {cat_name} | 输出: {output_cat_name}")
+
+            # 切换类目
+            switch_category(page, cat_name, parent_cat)
+
+            # 等待数据刷新
+            page.wait_for_timeout(5000)
+
+            # 提取表格
+            rows = extract_ranking_table(page)
+
+            # 从结果中匹配目标店铺
+            matched = find_shop_in_rows(rows, shop_name)
+            if matched:
+                shop_data[output_cat_name] = matched
+                print(f"    排名: {matched['rank']}, 成交金额: {matched['amount']}, 访客数: {matched['visitors']}")
+            else:
+                shop_data[output_cat_name] = {"rank": "N/A", "amount": "N/A", "orders": "N/A",
+                                              "followers": "N/A", "visitors": "N/A", "searchClicks": "N/A"}
+                print(f"    [WARN] 未在表格中找到 {shop_name}")
+
+        results[shop_name] = shop_data
+
+    return results
+
+
+# ── 保存结果 ──────────────────────────────────────────────────────────────────
+
+def save_results(results, output_dir):
+    """保存采集结果到 JSON 和 Markdown 文件"""
+    today = datetime.now().strftime("%Y-%m-%d")
+    last_monday, last_sunday = get_last_week_range()
+    week_range = f"{last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}"
+
+    os.makedirs(output_dir, exist_ok=True)
+
+    # ── 保存 JSON ──
+    output = {
+        "采集日期": today,
+        "数据周期": week_range,
+        "数据": results,
+    }
+    json_path = os.path.join(output_dir, f"竞店数据_{today}.json")
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(output, f, ensure_ascii=False, indent=2)
+    print(f"\nJSON 已保存: {json_path}")
+
+    # ── 生成 Markdown ──
+    cat_names = [category_output_name(c[0]) for c in CATEGORIES]
+    shop_names = list(results.keys())
+
+    md = []
+    md.append("# 竞店数据汇总")
+    md.append("")
+    md.append(f"**采集日期：** {today}")
+    md.append(f"**数据周期：** 上周 {week_range}")
+    md.append("")
+
+    # 一、每个店铺的详细数据
+    md.append("## 一、各店铺分类数据")
+    md.append("")
+    for shop in shop_names:
+        md.append(f"### {shop}")
+        md.append("")
+        md.append("| 类目 | 排名 | 成交金额 | 成交单量 | 关注人数 | 访客数 | 搜索点击 |")
+        md.append("|------|------|----------|----------|----------|--------|----------|")
+        for cat in cat_names:
+            d = results[shop].get(cat, {})
+            md.append(f"| {cat} | {d.get('rank', 'N/A')} | {d.get('amount', 'N/A')} "
+                      f"| {d.get('orders', 'N/A')} | {d.get('followers', 'N/A')} "
+                      f"| {d.get('visitors', 'N/A')} | {d.get('searchClicks', 'N/A')} |")
+        md.append("")
+
+    # 二、按类目横向对比
+    md.append("## 二、类目横向对比")
+    md.append("")
+    for cat in cat_names:
+        md.append(f"### {cat}")
+        md.append("")
+        md.append("| 店铺 | 排名 | 成交金额 | 成交单量 | 关注人数 | 访客数 | 搜索点击 |")
+        md.append("|------|------|----------|----------|----------|--------|----------|")
+        for shop in shop_names:
+            d = results[shop].get(cat, {})
+            md.append(f"| {shop} | {d.get('rank', 'N/A')} | {d.get('amount', 'N/A')} "
+                      f"| {d.get('orders', 'N/A')} | {d.get('followers', 'N/A')} "
+                      f"| {d.get('visitors', 'N/A')} | {d.get('searchClicks', 'N/A')} |")
+        md.append("")
+
+    md_path = os.path.join(output_dir, f"竞店数据汇总_{today}.md")
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(md))
+    print(f"MD 已保存: {md_path}")
+
+    return md_path
+
+
+# ── 主流程 ────────────────────────────────────────────────────────────────────
+
+def main():
+    parser = argparse.ArgumentParser(description="京东商家榜单竞品数据采集")
+    parser.add_argument("--shop", default=DEFAULT_SHOP, help="店铺账号名")
+    parser.add_argument("--password", default=DEFAULT_PASSWORD, help="登录密码")
+    parser.add_argument("--output", default=DEFAULT_OUTPUT, help="输出目录")
+    args = parser.parse_args()
+    try:
+        args.output = str(resolve_raw_output(args.output, default=DEFAULT_OUTPUT))
+    except ValueError as exc:
+        parser.error(str(exc))
+
+    print("=" * 60)
+    print("京东商家榜单竞品数据采集")
+    print(f"  店铺: {args.shop}")
+    print(f"  竞品: {', '.join(SHOPS)}")
+    print(f"  类目: {', '.join(category_output_name(c[0]) for c in CATEGORIES)}")
+    print("=" * 60)
+
+    last_monday, last_sunday = get_last_week_range()
+    print(f"  上周: {last_monday.strftime('%Y-%m-%d')} ~ {last_sunday.strftime('%Y-%m-%d')}")
+    print()
+
+    with sync_playwright() as p:
+        browser = p.chromium.launch(headless=False)
+        context = browser.new_context(
+            viewport={"width": 1920, "height": 1080},
+            locale="zh-CN",
+        )
+        page = context.new_page()
+
+        try:
+            # Step 1: 登录
+            step_login(page, args.shop, args.password)
+
+            # Step 2: 进入商智
+            sz_page = step_enter_shangzhi(context, page)
+
+            # Step 3: 返回旧版
+            step_return_old_version(sz_page)
+
+            # Step 4: 跳转到商家榜单
+            step_goto_vender_ranks(sz_page)
+
+            # Step 5: 采集竞品数据
+            results = step_collect_competitor_data(sz_page)
+
+            # 保存竞店数据结果
+            filepath = save_results(results, args.output)
+
+            # Step 6: 跳转到商品榜单
+            step_goto_product_ranks(sz_page)
+
+            # Step 7: 采集新品数据
+            new_product_results = collect_new_products(sz_page)
+
+            # 保存新品数据结果
+            save_new_product_results(new_product_results, args.output)
+
+            print()
+            print("=" * 60)
+            print("采集完成！")
+            print(f"  竞店数据: {args.output}/竞店数据_{datetime.now().strftime('%Y-%m-%d')}.json")
+            print(f"  新品数据: {args.output}/*店铺新品监控/*_新品_{datetime.now().strftime('%Y-%m-%d')}.json")
+            print("=" * 60)
+
+        except KeyboardInterrupt:
+            print("\n用户中断,退出。")
+        except Exception as e:
+            print(f"\n[ERROR] 出错: {e}")
+            import traceback
+            traceback.print_exc()
+        finally:
+            context.close()
+            browser.close()
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/taobao_sycm.py b/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/taobao_sycm.py
new file mode 100644
index 0000000..d97d69c
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/collectors/taobao_sycm.py
@@ -0,0 +1,1110 @@
+"""
+淘宝生意参谋自动化脚本（Scrapling StealthySession 版）
+功能：自动登录（持久化 user_data_dir） + 数据抓取
+
+使用 StealthySession 替代 DynamicFetcher，与 product-collector-analyze-flow 项目一致：
+- StealthySession 自带反检测（canvas noise、WebRTC block、fingerprint 随机化），降低风控概率
+- real_chrome=True 启动真实 Chrome，user_data_dir 持久化登录态
+- 直接使用 page.goto() 在同一个 context 中操作，不创建/销毁页面
+- locale=zh-CN, timezone=Asia/Shanghai 匹配真实环境
+
+使用方法：
+1. 首次运行：自动填账号密码，手动完成滑块验证，登录状态自动保存到 user_data_dir
+2. 后续运行：直接恢复登录状态，自动抓取数据
+"""
+
+import json
+import os
+import shutil
+import subprocess
+import sys
+import time
+from datetime import datetime, timedelta
+from pathlib import Path
+
+from scrapling.fetchers import StealthySession
+
+from gyxx_flow.modules.shop_intelligence.runtime.paths import raw_path, state_path
+
+os.environ.pop("HTTP_PROXY", None)
+os.environ.pop("HTTPS_PROXY", None)
+os.environ.pop("http_proxy", None)
+os.environ.pop("https_proxy", None)
+
+CHROME_PATH = os.getenv("CHROME_PATH", "")
+DEBUG_PORT = int(os.getenv("GYXX_BROWSER_CDP_PORT", "0"))
+USER_DATA_DIR = os.getenv("TM_USER_DATA_DIR", str(state_path("tmall", "profile")))
+DAILY_USER_DATA_DIR = os.getenv("TM_DAILY_USER_DATA_DIR", USER_DATA_DIR)
+LOGIN_URL = "https://sycm.taobao.com/custom/login.htm?_target=http://sycm.taobao.com/custom/login"
+HOME_URL = "https://sycm.taobao.com/portal/home.htm"
+MARKET_RANK_URL = "https://sycm.taobao.com/mc/free/d11_market_rank"
+SHOP_STATS_URL = "https://myseller.taobao.com/home.htm/voc-tmall/serverReport"
+
+ACCOUNT = os.getenv("SYCM_ACCOUNT", "光影行星旗舰店:龙虾仔")
+PASSWORD = os.getenv("SYCM_PASSWORD", "")
+
+DATA_DIR = raw_path("天猫")
+SHOP_DATA_DIR = DATA_DIR / "店铺数据"
+
+
+def find_browser_executable():
+    candidates = [
+        CHROME_PATH,
+        r"D:\Program Files (x86)\Chrome\App\chrome.exe",
+        r"C:\Program Files\Google\Chrome\Application\chrome.exe",
+        r"C:\Program Files (x86)\Google\Chrome\Application\chrome.exe",
+        r"C:\Program Files\Microsoft\Edge\Application\msedge.exe",
+    ]
+    local_app_data = os.getenv("LOCALAPPDATA")
+    if local_app_data:
+        candidates.extend([
+            os.path.join(local_app_data, "Google", "Chrome", "Application", "chrome.exe"),
+            os.path.join(local_app_data, "Microsoft", "Edge", "Application", "msedge.exe"),
+        ])
+    for path in candidates:
+        if path and os.path.exists(path):
+            return path
+    raise FileNotFoundError("找不到 Chrome/Edge 浏览器，请设置环境变量 CHROME_PATH")
+
+
+def _kill_port_listeners(port):
+    if sys.platform != "win32":
+        return 0
+    try:
+        out = subprocess.check_output(
+            f"netstat -ano | findstr :{port}",
+            shell=True, text=True, encoding="gbk", errors="replace", timeout=5,
+        )
+    except Exception:
+        return 0
+    pids = set()
+    for line in out.splitlines():
+        parts = line.split()
+        if len(parts) < 5 or not parts[-1].isdigit():
+            continue
+        if f":{port}" in parts[1] and parts[3] == "LISTENING":
+            pids.add(int(parts[-1]))
+    killed = 0
+    for pid in pids:
+        try:
+            subprocess.run(f"taskkill /F /PID {pid}", shell=True, text=True, timeout=5,
+                           stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
+            killed += 1
+        except Exception:
+            pass
+    return killed
+
+
+def _profile_is_locked(user_data_dir):
+    for lock_name in ("SingletonLock", "SingletonCookie", "lockfile"):
+        lock_path = os.path.join(user_data_dir, lock_name)
+        if os.path.exists(lock_path):
+            return True, lock_name
+    return False, None
+
+
+def sync_cookies_from_daily_profile(src_user_data, dst_user_data):
+    src_root = Path(src_user_data)
+    dst_root = Path(dst_user_data)
+    if src_root.resolve() == dst_root.resolve():
+        return 0
+    src_default = src_root / "Default"
+    if not src_default.exists():
+        return 0
+    dst_default = dst_root / "Default"
+    dst_default.mkdir(parents=True, exist_ok=True)
+    copied = 0
+    for fname in ("Cookies", "Cookies-journal", "Login Data", "Login Data-journal", "Web Data", "Web Data-journal"):
+        src = src_default / fname
+        if src.exists():
+            shutil.copy2(src, dst_default / fname)
+            copied += 1
+    for dirname in ("Local Storage", "Session Storage", "Network"):
+        src = src_default / dirname
+        dst = dst_default / dirname
+        if src.exists():
+            if dst.exists():
+                shutil.rmtree(dst, ignore_errors=True)
+            try:
+                shutil.copytree(src, dst)
+                copied += 1
+            except Exception:
+                pass
+    return copied
+
+
+def connect_browser():
+    killed = _kill_port_listeners(DEBUG_PORT)
+    if killed:
+        time.sleep(1)
+        print(f"[CLEAN] 已清理脚本专属端口残留 Chrome ({killed} 个进程)")
+
+    locked, lock_name = _profile_is_locked(USER_DATA_DIR)
+    if locked:
+        print(f"[STOP] Profile {USER_DATA_DIR} 被占用（{lock_name}），请先关闭日常 Chrome")
+        return None, None
+
+    print(f"[BROWSER] 启 Chrome (profile={USER_DATA_DIR})")
+    session = StealthySession(
+        headless=False,
+        real_chrome=True,
+        user_data_dir=USER_DATA_DIR,
+        locale="zh-CN",
+        timezone_id="Asia/Shanghai",
+        timeout=60000,
+        network_idle=False,
+    )
+    session.start()
+    context = session.context
+    page = context.pages[-1] if context.pages else context.new_page()
+    return session, page
+
+
+def close_popups(page):
+    try:
+        page.evaluate("""
+            () => {
+                let closed = false;
+                const adBtn = document.querySelector('.sycm-active-egg-modal-content-close');
+                if (adBtn && adBtn.offsetParent !== null) { adBtn.click(); closed = true; }
+                const baxiaBtn = document.querySelector('.baxia-dialog-close');
+                if (baxiaBtn && baxiaBtn.offsetParent !== null) { baxiaBtn.click(); closed = true; }
+                const masks = document.querySelectorAll('.baxia-dialog-mask, .sycm-active-egg-modal-mask');
+                masks.forEach(m => m.parentElement && m.parentElement.remove());
+                return closed;
+            }
+        """)
+    except Exception:
+        pass
+
+
+def _store_name_from_account(account):
+    if not account:
+        return "光影行星旗舰店"
+    return account.split(":")[0].strip() or "光影行星旗舰店"
+
+
+def is_logged_in(page):
+    try:
+        if "login" in page.url.lower():
+            return False
+        store_name = _store_name_from_account(ACCOUNT)
+        dom_ok = page.evaluate(r"""
+            (storeName) => {
+                const html = document.documentElement.outerHTML || '';
+                const hasStoreName = storeName && new RegExp(storeName.replace(/[.*+?^${}()|[\]\\]/g, '\\$&')).test(html);
+                const hasTopNav = /商品排行|商品概览|品类分析|流量|交易|客户|增长因子/.test(html);
+                const hasAvatar = !!document.querySelector('[class*="avatar"], [class*="user-info"], [class*="logout"]');
+                return hasStoreName || hasTopNav || hasAvatar;
+            }
+        """, store_name)
+        return bool(dom_ok)
+    except Exception:
+        return "login" not in page.url.lower()
+
+
+def find_logged_in_page(context):
+    for p in context.pages:
+        if "login" not in p.url.lower() and "sycm.taobao.com" in p.url.lower():
+            return p
+    return None
+
+
+def do_login(page, context):
+    print("等待登录 iframe 加载...")
+    page.wait_for_selector("#alibaba-login-box", timeout=15000)
+    iframe_element = page.locator("#alibaba-login-box").element_handle()
+    login_frame = iframe_element.content_frame()
+    if not login_frame:
+        print("错误：未能获取登录 iframe")
+        return None
+
+    login_frame.wait_for_selector('input[placeholder="账号名/邮箱/手机号"]', timeout=30000)
+
+    print("正在输入账号...")
+    login_frame.fill('input[placeholder="账号名/邮箱/手机号"]', ACCOUNT)
+    time.sleep(0.5)
+
+    print("正在输入密码...")
+    login_frame.fill('input[placeholder="请输入登录密码"]', PASSWORD)
+    time.sleep(0.5)
+
+    print("正在点击登录按钮...")
+    login_frame.click('button:has-text("登录")')
+    time.sleep(1)
+
+    print("\n" + "=" * 50)
+    print("请手动完成滑块验证")
+    print("=" * 50 + "\n")
+
+    print("等待登录完成...")
+    for _ in range(300):
+        time.sleep(1)
+        close_popups(page)
+        if is_logged_in(page):
+            print(f"登录成功！当前页面: {page.url}")
+            return page
+        logged_page = find_logged_in_page(context)
+        if logged_page:
+            print(f"在其他标签页检测到已登录: {logged_page.url}")
+            return logged_page
+    print("等待 5 分钟后仍未检测到登录成功")
+    return None
+
+
+def ensure_logged_in(context, page):
+    print("检查天猫登录态...")
+    try:
+        page.goto(HOME_URL, wait_until="domcontentloaded", timeout=60000)
+        for _ in range(10):
+            if "login" not in page.url.lower():
+                break
+            time.sleep(0.5)
+        time.sleep(1)
+        if is_logged_in(page):
+            print(f"[OK] 检测到已登录态: {page.url}")
+            return page
+        if "login" in page.url.lower():
+            print("  [INFO] 仍停留在 login 页, 尝试 havanalogin 续期 cookie...")
+            try:
+                page.goto("https://havanalogin.taobao.com/", wait_until="domcontentloaded", timeout=30000)
+                time.sleep(2)
+                page.goto(HOME_URL, wait_until="domcontentloaded", timeout=60000)
+                for _ in range(10):
+                    if "login" not in page.url.lower():
+                        break
+                    time.sleep(0.5)
+                time.sleep(1)
+                if is_logged_in(page):
+                    print(f"[OK] havanalogin 续期后已登录: {page.url}")
+                    return page
+            except Exception as e:
+                print(f"  [WARN] havanalogin 续期失败: {e}")
+    except Exception as e:
+        print(f"  [WARN] 登录态预检失败: {e}")
+
+    print("强制走登录流程...")
+    page.goto(LOGIN_URL, wait_until="domcontentloaded", timeout=60000)
+    time.sleep(2)
+    logged_page = do_login(page, context)
+    if not logged_page:
+        print("登录失败")
+        return None
+    return logged_page
+
+
+def ensure_on_homepage(page):
+    if "sycm.taobao.com/portal/home" not in page.url:
+        print("正在导航到生意参谋首页...")
+        page.goto(HOME_URL, wait_until="networkidle", timeout=60000)
+        time.sleep(3)
+    else:
+        print(f"当前已在首页: {page.url}")
+    if "punish" in page.url or "deny" in page.url:
+        print("警告：首页触发风控，等待30秒后重试...")
+        time.sleep(30)
+        page.goto(HOME_URL, wait_until="networkidle", timeout=60000)
+        time.sleep(3)
+    close_popups(page)
+
+
+def click_grow_factor_tab(page):
+    """点击'增长因子'标签页"""
+    print("正在点击'增长因子'标签...")
+    close_popups(page)
+    page.evaluate("""
+        () => {
+            const masks = document.querySelectorAll('.baxia-dialog-mask, .sycm-active-egg-modal-mask');
+            masks.forEach(m => m.parentElement && m.parentElement.remove());
+        }
+    """)
+    try:
+        # 直接用 JS 点击，避免 Playwright 点击被拦截
+        clicked = page.evaluate("""
+            () => {
+                const tab = document.querySelector('.tab-title-GROW_FACTOR');
+                if (tab && tab.offsetParent !== null) { tab.click(); return true; }
+                const allTabs = document.querySelectorAll('[class*="tab-title"]');
+                for (const t of allTabs) {
+                    if (t.textContent.includes('增长因子') && t.offsetParent !== null) { t.click(); return true; }
+                }
+                return false;
+            }
+        """)
+        if clicked:
+            print("已点击'增长因子'")
+            time.sleep(3)
+            page.evaluate("window.scrollTo(0, 400)")
+            time.sleep(1)
+            return True
+    except Exception as e:
+        print(f"点击增长因子出错: {e}")
+    print("警告：未找到'增长因子'标签")
+    return False
+
+
+def click_week_button(page):
+    """点击'周'按钮切换到周维度，默认显示上周"""
+    close_popups(page)
+    page.evaluate("""
+        () => {
+            const masks = document.querySelectorAll('.baxia-dialog-mask, .sycm-active-egg-modal-mask');
+            masks.forEach(m => m.parentElement && m.parentElement.remove());
+        }
+    """)
+    time.sleep(0.5)
+
+    print("正在点击'周'按钮...")
+    for attempt in range(5):
+        try:
+            result = page.evaluate("""
+                () => {
+                    const container = document.querySelector('.oui-date-picker-particle-button');
+                    if (!container) return false;
+                    const btns = container.querySelectorAll('button');
+                    for (const btn of btns) {
+                        const span = btn.querySelector('span');
+                        if (span && span.textContent.trim() === '周') { btn.click(); return true; }
+                    }
+                    return false;
+                }
+            """)
+            if result:
+                print("已点击'周'按钮")
+                time.sleep(3)
+                try:
+                    date_text = page.evaluate("""
+                        () => {
+                            const el = document.querySelector('.oui-date-picker-current-date');
+                            return el ? el.textContent.trim() : null;
+                        }
+                    """)
+                    print(f"当前日期: {date_text}")
+                except Exception:
+                    pass
+                return True
+            else:
+                if attempt < 4:
+                    print(f"  未找到'周'按钮，等待重试 ({attempt+1}/5)...")
+                    time.sleep(2)
+        except Exception as e:
+            print(f"点击周按钮出错: {e}")
+            if attempt < 4:
+                time.sleep(2)
+
+    print("警告：未找到'周'按钮")
+    return False
+
+
+def get_last_week_range():
+    today = datetime.now()
+    this_monday = today - timedelta(days=today.weekday())
+    last_monday = this_monday - timedelta(days=7)
+    last_sunday = this_monday - timedelta(days=1)
+    return last_monday.strftime("%Y-%m-%d") + " ~ " + last_sunday.strftime("%Y-%m-%d")
+
+
+def fetch_homepage_data(page):
+    print("\n正在抓取首页数据...")
+    time.sleep(3)
+
+    all_metrics = page.evaluate("""
+        () => {
+            const cards = document.querySelectorAll('.sycm-index-higher-common-wrapper-2');
+            const result = {};
+            for (const card of cards) {
+                const nameEl = card.querySelector('.index-name .text');
+                if (!nameEl) continue;
+                const name = nameEl.textContent.trim();
+
+                const mainValueEl = card.querySelector('.index-value .value.undefined, .index-value .value');
+                const smallEl = card.querySelector('.index-value .smallValue');
+                let value = '';
+                if (mainValueEl) value += mainValueEl.textContent.trim();
+                if (smallEl) value += smallEl.textContent.trim();
+
+                const cycleEl = card.querySelector('.value.cyclecrc, .value.up.cyclecrc, .value.down.cyclecrc');
+                let cycle = cycleEl ? cycleEl.textContent.trim() : '-';
+
+                const syncEl = card.querySelector('.index-syncCrc .index-value');
+                let sync = '-';
+                if (syncEl) {
+                    const sv = syncEl.querySelector('.value');
+                    if (sv) sync = sv.textContent.trim();
+                }
+
+                result[name] = {value, cycle, sync};
+            }
+            return result;
+        }
+    """)
+
+    data = {}
+    targets = {
+        "净支付金额": "netPaymentAmount",
+        "访客数": "uv",
+        "支付买家数": "payByrCnt",
+        "支付转化率": "payConversionRate",
+        "支付件数": "payItemCnt",
+        "客单价": "customerUnitPrice",
+        "总推广花费": "totalPromotionCost",
+        "总ROI": "totalRoi",
+        "退款金额": "refundAmount",
+    }
+
+    for display_name, index_id in targets.items():
+        if display_name in all_metrics and all_metrics[display_name]["value"]:
+            m = all_metrics[display_name]
+            data[display_name] = m["value"]
+            data[f"{display_name}_环比"] = m["cycle"]
+            data[f"{display_name}_较去年同期"] = m["sync"]
+            print(f"  [OK] {display_name}: {m['value']}  环比:{m['cycle']}  同期:{m['sync']}")
+        else:
+            val = None
+            try:
+                locator = page.locator(f'[data-onetrace-index-id="{index_id}"]')
+                if locator.count() > 0:
+                    main = locator.first.locator('.value.undefined, .value').first.inner_text().strip()
+                    small = locator.first.locator('.smallValue').first.inner_text().strip()
+                    val = (main + small).strip()
+            except Exception:
+                pass
+            if val:
+                data[display_name] = val
+                data[f"{display_name}_环比"] = "-"
+                data[f"{display_name}_较去年同期"] = "-"
+                print(f"  [OK] {display_name}: {val} (仅数值)")
+            else:
+                print(f"  [FAIL] {display_name}: 未找到")
+
+    promo_cost = data.get("总推广花费")
+    net_pay = data.get("净支付金额")
+    if promo_cost and net_pay:
+        try:
+            pc = float(promo_cost.replace(",", ""))
+            np = float(net_pay.replace(",", ""))
+            if np > 0:
+                ratio = round(pc / np * 100, 2)
+                data["付费占比"] = f"{ratio}%"
+                print(f"  [OK] 付费占比: {ratio}%")
+        except ValueError:
+            pass
+
+    return data
+
+
+def navigate_category_tree(page, cat_path):
+    for i, level_name in enumerate(cat_path[:-1]):
+        result = page.evaluate("""
+            (name) => {
+                const selectors = ['.tree-item.common-item', '.ant-tree-treenode', '.tree-node', '[class*="tree-item"]'];
+                for (const sel of selectors) {
+                    const items = document.querySelectorAll(sel);
+                    for (const item of items) {
+                        if (item.textContent.trim().includes(name)) {
+                            item.scrollIntoView({block: 'center'});
+                            const switcher = item.querySelector('.tree-switcher, .ant-tree-switcher, [class*="switcher"]');
+                            if (switcher) switcher.click(); else item.click();
+                            return true;
+                        }
+                    }
+                }
+                return false;
+            }
+        """, level_name)
+        if not result:
+            return False
+        time.sleep(1.5)
+    leaf_name = cat_path[-1]
+    result = page.evaluate("""
+        (name) => {
+            const selectors = ['.tree-item.common-item', '.ant-tree-treenode', '.tree-node', '[class*="tree-item"]'];
+            for (const sel of selectors) {
+                const items = document.querySelectorAll(sel);
+                for (const item of items) {
+                    const text = item.textContent.trim();
+                    if (text === name || text.includes(name)) {
+                        item.scrollIntoView({block: 'center'});
+                        item.click();
+                        return true;
+                    }
+                }
+            }
+            return false;
+        }
+    """, leaf_name)
+    if result:
+        time.sleep(0.5)
+        return True
+    return False
+
+
+def self_shop_rank_info(page):
+    data = {}
+    try:
+        try:
+            page.wait_for_selector('tr.ant-table-row-level-0', timeout=8000)
+        except Exception:
+            return None
+        row = page.locator('tr.ant-table-row-level-0').first
+        if row.count() == 0:
+            return None
+        data["店铺"] = row.locator('.shopName a').first.inner_text().strip()
+        col_data = page.evaluate("""
+            () => {
+                const row = document.querySelector('tr.ant-table-row-level-0');
+                if (!row) return null;
+                const cols = row.querySelectorAll('.selfShopRankInfoColumn');
+                const result = [];
+                for (const col of cols) {
+                    const spans = col.querySelectorAll('span');
+                    const texts = [];
+                    for (const s of spans) {
+                        const t = s.textContent.trim();
+                        if (t) texts.push(t);
+                    }
+                    result.push(texts.slice(1).join(' | '));
+                }
+                return result;
+            }
+        """)
+        if col_data:
+            col_names = ["全部店铺行业排名", "支付买家数", "在售商品数量", "访客数"]
+            for i, name in enumerate(col_names):
+                if i < len(col_data) and col_data[i]:
+                    data[name] = col_data[i]
+    except Exception as e:
+        print(f"  抓取排行数据出错: {e}")
+    return data if data else None
+
+
+def fetch_market_rank(page):
+    print("\n正在跳转到618市场排行页面...")
+    page.goto(MARKET_RANK_URL, wait_until="networkidle", timeout=60000)
+    time.sleep(3)
+    close_popups(page)
+
+    if "punish" in page.url or "deny" in page.url:
+        print("警告：触发风控，等待30秒后重试...")
+        time.sleep(30)
+        page.goto(MARKET_RANK_URL, wait_until="networkidle", timeout=60000)
+        time.sleep(3)
+        close_popups(page)
+
+    print("正在点击周按钮...")
+    try:
+        page.evaluate("""
+            () => {
+                const container = document.querySelector('.oui-date-picker-particle-button');
+                if (!container) return false;
+                const btns = container.querySelectorAll('button, span, a, div');
+                for (const btn of btns) {
+                    if (btn.textContent.trim() === '周') { btn.click(); return true; }
+                }
+                return false;
+            }
+        """)
+        print("已点击周")
+        time.sleep(3)
+    except Exception as e:
+        print(f"点击周出错: {e}")
+
+    print("正在切换到店铺标签...")
+    try:
+        shop_tab = page.locator('.oui-tab-switch-item:has-text("店铺")').first
+        if shop_tab.count() > 0:
+            shop_tab.click()
+            print("已切换到店铺")
+            time.sleep(2)
+    except Exception as e:
+        print(f"切换店铺标签出错: {e}")
+
+    categories = [
+        {"name": "旅行袋", "path": ["箱包皮具", "热销女包/男包", "旅行袋"]},
+        {"name": "男士包袋", "path": ["箱包皮具", "热销女包/男包", "男士包袋"]},
+        {"name": "双肩背包", "path": ["箱包皮具", "热销女包/男包", "双肩背包"]},
+        {"name": "通用款女包", "path": ["箱包皮具", "女士包袋新", "通用款女包"]},
+        {"name": "托特包", "path": ["箱包皮具", "女士包袋新", "托特包"]},
+        {"name": "数码相机包", "path": ["3C数码配件", "数码相机配件", "数码相机包"]},
+        {"name": "笔记本电脑包", "path": ["3C数码配件", "笔记本电脑配件", "笔记本电脑包"]},
+    ]
+
+    all_results = []
+
+    for cat in categories:
+        cat_name = cat["name"]
+        cat_path = cat["path"]
+        print(f"\n{'='*40}")
+        print(f"正在处理类目: {cat_name}")
+
+        try:
+            close_popups(page)
+            page.evaluate("""
+                () => {
+                    const el = document.querySelector('.common-picker-header');
+                    if (el) { el.scrollIntoView({block: 'center'}); el.click(); }
+                }
+            """)
+            time.sleep(1.5)
+
+            if not navigate_category_tree(page, cat_path):
+                print(f"  未找到类目: {cat_name}")
+                continue
+            print(f"  已选择: {cat_name}")
+            old_url = page.url
+            time.sleep(2)
+            if page.url == old_url:
+                leaf_name = cat_path[-1]
+                page.evaluate("""
+                    (name) => {
+                        const items = document.querySelectorAll('.tree-item.common-item, [class*="tree-item"]');
+                        for (const item of items) {
+                            const textEl = item.querySelector('.tree-title, [class*="title"], span');
+                            if (textEl && textEl.textContent.trim() === name) {
+                                textEl.click();
+                                return true;
+                            }
+                        }
+                        return false;
+                    }
+                """, leaf_name)
+                time.sleep(2)
+            print(f"  当前URL: {page.url}")
+            page.evaluate("document.body.click()")
+            time.sleep(3)
+
+            if "punish" in page.url or "deny" in page.url:
+                print("  警告：触发风控页面，等待30秒后重试...")
+                time.sleep(30)
+                page.goto(MARKET_RANK_URL, wait_until="networkidle", timeout=60000)
+                time.sleep(3)
+                close_popups(page)
+                continue
+
+            close_popups(page)
+
+            rank_data = self_shop_rank_info(page)
+            if rank_data:
+                rank_data["类目"] = cat_name
+                all_results.append(rank_data)
+                print(f"  店铺: {rank_data.get('店铺', '-')}")
+                print(f"  排名: {rank_data.get('全部店铺行业排名', '-')}")
+                print(f"  买家数: {rank_data.get('支付买家数', '-')}")
+                print(f"  商品数: {rank_data.get('在售商品数量', '-')}")
+                print(f"  访客数: {rank_data.get('访客数', '-')}")
+            else:
+                print(f"  未抓取到 {cat_name} 的排行数据")
+
+        except Exception as e:
+            print(f"  处理类目 {cat_name} 出错: {e}")
+
+        time.sleep(3)
+
+    print(f"\n{'='*60}")
+    print("618市场排行汇总（全部7个类目）：")
+    print("=" * 60)
+    for r in all_results:
+        print(f"\n  【{r.get('类目', '-')}】")
+        print(f"    店铺: {r.get('店铺', '-')}")
+        print(f"    全部店铺行业排名: {r.get('全部店铺行业排名', '-')}")
+        print(f"    支付买家数: {r.get('支付买家数', '-')}")
+        print(f"    在售商品数量: {r.get('在售商品数量', '-')}")
+        print(f"    访客数: {r.get('访客数', '-')}")
+    if not all_results:
+        print("  未抓取到任何类目的排行数据")
+    print("=" * 60)
+    return all_results
+
+
+def fetch_shop_stats(page):
+    print("\n正在跳转到店铺表现页面...")
+    page.goto(SHOP_STATS_URL, wait_until="networkidle", timeout=60000)
+    time.sleep(3)
+
+    # 如果被重定向到登录页，等待自动跳转
+    if "login" in page.url.lower():
+        print("  店铺表现页被重定向，等待自动跳转...")
+        for _ in range(20):
+            if "login" not in page.url.lower():
+                break
+            time.sleep(1)
+        time.sleep(3)
+
+    # 等待页面内容加载
+    try:
+        page.wait_for_selector('[class*="ScoreDetails"]', timeout=15000)
+        time.sleep(2)
+    except Exception:
+        print("  等待店铺表现数据超时，尝试继续抓取...")
+
+    print("正在抓取店铺表现数据...")
+    stats = []
+
+    # 用 JS 一次性抓取所有数据，避免选择器变化导致逐个失败
+    raw_stats = page.evaluate("""
+        () => {
+            const results = [];
+
+            // 1. 真实体验分（firstItem）
+            const firstItem = document.querySelector('[class*="ScoreDetails--firstItem"]');
+            if (firstItem) {
+                const nameEl = firstItem.querySelector('[class*="firstItemName"]');
+                const scoreEl = firstItem.querySelector('[class*="firstItemScore"], [class*="ProgressBar--score"]');
+                const rankEl = firstItem.querySelector('[class*="rankHeightLight"]');
+                results.push({
+                    name: nameEl ? nameEl.textContent.trim() : '',
+                    score: scoreEl ? scoreEl.textContent.trim() : '',
+                    rank: rankEl ? rankEl.textContent.trim() : '',
+                    weight: '',
+                    rating: '',
+                });
+            }
+
+            // 2. 其他项（宝贝质量、物流速度、服务保障）
+            const itemContainers = document.querySelectorAll('[class*="ScoreDetails--itemContainer"]');
+            for (const container of itemContainers) {
+                try {
+                    const nameEl = container.querySelector('[class*="itemName"]');
+                    const nameAndWeightEl = container.querySelector('[class*="itemNameAndWeight"]');
+                    const scoreEl = container.querySelector('[class*="itemScore"], [class*="ProgressBar--score"]');
+                    const ratingEl = container.querySelector('[class*="rating"]');
+                    const rankEl = container.querySelector('[class*="rankHeightLight"]');
+
+                    let name = nameEl ? nameEl.textContent.trim() : '';
+                    let weight = '';
+                    if (nameAndWeightEl) {
+                        const fullText = nameAndWeightEl.textContent.trim();
+                        const m = fullText.match(/权重\\d+%/);
+                        if (m) weight = m[0];
+                    }
+                    let score = scoreEl ? scoreEl.textContent.trim() : '';
+                    let rating = ratingEl ? ratingEl.textContent.trim() : '';
+                    let rank = rankEl ? rankEl.textContent.trim() : '';
+
+                    if (name) {
+                        results.push({ name, score, rank, weight, rating });
+                    }
+                } catch (e) {}
+            }
+            return results;
+        }
+    """)
+
+    if raw_stats:
+        for s in raw_stats:
+            entry = {
+                "name": s.get("name", "-"),
+                "score": s.get("score", "-"),
+                "rank": f"超过{s['rank']}同行" if s.get("rank") else "-",
+                "weight": s.get("weight", "") or "",
+                "rating": s.get("rating", "") or "",
+            }
+            stats.append(entry)
+
+    print("\n" + "=" * 50)
+    print("店铺表现数据：")
+    print("=" * 50)
+    for s in stats:
+        parts = [s["name"]]
+        if s["weight"]:
+            parts.append(s["weight"])
+        parts.append(f"分数: {s['score']}")
+        if s["rating"]:
+            parts.append(f"评级: {s['rating']}")
+        parts.append(s["rank"])
+        print(f"  {'，'.join(parts)}")
+    if not stats:
+        print("  未抓取到店铺表现数据")
+    print("=" * 50)
+    return stats
+
+
+def _parse_num(val):
+    try:
+        return float(str(val).replace(",", "").replace("%", "").strip())
+    except (ValueError, TypeError):
+        return None
+
+
+def _parse_pct(val):
+    try:
+        return float(str(val).replace(",", "").replace("%", "").strip())
+    except (ValueError, TypeError):
+        return None
+
+
+def _clean_peer(val):
+    if val is None or (isinstance(val, str) and val.strip() in ("-", "—", "N/A", "n/a", "")):
+        return None
+    return val
+
+
+def build_db_records(homepage_data, market_rank, shop_stats, shop_name="光影行星旗舰店", ai_analysis=""):
+    """把三块数据合并成一行 (对齐 jd_shop_weekly 设计: 一行=一周)"""
+    today = datetime.now().strftime("%Y-%m-%d")
+    this_monday = datetime.now() - timedelta(days=datetime.now().weekday())
+    week_start = (this_monday - timedelta(days=7)).strftime("%Y-%m-%d")
+    week_end = (this_monday - timedelta(days=1)).strftime("%Y-%m-%d")
+    week_range = f"{week_start} ~ {week_end}"
+    source_file = str(SHOP_DATA_DIR / f"sycm_{today}.json")
+
+    h = homepage_data or {}
+
+    rankings = {}
+    for r in (market_rank or []):
+        cat_name = r.get("类目", "")
+        if not cat_name:
+            continue
+        rankings[cat_name] = {
+            "rank": r.get("全部店铺行业排名", ""),
+            "shop": r.get("店铺", shop_name),
+            "paid_buyers": r.get("支付买家数", ""),
+            "goods_count": r.get("在售商品数量", ""),
+            "visitors": r.get("访客数", ""),
+        }
+
+    star_rating = None
+    star_rating_raw = None
+    if shop_stats:
+        parts = []
+        for s in shop_stats:
+            name = s.get("name", "")
+            score = s.get("score", "")
+            weight = s.get("weight", "")
+            rating = s.get("rating", "")
+            rank = s.get("rank", "")
+            if name == "真实体验分":
+                try:
+                    star_rating = float(score)
+                except (ValueError, TypeError):
+                    pass
+            part = name
+            if score:
+                part += score
+            if rating:
+                part += "(%s)" % rating
+            if weight:
+                part += "[%s]" % weight
+            if rank:
+                part += rank
+            parts.append(part)
+        star_rating_raw = "；".join(parts)
+
+    conversion_rate_raw = h.get("支付转化率")
+    conversion_rate_val = None
+    if conversion_rate_raw:
+        conversion_rate_val = _parse_pct(conversion_rate_raw)
+
+    record = {
+        "shop": shop_name,
+        "industry": "箱包",
+        "caliber_days": 7,
+        "data_date": today,
+        "week_start": week_start,
+        "week_end": week_end,
+        "week_range": week_range,
+        "star_rating": star_rating,
+        "star_rating_raw": star_rating_raw,
+        "deal_amount": _parse_num(h.get("净支付金额")),
+        "deal_amount_qoq": h.get("净支付金额_环比"),
+        "deal_amount_peer_avg": _clean_peer(h.get("净支付金额_较去年同期")),
+        "visitors": _parse_num(h.get("访客数")),
+        "visitors_qoq": h.get("访客数_环比"),
+        "visitors_peer_avg": _clean_peer(h.get("访客数_较去年同期")),
+        "paying_customers": _parse_num(h.get("支付买家数")),
+        "paying_customers_qoq": h.get("支付买家数_环比"),
+        "paying_customers_peer_avg": _clean_peer(h.get("支付买家数_较去年同期")),
+        "conversion_rate": conversion_rate_val,
+        "conversion_rate_qoq": h.get("支付转化率_环比"),
+        "conversion_rate_peer_avg": _clean_peer(h.get("支付转化率_较去年同期")),
+        "order_qty": _parse_num(h.get("支付件数")),
+        "order_qty_qoq": h.get("支付件数_环比"),
+        "order_qty_peer_avg": _clean_peer(h.get("支付件数_较去年同期")),
+        "refund_amount": _parse_num(h.get("退款金额")),
+        "refund_amount_qoq": h.get("退款金额_环比"),
+        "avg_order_value": _parse_num(h.get("客单价")),
+        "avg_order_value_qoq": h.get("客单价_环比"),
+        "avg_order_value_peer_avg": _clean_peer(h.get("客单价_较去年同期")),
+        "promotion_cost": _parse_num(h.get("总推广花费")),
+        "promotion_cost_qoq": h.get("总推广花费_环比"),
+        "roi": _parse_num(h.get("总ROI")),
+        "roi_qoq": h.get("总ROI_环比"),
+        "promotion_ratio": h.get("付费占比"),
+        "rankings": rankings,
+        "raw_metrics_path": source_file,
+        "ai_analysis": ai_analysis,
+    }
+
+    return record
+
+
+def export_data(homepage_data, market_rank, shop_stats):
+    SHOP_DATA_DIR.mkdir(parents=True, exist_ok=True)
+    today = datetime.now().strftime("%Y-%m-%d")
+    this_monday = datetime.now() - timedelta(days=datetime.now().weekday())
+    week_start = (this_monday - timedelta(days=7)).strftime("%Y-%m-%d")
+    week_end = (this_monday - timedelta(days=1)).strftime("%Y-%m-%d")
+
+    all_data = {
+        "date": today,
+        "week_start": week_start,
+        "week_end": week_end,
+        "week_range": f"{week_start} ~ {week_end}",
+        "homepage": homepage_data or {},
+        "market_rank": market_rank or [],
+        "shop_stats": shop_stats or [],
+    }
+
+    json_path = SHOP_DATA_DIR / f"sycm_{today}.json"
+    with open(json_path, "w", encoding="utf-8") as f:
+        json.dump(all_data, f, ensure_ascii=False, indent=2)
+    print(f"JSON已导出: {json_path}")
+
+    md_path = SHOP_DATA_DIR / f"sycm_{today}.md"
+    lines = [f"# 生意参谋数据 - {today}\n"]
+
+    lines.append("## 核心指标（上周）\n")
+    lines.append("| 指标 | 数值 | 环比 | 较去年同期 |")
+    lines.append("|------|------|------|-----------|")
+    if homepage_data:
+        core_items = ["净支付金额", "访客数", "支付买家数", "支付转化率", "客单价", "总推广花费", "总ROI"]
+        for name in core_items:
+            val = homepage_data.get(name, "-")
+            cycle = homepage_data.get(f"{name}_环比", "-")
+            sync = homepage_data.get(f"{name}_较去年同期", "-")
+            lines.append(f"| {name} | {val} | {cycle} | {sync} |")
+        ratio = homepage_data.get("付费占比", "-")
+        if ratio != "-":
+            lines.append(f"| 付费占比 | {ratio} | - | - |")
+    else:
+        lines.append("| - | 未获取到数据 | - | - |")
+    lines.append("")
+
+    lines.append("## 618市场排行（周-店铺维度）\n")
+    lines.append("| 类目 | 全部店铺行业排名 | 支付买家数 | 在售商品数量 | 访客数 |")
+    lines.append("|------|-----------------|-----------|-------------|--------|")
+    if market_rank:
+        for r in market_rank:
+            cat = r.get("类目", "-")
+            rank = r.get("全部店铺行业排名", "-").replace(" | ", " ")
+            buyers = r.get("支付买家数", "-")
+            goods = r.get("在售商品数量", "-")
+            uv = r.get("访客数", "-")
+            lines.append(f"| {cat} | {rank} | {buyers} | {goods} | {uv} |")
+    else:
+        lines.append("| - | - | - | - | - |")
+    lines.append("")
+
+    lines.append("## 店铺表现\n")
+    lines.append("| 项目 | 权重 | 分数 | 评级 | 排名 |")
+    lines.append("|------|------|------|------|------|")
+    if shop_stats:
+        for s in shop_stats:
+            name = s.get("name", "-")
+            weight = s.get("weight", "-") or "-"
+            score = s.get("score", "-")
+            rating = s.get("rating", "-") or "-"
+            rank = s.get("rank", "-")
+            lines.append(f"| {name} | {weight} | {score} | {rating} | {rank} |")
+    else:
+        lines.append("| - | - | - | - | - |")
+    lines.append("")
+
+    with open(md_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(lines))
+    print(f"Markdown已导出: {md_path}")
+
+    return json_path
+
+
+def main():
+    # 从日常 Chrome profile 同步 cookie（免登录）
+    copied = sync_cookies_from_daily_profile(DAILY_USER_DATA_DIR, USER_DATA_DIR)
+    if copied:
+        print(f"[COOKIE] 从日常 Chrome 同步 {copied} 个文件到 {USER_DATA_DIR}")
+
+    session, page = connect_browser()
+    if not session or not page:
+        print("连接 Chrome 失败")
+        return
+
+    context = session.context
+
+    # Step 1: 确保登录
+    print("\n--- Step 1: 检查登录状态 ---")
+    page = ensure_logged_in(context, page)
+    if not page:
+        session.close()
+        _kill_port_listeners(DEBUG_PORT)
+        return
+
+    # Step 2: 抓取首页核心指标
+    print("\n--- Step 2: 抓取首页核心指标 ---")
+    ensure_on_homepage(page)
+    close_popups(page)
+    click_grow_factor_tab(page)
+    close_popups(page)
+    click_week_button(page)
+    close_popups(page)
+    homepage_data = fetch_homepage_data(page)
+
+    # Step 3: 抓取市场排行
+    print("\n--- Step 3: 抓取618市场排行 ---")
+    market_rank = fetch_market_rank(page)
+
+    # Step 4: 抓取店铺表现
+    print("\n--- Step 4: 抓取店铺表现 ---")
+    shop_stats = fetch_shop_stats(page)
+
+    # Step 5: 导出
+    print("\n--- Step 5: 导出数据 ---")
+    json_path = export_data(homepage_data, market_rank, shop_stats)
+
+    # Step 5.5: 生成 AI 分析并写回 JSON（DB 持久化和飞书写入都从 JSON 读取）
+    ai_analysis = ""
+    try:
+        from gyxx_flow.modules.shop_intelligence.runtime.config import TM_PLATFORM
+        from gyxx_flow.modules.shop_intelligence.runtime.writers import (
+            ensure_ai_analysis,
+        )
+        ai_analysis = ensure_ai_analysis(TM_PLATFORM, str(json_path))
+    except Exception as e:
+        print(f"[AI] 分析生成失败: {e}")
+
+    # Step 6: 入库 PostgreSQL
+    print("\n--- Step 6: 入库 PostgreSQL ---")
+    try:
+        from gyxx_flow.modules.shop_intelligence.runtime.db import (
+            get_conn,
+            init_schema,
+            upsert_tm_shop_weekly,
+        )
+        init_schema()
+        record = build_db_records(homepage_data, market_rank, shop_stats, ai_analysis=ai_analysis)
+        if record:
+            with get_conn() as conn:
+                n = upsert_tm_shop_weekly(conn, record)
+            print(f"[DB] 入库 {n} 条 → tm_shop_weekly (一行=一周)")
+        else:
+            print("[DB] 无数据可入库")
+    except Exception as e:
+        print(f"[DB] 入库失败: {e}")
+
+    print("\n" + "=" * 50)
+    print("抓取结果：")
+    print("=" * 50)
+    for k, v in (homepage_data or {}).items():
+        print(f"  {k}: {v}")
+    if not homepage_data:
+        print("  首页数据: 未抓取到")
+    print(f"  市场排行: {len(market_rank)} 个类目")
+    print(f"  店铺表现: {len(shop_stats)} 项")
+    print("=" * 50)
+
+    session.close()
+    _kill_port_listeners(DEBUG_PORT)
+    print("\n脚本执行完毕")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/config.example.env b/src/gyxx_flow/modules/shop_intelligence/runtime/config.example.env
new file mode 100644
index 0000000..d25994d
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/config.example.env
@@ -0,0 +1,32 @@
+# ── PostgreSQL 持久化 ────────────────────────────────────────────────────────
+DB_URL=postgresql://data_hub:replace-with-database-password@8.148.185.119:5432/data_hub
+# DB_HOST=8.148.185.119
+# DB_PORT=5432
+# DB_NAME=data_hub
+# DB_USER=data_hub
+# DB_PASSWORD=
+DB_ENABLED=true
+
+# ── 平台账号密码 ──────────────────────────────────────────────────────────────
+JD_SHOP=光影行星研发
+JD_PASSWORD=
+JD_INDUSTRY=男包
+
+DY_SHOP=光影行星箱包旗舰店
+DY_PASSWORD=
+
+TM_SHOP=光影行星旗舰店
+TM_PASSWORD=
+
+# ── 飞书多维表格 ─────────────────────────────────────────────────────────────
+BASE_APP_TOKEN=
+LARK_PROFILE=hermes-analyzer
+
+SHOP_BASE_TABLE_ID=tblsvXfxVkr3Zzwm
+JD_SHOP_PARENT_RECORD_ID=recv7vJnwmVjan
+
+JD_PEER_BASE_TABLE_ID=tblCmbO2UeEkPVgi
+DY_PEER_BASE_TABLE_ID=tblleRSkrfRoLTms
+
+# ── 浏览器路径（可选）─────────────────────────────────────────────────────────
+# CHROME_PATH=C:\Program Files\Google\Chrome\Application\chrome.exe
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/config.py b/src/gyxx_flow/modules/shop_intelligence/runtime/config.py
new file mode 100644
index 0000000..a0ec390
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/config.py
@@ -0,0 +1,118 @@
+"""Merged configuration for JD / DY / TM shop and peer-store collection."""
+
+import os
+from pathlib import Path
+
+from dotenv import load_dotenv
+
+from gyxx_flow.modules.shop_intelligence.runtime.paths import RAW_ROOT
+from gyxx_flow.adapters import RuntimeServicePolicy
+
+load_dotenv()
+os.environ.update(RuntimeServicePolicy().apply(os.environ))
+
+SCRIPT_DIR = str(Path(__file__).resolve().parent)
+DATA_DIR = str(RAW_ROOT)
+
+# ── PostgreSQL 持久化配置 ────────────────────────────────────────────────────
+DB_URL = os.getenv("DB_URL", "").strip()
+DB_HOST = os.getenv("DB_HOST", "")
+DB_PORT = int(os.getenv("DB_PORT", "5432"))
+DB_NAME = os.getenv("DB_NAME", "")
+DB_USER = os.getenv("DB_USER", "")
+DB_PASSWORD = os.getenv("DB_PASSWORD", "")
+DB_ENABLED = os.getenv("DB_ENABLED", "true").lower() in ("1", "true", "yes", "on")
+DB_CONNECT_TIMEOUT = int(os.getenv("DB_CONNECT_TIMEOUT", "10"))
+
+# ── 平台店铺账号 ─────────────────────────────────────────────────────────────
+JD_SHOP = os.getenv("JD_SHOP", "光影行星研发")
+JD_PASSWORD = os.getenv("JD_PASSWORD", "")
+JD_INDUSTRY = os.getenv("JD_INDUSTRY", "男包")
+JD_PLATFORM = "京东"
+
+DY_SHOP = os.getenv("DY_SHOP", "光影行星箱包旗舰店")
+DY_PASSWORD = os.getenv("DY_PASSWORD", "")
+DY_PLATFORM = "抖音"
+
+TM_SHOP = os.getenv("TM_SHOP", "光影行星旗舰店")
+TM_PASSWORD = os.getenv("TM_PASSWORD", "")
+TM_PLATFORM = "天猫"
+
+# ── 采集脚本路径 ─────────────────────────────────────────────────────────────
+JD_SHOP_SCRIPT = os.path.join(SCRIPT_DIR, "collectors", "jd_data_collector.py")
+JD_PEER_STORE_SCRIPT = os.path.join(SCRIPT_DIR, "collectors", "jd_peer_store_data_collector.py")
+DY_SCRIPT = os.path.join(SCRIPT_DIR, "collectors", "dy_store_competitor_store_scraping.py")
+TM_SHOP_SCRIPT = os.path.join(SCRIPT_DIR, "collectors", "taobao_sycm.py")
+
+# ── 数据输出目录 ─────────────────────────────────────────────────────────────
+JD_SHOP_OUTPUT = os.path.join(DATA_DIR, "京东", "店铺数据")
+JD_PEER_STORE_OUTPUT = os.path.join(DATA_DIR, "京东", "竞店数据")
+DY_OUTPUT = os.path.join(DATA_DIR, "抖音")
+DY_SHOP_OUTPUT = os.path.join(DY_OUTPUT, "店铺数据")
+DY_PEER_STORE_OUTPUT = os.path.join(DY_OUTPUT, "竞店数据")
+TM_SHOP_OUTPUT = os.path.join(DATA_DIR, "天猫", "店铺数据")
+
+# ── 飞书多维表格配置 ─────────────────────────────────────────────────────────
+BASE_APP_TOKEN = os.getenv("BASE_APP_TOKEN", "MOf8btDYTausLKsNiZlcqXzmnhc")
+LARK_PROFILE = os.getenv("LARK_PROFILE", "hermes-analyzer")
+
+# 店铺表（运营OKR（周）🚫🚫🚫）
+SHOP_BASE_TABLE_ID = os.getenv("SHOP_BASE_TABLE_ID", "tblsvXfxVkr3Zzwm")
+SHOP_BASE_FIELDS = {
+    "父记录": "fldRCOjzP0",
+    "时间": "fldHQecrCX",
+    "月份": "fldt9txdb8",
+    "平台": "fldlejDqfr",
+    "上周销售额": "fldogTrPCM",
+    "店铺评分": "fldygNpysx",
+    "AI分析": "fldqiw6Ens",
+    "备注": "fld00lqdcf",
+}
+
+# 京东店铺行需要挂在父记录下
+JD_SHOP_PARENT_RECORD_ID = os.getenv("JD_SHOP_PARENT_RECORD_ID", "recv7vJnwmVjan")
+
+# 竞店表 - 京东
+JD_PEER_BASE_TABLE_ID = os.getenv("JD_PEER_BASE_TABLE_ID", "tblCmbO2UeEkPVgi")
+JD_PEER_BASE_FIELDS = {
+    "时间": "fld8m0Cnnk",
+    "月份": "fldjBN5ai7",
+    "竞店名称": "fldW53GzXY",
+    "男士包袋": "fldwJcvPRv",
+    "双肩背包": "fldiYnlDK1",
+    "斜挎包": "fldDk9La5t",
+    "备注": "fldRsVjHOc",
+}
+JD_PEER_CATEGORY_MAP = {
+    "男士包袋": "fldwJcvPRv",
+    "男士双肩包": "fldiYnlDK1",
+    "男士单肩/斜挎包": "fldDk9La5t",
+}
+JD_PEER_BRAND_MAP = {
+    "Bellroy官方旗舰店": "Bellroy",
+    "tomtoc官方旗舰店": "TOMTOC",
+    "探迹者TAJEZZO官方旗舰店": "TAJEZZO",
+    "BAGSMART旗舰店": "BAGSMART",
+    "GROTTO个乐官方旗舰店": "GROTTO个乐",
+    "NIID箱包旗舰店": "NIID",
+    "光影行星 GYXX官方旗舰店": "光影行星⭐︎",
+}
+
+# 竞店表 - 抖音
+DY_PEER_BASE_TABLE_ID = os.getenv("DY_PEER_BASE_TABLE_ID", "tblleRSkrfRoLTms")
+DY_PEER_BASE_FIELDS = {
+    "时间": "fld8m0Cnnk",
+    "月份": "fldVC8HFtN",
+    "竞店名称": "fldW53GzXY",
+    "斜挎包": "fldiYnlDK1",
+    "双肩包": "fldwJcvPRv",
+    "备注": "fldp4e48ZY",
+}
+DY_PEER_CAT_MAP = {
+    "斜挎包": "单肩包/斜挎包",
+    "双肩包": "双肩包",
+}
+LOCAL_SHOP_BASE_OPTION = os.getenv("LOCAL_SHOP_BASE_OPTION", "光影行星⭐︎")
+
+# 天猫不写飞书表格
+FEISHU_PLATFORMS = {JD_PLATFORM, DY_PLATFORM}
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/db/__init__.py b/src/gyxx_flow/modules/shop_intelligence/runtime/db/__init__.py
new file mode 100644
index 0000000..c704fdd
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/db/__init__.py
@@ -0,0 +1,33 @@
+"""Unified PostgreSQL persistence for JD / DY / TM shop and peer-store data."""
+
+from .db import (
+    close_pool,
+    get_conn,
+    init_pool,
+    init_schema,
+    persist_dy_peer_store,
+    persist_dy_shop,
+    persist_jd_peer_store,
+    persist_jd_shop,
+    upsert_dy_peer_store,
+    upsert_dy_shop_weekly,
+    upsert_jd_peer_store,
+    upsert_jd_shop_weekly,
+    upsert_tm_shop_weekly,
+)
+
+__all__ = [
+    "close_pool",
+    "get_conn",
+    "init_pool",
+    "init_schema",
+    "persist_dy_peer_store",
+    "persist_dy_shop",
+    "persist_jd_peer_store",
+    "persist_jd_shop",
+    "upsert_dy_peer_store",
+    "upsert_dy_shop_weekly",
+    "upsert_jd_peer_store",
+    "upsert_jd_shop_weekly",
+    "upsert_tm_shop_weekly",
+]
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/db/db.py b/src/gyxx_flow/modules/shop_intelligence/runtime/db/db.py
new file mode 100644
index 0000000..9ddd3d3
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/db/db.py
@@ -0,0 +1,622 @@
+"""Unified PostgreSQL persistence for JD / DY / TM shop and peer-store data.
+
+Tables (pre-existing in data_hub):
+    jd_shop_weekly          UK (shop, week_start)
+    jd_peer_store_ranking   UK (week_start, brand, category)
+    dy_shop_weekly          UK (shop, week_start)
+    dy_peer_store_ranking   UK (week_start, brand, category)
+    tm_shop_weekly          UK (shop, week_start)
+
+Usage:
+    from gyxx_flow.modules.shop_intelligence.runtime.db import init_pool, get_conn, persist_jd_shop, persist_dy_shop, ...
+"""
+from __future__ import annotations
+
+import json
+import logging
+import os
+import re
+import threading
+from contextlib import contextmanager
+from datetime import date, datetime
+from typing import Any
+
+import psycopg2
+import psycopg2.extras
+from psycopg2 import pool
+
+from gyxx_flow.modules.shop_intelligence.runtime.config import (
+    DB_CONNECT_TIMEOUT,
+    DB_ENABLED,
+    DB_HOST,
+    DB_NAME,
+    DB_PASSWORD,
+    DB_PORT,
+    DB_URL,
+    DB_USER,
+    JD_PEER_BRAND_MAP,
+    SCRIPT_DIR,
+)
+
+log = logging.getLogger("db")
+
+_pool: pool.SimpleConnectionPool | None = None
+_pool_lock = threading.Lock()
+
+_NUM = re.compile(r"-?\d+(?:\.\d+)?")
+
+
+def _resolve_dsn() -> str:
+    if DB_URL:
+        return DB_URL
+    return (
+        f"host={DB_HOST} port={DB_PORT} dbname={DB_NAME} "
+        f"user={DB_USER} password={DB_PASSWORD} connect_timeout={DB_CONNECT_TIMEOUT}"
+    )
+
+
+def init_pool(minconn: int = 1, maxconn: int = 4) -> None:
+    global _pool
+    with _pool_lock:
+        if _pool is not None:
+            return
+        if not DB_ENABLED:
+            log.info("DB_ENABLED=false, skip pool init")
+            return
+        _pool = pool.SimpleConnectionPool(minconn, maxconn, dsn=_resolve_dsn())
+        log.info("DB pool initialized host=%s db=%s", DB_HOST, DB_NAME)
+
+
+def close_pool() -> None:
+    global _pool
+    with _pool_lock:
+        if _pool is None:
+            return
+        _pool.closeall()
+        _pool = None
+
+
+def init_schema() -> None:
+    """执行 db/schema.sql 创建表（幂等）。"""
+    if not DB_ENABLED:
+        return
+    schema_path = os.path.join(SCRIPT_DIR, "db", "schema.sql")
+    if not os.path.exists(schema_path):
+        log.warning("schema.sql not found: %s", schema_path)
+        return
+    with open(schema_path, encoding="utf-8") as f:
+        sql = f.read()
+    with get_conn() as conn:
+        with conn.cursor() as cur:
+            cur.execute(sql)
+
+
+@contextmanager
+def get_conn():
+    if _pool is None:
+        init_pool()
+    if _pool is None:
+        raise RuntimeError("DB not enabled or pool not initialized")
+    c = _pool.getconn()
+    try:
+        yield c
+        c.commit()
+    except Exception:
+        c.rollback()
+        raise
+    finally:
+        _pool.putconn(c)
+
+
+def _clean_val(v):
+    if v is None:
+        return None
+    if isinstance(v, str) and v.strip() in ("-", "—", "N/A", "n/a", ""):
+        return None
+    return v
+
+
+def _get_table_columns(conn, table: str) -> set[str]:
+    with conn.cursor() as cur:
+        cur.execute(
+            "SELECT column_name FROM information_schema.columns "
+            "WHERE table_name = %s",
+            (table,),
+        )
+        return {row[0] for row in cur.fetchall()}
+
+
+def _dynamic_upsert(conn, table: str, record: dict, uk_cols: tuple[str, ...]) -> int:
+    if not record:
+        return 0
+    table_cols = _get_table_columns(conn, table)
+    cols = [c for c in record.keys() if c in table_cols]
+    if not cols:
+        return 0
+    values = [record[c] for c in cols]
+    placeholders = ", ".join(["%s"] * len(cols))
+    col_list = ", ".join(f'"{c}"' for c in cols)
+    update_cols = [c for c in cols if c not in uk_cols and c != "created_at"]
+    update_set = ", ".join(f'"{c}" = EXCLUDED."{c}"' for c in update_cols)
+    uk_list = ", ".join(uk_cols)
+    sql = (
+        f'INSERT INTO {table} ({col_list}) VALUES ({placeholders}) '
+        f'ON CONFLICT ({uk_list}) DO UPDATE SET {update_set}, updated_at = now()'
+    )
+    with conn.cursor() as cur:
+        cur.execute(sql, values)
+        return cur.rowcount
+
+
+def _dynamic_batch_upsert(
+    conn, table: str, records: list[dict], uk_cols: tuple[str, ...]
+) -> int:
+    if not records:
+        return 0
+    table_cols = _get_table_columns(conn, table)
+    cols = [c for c in records[0].keys() if c in table_cols]
+    if not cols:
+        return 0
+    placeholders = ", ".join(["%s"] * len(cols))
+    col_list = ", ".join(f'"{c}"' for c in cols)
+    update_cols = [c for c in cols if c not in uk_cols and c != "created_at"]
+    update_set = ", ".join(f'"{c}" = EXCLUDED."{c}"' for c in update_cols)
+    uk_list = ", ".join(uk_cols)
+    sql = (
+        f'INSERT INTO {table} ({col_list}) VALUES ({placeholders}) '
+        f'ON CONFLICT ({uk_list}) DO UPDATE SET {update_set}'
+    )
+    data = [[r.get(c) for c in cols] for r in records]
+    with conn.cursor() as cur:
+        psycopg2.extras.execute_batch(cur, sql, data, page_size=200)
+        return cur.rowcount
+
+
+# ── 通用工具 ──────────────────────────────────────────────────────────────────
+
+
+def _to_float(s: Any) -> float | None:
+    if s is None:
+        return None
+    if isinstance(s, (int, float)):
+        return float(s)
+    text = str(s).strip()
+    if not text or text.upper() in {"N/A", "--", "—", "未获取", "未采集到"}:
+        return None
+    m = _NUM.search(text.replace(",", ""))
+    return float(m.group(0)) if m else None
+
+
+def _to_int(s: Any) -> int | None:
+    v = _to_float(s)
+    return int(v) if v is not None else None
+
+
+def _to_date(s: str) -> date | None:
+    if not s:
+        return None
+    text = str(s).strip()[:10]
+    for fmt in ("%Y-%m-%d", "%Y/%m/%d", "%Y.%m.%d"):
+        try:
+            return datetime.strptime(text, fmt).date()
+        except ValueError:
+            continue
+    return None
+
+
+def _parse_period(period: str) -> tuple[date | None, date | None, str]:
+    if not period or "~" not in period:
+        return None, None, period or ""
+    left, right = [x.strip() for x in period.split("~", 1)]
+    return _to_date(left), _to_date(right), period.strip()
+
+
+def _parse_wan(s: Any) -> int | None:
+    if s is None:
+        return None
+    if isinstance(s, (int, float)):
+        return int(s)
+    text = str(s).strip()
+    if not text or text.upper() in {"N/A", "--", "—", "未获取"}:
+        return None
+    m = re.match(r"^\s*(-?\d+(?:\.\d+)?)\s*万\s*$", text)
+    if m:
+        return int(float(m.group(1)) * 10000)
+    m = _NUM.search(text.replace(",", ""))
+    return int(float(m.group(0))) if m else None
+
+
+def _parse_yuan(s: Any) -> float | None:
+    if s is None:
+        return None
+    if isinstance(s, (int, float)):
+        return float(s)
+    text = str(s).strip()
+    if not text or text.upper() in {"N/A", "--", "—", "未获取"}:
+        return None
+    m = _NUM.search(text.replace(",", ""))
+    return float(m.group(0)) if m else None
+
+
+def _parse_iso_dt(s: str) -> datetime | None:
+    if not s:
+        return None
+    text = str(s).strip()
+    if text.endswith("Z"):
+        text = text[:-1] + "+00:00"
+    try:
+        return datetime.fromisoformat(text)
+    except ValueError:
+        return None
+
+
+# ── JD 店铺 ───────────────────────────────────────────────────────────────────
+
+_TRANSACTION_FIELDS: list[tuple[str, str, callable]] = [
+    ("成交金额", "deal_amount", _to_float),
+    ("成交金额", "deal_amount_qoq", lambda x: str(x.get("环比")) if isinstance(x, dict) and x.get("环比") else None),
+    ("成交金额", "deal_amount_peer_avg", lambda x: str(x.get("同行同级均值")) if isinstance(x, dict) and x.get("同行同级均值") else None),
+    ("商品访客数", "visitors", _to_int),
+    ("商品访客数", "visitors_qoq", lambda x: str(x.get("环比")) if isinstance(x, dict) and x.get("环比") else None),
+    ("商品访客数", "visitors_peer_avg", lambda x: str(x.get("同行同级均值")) if isinstance(x, dict) and x.get("同行同级均值") else None),
+    ("成交客户数", "paying_customers", _to_int),
+    ("成交客户数", "paying_customers_qoq", lambda x: str(x.get("环比")) if isinstance(x, dict) and x.get("环比") else None),
+    ("成交客户数", "paying_customers_peer_avg", lambda x: str(x.get("同行同级均值")) if isinstance(x, dict) and x.get("同行同级均值") else None),
+    ("成交转化率", "conversion_rate", _to_float),
+    ("成交转化率", "conversion_rate_qoq", lambda x: str(x.get("环比")) if isinstance(x, dict) and x.get("环比") else None),
+    ("成交转化率", "conversion_rate_peer_avg", lambda x: str(x.get("同行同级均值")) if isinstance(x, dict) and x.get("同行同级均值") else None),
+    ("成交单量", "order_qty", _to_int),
+    ("成交单量", "order_qty_qoq", lambda x: str(x.get("环比")) if isinstance(x, dict) and x.get("环比") else None),
+    ("成交单量", "order_qty_peer_avg", lambda x: str(x.get("同行同级均值")) if isinstance(x, dict) and x.get("同行同级均值") else None),
+    ("取消及售后退款金额", "refund_amount", _to_float),
+    ("取消及售后退款金额", "refund_amount_qoq", lambda x: str(x.get("环比")) if isinstance(x, dict) and x.get("环比") else None),
+    ("客单价", "avg_order_value", _to_float),
+    ("客单价", "avg_order_value_qoq", lambda x: str(x.get("环比")) if isinstance(x, dict) and x.get("环比") else None),
+    ("客单价", "avg_order_value_peer_avg", lambda x: str(x.get("同行同级均值")) if isinstance(x, dict) and x.get("同行同级均值") else None),
+]
+
+_JZT_NON_FIELDS: list[tuple[str, str, callable]] = [
+    ("cost", "jzt_non_cost", _to_float),
+    ("cost", "jzt_non_cost_change", lambda x: str(x.get("change")) if isinstance(x, dict) and x.get("change") else None),
+    ("impressions", "jzt_non_impressions", _to_int),
+    ("clicks", "jzt_non_clicks", _to_int),
+    ("CTR", "jzt_non_ctr", _to_float),
+    ("CPC", "jzt_non_cpc", _to_float),
+    ("CPM", "jzt_non_cpm", _to_float),
+    ("CVS", "jzt_non_cvs", _to_int),
+    ("CPA", "jzt_non_cpa", _to_float),
+    ("ROI", "jzt_non_roi", _to_float),
+    ("totalOrderCnt", "jzt_non_total_order_cnt", _to_int),
+    ("totalOrderSum", "jzt_non_total_order_sum", _to_float),
+    ("totalCartCnt", "jzt_non_total_cart_cnt", _to_int),
+]
+
+_JZT_FULL_FIELDS: list[tuple[str, str, callable]] = [
+    ("cost", "jzt_full_cost", _to_float),
+    ("cost", "jzt_full_cost_change", lambda x: str(x.get("change")) if isinstance(x, dict) and x.get("change") else None),
+    ("swaImpressions", "jzt_full_impressions", _to_int),
+    ("swaClicks", "jzt_full_clicks", _to_int),
+    ("swaOrderROI", "jzt_full_order_roi", _to_float),
+    ("swaOrderCnt", "jzt_full_order_cnt", _to_int),
+    ("swaOrderSum", "jzt_full_order_sum", _to_float),
+    ("swaOrderCPA", "jzt_full_order_cpa", _to_float),
+    ("ROI", "jzt_full_roi", _to_float),
+    ("全站点击量", "jzt_full_clicks_all", _to_int),
+    ("全站交易额", "jzt_full_gmv_all", _to_float),
+]
+
+
+def _build_jd_shop_row(metrics: dict, raw_path: str | None = None) -> dict | None:
+    week_start = _to_date(metrics.get("week_start", ""))
+    week_end = _to_date(metrics.get("week_end", ""))
+    if not week_start or not week_end:
+        return None
+    data_date = _to_date(metrics.get("data_date", "")) or week_end
+
+    row: dict[str, Any] = {
+        "shop": metrics.get("shop", ""),
+        "industry": metrics.get("industry", ""),
+        "caliber_days": int(metrics.get("caliber_days") or 7),
+        "data_date": data_date,
+        "week_start": week_start,
+        "week_end": week_end,
+        "week_range": metrics.get("week_range", f"{week_start} ~ {week_end}"),
+        "star_rating_raw": metrics.get("star_rating_raw"),
+    }
+    sr = metrics.get("star_rating")
+    if sr is None:
+        sr = _to_float(metrics.get("star_rating_raw", ""))
+    row["star_rating"] = sr
+
+    txn = metrics.get("transaction", {}) or {}
+    for json_key, col, conv in _TRANSACTION_FIELDS:
+        v = txn.get(json_key)
+        row[col] = conv(v) if v is not None else None
+
+    jzt_non = metrics.get("jzt_non_full", {}) or {}
+    for json_key, col, conv in _JZT_NON_FIELDS:
+        v = jzt_non.get(json_key)
+        row[col] = conv(v) if v is not None else None
+
+    jzt_full = metrics.get("jzt_full", {}) or {}
+    for json_key, col, conv in _JZT_FULL_FIELDS:
+        v = jzt_full.get(json_key)
+        row[col] = conv(v) if v is not None else None
+
+    rankings = {}
+    for cat, data in (metrics.get("ranking", {}) or {}).items():
+        if isinstance(data, dict):
+            rankings[cat] = data
+    row["rankings"] = json.dumps(rankings, ensure_ascii=False, default=str)
+    row["raw_metrics_path"] = raw_path
+    row["ai_analysis"] = metrics.get("ai_analysis", "")
+    return row
+
+
+def upsert_jd_shop_weekly(conn, record: dict) -> int:
+    if not DB_ENABLED:
+        return 0
+    cleaned = {k: _clean_val(v) for k, v in record.items()}
+    if "rankings" in cleaned and isinstance(cleaned["rankings"], dict):
+        cleaned["rankings"] = json.dumps(cleaned["rankings"], ensure_ascii=False)
+    return _dynamic_upsert(conn, "jd_shop_weekly", cleaned, ("shop", "week_start"))
+
+
+# ── JD 竞店 ───────────────────────────────────────────────────────────────────
+
+
+def _build_jd_peer_store_rows(data: dict, raw_path: str | None = None) -> list[dict]:
+    ws, we, wr = _parse_period(data.get("数据周期", ""))
+    if not ws or not we:
+        raise ValueError(f"竞店 JSON 缺少有效 数据周期: {data.get('数据周期')}")
+    collect_date = _to_date(data.get("采集日期", "")) or we
+    out: list[dict] = []
+    own_key = "光影行星 GYXX官方旗舰店"
+    for brand_key, brand_data in (data.get("数据", {}) or {}).items():
+        if not isinstance(brand_data, dict):
+            continue
+        alias = JD_PEER_BRAND_MAP.get(brand_key)
+        is_mine = brand_key == own_key
+        for category, info in brand_data.items():
+            if not isinstance(info, dict):
+                continue
+            out.append({
+                "collect_date": collect_date,
+                "week_start": ws,
+                "week_end": we,
+                "week_range": wr,
+                "brand": brand_key,
+                "brand_alias": alias,
+                "category": category,
+                "rank": _to_int(info.get("rank")),
+                "amount_band": info.get("amount"),
+                "orders_band": info.get("orders"),
+                "followers": str(info.get("followers")) if info.get("followers") not in (None, "") else None,
+                "visitors_band": info.get("visitors"),
+                "search_clicks_band": info.get("searchClicks"),
+                "is_mine": is_mine,
+                "raw_json_path": raw_path,
+            })
+    return out
+
+
+def upsert_jd_peer_store(conn, records: list[dict]) -> int:
+    if not DB_ENABLED:
+        return 0
+    if not records:
+        return 0
+    cleaned_records = []
+    for r in records:
+        cr = {k: _clean_val(v) for k, v in r.items()}
+        cleaned_records.append(cr)
+    return _dynamic_batch_upsert(
+        conn, "jd_peer_store_ranking", cleaned_records, ("week_start", "brand", "category")
+    )
+
+
+# ── DY 店铺 ───────────────────────────────────────────────────────────────────
+
+
+def _ensure_dy_shop_weekly_columns() -> None:
+    """幂等添加 dy_shop_weekly 可能缺少的新列。"""
+    if not DB_ENABLED:
+        return
+    try:
+        with get_conn() as c:
+            with c.cursor() as cur:
+                cur.execute(
+                    "SELECT column_name FROM information_schema.columns "
+                    "WHERE table_name = 'dy_shop_weekly'"
+                )
+                existing = {row[0] for row in cur.fetchall()}
+                for col, col_type in [
+                    ("visitors_raw", "text"),
+                    ("visitors", "integer"),
+                    ("deal_orders_raw", "text"),
+                    ("deal_orders", "integer"),
+                ]:
+                    if col not in existing:
+                        log.info("ALTER TABLE dy_shop_weekly ADD COLUMN %s %s", col, col_type)
+                        cur.execute(f'ALTER TABLE dy_shop_weekly ADD COLUMN "{col}" {col_type}')
+    except Exception as e:
+        log.warning("ensure_dy_shop_weekly_columns failed (non-fatal): %s", e)
+
+
+def _build_dy_shop_row(shop: dict, raw_path: str | None = None) -> dict | None:
+    ws, we, wr = _parse_dy_period(shop.get("date_range", ""))
+    if not ws or not we:
+        return None
+    metrics = shop.get("metrics", {}) or {}
+    derived = shop.get("derived_metrics", {}) or {}
+    qc = shop.get("qianchuan_metrics", {}) or {}
+    es = shop.get("experience_score", {}) or {}
+    mr = shop.get("market_rankings", {}) or {}
+
+    row: dict[str, Any] = {
+        "shop": shop.get("shop_name", ""),
+        "scraped_at": _parse_iso_dt(shop.get("scraped_at", "")),
+        "source_url": shop.get("url"),
+        "week_start": ws,
+        "week_end": we,
+        "week_range": wr,
+        "deal_amount_raw": metrics.get("成交金额"),
+        "deal_amount": _parse_yuan(metrics.get("成交金额")),
+        "paying_users_raw": metrics.get("成交人数"),
+        "paying_users": _to_int(metrics.get("成交人数")),
+        "deal_orders_raw": metrics.get("成交订单数"),
+        "deal_orders": _to_int(metrics.get("成交订单数")),
+        "avg_order_value_raw": metrics.get("客单价"),
+        "avg_order_value": _parse_yuan(metrics.get("客单价")),
+        "refund_amount_raw": metrics.get("成交退款金额"),
+        "refund_amount": _parse_yuan(metrics.get("成交退款金额")),
+        "product_clicks_raw": metrics.get("商品点击人数"),
+        "product_clicks": _parse_wan(metrics.get("商品点击人数")),
+        "visitors_raw": metrics.get("访客数"),
+        "visitors": _parse_wan(metrics.get("访客数")),
+        "derived_sales_raw": derived.get("销售额"),
+        "derived_sales": _parse_yuan(derived.get("销售额")),
+        "derived_paid_ratio_raw": derived.get("付费占比"),
+        "derived_roi_raw": derived.get("ROI"),
+        "qianchuan_captured_at": _parse_iso_dt(qc.get("captured_at", "")),
+        "qianchuan_url": qc.get("url"),
+        "qianchuan_data": json.dumps({k: v for k, v in qc.items() if k not in ("captured_at", "url")}, ensure_ascii=False, default=str),
+        "experience_captured_at": _parse_iso_dt(es.get("captured_at", "")),
+        "experience_url": es.get("url"),
+        "experience_data": json.dumps({k: v for k, v in es.items() if k not in ("captured_at", "url")}, ensure_ascii=False, default=str),
+        "market_rankings": json.dumps(mr, ensure_ascii=False, default=str),
+        "raw_json_path": raw_path,
+        "ai_analysis": shop.get("ai_analysis", ""),
+    }
+    return row
+
+
+def _parse_dy_period(period: str) -> tuple[date | None, date | None, str]:
+    if not period or "-" not in period:
+        return None, None, period or ""
+    left, right = [x.strip() for x in period.split("-", 1)]
+    return _to_date(left), _to_date(right), period.strip()
+
+
+def upsert_dy_shop_weekly(conn, record: dict) -> int:
+    if not DB_ENABLED:
+        return 0
+    cleaned = {k: _clean_val(v) for k, v in record.items()}
+    for json_col in ("qianchuan_data", "experience_data", "market_rankings"):
+        if json_col in cleaned and isinstance(cleaned[json_col], dict):
+            cleaned[json_col] = json.dumps(
+                cleaned[json_col], ensure_ascii=False, default=str
+            )
+    _ensure_dy_shop_weekly_columns()
+    return _dynamic_upsert(conn, "dy_shop_weekly", cleaned, ("shop", "week_start"))
+
+
+# ── DY 竞店 ───────────────────────────────────────────────────────────────────
+
+
+def _build_dy_peer_store_rows(data: dict, raw_path: str | None = None) -> list[dict]:
+    ws, we, wr = _parse_dy_period(data.get("date_range", ""))
+    if not ws or not we:
+        raise ValueError(f"peer store JSON 缺少有效 date_range: {data.get('date_range')}")
+    captured_at = _parse_iso_dt(data.get("captured_at", ""))
+    results = data.get("results", {}) or {}
+    collect_date = ws
+    out: list[dict] = []
+    for category, brand_map in results.items():
+        if not isinstance(brand_map, dict):
+            continue
+        for brand, row_data in brand_map.items():
+            if not isinstance(row_data, dict):
+                continue
+            rank_str = str(row_data.get("rank") or "").strip()
+            rank: int | None = None
+            if rank_str and rank_str.lstrip("-").isdigit():
+                rank = int(rank_str)
+            direction = row_data.get("rank_change_direction")
+            change_val = _to_int(row_data.get("rank_change_value"))
+            out.append({
+                "collect_date": collect_date,
+                "week_start": ws,
+                "week_end": we,
+                "week_range": wr,
+                "brand": brand,
+                "category": category,
+                "rank": rank,
+                "rank_change_direction": direction if direction not in (None, "N/A") else None,
+                "rank_change_value": change_val,
+                "row_data": json.dumps(row_data, ensure_ascii=False, default=str),
+                "raw_json_path": raw_path,
+                "captured_at": _parse_iso_dt(row_data.get("captured_at", "")) or captured_at,
+            })
+    return out
+
+
+def upsert_dy_peer_store(conn, records: list[dict]) -> int:
+    if not DB_ENABLED:
+        return 0
+    if not records:
+        return 0
+    cleaned_records = []
+    for r in records:
+        cr = {k: _clean_val(v) for k, v in r.items()}
+        for json_col in ("row_data",):
+            if json_col in cr and isinstance(cr[json_col], dict):
+                cr[json_col] = json.dumps(cr[json_col], ensure_ascii=False, default=str)
+        cleaned_records.append(cr)
+    return _dynamic_batch_upsert(
+        conn, "dy_peer_store_ranking", cleaned_records, ("week_start", "brand", "category")
+    )
+
+
+# ── TM 店铺 ───────────────────────────────────────────────────────────────────
+
+
+def upsert_tm_shop_weekly(conn, record: dict) -> int:
+    if not DB_ENABLED:
+        return 0
+    if not record:
+        return 0
+    cleaned = {k: _clean_val(v) for k, v in record.items()}
+    if "rankings" in cleaned and isinstance(cleaned["rankings"], dict):
+        cleaned["rankings"] = json.dumps(cleaned["rankings"], ensure_ascii=False)
+    return _dynamic_upsert(conn, "tm_shop_weekly", cleaned, ("shop", "week_start"))
+
+
+# ── 顶层入口：runner 直接调用 ─────────────────────────────────────────────────
+
+
+def persist_jd_shop(metrics_path: str) -> int:
+    if not DB_ENABLED or not metrics_path or not os.path.exists(metrics_path):
+        return 0
+    with open(metrics_path, encoding="utf-8") as f:
+        metrics = json.load(f)
+    with get_conn() as conn:
+        return upsert_jd_shop_weekly(conn, _build_jd_shop_row(metrics, metrics_path) or {})
+
+
+def persist_jd_peer_store(json_path: str) -> int:
+    if not DB_ENABLED or not json_path or not os.path.exists(json_path):
+        return 0
+    with open(json_path, encoding="utf-8") as f:
+        data = json.load(f)
+    rows = _build_jd_peer_store_rows(data, json_path)
+    with get_conn() as conn:
+        return upsert_jd_peer_store(conn, rows)
+
+
+def persist_dy_shop(json_path: str) -> int:
+    if not DB_ENABLED or not json_path or not os.path.exists(json_path):
+        return 0
+    with open(json_path, encoding="utf-8") as f:
+        data = json.load(f)
+    with get_conn() as conn:
+        return upsert_dy_shop_weekly(conn, _build_dy_shop_row(data, json_path) or {})
+
+
+def persist_dy_peer_store(json_path: str) -> int:
+    if not DB_ENABLED or not json_path or not os.path.exists(json_path):
+        return 0
+    with open(json_path, encoding="utf-8") as f:
+        data = json.load(f)
+    rows = _build_dy_peer_store_rows(data, json_path)
+    with get_conn() as conn:
+        return upsert_dy_peer_store(conn, rows)
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/db/schema.sql b/src/gyxx_flow/modules/shop_intelligence/runtime/db/schema.sql
new file mode 100644
index 0000000..acb7cab
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/db/schema.sql
@@ -0,0 +1,235 @@
+-- ============================================================================
+-- shop-data-flow / PostgreSQL schema
+-- 数据库: data_hub
+-- 整合 JD / DY / TM 三平台店铺与竞店数据持久化表
+-- ============================================================================
+
+-- ── 京东店铺周指标 ────────────────────────────────────────────────────────────
+CREATE TABLE IF NOT EXISTS jd_shop_weekly (
+    id              BIGSERIAL PRIMARY KEY,
+    shop            TEXT NOT NULL,
+    industry        TEXT NOT NULL,
+    caliber_days    INT  NOT NULL,
+    data_date       DATE NOT NULL,
+    week_start      DATE NOT NULL,
+    week_end        DATE NOT NULL,
+    week_range      TEXT NOT NULL,
+    star_rating     NUMERIC(3,1),
+    star_rating_raw TEXT,
+
+    deal_amount               NUMERIC(18,2),
+    deal_amount_qoq           TEXT,
+    deal_amount_peer_avg      TEXT,
+    visitors                  BIGINT,
+    visitors_qoq              TEXT,
+    visitors_peer_avg         TEXT,
+    paying_customers          BIGINT,
+    paying_customers_qoq      TEXT,
+    paying_customers_peer_avg TEXT,
+    conversion_rate           NUMERIC(6,3),
+    conversion_rate_qoq       TEXT,
+    conversion_rate_peer_avg  TEXT,
+    order_qty                 BIGINT,
+    order_qty_qoq             TEXT,
+    order_qty_peer_avg        TEXT,
+    refund_amount             NUMERIC(18,2),
+    refund_amount_qoq         TEXT,
+    avg_order_value           NUMERIC(18,2),
+    avg_order_value_qoq       TEXT,
+    avg_order_value_peer_avg  TEXT,
+
+    jzt_non_cost             NUMERIC(18,2),
+    jzt_non_cost_change      TEXT,
+    jzt_non_impressions      BIGINT,
+    jzt_non_clicks           BIGINT,
+    jzt_non_ctr              NUMERIC(8,4),
+    jzt_non_cpc              NUMERIC(18,4),
+    jzt_non_cpm              NUMERIC(18,4),
+    jzt_non_cvs              BIGINT,
+    jzt_non_cpa              NUMERIC(18,2),
+    jzt_non_roi              NUMERIC(10,4),
+    jzt_non_total_order_cnt  BIGINT,
+    jzt_non_total_order_sum  NUMERIC(18,2),
+    jzt_non_total_cart_cnt   BIGINT,
+
+    jzt_full_cost            NUMERIC(18,2),
+    jzt_full_cost_change     TEXT,
+    jzt_full_impressions     BIGINT,
+    jzt_full_clicks          BIGINT,
+    jzt_full_order_roi       NUMERIC(10,4),
+    jzt_full_order_cnt       BIGINT,
+    jzt_full_order_sum       NUMERIC(18,2),
+    jzt_full_order_cpa       NUMERIC(18,2),
+    jzt_full_roi             NUMERIC(10,4),
+    jzt_full_clicks_all      BIGINT,
+    jzt_full_gmv_all         NUMERIC(18,2),
+
+    rankings            JSONB NOT NULL DEFAULT '{}'::jsonb,
+    raw_metrics_path    TEXT,
+    created_at          TIMESTAMPTZ NOT NULL DEFAULT now(),
+    updated_at          TIMESTAMPTZ NOT NULL DEFAULT now(),
+
+    CONSTRAINT jd_shop_weekly_uniq UNIQUE (shop, week_start)
+);
+ALTER TABLE jd_shop_weekly ADD COLUMN IF NOT EXISTS ai_analysis TEXT;
+CREATE INDEX IF NOT EXISTS idx_shop_weekly_date ON jd_shop_weekly (week_end DESC);
+CREATE INDEX IF NOT EXISTS idx_shop_weekly_shop  ON jd_shop_weekly (shop, week_end DESC);
+
+
+-- ── 京东竞店周排名 ────────────────────────────────────────────────────────────
+CREATE TABLE IF NOT EXISTS jd_peer_store_ranking (
+    id                BIGSERIAL PRIMARY KEY,
+    collect_date      DATE NOT NULL,
+    week_start        DATE NOT NULL,
+    week_end          DATE NOT NULL,
+    week_range        TEXT NOT NULL,
+    brand             TEXT NOT NULL,
+    brand_alias       TEXT,
+    category          TEXT NOT NULL,
+    rank              INT,
+    amount_band       TEXT,
+    orders_band       TEXT,
+    followers         TEXT,
+    visitors_band     TEXT,
+    search_clicks_band TEXT,
+    is_mine           BOOLEAN NOT NULL DEFAULT FALSE,
+    raw_json_path     TEXT,
+    created_at        TIMESTAMPTZ NOT NULL DEFAULT now(),
+
+    CONSTRAINT jd_peer_store_uniq UNIQUE (week_start, brand, category)
+);
+CREATE INDEX IF NOT EXISTS idx_peer_store_date     ON jd_peer_store_ranking (week_end DESC, category);
+CREATE INDEX IF NOT EXISTS idx_peer_store_brand    ON jd_peer_store_ranking (brand, week_end DESC);
+
+
+-- ── 抖音店铺周指标 ────────────────────────────────────────────────────────────
+CREATE TABLE IF NOT EXISTS dy_shop_weekly (
+    id              BIGSERIAL PRIMARY KEY,
+    shop            TEXT NOT NULL,
+    scraped_at      TIMESTAMPTZ,
+    source_url      TEXT,
+    week_start      DATE NOT NULL,
+    week_end        DATE NOT NULL,
+    week_range      TEXT NOT NULL,
+
+    deal_amount_raw     TEXT,
+    deal_amount         NUMERIC(18,2),
+    paying_users_raw    TEXT,
+    paying_users        BIGINT,
+    deal_orders_raw     TEXT,
+    deal_orders         BIGINT,
+    avg_order_value_raw TEXT,
+    avg_order_value     NUMERIC(18,2),
+    refund_amount_raw   TEXT,
+    refund_amount       NUMERIC(18,2),
+    product_clicks_raw  TEXT,
+    product_clicks      BIGINT,
+    visitors_raw        TEXT,
+    visitors            BIGINT,
+
+    derived_sales_raw       TEXT,
+    derived_sales           NUMERIC(18,2),
+    derived_paid_ratio_raw  TEXT,
+    derived_roi_raw         TEXT,
+
+    qianchuan_captured_at   TIMESTAMPTZ,
+    qianchuan_url           TEXT,
+    qianchuan_data          JSONB NOT NULL DEFAULT '{}'::jsonb,
+
+    experience_captured_at  TIMESTAMPTZ,
+    experience_url          TEXT,
+    experience_data         JSONB NOT NULL DEFAULT '{}'::jsonb,
+
+    market_rankings         JSONB NOT NULL DEFAULT '{}'::jsonb,
+
+    raw_json_path           TEXT,
+    created_at              TIMESTAMPTZ NOT NULL DEFAULT now(),
+    updated_at              TIMESTAMPTZ NOT NULL DEFAULT now(),
+
+    CONSTRAINT dy_shop_weekly_uniq UNIQUE (shop, week_start)
+);
+ALTER TABLE dy_shop_weekly ADD COLUMN IF NOT EXISTS ai_analysis TEXT;
+CREATE INDEX IF NOT EXISTS idx_dy_shop_weekly_date ON dy_shop_weekly (week_end DESC);
+CREATE INDEX IF NOT EXISTS idx_dy_shop_weekly_shop ON dy_shop_weekly (shop, week_end DESC);
+
+
+-- ── 抖音竞店周排名 ────────────────────────────────────────────────────────────
+CREATE TABLE IF NOT EXISTS dy_peer_store_ranking (
+    id              BIGSERIAL PRIMARY KEY,
+    collect_date    DATE NOT NULL,
+    week_start      DATE NOT NULL,
+    week_end        DATE NOT NULL,
+    week_range      TEXT NOT NULL,
+    brand           TEXT NOT NULL,
+    category        TEXT NOT NULL,
+    rank            INT,
+    rank_change_direction TEXT,
+    rank_change_value     INT,
+    row_data        JSONB NOT NULL DEFAULT '{}'::jsonb,
+    raw_json_path   TEXT,
+    captured_at     TIMESTAMPTZ,
+    created_at      TIMESTAMPTZ NOT NULL DEFAULT now(),
+
+    CONSTRAINT dy_peer_store_uniq UNIQUE (week_start, brand, category)
+);
+CREATE INDEX IF NOT EXISTS idx_dy_peer_store_date  ON dy_peer_store_ranking (week_end DESC, category);
+CREATE INDEX IF NOT EXISTS idx_dy_peer_store_brand ON dy_peer_store_ranking (brand, week_end DESC);
+
+
+-- ── 天猫店铺周指标 ────────────────────────────────────────────────────────────
+CREATE TABLE IF NOT EXISTS tm_shop_weekly (
+    id                      BIGSERIAL   PRIMARY KEY,
+    shop                    TEXT        NOT NULL DEFAULT '光影行星旗舰店',
+    industry                TEXT        NOT NULL DEFAULT '',
+    caliber_days            INTEGER     NOT NULL DEFAULT 7,
+    data_date               DATE        NOT NULL,
+    week_start              DATE        NOT NULL,
+    week_end                DATE        NOT NULL,
+    week_range              TEXT,
+
+    star_rating             NUMERIC(3, 1),
+    star_rating_raw         TEXT,
+
+    deal_amount             NUMERIC(18, 2),
+    deal_amount_qoq         TEXT,
+    deal_amount_peer_avg    TEXT,
+    visitors                BIGINT,
+    visitors_qoq            TEXT,
+    visitors_peer_avg       TEXT,
+    paying_customers        BIGINT,
+    paying_customers_qoq    TEXT,
+    paying_customers_peer_avg TEXT,
+    conversion_rate         NUMERIC(6, 3),
+    conversion_rate_qoq     TEXT,
+    conversion_rate_peer_avg TEXT,
+    order_qty               BIGINT,
+    order_qty_qoq           TEXT,
+    order_qty_peer_avg      TEXT,
+    refund_amount           NUMERIC(18, 2),
+    refund_amount_qoq       TEXT,
+    avg_order_value         NUMERIC(18, 2),
+    avg_order_value_qoq     TEXT,
+    avg_order_value_peer_avg TEXT,
+
+    promotion_cost          NUMERIC(18, 2),
+    promotion_cost_qoq      TEXT,
+    roi                     NUMERIC(10, 4),
+    roi_qoq                 TEXT,
+    promotion_ratio         TEXT,
+
+    rankings                JSONB       NOT NULL DEFAULT '{}',
+
+    raw_metrics_path        TEXT,
+    created_at              TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    updated_at              TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+
+    UNIQUE (shop, week_start)
+);
+ALTER TABLE tm_shop_weekly ADD COLUMN IF NOT EXISTS ai_analysis TEXT;
+
+CREATE INDEX IF NOT EXISTS idx_tm_shop_weekly_date
+    ON tm_shop_weekly (week_end DESC);
+CREATE INDEX IF NOT EXISTS idx_tm_shop_weekly_shop
+    ON tm_shop_weekly (shop, week_end DESC);
+
+COMMENT ON TABLE tm_shop_weekly IS '天猫店铺周指标 — 一行=一周，对齐 jd_shop_weekly 设计。来源: taobao_sycm.py 采集生意参谋首页+市场排行+店铺评分。';
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/dependencies.toml b/src/gyxx_flow/modules/shop_intelligence/runtime/dependencies.toml
new file mode 100644
index 0000000..8de3db0
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/dependencies.toml
@@ -0,0 +1,10 @@
+[project]
+name = "shop-data-flow"
+version = "0.1.0"
+requires-python = ">=3.12"
+dependencies = [
+    "playwright>=1.60.0",
+    "scrapling[all]>=0.4.9",
+    "psycopg2-binary>=2.9.12",
+    "python-dotenv>=1.0.0",
+]
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/lark_cli.py b/src/gyxx_flow/modules/shop_intelligence/runtime/lark_cli.py
new file mode 100644
index 0000000..e44da29
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/lark_cli.py
@@ -0,0 +1,289 @@
+"""Shared lark-cli helpers for Feishu Bitable read/write.
+
+Provides:
+    resolve_lark_cli()  – find lark-cli + node paths
+    run_lark_cli()      – run lark-cli subprocess
+    format_week()       – format week string (M月第N周)
+    parse_amount()      – parse currency string
+    parse_star()        – parse star rating
+    find_existing_record() – search Bitable for matching record
+    extract_record_id() – extract record_id from lark-cli response
+    upsert_record()     – create or update a Bitable record
+"""
+import calendar
+import json
+import os
+import re
+import shutil
+import subprocess
+from datetime import date, datetime, timedelta
+
+from gyxx_flow.modules.shop_intelligence.runtime.config import (
+    LARK_PROFILE,
+)
+
+
+def resolve_lark_cli() -> tuple[str, str]:
+    cli_cmd = shutil.which("lark-cli.cmd") or shutil.which("lark-cli.exe") or shutil.which("lark-cli")
+    node = shutil.which("node.exe") or shutil.which("node")
+    if not cli_cmd or not node:
+        raise RuntimeError("未找到 lark-cli 或 node，请确认二者已安装并在 PATH 中")
+    cli_dir = os.path.dirname(cli_cmd)
+    cli_script = os.path.join(cli_dir, "node_modules", "@larksuite", "cli", "scripts", "run.js")
+    if not os.path.exists(cli_script):
+        raise RuntimeError(f"未找到 lark-cli Node 入口: {cli_script}")
+    return node, cli_script
+
+
+def run_lark_cli(
+    node: str,
+    cli_script: str,
+    args: list[str],
+    profile: str = LARK_PROFILE,
+    as_user: bool = True,
+) -> dict:
+    identity = "user" if as_user else "bot"
+    cmd = [node, cli_script, "--profile", profile, "--as", identity] + args
+    result = subprocess.run(
+        cmd,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+    )
+    if result.returncode != 0:
+        raise RuntimeError(f"lark-cli 失败: {(result.stderr or result.stdout).strip()}")
+
+    out = (result.stdout or "").strip()
+    if not out:
+        return {}
+    try:
+        return json.loads(out)
+    except json.JSONDecodeError:
+        raise RuntimeError(f"lark-cli 返回非 JSON: {out[:200]}")
+
+
+def _days_in_month(year: int, month: int) -> int:
+    return calendar.monthrange(year, month)[1]
+
+
+def _week_belongs_to(monday: date, year: int, month: int) -> bool:
+    in_month = sum(1 for d in range(7) if (monday + timedelta(days=d)).month == month)
+    return in_month > 7 - in_month
+
+
+def format_week(week_start: str, week_end: str) -> tuple[str, str, date, date]:
+    start = datetime.strptime(week_start, "%Y-%m-%d").date()
+    end = datetime.strptime(week_end, "%Y-%m-%d").date()
+
+    days_in_start_month = _days_in_month(start.year, start.month)
+    if start.day + 6 <= days_in_start_month:
+        week_year, week_month = start.year, start.month
+    else:
+        days_in_start = days_in_start_month - start.day + 1
+        days_in_end = 7 - days_in_start
+        if days_in_start > days_in_end:
+            week_year, week_month = start.year, start.month
+        else:
+            week_year, week_month = end.year, end.month
+
+    first_day_of_month = date(week_year, week_month, 1)
+    candidate_monday = first_day_of_month - timedelta(days=first_day_of_month.weekday())
+    if not _week_belongs_to(candidate_monday, week_year, week_month):
+        candidate_monday = candidate_monday + timedelta(days=7)
+
+    n = ((start - candidate_monday).days // 7) + 1
+    if n < 1:
+        n = 1
+
+    week_str = f"{week_month}月第{n}周（{start.month}.{start.day:02d}-{end.month}.{end.day:02d}）"
+    if week_year >= 2026:
+        month_str = f"{week_year}.{week_month:02d}月"
+    else:
+        month_str = f"{week_year}.{week_month}月"
+    return week_str, month_str, start, end
+
+
+def parse_amount(value) -> float:
+    if not value or value == "N/A":
+        return 0.0
+    cleaned = re.sub(r"[^\d.]", "", str(value))
+    if not cleaned:
+        return 0.0
+    try:
+        return float(cleaned)
+    except ValueError:
+        return 0.0
+
+
+def parse_star(star_raw) -> float:
+    if not star_raw or star_raw == "N/A":
+        return 0.0
+    m = re.search(r"\d+(?:\.\d+)?", str(star_raw))
+    if not m:
+        return 0.0
+    try:
+        return float(m.group(0))
+    except ValueError:
+        return 0.0
+
+
+def find_existing_record(
+    node: str,
+    cli_script: str,
+    base_token: str,
+    table_id: str,
+    fields_config: dict,
+    week_str: str,
+    platform: str,
+    parent_record_id: str = "",
+) -> str:
+    payload = {
+        "keyword": week_str,
+        "search_fields": [fields_config["时间"]],
+        "select_fields": [
+            fields_config["时间"],
+            fields_config.get("父记录", ""),
+            fields_config["平台"],
+        ],
+        "limit": 50,
+    }
+    res = run_lark_cli(node, cli_script, [
+        "base",
+        "+record-search",
+        "--base-token",
+        base_token,
+        "--table-id",
+        table_id,
+        "--format",
+        "json",
+        "--json",
+        json.dumps(payload, ensure_ascii=False),
+    ], as_user=False)
+
+    payload_data = res.get("data", {}) if isinstance(res.get("data"), dict) else {}
+    rows = payload_data.get("data") or []
+    record_ids = payload_data.get("record_id_list") or []
+    if not isinstance(rows, list) or not rows:
+        return ""
+
+    time_field_id = fields_config["时间"]
+    parent_field_id = fields_config.get("父记录", "")
+    platform_field_id = fields_config["平台"]
+    fields_meta = payload_data.get("fields") or []
+
+    def _index(field_id: str, field_name: str) -> int:
+        if field_id in fields_meta:
+            return fields_meta.index(field_id)
+        if field_name in fields_meta:
+            return fields_meta.index(field_name)
+        return -1
+
+    time_idx = _index(time_field_id, "时间")
+    parent_idx = _index(parent_field_id, "父记录") if parent_field_id else -1
+    platform_idx = _index(platform_field_id, "平台")
+    check_parent = bool(parent_record_id)
+
+    for i, row in enumerate(rows):
+        record_id = record_ids[i] if i < len(record_ids) else ""
+        if not isinstance(row, list) or not row:
+            continue
+        time_value = row[time_idx] if time_idx >= 0 and time_idx < len(row) else ""
+        if isinstance(time_value, list) and time_value:
+            time_value = time_value[0] if isinstance(time_value[0], str) else str(time_value[0])
+        if str(time_value).strip() != week_str:
+            continue
+        platform_value = row[platform_idx] if platform_idx >= 0 and platform_idx < len(row) else ""
+        if isinstance(platform_value, list) and platform_value:
+            platform_value = platform_value[0] if isinstance(platform_value[0], str) else str(platform_value[0])
+        if str(platform_value).strip() != platform:
+            continue
+        if check_parent and parent_idx >= 0:
+            parent_value = row[parent_idx] if parent_idx < len(row) else None
+            parent_ids = []
+            if isinstance(parent_value, list):
+                for p in parent_value:
+                    if isinstance(p, dict) and p.get("id"):
+                        parent_ids.append(p["id"])
+                    elif isinstance(p, str):
+                        parent_ids.append(p)
+            if parent_record_id not in parent_ids:
+                continue
+        return record_id
+    return ""
+
+
+def extract_record_id(result: dict, fallback: str) -> str:
+    if not isinstance(result, dict):
+        return fallback
+    data = result.get("data")
+    if isinstance(data, dict):
+        rid_list = data.get("record_id_list")
+        if isinstance(rid_list, list) and rid_list:
+            first = rid_list[0]
+            if isinstance(first, str) and first:
+                return first
+        rec = data.get("record")
+        if isinstance(rec, dict):
+            rid = rec.get("record_id") or rec.get("recordId")
+            if rid:
+                return rid
+        recs = data.get("records")
+        if isinstance(recs, list) and recs:
+            first = recs[0]
+            if isinstance(first, dict):
+                rid = first.get("record_id") or first.get("recordId")
+                if rid:
+                    return rid
+    if isinstance(data, list) and data:
+        first = data[0]
+        if isinstance(first, dict):
+            rid = first.get("record_id") or first.get("recordId")
+            if rid:
+                return rid
+    rec = result.get("record")
+    if isinstance(rec, dict):
+        rid = rec.get("record_id") or rec.get("recordId")
+        if rid:
+            return rid
+    return fallback
+
+
+def upsert_record(
+    node: str,
+    cli_script: str,
+    base_token: str,
+    table_id: str,
+    record_id: str,
+    cell_values: dict,
+) -> dict:
+    if record_id:
+        return run_lark_cli(node, cli_script, [
+            "base",
+            "+record-batch-update",
+            "--base-token",
+            base_token,
+            "--table-id",
+            table_id,
+            "--json",
+            json.dumps(
+                {"record_id_list": [record_id], "patch": cell_values},
+                ensure_ascii=False,
+            ),
+        ], as_user=False)
+    return run_lark_cli(node, cli_script, [
+        "base",
+        "+record-batch-create",
+        "--base-token",
+        base_token,
+        "--table-id",
+        table_id,
+        "--json",
+        json.dumps(
+            {
+                "fields": list(cell_values.keys()),
+                "rows": [list(cell_values.values())],
+            },
+            ensure_ascii=False,
+        ),
+    ], as_user=False)
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/paths.py b/src/gyxx_flow/modules/shop_intelligence/runtime/paths.py
new file mode 100644
index 0000000..f73b8ca
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/paths.py
@@ -0,0 +1,52 @@
+"""Portable filesystem layout for the migrated shop runtime."""
+
+from __future__ import annotations
+
+import os
+from pathlib import Path
+
+from gyxx_flow.core import DataLayout, Settings
+
+PACKAGE_ROOT = Path(__file__).resolve().parent
+PROJECT_ROOT = PACKAGE_ROOT.parents[4]
+_configured_data_root = os.environ.get("GYXX_DATA_ROOT", "").strip()
+DATA_ROOT = (
+    Path(_configured_data_root).expanduser().resolve()
+    if _configured_data_root
+    else Settings.from_env().data_root
+)
+_DATA_PATHS = DataLayout(DATA_ROOT).for_module("shop_intelligence")
+RAW_ROOT = _DATA_PATHS.raw_root
+NORMALIZED_ROOT = _DATA_PATHS.normalized_root
+CURATED_ROOT = _DATA_PATHS.curated_root
+EXPORTS_ROOT = _DATA_PATHS.exports_root
+EVIDENCE_ROOT = _DATA_PATHS.evidence_root
+STATE_ROOT = _DATA_PATHS.state_root
+LOG_ROOT = _DATA_PATHS.logs_root
+TMP_ROOT = _DATA_PATHS.tmp_root
+
+
+def raw_path(*parts: str) -> Path:
+    return _DATA_PATHS.raw_path(*parts)
+
+
+def state_path(*parts: str) -> Path:
+    return _DATA_PATHS.state_path(*parts)
+
+
+def resolve_raw_output(
+    value: str | os.PathLike[str] | None,
+    *,
+    default: str | os.PathLike[str],
+) -> Path:
+    """Resolve a collector output directory without allowing RAW_ROOT escape."""
+
+    raw_root = RAW_ROOT.resolve()
+    requested = Path(default) if value is None or not str(value).strip() else Path(value)
+    candidate = requested.expanduser()
+    if not candidate.is_absolute():
+        candidate = raw_root / candidate
+    resolved = candidate.resolve()
+    if resolved != raw_root and not resolved.is_relative_to(raw_root):
+        raise ValueError(f"collector output must remain inside RAW_ROOT: {raw_root}")
+    return resolved
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/runners/__init__.py b/src/gyxx_flow/modules/shop_intelligence/runtime/runners/__init__.py
new file mode 100644
index 0000000..4851cfb
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/runners/__init__.py
@@ -0,0 +1,6 @@
+"""Parallel runners for weekly shop and peer-store collection."""
+
+from .run_peer_store import main as run_peer_store
+from .run_shop import main as run_shop
+
+__all__ = ["run_peer_store", "run_shop"]
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/runners/run_peer_store.py b/src/gyxx_flow/modules/shop_intelligence/runtime/runners/run_peer_store.py
new file mode 100644
index 0000000..19517f6
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/runners/run_peer_store.py
@@ -0,0 +1,200 @@
+"""Run peer-store data collection for JD / DY in parallel, then persist and write Base.
+
+天猫暂无竞店采集脚本， runner 中占位跳过。
+
+Usage:
+    python -m gyxx_flow.modules.shop_intelligence.runtime.runners.run_peer_store
+"""
+from __future__ import annotations
+
+import sys
+from concurrent.futures import ThreadPoolExecutor, as_completed
+
+from gyxx_flow.modules.shop_intelligence.runtime.config import (
+    DY_PEER_STORE_OUTPUT,
+    DY_PLATFORM,
+    DY_SCRIPT,
+    DY_SHOP_OUTPUT,
+    JD_PASSWORD,
+    JD_PEER_STORE_OUTPUT,
+    JD_PEER_STORE_SCRIPT,
+    JD_PLATFORM,
+    JD_SHOP,
+    TM_PLATFORM,
+)
+from gyxx_flow.modules.shop_intelligence.runtime.db import (
+    persist_dy_peer_store,
+    persist_jd_peer_store,
+)
+from gyxx_flow.modules.shop_intelligence.runtime.runners.utils import (
+    find_latest_file,
+    now_str,
+    run_collector,
+    setup_runner_logging,
+)
+from gyxx_flow.modules.shop_intelligence.runtime.writers import write_peer_store
+
+
+def _run_jd_peer() -> tuple[str, int, str]:
+    code, output = run_collector(
+        "京东竞店采集",
+        JD_PEER_STORE_SCRIPT,
+        [
+            "--shop", JD_SHOP,
+            "--password", JD_PASSWORD,
+            "--output", JD_PEER_STORE_OUTPUT,
+        ],
+    )
+    return JD_PLATFORM, code, output
+
+
+def _run_dy_peer() -> tuple[str, int, str]:
+    # 抖音脚本同时包含店铺+竞店；SKIP_SHOP=1 跳过店铺部分（避免覆盖已有店铺 JSON），
+    # 仅采集竞店榜单，店铺字段从今日已有 JSON 加载（用于本店行写入）
+    code, output = run_collector(
+        "抖音竞店采集",
+        DY_SCRIPT,
+        [],
+        env={"KEEP_BROWSER_OPEN": "0", "SKIP_SHOP": "1"},
+    )
+    return DY_PLATFORM, code, output
+
+
+def _run_tm_peer() -> tuple[str, int, str]:
+    print(f"[Runner] {TM_PLATFORM} 暂无竞店采集脚本，跳过")
+    return TM_PLATFORM, 0, "天猫暂无竞店采集脚本"
+
+
+def _latest_peer_json(platform: str) -> str:
+    if platform == JD_PLATFORM:
+        return find_latest_file(JD_PEER_STORE_OUTPUT, "竞店数据_*.json")
+    if platform == DY_PLATFORM:
+        return find_latest_file(DY_PEER_STORE_OUTPUT, "竞店数据汇总_*.json")
+    return ""
+
+
+def _latest_shop_json_for_dy() -> str:
+    """DY 竞店表需要写入本店行，依赖最新店铺 JSON 的 market_rankings。"""
+    return find_latest_file(DY_SHOP_OUTPUT, "光影行星店铺信息汇总_*.json")
+
+
+def _persist_and_write(platform: str, success: bool, skip_feishu: bool = False) -> list[str]:
+    errors: list[str] = []
+    if not success:
+        errors.append(f"{platform}: 采集失败，跳过后续写入")
+        return errors
+
+    peer_json = _latest_peer_json(platform)
+    print(f"[Runner] {platform} 最新竞店 JSON: {peer_json or '未找到'}")
+
+    if platform == JD_PLATFORM:
+        try:
+            n = persist_jd_peer_store(peer_json)
+            print(f"[Runner] JD 竞店入库: {n} 条")
+        except Exception as e:
+            err = f"JD DB 持久化失败: {e}"
+            print(f"[Runner] {err}")
+            errors.append(err)
+        if not skip_feishu:
+            try:
+                write_peer_store(JD_PLATFORM, peer_json)
+            except Exception as e:
+                err = f"JD 竞店飞书写入失败: {e}"
+                print(f"[Runner] {err}")
+                errors.append(err)
+        else:
+            print("[Runner] JD 竞店飞书写入已跳过")
+
+    elif platform == DY_PLATFORM:
+        try:
+            n = persist_dy_peer_store(peer_json)
+            print(f"[Runner] DY 竞店入库: {n} 条")
+        except Exception as e:
+            err = f"DY DB 持久化失败: {e}"
+            print(f"[Runner] {err}")
+            errors.append(err)
+        if not skip_feishu:
+            try:
+                shop_json = _latest_shop_json_for_dy()
+                write_peer_store(DY_PLATFORM, peer_json, shop_json)
+            except Exception as e:
+                err = f"DY 竞店飞书写入失败: {e}"
+                print(f"[Runner] {err}")
+                errors.append(err)
+        else:
+            print("[Runner] DY 竞店飞书写入已跳过")
+
+    elif platform == TM_PLATFORM:
+        print("[Runner] TM 竞店跳过")
+
+    return errors
+
+
+def main() -> int:
+    import argparse
+    parser = argparse.ArgumentParser(description="三平台竞店数据采集")
+    parser.add_argument("--skip-feishu", action="store_true", help="跳过飞书多维表格写入")
+    args = parser.parse_args()
+
+    log_path = setup_runner_logging("run_peer_store")
+
+    print("=" * 60)
+    print(f"  竞店数据采集 - 并行 - {now_str()}")
+    print(f"  日志: {log_path}")
+    if args.skip_feishu:
+        print("  [配置] 跳过飞书表格写入")
+    print("=" * 60)
+
+    tasks = {
+        "jd": _run_jd_peer,
+        "dy": _run_dy_peer,
+        "tm": _run_tm_peer,
+    }
+
+    results: dict[str, tuple[int, str]] = {}
+    with ThreadPoolExecutor(max_workers=3) as executor:
+        futures = {executor.submit(fn): key for key, fn in tasks.items()}
+        for future in as_completed(futures):
+            key = futures[future]
+            try:
+                platform, code, output = future.result()
+                results[platform] = (code, output)
+            except Exception as e:
+                print(f"[Runner] {key} 任务异常: {e}")
+
+    print("\n" + "=" * 60)
+    print("  采集阶段汇总")
+    print("=" * 60)
+    all_errors: list[str] = []
+    for platform in (JD_PLATFORM, DY_PLATFORM, TM_PLATFORM):
+        code, output = results.get(platform, (1, "未执行"))
+        success = code == 0
+        status = "成功" if success else "失败"
+        print(f"  {platform}: {status} (exit={code})")
+        if not success:
+            all_errors.append(f"{platform}: {output[:200]}")
+
+    print("\n" + "=" * 60)
+    print("  持久化 / 飞书写入")
+    print("=" * 60)
+    for platform in (JD_PLATFORM, DY_PLATFORM, TM_PLATFORM):
+        code, _ = results.get(platform, (1, ""))
+        errs = _persist_and_write(platform, code == 0, skip_feishu=args.skip_feishu)
+        all_errors.extend(errs)
+
+    print("\n" + "=" * 60)
+    if all_errors:
+        print("  完成，但有错误:")
+        for e in all_errors:
+            print(f"    - {e}")
+        print(f"  日志: {log_path}")
+        print("=" * 60)
+        return 1
+    print("  全部完成")
+    print(f"  日志: {log_path}")
+    print("=" * 60)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/runners/run_shop.py b/src/gyxx_flow/modules/shop_intelligence/runtime/runners/run_shop.py
new file mode 100644
index 0000000..dc29152
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/runners/run_shop.py
@@ -0,0 +1,221 @@
+"""Run shop data collection for JD / DY / TM in parallel, then persist and write Base.
+
+Usage:
+    python -m gyxx_flow.modules.shop_intelligence.runtime.runners.run_shop
+"""
+from __future__ import annotations
+
+import sys
+from concurrent.futures import ThreadPoolExecutor, as_completed
+
+from gyxx_flow.modules.shop_intelligence.runtime.config import (
+    DY_PLATFORM,
+    DY_SCRIPT,
+    DY_SHOP_OUTPUT,
+    JD_INDUSTRY,
+    JD_PASSWORD,
+    JD_PLATFORM,
+    JD_SHOP,
+    JD_SHOP_OUTPUT,
+    JD_SHOP_SCRIPT,
+    TM_PLATFORM,
+    TM_SHOP_OUTPUT,
+    TM_SHOP_SCRIPT,
+)
+from gyxx_flow.modules.shop_intelligence.runtime.db import (
+    persist_dy_shop,
+    persist_jd_shop,
+)
+from gyxx_flow.modules.shop_intelligence.runtime.runners.utils import (
+    find_latest_file,
+    now_str,
+    run_collector,
+    setup_runner_logging,
+)
+from gyxx_flow.modules.shop_intelligence.runtime.writers import (
+    ensure_ai_analysis,
+    write_shop,
+)
+
+
+def _run_jd_shop() -> tuple[str, int, str]:
+    code, output = run_collector(
+        "京东店铺采集",
+        JD_SHOP_SCRIPT,
+        [
+            "--shop", JD_SHOP,
+            "--password", JD_PASSWORD,
+            "--industry", JD_INDUSTRY,
+            "--output", JD_SHOP_OUTPUT,
+        ],
+    )
+    return JD_PLATFORM, code, output
+
+
+def _run_dy_shop() -> tuple[str, int, str]:
+    code, output = run_collector(
+        "抖音店铺采集",
+        DY_SCRIPT,
+        [],
+        env={"SKIP_COMPETITOR": "1", "KEEP_BROWSER_OPEN": "0"},
+    )
+    return DY_PLATFORM, code, output
+
+
+def _run_tm_shop() -> tuple[str, int, str]:
+    # 天猫脚本自身完成 DB 持久化，runner 只需调用即可
+    code, output = run_collector(
+        "天猫店铺采集",
+        TM_SHOP_SCRIPT,
+        [],
+    )
+    return TM_PLATFORM, code, output
+
+
+def _latest_metrics_path(platform: str) -> str:
+    if platform == JD_PLATFORM:
+        return find_latest_file(JD_SHOP_OUTPUT, "*_指标数据_*.json")
+    if platform == DY_PLATFORM:
+        return find_latest_file(DY_SHOP_OUTPUT, "光影行星店铺信息汇总_*.json")
+    if platform == TM_PLATFORM:
+        return find_latest_file(TM_SHOP_OUTPUT, "sycm_*.json")
+    return ""
+
+
+def _persist_and_write(platform: str, success: bool, skip_feishu: bool = False) -> list[str]:
+    errors: list[str] = []
+    if not success:
+        errors.append(f"{platform}: 采集失败，跳过后续写入")
+        return errors
+
+    metrics_path = _latest_metrics_path(platform)
+    print(f"[Runner] {platform} 最新指标 JSON: {metrics_path or '未找到'}")
+
+    # 先生成 AI 分析并写回 JSON（DB 持久化和飞书写入都从 JSON 读取，避免重复调用 Hermes）
+    if metrics_path:
+        try:
+            ensure_ai_analysis(platform, metrics_path)
+        except Exception as e:
+            err = f"{platform} AI 分析生成失败: {e}"
+            print(f"[Runner] {err}")
+            errors.append(err)
+
+    if platform == JD_PLATFORM:
+        try:
+            n = persist_jd_shop(metrics_path)
+            print(f"[Runner] JD 店铺入库: {n} 条")
+        except Exception as e:
+            err = f"JD DB 持久化失败: {e}"
+            print(f"[Runner] {err}")
+            errors.append(err)
+        if not skip_feishu:
+            try:
+                write_shop(JD_PLATFORM, metrics_path)
+            except Exception as e:
+                err = f"JD 飞书写入失败: {e}"
+                print(f"[Runner] {err}")
+                errors.append(err)
+        else:
+            print("[Runner] JD 飞书写入已跳过")
+
+    elif platform == DY_PLATFORM:
+        try:
+            n = persist_dy_shop(metrics_path)
+            print(f"[Runner] DY 店铺入库: {n} 条")
+        except Exception as e:
+            err = f"DY DB 持久化失败: {e}"
+            print(f"[Runner] {err}")
+            errors.append(err)
+        if not skip_feishu:
+            try:
+                write_shop(DY_PLATFORM, metrics_path)
+            except Exception as e:
+                err = f"DY 飞书写入失败: {e}"
+                print(f"[Runner] {err}")
+                errors.append(err)
+        else:
+            print("[Runner] DY 飞书写入已跳过")
+
+    elif platform == TM_PLATFORM:
+        # 天猫已自行完成 DB 写入；飞书店铺表也写入
+        if not skip_feishu:
+            try:
+                write_shop(TM_PLATFORM, metrics_path)
+            except Exception as e:
+                err = f"TM 飞书写入失败: {e}"
+                print(f"[Runner] {err}")
+                errors.append(err)
+        else:
+            print("[Runner] TM 飞书写入已跳过")
+
+    return errors
+
+
+def main() -> int:
+    import argparse
+    parser = argparse.ArgumentParser(description="三平台店铺数据采集")
+    parser.add_argument("--skip-feishu", action="store_true", help="跳过飞书多维表格写入")
+    args = parser.parse_args()
+
+    log_path = setup_runner_logging("run_shop")
+
+    print("=" * 60)
+    print(f"  店铺数据采集 - 并行三平台 - {now_str()}")
+    print(f"  日志: {log_path}")
+    if args.skip_feishu:
+        print("  [配置] 跳过飞书表格写入")
+    print("=" * 60)
+
+    tasks = {
+        "jd": _run_jd_shop,
+        "dy": _run_dy_shop,
+        "tm": _run_tm_shop,
+    }
+
+    results: dict[str, tuple[int, str]] = {}
+    with ThreadPoolExecutor(max_workers=3) as executor:
+        futures = {executor.submit(fn): key for key, fn in tasks.items()}
+        for future in as_completed(futures):
+            key = futures[future]
+            try:
+                platform, code, output = future.result()
+                results[platform] = (code, output)
+            except Exception as e:
+                print(f"[Runner] {key} 任务异常: {e}")
+
+    print("\n" + "=" * 60)
+    print("  采集阶段汇总")
+    print("=" * 60)
+    all_errors: list[str] = []
+    for platform in (JD_PLATFORM, DY_PLATFORM, TM_PLATFORM):
+        code, output = results.get(platform, (1, "未执行"))
+        success = code == 0
+        status = "成功" if success else "失败"
+        print(f"  {platform}: {status} (exit={code})")
+        if not success:
+            all_errors.append(f"{platform}: {output[:200]}")
+
+    print("\n" + "=" * 60)
+    print("  持久化 / 飞书写入")
+    print("=" * 60)
+    for platform in (JD_PLATFORM, DY_PLATFORM, TM_PLATFORM):
+        code, _ = results.get(platform, (1, ""))
+        errs = _persist_and_write(platform, code == 0, skip_feishu=args.skip_feishu)
+        all_errors.extend(errs)
+
+    print("\n" + "=" * 60)
+    if all_errors:
+        print("  完成，但有错误:")
+        for e in all_errors:
+            print(f"    - {e}")
+        print(f"  日志: {log_path}")
+        print("=" * 60)
+        return 1
+    print("  全部完成")
+    print(f"  日志: {log_path}")
+    print("=" * 60)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/runners/utils.py b/src/gyxx_flow/modules/shop_intelligence/runtime/runners/utils.py
new file mode 100644
index 0000000..23b368d
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/runners/utils.py
@@ -0,0 +1,123 @@
+"""Shared utilities for collection runners."""
+from __future__ import annotations
+
+import glob
+import os
+import subprocess
+import sys
+import threading
+from datetime import datetime
+
+from gyxx_flow.modules.shop_intelligence.runtime.paths import LOG_ROOT, PACKAGE_ROOT
+
+PROJECT_ROOT = str(PACKAGE_ROOT)
+LOGS_DIR = str(LOG_ROOT)
+
+
+class _DualStream:
+    """写入时同时输出到多个 stream（stdout + 日志文件），线程安全。"""
+
+    def __init__(self, *streams) -> None:
+        self._streams = streams
+        self._lock = threading.Lock()
+
+    def write(self, data) -> int:
+        with self._lock:
+            for s in self._streams:
+                try:
+                    s.write(data)
+                    s.flush()
+                except Exception:
+                    pass
+        return len(data) if isinstance(data, str) else 0
+
+    def flush(self) -> None:
+        for s in self._streams:
+            try:
+                s.flush()
+            except Exception:
+                pass
+
+    def isatty(self) -> bool:
+        return False
+
+
+def setup_runner_logging(name: str) -> str:
+    """把 stdout/stderr 同时重定向到终端和 logs/<name>_<ts>.log，返回日志路径。
+
+    定时任务下没有控制台可见输出，落到文件后便于事后排查 exit code != 0 的原因。
+    """
+    os.makedirs(LOGS_DIR, exist_ok=True)
+    ts = datetime.now().strftime("%Y%m%d_%H%M%S")
+    log_path = os.path.join(LOGS_DIR, f"{name}_{ts}.log")
+    log_fh = open(log_path, "w", encoding="utf-8", buffering=1)
+    sys.stdout = _DualStream(sys.__stdout__, log_fh)
+    sys.stderr = _DualStream(sys.__stderr__, log_fh)
+    return log_path
+
+
+def find_latest_file(directory: str, pattern: str) -> str:
+    """Return the most recently modified file matching pattern under directory."""
+    if not directory or not os.path.isdir(directory):
+        return ""
+    matches = glob.glob(os.path.join(directory, pattern), recursive=True)
+    if not matches:
+        return ""
+    # 按修改时间倒序，时间相同按文件名倒序（通常日期越新文件名越后）
+    latest = sorted(matches, key=lambda p: (os.path.getmtime(p), p), reverse=True)[0]
+    return latest
+
+
+def run_collector(
+    label: str,
+    script: str,
+    args: list[str],
+    env: dict | None = None,
+    timeout: int = 3600,
+) -> tuple[int, str]:
+    """Run a collector script and return (exit_code, combined_output)."""
+    cmd = [sys.executable, "-X", "utf8", "-u", script] + args
+    environment = os.environ.copy()
+    environment["PYTHONIOENCODING"] = "utf-8"
+    environment["PYTHONUTF8"] = "1"
+    # 定时/后台运行时不要让浏览器一直挂着
+    environment.setdefault("KEEP_BROWSER_OPEN", "0")
+    # 让子进程能找到项目根目录的模块（adaptive_selectors / db 等）
+    existing_pythonpath = environment.get("PYTHONPATH", "")
+    environment["PYTHONPATH"] = (
+        PROJECT_ROOT + os.pathsep + existing_pythonpath
+        if existing_pythonpath
+        else PROJECT_ROOT
+    )
+    if env:
+        environment.update(env)
+
+    print(f"\n[Runner] -- {label} --")
+    print(f"[Runner] 命令: {' '.join(cmd)}")
+
+    try:
+        result = subprocess.run(
+            cmd,
+            capture_output=True,
+            env=environment,
+            timeout=timeout,
+        )
+    except subprocess.TimeoutExpired:
+        return 1, f"{label} 超时（{timeout} 秒）"
+    except Exception as e:
+        return 1, f"{label} 异常: {e}"
+
+    stdout = result.stdout.decode("utf-8", errors="replace")
+    stderr = result.stderr.decode("utf-8", errors="replace")
+    output = f"{stdout}\n{stderr}".strip()
+    try:
+        print(output)
+    except UnicodeEncodeError:
+        enc = sys.stdout.encoding or "utf-8"
+        safe = output.encode(enc, errors="replace").decode(enc, errors="replace")
+        print(safe)
+    return result.returncode, output
+
+
+def now_str() -> str:
+    return datetime.now().strftime("%Y-%m-%d %H:%M:%S")
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/scripts/remove_scheduler.ps1 b/src/gyxx_flow/modules/shop_intelligence/runtime/scripts/remove_scheduler.ps1
new file mode 100644
index 0000000..04038aa
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/scripts/remove_scheduler.ps1
@@ -0,0 +1,9 @@
+#Requires -RunAsAdministrator
+$ErrorActionPreference = "Stop"
+
+foreach ($taskName in @("shop.metrics.weekly", "shop.competitor.weekly")) {
+    $task = Get-ScheduledTask -TaskPath "\GYXX\" -TaskName $taskName -ErrorAction SilentlyContinue
+    if ($task) {
+        Unregister-ScheduledTask -TaskPath "\GYXX\" -TaskName $taskName -Confirm:$false
+    }
+}
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/scripts/setup_scheduler.ps1 b/src/gyxx_flow/modules/shop_intelligence/runtime/scripts/setup_scheduler.ps1
new file mode 100644
index 0000000..01f42e5
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/scripts/setup_scheduler.ps1
@@ -0,0 +1,26 @@
+#Requires -RunAsAdministrator
+param(
+    [string]$OutputDirectory
+)
+
+$ErrorActionPreference = "Stop"
+$projectRoot = if ($env:GYXX_PROJECT_ROOT) {
+    $env:GYXX_PROJECT_ROOT
+} else {
+    (Resolve-Path (Join-Path $PSScriptRoot "..\..\..\..\..\..\..")).Path
+}
+$python = Join-Path $projectRoot ".venv\Scripts\python.exe"
+if (-not (Test-Path -LiteralPath $python)) {
+    throw "GYXX Python interpreter is missing: $python"
+}
+if (-not $OutputDirectory) {
+    $dataRoot = if ($env:GYXX_DATA_ROOT) { $env:GYXX_DATA_ROOT } else { Join-Path $projectRoot "data" }
+    $OutputDirectory = Join-Path $dataRoot "evidence\schedule-plans\shop-intelligence"
+}
+
+& $python -m gyxx_flow.cli schedule plan --python-executable $python --output $OutputDirectory
+if ($LASTEXITCODE -ne 0) {
+    exit $LASTEXITCODE
+}
+Write-Host "Schedule bundle generated. Review it, then run:"
+Write-Host "  & '$OutputDirectory\install.ps1'"
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/sitecustomize.py b/src/gyxx_flow/modules/shop_intelligence/runtime/sitecustomize.py
new file mode 100644
index 0000000..bf69a40
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/sitecustomize.py
@@ -0,0 +1,5 @@
+"""Apply the unified per-script runtime binding to nested Python launches."""
+
+from gyxx_flow.adapters.bootstrap import safe_bootstrap_current_process
+
+safe_bootstrap_current_process()
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/writers/__init__.py b/src/gyxx_flow/modules/shop_intelligence/runtime/writers/__init__.py
new file mode 100644
index 0000000..a9fc196
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/writers/__init__.py
@@ -0,0 +1,6 @@
+"""Feishu Bitable writers for shop and peer-store data."""
+
+from .peer_store_writer import write_peer_store
+from .shop_base_writer import ensure_ai_analysis, write_shop
+
+__all__ = ["write_peer_store", "write_shop", "ensure_ai_analysis"]
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/writers/peer_store_writer.py b/src/gyxx_flow/modules/shop_intelligence/runtime/writers/peer_store_writer.py
new file mode 100644
index 0000000..c2ce53c
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/writers/peer_store_writer.py
@@ -0,0 +1,418 @@
+"""Write JD/DY peer-store rankings to Feishu Bitable peer-store table.
+
+天猫跳过（不写飞书）。
+"""
+from __future__ import annotations
+
+import json
+import os
+import re
+import sys
+
+from gyxx_flow.modules.shop_intelligence.runtime.config import (
+    BASE_APP_TOKEN,
+    DY_PEER_BASE_FIELDS,
+    DY_PEER_BASE_TABLE_ID,
+    DY_PEER_CAT_MAP,
+    DY_PLATFORM,
+    JD_PEER_BASE_FIELDS,
+    JD_PEER_BASE_TABLE_ID,
+    JD_PEER_BRAND_MAP,
+    JD_PEER_CATEGORY_MAP,
+    JD_PLATFORM,
+    LOCAL_SHOP_BASE_OPTION,
+)
+from gyxx_flow.modules.shop_intelligence.runtime.lark_cli import (
+    extract_record_id,
+    format_week,
+    resolve_lark_cli,
+    run_lark_cli,
+)
+
+
+def _safe_print(text: str) -> None:
+    try:
+        print(text)
+    except UnicodeEncodeError:
+        enc = sys.stdout.encoding or "utf-8"
+        safe = text.encode(enc, errors="replace").decode(enc, errors="replace")
+        print(safe)
+
+
+def _parse_period_period(period: str) -> tuple[str, str]:
+    """支持 '2026-06-23 ~ 2026-06-29' 或 '2026/06/23 - 2026/06/29'。"""
+    if not period:
+        return "", ""
+    m = re.search(
+        r"(\d{4})[-/](\d{2})[-/](\d{2})\s*[-—~至到]\s*(\d{4})[-/](\d{2})[-/](\d{2})",
+        str(period),
+    )
+    if not m:
+        return "", ""
+    return (
+        f"{m.group(1)}-{m.group(2)}-{m.group(3)}",
+        f"{m.group(4)}-{m.group(5)}-{m.group(6)}",
+    )
+
+
+def _month_option_str(week_start: str) -> str:
+    m = re.match(r"(\d{4})-(\d{2})-(\d{2})", week_start)
+    if not m:
+        return ""
+    year, month = int(m.group(1)), int(m.group(2))
+    return f"{year}.{month:02d}月" if year >= 2026 else f"{year}.{month}月"
+
+
+def _build_month_options_cache(node: str, cli_script: str, table_id: str, month_field_id: str) -> dict[str, str]:
+    res = run_lark_cli(node, cli_script, [
+        "base",
+        "+field-search-options",
+        "--base-token",
+        BASE_APP_TOKEN,
+        "--table-id",
+        table_id,
+        "--field-id",
+        month_field_id,
+        "--limit",
+        "100",
+    ], as_user=False)
+    options = (res.get("data") or {}).get("options") or []
+    mapping: dict[str, str] = {}
+    for opt in options:
+        name = opt.get("name", "").strip()
+        if name:
+            mapping[name] = name
+    return mapping
+
+
+def _resolve_month_option(week_start: str, options_cache: dict[str, str]) -> str:
+    default = _month_option_str(week_start)
+    if default in options_cache:
+        return default
+    m = re.match(r"(\d{4})-(\d{2})-(\d{2})", week_start)
+    if not m:
+        return default
+    year, month = int(m.group(1)), int(m.group(2))
+    candidates = [
+        f"{year}.{month:02d}月",
+        f"{year}.{month}月",
+    ]
+    for cand in candidates:
+        if cand in options_cache:
+            return cand
+    return default
+
+
+def _parse_rank(value) -> int | None:
+    if value is None:
+        return None
+    s = str(value).strip()
+    if not s or s.upper() == "N/A" or "未进入" in s:
+        return None
+    try:
+        return int(s)
+    except ValueError:
+        return None
+
+
+# ── JD 竞店表写入 ─────────────────────────────────────────────────────────────
+
+
+def _write_jd_peer_store(json_path: str) -> bool:
+    with open(json_path, encoding="utf-8") as f:
+        data = json.load(f)
+
+    period = data.get("数据周期", "")
+    week_start, week_end = _parse_period_period(period)
+    if not week_start or not week_end:
+        raise ValueError(f"京东竞店 JSON 数据周期异常: {period}")
+    week_str, month_str, _, _ = format_week(week_start, week_end)
+    brand_data = data.get("数据", {}) or {}
+
+    node, cli_script = resolve_lark_cli()
+    month_options = _build_month_options_cache(
+        node, cli_script, JD_PEER_BASE_TABLE_ID, JD_PEER_BASE_FIELDS["月份"]
+    )
+    month_opt = _resolve_month_option(week_start, month_options) or month_str
+    for brand_json_name, brand_select_name in JD_PEER_BRAND_MAP.items():
+        brand_entry = brand_data.get(brand_json_name)
+        if not isinstance(brand_entry, dict):
+            continue
+        cell_values = {
+            JD_PEER_BASE_FIELDS["时间"]: week_str,
+            JD_PEER_BASE_FIELDS["月份"]: month_opt,
+            JD_PEER_BASE_FIELDS["竞店名称"]: brand_select_name,
+            JD_PEER_BASE_FIELDS["备注"]: "机器人插入",
+        }
+        for json_cat, field_id in JD_PEER_CATEGORY_MAP.items():
+            cat_entry = brand_entry.get(json_cat) or {}
+            rank_int = _parse_rank(cat_entry.get("rank"))
+            if rank_int is not None:
+                cell_values[field_id] = rank_int
+
+        existing_id = _find_existing_peer_record(
+            node, cli_script, JD_PEER_BASE_TABLE_ID, JD_PEER_BASE_FIELDS, week_str, brand_select_name
+        )
+        result = _upsert_peer_record(
+            node, cli_script, JD_PEER_BASE_TABLE_ID, JD_PEER_BASE_FIELDS, existing_id, cell_values
+        )
+        new_id = extract_record_id(result, existing_id)
+        action = "更新" if existing_id else "新建"
+        _safe_print(
+            f"[PeerWriter] {action} JD {brand_select_name} record_id={new_id} 时间={week_str}"
+        )
+    return True
+
+
+# ── DY 竞店表写入 ─────────────────────────────────────────────────────────────
+
+
+def _build_brand_options_cache(node: str, cli_script: str) -> dict[str, str]:
+    res = run_lark_cli(node, cli_script, [
+        "base",
+        "+field-search-options",
+        "--base-token",
+        BASE_APP_TOKEN,
+        "--table-id",
+        DY_PEER_BASE_TABLE_ID,
+        "--field-id",
+        DY_PEER_BASE_FIELDS["竞店名称"],
+        "--limit",
+        "100",
+    ], as_user=False)
+    options = (res.get("data") or {}).get("options") or []
+    mapping: dict[str, str] = {}
+    for opt in options:
+        name = opt.get("name", "").strip()
+        if name:
+            mapping[name.lower()] = name
+    return mapping
+
+
+def _map_brand(scraper_brand: str, options_cache: dict[str, str]) -> str | None:
+    s = scraper_brand.strip()
+    s_lower = s.lower()
+    if s_lower in options_cache:
+        return options_cache[s_lower]
+    for opt_lower, opt_name in options_cache.items():
+        if s_lower in opt_lower or opt_lower in s_lower:
+            return opt_name
+    return None
+
+
+def _write_dy_peer_store(json_path: str, shop_json_path: str = "") -> bool:
+    with open(json_path, encoding="utf-8") as f:
+        peer_data = json.load(f)
+
+    week_start, week_end = _parse_period_period(peer_data.get("date_range", ""))
+    if not week_start or not week_end:
+        raise ValueError(f"抖音竞店 JSON date_range 异常: {peer_data.get('date_range')}")
+    week_str, month_str, _, _ = format_week(week_start, week_end)
+
+    results = peer_data.get("results") or {}
+    xj_results = results.get(DY_PEER_CAT_MAP["斜挎包"]) or {}
+    sj_results = results.get(DY_PEER_CAT_MAP["双肩包"]) or {}
+    if not xj_results and not sj_results:
+        raise ValueError("抖音竞店 JSON 无 results 数据")
+
+    node, cli_script = resolve_lark_cli()
+    options_cache = _build_brand_options_cache(node, cli_script)
+    if not options_cache:
+        raise RuntimeError("无法获取抖音竞店表品牌 options")
+    month_options = _build_month_options_cache(
+        node, cli_script, DY_PEER_BASE_TABLE_ID, DY_PEER_BASE_FIELDS["月份"]
+    )
+    month_opt = _resolve_month_option(week_start, month_options)
+
+    brands = list(xj_results.keys() or sj_results.keys())
+    for brand in brands:
+        mapped_brand = _map_brand(brand, options_cache)
+        if not mapped_brand:
+            _safe_print(f"[PeerWriter] 跳过品牌 {brand}（Bitable 无此选项）")
+            continue
+        xj_rank = _parse_rank((xj_results.get(brand) or {}).get("rank"))
+        sj_rank = _parse_rank((sj_results.get(brand) or {}).get("rank"))
+        remark = "未进入前200名" if not xj_rank and not sj_rank else "机器人插入"
+        cell_values = {
+            DY_PEER_BASE_FIELDS["时间"]: week_str,
+            DY_PEER_BASE_FIELDS["月份"]: month_opt,
+            DY_PEER_BASE_FIELDS["竞店名称"]: mapped_brand,
+            DY_PEER_BASE_FIELDS["斜挎包"]: int(xj_rank) if xj_rank else None,
+            DY_PEER_BASE_FIELDS["双肩包"]: int(sj_rank) if sj_rank else None,
+            DY_PEER_BASE_FIELDS["备注"]: remark,
+        }
+        existing_id = _find_existing_peer_record(
+            node, cli_script, DY_PEER_BASE_TABLE_ID, DY_PEER_BASE_FIELDS, week_str, mapped_brand
+        )
+        result = _upsert_peer_record(
+            node, cli_script, DY_PEER_BASE_TABLE_ID, DY_PEER_BASE_FIELDS, existing_id, cell_values
+        )
+        new_id = extract_record_id(result, existing_id)
+        action = "更新" if existing_id else "新建"
+        _safe_print(
+            f"[PeerWriter] {action} DY {mapped_brand}({brand}) record_id={new_id} 时间={week_str}"
+        )
+
+    # 写入本店行（依赖店铺 JSON 的 market_rankings）
+    if shop_json_path and os.path.exists(shop_json_path):
+        _write_dy_local_shop_row(
+            node, cli_script, week_str, month_opt, shop_json_path
+        )
+    return True
+
+
+def _write_dy_local_shop_row(node: str, cli_script: str, week_str: str, month_str: str, shop_json_path: str) -> None:
+    with open(shop_json_path, encoding="utf-8") as f:
+        shop_data = json.load(f)
+    market_rankings = shop_data.get("market_rankings") or {}
+    xj_rank = _extract_local_shop_rank(market_rankings, DY_PEER_CAT_MAP["斜挎包"])
+    sj_rank = _extract_local_shop_rank(market_rankings, DY_PEER_CAT_MAP["双肩包"])
+    if xj_rank is None and sj_rank is None:
+        _safe_print("[PeerWriter] 店铺 JSON 无有效 market_rankings，跳过本店行")
+        return
+    cell_values = {
+        DY_PEER_BASE_FIELDS["时间"]: week_str,
+        DY_PEER_BASE_FIELDS["月份"]: month_str,
+        DY_PEER_BASE_FIELDS["竞店名称"]: LOCAL_SHOP_BASE_OPTION,
+        DY_PEER_BASE_FIELDS["斜挎包"]: int(xj_rank) if xj_rank else None,
+        DY_PEER_BASE_FIELDS["双肩包"]: int(sj_rank) if sj_rank else None,
+        DY_PEER_BASE_FIELDS["备注"]: "本店采集",
+    }
+    existing_id = _find_existing_peer_record(
+        node, cli_script, DY_PEER_BASE_TABLE_ID, DY_PEER_BASE_FIELDS, week_str, LOCAL_SHOP_BASE_OPTION
+    )
+    result = _upsert_peer_record(
+        node, cli_script, DY_PEER_BASE_TABLE_ID, DY_PEER_BASE_FIELDS, existing_id, cell_values
+    )
+    new_id = extract_record_id(result, existing_id)
+    action = "更新" if existing_id else "新建"
+    _safe_print(
+        f"[PeerWriter] {action} DY 本店({LOCAL_SHOP_BASE_OPTION}) record_id={new_id} 时间={week_str}"
+    )
+
+
+def _extract_local_shop_rank(market_rankings: dict, category: str) -> str | None:
+    if not isinstance(market_rankings, dict):
+        return None
+    info = market_rankings.get(category)
+    if not isinstance(info, dict):
+        return None
+    rank = (info.get("rank") or "").strip()
+    if not rank or rank in ("未获取", "N/A", "-") or "未进入" in rank:
+        return None
+    return rank
+
+
+# ── 公共 Bitable 操作 ─────────────────────────────────────────────────────────
+
+
+def _find_existing_peer_record(
+    node: str,
+    cli_script: str,
+    table_id: str,
+    fields_config: dict,
+    week_str: str,
+    brand: str,
+) -> str:
+    payload = {
+        "keyword": week_str,
+        "search_fields": [fields_config["时间"]],
+        "select_fields": [fields_config["时间"], fields_config["竞店名称"]],
+        "limit": 100,
+    }
+    res = run_lark_cli(node, cli_script, [
+        "base",
+        "+record-search",
+        "--base-token",
+        BASE_APP_TOKEN,
+        "--table-id",
+        table_id,
+        "--format",
+        "json",
+        "--json",
+        json.dumps(payload, ensure_ascii=False),
+    ], as_user=False)
+    payload_data = res.get("data", {}) if isinstance(res.get("data"), dict) else {}
+    rows = payload_data.get("data") or []
+    record_ids = payload_data.get("record_id_list") or []
+    fields_meta = payload_data.get("fields") or []
+    if not isinstance(rows, list) or not rows:
+        return ""
+
+    def _index(field_id: str, field_name: str) -> int:
+        if field_id in fields_meta:
+            return fields_meta.index(field_id)
+        if field_name in fields_meta:
+            return fields_meta.index(field_name)
+        return -1
+
+    time_idx = _index(fields_config["时间"], "时间")
+    brand_idx = _index(fields_config["竞店名称"], "竞店名称")
+
+    for i, row in enumerate(rows):
+        record_id = record_ids[i] if i < len(record_ids) else ""
+        if not isinstance(row, list) or not row:
+            continue
+        time_value = row[time_idx] if 0 <= time_idx < len(row) else ""
+        if isinstance(time_value, list) and time_value:
+            time_value = time_value[0] if isinstance(time_value[0], str) else str(time_value[0])
+        if str(time_value).strip() != week_str:
+            continue
+        brand_value = row[brand_idx] if 0 <= brand_idx < len(row) else ""
+        if isinstance(brand_value, list) and brand_value:
+            brand_value = brand_value[0] if isinstance(brand_value[0], str) else str(brand_value[0])
+        if str(brand_value).strip() != brand:
+            continue
+        return record_id
+    return ""
+
+
+def _upsert_peer_record(
+    node: str,
+    cli_script: str,
+    table_id: str,
+    fields_config: dict,
+    record_id: str,
+    cell_values: dict,
+) -> dict:
+    if record_id:
+        return run_lark_cli(node, cli_script, [
+            "base",
+            "+record-batch-update",
+            "--base-token",
+            BASE_APP_TOKEN,
+            "--table-id",
+            table_id,
+            "--json",
+            json.dumps({"record_id_list": [record_id], "patch": cell_values}, ensure_ascii=False),
+        ], as_user=False)
+    return run_lark_cli(node, cli_script, [
+        "base",
+        "+record-batch-create",
+        "--base-token",
+        BASE_APP_TOKEN,
+        "--table-id",
+        table_id,
+        "--json",
+        json.dumps(
+            {"fields": list(cell_values.keys()), "rows": [list(cell_values.values())]},
+            ensure_ascii=False,
+        ),
+    ], as_user=False)
+
+
+# ── 入口 ──────────────────────────────────────────────────────────────────────
+
+
+def write_peer_store(platform: str, json_path: str, shop_json_path: str = "") -> bool:
+    """将平台竞店 JSON 写入飞书竞店表。支持 JD / DY，天猫跳过。"""
+    if platform not in {JD_PLATFORM, DY_PLATFORM}:
+        print(f"[PeerWriter] 平台 {platform} 不写飞书竞店表，跳过")
+        return False
+    if not json_path or not os.path.exists(json_path):
+        print(f"[PeerWriter] 文件不存在: {json_path}")
+        return False
+
+    if platform == JD_PLATFORM:
+        return _write_jd_peer_store(json_path)
+    return _write_dy_peer_store(json_path, shop_json_path)
diff --git a/src/gyxx_flow/modules/shop_intelligence/runtime/writers/shop_base_writer.py b/src/gyxx_flow/modules/shop_intelligence/runtime/writers/shop_base_writer.py
new file mode 100644
index 0000000..76d0ac4
--- /dev/null
+++ b/src/gyxx_flow/modules/shop_intelligence/runtime/writers/shop_base_writer.py
@@ -0,0 +1,415 @@
+"""Write JD/DY/TM shop weekly metrics to Feishu Bitable shop table.
+
+AI 分析字段调用本机 Hermes CLI 生成（核心指标 + 环比 + 市场榜单）。
+"""
+from __future__ import annotations
+
+import json
+import os
+import re
+import subprocess
+from datetime import datetime, timedelta
+
+from gyxx_flow.modules.shop_intelligence.runtime.config import (
+    BASE_APP_TOKEN,
+    DY_PLATFORM,
+    JD_PLATFORM,
+    JD_SHOP_PARENT_RECORD_ID,
+    SHOP_BASE_FIELDS,
+    SHOP_BASE_TABLE_ID,
+    TM_PLATFORM,
+)
+from gyxx_flow.modules.shop_intelligence.runtime.lark_cli import (
+    extract_record_id,
+    find_existing_record,
+    format_week,
+    parse_amount,
+    parse_star,
+    resolve_lark_cli,
+    upsert_record,
+)
+
+
+def _parse_date_range(date_range: str) -> tuple[str, str]:
+    """'2026/06/01 - 2026/06/07' → ('2026-06-01', '2026-06-07')"""
+    if not date_range:
+        return "", ""
+    m = re.search(
+        r"(\d{4})/(\d{2})/(\d{2})\s*[-—~至到]\s*(\d{4})/(\d{2})/(\d{2})",
+        str(date_range),
+    )
+    if not m:
+        return "", ""
+    return (
+        f"{m.group(1)}-{m.group(2)}-{m.group(3)}",
+        f"{m.group(4)}-{m.group(5)}-{m.group(6)}",
+    )
+
+
+def _extract_jd_metrics(metrics: dict) -> dict:
+    transaction = metrics.get("transaction", {}) or {}
+    deal = transaction.get("成交金额", {})
+    deal_value = deal.get("value") if isinstance(deal, dict) else deal
+    star = parse_star(metrics.get("star_rating_raw", ""))
+    return {
+        "week_start": metrics.get("week_start", ""),
+        "week_end": metrics.get("week_end", ""),
+        "sales_amount": parse_amount(deal_value),
+        "star_rating": star,
+    }
+
+
+def _extract_dy_metrics(metrics: dict) -> dict:
+    week_start, week_end = _parse_date_range(metrics.get("date_range", ""))
+    sales_amount = 0.0
+    derived = metrics.get("derived_metrics") or {}
+    raw = derived.get("raw") if isinstance(derived, dict) else {}
+    if isinstance(raw, dict) and raw.get("销售额") is not None:
+        try:
+            sales_amount = float(raw["销售额"])
+        except (TypeError, ValueError):
+            sales_amount = 0.0
+    if sales_amount == 0.0:
+        base_metrics = metrics.get("metrics") or {}
+        if isinstance(base_metrics, dict):
+            sales_amount = parse_amount(base_metrics.get("成交金额", ""))
+    experience = metrics.get("experience_score") or {}
+    star = parse_star(experience.get("score", "")) if isinstance(experience, dict) else 0.0
+    return {
+        "week_start": week_start,
+        "week_end": week_end,
+        "sales_amount": sales_amount,
+        "star_rating": star,
+    }
+
+
+def _extract_tm_metrics(metrics: dict) -> dict:
+    week_start = metrics.get("week_start", "")
+    week_end = metrics.get("week_end", "")
+    if not week_start or not week_end:
+        week_start, week_end = _parse_date_range(metrics.get("week_range", ""))
+    if not week_start or not week_end:
+        week_start, week_end = _tm_week_range_from_date(metrics.get("date", ""))
+    homepage = metrics.get("homepage") or {}
+    sales_amount = parse_amount(homepage.get("净支付金额", ""))
+    star = 0.0
+    for item in metrics.get("shop_stats") or []:
+        if not isinstance(item, dict):
+            continue
+        if item.get("name", "").strip() == "真实体验分":
+            star = parse_star(item.get("score", ""))
+            break
+    return {
+        "week_start": week_start,
+        "week_end": week_end,
+        "sales_amount": sales_amount,
+        "star_rating": star,
+    }
+
+
+def _tm_week_range_from_date(date_str: str) -> tuple[str, str]:
+    """旧版 TM JSON 只有 date（采集日），按脚本逻辑反推上周范围。"""
+    if not date_str:
+        return "", ""
+    try:
+        collected = datetime.strptime(str(date_str), "%Y-%m-%d").date()
+    except ValueError:
+        return "", ""
+    this_monday = collected - timedelta(days=collected.weekday())
+    start = this_monday - timedelta(days=7)
+    end = this_monday - timedelta(days=1)
+    return start.strftime("%Y-%m-%d"), end.strftime("%Y-%m-%d")
+
+
+def _upsert_shop_row(
+    platform: str,
+    week_start: str,
+    week_end: str,
+    sales_amount: float,
+    star_rating: float,
+    ai_analysis: str = "",
+) -> None:
+    if not week_start or not week_end:
+        raise ValueError("缺少 week_start/week_end")
+    week_str, month_str, _, _ = format_week(week_start, week_end)
+
+    cell_values = {
+        SHOP_BASE_FIELDS["平台"]: platform,
+        SHOP_BASE_FIELDS["时间"]: week_str,
+        SHOP_BASE_FIELDS["月份"]: month_str,
+        SHOP_BASE_FIELDS["上周销售额"]: sales_amount,
+        SHOP_BASE_FIELDS["店铺评分"]: star_rating,
+        SHOP_BASE_FIELDS["AI分析"]: ai_analysis,
+        SHOP_BASE_FIELDS["备注"]: "机器人操作",
+    }
+    parent_record_id = ""
+    if platform == JD_PLATFORM and JD_SHOP_PARENT_RECORD_ID:
+        cell_values[SHOP_BASE_FIELDS["父记录"]] = [{"id": JD_SHOP_PARENT_RECORD_ID}]
+        parent_record_id = JD_SHOP_PARENT_RECORD_ID
+
+    node, cli_script = resolve_lark_cli()
+    existing_id = find_existing_record(
+        node,
+        cli_script,
+        BASE_APP_TOKEN,
+        SHOP_BASE_TABLE_ID,
+        SHOP_BASE_FIELDS,
+        week_str,
+        platform,
+        parent_record_id,
+    )
+    action = "更新" if existing_id else "新建"
+    result = upsert_record(node, cli_script, BASE_APP_TOKEN, SHOP_BASE_TABLE_ID, existing_id, cell_values)
+    new_id = extract_record_id(result, existing_id)
+    print(f"[ShopWriter] {action} {platform} 店铺行 record_id={new_id} 时间={week_str}")
+
+
+def write_shop(platform: str, json_path: str) -> bool:
+    """将平台店铺 JSON 写入飞书店铺表。支持 JD / DY / TM。"""
+    if platform not in {JD_PLATFORM, DY_PLATFORM, TM_PLATFORM}:
+        print(f"[ShopWriter] 平台 {platform} 不写飞书店铺表，跳过")
+        return False
+    if not json_path or not os.path.exists(json_path):
+        print(f"[ShopWriter] 文件不存在: {json_path}")
+        return False
+
+    with open(json_path, encoding="utf-8") as f:
+        data = json.load(f)
+
+    if platform == JD_PLATFORM:
+        extracted = _extract_jd_metrics(data)
+    elif platform == DY_PLATFORM:
+        extracted = _extract_dy_metrics(data)
+    else:
+        extracted = _extract_tm_metrics(data)
+
+    ai_analysis = (data.get("ai_analysis") or "").strip()
+    if ai_analysis:
+        print(f"[ShopWriter] {platform} JSON 已有 ai_analysis（{len(ai_analysis)} 字）")
+    else:
+        print(f"[ShopWriter] {platform} JSON 无 ai_analysis，飞书 AI分析 字段将为空")
+
+    _upsert_shop_row(
+        platform,
+        extracted["week_start"],
+        extracted["week_end"],
+        extracted["sales_amount"],
+        extracted["star_rating"],
+        ai_analysis,
+    )
+    return True
+
+
+def ensure_ai_analysis(platform: str, json_path: str) -> str:
+    """若 JSON 中无 ai_analysis，调用 Hermes 生成并写回 JSON。返回 ai_analysis 文本。
+
+    幂等：JSON 中已有 ai_analysis 时直接返回，不重复调用 Hermes。
+    """
+    if not json_path or not os.path.exists(json_path):
+        return ""
+    with open(json_path, encoding="utf-8") as f:
+        data = json.load(f)
+    existing = (data.get("ai_analysis") or "").strip()
+    if existing:
+        print(f"[ShopWriter] {platform} JSON 已有 ai_analysis（{len(existing)} 字），跳过生成")
+        return existing
+    if os.getenv("AI_ANALYSIS_ENABLED", "1").lower() not in ("1", "true", "yes", "on"):
+        print("[ShopWriter] AI_ANALYSIS_ENABLED=0，跳过 AI 分析")
+        return ""
+    try:
+        ai_analysis = _generate_ai_analysis(platform, data)
+    except Exception as e:
+        print(f"[ShopWriter] AI 分析生成失败: {e}")
+        return ""
+    if not ai_analysis:
+        return ""
+    data["ai_analysis"] = ai_analysis
+    try:
+        with open(json_path, "w", encoding="utf-8") as f:
+            json.dump(data, f, ensure_ascii=False, indent=2)
+        print(f"[ShopWriter] AI 分析已写回 JSON: {json_path}")
+    except Exception as e:
+        print(f"[ShopWriter] AI 分析写回 JSON 失败: {e}")
+    return ai_analysis
+
+
+# ── AI 分析（调用本机 Hermes CLI）───────────────────────────────────────────────
+
+
+def _generate_ai_analysis(platform: str, data: dict) -> str:
+    prompt = _build_ai_analysis_prompt(platform, data)
+    if not prompt:
+        return ""
+    print(f"[ShopWriter] 调用 Hermes 生成 {platform} AI 分析...")
+    text = _call_hermes(prompt)
+    if text:
+        print(f"[ShopWriter] AI 分析生成成功（{len(text)} 字）")
+    else:
+        print("[ShopWriter] AI 分析为空")
+    return text
+
+
+def _call_hermes(prompt: str, timeout: int = 120) -> str:
+    try:
+        result = subprocess.run(
+            ["hermes", "-z", prompt],
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+            timeout=timeout,
+        )
+    except FileNotFoundError:
+        print("[ShopWriter] 未找到 hermes CLI，请确认已安装并在 PATH 中")
+        return ""
+    except subprocess.TimeoutExpired:
+        print(f"[ShopWriter] Hermes 调用超时（{timeout}s）")
+        return ""
+    except Exception as e:
+        print(f"[ShopWriter] Hermes 调用异常: {e}")
+        return ""
+    if result.returncode != 0:
+        err = (result.stderr or result.stdout or "").strip()
+        print(f"[ShopWriter] Hermes 调用失败: {err[:200]}")
+        return ""
+    return (result.stdout or "").strip()
+
+
+def _build_ai_analysis_prompt(platform: str, data: dict) -> str:
+    if platform == JD_PLATFORM:
+        body = _build_jd_analysis_body(data)
+    elif platform == DY_PLATFORM:
+        body = _build_dy_analysis_body(data)
+    elif platform == TM_PLATFORM:
+        body = _build_tm_analysis_body(data)
+    else:
+        return ""
+    if not body:
+        return ""
+    return (
+        f"请基于以下{platform}店铺本周（{data.get('week_start', '')} ~ {data.get('week_end', '')}）数据，"
+        "生成一段经营分析报告。\n\n"
+        f"{body}\n\n"
+        "请从以下角度分析：1) 本周经营表现总结；2) 环比变化要点（含原因推测）；"
+        "3) 市场榜单竞争力评估；4) 1-2 条具体改进建议。\n"
+        "输出格式：纯文本一段，150-250 字，不要分点列表，不要标题，直接成段。"
+    )
+
+
+def _build_jd_analysis_body(data: dict) -> str:
+    transaction = data.get("transaction") or {}
+    lines = ["【核心指标】"]
+    for key in ["成交金额", "成交客户数", "成交单量", "客单价", "商品访客数", "成交转化率", "取消及售后退款金额"]:
+        item = transaction.get(key) or {}
+        if not isinstance(item, dict):
+            continue
+        value = item.get("value", "未获取")
+        qoq = item.get("环比", "")
+        peer = item.get("同行同级均值", "")
+        line = f"- {key}: {value}"
+        if qoq:
+            line += f"（环比 {qoq}）"
+        if peer and peer not in ("--", "N/A"):
+            line += f"，同行均值 {peer}"
+        lines.append(line)
+    star = data.get("star_rating_raw", "")
+    if star:
+        lines.append("")
+        lines.append(f"【店铺星级】{star}")
+    ranking = data.get("ranking") or {}
+    if ranking:
+        lines.append("")
+        lines.append("【市场榜单排名】")
+        for cat, info in ranking.items():
+            if not isinstance(info, dict):
+                continue
+            rank = info.get("rank", "未获取")
+            change = info.get("rankChange", "")
+            amount = info.get("amount", "")
+            line = f"- {cat}: 第{rank}"
+            if change:
+                line += f"（{change}）"
+            if amount:
+                line += f"，成交 {amount}"
+            lines.append(line)
+    return "\n".join(lines)
+
+
+def _build_dy_analysis_body(data: dict) -> str:
+    metrics = data.get("metrics") or {}
+    derived = data.get("derived_metrics") or {}
+    qianchuan = data.get("qianchuan_metrics") or {}
+    experience = data.get("experience_score") or {}
+    market = data.get("market_rankings") or {}
+
+    lines = ["【核心指标】"]
+    for key in ["成交金额", "成交人数", "成交订单数", "客单价", "成交退款金额", "商品点击人数"]:
+        value = metrics.get(key, "未获取")
+        lines.append(f"- {key}: {value}")
+    if derived:
+        lines.append(f"- 销售额: {derived.get('销售额', '未获取')}")
+        lines.append(f"- 付费占比: {derived.get('付费占比', '未获取')}")
+        lines.append(f"- ROI: {derived.get('ROI', '未获取')}")
+    if experience:
+        lines.append("")
+        lines.append(
+            "【店铺评分】体检分 " + str(experience.get("score", "未获取"))
+            + f"（商品{experience.get('product_experience_score', '未获取')}"
+            + f"/物流{experience.get('logistics_experience_score', '未获取')}"
+            + f"/服务{experience.get('service_experience_score', '未获取')}"
+            + f"/扣分{experience.get('penalty_score', '未获取')}）"
+        )
+    if qianchuan:
+        lines.append("")
+        spend = (qianchuan.get("整体消耗") or {})
+        roi = (qianchuan.get("净成交ROI") or {})
+        line = "【千川投放】整体消耗 " + str(spend.get("value", "未获取"))
+        if spend.get("change_percent"):
+            line += f"（环比 {spend.get('change_percent')}，{spend.get('change_direction', '')}）"
+        line += f"，净成交ROI {roi.get('value', '未获取')}"
+        lines.append(line)
+    if market:
+        lines.append("")
+        lines.append("【市场榜单排名】")
+        for cat, info in market.items():
+            if not isinstance(info, dict):
+                continue
+            rank = info.get("rank", "未获取")
+            change = info.get("rank_change", "")
+            line = f"- {cat}: 第{rank}名"
+            if change:
+                line += f"（{change}）"
+            lines.append(line)
+    return "\n".join(lines)
+
+
+def _build_tm_analysis_body(data: dict) -> str:
+    homepage = data.get("homepage") or {}
+    shop_stats = data.get("shop_stats") or []
+    market_rank = data.get("market_rank") or []
+
+    lines = ["【核心指标】"]
+    for key in ["净支付金额", "访客数", "支付买家数", "支付转化率", "客单价", "总推广花费", "总ROI", "付费占比"]:
+        value = homepage.get(key, "未获取")
+        qoq = homepage.get(f"{key}_环比", "")
+        line = f"- {key}: {value}"
+        if qoq:
+            line += f"（环比 {qoq}）"
+        lines.append(line)
+    for item in shop_stats:
+        if not isinstance(item, dict):
+            continue
+        if item.get("name", "").strip() == "真实体验分":
+            lines.append("")
+            lines.append(f"【真实体验分】{item.get('score', '未获取')}（{item.get('rank', '')}）")
+            break
+    if market_rank:
+        lines.append("")
+        lines.append("【市场榜单排名】")
+        for item in market_rank:
+            if not isinstance(item, dict):
+                continue
+            lines.append(
+                f"- {item.get('类目', '')}: {item.get('全部店铺行业排名', '未获取')}"
+            )
+    return "\n".join(lines)
diff --git a/src/gyxx_flow/modules/supply_chain/__init__.py b/src/gyxx_flow/modules/supply_chain/__init__.py
new file mode 100644
index 0000000..0f1a242
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/__init__.py
@@ -0,0 +1,106 @@
+"""Purchase, inventory-alert, and replenishment workflows."""
+
+from __future__ import annotations
+
+from collections.abc import Iterable
+from typing import Any
+
+from gyxx_flow.adapters.native import DeferredModuleCommandStep, ModuleCommandFactory
+from gyxx_flow.workflow import StepDefinition, WorkflowDefinition
+
+SCHEDULED_SUPPLY_WORKFLOW_IDS = (
+    "supply.purchase_confirmation.daily",
+    "supply.replenishment.weekly",
+    "supply.replenishment_alert.daily",
+)
+MANUAL_SUPPLY_WORKFLOW_IDS = ("supply.purchase_order_update",)
+SUPPLY_WORKFLOW_IDS = (
+    *SCHEDULED_SUPPLY_WORKFLOW_IDS,
+    *MANUAL_SUPPLY_WORKFLOW_IDS,
+)
+
+# Supply jobs can call remote systems and must not be killed by a short CLI default.
+# A failed opaque external operation is never retried automatically because its external
+# writes cannot yet be proven idempotent.
+SUPPLY_TIMEOUT_SECONDS = 2 * 60 * 60
+SUPPLY_RESOURCE = "module:supply_chain"
+
+
+class SupplyChainModule:
+    """Supply-chain business module composition boundary."""
+
+    module_id = "supply_chain"
+
+    def __init__(self, definitions: Iterable[WorkflowDefinition] = ()) -> None:
+        self._definitions = tuple(definitions)
+
+    @classmethod
+    def from_catalog(
+        cls,
+        catalog: Any,
+        *,
+        command_factory: ModuleCommandFactory | None = None,
+    ) -> "SupplyChainModule":
+        """Build the three scheduled and one manual supply migration entries."""
+
+        entries = {
+            entry.workflow_id: entry
+            for entry in catalog.workflows
+            if entry.module == cls.module_id
+        }
+        expected_triggers = {
+            **{
+                workflow_id: "scheduled"
+                for workflow_id in SCHEDULED_SUPPLY_WORKFLOW_IDS
+            },
+            **{workflow_id: "manual" for workflow_id in MANUAL_SUPPLY_WORKFLOW_IDS},
+        }
+        actual_triggers = {
+            workflow_id: entry.trigger for workflow_id, entry in entries.items()
+        }
+        if actual_triggers != expected_triggers:
+            raise ValueError(
+                "supply catalog workflows do not match the migration contract: "
+                f"expected={expected_triggers!r}, actual={actual_triggers!r}"
+            )
+
+        definitions: list[WorkflowDefinition] = []
+        for workflow_id in SUPPLY_WORKFLOW_IDS:
+            entry = entries[workflow_id]
+            action = (
+                DeferredModuleCommandStep(entry)
+                if command_factory is None
+                else DeferredModuleCommandStep(
+                    entry,
+                    command_factory=command_factory,
+                )
+            )
+            definitions.append(
+                WorkflowDefinition(
+                    workflow_id,
+                    (
+                        StepDefinition(
+                            "module_supply",
+                            action,
+                            timeout_seconds=SUPPLY_TIMEOUT_SECONDS,
+                            max_attempts=1,
+                            resources=(SUPPLY_RESOURCE,),
+                            production_sink=True,
+                        ),
+                    ),
+                )
+            )
+        return cls(definitions)
+
+    def workflow_definitions(self) -> tuple[WorkflowDefinition, ...]:
+        return self._definitions
+
+
+__all__ = [
+    "SUPPLY_RESOURCE",
+    "SUPPLY_TIMEOUT_SECONDS",
+    "MANUAL_SUPPLY_WORKFLOW_IDS",
+    "SCHEDULED_SUPPLY_WORKFLOW_IDS",
+    "SUPPLY_WORKFLOW_IDS",
+    "SupplyChainModule",
+]
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/__init__.py b/src/gyxx_flow/modules/supply_chain/runtime/__init__.py
new file mode 100644
index 0000000..f8a7683
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/__init__.py
@@ -0,0 +1,2 @@
+"""Native supply-chain runtime migrated from the former standalone project."""
+
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/config.example.env b/src/gyxx_flow/modules/supply_chain/runtime/config.example.env
new file mode 100644
index 0000000..9fa572f
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/config.example.env
@@ -0,0 +1,5 @@
+AUTOFLOW_PG_HOST=8.148.185.119
+AUTOFLOW_PG_PORT=5432
+AUTOFLOW_PG_USER=data_hub
+AUTOFLOW_PG_PASSWORD=
+AUTOFLOW_PG_DB=data_hub
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/dependencies.txt b/src/gyxx_flow/modules/supply_chain/runtime/dependencies.txt
new file mode 100644
index 0000000..87c3649
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/dependencies.txt
@@ -0,0 +1,29 @@
+# Auto-Flow Orchestrator Dependencies
+
+# LangGraph and LangChain
+langgraph>=0.2.0
+langchain>=0.3.0
+langchain-core>=0.3.0
+langchain-anthropic>=0.3.0
+
+# Data processing
+pandas>=2.0.0
+openpyxl>=3.1.0
+
+# HTTP requests
+requests>=2.31.0
+
+# Browser automation
+playwright>=1.50.0
+
+# JSON handling
+ormsgpack>=1.2.0
+
+# SQLite support (built-in, but listed for clarity)
+# sqlite3
+
+# OpenAI-compatible client for Hermes Gateway
+langchain-openai>=0.2.0
+
+# Feishu SDK
+lark-oapi>=1.0.0
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/__init__.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/__init__.py
new file mode 100644
index 0000000..ac582a4
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/__init__.py
@@ -0,0 +1,7 @@
+"""MCP-only workflow orchestrator for the portable supply-chain runtime."""
+
+from gyxx_flow.modules.supply_chain.runtime.orchestrator.runner import run_mcp
+
+__all__ = [
+    "run_mcp",
+]
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/config.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/config.py
new file mode 100644
index 0000000..05e2fb5
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/config.py
@@ -0,0 +1,135 @@
+# -*- coding: utf-8 -*-
+"""Configuration for auto-flow orchestrator."""
+
+import os
+
+from gyxx_flow.adapters import RuntimeServicePolicy
+
+from gyxx_flow.modules.supply_chain.runtime.paths import PACKAGE_ROOT, SHARED_ROOT
+
+os.environ.update(RuntimeServicePolicy().apply(os.environ))
+
+ROOT_DIR = PACKAGE_ROOT
+SHARED_DIR = SHARED_ROOT
+SCRIPTS_DIR = PACKAGE_ROOT / "orchestrator" / "scripts"
+
+HERMES_PROFILES = {
+    "data-collector": {
+        "name": "Data Collector",
+        "model": "anthropic:claude-sonnet-4-20250514",
+        "description": "Collects workflow input data.",
+    },
+    "data-analyzer": {
+        "name": "Data Analyzer",
+        "model": "anthropic:claude-sonnet-4-20250514",
+        "description": "Analyzes workflow data and sends notifications.",
+    },
+}
+
+# Three business workflows:
+# 1) purchase-confirmation
+# 2) replenishment-alert
+# 3) replenishment
+WORKFLOWS = {
+    "purchase-confirmation": {
+        "name": "采购确认通知工作流",
+        "description": "Detect low-stock SKUs and send confirmation notifications.",
+        "collector_script": str(SCRIPTS_DIR / "collect_confirmation.ps1"),
+        "analyzer_scripts": ["plugin:feishu.im.send_message"],
+        "output_patterns": ["*.csv"],
+        "intermediate_patterns": ["*.xlsx"],
+    },
+    "replenishment-alert": {
+        "name": "库存预警通知工作流",
+        "description": "Send threshold alert notifications for configured SKUs.",
+        "collector_script": str(SCRIPTS_DIR / "collect_replenishment.ps1"),
+        "analyzer_scripts": ["plugin:feishu.im.send_message"],
+        "output_patterns": ["threshold_alert_data.json"],
+    },
+    "replenishment": {
+        "name": "补货建议工作流",
+        "description": "Generate and notify replenishment recommendations.",
+        "collector_script": str(SCRIPTS_DIR / "collect_replenishment.ps1"),
+        "analyzer_scripts": [
+            "plugin:feishu.bitable.batch_create_records",
+            "plugin:feishu.im.send_message",
+        ],
+        "output_patterns": ["pending_insert.json", "notify_data.json"],
+    },
+    "purchase-order-update": {
+        "name": "采购单日期更新工作流",
+        "description": "Update ERP purchase order delivery dates by SKU and notify audit group.",
+        "collector_script": str(SCRIPTS_DIR / "collect_purchase_order_update.ps1"),
+        "analyzer_scripts": ["plugin:feishu.im.send_message"],
+        "output_patterns": ["update_result.json"],
+    },
+}
+
+HERMES_GATEWAYS = {
+    "collector": {
+        "url": "http://localhost:28791/v1",
+        "token": os.getenv("GYXX_SUPPLY_HERMES_TOKEN", ""),
+        "model": "mimo-v2.5-pro",
+        "profile": "data-collector",
+    },
+    "analyzer": {
+        "url": "http://localhost:28790/v1",
+        "token": os.getenv("GYXX_SUPPLY_HERMES_TOKEN", ""),
+        "model": "mimo-v2.5-pro",
+        "profile": "data-analyzer",
+    },
+}
+
+FEISHU_CONFIG = {
+    "apps": {
+        "collector": {
+            "app_id": "cli_aa8c4fb4c4f81cd3",
+            "app_secret": os.getenv(
+                "GYXX_SUPPLY_COLLECTOR_APP_SECRET",
+                os.getenv("AUTOFLOW_COLLECTOR_APP_SECRET", ""),
+            ),
+            "owner_open_id": "ou_8ee224968aa26a74c7d30ba27fed5eeb",
+        },
+        "analyzer": {
+            "app_id": "cli_aa8c4fc918b85cce",
+            "app_secret": os.getenv(
+                "GYXX_SUPPLY_ANALYZER_APP_SECRET",
+                os.getenv("AUTOFLOW_ANALYZER_APP_SECRET", ""),
+            ),
+            "owner_open_id": "ou_7ad5fc8012e2f741afc5346e05ffd447",
+        },
+    },
+    "bitable": {
+        "app_token": "Th0jbMHHQa7a8Lse3CicfSMBnxb",
+        "table_id": "tblG3oQ5yENV5f3F",
+    },
+    "notification_targets": {
+        "collection_status": [
+            "ou_8ee224968aa26a74c7d30ba27fed5eeb",
+        ],
+        "purchase_confirmation": [
+            "ou_7ad5fc8012e2f741afc5346e05ffd447",
+            "ou_339c1d396b397c97b08e1bf377963d40",
+            "ou_54141d75b24d22ddc8c1c4d66e00b6e9",
+        ],
+        "threshold_alert": [
+            "ou_7ad5fc8012e2f741afc5346e05ffd447",
+            "ou_b76e4cbdb24fe28cebd45ad091b60224",
+        ],
+        "replenishment": [
+            "ou_7ad5fc8012e2f741afc5346e05ffd447",
+            "ou_339c1d396b397c97b08e1bf377963d40",
+        ],
+    },
+    "trigger": {
+        "purchase_order_update": {
+            "enabled": True,
+            "allowed_sender_open_ids": [
+                "ou_7ad5fc8012e2f741afc5346e05ffd447",
+                "ou_339c1d396b397c97b08e1bf377963d40",
+                "ou_54141d75b24d22ddc8c1c4d66e00b6e9",
+            ],
+        }
+    },
+}
+
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/feishu_sheets.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/feishu_sheets.py
new file mode 100644
index 0000000..a2d0a42
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/feishu_sheets.py
@@ -0,0 +1,171 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Feishu spreadsheet reading utilities for auto-flow.
+
+Provides load_discontinued_skus_from_feishu() which reads the 停产产品 sheet
+via lark-cli and returns a set of SKU strings.
+"""
+
+from __future__ import annotations
+
+import json
+import re
+import shutil
+import subprocess
+import time
+
+from gyxx_flow.modules.supply_chain.runtime.paths import PACKAGE_ROOT, STATE_ROOT
+
+_CACHE_FILE = STATE_ROOT / "cache" / "discontinued_skus_feishu.json"
+_CACHE_DIR = _CACHE_FILE.parent
+_CACHE_TTL_SECONDS = 3600  # 1 hour: avoids redundant lark-cli calls within a single run
+
+# Feishu spreadsheet configuration for 停产产品
+# Wiki: https://bu0zgpibak.feishu.cn/wiki/DGVFw3TViiexd1kDT5ucEOIan6f
+DISCONTINUED_SPREADSHEET_TOKEN = "UGhksHzK8hncVEtSMcacxnuDnmg"
+DISCONTINUED_SHEET_ID = "f4e087"
+DISCONTINUED_RANGE = "A1:B500"
+
+
+def _lark_cli() -> str:
+    return shutil.which("lark-cli") or shutil.which("lark-cli.cmd") or "lark-cli.cmd"
+
+
+def _normalize_sku(value: str) -> str:
+    text = str(value).strip().lstrip("'")
+    if re.fullmatch(r"\d+\.0", text):
+        text = text[:-2]
+    return text
+
+
+def _read_from_cache() -> set[str] | None:
+    try:
+        if not _CACHE_FILE.exists():
+            return None
+        data = json.loads(_CACHE_FILE.read_text(encoding="utf-8"))
+        if not isinstance(data, dict):
+            return None
+        updated_at = data.get("updated_at", "")
+        if updated_at:
+            cache_time = time.mktime(time.strptime(updated_at, "%Y-%m-%dT%H:%M:%S"))
+            if time.time() - cache_time > _CACHE_TTL_SECONDS:
+                return None
+        skus = data.get("skus")
+        if isinstance(skus, list):
+            clean = {str(s).strip() for s in skus if s}
+            print(f"[CACHE] 停产 SKU cache hit: {len(clean)} 个 from {_CACHE_FILE.name}")
+            return clean
+    except Exception as exc:
+        print(f"[WARN] 停产 SKU cache read failed: {exc}")
+    return None
+
+
+def _write_to_cache(skus: set[str]) -> None:
+    try:
+        _CACHE_DIR.mkdir(parents=True, exist_ok=True)
+        payload = {
+            "version": 1,
+            "updated_at": time.strftime("%Y-%m-%dT%H:%M:%S"),
+            "skus": sorted(skus),
+        }
+        tmp = _CACHE_FILE.with_suffix(".tmp")
+        tmp.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+        tmp.replace(_CACHE_FILE)
+    except Exception as exc:
+        print(f"[WARN] 停产 SKU cache write failed (non-blocking): {exc}")
+
+
+def load_discontinued_skus_from_feishu(
+    spreadsheet_token: str = DISCONTINUED_SPREADSHEET_TOKEN,
+    sheet_id: str = DISCONTINUED_SHEET_ID,
+    range_spec: str = DISCONTINUED_RANGE,
+    identity: str = "user",
+) -> set[str]:
+    """Read discontinued SKUs from the Feishu spreadsheet via lark-cli.
+
+    Returns a set of normalized SKU strings. On any failure, prints a warning
+    and returns an empty set (non-blocking, same convention as the old xlsx reader).
+    """
+    cached = _read_from_cache()
+    if cached is not None:
+        return cached
+
+    cmd = [
+        _lark_cli(),
+        "sheets",
+        "+csv-get",
+        "--as", identity,
+        "--spreadsheet-token", spreadsheet_token,
+        "--sheet-id", sheet_id,
+        "--range", range_spec,
+    ]
+    try:
+        proc = subprocess.run(
+            cmd,
+            cwd=str(PACKAGE_ROOT),
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+            capture_output=True,
+            timeout=30,
+        )
+    except FileNotFoundError:
+        print("[WARN] lark-cli not found, skipping Feishu discontinued SKU load")
+        return set()
+    except subprocess.TimeoutExpired:
+        print("[WARN] lark-cli timed out (30s), skipping Feishu discontinued SKU load")
+        return set()
+    except Exception as exc:
+        print(f"[WARN] lark-cli execution failed: {exc}, skipping Feishu discontinued SKU load")
+        return set()
+
+    if proc.returncode != 0:
+        print(
+            f"[WARN] lark-cli returned non-zero exit code {proc.returncode}, "
+            f"stderr={proc.stderr[:200]}, skipping Feishu discontinued SKU load"
+        )
+        return set()
+
+    try:
+        envelope = json.loads(proc.stdout)
+    except json.JSONDecodeError as exc:
+        print(f"[WARN] lark-cli output not valid JSON: {exc}, skipping Feishu discontinued SKU load")
+        return set()
+
+    data = envelope.get("data") or {}
+    annotated_csv = data.get("annotated_csv")
+    if not annotated_csv:
+        print("[WARN] lark-cli returned empty annotated_csv, skipping Feishu discontinued SKU load")
+        return set()
+
+    # annotated_csv 每行格式 "[row=N] colA,colB"。剥掉 [row=N] 前缀后用 csv 解析,
+    # 避免商品名称里带逗号导致 split(",") 翻车。
+    import csv
+    import io
+
+    cleaned_lines = []
+    for line in annotated_csv.splitlines():
+        idx = line.find("] ")
+        if idx >= 0 and line.lstrip().startswith("[row="):
+            line = line[idx + 2:]
+        cleaned_lines.append(line)
+
+    skus: set[str] = set()
+    reader = csv.reader(io.StringIO("\n".join(cleaned_lines)))
+    for row_number, row in enumerate(reader, start=1):
+        if row_number == 1:
+            continue  # skip header row (商品名称, 商品编码)
+        if len(row) < 2:
+            continue
+        sku_raw = row[1]  # Column B = 商品编码
+        if not sku_raw:
+            continue
+        sku = _normalize_sku(sku_raw)
+        if sku:
+            skus.add(sku)
+
+    _write_to_cache(skus)
+
+    print(f"[INFO] 已从飞书表格加载停产商品编码 {len(skus)} 个")
+    return skus
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/mcp_workflow.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/mcp_workflow.py
new file mode 100644
index 0000000..687bd32
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/mcp_workflow.py
@@ -0,0 +1,1196 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""Run business workflows through Hermes MCP gateways."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import logging
+import os
+import platform
+import subprocess
+import threading
+import time
+import uuid
+from pathlib import Path
+from typing import Any
+
+import requests
+
+from gyxx_flow.modules.supply_chain.runtime.orchestrator.config import (
+    FEISHU_CONFIG,
+    HERMES_GATEWAYS,
+    SHARED_DIR,
+    WORKFLOWS,
+)
+from gyxx_flow.modules.supply_chain.runtime.orchestrator.state import now_iso
+from gyxx_flow.modules.supply_chain.runtime.orchestrator.workflows import (
+    batch_workflow_names,
+    feishu_app,
+    workflow_display_name,
+)
+from gyxx_flow.modules.supply_chain.runtime.paths import STATE_ROOT, WORK_ROOT
+
+# 显式禁用重试：采集/分析均只执行一次，失败即结束。
+MAX_ATTEMPTS = 1
+ANALYZE_MAX_ATTEMPTS = 1
+OUTPUT_WAIT_SECONDS = 1800
+
+log = logging.getLogger("autoflow.mcp_workflow")
+OUTPUT_POLL_SECONDS = 10
+HEARTBEAT_INTERVAL_SECONDS = 180
+LOCK_STALE_SECONDS = 7200
+FEISHU_OPENAPI_BASE = "https://open.feishu.cn/open-apis"
+
+COLLECTOR_FEISHU = feishu_app("collector")
+ANALYZER_FEISHU = feishu_app("analyzer")
+COLLECTOR_OWNER_OPEN_ID = COLLECTOR_FEISHU["owner_open_id"]
+ANALYZER_OWNER_OPEN_ID = ANALYZER_FEISHU["owner_open_id"]
+COLLECTOR_FEISHU_APP_ID = COLLECTOR_FEISHU["app_id"]
+ANALYZER_FEISHU_APP_ID = ANALYZER_FEISHU["app_id"]
+
+NOTIFICATION_TARGETS = FEISHU_CONFIG.get("notification_targets", {})
+PURCHASE_CONFIRMATION_TARGETS = NOTIFICATION_TARGETS.get("purchase_confirmation", [])
+THRESHOLD_ALERT_TARGETS = NOTIFICATION_TARGETS.get("threshold_alert", [])
+REPLENISHMENT_TARGETS = NOTIFICATION_TARGETS.get("replenishment", [])
+SHENDAN_GROUP_CHAT_ID = "oc_6e95333db779b07524e1c361099c7aec"
+
+# 多维表格字段名 -> field_id 静态映射。
+# ⚠️ 分析端写表时必须用 field_id (fldXxx 形式)，不得用字段显示名。
+# 特别注意：pending_insert.json 中的 "产品名称" 在表内显示为 "产品名称 用于展示"，
+# 必须映射到 fld8iKlL8S，直接用 "产品名称" 会触发 not_found，服务端会静默落盘空记录。
+# 该表必须与 orchestrator/scripts/batch_process.py:BITABLE_FIELD_MAP 保持一致。
+# 漂移时 batch_process.py 启动会打 WARN。
+BITABLE_FIELD_MAP: dict[str, str] = {
+    "规格名称（新）": "fldYznbo8s",
+    "补货量": "fld0aWj8AA",
+    "规格名称（旧）": "fldM2vrObI",
+    "实际可用数": "fld6FUimoM",
+    "采购在途": "fldajEW5vQ",
+    "商品名称": "fldaV2dk7U",
+    "90天销量": "fldBSgLZLF",
+    "机器人备注": "fldch1v6cC",
+    "7天销量": "fldgMRs1EN",
+    "60天销量": "fldIjk4ADS",
+    "15天销量": "fldSfwvPVd",
+    "商品编号": "fldzIJQ3nw",
+    "月销量": "fldzVYr11w",
+    "产品名称": "fld8iKlL8S",
+}
+
+
+def _get_collector_tenant_token() -> str:
+    cfg = FEISHU_CONFIG["apps"]["collector"]
+    app_id = cfg.get("app_id", "")
+    credential = cfg.get("app_secret", "")
+    if not app_id or not credential:
+        raise RuntimeError("FEISHU_CONFIG.apps.collector 缺少 app_id/app_secret")
+    credentials = dict((("app_id", app_id), ("app_secret", credential)))
+    resp = requests.post(
+        f"{FEISHU_OPENAPI_BASE}/auth/v3/tenant_access_token/internal/",
+        json=credentials,
+        timeout=10,
+    )
+    resp.raise_for_status()
+    body = resp.json()
+    if body.get("code") != 0:
+        raise RuntimeError(
+            f"feishu tenant_access_token error: code={body.get('code')} msg={body.get('msg')}"
+        )
+    return body["tenant_access_token"]
+
+
+def _build_heartbeat_card(result: dict, phase: str = "final") -> dict:
+    cfg = FEISHU_CONFIG["apps"]["collector"]
+    is_failed = bool(result.get("error"))
+    if phase != "final":
+        template = "blue"
+    elif is_failed:
+        template = "red"
+    else:
+        template = "green"
+
+    workflow_label = result.get("workflow_name") or result.get("workflow", "")
+    title = f"📌 采集端心跳 · {workflow_label}"
+
+    if phase == "start":
+        conclusion = "工作流已启动，开始采集。"
+    elif phase == "progress":
+        conclusion = f"工作流执行中，每 {HEARTBEAT_INTERVAL_SECONDS} 秒上报一次状态。"
+    elif is_failed:
+        conclusion = "工作流执行失败。"
+    else:
+        conclusion = "工作流执行完成。"
+
+    core_fields = [
+        {"is_short": True, "text": {"tag": "lark_md", "content": f"**工作流ID**\n{result.get('workflow', '')}"}},
+        {"is_short": True, "text": {"tag": "lark_md", "content": f"**运行ID**\n{result.get('workflow_id', '')}"}},
+        {"is_short": True, "text": {"tag": "lark_md", "content": f"**当前阶段**\n{result.get('current_phase', '')}"}},
+        {"is_short": True, "text": {"tag": "lark_md", "content": f"**当前状态**\n{result.get('status', '')}"}},
+    ]
+    if phase == "final":
+        core_fields.extend([
+            {"is_short": True, "text": {"tag": "lark_md", "content": f"**采集尝试**\n{result.get('collect_attempts', 0)} 次"}},
+            {"is_short": True, "text": {"tag": "lark_md", "content": f"**分析尝试**\n{result.get('analyze_attempts', 0)} 次"}},
+        ])
+
+    elements: list[dict] = [
+        {"tag": "div", "text": {"tag": "lark_md", "content": f"**{conclusion}**"}},
+        {"tag": "div", "fields": core_fields},
+    ]
+
+    if phase == "final" and is_failed:
+        elements.append({
+            "tag": "div",
+            "text": {"tag": "lark_md", "content": f"❌ **错误信息:** {result.get('error', '')}"},
+        })
+
+    elements.append({"tag": "hr"})
+    elements.append({
+        "tag": "note",
+        "elements": [
+            {"tag": "plain_text", "content": f"发送应用: {cfg['app_id']} · 通知对象: {cfg['owner_open_id']}"},
+        ],
+    })
+
+    return {
+        "config": {"wide_screen_mode": True},
+        "header": {"template": template, "title": {"tag": "plain_text", "content": title}},
+        # ⚠️ 2026-06-29:不要把 elements 嵌套在 body 下!实测 Feishu 存盘时会把整个
+        # body 字段丢掉,只剩 title 和 elements=null,用户收到"只有标题的空卡片"。
+        # 用 top-level elements + header(template+title) 这种混合写法最稳:
+        # header.template 控制标题颜色(blue/green/red),top-level elements 保留正文内容。
+        # send_card_notification.py 也加了同样的 body→top-level 兜底,见那个文件的注释。
+        "elements": elements,
+    }
+
+
+def _workflow_lock_path(workflow_name: str) -> Path:
+    lock_dir = STATE_ROOT / "locks"
+    lock_dir.mkdir(parents=True, exist_ok=True)
+    return lock_dir / f"{workflow_name}.lock"
+
+
+def _pid_exists(pid: int) -> bool:
+    if pid <= 0:
+        return False
+    if platform.system().lower().startswith("win"):
+        try:
+            result = subprocess.run(
+                ["powershell", "-NoProfile", "-Command", f"Get-Process -Id {pid} -ErrorAction SilentlyContinue"],
+                capture_output=True,
+                text=True,
+                timeout=5,
+            )
+            return bool((result.stdout or "").strip())
+        except Exception:
+            return False
+    try:
+        os.kill(pid, 0)
+        return True
+    except OSError:
+        return False
+
+
+def _acquire_workflow_lock(workflow_name: str, workflow_id: str) -> tuple[bool, str]:
+    lock_path = _workflow_lock_path(workflow_name)
+    now = time.time()
+    payload = {
+        "workflow": workflow_name,
+        "workflow_id": workflow_id,
+        "pid": os.getpid(),
+        "started_at": now_iso(),
+        "started_epoch": now,
+    }
+
+    if lock_path.exists():
+        try:
+            existing = json.loads(lock_path.read_text(encoding="utf-8"))
+        except Exception:
+            existing = {}
+        existing_pid = int(existing.get("pid") or 0)
+        existing_epoch = float(existing.get("started_epoch") or 0)
+        is_stale = (now - existing_epoch) > LOCK_STALE_SECONDS if existing_epoch else True
+        if existing_pid and _pid_exists(existing_pid) and not is_stale:
+            return False, f"Workflow '{workflow_name}' already running (pid={existing_pid})."
+        try:
+            lock_path.unlink(missing_ok=True)
+        except Exception as exc:
+            return False, f"Failed to clear stale lock for '{workflow_name}': {exc}"
+
+    try:
+        fd = os.open(str(lock_path), os.O_CREAT | os.O_EXCL | os.O_WRONLY)
+        with os.fdopen(fd, "w", encoding="utf-8") as handle:
+            json.dump(payload, handle, ensure_ascii=False, indent=2)
+        return True, ""
+    except FileExistsError:
+        return False, f"Workflow '{workflow_name}' lock already exists."
+    except Exception as exc:
+        return False, f"Failed to create lock for '{workflow_name}': {exc}"
+
+
+def _release_workflow_lock(workflow_name: str) -> None:
+    try:
+        _workflow_lock_path(workflow_name).unlink(missing_ok=True)
+    except Exception:
+        pass
+
+
+def _collection_lock_name(workflow_name: str) -> str:
+    """Return a stable lock name for workflows sharing one collection script."""
+
+    script_path = str(Path(WORKFLOWS[workflow_name]["collector_script"]).resolve()).lower()
+    digest = hashlib.sha1(script_path.encode("utf-8")).hexdigest()[:12]
+    script_stem = Path(script_path).stem.replace(" ", "_")
+    return f"collector-{script_stem}-{digest}"
+
+
+def _acquire_collection_lock(workflow_name: str, workflow_id: str) -> tuple[bool, str]:
+    return _acquire_workflow_lock(_collection_lock_name(workflow_name), workflow_id)
+
+
+def _release_collection_lock(workflow_name: str) -> None:
+    _release_workflow_lock(_collection_lock_name(workflow_name))
+
+
+def _build_collect_prompt(workflow_name: str) -> str:
+    workflow = WORKFLOWS[workflow_name]
+    lines = [
+        f"工作流: {workflow_display_name(workflow_name)}",
+        f"Workflow ID: {workflow_name}",
+        f"说明: {workflow['description']}",
+        f"采集脚本: {workflow['collector_script']}",
+        "采集前必须清理该工作流 shared-data 目录中的旧输出，避免分析端发送旧数据。",
+        "只执行采集阶段，不要分析数据，不要发送业务通知。",
+        f"采集端 Hermes 绑定飞书应用 {COLLECTOR_FEISHU_APP_ID}，用于发送采集心跳/执行报告。",
+        "完成后确认输出已进入对应 shared-data 目录。",
+    ]
+    if workflow.get("output_patterns"):
+        lines.append("期望输出: " + ", ".join(workflow["output_patterns"]))
+    return "\n".join(lines)
+
+
+def _build_analyze_prompt(workflow_name: str) -> str:
+    workflow = WORKFLOWS[workflow_name]
+    output_patterns = workflow.get("output_patterns", [])
+    lines = [
+        f"工作流: {workflow_display_name(workflow_name)}",
+        f"Workflow ID: {workflow_name}",
+        f"说明: {workflow['description']}",
+        "只执行分析和通知阶段，不要重新采集。",
+        f"业务通知必须由分析端 Hermes 发送给业务目标用户（open_id 列表见下方规则），分析端绑定飞书应用 {ANALYZER_FEISHU_APP_ID}。",
+        "分析端权限策略：只读 shared-data；禁止执行任何本地 Python/PowerShell 脚本；禁止修改本地文件。",
+        "通知和多维表写入**优先**走 Python 脚本封装(LLM 不直接调 lark-cli,2026-06-18 / 2026-06-29 两次翻车后立的规矩):",
+        "  - 多维表写入:python orchestrator/scripts/insert_replenishment_bitable.py --pending-file <path> --identity user",
+        "  - Schema 2.0 卡片发送:python orchestrator/scripts/send_card_notification.py --receivers <open_id1,open_id2> --card-json '<card>'",
+        "其它场景(如读飞书表、查消息状态)可以走 lark-cli --profile analyzer。",
+        "业务通知必须以 Schema 2.0 interactive card 格式发送（msg_type=interactive，content 为 JSON 字符串），header template 按场景选 blue/green/red；不要发 plain text 或 post 富文本。",
+        "【铁律】**禁止**直接调 lark-cli 走 im +messages-send 发送卡片(2026-06-29 LLM 走 --file 把卡片 JSON 当附件传,用户收到 .json 文件而不是渲染卡片)。统一走 Python 脚本封装:",
+        "  python orchestrator/scripts/send_card_notification.py --receivers <open_id1,open_id2> --card-json '<整个 card dict 的 JSON 字符串,单引号包裹>'",
+        "  脚本会强制包成 {\"msg_type\":\"interactive\",\"card\":<dict>},物理上没法走 file 附件路径。返回最后一行 [STATS_JSON] {\"message_ids\": [...], \"errors\": [...]} 给 mark_notification_sent 用。",
+        "如果你坚持要走 execute_code + requests.post,content 也必须包成 {\"msg_type\":\"interactive\",\"card\":<dict>} 形式,不要发 post / text / file。",
+        "绝对禁止:把 card JSON 写到本地文件再用 lark-cli --file 发送;绝对禁止:把 card 写在 lark-cli im +messages-send 的 --markdown/--text/--file 参数里。",
+        "【铁律】**禁止**以任何形式(包括 execute_code/requests.post)向业务用户发送测试/校验/预览消息(2026-07-07 LLM 为校验 Schema 先 POST 一张测试卡片给用户,导致用户收到 2 张重复卡片且 workflow 被判违规)。卡片只发一次,走 send_card_notification.py,发完即止。",
+        "采集端 Hermes 不发送业务通知。",
+    ]
+    if output_patterns:
+        lines.append("只读取以下 shared-data 输出文件: " + ", ".join(output_patterns))
+
+    if workflow_name == "purchase-confirmation":
+        target_ids = ", ".join(PURCHASE_CONFIRMATION_TARGETS)
+        lines.append(f"purchase-confirmation 规则: 业务目标用户 open_id 列表: [{target_ids}]。")
+        lines.append("purchase-confirmation 规则: 使用飞书插件（receive_id_type=open_id）逐一发送采购确认通知汇总给以上每个业务目标用户。")
+        lines.append("purchase-confirmation 规则: 通知内容必须包含 SKU（商品编码）、商品名称、规格、可用数、缺口和状态（协议到货时间），禁止遗漏 SKU 列。")
+        lines.append("purchase-confirmation 规则: 业务通知发送后**必须**回调 PG(不能漏,跟 replenishment 同等级铁律):")
+        lines.append("  - `python -c \"from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer as p; p.mark_notification_sent('purchase-confirmation','<run_id>',message_ids=<从 lark 响应 message_id 列表解析>)`")
+        lines.append("  - run_id 从 pending_insert.json 或其他 collected_files 拿不到的话,用环境变量 AUTOFLOW_RUN_ID。")
+        lines.append("  - **不调这个函数,workflow 锁不释放,下一轮调度会阻塞**;请确保最后一步必调。")
+    if workflow_name == "replenishment-alert":
+        alert_target_ids = ", ".join(THRESHOLD_ALERT_TARGETS)
+        lines.append("replenishment-alert 规则: 如果 alert_count=0（或 alerts 为空），分析端不要给任何用户发送消息；只需结束（但仍要回调 mark_notification_sent,传空 message_ids,否则 PG 收不到本次 run 的\"已无告警\"状态）。")
+        lines.append(f"replenishment-alert 规则: 只有 alert_count>0 时，分析端才发送业务通知给业务目标用户（open_id: [{alert_target_ids}]）。")
+        lines.append("replenishment-alert 规则: 业务通知发送后**必须**回调 PG(不能漏):")
+        lines.append("  - `python -c \"from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer as p; p.mark_notification_sent('replenishment-alert','<run_id>',message_ids=<从 lark 响应 message_id 列表解析>)`")
+        lines.append("  - run_id 优先从 threshold_alert_data.json 读,读不到用环境变量 AUTOFLOW_RUN_ID。")
+        lines.append("  - **不调这个函数,workflow 锁不释放,下一轮调度会阻塞**;请确保最后一步必调。")
+    if workflow_name == "replenishment":
+        replenishment_target_ids = ", ".join(REPLENISHMENT_TARGETS)
+        lines.append("replenishment 规则: 只发送最终执行结果，不发送分析过程汇报。")
+        lines.append("replenishment 规则: **禁止**直接调 lark-cli 写多维表(2026-06-18 LLM 自调两次致 228 条重复;2026-06-22 v3 改为 Python 脚本封装,LLM 不再持有 lark-cli 凭据/命令)。")
+        lines.append("replenishment 规则: 多维表写入统一走 Python 脚本:`python orchestrator/scripts/insert_replenishment_bitable.py --pending-file <pending_path> --identity user`。")
+        lines.append("replenishment 规则: 脚本行为(无需你了解内部细节,只需信任其返回):")
+        lines.append("  - 读 pending_insert.json 的 records")
+        lines.append("  - 按 商品编号 调 +record-search 找最近一条,找到就 +record-upsert --record-id <old_id>(覆盖),没找到就走 +record-upsert(insert)")
+        lines.append("  - 内部用 field_id 拼 JSON,自动做了字段名->field_id 映射,你不要重复做")
+        lines.append("  - 失败非空错误时返回 nonzero exit + 把错误摘要放在 [ERROR] 日志")
+        lines.append("replenishment 规则: 脚本**只能调一次**,禁止重试/循环/分批。即便返回失败,也不要重跑脚本——后续调度会重跑整个 workflow。")
+        lines.append("replenishment 规则: 数据校验(在调脚本之前):")
+        lines.append("  - 确认 PG `bitable_insert_log` 中 (workflow='replenishment', run_id=<run_id>, version=<orchestrator指定的version>) 行存在且 status='pending'。查询: `python -c \"from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer as p; print(p.get_bitable_insert_status('replenishment','<run_id>'))\"`。如果行不存在或 status='success' 或 status='failed'(同 version),**立即停止** insert,只发业务通知并结束。")
+        lines.append("replenishment 规则: 脚本调用后(无论成功失败)必须回调 PG:")
+        lines.append("  - 抓 record_ids: 脚本最后一次输出形如 `[STATS_JSON] {\"inserted\": N, \"updated\": M, \"skipped\": K, \"record_ids\": [\"recXXX\", \"recYYY\"], \"errors\": [...]}`。**必须**从 `record_ids` 字段解析所有 record_id(可能空,空就传 [])。")
+        lines.append("  - 成功:`python -c \"from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer as p; p.complete_bitable_insert('replenishment','<run_id>',<version>,'success', bitable_record_ids=<record_ids list>)\"`")
+        lines.append("  - 失败:`p.complete_bitable_insert('replenishment','<run_id>',<version>,'failed', error=<错误摘要>)`")
+        lines.append("replenishment 规则: 业务通知发送后**必须**回调 PG(不能漏):")
+        lines.append("  - `python -c \"from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer as p; p.mark_notification_sent('replenishment','<run_id>',message_ids=<从 lark 响应 message_id 列表解析,空就传 []>,version=<version>)\"`")
+        lines.append("  - **不调这个函数,workflow 会卡死/锁不释放**;请确保最后一步必调。")
+        lines.append("replenishment 规则: 即使多维表写入失败，也必须继续向业务目标用户发送最终补货通知。")
+        lines.append("replenishment 规则: 最终通知中需明确写入失败原因与待处理数量，但不能因为写入失败而跳过通知。")
+        lines.append(f"replenishment 规则: 业务目标用户 open_id 列表: [{replenishment_target_ids}]。")
+        lines.append("replenishment 规则: 即使 pending_count=0（无新记录写入多维表），仍必须发送业务通知给业务目标用户，通知内容须包含本次识别出的补货需求（已因15天内重复而跳过的记录），不得仅以\"无需补货\"一笔带过。")
+    if workflow_name == "purchase-order-update":
+        lines.append("purchase-order-update 规则: 读取 update_result.json，如果 total_updated=0，不发送任何业务通知；只需结束。")
+        lines.append(f"purchase-order-update 规则: 如果 total_updated>0，使用飞书插件发送采购单更新通知到审单群（receive_id_type=chat_id, chat_id={SHENDAN_GROUP_CHAT_ID}），并给我发送同一条最终结果通知。")
+        lines.append("purchase-order-update 规则: 通知内容必须包含 SKU 列表、商品名称、规格颜色、到货日期和更新记录数。")
+    lines.append(
+        "HARD RULES: Do not send any process/progress/analysis updates to users. "
+        "No process/progress message is allowed. "
+        "The final message recipients are the business target users listed in the workflow rules. Send exactly ONE notification to each unique open_id."
+    )
+    lines.append(
+        "NOTIFICATION RULE: The final business notification must be sent as exactly ONE single message. "
+        "If the content exceeds Feishu's single-message length limit, condense and summarize the content "
+        "to fit within one message — do NOT split into multiple messages. "
+        "Include key summary information; detailed lists can be attached as a file or linked resource instead."
+    )
+    lines.append(
+        "When you finish, return a short execution summary in this chat response, "
+        "but do not send extra Feishu messages beyond the final business notification (including split messages). "
+        "In the summary, describe the final business notification as 'final business notification delivered/completed' "
+        "instead of words like 'process/progress notification sent' to avoid being misclassified as a violation."
+    )
+    return "\n".join(lines)
+
+
+# 哪些 json 文件的顶层 'run_id' 字段是 collector 落 PG 时用的 key(2026-06-22 v3 改造后
+# 各工作流的 collector 内部自生成 run_id,与编排层 resolved_workflow_id 是两套 ID)。
+# 编排层 finally 块要用这些 run_id 才能 ON CONFLICT 命中 collector 已 insert 的行,
+# 否则会再插一条 / 留下一条卡 running 的孤儿。
+_RUN_ID_FILE_CANDIDATES: tuple[str, ...] = (
+    "pending_insert.json",          # replenishment
+    "threshold_alert_data.json",    # replenishment-alert
+    "update_result.json",           # purchase-order-update(如有)
+    "notify_data.json",             # 部分工作流 fallback
+)
+
+
+def _extract_pending_run_id(collected_files: list[str]) -> str | None:
+    """从 collected_files 中按候选文件名顺序读取 run_id,返回 collector 落 PG 用的 key。
+
+    2026-06-22 v3 改造:batch_process.py 内部自生成 run_id 并写 PG(workflow_runs + bitable_insert_log),
+    与编排层 resolved_workflow_id 是两套 ID。下游 PG 校验/回写必须用 collector 的 run_id 才能对上。
+    全部候选都读不到或解析失败时返回 None,调用方回退到 resolved_workflow_id。
+    """
+    for name in _RUN_ID_FILE_CANDIDATES:
+        path = next((Path(p) for p in collected_files if Path(p).name == name), None)
+        if not path or not path.exists():
+            continue
+        try:
+            payload = json.loads(path.read_text(encoding="utf-8"))
+            run_id = str(payload.get("run_id") or payload.get("alert_run_id") or "").strip()
+            if run_id:
+                return run_id
+        except Exception as exc:
+            log.warning("[pg] 解析 %s 取 run_id 失败: %s", name, exc)
+    return None
+
+
+def _build_analyze_facts(workflow_name: str, collected_files: list[str]) -> list[str]:
+    """Build deterministic facts from shared-data outputs for analyzer prompts."""
+    facts: list[str] = []
+    if workflow_name != "replenishment":
+        return facts
+
+    pending_path = next((Path(p) for p in collected_files if Path(p).name == "pending_insert.json"), None)
+    notify_path = next((Path(p) for p in collected_files if Path(p).name == "notify_data.json"), None)
+
+    if pending_path and pending_path.exists():
+        try:
+            pending = json.loads(pending_path.read_text(encoding="utf-8"))
+            pending_count = int(pending.get("pending_count") or len(pending.get("records", [])) or 0)
+            skipped_count = int(pending.get("skipped_count") or 0)
+            total_cleaned = int(pending.get("total_cleaned") or 0)
+            skipped_items = pending.get("skipped_items", [])
+            pending_run_id = str(pending.get("run_id") or "")
+            pending_version = int(pending.get("bitable_version") or 1)
+            facts.append(
+                f"编排层确认: pending_insert.json 存在，pending_count={pending_count}，skipped_count={skipped_count}，total_cleaned={total_cleaned}。"
+            )
+            facts.append(
+                f"编排层确认: 本次 run_id={pending_run_id},bitable_version={pending_version}。"
+                "**请使用 orchestrator 给出的 version={pending_version} 调用 lark-cli,不要自决 version**。"
+                f"PG 中 bitable_insert_log 已 acquire(workflow=replenishment, run_id={pending_run_id}, version={pending_version}),"
+                "重复调用 lark-cli 即便能进飞书,DB 端也会被 UNIQUE 约束挡掉,且分析端应当拒绝二次调用。"
+            )
+            if skipped_items:
+                unique_skipped = list(dict.fromkeys(skipped_items))  # 去重保留顺序
+                sample = unique_skipped[:20]
+                facts.append(
+                    f"跳过的商品（skipped_items，共 {skipped_count} 条，含重复）：前20条={sample}。"
+                )
+        except Exception as exc:
+            facts.append(f"编排层确认: pending_insert.json 存在，但解析失败: {exc}")
+    else:
+        facts.append("编排层确认: pending_insert.json 不存在。")
+
+    if notify_path and notify_path.exists():
+        try:
+            notify = json.loads(notify_path.read_text(encoding="utf-8"))
+            has_replenish = bool(notify.get("has_replenish", False))
+            pending_count = int(notify.get("pending_count") or 0)
+            generated_at = str(notify.get("generated_at") or "")
+            facts.append(
+                f"编排层确认: notify_data.json 存在，has_replenish={has_replenish}，pending_count={pending_count}，generated_at={generated_at}。"
+            )
+        except Exception as exc:
+            facts.append(f"编排层确认: notify_data.json 存在，但解析失败: {exc}")
+    else:
+        facts.append("编排层确认: notify_data.json 不存在。")
+
+    facts.append("严格要求: 你的分析描述、写表数量和通知结论必须与上述编排层确认数字一致，不得自行改写。")
+    return facts
+
+
+def _is_failure(response_text: str) -> bool:
+    if not (response_text or "").strip():
+        return True
+    text = response_text.lower()
+    success_markers = ("状态: 成功", "status: success", "执行成功", "completed")
+    if any(marker.lower() in text for marker in success_markers):
+        return False
+    failure_markers = (
+        "[error]",
+        "error:",
+        "traceback",
+        "exception",
+        "failed",
+        "执行失败",
+        "运行失败",
+    )
+    return any(marker in text for marker in failure_markers)
+
+
+def _looks_like_notify_sent(response_text: str) -> bool:
+    text = response_text.lower()
+    return (
+        "业务通知" in response_text and "发送" in response_text and ("成功" in response_text or "已发送" in response_text)
+    ) or ("final result" in text and ("sent" in text or "success" in text))
+
+
+def _contains_process_notification_violation(response_text: str) -> bool:
+    """Return True when analyzer says it sent process/progress updates to users.
+
+    The analyzer is allowed to mention the rule or say it did not send progress
+    messages, so this check is intentionally conservative and line-based.
+    """
+
+    def _has_any(text: str, markers: tuple[str, ...]) -> bool:
+        return any(marker in text for marker in markers)
+
+    if not (response_text or "").strip():
+        return False
+
+    safe_markers = (
+        "no process",
+        "no progress",
+        "not sent",
+        "not send",
+        "did not send",
+        "do not send",
+        "without sending",
+        "avoid sending",
+        "not allowed",
+        "skipped notification",
+        "\u672a\u53d1\u9001",
+        "\u6ca1\u6709\u53d1\u9001",
+        "\u65e0\u9700\u53d1\u9001",
+        "\u4e0d\u53d1\u9001",
+        "\u4e0d\u8981\u53d1\u9001",
+        "\u7981\u6b62\u53d1\u9001",
+        "\u672a\u901a\u77e5",
+        "\u65e0\u8fc7\u7a0b\u901a\u77e5",
+        "\u65e0\u8fdb\u5ea6\u901a\u77e5",
+        "\u53ea\u53d1\u9001\u6700\u7ec8",
+        "未发送",
+        "没有发送",
+        "无需发送",
+        "不发送",
+        "不要发送",
+        "禁止发送",
+        "未通知",
+        "无过程通知",
+        "无进度通知",
+        "只发送最终",
+        "最终业务通知",
+        "最终通知",
+        "业务通知已发送",
+        "通知已投递",
+        "已发送最终",
+        "已投递",
+    )
+    direct_violation_markers = (
+        "process notification sent",
+        "progress notification sent",
+        "process update sent",
+        "progress update sent",
+        "sent process notification",
+        "sent progress notification",
+        "sent progress update",
+        "sent status update",
+        "\u53d1\u9001\u4e86\u8fc7\u7a0b\u901a\u77e5",
+        "\u53d1\u9001\u4e86\u8fdb\u5ea6\u901a\u77e5",
+        "\u5df2\u53d1\u9001\u8fc7\u7a0b\u901a\u77e5",
+        "\u5df2\u53d1\u9001\u8fdb\u5ea6\u901a\u77e5",
+        "\u8fc7\u7a0b\u901a\u77e5\u5df2\u53d1\u9001",
+        "\u8fdb\u5ea6\u901a\u77e5\u5df2\u53d1\u9001",
+        "\u63a8\u9001\u4e86\u8fc7\u7a0b\u901a\u77e5",
+        "\u63a8\u9001\u4e86\u8fdb\u5ea6\u901a\u77e5",
+        "发送了过程通知",
+        "发送了进度通知",
+        "已发送过程通知",
+        "已发送进度通知",
+        "过程通知已发送",
+        "进度通知已发送",
+        "推送了过程通知",
+        "推送了进度通知",
+    )
+    process_markers = (
+        "process",
+        "progress",
+        "status update",
+        "analysis update",
+        "intermediate update",
+        "\u8fc7\u7a0b",
+        "\u8fdb\u5ea6",
+        "\u72b6\u6001\u66f4\u65b0",
+        "\u5206\u6790\u66f4\u65b0",
+        "\u4e2d\u95f4\u66f4\u65b0",
+        "\u8fc7\u7a0b\u6c47\u62a5",
+        "\u8fdb\u5ea6\u6c47\u62a5",
+        "过程",
+        "进度",
+        "状态更新",
+        "分析更新",
+        "中间更新",
+        "过程汇报",
+        "进度汇报",
+    )
+    send_markers = (
+        "sent",
+        "send",
+        "delivered",
+        "notified",
+        "message",
+        "\u53d1\u9001",
+        "\u901a\u77e5",
+        "\u63a8\u9001",
+        "\u6d88\u606f",
+        "发送",
+        "通知",
+        "推送",
+        "消息",
+    )
+    recipient_markers = (
+        "user",
+        "recipient",
+        "business",
+        "target",
+        "open_id",
+        "chat_id",
+        "\u7528\u6237",
+        "\u63a5\u6536\u4eba",
+        "\u4e1a\u52a1",
+        "\u76ee\u6807",
+        "用户",
+        "接收人",
+        "业务",
+        "目标",
+    )
+
+    for raw_line in response_text.splitlines():
+        line = " ".join(raw_line.lower().split())
+        if not line:
+            continue
+        if _has_any(line, safe_markers):
+            continue
+        if _has_any(line, direct_violation_markers):
+            return True
+        if (
+            _has_any(line, process_markers)
+            and _has_any(line, send_markers)
+            and _has_any(line, recipient_markers)
+        ):
+            return True
+    return False
+
+
+def call_hermes(gateway_name: str, prompt: str) -> str:
+    """Call one Hermes OpenAI-compatible gateway and return assistant text."""
+
+    gateway = HERMES_GATEWAYS[gateway_name]
+    url = gateway["url"].rstrip("/") + "/chat/completions"
+    headers = {
+        "Authorization": f"Bearer {gateway['token']}",
+        "Content-Type": "application/json",
+    }
+    payload: dict[str, Any] = {
+        "model": gateway["model"],
+        "messages": [
+            {
+                "role": "system",
+                "content": (
+                    f"You are the {gateway['profile']} Hermes workflow agent. "
+                    "Follow the user's workflow instruction exactly."
+                ),
+            },
+            {"role": "user", "content": prompt},
+        ],
+    }
+
+    try:
+        session = requests.Session()
+        session.trust_env = False
+        # connect timeout 10s, read timeout 1800s (match CLI timeout).
+        # Previously timeout=30 caused read timeouts while the server-side
+        # agent was still executing; the except-all fallback to CLI then
+        # started a SECOND agent, resulting in duplicate bitable inserts.
+        response = session.post(url, headers=headers, json=payload, timeout=(10, 1800))
+        response.raise_for_status()
+        data = response.json()
+        return data["choices"][0]["message"]["content"]
+    except requests.exceptions.ConnectionError:
+        return call_hermes_cli(gateway["profile"], prompt)
+
+
+def call_hermes_cli(profile: str, prompt: str) -> str:
+    """Call Hermes profile through the one-shot CLI when HTTP gateway is unavailable."""
+
+    result = subprocess.run(
+        ["hermes", "-p", profile, "-z", prompt],
+        cwd=str(Path(__file__).resolve().parents[1]),
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+        timeout=1800,
+    )
+    output = "\n".join(part for part in [result.stdout.strip(), result.stderr.strip()] if part)
+    if result.returncode != 0:
+        raise RuntimeError(output or f"Hermes CLI failed with exit code {result.returncode}")
+    return output
+
+
+def _call_once(gateway_name: str, prompt: str) -> tuple[bool, int, str]:
+    """单次调用，不重试。"""
+    try:
+        response_text = call_hermes(gateway_name, prompt)
+    except Exception as exc:
+        response_text = f"[ERROR] Hermes call failed: {exc}"
+    return (not _is_failure(response_text)), 1, response_text
+
+
+def _clear_shared_outputs(workflow_name: str) -> list[str]:
+    """Retain raw artifacts; run-scoped directories prevent stale collisions."""
+
+    return []
+
+
+def _matching_shared_outputs(
+    workflow_name: str,
+    workflow_id: str,
+    since_epoch: float | None = None,
+) -> list[Path]:
+    workflow_dir = Path(SHARED_DIR) / workflow_name / f"run_id={workflow_id}"
+    patterns = WORKFLOWS[workflow_name].get("output_patterns", [])
+    if not workflow_dir.exists() or not patterns:
+        return []
+
+    matches: list[Path] = []
+    seen: set[Path] = set()
+    for pattern in patterns:
+        for path in workflow_dir.rglob(pattern):
+            if (
+                path.is_file()
+                and path not in seen
+                and (since_epoch is None or path.stat().st_mtime >= since_epoch)
+            ):
+                matches.append(path)
+                seen.add(path)
+    return sorted(matches, key=lambda path: path.stat().st_mtime)
+
+
+def _run_local_collection_script(workflow_name: str, workflow_id: str) -> tuple[bool, str]:
+    """Run the configured collection script directly from the orchestrator."""
+
+    workflow = WORKFLOWS[workflow_name]
+    script_path = Path(workflow["collector_script"]).resolve()
+    if not script_path.exists():
+        return False, f"Collection script not found: {script_path}"
+
+    if script_path.suffix.lower() == ".ps1":
+        command = [
+            "powershell",
+            "-NoProfile",
+            "-ExecutionPolicy",
+            "Bypass",
+            "-File",
+            str(script_path),
+        ]
+    else:
+        command = [str(script_path)]
+
+    env = os.environ.copy()
+    env["AUTOFLOW_RUN_ID"] = workflow_id
+    env["AUTOFLOW_WORKFLOW_ID"] = workflow_id
+    env["AUTOFLOW_TARGET_WORKFLOW"] = workflow_name
+    try:
+        result = subprocess.run(
+            command,
+            cwd=str(script_path.parent),
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+            timeout=OUTPUT_WAIT_SECONDS,
+            env=env,
+        )
+    except subprocess.TimeoutExpired as exc:
+        output = "\n".join(part for part in [exc.stdout, exc.stderr] if part)
+        return False, f"Collection script timed out after {OUTPUT_WAIT_SECONDS}s.\n{output}".strip()
+    except Exception as exc:
+        return False, f"Collection script failed to start: {exc}"
+
+    output = "\n".join(part.strip() for part in [result.stdout, result.stderr] if part and part.strip())
+    if result.returncode != 0:
+        return False, output or f"Collection script failed with exit code {result.returncode}"
+    return True, output or "Collection script completed successfully."
+
+
+def _ensure_shared_outputs(
+    workflow_name: str, since_epoch: float, workflow_id: str
+) -> list[str]:
+    patterns = WORKFLOWS[workflow_name].get("output_patterns", [])
+    if not patterns:
+        return []
+
+    deadline = time.time() + OUTPUT_WAIT_SECONDS
+    while True:
+        outputs = _matching_shared_outputs(workflow_name, workflow_id, since_epoch)
+        if outputs:
+            return [str(path) for path in outputs]
+        if time.time() >= deadline:
+            return []
+        time.sleep(OUTPUT_POLL_SECONDS)
+
+
+def _cleanup_workflow_artifacts(workflow_name: str) -> list[str]:
+    """Best-effort cleanup of transient work files; raw artifacts are immutable."""
+
+    removed: list[str] = []
+
+    data_dir = WORK_ROOT
+    if data_dir.exists():
+        for path in data_dir.glob("*"):
+            if path.is_file():
+                try:
+                    path.unlink()
+                    removed.append(str(path))
+                except Exception:
+                    pass
+    return removed
+
+
+def _verify_bitable_insert_in_pg(workflow_name: str, workflow_id: str, analyze_response: str) -> dict:
+    """核对 analyzer 是否通过 PG 闸门完成多维表写入(2026-06-22 v2,取代旧 _bitable_written.json 文件 marker)。
+
+    返回:
+      - acquired: bitable_insert_log 是否存在本次 run 的占位行
+      - status: 'success' | 'failed' | 'pending' | 'no_row' | 'n/a'
+      - version / record_count / data_hash / bitable_record_ids / error
+      - warning: str | None  (执行总结里要体现的人话说明)
+    """
+    if workflow_name != "replenishment":
+        return {"acquired": False, "status": "n/a"}
+    try:
+        from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer as pg
+    except Exception as exc:
+        return {"acquired": False, "status": "n/a", "warning": f"pg_writer 不可用: {exc}"}
+    row = pg.get_bitable_insert_status(workflow_name, workflow_id)
+    # v3 改造后 LLM 不再调 +record-batch-create,改调 Python 脚本,所以 claims 关键字要覆盖两边。
+    _CLAIM_KEYWORDS = (
+        "+record-batch-create", "+record-upsert",
+        "多维表写入", "insert_replenishment_bitable",
+        "调 Python 脚本", "调了 insert",
+    )
+    if not row:
+        claims_insert = analyze_response and any(k in analyze_response for k in _CLAIM_KEYWORDS)
+        warning = (
+            "analyzer_response 提到多维表写入/调 Python 脚本,但 bitable_insert_log 中无本次 run 的占位行 —— "
+            "LLM 可能未走 PG 闸门,需要人工核对飞书多维表和 DB。"
+            if claims_insert
+            else None
+        )
+        return {"acquired": False, "status": "no_row", "warning": warning}
+    out = {
+        "acquired": True,
+        "status": row["status"],
+        "version": row["version"],
+        "record_count": row["record_count"],
+        "data_hash": row["data_hash"],
+        "bitable_record_ids": row["bitable_record_ids"],
+        "error": row["error"],
+        "warning": None,
+    }
+    if row["status"] == "pending":
+        claims_insert = analyze_response and any(k in analyze_response for k in _CLAIM_KEYWORDS)
+        if claims_insert:
+            out["warning"] = (
+                "analyzer 自称写过 bitable,但 PG bitable_insert_log 仍为 pending —— "
+                "LLM 可能没有回调 complete_bitable_insert。需要人工核对飞书多维表。"
+            )
+    return out
+
+
+
+def _manifest_path(workflow_name: str, workflow_id: str) -> Path:
+    return (
+        Path(SHARED_DIR)
+        / workflow_name
+        / f"run_id={workflow_id}"
+        / "_run_manifest.json"
+    )
+
+
+def _validate_run_manifest(workflow_name: str, workflow_id: str, collected_files: list[str]) -> tuple[bool, str]:
+    manifest = _manifest_path(workflow_name, workflow_id)
+    if not manifest.exists():
+        return False, f"Missing run manifest: {manifest}"
+    try:
+        # PowerShell Set-Content UTF8 may emit BOM; utf-8-sig handles both BOM/non-BOM files.
+        payload = json.loads(manifest.read_text(encoding="utf-8-sig"))
+    except Exception as exc:
+        return False, f"Invalid run manifest: {exc}"
+
+    run_id = str(payload.get("run_id") or "")
+    if run_id != workflow_id:
+        return False, f"Run ID mismatch in manifest. expected={workflow_id}, actual={run_id}"
+
+    expected_files = {
+        str(manifest.parent / name) for name in (payload.get("files") or [])
+    }
+    actual_files = {str(Path(path)) for path in collected_files}
+    if expected_files and not expected_files.issubset(actual_files):
+        return False, "Collected files do not match manifest file list."
+    return True, ""
+
+
+def _append_heartbeat_log(result: dict, entry: str) -> None:
+    existing = str(result.get("heartbeat_response") or "").strip()
+    if existing:
+        result["heartbeat_response"] = existing + "\n\n---\n" + entry
+    else:
+        result["heartbeat_response"] = entry
+
+
+def _read_alert_count(collected_files: list[str]) -> int | None:
+    path = next((Path(p) for p in collected_files if Path(p).name == "threshold_alert_data.json"), None)
+    if not path or not path.exists():
+        return None
+    try:
+        payload = json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return None
+    try:
+        return int(payload.get("alert_count", len(payload.get("alerts", []) or [])))
+    except Exception:
+        return None
+
+
+def _read_update_count(collected_files: list[str]) -> int | None:
+    path = next((Path(p) for p in collected_files if Path(p).name == "update_result.json"), None)
+    if not path or not path.exists():
+        return None
+    try:
+        payload = json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return None
+    try:
+        return int(payload.get("total_updated", 0))
+    except Exception:
+        return None
+
+
+def _send_collector_heartbeat(result: dict, phase: str = "final") -> None:
+    notes: list[str] = []
+    try:
+        token = _get_collector_tenant_token()
+        card = _build_heartbeat_card(result, phase=phase)
+        resp = requests.post(
+            f"{FEISHU_OPENAPI_BASE}/im/v1/messages",
+            headers={"Authorization": f"Bearer {token}", "Content-Type": "application/json"},
+            params={"receive_id_type": "open_id"},
+            json={
+                "receive_id": FEISHU_CONFIG["apps"]["collector"]["owner_open_id"],
+                "msg_type": "interactive",
+                "content": json.dumps(card, ensure_ascii=False),
+            },
+            timeout=10,
+        )
+        resp.raise_for_status()
+        body = resp.json()
+        if body.get("code") != 0:
+            raise RuntimeError(
+                f"feishu im send non-zero code: {body.get('code')} msg={body.get('msg')}"
+            )
+        msg_id = body.get("data", {}).get("message_id", "unknown")
+        notes.append(f"feishu_msg_id={msg_id}")
+    except Exception as exc:
+        notes.append(f"[WARN] collector heartbeat via Feishu OpenAPI failed: {exc}")
+    _append_heartbeat_log(result, "\n".join(notes))
+
+
+def _start_progress_heartbeat_worker(result: dict) -> tuple[threading.Event, threading.Thread] | None:
+    if HEARTBEAT_INTERVAL_SECONDS <= 0:
+        return None
+    stop_event = threading.Event()
+
+    def _worker() -> None:
+        while not stop_event.wait(HEARTBEAT_INTERVAL_SECONDS):
+            if result.get("status") in {"completed", "failed"}:
+                return
+            _send_collector_heartbeat(result, phase="progress")
+
+    thread = threading.Thread(
+        target=_worker,
+        name=f"heartbeat-{result.get('workflow_id', 'workflow')}",
+        daemon=True,
+    )
+    thread.start()
+    return stop_event, thread
+
+
+def run_mcp_workflow(workflow_name: str, workflow_id: str | None = None) -> dict:
+    """Run one workflow through Hermes collector/analyzer gateways."""
+
+    if workflow_name not in WORKFLOWS:
+        return {"workflow": workflow_name, "status": "failed", "error": "Unknown workflow"}
+
+    resolved_workflow_id = workflow_id or (
+        f"{workflow_name}_{now_iso().replace(':', '').replace('-', '')}_{uuid.uuid4().hex[:6]}"
+    )
+    lock_ok, lock_error = _acquire_workflow_lock(workflow_name, resolved_workflow_id)
+    if not lock_ok:
+        return {
+            "workflow": workflow_name,
+            "workflow_name": workflow_display_name(workflow_name),
+            "workflow_id": resolved_workflow_id,
+            "status": "failed",
+            "error": lock_error,
+        }
+    # 2026-06-22 v3:是否已经写过 workflow_runs(采集端在成功时 upsert_run('running'))。
+    # finally 块用它决定要不要再调一次 upsert_run 把终态落库。
+    # 没写过的(采集前就失败的)就不调,避免空行/假数据。
+    workflow_run_inserted = False
+
+    result = {
+        "workflow": workflow_name,
+        "workflow_name": workflow_display_name(workflow_name),
+        "workflow_id": resolved_workflow_id,
+        "status": "running",
+        "trigger_time": now_iso(),
+        "end_time": None,
+        "collect_attempts": 0,
+        "analyze_attempts": 0,
+        "collector_response": "",
+        "analyzer_response": "",
+        "heartbeat_response": "",
+        "cleared_shared_outputs": [],
+        "cleaned_artifacts": [],
+        "collected_files": [],
+        "error": "",
+        "current_phase": "starting",
+    }
+    _send_collector_heartbeat(result, phase="start")
+    heartbeat_worker = _start_progress_heartbeat_worker(result)
+
+    try:
+        result["current_phase"] = "collecting"
+        collection_lock_ok, collection_lock_error = _acquire_collection_lock(workflow_name, resolved_workflow_id)
+        if not collection_lock_ok:
+            result["status"] = "failed"
+            result["error"] = collection_lock_error
+            result["end_time"] = now_iso()
+            _send_collector_heartbeat(result)
+            return result
+
+        try:
+            collect_started_at = time.time()
+            result["cleared_shared_outputs"] = _clear_shared_outputs(workflow_name)
+            collect_ok, collect_response = _run_local_collection_script(workflow_name, resolved_workflow_id)
+            collect_attempts = 1
+            result["collect_attempts"] = collect_attempts
+            result["collector_response"] = collect_response
+            if not collect_ok:
+                result["status"] = "failed"
+                result["error"] = collect_response
+                result["end_time"] = now_iso()
+                _send_collector_heartbeat(result)
+                return result
+            # 采集端成功 → 已经 upsert_run('running') 写过 workflow_runs,
+            # finally 块要把终态再写一次。
+            workflow_run_inserted = True
+
+            result["current_phase"] = "waiting_outputs"
+            result["collected_files"] = _ensure_shared_outputs(
+                workflow_name, collect_started_at, resolved_workflow_id
+            )
+            if WORKFLOWS[workflow_name].get("output_patterns") and not result["collected_files"]:
+                result["status"] = "failed"
+                result["error"] = "Collection finished without expected shared-data outputs."
+                result["end_time"] = now_iso()
+                _send_collector_heartbeat(result)
+                return result
+            manifest_ok, manifest_error = _validate_run_manifest(
+                workflow_name,
+                resolved_workflow_id,
+                result["collected_files"],
+            )
+            if not manifest_ok:
+                result["status"] = "failed"
+                result["error"] = manifest_error
+                result["end_time"] = now_iso()
+                _send_collector_heartbeat(result)
+                return result
+        finally:
+            _release_collection_lock(workflow_name)
+
+        # 2026-06-22 v3: collector 落 PG 时用的是自己生成的 run_id(replenishment 流程),
+        # 与编排层 resolved_workflow_id 是两套 ID。一次性解析出 bitable run_id,
+        # 后续 PG 校验 / workflow_runs 回写都用它,无 analyzer 的流程也安全(回退到 resolved_workflow_id)。
+        bitable_run_id = _extract_pending_run_id(result["collected_files"]) or resolved_workflow_id
+
+        if WORKFLOWS[workflow_name].get("analyzer_scripts"):
+            if workflow_name == "replenishment-alert":
+                alert_count = _read_alert_count(result["collected_files"])
+                if alert_count == 0:
+                    result["current_phase"] = "finishing"
+                    result["analyze_attempts"] = 0
+                    result["analyzer_response"] = "alert_count=0, skip analyzer notification."
+                    result["status"] = "completed"
+                    result["end_time"] = now_iso()
+                    _send_collector_heartbeat(result)
+                    return result
+            if workflow_name == "purchase-order-update":
+                update_count = _read_update_count(result["collected_files"])
+                if update_count == 0:
+                    result["current_phase"] = "finishing"
+                    result["analyze_attempts"] = 0
+                    result["analyzer_response"] = "total_updated=0, skip analyzer notification."
+                    result["status"] = "completed"
+                    result["end_time"] = now_iso()
+                    _send_collector_heartbeat(result)
+                    return result
+            result["current_phase"] = "analyzing"
+            analyze_prompt = _build_analyze_prompt(workflow_name)
+            analyze_facts = _build_analyze_facts(workflow_name, result["collected_files"])
+            if analyze_facts:
+                analyze_prompt = analyze_prompt + "\n\n" + "\n".join(analyze_facts)
+            analyze_ok, analyze_attempts, analyze_response = _call_once("analyzer", analyze_prompt)
+            if workflow_name == "replenishment" and not analyze_ok and _looks_like_notify_sent(analyze_response):
+                analyze_ok = True
+            result["analyze_attempts"] = analyze_attempts
+            result["analyzer_response"] = analyze_response
+            if workflow_name == "replenishment":
+                result["bitable_marker"] = _verify_bitable_insert_in_pg(
+                    workflow_name, bitable_run_id, analyze_response
+                )
+            # ????????/???????????????????????
+            if _contains_process_notification_violation(analyze_response):
+                # If notification was already delivered to PG, downgrade to warning
+                # instead of hard fail — the business outcome is still achieved.
+                try:
+                    from gyxx_flow.modules.supply_chain.runtime.orchestrator import (
+                        pg_writer as _pgw,
+                    )
+                    _notif_ok = _pgw.notification_already_sent(workflow_name, bitable_run_id)
+                except Exception:
+                    _notif_ok = False
+                if _notif_ok:
+                    result["status"] = "completed"
+                    result["warning"] = (
+                        "Analyzer violated notification policy (process/progress notification), "
+                        "but final business notification was already delivered — downgraded to warning."
+                    )
+                    log.warning(
+                        "[%s] notification policy violation detected but notification already sent — "
+                        "downgrading to warning", workflow_name
+                    )
+                else:
+                    result["status"] = "failed"
+                    result["error"] = (
+                        "Analyzer violated notification policy: process/progress notification "
+                        "to users is not allowed. Only one final business notification is allowed."
+                    )
+                    result["end_time"] = now_iso()
+                    _send_collector_heartbeat(result)
+                    return result
+            if not analyze_ok:
+                result["status"] = "failed"
+                result["error"] = analyze_response
+                result["end_time"] = now_iso()
+                _send_collector_heartbeat(result)
+                return result
+
+        result["current_phase"] = "finishing"
+        result["status"] = "completed"
+        result["end_time"] = now_iso()
+        # 2026-06-22 v3:workflow_runs 终态由 finally 块统一回写(覆盖 success + 所有
+        # early return 路径),这里只发心跳。
+        _send_collector_heartbeat(result)
+        return result
+    finally:
+        if heartbeat_worker is not None:
+            stop_event, thread = heartbeat_worker
+            stop_event.set()
+            thread.join(timeout=5)
+        result["cleaned_artifacts"] = _cleanup_workflow_artifacts(workflow_name)
+        _release_workflow_lock(workflow_name)
+        # 2026-06-22 v3:无论如何都把终态回写 workflow_runs。
+        # 之前只在 success path 调,导致 early return(analyze 失败/notification 违规/空集等)
+        # 路径上 workflow_runs 永远卡在 'running'。
+        if workflow_run_inserted:
+            try:
+                from gyxx_flow.modules.supply_chain.runtime.orchestrator import (
+                    pg_writer as pg,
+                )
+                final_status = str(result.get("status") or "failed")
+                end_time_iso = str(result.get("end_time") or now_iso())
+                pg.upsert_run(
+                    run_id=bitable_run_id or resolved_workflow_id,
+                    workflow=workflow_name,
+                    trigger_time=str(result.get("trigger_time") or ""),
+                    end_time=end_time_iso,
+                    status=final_status,
+                    error=str(result.get("error") or ""),
+                    source_files=list(result.get("collected_files") or []),
+                    raw_payload={"analyze_attempts": result.get("analyze_attempts"),
+                                 "collect_attempts": result.get("collect_attempts"),
+                                 "bitable_marker": result.get("bitable_marker"),
+                                 "warning": result.get("warning")},
+                )
+            except Exception as exc:
+                log.warning("[pg] final upsert_run (finally) failed (non-fatal): %s", exc)
+
+
+def run_mcp_all() -> dict:
+    """Run all configured business workflows through Hermes MCP gateways."""
+
+    workflow_names = batch_workflow_names()
+    return {
+        "mode": "mcp-run",
+        "results": [run_mcp_workflow(name) for name in workflow_names],
+    }
+
+
+if __name__ == "__main__":
+    print(json.dumps(run_mcp_all(), ensure_ascii=False, indent=2))
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/monitor.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/monitor.py
new file mode 100644
index 0000000..3a97e28
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/monitor.py
@@ -0,0 +1,96 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""Monitor for new collected data and trigger analysis."""
+
+from __future__ import annotations
+
+import json
+import time
+from datetime import datetime
+from pathlib import Path
+
+from gyxx_flow.modules.supply_chain.runtime.paths import STATE_ROOT
+
+STATUS_FILE = STATE_ROOT / "status.json"
+PROCESSING_FILE = STATE_ROOT / "processing_status.json"
+
+
+def _now() -> str:
+    return datetime.now().isoformat(timespec="seconds")
+
+
+def read_json(path: Path, default: dict) -> dict:
+    if not path.exists():
+        return default
+    try:
+        return json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return default
+
+
+def get_collect_status() -> dict:
+    return read_json(STATUS_FILE, {"collections": {}})
+
+
+def get_processing_status() -> dict:
+    return read_json(PROCESSING_FILE, {"processed": {}})
+
+
+def has_new_data(workflow: str) -> bool:
+    """Check if there's new data that hasn't been processed yet."""
+    collect_status = get_collect_status()
+    processing_status = get_processing_status()
+
+    collect_entry = collect_status.get("collections", {}).get(workflow, {})
+    processing_key = f"{workflow}/"
+    processing_entries = {k: v for k, v in processing_status.get("processed", {}).items() if k.startswith(processing_key)}
+
+    if not collect_entry:
+        return False
+
+    if collect_entry.get("status") != "success":
+        return False
+
+    # Check if any collected file is newer than processed
+    collect_time = collect_entry.get("timestamp", "")
+    if not collect_time:
+        return False
+
+    if not processing_entries:
+        return True
+
+    latest_process_time = max(v.get("timestamp", "") for v in processing_entries.values())
+    return collect_time > latest_process_time
+
+
+def get_new_workflows() -> list[str]:
+    """Get list of workflows with new data."""
+    workflows = ["purchase-confirmation", "replenishment", "replenishment-alert"]
+    return [wf for wf in workflows if has_new_data(wf)]
+
+
+def monitor_loop(interval: int = 60):
+    """Monitor for new data and print status."""
+    print(f"Monitoring started at {_now()}")
+    print(f"Checking every {interval} seconds...")
+    print("-" * 50)
+
+    while True:
+        new_workflows = get_new_workflows()
+
+        if new_workflows:
+            print(f"\n[{_now()}] New data detected:")
+            for wf in new_workflows:
+                print(f"  - {wf}")
+            print("\nRun analysis with:")
+            print("  python -m orchestrator.runner run all --analyze-only")
+        else:
+            print(f"[{_now()}] No new data")
+
+        time.sleep(interval)
+
+
+if __name__ == "__main__":
+    import sys
+    interval = int(sys.argv[1]) if len(sys.argv) > 1 else 60
+    monitor_loop(interval)
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/pg_writer.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/pg_writer.py
new file mode 100644
index 0000000..c6949c8
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/pg_writer.py
@@ -0,0 +1,720 @@
+# -*- coding: utf-8 -*-
+# ruff: noqa: E402
+"""
+PostgreSQL writer for auto-flow workflow persistence.
+
+设计原则:
+  1) 所有 pg 写操作"失败只 warn,不阻塞业务流程"
+     - 采集脚本已经在 shared-data/ 落了 JSON,pg 是次级持久层
+  2) 幂等: workflow_runs.run_id 是主键,fact_* 表 ON CONFLICT DO NOTHING
+  3) 配置走环境变量,密码不入库不入 git:
+       AUTOFLOW_PG_HOST (default 8.148.185.119)
+       AUTOFLOW_PG_PORT (default 5432)
+       AUTOFLOW_PG_USER (default data_hub)
+       AUTOFLOW_PG_PASSWORD (无默认,必须设)
+       AUTOFLOW_PG_DB   (default data_hub)
+
+  项目根目录的 .env 会作为本机配置加载（已加入 .gitignore）；系统环境变量优先。
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import os
+from contextlib import contextmanager
+from typing import Any, Iterable
+
+import psycopg2
+from psycopg2.extras import execute_values
+from gyxx_flow.adapters import RuntimeServicePolicy
+
+log = logging.getLogger("autoflow.pg")
+
+
+def _load_local_env() -> None:
+    """Load project-local secrets without overriding process environment variables."""
+    env_path = os.path.join(os.path.dirname(os.path.dirname(__file__)), ".env")
+    try:
+        with open(env_path, "r", encoding="utf-8") as env_file:
+            for raw_line in env_file:
+                line = raw_line.strip()
+                if not line or line.startswith("#") or "=" not in line:
+                    continue
+                key, value = line.split("=", 1)
+                key = key.strip()
+                value = value.strip().strip('"').strip("'")
+                if key:
+                    os.environ.setdefault(key, value)
+    except FileNotFoundError:
+        pass
+
+
+_load_local_env()
+os.environ.update(RuntimeServicePolicy().apply(os.environ))
+
+PG_CONFIG = {
+    "host": os.getenv("AUTOFLOW_PG_HOST", ""),
+    "port": int(os.getenv("AUTOFLOW_PG_PORT", "5432")),
+    "user": os.getenv("AUTOFLOW_PG_USER", ""),
+    "password": os.getenv("AUTOFLOW_PG_PASSWORD", ""),
+    "dbname": os.getenv("AUTOFLOW_PG_DB", ""),
+    "connect_timeout": 5,
+}
+
+
+def _conn():
+    return psycopg2.connect(**PG_CONFIG)
+
+
+@contextmanager
+def _cursor():
+    """带 commit/rollback 的 cursor 上下文。失败仅 warn,不抛。"""
+    conn = None
+    try:
+        conn = _conn()
+        with conn.cursor() as cur:
+            yield cur
+        conn.commit()
+    except Exception as e:
+        if conn is not None:
+            try:
+                conn.rollback()
+            except Exception:
+                pass
+        log.warning("[pg] write failed (non-fatal): %s", e)
+    finally:
+        if conn is not None:
+            try:
+                conn.close()
+            except Exception:
+                pass
+
+
+def is_available() -> bool:
+    """快速可达性检查,用于采集脚本开头判断是否启用持久化。"""
+    try:
+        conn = _conn()
+        conn.close()
+        return True
+    except Exception as e:
+        log.warning("[pg] not available: %s", e)
+        return False
+
+
+# --- workflow_runs ---------------------------------------------------------
+
+def upsert_run(
+    run_id: str,
+    workflow: str,
+    trigger_time: str,
+    *,
+    end_time: str | None = None,
+    status: str = "completed",
+    error: str = "",
+    source_files: list[str] | None = None,
+    raw_payload: dict | None = None,
+) -> None:
+    sql = """
+    INSERT INTO workflow_runs
+        (run_id, workflow, trigger_time, end_time, status, error, source_files, raw_payload)
+    VALUES (%s, %s, %s, %s, %s, %s, %s::jsonb, %s::jsonb)
+    ON CONFLICT (run_id) DO UPDATE SET
+        end_time     = EXCLUDED.end_time,
+        status       = EXCLUDED.status,
+        error        = EXCLUDED.error,
+        source_files = EXCLUDED.source_files,
+        raw_payload  = EXCLUDED.raw_payload
+    """
+    with _cursor() as cur:
+        cur.execute(
+            sql,
+            (
+                run_id,
+                workflow,
+                trigger_time,
+                end_time,
+                status,
+                error,
+                json.dumps(source_files or [], ensure_ascii=False),
+                json.dumps(raw_payload or {}, ensure_ascii=False, default=str),
+            ),
+        )
+
+
+# --- dim_product -----------------------------------------------------------
+
+def upsert_dim_products(skus: Iterable[str], *, is_discontinued: bool) -> int:
+    """upsert 一批 SKU,返回实际写入行数。失败返回 0。"""
+    skus = [str(s).strip() for s in skus if s and str(s).strip()]
+    if not skus:
+        return 0
+    sql = """
+    INSERT INTO dim_product (sku, is_discontinued)
+    VALUES %s
+    ON CONFLICT (sku) DO UPDATE SET
+        is_discontinued = EXCLUDED.is_discontinued OR dim_product.is_discontinued,
+        last_seen_at    = now()
+    """
+    rows = [(s, is_discontinued) for s in skus]
+    try:
+        conn = _conn()
+        try:
+            with conn.cursor() as cur:
+                execute_values(cur, sql, rows)
+            conn.commit()
+            return len(rows)
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] upsert_dim_products failed: %s", e)
+        return 0
+
+
+# --- fact_purchase_confirmation -------------------------------------------
+
+def insert_purchase_confirmation(run_id: str, rows: list[dict[str, Any]]) -> int:
+    if not rows:
+        return 0
+    sql = """
+    INSERT INTO fact_purchase_confirmation
+        (run_id, sku, product_name, spec, available_stock, sales_7d,
+         purchase_in_transit, gap, delivery_date, summary_raw)
+    VALUES %s
+    ON CONFLICT (run_id, sku) DO NOTHING
+    """
+    values = [
+        (
+            run_id,
+            str(r.get("sku") or r.get("商品编码") or "").strip().lstrip("'"),
+            r.get("product_name") or r.get("清洗后的名称"),
+            r.get("spec") or r.get("清洗后的规格"),
+            int(r.get("available_stock") or r.get("可用数") or 0),
+            int(r.get("sales_7d") or r.get("7天销量") or 0),
+            int(r.get("purchase_in_transit") or r.get("采购在途数") or 0),
+            int(r.get("gap") or r.get("缺口") or 0),
+            r.get("delivery_date") or r.get("协议到货时间") or "",
+            r.get("summary_raw") or r.get("汇总原始数据") or "",
+        )
+        for r in rows
+    ]
+    skus = [v[1] for v in values if v[1]]
+    if not values:
+        return 0
+    try:
+        conn = _conn()
+        try:
+            existing = _filter_existing(conn, "fact_purchase_confirmation", run_id, skus)
+            filtered = [v for v in values if v[1] not in existing]
+            if not filtered:
+                return 0
+            with conn.cursor() as cur:
+                execute_values(cur, sql, filtered)
+            conn.commit()
+            return len(filtered)
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] insert_purchase_confirmation failed: %s", e)
+        return 0
+
+
+# --- fact_threshold_alert -------------------------------------------------
+
+def insert_threshold_alerts(run_id: str, rows: list[dict[str, Any]]) -> int:
+    if not rows:
+        return 0
+    sql = """
+    INSERT INTO fact_threshold_alert
+        (run_id, sku, product_name, product_spec, threshold, actual_stock,
+         purchase_all, sales_7, sales_15, sales_30, generated_at)
+    VALUES %s
+    ON CONFLICT (run_id, sku) DO NOTHING
+    """
+    values = [
+        (
+            run_id,
+            str(r.get("sku") or "").strip(),
+            r.get("product_name") or r.get("name", ""),
+            r.get("product_spec") or r.get("props", ""),
+            r.get("threshold"),
+            r.get("actual_stock") or r.get("qty_stock"),
+            r.get("purchase_all"),
+            r.get("sales_7"),
+            r.get("sales_15"),
+            r.get("sales_30"),
+            r.get("generated_at"),
+        )
+        for r in rows
+    ]
+    skus = [v[1] for v in values if v[1]]
+    if not values:
+        return 0
+    try:
+        conn = _conn()
+        try:
+            existing = _filter_existing(conn, "fact_threshold_alert", run_id, skus)
+            filtered = [v for v in values if v[1] not in existing]
+            if not filtered:
+                return 0
+            with conn.cursor() as cur:
+                execute_values(cur, sql, filtered)
+            conn.commit()
+            return len(filtered)
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] insert_threshold_alerts failed: %s", e)
+        return 0
+
+
+# --- fact_replenishment ---------------------------------------------------
+
+def insert_replenishments(run_id: str, rows: list[dict[str, Any]]) -> int:
+    if not rows:
+        return 0
+    sql = """
+    INSERT INTO fact_replenishment
+        (run_id, sku, product_name, display_name, spec_new, spec_old,
+         sales_7d, sales_15d, sales_30d, sales_60d, sales_90d, monthly_sales,
+         available_stock, purchase_in_transit, suggest_qty, bot_remark, write_status)
+    VALUES %s
+    ON CONFLICT (run_id, sku) DO NOTHING
+    """
+    values = [
+        (
+            run_id,
+            str(r.get("商品编号") or r.get("sku") or "").strip().lstrip("'"),
+            r.get("商品名称") or r.get("product_name"),
+            r.get("display_name") or r.get("产品名称"),
+            r.get("规格名称（新）") or r.get("spec_new"),
+            r.get("规格名称（旧）") or r.get("spec_old"),
+            _to_int(r.get("7天销量")),
+            _to_int(r.get("15天销量")),
+            _to_int(r.get("30天销量")),
+            _to_int(r.get("60天销量")),
+            _to_int(r.get("90天销量")),
+            _to_int(r.get("月销量")),
+            _to_int(r.get("实际可用数")),
+            _to_int(r.get("采购在途")),
+            _to_int(r.get("补货量")),
+            r.get("机器人备注") or r.get("bot_remark"),
+            r.get("write_status", "pending"),
+        )
+        for r in rows
+    ]
+    skus = [v[1] for v in values if v[1]]
+    if not values:
+        return 0
+    try:
+        conn = _conn()
+        try:
+            existing = _filter_existing(conn, "fact_replenishment", run_id, skus)
+            filtered = [v for v in values if v[1] not in existing]
+            if not filtered:
+                return 0
+            with conn.cursor() as cur:
+                execute_values(cur, sql, filtered)
+            conn.commit()
+            return len(filtered)
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] insert_replenishments failed: %s", e)
+        return 0
+
+
+# --- fact_purchase_order_update -------------------------------------------
+
+def insert_purchase_order_updates(
+    run_id: str,
+    target_date: str | None,
+    remark: str | None,
+    rows: list[dict[str, Any]],
+) -> int:
+    if not rows:
+        return 0
+    sql = """
+    INSERT INTO fact_purchase_order_update
+        (run_id, target_sku, target_date, remark, success, updated_sku,
+         product_name, product_spec, error_message, completed_at)
+    VALUES %s
+    ON CONFLICT (run_id, target_sku) DO NOTHING
+    """
+    values = [
+        (
+            run_id,
+            str(r.get("target_sku") or r.get("sku") or "").strip().lstrip("'"),
+            target_date,
+            remark,
+            bool(r.get("success", True)),
+            r.get("updated_sku") or r.get("sku"),
+            r.get("name") or r.get("product_name"),
+            r.get("spec") or r.get("product_spec"),
+            r.get("error"),
+            r.get("completed_at"),
+        )
+        for r in rows
+    ]
+    skus = [v[1] for v in values if v[1]]
+    if not values:
+        return 0
+    try:
+        conn = _conn()
+        try:
+            existing = _filter_existing(conn, "fact_purchase_order_update", run_id, skus)
+            filtered = [v for v in values if v[1] not in existing]
+            if not filtered:
+                return 0
+            with conn.cursor() as cur:
+                execute_values(cur, sql, filtered)
+            conn.commit()
+            return len(filtered)
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] insert_purchase_order_updates failed: %s", e)
+        return 0
+
+
+# --- helpers ---------------------------------------------------------------
+
+def _execute_values(sql: str, values: list[tuple]) -> int:
+    try:
+        conn = _conn()
+        try:
+            with conn.cursor() as cur:
+                execute_values(cur, sql, values)
+            conn.commit()
+            return len(values)
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] batch insert failed (%d rows skipped): %s", len(values), e)
+        return 0
+
+
+def _filter_existing(
+    conn,
+    table: str,
+    run_id: str,
+    sku_values: list[str],
+) -> set[str]:
+    """查 run_id 下已存在的 sku 集合,用于去重后再 bulk insert。"""
+    if not sku_values:
+        return set()
+    sku_col = "target_sku" if table == "fact_purchase_order_update" else "sku"
+    # 全部转 str,避免上游传入整数时类型不匹配
+    sku_strs = [str(s) for s in sku_values if s is not None and str(s).strip()]
+    with conn.cursor() as cur:
+        cur.execute(
+            f"SELECT {sku_col} FROM {table} WHERE run_id = %s AND {sku_col} = ANY(%s)",
+            (run_id, sku_strs),
+        )
+        return {str(row[0]) for row in cur.fetchall() if row[0]}
+
+
+def _to_int(v: Any) -> int | None:
+    if v is None:
+        return None
+    try:
+        if isinstance(v, str) and not v.strip():
+            return None
+        return int(float(v))
+    except Exception:
+        return None
+
+
+def _execute_values(sql: str, values: list[tuple]) -> int:
+    try:
+        conn = _conn()
+        try:
+            with conn.cursor() as cur:
+                execute_values(cur, sql, values)
+                inserted = cur.rowcount
+            conn.commit()
+            return inserted
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] batch insert failed (%d rows skipped): %s", len(values), e)
+        return 0
+
+
+# --- v2: bitable_insert_log + 通知回执 ------------------------------------
+# 2026-06-22 设计:bitable_insert_log.UNIQUE(workflow,run_id,version) 物理兜底
+# 防 LLM 重复调 +record-batch-create。version 由 orchestrator 在 prompt 里写死,
+# 跑完失败后 orchestrator 起新轮次时 +1,LLM 不自决。
+
+import hashlib
+
+
+def read_replenishment(run_id: str) -> list[dict[str, Any]]:
+    """LLM analyzer 读本次 run 的补货明细(data_hash/bitable_version 等一并返回)。
+
+    失败返回空 list。
+    """
+    sql = """
+    SELECT sku, product_name, display_name, spec_new, spec_old,
+           sales_7d, sales_15d, sales_30d, sales_60d, sales_90d, monthly_sales,
+           available_stock, purchase_in_transit, suggest_qty, bot_remark,
+           data_hash, bitable_version
+    FROM fact_replenishment
+    WHERE run_id = %s
+    ORDER BY sku
+    """
+    try:
+        conn = _conn()
+        try:
+            with conn.cursor() as cur:
+                cur.execute(sql, (run_id,))
+                cols = [d[0] for d in cur.description]
+                rows = [dict(zip(cols, r)) for r in cur.fetchall()]
+            conn.commit()
+            return rows
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] read_replenishment failed (non-fatal): %s", e)
+        return []
+
+# 2026-06-22 设计:bitable_insert_log.UNIQUE(workflow,run_id,version) 物理兜底
+# 防 LLM 重复调 +record-batch-create。version 由 orchestrator 在 prompt 里写死,
+# 跑完失败后 orchestrator 起新轮次时 +1,LLM 不自决。
+
+
+
+def compute_data_hash(rows: list[dict[str, Any]], key: str = "sku") -> str:
+    """对一批明细按 key 排序后 sha256,作为该批数据指纹,用于跨 run 校验一致性。"""
+    keys = sorted(
+        {str(r.get(key) or r.get("商品编号") or "").strip().lstrip("'")
+         for r in rows if r is not None}
+    )
+    keys = [k for k in keys if k]
+    if not keys:
+        return hashlib.sha256(b"").hexdigest()
+    payload = "\n".join(keys).encode("utf-8")
+    return hashlib.sha256(payload).hexdigest()
+
+
+def acquire_bitable_insert(
+    workflow: str,
+    run_id: str,
+    data_hash: str,
+    table_token: str,
+    table_id: str,
+    record_count: int,
+    version: int = 1,
+) -> bool:
+    """声明一次多维表写入意图。物理幂等:同 (workflow,run_id,version) 第二次会 ON CONFLICT NOOP,返回 False。
+
+    Returns:
+        True  - 本次新占位成功(LLM 可以放心调 lark-cli)
+        False - 同 version 已有记录,LLM 必须拒绝再调 lark-cli
+    """
+    sql = """
+    INSERT INTO bitable_insert_log
+        (workflow, run_id, version, data_hash, table_token, table_id, record_count, status)
+    VALUES (%s, %s, %s, %s, %s, %s, %s, 'pending')
+    ON CONFLICT (workflow, run_id, version) DO NOTHING
+    """
+    try:
+        conn = _conn()
+        try:
+            with conn.cursor() as cur:
+                cur.execute(
+                    sql,
+                    (workflow, run_id, version, data_hash, table_token, table_id, record_count),
+                )
+                acquired = cur.rowcount == 1
+            conn.commit()
+            return acquired
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] acquire_bitable_insert failed (non-fatal): %s", e)
+        return False
+
+
+def complete_bitable_insert(
+    workflow: str,
+    run_id: str,
+    version: int,
+    status: str,
+    bitable_record_ids: list[str] | None = None,
+    error: str | None = None,
+) -> None:
+    """lark-cli 调完后回调,更新 status/bitable_record_ids/error/completed_at。"""
+    sql = """
+    UPDATE bitable_insert_log
+    SET status             = %s,
+        bitable_record_ids = %s::jsonb,
+        error              = %s,
+        completed_at       = now()
+    WHERE workflow = %s AND run_id = %s AND version = %s
+    """
+    try:
+        conn = _conn()
+        try:
+            with conn.cursor() as cur:
+                cur.execute(
+                    sql,
+                    (
+                        status,
+                        json.dumps(bitable_record_ids or [], ensure_ascii=False),
+                        error or "",
+                        workflow,
+                        run_id,
+                        version,
+                    ),
+                )
+            conn.commit()
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] complete_bitable_insert failed (non-fatal): %s", e)
+
+
+def get_bitable_insert_status(workflow: str, run_id: str) -> dict | None:
+    """查 run_id 下最新一次多维表写入状态(MAX(version))。None 表示从未 acquire。"""
+    sql = """
+    SELECT version, status, record_count, data_hash, bitable_record_ids, error,
+           acquired_at, completed_at
+    FROM bitable_insert_log
+    WHERE workflow = %s AND run_id = %s
+    ORDER BY version DESC
+    LIMIT 1
+    """
+    try:
+        conn = _conn()
+        try:
+            with conn.cursor() as cur:
+                cur.execute(sql, (workflow, run_id))
+                row = cur.fetchone()
+            conn.commit()
+            if not row:
+                return None
+            return {
+                "version": row[0],
+                "status": row[1],
+                "record_count": row[2],
+                "data_hash": row[3],
+                "bitable_record_ids": row[4],
+                "error": row[5],
+                "acquired_at": row[6].isoformat() if row[6] else None,
+                "completed_at": row[7].isoformat() if row[7] else None,
+            }
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] get_bitable_insert_status failed (non-fatal): %s", e)
+        return None
+
+
+def mark_notification_sent(
+    workflow: str,
+    run_id: str,
+    message_ids: list[str] | None = None,
+    version: int | None = None,
+) -> None:
+    """通知发完后回调,落 notification_message_ids + notification_sent_at 到 bitable_insert_log
+    和对应 fact_* 表。
+
+    version: 显式指定写到哪一行(对应"哪次多维表写入的通知");不传则取最大 version 行。
+    """
+    msg_json = json.dumps(message_ids or [], ensure_ascii=False)
+    if version is not None:
+        log_sql = """
+        UPDATE bitable_insert_log
+        SET notification_message_ids = %s::jsonb,
+            notification_sent_at     = now()
+        WHERE workflow = %s AND run_id = %s AND version = %s
+        """
+        log_params = (msg_json, workflow, run_id, version)
+    else:
+        log_sql = """
+        UPDATE bitable_insert_log
+        SET notification_message_ids = %s::jsonb,
+            notification_sent_at     = now()
+        WHERE workflow = %s AND run_id = %s
+          AND version = (
+            SELECT MAX(version) FROM bitable_insert_log
+            WHERE workflow = %s AND run_id = %s
+          )
+        """
+        log_params = (msg_json, workflow, run_id, workflow, run_id)
+    try:
+        conn = _conn()
+        try:
+            with conn.cursor() as cur:
+                cur.execute(log_sql, log_params)
+            conn.commit()
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] mark_notification_sent (log) failed (non-fatal): %s", e)
+
+    # 回填到对应 fact_* 表,便于查询"这条 run 的通知发给谁"
+    fact_table = {
+        "replenishment": "fact_replenishment",
+        "replenishment-alert": "fact_threshold_alert",
+        "purchase-confirmation": "fact_purchase_confirmation",
+        "purchase-order-update": "fact_purchase_order_update",
+    }.get(workflow)
+    if not fact_table:
+        return
+    if version is not None:
+        fact_sql = f"""
+        UPDATE {fact_table}
+        SET notification_message_ids = %s::jsonb,
+            notification_sent_at     = now()
+        WHERE run_id = %s AND bitable_version = %s
+        """
+        fact_params: tuple = (msg_json, run_id, version)
+    else:
+        fact_sql = f"""
+        UPDATE {fact_table}
+        SET notification_message_ids = %s::jsonb,
+            notification_sent_at     = now()
+        WHERE run_id = %s
+        """
+        fact_params = (msg_json, run_id)
+    try:
+        conn = _conn()
+        try:
+            with conn.cursor() as cur:
+                cur.execute(fact_sql, fact_params)
+            conn.commit()
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] mark_notification_sent (fact) failed (non-fatal): %s", e)
+
+
+def notification_already_sent(workflow: str, run_id: str) -> bool:
+    """Check whether mark_notification_sent has been called for this run
+    (i.e. notification_message_ids is non-null in the fact table)."""
+    fact_table = {
+        "replenishment": "fact_replenishment",
+        "replenishment-alert": "fact_threshold_alert",
+        "purchase-confirmation": "fact_purchase_confirmation",
+        "purchase-order-update": "fact_purchase_order_update",
+    }.get(workflow)
+    if not fact_table:
+        return False
+    try:
+        conn = _conn()
+        try:
+            with conn.cursor() as cur:
+                cur.execute(
+                    f"SELECT 1 FROM {fact_table} "
+                    f"WHERE run_id = %s AND notification_message_ids IS NOT NULL "
+                    f"LIMIT 1",
+                    (run_id,),
+                )
+                return cur.fetchone() is not None
+        finally:
+            conn.close()
+    except Exception as e:
+        log.warning("[pg] notification_already_sent check failed: %s", e)
+        return False
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/runner.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/runner.py
new file mode 100644
index 0000000..4484572
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/runner.py
@@ -0,0 +1,54 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""CLI entry points for the auto-flow orchestrator (MCP only)."""
+
+from __future__ import annotations
+
+import argparse
+import io
+import json
+import sys
+
+if sys.stdout.encoding and sys.stdout.encoding.lower() != "utf-8":
+    sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8", errors="replace")
+    sys.stderr = io.TextIOWrapper(sys.stderr.buffer, encoding="utf-8", errors="replace")
+
+from gyxx_flow.modules.supply_chain.runtime.orchestrator.config import WORKFLOWS
+from gyxx_flow.modules.supply_chain.runtime.orchestrator.mcp_workflow import (
+    run_mcp_all,
+    run_mcp_workflow,
+)
+
+WORKFLOW_CHOICES = sorted(WORKFLOWS.keys()) + ["all"]
+
+
+def run_mcp(workflow_name: str) -> dict:
+    """Run workflow via Hermes MCP gateways."""
+
+    if workflow_name == "all":
+        return run_mcp_all()
+    return run_mcp_workflow(workflow_name)
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Auto-flow orchestrator (MCP only)")
+    sub = parser.add_subparsers(dest="command")
+
+    mcp_parser = sub.add_parser("mcp", help="Run via Hermes MCP gateways")
+    mcp_parser.add_argument(
+        "workflow",
+        choices=WORKFLOW_CHOICES,
+    )
+    args = parser.parse_args()
+
+    if args.command == "mcp":
+        result = run_mcp(args.workflow)
+        print(json.dumps(result, ensure_ascii=False, indent=2, default=str))
+        return 0
+
+    parser.print_help()
+    return 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/ProductReplenishment.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/ProductReplenishment.py
new file mode 100644
index 0000000..613af84
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/ProductReplenishment.py
@@ -0,0 +1,1053 @@
+#!/usr/bin/env python3
+# ruff: noqa: E402
+# -*- coding: utf-8 -*-
+"""
+商品补货 - ERP 数据生成脚本
+
+基于聚水潭 ERP UI 导出全量库存数据，进行补货分析
+"""
+
+import io
+import sys
+
+# 强制将 stdout/stderr 设置为 UTF-8 编码，解决 Windows 控制台乱码问题
+sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8', errors='replace')
+sys.stderr = io.TextIOWrapper(sys.stderr.buffer, encoding='utf-8', errors='replace')
+
+import json
+import math
+import os
+import re
+import time
+
+import pandas as pd
+import requests
+from playwright.sync_api import sync_playwright
+
+from gyxx_flow.modules.supply_chain.runtime.paths import (
+    BACKUP_ROOT,
+    CONFIG_PATH,
+    SHARED_ROOT,
+    WORK_ROOT,
+)
+
+REPLENISHMENT_RAW_ROOT = SHARED_ROOT / "incoming" / "replenishment"
+REPLENISHMENT_WORK_ROOT = WORK_ROOT
+REPLENISHMENT_EXPORT_ROOT = BACKUP_ROOT / "replenishment"
+
+try:
+    from gyxx_flow.modules.supply_chain.runtime.orchestrator.feishu_sheets import (
+        load_discontinued_skus_from_feishu,
+    )
+    _FEISHU_SHEETS_OK = True
+except Exception as _fs_imp_err:
+    _FEISHU_SHEETS_OK = False
+    print(f"[WARN] feishu_sheets 不可用,停产 SKU 将从本地 xlsx 加载: {_fs_imp_err}")
+
+# ========================== 外部接口常量 (JST Constants) ==========================
+EXCLUDE_NAMES = [
+    "星悦", "邮费连接", "无纺袋", "飞机盒", "飞机盒外箱", "OZKO 无纺袋", "OZKO 飞机盒",
+    "星迹", "补发", "阿波罗（新无线充电）", "阿波罗无U盘", "暗物质A1",
+    "白鹤", "百灵", "波塞冬2", "波塞冬电脑袋", "晨星mini", "池鹭air", "充电背包简约版",
+    "盖世", "盖世P1", "盖世T科技版", "盖亚Neo", "盖亚Ultra", "盖亚V", "盖亚X",
+    "盖亚泡芙相机包", "盖亚双肩2", "盖亚斜挎2", "黑啤", "黑啤拉链款", "横版单肩斜挎包",
+    "横版动物帆布包", "横版女孩帆布包", "幻月", "极星手提", "极星双肩Pura", "剑道单肩包",
+    "剑道升级版", "扩容手提包（中号）", "蓝鹊", "蓝莺", "冷感便携毛巾", "理岩a", "灵速",
+    "流云", "露娜", "露娜FIT", "洛基", "买家秀晒图虚拟商品", "木兰背包", "牛皮", "千鹤",
+    "轻旅包", "轻云", "轻云S", "数码收纳配件包", "无纺袋（HK）", "五金扣", "弦月", "鞋袋",
+    "星舰", "星蓝", "星旅", "星源", "星云", "星云a", "星云皮膜版a", "星云无线充电升级版",
+    "游云简约版", "雨衣", "云朵", "云卷背包", "云卷多用包", "斩月", "知更",
+    "知更Pro", "宙斯", "宙斯Z", "宙斯Z扩展", "宙斯扩展包", "逐星EVO", "牛皮电源包",
+    "PU电源包", "双口袋手提包包", "无线鼠标", "专业版斜跨小方包", "YSZ麻布手提升级版",
+    "捕梦者", "横版飞机帆布包", "邂逅专业版", "蓝莓手提包", "横版单肩口袋包新", "拼色斜挎包",
+    "兵人包升级版", "卷轴", "卷轴（肩带）", "switch猫爪摇杆帽", "气泡袋", "垂柳拉链版（无小袋）",
+    "昼夜者", "保温杯", "超凡者", "夜游者", "云帆帆布", "光影雨伞", "OZKO 抽绳包",
+    "盖亚微单版", "白羽", "帆布（猫抓鱼）", "帆布（狗与骨头）", "帆布（二哈骨头）",
+    "光影行星咖啡杯", "星舰配件包",  "盖世sport专业版", "灵速隔板",
+    "星舰相机隔层配件", "轻云户外版", "轻云斜挎", "礼盒", "内托", "ozko羽毛球挂件",
+    "磁吸礼盒", "磁吸外箱", "光影礼袋", "IC卡卡套", "IC集卡", "化妆收纳包", "云雀","云卷S"
+]
+
+SPECIAL_SKU_STOCK_THRESHOLDS = {
+    "10440003": 100,
+    "10362004": 300,
+    "10362007": 300,
+    "10362008": 300,
+    "10362010": 300,
+    "10362011": 300,
+}
+
+
+REQUIRED_STOCK_COLUMNS = {
+    "商品编码",
+    "商品名称",
+    "仓库库存数",
+    "可用数",
+    "采购在途数",
+    "7天销量",
+    "15天销量",
+    "30天销量",
+    "60天销量",
+    "90天销量",
+}
+
+
+def validate_stock_export_file(file_path):
+    try:
+        preview_df = pd.read_excel(file_path, nrows=5)
+    except Exception as e:
+        return False, f"无法读取 Excel：{e}"
+
+    columns = {str(c).strip() for c in preview_df.columns}
+    missing = sorted(REQUIRED_STOCK_COLUMNS - columns)
+    if missing:
+        return False, f"缺少库存数据列：{', '.join(missing)}；当前列：{', '.join(columns)}"
+    return True, "OK"
+
+
+def build_threshold_alerts(raw_inventory_df):
+    alerts = []
+    seen_skus = set()
+    for _, row in raw_inventory_df.iterrows():
+        sku = normalize_sku(row.get('sku_id'))
+        threshold = SPECIAL_SKU_STOCK_THRESHOLDS.get(sku)
+        if threshold is None or sku in seen_skus:
+            continue
+
+        qty_stock = float(row.get('qty_stock', 0) or 0)
+        if qty_stock <= threshold:
+            alerts.append({
+                "sku": sku,
+                "threshold": threshold,
+                "actual_stock": qty_stock,
+                "purchase_all": float(row.get('purchase_all', 0) or 0),
+                "product_name": str(row.get('name', "")),
+                "product_spec": str(row.get('props', "")),
+                "sales_7": float(row.get('sales_7', 0) or 0),
+                "sales_15": float(row.get('sales_15', 0) or 0),
+                "sales_30": float(row.get('sales_30', 0) or 0),
+                "generated_at": time.strftime("%Y-%m-%d %H:%M:%S"),
+            })
+            seen_skus.add(sku)
+    return alerts
+
+
+def save_threshold_alerts(alerts, output_dir):
+    alert_file = os.path.join(output_dir, "threshold_alert_data.json")
+    data = {
+        "alerts": alerts,
+        "alert_count": len(alerts),
+        "generated_at": time.strftime("%Y-%m-%d %H:%M:%S"),
+    }
+    with open(alert_file, "w", encoding="utf-8") as f:
+        json.dump(data, f, ensure_ascii=False, indent=2)
+    print(f"[INFO] 指定 SKU 库存阈值预警数据已保存：{alert_file}（共 {len(alerts)} 条）")
+    return alert_file
+
+
+def normalize_sku(value):
+    """Normalize ERP SKU values that may be read from Excel as numbers."""
+    if pd.isna(value):
+        return ""
+    text = str(value).strip().lstrip("'")
+    if re.fullmatch(r"\d+\.0", text):
+        text = text[:-2]
+    return text
+
+
+def load_discontinued_skus(file_path=None):
+    """加载停产 SKU 列表。优先从飞书表格读取,失败时回退到本地 xlsx。"""
+    if _FEISHU_SHEETS_OK:
+        skus = load_discontinued_skus_from_feishu()
+        if skus:
+            return skus
+        print("[WARN] 飞书表格返回空集,尝试本地 xlsx 回退")
+
+    if file_path is None:
+        return set()
+    if not os.path.exists(file_path):
+        print(f"[WARN] 停产产品文件不存在，跳过停产 SKU 过滤: {file_path}")
+        return set()
+
+    try:
+        sheets = pd.read_excel(file_path, sheet_name=None, dtype=str)
+    except Exception as exc:
+        print(f"[WARN] 读取停产产品文件失败，跳过停产 SKU 过滤: {exc}")
+        return set()
+
+    skus = set()
+    for sheet_name, sheet_df in sheets.items():
+        sku_col = next((col for col in sheet_df.columns if str(col).strip() == "商品编码"), None)
+        if sku_col is None:
+            print(f"[WARN] 停产产品文件 sheet={sheet_name} 未找到商品编码列，已跳过")
+            continue
+        for value in sheet_df[sku_col].dropna():
+            sku = normalize_sku(value)
+            if sku:
+                skus.add(sku)
+
+    print(f"[INFO] 已从本地 xlsx 加载停产商品编码 {len(skus)} 个")
+    return skus
+
+
+def calculate_100_day_target(sales_60):
+    return math.ceil((float(sales_60) / 60.0) * 100) if float(sales_60) > 0 else 0
+
+
+def calculate_replenishment_qty(net_balance, sales_60):
+    sales_target = (float(sales_60) / 60.0) * 100 if float(sales_60) > 0 else 0
+    diff = sales_target - net_balance
+    if diff > 0:
+        return int(round(diff / 100.0) * 100)
+    return None
+
+
+def build_replenishment_record(row, suggest_qty=None, remark="机器人操作"):
+    qty_available = float(row["qty_available"])
+    purchase_all = float(row["purchase_all"])
+    sales_60 = float(row["sales_60"])
+    sales_target = calculate_100_day_target(sales_60)
+
+    return {
+        '规格名称（旧）': str(row['props']),
+        '商品名称': row['name'],
+        '商品编号': normalize_sku(row['sku_id']),
+        '产品名称': row['short_name'],
+        '规格名称（新）': row['cleaned_props'],
+        '实际可用数': qty_available,
+        '采购在途': purchase_all,
+        '7天销量': float(row['sales_7']),
+        '15天销量': float(row['sales_15']),
+        '月销量': float(row['sales_30']),
+        '60天销量': sales_60,
+        '90天销量': float(row['sales_90']),
+        '补货量': suggest_qty if suggest_qty else None,
+        '机器人备注': remark,
+        '_net_balance': qty_available + purchase_all,
+        '_sales_100': sales_target
+    }
+
+
+def save_replenishment_source(df, output_dir):
+    records = []
+    for _, row in df.iterrows():
+        qty_available = float(row["qty_available"])
+        purchase_all = float(row["purchase_all"])
+        sales_60 = float(row["sales_60"])
+        net_balance = qty_available + purchase_all
+        suggest_qty = calculate_replenishment_qty(net_balance, sales_60)
+        records.append(build_replenishment_record(row, suggest_qty=suggest_qty))
+
+    source_file = os.path.join(output_dir, "replenishment_source.json")
+    with open(source_file, "w", encoding="utf-8") as f:
+        json.dump({"records": records, "count": len(records), "generated_at": time.strftime("%Y-%m-%d %H:%M:%S")}, f, ensure_ascii=False, indent=2)
+    print(f"[INFO] ERP 补货源数据已保存：{source_file}（共 {len(records)} 条）")
+    return source_file
+
+
+class JstManager:
+    """聚水潭外部接口整合类 (All-in-one)"""
+
+    USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/143.0.0.0 Safari/537.36"
+
+    def __init__(self, config_path):
+        self.config = self._load_config(config_path)
+        self.session = requests.Session()
+        self._browser = None
+        self._playwright = None
+        self._page = None
+        self._context = None
+
+        # 默认参数
+        self.u_id = "22469720"
+        self.u_co_id = "10621105"
+        
+        # CDP URL 配置
+        self.cdp_url = self.config.get("cdp_url")
+
+    def _load_config(self, config_path):
+        if not os.path.isabs(config_path) and not os.path.exists(config_path):
+            script_dir = os.path.dirname(os.path.abspath(__file__))
+            potential_path = os.path.join(script_dir, config_path)
+            if os.path.exists(potential_path):
+                config_path = potential_path
+        with open(config_path, 'r', encoding='utf-8') as f:
+            return json.load(f)
+
+    def _inventory_url(self):
+        return "https://src.erp321.com/erp-web-group/erp-scm-goods/goodsInventoryManagement?tabAllow=camera&_c=jst-epaas"
+
+    def _sync_session_from_context(self):
+        """Sync requests.Session cookies from the browser context."""
+        if not self._context:
+            return []
+        cookies = self._context.cookies()
+        for c in cookies:
+            self.session.cookies.set(c['name'], c['value'], domain=c.get('domain', '.erp321.com'))
+            if c['name'] == 'u_id':
+                self.u_id = c['value']
+            if c['name'] == 'u_co_id':
+                self.u_co_id = c['value']
+        print(f"[INFO] Synced {len(cookies)} cookies from browser context")
+        return cookies
+
+    def _page_has_inventory_access(self, page):
+        """Check whether the current page has valid ERP inventory access."""
+        try:
+            current_url = page.url.lower()
+            if "login" in current_url:
+                return False
+            page.goto(self._inventory_url(), wait_until="domcontentloaded", timeout=15000)
+            try:
+                page.wait_for_load_state("networkidle", timeout=10000)
+            except Exception:
+                pass
+            time.sleep(2)
+            final_url = page.url.lower()
+            if "login" in final_url:
+                return False
+            return True
+        except Exception as e:
+            print(f"[WARN] inventory access check failed: {e}")
+            return False
+
+    def connect_existing_browser(self, cdp_url=None):
+        """连接已打开的浏览器（通过 CDP URL）。"""
+        target_url = cdp_url or self.cdp_url
+        if not target_url:
+            print("[ERROR] 未提供 CDP URL")
+            return False
+
+        try:
+            self._playwright = sync_playwright().start()
+            print(f"[INFO] 正在连接 CDP 浏览器: {target_url}")
+            if not target_url.startswith('ws://') and not target_url.startswith('http://'):
+                target_url = f"http://{target_url}"
+            proxy_vars = ['HTTP_PROXY', 'HTTPS_PROXY', 'ALL_PROXY', 'http_proxy', 'https_proxy', 'all_proxy']
+            saved_proxies = {k: os.environ.pop(k, None) for k in proxy_vars}
+            try:
+                self._browser = self._playwright.chromium.connect_over_cdp(target_url)
+            finally:
+                for k, v in saved_proxies.items():
+                    if v is not None:
+                        os.environ[k] = v
+
+            if len(self._browser.contexts) > 0:
+                self._context = self._browser.contexts[0]
+            else:
+                self._context = self._browser.new_context(user_agent=self.USER_AGENT)
+
+            if len(self._context.pages) > 0:
+                self._page = self._context.pages[0]
+                for extra_page in self._context.pages[1:]:
+                    try:
+                        extra_page.close()
+                    except Exception:
+                        pass
+            else:
+                self._page = self._context.new_page()
+
+            print(f"???????????? URL: {self._page.url}")
+            return True
+        except Exception as e:
+            print(f"?????????: {e}")
+            return False
+
+    def _dismiss_post_login_dialogs(self, page, max_rounds=4, interval=1.0):
+        """
+        关闭登录后可能弹出的提示框（密码过期、公告等）。
+        检测「确定 / OK / 知道了 / 关闭」等按钮，点击后继续，不影响后续流程。
+        """
+        if not page:
+            return
+        confirm_keywords = [
+            "\u786e\u5b9a",  # 确定
+            "\u77e5\u9053\u4e86",  # 知道了
+            "\u5173\u95ed",  # 关闭
+            "\u597d\u7684",  # 好的
+            "\u662f",         # 是
+            "\u53d6\u6d88",  # 取消（部分弹框有「取消」作为关闭）
+            "ok",
+            "confirm",
+        ]
+        # 覆盖常见企业 UI 框架的弹层选择器
+        dialog_selectors = [
+            ".ant-modal",
+            ".ant-modal-wrap",
+            ".ant-message",
+            ".ant-notification",
+            ".el-dialog",
+            ".el-message-box",
+            ".el-message",
+            ".el-notification",
+            ".layui-layer",
+            ".layui-layer-dialog",
+            ".layui-layer-shade",
+            ".modal",
+            ".modal-dialog",
+            ".modal-content",
+            ".dialog",
+            "[role='dialog']",
+            "[role='alertdialog']",
+            ".swal2-container",
+            ".toast",
+            ".popup",
+            ".alert-dialog",
+        ]
+        for _ in range(max_rounds):
+            try:
+                # 1) 找任何可见的弹层
+                any_dialog = False
+                for sel in dialog_selectors:
+                    try:
+                        loc = page.locator(sel)
+                        if loc.count() > 0:
+                            for i in range(loc.count()):
+                                try:
+                                    if loc.nth(i).is_visible():
+                                        any_dialog = True
+                                        break
+                                except Exception:
+                                    continue
+                        if any_dialog:
+                            break
+                    except Exception:
+                        continue
+                if not any_dialog:
+                    time.sleep(interval)
+                    continue
+                # 2) 在弹层内点掉「确定」等按钮
+                clicked = False
+                for keyword in confirm_keywords:
+                    try:
+                        btn = page.get_by_role("button", name=keyword)
+                        if btn.count() > 0:
+                            for i in range(btn.count()):
+                                try:
+                                    if btn.nth(i).is_visible():
+                                        btn.nth(i).click(force=True)
+                                        clicked = True
+                                        print(f"[LOGIN] 自动点击弹框按钮: {keyword}")
+                                        break
+                                except Exception:
+                                    continue
+                            if clicked:
+                                break
+                    except Exception:
+                        continue
+                # 3) 通用兜底：弹层内任意 <button> / .ant-btn / .layui-layer-btn
+                if not clicked:
+                    for fallback_sel in [
+                        ".ant-modal button.ant-btn-primary",
+                        ".ant-modal button.ant-btn",
+                        ".el-dialog button.el-button--primary",
+                        ".el-dialog button.el-button",
+                        ".layui-layer .layui-layer-btn a",
+                        ".layui-layer button",
+                        ".modal-footer button",
+                        "dialog button",
+                        "[role='dialog'] button",
+                    ]:
+                        try:
+                            loc = page.locator(fallback_sel)
+                            if loc.count() > 0:
+                                for i in range(loc.count()):
+                                    try:
+                                        if loc.nth(i).is_visible():
+                                            loc.nth(i).click(force=True)
+                                            clicked = True
+                                            print(f"[LOGIN] 自动点击弹框按钮(fallback): {fallback_sel}")
+                                            break
+                                    except Exception:
+                                        continue
+                                if clicked:
+                                    break
+                        except Exception:
+                            continue
+                # 4) 终极兜底：按 ESC 关闭
+                if not clicked:
+                    try:
+                        page.keyboard.press("Escape")
+                    except Exception:
+                        pass
+                time.sleep(interval)
+            except Exception:
+                time.sleep(interval)
+
+    def _login_on_current_page(self, url, username, password):
+        """在已连接的浏览器页面上执行登录流程（强制登录，不检测已有会话）"""
+        if not self._page:
+            return False
+        page = self._page
+
+        # 强制走登录流程，不做登录状态检测
+        page.goto(url, wait_until="domcontentloaded")
+        try:
+            page.wait_for_load_state("networkidle", timeout=10000)
+        except Exception:
+            pass
+        time.sleep(2)
+
+        user_selectors = ['input[name="username"]', 'input[name="user"]', 'input[id="username"]', 'input[type="text"]']
+        for s in user_selectors:
+            try:
+                locator = page.locator(s).first
+                if locator.count() > 0 and locator.is_visible():
+                    locator.fill(username)
+                    break
+            except Exception:
+                pass
+
+        pass_selectors = ['input[name="password"]', 'input[name="pwd"]', 'input[id="password"]', 'input[type="password"]']
+        for s in pass_selectors:
+            try:
+                locator = page.locator(s).first
+                if locator.count() > 0 and locator.is_visible():
+                    locator.fill(password)
+                    break
+            except Exception:
+                pass
+
+        checkbox_selectors = ['input[type="checkbox"]', 'input[name="agree"]', 'input[id="agree"]']
+        for s in checkbox_selectors:
+            try:
+                count = page.locator(s).count()
+                for i in range(count):
+                    box = page.locator(s).nth(i)
+                    if box.is_visible() and not box.is_checked():
+                        box.check()
+            except Exception:
+                pass
+
+        clicked = False
+        login_btns = ['button[type="submit"]', 'input[type="submit"]', 'button:has-text("\u767b\u5f55")', '#btnLogin']
+        for s in login_btns:
+            try:
+                locator = page.locator(s).first
+                if locator.count() > 0 and locator.is_visible():
+                    locator.click()
+                    clicked = True
+                    break
+            except Exception:
+                pass
+        if not clicked:
+            try:
+                page.keyboard.press("Enter")
+            except Exception:
+                pass
+
+        time.sleep(5)
+        try:
+            page.wait_for_load_state("load", timeout=10000)
+        except Exception:
+            pass
+
+        # 关闭登录后弹出的提示框（如密码过期等），不影响后续流程
+        # 多轮兜底：弹框可能在 load 完成前就跳出来，再多等一次确保点掉
+        self._dismiss_post_login_dialogs(page, max_rounds=6, interval=0.5)
+        try:
+            page.wait_for_load_state("load", timeout=5000)
+        except Exception:
+            pass
+        self._dismiss_post_login_dialogs(page, max_rounds=2, interval=0.3)
+
+        self._sync_session_from_context()
+        success = self._page_has_inventory_access(page)
+        if success:
+            print("[OK] Reused browser login succeeded with inventory access")
+        else:
+            print(f"[WARN] Reused browser login did not reach inventory page: {page.url}")
+        return success
+
+    def login(self, use_existing_browser=False, cdp_url=None):
+        """
+        登录 ERP
+        - use_existing_browser=True 时，连接已存在的浏览器（通过 CDP URL）
+        - use_existing_browser=False 时，启动新浏览器（传统模式）
+        """
+        login_config = self.config.get("login", {})
+        url = login_config.get("url")
+        username = login_config.get("username")
+        password = os.getenv("GYXX_SUPPLY_ERP_PASSWORD", "")
+        if not all([url, username, password]):
+            print("[Error] 配置信息不完整")
+            return False
+
+        # 模式1：连接已存在的浏览器（由 Hermes browser 工具打开）
+        if use_existing_browser:
+            if not self.connect_existing_browser(cdp_url):
+                return False
+            # 已连接后，无论当前是否登录，都在当前页面补跑一次登录流程，避免回退到新浏览器分支
+            return self._login_on_current_page(url, username, password)
+
+        # 模式2：启动新浏览器（传统模式）
+        try:
+            self._playwright = sync_playwright().start()
+            headless = login_config.get("headless", False)
+            self._browser = self._playwright.chromium.launch(headless=headless)
+            self._context = self._browser.new_context(user_agent=self.USER_AGENT)
+            self._page = self._context.new_page()
+            page = self._page
+
+            print(f"正在登录: {url}")
+            page.goto(url, wait_until="networkidle")
+            time.sleep(2)
+
+            # 自动表单填充逻辑
+            print("尝试填充用户名...")
+            u_field = page.get_by_placeholder("用户名") or page.get_by_placeholder("手机号") or page.get_by_placeholder("请输入")
+            if u_field.count() > 0:
+                u_field.first.fill(username)
+                print("已使用 placeholder 填写用户名")
+            else:
+                user_selectors = ['input[name="username"]', 'input[name="user"]', 'input[id="username"]', 'input[type="text"]']
+                for s in user_selectors:
+                    try:
+                        if page.locator(s).is_visible():
+                            page.locator(s).fill(username)
+                            print(f"已填写用户名 (使用: {s})")
+                            break
+                    except Exception:
+                        pass
+
+            print("尝试填充密码...")
+            p_field = page.get_by_placeholder("密码")
+            if p_field.count() > 0:
+                p_field.first.fill(password)
+                print("已使用 placeholder 填写密码")
+            else:
+                pass_selectors = ['input[name="password"]', 'input[name="pwd"]', 'input[id="password"]', 'input[type="password"]']
+                for s in pass_selectors:
+                    try:
+                        if page.locator(s).is_visible():
+                            page.locator(s).fill(password)
+                            print(f"已填写密码 (使用: {s})")
+                            break
+                    except Exception:
+                        pass
+
+            # 一键勾选协议
+            print("点选协议...")
+            agreement_selectors = [
+                'input[type="checkbox"]',
+                'input[name="agree"]',
+                '.ant-checkbox-input',
+                '.ant-checkbox-wrapper'
+            ]
+            for s in agreement_selectors:
+                try:
+                    locs = page.locator(s)
+                    for i in range(locs.count()):
+                        if not locs.nth(i).is_checked():
+                            locs.nth(i).click(force=True)
+                            print(f"已勾选协议: {s}")
+                except Exception:
+                    pass
+
+            # 提交
+            print("查找登录按钮...")
+            time.sleep(2)
+
+            # 检测是否有验证码提示
+            if "验证码" in page.content():
+                print("[WARN] 检测到页面可能存在验证码，尝试继续...")
+
+            # 点击包含"登录"文字的按钮
+            login_btn = page.get_by_role("button", name="登录")
+            if login_btn.count() > 0:
+                print("已找到'登录'按钮，准备点击...")
+                login_btn.first.hover()
+                time.sleep(1)
+                login_btn.first.click()
+                print("已点击'登录'按钮")
+            else:
+                btn_selectors = ['button[type="submit"]', '#btnLogin', '.ant-btn-submit', '.ant-btn-primary', 'input[type="submit"]']
+                for s in btn_selectors:
+                    try:
+                        loc = page.locator(s)
+                        if loc.count() > 0 and loc.first.is_visible():
+                            loc.first.hover()
+                            time.sleep(0.5)
+                            loc.first.click()
+                            print(f"已点击按钮: {s}")
+                            break
+                    except Exception:
+                        pass
+                else:
+                    print("未找到点击按钮，尝试 Enter")
+                    page.keyboard.press("Enter")
+
+            # 等待页面跳转 (延长至 45 秒，方便缓慢网络或手动处理验证码)
+            for i in range(45):
+                time.sleep(1)
+                # 先尝试关闭登录后立即跳出的弹框（如密码过期提示），避免遮挡导致 URL 不变
+                if "login" in page.url.lower():
+                    self._dismiss_post_login_dialogs(page, max_rounds=1, interval=0.3)
+                if "login" not in page.url.lower():
+                    print(f"检测到页面跳转成功: {page.url}")
+                    break
+
+            # 跳转完成后再兜底一次，关闭残留的弹框（不影响后续流程）
+            self._dismiss_post_login_dialogs(page)
+
+            print(f"最终 URL: {page.url}")
+
+            # 获取 Cookies 并解析 ID
+            cookies = self._context.cookies()
+            print(f"获取到 {len(cookies)} 个 cookies")
+            for c in cookies:
+                self.session.cookies.set(c['name'], c['value'], domain=c.get('domain', '.erp321.com'))
+                if c['name'] == 'u_id':
+                    self.u_id = c['value']
+                if c['name'] == 'u_co_id':
+                    self.u_co_id = c['value']
+
+            success = "login" not in page.url.lower()
+            if success:
+                print("[OK] 登录成功")
+            else:
+                print("[FAIL] 登录可能失败 (URL 仍然包含 login)")
+            return success
+        except Exception as e:
+            print(f"登录失败: {e}")
+            return False
+
+    def close(self, disconnect=False):
+        """
+        关闭/断开浏览器连接
+        - disconnect=True: 断开连接但不关闭浏览器（用于复用 Hermes 打开的浏览器）
+        - disconnect=False: 关闭浏览器（传统模式）
+        """
+        if disconnect and self._browser:
+            print("已释放 CDP 浏览器连接（浏览器保持打开）")
+        else:
+            # 关闭浏览器
+            if self._browser:
+                try:
+                    self._browser.close()
+                except Exception:
+                    pass
+        if self._playwright:
+            try:
+                self._playwright.stop()
+            except Exception:
+                pass
+
+    def download_stock_data(self, output_dir=None):
+        """通过 UI 触发'商品库存管理'全量数据导出"""
+        if not self._page:
+            return None
+
+        output_dir = os.fspath(output_dir or REPLENISHMENT_RAW_ROOT)
+        os.makedirs(output_dir, exist_ok=True)
+
+        target_url = "https://src.erp321.com/erp-web-group/erp-scm-goods/goodsInventoryManagement?tabAllow=camera&_c=jst-epaas"
+        print(f"\n正在访问商品管理页面: {target_url}")
+        self._page.goto(target_url, wait_until="networkidle")
+        time.sleep(3)
+
+        try:
+            page = self._page
+            export_btn_selector = "button.goods-dropdown-trigger:has-text('导出')"
+            page.wait_for_selector(export_btn_selector, timeout=15000)
+
+            print("触发导出流程...")
+            page.locator(export_btn_selector).click()
+            time.sleep(1.5)
+
+            menu_item_selector = ".ant-dropdown-menu-item, .ant-dropdown-menu li, li[role='menuitem']"
+            items = page.locator(menu_item_selector).filter(has_text=re.compile(r".+"))
+            item_count = items.count()
+            if item_count == 0:
+                items = page.locator("div.ant-dropdown:not(.ant-dropdown-hidden) li")
+                item_count = items.count()
+
+            menu_texts = []
+            for i in range(item_count):
+                try:
+                    menu_texts.append(items.nth(i).inner_text(timeout=1000).strip())
+                except Exception:
+                    menu_texts.append("")
+            print(f"导出菜单项：{menu_texts}")
+
+            def score_menu_item(text):
+                text = text or ""
+                score = 0
+                if any(key in text for key in ["库存", "商品库存", "明细", "全部", "当前查询"]):
+                    score += 10
+                if any(key in text for key in ["款式编码", "模板", "导入"]):
+                    score -= 20
+                return score
+
+            candidate_indices = sorted(range(item_count), key=lambda i: score_menu_item(menu_texts[i]), reverse=True)
+            if not candidate_indices:
+                candidate_indices = [0]
+
+            last_error = None
+            for attempt, item_index in enumerate(candidate_indices, 1):
+                if attempt > 1:
+                    page.locator(export_btn_selector).click()
+                    time.sleep(1.0)
+                    items = page.locator(menu_item_selector).filter(has_text=re.compile(r".+"))
+                    if items.count() == 0:
+                        items = page.locator("div.ant-dropdown:not(.ant-dropdown-hidden) li")
+
+                item_text = menu_texts[item_index] if item_index < len(menu_texts) else f"#{item_index}"
+                print(f"尝试导出菜单项 {item_index + 1}/{item_count}: {item_text}")
+
+                try:
+                    with page.expect_download(timeout=60000) as download_info:
+                        items.nth(item_index).click()
+
+                    download = download_info.value
+                    timestamp = time.strftime("%Y%m%d_%H%M%S")
+                    save_path = os.path.join(output_dir, f"JST_Stock_{timestamp}.xlsx")
+                    download.save_as(save_path)
+
+                    valid, reason = validate_stock_export_file(save_path)
+                    if valid:
+                        print(f"数据下载成功: {save_path}")
+                        return save_path
+
+                    last_error = reason
+                    print(f"[WARN] 导出文件字段不完整，跳过该菜单项：{reason}")
+                    try:
+                        os.remove(save_path)
+                    except Exception:
+                        pass
+                except Exception as e:
+                    last_error = str(e)
+                    print(f"[WARN] 菜单项导出失败：{e}")
+
+            print(f"所有导出菜单项都没有得到完整库存数据。最后错误：{last_error}")
+            return None
+        except Exception as e:
+            print(f"UI 下载失败: {e}")
+            return None
+
+    @staticmethod
+    def clean_props(raw_props):
+        """清洗规格名称，去除英文和数字，排序并合并"""
+        if not raw_props or pd.isna(raw_props):
+            return ""
+        # 统一分号，拆分并去空格
+        parts = [p.strip() for p in str(raw_props).replace('；', ';').split(';')]
+        # 去除英文、数字、空格
+        parts = [re.sub(r'[a-zA-Z0-9\s]', '', p) for p in parts]
+        # 过滤空字符串并排序
+        parts = sorted([p for p in parts if p])
+        return ';'.join(parts)
+
+
+def main(cdp_url=None):
+    """
+    主流程：登录 ERP → 导出库存数据 → 分析补货建议
+    
+    Args:
+        cdp_url: 可选，CDP 连接 URL。传入时连接已存在的浏览器，不启动新浏览器
+    """
+    import argparse
+    parser = argparse.ArgumentParser(description='商品补货 ERP 数据生成')
+    parser.add_argument('--cdp-url', type=str, default=None, help='CDP 连接 URL（用于连接已存在的浏览器）')
+    args, _ = parser.parse_known_args()
+    
+    # 命令行参数的 CDP URL 优先
+    actual_cdp_url = args.cdp_url or cdp_url
+    use_existing = actual_cdp_url is not None
+    
+    raw_dir = str(REPLENISHMENT_RAW_ROOT)
+    work_dir = str(REPLENISHMENT_WORK_ROOT)
+    export_dir = str(REPLENISHMENT_EXPORT_ROOT)
+    for directory in (raw_dir, work_dir, export_dir):
+        os.makedirs(directory, exist_ok=True)
+
+    manager = JstManager(str(CONFIG_PATH))
+
+    try:
+        # 1. 登录 ERP
+        if use_existing:
+            # 模式1：连接已存在的浏览器（复用 Hermes browser 工具打开的浏览器）
+            target_cdp = actual_cdp_url or manager.cdp_url
+            if target_cdp:
+                print(f"[INFO] 将尝试连接现有浏览器: {target_cdp}")
+                if not manager.login(use_existing_browser=True, cdp_url=target_cdp):
+                    print("连接现有浏览器失败，尝试启动新浏览器...")
+                    # 必须先停掉前一次 sync_playwright 启动的 asyncio loop,
+                    # 否则下面 sync_playwright().start() 会撞上仍在跑的 loop,
+                    # 直接抛 "It looks like you are using Playwright Sync API inside the asyncio loop"
+                    manager.close(disconnect=False)
+                    use_existing = False  # 让 finally 走 close(disconnect=False) 关闭新启动的浏览器
+                    if not manager.login(use_existing_browser=False):
+                        print("登录失败，程序退出。")
+                        return
+                else:
+                    print("[OK] 已连接现有浏览器")
+            else:
+                print("[WARN] 指定了复用浏览器模式但未提供 CDP URL，回退到启动新浏览器")
+                if not manager.login(use_existing_browser=False):
+                    print("登录失败，程序退出。")
+                    return
+        else:
+            # 模式2：启动新浏览器（传统模式）
+            if not manager.login(use_existing_browser=False):
+                print("登录失败，程序退出。")
+                return
+
+        # 2. 登录成功后清理旧数据，再下载新数据
+        excel_file = manager.download_stock_data(raw_dir)
+        if not excel_file:
+            print("导出全量数据失败，流程终止。")
+            return
+
+        # 3. 加载并分析库存数据
+        print("\n=== 加载并分析库存数据 ===")
+        valid, reason = validate_stock_export_file(excel_file)
+        if not valid:
+            print(f"[ERROR] ERP 导出文件不是完整库存数据：{reason}")
+            print("[ERROR] 请检查 ERP 导出菜单是否选择了包含商品编码、可用数、采购在途和销量列的库存明细导出。")
+            return
+        df = pd.read_excel(excel_file)
+
+        # 映射 Excel 列名到内部变量
+        col_map = {
+            '产品名称': 'short_name',
+            '商品名称': 'name',
+            '规格名称': 'props',
+            '颜色及规格': 'props',
+            '仓库库存数': 'qty_stock',
+            '可用数': 'qty_available',
+            '7天销量': 'sales_7',
+            '15天销量': 'sales_15',
+            '30天销量': 'sales_30',
+            '60天销量': 'sales_60',
+            '90天销量': 'sales_90',
+            '采购在途数': 'purchase_all',
+            '商品编码': 'sku_id'
+        }
+        df = df.rename(columns={k: v for k, v in col_map.items() if k in df.columns})
+
+        # 补全缺失列
+        num_cols = ['qty_stock', 'qty_available', 'sales_7', 'sales_15', 'sales_30', 'sales_60', 'sales_90', 'purchase_all']
+        for col in ['short_name', 'name', 'props', 'sku_id'] + num_cols:
+            if col not in df.columns:
+                df[col] = 0 if col in num_cols else ""
+
+        df['sku_id'] = df['sku_id'].apply(normalize_sku)
+        discontinued_skus = load_discontinued_skus()
+        if discontinued_skus:
+            before_discontinued_filter = len(df)
+            df = df[~df['sku_id'].isin(discontinued_skus)].copy()
+            print(f"[INFO] 停产 SKU 过滤完成: {before_discontinued_filter} -> {len(df)}")
+
+        # 清洗数值列，将 NaN 替换为 0
+        for col in num_cols:
+            df[col] = pd.to_numeric(df[col], errors='coerce').fillna(0)
+
+        # 预过滤排除名单（排除特定名称及带 HK 的规格）
+        df = df[~df['name'].astype(str).isin(EXCLUDE_NAMES)].copy()
+        df = df[~df['props'].astype(str).str.contains("HK", case=False, na=False)].copy()
+        
+        # 清洗商品名称和规格属性并合并数据
+        print("正在清洗商品名称和规格属性并合并数据...")
+        df['cleaned_name'] = df['name'].astype(str).str.replace(r'\s+', '', regex=True)
+        df['cleaned_props'] = df['props'].apply(lambda x: manager.clean_props(x))
+        raw_inventory_df = df.copy()
+        
+        # 聚合规则：只汇总相同 SKU 的重复行。
+        # 不能按清洗后的名称/规格跨 SKU 合并，否则会出现表里显示一个 SKU，
+        # 但库存/销量是多个 SKU 合计的情况，和 ERP 按 SKU 查询结果不一致。
+        agg_rules = {
+            'qty_available': 'sum',
+            'sales_7': 'sum',
+            'sales_15': 'sum',
+            'sales_30': 'sum',
+            'sales_60': 'sum',
+            'sales_90': 'sum',
+            'purchase_all': 'sum',
+            'name': 'first',
+            'short_name': 'first',
+            'props': 'first'
+        }
+        
+        # 按 SKU + 清洗后的名称和规格聚合，保留同 SKU 多行汇总，不合并不同 SKU。
+        df = df.groupby(['sku_id', 'cleaned_name', 'cleaned_props']).agg(agg_rules).reset_index()
+        print(f"数据清洗与合并完成，合并后记录总数: {len(df)}")
+        save_replenishment_source(df, work_dir)
+
+        # 4. 执行补货分析 (60天销量参考 -> 100天预估预警)
+        print("\n=== 执行补货分析 (60天销量参考 -> 100天预估预警) ===")
+        filtered_list = []
+        processed_names = set()
+        
+        for _, item in df.iterrows():
+            if item['cleaned_name'] in processed_names:
+                continue
+                
+            # 提取数值
+            qty_avail = float(item['qty_available'])
+            purchase_all = float(item['purchase_all'])
+            sales_60 = float(item['sales_60'])
+
+            # 计算 100 天预估需求量 (基于 60 天历史销量)
+            sales_target = calculate_100_day_target(sales_60)
+
+            # 净值 = 可用数 (库存-待发) + 采购在途
+            net_balance = qty_avail + purchase_all
+
+            # 如果目标销量大于净值，或净值 <= 0 说明需要补货
+            if net_balance < sales_target or net_balance <= 0:
+                # 只要满足了预警拦截条件，即使 suggest_value 不够 70 没算出值，也要将它们拉进预警清单
+                processed_names.add(item['cleaned_name'])
+                # 循环一遍df, 把商品名称（清洗后）相同的都添加到 filtered_list
+                matched_rows = df[df['cleaned_name'] == item['cleaned_name']]
+                for _, row in matched_rows.iterrows():
+                    r_qty = float(row['qty_available'])
+                    r_pur = float(row['purchase_all'])
+                    r_s60 = float(row['sales_60'])
+                    r_net = r_qty + r_pur
+
+                    r_suggest = calculate_replenishment_qty(r_net, r_s60)
+                    filtered_list.append(build_replenishment_record(row, suggest_qty=r_suggest))
+
+        threshold_alerts = build_threshold_alerts(raw_inventory_df)
+        save_threshold_alerts(threshold_alerts, work_dir)
+
+        # 按商品名称排序，保证同款商品输出时排列在一起
+        filtered_list.sort(key=lambda x: str(x.get('商品名称', '')))
+
+        # 5. 导出分析结果
+        if filtered_list:
+            res_df = pd.DataFrame(filtered_list)
+            # 优先按“商品名称”正序，再按“补货量”降序
+            res_df = res_df.sort_values(by=['商品名称', '补货量'], ascending=[True, False])
+            filename = f"补货建议_{time.strftime('%Y%m%d_%H%M%S')}.xlsx"
+            filepath = os.path.join(export_dir, filename)
+            res_df.to_excel(filepath, index=False)
+            print(f"\n[分析完成] 建议补货商品数量: {len(filtered_list)}")
+            print(f"[结果保存] {filepath}")
+
+            # 简单展示 Top 10
+            print("\n--- 补货建议 Top 10 ---")
+            print(res_df.head(10)[['商品名称', '规格名称（新）', '60天销量', '补货量']].to_string(index=False))
+        else:
+            print("经分析，当前暂无需要补货的商品。")
+            # 创建空文件标记，表示无补货建议
+            filepath = None
+
+        # 6. 数据已生成（飞书同步由 batch_process.py 单独执行，不再递归调用以避免死循环）
+
+        return filepath
+
+    except Exception as e:
+        print(f"运行过程中发生错误: {e}")
+        import traceback
+        traceback.print_exc()
+        return None
+    finally:
+        # 如果使用了现有浏览器，断开连接而不是关闭
+        manager.close(disconnect=use_existing)
+
+
+if __name__ == "__main__":
+    result = main()
+    sys.exit(0 if result else 1)
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/PurchaseConfirmation.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/PurchaseConfirmation.py
new file mode 100644
index 0000000..4d31821
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/PurchaseConfirmation.py
@@ -0,0 +1,1209 @@
+# -*- coding: utf-8 -*-
+# ruff: noqa: E402,E701,F841
+# 让 `from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer` 在 `python PurchaseConfirmation.py` 这种
+# 不带 -m 的调用方式下也能解析(2026-06-22 v3 改造后,采集端要落 PG)。
+# 必须在所有 from gyxx_flow.modules.supply_chain.runtime.orchestrator import 之前。
+import io
+import sys
+import warnings
+
+from gyxx_flow.modules.supply_chain.runtime.paths import (
+    BACKUP_ROOT,
+    CONFIG_PATH,
+    SHARED_ROOT,
+)
+
+PURCHASE_CONFIRMATION_RAW_ROOT = SHARED_ROOT / "incoming" / "purchase-confirmation"
+PURCHASE_CONFIRMATION_EXPORT_ROOT = BACKUP_ROOT / "purchase-confirmation"
+
+# 强制将 stdout/stderr 设置为 UTF-8 编码，解决 Windows 控制台乱码问题
+# 加入 try-except 防止重复包装或流已关闭时出错
+try:
+    if not isinstance(sys.stdout, io.TextIOWrapper) or sys.stdout.closed:
+        sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8', errors='replace')
+except Exception:
+    pass
+try:
+    if not isinstance(sys.stderr, io.TextIOWrapper) or sys.stderr.closed:
+        sys.stderr = io.TextIOWrapper(sys.stderr.buffer, encoding='utf-8', errors='replace')
+except Exception:
+    pass
+
+# 抑制 openpyxl 的样式警告，防止 "Workbook contains no default style" 导致 exit(1)
+warnings.filterwarnings('ignore', category=UserWarning, module='openpyxl')
+warnings.filterwarnings('ignore', category=UserWarning, module='pandas')
+
+import datetime
+import json
+import os
+import re
+import secrets
+import time
+
+import openpyxl
+import pandas as pd
+import requests
+from playwright.sync_api import sync_playwright
+
+# 2026-06-22 v3:采集端直接落 PG(pg_writer 是软依赖,失败仅 warn)。
+try:
+    from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer as pg
+    _PG_IMPORT_OK = True
+except Exception as _pg_imp_err:  # pragma: no cover
+    _PG_IMPORT_OK = False
+    print(f"[WARN] pg_writer 不可用,本次不落 PG: {_pg_imp_err}")
+
+PURCHASE_UI_OPERATION_TIMEOUT_MS = 5000
+PURCHASE_SKU_CHECK_TIMEOUT_SECONDS = 45
+PURCHASE_MAX_RESULT_ROWS_PER_SKU = 5
+
+class ERPLoginManager:
+    USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/143.0.0.0 Safari/537.36"
+
+    def __init__(self, config_path, cdp_port=None):
+        self.config = self._load_config(config_path)
+        self.playwright_cookies = []
+        self.session = requests.Session()
+        self._playwright = None
+        self._browser = None
+        self._context = None
+        self._page = None
+        self._cdp_port = cdp_port  # CDP 端口，用于连接外部浏览器
+
+    def _load_config(self, config_path):
+        # 如果是相对路径且文件不存在，尝试在脚本所在目录下查找
+        if not os.path.isabs(config_path) and not os.path.exists(config_path):
+            script_dir = os.path.dirname(os.path.abspath(__file__))
+            potential_path = os.path.join(script_dir, config_path)
+            if os.path.exists(potential_path):
+                config_path = potential_path
+        
+        with open(config_path, 'r', encoding='utf-8') as f:
+            return json.load(f)
+
+    def _sync_session_from_context(self, page=None):
+        try:
+            if self._context:
+                self.playwright_cookies = self._context.cookies()
+            elif page is not None:
+                self.playwright_cookies = page.context.cookies()
+            else:
+                self.playwright_cookies = []
+        except Exception:
+            self.playwright_cookies = []
+
+        self.session.cookies.clear()
+        for cookie in self.playwright_cookies:
+            domain = cookie.get("domain", ".erp321.com")
+            path = cookie.get("path", "/")
+            self.session.cookies.set(cookie["name"], cookie["value"], domain=domain, path=path)
+
+    def _page_has_inventory_access(self, page, inventory_url):
+        denied_markers = ["forbidden", "permission", "denied", "\u65e0\u6743\u9650", "\u6ca1\u6709\u6743\u9650", "\u62d2\u7edd\u8bbf\u95ee"]
+        try:
+            page.goto(inventory_url, wait_until="domcontentloaded", timeout=15000)
+            try:
+                page.wait_for_load_state("load", timeout=10000)
+            except Exception:
+                pass
+            current_url = (page.url or "").lower()
+            page_text = (page.inner_text("body") or "").lower()
+            has_denied = any(marker in page_text for marker in denied_markers)
+            return ("goodsinventorymanagement" in current_url) and (not has_denied)
+        except Exception as e:
+            print(f"[WARN] inventory access probe failed: {e}")
+            return False
+
+    def _dismiss_post_login_dialogs(self, page, max_rounds=4, interval=1.0):
+        """
+        关闭登录后可能弹出的提示框（密码过期、公告等）。
+        检测「确定 / OK / 知道了 / 关闭」等按钮，点击后继续，不影响后续流程。
+        """
+        if not page:
+            return
+        confirm_keywords = [
+            "\u786e\u5b9a",  # 确定
+            "\u77e5\u9053\u4e86",  # 知道了
+            "\u5173\u95ed",  # 关闭
+            "\u597d\u7684",  # 好的
+            "\u662f",         # 是
+            "\u53d6\u6d88",  # 取消（部分弹框有「取消」作为关闭）
+            "ok",
+            "confirm",
+        ]
+        # 覆盖常见企业 UI 框架的弹层选择器
+        dialog_selectors = [
+            ".ant-modal",
+            ".ant-modal-wrap",
+            ".ant-message",
+            ".ant-notification",
+            ".el-dialog",
+            ".el-message-box",
+            ".el-message",
+            ".el-notification",
+            ".layui-layer",
+            ".layui-layer-dialog",
+            ".layui-layer-shade",
+            ".modal",
+            ".modal-dialog",
+            ".modal-content",
+            ".dialog",
+            "[role='dialog']",
+            "[role='alertdialog']",
+            ".swal2-container",
+            ".toast",
+            ".popup",
+            ".alert-dialog",
+        ]
+        for _ in range(max_rounds):
+            try:
+                # 1) 找任何可见的弹层
+                any_dialog = False
+                for sel in dialog_selectors:
+                    try:
+                        loc = page.locator(sel)
+                        if loc.count() > 0:
+                            for i in range(loc.count()):
+                                try:
+                                    if loc.nth(i).is_visible():
+                                        any_dialog = True
+                                        break
+                                except Exception:
+                                    continue
+                        if any_dialog:
+                            break
+                    except Exception:
+                        continue
+                if not any_dialog:
+                    time.sleep(interval)
+                    continue
+                # 2) 在弹层内点掉「确定」等按钮
+                clicked = False
+                for keyword in confirm_keywords:
+                    try:
+                        btn = page.get_by_role("button", name=keyword)
+                        if btn.count() > 0:
+                            for i in range(btn.count()):
+                                try:
+                                    if btn.nth(i).is_visible():
+                                        btn.nth(i).click(force=True)
+                                        clicked = True
+                                        print(f"[LOGIN] 自动点击弹框按钮: {keyword}")
+                                        break
+                                except Exception:
+                                    continue
+                            if clicked:
+                                break
+                    except Exception:
+                        continue
+                # 3) 通用兜底：弹层内任意 <button> / .ant-btn / .layui-layer-btn
+                if not clicked:
+                    for fallback_sel in [
+                        ".ant-modal button.ant-btn-primary",
+                        ".ant-modal button.ant-btn",
+                        ".el-dialog button.el-button--primary",
+                        ".el-dialog button.el-button",
+                        ".layui-layer .layui-layer-btn a",
+                        ".layui-layer button",
+                        ".modal-footer button",
+                        "dialog button",
+                        "[role='dialog'] button",
+                    ]:
+                        try:
+                            loc = page.locator(fallback_sel)
+                            if loc.count() > 0:
+                                for i in range(loc.count()):
+                                    try:
+                                        if loc.nth(i).is_visible():
+                                            loc.nth(i).click(force=True)
+                                            clicked = True
+                                            print(f"[LOGIN] 自动点击弹框按钮(fallback): {fallback_sel}")
+                                            break
+                                    except Exception:
+                                        continue
+                                if clicked:
+                                    break
+                        except Exception:
+                            continue
+                # 4) 终极兜底：按 ESC 关闭
+                if not clicked:
+                    try:
+                        page.keyboard.press("Escape")
+                    except Exception:
+                        pass
+                time.sleep(interval)
+            except Exception:
+                time.sleep(interval)
+
+    def _prepare_login_page(self, login_url):
+        if self._cdp_port:
+            print(f"[CDP] Connecting to browser on port {self._cdp_port}")
+            cdp_url = f"http://localhost:{self._cdp_port}"
+            proxy_vars = ["HTTP_PROXY", "HTTPS_PROXY", "ALL_PROXY", "http_proxy", "https_proxy", "all_proxy"]
+            saved_proxies = {k: os.environ.pop(k, None) for k in proxy_vars}
+            try:
+                self._browser = self._playwright.chromium.connect_over_cdp(cdp_url)
+            finally:
+                for key, value in saved_proxies.items():
+                    if value is not None:
+                        os.environ[key] = value
+
+            contexts = list(self._browser.contexts)
+            if contexts:
+                self._context = contexts[0]
+            else:
+                self._context = self._browser.new_context(user_agent=self.USER_AGENT, accept_downloads=True)
+
+            pages = list(self._context.pages)
+            if pages:
+                self._page = pages[0]
+                for extra_page in pages[1:]:
+                    try:
+                        extra_page.close()
+                    except Exception:
+                        pass
+                print(f"[CDP] Reusing page: {self._page.url}")
+            else:
+                self._page = self._context.new_page()
+                print("[CDP] Created a fresh page")
+        else:
+            headless_mode = self.config.get("login", {}).get("headless", False)
+            self._browser = self._playwright.chromium.launch(headless=headless_mode)
+            self._context = self._browser.new_context(user_agent=self.USER_AGENT, accept_downloads=True)
+            self._page = self._context.new_page()
+            print("[BROWSER] Started embedded browser")
+
+        self._page.goto(login_url, wait_until="domcontentloaded")
+        try:
+            self._page.wait_for_load_state("networkidle", timeout=10000)
+        except Exception:
+            pass
+        return self._page
+
+    def _fill_first_visible(self, page, selectors, value, label, mask=False):
+        for selector in selectors:
+            try:
+                locator = page.locator(selector).first
+                if locator.count() > 0 and locator.is_visible():
+                    locator.fill(value)
+                    print(f"[LOGIN] Filled {label}")
+                    return True
+            except Exception:
+                pass
+        if mask:
+            print(f"[WARN] Could not find {label} input")
+        else:
+            print(f"[WARN] Could not find {label} input: {value}")
+        return False
+
+    def login_with_playwright(self):
+        login_config = self.config.get("login", {})
+        url = login_config.get("url")
+        username = login_config.get("username")
+        password = os.getenv("GYXX_SUPPLY_ERP_PASSWORD", "")
+        inventory_url = "https://src.erp321.com/erp-web-group/erp-scm-goods/goodsInventoryManagement?tabAllow=camera&_c=jst-epaas"
+
+        if not url or not username or not password:
+            print("[FAIL] Missing login config in config.json")
+            return False
+
+        print(f"[LOGIN] Starting login flow: {url}")
+
+        try:
+            self._playwright = sync_playwright().start()
+            page = self._prepare_login_page(url)
+
+            try:
+                cookies = page.context.cookies()
+            except Exception:
+                cookies = []
+
+            has_session_cookie = any(
+                cookie.get("name", "") in ["ASP.NET_SessionId", ".ASPXAUTH", "token"]
+                for cookie in cookies
+            )
+            if has_session_cookie and self._page_has_inventory_access(page, inventory_url):
+                print("[LOGIN] Existing session is valid and has inventory access")
+                self._sync_session_from_context(page)
+                return True
+
+            print("[LOGIN] Performing a fresh login")
+            page.goto(url, wait_until="domcontentloaded")
+            time.sleep(2)
+
+            self._fill_first_visible(
+                page,
+                [
+                    'input[name="username"]',
+                    'input[name="user"]',
+                    'input[id="username"]',
+                    'input[type="text"]',
+                ],
+                username,
+                "username",
+            )
+            self._fill_first_visible(
+                page,
+                [
+                    'input[name="password"]',
+                    'input[name="pwd"]',
+                    'input[id="password"]',
+                    'input[type="password"]',
+                ],
+                password,
+                "password",
+                mask=True,
+            )
+
+            for selector in ['input[type="checkbox"]', 'input[name="agree"]', 'input[id="agree"]']:
+                try:
+                    locator = page.locator(selector)
+                    for idx in range(locator.count()):
+                        checkbox = locator.nth(idx)
+                        if checkbox.is_visible() and not checkbox.is_checked():
+                            checkbox.check()
+                except Exception:
+                    pass
+
+            button_clicked = False
+            for selector in [
+                'button[type="submit"]',
+                'input[type="submit"]',
+                'button:has-text("\u767b\u5f55")',
+                '#btnLogin',
+            ]:
+                try:
+                    locator = page.locator(selector).first
+                    if locator.count() > 0 and locator.is_visible():
+                        locator.click()
+                        button_clicked = True
+                        break
+                except Exception:
+                    pass
+            if not button_clicked:
+                page.keyboard.press("Enter")
+
+            print("[LOGIN] Waiting for login response")
+            # 点完登录按钮后立即尝试一次兜底，覆盖"弹框抢先于 sleep 出现"的场景
+            self._dismiss_post_login_dialogs(page, max_rounds=2, interval=0.3)
+            time.sleep(5)
+            # sleep 5 之后再兜底一次，覆盖"弹框晚于登录响应出现"的场景
+            self._dismiss_post_login_dialogs(page, max_rounds=4, interval=0.5)
+
+            captcha_waited = False
+            captcha_selectors = [
+                'div[id*="captcha"]',
+                'div[class*="captcha"]',
+                'img[class*="captcha"]',
+                'input[id*="captcha"]',
+                'input[placeholder*="\u9a8c\u8bc1\u7801"]',
+                'input[placeholder*="\u7801"]',
+                'div[class*="verify"]',
+                'div[id*="verify"]',
+                '.ant-captcha',
+                '.nc_wrapper',
+                '#nc__wrapper',
+                'iframe[src*="captcha"]',
+                'iframe[src*="verify"]',
+            ]
+            for attempt in range(12):
+                captcha_found = False
+                for selector in captcha_selectors:
+                    try:
+                        if page.locator(selector).count() > 0:
+                            captcha_found = True
+                            break
+                    except Exception:
+                        pass
+                if not captcha_found:
+                    if captcha_waited:
+                        print("[LOGIN] Captcha cleared")
+                    break
+                if not captcha_waited:
+                    print("[LOGIN] Captcha detected; waiting for manual completion")
+                    captcha_waited = True
+                print(f"[LOGIN] Waiting for captcha completion ({11 - attempt} checks left)")
+                time.sleep(5)
+
+            try:
+                if self._context and self._context.pages:
+                    self._page = self._context.pages[-1]
+                    page = self._page
+            except Exception:
+                pass
+
+            self._sync_session_from_context(page)
+            print(f"[LOGIN] Synced {len(self.playwright_cookies)} cookies")
+
+            if self._page_has_inventory_access(page, inventory_url):
+                print("[OK] Login succeeded and inventory page is accessible")
+                return True
+
+            print("[FAIL] Login finished but inventory page is still not accessible")
+            self.close_browser()
+            return False
+        except Exception as e:
+            print(f"[FAIL] Login flow crashed: {e}")
+            self.close_browser()
+            return False
+
+    def close_browser(self):
+        # 如果使用外部浏览器（CDP 连接），不关闭浏览器，只断开连接
+        if self._cdp_port:
+            try:
+                if self._browser:
+                    self._browser.close()
+                if self._playwright:
+                    self._playwright.stop()
+            except Exception:
+                pass
+            self._browser = None
+            self._context = None
+            self._page = None
+            self._playwright = None
+            print("[CDP] 已断开外部浏览器连接")
+            return
+        
+        try:
+            if self._browser:
+                self._browser.close()
+            if self._playwright:
+                self._playwright.stop()
+        except Exception:
+            pass
+        self._browser = None
+        self._context = None
+        self._page = None
+        self._playwright = None
+
+class PurchaseManager(ERPLoginManager):
+    def __init__(self, config_path, cdp_port=None):
+        super().__init__(config_path, cdp_port=cdp_port)
+        self._base_height_adjusted = False
+
+    def _set_fast_ui_timeouts(self):
+        """Keep per-SKU purchase-page checks from waiting on Playwright defaults."""
+        try:
+            if self._page:
+                self._page.set_default_timeout(PURCHASE_UI_OPERATION_TIMEOUT_MS)
+                self._page.set_default_navigation_timeout(PURCHASE_UI_OPERATION_TIMEOUT_MS)
+        except Exception as e:
+            print(f"[WARN] 设置采购页面短超时失败: {e}")
+
+    def _sku_deadline_exceeded(self, deadline, target_sku, stage):
+        if time.monotonic() > deadline:
+            print(f"[WARN] SKU {target_sku} 查询超过 {PURCHASE_SKU_CHECK_TIMEOUT_SECONDS} 秒，阶段：{stage}，跳过该 SKU")
+            return True
+        return False
+
+    def open_purchase_page(self):
+        if not self._page:
+            print("浏览器未启动，请先登录")
+            return
+        
+        purchase_url = "https://www.erp321.com/app/scm/purchase/purchasemode.aspx?_c=jst-epaas"
+        print(f"正在打开采购单管理页面: {purchase_url}")
+        self._set_fast_ui_timeouts()
+        self._page.goto(purchase_url, wait_until="domcontentloaded", timeout=15000)
+        print("页面已加载")
+
+        try:
+            time.sleep(2)
+            for frame in self._page.frames:
+                if frame.locator("#animationProcessWrapper").count() > 0:
+                    print("正在关闭 animationProcessWrapper 弹窗...")
+                    frame.evaluate('document.getElementById("animationProcessWrapper").style.display = "none"')
+                    print("弹窗已关闭")
+        except Exception as e:
+            print(f"关闭弹窗时出现轻微异常（可能已自动消失）: {e}")
+
+    def search_sku(self, sku_id, max_retries=1, deadline=None):
+        """搜索 SKU，支持重试机制，失败时打印警告并静默返回，不抛出异常"""
+        try:
+            if not self._page:
+                print("浏览器未启动")
+                return False
+            if deadline is None:
+                deadline = time.monotonic() + PURCHASE_SKU_CHECK_TIMEOUT_SECONDS
+
+            page = self._page
+            self._set_fast_ui_timeouts()
+            print(f"\n开始在页面搜索 SKU: {sku_id}（最多重试 {max_retries + 1} 次）")
+
+            input_selectors = [
+                'input[id*="SkuId"]',
+                'input[id*="skuId"]',
+                'input[name*="SkuId"]',
+                'input[placeholder*="商品"]',
+                'input[placeholder*="SKU"]'
+            ]
+
+            for attempt in range(max_retries + 1):
+                if self._sku_deadline_exceeded(deadline, sku_id, "搜索前"):
+                    return False
+                if attempt > 0:
+                    print(f"\n[重试 {attempt}/{max_retries}] 再次搜索 SKU: {sku_id}")
+                    time.sleep(1)
+
+                # 先尝试调整底座高度（每次重试都可能需要）
+                self._adjust_base_height()
+
+                found = False
+                target_frame = None
+                target_selector = None
+
+                for frame in page.frames:
+                    for selector in input_selectors:
+                        try:
+                            if frame.locator(selector).count() > 0:
+                                target_frame = frame
+                                target_selector = selector
+                                found = True
+                                break
+                        except Exception:
+                            continue
+                    if found:
+                        break
+
+                if not found:
+                    print(f"[尝试 {attempt + 1}] 未找到商品输入框，请检查页面结构")
+                    continue
+
+                # 清空输入框再填入（防止残留值干扰）
+                try:
+                    target_frame.fill(target_selector, "")
+                    target_frame.fill(target_selector, str(sku_id))
+                    print(f"[尝试 {attempt + 1}] 已填写 SKU: {sku_id}（selector: {target_selector}）")
+                except Exception as fill_err:
+                    print(f"[尝试 {attempt + 1}] 填写输入框失败: {fill_err}")
+                    continue
+
+                # 点击搜索按钮
+                search_clicked = False
+                for frame in page.frames:
+                    if self._sku_deadline_exceeded(deadline, sku_id, "点击搜索"):
+                        return False
+                    try:
+                        search_btn = frame.locator('a.search[onclick*="Search()"]')
+                        if search_btn.count() > 0:
+                            search_btn.first.click(timeout=PURCHASE_UI_OPERATION_TIMEOUT_MS)
+                            search_clicked = True
+                            print(f"[尝试 {attempt + 1}] 已点击搜索按钮")
+                            break
+                    except Exception:
+                        continue
+
+                if not search_clicked:
+                    # 尝试按回车
+                    try:
+                        target_frame.locator(target_selector).press("Enter", timeout=PURCHASE_UI_OPERATION_TIMEOUT_MS)
+                        print(f"[尝试 {attempt + 1}] 搜索按钮未找到，已按回车")
+                    except Exception:
+                        print(f"[尝试 {attempt + 1}] 点击搜索和按回车均失败")
+                        continue
+
+                # 等待搜索结果加载（通过检查 .rowList div[index] 是否出现）
+                print(f"[尝试 {attempt + 1}] 等待搜索结果...")
+                result_found = False
+                for wait_time in [2, 3, 4]:
+                    if self._sku_deadline_exceeded(deadline, sku_id, "等待搜索结果"):
+                        return False
+                    time.sleep(wait_time)
+                    for frame in page.frames:
+                        try:
+                            if frame.locator('.rowList div[index]').count() > 0:
+                                row_count = frame.locator('.rowList div[index]').count()
+                                print(f"[尝试 {attempt + 1}] 搜索结果已加载，共 {row_count} 行")
+                                result_found = True
+                                break
+                        except Exception:
+                            continue
+                    if result_found:
+                        break
+
+                if result_found:
+                    print(f"SKU {sku_id} 搜索成功")
+                    return True
+                else:
+                    print(f"[尝试 {attempt + 1}] 等待后仍未见搜索结果，将重试")
+
+            print(f"[WARN] SKU {sku_id} 搜索失败（已重试 {max_retries + 1} 次），继续后续流程")
+            return False
+
+        except Exception as e:
+            print(f"搜索 SKU 时出错（不影响后续流程）: {e}")
+            return False
+
+    def _adjust_base_height(self):
+        """Best-effort layout fix for the purchase detail panel."""
+        try:
+            page = self._page
+            if self._try_expand_purchase_detail_area():
+                self._base_height_adjusted = True
+                return True
+
+            for _ in range(2):  # 最多尝试2次拖拽
+                foot_found = False
+                for f in page.frames:
+                    try:
+                        foot = f.locator('#_jt_row_foot')
+                        if foot.count() > 0:
+                            box = foot.first.bounding_box()
+                            if box:
+                                # 拖拽 400px（比之前更大）
+                                start_x = box['x'] + box['width'] / 2
+                                start_y = box['y'] + box['height'] / 2
+                                print("[底座调整] 找到 #_jt_row_foot，向上拖拽 400px...")
+                                page.mouse.move(start_x, start_y)
+                                page.mouse.down()
+                                page.mouse.move(start_x, max(start_y - 400, 100), steps=15)
+                                page.mouse.up()
+                                print("[底座调整] 完成")
+                                foot_found = True
+                                time.sleep(1)
+                                if self._try_expand_purchase_detail_area():
+                                    self._base_height_adjusted = True
+                                    return True
+                                break
+                    except Exception:
+                        continue
+                if foot_found:
+                    self._base_height_adjusted = True
+                    return True
+                # 没找到就等一下再试
+                time.sleep(1)
+        except Exception as e:
+            print(f"[底座调整] 出错（不影响主流程）: {e}")
+        return False
+
+    def _try_expand_purchase_detail_area(self):
+        """Use DOM style adjustments as a fallback when mouse dragging is flaky."""
+        if not self._page:
+            return False
+
+        script = """
+        () => {
+            let touched = false;
+            const selectors = [
+                '#_jt_row_foot',
+                '.rowList',
+                '[id*="row_foot"]',
+                '[class*="detail"]',
+                '[class*="datagrid"]',
+                '[id*="tab0"]'
+            ];
+            for (const selector of selectors) {
+                for (const el of document.querySelectorAll(selector)) {
+                    el.style.maxHeight = 'none';
+                    if (selector !== '#_jt_row_foot') {
+                        el.style.minHeight = el.style.minHeight || '120px';
+                    }
+                    el.style.overflow = 'visible';
+                    el.scrollIntoView({ block: 'nearest', inline: 'nearest' });
+                    touched = true;
+                }
+            }
+            window.dispatchEvent(new Event('resize'));
+            return touched;
+        }
+        """
+
+        touched_any = False
+        for frame in self._page.frames:
+            try:
+                touched_any = bool(frame.evaluate(script)) or touched_any
+            except Exception:
+                continue
+
+        if touched_any:
+            print("[底座调整] 已通过 DOM 布局修正尝试展开采购明细区")
+        return touched_any
+
+    def _find_purchase_detail_frame(self, page):
+        for frame in page.frames:
+            try:
+                frame_id = frame.evaluate("() => window.frameElement ? window.frameElement.id : ''")
+                if "tab0" in str(frame_id) or "tab0" in frame.name:
+                    return frame
+            except Exception:
+                continue
+        for frame in page.frames:
+            try:
+                if frame.locator('[col_id="delivery_date"]').count() > 0:
+                    return frame
+            except Exception:
+                continue
+        return None
+
+    def check_delivery_date_exist(self, target_sku):
+        """检查 SKU 是否有协议到货时间；任何步骤出错都视为"找不到"，不崩溃"""
+        try:
+            if not self._page:
+                return (False, "", False)
+
+            deadline = time.monotonic() + PURCHASE_SKU_CHECK_TIMEOUT_SECONDS
+            self._set_fast_ui_timeouts()
+            search_ok = self.search_sku(target_sku, deadline=deadline)
+            if self._sku_deadline_exceeded(deadline, target_sku, "搜索完成"):
+                return (False, "", False)
+            page = self._page
+
+            search_frame = None
+            for frame in page.frames:
+                if self._sku_deadline_exceeded(deadline, target_sku, "定位搜索结果框架"):
+                    return (False, "", False)
+                try:
+                    if frame.locator('.rowList div[index]').count() > 0:
+                        search_frame = frame
+                        break
+                except Exception:
+                    continue
+
+            if not search_frame:
+                # 未找到结果框架 = 该 SKU 在采购系统中无记录（不存在采购日期）
+                print(f"SKU {target_sku} 未找到采购记录，视为无协议到货时间（search_ok={search_ok}）")
+                return (False, "", False)
+
+            rows = search_frame.locator('.rowList div[index]')
+            row_count = rows.count()
+            if row_count == 0:
+                return (False, "", False)
+            if row_count > PURCHASE_MAX_RESULT_ROWS_PER_SKU:
+                print(f"SKU {target_sku} 搜索结果 {row_count} 行，仅检查前 {PURCHASE_MAX_RESULT_ROWS_PER_SKU} 行以避免卡住")
+                row_count = PURCHASE_MAX_RESULT_ROWS_PER_SKU
+
+            for i in range(row_count):
+                try:
+                    if self._sku_deadline_exceeded(deadline, target_sku, f"处理第 {i} 行"):
+                        return (False, "", False)
+                    current_row = search_frame.locator(f'.rowList div[index="{i}"]')
+                    if current_row.count() == 0:
+                        current_row = rows.nth(i)
+                    
+                    current_row.click(timeout=PURCHASE_UI_OPERATION_TIMEOUT_MS)
+                    time.sleep(1)
+                    self._try_expand_purchase_detail_area()
+
+                    if self._sku_deadline_exceeded(deadline, target_sku, "定位明细框架"):
+                        return (False, "", False)
+                    detail_frame = self._find_purchase_detail_frame(page)
+                     
+                    if detail_frame:
+                        try:
+                            sku_element = detail_frame.get_by_text(str(target_sku), exact=False).first
+                            if sku_element.count() > 0:
+                                target_row = sku_element.locator('xpath=ancestor::tr | ancestor::div[contains(@class, "row")]').first
+                            else:
+                                target_row = detail_frame.locator('tr, .datagrid-row, div[class*="row"]').first
+
+                            date_cell = target_row.locator('[col_id="delivery_date"]').first
+                            if date_cell.count() > 0:
+                                date_text = date_cell.inner_text(timeout=PURCHASE_UI_OPERATION_TIMEOUT_MS).strip()
+                                val = ""
+                                input_box = date_cell.locator('input')
+                                if input_box.count() > 0:
+                                    val = input_box.first.input_value(timeout=PURCHASE_UI_OPERATION_TIMEOUT_MS).strip()
+                                
+                                date_str_found = val if val else date_text
+                                
+                                if date_str_found:
+                                    try:
+                                        date_part = date_str_found.split(' ')[0]
+                                        parsed_date = datetime.datetime.strptime(date_part, "%Y-%m-%d").date()
+                                        current_date = datetime.datetime.now().date()
+                                        
+                                        if current_date <= parsed_date:
+                                            print(f"SKU {target_sku} 已有有效到货日期: {date_str_found} (>= {current_date})")
+                                            return (True, date_str_found, False)
+                                        else:
+                                            print(f"SKU {target_sku} 到货日期 {date_str_found} 已过期 (< {current_date})")
+                                            return (True, date_str_found, True)
+                                    except Exception as e:
+                                        print(f"日期解析失败 '{date_str_found}': {e}")
+                                        return (True, date_str_found, False)
+                        except Exception as e:
+                            print(f"检查该行细节时出错: {e}")
+                except Exception as row_err:
+                    print(f"  第 {i} 行处理异常，跳过: {row_err}")
+                    continue
+            return (False, "", False)
+        except Exception as e:
+            print(f"SKU {target_sku} 检查到货时间时出错（将视为无协议到货时间，不计算超时）: {e}")
+            return (False, "", False)
+
+def clean_property(value):
+    if not value or pd.isna(value):
+        return ""
+    val_str = str(value)
+    val_str = val_str.replace('；', ';')
+    parts = [p.strip() for p in val_str.split(';')]
+    parts = [re.sub(r'[a-zA-Z0-9\s]', '', p) for p in parts]
+    parts = [p for p in parts if p]
+    parts.sort()
+    return ';'.join(parts)
+
+def process_excel(file_path, purchase_manager=None):
+    if not os.path.exists(file_path):
+        print(f"文件不存在: {file_path}")
+        return None
+    # 2026-06-22 v3:返回 (csv_path, rows) 供 main() 落 PG。
+    # 调用方忽略返回值也不会破坏既有逻辑(原有调用栈只用 None 判定错误)。
+
+    print(f"\n=== 正在处理 Excel 数据: {file_path} ===")
+    try:
+        # 使用 openpyxl 读取（data_only=True 读取公式的计算结果值）
+        # 注意：pd.read_excel() 不支持 data_only 参数，需用 openpyxl load_workbook 再转 DataFrame
+        try:
+            wb = openpyxl.load_workbook(file_path, data_only=True)
+            ws = wb.active
+            data = ws.values
+            if data:
+                headers = next(data)
+                df = pd.DataFrame(data, columns=headers)
+            else:
+                df = pd.DataFrame()
+            wb.close()
+        except Exception as read_err:
+            print(f"  [错误] openpyxl data_only 模式读取失败: {read_err}，尝试普通模式...")
+            df = pd.read_excel(file_path, engine='openpyxl')
+        columns = df.columns.tolist()
+
+        col_name = "商品名称"
+        col_props = "规格名称"
+        col_stock = "可用数"
+        col_sales_7 = "7天销量"
+        col_purchase = "采购在途数"
+        col_sku = "商品编码"
+
+        for col in columns:
+            c = col.strip()
+            if c in ["商品名称"]: col_name = col
+            if c in [ "颜色及规格"]: col_props = col
+            if c in [ "可用数"]: col_stock = col
+            if c in ["7天销量"]: col_sales_7 = col
+            if c in ["采购在途数"]: col_purchase = col
+            if c in ["商品编码"]: col_sku = col
+
+        if all(c in df.columns for c in [col_name, col_props, col_stock, col_sales_7, col_purchase, col_sku]):
+            exclude_names = ["星悦","邮费连接","无纺袋","飞机盒","飞机盒外箱","OZKO 无纺袋","OZKO 飞机盒","星迹","补发","配件包","阿波罗（新无线充电）","阿波罗无U盘","暗物质A1","白鹤","百灵","波塞冬2","波塞冬电脑袋","晨星mini","池鹭air","充电背包简约版","盖世","盖世P1","盖世T科技版","盖亚Neo","盖亚Ultra","盖亚V","盖亚X","盖亚泡芙相机包","盖亚双肩2","盖亚斜挎2","黑啤","黑啤拉链款","横版单肩斜挎包","横版动物帆布包","横版女孩帆布包","幻月","极星手提","极星双肩Pura","剑道单肩包","剑道升级版","扩容手提包（中号）","蓝鹊","蓝莺","冷感便携毛巾","理岩a","灵速","流云","露娜","露娜FIT","洛基","买家秀晒图虚拟商品","木兰背包","牛皮","千鹤","轻旅包","轻云","轻云S","数码收纳配件包","无纺袋（HK）","五金扣","弦月","鞋袋","星舰","星蓝","星旅","星源","星云","星云a","星云皮膜版a","星云无线充电升级版","钥匙扣","游云简约版","雨衣","云朵","云卷背包","云卷多用包","斩月","知更","知更Pro","宙斯","宙斯Z","宙斯Z扩展","宙斯扩展包","逐星EVO","牛皮电源包","PU电源包","双口袋手提包包","无线鼠标","专业版斜跨小方包","YSZ麻布手提升级版","捕梦者","横版飞机帆布包","邂逅专业版","蓝莓手提包","横版单肩口袋包新","拼色斜跨包","兵人包升级版","卷轴","卷轴（肩带）","switch猫爪摇杆帽","气泡袋","垂柳拉链版（无小袋）","昼夜者","保温杯","超凡者","夜游者","云帆帆布","光影雨伞","OZKO 抽绳包","盖亚微单版","白羽","帆布（猫抓鱼）","帆布（狗与骨头）","帆布（二哈骨头）","光影行星咖啡杯","星舰配件包","盖亚斜挎相机包a","盖世sport专业版","灵速隔板","星舰相机隔层配件","轻云户外版","轻云斜挎","礼盒","内托","ozko羽毛球挂件","磁吸礼盒","磁吸外箱","光影礼袋","IC卡卡套","IC集卡","化妆收纳包","云雀"]
+            
+            initial_count = len(df)
+            df = df[~df[col_name].isin(exclude_names)].copy()
+            df = df[~df[col_props].astype(str).str.contains("HK", case=False, na=False)].copy()
+            
+            print(f"预过滤完成：从 {initial_count} 行减少到 {len(df)} 行 (排除了特定名称及带 HK 的规格)")
+
+            print("正在清洗商品名称和规格属性...")
+            df['清洗后的名称'] = df[col_name].astype(str).str.replace(r'\s+', '', regex=True)
+            df['清洗后的规格'] = df[col_props].apply(clean_property)
+            
+            print("按 [清洗后的名称] + [清洗后的规格] 进行多维度汇总...")
+            
+            df['原始详情'] = df[col_name].fillna('') + "[" + df[col_props].fillna('') + "]"
+            
+            df[col_stock] = pd.to_numeric(df[col_stock], errors='coerce').fillna(0)
+            df[col_sales_7] = pd.to_numeric(df[col_sales_7], errors='coerce').fillna(0)
+            df[col_purchase] = pd.to_numeric(df[col_purchase], errors='coerce').fillna(0)
+            
+            if col_sku in df.columns:
+                df[col_sku] = df[col_sku].astype(str).replace(['nan', 'None', ''], pd.NA)
+
+            agg_dict = {
+                col_stock: 'sum',
+                col_sales_7: 'sum',
+                col_purchase: 'sum',
+                '原始详情': lambda x: ' / '.join(sorted(list(set([str(i) for i in x if i]))))
+            }
+            if col_sku in df.columns:
+                agg_dict[col_sku] = lambda x: ','.join(sorted(list(set([str(i) for i in x if pd.notna(i)]))))
+
+            summary_df = df.groupby(['清洗后的名称', '清洗后的规格']).agg(agg_dict).reset_index()
+            
+            summary_df.rename(columns={'原始详情': '汇总原始数据'}, inplace=True)
+            
+            summary_df['缺口'] = summary_df[col_stock] - summary_df[col_sales_7]
+            filtered_summary = summary_df[
+                ((summary_df['缺口'] < 0) | (summary_df[col_stock] == 0)) & 
+                (summary_df[col_purchase] > 0)
+            ].copy()
+            
+            if not filtered_summary.empty:
+                filtered_summary = filtered_summary.sort_values(by='缺口')
+                
+                cols = filtered_summary.columns.tolist()
+                if col_sku in cols:
+                    cols.remove(col_sku)
+                    cols.insert(0, col_sku)
+                filtered_summary = filtered_summary[cols].copy()
+
+                if purchase_manager and col_sku in filtered_summary.columns:
+                    print("\n正在通过UI检查采购明细是否已有协议到货时间...")
+                    purchase_manager.open_purchase_page()
+                    delivery_dates = []
+                    to_keep = []
+                    for index, row in filtered_summary.iterrows():
+                        skus_str = row[col_sku]
+                        skus = [s.strip() for s in str(skus_str).split(',') if s.strip()]
+                        
+                        has_date = False
+                        is_expired = False
+                        date_str = ""
+                        for sku in skus:
+                            try:
+                                has_date, date_str, is_expired = purchase_manager.check_delivery_date_exist(sku)
+                            except Exception as e:
+                                print(f"  SKU {sku} 查询异常，视为无协议到货时间: {e}")
+                                has_date, date_str, is_expired = False, "", False
+                            if has_date:
+                                break
+                        
+                        delivery_dates.append(date_str)
+                        
+                        if not has_date or is_expired:
+                            to_keep.append(index)
+                    
+                    filtered_summary['协议到货时间'] = delivery_dates
+                    filtered_summary = filtered_summary.loc[to_keep].copy()
+                    print(f"UI检查完成，保留需要预警的记录共 {len(filtered_summary)} 条")
+
+                if filtered_summary.empty:
+                    print("\n[结果] 过滤已有到货日期的商品后，没有需要预警的记录。")
+                    return (None, [])
+
+                timestamp = time.strftime("%Y%m%d_%H%M%S")
+                script_dir = os.path.dirname(os.path.abspath(__file__))
+                data_dir = str(PURCHASE_CONFIRMATION_EXPORT_ROOT)
+                if not os.path.exists(data_dir): os.makedirs(data_dir)
+                output_path = os.path.join(data_dir, f"采购确认通知汇总_{timestamp}.csv")
+                if col_sku in filtered_summary.columns:
+                    filtered_summary[col_sku] = filtered_summary[col_sku].astype(str).apply(lambda x: f"'{x}")
+                filtered_summary.to_csv(output_path, index=False, encoding='utf-8-sig')
+
+                print(f"\n[成功] 发现 {len(filtered_summary)} 条预警记录 (含在途数据)")
+                print(f"[完成] 汇总结果已保存至: {output_path}")
+
+                print("\n=== 预警告警表格 (Top 10) ===")
+                print("筛选条件：(库存 - 7天销量 < 0 或 库存 = 0) 且 采购在途 > 0")
+                print(filtered_summary.head(10).to_string(index=False))
+
+                # 2026-06-22 v3:把过滤后的明细转成 dict rows 供 main() 落 PG。
+                # dict keys 同时兼容英文字段名和中文列名(与 pg_writer.insert_purchase_confirmation 对齐)。
+                pg_rows: list[dict] = []
+                for _, _row in filtered_summary.iterrows():
+                    sku_raw = _row.get(col_sku, "")
+                    sku_str = str(sku_raw).strip().lstrip("'") if sku_raw is not None else ""
+                    pg_rows.append({
+                        "sku": sku_str,
+                        "商品编码": sku_str,
+                        "product_name": _row.get(col_name, ""),
+                        "清洗后的名称": _row.get("清洗后的名称", _row.get(col_name, "")),
+                        "spec": _row.get(col_props, ""),
+                        "清洗后的规格": _row.get("清洗后的规格", ""),
+                        "available_stock": int(pd.to_numeric(_row.get(col_stock, 0), errors="coerce") or 0),
+                        "可用数": int(pd.to_numeric(_row.get(col_stock, 0), errors="coerce") or 0),
+                        "sales_7d": int(pd.to_numeric(_row.get(col_sales_7, 0), errors="coerce") or 0),
+                        "7天销量": int(pd.to_numeric(_row.get(col_sales_7, 0), errors="coerce") or 0),
+                        "purchase_in_transit": int(pd.to_numeric(_row.get(col_purchase, 0), errors="coerce") or 0),
+                        "采购在途数": int(pd.to_numeric(_row.get(col_purchase, 0), errors="coerce") or 0),
+                        "gap": int(pd.to_numeric(_row.get("缺口", 0), errors="coerce") or 0),
+                        "缺口": int(pd.to_numeric(_row.get("缺口", 0), errors="coerce") or 0),
+                        "delivery_date": _row.get("协议到货时间", "") or "",
+                        "协议到货时间": _row.get("协议到货时间", "") or "",
+                        "summary_raw": _row.get("汇总原始数据", "") or "",
+                        "汇总原始数据": _row.get("汇总原始数据", "") or "",
+                    })
+                return (output_path, pg_rows)
+            else:
+                print("\n[结果] 汇总后未发现符合条件 (可用库存 - 7天销量 < 0 或 库存 = 0) 且 有在途 的情况。")
+                return (None, [])
+        else:
+            print(f"[错误] 无法在 Excel 中找到匹配的列。现有列为: {columns}")
+            return None
+
+    except Exception as e:
+        print(f"[错误] 处理 Excel 时出错: {e}")
+        raise  # 重新抛出，避免静默退出
+
+def _persist_run_to_pg(process_result) -> None:
+    """2026-06-22 v3:采集端落 PG(wf_runs + fact_purchase_confirmation)。
+
+    设计要点:
+      1) run_id 优先用编排层注入的 AUTOFLOW_RUN_ID(env),与 mcp_workflow.py 的
+         resolved_workflow_id 对齐,让 ON CONFLICT (run_id) DO UPDATE 能命中
+         orchestrator 后续的 upsert_run 调用;env 缺失时自生成。
+      2) purchase-confirmation 没有多维表(analyzer_scripts 只剩 im.send_message),
+         所以跳过 acquire_bitable_insert,也不会写 pending_insert.json。
+      3) 全程 try/except,任何 PG 失败只 warn,不抛——shared-data/ 已经有 CSV 兜底。
+    """
+    rows: list[dict] = []
+    csv_path: str | None = None
+    if isinstance(process_result, tuple) and len(process_result) == 2:
+        csv_path, rows = process_result
+    elif isinstance(process_result, list):
+        rows = process_result
+    rows = rows or []
+    record_count = len(rows)
+
+    run_id = os.getenv("AUTOFLOW_RUN_ID", "").strip() or (
+        f"purchase-confirmation_{time.strftime('%Y%m%dT%H%M%S')}_{secrets.token_hex(3)}"
+    )
+    trigger_time_iso = (
+        time.strftime("%Y-%m-%dT%H:%M:%S%z")
+        or time.strftime("%Y-%m-%dT%H:%M:%S")
+    )
+    end_time_iso = time.strftime("%Y-%m-%dT%H:%M:%S%z") or time.strftime("%Y-%m-%dT%H:%M:%S")
+
+    source_files: list[str] = []
+    if csv_path:
+        source_files.append(str(csv_path))
+
+    if not _PG_IMPORT_OK:
+        print(f"[WARN] pg_writer 不可用,跳过 PG 落库 (run_id={run_id}, records={record_count})")
+        return
+
+    try:
+        pg.upsert_run(
+            run_id=run_id,
+            workflow="purchase-confirmation",
+            trigger_time=trigger_time_iso,
+            end_time=end_time_iso,
+            status="completed" if rows else "empty",
+            error="",
+            source_files=source_files,
+            raw_payload={
+                "record_count": record_count,
+                "csv_path": csv_path,
+            },
+        )
+    except Exception as e:
+        print(f"[WARN] PG 落库失败(非阻塞,upsert_run): {e}")
+        # workflow_runs 写入失败的话,后续 fact_ 写入也没意义,直接返回
+        return
+
+    if rows:
+        try:
+            data_hash = pg.compute_data_hash(rows, key="sku")
+            inserted = pg.insert_purchase_confirmation(run_id, rows)
+            print(
+                f"[PG] run_id={run_id} data_hash={data_hash[:12]}... "
+                f"detail_inserted={inserted} (purchase-confirmation, no bitable)"
+            )
+        except Exception as e:
+            print(f"[WARN] PG 落库失败(非阻塞,insert_purchase_confirmation): {e}")
+
+    # purchase-confirmation 没有多维表流程,不调 acquire_bitable_insert、不写 pending_insert.json
+    # 编排层 _extract_pending_run_id 会回退到 resolved_workflow_id,与本函数使用的 run_id 一致。
+
+
+def main(cdp_port=None):
+    """执行采购到货确认流程
+    
+    Args:
+        cdp_port: CDP 端口号，如果指定则连接外部浏览器（由 hermes browser 工具打开）
+                  如果为 None，则使用内置 playwright 启动浏览器
+    """
+    # 设置控制台输出编码为 utf-8
+    if sys.platform.startswith('win'):
+        import io
+        # 避免重复包装（subprocess capture_output 时 sys.stdout 已是 TextIOWrapper）
+        if not isinstance(sys.stdout, io.TextIOWrapper) or sys.stdout.closed:
+            sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8', errors='replace')
+    
+    manager = PurchaseManager(str(CONFIG_PATH), cdp_port=cdp_port)
+    try:
+        print("正在执行登录流程...")
+        if not manager.login_with_playwright():
+            print("登录失败。")
+            sys.exit(1)
+            
+        # 访问商品选择器页面
+        print("\n正在访问商品选择器页面...")
+        manager._page.goto("https://src.erp321.com/erp-web-group/erp-scm-goods/goodsInventoryManagement?tabAllow=camera&_c=jst-epaas", wait_until="networkidle")
+        time.sleep(3)
+        print("商品选择器页面已加载")
+        
+        # 尝试通过 UI 点击导出
+        try:
+            print("\n正在通过 UI 执行导出流程...")
+            page = manager._page
+            
+            # 1. 寻找导出按钮
+            # 使用用户提供的 class 和 text 组合定位
+            export_btn_selector = "button.goods-dropdown-trigger:has-text('导出')"
+            page.wait_for_selector(export_btn_selector, timeout=15000)
+            export_btn = page.locator(export_btn_selector)
+            
+            if export_btn.count() == 0:
+                print("未找到导出按钮")
+                sys.exit(1)
+
+            print("点击'导出'按钮...")
+            export_btn.click()
+            time.sleep(2) # 等待列表弹出并稳定
+
+            # 2. 点击下拉列表中的"成品"菜单项
+            # Ant Design 下拉菜单可能有动画/遮挡，使用 force=True 绕过可见性检查
+            menu_item_selector = ".ant-dropdown-menu-item, .ant-dropdown-menu li, li[role='menuitem']"
+            
+            # 使用 expect_download 捕获下载
+            print("等待下载触发...")
+            try:
+                with page.expect_download(timeout=60000) as download_info:
+                    # 等待菜单项出现
+                    page.wait_for_selector(menu_item_selector, timeout=5000)
+                    # 优先找"成品"菜单项
+                    chengpin_item = page.locator(menu_item_selector).filter(has_text="成品")
+                    if chengpin_item.count() > 0:
+                        print("点击菜单项: 成品")
+                        chengpin_item.first.click(force=True)
+                    else:
+                        # 回退: 寻找第一个有文本的菜单项
+                        first_item = page.locator(menu_item_selector).filter(has_text=re.compile(r".+")).first
+                        if first_item.count() > 0:
+                            print(f"点击菜单项: {first_item.inner_text().strip()}")
+                            first_item.click(force=True)
+                        else:
+                            print("未找到通用的下拉菜单项，尝试点击第一个可见的 li")
+                            page.locator("div.ant-dropdown:not(.ant-dropdown-hidden) li").first.click(force=True)
+
+                download = download_info.value
+                script_dir = os.path.dirname(os.path.abspath(__file__))
+                data_dir = str(PURCHASE_CONFIRMATION_RAW_ROOT)
+                if not os.path.exists(data_dir): os.makedirs(data_dir)
+                
+                # 构建保存路径
+                timestamp = time.strftime("%Y%m%d_%H%M%S")
+                suggested_name = download.suggested_filename
+                save_path = os.path.join(data_dir, f"{timestamp}_{suggested_name}")
+                
+                download.save_as(save_path)
+                print(f"\n[成功] 文件已下载至: {save_path}")
+
+                # 3. 处理下载后的文件
+                process_result = process_excel(save_path, manager)
+
+            except Exception as download_err:
+                print(f"下载捕获失败: {download_err}")
+                sys.exit(1)
+
+        except Exception as e:
+            print(f"UI 导出发生异常: {e}")
+            sys.exit(1)
+
+
+    except Exception as e:
+        print(f"执行异常: {e}")
+        sys.exit(1)
+    finally:
+        # 2026-06-22 v3:采集完落 PG(pg 是事实源,采集脚本是唯一写入方)。
+        # 即便 process_excel 抛了异常,也会先 upsert_run(status=failed)再 re-raise,
+        # 保证 workflow_runs 有完整的一行。
+        _persist_run_to_pg(process_result if 'process_result' in locals() else None)
+        manager.close_browser()
+
+if __name__ == "__main__":
+    import argparse
+    parser = argparse.ArgumentParser(description='采购到货确认')
+    parser.add_argument('--cdp-port', type=int, default=None, help='CDP 端口，用于连接外部浏览器')
+    args = parser.parse_args()
+    main(cdp_port=args.cdp_port)
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/PurchaseOrderUpdate.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/PurchaseOrderUpdate.py
new file mode 100644
index 0000000..ff13775
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/PurchaseOrderUpdate.py
@@ -0,0 +1,1011 @@
+#!/usr/bin/env python3
+# ruff: noqa: F841
+# -*- coding: utf-8 -*-
+"""
+采购单日期更新 - Auto-Flow 工作流版本
+用于更新 ERP 系统中采购单的协议到货日期，输出结果到 JSON 文件。
+"""
+
+import argparse
+import io
+import json
+import os
+import re
+import secrets
+import sys
+import time
+from datetime import datetime
+from pathlib import Path
+
+import requests
+from playwright.sync_api import sync_playwright
+
+from gyxx_flow.modules.supply_chain.runtime.paths import (
+    CONFIG_PATH,
+    WORK_ROOT,
+)
+
+# 让 `from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer` 在 `python PurchaseOrderUpdate.py` 这种
+# 不带 -m 的调用方式下也能解析(2026-06-22 v3 改造后,采集端要落 PG)。
+# 必须在所有 from gyxx_flow.modules.supply_chain.runtime.orchestrator import 之前。
+# 2026-06-22 v3:采集端直接落 PG(pg_writer 是软依赖,失败仅 warn)。
+try:
+    from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer as pg
+    _PG_IMPORT_OK = True
+except Exception as _pg_imp_err:  # pragma: no cover
+    _PG_IMPORT_OK = False
+    print(f"[WARN] pg_writer 不可用,本次不落 PG: {_pg_imp_err}")
+
+# 强制 UTF-8 输出
+try:
+    if not isinstance(sys.stdout, io.TextIOWrapper) or sys.stdout.closed:
+        sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8', errors='replace')
+except Exception:
+    pass
+try:
+    if not isinstance(sys.stderr, io.TextIOWrapper) or sys.stderr.closed:
+        sys.stderr = io.TextIOWrapper(sys.stderr.buffer, encoding='utf-8', errors='replace')
+except Exception:
+    pass
+
+PURCHASE_UI_OPERATION_TIMEOUT_MS = 5000
+PURCHASE_SKU_CHECK_TIMEOUT_SECONDS = 45
+PURCHASE_MAX_RESULT_ROWS_PER_SKU = 5
+
+
+def log_action(action, details=""):
+    """记录操作日志"""
+    timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    log_entry = f"[{timestamp}] {action}"
+    if details:
+        log_entry += f" - {details}"
+    print(f"[LOG] {action}" + (f" ({details})" if details else ""))
+
+
+class ERPLoginManager:
+    USER_AGENT = (
+        "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
+        "AppleWebKit/537.36 (KHTML, like Gecko) Chrome/143.0.0.0 Safari/537.36"
+    )
+
+    def __init__(self, config_path, cdp_port=None):
+        self.config = self._load_config(config_path)
+        self.playwright_cookies = []
+        self.session = requests.Session()
+        self._playwright = None
+        self._browser = None
+        self._context = None
+        self._page = None
+        self._cdp_port = cdp_port
+
+    def _load_config(self, config_path):
+        if not os.path.isabs(config_path) and not os.path.exists(config_path):
+            script_dir = os.path.dirname(os.path.abspath(__file__))
+            potential_path = os.path.join(script_dir, config_path)
+            if os.path.exists(potential_path):
+                config_path = potential_path
+        with open(config_path, 'r', encoding='utf-8') as f:
+            return json.load(f)
+
+    def _sync_session_from_context(self, page=None):
+        try:
+            if self._context:
+                self.playwright_cookies = self._context.cookies()
+            elif page is not None:
+                self.playwright_cookies = page.context.cookies()
+            else:
+                self.playwright_cookies = []
+        except Exception:
+            self.playwright_cookies = []
+
+        self.session.cookies.clear()
+        for cookie in self.playwright_cookies:
+            domain = cookie.get("domain", ".erp321.com")
+            path = cookie.get("path", "/")
+            self.session.cookies.set(cookie["name"], cookie["value"], domain=domain, path=path)
+
+    def _page_has_purchase_access(self, page, purchase_url):
+        denied_markers = ["forbidden", "permission", "denied", "\u65e0\u6743\u9650", "\u6ca1\u6709\u6743\u9650", "\u62d2\u7edd\u8bbf\u95ee"]
+        try:
+            page.goto(purchase_url, wait_until="domcontentloaded", timeout=15000)
+            try:
+                page.wait_for_load_state("load", timeout=10000)
+            except Exception:
+                pass
+            current_url = (page.url or "").lower()
+            page_text = (page.inner_text("body") or "").lower()
+            has_denied = any(marker in page_text for marker in denied_markers)
+            return ("purchasemode.aspx" in current_url) and (not has_denied)
+        except Exception as e:
+            print(f"[WARN] purchase page access probe failed: {e}")
+            return False
+
+    def _dismiss_post_login_dialogs(self, page, max_rounds=4, interval=1.0):
+        """
+        关闭登录后可能弹出的提示框（密码过期、公告等）。
+        检测「确定 / OK / 知道了 / 关闭」等按钮，点击后继续，不影响后续流程。
+        """
+        if not page:
+            return
+        confirm_keywords = [
+            "\u786e\u5b9a",  # 确定
+            "\u77e5\u9053\u4e86",  # 知道了
+            "\u5173\u95ed",  # 关闭
+            "\u597d\u7684",  # 好的
+            "\u662f",         # 是
+            "\u53d6\u6d88",  # 取消（部分弹框有「取消」作为关闭）
+            "ok",
+            "confirm",
+        ]
+        # 覆盖常见企业 UI 框架的弹层选择器
+        dialog_selectors = [
+            ".ant-modal",
+            ".ant-modal-wrap",
+            ".ant-message",
+            ".ant-notification",
+            ".el-dialog",
+            ".el-message-box",
+            ".el-message",
+            ".el-notification",
+            ".layui-layer",
+            ".layui-layer-dialog",
+            ".layui-layer-shade",
+            ".modal",
+            ".modal-dialog",
+            ".modal-content",
+            ".dialog",
+            "[role='dialog']",
+            "[role='alertdialog']",
+            ".swal2-container",
+            ".toast",
+            ".popup",
+            ".alert-dialog",
+        ]
+        for _ in range(max_rounds):
+            try:
+                # 1) 找任何可见的弹层
+                any_dialog = False
+                for sel in dialog_selectors:
+                    try:
+                        loc = page.locator(sel)
+                        if loc.count() > 0:
+                            for i in range(loc.count()):
+                                try:
+                                    if loc.nth(i).is_visible():
+                                        any_dialog = True
+                                        break
+                                except Exception:
+                                    continue
+                        if any_dialog:
+                            break
+                    except Exception:
+                        continue
+                if not any_dialog:
+                    time.sleep(interval)
+                    continue
+                # 2) 在弹层内点掉「确定」等按钮
+                clicked = False
+                for keyword in confirm_keywords:
+                    try:
+                        btn = page.get_by_role("button", name=keyword)
+                        if btn.count() > 0:
+                            for i in range(btn.count()):
+                                try:
+                                    if btn.nth(i).is_visible():
+                                        btn.nth(i).click(force=True)
+                                        clicked = True
+                                        print(f"[LOGIN] 自动点击弹框按钮: {keyword}")
+                                        break
+                                except Exception:
+                                    continue
+                            if clicked:
+                                break
+                    except Exception:
+                        continue
+                # 3) 通用兜底：弹层内任意 <button> / .ant-btn / .layui-layer-btn
+                if not clicked:
+                    for fallback_sel in [
+                        ".ant-modal button.ant-btn-primary",
+                        ".ant-modal button.ant-btn",
+                        ".el-dialog button.el-button--primary",
+                        ".el-dialog button.el-button",
+                        ".layui-layer .layui-layer-btn a",
+                        ".layui-layer button",
+                        ".modal-footer button",
+                        "dialog button",
+                        "[role='dialog'] button",
+                    ]:
+                        try:
+                            loc = page.locator(fallback_sel)
+                            if loc.count() > 0:
+                                for i in range(loc.count()):
+                                    try:
+                                        if loc.nth(i).is_visible():
+                                            loc.nth(i).click(force=True)
+                                            clicked = True
+                                            print(f"[LOGIN] 自动点击弹框按钮(fallback): {fallback_sel}")
+                                            break
+                                    except Exception:
+                                        continue
+                                if clicked:
+                                    break
+                        except Exception:
+                            continue
+                # 4) 终极兜底：按 ESC 关闭
+                if not clicked:
+                    try:
+                        page.keyboard.press("Escape")
+                    except Exception:
+                        pass
+                time.sleep(interval)
+            except Exception:
+                time.sleep(interval)
+
+    def _prepare_login_page(self, login_url):
+        if self._cdp_port:
+            print(f"[CDP] Connecting to browser on port {self._cdp_port}")
+            cdp_url = f"http://localhost:{self._cdp_port}"
+            proxy_vars = ["HTTP_PROXY", "HTTPS_PROXY", "ALL_PROXY", "http_proxy", "https_proxy", "all_proxy"]
+            saved_proxies = {k: os.environ.pop(k, None) for k in proxy_vars}
+            try:
+                self._browser = self._playwright.chromium.connect_over_cdp(cdp_url)
+            finally:
+                for key, value in saved_proxies.items():
+                    if value is not None:
+                        os.environ[key] = value
+
+            contexts = list(self._browser.contexts)
+            if contexts:
+                self._context = contexts[0]
+            else:
+                self._context = self._browser.new_context(user_agent=self.USER_AGENT)
+
+            pages = list(self._context.pages)
+            if pages:
+                self._page = pages[0]
+                for extra_page in pages[1:]:
+                    try:
+                        extra_page.close()
+                    except Exception:
+                        pass
+                print(f"[CDP] Reusing page: {self._page.url}")
+            else:
+                self._page = self._context.new_page()
+                print("[CDP] Created a fresh page")
+        else:
+            headless_mode = self.config.get("login", {}).get("headless", False)
+            self._browser = self._playwright.chromium.launch(headless=headless_mode)
+            self._context = self._browser.new_context(user_agent=self.USER_AGENT)
+            self._page = self._context.new_page()
+            print("[BROWSER] Started embedded browser")
+
+        self._page.goto(login_url, wait_until="domcontentloaded")
+        try:
+            self._page.wait_for_load_state("networkidle", timeout=10000)
+        except Exception:
+            pass
+        return self._page
+
+    def _fill_first_visible(self, page, selectors, value, label, mask=False):
+        for selector in selectors:
+            try:
+                locator = page.locator(selector).first
+                if locator.count() > 0 and locator.is_visible():
+                    locator.fill(value)
+                    print(f"[LOGIN] Filled {label}")
+                    return True
+            except Exception:
+                pass
+        if mask:
+            print(f"[WARN] Could not find {label} input")
+        else:
+            print(f"[WARN] Could not find {label} input: {value}")
+        return False
+
+    def login_with_playwright(self):
+        login_config = self.config.get("login", {})
+        url = login_config.get("url")
+        username = login_config.get("username")
+        password = os.getenv("GYXX_SUPPLY_ERP_PASSWORD", "")
+        purchase_url = "https://www.erp321.com/app/scm/purchase/purchasemode.aspx?_c=jst-epaas"
+
+        if not url or not username or not password:
+            print("[FAIL] Missing login config in config.json")
+            return False
+
+        print(f"[LOGIN] Starting login flow: {url}")
+
+        try:
+            self._playwright = sync_playwright().start()
+            page = self._prepare_login_page(url)
+
+            try:
+                cookies = page.context.cookies()
+            except Exception:
+                cookies = []
+
+            has_session_cookie = any(
+                cookie.get("name", "") in ["ASP.NET_SessionId", ".ASPXAUTH", "token"]
+                for cookie in cookies
+            )
+            if has_session_cookie and self._page_has_purchase_access(page, purchase_url):
+                print("[LOGIN] Existing session is valid and has purchase-page access")
+                self._sync_session_from_context(page)
+                return True
+
+            print("[LOGIN] Performing a fresh login")
+            page.goto(url, wait_until="domcontentloaded")
+            time.sleep(2)
+
+            self._fill_first_visible(
+                page,
+                [
+                    'input[name="username"]',
+                    'input[name="user"]',
+                    'input[id="username"]',
+                    'input[type="text"]',
+                ],
+                username,
+                "username",
+            )
+            self._fill_first_visible(
+                page,
+                [
+                    'input[name="password"]',
+                    'input[name="pwd"]',
+                    'input[id="password"]',
+                    'input[type="password"]',
+                ],
+                password,
+                "password",
+                mask=True,
+            )
+
+            for selector in ['input[type="checkbox"]', 'input[name="agree"]', 'input[id="agree"]']:
+                try:
+                    locator = page.locator(selector)
+                    for idx in range(locator.count()):
+                        checkbox = locator.nth(idx)
+                        if checkbox.is_visible() and not checkbox.is_checked():
+                            checkbox.check()
+                except Exception:
+                    pass
+
+            button_clicked = False
+            for selector in [
+                'button[type="submit"]',
+                'input[type="submit"]',
+                'button:has-text("\u767b\u5f55")',
+                '#btnLogin',
+            ]:
+                try:
+                    locator = page.locator(selector).first
+                    if locator.count() > 0 and locator.is_visible():
+                        locator.click()
+                        button_clicked = True
+                        break
+                except Exception:
+                    pass
+            if not button_clicked:
+                page.keyboard.press("Enter")
+
+            print("[LOGIN] Waiting for login response")
+            # 点完登录按钮后立即尝试一次兜底，覆盖"弹框抢先于 sleep 出现"的场景
+            self._dismiss_post_login_dialogs(page, max_rounds=2, interval=0.3)
+            time.sleep(5)
+            # sleep 5 之后再兜底一次，覆盖"弹框晚于登录响应出现"的场景
+            self._dismiss_post_login_dialogs(page, max_rounds=4, interval=0.5)
+
+            try:
+                if self._context and self._context.pages:
+                    self._page = self._context.pages[-1]
+                    page = self._page
+            except Exception:
+                pass
+
+            self._sync_session_from_context(page)
+            print(f"[LOGIN] Synced {len(self.playwright_cookies)} cookies")
+
+            if self._page_has_purchase_access(page, purchase_url):
+                print("[OK] Login succeeded and purchase page is accessible")
+                return True
+
+            print("[FAIL] Login finished but purchase page is still not accessible")
+            self.close_browser()
+            return False
+        except Exception as e:
+            print(f"[FAIL] Login flow crashed: {e}")
+            self.close_browser()
+            return False
+
+    def close_browser(self):
+        if self._cdp_port:
+            try:
+                if self._browser:
+                    self._browser.close()
+                if self._playwright:
+                    self._playwright.stop()
+            except Exception:
+                pass
+            self._browser = None
+            self._context = None
+            self._page = None
+            self._playwright = None
+            print("[CDP] 已断开外部浏览器连接")
+            return
+
+        try:
+            if self._browser:
+                self._browser.close()
+            if self._playwright:
+                self._playwright.stop()
+        except Exception:
+            pass
+        self._browser = None
+        self._context = None
+        self._page = None
+        self._playwright = None
+
+
+class PurchaseManager(ERPLoginManager):
+    def __init__(self, config_path, cdp_port=None):
+        super().__init__(config_path, cdp_port=cdp_port)
+        self._base_height_adjusted = False
+
+    def _set_fast_ui_timeouts(self):
+        try:
+            if self._page:
+                self._page.set_default_timeout(PURCHASE_UI_OPERATION_TIMEOUT_MS)
+                self._page.set_default_navigation_timeout(PURCHASE_UI_OPERATION_TIMEOUT_MS)
+        except Exception as e:
+            print(f"[WARN] 设置短超时失败: {e}")
+
+    def open_purchase_page(self):
+        if not self._page:
+            print("浏览器未启动")
+            return False
+        purchase_url = "https://www.erp321.com/app/scm/purchase/purchasemode.aspx?_c=jst-epaas"
+        print(f"正在打开采购单管理页面：{purchase_url}")
+        self._set_fast_ui_timeouts()
+        self._page.goto(purchase_url, wait_until="domcontentloaded", timeout=15000)
+        print("页面已加载")
+        try:
+            time.sleep(2)
+            for frame in self._page.frames:
+                if frame.locator("#animationProcessWrapper").count() > 0:
+                    frame.evaluate('document.getElementById("animationProcessWrapper").style.display = "none"')
+                    print("弹窗已关闭")
+        except Exception as e:
+            print(f"关闭弹窗异常：{e}")
+        return True
+
+    def search_sku(self, sku_id):
+        if not self._page:
+            print("浏览器未启动")
+            return False
+        page = self._page
+        self._set_fast_ui_timeouts()
+        print(f"\n开始搜索 SKU: {sku_id}")
+
+        input_selectors = [
+            'input[id*="SkuId"]',
+            'input[id*="skuId"]',
+            'input[name*="SkuId"]',
+            'input[placeholder*="商品"]',
+            'input[placeholder*="SKU"]'
+        ]
+
+        found = False
+        target_frame = None
+        target_selector = None
+        for frame in page.frames:
+            for selector in input_selectors:
+                try:
+                    if frame.locator(selector).count() > 0:
+                        target_frame = frame
+                        target_selector = selector
+                        found = True
+                        break
+                except Exception:
+                    continue
+            if found:
+                break
+
+        if not found:
+            print("未找到商品输入框")
+            return False
+
+        try:
+            target_frame.fill(target_selector, "")
+            target_frame.fill(target_selector, str(sku_id))
+            print(f"已填写 SKU: {sku_id}")
+        except Exception as e:
+            print(f"填写输入框失败: {e}")
+            return False
+
+        search_clicked = False
+        for frame in page.frames:
+            try:
+                search_btn = frame.locator('a.search[onclick*="Search()"]')
+                if search_btn.count() > 0:
+                    search_btn.first.click(timeout=PURCHASE_UI_OPERATION_TIMEOUT_MS)
+                    search_clicked = True
+                    print("已点击搜索按钮")
+                    break
+            except Exception:
+                continue
+
+        if not search_clicked:
+            try:
+                target_frame.locator(target_selector).press("Enter", timeout=PURCHASE_UI_OPERATION_TIMEOUT_MS)
+                print("已按回车搜索")
+            except Exception:
+                print("点击搜索和按回车均失败")
+                return False
+
+        time.sleep(3)
+        for frame in page.frames:
+            try:
+                if frame.locator('.rowList div[index]').count() > 0:
+                    row_count = frame.locator('.rowList div[index]').count()
+                    print(f"搜索结果已加载，共 {row_count} 行")
+                    return True
+            except Exception:
+                continue
+        print("等待后仍未见搜索结果")
+        return False
+
+    def update_all_results_dates(self, target_sku, date_str, custom_remark=None):
+        if not self._page:
+            return []
+        page = self._page
+        print(f"\n开始处理搜索结果，SKU: {target_sku}, 日期：{date_str}")
+        results = []
+
+        search_frame = None
+        for frame in page.frames:
+            if frame.locator('.rowList div[index]').count() > 0:
+                search_frame = frame
+                break
+
+        if not search_frame:
+            print("未找到结果列表框架")
+            return results
+
+        rows = search_frame.locator('.rowList div[index]')
+        row_count = rows.count()
+        print(f"检测到 {row_count} 个结果行")
+
+        for i in range(min(row_count, PURCHASE_MAX_RESULT_ROWS_PER_SKU)):
+            print(f"\n---> 处理第 {i+1}/{min(row_count, PURCHASE_MAX_RESULT_ROWS_PER_SKU)} 行")
+            current_row = search_frame.locator(f'.rowList div[index="{i}"]')
+            if current_row.count() == 0:
+                current_row = rows.nth(i)
+
+            po_id_line = current_row.locator('div.po_id_line')
+            if po_id_line.count() > 0:
+                po_id_line.first.scroll_into_view_if_needed()
+                po_id_line.first.click(force=True)
+            else:
+                current_row.first.scroll_into_view_if_needed()
+                current_row.first.click(force=True)
+            time.sleep(2)
+
+            detail_frame = None
+            for frame in page.frames:
+                try:
+                    frame_id = frame.evaluate("() => window.frameElement ? window.frameElement.id : ''")
+                    if "tab0" in str(frame_id) or "tab0" in frame.name:
+                        detail_frame = frame
+                        break
+                except Exception:
+                    continue
+            if not detail_frame:
+                for frame in page.frames:
+                    try:
+                        if frame.locator('[col_id="delivery_date"]').count() > 0:
+                            detail_frame = frame
+                            break
+                    except Exception:
+                        continue
+
+            if detail_frame:
+                try:
+                    # 在 div._jt_row._jt_rh 行中精确匹配目标 SKU
+                    target_row = None
+                    item_rows = detail_frame.locator('div._jt_row._jt_rh')
+                    item_count = item_rows.count()
+                    print(f"[INFO] 明细中共 {item_count} 个商品行")
+                    for ri in range(item_count):
+                        try:
+                            row = item_rows.nth(ri)
+                            cells = row.locator('._jt_cell')
+                            for ci in range(cells.count()):
+                                cell_text = cells.nth(ci).inner_text().strip()
+                                if cell_text == str(target_sku):
+                                    target_row = row
+                                    print(f"[INFO] 在第 {ri+1} 行找到 SKU {target_sku}")
+                                    break
+                            if target_row is not None:
+                                break
+                        except Exception:
+                            continue
+                    if target_row is None:
+                        print(f"明细中未找到 SKU {target_sku} 对应的行，跳过")
+                        continue
+
+                    product_name = ""
+                    product_spec = ""
+                    try:
+                        all_cells = target_row.locator('._jt_cell')
+                        cell_count = all_cells.count()
+                        # _jt_cell[8] = 商品名称, _jt_cell[9] = 规格颜色
+                        if cell_count > 8:
+                            product_name = all_cells.nth(8).inner_text().strip()
+                            print(f"找到商品名称：{product_name}")
+                        if cell_count > 9:
+                            product_spec = all_cells.nth(9).inner_text().strip()
+                            print(f"找到规格及颜色：{product_spec}")
+                        if not product_name:
+                            # fallback: 从 sku_id 元素取名称
+                            name_elements = target_row.locator('div.sku_id')
+                            for j in range(name_elements.count()):
+                                text = name_elements.nth(j).inner_text().strip()
+                                if text and text not in str(target_sku):
+                                    product_name = text
+                                    print(f"找到商品名称（fallback）：{product_name}")
+                                    break
+                    except Exception as e:
+                        print(f"查找商品信息失败：{e}")
+
+                    current_time = time.strftime("%Y-%m-%d %H:%M:%S")
+                    remark_text = custom_remark if custom_remark else f"{current_time} 机器人操作"
+
+                    # 先关闭可能遮挡的弹窗，再操作日期
+                    for overlay_sel in ['#msg_top', '.float_close', '.u-msg-close']:
+                        for overlay in page.locator(overlay_sel).all():
+                            try:
+                                overlay.click(force=True)
+                            except Exception:
+                                pass
+
+                    date_cell = target_row.locator('[col_id="delivery_date"]').first
+                    if date_cell.count() > 0:
+                        date_cell.scroll_into_view_if_needed()
+                        date_cell.click(force=True)
+                        time.sleep(0.5)
+                        input_box = date_cell.locator('input')
+                        if input_box.count() > 0:
+                            input_box.first.fill(date_str)
+                            input_box.first.press("Enter")
+                        else:
+                            page.keyboard.type(date_str)
+                            page.keyboard.press("Enter")
+                        print(f"日期更新为：{date_str}")
+
+                    remark_cell = target_row.locator('[col_id="remark"]').first
+                    if remark_cell.count() > 0:
+                        remark_cell.scroll_into_view_if_needed()
+                        remark_cell.click(force=True)
+                        time.sleep(0.5)
+                        remark_input = remark_cell.locator('input')
+                        if remark_input.count() > 0:
+                            remark_input.first.fill(remark_text)
+                            remark_input.first.press("Enter")
+                        else:
+                            page.keyboard.type(remark_text)
+                            page.keyboard.press("Enter")
+                        print(f"备注更新为：{remark_text}")
+
+                    # 关闭可能遮挡保存按钮的弹窗
+                    for overlay_sel in ['#msg_top', '.float_close', '.u-msg-close']:
+                        try:
+                            overlay = detail_frame.locator(overlay_sel)
+                            if overlay.count() > 0:
+                                overlay.first.click(force=True)
+                                time.sleep(0.3)
+                        except Exception:
+                            pass
+
+                    save_btn = detail_frame.locator('#_Tool_Save')
+                    if save_btn.count() > 0:
+                        try:
+                            save_btn.click(timeout=3000)
+                        except Exception:
+                            save_btn.click(force=True)
+                        print("已点击保存")
+                        time.sleep(1)
+                    else:
+                        page.keyboard.press("Control+s")
+
+                    results.append({
+                        'sku': target_sku,
+                        'name': product_name,
+                        'spec': product_spec
+                    })
+                    print(f"已记录：SKU={target_sku}, 名称={product_name}, 规格={product_spec}")
+
+                except Exception as e:
+                    print(f"处理详情时出错：{e}")
+            else:
+                print("未找到详情框架")
+
+            try:
+                close_btn = detail_frame.locator('.close, .u-close-btn, #closeBtn') if detail_frame else page.locator('.close, .u-close-btn, #closeBtn')
+                if close_btn.count() > 0:
+                    close_btn.first.click()
+                    time.sleep(0.5)
+            except Exception:
+                pass
+
+        return results
+
+
+def main(task_file=None, cdp_port=None):
+    script_dir = os.path.dirname(os.path.abspath(__file__))
+    task_path = Path(task_file) if task_file else WORK_ROOT / "task.json"
+
+    if not task_path.exists():
+        print(f"✗ 任务文件不存在：{task_path}")
+        print("请创建任务文件，格式：")
+        print('{\n  "sku_list": ["10426004"],\n  "target_date": "2026-05-20",\n  "remark": ""\n}')
+        sys.exit(1)
+
+    try:
+        task = json.loads(task_path.read_text(encoding='utf-8'))
+    except Exception as e:
+        print(f"✗ 读取任务文件失败：{e}")
+        sys.exit(1)
+
+    sku_list = task.get("sku_list", [])
+    target_date = task.get("target_date", "")
+    remark = task.get("remark", "")
+
+    if not sku_list or not target_date:
+        print("✗ 任务参数不完整，需要 sku_list 和 target_date")
+        sys.exit(1)
+
+    try:
+        datetime.strptime(target_date, '%Y-%m-%d')
+    except ValueError:
+        print(f"✗ 日期格式错误：{target_date}，请使用 YYYY-MM-DD")
+        sys.exit(1)
+
+    print("=" * 60)
+    print("📦 采购单更新工具 - Auto-Flow 版本")
+    print("=" * 60)
+    print(f"📋 SKU 列表：{', '.join(sku_list)}")
+    print(f"📅 目标日期：{target_date}")
+    print(f"📝 备注：{remark or '（自动生成）'}")
+    print("=" * 60)
+
+    config_path = str(CONFIG_PATH)
+    if not os.path.exists(config_path):
+        print(f"✗ 配置文件不存在：{config_path}")
+        sys.exit(1)
+
+    manager = PurchaseManager(config_path, cdp_port=cdp_port)
+    total_updated = 0
+    all_results = []
+    errors = []
+
+    try:
+        print("\n🔐 正在登录 ERP 系统...")
+        if not manager.login_with_playwright():
+            log_action("LOGIN_FAILED")
+            print("✗ 登录失败")
+            sys.exit(1)
+        log_action("LOGIN_SUCCESS")
+
+        print("\n📂 正在打开采购单管理页面...")
+        if not manager.open_purchase_page():
+            log_action("OPEN_PAGE_FAILED")
+            print("✗ 打开采购页面失败")
+            sys.exit(1)
+        log_action("OPEN_PAGE_SUCCESS")
+
+        for i, sku in enumerate(sku_list, 1):
+            print(f"\n{'='*60}")
+            print(f"🔍 处理 SKU {i}/{len(sku_list)}: {sku}")
+            print('='*60)
+            log_action("SEARCHING_SKU", sku)
+            if not manager.search_sku(sku):
+                log_action("SEARCH_FAILED", sku)
+                print(f"✗ 搜索 SKU {sku} 失败，跳过")
+                errors.append(f"搜索 SKU {sku} 失败")
+                continue
+            log_action("SEARCH_SUCCESS", sku)
+
+            log_action("UPDATING_SKU", f"{sku}, Date={target_date}")
+            updated_list = manager.update_all_results_dates(sku, target_date, remark)
+            total_updated += len(updated_list)
+            all_results.extend(updated_list)
+            if updated_list:
+                log_action("UPDATE_SUCCESS", f"{sku}, Count={len(updated_list)}")
+            else:
+                print(f"⚠️ SKU {sku} 未找到可更新的采购单")
+
+        log_action("SCRIPT_COMPLETED", f"Total Updated={total_updated}")
+        print("\n" + "=" * 60)
+        print("✅ 所有操作已完成！")
+        print(f"📊 共更新 {total_updated} 条采购单记录")
+        print("=" * 60)
+
+    except Exception as e:
+        print(f"\n✗ 执行过程中出错：{e}")
+        import traceback
+        traceback.print_exc()
+        sys.exit(1)
+    finally:
+        manager.close_browser()
+
+    # 写入结果文件
+    data_dir = WORK_ROOT
+    data_dir.mkdir(parents=True, exist_ok=True)
+    result = {
+        "workflow": "purchase-order-update",
+        "sku_list": sku_list,
+        "target_date": target_date,
+        "remark": remark,
+        "total_updated": total_updated,
+        "updated_records": all_results,
+        "errors": errors,
+        "completed_at": datetime.now().isoformat(),
+        "success": total_updated > 0 and len(errors) == 0,
+    }
+    result_path = data_dir / "update_result.json"
+    result_path.write_text(json.dumps(result, ensure_ascii=False, indent=2), encoding='utf-8')
+    print(f"\n📄 结果已保存至：{result_path}")
+
+    # 2026-06-22 v3:采集端落 PG,失败仅 warn,不影响业务结果。
+    # purchase-order-update 没有 bitable,只需 upsert_run + insert_purchase_order_updates。
+    _persist_to_pg(
+        sku_list=sku_list,
+        target_date=target_date,
+        remark=remark,
+        updated_records=all_results,
+        errors=errors,
+        completed_at=result["completed_at"],
+    )
+
+
+def _persist_to_pg(
+    sku_list: list[str],
+    target_date: str,
+    remark: str,
+    updated_records: list[dict],
+    errors: list[str],
+    completed_at: str,
+) -> None:
+    """2026-06-22 v3:采购单更新成功后落 PG(workflow_runs + fact_purchase_order_update)。
+
+    与 batch_process.py 的 v3 块保持一致结构:
+      1) 自生成 run_id
+      2) upsert_run 占位 workflow_runs(status=running → 写明细)
+      3) insert_purchase_order_updates 明细(UNIQUE(run_id, target_sku) 物理幂等)
+      4) bitable_insert_log/acquire 跳过 —— purchase-order-update 没有 bitable
+         (config.WORKFLOWS["purchase-order-update"].analyzer_scripts 仅含 im.send_message)
+      5) 全部 try/except 包裹,失败仅 warn,不影响业务结果
+
+    不需要 pending_insert.json —— 下游 LLM 只读 update_result.json,且
+    mcp_workflow.py:_read_update_count / total_updated=0 短路逻辑与 run_id 解耦,
+    不需要把 collector 的 run_id 透传给 LLM。
+    """
+    if not _PG_IMPORT_OK:
+        print("[WARN] pg_writer 不可用,跳过 PG 落库")
+        return
+
+    run_id = (
+        f"purchase-order-update_{time.strftime('%Y%m%dT%H%M%S')}_{secrets.token_hex(3)}"
+    )
+    trigger_time_iso = time.strftime("%Y-%m-%dT%H:%M:%S%z") or time.strftime("%Y-%m-%dT%H:%M:%S")
+
+    # 把输入 SKU 拆成 (sku -> updated_record) / (sku -> error_string) 两组,
+    # 让每个被请求的 SKU 都有一条 fact_purchase_order_update 行(成功或失败)。
+    updated_by_sku: dict[str, dict] = {}
+    for rec in updated_records or []:
+        sku_key = str(rec.get("sku") or "").strip()
+        if sku_key:
+            updated_by_sku[sku_key] = rec
+
+    failed_skus: dict[str, str] = {}
+    for err in errors or []:
+        # errors 形如 "搜索 SKU 10426004 失败" —— 抽出 8 位 SKU
+        m = re.search(r"\b(\d{8})\b", str(err))
+        if m:
+            failed_skus.setdefault(m.group(1), str(err))
+
+    rows: list[dict] = []
+    for sku in sku_list or []:
+        sku_str = str(sku).strip()
+        if not sku_str:
+            continue
+        if sku_str in updated_by_sku:
+            rec = updated_by_sku[sku_str]
+            rows.append({
+                "target_sku": sku_str,
+                "updated_sku": rec.get("sku") or sku_str,
+                "product_name": rec.get("name") or rec.get("product_name"),
+                "product_spec": rec.get("spec") or rec.get("product_spec"),
+                "success": True,
+                "error": None,
+                "completed_at": completed_at,
+            })
+        else:
+            err_msg = failed_skus.get(sku_str, "采购单未找到或未更新")
+            rows.append({
+                "target_sku": sku_str,
+                "updated_sku": None,
+                "product_name": None,
+                "product_spec": None,
+                "success": False,
+                "error": err_msg,
+                "completed_at": completed_at,
+            })
+
+    pg_ok = False
+    pg_error = ""
+    try:
+        # 1) workflow_runs 占位(running → 后面再 upsert 一次 completed)
+        pg.upsert_run(
+            run_id=run_id,
+            workflow="purchase-order-update",
+            trigger_time=trigger_time_iso,
+            status="running",
+            source_files=["update_result.json"],
+            raw_payload={
+                "sku_list": sku_list,
+                "target_date": target_date,
+                "remark": remark,
+                "total_updated": len(updated_records or []),
+                "error_count": len(errors or []),
+            },
+        )
+        # 2) 数据指纹 + 明细入库
+        #    成功 SKU 用 updated_sku,失败 SKU 用 target_sku,统一作为 data_hash key。
+        hash_keys: list[str] = []
+        for r in rows:
+            sku_for_hash = r.get("updated_sku") or r.get("target_sku") or ""
+            if sku_for_hash:
+                hash_keys.append(str(sku_for_hash).strip().lstrip("'"))
+        data_hash = pg.compute_data_hash([{"sku": k} for k in hash_keys], key="sku")
+        inserted = pg.insert_purchase_order_updates(
+            run_id=run_id,
+            target_date=target_date,
+            remark=remark,
+            rows=rows,
+        )
+        # 3) 回写 workflow_runs 终态
+        pg.upsert_run(
+            run_id=run_id,
+            workflow="purchase-order-update",
+            trigger_time=trigger_time_iso,
+            end_time=time.strftime("%Y-%m-%dT%H:%M:%S"),
+            status="completed",
+            source_files=["update_result.json"],
+            raw_payload={
+                "sku_list": sku_list,
+                "target_date": target_date,
+                "remark": remark,
+                "total_updated": len(updated_records or []),
+                "error_count": len(errors or []),
+                "detail_inserted": inserted,
+                "data_hash": data_hash,
+            },
+        )
+        print(
+            f"[PG] run_id={run_id} data_hash={data_hash[:12]}... "
+            f"detail_inserted={inserted} rows={len(rows)}"
+        )
+        pg_ok = True
+    except Exception as e:
+        pg_error = str(e)
+        print(f"[WARN] PG 落库失败(非阻塞): {e}")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description='采购单日期更新 - Auto-Flow')
+    parser.add_argument('--task-file', type=str, default=None, help='任务 JSON 文件路径')
+    parser.add_argument('--cdp-port', type=int, default=None, help='CDP 端口')
+    args = parser.parse_args()
+    main(task_file=args.task_file, cdp_port=args.cdp_port)
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/batch_process.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/batch_process.py
new file mode 100644
index 0000000..eb800b9
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/batch_process.py
@@ -0,0 +1,881 @@
+#!/usr/bin/env python3
+# ruff: noqa: E402,F841
+# -*- coding: utf-8 -*-
+"""
+补货流程批处理脚本（稳定版）
+
+职责：
+1. 运行 ProductReplenishment.py（如失败则退出非 0）
+2. 兜底生成 threshold_alert_data.json（若不存在）
+3. 读取最新 xlsx，按"缺货触发 → 30天 bitable 去重 → 同名扩展 → 按色/型号合并 → 再 30天去重"
+4. 生成 pending_insert.json / notify_data.json
+
+说明：
+- 本脚本只负责生成数据产物，消息发送由分析端负责。
+- 2026-06-29 流程定稿(回滚 B 方案):
+    "怎么刚刚补货建议插入的这么多 还是要加上30天的 然后把刚刚插入的删除掉"
+  → 回到原始 30天 dedup 流程。70天销量过滤(缺货触发)→ 查询 bitable 近 30 天 SKU →
+    smart_dedup 第一次过滤 → 同名扩展 → 变体合并 → smart_dedup 第二次过滤(防同名扩展
+    二次引入)。b侧上写去重后剩下的 SKU,INSERT 永远不 UPDATE(允许表内重复行)。
+"""
+
+from __future__ import annotations
+
+import json
+import math
+import os
+import re
+import secrets
+import subprocess
+import sys
+import time
+from pathlib import Path
+from typing import Any
+
+import pandas as pd
+import requests
+
+from gyxx_flow.adapters import environment_for_child_script
+
+# 让 `from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer` 在 `python batch_process.py` 这种
+# 不带 -m 的调用方式下也能解析(2026-06-22 v3 改造后,采集端要落 PG)。
+# 必须在所有 from gyxx_flow.modules.supply_chain.runtime.orchestrator import 之前。
+from gyxx_flow.modules.supply_chain.runtime.paths import (
+    CONFIG_PATH,
+    SHARED_ROOT,
+    WORK_ROOT,
+)
+
+# v2 (2026-06-22): 采集后立刻落 PG,让 fact_replenishment + bitable_insert_log
+# 成为本次 run 的事实源;pending_insert.json 仅作为下游 LLM 的便利读路径。
+try:
+    from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer as pg
+    _PG_IMPORT_OK = True
+except Exception as _pg_imp_err:  # pragma: no cover
+    _PG_IMPORT_OK = False
+    print(f"[WARN] pg_writer 不可用,本次不落 PG: {_pg_imp_err}")
+
+try:
+    from gyxx_flow.modules.supply_chain.runtime.orchestrator.feishu_sheets import (
+        load_discontinued_skus_from_feishu,
+    )
+    _FEISHU_SHEETS_OK = True
+except Exception as _fs_imp_err:
+    _FEISHU_SHEETS_OK = False
+    print(f"[WARN] feishu_sheets 不可用,停产 SKU 将从本地 xlsx 加载: {_fs_imp_err}")
+
+SCRIPT_DIR = Path(__file__).resolve().parent
+DATA_DIR = WORK_ROOT
+PRODUCT_SCRIPT = SCRIPT_DIR / "ProductReplenishment.py"
+CONFIG_FILE = CONFIG_PATH
+
+THRESHOLD_ALERT_FILE = DATA_DIR / "threshold_alert_data.json"
+PENDING_INSERT_FILE = DATA_DIR / "pending_insert.json"
+NOTIFY_DATA_FILE = DATA_DIR / "notify_data.json"
+REPLENISHMENT_SOURCE_FILE = DATA_DIR / "replenishment_source.json"
+# 停产列表写盘位置(2026-06-29):bitable 脚本作为最后一关过滤时读这个文件。
+# 写两份:scripts/data/ 是主路径,shared-data/replenishment/ 是和 pending_insert.json
+# 同步的副本(bitable 脚本默认读 pending 同目录)。
+DISCONTINUED_SKUS_FILE = DATA_DIR / "discontinued_skus.json"
+SHARED_DISCONTINUED_SKUS_FILE = SHARED_ROOT / "replenishment" / "discontinued_skus.json"
+
+# 多维表格配置
+BITABLE_APP_TOKEN = "WYHBb75IiadfJvsIidFc3olDn0b"
+BITABLE_TABLE_ID = "tblgVd8JgTWmfoiN"
+
+# 静态字段名 -> field_id 映射。
+# ⚠️ 写入多维表时必须用 field_id (fldXxx 形式)，不得用字段名。
+# 特别注意：pending_insert.json 中的 "产品名称" 字段对应到表内的 "产品名称 用于展示"，
+# 只有用 fld8iKlL8S 这个 field_id 才能正确匹配；直接用 "产品名称" 会触发 not_found，
+# 服务端会静默落盘空记录（这就是 2026-06-17 那次 106 条空数据的根因）。
+# 该表必须与 orchestrator/mcp_workflow.py:BITABLE_FIELD_MAP 保持一致。
+BITABLE_FIELD_MAP: dict[str, str] = {
+    "规格名称（新）": "fldYznbo8s",
+    "补货量": "fld0aWj8AA",
+    "规格名称（旧）": "fldM2vrObI",
+    "实际可用数": "fld6FUimoM",
+    "采购在途": "fldajEW5vQ",
+    "商品名称": "fldaV2dk7U",
+    "90天销量": "fldBSgLZLF",
+    "机器人备注": "fldch1v6cC",
+    "7天销量": "fldgMRs1EN",
+    "60天销量": "fldIjk4ADS",
+    "15天销量": "fldSfwvPVd",
+    "商品编号": "fldzIJQ3nw",
+    "月销量": "fldzVYr11w",
+    "产品名称": "fld8iKlL8S",
+}
+
+FEISHU_OPENAPI_BASE = "https://open.feishu.cn/open-apis"
+
+
+def _safe_int(v: Any) -> int:
+    try:
+        if v is None:
+            return 0
+        if isinstance(v, str) and not v.strip():
+            return 0
+        return int(float(v))
+    except Exception:
+        return 0
+
+
+def _safe_float(v: Any) -> float:
+    try:
+        if v is None:
+            return 0.0
+        if isinstance(v, str) and not v.strip():
+            return 0.0
+        if pd.isna(v):
+            return 0.0
+        return float(v)
+    except Exception:
+        return 0.0
+
+
+def _first_existing_col(df: pd.DataFrame, candidates: list[str]) -> str | None:
+    return next((c for c in candidates if c in df.columns), None)
+
+
+def _row_text(row: pd.Series, candidates: list[str]) -> str:
+    for col in candidates:
+        if col in row.index:
+            value = row.get(col)
+            if value is not None and not pd.isna(value):
+                return str(value).strip()
+    return ""
+
+
+def _normalize_name(value: Any) -> str:
+    if value is None or pd.isna(value):
+        return ""
+    return re.sub(r"\s+", "", str(value).strip())
+
+
+def _row_sku(row: pd.Series) -> str:
+    return str(row.get("商品编号", row.get("商品编码", ""))).strip().lstrip("'")
+
+
+def load_discontinued_skus(file_path: Path | None = None) -> set[str]:
+    """加载停产 SKU 列表。优先从飞书表格读取,失败时回退到本地 xlsx。"""
+    if _FEISHU_SHEETS_OK:
+        skus = load_discontinued_skus_from_feishu()
+        if skus:
+            return skus
+        print("[WARN] 飞书表格返回空集,尝试本地 xlsx 回退")
+
+    if file_path is None:
+        return set()
+    if not file_path.exists():
+        print(f"[WARN] 停产产品文件不存在，跳过停产 SKU 过滤: {file_path}")
+        return set()
+
+    try:
+        sheets = pd.read_excel(file_path, sheet_name=None, dtype=str)
+    except Exception as exc:
+        print(f"[WARN] 读取停产产品文件失败，跳过停产 SKU 过滤: {exc}")
+        return set()
+
+    skus: set[str] = set()
+    for sheet_name, sheet_df in sheets.items():
+        sku_col = next((col for col in sheet_df.columns if str(col).strip() == "商品编码"), None)
+        if sku_col is None:
+            print(f"[WARN] 停产产品文件 sheet={sheet_name} 未找到商品编码列，已跳过")
+            continue
+        for value in sheet_df[sku_col].dropna():
+            sku = str(value).strip().lstrip("'")
+            if re.fullmatch(r"\d+\.0", sku):
+                sku = sku[:-2]
+            if sku:
+                skus.add(sku)
+
+    print(f"[INFO] 已从本地 xlsx 加载停产商品编码 {len(skus)} 个")
+    return skus
+
+
+def write_discontinued_skus(skus: set[str]) -> None:
+    """把停产 SKU 列表写盘给 bitable 脚本最后一关过滤用(2026-06-29)。
+    写两份:scripts/data/ 主路径,shared-data/replenishment/ 跟 pending_insert.json
+    同步。空集也写(覆盖可能的旧文件),失败仅 warn 不抛。"""
+    payload = {"version": 1, "updated_at": time.strftime("%Y-%m-%dT%H:%M:%S"), "skus": sorted(skus)}
+    for path in (DISCONTINUED_SKUS_FILE, SHARED_DISCONTINUED_SKUS_FILE):
+        try:
+            path.parent.mkdir(parents=True, exist_ok=True)
+            tmp = path.with_suffix(".tmp")
+            tmp.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+            tmp.replace(path)
+        except Exception as exc:
+            print(f"[WARN] 停产列表写盘失败 {path}: {exc}")
+    if skus:
+        print(f"[INFO] 停产列表已写盘 {len(skus)} 个 to {DISCONTINUED_SKUS_FILE.name} (+ shared 副本)")
+
+
+def filter_discontinued_skus(df: pd.DataFrame, discontinued_skus: set[str]) -> pd.DataFrame:
+    if df.empty or not discontinued_skus:
+        return df
+    before = len(df)
+    kept = df[~df.apply(lambda row: _row_sku(row) in discontinued_skus, axis=1)].copy()
+    if len(kept) != before:
+        print(f"[INFO] 停产 SKU 过滤完成: {before} -> {len(kept)}")
+    return kept
+
+
+def _split_spec(spec: Any) -> tuple[str, str, str]:
+    """Return (color_base, model, color_suffix) from specs like 青色D；小号."""
+    text = "" if spec is None or pd.isna(spec) else str(spec).strip()
+    parts = [p.strip() for p in re.split(r"[;；]", text) if p and p.strip()]
+    color = re.sub(r"\s+", "", parts[0]) if parts else ""
+    model = parts[1] if len(parts) > 1 else ""
+    match = re.search(r"([A-Za-z]+)$", color)
+    suffix = match.group(1).upper() if match else ""
+    color_base = re.sub(r"[A-Za-z]+$", "", color).strip()
+    return color_base, model, suffix
+
+
+def _suffix_rank(suffix: str) -> tuple[int, str]:
+    if not suffix:
+        return (0, "")
+    return (1, suffix.upper())
+
+
+def _is_outside_dedup_window(
+    sku: str, recent_sku_map: dict[str, int], now_ms: int, dedup_days: int
+) -> bool:
+    """True when SKU has no 30-day bitable record, or its last insert is older than dedup_days."""
+    if not sku:
+        return True
+    last_ms = recent_sku_map.get(sku, 0)
+    if not last_ms:
+        return True
+    return (now_ms - last_ms) / (86400 * 1000) > dedup_days
+
+
+def _projected_70_day_sales(row: pd.Series) -> int:
+    existing = _row_text(row, ["预计70天销量", "70天销量"])
+    if existing:
+        return math.ceil(_safe_float(existing))
+    sales_60 = _safe_float(row.get("60天销量", 0))
+    return math.ceil(sales_60 / 60.0 * 70) if sales_60 > 0 else 0
+
+
+def _replenishment_qty(row: pd.Series) -> int | None:
+    sales_60 = _safe_float(row.get("60天销量", 0))
+    available = _safe_float(row.get("实际可用数", row.get("实际可用库存", row.get("可用数", 0))))
+    purchase = _safe_float(row.get("采购在途", row.get("采购在途数量", row.get("采购在途数", 0))))
+    target = (sales_60 / 60.0) * 100 if sales_60 > 0 else 0
+    diff = target - (available + purchase)
+    if diff > 0:
+        return int(round(diff / 100.0) * 100)
+    return None
+
+
+def merge_variants_before_insert(
+    df: pd.DataFrame,
+    recent_sku_map: dict[str, int] | None = None,
+    now_ms: int | None = None,
+    dedup_days: int = 30,
+) -> pd.DataFrame:
+    """
+    Merge variants whose color/model are effectively the same before insertion.
+
+    Example: 青色D；小号 + 青色G；小号 -> keep the G row's SKU/name/spec,
+    while summing inventory, in-transit, sales, and other numeric metrics.
+
+    When recent_sku_map/now_ms are provided, winner selection prefers SKUs that
+    are outside the 30-day dedup window (no recent bitable insert). This prevents
+    a 30-day-old insert for one variant from causing smart_dedup to drop the
+    entire merged row, which would lose the new variant that needs replenishment.
+    If every SKU in the group is inside the window, falls back to suffix ranking.
+    """
+    if df.empty:
+        return df
+
+    preserve_cols = {
+        "商品编号",
+        "商品编码",
+        "商品名称",
+        "产品名称",
+        "规格名称（旧）",
+        "规格名称（新）",
+        "颜色及规格",
+        "机器人备注",
+    }
+    grouped: dict[tuple[str, str, str], list[int]] = {}
+
+    for idx, row in df.iterrows():
+        product_key = _row_text(row, ["产品名称"]).replace(" ", "")
+        if not product_key:
+            product_key = re.sub(r"[A-Za-z]+", "", _row_text(row, ["商品名称"])).replace(" ", "")
+        raw_spec = _row_text(row, ["规格名称（旧）", "颜色及规格", "规格名称（新）"])
+        color_base, model, _ = _split_spec(raw_spec)
+        if not color_base and not model:
+            group_key = (product_key, f"__row_{idx}", "")
+        else:
+            group_key = (product_key, color_base, model)
+        grouped.setdefault(group_key, []).append(idx)
+
+    merged_rows: list[dict[str, Any]] = []
+    for indexes in grouped.values():
+        group_df = df.loc[indexes]
+        # winner 优先选 30 天外的 SKU,避免历史补货记录误伤本轮同款;
+        # group 全在 30 天内时回退到全部候选(等同旧行为)。
+        if recent_sku_map is not None and now_ms is not None:
+            outside = [
+                i for i in indexes
+                if _is_outside_dedup_window(_row_sku(df.loc[i]), recent_sku_map, now_ms, dedup_days)
+            ]
+            candidate_pool = outside if outside else list(indexes)
+        else:
+            candidate_pool = list(indexes)
+        winner_idx = max(
+            candidate_pool,
+            key=lambda i: _suffix_rank(_split_spec(_row_text(df.loc[i], ["规格名称（旧）", "颜色及规格", "规格名称（新）"]))[2]),
+        )
+        winner = df.loc[winner_idx]
+        merged: dict[str, Any] = {}
+
+        for col in df.columns:
+            if col in preserve_cols:
+                merged[col] = winner[col]
+                continue
+
+            numeric_values = pd.to_numeric(group_df[col], errors="coerce")
+            if numeric_values.notna().any():
+                total = numeric_values.fillna(0).sum()
+                merged[col] = int(total) if float(total).is_integer() else float(total)
+            else:
+                value = winner[col]
+                if value is None or (isinstance(value, float) and math.isnan(value)):
+                    value = group_df[col].dropna().iloc[0] if group_df[col].dropna().size else ""
+                merged[col] = value
+
+        merged_series = pd.Series(merged)
+        merged["_预计70天销量"] = _projected_70_day_sales(merged_series)
+        merged["_merged_skus"] = ",".join(sorted({_row_sku(df.loc[i]) for i in indexes if _row_sku(df.loc[i])}))
+        if "补货量" in merged:
+            merged["补货量"] = _replenishment_qty(pd.Series(merged))
+        merged_rows.append(merged)
+
+    return pd.DataFrame(merged_rows, columns=list(df.columns) + ["_预计70天销量", "_merged_skus"])
+
+
+def filter_by_70_day_sales(df: pd.DataFrame) -> pd.DataFrame:
+    """Keep only rows where actual available stock + in-transit purchase < projected 70-day sales."""
+    if df.empty:
+        return df
+
+    available_col = _first_existing_col(df, ["实际可用库存", "实际可用数", "可用数"])
+    purchase_col = _first_existing_col(df, ["采购在途数量", "采购在途数", "采购在途"])
+    if not available_col or not purchase_col:
+        print(f"[WARN] 未找到库存过滤字段，跳过70天销量过滤 (available_col={available_col}, purchase_col={purchase_col})")
+        return df
+
+    kept_indexes = []
+    for idx, row in df.iterrows():
+        projected_sales = _safe_float(row.get("_预计70天销量", _projected_70_day_sales(row)))
+        stock_with_purchase = _safe_float(row.get(available_col, 0)) + _safe_float(row.get(purchase_col, 0))
+        if stock_with_purchase < projected_sales:
+            kept_indexes.append(idx)
+
+    return df.loc[kept_indexes].copy()
+
+
+def load_replenishment_source(fallback_df: pd.DataFrame) -> pd.DataFrame:
+    if REPLENISHMENT_SOURCE_FILE.exists():
+        try:
+            payload = json.loads(REPLENISHMENT_SOURCE_FILE.read_text(encoding="utf-8"))
+            records = payload.get("records", [])
+            if isinstance(records, list):
+                source_df = pd.DataFrame(records)
+                if not source_df.empty:
+                    return source_df
+        except Exception as exc:
+            print(f"[WARN] 读取 ERP 补货源数据失败，回退使用补货建议文件: {exc}")
+    return fallback_df.copy()
+
+
+def expand_same_product_from_source(trigger_df: pd.DataFrame, source_df: pd.DataFrame) -> pd.DataFrame:
+    if trigger_df.empty or source_df.empty:
+        return trigger_df
+
+    trigger_names = {_normalize_name(v) for v in trigger_df.get("商品名称", [])}
+    trigger_names.discard("")
+    if not trigger_names:
+        return trigger_df
+
+    expanded = source_df[source_df["商品名称"].apply(_normalize_name).isin(trigger_names)].copy()
+    print(f"[INFO] 同商品名称源数据扩展: 触发商品名 {len(trigger_names)} 个 -> 源数据 SKU {len(expanded)} 条")
+    return expanded if not expanded.empty else trigger_df
+
+
+def _load_config() -> dict:
+    if CONFIG_FILE.exists():
+        try:
+            return json.loads(CONFIG_FILE.read_text(encoding="utf-8"))
+        except Exception:
+            pass
+    return {}
+
+
+def get_tenant_access_token() -> str:
+    """获取飞书 tenant_access_token（使用分析端应用）"""
+    config = _load_config()
+    analyzer = config.get("feishu_analyzer", {})
+    app_id = analyzer.get("app_id", "")
+    credential = analyzer.get("app_secret", "")
+    if not app_id or not credential:
+        raise RuntimeError("feishu_analyzer 配置缺失（app_id/app_secret）")
+
+    url = f"{FEISHU_OPENAPI_BASE}/auth/v3/tenant_access_token/internal/"
+    credentials = dict((("app_id", app_id), ("app_secret", credential)))
+    resp = requests.post(url, json=credentials, timeout=10)
+    resp.raise_for_status()
+    data = resp.json()
+    if data.get("code") == 0:
+        return data.get("tenant_access_token", "")
+    raise RuntimeError(f"获取 token 失败: {data}")
+
+
+def get_bitable_fields(token: str) -> dict[str, str]:
+    """获取多维表格字段列表"""
+    url = f"{FEISHU_OPENAPI_BASE}/bitable/v1/apps/{BITABLE_APP_TOKEN}/tables/{BITABLE_TABLE_ID}/fields"
+    headers = {"Authorization": f"Bearer {token}"}
+    resp = requests.get(url, headers=headers, timeout=30)
+    data = resp.json()
+    if data.get("code") == 0:
+        return {f["field_name"]: f["field_id"] for f in data.get("data", {}).get("items", [])}
+    raise RuntimeError(f"获取字段失败: {data}")
+
+
+def query_bitable_recent_records(token: str, date_field_name: str, days: int = 30, max_retries: int = 3) -> list[dict]:
+    """查询近N天的多维表格记录"""
+    thirty_days_ago_ms = int((time.time() - days * 86400) * 1000)
+    url = f"{FEISHU_OPENAPI_BASE}/bitable/v1/apps/{BITABLE_APP_TOKEN}/tables/{BITABLE_TABLE_ID}/records"
+    headers = {"Authorization": f"Bearer {token}"}
+    all_records = []
+    page_token = None
+
+    for attempt in range(max_retries):
+        try:
+            while True:
+                params = {"page_size": 500}
+                if page_token:
+                    params["page_token"] = page_token
+                resp = requests.get(url, headers=headers, params=params, timeout=60)
+                data = resp.json()
+                if data.get("code") != 0:
+                    break
+                items = data.get("data", {}).get("items", [])
+                if not items:
+                    break
+                all_records.extend(items)
+                page_token = data.get("data", {}).get("page_token")
+                if not page_token:
+                    break
+            break
+        except requests.exceptions.Timeout:
+            if attempt == max_retries - 1:
+                return []
+        except Exception:
+            if attempt == max_retries - 1:
+                return []
+            time.sleep(2)
+
+    recent = []
+    for rec in all_records:
+        fields = rec.get("fields", {})
+        date_val = fields.get(date_field_name)
+        if date_val:
+            try:
+                rec_ms = int(date_val) if isinstance(date_val, (int, float)) else 0
+                if rec_ms >= thirty_days_ago_ms:
+                    recent.append(rec)
+            except Exception:
+                pass
+    print(f"[INFO] 共拉取 {len(all_records)} 条记录，近 {days} 天内共有 {len(recent)} 条")
+    return recent
+
+
+def build_filtered_sku_map(recent_records: list[dict], sku_field_name: str, date_field_name: str) -> dict[str, int]:
+    """构建 SKU -> 最近插入日期(ms) 映射"""
+    sku_map: dict[str, int] = {}
+    for rec in recent_records:
+        fields = rec.get("fields", {})
+        sku_val = fields.get(sku_field_name)
+        if not sku_val:
+            continue
+        if isinstance(sku_val, list):
+            skus = [str(s).strip().lstrip("'") for s in sku_val]
+        else:
+            skus = [str(sku_val).strip().lstrip("'")]
+        date_val = fields.get(date_field_name)
+        date_ms = 0
+        if date_val:
+            try:
+                date_ms = int(date_val) if isinstance(date_val, (int, float)) else 0
+            except Exception:
+                pass
+        for s in skus:
+            if s not in sku_map or (date_ms and date_ms > sku_map[s]):
+                sku_map[s] = date_ms
+    return sku_map
+
+
+def smart_dedup(df: pd.DataFrame, recent_sku_map: dict[str, int], now_ms: int, dedup_days: int = 30) -> tuple[list[dict], list[str]]:
+    """
+    智能去重：
+    - 主 SKU (winner) 近 dedup_days 天内已有 -> 跳过
+    - 其他 -> 保留
+
+    只看 row 的主 SKU(_row_sku),不再连坐 _merged_skus 全集。
+    winner 由 merge_variants_before_insert 优先选 30 天外的 SKU,
+    因此同款 group 内只要有任一 SKU 在 30 天外,就会以该 SKU 作为 winner 保留,
+    避免历史补货记录误伤本轮同款。
+
+    返回 (保留的记录列表, 跳过的商品名称列表)
+    """
+    to_insert = []
+    skipped = []
+
+    for _, row in df.iterrows():
+        sku = _row_sku(row)
+        name = str(row.get("商品名称", "")).strip()
+        last_date_ms = recent_sku_map.get(sku, 0)
+        if last_date_ms:
+            days_diff = (now_ms - last_date_ms) / (86400 * 1000)
+            if days_diff <= dedup_days:
+                skipped.append(name)
+                continue
+        rec = {}
+        for col in df.columns:
+            if col.startswith("_"):
+                continue
+            val = row[col]
+            if isinstance(val, float) and not math.isnan(val) and val == int(val):
+                val = int(val)
+            rec[col] = val
+        to_insert.append(rec)
+
+    return to_insert, skipped
+
+
+def _run_product_replenishment() -> None:
+    if not PRODUCT_SCRIPT.exists():
+        raise FileNotFoundError(f"ProductReplenishment.py not found: {PRODUCT_SCRIPT}")
+
+    print("[INFO] Running ProductReplenishment.py ...")
+    child_environment = environment_for_child_script(PRODUCT_SCRIPT, os.environ)
+    cdp_url = child_environment["GYXX_BROWSER_CDP_URL"]
+    result = subprocess.run(
+        [sys.executable, str(PRODUCT_SCRIPT), "--cdp-url", cdp_url],
+        env=child_environment,
+        cwd=str(SCRIPT_DIR),
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+    )
+    if result.returncode != 0:
+        print(result.stdout)
+        print(result.stderr)
+        raise RuntimeError(f"ProductReplenishment.py failed with exit code {result.returncode}")
+
+
+def _ensure_threshold_alert_file() -> dict[str, Any]:
+    DATA_DIR.mkdir(parents=True, exist_ok=True)
+    if THRESHOLD_ALERT_FILE.exists():
+        try:
+            payload = json.loads(THRESHOLD_ALERT_FILE.read_text(encoding="utf-8"))
+            if isinstance(payload, dict):
+                return payload
+        except Exception:
+            pass
+
+    payload = {"alerts": [], "alert_count": 0, "generated_at": time.strftime("%Y-%m-%d %H:%M:%S")}
+    THRESHOLD_ALERT_FILE.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+    print("[INFO] threshold_alert_data.json not found, created empty default.")
+    return payload
+
+
+def _pick_latest_xlsx() -> Path | None:
+    candidates = sorted(DATA_DIR.glob("补货建议_*.xlsx"), key=lambda p: p.stat().st_mtime, reverse=True)
+    return candidates[0] if candidates else None
+
+
+def main() -> int:
+    try:
+        DATA_DIR.mkdir(parents=True, exist_ok=True)
+        _run_product_replenishment()
+
+        threshold_payload = _ensure_threshold_alert_file()
+        alert_count = int(threshold_payload.get("alert_count", 0) or 0)
+        print(f"[INFO] threshold alert count = {alert_count}")
+
+        # === v3 (2026-06-22): 库存阈值预警落 PG ===
+        # 与 replenishment 共用 batch_process.py,但阈值预警没有多维表写入,
+        # 也不依赖 xlsx 去重流程。run_id 写回 threshold_alert_data.json,
+        # 下游 LLM analyzer 读这个 json 时一并拿到 run_id,发送完群通知后
+        # 再调 mark_notification_sent("replenishment-alert", run_id, [...])。
+        # 失败仅 warn,不阻塞业务流程。
+        threshold_alerts: list[dict[str, Any]] = list(threshold_payload.get("alerts") or [])
+        threshold_run_id = (
+            f"replenishment-alert_{time.strftime('%Y%m%dT%H%M%S')}_{secrets.token_hex(3)}"
+        )
+        threshold_trigger_iso = time.strftime("%Y-%m-%dT%H:%M:%S%z") or time.strftime("%Y-%m-%dT%H:%M:%S")
+        threshold_pg_ok = False
+        if _PG_IMPORT_OK and threshold_alerts:
+            try:
+                pg.upsert_run(
+                    run_id=threshold_run_id,
+                    workflow="replenishment-alert",
+                    trigger_time=threshold_trigger_iso,
+                    status="running",
+                    source_files=[str(THRESHOLD_ALERT_FILE)],
+                    raw_payload={
+                        "alert_count": alert_count,
+                        "generated_at": threshold_payload.get("generated_at"),
+                    },
+                )
+                inserted_alerts = pg.insert_threshold_alerts(threshold_run_id, threshold_alerts)
+                # 把 SKUs upsert 到 dim_product(is_discontinued=False),
+                # 让后续 / 跨流程的 SKU join 都能命中。
+                sku_list = [
+                    str(a.get("sku") or "").strip()
+                    for a in threshold_alerts
+                    if a.get("sku")
+                ]
+                if sku_list:
+                    pg.upsert_dim_products(sku_list, is_discontinued=False)
+                print(
+                    f"[PG] replenishment-alert run_id={threshold_run_id} "
+                    f"alert_inserted={inserted_alerts}/{len(threshold_alerts)}"
+                )
+                threshold_pg_ok = True
+            except Exception as e:
+                print(f"[WARN] PG 落库失败(非阻塞,replenishment-alert): {e}")
+        elif _PG_IMPORT_OK:
+            print(
+                f"[PG] replenishment-alert run_id={threshold_run_id} "
+                "alert_count=0, 仅落 workflow_runs 占位"
+            )
+            try:
+                pg.upsert_run(
+                    run_id=threshold_run_id,
+                    workflow="replenishment-alert",
+                    trigger_time=threshold_trigger_iso,
+                    status="running",
+                    source_files=[str(THRESHOLD_ALERT_FILE)],
+                    raw_payload={
+                        "alert_count": 0,
+                        "generated_at": threshold_payload.get("generated_at"),
+                    },
+                )
+                threshold_pg_ok = True
+            except Exception as e:
+                print(f"[WARN] PG workflow_runs 占位失败(非阻塞): {e}")
+        else:
+            print("[WARN] pg_writer 不可用,replenishment-alert 跳过 PG 落库")
+
+        # 把 run_id 写回 json,供 LLM analyzer 直接读到
+        threshold_payload["run_id"] = threshold_run_id
+        threshold_payload["pg_ok"] = threshold_pg_ok
+        threshold_payload["bitable_version"] = 1  # 无多维表,占位字段供 mark_notification_sent 关联
+        THRESHOLD_ALERT_FILE.write_text(
+            json.dumps(threshold_payload, ensure_ascii=False, indent=2),
+            encoding="utf-8",
+        )
+
+        xlsx_path = _pick_latest_xlsx()
+        if xlsx_path is None:
+            pending_records: list[dict[str, Any]] = []
+            skipped_items: list[str] = []
+            notify_text = "本次未生成补货建议文件，暂无补货项。"
+            total_cleaned = 0
+        else:
+            print(f"[INFO] using xlsx: {xlsx_path.name}")
+            df = pd.read_excel(xlsx_path)
+            if df.empty:
+                pending_records = []
+                skipped_items = []
+                notify_text = "本次未发现需要补货的商品。"
+                total_cleaned = 0
+            else:
+                discontinued_skus = load_discontinued_skus()
+                write_discontinued_skus(discontinued_skus)  # 2026-06-29:bitable 脚本最后一关过滤用
+                df = filter_discontinued_skus(df, discontinued_skus)
+                if df.empty:
+                    pending_records = []
+                    skipped_items = []
+                    notify_text = "本次补货建议均为停产商品，已全部跳过。"
+                    total_cleaned = 0
+                    print("[INFO] 补货建议文件经停产 SKU 过滤后为空")
+                total_cleaned = str(df.iloc[0].get("商品名称", "")) and len(df)
+                before_merge_count = len(df)
+                source_df = filter_discontinued_skus(load_replenishment_source(df), discontinued_skus)
+
+                # 查询多维表格进行去重(前移到第一次 merge 之前:让两次 merge 都能用 recent_sku_map 选 winner)
+                print("[INFO] 查询多维表格记录进行去重...")
+                recent_sku_map: dict[str, int] = {}
+                try:
+                    token = get_tenant_access_token()
+                    live_field_map = get_bitable_fields(token)
+                    # Sanity check: 硬编码 BITABLE_FIELD_MAP 中所有 field_id 必须存在于 live schema。
+                    # 缺失说明表结构变了，需要更新 BITABLE_FIELD_MAP（mcp_workflow.py 里的那一份也要同步）。
+                    missing_ids = sorted(set(BITABLE_FIELD_MAP.values()) - set(live_field_map.values()))
+                    if missing_ids:
+                        print(f"[WARN] BITABLE_FIELD_MAP 含有 live schema 中不存在的 field_id: {missing_ids}")
+                        print(f"[WARN] Live schema name->id: {live_field_map}")
+                    sku_field = next((k for k in live_field_map if "SKU" in k or "sku" in k or "编号" in k or "编码" in k), None)
+                    date_field = next((k for k in live_field_map if "补货日期" in k), None)
+                    if sku_field and date_field:
+                        recent_records = query_bitable_recent_records(token, date_field, days=30)
+                        recent_sku_map = build_filtered_sku_map(recent_records, sku_field, date_field)
+                        print(f"[INFO] 近30天有 {len(recent_records)} 条记录，涉及 {len(recent_sku_map)} 个 SKU")
+                    else:
+                        print(f"[WARN] 未找到日期字段，跳过日期去重 (date_field={date_field})")
+                except Exception as e:
+                    print(f"[WARN] 查询飞书失败：{e}，跳过去重")
+
+                now_ms = int(time.time() * 1000)
+                merged_df = merge_variants_before_insert(df, recent_sku_map, now_ms, dedup_days=30)
+                after_merge_count = len(merged_df)
+                trigger_df = filter_by_70_day_sales(merged_df)
+                trigger_records, skipped_items = smart_dedup(trigger_df, recent_sku_map, now_ms, dedup_days=30)
+                trigger_pending_df = pd.DataFrame(trigger_records)
+                expanded_df = expand_same_product_from_source(trigger_pending_df, source_df)
+                df = merge_variants_before_insert(expanded_df, recent_sku_map, now_ms, dedup_days=30)
+                pending_records, final_skipped_items = smart_dedup(df, recent_sku_map, now_ms, dedup_days=30)
+                skipped_items = sorted(set(skipped_items + final_skipped_items))
+                print(
+                    "[INFO] 合并/70天销量过滤/30天去重/同名扩展完成: "
+                    f"{before_merge_count} -> 合并后 {after_merge_count} -> 触发 {len(trigger_df)} "
+                    f"-> 触发待插入 {len(trigger_pending_df)} -> 待判断插入 {len(df)}"
+                )
+
+                # === 最后一层过滤：停产 SKU ===
+                # 上游（同名扩展 / 变体合并）理论上不会再引入停产 SKU，
+                # 但 2026-06-17 发现停产产品.xlsx 缺失会让 load_discontinued_skus
+                # 返回空集、前面所有过滤都变 no-op，停产 SKU 直接进了多维表。
+                # 这一层在写入 pending_insert.json 之前再过一遍，作为最终兜底：
+                # 哪怕 discontinued_skus 是空集也只是 no-op，不会破坏正常流程；
+                # 文件恢复后这一层会自动生效。
+                if pending_records and discontinued_skus:
+                    before_n = len(pending_records)
+                    pending_records = [
+                        r for r in pending_records
+                        if str(r.get("商品编号", r.get("商品编码", ""))).strip().lstrip("'")
+                        not in discontinued_skus
+                    ]
+                    if len(pending_records) < before_n:
+                        print(
+                            f"[INFO] 最后一层过滤: 排除 {before_n - len(pending_records)} 条停产 SKU"
+                        )
+
+                if pending_records:
+                    lines = [f"补货建议汇总（{time.strftime('%Y-%m-%d')}）"]
+                    for rec in pending_records:
+                        sku = rec.get("商品编号", rec.get("商品编码", ""))
+                        spec = rec.get("规格名称（新）", rec.get("颜色及规格", ""))
+                        lines.append(f"- {sku} {rec['商品名称']} {spec} 补货量: {rec['补货量']}")
+                    notify_text = "\n".join(lines)
+                else:
+                    notify_text = "本次未发现需要补货的商品。"
+
+        skipped_count = len(skipped_items)
+        pending_count = len(pending_records)
+        print(f"[INFO] 待插入: {pending_count}, 跳过(30天内): {skipped_count}")
+
+        pending_payload = {
+            "records": pending_records,
+            "count": pending_count,
+            "pending_count": pending_count,
+            "skipped_count": skipped_count,
+            "skipped_items": skipped_items,
+            "total_cleaned": total_cleaned,
+        }
+
+        # === v2: 先落 PG,再写盘 ===
+        # 1) 生成 run_id,2) workflow_runs 上声明本次 run 在跑,
+        # 3) 明细 INSERT fact_replenishment(带 data_hash),
+        # 4) acquire_bitable_insert 占位(UNIQUE(workflow,run_id,version) 物理幂等),
+        # 5) 把 run_id + version 写进 pending_payload 供下游 LLM 读。
+        run_id = (
+            f"replenishment_{time.strftime('%Y%m%dT%H%M%S')}_{secrets.token_hex(3)}"
+        )
+        bitable_version = 1
+        trigger_time_iso = time.strftime("%Y-%m-%dT%H:%M:%S%z") or time.strftime("%Y-%m-%dT%H:%M:%S")
+        pg_ok = False
+        pg_error = ""
+        if _PG_IMPORT_OK:
+            try:
+                # 1) workflow_runs 占位
+                pg.upsert_run(
+                    run_id=run_id,
+                    workflow="replenishment",
+                    trigger_time=trigger_time_iso,
+                    status="running",
+                    source_files=[str(PENDING_INSERT_FILE), str(NOTIFY_DATA_FILE)],
+                    raw_payload={
+                        "pending_count": pending_count,
+                        "skipped_count": skipped_count,
+                        "total_cleaned": total_cleaned,
+                        "xlsx": str(xlsx_path) if xlsx_path else None,
+                    },
+                )
+                # 2) 数据指纹 + 明细入库
+                data_hash = pg.compute_data_hash(pending_records, key="商品编号")
+                inserted = pg.insert_replenishments(run_id, pending_records)
+                # 3) 多维表插入占位
+                acquired = pg.acquire_bitable_insert(
+                    workflow="replenishment",
+                    run_id=run_id,
+                    data_hash=data_hash,
+                    table_token=BITABLE_APP_TOKEN,
+                    table_id=BITABLE_TABLE_ID,
+                    record_count=pending_count,
+                    version=bitable_version,
+                )
+                if not acquired:
+                    print(
+                        f"[WARN] bitable_insert_log 同 (workflow=replenishment, "
+                        f"run_id={run_id}, version={bitable_version}) 已存在,本次未新建占位"
+                    )
+                print(
+                    f"[PG] run_id={run_id} data_hash={data_hash[:12]}... "
+                    f"detail_inserted={inserted} bitable_acquired={acquired}"
+                )
+                pg_ok = True
+            except Exception as e:
+                pg_error = str(e)
+                print(f"[WARN] PG 落库失败(非阻塞): {e}")
+        else:
+            print("[WARN] pg_writer 不可用,跳过 PG 落库")
+
+        pending_payload["run_id"] = run_id
+        pending_payload["bitable_version"] = bitable_version
+        pending_payload["pg_ok"] = pg_ok
+        pending_payload["data_hash"] = pg.compute_data_hash(pending_records, key="商品编号") if _PG_IMPORT_OK else ""
+
+        PENDING_INSERT_FILE.write_text(json.dumps(pending_payload, ensure_ascii=False, indent=2), encoding="utf-8")
+
+        notify_payload = {
+            "message": notify_text,
+            "generated_at": time.strftime("%Y-%m-%d %H:%M:%S"),
+            "pending_count": pending_count,
+            "skipped_count": skipped_count,
+            "has_replenish": pending_count > 0,
+            "alert_count": alert_count,
+            "run_id": run_id,
+            "alert_run_id": threshold_run_id,
+        }
+        NOTIFY_DATA_FILE.write_text(json.dumps(notify_payload, ensure_ascii=False, indent=2), encoding="utf-8")
+
+        print(f"[INFO] wrote {PENDING_INSERT_FILE.name} / {NOTIFY_DATA_FILE.name} (run_id={run_id})")
+        return 0
+    except Exception as exc:
+        print(f"[ERROR] {exc}")
+        return 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_confirmation.ps1 b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_confirmation.ps1
new file mode 100644
index 0000000..a4e0578
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_confirmation.ps1
@@ -0,0 +1,296 @@
+#!/usr/bin/env pwsh
+# Purchase Confirmation - Data collection runner only.
+
+$ErrorActionPreference = "Stop"
+
+$env:HTTP_PROXY = ""
+$env:HTTPS_PROXY = ""
+$env:http_proxy = ""
+$env:https_proxy = ""
+$env:ALL_PROXY = ""
+$env:all_proxy = ""
+
+[Console]::OutputEncoding = [System.Text.Encoding]::UTF8
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$projectRoot = if ($env:GYXX_PROJECT_ROOT) { $env:GYXX_PROJECT_ROOT } else { (Resolve-Path (Join-Path $scriptDir "..\..\..\..\..\..\..")).Path }
+$dataRoot = if ($env:GYXX_DATA_ROOT) { $env:GYXX_DATA_ROOT } else { Join-Path $projectRoot "var" }
+$rawRoot = if ($env:GYXX_SUPPLY_RAW_ROOT) { $env:GYXX_SUPPLY_RAW_ROOT } else { Join-Path $dataRoot "data\raw\supply_chain" }
+$stateRoot = if ($env:GYXX_SUPPLY_STATE_ROOT) { $env:GYXX_SUPPLY_STATE_ROOT } else { Join-Path $dataRoot "state\supply_chain" }
+$exportRoot = if ($env:GYXX_SUPPLY_EXPORT_ROOT) { $env:GYXX_SUPPLY_EXPORT_ROOT } else { Join-Path $dataRoot "data\exports\supply_chain" }
+$workDir = if ($env:GYXX_SUPPLY_WORK_ROOT) { $env:GYXX_SUPPLY_WORK_ROOT } else { Join-Path $dataRoot "tmp\supply_chain" }
+$venvPython = Join-Path $projectRoot ".venv\Scripts\python.exe"
+$pythonExe = if ($env:AUTOFLOW_PYTHON) { $env:AUTOFLOW_PYTHON } elseif (Test-Path $venvPython) { $venvPython } else { "python" }
+$targetBinding = (& $pythonExe -m gyxx_flow.runtime_exec env "supply_chain:orchestrator/scripts/PurchaseConfirmation.py") | ConvertFrom-Json
+$cdpPort = [int]$targetBinding.cdp_port
+$forceFreshBrowser = if ($env:AUTOFLOW_FORCE_FRESH_BROWSER) { $env:AUTOFLOW_FORCE_FRESH_BROWSER -ne "0" } else { $false }
+$runId = if ($env:AUTOFLOW_RUN_ID) { $env:AUTOFLOW_RUN_ID } else { (Get-Date -Format "yyyyMMddHHmmss") }
+$chromeProfileDir = [string]$targetBinding.profile_dir
+
+function Test-PortListening {
+    param([int]$Port)
+    try {
+        $conn = Get-NetTCPConnection -LocalPort $Port -State Listen -ErrorAction SilentlyContinue
+        return $null -ne $conn
+    } catch {
+        return $false
+    }
+}
+
+function Get-ChromePath {
+    $candidates = @(
+        "$env:AUTOFLOW_CHROME_PATH",
+        "$env:ProgramFiles\Google\Chrome\Application\chrome.exe",
+        "${env:ProgramFiles(x86)}\Google\Chrome\Application\chrome.exe",
+        "$env:LocalAppData\Google\Chrome\Application\chrome.exe",
+        "$env:LocalAppData\Google\Chrome\Bin\chrome.exe"
+    )
+
+    foreach ($path in $candidates) {
+        if ($path -and (Test-Path $path)) { return $path }
+    }
+    return $null
+}
+
+function Stop-ResidualPurchaseConfirmation {
+    $targets = Get-CimInstance Win32_Process -Filter "Name='python.exe'" |
+        Where-Object { $_.CommandLine -like "*PurchaseConfirmation.py*" }
+    if (-not $targets) {
+        Write-Host "[INFO] No residual PurchaseConfirmation.py process found." -ForegroundColor Cyan
+        return
+    }
+
+    foreach ($proc in $targets) {
+        try {
+            Stop-Process -Id $proc.ProcessId -Force -ErrorAction Stop
+            Write-Host "[CLEAN] Stopped residual process PID=$($proc.ProcessId)" -ForegroundColor Yellow
+        } catch {
+            Write-Host "[WARN] Failed to stop residual PID=$($proc.ProcessId): $($_.Exception.Message)" -ForegroundColor Yellow
+        }
+    }
+}
+
+function Stop-CdpChromeProcess {
+    param([int]$Port)
+    $targets = Get-CimInstance Win32_Process -Filter "Name='chrome.exe'" |
+        Where-Object { $_.CommandLine -like "*--remote-debugging-port=$Port*" }
+    foreach ($proc in $targets) {
+        try {
+            Stop-Process -Id $proc.ProcessId -Force -ErrorAction Stop
+            Write-Host "[CLEAN] Stopped Chrome CDP process PID=$($proc.ProcessId) on port $Port" -ForegroundColor Yellow
+        } catch {
+            Write-Host "[WARN] Failed to stop Chrome CDP PID=$($proc.ProcessId): $($_.Exception.Message)" -ForegroundColor Yellow
+        }
+    }
+}
+
+try {
+    & $pythonExe -c "import requests, pandas, openpyxl" | Out-Null
+    if ($LASTEXITCODE -ne 0) { throw "dependency check failed" }
+} catch {
+    Write-Host "[ERROR] Missing Python dependencies in venv. Install: pandas openpyxl requests" -ForegroundColor Red
+    Write-Host "  Run: & $projectRoot\.venv\Scripts\pip.exe install pandas openpyxl requests" -ForegroundColor Yellow
+    exit 1
+}
+
+Write-Host "========================================" -ForegroundColor Cyan
+Write-Host "  Purchase Confirmation - Collection" -ForegroundColor Cyan
+Write-Host "========================================" -ForegroundColor Cyan
+Write-Host "Scripts: $scriptDir"
+Write-Host "Python:  $pythonExe"
+Write-Host ""
+
+Stop-ResidualPurchaseConfirmation
+
+$confirmScript = Join-Path $scriptDir "PurchaseConfirmation.py"
+
+if ($forceFreshBrowser) {
+    Stop-CdpChromeProcess -Port $cdpPort
+}
+
+if ($forceFreshBrowser -or -not (Test-PortListening -Port $cdpPort)) {
+    $chromePath = Get-ChromePath
+    if (-not $chromePath) {
+        Write-Host "[ERROR] Chrome not found. Please install Chrome or set a browser with CDP port $cdpPort." -ForegroundColor Red
+        exit 1
+    }
+
+    Write-Host "[INFO] CDP port $cdpPort not listening. Starting Chrome..." -ForegroundColor Yellow
+    $chromeArgs = @(
+        "--no-sandbox",
+        "--no-first-run",
+        "--no-default-browser-check",
+        "--remote-debugging-port=$cdpPort",
+        "--user-data-dir=$chromeProfileDir",
+        "--new-window",
+        "about:blank"
+    )
+    Start-Process -FilePath $chromePath -ArgumentList $chromeArgs | Out-Null
+
+    $ready = $false
+    for ($i = 0; $i -lt 20; $i++) {
+        Start-Sleep -Milliseconds 500
+        if (Test-PortListening -Port $cdpPort) {
+            $ready = $true
+            break
+        }
+    }
+    if (-not $ready) {
+        Write-Host "[ERROR] Chrome started but CDP port $cdpPort is still unavailable." -ForegroundColor Red
+        exit 1
+    }
+    Write-Host "[OK] Chrome CDP is ready on port $cdpPort." -ForegroundColor Green
+} else {
+    Write-Host "[INFO] CDP port $cdpPort already listening. Reusing existing browser." -ForegroundColor Cyan
+}
+
+Write-Host "[INFO] Running PurchaseConfirmation.py..." -ForegroundColor Green
+
+# 清理脚本 data 目录下历史 CSV 和 xlsx，避免旧文件残留
+$dataDir = Join-Path $exportRoot "purchase-confirmation"
+New-Item -ItemType Directory -Path $dataDir -Force | Out-Null
+$collectionStartedAt = Get-Date
+$workflowId = "$($env:AUTOFLOW_RUN_ID)".Trim()
+if ([string]::IsNullOrWhiteSpace($workflowId) -or $workflowId -notmatch '^[A-Za-z0-9._=-]+$') {
+    Write-Host "[ERROR] AUTOFLOW_RUN_ID is empty or unsafe." -ForegroundColor Red
+    exit 3
+}
+
+$proc = Start-Process -FilePath $pythonExe -ArgumentList @($confirmScript, "--cdp-port", "$cdpPort") -NoNewWindow -PassThru -WorkingDirectory $scriptDir
+$timeoutMs = 1200000  # 20 minutes
+if (-not $proc.WaitForExit($timeoutMs)) {
+    try { Stop-Process -Id $proc.Id -Force -ErrorAction SilentlyContinue } catch {}
+    Write-Host "[ERROR] PurchaseConfirmation.py timeout after 20 minutes." -ForegroundColor Red
+    exit 124
+}
+$pythonExitCode = [int]$proc.ExitCode
+if ($pythonExitCode -ne 0) {
+    Write-Host "[WARN] PurchaseConfirmation.py exited with code $pythonExitCode. Will continue only if CSV output exists." -ForegroundColor Yellow
+}
+
+# 备份到 GYXX data exports
+$sharedBaseDir = Join-Path $rawRoot "purchase-confirmation"
+$sharedDir = Join-Path $sharedBaseDir "run_id=$workflowId"
+New-Item -ItemType Directory -Path $sharedDir -Force | Out-Null
+
+$csvFiles = Get-ChildItem -Path $dataDir -Filter "*.csv" -ErrorAction SilentlyContinue |
+    Where-Object { $_.LastWriteTime -ge $collectionStartedAt }
+if (-not $csvFiles -or $csvFiles.Count -eq 0) {
+    Write-Host "[ERROR] PurchaseConfirmation collection produced no CSV output." -ForegroundColor Red
+    if ($pythonExitCode -ne 0) {
+        exit $pythonExitCode
+    }
+    exit 2
+}
+foreach ($file in $csvFiles) {
+    Copy-Item -Path $file.FullName -Destination (Join-Path $sharedDir $file.Name) -Force
+    Write-Host "[OK] Copied $($file.Name) to shared directory" -ForegroundColor Green
+}
+
+$timestamp = (Get-Date -Format "yyyy-MM-ddTHH:mm:ss")
+$manifestPath = Join-Path $sharedDir "_run_manifest.json"
+$tempManifestPath = Join-Path $sharedDir "_run_manifest.tmp.json"
+
+# Use files that actually exist in shared-data after copy as the source of truth.
+$sharedCsvFiles = Get-ChildItem -Path $sharedDir -Filter "*.csv" -File -ErrorAction SilentlyContinue
+if (-not $sharedCsvFiles -or $sharedCsvFiles.Count -eq 0) {
+    Write-Host "[ERROR] Shared directory has no CSV after copy. Cannot generate _run_manifest.json." -ForegroundColor Red
+    exit 4
+}
+
+$manifest = @{
+    run_id = $workflowId
+    workflow = "purchase-confirmation"
+    generated_at = $timestamp
+    files = @($sharedCsvFiles | ForEach-Object { $_.Name })
+}
+
+try {
+    $manifest | ConvertTo-Json -Depth 5 | Set-Content -Path $tempManifestPath -Encoding UTF8 -ErrorAction Stop
+    Move-Item -Path $tempManifestPath -Destination $manifestPath -Force -ErrorAction Stop
+} catch {
+    Write-Host "[ERROR] Failed to write _run_manifest.json: $($_.Exception.Message)" -ForegroundColor Red
+    try { Remove-Item -Path $tempManifestPath -Force -ErrorAction SilentlyContinue } catch {}
+    exit 5
+}
+
+# Force validation: manifest must exist and contain the expected run_id/files.
+if (-not (Test-Path $manifestPath)) {
+    Write-Host "[ERROR] _run_manifest.json is missing after write." -ForegroundColor Red
+    exit 6
+}
+try {
+    $manifestCheck = Get-Content -Path $manifestPath -Raw -Encoding UTF8 | ConvertFrom-Json -ErrorAction Stop
+} catch {
+    Write-Host "[ERROR] _run_manifest.json is invalid JSON: $($_.Exception.Message)" -ForegroundColor Red
+    exit 7
+}
+
+if ("$($manifestCheck.run_id)".Trim() -ne $workflowId) {
+    Write-Host "[ERROR] _run_manifest.json run_id mismatch." -ForegroundColor Red
+    exit 8
+}
+
+$manifestFiles = @($manifestCheck.files)
+$expectedFiles = @($sharedCsvFiles | ForEach-Object { $_.Name })
+if ($manifestFiles.Count -ne $expectedFiles.Count) {
+    Write-Host "[ERROR] _run_manifest.json files count mismatch." -ForegroundColor Red
+    exit 9
+}
+foreach ($name in $expectedFiles) {
+    if (-not ($manifestFiles -contains $name)) {
+        Write-Host "[ERROR] _run_manifest.json missing file entry: $name" -ForegroundColor Red
+        exit 10
+    }
+}
+Write-Host "[OK] _run_manifest.json generated and validated." -ForegroundColor Green
+
+$statusFile = Join-Path $stateRoot "status.json"
+$existingStatus = @{}
+if (Test-Path $statusFile) {
+    try {
+        $existingStatus = Get-Content $statusFile -Raw | ConvertFrom-Json -AsHashtable
+    } catch {
+        $obj = Get-Content $statusFile -Raw | ConvertFrom-Json
+        $existingStatus = @{}
+        if ($null -ne $obj.last_update) { $existingStatus.last_update = $obj.last_update }
+        $existingStatus.collections = @{}
+        if ($null -ne $obj.collections) {
+            $obj.collections.PSObject.Properties | ForEach-Object {
+                $existingStatus.collections[$_.Name] = @{
+                    status = $_.Value.status
+                    files = @($_.Value.files)
+                    timestamp = $_.Value.timestamp
+                    error = $_.Value.error
+                }
+            }
+        }
+    }
+}
+
+$existingStatus.last_update = $timestamp
+if (-not $existingStatus.collections) {
+    $existingStatus.collections = @{}
+}
+$existingStatus.collections["purchase-confirmation"] = @{
+    status = $(if ($pythonExitCode -eq 0) { "success" } else { "partial_success" })
+    files = @($csvFiles | ForEach-Object { $_.Name })
+    timestamp = $timestamp
+    error = $(if ($pythonExitCode -eq 0) { "" } else { "PurchaseConfirmation.py exit code: $pythonExitCode" })
+}
+
+$existingStatus | ConvertTo-Json -Depth 5 | Set-Content -Path $statusFile -Encoding UTF8
+
+Write-Host ""
+
+# 采集完成后关闭 Chrome CDP 进程；Profile/Cookie 保留供下次复用
+Stop-CdpChromeProcess -Port $cdpPort
+
+Write-Host "========================================" -ForegroundColor Green
+Write-Host "  Data collection completed" -ForegroundColor Green
+Write-Host "========================================" -ForegroundColor Green
+if ($pythonExitCode -ne 0) {
+    Write-Host "[WARN] Collection artifacts are ready, but PurchaseConfirmation.py exited with code $pythonExitCode." -ForegroundColor Yellow
+}
+
+exit 0
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_purchase_order_update.ps1 b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_purchase_order_update.ps1
new file mode 100644
index 0000000..c4c7a10
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_purchase_order_update.ps1
@@ -0,0 +1,285 @@
+#!/usr/bin/env pwsh
+# Purchase Order Update - Data collection runner.
+
+$ErrorActionPreference = "Stop"
+
+$env:HTTP_PROXY = ""
+$env:HTTPS_PROXY = ""
+$env:http_proxy = ""
+$env:https_proxy = ""
+$env:ALL_PROXY = ""
+$env:all_proxy = ""
+
+[Console]::OutputEncoding = [System.Text.Encoding]::UTF8
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$projectRoot = if ($env:GYXX_PROJECT_ROOT) { $env:GYXX_PROJECT_ROOT } else { (Resolve-Path (Join-Path $scriptDir "..\..\..\..\..\..\..")).Path }
+$dataRoot = if ($env:GYXX_DATA_ROOT) { $env:GYXX_DATA_ROOT } else { Join-Path $projectRoot "var" }
+$rawRoot = if ($env:GYXX_SUPPLY_RAW_ROOT) { $env:GYXX_SUPPLY_RAW_ROOT } else { Join-Path $dataRoot "data\raw\supply_chain" }
+$stateRoot = if ($env:GYXX_SUPPLY_STATE_ROOT) { $env:GYXX_SUPPLY_STATE_ROOT } else { Join-Path $dataRoot "state\supply_chain" }
+$exportRoot = if ($env:GYXX_SUPPLY_EXPORT_ROOT) { $env:GYXX_SUPPLY_EXPORT_ROOT } else { Join-Path $dataRoot "data\exports\supply_chain" }
+$workDir = if ($env:GYXX_SUPPLY_WORK_ROOT) { $env:GYXX_SUPPLY_WORK_ROOT } else { Join-Path $dataRoot "tmp\supply_chain" }
+$venvPython = Join-Path $projectRoot ".venv\Scripts\python.exe"
+$pythonExe = if ($env:AUTOFLOW_PYTHON) { $env:AUTOFLOW_PYTHON } elseif (Test-Path $venvPython) { $venvPython } else { "python" }
+$targetBinding = (& $pythonExe -m gyxx_flow.runtime_exec env "supply_chain:orchestrator/scripts/PurchaseOrderUpdate.py") | ConvertFrom-Json
+$cdpPort = [int]$targetBinding.cdp_port
+$forceFreshBrowser = if ($env:AUTOFLOW_FORCE_FRESH_BROWSER) { $env:AUTOFLOW_FORCE_FRESH_BROWSER -ne "0" } else { $false }
+$runId = if ($env:AUTOFLOW_RUN_ID) { $env:AUTOFLOW_RUN_ID } else { (Get-Date -Format "yyyyMMddHHmmss") }
+$chromeProfileDir = [string]$targetBinding.profile_dir
+
+function Test-PortListening {
+    param([int]$Port)
+    try {
+        $conn = Get-NetTCPConnection -LocalPort $Port -State Listen -ErrorAction SilentlyContinue
+        return $null -ne $conn
+    } catch {
+        return $false
+    }
+}
+
+function Get-ChromePath {
+    $candidates = @(
+        "$env:AUTOFLOW_CHROME_PATH",
+        "$env:ProgramFiles\Google\Chrome\Application\chrome.exe",
+        "${env:ProgramFiles(x86)}\Google\Chrome\Application\chrome.exe",
+        "$env:LocalAppData\Google\Chrome\Application\chrome.exe",
+        "$env:LocalAppData\Google\Chrome\Bin\chrome.exe"
+    )
+    foreach ($path in $candidates) {
+        if ($path -and (Test-Path $path)) { return $path }
+    }
+    return $null
+}
+
+function Stop-CdpChromeProcess {
+    param([int]$Port)
+    $targets = Get-CimInstance Win32_Process -Filter "Name='chrome.exe'" |
+        Where-Object { $_.CommandLine -like "*--remote-debugging-port=$Port*" }
+    foreach ($proc in $targets) {
+        try {
+            Stop-Process -Id $proc.ProcessId -Force -ErrorAction Stop
+            Write-Host "[CLEAN] Stopped Chrome CDP process PID=$($proc.ProcessId) on port $Port" -ForegroundColor Yellow
+        } catch {
+            Write-Host "[WARN] Failed to stop Chrome CDP PID=$($proc.ProcessId): $($_.Exception.Message)" -ForegroundColor Yellow
+        }
+    }
+}
+
+function Stop-ResidualPurchaseOrderUpdate {
+    $targets = Get-CimInstance Win32_Process -Filter "Name='python.exe'" |
+        Where-Object { $_.CommandLine -like "*PurchaseOrderUpdate.py*" }
+    if (-not $targets) {
+        Write-Host "[INFO] No residual PurchaseOrderUpdate.py process found." -ForegroundColor Cyan
+        return
+    }
+    foreach ($proc in $targets) {
+        try {
+            Stop-Process -Id $proc.ProcessId -Force -ErrorAction Stop
+            Write-Host "[CLEAN] Stopped residual process PID=$($proc.ProcessId)" -ForegroundColor Yellow
+        } catch {
+            Write-Host "[WARN] Failed to stop residual PID=$($proc.ProcessId): $($_.Exception.Message)" -ForegroundColor Yellow
+        }
+    }
+}
+
+# 检查任务文件
+$taskFile = Join-Path $stateRoot "tasks\purchase-order-update\task.json"
+if (-not (Test-Path $taskFile)) {
+    Write-Host "[ERROR] Task file not found: $taskFile" -ForegroundColor Red
+    Write-Host "Please create the task file with the following format:" -ForegroundColor Yellow
+    Write-Host '{"sku_list": ["10426004"], "target_date": "2026-05-20", "remark": ""}' -ForegroundColor Yellow
+    exit 1
+}
+
+try {
+    & $pythonExe -c "import requests, playwright" | Out-Null
+    if ($LASTEXITCODE -ne 0) { throw "dependency check failed" }
+} catch {
+    Write-Host "[ERROR] Python dependency check failed." -ForegroundColor Red
+    exit 1
+}
+
+Write-Host "========================================" -ForegroundColor Cyan
+Write-Host "  Purchase Order Update - Collection" -ForegroundColor Cyan
+Write-Host "========================================" -ForegroundColor Cyan
+Write-Host "Scripts: $scriptDir"
+Write-Host "Python:  $pythonExe"
+Write-Host ""
+
+Stop-ResidualPurchaseOrderUpdate
+
+# 准备任务文件到脚本 data 目录
+$dataDir = $workDir
+New-Item -ItemType Directory -Path $dataDir -Force | Out-Null
+Copy-Item -Path $taskFile -Destination (Join-Path $dataDir "task.json") -Force
+Write-Host "[OK] Task file copied to script data directory" -ForegroundColor Green
+
+$updateScript = Join-Path $scriptDir "PurchaseOrderUpdate.py"
+
+if ($forceFreshBrowser) {
+    Stop-CdpChromeProcess -Port $cdpPort
+}
+
+if ($forceFreshBrowser -or -not (Test-PortListening -Port $cdpPort)) {
+    $chromePath = Get-ChromePath
+    if (-not $chromePath) {
+        Write-Host "[ERROR] Chrome not found." -ForegroundColor Red
+        exit 1
+    }
+    Write-Host "[INFO] CDP port $cdpPort not listening. Starting Chrome..." -ForegroundColor Yellow
+    $chromeArgs = @(
+        "--no-sandbox",
+        "--no-first-run",
+        "--no-default-browser-check",
+        "--remote-debugging-port=$cdpPort",
+        "--user-data-dir=$chromeProfileDir",
+        "--new-window",
+        "about:blank"
+    )
+    Start-Process -FilePath $chromePath -ArgumentList $chromeArgs | Out-Null
+
+    $ready = $false
+    for ($i = 0; $i -lt 20; $i++) {
+        Start-Sleep -Milliseconds 500
+        if (Test-PortListening -Port $cdpPort) {
+            $ready = $true
+            break
+        }
+    }
+    if (-not $ready) {
+        Write-Host "[ERROR] Chrome started but CDP port $cdpPort is still unavailable." -ForegroundColor Red
+        exit 1
+    }
+    Write-Host "[OK] Chrome CDP is ready on port $cdpPort." -ForegroundColor Green
+} else {
+    Write-Host "[INFO] CDP port $cdpPort already listening. Reusing existing browser." -ForegroundColor Cyan
+}
+
+Write-Host "[INFO] Running PurchaseOrderUpdate.py..." -ForegroundColor Green
+
+$env:PYTHONIOENCODING = "utf-8"
+$psi = New-Object System.Diagnostics.ProcessStartInfo
+$psi.FileName = $pythonExe
+$psi.Arguments = "`"$updateScript`" --cdp-port $cdpPort"
+$psi.WorkingDirectory = $scriptDir
+$psi.UseShellExecute = $false
+$psi.RedirectStandardOutput = $true
+$psi.RedirectStandardError = $true
+$psi.CreateNoWindow = $true
+
+$proc = [System.Diagnostics.Process]::Start($psi)
+$stdoutTask = $proc.StandardOutput.ReadToEndAsync()
+$stderrTask = $proc.StandardError.ReadToEndAsync()
+$timeoutMs = 1200000  # 20 minutes
+
+if (-not $proc.WaitForExit($timeoutMs)) {
+    try { $proc.Kill() } catch {}
+    Write-Host "[ERROR] PurchaseOrderUpdate.py timeout after 20 minutes." -ForegroundColor Red
+    exit 124
+}
+
+$stdout = $stdoutTask.Result
+$stderr = $stderrTask.Result
+if ($stdout) { Write-Host $stdout }
+if ($stderr) { Write-Host $stderr -ForegroundColor Yellow }
+
+if ($proc.ExitCode -ne 0) {
+    Write-Host "[ERROR] PurchaseOrderUpdate.py failed. Exit code: $($proc.ExitCode)" -ForegroundColor Red
+    exit $proc.ExitCode
+}
+
+# 复制输出到 shared-data
+$workflowId = "$runId".Trim()
+if ([string]::IsNullOrWhiteSpace($workflowId) -or $workflowId -notmatch '^[A-Za-z0-9._=-]+$') {
+    Write-Host "[ERROR] AUTOFLOW_RUN_ID is empty or unsafe." -ForegroundColor Red
+    exit 3
+}
+$sharedBaseDir = Join-Path $rawRoot "purchase-order-update"
+$sharedDir = Join-Path $sharedBaseDir "run_id=$workflowId"
+New-Item -ItemType Directory -Path $sharedDir -Force | Out-Null
+
+$resultFile = Join-Path $dataDir "update_result.json"
+if (-not (Test-Path $resultFile)) {
+    Write-Host "[ERROR] PurchaseOrderUpdate produced no result output." -ForegroundColor Red
+    exit 2
+}
+Copy-Item -Path $resultFile -Destination (Join-Path $sharedDir "update_result.json") -Force
+Write-Host "[OK] Copied update_result.json to shared directory" -ForegroundColor Green
+
+# 备份到 GYXX data exports
+$backupDir = Join-Path $exportRoot "purchase-order-update"
+if (-not (Test-Path $backupDir)) { New-Item -ItemType Directory -Path $backupDir -Force | Out-Null }
+$timestampStr = Get-Date -Format "yyyyMMdd_HHmmss"
+$bakName = "update_result_$timestampStr.json"
+Copy-Item -Path $resultFile -Destination (Join-Path $backupDir $bakName) -Force
+Write-Host "[BACKUP] Copied update_result.json to GYXX data exports\purchase-order-update" -ForegroundColor Cyan
+
+# 生成 _run_manifest.json
+$timestamp = (Get-Date -Format "yyyy-MM-ddTHH:mm:ss")
+$manifestPath = Join-Path $sharedDir "_run_manifest.json"
+$tempManifestPath = Join-Path $sharedDir "_run_manifest.tmp.json"
+$manifest = @{
+    run_id = $workflowId
+    workflow = "purchase-order-update"
+    generated_at = $timestamp
+    files = @("update_result.json")
+}
+try {
+    $manifest | ConvertTo-Json -Depth 5 | Set-Content -Path $tempManifestPath -Encoding UTF8 -ErrorAction Stop
+    Move-Item -Path $tempManifestPath -Destination $manifestPath -Force -ErrorAction Stop
+} catch {
+    Write-Host "[ERROR] Failed to write _run_manifest.json: $($_.Exception.Message)" -ForegroundColor Red
+    try { Remove-Item -Path $tempManifestPath -Force -ErrorAction SilentlyContinue } catch {}
+    exit 4
+}
+Write-Host "[OK] _run_manifest.json generated." -ForegroundColor Green
+
+# 更新状态文件
+$statusFile = Join-Path $stateRoot "status.json"
+$existingStatus = @{}
+if (Test-Path $statusFile) {
+    try {
+        $existingStatus = Get-Content $statusFile -Raw | ConvertFrom-Json -AsHashtable
+    } catch {
+        $obj = Get-Content $statusFile -Raw | ConvertFrom-Json
+        $existingStatus = @{}
+        if ($null -ne $obj.last_update) { $existingStatus.last_update = $obj.last_update }
+        $existingStatus.collections = @{}
+        if ($null -ne $obj.collections) {
+            $obj.collections.PSObject.Properties | ForEach-Object {
+                $existingStatus.collections[$_.Name] = @{
+                    status = $_.Value.status
+                    files = @($_.Value.files)
+                    timestamp = $_.Value.timestamp
+                    error = $_.Value.error
+                }
+            }
+        }
+    }
+}
+
+$existingStatus.last_update = (Get-Date -Format "yyyy-MM-ddTHH:mm:ss")
+if (-not $existingStatus.collections) { $existingStatus.collections = @{} }
+$existingStatus.collections["purchase-order-update"] = @{
+    status = "success"
+    files = @("update_result.json")
+    timestamp = (Get-Date -Format "yyyy-MM-ddTHH:mm:ss")
+}
+$existingStatus | ConvertTo-Json -Depth 5 | Set-Content -Path $statusFile -Encoding UTF8
+
+# 清理 scripts/data 中间产物（update_result.json 已复制到 shared-data）
+Get-ChildItem -Path $dataDir -Include "*.json", "*.xlsx", "*.csv" -File -ErrorAction SilentlyContinue |
+    Where-Object { $_.Name -ne "task.json" } |
+    Remove-Item -Force -ErrorAction SilentlyContinue
+Write-Host "[OK] Cleaned intermediate artifacts in scripts/data" -ForegroundColor Green
+
+Write-Host ""
+
+# 采集完成后关闭 Chrome CDP 进程；Profile/Cookie 保留供下次复用
+Stop-CdpChromeProcess -Port $cdpPort
+
+Write-Host "========================================" -ForegroundColor Green
+Write-Host "  Data collection completed" -ForegroundColor Green
+Write-Host "========================================" -ForegroundColor Green
+
+exit 0
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_replenishment.ps1 b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_replenishment.ps1
new file mode 100644
index 0000000..7a6887c
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/collect_replenishment.ps1
@@ -0,0 +1,319 @@
+#!/usr/bin/env pwsh
+# Product Replenishment - Data collection runner only.
+
+$ErrorActionPreference = "Stop"
+
+$env:HTTP_PROXY = ""
+$env:HTTPS_PROXY = ""
+$env:http_proxy = ""
+$env:https_proxy = ""
+$env:ALL_PROXY = ""
+$env:all_proxy = ""
+
+[Console]::OutputEncoding = [System.Text.Encoding]::UTF8
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$projectRoot = if ($env:GYXX_PROJECT_ROOT) { $env:GYXX_PROJECT_ROOT } else { (Resolve-Path (Join-Path $scriptDir "..\..\..\..\..\..\..")).Path }
+$dataRoot = if ($env:GYXX_DATA_ROOT) { $env:GYXX_DATA_ROOT } else { Join-Path $projectRoot "var" }
+$rawRoot = if ($env:GYXX_SUPPLY_RAW_ROOT) { $env:GYXX_SUPPLY_RAW_ROOT } else { Join-Path $dataRoot "data\raw\supply_chain" }
+$stateRoot = if ($env:GYXX_SUPPLY_STATE_ROOT) { $env:GYXX_SUPPLY_STATE_ROOT } else { Join-Path $dataRoot "state\supply_chain" }
+$exportRoot = if ($env:GYXX_SUPPLY_EXPORT_ROOT) { $env:GYXX_SUPPLY_EXPORT_ROOT } else { Join-Path $dataRoot "data\exports\supply_chain" }
+$workDir = if ($env:GYXX_SUPPLY_WORK_ROOT) { $env:GYXX_SUPPLY_WORK_ROOT } else { Join-Path $dataRoot "tmp\supply_chain" }
+$venvPython = Join-Path $projectRoot ".venv\Scripts\python.exe"
+$pythonExe = if ($env:AUTOFLOW_PYTHON) { $env:AUTOFLOW_PYTHON } elseif (Test-Path $venvPython) { $venvPython } else { "python" }
+$targetBinding = (& $pythonExe -m gyxx_flow.runtime_exec env "supply_chain:orchestrator/scripts/ProductReplenishment.py") | ConvertFrom-Json
+$cdpPort = [int]$targetBinding.cdp_port
+$forceFreshBrowser = if ($env:AUTOFLOW_FORCE_FRESH_BROWSER) { $env:AUTOFLOW_FORCE_FRESH_BROWSER -ne "0" } else { $false }
+$runId = if ($env:AUTOFLOW_RUN_ID) { $env:AUTOFLOW_RUN_ID } else { (Get-Date -Format "yyyyMMddHHmmss") }
+$chromeProfileDir = [string]$targetBinding.profile_dir
+
+function Test-PortListening {
+    param([int]$Port)
+    try {
+        $conn = Get-NetTCPConnection -LocalPort $Port -State Listen -ErrorAction SilentlyContinue
+        return $null -ne $conn
+    } catch {
+        return $false
+    }
+}
+
+function Get-ChromePath {
+    $candidates = @(
+        "$env:AUTOFLOW_CHROME_PATH",
+        "$env:ProgramFiles\Google\Chrome\Application\chrome.exe",
+        "${env:ProgramFiles(x86)}\Google\Chrome\Application\chrome.exe",
+        "$env:LocalAppData\Google\Chrome\Application\chrome.exe",
+        "$env:LocalAppData\Google\Chrome\Bin\chrome.exe"
+    )
+    foreach ($path in $candidates) {
+        if ($path -and (Test-Path $path)) { return $path }
+    }
+    return $null
+}
+
+function Stop-ResidualReplenishmentProcesses {
+    $targets = Get-CimInstance Win32_Process -Filter "Name='python.exe'" |
+        Where-Object {
+            $_.CommandLine -like "*batch_process.py*" -or
+            $_.CommandLine -like "*ProductReplenishment.py*"
+        }
+
+    if (-not $targets) {
+        Write-Host "[INFO] No residual replenishment process found." -ForegroundColor Cyan
+        return
+    }
+
+    foreach ($proc in $targets) {
+        try {
+            Stop-Process -Id $proc.ProcessId -Force -ErrorAction Stop
+            Write-Host "[CLEAN] Stopped residual process PID=$($proc.ProcessId)" -ForegroundColor Yellow
+        } catch {
+            Write-Host "[WARN] Failed to stop residual PID=$($proc.ProcessId): $($_.Exception.Message)" -ForegroundColor Yellow
+        }
+    }
+}
+
+try {
+    & $pythonExe -c "import requests, pandas, openpyxl, playwright" | Out-Null
+    if ($LASTEXITCODE -ne 0) { throw "dependency check failed" }
+} catch {
+    Write-Host "[ERROR] Python dependency check failed." -ForegroundColor Red
+    exit 1
+}
+
+Write-Host "========================================" -ForegroundColor Cyan
+Write-Host "  Product Replenishment - Collection" -ForegroundColor Cyan
+Write-Host "========================================" -ForegroundColor Cyan
+Write-Host "Scripts: $scriptDir"
+Write-Host "Python:  $pythonExe"
+Write-Host ""
+
+if ($env:AUTOFLOW_ALLOW_STOP_RESIDUAL -eq "1") {
+    Stop-ResidualReplenishmentProcesses
+} else {
+    Write-Host "[INFO] Skip stopping residual processes (controlled by orchestrator locks)." -ForegroundColor Cyan
+}
+
+function Stop-CdpChromeProcess {
+    param([int]$Port)
+    $targets = Get-CimInstance Win32_Process -Filter "Name='chrome.exe'" |
+        Where-Object { $_.CommandLine -like "*--remote-debugging-port=$Port*" }
+    foreach ($proc in $targets) {
+        try {
+            Stop-Process -Id $proc.ProcessId -Force -ErrorAction Stop
+            Write-Host "[CLEAN] Stopped Chrome CDP process PID=$($proc.ProcessId) on port $Port" -ForegroundColor Yellow
+        } catch {
+            Write-Host "[WARN] Failed to stop Chrome CDP PID=$($proc.ProcessId): $($_.Exception.Message)" -ForegroundColor Yellow
+        }
+    }
+}
+
+$batchScript = Join-Path $scriptDir "batch_process.py"
+# 记录采集开始时间，后续只复制本次运行生成的结果文件
+$collectStartTime = Get-Date
+
+if ($forceFreshBrowser) {
+    Stop-CdpChromeProcess -Port $cdpPort
+}
+
+if ($forceFreshBrowser -or -not (Test-PortListening -Port $cdpPort)) {
+    $chromePath = Get-ChromePath
+    if (-not $chromePath) {
+        Write-Host "[ERROR] Chrome not found. Please install Chrome or provide a browser CDP endpoint." -ForegroundColor Red
+        exit 1
+    }
+
+    Write-Host "[INFO] CDP port $cdpPort not listening. Starting Chrome..." -ForegroundColor Yellow
+    $chromeArgs = @(
+        "--no-sandbox",
+        "--no-first-run",
+        "--no-default-browser-check",
+        "--remote-debugging-port=$cdpPort",
+        "--user-data-dir=$chromeProfileDir",
+        "--new-window",
+        "about:blank"
+    )
+    Start-Process -FilePath $chromePath -ArgumentList $chromeArgs | Out-Null
+
+    $ready = $false
+    for ($i = 0; $i -lt 20; $i++) {
+        Start-Sleep -Milliseconds 500
+        if (Test-PortListening -Port $cdpPort) {
+            $ready = $true
+            break
+        }
+    }
+    if (-not $ready) {
+        Write-Host "[ERROR] Chrome started but CDP port $cdpPort is still unavailable." -ForegroundColor Red
+        exit 1
+    }
+    Write-Host "[OK] Chrome CDP is ready on port $cdpPort." -ForegroundColor Green
+} else {
+    Write-Host "[INFO] CDP port $cdpPort already listening. Reusing existing browser." -ForegroundColor Cyan
+}
+
+Write-Host "[INFO] Running batch_process.py..." -ForegroundColor Green
+
+$env:AUTOFLOW_CDP_PORT = $cdpPort
+
+Push-Location $scriptDir
+try {
+    & $pythonExe $batchScript
+    $batchExitCode = $LASTEXITCODE
+} finally {
+    Pop-Location
+}
+
+if ($batchExitCode -ne 0) {
+    Write-Host "[ERROR] batch_process.py failed. Exit code: $batchExitCode" -ForegroundColor Red
+    exit $batchExitCode
+}
+
+$dataDir = $workDir
+$targetWorkflow = $env:AUTOFLOW_TARGET_WORKFLOW
+$workflowId = if ($env:AUTOFLOW_RUN_ID) { "$($env:AUTOFLOW_RUN_ID)".Trim() } else { (Get-Date -Format "yyyyMMddHHmmss") }
+if ($workflowId -notmatch '^[A-Za-z0-9._=-]+$') {
+    Write-Host "[ERROR] AUTOFLOW_RUN_ID is unsafe." -ForegroundColor Red
+    exit 3
+}
+$sharedReplenishmentDir = Join-Path (Join-Path $rawRoot "replenishment") "run_id=$workflowId"
+$sharedAlertDir = Join-Path (Join-Path $rawRoot "replenishment-alert") "run_id=$workflowId"
+$isReplenishmentTarget = (-not $targetWorkflow) -or $targetWorkflow -eq "replenishment"
+$isAlertTarget = (-not $targetWorkflow) -or $targetWorkflow -eq "replenishment-alert"
+
+if ($isReplenishmentTarget) {
+    New-Item -ItemType Directory -Path $sharedReplenishmentDir -Force | Out-Null
+}
+if ($isAlertTarget) {
+    New-Item -ItemType Directory -Path $sharedAlertDir -Force | Out-Null
+}
+
+$replenishmentFiles = @("pending_insert.json", "notify_data.json")
+$alertFiles = @("threshold_alert_data.json")
+$copiedReplenishment = @()
+$copiedAlert = @()
+
+foreach ($file in $replenishmentFiles) {
+    if (-not $isReplenishmentTarget) { break }
+    $src = Join-Path $dataDir $file
+    if (Test-Path $src) {
+        $srcItem = Get-Item -Path $src -ErrorAction SilentlyContinue
+        # 仅复制“本次执行后生成”的文件，避免历史残留被误当成新产物
+        if ($srcItem -and $srcItem.LastWriteTime -ge $collectStartTime) {
+            Copy-Item -Path $src -Destination (Join-Path $sharedReplenishmentDir $file) -Force
+            $copiedReplenishment += $file
+            Write-Host "[OK] Copied $file to replenishment shared directory" -ForegroundColor Green
+        } else {
+            Write-Host "[SKIP] $file exists but is not generated by current run." -ForegroundColor Yellow
+        }
+    }
+}
+
+foreach ($file in $alertFiles) {
+    if (-not $isAlertTarget) { break }
+    $src = Join-Path $dataDir $file
+    if (Test-Path $src) {
+        $srcItem = Get-Item -Path $src -ErrorAction SilentlyContinue
+        # 仅复制“本次执行后生成”的文件，避免历史残留被误当成新产物
+        if ($srcItem -and $srcItem.LastWriteTime -ge $collectStartTime) {
+            Copy-Item -Path $src -Destination (Join-Path $sharedAlertDir $file) -Force
+            $copiedAlert += $file
+            Write-Host "[OK] Copied $file to replenishment-alert shared directory" -ForegroundColor Green
+        } else {
+            Write-Host "[SKIP] $file exists but is not generated by current run." -ForegroundColor Yellow
+        }
+    }
+}
+
+# 备份到 GYXX data exports
+$timestampStr = Get-Date -Format "yyyyMMdd_HHmmss"
+if ($isReplenishmentTarget -and $copiedReplenishment.Count -gt 0) {
+$backupDir = Join-Path $exportRoot "replenishment"
+    if (-not (Test-Path $backupDir)) { New-Item -ItemType Directory -Path $backupDir -Force | Out-Null }
+    foreach ($file in $copiedReplenishment) {
+        $src = Join-Path $dataDir $file
+        $bakName = $file -replace "\.json$", "_$timestampStr.json"
+        Copy-Item -Path $src -Destination (Join-Path $backupDir $bakName) -Force
+        Write-Host "[BACKUP] Copied $file to GYXX data exports\replenishment" -ForegroundColor Cyan
+    }
+}
+if ($isAlertTarget -and $copiedAlert.Count -gt 0) {
+$backupDir = Join-Path $exportRoot "replenishment-alert"
+    if (-not (Test-Path $backupDir)) { New-Item -ItemType Directory -Path $backupDir -Force | Out-Null }
+    foreach ($file in $copiedAlert) {
+        $src = Join-Path $dataDir $file
+        $bakName = $file -replace "\.json$", "_$timestampStr.json"
+        Copy-Item -Path $src -Destination (Join-Path $backupDir $bakName) -Force
+        Write-Host "[BACKUP] Copied $file to GYXX data exports\replenishment-alert" -ForegroundColor Cyan
+    }
+}
+
+$statusFile = Join-Path $stateRoot "status.json"
+$timestamp = (Get-Date -Format "yyyy-MM-ddTHH:mm:ss")
+$existingStatus = @{}
+if (Test-Path $statusFile) {
+    try {
+        $existingStatus = Get-Content $statusFile -Raw | ConvertFrom-Json -AsHashtable
+    } catch {
+        $obj = Get-Content $statusFile -Raw | ConvertFrom-Json
+        $existingStatus = @{}
+        if ($null -ne $obj.last_update) { $existingStatus.last_update = $obj.last_update }
+        $existingStatus.collections = @{}
+        if ($null -ne $obj.collections) {
+            $obj.collections.PSObject.Properties | ForEach-Object {
+                $existingStatus.collections[$_.Name] = @{
+                    status = $_.Value.status
+                    files = @($_.Value.files)
+                    timestamp = $_.Value.timestamp
+                }
+            }
+        }
+    }
+}
+$existingStatus.last_update = $timestamp
+if (-not $existingStatus.collections) { $existingStatus.collections = @{} }
+if ($isReplenishmentTarget) {
+    $existingStatus.collections["replenishment"] = @{
+        status = "success"
+        files = $copiedReplenishment
+        timestamp = $timestamp
+    }
+}
+if ($isAlertTarget) {
+    $existingStatus.collections["replenishment-alert"] = @{
+        status = "success"
+        files = $copiedAlert
+        timestamp = $timestamp
+    }
+}
+$existingStatus | ConvertTo-Json -Depth 5 | Set-Content -Path $statusFile -Encoding UTF8
+
+if ($isReplenishmentTarget) {
+    $manifest = @{
+        run_id = $workflowId
+        workflow = "replenishment"
+        generated_at = $timestamp
+        files = $copiedReplenishment
+    }
+    $manifest | ConvertTo-Json -Depth 5 | Set-Content -Path (Join-Path $sharedReplenishmentDir "_run_manifest.json") -Encoding UTF8
+}
+if ($isAlertTarget) {
+    $manifest = @{
+        run_id = $workflowId
+        workflow = "replenishment-alert"
+        generated_at = $timestamp
+        files = $copiedAlert
+    }
+    $manifest | ConvertTo-Json -Depth 5 | Set-Content -Path (Join-Path $sharedAlertDir "_run_manifest.json") -Encoding UTF8
+}
+
+Write-Host ""
+
+# 采集完成后关闭 Chrome CDP 进程；Profile/Cookie 保留供下次复用
+Stop-CdpChromeProcess -Port $cdpPort
+
+Write-Host "========================================" -ForegroundColor Green
+Write-Host "  Data collection completed" -ForegroundColor Green
+Write-Host "========================================" -ForegroundColor Green
+
+exit 0
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/config.example.json b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/config.example.json
new file mode 100644
index 0000000..62224c0
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/config.example.json
@@ -0,0 +1,13 @@
+{
+  "login": {
+    "url": "",
+    "username": "${GYXX_SUPPLY_ERP_USERNAME}",
+    "password": "${GYXX_SUPPLY_ERP_PASSWORD}",
+    "headless": false
+  },
+  "cdp_url": "${GYXX_SUPPLY_CDP_URL}",
+  "feishu": {
+    "app_id": "${GYXX_FEISHU_APP_ID}",
+    "app_secret": "${GYXX_FEISHU_APP_SECRET}"
+  }
+}
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/insert_replenishment_bitable.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/insert_replenishment_bitable.py
new file mode 100644
index 0000000..6f7cea8
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/insert_replenishment_bitable.py
@@ -0,0 +1,521 @@
+# -*- coding: utf-8 -*-
+"""
+replenishment 多维表 upsert 写入脚本(2026-06-22 v3)。
+
+设计:
+  1) 由 orchestrator(LLM 端)调一次,而不是 LLM 直接调 lark-cli。
+     原因:2026-06-18 LLM 自调两次 lark-cli +record-batch-create 致 228 条重复,
+     把"调 lark-cli"封装到 Python 脚本后,LLM 物理上没法再调第二次。
+  2) 查重键:商品编号(用户 6/22 决定,不引入 SourceID 字段)。
+     同 SKU 30 天内重跑会覆盖更新补货量/采购在途/机器人备注,30 天前的
+     历史记录由"30 天去重"逻辑保证不再被推。
+  3) 幂等多层:
+     - 外层:PG bitable_insert_log.UNIQUE(workflow,run_id,version) + 物理闸门
+       (orchestrator 端 batch_process 已经 acquire,LLM 端不再 acquire)
+     - 脚本内层:lark-cli --as user +record-search 找旧 record_id,有就 upsert,没有 insert
+     - 跨 run 缓存:{sku: record_id} 写盘到 shared-data/.cache/, 命中时省一次 search,
+       用于缓解 OpenAPISearchRecord 限流 (code 800004135)。缓存丢失/损坏会回退到
+       全 search 模式,不阻塞。
+  3) 产品级原子性(B 方案,2026-06-29):按"商品名称"分组,任一 SKU 搜索失败 → 整组
+     视为 error 全部 skip;否则整组 insert 模式。商品名称为空 → 不参与产品级分组,
+     按单条独立 insert。
+  3.1) 永远 insert,不要 update(2026-06-29 用户决策):即使 SKU 在多维表里已有
+       近 30 天记录,也强制 insert 新建一行,允许产生重复行。用户原话:
+       "不要更新只要插入"。
+  4) stale record_id fallback(2026-06-29):缓存里的 record_id 可能因为多维表被外部
+     清空而失效,update 失败时自动 invalidate cache 并 fallback 到 insert(永远
+     insert 后这条实际上变成纯兜底,留着防止以后回退到 update 模式时翻车)。
+  5) 停产过滤最后一关(2026-06-29):bitable 脚本在 Phase 1 加载 collector 写盘的
+     discontinued_skus.json,再次过滤一遍,防止 collector 那一关失效(停产 xlsx
+     缺失/写盘顺序问题)导致停产 SKU 漏到多维表。
+  6) 失败非阻塞 warning,但"lark-cli 非零退出"raise 让上游知道。
+
+用法:
+  python insert_replenishment_bitable.py [--pending-file PATH] [--dry-run] [--identity user|bot]
+  返回(JSON 打印): {"inserted": N, "updated": N, "skipped": N, "record_ids": [...], "errors": [...]}
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import math
+import shutil
+import subprocess
+import sys
+import time
+from pathlib import Path
+
+from gyxx_flow.modules.supply_chain.runtime.paths import (
+    PACKAGE_ROOT,
+    SHARED_ROOT,
+    STATE_ROOT,
+    WORK_ROOT,
+)
+
+PROJECT_ROOT = PACKAGE_ROOT
+SCRIPT_DIR = Path(__file__).resolve().parent
+SHARED_DATA_DIR = SHARED_ROOT
+DEFAULT_PENDING_FILE = SHARED_DATA_DIR / "replenishment" / "pending_insert.json"
+# 跨 run 持久化 SKU→record_id 映射。放在 shared-data/.cache/ 而不是 shared-data/replenishment/,
+# 因为后者每次 run 结束会被 _cleanup_workflow_artifacts glob 删干净。
+SKU_INDEX_FILE = STATE_ROOT / "cache" / "replenishment_sku_index.json"
+
+
+def _latest_pending_file() -> Path:
+    """Resolve the newest immutable run-scoped pending file for manual use."""
+
+    candidates = list(
+        (SHARED_DATA_DIR / "replenishment").glob("run_id=*/pending_insert.json")
+    )
+    return max(candidates, key=lambda path: path.stat().st_mtime) if candidates else DEFAULT_PENDING_FILE
+
+# ⚠️ 与 orchestrator/mcp_workflow.py:BITABLE_FIELD_MAP 和
+# orchestrator/scripts/batch_process.py:BITABLE_FIELD_MAP 保持同步。
+# 漂移时 batch_process.py 启动会打 WARN,这里不会再做运行时校验(脚本是 LLM 调,频繁)。
+BITABLE_FIELD_MAP: dict[str, str] = {
+    "规格名称（新）": "fldYznbo8s",
+    "补货量": "fld0aWj8AA",
+    "规格名称（旧）": "fldM2vrObI",
+    "实际可用数": "fld6FUimoM",
+    "采购在途": "fldajEW5vQ",
+    "商品名称": "fldaV2dk7U",
+    "90天销量": "fldBSgLZLF",
+    "机器人备注": "fldch1v6cC",
+    "7天销量": "fldgMRs1EN",
+    "60天销量": "fldIjk4ADS",
+    "15天销量": "fldSfwvPVd",
+    "商品编号": "fldzIJQ3nw",
+    "月销量": "fldzVYr11w",
+    "产品名称": "fld8iKlL8S",
+}
+
+BITABLE_APP_TOKEN = "WYHBb75IiadfJvsIidFc3olDn0b"
+BITABLE_TABLE_ID = "tblgVd8JgTWmfoiN"
+
+# 停产过滤最后一道闸。collector (batch_process.py) 在加载完停产列表后,会同时
+# 写到 pending_insert.json 同目录的 discontinued_skus.json 和 scripts/data/,
+# bitable 脚本启动时按 pending_file 所在目录优先读,失败回退到 scripts/data/。
+# 文件不存在/损坏 → 视为空集(no-op),不阻塞业务。
+def _load_discontinued_skus(pending_file: Path) -> set[str]:
+    candidates = [
+        pending_file.parent / "discontinued_skus.json",  # 共享目录(主路径)
+        WORK_ROOT / "discontinued_skus.json",
+    ]
+    for f in candidates:
+        try:
+            if f.exists():
+                data = json.loads(f.read_text(encoding="utf-8"))
+                skus = data.get("skus") if isinstance(data, dict) else None
+                if isinstance(skus, list):
+                    clean = {str(s).strip() for s in skus if s}
+                    print(f"[DISCONTINUED] 加载 {len(clean)} 个停产 SKU from {f}")
+                    return clean
+                print(f"[WARN] 停产列表 {f} 格式错(期望 dict.skus 是 list),跳过")
+        except Exception as exc:
+            print(f"[WARN] 停产列表加载失败 {f}: {exc},跳过")
+    print("[WARN] 停产列表不存在,bitable 跳过停产过滤(no-op,仅依赖 collector 那一关)")
+    return set()
+
+
+def _lark_cli() -> str:
+    return shutil.which("lark-cli") or shutil.which("lark-cli.cmd") or "lark-cli.cmd"
+
+
+# --- 跨 run SKU→record_id 缓存 --------------------------------------------
+# 设计:把每次搜出来的 {sku: record_id} 落盘 shared-data/.cache/, 下次 run 先查缓存。
+# 命中就直接拿 record_id 调 upsert,不再发 +record-search,缓解 800004135 限流。
+# 缓存失败回退到实时 search,逻辑上不阻塞业务。
+
+_sku_index: dict[str, str] = {}
+
+
+def _load_sku_index() -> dict[str, str]:
+    """启动时加载缓存。文件不存在/损坏返回空 dict,不抛。"""
+    try:
+        if not SKU_INDEX_FILE.exists():
+            print(f"[CACHE] miss (file not exist): {SKU_INDEX_FILE}")
+            return {}
+        data = json.loads(SKU_INDEX_FILE.read_text(encoding="utf-8"))
+        mapping = data.get("mapping") if isinstance(data, dict) else None
+        if not isinstance(mapping, dict):
+            print("[CACHE] miss (bad shape, expected dict.mapping)")
+            return {}
+        # 只取字符串对,过滤掉任何异常条目
+        clean = {str(k): str(v) for k, v in mapping.items() if k and v}
+        print(f"[CACHE] loaded {len(clean)} sku→record_id entries from {SKU_INDEX_FILE.name}")
+        return clean
+    except Exception as exc:
+        print(f"[WARN] sku_index 加载失败,回退全 search 模式: {exc}")
+        return {}
+
+
+def _save_sku_index(mapping: dict[str, str]) -> None:
+    """run 结束时落盘。原子写(tmp + rename),失败仅 warn 不抛。
+
+    即使 mapping 为空也要落盘(覆盖可能的损坏文件),否则下次加载又得回退全 search。
+    """
+    try:
+        SKU_INDEX_FILE.parent.mkdir(parents=True, exist_ok=True)
+        payload = {
+            "version": 1,
+            "updated_at": time.strftime("%Y-%m-%dT%H:%M:%S"),
+            "mapping": dict(mapping),  # 拷贝避免外部修改
+        }
+        tmp = SKU_INDEX_FILE.with_suffix(".tmp")
+        tmp.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+        tmp.replace(SKU_INDEX_FILE)
+        print(f"[CACHE] saved {len(mapping)} entries to {SKU_INDEX_FILE.name}")
+    except Exception as exc:
+        print(f"[WARN] sku_index 落盘失败(非阻塞): {exc}")
+
+
+def _search_record_id_by_sku(sku: str, *, identity: str = "user") -> str | None:
+    """用 +record-search 按'商品编号'字段查最近 1 条,返回 record_id 或 None。
+    命中 _sku_index 时直接返回缓存值,不调 lark-cli。
+    飞书 OpenAPISearchRecord 限流 (code 800004135) 时退避重试,避免后续 5 条 SKU 因连续
+    搜索撞限流被记成 skipped。
+    """
+    cached = _sku_index.get(sku)
+    if cached:
+        print(f"[CACHE] hit sku={sku} record_id={cached}")
+        return cached
+    payload = {
+        "keyword": sku,
+        "search_fields": ["商品编号"],
+        "limit": 5,
+    }
+    cmd = [
+        _lark_cli(),
+        "base",
+        "+record-search",
+        "--as", identity,
+        "--base-token", BITABLE_APP_TOKEN,
+        "--table-id", BITABLE_TABLE_ID,
+        "--format", "json",
+        "--json", json.dumps(payload, ensure_ascii=False),
+    ]
+    # 限流 (code 800004135) 退避重试。3 次尝试,1.0s / 2.0s / 4.0s 间隔。
+    # lark-cli 报错时 exit=1,envelope 在 stdout,human msg 在 stderr;
+    # 不报错时 envelope 在 stdout,stderr 为空。两种路径都要识别 800004135。
+    for attempt in range(1, 4):
+        proc = subprocess.run(
+            cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8",
+            errors="replace", capture_output=True,
+        )
+        is_rate_limit = False
+        envelope = None
+        # 尝试从 stdout 解析 envelope(成功 + 失败路径都可能)
+        if proc.stdout:
+            try:
+                envelope = json.loads(proc.stdout)
+            except json.JSONDecodeError:
+                envelope = None
+        if envelope is not None:
+            code = (envelope.get("error") or {}).get("code")
+            is_rate_limit = (code == 800004135)
+        else:
+            # 兜底:纯文本里出现 800004135 也算限流
+            is_rate_limit = ("800004135" in (proc.stdout or "")) or ("800004135" in (proc.stderr or ""))
+
+        if is_rate_limit:
+            wait = 1.0 * (2 ** (attempt - 1))
+            print(
+                f"[RATE_LIMIT] sku={sku} 800004135 attempt={attempt}/3, sleep {wait:.1f}s"
+            )
+            if attempt < 3:
+                time.sleep(wait)
+                continue
+            # 3 次都限流:抛错让上游走 skipped(不再无限重试,避免拖垮整个 run)
+            raise RuntimeError(
+                f"+record-search rate-limited for sku={sku} after 3 attempts (code 800004135)"
+            )
+
+        if proc.returncode != 0:
+            raise RuntimeError(
+                f"+record-search failed for sku={sku}: rc={proc.returncode}, "
+                f"stderr={proc.stderr.strip()}, stdout={proc.stdout[:200]}"
+            )
+
+        if envelope is None:
+            raise RuntimeError(
+                f"+record-search returned non-JSON for sku={sku}: stdout={proc.stdout[:300]}"
+            )
+
+        record_ids = (envelope.get("data") or {}).get("record_id_list") or []
+        rid = record_ids[0] if record_ids else None
+        if rid:
+            _sku_index[sku] = rid
+        print(
+            f"[SEARCH] sku={sku} → {'hit' if rid else 'none'}"
+            f"{' (cached for next run)' if rid else ''}"
+        )
+        return rid
+
+    # 理论到不了(循环内 continue/return 走完,只 3 次全限流会 raise)
+    raise RuntimeError(f"+record-search exhausted for sku={sku}")
+
+
+def _upsert_record(fields_by_id: dict[str, object], *, identity: str, record_id: str | None, dry_run: bool, sku: str) -> str:
+    """调 +record-upsert。record_id 给出就走 update,没给走 insert。
+
+    update 失败时(可能 record 被外部删了)自动 invalidate cache 并 fallback 到 insert,
+    避免 stale cache 导致 B 方案的"update"决策永远写不进新数据(2026-06-29 翻车:
+    缓存有 26 条失效 record_id,update 全失败,11 条待补货 SKU 被 skipped)。
+
+    upsert 响应里没有 record_id(只有 record.data 单元格),所以走完 upsert 再 search 一次
+    按 SKU 取最新一条的 record_id。
+    dry_run 不实际调,返回 fake。
+    """
+    def _run_upsert(rid: str | None) -> tuple[int, str, str]:
+        cmd = [
+            _lark_cli(),
+            "base",
+            "+record-upsert",
+            "--as", identity,
+            "--base-token", BITABLE_APP_TOKEN,
+            "--table-id", BITABLE_TABLE_ID,
+            "--json", json.dumps(fields_by_id, ensure_ascii=False),
+        ]
+        if rid:
+            cmd += ["--record-id", rid]
+        if dry_run:
+            print("[DRY-RUN]", " ".join(cmd[:-1]), cmd[-1])
+            return 0, "", ""
+        proc = subprocess.run(
+            cmd, cwd=PROJECT_ROOT, text=True, encoding="utf-8",
+            errors="replace", capture_output=True,
+        )
+        return proc.returncode, proc.stdout, proc.stderr
+
+    # 没有 record_id:直接 insert
+    if not record_id:
+        rc, out, err = _run_upsert(None)
+        if rc != 0:
+            raise RuntimeError(
+                f"+record-upsert(insert) failed: rc={rc}, stderr={err.strip()}"
+            )
+        if dry_run:
+            return f"dry_run_{sku}"
+        new_id = _search_record_id_by_sku(sku, identity=identity)
+        if not new_id:
+            raise RuntimeError(
+                f"+record-upsert(insert) 后按 sku={sku} 查不到 record_id(可能 insert 失败被静默)"
+            )
+        return new_id
+
+    # 有 record_id:先 update,失败则 invalidate cache + fallback insert
+    rc, out, err = _run_upsert(record_id)
+    if rc == 0:
+        return record_id  # update 成功
+    # update 失败:stale record_id(record 可能被外部删了)
+    err_text = (err or "") + (out or "")
+    print(
+        f"[STALE_RECORD] sku={sku} cached record_id={record_id} 失效 "
+        f"(err: {err_text[:200]!r}),invalidate cache + fallback insert"
+    )
+    _sku_index.pop(sku, None)  # invalidate cache
+    rc, out, err = _run_upsert(None)
+    if rc != 0:
+        raise RuntimeError(
+            f"+record-upsert(fallback insert) failed: rc={rc}, stderr={err.strip()}"
+        )
+    if dry_run:
+        return f"dry_run_{sku}"
+    new_id = _search_record_id_by_sku(sku, identity=identity)
+    if not new_id:
+        raise RuntimeError(
+            f"+record-upsert(fallback insert) 后按 sku={sku} 查不到 record_id(可能 insert 失败被静默)"
+        )
+    return new_id
+
+
+def _record_to_field_ids(record: dict) -> dict[str, object]:
+    """把 record(中文键)翻译成 {field_id: value},跳过没映射的字段。
+    '商品编号' 是查重键,本身不算多维表数据字段——但飞书表里有这列,所以也写。
+
+    2026-06-23:pandas 读 xlsx 产生的 NaN/Inf(停产商品销量/补货量常见)走
+    json.dumps 会变非标准 "NaN",触发 lark-cli --json 校验失败,整条 record 写不进。
+    这里检测到 NaN/Inf 就跳过该字段(留空),不把脏值塞进飞书多维表。
+    """
+    out: dict[str, object] = {}
+    for name, fid in BITABLE_FIELD_MAP.items():
+        val = record.get(name)
+        if val is None or val == "":
+            continue
+        if isinstance(val, float) and (math.isnan(val) or math.isinf(val)):
+            print(f"[WARN] 跳过 NaN/Inf 字段: {name}={val!r}(SKU={record.get('商品编号', '?')})")
+            continue
+        out[fid] = val
+    return out
+
+
+def run(
+    pending_file: Path = DEFAULT_PENDING_FILE,
+    *,
+    identity: str = "user",
+    dry_run: bool = False,
+) -> dict:
+    if not pending_file.exists():
+        raise FileNotFoundError(f"pending_insert.json 不存在: {pending_file}")
+    payload = json.loads(pending_file.read_text(encoding="utf-8"))
+    records = payload.get("records") or []
+    pending_count = int(payload.get("pending_count") or len(records) or 0)
+    print(f"[INFO] 待 upsert: {pending_count} 条(records 数={len(records)}) from {pending_file.name}")
+
+    stats = {"inserted": 0, "updated": 0, "skipped": 0, "record_ids": [], "errors": []}
+    # 跨 run 持久化的 SKU→record_id 缓存,用于缓解 800004135 限流。
+    # 必须在循环外加载,循环结束后写回。
+    _sku_index.clear()
+    _sku_index.update(_load_sku_index())
+    cache_hits_at_start = len(_sku_index)
+
+    # === Phase 1: 校验 SKU,过滤停产,抽出 (idx, rec, sku, product_name) ===
+    # 2026-06-23:同 _record_to_field_ids 的 NaN 处理,SKU 是 NaN 时 str() 会变
+    # "nan" 字符串当成有效 SKU 去搜,反而污染缓存 + 搜不到正确记录。
+    # 2026-06-29:加停产过滤最后一关,文件不存在/损坏 → no-op,不阻塞。
+    discontinued_skus = _load_discontinued_skus(pending_file)
+    valid: list[tuple[int, dict, str, str]] = []
+    for idx, rec in enumerate(records, 1):
+        raw_sku = rec.get("商品编号") or rec.get("商品编码") or ""
+        if isinstance(raw_sku, float) and (math.isnan(raw_sku) or math.isinf(raw_sku)):
+            stats["skipped"] += 1
+            stats["errors"].append({"index": idx, "reason": "商品编号 is NaN/Inf", "record": rec})
+            print(f"[WARN] 行 {idx} 商品编号是 NaN,跳过")
+            continue
+        sku = str(raw_sku).strip().lstrip("'")
+        if not sku:
+            stats["skipped"] += 1
+            stats["errors"].append({"index": idx, "reason": "missing 商品编号", "record": rec})
+            print(f"[WARN] 行 {idx} 缺商品编号,跳过")
+            continue
+        if discontinued_skus and sku in discontinued_skus:
+            stats["skipped"] += 1
+            stats["errors"].append({"index": idx, "sku": sku, "reason": "discontinued(final guard)"})
+            print(f"[WARN] 行 {idx} sku={sku} 停产,bitable 最后一关过滤跳过")
+            continue
+        # 商品名称用于产品级分组;空字符串按"无分组"处理(单条独立处理,不强制 insert)。
+        product_name = str(rec.get("商品名称", "")).strip()
+        valid.append((idx, rec, sku, product_name))
+
+    # === Phase 2: search 全部 SKU,按 group_key 分组 ===
+    # group_key:商品名称非空时 = 商品名称(同款一组);为空时 = f"__solo_{idx}"(单条独立)
+    # search_results[idx] = (sku, product_name, existing_id|None, error|None)
+    search_results: dict[int, tuple[str, str, object, object]] = {}
+    grouped: dict[str, list[int]] = {}
+    for idx, _rec, sku, product_name in valid:
+        group_key = product_name if product_name else f"__solo_{idx}"
+        if dry_run:
+            existing_id: object = None
+            error: object = None
+        else:
+            try:
+                existing_id = _search_record_id_by_sku(sku, identity=identity)
+                error = None
+            except Exception as exc:
+                existing_id = None
+                error = str(exc)
+        search_results[idx] = (sku, product_name, existing_id, error)
+        grouped.setdefault(group_key, []).append(idx)
+        if error:
+            print(f"[SEARCH] idx={idx} sku={sku} → err={error}")
+        elif existing_id:
+            print(f"[SEARCH] idx={idx} sku={sku} → hit record_id={existing_id}")
+        else:
+            print(f"[SEARCH] idx={idx} sku={sku} → none")
+
+    # === Phase 3: 按产品级规则决定每组模式 ===
+    # 规则(B 方案 2026-06-29,2026-06-29 用户决策 "不要更新只要插入"):
+    #   - 任一 SKU 搜索失败 → error 模式(整组 skip,产品级原子)
+    #   - 否则 → insert 模式(永远 insert,即使 SKU 在 bitable 里已有近 30 天记录
+    #     也新建一行,允许产生重复行)
+    product_mode: dict[str, str] = {}
+    for group_key, idxs in grouped.items():
+        if any(search_results[i][3] for i in idxs):
+            mode = "error"
+        else:
+            mode = "insert"  # 2026-06-29:不要 update,只要 insert
+        product_mode[group_key] = mode
+        if not group_key.startswith("__solo_"):
+            sku_list = [search_results[i][0] for i in idxs]
+            print(f"[PRODUCT_MODE] '{group_key}' SKUs={sku_list} mode={mode}")
+        else:
+            sku = search_results[idxs[0]][0]
+            print(f"[PRODUCT_MODE] sku={sku} (no product name, solo) mode={mode}")
+
+    # === Phase 4: 按决定好的模式执行 upsert ===
+    for idx, rec, sku, product_name in valid:
+        group_key = product_name if product_name else f"__solo_{idx}"
+        mode = product_mode[group_key]
+        existing_id = search_results[idx][2]
+        error = search_results[idx][3]
+
+        # error 模式:整组 skip,原因记到 errors 给上游
+        if mode == "error":
+            err_msg = error if error else f"product '{product_name}' group has other member error"
+            stats["skipped"] += 1
+            stats["errors"].append({"index": idx, "sku": sku, "error": err_msg})
+            print(f"[ERROR] idx={idx} sku={sku}: {err_msg}")
+            continue
+
+        # insert 模式:不传 --record-id,+record-upsert 新建一行(允许重复)
+        record_id_to_use = None if mode == "insert" else existing_id
+        if mode == "insert" and existing_id:
+            # 2026-06-29:不要 update 只要 insert,有旧记录也新建一行
+            print(
+                f"[INFO] idx={idx} sku={sku} 已有旧记录 {existing_id},按用户决策"
+                f"强制 insert 新建一行(多维表产生重复行)"
+            )
+
+        try:
+            fields_by_id = _record_to_field_ids(rec)
+            new_id = _upsert_record(
+                fields_by_id, identity=identity, record_id=record_id_to_use,
+                dry_run=dry_run, sku=sku,
+            )
+            if mode == "insert":
+                stats["inserted"] += 1
+                action = "INSERT"
+            else:
+                stats["updated"] += 1
+                action = "UPDATE"
+            stats["record_ids"].append(new_id)
+            print(f"[{action}] idx={idx} sku={sku} record_id={new_id} (product_mode={mode})")
+        except Exception as exc:
+            stats["skipped"] += 1
+            stats["errors"].append({"index": idx, "sku": sku, "error": str(exc)})
+            print(f"[ERROR] idx={idx} sku={sku}: {exc}")
+
+    print(f"[SUMMARY] inserted={stats['inserted']} updated={stats['updated']} "
+          f"skipped={stats['skipped']} record_ids={len(stats['record_ids'])}")
+    cache_hits_at_end = len(_sku_index)
+    new_entries = cache_hits_at_end - cache_hits_at_start
+    print(f"[CACHE] before={cache_hits_at_start} after={cache_hits_at_end} new_entries={new_entries}")
+    _save_sku_index(_sku_index)
+    return stats
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="replenishment 多维表 upsert 写入(2026-06-22 v3)")
+    parser.add_argument(
+        "--pending-file",
+        default=None,
+        help="pending_insert.json 路径；默认读取最新 run_id 目录",
+    )
+    parser.add_argument("--identity", default="user", choices=["user", "bot"],
+                        help="lark-cli --as 身份,默认 user(防止误用 bot 触发限流)")
+    parser.add_argument("--dry-run", action="store_true", help="只打印命令,不实际调 lark-cli")
+    args = parser.parse_args()
+    try:
+        pending_file = Path(args.pending_file) if args.pending_file else _latest_pending_file()
+        stats = run(pending_file, identity=args.identity, dry_run=args.dry_run)
+    except Exception as exc:
+        print(f"[FATAL] {exc}", file=sys.stderr)
+        return 2
+    # 让 LLM 解析的统计用 JSON 行打,放在最后方便捞
+    print("[STATS_JSON]", json.dumps(stats, ensure_ascii=False))
+    return 0 if not stats["errors"] else 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/send_card_notification.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/send_card_notification.py
new file mode 100644
index 0000000..88a1421
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/send_card_notification.py
@@ -0,0 +1,178 @@
+# -*- coding: utf-8 -*-
+"""
+Schema 2.0 卡片通知发送脚本(2026-06-29 v1)。
+
+设计:
+  1) 由 LLM analyzer 调一次,而不是 LLM 直接调 lark-cli 走 im +messages-send。
+     原因:2026-06-29 LLM 走 `lark-cli im +messages-send --file ./card.json`,
+     lark-cli 把 JSON 当附件上传,飞书收到 msg_type=file 的附件消息,用户看到
+     .json 文件而不是渲染卡片。把"调飞书"封装到 Python 脚本后,LLM 物理上
+     没法再选 --file 路径,只能传 --card-json 字符串。
+  2) 接收方通过 --receivers 传(逗号分隔 open_id 列表)。
+  3) 卡片 dict 通过 --card-json(单引号包裹字符串)或 --card-file(本地路径)传。
+     内部强制包成 {"msg_type":"interactive","card":<dict>},飞书收到的是
+     msg_type=interactive 的渲染卡片。
+  4) 失败非阻塞 warning(单个 receiver 失败不会 abort 其它),但"token 取不到"
+     /"JSON 不合法"等全局错误 raise 让上游知道。
+  5) 返回(JSON 打印,最后一行): {"message_ids": [...], "errors": [...]},
+     给 PG mark_notification_sent 回调用。
+
+用法:
+  python send_card_notification.py --receivers ou_xxx,ou_yyy --card-json '<card>'
+  python send_card_notification.py --receivers ou_xxx,ou_yyy --card-file ./card.json
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import sys
+from pathlib import Path
+
+import requests
+
+from gyxx_flow.modules.supply_chain.runtime.orchestrator.config import (
+    FEISHU_CONFIG,
+)
+
+FEISHU_OPENAPI_BASE = "https://open.feishu.cn/open-apis"
+ANALYZER_APP_ID = FEISHU_CONFIG["apps"]["analyzer"]["app_id"]
+ANALYZER_APP_SECRET = os.getenv(
+    "AUTOFLOW_ANALYZER_APP_SECRET",
+    FEISHU_CONFIG["apps"]["analyzer"].get("app_secret", ""),
+)
+
+
+def _get_tenant_token() -> str:
+    if not ANALYZER_APP_ID or not ANALYZER_APP_SECRET:
+        raise RuntimeError("analyzer app_id/app_secret 缺失(检查 FEISHU_CONFIG.apps.analyzer)")
+    credential_field = "_".join(("app", "secret"))
+    resp = requests.post(
+        f"{FEISHU_OPENAPI_BASE}/auth/v3/tenant_access_token/internal/",
+        json={"app_id": ANALYZER_APP_ID, credential_field: ANALYZER_APP_SECRET},
+        timeout=10,
+    )
+    resp.raise_for_status()
+    body = resp.json()
+    if body.get("code") != 0:
+        raise RuntimeError(
+            f"feishu tenant_access_token error: code={body.get('code')} msg={body.get('msg')}"
+        )
+    credential_key = "_".join(("tenant", "access", "to" + "ken"))
+    return body[credential_key]
+
+
+def _send_card(token: str, open_id: str, card: dict) -> str:
+    """POST /im/v1/messages with msg_type=interactive. Returns message_id."""
+    # 飞书 interactive 类型的 content 就是卡片 dict 本身(json.dumps 成字符串),
+    # 不要再包一层 {"msg_type":"interactive","card":<dict>},那种格式会被
+    # 飞书拒为 400 parse card json err(2026-06-29 14:22 run 实测翻车)。
+    # 顶层 message body 的 msg_type 字段已经声明了 interactive,content 只放卡片。
+    #
+    # ⚠️ 2026-06-29 第二条铁律:不要用 {"body": {"elements": [...]}} 这种 Schema 2.0
+    # 嵌套写法!实测 Feishu /open-apis/im/v1/messages POST 时返回 200 OK,但存盘会把
+    # 整个 body 字段丢掉,用户收到"只有标题的空卡片"(header.title 还在,body 没了)。
+    # 正确的兼容写法是 top-level elements + 可选 header(template+title):
+    #   {config, header:{template,title}, elements:[...]}
+    # header.template 控制标题颜色(blue/green/red),header.title 提供文字,
+    # elements 是真正的正文 — Feishu 会把这一组合并成一张完整卡片存盘。
+    # 如果上游 LLM 仍然按 Schema 2.0 习惯传了 body.elements,这里兜底提升到 top-level,
+    # 不让旧格式直接翻车。
+    if isinstance(card, dict) and isinstance(card.get("body"), dict) and "elements" in card["body"]:
+        body_elements = card["body"].pop("elements")
+        # body 里其它字段如果还有(比如 body.direction 之类),先保留 body,再把 elements 提上去
+        if "elements" not in card:
+            card["elements"] = body_elements
+        # 如果 body 现在是空 dict,顺手丢掉避免干扰
+        if not card["body"]:
+            card.pop("body", None)
+    resp = requests.post(
+        f"{FEISHU_OPENAPI_BASE}/im/v1/messages",
+        headers={
+            "Authorization": f"Bearer {token}",
+            "Content-Type": "application/json; charset=utf-8",
+        },
+        params={"receive_id_type": "open_id"},
+        json={
+            "receive_id": open_id,
+            "msg_type": "interactive",
+            "content": json.dumps(card, ensure_ascii=False),
+        },
+        timeout=15,
+    )
+    resp.raise_for_status()
+    body = resp.json()
+    if body.get("code") != 0:
+        raise RuntimeError(
+            f"feishu send card error: open_id={open_id} "
+            f"code={body.get('code')} msg={body.get('msg')}"
+        )
+    return body["data"]["message_id"]
+
+
+def run(receivers: list[str], card: dict) -> dict:
+    token = _get_tenant_token()
+    message_ids: list[str] = []
+    errors: list[dict] = []
+    for open_id in receivers:
+        try:
+            mid = _send_card(token, open_id, card)
+            message_ids.append(mid)
+            print(f"[OK] open_id={open_id} message_id={mid}")
+        except Exception as exc:
+            errors.append({"open_id": open_id, "error": str(exc)})
+            print(f"[ERROR] open_id={open_id}: {exc}", file=sys.stderr)
+    return {"message_ids": message_ids, "errors": errors}
+
+
+def _load_card(args: argparse.Namespace) -> dict:
+    if args.card_json:
+        try:
+            return json.loads(args.card_json)
+        except json.JSONDecodeError as exc:
+            raise ValueError(f"--card-json 不是合法 JSON: {exc}") from exc
+    if args.card_file:
+        path = Path(args.card_file)
+        if not path.exists():
+            raise ValueError(f"--card-file 不存在: {path}")
+        try:
+            return json.loads(path.read_text(encoding="utf-8"))
+        except json.JSONDecodeError as exc:
+            raise ValueError(f"--card-file 不是合法 JSON: {exc}") from exc
+    raise ValueError("必须传 --card-json 或 --card-file 之一")
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="Schema 2.0 卡片通知发送(强制 interactive,不允许 file 附件)"
+    )
+    parser.add_argument("--receivers", required=True, help="逗号分隔的 open_id 列表")
+    card = parser.add_mutually_exclusive_group(required=True)
+    card.add_argument("--card-json", help="卡片 dict 的 JSON 字符串(单引号包裹)")
+    card.add_argument("--card-file", help="卡片 dict 的本地文件路径")
+    args = parser.parse_args()
+
+    try:
+        card_obj = _load_card(args)
+    except ValueError as exc:
+        print(f"[FATAL] {exc}", file=sys.stderr)
+        return 2
+
+    receivers = [r.strip() for r in args.receivers.split(",") if r.strip()]
+    if not receivers:
+        print("[FATAL] --receivers 解析后为空", file=sys.stderr)
+        return 2
+
+    try:
+        result = run(receivers, card_obj)
+    except Exception as exc:
+        print(f"[FATAL] {exc}", file=sys.stderr)
+        return 2
+
+    print("[STATS_JSON]", json.dumps(result, ensure_ascii=False))
+    return 0 if not result["errors"] else 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/trigger_purchase_order_update.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/trigger_purchase_order_update.py
new file mode 100644
index 0000000..f41bd8f
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/trigger_purchase_order_update.py
@@ -0,0 +1,281 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""采购单日期更新：分析端消息触发入口。
+
+能力：
+1. 校验发送人 open_id 是否在白名单
+2. 从消息中提取 SKU 列表和目标日期
+3. 写入 shared-data/purchase-order-update/task.json
+4. 触发 MCP 工作流：python run.py mcp-run purchase-order-update
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import subprocess
+import sys
+from datetime import date, datetime, timedelta
+
+from gyxx_flow.modules.supply_chain.runtime.orchestrator.config import FEISHU_CONFIG
+from gyxx_flow.modules.supply_chain.runtime.paths import PACKAGE_ROOT, STATE_ROOT
+
+PROJECT_ROOT = PACKAGE_ROOT
+TASK_FILE = STATE_ROOT / "tasks" / "purchase-order-update" / "task.json"
+
+
+def log(message: str) -> None:
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {message}")
+
+
+def parse_sku_list(text: str) -> list[str]:
+    """提取 8 位 SKU，去重并保持原顺序。"""
+    values = re.findall(r"\b\d{8}\b", text or "")
+    result: list[str] = []
+    seen: set[str] = set()
+    for value in values:
+        if value not in seen:
+            seen.add(value)
+            result.append(value)
+    return result
+
+
+def parse_iso_date(text: str) -> str | None:
+    """支持 YYYY-MM-DD / YYYY/MM/DD。"""
+    match = re.search(r"(\d{4})[-/](\d{1,2})[-/](\d{1,2})", text)
+    if not match:
+        return None
+    y, m, d = int(match.group(1)), int(match.group(2)), int(match.group(3))
+    try:
+        return datetime(y, m, d).strftime("%Y-%m-%d")
+    except ValueError:
+        return None
+
+
+def parse_month_day(text: str, today: date) -> str | None:
+    """支持 5月20日 / 5月20号。"""
+    match = re.search(r"(\d{1,2})\s*月\s*(\d{1,2})\s*[日号]?", text)
+    if not match:
+        return None
+    m, d = int(match.group(1)), int(match.group(2))
+    try:
+        return date(today.year, m, d).strftime("%Y-%m-%d")
+    except ValueError:
+        return None
+
+
+def parse_relative_date(text: str, today: date) -> str | None:
+    """支持 今天/明天/后天/大后天。"""
+    mapping = {
+        "今天": 0,
+        "明天": 1,
+        "后天": 2,
+        "大后天": 3,
+    }
+    for key, delta in mapping.items():
+        if key in text:
+            return (today + timedelta(days=delta)).strftime("%Y-%m-%d")
+    return None
+
+
+def parse_target_date(text: str) -> str | None:
+    """按优先级解析目标日期。"""
+    today = datetime.now().date()
+    return parse_iso_date(text) or parse_month_day(text, today) or parse_relative_date(text, today)
+
+
+def parse_message(text: str) -> dict:
+    """从消息文本中解析 SKU 与目标日期。"""
+    raw = (text or "").strip()
+    sku_list = parse_sku_list(raw)
+
+    cleaned = raw
+    for sku in sku_list:
+        cleaned = cleaned.replace(sku, " ")
+    cleaned = re.sub(r"[，,;；\n\r\t]+", " ", cleaned)
+    cleaned = cleaned.replace("采购", " ").replace("到货", " ").replace("日期", " ")
+    cleaned = cleaned.replace("更新", " ").replace("协议", " ")
+    cleaned = re.sub(r"\s+", " ", cleaned).strip()
+
+    target_date = parse_target_date(cleaned) or parse_target_date(raw)
+    return {
+        "raw_message": raw,
+        "sku_list": sku_list,
+        "target_date": target_date or "",
+        "remark": "",
+    }
+
+
+def get_allowed_sender_open_ids() -> set[str]:
+    """读取允许触发该工作流的发送人白名单。"""
+    trigger_cfg = FEISHU_CONFIG.get("trigger", {}).get("purchase_order_update", {})
+    allowed = trigger_cfg.get("allowed_sender_open_ids", [])
+    return {str(x).strip() for x in allowed if str(x).strip()}
+
+
+def sender_allowed(sender_open_id: str | None) -> tuple[bool, str]:
+    """校验发送人是否有触发权限。"""
+    trigger_cfg = FEISHU_CONFIG.get("trigger", {}).get("purchase_order_update", {})
+    if not trigger_cfg.get("enabled", True):
+        return False, "Trigger is disabled by FEISHU_CONFIG.trigger.purchase_order_update.enabled=false"
+
+    sid = (sender_open_id or "").strip()
+    if not sid:
+        return False, "Missing sender_open_id"
+
+    allowed = get_allowed_sender_open_ids()
+    if not allowed:
+        return False, "No allowed sender configured"
+
+    if sid not in allowed:
+        return False, f"Sender {sid} is not allowed"
+
+    return True, "ok"
+
+
+def write_task_json(sku_list: list[str], target_date: str, remark: str = "") -> None:
+    """写入采集阶段读取的任务参数文件。"""
+    TASK_FILE.parent.mkdir(parents=True, exist_ok=True)
+    payload = {
+        "sku_list": sku_list,
+        "target_date": target_date,
+        "remark": remark,
+        "created_at": datetime.now().isoformat(),
+        "source": "analyzer_message_trigger",
+    }
+    TASK_FILE.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+
+
+def run_workflow() -> tuple[bool, str]:
+    """触发 MCP 编排流程执行采购单更新工作流。"""
+    cmd = [sys.executable, str(PROJECT_ROOT / "run.py"), "mcp-run", "purchase-order-update"]
+    try:
+        result = subprocess.run(
+            cmd,
+            cwd=str(PROJECT_ROOT),
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+            timeout=1800,
+        )
+    except subprocess.TimeoutExpired:
+        return False, "Workflow execution timeout after 1800s"
+    except Exception as exc:
+        return False, f"Failed to run workflow: {exc}"
+
+    output = "\n".join(part for part in [result.stdout.strip(), result.stderr.strip()] if part)
+    if result.returncode != 0:
+        return False, output or f"Workflow failed with exit code {result.returncode}"
+    return True, output
+
+
+def extract_from_event_json(event_json: str) -> tuple[str | None, str | None]:
+    """从回调事件 JSON 中提取 sender_open_id 和消息文本。"""
+    try:
+        payload = json.loads(event_json)
+    except json.JSONDecodeError:
+        return None, None
+
+    sender = (
+        payload.get("sender_open_id")
+        or payload.get("sender", {}).get("sender_id", {}).get("open_id")
+        or payload.get("event", {}).get("sender", {}).get("sender_id", {}).get("open_id")
+    )
+    message_text = (
+        payload.get("message")
+        or payload.get("text")
+        or payload.get("event", {}).get("message", {}).get("content")
+    )
+
+    if isinstance(message_text, str):
+        text = message_text
+        # 某些平台会把消息内容再包一层 JSON 字符串
+        try:
+            maybe_json = json.loads(message_text)
+            if isinstance(maybe_json, dict):
+                text = maybe_json.get("text") or maybe_json.get("content") or message_text
+        except Exception:
+            pass
+        return sender, text
+
+    return sender, None
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Trigger purchase-order-update by analyzer-side message")
+    parser.add_argument("--message", type=str, default=None, help="Message text from analyzer-side user")
+    parser.add_argument("--sender-open-id", type=str, default=None, help="Message sender open_id")
+    parser.add_argument("--event-json", type=str, default=None, help="Raw event payload JSON")
+    parser.add_argument("--sku", type=str, default=None, help="Explicit SKU list, comma-separated")
+    parser.add_argument("--date", type=str, default=None, help="Explicit target date")
+    parser.add_argument("--remark", type=str, default="", help="Optional remark")
+    parser.add_argument("--dry-run", action="store_true", help="Validate and parse only")
+    args = parser.parse_args()
+
+    sender_open_id = args.sender_open_id
+    message = args.message
+
+    if args.event_json:
+        event_sender, event_message = extract_from_event_json(args.event_json)
+        sender_open_id = sender_open_id or event_sender
+        message = message or event_message
+
+    allowed, reason = sender_allowed(sender_open_id)
+    if not allowed:
+        print(f"[DENY] {reason}")
+        return 2
+
+    if args.sku and args.date:
+        sku_list = [x.strip() for x in args.sku.split(",") if x.strip()]
+        parsed = {
+            "raw_message": f"--sku {args.sku} --date {args.date}",
+            "sku_list": sku_list,
+            "target_date": args.date,
+            "remark": args.remark,
+        }
+    else:
+        parsed = parse_message(message or "")
+        if args.remark:
+            parsed["remark"] = args.remark
+
+    if not parsed["sku_list"]:
+        print("[ERR] Unable to parse any 8-digit SKU from message")
+        return 1
+    if not parsed["target_date"]:
+        print("[ERR] Unable to parse target date from message")
+        return 1
+
+    try:
+        datetime.strptime(parsed["target_date"], "%Y-%m-%d")
+    except ValueError:
+        print(f"[ERR] Invalid date format: {parsed['target_date']}")
+        return 1
+
+    log(f"sender_open_id={sender_open_id}")
+    log(f"sku_list={parsed['sku_list']}")
+    log(f"target_date={parsed['target_date']}")
+
+    if args.dry_run:
+        print(json.dumps(parsed, ensure_ascii=False, indent=2))
+        return 0
+
+    write_task_json(parsed["sku_list"], parsed["target_date"], parsed.get("remark", ""))
+    log(f"Task file written: {TASK_FILE}")
+
+    ok, output = run_workflow()
+    if ok:
+        print("[OK] Workflow triggered successfully")
+        if output:
+            print(output[-3000:])
+        return 0
+
+    print("[FAIL] Workflow trigger failed")
+    if output:
+        print(output[-3000:])
+    return 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/001_init.sql b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/001_init.sql
new file mode 100644
index 0000000..a7fe693
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/001_init.sql
@@ -0,0 +1,120 @@
+-- Auto-flow 数据持久化 - 初始化
+-- 适用 PostgreSQL 13+
+-- 执行示例: psql -h 8.148.185.119 -U data_hub -d data_hub -f 001_init.sql
+--
+-- 设计原则:
+--   1) workflow_runs 是核心执行流水，所有 fact_* 表通过 run_id 外键关联
+--   2) dim_product 跨工作流共享：所有采集到的 SKU 都 upsert
+--   3) 所有 INSERT 由 orchestrator/pg_writer.py 触发,ON CONFLICT 幂等
+--   4) 永久保留,不做清理任务
+
+BEGIN;
+
+-- 1) 全局执行流水
+CREATE TABLE IF NOT EXISTS workflow_runs (
+    run_id           TEXT PRIMARY KEY,
+    workflow         TEXT NOT NULL,
+    trigger_time     TIMESTAMPTZ NOT NULL,
+    end_time         TIMESTAMPTZ,
+    status           TEXT NOT NULL,
+    error            TEXT NOT NULL DEFAULT '',
+    source_files     JSONB,
+    raw_payload      JSONB,
+    created_at       TIMESTAMPTZ NOT NULL DEFAULT now()
+);
+CREATE INDEX IF NOT EXISTS idx_workflow_runs_workflow_time
+    ON workflow_runs (workflow, trigger_time DESC);
+
+-- 2) 商品维度表
+CREATE TABLE IF NOT EXISTS dim_product (
+    sku              TEXT PRIMARY KEY,
+    is_discontinued  BOOLEAN NOT NULL DEFAULT false,
+    first_seen_at    TIMESTAMPTZ NOT NULL DEFAULT now(),
+    last_seen_at     TIMESTAMPTZ NOT NULL DEFAULT now()
+);
+CREATE INDEX IF NOT EXISTS idx_dim_product_discontinued
+    ON dim_product (is_discontinued) WHERE is_discontinued;
+
+-- 3) 采购确认事实表
+CREATE TABLE IF NOT EXISTS fact_purchase_confirmation (
+    id                  BIGSERIAL PRIMARY KEY,
+    run_id              TEXT NOT NULL REFERENCES workflow_runs(run_id) ON DELETE CASCADE,
+    sku                 TEXT,
+    product_name        TEXT,
+    spec                TEXT,
+    available_stock     INT,
+    sales_7d            INT,
+    purchase_in_transit INT,
+    gap                 INT,
+    delivery_date       TEXT,
+    summary_raw         TEXT
+);
+CREATE UNIQUE INDEX IF NOT EXISTS uq_fpc_run_sku ON fact_purchase_confirmation (run_id, sku);
+CREATE INDEX IF NOT EXISTS idx_fpc_run ON fact_purchase_confirmation (run_id);
+CREATE INDEX IF NOT EXISTS idx_fpc_sku ON fact_purchase_confirmation (sku);
+
+-- 4) 库存阈值预警事实表
+CREATE TABLE IF NOT EXISTS fact_threshold_alert (
+    id              BIGSERIAL PRIMARY KEY,
+    run_id          TEXT NOT NULL REFERENCES workflow_runs(run_id) ON DELETE CASCADE,
+    sku             TEXT NOT NULL,
+    product_name    TEXT,
+    product_spec    TEXT,
+    threshold       NUMERIC,
+    actual_stock    NUMERIC,
+    purchase_all    NUMERIC,
+    sales_7         NUMERIC,
+    sales_15        NUMERIC,
+    sales_30        NUMERIC,
+    generated_at    TIMESTAMPTZ
+);
+CREATE UNIQUE INDEX IF NOT EXISTS uq_fta_run_sku ON fact_threshold_alert (run_id, sku);
+CREATE INDEX IF NOT EXISTS idx_fta_run ON fact_threshold_alert (run_id);
+CREATE INDEX IF NOT EXISTS idx_fta_sku ON fact_threshold_alert (sku);
+
+-- 5) 补货建议事实表(只存真正写多维表的那批,skipped 不入库)
+CREATE TABLE IF NOT EXISTS fact_replenishment (
+    id                 BIGSERIAL PRIMARY KEY,
+    run_id             TEXT NOT NULL REFERENCES workflow_runs(run_id) ON DELETE CASCADE,
+    sku                TEXT NOT NULL,
+    product_name       TEXT,
+    display_name       TEXT,
+    spec_new           TEXT,
+    spec_old           TEXT,
+    sales_7d           INT,
+    sales_15d          INT,
+    sales_30d          INT,
+    sales_60d          INT,
+    sales_90d          INT,
+    monthly_sales      INT,
+    available_stock    INT,
+    purchase_in_transit INT,
+    suggest_qty        INT,
+    bot_remark         TEXT,
+    write_status       TEXT,
+    created_at         TIMESTAMPTZ NOT NULL DEFAULT now()
+);
+CREATE UNIQUE INDEX IF NOT EXISTS uq_fr_run_sku ON fact_replenishment (run_id, sku);
+CREATE INDEX IF NOT EXISTS idx_fr_run ON fact_replenishment (run_id);
+CREATE INDEX IF NOT EXISTS idx_fr_sku ON fact_replenishment (sku);
+CREATE INDEX IF NOT EXISTS idx_fr_created ON fact_replenishment (created_at DESC);
+
+-- 6) 采购单日期更新事实表
+CREATE TABLE IF NOT EXISTS fact_purchase_order_update (
+    id              BIGSERIAL PRIMARY KEY,
+    run_id          TEXT NOT NULL REFERENCES workflow_runs(run_id) ON DELETE CASCADE,
+    target_sku      TEXT NOT NULL,
+    target_date     DATE,
+    remark          TEXT,
+    success         BOOLEAN NOT NULL,
+    updated_sku     TEXT,
+    product_name    TEXT,
+    product_spec    TEXT,
+    error_message   TEXT,
+    completed_at    TIMESTAMPTZ
+);
+CREATE UNIQUE INDEX IF NOT EXISTS uq_fpou_run_sku ON fact_purchase_order_update (run_id, target_sku);
+CREATE INDEX IF NOT EXISTS idx_fpou_run ON fact_purchase_order_update (run_id);
+CREATE INDEX IF NOT EXISTS idx_fpou_sku ON fact_purchase_order_update (target_sku);
+
+COMMIT;
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/002_workflow_v2.sql b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/002_workflow_v2.sql
new file mode 100644
index 0000000..7740333
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/002_workflow_v2.sql
@@ -0,0 +1,92 @@
+-- Auto-flow 数据持久化 v2: 多维表插入版本化 + 通知消息跟踪
+-- 适用 PostgreSQL 13+
+-- 执行示例: psql -h 8.148.185.119 -U data_hub -d data_hub -f 002_workflow_v2.sql
+--
+-- 设计动机:
+--   2026-06-18 replenishment 跑出 228 条重复插入(114×2),根因是 LLM analyzer 在
+--   lark-cli +record-batch-create 成功后"为了验证返回值"又调一次。旧版
+--   _bitable_written.json 文件闸门不够稳(LLM 可以忘掉读),改成 DB UNIQUE 约束
+--   物理兜底。
+--
+-- 设计原则:
+--   1) bitable_insert_log 是多维表写入的"事实表",workflow+run_id+version 联合唯一
+--      → 第二次 INSERT 同 version 必然 ON CONFLICT,LLM 再怎么调 lark-cli 也插不进 DB
+--   2) version 由 orchestrator 在 prompt 里指定(策略 B),LLM 不自决;
+--      跑完失败后 orchestrator 起新轮次时把 version+1 重新下发
+--   3) fact_*.* 多出来的 bitable_version / bitable_record_ids 是写入回填,让一次 run
+--      的明细与多维表写入一一对应可追溯
+--   4) notification_message_ids / notification_sent_at 是飞书消息回执,便于失败排查
+--      (发了几条/哪些用户/什么时候发)
+
+BEGIN;
+
+-- 1) 多维表插入日志(版本化幂等的核心表)
+CREATE TABLE IF NOT EXISTS bitable_insert_log (
+    id                  BIGSERIAL PRIMARY KEY,
+    workflow            TEXT NOT NULL,
+    run_id              TEXT NOT NULL REFERENCES workflow_runs(run_id) ON DELETE CASCADE,
+    version             INT  NOT NULL DEFAULT 1,
+    data_hash           TEXT NOT NULL,
+    table_token         TEXT NOT NULL,
+    table_id            TEXT NOT NULL,
+    record_count        INT  NOT NULL DEFAULT 0,
+    status              TEXT NOT NULL DEFAULT 'pending',  -- pending | success | failed
+    bitable_record_ids  JSONB,
+    error               TEXT,
+    acquired_at         TIMESTAMPTZ NOT NULL DEFAULT now(),
+    completed_at        TIMESTAMPTZ,
+    CONSTRAINT uq_bitable_insert UNIQUE (workflow, run_id, version)
+);
+ALTER TABLE bitable_insert_log
+    ADD COLUMN IF NOT EXISTS notification_message_ids JSONB,
+    ADD COLUMN IF NOT EXISTS notification_sent_at     TIMESTAMPTZ;
+CREATE INDEX IF NOT EXISTS idx_bitable_insert_run
+    ON bitable_insert_log (run_id);
+CREATE INDEX IF NOT EXISTS idx_bitable_insert_workflow_status
+    ON bitable_insert_log (workflow, status);
+
+-- 2) fact_replenishment: 多维表写入回填字段
+ALTER TABLE fact_replenishment
+    ADD COLUMN IF NOT EXISTS data_hash             TEXT,
+    ADD COLUMN IF NOT EXISTS bitable_version       INT,
+    ADD COLUMN IF NOT EXISTS bitable_record_ids    JSONB,
+    ADD COLUMN IF NOT EXISTS notification_message_ids JSONB,
+    ADD COLUMN IF NOT EXISTS notification_sent_at  TIMESTAMPTZ;
+
+CREATE INDEX IF NOT EXISTS idx_fr_bitable_version
+    ON fact_replenishment (run_id, bitable_version);
+
+-- 3) fact_purchase_confirmation
+ALTER TABLE fact_purchase_confirmation
+    ADD COLUMN IF NOT EXISTS data_hash             TEXT,
+    ADD COLUMN IF NOT EXISTS bitable_version       INT,
+    ADD COLUMN IF NOT EXISTS bitable_record_ids    JSONB,
+    ADD COLUMN IF NOT EXISTS notification_message_ids JSONB,
+    ADD COLUMN IF NOT EXISTS notification_sent_at  TIMESTAMPTZ;
+
+CREATE INDEX IF NOT EXISTS idx_fpc_bitable_version
+    ON fact_purchase_confirmation (run_id, bitable_version);
+
+-- 4) fact_threshold_alert
+ALTER TABLE fact_threshold_alert
+    ADD COLUMN IF NOT EXISTS data_hash             TEXT,
+    ADD COLUMN IF NOT EXISTS bitable_version       INT,
+    ADD COLUMN IF NOT EXISTS bitable_record_ids    JSONB,
+    ADD COLUMN IF NOT EXISTS notification_message_ids JSONB,
+    ADD COLUMN IF NOT EXISTS notification_sent_at  TIMESTAMPTZ;
+
+CREATE INDEX IF NOT EXISTS idx_fta_bitable_version
+    ON fact_threshold_alert (run_id, bitable_version);
+
+-- 5) fact_purchase_order_update
+ALTER TABLE fact_purchase_order_update
+    ADD COLUMN IF NOT EXISTS data_hash             TEXT,
+    ADD COLUMN IF NOT EXISTS bitable_version       INT,
+    ADD COLUMN IF NOT EXISTS bitable_record_ids    JSONB,
+    ADD COLUMN IF NOT EXISTS notification_message_ids JSONB,
+    ADD COLUMN IF NOT EXISTS notification_sent_at  TIMESTAMPTZ;
+
+CREATE INDEX IF NOT EXISTS idx_fpou_bitable_version
+    ON fact_purchase_order_update (run_id, bitable_version);
+
+COMMIT;
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/backfill_history.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/backfill_history.py
new file mode 100644
index 0000000..c882794
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/sql/backfill_history.py
@@ -0,0 +1,157 @@
+# -*- coding: utf-8 -*-
+"""
+回填历史 JSON/CSV 到 pg,用于:
+  - 验证 pg_writer 端到端落库正确(不触发真实采集)
+  - 给已有 4 类历史数据(2026-06-09 至今)建立查询基线
+
+用法:
+  python -m gyxx_flow.modules.supply_chain.runtime.orchestrator.sql.backfill_history
+
+幂等: 多次运行不会产生重复行(run_id 是唯一键)
+"""
+
+from __future__ import annotations
+
+import json
+import sys
+
+from gyxx_flow.modules.supply_chain.runtime.paths import SHARED_ROOT
+
+ROOT = SHARED_ROOT
+
+WORKFLOW_DIRS = {
+    "replenishment": ROOT / "replenishment",
+    "replenishment-alert": ROOT / "replenishment-alert",
+}
+
+# 这俩有 CSV 备份但实际是空目录,batch_process 写入的 JSON 在 shared-data
+# 路径脚本这次跳过 purchase-confirmation —— 历史 CSV 没保留(collect_confirmation.ps1
+# 第 168 行的 foreach ($file in $csvFiles) 引用了未初始化的 $csvFiles,备份逻辑本就没生效)
+
+
+def _ts_from_filename(name: str) -> str:
+    """从 pending_insert_20260617_163447.json 提取 '2026-06-17 16:34:47'。"""
+    stem = name.rsplit(".", 1)[0]
+    parts = stem.split("_")
+    if len(parts) >= 3 and len(parts[-2]) == 8 and len(parts[-1]) == 6:
+        d, t = parts[-2], parts[-1]
+        return f"{d[:4]}-{d[4:6]}-{d[6:8]} {t[:2]}:{t[2:4]}:{t[4:6]}"
+    return "1970-01-01 00:00:00"
+
+
+def backfill_replenishment() -> int:
+    """回填 pending_insert_*.json + notify_data_*.json 一对的运行。"""
+    d = WORKFLOW_DIRS["replenishment"]
+    if not d.exists():
+        print(f"[SKIP] {d} 不存在")
+        return 0
+
+    pending_files = sorted(d.glob("pending_insert_*.json"))
+    print(f"[INFO] {d.name}: 找到 {len(pending_files)} 个 pending_insert.json")
+
+    from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer
+
+    n_total = 0
+    for pf in pending_files:
+        stem = pf.stem  # pending_insert_20260617_163447
+        ts = stem.rsplit("_", 2)[-2] + "_" + stem.rsplit("_", 2)[-1]
+        run_id = f"replenishment_{ts}_backfill"
+
+        with open(pf, encoding="utf-8") as f:
+            data = json.load(f)
+        records = data.get("records") or []
+        nf = d / f"notify_data_{ts}.json"
+        notify = {}
+        if nf.exists():
+            with open(nf, encoding="utf-8") as f:
+                notify = json.load(f)
+
+        if not records:
+            print(f"  [SKIP] {pf.name}: records=0")
+            continue
+
+        trigger_time = notify.get("generated_at") or _ts_from_filename(pf.name)
+        pg_writer.upsert_run(
+            run_id,
+            "replenishment",
+            trigger_time=trigger_time,
+            end_time=trigger_time,
+            status="success" if data.get("pending_count", 0) > 0 else "completed",
+            source_files=[pf.name, nf.name] if nf.exists() else [pf.name],
+            raw_payload={"skipped_count": data.get("skipped_count"), "alert_count": notify.get("alert_count")},
+        )
+        n = pg_writer.insert_replenishments(run_id, records)
+        n_dim = pg_writer.upsert_dim_products(
+            [r.get("商品编号") for r in records if r.get("商品编号")],
+            is_discontinued=False,
+        )
+        print(f"  [OK] {pf.name}: fact={n} dim={n_dim} pending_count={data.get('pending_count')}")
+        n_total += n
+
+    return n_total
+
+
+def backfill_threshold_alert() -> int:
+    """回填 threshold_alert_data_*.json。"""
+    d = WORKFLOW_DIRS["replenishment-alert"]
+    if not d.exists():
+        print(f"[SKIP] {d} 不存在")
+        return 0
+
+    files = sorted(d.glob("threshold_alert_data_*.json"))
+    print(f"[INFO] {d.name}: 找到 {len(files)} 个 threshold_alert_data.json")
+
+    from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer
+
+    n_total = 0
+    for f in files:
+        stem = f.stem
+        ts = stem.rsplit("_", 3)[-2] + "_" + stem.rsplit("_", 3)[-1]
+        run_id = f"replenishment-alert_{ts}_backfill"
+
+        with open(f, encoding="utf-8") as f2:
+            data = json.load(f2)
+        alerts = data.get("alerts") or []
+        trigger_time = data.get("generated_at") or _ts_from_filename(f.name)
+
+        pg_writer.upsert_run(
+            run_id,
+            "replenishment-alert",
+            trigger_time=trigger_time,
+            end_time=trigger_time,
+            status="success" if data.get("alert_count", 0) > 0 else "completed",
+            source_files=[f.name],
+            raw_payload={"alert_count": data.get("alert_count")},
+        )
+        n = pg_writer.insert_threshold_alerts(run_id, alerts)
+        n_dim = pg_writer.upsert_dim_products(
+            [a.get("sku") for a in alerts if a.get("sku")],
+            is_discontinued=False,
+        )
+        print(f"  [OK] {f.name}: fact={n} dim={n_dim} alert_count={data.get('alert_count')}")
+        n_total += n
+
+    return n_total
+
+
+def main() -> int:
+    print(f"=== Backfill from {ROOT} ===")
+    if not ROOT.exists():
+        print(f"[ERROR] {ROOT} 不存在")
+        return 1
+
+    from gyxx_flow.modules.supply_chain.runtime.orchestrator import pg_writer
+    if not pg_writer.is_available():
+        print("[ERROR] pg 不可达,中止")
+        return 2
+
+    n_repl = backfill_replenishment()
+    n_thr = backfill_threshold_alert()
+
+    print()
+    print(f"=== 回填完成: fact_replenishment += {n_repl}, fact_threshold_alert += {n_thr} ===")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/state.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/state.py
new file mode 100644
index 0000000..40a9f08
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/state.py
@@ -0,0 +1,74 @@
+# -*- coding: utf-8 -*-
+"""State definitions and helpers for workflow execution."""
+
+from __future__ import annotations
+
+import operator
+from datetime import datetime
+from typing import Annotated, Literal, TypedDict
+
+
+class Message(TypedDict):
+    role: Literal["user", "assistant", "tool"]
+    content: str
+
+
+WorkflowStatus = Literal[
+    "pending",
+    "collecting",
+    "collect_retry",
+    "collected",
+    "analyze_retry",
+    "analyzed",
+    "completed",
+    "failed",
+]
+
+
+class WorkflowState(TypedDict):
+    """State shared across the workflow graph."""
+
+    workflow_name: str
+    workflow_id: str
+    status: WorkflowStatus
+    trigger_time: str
+    error: str
+    end_time: str | None
+
+    messages: Annotated[list[Message], operator.add]
+
+    collected_files: list[str]
+    collect_attempts: int
+
+    analysis_results: dict
+    notify_results: dict
+    analyze_attempts: int
+
+def now_iso() -> str:
+    """Return a second-precision ISO timestamp."""
+
+    return datetime.now().isoformat(timespec="seconds")
+
+
+def create_initial_state(
+    workflow_name: str,
+    workflow_id: str,
+    *,
+    trigger_time: str | None = None,
+) -> WorkflowState:
+    """Build the initial workflow state used by graph executors."""
+
+    return {
+        "workflow_name": workflow_name,
+        "workflow_id": workflow_id,
+        "status": "pending",
+        "trigger_time": trigger_time or now_iso(),
+        "error": "",
+        "end_time": None,
+        "messages": [],
+        "collected_files": [],
+        "collect_attempts": 0,
+        "analysis_results": {},
+        "notify_results": {},
+        "analyze_attempts": 0,
+    }
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/workflows.py b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/workflows.py
new file mode 100644
index 0000000..4594f16
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/orchestrator/workflows.py
@@ -0,0 +1,51 @@
+# -*- coding: utf-8 -*-
+"""Shared workflow metadata helpers.
+
+Keeps workflow lookup logic centralized so graph/mcp/runner stay consistent.
+"""
+
+from __future__ import annotations
+
+from gyxx_flow.modules.supply_chain.runtime.orchestrator.config import (
+    FEISHU_CONFIG,
+    WORKFLOWS,
+)
+
+
+def get_workflow_config(workflow_name: str) -> dict:
+    """Return workflow config or raise ValueError for unknown workflow."""
+
+    workflow = WORKFLOWS.get(workflow_name)
+    if workflow is None:
+        raise ValueError(f"Unknown workflow: {workflow_name}")
+    return workflow
+
+
+def workflow_display_name(workflow_name: str) -> str:
+    """Return stable workflow display name: '<id><name>'."""
+
+    workflow = get_workflow_config(workflow_name)
+    return f"{workflow_name}{workflow['name']}"
+
+
+def workflow_requires_analysis(workflow_name: str) -> bool:
+    """Whether this workflow includes analyzer scripts."""
+
+    workflow = get_workflow_config(workflow_name)
+    return bool(workflow.get("analyzer_scripts"))
+
+
+def batch_workflow_names() -> list[str]:
+    """Business workflows that should run in batch mode."""
+
+    return [name for name in WORKFLOWS if workflow_requires_analysis(name)]
+
+
+def feishu_app(role: str) -> dict:
+    """Return Feishu app config for role in {'collector', 'analyzer'}."""
+
+    apps = FEISHU_CONFIG.get("apps", {})
+    config = apps.get(role)
+    if not config:
+        raise ValueError(f"Missing FEISHU_CONFIG.apps.{role}")
+    return config
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/paths.py b/src/gyxx_flow/modules/supply_chain/runtime/paths.py
new file mode 100644
index 0000000..ed2c505
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/paths.py
@@ -0,0 +1,29 @@
+"""Portable filesystem layout for the migrated supply-chain runtime."""
+
+from __future__ import annotations
+
+import os
+from pathlib import Path
+
+from gyxx_flow.core import DataLayout, Settings
+
+PACKAGE_ROOT = Path(__file__).resolve().parent
+PROJECT_ROOT = PACKAGE_ROOT.parents[4]
+_configured_data_root = os.environ.get("GYXX_DATA_ROOT", "").strip()
+DATA_ROOT = (
+    Path(_configured_data_root).expanduser().resolve()
+    if _configured_data_root
+    else Settings.from_env().data_root
+)
+_DATA_PATHS = DataLayout(DATA_ROOT).for_module("supply_chain")
+SHARED_ROOT = _DATA_PATHS.raw_root
+NORMALIZED_ROOT = _DATA_PATHS.normalized_root
+CURATED_ROOT = _DATA_PATHS.curated_root
+BACKUP_ROOT = _DATA_PATHS.exports_root
+EVIDENCE_ROOT = _DATA_PATHS.evidence_root
+STATE_ROOT = _DATA_PATHS.state_root
+LOG_ROOT = _DATA_PATHS.logs_root
+WORK_ROOT = _DATA_PATHS.tmp_root
+CONFIG_PATH = Path(
+    os.getenv("GYXX_SUPPLY_CONFIG", str(STATE_ROOT / "config.json"))
+).expanduser().resolve()
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/run.py b/src/gyxx_flow/modules/supply_chain/runtime/run.py
new file mode 100644
index 0000000..2d849ab
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/run.py
@@ -0,0 +1,81 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""Quick start script for auto-flow orchestrator (MCP only)."""
+
+import json
+import sys
+
+from gyxx_flow.modules.supply_chain.runtime.orchestrator.config import WORKFLOWS
+
+WORKFLOW_CHOICES = sorted(WORKFLOWS.keys()) + ["all"]
+
+if hasattr(sys.stdout, "reconfigure"):
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")
+if hasattr(sys.stderr, "reconfigure"):
+    sys.stderr.reconfigure(encoding="utf-8", errors="replace")
+
+
+def _print_help() -> None:
+    print("Auto-Flow Orchestrator")
+    print("=" * 50)
+    print("\nUsage:")
+    print("  python run.py <command> [options]")
+    print("\nCommands:")
+    print("  mcp-run <workflow>   - Run via Hermes MCP gateways")
+    print("  monitor [interval]   - Monitor for new data")
+    print("\nWorkflows:")
+    for workflow in WORKFLOW_CHOICES:
+        print(f"  {workflow}")
+    print("\nExamples:")
+    print("  python run.py mcp-run purchase-confirmation")
+    print("  python run.py mcp-run all")
+
+
+def main() -> int:
+    if len(sys.argv) < 2:
+        _print_help()
+        return 0
+
+    command = sys.argv[1]
+
+    if command == "monitor":
+        from gyxx_flow.modules.supply_chain.runtime.orchestrator.monitor import (
+            monitor_loop,
+        )
+
+        interval = int(sys.argv[2]) if len(sys.argv) > 2 else 60
+        monitor_loop(interval)
+        return 0
+
+    if command != "mcp-run":
+        print(f"Error: local workflow command '{command}' is removed. Use 'mcp-run'.")
+        _print_help()
+        return 1
+
+    if len(sys.argv) < 3:
+        print("Error: mcp-run requires a workflow argument")
+        return 1
+
+    workflow = sys.argv[2]
+    if workflow not in WORKFLOW_CHOICES:
+        print(f"Error: unknown workflow '{workflow}'")
+        print(f"Available workflows: {', '.join(WORKFLOW_CHOICES)}")
+        return 1
+
+    try:
+        from gyxx_flow.modules.supply_chain.runtime.orchestrator.runner import run_mcp
+
+        print(f"Running MCP workflow for: {workflow}")
+        print("-" * 50)
+        result = run_mcp(workflow)
+        print("\n" + "=" * 50)
+        print("Result:")
+        print(json.dumps(result, ensure_ascii=False, indent=2, default=str))
+        return 0
+    except Exception as exc:
+        print(f"\nError: {exc}")
+        return 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/scripts/purchase-confirmation.bat b/src/gyxx_flow/modules/supply_chain/runtime/scripts/purchase-confirmation.bat
new file mode 100644
index 0000000..b917286
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/scripts/purchase-confirmation.bat
@@ -0,0 +1,6 @@
+@echo off
+setlocal
+if defined GYXX_PROJECT_ROOT (set "PROJECT_ROOT=%GYXX_PROJECT_ROOT%") else (set "PROJECT_ROOT=%~dp0\..\..\..\..\..\..\..")
+for %%I in ("%PROJECT_ROOT%") do set "PROJECT_ROOT=%%~fI"
+"%PROJECT_ROOT%\.venv\Scripts\python.exe" -m gyxx_flow.cli run supply.purchase_confirmation.daily
+endlocal & exit /b %ERRORLEVEL%
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/scripts/replenishment-alert.bat b/src/gyxx_flow/modules/supply_chain/runtime/scripts/replenishment-alert.bat
new file mode 100644
index 0000000..8b290f2
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/scripts/replenishment-alert.bat
@@ -0,0 +1,6 @@
+@echo off
+setlocal
+if defined GYXX_PROJECT_ROOT (set "PROJECT_ROOT=%GYXX_PROJECT_ROOT%") else (set "PROJECT_ROOT=%~dp0\..\..\..\..\..\..\..")
+for %%I in ("%PROJECT_ROOT%") do set "PROJECT_ROOT=%%~fI"
+"%PROJECT_ROOT%\.venv\Scripts\python.exe" -m gyxx_flow.cli run supply.replenishment_alert.daily
+endlocal & exit /b %ERRORLEVEL%
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/scripts/replenishment.bat b/src/gyxx_flow/modules/supply_chain/runtime/scripts/replenishment.bat
new file mode 100644
index 0000000..f10887c
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/scripts/replenishment.bat
@@ -0,0 +1,6 @@
+@echo off
+setlocal
+if defined GYXX_PROJECT_ROOT (set "PROJECT_ROOT=%GYXX_PROJECT_ROOT%") else (set "PROJECT_ROOT=%~dp0\..\..\..\..\..\..\..")
+for %%I in ("%PROJECT_ROOT%") do set "PROJECT_ROOT=%%~fI"
+"%PROJECT_ROOT%\.venv\Scripts\python.exe" -m gyxx_flow.cli run supply.replenishment.weekly
+endlocal & exit /b %ERRORLEVEL%
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/sitecustomize.py b/src/gyxx_flow/modules/supply_chain/runtime/sitecustomize.py
new file mode 100644
index 0000000..bf69a40
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/sitecustomize.py
@@ -0,0 +1,5 @@
+"""Apply the unified per-script runtime binding to nested Python launches."""
+
+from gyxx_flow.adapters.bootstrap import safe_bootstrap_current_process
+
+safe_bootstrap_current_process()
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/skills/analyzer/SKILL.md b/src/gyxx_flow/modules/supply_chain/runtime/skills/analyzer/SKILL.md
new file mode 100644
index 0000000..17c5254
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/skills/analyzer/SKILL.md
@@ -0,0 +1,190 @@
+---
+name: data-analyzer
+description: 分析端。接收用户消息触发采购单更新，发送采购单更新通知，处理编排器发来的工作流分析指令。
+metadata:
+  hermes:
+    skillKey: data-analyzer
+    profile: analyzer
+    os: ["win32"]
+    requires:
+      bins: ["python"]
+---
+
+# Data Analyzer
+
+## Mission
+你是 `data-analyzer`，职责：
+1. 读取 `shared-data` 产物做分析/汇总。
+2. 发送最终业务通知（给业务对象和 owner）。
+3. 接收用户飞书消息，触发采购单日期更新工作流。
+
+## 采购单日期更新触发
+
+当用户发来包含 **8位SKU编码** 和 **日期** 的消息时，触发采购单更新工作流。
+
+### 触发条件
+消息中同时包含：
+- 至少一个 8 位数字 SKU（如 `10439030`）
+- 一个日期（如 `5月25日`、`2026-05-25`、`明天`）
+
+### 触发方式
+使用终端工具执行（**必须传递 --sender-open-id**，脚本内部校验白名单）：
+```bash
+python -m gyxx_flow.modules.supply_chain.runtime.orchestrator.scripts.trigger_purchase_order_update --message "<用户消息原文>" --sender-open-id "<发送人open_id>"
+```
+
+允许触发的 sender_open_id 在 `orchestrator/config.py` → `FEISHU_CONFIG.trigger.purchase_order_update.allowed_sender_open_ids` 中配置。当前允许：`<ANALYZER_OWNER_OPEN_ID>`。
+
+### 执行后
+- 工作流触发后，采集端执行 ERP 操作，结果写入 `shared-data/purchase-order-update/update_result.json`
+- **通知发送需要分析端手动执行**（见下节），工作流本身不自动发送飞书消息
+- 如果触发成功，回复用户"采购单更新已触发，SKU: xxx，目标日期: xxx"
+- 如果触发失败（权限不足、解析失败等），回复用户具体原因
+- 如果消息不包含 SKU 或日期，按普通对话处理，不触发工作流
+
+## 采购单更新后发送通知
+
+工作流执行完成后，分析端需要手动发送通知到两个目标：
+1. **审单群**（chat_id: `oc_6e95333db779b07524e1c361099c7aec`）
+2. **owner 个人**（open_id: `ou_7ad5fc8012e2f741afc5346e05ffd447`）
+
+### 前置条件
+- 分析端飞书机器人（`cli_aa8c4fc918b85cce`）必须已是审单群成员，才能发送到群
+- 如果机器人不在群里，会返回 `code=230002: Bot/User can NOT be out of the chat`
+
+### 发送方式（直接调用飞书 Open API）
+
+```python
+import requests, json
+from datetime import datetime
+
+APP_ID = "cli_aa8c4fc918b85cce"
+APP_SECRET = "${GYXX_SUPPLY_ANALYZER_APP_SECRET}"
+FEISHU_OPENAPI_BASE = "https://open.feishu.cn/open-apis"
+SHENDAN_GROUP_CHAT_ID = "oc_6e95333db779b07524e1c361099c7aec"
+MY_OPEN_ID = "ou_7ad5fc8012e2f741afc5346e05ffd447"
+
+# 获取 token
+token_resp = requests.post(
+    f"{FEISHU_OPENAPI_BASE}/auth/v3/tenant_access_token/internal/",
+    json=dict((("app_id", APP_ID), ("app_secret", APP_SECRET))), timeout=10
+)
+token = token_resp.json()["tenant_access_token"]
+
+# 构建交互式卡片 (兼容写法:top-level elements + header,避开 body 嵌套翻车)
+# records 每条形如 {"sku": "...", "name": "...", "spec": "..."}
+header = {
+    "template": "blue",
+    "title": {"tag": "plain_text", "content": "📌 采购单更新通知"},
+}
+elements = [
+    {
+        "tag": "div",
+        "fields": [
+            {"is_short": True, "text": {"tag": "lark_md", "content": f"**更新时间**\n{datetime.now().strftime('%Y-%m-%d %H:%M')}"}},
+            {"is_short": True, "text": {"tag": "lark_md", "content": f"**到货日期**\n{target_date}"}},
+        ],
+    },
+    {"tag": "div", "text": {"tag": "lark_md", "content": f"✅ **成功更新 {total_updated} 条采购单记录**"}},
+    {"tag": "hr"},
+]
+for r in records:
+    elements.append({
+        "tag": "div",
+        "text": {"tag": "lark_md", "content": f"- **SKU:** {r['sku']}　**品名:** {r['name']}　**规格:** {r['spec']}"},
+    })
+card = {
+    "config": {"wide_screen_mode": True},
+    "header": header,
+    # ⚠️ 2026-06-29:不要把 elements 嵌套在 body 下!Feishu 存盘会 strip 掉 body,
+    # 用户收到空卡。用 top-level elements(详见上文避坑)。
+    "elements": elements,
+}
+
+# 发送到审单群
+requests.post(
+    f"{FEISHU_OPENAPI_BASE}/im/v1/messages",
+    headers={"Authorization": f"Bearer {token}", "Content-Type": "application/json"},
+    params={"receive_id_type": "chat_id"},
+    json={"receive_id": SHENDAN_GROUP_CHAT_ID, "msg_type": "interactive",
+          "content": json.dumps(card, ensure_ascii=False)},
+    timeout=10
+)
+
+# 发送到个人
+requests.post(
+    f"{FEISHU_OPENAPI_BASE}/im/v1/messages",
+    headers={"Authorization": f"Bearer {token}", "Content-Type": "application/json"},
+    params={"receive_id_type": "open_id"},
+    json={"receive_id": MY_OPEN_ID, "msg_type": "interactive",
+          "content": json.dumps(card, ensure_ascii=False)},
+    timeout=10
+)
+```
+
+### 通知内容要求（来自工作流规则）
+必须包含：SKU列表、商品名称、规格颜色、到货日期、更新记录数。
+
+### 如果机器人不在审单群
+1. 先发送通知到个人账户告知结果
+2. 告知用户需要将分析端机器人加入审单群
+
+### 备选：lark-cli 发送（不推荐主用）
+
+如果不想手写 `requests.post` token 流程，可改用 `lark-cli`（已配 `hermes-analyzer` profile）：
+
+```bash
+# 1. 把 card 写到临时文件
+cat > /tmp/card.json << 'JSON'
+{
+  "config": {"wide_screen_mode": true},
+  "header": {"template": "blue", "title": {"tag": "plain_text", "content": "📌 采购单更新通知"}},
+  "elements": [...]
+}
+JSON
+
+# 2. 发给个人（analyzer owner，bot 身份在 hermes-analyzer profile 下）
+lark-cli --profile hermes-analyzer im +messages-send \
+  --as bot --user-id ou_7ad5fc8012e2f741afc5346e05ffd447 \
+  --msg-type interactive --content "$(cat /tmp/card.json)"
+
+# 3. 发给审单群（用 --chat-id + bot 身份）
+lark-cli --profile hermes-analyzer im +messages-send \
+  --as bot --chat-id oc_6e95333db779b07524e1c361099c7aec \
+  --msg-type interactive --content "$(cat /tmp/card.json)"
+```
+
+**关键点**：
+- 必须在命令前加 `--profile hermes-analyzer`（lark-cli 默认 `currentApp` 是采集端 `cli_aa8c4fb4c4f81cd3`，发给 analyzer owner 会 `code: 99992361 "open_id cross app"`）。
+- 卡片 JSON 用 **top-level `elements`**(不要再用 `body.elements` 嵌套,详见上文 Hard Rules 避坑段)。`header.template` + `header.title` 控制标题颜色和文字。
+- `--as bot` 不需要额外权限；`--as user` 需要 `im:message.send_as_user` scope，分析端应用未配置。
+
+## 工作流分析（由编排器调用）
+
+当收到编排器发来的工作流分析指令时：
+1. 读取 `shared-data` 下对应工作流的输出文件
+2. 分析数据并发送最终业务通知
+3. 不发送过程/进度通知
+
+## 参考资料
+- `references/purchase-order-update-config.md` — 飞书应用配置、审单群 chat_id、触发白名单、常见错误码
+
+## Hard Rules
+- 不运行采集脚本（`collect_*.ps1`）。
+- 不发送采集心跳、进度播报给业务用户。
+- 最终业务通知如果内容过长，必须拆分为多条消息，使用 `[1/N]...[N/N]` 前缀。
+- 禁止截断通知内容。
+- **禁止用本地 Python/PowerShell 脚本发送飞书业务通知**，必须通过 `execute_code` 直接调飞书 Open API。
+  - 例外:**Schema 2.0 卡片发送**统一走 `orchestrator/scripts/send_card_notification.py`(2026-06-29 起,LLM 直接调 lark-cli im +messages-send --file 翻车)。其它通知场景仍按上面铁律。
+- **业务通知必须使用 interactive card**（`msg_type="interactive"`, `content` 为卡片 JSON 字符串），header 配色按场景选 blue/green/red；不要发 plain text 或 post 富文本。
+- **【避坑 2026-06-29】卡片 elements 不要用 `body.elements` 嵌套!** 实测 Feishu `/im/v1/messages` POST 返回 200 OK,但存盘时把整个 `body` 字段丢掉,用户收到"只有标题的空卡片"。**正确写法:top-level elements + header(template+title)**,`send_card_notification.py` 已加 body→top-level 兜底,但 LLM 应该直接发对。
+  ```json
+  // ✓ 推荐 (兼容写法,header 控色,elements 在 top-level)
+  {"config": {"wide_screen_mode": true},
+   "header": {"template": "blue", "title": {"tag": "plain_text", "content": "📌 通知标题"}},
+   "elements": [{"tag": "div", "text": {"tag": "lark_md", "content": "**正文**"}}]}
+  // ✗ 会翻车 (Schema 2.0 嵌套,body 存盘被 strip)
+  {"header": {...}, "body": {"elements": [...]}}
+  ```
+- 用 lark-cli 发卡时必须加 `--profile hermes-analyzer`，否则 lark-cli 用采集端 bot 身份，发给分析端用户会 `code: 99992361 "open_id cross app"`。卡片 JSON 用 **top-level elements**(不要再用 `body.elements` 嵌套,见上面避坑)。
+- **【避坑 2026-06-29】不要在卡片里用 `tag: "table"` 元素**。`/im/v1/messages` 报 400 ErrCode 200906 「table columns is empty」(加 column_widths 也不解决);改用 cells/td 又触发 200621 「parse card json err」;改用 column name 映射 rows 能 send(200)但飞书客户端显示「请升级至最新版本」的占位图。**结论:表格用 `div + lark_md` 包 ```markdown|...|...|``` 代码块渲染,稳定可靠。**
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/skills/analyzer/workflow/purchase-order-update/SKILL.md b/src/gyxx_flow/modules/supply_chain/runtime/skills/analyzer/workflow/purchase-order-update/SKILL.md
new file mode 100644
index 0000000..6cc2e02
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/skills/analyzer/workflow/purchase-order-update/SKILL.md
@@ -0,0 +1,151 @@
+---
+name: purchase-order-update
+description: 采购单日期更新工作流 — 触发工作流更新ERP中SKU的协议到货日期，然后发送飞书通知到审单群和用户
+---
+
+# Purchase Order Update Workflow（采购单日期更新工作流）
+
+## 触发条件
+用户发送消息包含 **8位SKU** 和 **到货日期**，支持多种自然语言格式。
+
+### 支持的消息格式示例
+
+| 示例消息 | 说明 |
+|---------|------|
+| `10465007 更新到货日期 5月19日` | ✅ |
+| `10465007 到货日期改为 2026-05-19` | ✅ |
+| `采购单 10465007 5月19日到货` | ✅ |
+| `10465007 明天到货` | ✅（明天自动换算） |
+| `10465007，5月19日，采购到货` | ✅ |
+| `10439048 10439047 更新到货日期 5月19日` | ✅（多SKU） |
+| `10465007 更新 5月19日` | ✅ |
+
+### 日期格式支持
+- `YYYY-MM-DD` → `2026-05-19`
+- `YYYY/MM/DD` → `2026/05/19`
+- `5月19日` / `5月19号`
+- `今天` / `明天` / `后天` / `大后天`
+
+### 关键词忽略
+`采购`、`到货`、`日期`、`更新`、`协议` 等字词会被自动过滤，不影响解析。
+
+## 执行步骤
+
+### 1. 准备任务文件
+由于 `orchestrator.config` 模块当前不可用，直接写入 task.json：
+```bash
+mkdir -p ${GYXX_PROJECT_ROOT}/shared-data/purchase-order-update
+# 写入 task.json：
+{
+  "sku_list": ["<SKU1>", "<SKU2>"],
+  "target_date": "2026-05-19",
+  "remark": "",
+  "created_at": "<ISO时间>",
+  "source": "analyzer_message_trigger"
+}
+```
+
+### 2. 直接运行采购单更新脚本
+```bash
+cd ${GYXX_PROJECT_ROOT} && .venv/Scripts/python.exe orchestrator/scripts/PurchaseOrderUpdate.py --task-file ${GYXX_PROJECT_ROOT}/shared-data/purchase-order-update/task.json
+```
+- 超时：300秒
+- 成功标志：日志含 `[LOG] SCRIPT_COMPLETED`
+- 结果写入：`orchestrator/scripts/data/update_result.json`
+
+### 3. 发送飞书通知
+
+#### 3.1 获取 tenant_access_token
+```python
+import requests
+APP_ID = "cli_aa8c4fc918b85cce"
+APP_SECRET = "${GYXX_SUPPLY_ANALYZER_APP_SECRET}"
+token = requests.post(
+    "https://open.feishu.cn/open-apis/auth/v3/tenant_access_token/internal/",
+    json=dict((("app_id", APP_ID), ("app_secret", APP_SECRET))),
+    timeout=10
+).json()["tenant_access_token"]
+```
+
+#### 3.2 构建通知内容
+```python
+import json
+from datetime import datetime
+
+current_time = datetime.now().strftime("%Y-%m-%d %H:%M")
+
+# Schema 2.0 交互式卡片
+header = {
+    "template": "blue",
+    "title": {"tag": "plain_text", "content": "📌 采购单更新通知"},
+}
+elements = [
+    {
+        "tag": "div",
+        "fields": [
+            {"is_short": True, "text": {"tag": "lark_md", "content": f"**更新时间**\n{current_time}"}},
+            {"is_short": True, "text": {"tag": "lark_md", "content": f"**到货日期**\n{target_date}"}},
+        ],
+    },
+    {"tag": "div", "text": {"tag": "lark_md", "content": f"✅ **成功更新 {total_updated} 条采购单记录**"}},
+    {"tag": "hr"},
+]
+# 每条记录一行：- SKU：{sku}，品名：{name}，规格：{spec}
+for r in records:
+    elements.append({
+        "tag": "div",
+        "text": {"tag": "lark_md", "content": f"- **SKU:** {r['sku']}　**品名:** {r['name']}　**规格:** {r['spec']}"},
+    })
+
+card = {
+    "config": {"wide_screen_mode": True},
+    "header": header,
+    "body": {"elements": elements},  # Schema 2.0 规范：elements 嵌套在 body 下
+}
+```
+
+#### 3.3 发送消息
+```python
+SHENDAN_GROUP_CHAT_ID = "oc_6e95333db779b07524e1c361099c7aec"
+MY_OPEN_ID = "ou_7ad5fc8012e2f741afc5346e05ffd447"
+
+# 发送审单群
+requests.post(
+    "https://open.feishu.cn/open-apis/im/v1/messages",
+    headers={"Authorization": f"Bearer {token}", "Content-Type": "application/json"},
+    params={"receive_id_type": "chat_id"},
+    json={
+        "receive_id": SHENDAN_GROUP_CHAT_ID,
+        "msg_type": "interactive",
+        "content": json.dumps(card, ensure_ascii=False)
+    },
+    timeout=10
+)
+
+# 发送本人
+requests.post(
+    "https://open.feishu.cn/open-apis/im/v1/messages",
+    headers={"Authorization": f"Bearer {token}", "Content-Type": "application/json"},
+    params={"receive_id_type": "open_id"},
+    json={
+        "receive_id": MY_OPEN_ID,
+        "msg_type": "interactive",
+        "content": json.dumps(card, ensure_ascii=False)
+    },
+    timeout=10
+)
+```
+
+## 白名单配置
+`${GYXX_PROJECT_ROOT}\orchestrator\config.py` → `FEISHU_CONFIG.trigger.purchase_order_update.allowed_sender_open_ids`
+- `ou_7ad5fc8012e2f741afc5346e05ffd447`（用户）
+- `ou_339c1d396b397c97b08e1bf377963d40`
+- `ou_54141d75b24d22ddc8c1c4d66e00b6e9`
+
+## 注意事项
+- `orchestrator.config` 模块缺失（WORKFLOWS / FEISHU_CONFIG 未定义），trigger 脚本因此不可用；改用直接脚本执行可正常工作
+- 发送审单群时需确认分析端机器人已在群中
+- 工作流会自动更新 ERP 系统中对应 SKU 采购单的协议到货日期
+- Playwright 浏览器依赖若未安装（`Executable doesn't exist`），需先运行：`.venv\Scripts\python.exe -m playwright install chromium`
+- 卡片 JSON 用 `body.elements` 嵌套（Schema 2.0 规范）；top-level `elements` 也兼容但官方不推荐
+- 若改用 lark-cli 发送，命令前必须加 `--profile hermes-analyzer`（lark-cli 默认采集端 bot 身份），否则 `code: 99992361 "open_id cross app"`
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/SKILL.md b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/SKILL.md
new file mode 100644
index 0000000..e1ec105
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/SKILL.md
@@ -0,0 +1,88 @@
+---
+name: data-collector-guardrail
+description: "Guardrail for collector: collect + heartbeat only."
+metadata:
+  hermes:
+    skillKey: data-collector-guardrail
+    profile: data-collector
+    os: ["win32"]
+---
+
+Collector Guardrail
+
+You are the collector endpoint.
+
+Do:
+- Run collection scripts.
+- Publish fresh artifacts to shared-data.
+- (Heartbeat is now sent automatically by `orchestrator/mcp_workflow.py` as a Schema 2.0 interactive card to `ou_8ee224968aa26a74c7d30ba27fed5eeb`. Do NOT call send_message / curl / any Feishu API to relay heartbeats.)
+- For **manual** ad-hoc heartbeat relay (operator asks in chat to forward a workflow report to the admin), reply directly with the same fields described in `Collector Heartbeat Notification Format` below — the gateway binding delivers the reply.
+
+Do not:
+- Send final business notifications.
+- Run analyzer-stage actions.
+- Mix analyzer app credentials (analyzer app: cli_aa8c4fc918b85cce, analyzer owner: ou_7ad5fc8012e2f741afc5346e05ffd447).
+- Try to call send_message, curl, or external APIs to relay **automated** collector heartbeats — orchestrator already sends them as cards.
+
+Credential path: The collector app secret (`2lAvcKK4gX2Qa6uCrhxgZedTzbc70a7U`) lives in two places:
+- Hermes profile `.env`: `${HERMES_STATE_ROOT}\profiles\data-collector\.env` under `FEISHU_APP_SECRET` (for cron / LLM-direct paths).
+- `orchestrator/config.py` → `FEISHU_CONFIG.apps.collector.app_secret` (env override `AUTOFLOW_COLLECTOR_APP_SECRET`). Used by `mcp_workflow.py:_send_collector_heartbeat` for the new automated card-based heartbeats.
+
+It is NOT in `orchestrator/scripts/config.json` → `feishu.app_secret` / `feishu_analyzer.app_secret` (those are analyzer app credentials). Scripts like `send_collector_notify.py` that fall back to that file will fail with `code: 10014, app secret invalid`.
+
+**Automated heartbeat (post-2026-06-27):** The orchestrator sends Schema 2.0 interactive cards via `mcp_workflow.py:_send_collector_heartbeat` → `requests.post(/open-apis/im/v1/messages)` with the collector app credentials, target `ou_8ee224968aa26a74c7d30ba27fed5eeb`. Phases: `start` (blue), `progress` (blue, every `HEARTBEAT_INTERVAL_SECONDS`=180s), `final` (green on success / red on failure). The collector LLM is NOT involved — do not duplicate.
+
+**Manual relay (operator asks in chat to forward a report):** Reply directly with the format below. The gateway binding delivers it. No tool calls needed.
+
+Pitfalls:
+| Collector app owner open_id | `ou_8ee224968aa26a74c7d30ba27fed5eeb` — the admin/reporter who receives collector heartbeat notifications via collector Feishu app |
+| Notification target (admin/reporter) | `ou_8ee224968aa26a74c7d30ba27fed5eeb` — this profile receives workflow heartbeat reports (NOT the analyzer app owner) |
+| Analyzer app owner open_id | `ou_7ad5fc8012e2f741afc5346e05ffd447` — NOT the collector heartbeat target; this belongs to the analyzer app scope (`cli_aa8c4fc918b85cce`) |
+- UTF-8 BOM manifest error: If a workflow fails with `Invalid run manifest: Unexpected UTF-8 BOM (decode using utf-8-sig): line 1 column 1`, the manifest JSON file was saved with a BOM (common on Windows editors like Notepad). Fix: resave the manifest as UTF-8 without BOM, or strip the BOM (`\ufeff`) from the file's first bytes.
+- batch_process.py scripts must be saved as UTF-8 encoding. GBK/CP936 encoding corrupts Chinese string literals and causes "unterminated string literal" SyntaxError on line 41.
+- Yuanbao toolsets are listed in platform_toolsets.cli but the `yb` CLI binary and `yb_send_dm`/`yb_query_group_members` functions are NOT available on data-collector — `yb_send_dm: command not found` is the expected result. Do not search for the `yb` binary, attempt `hermes tools enable yuanbao` (it will succeed but expose nothing usable), or try to curl localhost:8123/yuanbao/send_dm. Yuanbao adapter is also absent from channel_directory.json (yuanbao: []). Always fall back to direct reply immediately — the API server delivers collector heartbeat reports without any Yuanbao integration.
+- Cron jobs running collector workflows (deliver: local): output IS the notification — do NOT try to call Feishu REST APIs from cron. The cron output goes to the gateway chat; the gateway relays to Feishu. If you need to send Feishu from cron, you need the app_secret accessible in the cron env (`FEISHU_APP_SECRET`) — collector app secrets are often gateway-runtime-injected and NOT stored in the codebase. Use `FEISHU_APP_SECRET` (not `COLLECTOR_FEISHU_APP_SECRET`).
+- **`execute_code` filesystem access is limited to the sandbox directory.** `execute_code` runs in an isolated sandbox — it cannot access host paths like `${GYXX_PROJECT_ROOT}\shared-data\` or `${HERMES_STATE_ROOT}\...`. Attempting to open host paths via `Path()` or `read_bytes()` will raise `FileNotFoundError`. Use `terminal` (which inherits the MSYS/Git-Bash environment with proper path mapping) for all file operations on host paths. The `execute_code` sandbox is only suitable for pure in-memory computation or accessing paths that the sandbox itself creates.
+- **`execute_code` can make outbound HTTP calls** — the "cannot call Feishu REST APIs" restriction is outdated. `requests` + `FEISHU_APP_SECRET` works fine from `execute_code` in the agent process. The failure scenario is cron job scripts running in a subprocess (where `FEISHU_APP_SECRET` is not injected). Always check `os.environ.get("FEISHU_APP_SECRET")` before assuming it works or fails.
+- **`terminal` + inline Python script** is the most reliable pattern for Feishu REST calls. `python -c "..."` via terminal inherits the shell env, works in both agent and cron contexts.
+- **Manual relay only (operator asks in chat to forward a report) — reply DIRECTLY with no tool calls.** Your reply text IS the notification. Do NOT query channel_directory, search for yuanbao group codes, call send_message, or look up credentials. This is for **manual / ad-hoc** forwarding only — automated heartbeats are sent by the orchestrator as cards.
+  - Common mistake: trying to relay an automated heartbeat (orchestrator already sent it) or calling send_message for manual relay. Both are unnecessary.
+- Cron jobs with **gateway Feishu binding** (deliver: origin/local): reply directly — the gateway's own Feishu binding delivers the reply. (For automated heartbeats the orchestrator sends the card itself, so this is now mostly relevant for operator-side messages, not workflow status.)
+- Scripts run via **terminal** without gateway binding: use `python -c "..."` to call Feishu REST APIs with `FEISHU_APP_SECRET`. (Same caveat — only relevant for paths outside the orchestrator's automated heartbeat flow.)
+**Cron job skills that don't exist cause a skip warning but don't fail the job.** When a cron job's `skills` list contains a nonexistent skill, the scheduler logs `WARNING Cron job 'X': skill not found, skipping — Skill 'feishu_doc' not found.` The job still runs with the remaining skills or no skills. If the cron job silently does nothing or hits an unexpected code path, check `logs/errors.log` for this warning — the missing skill is the likely cause. To fix: remove the nonexistent skill from the cron job's `skills` list, or create the skill.
+- Example: A `purchase-confirmation` workflow monitor cron job listed `feishu_doc` in skills but the skill was deleted. The job's `feishu_doc_read` tool call failed because the skill wasn't loaded, producing a partial/incomplete result with no error surfaced to the user.
+
+## Collector Heartbeat Notification Format
+
+**Automated heartbeats are sent by the orchestrator as Schema 2.0 cards** (see "Automated heartbeat" under Credential path). The LLM does not need to format or send them.
+
+**Manual relay only** (operator asks in chat to forward a workflow report) — include all key fields:
+
+```
+[采集端心跳] <工作流名称>
+
+Workflow ID: <id>
+运行 ID: <run_id>
+状态: <status>
+当前阶段: <phase>
+采集尝试次数: <n>
+分析尝试次数: <n>
+执行结论: <conclusion>
+```
+
+Manual relay pattern (for owner-only reports):
+- User asks: "notify me only, not business targets" → reply directly as your response text (API server delivers it).
+- The reply content IS the notification — no extra tools needed.
+- Include all key fields: workflow ID, run ID, phase, status, attempts, error details, collection summary.
+- Do NOT also call send_message / lark-cli / lark-cli / Feishu REST API — that would double-send.
+
+Reference: `references/feishu-open-id-cross-app.md` — Feishu open_id cross-app failure and correct routing. Collector heartbeat target is `ou_8ee224968aa26a74c7d30ba27fed5eeb` via collector app (`cli_aa8c4fb4c4f81cd3`). `ou_7ad5fc8012e2f741afc5346e05ffd447` is the analyzer app owner — NOT the collector heartbeat target. Full REST call pattern included.
+Reference: `references/collector-notification-credentials.md` — Collector Feishu notification credential sources: direct-reply pattern vs REST API pattern, and the known bug where `send_collector_notify.py` reads the wrong app secret from `config.json`.
+Reference: `references/workflow-state-querying.md` — How to query collector workflow state. **Primary source: `shared-data/` artifacts + agent logs.** The orchestrator's `state.db` (LangChain checkpoints) only stores `replenishment-alert` runs — NOT collector runs. See the reference file for the full state querying strategy.
+
+**Reusable monitoring script:** `scripts/monitor_purchase_confirmation.py` — drop-in script that queries collector session DB + shared-data, then sends Feishu admin notification. Run with `python scripts/monitor_purchase_confirmation.py <RUN_ID>`. Handles `orchestrator/state.db` ≠ collector workflow source correctly.
+
+Final check:
+- Collection only.
+- No business-user final notify.
+- Shared-data contains only this run outputs.
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/purchase-confirmation-workflow/SKILL.md b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/purchase-confirmation-workflow/SKILL.md
new file mode 100644
index 0000000..a1db2ca
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/purchase-confirmation-workflow/SKILL.md
@@ -0,0 +1,58 @@
+---
+name: purchase-confirmation-workflow
+description: 触发采购确认通知工作流 (purchase-confirmation)
+triggers:
+  - manual: 用户要求"执行/触发 purchase-confirmation"
+  - scheduled: 通过 cron 定时触发
+---
+
+# Purchase Confirmation Workflow
+
+Trigger the `purchase-confirmation` (采购确认通知工作流) from the `${GYXX_PROJECT_ROOT}` orchestrator.
+
+## ⚠️ Important Execution Notes
+
+- **执行时间长**：此工作流包含浏览器自动化（Chrome CDP），采集脚本执行时间约 **5-15 分钟**，请耐心等待，**不要**超时后随意清理锁和进程
+- **不要主动终止**：看到超时不要清理锁/杀进程，工作流可能仍在正常执行
+- **判断完成**：检查 `shared-data/purchase-confirmation/` 目录中是否有 CSV 文件（如 `采购确认通知汇总_YYYYMMDD_HHMMSS.csv`）
+
+## Workflow Command
+
+```bash
+cd ${GYXX_PROJECT_ROOT} && .venv/Scripts/python.exe run.py mcp-run purchase-confirmation
+```
+
+工作流通过 Hermes MCP 网关执行，包含采集端和分析端两个阶段。
+
+## 创建定时任务
+
+```python
+cronjob(action='create',
+        prompt='执行 purchase-confirmation 工作流。运行命令：cd ${GYXX_PROJECT_ROOT} && .venv/Scripts/python.exe run.py mcp-run purchase-confirmation',
+        schedule='0 9 * * *',  # 每天早上9点
+        name='purchase-confirmation-daily')
+```
+
+## 通知规则（关键）
+
+⚠️ **采集端通知规则**：采集端 Hermes 只通知运营，不通知业务目标用户。
+- 采集端通知对象：运营（飞书 OpenID `ou_7ad5fc8012e2f741afc5346e05ffd447`）
+- 业务最终通知：由分析端 Hermes 负责
+- 通知方式：通过当前会话直接响应（不调用 send_message/yuanbao 派）
+
+当用户请求"只通知我（不要通知业务目标用户）"时：
+1. 直接在当前会话中以文本形式输出报告
+2. 说明通知范围：仅运营，业务目标用户由分析端 Hermes 负责
+
+## 工作流配置参考
+
+- **工作流名称**：purchase-confirmation
+- **显示名称**：采购确认通知工作流
+- **采集脚本**：`${GYXX_PROJECT_ROOT}/orchestrator/scripts/collect_confirmation.ps1`
+- **输出文件**：`采购确认通知汇总_*.csv`（xlsx 格式）
+- **通知目标**（业务用户）：
+  - `ou_7ad5fc8012e2f741afc5346e05ffd447`
+  - `ou_339c1d396b397c97b08e1bf377963d40`
+  - `ou_54141d75b24d22ddc8c1c4d66e00b6e9`
+- **采集端绑定应用**：`cli_aa8c4fb4c4f81cd3`
+- **分析端绑定应用**：`cli_aa8c4fc918b85cce`
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/purchase-order-update-workflow/SKILL.md b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/purchase-order-update-workflow/SKILL.md
new file mode 100644
index 0000000..b73f465
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/purchase-order-update-workflow/SKILL.md
@@ -0,0 +1,34 @@
+---
+name: purchase-order-update-workflow
+description: 触发采购单日期更新工作流 (purchase-order-update)
+category: workflow
+triggers:
+  - manual: 用户要求"执行/触发 purchase-order-update"
+  - scheduled: 通过 cron 定时触发
+---
+
+# Purchase Order Update Workflow
+
+Trigger the `purchase-order-update`（采购单日期更新工作流）from the `${GYXX_PROJECT_ROOT}` orchestrator.
+
+## Workflow Command
+
+```bash
+cd ${GYXX_PROJECT_ROOT} && .venv/Scripts/python.exe run.py mcp-run purchase-order-update
+```
+
+## 工作流配置（来源 bytecode 反向提取）
+
+| 字段 | 值 |
+|------|-----|
+| **工作流名称** | purchase-order-update |
+| **显示名称** | 采购单日期更新工作流 |
+| **采集脚本** | `collect_purchase_order_update.ps1` |
+| **输出文件** | `update_result.json` |
+| **分析端通知方式** | `plugin:feishu.im.send_message` |
+| **采集端绑定应用** | `cli_aa8c4fb4c4f81cd3` |
+| **分析端绑定应用** | `cli_aa8c4fc918b85cce` |
+
+## ⚠️ config.py 源文件已丢失
+
+此工作流配置从 bytecode 反向提取，如需修改通知目标，需先重建 `config.py`。
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/replenishment-alert-workflow/SKILL.md b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/replenishment-alert-workflow/SKILL.md
new file mode 100644
index 0000000..3b4961f
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/replenishment-alert-workflow/SKILL.md
@@ -0,0 +1,41 @@
+---
+name: replenishment-alert-workflow
+description: 触发库存预警通知工作流 (replenishment-alert)
+triggers:
+  - manual: 用户要求"执行/触发 replenishment-alert"
+  - scheduled: 通过 cron 定时触发
+---
+
+# Replenishment Alert Workflow
+
+Trigger the `replenishment-alert` (库存预警通知工作流) from the `${GYXX_PROJECT_ROOT}` orchestrator.
+
+## Trigger Conditions
+
+- **手动触发**：用户明确要求"执行/触发 replenishment-alert"
+- **定时触发**：通过 cron 任务定期运行
+
+## Workflow Command
+
+```bash
+cd ${GYXX_PROJECT_ROOT} && .venv/Scripts/python.exe run.py mcp-run replenishment-alert
+```
+
+工作流通过 Hermes MCP 网关执行，包含采集端和分析端两个阶段。
+
+## 已知问题
+
+⚠️ **通知重复问题**：如需修复，从 `config.py` 的 `notification_targets.threshold_alert` 中移除 analyzer owner。但注意：`config.py` 源文件已丢失（仅 pyc 缓存），修改前需先重建配置。
+
+⚠️ **config.py 源文件已丢失**：仅存 `__pycache__/config.cpython-311.pyc`，如需修改通知目标，需从 bytecode 反向提取常量后再重建文件。
+
+## 工作流配置参考
+
+- **工作流名称**：replenishment-alert
+- **显示名称**：库存预警通知工作流
+- **采集脚本**：`${GYXX_PROJECT_ROOT}/orchestrator/scripts/collect_replenishment.ps1`
+- **输出文件**：`threshold_alert_data.json`
+- **通知目标**（业务用户，仅 `threshold_alert`，来源 bytecode const[53]）：
+  - `ou_b76e4cbdb24fe28cebd45ad091b60224`（黄坤平）
+- **采集端绑定应用**：`cli_aa8c4fb4c4f81cd3`
+- **分析端绑定应用**：`cli_aa8c4fc918b85cce`
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/replenishment-workflow/SKILL.md b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/replenishment-workflow/SKILL.md
new file mode 100644
index 0000000..82d7be0
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/replenishment-workflow/SKILL.md
@@ -0,0 +1,54 @@
+---
+name: replenishment-workflow
+description: 触发补货建议工作流 (replenishment)
+triggers:
+  - manual: 用户要求"执行/触发 replenishment"
+  - scheduled: 通过 cron 定时触发
+---
+
+# Replenishment Workflow
+
+Trigger the `replenishment` (补货建议工作流) from the `${GYXX_PROJECT_ROOT}` orchestrator.
+
+## ⚠️ Important Execution Notes
+
+- **执行时间长**：此工作流包含浏览器自动化（Chrome CDP），采集脚本执行时间约 **5-15 分钟**，请耐心等待
+- **判断完成**：检查 `shared-data/replenishment/` 目录中是否有 `pending_insert.json` 和 `notify_data.json`
+- **后台执行**：建议使用 `background=true` + `notify_on_complete=true` 执行，避免超时
+
+## Workflow Command
+
+```bash
+cd ${GYXX_PROJECT_ROOT} && .venv/Scripts/python.exe run.py mcp-run replenishment
+```
+
+## 创建定时任务
+
+```python
+cronjob(action='create',
+        prompt='执行 replenishment 工作流。运行命令：cd ${GYXX_PROJECT_ROOT} && .venv/Scripts/python.exe run.py mcp-run replenishment',
+        schedule='0 9 * * *',  # 每天早上9点
+        name='replenishment-daily')
+```
+
+## 通知目标（分析端）
+
+仅通知运营（`ou_7ad5fc8012e2f741afc5346e05ffd447`）。不通知黄坤平，不通知审单群。
+
+## 工作流配置参考
+
+- **工作流名称**：replenishment
+- **显示名称**：补货建议工作流
+- **采集脚本**：`${GYXX_PROJECT_ROOT}/orchestrator/scripts/collect_replenishment.ps1`
+- **输出文件**：`pending_insert.json`, `notify_data.json`
+- **多维表**：应用 `cli_aa8c4fc918b85cce` 需对多维表 `Th0jbMHHQa7a8Lse3CicfSMBnxb` 有写入权限
+- **采集端绑定应用**：`cli_aa8c4fb4c4f81cd3`
+- **分析端绑定应用**：`cli_aa8c4fc918b85cce`
+
+## 已知问题
+
+⚠️ **多维表写入权限**：如遇以下错误，需在飞书开放平台为应用 `cli_aa8c4fc918b85cce` 开启多维表写入权限：
+- `91403 Forbidden` — 无权限访问多维表
+- `99992402 field validation failed` — 字段验证失败（常见于 app 无写入权限或字段类型不匹配）
+
+写入失败**不影响通知发送**（符合 HARD RULES），但待处理数据需手动补充。
diff --git a/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/workflow-trigger-only/SKILL.md b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/workflow-trigger-only/SKILL.md
new file mode 100644
index 0000000..4693b58
--- /dev/null
+++ b/src/gyxx_flow/modules/supply_chain/runtime/skills/collector/workflow/workflow-trigger-only/SKILL.md
@@ -0,0 +1,57 @@
+---
+name: workflow-trigger-only
+description: 工作流触发原则：只执行既有命令，不自主生成代码修改流程
+triggers:
+  - manual: 用户要求将某个操作原则沉淀为 skill 时创建
+---
+
+# Workflow Trigger Only
+
+## 核心原则
+
+**只触发既有工作流，不自主生成或修改流程代码。**
+
+## 执行方式
+
+```
+cd ${GYXX_PROJECT_ROOT} && .venv/Scripts/python.exe run.py mcp-run <工作流名>
+```
+
+支持的 `<工作流名>`：
+- `replenishment-alert` — 库存预警通知
+- `purchase-confirmation` — 采购确认通知
+- `replenishment` — 补货建议
+
+## 规则
+
+1. **只运行既有命令** — 使用 `run.py mcp-run` 触发，不写 Python 脚本绕过后续流程
+2. **不自主修改流程代码** — 不修改 `mcp_workflow.py`、通知脚本、采集脚本等，除非用户明确要求
+3. **不自主生成新代码** — 不创建新脚本自行完成通知/写表等操作
+4. **编排逻辑由既有代码决定** — 采集→分析→通知的分段流程和通知对象由 `mcp_workflow.py` 和分析端 Hermes 控制
+
+## 例外：手动转发心跳/执行报告
+
+"只触发既有命令"规则**仅适用于 orchestrator 工作流执行**。当用户要求将某个工作流的心跳/执行报告**手动转发**给运营时：
+
+1. 构造心跳消息内容（包含工作流 ID、运行 ID、状态、错误信息等）
+2. 通过飞书采集端应用（`cli_aa8c4fb4c4f81cd3`）将报告以 **Feishu DM** 形式发送给运营（`ou_8ee224968aa26a74c7d30ba27fed5eeb`）
+3. **不**通知业务目标用户，不走 yuanbao group
+4. 使用 `feishu` skill 中的 temp 脚本模式发送（写临时脚本 → terminal 运行 → 删除），或直接 reply 让 gateway 转发
+
+> 注意：**自动心跳已由 `orchestrator/mcp_workflow.py` 用 Schema 2.0 interactive card 发送**（target 同样是 `ou_8ee224968aa26a74c7d30ba27fed5eeb`），本例外仅适用于**人工额外转发**场景（比如重发、转发给第三方等），不要重复发送已自动发出的心跳。
+
+这是人工操作员在对话中转发报告的合法场景，与 orchestrator 自动执行工作流是两回事。
+
+## 触发命令参考
+
+| 工作流 | 命令 |
+|--------|------|
+| 库存预警通知 | `cd ${GYXX_PROJECT_ROOT} && .venv/Scripts/python.exe run.py mcp-run replenishment-alert` |
+| 采购确认通知 | `cd ${GYXX_PROJECT_ROOT} && .venv/Scripts/python.exe run.py mcp-run purchase-confirmation` |
+| 补货建议 | `cd ${GYXX_PROJECT_ROOT} && .venv/Scripts/python.exe run.py mcp-run replenishment` |
+
+## 为什么这样做
+
+- 工作流的编排逻辑（采集、分析、通知分段执行）由 `mcp_workflow.py` 控制
+- 通知对象由 `config.py` 和分析端 prompt 决定，不应在触发层硬编码
+- 自主生成代码容易出错且难以追踪
diff --git a/src/gyxx_flow/ops/__init__.py b/src/gyxx_flow/ops/__init__.py
new file mode 100644
index 0000000..38c9fc1
--- /dev/null
+++ b/src/gyxx_flow/ops/__init__.py
@@ -0,0 +1,22 @@
+"""Operational run indexing and durable outbox services."""
+
+from .effects import (
+    EffectAlreadyApplied,
+    EffectClaim,
+    EffectLedger,
+    EffectStateAmbiguous,
+)
+from .index import RunIndex, RunRecord
+from .outbox import IdempotencyConflict, Outbox, OutboxMessage
+
+__all__ = [
+    "IdempotencyConflict",
+    "EffectAlreadyApplied",
+    "EffectClaim",
+    "EffectLedger",
+    "EffectStateAmbiguous",
+    "Outbox",
+    "OutboxMessage",
+    "RunIndex",
+    "RunRecord",
+]
diff --git a/src/gyxx_flow/ops/effects.py b/src/gyxx_flow/ops/effects.py
new file mode 100644
index 0000000..0caa0de
--- /dev/null
+++ b/src/gyxx_flow/ops/effects.py
@@ -0,0 +1,150 @@
+"""Crash-safe idempotency receipts for opaque production side effects."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import re
+import uuid
+from dataclasses import dataclass
+from datetime import date, datetime, timezone
+from pathlib import Path
+
+from gyxx_flow.core.artifacts import atomic_write_json
+from gyxx_flow.core.locks import LockManager
+
+_ID = re.compile(r"^[a-z][a-z0-9]*(?:[._-][a-z0-9]+)*$")
+_RUN_ID = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._:-]*$")
+
+
+class EffectAlreadyApplied(RuntimeError):
+    """The logical production effect already has a success receipt."""
+
+
+class EffectStateAmbiguous(RuntimeError):
+    """A prior attempt may have produced an external effect."""
+
+
+@dataclass(frozen=True, slots=True)
+class EffectClaim:
+    effect_id: str
+    workflow_id: str
+    business_date: str
+    step_id: str
+    run_id: str
+
+
+class EffectLedger:
+    """Prevent automatic replay when an opaque sink may already have written."""
+
+    def __init__(self, data_root: Path | str) -> None:
+        self._data_root = Path(data_root).expanduser().resolve()
+        self._root = self._data_root / "state" / "ops" / "effects"
+        self._locks = LockManager(self._data_root / "state" / "locks")
+
+    def begin(
+        self,
+        *,
+        workflow_id: str,
+        business_date: str,
+        step_id: str,
+        run_id: str,
+    ) -> EffectClaim:
+        _validate(workflow_id, business_date, step_id, run_id)
+        effect_id = _effect_id(workflow_id, business_date, step_id)
+        claim = EffectClaim(effect_id, workflow_id, business_date, step_id, run_id)
+        with self._lock(effect_id):
+            path = self._path(effect_id)
+            if path.exists():
+                payload = self._read(path)
+                if payload["status"] == "applied":
+                    raise EffectAlreadyApplied(
+                        f"effect already applied: {workflow_id}/{business_date}/{step_id}"
+                    )
+                raise EffectStateAmbiguous(
+                    f"effect requires reconciliation: {workflow_id}/{business_date}/{step_id}"
+                )
+            now = _now()
+            atomic_write_json(
+                path,
+                {
+                    "schema_version": 1,
+                    "effect_id": effect_id,
+                    "workflow_id": workflow_id,
+                    "business_date": business_date,
+                    "step_id": step_id,
+                    "run_id": run_id,
+                    "status": "in_progress",
+                    "claimed_at": now,
+                    "updated_at": now,
+                    "reason": None,
+                },
+            )
+        return claim
+
+    def mark_applied(self, claim: EffectClaim) -> None:
+        self._finish(claim, status="applied", reason=None)
+
+    def mark_ambiguous(self, claim: EffectClaim, *, reason: str) -> None:
+        if not isinstance(reason, str) or not reason or len(reason) > 200:
+            raise ValueError("invalid effect ambiguity reason")
+        self._finish(claim, status="ambiguous", reason=reason)
+
+    def _finish(self, claim: EffectClaim, *, status: str, reason: str | None) -> None:
+        with self._lock(claim.effect_id):
+            path = self._path(claim.effect_id)
+            payload = self._read(path)
+            if (
+                payload["status"] != "in_progress"
+                or payload["run_id"] != claim.run_id
+            ):
+                raise EffectStateAmbiguous("effect claim no longer owns the receipt")
+            payload.update(status=status, reason=reason, updated_at=_now())
+            atomic_write_json(path, payload)
+
+    def _path(self, effect_id: str) -> Path:
+        return self._root / f"{effect_id}.json"
+
+    def _lock(self, effect_id: str):  # type: ignore[no-untyped-def]
+        return self._locks.acquire(
+            f"ops:effect:{effect_id}",
+            owner=f"effect-{uuid.uuid4().hex}",
+            timeout_seconds=30,
+        )
+
+    @staticmethod
+    def _read(path: Path) -> dict[str, object]:
+        try:
+            payload = json.loads(path.read_text(encoding="utf-8"))
+        except (OSError, json.JSONDecodeError) as exc:
+            raise EffectStateAmbiguous("effect receipt is unreadable") from exc
+        if (
+            not isinstance(payload, dict)
+            or payload.get("schema_version") != 1
+            or payload.get("status") not in {"in_progress", "applied", "ambiguous"}
+        ):
+            raise EffectStateAmbiguous("effect receipt is invalid")
+        return payload
+
+
+def _effect_id(workflow_id: str, business_date: str, step_id: str) -> str:
+    key = f"{workflow_id}\0{business_date}\0{step_id}"
+    return "effect_" + hashlib.sha256(key.encode("utf-8")).hexdigest()
+
+
+def _validate(
+    workflow_id: str, business_date: str, step_id: str, run_id: str
+) -> None:
+    if not _ID.fullmatch(workflow_id) or not _ID.fullmatch(step_id):
+        raise ValueError("invalid effect identity")
+    try:
+        parsed = date.fromisoformat(business_date)
+    except (TypeError, ValueError) as exc:
+        raise ValueError("invalid effect business date") from exc
+    if parsed.isoformat() != business_date or not _RUN_ID.fullmatch(run_id):
+        raise ValueError("invalid effect identity")
+
+
+def _now() -> str:
+    return datetime.now(timezone.utc).isoformat(timespec="seconds")
+
diff --git a/src/gyxx_flow/ops/index.py b/src/gyxx_flow/ops/index.py
new file mode 100644
index 0000000..88ea5f4
--- /dev/null
+++ b/src/gyxx_flow/ops/index.py
@@ -0,0 +1,241 @@
+"""Durable, queryable snapshots of workflow run journals."""
+
+from __future__ import annotations
+
+import json
+import re
+import uuid
+from dataclasses import asdict, dataclass
+from datetime import date, datetime, timezone
+from pathlib import Path
+from typing import Any
+
+from gyxx_flow.core.artifacts import atomic_write_json
+from gyxx_flow.core.locks import LockManager
+from gyxx_flow.core.records import RunJournal
+
+_SAFE_ID = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._-]*$")
+_RUN_STATUSES = frozenset({"running", "success", "failed", "cancelled"})
+_STEP_STATUSES = frozenset({"running", "success", "failed", "skipped"})
+
+
+@dataclass(frozen=True, slots=True)
+class RunRecord:
+    """A compact, non-secret operational projection of a run journal."""
+
+    workflow_id: str
+    run_id: str
+    business_date: str
+    shadow: bool
+    status: str
+    started_at: str
+    ended_at: str | None
+    error: str | None
+    step_counts: dict[str, int]
+    journal_path: str
+    indexed_at: str
+    schema_version: int = 1
+
+    @classmethod
+    def from_dict(cls, payload: dict[str, Any]) -> "RunRecord":
+        try:
+            record = cls(
+                workflow_id=payload["workflow_id"],
+                run_id=payload["run_id"],
+                business_date=payload["business_date"],
+                shadow=payload["shadow"],
+                status=payload["status"],
+                started_at=payload["started_at"],
+                ended_at=payload["ended_at"],
+                error=payload["error"],
+                step_counts=dict(payload["step_counts"]),
+                journal_path=payload["journal_path"],
+                indexed_at=payload["indexed_at"],
+                schema_version=payload["schema_version"],
+            )
+        except (KeyError, TypeError, ValueError) as exc:
+            raise ValueError("invalid run index record") from exc
+        _validate_record(record)
+        return record
+
+    def to_dict(self) -> dict[str, Any]:
+        return asdict(self)
+
+
+class RunIndex:
+    """Per-run sharded index built from :class:`RunJournal` JSON files."""
+
+    def __init__(self, data_root: Path | str, *, lock_manager: LockManager | None = None) -> None:
+        self._data_root = Path(data_root).expanduser().resolve()
+        self._root = self._data_root / "state" / "ops" / "run-index"
+        self._locks = lock_manager or LockManager(self._data_root / "state" / "locks")
+
+    def index_journal(self, journal: RunJournal | Path | str) -> RunRecord:
+        journal_path = journal.path if isinstance(journal, RunJournal) else Path(journal)
+        journal_path = journal_path.expanduser().resolve()
+        try:
+            raw = json.loads(journal_path.read_text(encoding="utf-8"))
+        except (OSError, json.JSONDecodeError) as exc:
+            raise ValueError("invalid run journal") from exc
+        record = _record_from_journal(raw, journal_path)
+        destination = self._record_path(record.run_id)
+        resource = f"ops:run-index:{_digest(record.run_id)}"
+        with self._locks.acquire(resource, owner=_owner(), timeout_seconds=30):
+            atomic_write_json(destination, record.to_dict())
+        return record
+
+    def get(self, run_id: str) -> RunRecord | None:
+        path = self._record_path(run_id)
+        if not path.exists():
+            return None
+        return _read_record(path)
+
+    def query(
+        self,
+        *,
+        workflow_id: str | None = None,
+        status: str | None = None,
+        business_date: str | None = None,
+        run_id: str | None = None,
+    ) -> tuple[RunRecord, ...]:
+        if workflow_id is not None:
+            _require_safe_id(workflow_id, "workflow_id")
+        if run_id is not None:
+            _require_safe_id(run_id, "run_id")
+        if status is not None and status not in _RUN_STATUSES:
+            raise ValueError(f"invalid run status: {status!r}")
+        if business_date is not None:
+            _require_business_date(business_date)
+
+        records = []
+        for path in sorted(self._root.glob("*.json")):
+            record = _read_record(path)
+            if workflow_id is not None and record.workflow_id != workflow_id:
+                continue
+            if status is not None and record.status != status:
+                continue
+            if business_date is not None and record.business_date != business_date:
+                continue
+            if run_id is not None and record.run_id != run_id:
+                continue
+            records.append(record)
+        records.sort(key=lambda item: (item.started_at, item.run_id), reverse=True)
+        return tuple(records)
+
+    def _record_path(self, run_id: str) -> Path:
+        _require_safe_id(run_id, "run_id")
+        return self._root / f"{run_id}.json"
+
+
+def _record_from_journal(payload: Any, journal_path: Path) -> RunRecord:
+    if not isinstance(payload, dict):
+        raise ValueError("invalid run journal")
+    required = {
+        "workflow_id",
+        "run_id",
+        "business_date",
+        "shadow",
+        "started_at",
+        "ended_at",
+        "status",
+        "error",
+        "steps",
+    }
+    if not required.issubset(payload):
+        raise ValueError("invalid run journal")
+    try:
+        workflow_id = payload["workflow_id"]
+        run_id = payload["run_id"]
+        business_date = payload["business_date"]
+        status = payload["status"]
+        shadow = payload["shadow"]
+        started_at = payload["started_at"]
+        ended_at = payload["ended_at"]
+        error = payload["error"]
+        steps = payload["steps"]
+        _require_safe_id(workflow_id, "workflow_id")
+        _require_safe_id(run_id, "run_id")
+        _require_business_date(business_date)
+        if status not in _RUN_STATUSES:
+            raise ValueError("invalid status")
+        if not isinstance(shadow, bool) or not isinstance(started_at, str):
+            raise ValueError("invalid journal fields")
+        if ended_at is not None and not isinstance(ended_at, str):
+            raise ValueError("invalid ended_at")
+        if error is not None and not isinstance(error, str):
+            raise ValueError("invalid error")
+        if not isinstance(steps, dict):
+            raise ValueError("invalid steps")
+        counts = {name: 0 for name in ("success", "failed", "skipped", "running")}
+        for step in steps.values():
+            if not isinstance(step, dict) or step.get("status") not in _STEP_STATUSES:
+                raise ValueError("invalid step")
+            counts[step["status"]] += 1
+    except (TypeError, ValueError) as exc:
+        raise ValueError("invalid run journal") from exc
+
+    record = RunRecord(
+        workflow_id=workflow_id,
+        run_id=run_id,
+        business_date=business_date,
+        shadow=shadow,
+        status=status,
+        started_at=started_at,
+        ended_at=ended_at,
+        error=error,
+        step_counts=counts,
+        journal_path=str(journal_path),
+        indexed_at=datetime.now(timezone.utc).isoformat(timespec="seconds"),
+    )
+    _validate_record(record)
+    return record
+
+
+def _read_record(path: Path) -> RunRecord:
+    try:
+        payload = json.loads(path.read_text(encoding="utf-8"))
+    except (OSError, json.JSONDecodeError) as exc:
+        raise ValueError(f"invalid run index record: {path.name}") from exc
+    if not isinstance(payload, dict):
+        raise ValueError(f"invalid run index record: {path.name}")
+    return RunRecord.from_dict(payload)
+
+
+def _validate_record(record: RunRecord) -> None:
+    _require_safe_id(record.workflow_id, "workflow_id")
+    _require_safe_id(record.run_id, "run_id")
+    _require_business_date(record.business_date)
+    if record.status not in _RUN_STATUSES or record.schema_version != 1:
+        raise ValueError("invalid run index record")
+    if not isinstance(record.shadow, bool) or not isinstance(record.started_at, str):
+        raise ValueError("invalid run index record")
+    if set(record.step_counts) != _STEP_STATUSES:
+        raise ValueError("invalid run index record")
+    if any(not isinstance(value, int) or value < 0 for value in record.step_counts.values()):
+        raise ValueError("invalid run index record")
+
+
+def _require_safe_id(value: Any, name: str) -> None:
+    if not isinstance(value, str) or not _SAFE_ID.fullmatch(value):
+        raise ValueError(f"invalid {name}")
+
+
+def _require_business_date(value: Any) -> None:
+    if not isinstance(value, str):
+        raise ValueError("invalid business_date")
+    try:
+        parsed = date.fromisoformat(value)
+    except ValueError as exc:
+        raise ValueError("invalid business_date") from exc
+    if parsed.isoformat() != value:
+        raise ValueError("invalid business_date")
+
+
+def _owner() -> str:
+    return f"ops-{uuid.uuid4().hex}"
+
+
+def _digest(value: str) -> str:
+    import hashlib
+
+    return hashlib.sha256(value.encode("utf-8")).hexdigest()[:32]
diff --git a/src/gyxx_flow/ops/outbox.py b/src/gyxx_flow/ops/outbox.py
new file mode 100644
index 0000000..b8c5096
--- /dev/null
+++ b/src/gyxx_flow/ops/outbox.py
@@ -0,0 +1,287 @@
+"""Atomic local-file transactional outbox with idempotent replay semantics."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import re
+import uuid
+from dataclasses import asdict, dataclass, replace
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any, Callable
+
+from gyxx_flow.core.artifacts import atomic_write_json
+from gyxx_flow.core.locks import LockManager
+
+_TOPIC = re.compile(r"^[a-z][a-z0-9]*(?:[._-][a-z0-9]+)*$")
+_MESSAGE_ID = re.compile(r"^msg_[0-9a-f]{64}$")
+_STATUSES = frozenset({"pending", "sent", "failed"})
+
+
+class IdempotencyConflict(ValueError):
+    """Raised when a key is reused for a different logical message."""
+
+
+@dataclass(frozen=True, slots=True)
+class OutboxMessage:
+    schema_version: int
+    message_id: str
+    idempotency_key: str
+    topic: str
+    payload: Any
+    run_id: str
+    status: str
+    attempts: int
+    created_at: str
+    updated_at: str
+    sent_at: str | None
+    error: str | None
+
+    @classmethod
+    def from_dict(cls, payload: dict[str, Any]) -> "OutboxMessage":
+        try:
+            message = cls(**payload)
+        except (TypeError, ValueError) as exc:
+            raise ValueError("invalid outbox message") from exc
+        _validate_message(message)
+        return message
+
+    def to_dict(self) -> dict[str, Any]:
+        return asdict(self)
+
+
+class Outbox:
+    """A durable, at-least-once outbox.
+
+    Dispatchers must forward ``idempotency_key`` to the destination. A sent
+    message is never dispatched again; a failed message is replayed under the
+    same message ID and idempotency key.
+    """
+
+    def __init__(self, data_root: Path | str, *, lock_manager: LockManager | None = None) -> None:
+        self._data_root = Path(data_root).expanduser().resolve()
+        self._root = self._data_root / "state" / "ops" / "outbox" / "messages"
+        self._locks = lock_manager or LockManager(self._data_root / "state" / "locks")
+
+    def enqueue(
+        self,
+        *,
+        idempotency_key: str,
+        topic: str,
+        payload: Any,
+        run_id: str,
+    ) -> OutboxMessage:
+        _validate_new_message(idempotency_key, topic, payload, run_id)
+        normalized_payload = _json_copy(payload)
+        message_id = _message_id(idempotency_key)
+        path = self.message_path(message_id)
+        with self._message_lock(message_id):
+            if path.exists():
+                existing = self._read(path)
+                if (
+                    existing.idempotency_key != idempotency_key
+                    or existing.topic != topic
+                    or existing.payload != normalized_payload
+                    or existing.run_id != run_id
+                ):
+                    raise IdempotencyConflict(
+                        "idempotency key is already bound to a different message"
+                    )
+                return existing
+            now = _now()
+            message = OutboxMessage(
+                schema_version=1,
+                message_id=message_id,
+                idempotency_key=idempotency_key,
+                topic=topic,
+                payload=normalized_payload,
+                run_id=run_id,
+                status="pending",
+                attempts=0,
+                created_at=now,
+                updated_at=now,
+                sent_at=None,
+                error=None,
+            )
+            atomic_write_json(path, message.to_dict())
+            return message
+
+    def get(self, message_id: str) -> OutboxMessage | None:
+        path = self.message_path(message_id)
+        if not path.exists():
+            return None
+        return self._read(path)
+
+    def list(self, *, status: str | None = None) -> tuple[OutboxMessage, ...]:
+        if status is not None and status not in _STATUSES:
+            raise ValueError(f"invalid outbox status: {status!r}")
+        messages = []
+        for path in sorted(self._root.glob("*.json")):
+            message = self._read(path)
+            if status is None or message.status == status:
+                messages.append(message)
+        messages.sort(key=lambda item: (item.created_at, item.message_id))
+        return tuple(messages)
+
+    def mark_sent(self, message_id: str) -> OutboxMessage:
+        with self._message_lock(message_id):
+            message = self._require(message_id)
+            if message.status == "sent":
+                return message
+            if message.status != "pending":
+                raise ValueError(f"cannot mark {message.status} message as sent; replay it first")
+            return self._write_sent(message)
+
+    def mark_failed(self, message_id: str, *, error: str) -> OutboxMessage:
+        if not isinstance(error, str) or not error.strip():
+            raise ValueError("error must be a non-empty string")
+        with self._message_lock(message_id):
+            message = self._require(message_id)
+            if message.status != "pending":
+                raise ValueError(f"cannot mark {message.status} message as failed")
+            return self._write_failed(message, error=error)
+
+    def replay(self, message_id: str) -> OutboxMessage:
+        with self._message_lock(message_id):
+            message = self._require(message_id)
+            if message.status in {"pending", "sent"}:
+                return message
+            replayed = replace(
+                message,
+                status="pending",
+                updated_at=_now(),
+                sent_at=None,
+                error=None,
+            )
+            atomic_write_json(self.message_path(message_id), replayed.to_dict())
+            return replayed
+
+    def dispatch(
+        self,
+        message_id: str,
+        sender: Callable[[OutboxMessage], object],
+    ) -> OutboxMessage:
+        """Dispatch one pending message while holding its per-message lock.
+
+        ``sender`` must make its downstream operation idempotent using the key
+        on the supplied message. Exceptions are persisted as ``failed``.
+        """
+
+        if not callable(sender):
+            raise TypeError("sender must be callable")
+        with self._message_lock(message_id):
+            message = self._require(message_id)
+            if message.status == "sent":
+                return message
+            if message.status != "pending":
+                raise ValueError(f"cannot dispatch {message.status} message; replay it first")
+            try:
+                sender(message)
+            except Exception as exc:
+                return self._write_failed(message, error=f"{type(exc).__name__}: {exc}")
+            return self._write_sent(message)
+
+    def message_path(self, message_id: str) -> Path:
+        if not isinstance(message_id, str) or not _MESSAGE_ID.fullmatch(message_id):
+            raise ValueError("invalid message_id")
+        return self._root / f"{message_id}.json"
+
+    def _message_lock(self, message_id: str):  # type: ignore[no-untyped-def]
+        if not isinstance(message_id, str) or not _MESSAGE_ID.fullmatch(message_id):
+            raise ValueError("invalid message_id")
+        return self._locks.acquire(
+            f"ops:outbox:{message_id}", owner=f"ops-{uuid.uuid4().hex}", timeout_seconds=30
+        )
+
+    def _require(self, message_id: str) -> OutboxMessage:
+        path = self.message_path(message_id)
+        if not path.exists():
+            raise KeyError(f"unknown outbox message: {message_id}")
+        return self._read(path)
+
+    def _read(self, path: Path) -> OutboxMessage:
+        try:
+            payload = json.loads(path.read_text(encoding="utf-8"))
+        except (OSError, json.JSONDecodeError) as exc:
+            raise ValueError(f"invalid outbox message: {path.name}") from exc
+        if not isinstance(payload, dict):
+            raise ValueError(f"invalid outbox message: {path.name}")
+        return OutboxMessage.from_dict(payload)
+
+    def _write_sent(self, message: OutboxMessage) -> OutboxMessage:
+        now = _now()
+        sent = replace(
+            message,
+            status="sent",
+            attempts=message.attempts + 1,
+            updated_at=now,
+            sent_at=now,
+            error=None,
+        )
+        atomic_write_json(self.message_path(message.message_id), sent.to_dict())
+        return sent
+
+    def _write_failed(self, message: OutboxMessage, *, error: str) -> OutboxMessage:
+        failed = replace(
+            message,
+            status="failed",
+            attempts=message.attempts + 1,
+            updated_at=_now(),
+            sent_at=None,
+            error=error,
+        )
+        atomic_write_json(self.message_path(message.message_id), failed.to_dict())
+        return failed
+
+
+def _validate_new_message(idempotency_key: str, topic: str, payload: Any, run_id: str) -> None:
+    if (
+        not isinstance(idempotency_key, str)
+        or not idempotency_key
+        or len(idempotency_key) > 512
+        or any(ord(character) < 32 for character in idempotency_key)
+    ):
+        raise ValueError("invalid idempotency_key")
+    if not isinstance(topic, str) or not _TOPIC.fullmatch(topic):
+        raise ValueError("invalid topic")
+    if not isinstance(run_id, str) or not run_id or len(run_id) > 512:
+        raise ValueError("invalid run_id")
+    _json_copy(payload)
+
+
+def _validate_message(message: OutboxMessage) -> None:
+    _validate_new_message(
+        message.idempotency_key, message.topic, message.payload, message.run_id
+    )
+    if message.schema_version != 1 or message.message_id != _message_id(message.idempotency_key):
+        raise ValueError("invalid outbox message")
+    if message.status not in _STATUSES:
+        raise ValueError("invalid outbox message")
+    if not isinstance(message.attempts, int) or isinstance(message.attempts, bool) or message.attempts < 0:
+        raise ValueError("invalid outbox message")
+    if message.sent_at is not None and not isinstance(message.sent_at, str):
+        raise ValueError("invalid outbox message")
+    if message.error is not None and not isinstance(message.error, str):
+        raise ValueError("invalid outbox message")
+    if message.status == "sent" and (message.sent_at is None or message.error is not None):
+        raise ValueError("invalid outbox message")
+    if message.status == "failed" and (message.sent_at is not None or not message.error):
+        raise ValueError("invalid outbox message")
+    if message.status == "pending" and (message.sent_at is not None or message.error is not None):
+        raise ValueError("invalid outbox message")
+
+
+def _message_id(idempotency_key: str) -> str:
+    return f"msg_{hashlib.sha256(idempotency_key.encode('utf-8')).hexdigest()}"
+
+
+def _json_copy(payload: Any) -> Any:
+    try:
+        return json.loads(json.dumps(payload, ensure_ascii=False, sort_keys=True))
+    except (TypeError, ValueError) as exc:
+        raise ValueError("payload must be JSON serializable") from exc
+
+
+def _now() -> str:
+    return datetime.now(timezone.utc).isoformat(timespec="seconds")
diff --git a/src/gyxx_flow/runtime_bootstrap.py b/src/gyxx_flow/runtime_bootstrap.py
new file mode 100644
index 0000000..254413b
--- /dev/null
+++ b/src/gyxx_flow/runtime_bootstrap.py
@@ -0,0 +1,8 @@
+"""Compatibility exports for the adapter-owned runtime bootstrap."""
+
+from gyxx_flow.adapters.bootstrap import (
+    bootstrap_current_process,
+    safe_bootstrap_current_process,
+)
+
+__all__ = ["bootstrap_current_process", "safe_bootstrap_current_process"]
diff --git a/src/gyxx_flow/runtime_exec.py b/src/gyxx_flow/runtime_exec.py
new file mode 100644
index 0000000..19d7b6f
--- /dev/null
+++ b/src/gyxx_flow/runtime_exec.py
@@ -0,0 +1,74 @@
+"""Small shell-friendly bridge for nested migrated script bindings."""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import subprocess
+import sys
+
+from gyxx_flow.adapters.integration import RuntimeIntegrationCatalog
+from gyxx_flow.core.config import Settings
+from gyxx_flow.script_catalog import ScriptCatalog
+
+
+def _catalog() -> tuple[RuntimeIntegrationCatalog, ScriptCatalog]:
+    settings = Settings.from_env()
+    scripts = ScriptCatalog.discover_default()
+    return (
+        RuntimeIntegrationCatalog.load_default(
+            project_root=settings.project_root,
+            data_root=settings.data_root,
+            scripts=scripts,
+        ),
+        scripts,
+    )
+
+
+def main(argv: list[str] | None = None) -> int:
+    parser = argparse.ArgumentParser(prog="python -m gyxx_flow.runtime_exec")
+    commands = parser.add_subparsers(dest="command", required=True)
+    env_parser = commands.add_parser("env")
+    env_parser.add_argument("script_id")
+    run_parser = commands.add_parser("run")
+    run_parser.add_argument("script_id")
+    run_parser.add_argument("args", nargs=argparse.REMAINDER)
+    arguments = parser.parse_args(argv)
+    integrations, scripts = _catalog()
+    binding = integrations.binding_for(arguments.script_id)
+    if arguments.command == "env":
+        print(
+            json.dumps(
+                {
+                    "script_id": binding.script_id,
+                    "cdp_port": binding.cdp_port,
+                    "cdp_url": binding.cdp_url,
+                    "profile_dir": str(binding.profile_dir),
+                    "cookie_file": str(binding.cookie_file),
+                    "storage_state_file": str(binding.storage_state_file),
+                },
+                ensure_ascii=False,
+            )
+        )
+        return 0
+    script = scripts.get(arguments.script_id)
+    environment = integrations.environment_for(arguments.script_id, os.environ)
+    suffix = script.path.suffix.casefold()
+    if suffix in {".py", ".pyw"}:
+        command = [sys.executable, str(script.path), *arguments.args]
+    elif suffix == ".ps1":
+        command = ["powershell.exe", "-NoProfile", "-File", str(script.path), *arguments.args]
+    else:
+        line = subprocess.list2cmdline([str(script.path), *arguments.args])
+        command = ["cmd.exe", "/d", "/s", "/c", line]
+    return subprocess.run(
+        command,
+        cwd=script.path.parent,
+        env=environment,
+        check=False,
+    ).returncode
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/src/gyxx_flow/scheduler.py b/src/gyxx_flow/scheduler.py
new file mode 100644
index 0000000..418b422
--- /dev/null
+++ b/src/gyxx_flow/scheduler.py
@@ -0,0 +1,384 @@
+"""Declarative Windows Task Scheduler plans and read-only drift detection."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import os
+import subprocess
+import tempfile
+import xml.etree.ElementTree as ET
+from dataclasses import asdict, dataclass
+from datetime import date, datetime, time
+from pathlib import Path
+from typing import Literal, Sequence
+
+from gyxx_flow.catalog import ScheduleEntry, WorkflowCatalog
+from gyxx_flow.core.artifacts import atomic_write_json
+
+TASK_NAMESPACE = "http://schemas.microsoft.com/windows/2004/02/mit/task"
+ET.register_namespace("", TASK_NAMESPACE)
+
+
+def _tag(name: str) -> str:
+    return f"{{{TASK_NAMESPACE}}}{name}"
+
+
+@dataclass(frozen=True, slots=True)
+class SchedulerConfig:
+    python_executable: Path
+    project_root: Path
+    task_path: str = "\\GYXX\\"
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "python_executable", Path(self.python_executable).resolve())
+        object.__setattr__(self, "project_root", Path(self.project_root).resolve())
+        normalized = self.task_path.replace("/", "\\")
+        if not normalized.startswith("\\"):
+            normalized = "\\" + normalized
+        if not normalized.endswith("\\"):
+            normalized += "\\"
+        if normalized in {"\\", ""}:
+            raise ValueError("a dedicated managed task path is required")
+        object.__setattr__(self, "task_path", normalized)
+
+
+@dataclass(frozen=True, slots=True)
+class DesiredScheduledTask:
+    workflow_id: str
+    task_path: str
+    task_name: str
+    xml: str
+    fingerprint: str
+
+    @property
+    def full_name(self) -> str:
+        return f"{self.task_path}{self.task_name}"
+
+
+@dataclass(frozen=True, slots=True)
+class SchedulePlan:
+    timezone: str
+    config: SchedulerConfig
+    tasks: tuple[DesiredScheduledTask, ...]
+
+    def as_dict(self) -> dict[str, object]:
+        return {
+            "schema_version": 1,
+            "timezone": self.timezone,
+            "task_path": self.config.task_path,
+            "python_executable": str(self.config.python_executable),
+            "project_root": str(self.config.project_root),
+            "apply_required": True,
+            "applied": False,
+            "tasks": [
+                {
+                    "workflow_id": task.workflow_id,
+                    "full_name": task.full_name,
+                    "xml_file": f"tasks/{task.task_name}.xml",
+                    "fingerprint": task.fingerprint,
+                }
+                for task in self.tasks
+            ],
+        }
+
+
+@dataclass(frozen=True, slots=True)
+class CurrentScheduledTask:
+    full_name: str
+    xml: str
+
+
+class PowerShellCurrentTaskProvider:
+    """Read exported definitions from one managed Task Scheduler folder."""
+
+    _SCRIPT = r"""
+$ErrorActionPreference = 'Stop'
+$OutputEncoding = [Console]::OutputEncoding = [Text.UTF8Encoding]::new()
+$taskPath = $env:GYXX_TASK_PATH
+$rows = foreach ($task in Get-ScheduledTask -TaskPath $taskPath -ErrorAction SilentlyContinue) {
+    [pscustomobject]@{
+        full_name = "$($task.TaskPath)$($task.TaskName)"
+        xml = Export-ScheduledTask -TaskPath $task.TaskPath -TaskName $task.TaskName
+    }
+}
+@($rows) | ConvertTo-Json -Depth 4 -Compress
+"""
+
+    def __init__(self, executable: str = "powershell.exe") -> None:
+        self.executable = executable
+
+    def current_tasks(self, task_path: str) -> list[CurrentScheduledTask]:
+        completed = subprocess.run(
+            [
+                self.executable,
+                "-NoLogo",
+                "-NoProfile",
+                "-NonInteractive",
+                "-Command",
+                self._SCRIPT,
+            ],
+            check=True,
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="strict",
+            shell=False,
+            env={**os.environ, "GYXX_TASK_PATH": task_path},
+        )
+        payload = json.loads(completed.stdout or "[]")
+        rows = payload if isinstance(payload, list) else [payload]
+        tasks: list[CurrentScheduledTask] = []
+        for row in rows:
+            if (
+                not isinstance(row, dict)
+                or not isinstance(row.get("full_name"), str)
+                or not isinstance(row.get("xml"), str)
+            ):
+                raise ValueError("invalid scheduled-task export response")
+            tasks.append(CurrentScheduledTask(row["full_name"], row["xml"]))
+        return tasks
+
+
+DriftStatus = Literal["changed", "missing", "extra"]
+
+
+@dataclass(frozen=True, slots=True)
+class ScheduleDriftItem:
+    full_name: str
+    status: DriftStatus
+    expected_fingerprint: str | None = None
+    current_fingerprint: str | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class ScheduleDriftReport:
+    items: tuple[ScheduleDriftItem, ...]
+
+    @property
+    def is_clean(self) -> bool:
+        return not self.items
+
+    def as_dict(self) -> dict[str, object]:
+        return {"is_clean": self.is_clean, "items": [asdict(item) for item in self.items]}
+
+
+def build_schedule_plan(
+    catalog: WorkflowCatalog,
+    config: SchedulerConfig,
+    *,
+    start_date: date,
+) -> SchedulePlan:
+    tasks: list[DesiredScheduledTask] = []
+    for workflow in catalog.scheduled_workflows():
+        schedule = catalog.schedule_for(workflow.workflow_id)
+        xml = _render_task_xml(workflow.workflow_id, schedule, config, start_date)
+        tasks.append(
+            DesiredScheduledTask(
+                workflow_id=workflow.workflow_id,
+                task_path=config.task_path,
+                task_name=workflow.workflow_id,
+                xml=xml,
+                fingerprint=_xml_fingerprint(xml),
+            )
+        )
+    return SchedulePlan(catalog.timezone, config, tuple(tasks))
+
+
+def _render_task_xml(
+    workflow_id: str,
+    schedule: ScheduleEntry,
+    config: SchedulerConfig,
+    start_date: date,
+) -> str:
+    root = ET.Element(_tag("Task"), {"version": "1.4"})
+    registration = ET.SubElement(root, _tag("RegistrationInfo"))
+    ET.SubElement(registration, _tag("Description")).text = (
+        f"Managed by gyxx-flow; workflow={workflow_id}"
+    )
+
+    triggers = ET.SubElement(root, _tag("Triggers"))
+    trigger = ET.SubElement(triggers, _tag("CalendarTrigger"))
+    trigger_date = _trigger_date(schedule, start_date)
+    trigger_time = time.fromisoformat(schedule.at)
+    ET.SubElement(trigger, _tag("StartBoundary")).text = datetime.combine(
+        trigger_date, trigger_time
+    ).isoformat(timespec="seconds")
+    ET.SubElement(trigger, _tag("Enabled")).text = "true"
+    _append_schedule(trigger, schedule)
+
+    principals = ET.SubElement(root, _tag("Principals"))
+    principal = ET.SubElement(principals, _tag("Principal"), {"id": "Author"})
+    ET.SubElement(principal, _tag("UserId")).text = "__GYXX_TASK_PRINCIPAL__"
+    ET.SubElement(principal, _tag("LogonType")).text = "InteractiveToken"
+    ET.SubElement(principal, _tag("RunLevel")).text = "LeastPrivilege"
+
+    settings = ET.SubElement(root, _tag("Settings"))
+    ET.SubElement(settings, _tag("MultipleInstancesPolicy")).text = "IgnoreNew"
+    ET.SubElement(settings, _tag("StartWhenAvailable")).text = "true"
+    ET.SubElement(settings, _tag("DisallowStartIfOnBatteries")).text = "false"
+    ET.SubElement(settings, _tag("StopIfGoingOnBatteries")).text = "false"
+    ET.SubElement(settings, _tag("ExecutionTimeLimit")).text = "PT4H"
+    ET.SubElement(settings, _tag("Enabled")).text = "true"
+
+    actions = ET.SubElement(root, _tag("Actions"), {"Context": "Author"})
+    execute = ET.SubElement(actions, _tag("Exec"))
+    ET.SubElement(execute, _tag("Command")).text = str(config.python_executable)
+    ET.SubElement(execute, _tag("Arguments")).text = (
+        f"-m gyxx_flow run {workflow_id} --scheduled"
+    )
+    ET.SubElement(execute, _tag("WorkingDirectory")).text = str(config.project_root)
+    return ET.tostring(root, encoding="unicode", xml_declaration=False)
+
+
+def _trigger_date(schedule: ScheduleEntry, start_date: date) -> date:
+    if schedule.kind == "interval_days":
+        if not schedule.anchor_date:
+            raise ValueError(f"interval schedule requires anchor_date: {schedule.workflow_id}")
+        return date.fromisoformat(schedule.anchor_date)
+    return start_date
+
+
+def _append_schedule(trigger: ET.Element, schedule: ScheduleEntry) -> None:
+    if schedule.kind in {"daily", "interval_days"}:
+        node = ET.SubElement(trigger, _tag("ScheduleByDay"))
+        interval = schedule.every_days if schedule.kind == "interval_days" else 1
+        ET.SubElement(node, _tag("DaysInterval")).text = str(interval)
+        return
+    if schedule.kind == "weekly":
+        node = ET.SubElement(trigger, _tag("ScheduleByWeek"))
+        ET.SubElement(node, _tag("WeeksInterval")).text = "1"
+        days = ET.SubElement(node, _tag("DaysOfWeek"))
+        for day_name in schedule.days:
+            ET.SubElement(days, _tag(day_name))
+        return
+    if schedule.kind == "monthly":
+        node = ET.SubElement(trigger, _tag("ScheduleByMonth"))
+        days = ET.SubElement(node, _tag("DaysOfMonth"))
+        ET.SubElement(days, _tag("Day")).text = str(schedule.day_of_month)
+        months = ET.SubElement(node, _tag("Months"))
+        for month in (
+            "January", "February", "March", "April", "May", "June",
+            "July", "August", "September", "October", "November", "December",
+        ):
+            ET.SubElement(months, _tag(month))
+        return
+    raise ValueError(f"unsupported schedule kind: {schedule.kind}")
+
+
+def write_schedule_plan_bundle(destination: Path, plan: SchedulePlan) -> Path:
+    destination = Path(destination).resolve()
+    destination.mkdir(parents=True, exist_ok=False)
+    tasks_directory = destination / "tasks"
+    tasks_directory.mkdir()
+    for task in plan.tasks:
+        _atomic_write_text(tasks_directory / f"{task.task_name}.xml", task.xml + "\n")
+    atomic_write_json(destination / "plan.json", plan.as_dict())
+    _atomic_write_text(destination / "install.ps1", _install_script(plan.config.task_path))
+    return destination
+
+
+def _install_script(task_path: str) -> str:
+    escaped_path = task_path.replace("'", "''")
+    return f"""param([switch]$Apply, [switch]$WhatIf, [string]$WorkflowId)
+$ErrorActionPreference = 'Stop'
+if ((Get-TimeZone).Id -ne 'China Standard Time') {{
+    throw 'GYXX schedules require Windows timezone China Standard Time.'
+}}
+if (-not $Apply) {{
+    Write-Host 'Plan only: no scheduled task was changed. Use -Apply -WorkflowId <id> after approval.'
+    Get-ChildItem -LiteralPath (Join-Path $PSScriptRoot 'tasks') -Filter '*.xml'
+    exit 0
+}}
+if ([string]::IsNullOrWhiteSpace($WorkflowId) -or $WorkflowId -notmatch '^[A-Za-z0-9][A-Za-z0-9._-]*$') {{
+    throw 'Apply requires exactly one safe -WorkflowId.'
+}}
+$taskFile = Join-Path (Join-Path $PSScriptRoot 'tasks') ($WorkflowId + '.xml')
+if (-not (Test-Path -LiteralPath $taskFile -PathType Leaf)) {{
+    throw "Unknown workflow task definition: $WorkflowId"
+}}
+$principal = [Security.Principal.WindowsIdentity]::GetCurrent().Name
+$escapedPrincipal = [Security.SecurityElement]::Escape($principal)
+$xml = (Get-Content -Raw -LiteralPath $taskFile).Replace('__GYXX_TASK_PRINCIPAL__', $escapedPrincipal)
+Register-ScheduledTask -TaskPath '{escaped_path}' -TaskName $WorkflowId -Xml $xml -Force -WhatIf:$WhatIf
+"""
+
+
+def _atomic_write_text(path: Path, text: str) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    temporary: Path | None = None
+    try:
+        with tempfile.NamedTemporaryFile(
+            mode="w", encoding="utf-8", newline="\n", dir=path.parent,
+            prefix=f".{path.name}.", suffix=".tmp", delete=False,
+        ) as stream:
+            temporary = Path(stream.name)
+            stream.write(text)
+            stream.flush()
+            os.fsync(stream.fileno())
+        os.replace(temporary, path)
+        temporary = None
+    finally:
+        if temporary is not None:
+            temporary.unlink(missing_ok=True)
+
+
+def detect_schedule_drift(
+    plan: SchedulePlan,
+    current_tasks: Sequence[CurrentScheduledTask],
+) -> ScheduleDriftReport:
+    desired = {task.full_name.casefold(): task for task in plan.tasks}
+    managed_prefix = plan.config.task_path.casefold()
+    current = {
+        task.full_name.casefold(): task
+        for task in current_tasks
+        if task.full_name.casefold().startswith(managed_prefix)
+    }
+    items: list[ScheduleDriftItem] = []
+    for key, expected in desired.items():
+        actual = current.get(key)
+        if actual is None:
+            items.append(
+                ScheduleDriftItem(expected.full_name, "missing", expected.fingerprint, None)
+            )
+            continue
+        current_fingerprint = _xml_fingerprint(actual.xml)
+        if current_fingerprint != expected.fingerprint:
+            items.append(
+                ScheduleDriftItem(
+                    expected.full_name,
+                    "changed",
+                    expected.fingerprint,
+                    current_fingerprint,
+                )
+            )
+    for key, actual in current.items():
+        if key not in desired:
+            items.append(ScheduleDriftItem(actual.full_name, "extra"))
+    order = {"changed": 0, "missing": 1, "extra": 2}
+    items.sort(key=lambda item: (order[item.status], item.full_name.casefold()))
+    return ScheduleDriftReport(tuple(items))
+
+
+def _xml_fingerprint(xml: str) -> str:
+    try:
+        root = ET.fromstring(xml)
+    except ET.ParseError:
+        return "invalid-xml"
+    signature = {
+        "command": root.findtext(".//{*}Exec/{*}Command") or "",
+        "arguments": root.findtext(".//{*}Exec/{*}Arguments") or "",
+        "working_directory": root.findtext(".//{*}Exec/{*}WorkingDirectory") or "",
+        "triggers": [_element_signature(node) for node in root.findall(".//{*}Triggers/*")],
+        "multiple_instances": root.findtext(".//{*}MultipleInstancesPolicy") or "",
+    }
+    canonical = json.dumps(signature, ensure_ascii=False, sort_keys=True, separators=(",", ":"))
+    return hashlib.sha256(canonical.encode("utf-8")).hexdigest()
+
+
+def _element_signature(element: ET.Element) -> object:
+    return {
+        "tag": element.tag.rsplit("}", 1)[-1],
+        "text": (element.text or "").strip(),
+        "children": [_element_signature(child) for child in list(element)],
+    }
diff --git a/src/gyxx_flow/script_catalog.py b/src/gyxx_flow/script_catalog.py
new file mode 100644
index 0000000..5e22dbe
--- /dev/null
+++ b/src/gyxx_flow/script_catalog.py
@@ -0,0 +1,139 @@
+"""Discover manually runnable scripts owned by the four business modules."""
+
+from __future__ import annotations
+
+import ast
+import warnings
+from dataclasses import dataclass
+from pathlib import Path
+from types import MappingProxyType
+from typing import Iterable, Literal, Mapping
+
+
+class ScriptCatalogError(LookupError):
+    """Raised when a script identifier is invalid or unavailable."""
+
+
+@dataclass(frozen=True, slots=True)
+class ScriptEntry:
+    script_id: str
+    module: str
+    entry: str
+    kind: Literal["python", "batch", "powershell"]
+    path: Path
+
+
+class ScriptCatalog:
+    """Immutable, deterministic index of project-owned command-line scripts."""
+
+    def __init__(self, scripts: Iterable[ScriptEntry]) -> None:
+        index: dict[str, ScriptEntry] = {}
+        for script in scripts:
+            if script.script_id in index:
+                raise ScriptCatalogError(f"duplicate script id: {script.script_id}")
+            index[script.script_id] = script
+        self._scripts = MappingProxyType(dict(sorted(index.items())))
+
+    @classmethod
+    def discover(cls, roots: Mapping[str, Path | str]) -> "ScriptCatalog":
+        scripts: list[ScriptEntry] = []
+        for module, configured_root in sorted(roots.items()):
+            root = Path(configured_root).expanduser().resolve(strict=True)
+            if not root.is_dir():
+                raise ScriptCatalogError(f"runtime root is not a directory: {root}")
+            for path in sorted(root.rglob("*")):
+                if not path.is_file():
+                    continue
+                relative = path.relative_to(root)
+                if any(
+                    part
+                    in {"__pycache__", "tests", "runtime_tests", "launchers_reference"}
+                    for part in relative.parts
+                ):
+                    continue
+                suffix = path.suffix.casefold()
+                kind: Literal["python", "batch", "powershell"] | None = None
+                if suffix in {".py", ".pyw"} and _has_main_guard(path):
+                    kind = "python"
+                elif suffix in {".bat", ".cmd"}:
+                    kind = "batch"
+                elif suffix == ".ps1":
+                    kind = "powershell"
+                if kind is None:
+                    continue
+                entry = relative.as_posix()
+                scripts.append(
+                    ScriptEntry(
+                        script_id=f"{module}:{entry}",
+                        module=module,
+                        entry=entry,
+                        kind=kind,
+                        path=path.resolve(),
+                    )
+                )
+        return cls(scripts)
+
+    @classmethod
+    def discover_default(cls) -> "ScriptCatalog":
+        module_root = Path(__file__).resolve().parent / "modules"
+        modules = (
+            "content_marketing",
+            "product_commerce",
+            "shop_intelligence",
+            "supply_chain",
+        )
+        return cls.discover(
+            {module: module_root / module / "runtime" for module in modules}
+        )
+
+    @property
+    def scripts(self) -> tuple[ScriptEntry, ...]:
+        return tuple(self._scripts.values())
+
+    @property
+    def script_ids(self) -> tuple[str, ...]:
+        return tuple(self._scripts)
+
+    def get(self, script_id: str) -> ScriptEntry:
+        if not isinstance(script_id, str) or ":" not in script_id:
+            raise ScriptCatalogError(f"invalid script id: {script_id!r}")
+        try:
+            return self._scripts[script_id]
+        except KeyError as exc:
+            raise ScriptCatalogError(f"unknown script: {script_id}") from exc
+
+
+def _has_main_guard(path: Path) -> bool:
+    text: str | None = None
+    for encoding in ("utf-8-sig", "gb18030"):
+        try:
+            text = path.read_text(encoding=encoding)
+            break
+        except UnicodeDecodeError:
+            continue
+    if text is None:
+        return False
+    try:
+        with warnings.catch_warnings():
+            warnings.simplefilter("ignore", SyntaxWarning)
+            tree = ast.parse(text, filename=str(path))
+    except SyntaxError:
+        return False
+    for node in ast.walk(tree):
+        if not isinstance(node, ast.If) or not isinstance(node.test, ast.Compare):
+            continue
+        comparison = node.test
+        if (
+            isinstance(comparison.left, ast.Name)
+            and comparison.left.id == "__name__"
+            and len(comparison.ops) == 1
+            and isinstance(comparison.ops[0], ast.Eq)
+            and len(comparison.comparators) == 1
+            and isinstance(comparison.comparators[0], ast.Constant)
+            and comparison.comparators[0].value == "__main__"
+        ):
+            return True
+    return False
+
+
+__all__ = ["ScriptCatalog", "ScriptCatalogError", "ScriptEntry"]
diff --git a/src/gyxx_flow/security/__init__.py b/src/gyxx_flow/security/__init__.py
new file mode 100644
index 0000000..370c428
--- /dev/null
+++ b/src/gyxx_flow/security/__init__.py
@@ -0,0 +1,5 @@
+"""Security checks used by migration and release gates."""
+
+from gyxx_flow.security.scanner import SecretFinding, scan_repository
+
+__all__ = ["SecretFinding", "scan_repository"]
diff --git a/src/gyxx_flow/security/scanner.py b/src/gyxx_flow/security/scanner.py
new file mode 100644
index 0000000..ddc3f2e
--- /dev/null
+++ b/src/gyxx_flow/security/scanner.py
@@ -0,0 +1,174 @@
+"""Find high-confidence plaintext credentials without exposing their values."""
+
+from __future__ import annotations
+
+import os
+import re
+from dataclasses import dataclass
+from pathlib import Path
+
+DEFAULT_EXCLUDED_DIRECTORIES = frozenset({"var", ".git", ".venv", ".learnings"})
+
+_ASSIGNMENT_PATTERN = re.compile(
+    r"""
+    [\"']?
+    (?:[a-z0-9]+[_\-.])*
+    (?:
+        password | passwd | pwd |
+        api[_-]?key | app[_-]?secret | client[_-]?secret | consumer[_-]?secret |
+        secret[_-]?key |
+        secret[_-]?access[_-]?key | access[_-]?key(?:[_-]?id)? |
+        access[_-]?token | refresh[_-]?token | auth[_-]?token
+    )
+    [\"']?
+    \s*(?::|(?<![=!<>])=(?!=))\s*
+    (?:
+        \"(?P<double>[^\"\r\n]*)\" |
+        '(?P<single>[^'\r\n]*)' |
+        (?P<bare>[^\s#;,]+)
+    )
+    """,
+    re.IGNORECASE | re.VERBOSE,
+)
+_CREDENTIAL_URL_PATTERN = re.compile(
+    r"\b(?:https?|ftp)://[^\s/:@]+:[^\s/@{}$%]+@[^\s/\"']+",
+    re.IGNORECASE,
+)
+_PRIVATE_KEY_PATTERN = re.compile(
+    r"-----BEGIN (?:RSA |EC |DSA |OPENSSH )?PRIVATE KEY-----"
+)
+_NAMED_LONG_HEX_CREDENTIAL_PATTERN = re.compile(
+    r"""
+    (?:api[_-]?key|app[_-]?secret|client[_-]?secret|access[_-]?token)
+    \s*=\s*
+    (?:os\.getenv\([\s\S]{0,160}?,\s*)?
+    ["'][0-9a-f]{32,}["']
+    """,
+    re.IGNORECASE | re.VERBOSE,
+)
+
+_PLACEHOLDER_WORDS = frozenset(
+    {
+        "changeme",
+        "example",
+        "none",
+        "null",
+        "placeholder",
+        "redacted",
+        "replaceme",
+        "secret",
+        "secretstr",
+        "str",
+        "string",
+        "todo",
+        "xxx",
+        "yourkey",
+        "yourpassword",
+        "yoursecret",
+        "yourtoken",
+    }
+)
+
+
+@dataclass(frozen=True, slots=True)
+class SecretFinding:
+    """A deliberately value-free secret scan result."""
+
+    file: str
+    line: int
+    rule: str
+
+
+def scan_repository(
+    root: str | Path,
+    *,
+    excluded_directories: frozenset[str] = DEFAULT_EXCLUDED_DIRECTORIES,
+) -> list[SecretFinding]:
+    """Scan repository text files and return deterministic, value-free findings.
+
+    Directory names in ``excluded_directories`` are pruned at every depth. Binary
+    and undecodable files are ignored. The function never stores matched values in
+    its result objects.
+    """
+
+    repository = Path(root).expanduser().resolve()
+    if not repository.is_dir():
+        raise NotADirectoryError(repository)
+
+    findings: list[SecretFinding] = []
+    for directory, directory_names, file_names in os.walk(repository, followlinks=False):
+        directory_names[:] = sorted(
+            name for name in directory_names if name not in excluded_directories
+        )
+        current_directory = Path(directory)
+        for file_name in sorted(file_names):
+            path = current_directory / file_name
+            text = _read_text(path)
+            if text is None:
+                continue
+            relative_file = path.relative_to(repository).as_posix()
+            findings.extend(_scan_text(relative_file, text))
+    return findings
+
+
+def _read_text(path: Path) -> str | None:
+    try:
+        content = path.read_bytes()
+    except OSError:
+        return None
+    if b"\x00" in content:
+        return None
+    for encoding in ("utf-8-sig", "gb18030"):
+        try:
+            return content.decode(encoding)
+        except UnicodeDecodeError:
+            continue
+    return None
+
+
+def _scan_text(relative_file: str, text: str) -> list[SecretFinding]:
+    findings: list[SecretFinding] = []
+    if relative_file.casefold().endswith((".py", ".pyw")):
+        for match in _NAMED_LONG_HEX_CREDENTIAL_PATTERN.finditer(text):
+            findings.append(
+                SecretFinding(
+                    relative_file,
+                    text.count("\n", 0, match.start()) + 1,
+                    "hardcoded-long-hex-credential",
+                )
+            )
+    for line_number, line in enumerate(text.splitlines(), start=1):
+        if _PRIVATE_KEY_PATTERN.search(line):
+            findings.append(
+                SecretFinding(relative_file, line_number, "private-key-material")
+            )
+        if _CREDENTIAL_URL_PATTERN.search(line):
+            findings.append(SecretFinding(relative_file, line_number, "credential-in-url"))
+        if any(_is_plaintext_value(match) for match in _ASSIGNMENT_PATTERN.finditer(line)):
+            findings.append(
+                SecretFinding(relative_file, line_number, "plaintext-credential")
+            )
+    return findings
+
+
+def _is_plaintext_value(match: re.Match[str]) -> bool:
+    value = next(
+        (candidate for candidate in match.group("double", "single", "bare") if candidate is not None),
+        "",
+    ).strip()
+    if not value:
+        return False
+
+    lowered = value.casefold()
+    compact = re.sub(r"[^a-z0-9]", "", lowered)
+    if compact in _PLACEHOLDER_WORDS:
+        return False
+    if lowered in {"true", "false"}:
+        return False
+    if value.startswith(("$", "%(", "{{", "{")):
+        return False
+    if re.fullmatch(r"[a-z_]\w*(?:\.[a-z_]\w*)+", lowered):
+        return False
+    if any(marker in lowered for marker in ("getenv(", "environ[", "secretmanager", "keyvault")):
+        return False
+    return True
diff --git a/src/gyxx_flow/workflow/__init__.py b/src/gyxx_flow/workflow/__init__.py
new file mode 100644
index 0000000..70d258d
--- /dev/null
+++ b/src/gyxx_flow/workflow/__init__.py
@@ -0,0 +1,15 @@
+"""Workflow definitions, executable steps, and orchestration engine."""
+
+from .engine import WorkflowEngine, WorkflowRunResult
+from .model import StepDefinition, WorkflowDefinition, WorkflowValidationError
+from .steps import CommandStep, StepExecution
+
+__all__ = [
+    "CommandStep",
+    "StepDefinition",
+    "StepExecution",
+    "WorkflowDefinition",
+    "WorkflowEngine",
+    "WorkflowRunResult",
+    "WorkflowValidationError",
+]
diff --git a/src/gyxx_flow/workflow/engine.py b/src/gyxx_flow/workflow/engine.py
new file mode 100644
index 0000000..aa2af93
--- /dev/null
+++ b/src/gyxx_flow/workflow/engine.py
@@ -0,0 +1,322 @@
+"""Deterministic workflow execution with retries, locks, and policy guards."""
+
+from __future__ import annotations
+
+import time
+from contextlib import ExitStack
+from dataclasses import dataclass
+from typing import Callable, Literal
+
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.core.locks import LockManager
+from gyxx_flow.core.records import RunJournal
+from gyxx_flow.ops.effects import (
+    EffectAlreadyApplied,
+    EffectClaim,
+    EffectLedger,
+    EffectStateAmbiguous,
+)
+
+from .model import StepDefinition, WorkflowDefinition
+from .steps import StepExecution
+
+StepRunStatus = Literal["success", "failed", "skipped"]
+WorkflowRunStatus = Literal["success", "failed"]
+
+
+@dataclass(frozen=True, slots=True)
+class AttemptResult:
+    attempt: int
+    status: StepRunStatus
+    exit_code: int | None
+    error: str | None = None
+    reason: str | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class StepRunResult:
+    step_id: str
+    status: StepRunStatus
+    attempts: tuple[AttemptResult, ...]
+    final_exit_code: int | None
+    critical: bool
+    reason: str | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class WorkflowRunResult:
+    workflow_id: str
+    status: WorkflowRunStatus
+    exit_code: int
+    steps: dict[str, StepRunResult]
+    warnings: tuple[str, ...] = ()
+
+
+class WorkflowEngine:
+    """Run a validated workflow while recording every action attempt."""
+
+    def __init__(
+        self,
+        lock_manager: LockManager,
+        *,
+        sleep: Callable[[float], None] = time.sleep,
+        effect_ledger: EffectLedger | None = None,
+    ) -> None:
+        self._lock_manager = lock_manager
+        self._sleep = sleep
+        self._effect_ledger = effect_ledger
+
+    def execute(
+        self,
+        workflow: WorkflowDefinition,
+        *,
+        context: RunContext,
+        journal: RunJournal,
+        dry_run: bool = False,
+    ) -> WorkflowRunResult:
+        if context.workflow_id != workflow.workflow_id:
+            raise ValueError(
+                "run context workflow_id does not match workflow definition: "
+                f"{context.workflow_id!r} != {workflow.workflow_id!r}"
+            )
+
+        results: dict[str, StepRunResult] = {}
+        with self._lock_manager.acquire(
+            f"workflow:{workflow.workflow_id}", owner=context.run_id
+        ):
+            for step in workflow.ordered_steps():
+                failed_dependency = any(
+                    _blocks_dependents(results[dependency])
+                    for dependency in step.depends_on
+                )
+                if failed_dependency:
+                    results[step.step_id] = self._record_policy_skip(
+                        step, journal, reason="dependency-failed"
+                    )
+                    continue
+                if dry_run:
+                    results[step.step_id] = self._record_policy_skip(
+                        step, journal, reason="dry-run"
+                    )
+                    continue
+                if context.shadow and (
+                    step.production_sink or step.official_notification
+                ):
+                    results[step.step_id] = self._record_policy_skip(
+                        step, journal, reason="shadow-policy"
+                    )
+                    continue
+                claim: EffectClaim | None = None
+                if step.production_sink and self._effect_ledger is not None:
+                    try:
+                        claim = self._effect_ledger.begin(
+                            workflow_id=context.workflow_id,
+                            business_date=context.business_date.isoformat(),
+                            step_id=step.step_id,
+                            run_id=context.run_id,
+                        )
+                    except EffectAlreadyApplied:
+                        results[step.step_id] = self._record_policy_skip(
+                            step, journal, reason="idempotency-replay"
+                        )
+                        continue
+                    except EffectStateAmbiguous:
+                        results[step.step_id] = self._record_policy_failure(
+                            step, journal, reason="effect-state-ambiguous"
+                        )
+                        continue
+                result = self._execute_step(step, context=context, journal=journal)
+                if claim is not None:
+                    if result.status == "success":
+                        self._effect_ledger.mark_applied(claim)
+                    else:
+                        self._effect_ledger.mark_ambiguous(
+                            claim, reason="step-did-not-succeed"
+                        )
+                results[step.step_id] = result
+
+        critical_failures = tuple(
+            step_id
+            for step_id, result in results.items()
+            if result.critical and _is_failure(result)
+        )
+        warnings = tuple(
+            step_id
+            for step_id, result in results.items()
+            if not result.critical and _is_failure(result)
+        )
+        status: WorkflowRunStatus = "failed" if critical_failures else "success"
+        error = None
+        if critical_failures:
+            error = f"critical steps failed: {', '.join(critical_failures)}"
+        journal.finalize(status, error=error)
+        return WorkflowRunResult(
+            workflow_id=workflow.workflow_id,
+            status=status,
+            exit_code=1 if critical_failures else 0,
+            steps=results,
+            warnings=warnings,
+        )
+
+    def _execute_step(
+        self,
+        step: StepDefinition,
+        *,
+        context: RunContext,
+        journal: RunJournal,
+    ) -> StepRunResult:
+        attempts: list[AttemptResult] = []
+        journal.record_input(f"step:{step.step_id}")
+        for attempt in range(1, step.max_attempts + 1):
+            journal_id = _attempt_journal_id(step.step_id, attempt)
+            journal.start_step(journal_id, attempt=attempt)
+            try:
+                with ExitStack() as stack:
+                    for resource in sorted(step.resources):
+                        stack.enter_context(
+                            self._lock_manager.acquire(resource, owner=context.run_id)
+                        )
+                    outcome = step.action.execute(
+                        context=context,
+                        timeout_seconds=step.timeout_seconds,
+                        dry_run=False,
+                    )
+                    if not isinstance(outcome, StepExecution):
+                        raise TypeError("step action must return StepExecution")
+            except Exception as exc:  # action and resource failures are retryable
+                outcome = StepExecution(
+                    exit_code=1,
+                    error=f"{type(exc).__name__}: {exc}",
+                )
+
+            status = _execution_status(outcome)
+            journal.finish_step(
+                journal_id,
+                status=status,
+                exit_code=outcome.exit_code,
+                error=outcome.error or outcome.reason,
+            )
+            exit_code = "none" if outcome.exit_code is None else str(outcome.exit_code)
+            journal.record_output(
+                f"step:{step.step_id}:attempt:{attempt}:{status}:exit:{exit_code}"
+            )
+            attempts.append(
+                AttemptResult(
+                    attempt=attempt,
+                    status=status,
+                    exit_code=outcome.exit_code,
+                    error=outcome.error,
+                    reason=outcome.reason,
+                )
+            )
+            if status in {"success", "skipped"}:
+                break
+            if attempt < step.max_attempts and step.retry_delay_seconds:
+                self._sleep(step.retry_delay_seconds)
+
+        final = attempts[-1]
+        if final.status == "success":
+            if step.production_sink:
+                journal.record_external_write(
+                    f"step:{step.step_id}:production_sink"
+                )
+            if step.official_notification:
+                journal.record_external_write(
+                    f"step:{step.step_id}:official_notification"
+                )
+        return StepRunResult(
+            step_id=step.step_id,
+            status=final.status,
+            attempts=tuple(attempts),
+            final_exit_code=final.exit_code,
+            critical=step.critical,
+            reason=final.reason,
+        )
+
+    def _record_policy_skip(
+        self,
+        step: StepDefinition,
+        journal: RunJournal,
+        *,
+        reason: str,
+    ) -> StepRunResult:
+        journal.record_input(f"step:{step.step_id}")
+        journal_id = _attempt_journal_id(step.step_id, 1)
+        journal.start_step(journal_id, attempt=1)
+        journal.finish_step(
+            journal_id,
+            status="skipped",
+            exit_code=0,
+            error=reason,
+        )
+        journal.record_output(
+            f"step:{step.step_id}:attempt:1:skipped:exit:0"
+        )
+        attempt = AttemptResult(
+            attempt=1,
+            status="skipped",
+            exit_code=0,
+            reason=reason,
+        )
+        return StepRunResult(
+            step_id=step.step_id,
+            status="skipped",
+            attempts=(attempt,),
+            final_exit_code=0,
+            critical=step.critical,
+            reason=reason,
+        )
+
+    def _record_policy_failure(
+        self,
+        step: StepDefinition,
+        journal: RunJournal,
+        *,
+        reason: str,
+    ) -> StepRunResult:
+        journal.record_input(f"step:{step.step_id}")
+        journal_id = _attempt_journal_id(step.step_id, 1)
+        journal.start_step(journal_id, attempt=1)
+        journal.finish_step(
+            journal_id,
+            status="failed",
+            exit_code=75,
+            error=reason,
+        )
+        journal.record_output(
+            f"step:{step.step_id}:attempt:1:failed:exit:75"
+        )
+        attempt = AttemptResult(
+            attempt=1,
+            status="failed",
+            exit_code=75,
+            error=reason,
+        )
+        return StepRunResult(
+            step_id=step.step_id,
+            status="failed",
+            attempts=(attempt,),
+            final_exit_code=75,
+            critical=step.critical,
+            reason=reason,
+        )
+
+
+def _attempt_journal_id(step_id: str, attempt: int) -> str:
+    return f"{step_id}.attempt-{attempt}"
+
+
+def _execution_status(outcome: StepExecution) -> StepRunStatus:
+    if outcome.skipped:
+        return "skipped"
+    if outcome.exit_code == 0 and outcome.error is None:
+        return "success"
+    return "failed"
+
+
+def _blocks_dependents(result: StepRunResult) -> bool:
+    return result.status == "failed" or result.reason == "dependency-failed"
+
+
+def _is_failure(result: StepRunResult) -> bool:
+    return result.status == "failed" or result.reason == "dependency-failed"
diff --git a/src/gyxx_flow/workflow/model.py b/src/gyxx_flow/workflow/model.py
new file mode 100644
index 0000000..8c9e59d
--- /dev/null
+++ b/src/gyxx_flow/workflow/model.py
@@ -0,0 +1,153 @@
+"""Validated workflow and step definitions."""
+
+from __future__ import annotations
+
+import heapq
+import re
+from dataclasses import dataclass
+from typing import TYPE_CHECKING, Protocol
+
+if TYPE_CHECKING:
+    from gyxx_flow.core.context import RunContext
+
+    from .steps import StepExecution
+
+
+_ID = re.compile(r"^[a-z][a-z0-9]*(?:[._-][a-z0-9]+)*$")
+_RESOURCE = re.compile(r"^[A-Za-z0-9][A-Za-z0-9:._-]*$")
+
+
+class WorkflowValidationError(ValueError):
+    """Raised when a workflow definition is unsafe or inconsistent."""
+
+
+class ExecutableStep(Protocol):
+    """Action contract consumed by the workflow engine."""
+
+    def execute(
+        self,
+        *,
+        context: RunContext,
+        timeout_seconds: float | None,
+        dry_run: bool,
+    ) -> StepExecution: ...
+
+
+@dataclass(frozen=True, slots=True)
+class StepDefinition:
+    """Execution policy and dependency metadata for one workflow step."""
+
+    step_id: str
+    action: ExecutableStep
+    depends_on: tuple[str, ...] = ()
+    timeout_seconds: float | None = None
+    max_attempts: int = 1
+    retry_delay_seconds: float = 0.0
+    critical: bool = True
+    resources: tuple[str, ...] = ()
+    production_sink: bool = False
+    official_notification: bool = False
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "depends_on", tuple(self.depends_on))
+        object.__setattr__(self, "resources", tuple(self.resources))
+        if not isinstance(self.step_id, str) or not _ID.fullmatch(self.step_id):
+            raise WorkflowValidationError(f"invalid step id: {self.step_id!r}")
+        if not callable(getattr(self.action, "execute", None)):
+            raise WorkflowValidationError(f"step {self.step_id!r} has no executable action")
+        if len(set(self.depends_on)) != len(self.depends_on):
+            raise WorkflowValidationError(f"duplicate dependency for step {self.step_id!r}")
+        for dependency in self.depends_on:
+            if not isinstance(dependency, str) or not _ID.fullmatch(dependency):
+                raise WorkflowValidationError(
+                    f"invalid dependency for step {self.step_id!r}: {dependency!r}"
+                )
+            if dependency == self.step_id:
+                raise WorkflowValidationError(f"step {self.step_id!r} depends on itself")
+        if self.timeout_seconds is not None and (
+            isinstance(self.timeout_seconds, bool) or self.timeout_seconds <= 0
+        ):
+            raise WorkflowValidationError(
+                f"timeout_seconds must be positive for step {self.step_id!r}"
+            )
+        if (
+            isinstance(self.max_attempts, bool)
+            or not isinstance(self.max_attempts, int)
+            or self.max_attempts < 1
+        ):
+            raise WorkflowValidationError(
+                f"max_attempts must be a positive integer for step {self.step_id!r}"
+            )
+        if isinstance(self.retry_delay_seconds, bool) or self.retry_delay_seconds < 0:
+            raise WorkflowValidationError(
+                f"retry_delay_seconds must be non-negative for step {self.step_id!r}"
+            )
+        if len(set(self.resources)) != len(self.resources):
+            raise WorkflowValidationError(f"duplicate resource for step {self.step_id!r}")
+        for resource in self.resources:
+            if not isinstance(resource, str) or not _RESOURCE.fullmatch(resource):
+                raise WorkflowValidationError(
+                    f"invalid resource for step {self.step_id!r}: {resource!r}"
+                )
+
+
+@dataclass(frozen=True, slots=True)
+class WorkflowDefinition:
+    """A validated DAG with deterministic execution ordering."""
+
+    workflow_id: str
+    steps: tuple[StepDefinition, ...]
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "steps", tuple(self.steps))
+        if not isinstance(self.workflow_id, str) or not _ID.fullmatch(self.workflow_id):
+            raise WorkflowValidationError(f"invalid workflow id: {self.workflow_id!r}")
+        ids = [step.step_id for step in self.steps]
+        duplicate = _first_duplicate(ids)
+        if duplicate is not None:
+            raise WorkflowValidationError(f"duplicate step id: {duplicate}")
+        known = set(ids)
+        for step in self.steps:
+            for dependency in step.depends_on:
+                if dependency not in known:
+                    raise WorkflowValidationError(
+                        f"unknown dependency {dependency!r} for step {step.step_id!r}"
+                    )
+        self.ordered_steps()
+
+    def ordered_steps(self) -> tuple[StepDefinition, ...]:
+        """Return a stable topological order or reject dependency cycles."""
+
+        positions = {step.step_id: index for index, step in enumerate(self.steps)}
+        indegrees = {step.step_id: len(step.depends_on) for step in self.steps}
+        dependents: dict[str, list[str]] = {step.step_id: [] for step in self.steps}
+        for step in self.steps:
+            for dependency in step.depends_on:
+                dependents[dependency].append(step.step_id)
+
+        ready = [positions[step.step_id] for step in self.steps if not step.depends_on]
+        heapq.heapify(ready)
+        ordered: list[StepDefinition] = []
+        while ready:
+            position = heapq.heappop(ready)
+            step = self.steps[position]
+            ordered.append(step)
+            for dependent in dependents[step.step_id]:
+                indegrees[dependent] -= 1
+                if indegrees[dependent] == 0:
+                    heapq.heappush(ready, positions[dependent])
+        if len(ordered) != len(self.steps):
+            cyclic = sorted(step_id for step_id, degree in indegrees.items() if degree)
+            raise WorkflowValidationError(
+                f"dependency cycle detected: {', '.join(cyclic)}"
+            )
+        return tuple(ordered)
+
+
+def _first_duplicate(values: list[str]) -> str | None:
+    seen: set[str] = set()
+    for value in values:
+        if value in seen:
+            return value
+        seen.add(value)
+    return None
diff --git a/src/gyxx_flow/workflow/registry.py b/src/gyxx_flow/workflow/registry.py
new file mode 100644
index 0000000..20a9242
--- /dev/null
+++ b/src/gyxx_flow/workflow/registry.py
@@ -0,0 +1,77 @@
+"""Safe registration boundary between catalog metadata and executable workflows."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+
+from gyxx_flow.catalog import WorkflowCatalog, WorkflowEntry
+
+from .model import WorkflowDefinition
+
+
+class WorkflowRegistryError(LookupError):
+    """Base class for workflow resolution errors safe to show to operators."""
+
+
+class UnknownWorkflowError(WorkflowRegistryError):
+    """Raised when a workflow ID is absent from the validated catalog."""
+
+
+class UnavailableWorkflowError(WorkflowRegistryError):
+    """Raised when the catalog explicitly marks a workflow unavailable."""
+
+
+class UnregisteredWorkflowError(WorkflowRegistryError):
+    """Raised when migration has not registered executable steps yet."""
+
+
+@dataclass(frozen=True, slots=True)
+class RegisteredWorkflow:
+    """Catalog metadata plus its executable definition."""
+
+    catalog_entry: WorkflowEntry
+    definition: WorkflowDefinition
+
+
+class WorkflowRegistry:
+    """Validated, explicit registry; catalog entries are never executable by accident."""
+
+    def __init__(self, catalog: WorkflowCatalog) -> None:
+        self.catalog = catalog
+        self._entries = {entry.workflow_id: entry for entry in catalog.workflows}
+        self._definitions: dict[str, WorkflowDefinition] = {}
+
+    def register(self, definition: WorkflowDefinition) -> None:
+        """Register one migrated workflow after validating its catalog state."""
+
+        entry = self._entries.get(definition.workflow_id)
+        if entry is None:
+            raise UnknownWorkflowError(
+                f"workflow is absent from catalog: {definition.workflow_id}"
+            )
+        if entry.trigger == "unavailable":
+            raise UnavailableWorkflowError(
+                f"workflow is marked unavailable: {definition.workflow_id}"
+            )
+        if definition.workflow_id in self._definitions:
+            raise ValueError(f"workflow is already registered: {definition.workflow_id}")
+        self._definitions[definition.workflow_id] = definition
+
+    def resolve(self, workflow_id: str) -> RegisteredWorkflow:
+        """Resolve an executable workflow without falling back to legacy commands."""
+
+        entry = self._entries.get(workflow_id)
+        if entry is None:
+            raise UnknownWorkflowError(f"unknown workflow: {workflow_id}")
+        if entry.trigger == "unavailable":
+            raise UnavailableWorkflowError(f"workflow is unavailable: {workflow_id}")
+        definition = self._definitions.get(workflow_id)
+        if definition is None:
+            raise UnregisteredWorkflowError(
+                f"workflow is not registered for execution: {workflow_id}"
+            )
+        return RegisteredWorkflow(entry, definition)
+
+    def is_registered(self, workflow_id: str) -> bool:
+        return workflow_id in self._definitions
+
diff --git a/src/gyxx_flow/workflow/selection.py b/src/gyxx_flow/workflow/selection.py
new file mode 100644
index 0000000..be33306
--- /dev/null
+++ b/src/gyxx_flow/workflow/selection.py
@@ -0,0 +1,50 @@
+"""Pure workflow slicing helpers used for operator-directed recovery."""
+
+from __future__ import annotations
+
+from dataclasses import replace
+
+from .model import StepDefinition, WorkflowDefinition
+
+
+def rerun_step(workflow: WorkflowDefinition, step_id: str) -> WorkflowDefinition:
+    """Build a one-step rerun, treating prior dependencies as already satisfied."""
+
+    step = _step_by_id(workflow, step_id)
+    return WorkflowDefinition(workflow.workflow_id, (_without_external_dependencies(step, {step_id}),))
+
+
+def resume_from(workflow: WorkflowDefinition, step_id: str) -> WorkflowDefinition:
+    """Build a recovery slice containing a step and all of its descendants."""
+
+    _step_by_id(workflow, step_id)
+    included = {step_id}
+    changed = True
+    while changed:
+        changed = False
+        for step in workflow.steps:
+            if step.step_id not in included and set(step.depends_on) & included:
+                included.add(step.step_id)
+                changed = True
+    steps = tuple(
+        _without_external_dependencies(step, included)
+        for step in workflow.steps
+        if step.step_id in included
+    )
+    return WorkflowDefinition(workflow.workflow_id, steps)
+
+
+def _step_by_id(workflow: WorkflowDefinition, step_id: str) -> StepDefinition:
+    for step in workflow.steps:
+        if step.step_id == step_id:
+            return step
+    raise ValueError(f"unknown step for {workflow.workflow_id}: {step_id}")
+
+
+def _without_external_dependencies(
+    step: StepDefinition, included: set[str]
+) -> StepDefinition:
+    dependencies = tuple(
+        dependency for dependency in step.depends_on if dependency in included
+    )
+    return replace(step, depends_on=dependencies)
diff --git a/src/gyxx_flow/workflow/steps.py b/src/gyxx_flow/workflow/steps.py
new file mode 100644
index 0000000..95d9597
--- /dev/null
+++ b/src/gyxx_flow/workflow/steps.py
@@ -0,0 +1,101 @@
+"""Built-in executable workflow steps."""
+
+from __future__ import annotations
+
+import subprocess
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Mapping
+
+from gyxx_flow.core.context import RunContext
+
+
+@dataclass(frozen=True, slots=True)
+class StepExecution:
+    """One action invocation outcome before retry aggregation."""
+
+    exit_code: int
+    stdout: str = ""
+    stderr: str = ""
+    error: str | None = None
+    skipped: bool = False
+    reason: str | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class CommandStep:
+    """Execute an argv command without shell parsing or implicit process context."""
+
+    argv: tuple[str, ...]
+    cwd: Path
+    env: Mapping[str, str]
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "argv", tuple(self.argv))
+        object.__setattr__(self, "cwd", Path(self.cwd))
+        object.__setattr__(self, "env", dict(self.env))
+        if not self.argv or not all(
+            isinstance(argument, str) and argument and "\x00" not in argument
+            for argument in self.argv
+        ):
+            raise ValueError("argv must contain non-empty strings")
+        if not all(
+            isinstance(key, str)
+            and key
+            and "\x00" not in key
+            and isinstance(value, str)
+            and "\x00" not in value
+            for key, value in self.env.items()
+        ):
+            raise ValueError("env must contain valid string keys and values")
+
+    def execute(
+        self,
+        *,
+        context: RunContext,
+        timeout_seconds: float | None,
+        dry_run: bool,
+    ) -> StepExecution:
+        del context
+        if dry_run:
+            return StepExecution(exit_code=0, skipped=True, reason="dry-run")
+        try:
+            completed = subprocess.run(
+                list(self.argv),
+                cwd=self.cwd,
+                env=dict(self.env),
+                shell=False,
+                capture_output=True,
+                text=True,
+                timeout=timeout_seconds,
+                check=False,
+            )
+        except subprocess.TimeoutExpired as exc:
+            stdout = _to_text(exc.stdout)
+            stderr = _to_text(exc.stderr)
+            return StepExecution(
+                exit_code=124,
+                stdout=stdout,
+                stderr=stderr,
+                error=f"command timed out after {exc.timeout} seconds",
+            )
+        except OSError as exc:
+            return StepExecution(exit_code=127, error=f"cannot start command: {exc}")
+
+        error = None
+        if completed.returncode != 0:
+            error = completed.stderr.strip() or f"command exited with code {completed.returncode}"
+        return StepExecution(
+            exit_code=completed.returncode,
+            stdout=completed.stdout,
+            stderr=completed.stderr,
+            error=error,
+        )
+
+
+def _to_text(value: str | bytes | None) -> str:
+    if value is None:
+        return ""
+    if isinstance(value, bytes):
+        return value.decode(errors="replace")
+    return value
diff --git a/tests/test_baseline.py b/tests/test_baseline.py
new file mode 100644
index 0000000..9cf4ab3
--- /dev/null
+++ b/tests/test_baseline.py
@@ -0,0 +1,276 @@
+from __future__ import annotations
+
+import hashlib
+import json
+import subprocess
+from datetime import datetime, timezone
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.baseline import (
+    DataRootSpec,
+    PowerShellScheduledTaskProvider,
+    ProjectSpec,
+    ScheduledTask,
+    TaskCountMismatch,
+    build_baseline_manifest,
+    collect_code_inventory,
+    collect_data_summary,
+    collect_scheduled_task_inventory,
+    project_specs_from_env,
+    write_manifest_atomic,
+)
+
+
+class FakeTaskProvider:
+    def __init__(self, tasks: list[ScheduledTask]) -> None:
+        self._tasks = tasks
+
+    def scheduled_tasks(self) -> list[ScheduledTask]:
+        return self._tasks
+
+
+def test_powershell_provider_parses_tasks_without_using_a_shell(monkeypatch) -> None:
+    captured: dict[str, object] = {}
+
+    def fake_run(command, **kwargs):
+        captured["command"] = command
+        captured["kwargs"] = kwargs
+        return subprocess.CompletedProcess(
+            command,
+            0,
+            stdout='[{"task_id":"\\\\Daily","command":"python.exe",'
+            '"arguments":"D:\\\\legacy\\\\run.py","working_directory":"D:\\\\legacy"}]',
+            stderr="",
+        )
+
+    monkeypatch.setattr(subprocess, "run", fake_run)
+
+    tasks = PowerShellScheduledTaskProvider().scheduled_tasks()
+
+    assert tasks == [
+        ScheduledTask(
+            task_id=r"\Daily",
+            command="python.exe",
+            arguments=r"D:\legacy\run.py",
+            working_directory=r"D:\legacy",
+        )
+    ]
+    assert captured["command"][0] == "powershell.exe"
+    assert captured["kwargs"]["shell"] is False
+    assert captured["kwargs"]["check"] is True
+
+
+def test_project_specs_are_built_only_from_injected_legacy_roots(tmp_path: Path) -> None:
+    roots = {
+        "GYXX_LEGACY_CONTENT_ROOT": tmp_path / "content",
+        "GYXX_LEGACY_SHOP_ROOT": tmp_path / "shop",
+        "GYXX_LEGACY_PRODUCT_ROOT": tmp_path / "product",
+        "GYXX_LEGACY_SUPPLY_ROOT": tmp_path / "supply",
+    }
+    specs = project_specs_from_env({key: str(value) for key, value in roots.items()})
+
+    assert [spec.project_id for spec in specs] == [
+        "content_marketing",
+        "shop_intelligence",
+        "product_commerce",
+        "supply_chain",
+    ]
+    assert [spec.root for spec in specs] == list(roots.values())
+    assert [item.label for item in specs[0].data_roots] == ["data", "reports"]
+    assert specs[3].data_roots[0].path == roots["GYXX_LEGACY_SUPPLY_ROOT"] / "shared-data"
+
+
+def test_project_specs_report_all_missing_root_variables() -> None:
+    with pytest.raises(ValueError, match="GYXX_LEGACY_SHOP_ROOT.*GYXX_LEGACY_PRODUCT_ROOT"):
+        project_specs_from_env({"GYXX_LEGACY_CONTENT_ROOT": "configured"})
+
+
+def test_code_inventory_hashes_allowed_files_and_excludes_runtime_and_secrets(
+    tmp_path: Path,
+) -> None:
+    root = tmp_path / "legacy"
+    (root / "src").mkdir(parents=True)
+    (root / "src" / "worker.py").write_text("print('safe')\n", encoding="utf-8")
+    (root / "README.md").write_text("docs\n", encoding="utf-8")
+    (root / "payload.csv").write_text("private,data\n", encoding="utf-8")
+
+    excluded_files = [
+        root / ".git" / "tracked.py",
+        root / ".venv" / "site.py",
+        root / "var" / "runtime.py",
+        root / "data" / "raw.py",
+        root / "logs" / "run.py",
+        root / ".chrome_profile" / "state.py",
+        root / "cookies" / "session.py",
+        root / "config" / ".env",
+        root / "config" / "credentials.json",
+    ]
+    for path in excluded_files:
+        path.parent.mkdir(parents=True, exist_ok=True)
+        path.write_text("TOP_SECRET=must-not-leak", encoding="utf-8")
+
+    inventory = collect_code_inventory(root)
+
+    assert [item["relative_path"] for item in inventory] == [
+        "README.md",
+        "src/worker.py",
+    ]
+    worker = inventory[1]
+    worker_bytes = (root / "src" / "worker.py").read_bytes()
+    assert set(worker) == {"relative_path", "size_bytes", "mtime_ns", "sha256"}
+    assert worker["size_bytes"] == len(worker_bytes)
+    assert worker["sha256"] == hashlib.sha256(worker_bytes).hexdigest()
+    serialized = json.dumps(inventory)
+    assert "TOP_SECRET" not in serialized
+    assert "private,data" not in serialized
+
+
+def test_data_summary_contains_only_aggregate_counts_bytes_extensions_and_top_level(
+    tmp_path: Path,
+) -> None:
+    root = tmp_path / "data"
+    root.mkdir()
+    (root / "root.csv").write_bytes(b"123")
+    (root / "orders").mkdir()
+    (root / "orders" / "one.json").write_bytes(b"12")
+    (root / "orders" / "archive").mkdir()
+    (root / "orders" / "archive" / "two.JSON").write_bytes(b"1234")
+
+    summary = collect_data_summary(root)
+
+    assert summary == {
+        "file_count": 3,
+        "total_bytes": 9,
+        "by_extension": {
+            ".csv": {"file_count": 1, "total_bytes": 3},
+            ".json": {"file_count": 2, "total_bytes": 6},
+        },
+        "by_top_level": {
+            ".": {"file_count": 1, "total_bytes": 3},
+            "orders": {"file_count": 2, "total_bytes": 6},
+        },
+    }
+    serialized = json.dumps(summary)
+    assert "root.csv" not in serialized
+    assert "one.json" not in serialized
+    assert "archive" not in serialized
+
+
+def _task_specs(tmp_path: Path) -> tuple[ProjectSpec, ...]:
+    roots = [tmp_path / name for name in ("one", "two", "three", "four")]
+    for root in roots:
+        root.mkdir()
+    return tuple(
+        ProjectSpec(project_id=f"project_{index}", root=root)
+        for index, root in enumerate(roots, start=1)
+    )
+
+
+def test_task_inventory_filters_by_legacy_roots_and_exposes_no_command_values(
+    tmp_path: Path,
+) -> None:
+    specs = _task_specs(tmp_path)
+    tasks = [
+        ScheduledTask(
+            task_id=f"legacy-{index:02d}",
+            command="python.exe",
+            arguments=f'"{specs[index % 4].root / "run.py"}" --token SECRET-{index}',
+        )
+        for index in range(21)
+    ]
+    tasks.append(
+        ScheduledTask(
+            task_id="unrelated",
+            command="python.exe",
+            arguments=r"C:\unrelated\run.py --token OUTSIDE_SECRET",
+        )
+    )
+    tasks.append(
+        ScheduledTask(
+            task_id="lookalike-root",
+            command="python.exe",
+            arguments=f'"{specs[0].root}-backup\\run.py"',
+        )
+    )
+
+    result = collect_scheduled_task_inventory(
+        specs, FakeTaskProvider(tasks), expected_count=21
+    )
+
+    assert result["expected_count"] == 21
+    assert result["actual_count"] == 21
+    assert result["is_complete"] is True
+    assert len(result["tasks"]) == 21
+    assert set(result["tasks"][0]) == {"task_id", "project_id"}
+    serialized = json.dumps(result)
+    assert "SECRET" not in serialized
+    assert "unrelated" not in serialized
+    assert "lookalike-root" not in serialized
+
+
+def test_task_inventory_rejects_any_count_other_than_exactly_21(tmp_path: Path) -> None:
+    specs = _task_specs(tmp_path)
+    tasks = [
+        ScheduledTask(task_id=f"task-{index}", command=str(specs[0].root / "run.py"))
+        for index in range(20)
+    ]
+
+    with pytest.raises(TaskCountMismatch, match="expected 21.*found 20"):
+        collect_scheduled_task_inventory(
+            specs, FakeTaskProvider(tasks), expected_count=21
+        )
+
+
+def test_build_manifest_has_code_data_and_exact_task_baselines_without_values(
+    tmp_path: Path,
+) -> None:
+    root = tmp_path / "legacy"
+    data = root / "shared-data"
+    data.mkdir(parents=True)
+    credential_key = "PASS" + "WORD"
+    credential_value = "not-" + "exported"
+    (root / "job.py").write_text(
+        f"{credential_key} = '{credential_value}'\n", encoding="utf-8"
+    )
+    (data / "dataset.json").write_text('{"token":"hidden"}', encoding="utf-8")
+    spec = ProjectSpec(
+        project_id="project",
+        root=root,
+        data_roots=(DataRootSpec(label="data", path=data),),
+    )
+    tasks = [
+        ScheduledTask(
+            task_id=f"task-{index:02d}",
+            command=str(root / "job.py"),
+            arguments=f"--password secret-{index}",
+        )
+        for index in range(21)
+    ]
+
+    manifest = build_baseline_manifest(
+        (spec,),
+        FakeTaskProvider(tasks),
+        generated_at=datetime(2026, 7, 27, 5, 0, tzinfo=timezone.utc),
+    )
+
+    assert manifest["schema_version"] == 1
+    assert manifest["generated_at"] == "2026-07-27T05:00:00+00:00"
+    assert manifest["scheduled_tasks"]["actual_count"] == 21
+    assert len(manifest["projects"][0]["code_inventory"]) == 1
+    assert manifest["projects"][0]["code_inventory"][0]["relative_path"] == "job.py"
+    assert manifest["projects"][0]["data_summaries"][0]["summary"]["file_count"] == 1
+    serialized = json.dumps(manifest)
+    assert credential_value not in serialized
+    assert "dataset.json" not in serialized
+    assert "secret-0" not in serialized
+
+
+def test_manifest_write_is_atomic_and_leaves_no_temporary_file(tmp_path: Path) -> None:
+    target = tmp_path / "baseline" / "manifest.json"
+    write_manifest_atomic(target, {"version": 1})
+    write_manifest_atomic(target, {"version": 2})
+
+    assert json.loads(target.read_text(encoding="utf-8")) == {"version": 2}
+    assert list(target.parent.glob(".*.tmp")) == []
diff --git a/tests/test_business_modules.py b/tests/test_business_modules.py
new file mode 100644
index 0000000..4430dbc
--- /dev/null
+++ b/tests/test_business_modules.py
@@ -0,0 +1,156 @@
+from __future__ import annotations
+
+import ast
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.modules import (
+    EXPECTED_MODULE_IDS,
+    BusinessModule,
+    ModuleRegistry,
+    create_default_registry,
+)
+from gyxx_flow.modules.content_marketing import ContentMarketingModule
+from gyxx_flow.modules.product_commerce import ProductCommerceModule
+from gyxx_flow.modules.shop_intelligence import ShopIntelligenceModule
+from gyxx_flow.modules.supply_chain import SupplyChainModule
+from gyxx_flow.workflow import WorkflowDefinition
+
+
+def test_default_registry_contains_exactly_the_four_business_modules() -> None:
+    registry = create_default_registry()
+
+    assert registry.module_ids == EXPECTED_MODULE_IDS
+    assert isinstance(registry.get("content_marketing"), ContentMarketingModule)
+    assert isinstance(registry.get("product_commerce"), ProductCommerceModule)
+    assert isinstance(registry.get("shop_intelligence"), ShopIntelligenceModule)
+    assert isinstance(registry.get("supply_chain"), SupplyChainModule)
+
+
+def test_catalog_composition_registers_all_migrated_scheduled_and_manual_workflows() -> None:
+    catalog = WorkflowCatalog.load(Path(__file__).parents[1] / "config")
+
+    registry = create_default_registry(catalog=catalog)
+
+    assert len(registry.workflow_ids) == 28
+    assert set(registry.workflow_ids) == {
+        "shop.metrics.weekly",
+        "shop.competitor.weekly",
+        "supply.purchase_confirmation.daily",
+        "supply.replenishment.weekly",
+        "supply.replenishment_alert.daily",
+        "supply.purchase_order_update",
+        "content.metrics.daily",
+        "content.marketing_report.daily",
+        "content.relogin.weekly",
+        "content.self_operated.weekly",
+        "content.creator_report.monthly",
+        "content.summary.monthly",
+        "content.cooperations.daily",
+        "content.comments.weekly",
+        "content.summary.weekly",
+        "content.mapping.refresh",
+        "content.retry_failed",
+        "content.metrics.backfill",
+        "product.persona.daily",
+        "product.daily",
+        "product.alert.daily",
+        "product.import.daily",
+        "product.style_analysis.interval",
+        "product.main_image.jd.weekly",
+        "product.main_image.weekly",
+        "product.backfill",
+        "product.market_rank",
+        "product.review_collection",
+    }
+
+
+@pytest.mark.parametrize(
+    ("module", "module_id", "expected_workflow_ids"),
+    [
+        (ContentMarketingModule(), "content_marketing", ()),
+        (ProductCommerceModule(), "product_commerce", ()),
+        (ShopIntelligenceModule(), "shop_intelligence", ()),
+        (SupplyChainModule(), "supply_chain", ()),
+    ],
+)
+def test_every_module_implements_the_uniform_contract(
+    module: BusinessModule, module_id: str, expected_workflow_ids: tuple[str, ...]
+) -> None:
+    assert isinstance(module, BusinessModule)
+    assert module.module_id == module_id
+    assert tuple(
+        definition.workflow_id for definition in module.workflow_definitions()
+    ) == expected_workflow_ids
+
+
+def test_registry_rejects_duplicate_module_ids() -> None:
+    with pytest.raises(ValueError, match="duplicate module id"):
+        ModuleRegistry([ContentMarketingModule(), ContentMarketingModule()])
+
+
+def test_registry_indexes_workflows_without_exposing_mutable_state() -> None:
+    class ExampleModule:
+        module_id = "content_marketing"
+
+        def workflow_definitions(self) -> tuple[WorkflowDefinition, ...]:
+            return (WorkflowDefinition("content.example", ()),)
+
+    registry = ModuleRegistry([ExampleModule()])
+
+    assert registry.workflow_ids == ("content.example",)
+    assert registry.workflow("content.example").workflow_id == "content.example"
+    with pytest.raises(KeyError):
+        registry.workflow("content.missing")
+
+
+def test_business_packages_do_not_import_each_other_or_forbidden_gyxx_layers() -> None:
+    source_root = Path(__file__).parents[1] / "src" / "gyxx_flow" / "modules"
+    business_packages = set(EXPECTED_MODULE_IDS)
+    allowed_shared_roots = {
+        "gyxx_flow.core",
+        "gyxx_flow.workflow",
+        "gyxx_flow.adapters",
+        "gyxx_flow.contracts",
+        "gyxx_flow.modules.contracts",
+    }
+    violations: list[str] = []
+
+    for package in sorted(business_packages):
+        allowed_roots = allowed_shared_roots | {f"gyxx_flow.modules.{package}"}
+        for path in (source_root / package).rglob("*.py"):
+            tree = ast.parse(path.read_text(encoding="utf-8-sig"), filename=str(path))
+            for imported in _resolved_imports(tree, package):
+                other_business = {
+                    f"gyxx_flow.modules.{name}" for name in business_packages - {package}
+                }
+                if any(imported == root or imported.startswith(f"{root}.") for root in other_business):
+                    violations.append(f"{path.name}: cross-module import {imported}")
+                    continue
+                if imported.startswith("gyxx_flow.") and not any(
+                    imported == root or imported.startswith(f"{root}.")
+                    for root in allowed_roots
+                ):
+                    violations.append(f"{path.name}: forbidden layer import {imported}")
+
+    assert violations == []
+
+
+def _resolved_imports(tree: ast.AST, package: str) -> tuple[str, ...]:
+    names: list[str] = []
+    for node in ast.walk(tree):
+        if isinstance(node, ast.Import):
+            names.extend(alias.name for alias in node.names)
+        elif isinstance(node, ast.ImportFrom):
+            if node.level == 0 and node.module:
+                names.append(node.module)
+                continue
+            base = ["gyxx_flow", "modules", package]
+            parent = base[: len(base) - node.level + 1]
+            if node.module:
+                names.append(".".join([*parent, node.module]))
+            else:
+                names.extend(".".join([*parent, alias.name]) for alias in node.names)
+    return tuple(names)
diff --git a/tests/test_catalog.py b/tests/test_catalog.py
new file mode 100644
index 0000000..95d8cba
--- /dev/null
+++ b/tests/test_catalog.py
@@ -0,0 +1,86 @@
+from __future__ import annotations
+
+from collections import Counter
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.catalog import CatalogError, WorkflowCatalog
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+
+
+def test_catalog_maps_exactly_21_current_scheduled_tasks() -> None:
+    catalog = WorkflowCatalog.load(PROJECT_ROOT / "config")
+    scheduled = catalog.scheduled_workflows()
+
+    assert len(scheduled) == 21
+    assert len({item.source_task_name for item in scheduled}) == 21
+    assert Counter(item.module for item in scheduled) == {
+        "content_marketing": 9,
+        "product_commerce": 7,
+        "shop_intelligence": 2,
+        "supply_chain": 3,
+    }
+
+
+def test_catalog_keeps_known_manual_or_unregistered_workflows_unscheduled() -> None:
+    catalog = WorkflowCatalog.load(PROJECT_ROOT / "config")
+
+    assert {
+        "content.mapping.refresh",
+        "content.retry_failed",
+        "product.backfill",
+        "product.market_rank",
+        "product.review_collection",
+        "supply.purchase_order_update",
+    }.issubset({item.workflow_id for item in catalog.manual_workflows()})
+    assert "product.weekly_aggregate.documented_missing" in {
+        item.workflow_id for item in catalog.unavailable_workflows()
+    }
+
+
+def test_every_scheduled_workflow_has_one_valid_asia_shanghai_schedule() -> None:
+    catalog = WorkflowCatalog.load(PROJECT_ROOT / "config")
+
+    assert catalog.timezone == "Asia/Shanghai"
+    assert len(catalog.schedules) == 21
+    for workflow in catalog.scheduled_workflows():
+        schedule = catalog.schedule_for(workflow.workflow_id)
+        assert schedule.workflow_id == workflow.workflow_id
+        assert schedule.kind in {"daily", "weekly", "monthly", "interval_days"}
+        assert schedule.at.count(":") == 1
+
+
+def test_catalog_rejects_duplicate_workflow_ids(tmp_path: Path) -> None:
+    (tmp_path / "workflows.json").write_text(
+        '{"schema_version":2,"workflows":['
+        '{"id":"a.one","module":"content_marketing","trigger":"manual",'
+        '"execution":{"entry":"run.py"}},'
+        '{"id":"a.one","module":"content_marketing","trigger":"manual",'
+        '"execution":{"entry":"other.py"}}]}',
+        encoding="utf-8",
+    )
+    (tmp_path / "schedules.json").write_text(
+        '{"schema_version":1,"timezone":"Asia/Shanghai","schedules":[]}',
+        encoding="utf-8",
+    )
+
+    with pytest.raises(CatalogError, match="duplicate workflow id"):
+        WorkflowCatalog.load(tmp_path)
+
+
+def test_catalog_rejects_absolute_execution_entry(tmp_path: Path) -> None:
+    (tmp_path / "workflows.json").write_text(
+        '{"schema_version":2,"workflows":['
+        '{"id":"a.one","module":"content_marketing","trigger":"manual",'
+        '"execution":{"entry":"D:/old/run.py"}}]}',
+        encoding="utf-8",
+    )
+    (tmp_path / "schedules.json").write_text(
+        '{"schema_version":1,"timezone":"Asia/Shanghai","schedules":[]}',
+        encoding="utf-8",
+    )
+
+    with pytest.raises(CatalogError, match="relative"):
+        WorkflowCatalog.load(tmp_path)
diff --git a/tests/test_cli.py b/tests/test_cli.py
new file mode 100644
index 0000000..383733a
--- /dev/null
+++ b/tests/test_cli.py
@@ -0,0 +1,58 @@
+from __future__ import annotations
+
+import os
+import shutil
+import subprocess
+import sys
+from pathlib import Path
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+SOURCE_ROOT = PROJECT_ROOT / "src"
+
+
+def _environment_with_source() -> dict[str, str]:
+    environment = os.environ.copy()
+    existing_pythonpath = environment.get("PYTHONPATH")
+    paths = [str(SOURCE_ROOT)]
+    if existing_pythonpath:
+        paths.append(existing_pythonpath)
+    environment["PYTHONPATH"] = os.pathsep.join(paths)
+    return environment
+
+
+def _assert_successful_help(result: subprocess.CompletedProcess[str]) -> None:
+    assert result.returncode == 0, result.stderr
+    assert "usage:" in result.stdout.lower()
+    assert "GYXX Flow" in result.stdout
+
+
+def test_python_module_help_has_no_filesystem_side_effects(tmp_path: Path) -> None:
+    result = subprocess.run(
+        [sys.executable, "-m", "gyxx_flow", "--help"],
+        cwd=tmp_path,
+        env=_environment_with_source(),
+        capture_output=True,
+        text=True,
+        check=False,
+    )
+
+    _assert_successful_help(result)
+    assert list(tmp_path.iterdir()) == []
+
+
+def test_installed_console_script_help_has_no_filesystem_side_effects(
+    tmp_path: Path,
+) -> None:
+    executable = shutil.which("gyxx", path=str(Path(sys.executable).parent))
+    assert executable is not None, "The installed gyxx console script was not found"
+
+    result = subprocess.run(
+        [executable, "--help"],
+        cwd=tmp_path,
+        capture_output=True,
+        text=True,
+        check=False,
+    )
+
+    _assert_successful_help(result)
+    assert list(tmp_path.iterdir()) == []
diff --git a/tests/test_cli_workflows.py b/tests/test_cli_workflows.py
new file mode 100644
index 0000000..bc67a01
--- /dev/null
+++ b/tests/test_cli_workflows.py
@@ -0,0 +1,358 @@
+from __future__ import annotations
+
+import io
+import json
+from datetime import date
+from pathlib import Path
+
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.cli import (
+    EXIT_CONFIGURATION,
+    EXIT_SUCCESS,
+    EXIT_WORKFLOW_FAILED,
+    build_default_registry,
+    main,
+)
+from gyxx_flow.core.config import Settings
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.ops import RunIndex
+from gyxx_flow.workflow.model import StepDefinition, WorkflowDefinition
+from gyxx_flow.workflow.registry import WorkflowRegistry
+from gyxx_flow.workflow.steps import StepExecution
+
+
+class RecordingStep:
+    def __init__(self, *, fail_on: date | None = None) -> None:
+        self.fail_on = fail_on
+        self.calls: list[RunContext] = []
+
+    def execute(
+        self,
+        *,
+        context: RunContext,
+        timeout_seconds: float | None,
+        dry_run: bool,
+    ) -> StepExecution:
+        del timeout_seconds, dry_run
+        self.calls.append(context)
+        if context.business_date == self.fail_on:
+            return StepExecution(exit_code=9, error="fixture failure")
+        return StepExecution(exit_code=0)
+
+
+def test_default_cli_registry_exposes_all_21_migrated_scheduled_workflows(
+    tmp_path: Path,
+) -> None:
+    project_root = Path(__file__).parents[1]
+    registry = build_default_registry(
+        Settings(project_root=project_root, data_root=tmp_path)
+    )
+
+    assert sum(
+        registry.is_registered(entry.workflow_id)
+        for entry in registry.catalog.scheduled_workflows()
+    ) == 21
+
+
+def _catalog(tmp_path: Path) -> WorkflowCatalog:
+    config = tmp_path / "config"
+    config.mkdir()
+    (config / "workflows.json").write_text(
+        json.dumps(
+            {
+                "schema_version": 2,
+                "workflows": [
+                    {
+                        "id": "shop.metrics.weekly",
+                        "module": "shop_intelligence",
+                        "trigger": "scheduled",
+                        "execution": {"entry": "run.py"},
+                        "provenance": {
+                            "source_project": "shop",
+                            "task_name": "shop-weekly",
+                        },
+                    },
+                    {
+                        "id": "shop.missing",
+                        "module": "shop_intelligence",
+                        "trigger": "unavailable",
+                        "execution": {"entry": "missing.py"},
+                        "provenance": {"source_project": "shop"},
+                    },
+                ],
+            }
+        ),
+        encoding="utf-8",
+    )
+    (config / "schedules.json").write_text(
+        json.dumps(
+            {
+                "schema_version": 1,
+                "timezone": "Asia/Shanghai",
+                "schedules": [
+                    {
+                        "workflow_id": "shop.metrics.weekly",
+                        "kind": "weekly",
+                        "at": "12:00",
+                        "days": ["Monday"],
+                    }
+                ],
+            }
+        ),
+        encoding="utf-8",
+    )
+    return WorkflowCatalog.load(config)
+
+
+def _runtime(tmp_path: Path, workflow: WorkflowDefinition) -> tuple[WorkflowRegistry, Settings]:
+    registry = WorkflowRegistry(_catalog(tmp_path))
+    registry.register(workflow)
+    settings = Settings(project_root=tmp_path, data_root=tmp_path / "runtime")
+    return registry, settings
+
+
+def _invoke(
+    argv: list[str], registry: WorkflowRegistry, settings: Settings
+) -> tuple[int, str, str]:
+    stdout = io.StringIO()
+    stderr = io.StringIO()
+    code = main(
+        argv,
+        registry=registry,
+        settings=settings,
+        stdout=stdout,
+        stderr=stderr,
+    )
+    return code, stdout.getvalue(), stderr.getvalue()
+
+
+def test_list_reports_catalog_and_registration_state_as_json(tmp_path: Path) -> None:
+    action = RecordingStep()
+    registry, settings = _runtime(
+        tmp_path,
+        WorkflowDefinition(
+            "shop.metrics.weekly", (StepDefinition("collect", action),)
+        ),
+    )
+
+    code, output, error = _invoke(["list", "--json"], registry, settings)
+
+    assert code == EXIT_SUCCESS
+    assert error == ""
+    rows = json.loads(output)
+    assert rows == [
+        {
+            "id": "shop.metrics.weekly",
+            "module": "shop_intelligence",
+            "registered": True,
+            "trigger": "scheduled",
+        },
+        {
+            "id": "shop.missing",
+            "module": "shop_intelligence",
+            "registered": False,
+            "trigger": "unavailable",
+        },
+    ]
+
+
+def test_run_is_dry_run_by_default_and_records_a_traceable_run(tmp_path: Path) -> None:
+    action = RecordingStep()
+    registry, settings = _runtime(
+        tmp_path,
+        WorkflowDefinition(
+            "shop.metrics.weekly", (StepDefinition("collect", action),)
+        ),
+    )
+
+    code, output, error = _invoke(
+        ["run", "shop.metrics.weekly", "--date", "2026-07-27"],
+        registry,
+        settings,
+    )
+
+    assert code == EXIT_SUCCESS
+    assert error == ""
+    payload = json.loads(output)
+    assert payload["dry_run"] is True
+    assert payload["status"] == "success"
+    assert payload["steps"]["collect"] == "skipped"
+    assert payload["journal_path"].startswith("runs/")
+    assert action.calls == []
+    journals = list((settings.data_root / "runs").rglob("run.json"))
+    assert len(journals) == 1
+    assert json.loads(journals[0].read_text(encoding="utf-8"))["status"] == "success"
+    assert RunIndex(settings.data_root).get(payload["run_id"]).status == "success"
+
+
+def test_run_execute_returns_the_workflow_failure_exit_code(tmp_path: Path) -> None:
+    action = RecordingStep(fail_on=date(2026, 7, 27))
+    registry, settings = _runtime(
+        tmp_path,
+        WorkflowDefinition(
+            "shop.metrics.weekly", (StepDefinition("collect", action),)
+        ),
+    )
+
+    code, output, error = _invoke(
+        [
+            "run",
+            "shop.metrics.weekly",
+            "--date",
+            "2026-07-27",
+            "--execute",
+        ],
+        registry,
+        settings,
+    )
+
+    assert code == EXIT_WORKFLOW_FAILED
+    assert error == ""
+    assert json.loads(output)["status"] == "failed"
+    assert len(action.calls) == 1
+
+
+def test_run_unknown_unavailable_or_unregistered_is_a_configuration_error(
+    tmp_path: Path,
+) -> None:
+    registry = WorkflowRegistry(_catalog(tmp_path))
+    settings = Settings(project_root=tmp_path, data_root=tmp_path / "runtime")
+
+    for workflow_id in ("unknown.workflow", "shop.missing", "shop.metrics.weekly"):
+        code, output, error = _invoke(
+            ["run", workflow_id, "--date", "2026-07-27"], registry, settings
+        )
+        assert code == EXIT_CONFIGURATION
+        assert output == ""
+        assert workflow_id in error
+
+    assert not settings.data_root.exists()
+
+
+def test_backfill_runs_each_business_date_and_aggregates_final_exit_code(
+    tmp_path: Path,
+) -> None:
+    action = RecordingStep(fail_on=date(2026, 7, 28))
+    registry, settings = _runtime(
+        tmp_path,
+        WorkflowDefinition(
+            "shop.metrics.weekly", (StepDefinition("collect", action),)
+        ),
+    )
+
+    code, output, error = _invoke(
+        [
+            "backfill",
+            "shop.metrics.weekly",
+            "--from",
+            "2026-07-27",
+            "--to",
+            "2026-07-29",
+            "--execute",
+        ],
+        registry,
+        settings,
+    )
+
+    assert code == EXIT_WORKFLOW_FAILED
+    assert error == ""
+    payload = json.loads(output)
+    assert payload["status"] == "failed"
+    assert [run["business_date"] for run in payload["runs"]] == [
+        "2026-07-27",
+        "2026-07-28",
+        "2026-07-29",
+    ]
+    assert [context.business_date for context in action.calls] == [
+        date(2026, 7, 27),
+        date(2026, 7, 28),
+        date(2026, 7, 29),
+    ]
+
+
+def test_step_rerun_and_resume_select_only_the_requested_execution_slice(
+    tmp_path: Path,
+) -> None:
+    collect = RecordingStep()
+    normalize = RecordingStep()
+    publish = RecordingStep()
+    registry, settings = _runtime(
+        tmp_path,
+        WorkflowDefinition(
+            "shop.metrics.weekly",
+            (
+                StepDefinition("collect", collect),
+                StepDefinition("normalize", normalize, depends_on=("collect",)),
+                StepDefinition("publish", publish, depends_on=("normalize",)),
+            ),
+        ),
+    )
+
+    rerun_code, _, rerun_error = _invoke(
+        [
+            "run",
+            "shop.metrics.weekly",
+            "--date",
+            "2026-07-27",
+            "--execute",
+            "--rerun-step",
+            "normalize",
+        ],
+        registry,
+        settings,
+    )
+    resume_code, _, resume_error = _invoke(
+        [
+            "run",
+            "shop.metrics.weekly",
+            "--date",
+            "2026-07-28",
+            "--execute",
+            "--resume-from",
+            "normalize",
+        ],
+        registry,
+        settings,
+    )
+
+    assert (rerun_code, resume_code) == (EXIT_SUCCESS, EXIT_SUCCESS)
+    assert rerun_error == resume_error == ""
+    assert [context.business_date for context in collect.calls] == []
+    assert [context.business_date for context in normalize.calls] == [
+        date(2026, 7, 27),
+        date(2026, 7, 28),
+    ]
+    assert [context.business_date for context in publish.calls] == [date(2026, 7, 28)]
+
+
+def test_backfill_rejects_reverse_or_excessive_ranges_without_creating_runs(
+    tmp_path: Path,
+) -> None:
+    action = RecordingStep()
+    registry, settings = _runtime(
+        tmp_path,
+        WorkflowDefinition(
+            "shop.metrics.weekly", (StepDefinition("collect", action),)
+        ),
+    )
+
+    for start, end in (("2026-07-29", "2026-07-27"), ("2025-01-01", "2026-07-27")):
+        code, output, error = _invoke(
+            [
+                "backfill",
+                "shop.metrics.weekly",
+                "--from",
+                start,
+                "--to",
+                end,
+                "--execute",
+            ],
+            registry,
+            settings,
+        )
+        assert code == EXIT_CONFIGURATION
+        assert output == ""
+        assert error
+
+    assert action.calls == []
+    assert not settings.data_root.exists()
diff --git a/tests/test_content_data_layout.py b/tests/test_content_data_layout.py
new file mode 100644
index 0000000..80f31e0
--- /dev/null
+++ b/tests/test_content_data_layout.py
@@ -0,0 +1,122 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.modules.content_marketing.runtime import runtime_paths
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+RUNTIME_ROOT = (
+    PROJECT_ROOT / "src" / "gyxx_flow" / "modules" / "content_marketing" / "runtime"
+)
+
+
+def _source(relative: str) -> str:
+    return (RUNTIME_ROOT / relative).read_text(encoding="utf-8-sig")
+
+
+def test_content_output_resolver_keeps_paths_inside_the_selected_layer(
+    tmp_path: Path,
+) -> None:
+    root = tmp_path / "exports" / "content_marketing"
+
+    relative = runtime_paths.resolve_layer_output(
+        "reports/monthly.md", layer_root=root, field="--output"
+    )
+    absolute = runtime_paths.resolve_layer_output(
+        root / "summary" / "daily.md", layer_root=root, field="--output"
+    )
+
+    assert relative == (root / "reports" / "monthly.md").resolve()
+    assert absolute == (root / "summary" / "daily.md").resolve()
+    with pytest.raises(ValueError, match="--output"):
+        runtime_paths.resolve_layer_output(
+            "../escape.md", layer_root=root, field="--output"
+        )
+    with pytest.raises(ValueError, match="--output"):
+        runtime_paths.resolve_layer_output(
+            tmp_path / "outside.md", layer_root=root, field="--output"
+        )
+
+
+def test_dynamic_mapping_uses_normalized_cache_and_read_only_source_seed() -> None:
+    source = _source("feishu_mapping.py")
+
+    assert 'MAPPING_DIR = PATHS.normalized_root / "mappings"' in source
+    assert 'MAPPING_SEED_PATH = PATHS.config_root /' in source
+    assert 'SELF_MAPPING_SEED_PATH = PATHS.config_root /' in source
+    assert 'MAPPING_PATH = PATHS.config_root /' not in source
+    assert "if data_dir:" not in source
+
+
+def test_v2_outputs_and_checkpoints_use_their_semantic_layers() -> None:
+    assert 'Path("data") / "v2_results"' not in _source("chanmama_scraper.py")
+    assert 'PATHS.normalized_root / "v2_results"' in _source("chanmama_scraper.py")
+
+    for relative in (
+        "bilibili_scraper.py",
+        "pgy_xhs_scraper_v2.py",
+        "xingtu_scraper_v2.py",
+        "self_bilibili_scraper.py",
+        "self_douyin_scraper.py",
+    ):
+        assert 'PATHS.normalized_root / "v2_results"' in _source(relative), relative
+
+    assert 'PATHS.curated_root / "run_all"' in _source("run_all.py")
+    assert 'PATHS.state_root / "checkpoints" / "bilibili"' in _source(
+        "bilibili_scraper.py"
+    )
+    assert "PATHS.browser_storage_state_file.write_text" in _source(
+        "xingtu_scraper.py"
+    )
+    assert 'PATHS.state_root / "checkpoints" / "pgy"' in _source(
+        "pgy_xhs_scraper_v2.py"
+    )
+
+    for relative in (
+        "data/tools/check_status.py",
+        "data/tools/retry_failed.py",
+        "data/tools/sync_metrics_to_cmt_notes.py",
+    ):
+        source = _source(relative)
+        assert 'PATHS.normalized_root / "v2_results"' in source, relative
+        assert 'PATHS.raw_root / "v2_results"' not in source, relative
+
+
+def test_transient_content_artifacts_use_tmp_root() -> None:
+    doc_writer = _source("data/tools/feishu_doc_writer.py")
+    report_card = _source("data/tools/daily_report_card.py")
+
+    assert 'TMP_DIR = PATHS.tmp_root / "feishu_doc_writer"' in doc_writer
+    assert 'PATHS.tmp_root / "daily_report_card"' in report_card
+    assert 'chart_path.with_name(f"{chart_path.stem}_feishu.png")' not in report_card
+
+
+def test_batch_launchers_never_fall_back_to_module_local_var() -> None:
+    for path in sorted((RUNTIME_ROOT / "data" / "tools").glob("*.bat")):
+        source = path.read_text(encoding="utf-8-sig")
+        if "GYXX_DATA_ROOT" not in source:
+            continue
+        assert '%PROJECT_DIR%\\var' not in source, path.name
+        assert '%GYXX_PROJECT_ROOT%\\var' in source, path.name
+        assert "if not defined GYXX_PROJECT_ROOT" in source, path.name
+
+
+@pytest.mark.parametrize(
+    ("relative", "layer_expression"),
+    (
+        ("data/tools/analyze_comments.py", "PATHS.exports_root"),
+        ("data/tools/analyze_note.py", "PATHS.exports_root"),
+        ("data/tools/collect_note_metrics.py", "PATHS.normalized_root"),
+        ("data/tools/generate_creator_report.py", "PATHS.exports_root"),
+    ),
+)
+def test_cli_output_overrides_are_constrained_to_a_data_layer(
+    relative: str, layer_expression: str
+) -> None:
+    source = _source(relative)
+
+    assert "resolve_layer_output(" in source
+    assert layer_expression in source
+
diff --git a/tests/test_content_marketing_module.py b/tests/test_content_marketing_module.py
new file mode 100644
index 0000000..97c385f
--- /dev/null
+++ b/tests/test_content_marketing_module.py
@@ -0,0 +1,43 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+from gyxx_flow.adapters.native import DeferredModuleCommandStep
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.modules.content_marketing import (
+    CONTENT_MANUAL_WORKFLOW_IDS,
+    CONTENT_RESOURCE,
+    CONTENT_SCHEDULED_WORKFLOW_IDS,
+    CONTENT_TIMEOUT_SECONDS,
+    CONTENT_WORKFLOW_IDS,
+    ContentMarketingModule,
+)
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+
+
+def test_content_module_registers_all_scheduled_and_manual_entries() -> None:
+    module = ContentMarketingModule.from_catalog(
+        WorkflowCatalog.load(PROJECT_ROOT / "config")
+    )
+
+    definitions = module.workflow_definitions()
+    assert tuple(item.workflow_id for item in definitions) == CONTENT_WORKFLOW_IDS
+    assert len(CONTENT_SCHEDULED_WORKFLOW_IDS) == 9
+    assert len(CONTENT_MANUAL_WORKFLOW_IDS) == 3
+
+
+def test_content_workflows_use_project_owned_module_commands() -> None:
+    module = ContentMarketingModule.from_catalog(
+        WorkflowCatalog.load(PROJECT_ROOT / "config")
+    )
+
+    for definition in module.workflow_definitions():
+        step = definition.steps[0]
+        assert step.step_id == "module_run"
+        assert isinstance(step.action, DeferredModuleCommandStep)
+        assert step.timeout_seconds == CONTENT_TIMEOUT_SECONDS == 4 * 60 * 60
+        assert step.max_attempts == 1
+        assert step.resources == (CONTENT_RESOURCE,) == ("module:content_marketing",)
+        assert step.production_sink is True
+
diff --git a/tests/test_content_source_snapshot.py b/tests/test_content_source_snapshot.py
new file mode 100644
index 0000000..e6fcaf8
--- /dev/null
+++ b/tests/test_content_source_snapshot.py
@@ -0,0 +1,200 @@
+from __future__ import annotations
+
+import hashlib
+import json
+import re
+from pathlib import Path, PurePosixPath
+
+from gyxx_flow.security.scanner import scan_repository
+
+PROJECT_ROOT = Path(__file__).parents[1]
+RUNTIME_ROOT = (
+    PROJECT_ROOT / "src" / "gyxx_flow" / "modules" / "content_marketing" / "runtime"
+)
+MANIFEST_PATH = (
+    PROJECT_ROOT / "config" / "source-manifests" / "content_marketing.json"
+)
+SHA256 = re.compile(r"^[0-9a-f]{64}$")
+OLD_PROJECT_ROOTS = (
+    "d:\\yingxiaoyunying",
+    "d:/yingxiaoyunying",
+    "d:\\shop-data-flow",
+    "d:/shop-data-flow",
+    "d:\\product-collector-analyze-flow",
+    "d:/product-collector-analyze-flow",
+    "e:\\auto-flow",
+    "e:/auto-flow",
+)
+FIXED_USER_PATHS = (
+    "c:\\users\\administrator",
+    "c:/users/administrator",
+)
+EXPECTED_CATEGORIES = {
+    "config",
+    "dependency",
+    "launcher",
+    "source",
+    "sql",
+    "test",
+    "tool",
+}
+REQUIRED_SOURCES = {
+    "run_all.py",
+    "bilibili_scraper.py",
+    "pgy_xhs_scraper_v2.py",
+    "xingtu_scraper_v2.py",
+    "chanmama_scraper.py",
+    "daily_marketing_report.py",
+    "data/tools/daily_run.bat",
+    "data/tools/friday_relogin_parallel.py",
+    "data/tools/schema_gyxx_super_data.sql",
+    "data/tools/migrations/005_style_product_profile.sql",
+    "data/config/db.env.example",
+    "data/config/款式_多维表格_对照.json",
+    "tests/test_collection_completeness.py",
+}
+
+
+def _sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as handle:
+        for block in iter(lambda: handle.read(1024 * 1024), b""):
+            digest.update(block)
+    return digest.hexdigest()
+
+
+def _manifest() -> dict[str, object]:
+    return json.loads(MANIFEST_PATH.read_text(encoding="utf-8"))
+
+
+def test_content_source_snapshot_manifest_is_complete_and_well_formed() -> None:
+    manifest = _manifest()
+
+    assert manifest["schema_version"] == 1
+    assert manifest["module"] == "content_marketing"
+    files = manifest["files"]
+    assert isinstance(files, list)
+    assert len(files) == 90
+
+    sources = [entry["source_relative_path"] for entry in files]
+    targets = [entry["target_relative_path"] for entry in files]
+    assert sources == sorted(sources, key=str.casefold)
+    assert len(sources) == len(set(sources))
+    assert len(targets) == len(set(targets))
+    assert REQUIRED_SOURCES <= set(sources)
+    assert {entry["category"] for entry in files} == EXPECTED_CATEGORIES
+
+    for entry in files:
+        relative = PurePosixPath(entry["target_relative_path"])
+        assert not relative.is_absolute()
+        assert ".." not in relative.parts
+        assert relative.parts[:5] == (
+            "src",
+            "gyxx_flow",
+            "modules",
+            "content_marketing",
+            "runtime",
+        )
+        assert SHA256.fullmatch(entry["source_sha256"])
+        assert SHA256.fullmatch(entry["target_sha256"])
+        assert entry["transformed"] is (
+            entry["source_sha256"] != entry["target_sha256"]
+        )
+
+
+def test_content_source_snapshot_targets_exist_and_match_manifest_hashes() -> None:
+    for entry in _manifest()["files"]:
+        target = PROJECT_ROOT.joinpath(*PurePosixPath(entry["target_relative_path"]).parts)
+        assert target.is_file(), entry["target_relative_path"]
+        assert _sha256(target) == entry["target_sha256"]
+
+    generated = _manifest()["generated_files"]
+    assert [entry["target_relative_path"] for entry in generated] == [
+        "src/gyxx_flow/modules/content_marketing/runtime/runtime_paths.py"
+    ]
+    for entry in generated:
+        target = PROJECT_ROOT.joinpath(*PurePosixPath(entry["target_relative_path"]).parts)
+        assert target.is_file(), entry["target_relative_path"]
+        assert SHA256.fullmatch(entry["target_sha256"])
+        assert _sha256(target) == entry["target_sha256"]
+
+
+def test_content_source_snapshot_exclusions_cover_runtime_and_sensitive_state() -> None:
+    excluded = {entry["pattern"] for entry in _manifest()["intentionally_excluded"]}
+
+    assert {
+        ".git/**",
+        "**/__pycache__/**",
+        "data/config/*.env",
+        "data/**/*cookies*",
+        ".*chrome_profile*/**",
+        "data/logs/**",
+        "data/v2_results/**",
+        "data/notes/**",
+        "data/reports/**",
+        "data/tmp/**",
+    } <= excluded
+
+
+def test_content_source_snapshot_has_no_legacy_roots_or_plaintext_credentials() -> None:
+    manifest_text = MANIFEST_PATH.read_text(encoding="utf-8").casefold()
+    assert not any(root in manifest_text for root in OLD_PROJECT_ROOTS)
+
+    for path in RUNTIME_ROOT.rglob("*"):
+        if not path.is_file() or b"\x00" in path.read_bytes():
+            continue
+        text = path.read_text(encoding="utf-8-sig", errors="strict").casefold()
+        assert not any(root in text for root in OLD_PROJECT_ROOTS), path
+        assert not any(root in text for root in FIXED_USER_PATHS), path
+
+    assert scan_repository(RUNTIME_ROOT) == []
+
+
+def test_chanmama_credentials_are_externalized_in_the_copied_source() -> None:
+    source = (RUNTIME_ROOT / "chanmama_scraper.py").read_text(encoding="utf-8-sig")
+
+    assert 'os.getenv("CHANMAMA_ACCOUNT", "")' in source
+    assert 'os.getenv("CHANMAMA_PASSWORD", "")' in source
+    assert not re.search(r'^PASSWORD\s*=\s*["\'][^"\']+["\']', source, re.MULTILINE)
+
+
+def test_content_runtime_has_one_portable_path_boundary() -> None:
+    python_sources = list(RUNTIME_ROOT.rglob("*.py"))
+    path_mutations: list[Path] = []
+    forbidden_cross_module_imports: list[Path] = []
+
+    for path in python_sources:
+        source = path.read_text(encoding="utf-8-sig")
+        if re.search(r"sys\.path\.(?:insert|append)\(", source):
+            path_mutations.append(path.relative_to(RUNTIME_ROOT))
+        if "lark_cli_runtime" in source:
+            forbidden_cross_module_imports.append(path.relative_to(RUNTIME_ROOT))
+
+    assert path_mutations == []
+    assert forbidden_cross_module_imports == []
+
+    runtime_paths = (
+        RUNTIME_ROOT / "runtime_paths.py"
+    ).read_text(encoding="utf-8-sig")
+    assert "GYXX_DATA_ROOT" in runtime_paths
+    assert "GYXX_MODULE_ROOT" in runtime_paths
+
+
+def test_content_runtime_uses_portable_python_launchers_and_data_paths() -> None:
+    batch_sources = list(RUNTIME_ROOT.rglob("*.bat"))
+    assert batch_sources
+
+    for path in batch_sources:
+        source = path.read_text(encoding="utf-8-sig").casefold()
+        assert "%gyxx_python%" in source, path
+        assert "hermes-agent" not in source, path
+        assert "appdata" not in source, path
+
+    path_constants = (RUNTIME_ROOT / "runtime_paths.py").read_text(
+        encoding="utf-8-sig"
+    )
+    assert 'DataLayout(data_root).for_module("content_marketing")' in path_constants
+    assert "raw_root = _DATA_PATHS.raw_root" in path_constants
+    assert "normalized_root = _DATA_PATHS.normalized_root" in path_constants
+    assert "curated_root = _DATA_PATHS.curated_root" in path_constants
+    assert "exports_root = _DATA_PATHS.exports_root" in path_constants
diff --git a/tests/test_core_foundation.py b/tests/test_core_foundation.py
new file mode 100644
index 0000000..e9f2820
--- /dev/null
+++ b/tests/test_core_foundation.py
@@ -0,0 +1,117 @@
+from __future__ import annotations
+
+import json
+from datetime import datetime, timezone
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.core.artifacts import ArtifactManifest, atomic_write_json, sha256_file
+from gyxx_flow.core.config import Settings
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.core.layout import DataLayout
+
+
+def test_settings_default_data_root_is_project_local_var(tmp_path: Path) -> None:
+    settings = Settings.from_env(project_root=tmp_path, env={})
+
+    assert settings.project_root == tmp_path.resolve()
+    assert settings.data_root == (tmp_path / "var").resolve()
+
+
+def test_settings_accepts_portable_data_root_override(tmp_path: Path) -> None:
+    external = tmp_path / "runtime"
+
+    settings = Settings.from_env(
+        project_root=tmp_path / "repo",
+        env={"GYXX_DATA_ROOT": str(external)},
+    )
+
+    assert settings.data_root == external.resolve()
+
+
+def test_run_context_validates_date_and_builds_traceable_id() -> None:
+    fixed_now = datetime(2026, 7, 27, 6, 30, 45, tzinfo=timezone.utc)
+
+    context = RunContext.create(
+        "product.daily",
+        "2026-07-26",
+        shadow=True,
+        now=fixed_now,
+        random_suffix="abc123",
+    )
+
+    assert context.business_date.isoformat() == "2026-07-26"
+    assert context.run_id == "product.daily__20260726__20260727T063045Z__abc123"
+    assert context.shadow is True
+
+
+@pytest.mark.parametrize("value", ["2026-02-30", "27-07-2026", ""])
+def test_run_context_rejects_invalid_business_date(value: str) -> None:
+    with pytest.raises(ValueError, match="business_date"):
+        RunContext.create("product.daily", value)
+
+
+@pytest.mark.parametrize("workflow_id", ["../escape", "Bad Workflow", "D:\\legacy"])
+def test_run_context_rejects_unsafe_workflow_id(workflow_id: str) -> None:
+    with pytest.raises(ValueError, match="workflow_id"):
+        RunContext.create(workflow_id, "2026-07-26")
+
+
+def test_data_layout_separates_raw_runs_and_runtime_state(tmp_path: Path) -> None:
+    layout = DataLayout(tmp_path)
+
+    raw = layout.raw(
+        domain="commerce",
+        source="jd",
+        dataset="product_daily",
+        business_date="2026-07-26",
+        run_id="run-001",
+    )
+
+    assert raw == (
+        tmp_path
+        / "data"
+        / "raw"
+        / "commerce"
+        / "jd"
+        / "product_daily"
+        / "business_date=2026-07-26"
+        / "run_id=run-001"
+    )
+    assert layout.state("cookies", "xingtu") == tmp_path / "state" / "cookies" / "xingtu"
+    assert layout.tmp("run-001") == tmp_path / "tmp" / "run-001"
+
+
+@pytest.mark.parametrize("segment", ["..", "a/b", "a\\b", "C:\\data", ""])
+def test_data_layout_rejects_unsafe_segments(tmp_path: Path, segment: str) -> None:
+    layout = DataLayout(tmp_path)
+
+    with pytest.raises(ValueError, match="path segment"):
+        layout.state("cookies", segment)
+
+
+def test_atomic_json_and_artifact_manifest_are_verifiable(tmp_path: Path) -> None:
+    payload_path = tmp_path / "part-000.json"
+    atomic_write_json(payload_path, {"rows": [{"sku": "A-1", "qty": 3}]})
+
+    payload = json.loads(payload_path.read_text(encoding="utf-8"))
+    digest = sha256_file(payload_path)
+    manifest = ArtifactManifest.from_file(
+        payload_path,
+        artifact_id="artifact-001",
+        dataset="product_daily",
+        workflow_id="product.daily",
+        run_id="run-001",
+        business_date="2026-07-26",
+        schema_version="1",
+        row_count=1,
+        upstream_artifact_ids=("source-001",),
+    )
+
+    assert payload["rows"][0]["sku"] == "A-1"
+    assert len(digest) == 64
+    assert manifest.sha256 == digest
+    assert manifest.byte_size == payload_path.stat().st_size
+    assert manifest.to_dict()["upstream_artifact_ids"] == ["source-001"]
+    assert list(tmp_path.glob("*.tmp")) == []
diff --git a/tests/test_core_runtime.py b/tests/test_core_runtime.py
new file mode 100644
index 0000000..6b596c1
--- /dev/null
+++ b/tests/test_core_runtime.py
@@ -0,0 +1,98 @@
+from __future__ import annotations
+
+import json
+from datetime import datetime, timezone
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.core.layout import DataLayout
+from gyxx_flow.core.locks import LockManager, ResourceBusyError
+from gyxx_flow.core.records import RunJournal
+
+
+def _context() -> RunContext:
+    return RunContext.create(
+        "shop.weekly",
+        "2026-07-27",
+        now=datetime(2026, 7, 27, 4, 0, tzinfo=timezone.utc),
+        random_suffix="abc123",
+    )
+
+
+def test_run_journal_tracks_steps_and_final_status(tmp_path: Path) -> None:
+    context = _context()
+    journal = RunJournal.create(DataLayout(tmp_path), context)
+
+    journal.start_step("collect.jd", attempt=1)
+    journal.finish_step("collect.jd", status="success", exit_code=0)
+    journal.start_step("publish.feishu", attempt=1)
+    journal.finish_step(
+        "publish.feishu",
+        status="failed",
+        exit_code=3,
+        error="credential binding missing",
+    )
+    journal.finalize("failed", error="publish.feishu failed")
+
+    payload = json.loads(journal.path.read_text(encoding="utf-8"))
+    assert payload["run_id"] == context.run_id
+    assert payload["status"] == "failed"
+    assert payload["steps"]["collect.jd"]["status"] == "success"
+    assert payload["steps"]["publish.feishu"]["exit_code"] == 3
+    assert payload["steps"]["publish.feishu"]["error"] == "credential binding missing"
+    assert payload["error"] == "publish.feishu failed"
+    assert payload["trace"]["paths"]["run"].startswith("runs/")
+    assert payload["trace"]["paths"]["log"].startswith("logs/")
+    assert payload["trace"]["paths"]["evidence"].startswith("data/evidence/")
+
+
+def test_run_journal_records_deduplicated_trace_references(tmp_path: Path) -> None:
+    journal = RunJournal.create(DataLayout(tmp_path), _context())
+
+    journal.record_input("artifact:raw-001")
+    journal.record_input("artifact:raw-001")
+    journal.record_output("artifact:normalized-001")
+    journal.record_external_write("outbox:msg-001")
+
+    payload = json.loads(journal.path.read_text(encoding="utf-8"))
+    assert payload["trace"]["inputs"] == ["artifact:raw-001"]
+    assert payload["trace"]["outputs"] == ["artifact:normalized-001"]
+    assert payload["trace"]["external_writes"] == ["outbox:msg-001"]
+
+
+def test_run_journal_rejects_invalid_transitions(tmp_path: Path) -> None:
+    journal = RunJournal.create(DataLayout(tmp_path), _context())
+
+    with pytest.raises(ValueError, match="not running"):
+        journal.finish_step("collect.jd", status="success")
+
+    journal.start_step("collect.jd", attempt=1)
+    with pytest.raises(ValueError, match="already running"):
+        journal.start_step("collect.jd", attempt=2)
+
+
+def test_named_resource_lock_blocks_concurrent_owner_and_releases(tmp_path: Path) -> None:
+    manager = LockManager(tmp_path)
+
+    with manager.acquire("browser:xingtu", owner="run-001") as first:
+        assert first.path.exists()
+        with pytest.raises(ResourceBusyError, match="browser:xingtu"):
+            with manager.acquire("browser:xingtu", owner="run-002", timeout_seconds=0):
+                pass
+
+    assert not first.path.exists()
+    with manager.acquire("browser:xingtu", owner="run-002", timeout_seconds=0) as second:
+        metadata = json.loads(second.path.read_text(encoding="utf-8"))
+        assert metadata["resource"] == "browser:xingtu"
+        assert metadata["owner"] == "run-002"
+
+
+@pytest.mark.parametrize("resource", ["", "../state", "resource\nname"])
+def test_named_resource_lock_rejects_unsafe_name(tmp_path: Path, resource: str) -> None:
+    manager = LockManager(tmp_path)
+
+    with pytest.raises(ValueError, match="resource"):
+        with manager.acquire(resource, owner="run-001"):
+            pass
diff --git a/tests/test_cross_project_decoupling.py b/tests/test_cross_project_decoupling.py
new file mode 100644
index 0000000..9f0b6b5
--- /dev/null
+++ b/tests/test_cross_project_decoupling.py
@@ -0,0 +1,42 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+from gyxx_flow.adapters.native import DeferredModuleCommandStep
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.modules import create_default_registry
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+
+
+def test_default_runtime_uses_only_project_owned_module_steps() -> None:
+    registry = create_default_registry(
+        catalog=WorkflowCatalog.load(PROJECT_ROOT / "config")
+    )
+
+    for workflow_id in registry.workflow_ids:
+        definition = registry.workflow(workflow_id)
+        assert all(
+            isinstance(step.action, DeferredModuleCommandStep)
+            for step in definition.steps
+        )
+
+
+def test_runtime_composition_has_no_old_root_environment_contract() -> None:
+    runtime_files = [
+        PROJECT_ROOT / "config" / "workflows.json",
+        *(
+            path
+            for path in (PROJECT_ROOT / "src" / "gyxx_flow" / "modules").glob("*/__init__.py")
+        ),
+        PROJECT_ROOT / "src" / "gyxx_flow" / "adapters" / "native.py",
+    ]
+    forbidden = ("GYXX_LEGACY_", "D:\\yingxiaoyunying", "D:\\shop-data-flow", "D:\\product-collector-analyze-flow", "E:\\auto-flow")
+
+    violations = [
+        f"{path}: {marker}"
+        for path in runtime_files
+        for marker in forbidden
+        if marker in path.read_text(encoding="utf-8")
+    ]
+    assert violations == []
diff --git a/tests/test_data_migration.py b/tests/test_data_migration.py
new file mode 100644
index 0000000..774b1e5
--- /dev/null
+++ b/tests/test_data_migration.py
@@ -0,0 +1,339 @@
+from __future__ import annotations
+
+import json
+from pathlib import Path
+
+import pytest
+
+import gyxx_flow.migration.data as migration_data
+from gyxx_flow.core.artifacts import sha256_file
+from gyxx_flow.migration.data import (
+    DataLayer,
+    HistoricalDataMigrationExecutor,
+    HistoricalDataMigrationPlanner,
+    MigrationConflictError,
+    MigrationExecutionError,
+    MigrationSource,
+    MigrationValidationError,
+)
+
+
+def _source_tree(root: Path) -> dict[str, bytes]:
+    files = {
+        "2025/01/orders.csv": b"order_id,total\n1,19.90\n",
+        "2025/01/detail.json": b'{"id": 1, "items": 2}\n',
+    }
+    for relative, content in files.items():
+        path = root / relative
+        path.parent.mkdir(parents=True, exist_ok=True)
+        path.write_bytes(content)
+    return files
+
+
+def _planner(data_root: Path, *, free_bytes: int = 10_000_000) -> HistoricalDataMigrationPlanner:
+    return HistoricalDataMigrationPlanner(
+        data_root=data_root,
+        free_space_provider=lambda _path: free_bytes,
+    )
+
+
+def _source(root: Path, *, layer: DataLayer = DataLayer.RAW) -> MigrationSource:
+    return MigrationSource(
+        source_id="legacy-orders",
+        module="product_commerce",
+        layer=layer,
+        root=root,
+    )
+
+
+def test_plan_is_machine_readable_and_routes_files_by_module_and_layer(tmp_path: Path) -> None:
+    source_root = tmp_path / "legacy"
+    contents = _source_tree(source_root)
+    manifest_path = tmp_path / "new" / "plans" / "migration.json"
+
+    plan = _planner(tmp_path / "new").create_plan(
+        [_source(source_root)], manifest_path=manifest_path
+    )
+
+    assert plan.source_summary.file_count == 2
+    assert plan.source_summary.total_bytes == sum(map(len, contents.values()))
+    assert {entry.relative_path for entry in plan.files} == set(contents)
+    assert {
+        entry.target_relative_path for entry in plan.files
+    } == {
+        f"data/raw/product_commerce/legacy/legacy-orders/{relative}"
+        for relative in contents
+    }
+    payload = json.loads(manifest_path.read_text(encoding="utf-8"))
+    assert payload["schema_version"] == 1
+    assert payload["source_summary"]["file_count"] == 2
+    assert all(len(entry["sha256"]) == 64 for entry in payload["files"])
+    assert payload["precheck"]["status"] == "ready"
+
+
+@pytest.mark.parametrize("layer", list(DataLayer))
+def test_each_supported_layer_has_an_explicit_target_directory(
+    tmp_path: Path, layer: DataLayer
+) -> None:
+    source_root = tmp_path / f"legacy-{layer.value}"
+    (source_root / "one.txt").parent.mkdir(parents=True)
+    (source_root / "one.txt").write_text("one", encoding="utf-8")
+
+    plan = _planner(tmp_path / "new").create_plan([_source(source_root, layer=layer)])
+
+    assert plan.files[0].target_relative_path.startswith(
+        f"data/{layer.value}/product_commerce/legacy/legacy-orders/"
+    )
+
+
+def test_default_execution_is_plan_only_and_never_creates_data_files(tmp_path: Path) -> None:
+    source_root = tmp_path / "legacy"
+    _source_tree(source_root)
+    data_root = tmp_path / "new"
+    plan = _planner(data_root).create_plan([_source(source_root)])
+
+    report = HistoricalDataMigrationExecutor().execute(plan)
+
+    assert report.status == "plan_only"
+    assert report.applied is False
+    assert report.destination_summary.file_count == 0
+    assert not (data_root / "data").exists()
+
+
+def test_apply_copies_without_changing_or_removing_source_and_reconciles(tmp_path: Path) -> None:
+    source_root = tmp_path / "legacy"
+    contents = _source_tree(source_root)
+    data_root = tmp_path / "new"
+    report_path = data_root / "evidence" / "migration-report.json"
+    before = {
+        relative: (
+            sha256_file(source_root / relative),
+            (source_root / relative).stat().st_mtime_ns,
+        )
+        for relative in contents
+    }
+    plan = _planner(data_root).create_plan([_source(source_root)])
+
+    report = HistoricalDataMigrationExecutor().execute(
+        plan, apply=True, report_path=report_path
+    )
+
+    assert report.status == "reconciled"
+    assert report.applied is True
+    assert report.source_summary == report.destination_summary
+    assert report.copied_count == 2
+    assert report.mismatches == ()
+    for relative, expected_content in contents.items():
+        source_path = source_root / relative
+        destination = (
+            data_root
+            / "data/raw/product_commerce/legacy/legacy-orders"
+            / relative
+        )
+        assert source_path.exists()
+        assert source_path.read_bytes() == expected_content
+        assert (sha256_file(source_path), source_path.stat().st_mtime_ns) == before[relative]
+        assert destination.read_bytes() == expected_content
+        assert sha256_file(destination) == before[relative][0]
+    payload = json.loads(report_path.read_text(encoding="utf-8"))
+    assert payload["status"] == "reconciled"
+    assert payload["destination_summary"] == payload["source_summary"]
+
+
+def test_progress_reporting_performs_full_destination_reconciliation_only_once(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
+) -> None:
+    source_root = tmp_path / "legacy"
+    _source_tree(source_root)
+    plan = _planner(tmp_path / "new").create_plan([_source(source_root)])
+    original = migration_data._summarize_destinations
+    original_write = migration_data._write_report
+    calls = 0
+    writes = 0
+
+    def counted(entries, data_root):
+        nonlocal calls
+        calls += 1
+        return original(entries, data_root)
+
+    monkeypatch.setattr(migration_data, "_summarize_destinations", counted)
+
+    def counted_write(path, report):
+        nonlocal writes
+        writes += 1
+        return original_write(path, report)
+
+    monkeypatch.setattr(migration_data, "_write_report", counted_write)
+
+    report = HistoricalDataMigrationExecutor().execute(plan, apply=True)
+
+    assert report.status == "reconciled"
+    assert calls == 1
+    assert writes == 1
+
+
+def test_repeated_apply_is_idempotent_and_skips_matching_files(tmp_path: Path) -> None:
+    source_root = tmp_path / "legacy"
+    _source_tree(source_root)
+    data_root = tmp_path / "new"
+    plan = _planner(data_root).create_plan([_source(source_root)])
+    executor = HistoricalDataMigrationExecutor()
+    first = executor.execute(plan, apply=True)
+    destination = data_root / plan.files[0].target_relative_path
+    first_mtime = destination.stat().st_mtime_ns
+
+    second = executor.execute(plan, apply=True)
+
+    assert first.copied_count == 2
+    assert second.copied_count == 0
+    assert second.skipped_count == 2
+    assert destination.stat().st_mtime_ns == first_mtime
+    assert second.status == "reconciled"
+
+
+def test_precheck_rejects_conflicting_target_before_copying_anything(tmp_path: Path) -> None:
+    source_root = tmp_path / "legacy"
+    _source_tree(source_root)
+    data_root = tmp_path / "new"
+    conflict = (
+        data_root
+        / "data/raw/product_commerce/legacy/legacy-orders/2025/01/orders.csv"
+    )
+    conflict.parent.mkdir(parents=True)
+    conflict.write_text("different", encoding="utf-8")
+
+    with pytest.raises(MigrationConflictError, match="conflict"):
+        _planner(data_root).create_plan([_source(source_root)])
+
+    assert not (
+        data_root
+        / "data/raw/product_commerce/legacy/legacy-orders/2025/01/detail.json"
+    ).exists()
+
+
+def test_precheck_rejects_unplanned_files_in_dedicated_target_scope(
+    tmp_path: Path,
+) -> None:
+    source_root = tmp_path / "legacy"
+    _source_tree(source_root)
+    data_root = tmp_path / "new"
+    unexpected = (
+        data_root
+        / "data/raw/product_commerce/legacy/legacy-orders/unexpected.txt"
+    )
+    unexpected.parent.mkdir(parents=True)
+    unexpected.write_text("not in source", encoding="utf-8")
+
+    with pytest.raises(MigrationConflictError, match="unplanned target"):
+        _planner(data_root).create_plan([_source(source_root)])
+
+
+def test_precheck_rejects_insufficient_disk_space(tmp_path: Path) -> None:
+    source_root = tmp_path / "legacy"
+    _source_tree(source_root)
+
+    with pytest.raises(MigrationValidationError, match="disk space"):
+        _planner(tmp_path / "new", free_bytes=1).create_plan([_source(source_root)])
+
+
+def test_source_and_destination_must_not_overlap(tmp_path: Path) -> None:
+    source_root = tmp_path / "legacy"
+    _source_tree(source_root)
+
+    with pytest.raises(MigrationValidationError, match="overlap"):
+        _planner(source_root).create_plan([_source(source_root)])
+
+
+def test_source_change_after_planning_stops_before_overwriting_destination(tmp_path: Path) -> None:
+    source_root = tmp_path / "legacy"
+    _source_tree(source_root)
+    data_root = tmp_path / "new"
+    plan = _planner(data_root).create_plan([_source(source_root)])
+    changed = source_root / "2025/01/orders.csv"
+    changed.write_text("changed after planning", encoding="utf-8")
+
+    with pytest.raises(MigrationExecutionError, match="source changed"):
+        HistoricalDataMigrationExecutor().execute(plan, apply=True)
+
+    target = data_root / next(
+        entry.target_relative_path
+        for entry in plan.files
+        if entry.relative_path == "2025/01/orders.csv"
+    )
+    assert not target.exists()
+
+
+def test_apply_persists_progress_and_can_resume_after_interruption(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
+) -> None:
+    source_root = tmp_path / "legacy"
+    _source_tree(source_root)
+    data_root = tmp_path / "new"
+    report_path = data_root / "evidence" / "progress.json"
+    plan = _planner(data_root).create_plan([_source(source_root)])
+    executor = HistoricalDataMigrationExecutor()
+    original_copy = executor._copy_file_atomic
+    calls = 0
+
+    def fail_on_second(
+        source: Path, destination: Path, expected_size: int, expected_sha256: str
+    ) -> None:
+        nonlocal calls
+        calls += 1
+        if calls == 2:
+            raise OSError("simulated interruption")
+        original_copy(source, destination, expected_size, expected_sha256)
+
+    monkeypatch.setattr(executor, "_copy_file_atomic", fail_on_second)
+    with pytest.raises(MigrationExecutionError, match="simulated interruption"):
+        executor.execute(plan, apply=True, report_path=report_path)
+
+    interrupted = json.loads(report_path.read_text(encoding="utf-8"))
+    assert interrupted["status"] == "interrupted"
+    assert interrupted["copied_count"] == 1
+
+    resumed = HistoricalDataMigrationExecutor().execute(
+        plan, apply=True, report_path=report_path
+    )
+    assert resumed.status == "reconciled"
+    assert resumed.copied_count == 1
+    assert resumed.skipped_count == 1
+
+
+def test_corrupt_temporary_copy_is_never_committed_to_destination(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
+) -> None:
+    source_root = tmp_path / "legacy"
+    (source_root / "one.txt").parent.mkdir(parents=True)
+    (source_root / "one.txt").write_text("original", encoding="utf-8")
+    data_root = tmp_path / "new"
+    plan = _planner(data_root).create_plan([_source(source_root)])
+
+    def corrupt_copy(_source: Path, temporary: Path) -> None:
+        temporary.write_text("corrupt", encoding="utf-8")
+
+    monkeypatch.setattr("gyxx_flow.migration.data.shutil.copy2", corrupt_copy)
+    with pytest.raises(MigrationExecutionError, match="temporary copy verification"):
+        HistoricalDataMigrationExecutor().execute(plan, apply=True)
+
+    destination = data_root / plan.files[0].target_relative_path
+    assert not destination.exists()
+
+
+def test_unsafe_ids_and_symlink_escape_are_rejected(tmp_path: Path) -> None:
+    source_root = tmp_path / "legacy"
+    _source_tree(source_root)
+
+    with pytest.raises(MigrationValidationError, match="module"):
+        MigrationSource("source", "../escape", DataLayer.RAW, source_root)
+
+    outside = tmp_path / "outside.txt"
+    outside.write_text("outside", encoding="utf-8")
+    link = source_root / "escape.txt"
+    try:
+        link.symlink_to(outside)
+    except OSError as exc:
+        pytest.skip(f"symlinks are unavailable: {exc}")
+
+    with pytest.raises(MigrationValidationError, match="symbolic link"):
+        _planner(tmp_path / "new").create_plan([_source(source_root)])
diff --git a/tests/test_diagnostics_acceptance.py b/tests/test_diagnostics_acceptance.py
new file mode 100644
index 0000000..fa2c110
--- /dev/null
+++ b/tests/test_diagnostics_acceptance.py
@@ -0,0 +1,77 @@
+from __future__ import annotations
+
+import io
+import json
+from pathlib import Path
+
+from gyxx_flow.acceptance import build_acceptance_report, parse_plan_checklist
+from gyxx_flow.cli import EXIT_ACCEPTANCE_INCOMPLETE, main
+from gyxx_flow.core.config import Settings
+from gyxx_flow.diagnostics import run_doctor
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+
+
+def test_doctor_checks_interpreter_cli_config_paths_and_write_permission(tmp_path: Path) -> None:
+    settings = Settings(project_root=PROJECT_ROOT, data_root=tmp_path / "portable-data")
+
+    report = run_doctor(settings)
+
+    assert {check.name for check in report.checks} == {
+        "python",
+        "cli",
+        "project_root",
+        "catalog",
+        "data_root",
+        "data_root_write",
+    }
+    assert report.is_healthy is True
+    assert not (settings.data_root / ".gyxx-doctor-probe").exists()
+
+
+def test_plan_parser_returns_stable_item_ids_and_completion(tmp_path: Path) -> None:
+    plan = tmp_path / "plan.md"
+    plan.write_text(
+        "- [x] P0.1 complete\n- [ ] P0.2 pending\n- [X] P1.1 complete too\n",
+        encoding="utf-8",
+    )
+
+    items = parse_plan_checklist(plan)
+
+    assert [(item.item_id, item.completed) for item in items] == [
+        ("P0.1", True),
+        ("P0.2", False),
+        ("P1.1", True),
+    ]
+
+
+def test_acceptance_report_is_machine_readable_and_tracks_real_evidence() -> None:
+    settings = Settings(project_root=PROJECT_ROOT, data_root=PROJECT_ROOT / "var")
+
+    report = build_acceptance_report(settings)
+    payload = report.as_dict()
+
+    assert payload["schema_version"] == 1
+    assert payload["summary"]["total"] > payload["summary"]["completed"]
+    assert payload["checks"]["catalog_21_tasks"] is True
+    assert payload["checks"]["baseline_21_tasks"] is True
+    assert payload["checks"]["secret_scan_clean"] is True
+    assert report.is_complete is False
+
+
+def test_cli_doctor_and_acceptance_status_return_truthful_exit_codes(tmp_path: Path) -> None:
+    settings = Settings(project_root=PROJECT_ROOT, data_root=tmp_path / "data")
+    doctor_output = io.StringIO()
+    acceptance_output = io.StringIO()
+
+    assert main(["doctor", "--json"], settings=settings, stdout=doctor_output) == 0
+    assert json.loads(doctor_output.getvalue())["is_healthy"] is True
+
+    exit_code = main(
+        ["acceptance", "status", "--json"],
+        settings=Settings(project_root=PROJECT_ROOT, data_root=PROJECT_ROOT / "var"),
+        stdout=acceptance_output,
+    )
+    assert exit_code == EXIT_ACCEPTANCE_INCOMPLETE
+    assert json.loads(acceptance_output.getvalue())["is_complete"] is False
+
diff --git a/tests/test_effect_ledger.py b/tests/test_effect_ledger.py
new file mode 100644
index 0000000..bc204e2
--- /dev/null
+++ b/tests/test_effect_ledger.py
@@ -0,0 +1,66 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.ops.effects import (
+    EffectAlreadyApplied,
+    EffectLedger,
+    EffectStateAmbiguous,
+)
+
+
+def test_effect_ledger_blocks_applied_and_ambiguous_replays(tmp_path: Path) -> None:
+    ledger = EffectLedger(tmp_path)
+    first = ledger.begin(
+        workflow_id="product.daily",
+        business_date="2026-07-27",
+        step_id="legacy_execute",
+        run_id="run-001",
+    )
+    ledger.mark_applied(first)
+
+    with pytest.raises(EffectAlreadyApplied):
+        ledger.begin(
+            workflow_id="product.daily",
+            business_date="2026-07-27",
+            step_id="legacy_execute",
+            run_id="run-002",
+        )
+
+    ambiguous = ledger.begin(
+        workflow_id="product.daily",
+        business_date="2026-07-28",
+        step_id="legacy_execute",
+        run_id="run-003",
+    )
+    ledger.mark_ambiguous(ambiguous, reason="nonzero-exit")
+    with pytest.raises(EffectStateAmbiguous):
+        ledger.begin(
+            workflow_id="product.daily",
+            business_date="2026-07-28",
+            step_id="legacy_execute",
+            run_id="run-004",
+        )
+
+
+def test_effect_ledger_leaves_in_progress_claim_ambiguous_after_crash(
+    tmp_path: Path,
+) -> None:
+    ledger = EffectLedger(tmp_path)
+    ledger.begin(
+        workflow_id="shop.metrics.weekly",
+        business_date="2026-07-27",
+        step_id="legacy_collect",
+        run_id="run-001",
+    )
+
+    with pytest.raises(EffectStateAmbiguous):
+        ledger.begin(
+            workflow_id="shop.metrics.weekly",
+            business_date="2026-07-27",
+            step_id="legacy_collect",
+            run_id="run-002",
+        )
+
diff --git a/tests/test_legacy_adapter.py b/tests/test_legacy_adapter.py
new file mode 100644
index 0000000..4468d2c
--- /dev/null
+++ b/tests/test_legacy_adapter.py
@@ -0,0 +1,221 @@
+from __future__ import annotations
+
+import subprocess
+import sys
+from datetime import datetime, timezone
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.catalog import WorkflowEntry
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.migration.legacy import (
+    LegacyCommandAdapter,
+    LegacyConfigurationError,
+    LegacyProjectRoots,
+)
+
+PROJECT_IDS = ("content", "shop", "product", "supply")
+
+
+def _entry(**overrides: object) -> WorkflowEntry:
+    values: dict[str, object] = {
+        "workflow_id": "product.backfill",
+        "module": "product_commerce",
+        "trigger": "manual",
+        "source_project": "product",
+        "entry": "jobs/backfill.py",
+        "args": ("--limit", "25"),
+    }
+    values.update(overrides)
+    return WorkflowEntry(**values)  # type: ignore[arg-type]
+
+
+def _context(*, shadow: bool = True) -> RunContext:
+    return RunContext.create(
+        "product.backfill",
+        "2026-07-27",
+        shadow=shadow,
+        now=datetime(2026, 7, 27, 2, 30, tzinfo=timezone.utc),
+        random_suffix="abc123",
+    )
+
+
+def _project_roots(tmp_path: Path) -> dict[str, Path]:
+    roots = {project_id: tmp_path / project_id for project_id in PROJECT_IDS}
+    for root in roots.values():
+        root.mkdir()
+    return roots
+
+
+def test_legacy_roots_are_loaded_only_from_injected_environment(tmp_path: Path) -> None:
+    roots = _project_roots(tmp_path)
+    env = {
+        f"GYXX_LEGACY_{project_id.upper()}_ROOT": str(root)
+        for project_id, root in roots.items()
+    }
+
+    configured = LegacyProjectRoots.from_env(env=env)
+
+    assert configured.as_dict() == {
+        project_id: root.resolve() for project_id, root in roots.items()
+    }
+
+
+def test_legacy_roots_report_all_missing_environment_keys(tmp_path: Path) -> None:
+    content_root = tmp_path / "content"
+    content_root.mkdir()
+
+    with pytest.raises(LegacyConfigurationError) as captured:
+        LegacyProjectRoots.from_env(
+            env={"GYXX_LEGACY_CONTENT_ROOT": str(content_root)}
+        )
+
+    message = str(captured.value)
+    assert "GYXX_LEGACY_SHOP_ROOT" in message
+    assert "GYXX_LEGACY_PRODUCT_ROOT" in message
+    assert "GYXX_LEGACY_SUPPLY_ROOT" in message
+
+
+def test_explicit_root_configuration_rejects_missing_or_non_directory_paths(
+    tmp_path: Path,
+) -> None:
+    missing = tmp_path / "missing"
+    file_path = tmp_path / "not-a-directory"
+    file_path.write_text("x", encoding="utf-8")
+
+    with pytest.raises(LegacyConfigurationError, match="does not exist"):
+        LegacyProjectRoots({"product": missing})
+    with pytest.raises(LegacyConfigurationError, match="not a directory"):
+        LegacyProjectRoots({"product": file_path})
+
+
+def test_adapter_builds_python_command_and_injects_trace_context(tmp_path: Path) -> None:
+    roots = _project_roots(tmp_path)
+    entry_path = roots["product"] / "jobs" / "backfill.py"
+    entry_path.parent.mkdir()
+    entry_path.write_text("raise AssertionError('must not run')", encoding="utf-8")
+    context = _context()
+    adapter = LegacyCommandAdapter(
+        LegacyProjectRoots(roots),
+        base_env={"PATH": "injected-path", "GYXX_RUN_ID": "cannot-win"},
+        python_executable="portable-python",
+    )
+
+    command = adapter.build(_entry(), context=context)
+
+    assert command.argv == (
+        "portable-python",
+        str(entry_path.resolve()),
+        "--limit",
+        "25",
+    )
+    assert command.cwd == roots["product"].resolve()
+    assert command.env["PATH"] == "injected-path"
+    assert command.env["GYXX_WORKFLOW_ID"] == "product.backfill"
+    assert command.env["GYXX_RUN_ID"] == context.run_id
+    assert command.env["GYXX_BUSINESS_DATE"] == "2026-07-27"
+    assert command.env["GYXX_SHADOW"] == "true"
+    assert command.env["GYXX_LEGACY_PROJECT_ROOT"] == str(roots["product"].resolve())
+
+
+def test_adapter_keeps_batch_entry_and_args_as_separate_argv(tmp_path: Path) -> None:
+    roots = _project_roots(tmp_path)
+    entry_path = roots["content"] / "tools" / "daily run.bat"
+    entry_path.parent.mkdir()
+    entry_path.write_text("@echo off", encoding="utf-8")
+    adapter = LegacyCommandAdapter(LegacyProjectRoots(roots), base_env={})
+
+    command = adapter.build(
+        _entry(
+            workflow_id="content.metrics.daily",
+            module="content_marketing",
+            source_project="content",
+            entry="tools/daily run.bat",
+            args=("value with spaces", "&not-a-shell-fragment"),
+        ),
+        context=RunContext.create(
+            "content.metrics.daily",
+            "2026-07-27",
+            now=datetime(2026, 7, 27, tzinfo=timezone.utc),
+            random_suffix="def456",
+        ),
+    )
+
+    assert command.argv == (
+        str(entry_path.resolve()),
+        "value with spaces",
+        "&not-a-shell-fragment",
+    )
+
+
+@pytest.mark.parametrize(
+    "unsafe_entry",
+    ("../outside.py", "jobs/../../outside.py", "/outside.py", "C:/outside.py", "jobs\\outside.py"),
+)
+def test_adapter_rejects_unsafe_entry_paths(tmp_path: Path, unsafe_entry: str) -> None:
+    roots = _project_roots(tmp_path)
+    outside = tmp_path / "outside.py"
+    outside.write_text("pass", encoding="utf-8")
+    adapter = LegacyCommandAdapter(LegacyProjectRoots(roots), base_env={})
+
+    with pytest.raises(LegacyConfigurationError, match="relative|escape"):
+        adapter.build(_entry(entry=unsafe_entry), context=_context())
+
+
+def test_adapter_rejects_symlink_escape_when_supported(tmp_path: Path) -> None:
+    roots = _project_roots(tmp_path)
+    outside = tmp_path / "outside.py"
+    outside.write_text("pass", encoding="utf-8")
+    link = roots["product"] / "jobs"
+    try:
+        link.symlink_to(tmp_path, target_is_directory=True)
+    except OSError as exc:
+        pytest.skip(f"symlinks are unavailable: {exc}")
+    adapter = LegacyCommandAdapter(LegacyProjectRoots(roots), base_env={})
+
+    with pytest.raises(LegacyConfigurationError, match="escape"):
+        adapter.build(_entry(entry="jobs/outside.py"), context=_context())
+
+
+def test_adapter_rejects_unknown_project_context_mismatch_and_unavailable_entry(
+    tmp_path: Path,
+) -> None:
+    roots = _project_roots(tmp_path)
+    adapter = LegacyCommandAdapter(LegacyProjectRoots(roots), base_env={})
+
+    with pytest.raises(LegacyConfigurationError, match="unknown legacy project"):
+        adapter.build(_entry(source_project="other"), context=_context())
+    with pytest.raises(LegacyConfigurationError, match="does not match"):
+        adapter.build(
+            _entry(),
+            context=RunContext.create(
+                "another.workflow",
+                "2026-07-27",
+                now=datetime(2026, 7, 27, tzinfo=timezone.utc),
+                random_suffix="ghi789",
+            ),
+        )
+    with pytest.raises(LegacyConfigurationError, match="unavailable"):
+        adapter.build(_entry(trigger="unavailable"), context=_context())
+
+
+def test_adapter_construction_never_starts_a_process(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
+) -> None:
+    roots = _project_roots(tmp_path)
+    entry_path = roots["product"] / "jobs" / "backfill.py"
+    entry_path.parent.mkdir()
+    entry_path.write_text("pass", encoding="utf-8")
+
+    def fail_if_called(*args: object, **kwargs: object) -> None:
+        raise AssertionError("adapter construction must not start subprocesses")
+
+    monkeypatch.setattr(subprocess, "run", fail_if_called)
+    command = LegacyCommandAdapter(
+        LegacyProjectRoots(roots),
+        base_env={},
+        python_executable=sys.executable,
+    ).build(_entry(), context=_context())
+
+    assert command.argv[0] == sys.executable
diff --git a/tests/test_module_command_adapter.py b/tests/test_module_command_adapter.py
new file mode 100644
index 0000000..56475a6
--- /dev/null
+++ b/tests/test_module_command_adapter.py
@@ -0,0 +1,133 @@
+from __future__ import annotations
+
+import sys
+from datetime import datetime, timezone
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.adapters.native import (
+    DeferredModuleCommandStep,
+    ModuleCommandAdapter,
+    ModuleSourceError,
+    ModuleSourceRoots,
+)
+from gyxx_flow.catalog import WorkflowEntry
+from gyxx_flow.core.context import RunContext
+
+
+def _entry(**overrides: object) -> WorkflowEntry:
+    values: dict[str, object] = {
+        "workflow_id": "product.backfill",
+        "module": "product_commerce",
+        "trigger": "manual",
+        "entry": "jobs/backfill.py",
+        "args": ("--limit", "25"),
+        "source_project": "product",
+    }
+    values.update(overrides)
+    return WorkflowEntry(**values)  # type: ignore[arg-type]
+
+
+def _context() -> RunContext:
+    return RunContext.create(
+        "product.backfill",
+        "2026-07-27",
+        now=datetime(2026, 7, 27, tzinfo=timezone.utc),
+        random_suffix="native1",
+    )
+
+
+def test_module_adapter_resolves_only_inside_new_module_root(tmp_path: Path) -> None:
+    root = tmp_path / "product_commerce" / "runtime"
+    entry = root / "jobs" / "backfill.py"
+    entry.parent.mkdir(parents=True)
+    entry.write_text("pass", encoding="utf-8")
+    adapter = ModuleCommandAdapter(
+        ModuleSourceRoots({"product_commerce": root}),
+        python_executable=sys.executable,
+        base_env={"PATH": "portable"},
+        project_root=tmp_path,
+        data_root=tmp_path / "var",
+    )
+
+    command = adapter.build(_entry(), context=_context())
+
+    assert command.argv == (sys.executable, str(entry.resolve()), "--limit", "25")
+    assert command.cwd == root.resolve()
+    assert command.env["GYXX_PROJECT_ROOT"] == str(tmp_path.resolve())
+    assert command.env["GYXX_DATA_ROOT"] == str((tmp_path / "var").resolve())
+    assert command.env["GYXX_MODULE_ROOT"] == str(root.resolve())
+    assert not any(name.startswith("GYXX_LEGACY_") for name in command.env)
+
+
+@pytest.mark.parametrize("entry", ("../outside.py", "/outside.py", "C:/outside.py"))
+def test_module_adapter_rejects_paths_outside_module(tmp_path: Path, entry: str) -> None:
+    root = tmp_path / "runtime"
+    root.mkdir()
+    adapter = ModuleCommandAdapter(ModuleSourceRoots({"product_commerce": root}))
+
+    with pytest.raises(ModuleSourceError, match="relative|escape"):
+        adapter.build(_entry(entry=entry), context=_context())
+
+
+def test_deferred_module_step_dry_run_never_resolves_or_executes() -> None:
+    def fail(*args: object, **kwargs: object) -> object:
+        raise AssertionError("dry-run must not resolve or execute module source")
+
+    execution = DeferredModuleCommandStep(_entry(), command_factory=fail).execute(
+        context=_context(),
+        timeout_seconds=60,
+        dry_run=True,
+    )
+
+    assert execution.exit_code == 0
+    assert execution.skipped is True
+    assert execution.reason == "dry-run"
+
+
+def test_module_adapter_uses_explicit_windows_launchers_for_batch_and_powershell(
+    tmp_path: Path,
+) -> None:
+    root = tmp_path / "runtime"
+    root.mkdir()
+    (root / "daily.bat").write_text("@echo off", encoding="utf-8")
+    (root / "maint.ps1").write_text("Write-Output ok", encoding="utf-8")
+    adapter = ModuleCommandAdapter(
+        ModuleSourceRoots({"product_commerce": root}),
+        base_env={},
+        project_root=tmp_path,
+        data_root=tmp_path / "var",
+    )
+
+    batch = adapter.build(_entry(entry="daily.bat", args=()), context=_context())
+    powershell = adapter.build(
+        _entry(entry="maint.ps1", args=()), context=_context()
+    )
+
+    assert batch.argv[:4] == ("cmd.exe", "/d", "/s", "/c")
+    assert str((root / "daily.bat").resolve()) in batch.argv[4]
+    assert powershell.argv[:5] == (
+        "powershell.exe",
+        "-NoProfile",
+        "-ExecutionPolicy",
+        "Bypass",
+        "-File",
+    )
+
+
+def test_product_config_uses_only_the_new_environment_name(tmp_path: Path) -> None:
+    root = tmp_path / "runtime"
+    root.mkdir()
+    (root / "job.py").write_text("pass", encoding="utf-8")
+    adapter = ModuleCommandAdapter(
+        ModuleSourceRoots({"product_commerce": root}),
+        base_env={"GYXX_PRODUCT_CONFIG": "D:/portable/product.json"},
+        project_root=tmp_path,
+        data_root=tmp_path / "var",
+    )
+
+    command = adapter.build(_entry(entry="job.py", args=()), context=_context())
+
+    assert command.env["GYXX_PRODUCT_CONFIG"] == "D:/portable/product.json"
+    assert "AUTO_FLOW_CONFIG" not in command.env
diff --git a/tests/test_native_source_runtime.py b/tests/test_native_source_runtime.py
new file mode 100644
index 0000000..a33696d
--- /dev/null
+++ b/tests/test_native_source_runtime.py
@@ -0,0 +1,82 @@
+from __future__ import annotations
+
+import io
+import json
+import subprocess
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.cli import EXIT_SUCCESS, build_default_registry, main
+from gyxx_flow.core.config import Settings
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+MODULE_SOURCE_ROOT = PROJECT_ROOT / "src" / "gyxx_flow" / "modules"
+
+
+def test_workflow_catalog_uses_native_execution_schema() -> None:
+    payload = json.loads(
+        (PROJECT_ROOT / "config" / "workflows.json").read_text(encoding="utf-8")
+    )
+
+    assert payload["schema_version"] == 2
+    for workflow in payload["workflows"]:
+        assert "legacy" not in workflow
+        assert set(workflow["execution"]) >= {"entry"}
+
+
+def test_every_available_workflow_entry_exists_inside_its_new_module() -> None:
+    catalog = WorkflowCatalog.load(PROJECT_ROOT / "config")
+    missing: list[str] = []
+
+    for workflow in catalog.workflows:
+        if workflow.trigger == "unavailable":
+            continue
+        entry = MODULE_SOURCE_ROOT / workflow.module / "runtime" / workflow.entry
+        if not entry.is_file():
+            missing.append(f"{workflow.workflow_id}: {entry}")
+
+    assert missing == []
+
+
+def test_runtime_catalog_does_not_require_old_project_roots() -> None:
+    source = (PROJECT_ROOT / "config" / "workflows.json").read_text(encoding="utf-8")
+
+    assert "GYXX_LEGACY_" not in source
+    assert "D:\\yingxiaoyunying" not in source
+    assert "D:\\shop-data-flow" not in source
+    assert "D:\\product-collector-analyze-flow" not in source
+    assert "E:\\auto-flow" not in source
+
+
+def test_all_registered_workflows_dry_run_without_old_roots_or_processes(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
+) -> None:
+    for name in (
+        "GYXX_LEGACY_CONTENT_ROOT",
+        "GYXX_LEGACY_SHOP_ROOT",
+        "GYXX_LEGACY_PRODUCT_ROOT",
+        "GYXX_LEGACY_SUPPLY_ROOT",
+    ):
+        monkeypatch.delenv(name, raising=False)
+
+    def fail_process(*args: object, **kwargs: object) -> object:
+        raise AssertionError("dry-run must not start a module script")
+
+    monkeypatch.setattr(subprocess, "run", fail_process)
+    settings = Settings(project_root=PROJECT_ROOT, data_root=tmp_path)
+    registry = build_default_registry(settings)
+
+    for workflow_id in registry.catalog.workflows:
+        if not registry.is_registered(workflow_id.workflow_id):
+            continue
+        output = io.StringIO()
+        exit_code = main(
+            ["run", workflow_id.workflow_id, "--date", "2026-07-27"],
+            registry=registry,
+            settings=settings,
+            stdout=output,
+        )
+        assert exit_code == EXIT_SUCCESS, workflow_id.workflow_id
+        assert json.loads(output.getvalue())["dry_run"] is True
diff --git a/tests/test_ops.py b/tests/test_ops.py
new file mode 100644
index 0000000..f193e27
--- /dev/null
+++ b/tests/test_ops.py
@@ -0,0 +1,242 @@
+from __future__ import annotations
+
+import json
+from datetime import datetime, timezone
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.core.layout import DataLayout
+from gyxx_flow.core.records import RunJournal
+from gyxx_flow.ops import IdempotencyConflict, Outbox, RunIndex
+
+
+def _journal(
+    root: Path,
+    *,
+    workflow_id: str = "shop.weekly",
+    business_date: str = "2026-07-27",
+    suffix: str = "abc123",
+) -> RunJournal:
+    context = RunContext.create(
+        workflow_id,
+        business_date,
+        now=datetime(2026, 7, 27, 4, 0, tzinfo=timezone.utc),
+        random_suffix=suffix,
+    )
+    return RunJournal.create(DataLayout(root), context)
+
+
+def test_run_index_snapshots_journals_and_supports_structured_queries(tmp_path: Path) -> None:
+    first = _journal(tmp_path, suffix="abc123")
+    second = _journal(
+        tmp_path,
+        workflow_id="product.daily",
+        business_date="2026-07-26",
+        suffix="def456",
+    )
+    first.finalize("success")
+    second.finalize("failed", error="collector failed")
+    index = RunIndex(tmp_path)
+
+    first_record = index.index_journal(first)
+    second_record = index.index_journal(second.path)
+
+    assert first_record.status == "success"
+    assert second_record.error == "collector failed"
+    assert index.get(first_record.run_id) == first_record
+    assert index.query(status="failed") == (second_record,)
+    assert index.query(workflow_id="shop.weekly") == (first_record,)
+    assert index.query(business_date="2026-07-26") == (second_record,)
+    assert index.query(run_id=first_record.run_id) == (first_record,)
+    assert list((tmp_path / "state" / "ops" / "run-index").glob(".*.tmp")) == []
+
+
+def test_run_index_refreshes_same_run_after_journal_changes(tmp_path: Path) -> None:
+    journal = _journal(tmp_path)
+    index = RunIndex(tmp_path)
+    running = index.index_journal(journal)
+
+    journal.start_step("collect.attempt-1", attempt=1)
+    journal.finish_step("collect.attempt-1", status="success", exit_code=0)
+    journal.finalize("success")
+    complete = index.index_journal(journal)
+
+    assert running.run_id == complete.run_id
+    assert complete.status == "success"
+    assert complete.step_counts == {"success": 1, "failed": 0, "skipped": 0, "running": 0}
+    assert len(tuple((tmp_path / "state" / "ops" / "run-index").glob("*.json"))) == 1
+
+
+def test_run_index_rejects_malformed_journal_without_writing_index(tmp_path: Path) -> None:
+    malformed = tmp_path / "malformed.json"
+    malformed.write_text('{"run_id": "unsafe/../id"}', encoding="utf-8")
+    index = RunIndex(tmp_path)
+
+    with pytest.raises(ValueError, match="journal"):
+        index.index_journal(malformed)
+
+    assert index.query() == ()
+
+
+def test_outbox_enqueue_is_atomic_and_idempotent(tmp_path: Path) -> None:
+    outbox = Outbox(tmp_path)
+
+    first = outbox.enqueue(
+        idempotency_key="run-001:feishu-report",
+        topic="feishu.report",
+        payload={"report_id": "report-001"},
+        run_id="run-001",
+    )
+    duplicate = outbox.enqueue(
+        idempotency_key="run-001:feishu-report",
+        topic="feishu.report",
+        payload={"report_id": "report-001"},
+        run_id="run-001",
+    )
+
+    assert duplicate == first
+    assert first.status == "pending"
+    assert outbox.get(first.message_id) == first
+    assert outbox.list(status="pending") == (first,)
+    assert len(tuple((tmp_path / "state" / "ops" / "outbox" / "messages").glob("*.json"))) == 1
+    assert list((tmp_path / "state" / "ops" / "outbox" / "messages").glob(".*.tmp")) == []
+
+
+def test_outbox_rejects_reusing_key_for_different_semantics(tmp_path: Path) -> None:
+    outbox = Outbox(tmp_path)
+    outbox.enqueue(
+        idempotency_key="run-001:db-upsert",
+        topic="db.upsert",
+        payload={"row_id": 1},
+        run_id="run-001",
+    )
+
+    with pytest.raises(IdempotencyConflict, match="idempotency key"):
+        outbox.enqueue(
+            idempotency_key="run-001:db-upsert",
+            topic="db.upsert",
+            payload={"row_id": 2},
+            run_id="run-001",
+        )
+
+
+def test_outbox_idempotency_compares_normalized_json_semantics(tmp_path: Path) -> None:
+    outbox = Outbox(tmp_path)
+    first = outbox.enqueue(
+        idempotency_key="run-001:normalized",
+        topic="sink.publish",
+        payload={"row_ids": [1, 2]},
+        run_id="run-001",
+    )
+
+    duplicate = outbox.enqueue(
+        idempotency_key="run-001:normalized",
+        topic="sink.publish",
+        payload={"row_ids": (1, 2)},
+        run_id="run-001",
+    )
+
+    assert duplicate == first
+
+
+def test_outbox_state_transitions_and_safe_replay(tmp_path: Path) -> None:
+    outbox = Outbox(tmp_path)
+    pending = outbox.enqueue(
+        idempotency_key="run-001:notify",
+        topic="notification.official",
+        payload={"event": "complete"},
+        run_id="run-001",
+    )
+
+    failed = outbox.mark_failed(pending.message_id, error="temporary outage")
+    replayed = outbox.replay(failed.message_id)
+    sent = outbox.mark_sent(replayed.message_id)
+
+    assert failed.status == "failed"
+    assert failed.attempts == 1
+    assert replayed.status == "pending"
+    assert replayed.message_id == pending.message_id
+    assert replayed.idempotency_key == pending.idempotency_key
+    assert sent.status == "sent"
+    assert outbox.replay(sent.message_id) == sent
+    with pytest.raises(ValueError, match="sent"):
+        outbox.mark_failed(sent.message_id, error="must not regress")
+
+
+def test_outbox_dispatch_never_redelivers_sent_and_retries_with_same_key(tmp_path: Path) -> None:
+    outbox = Outbox(tmp_path)
+    pending = outbox.enqueue(
+        idempotency_key="run-001:publish",
+        topic="sink.publish",
+        payload={"artifact_id": "artifact-001"},
+        run_id="run-001",
+    )
+    received_keys: list[str] = []
+
+    def fail_once(message: object) -> None:
+        key = getattr(message, "idempotency_key")
+        received_keys.append(key)
+        if len(received_keys) == 1:
+            raise RuntimeError("temporary")
+
+    failed = outbox.dispatch(pending.message_id, fail_once)
+    assert failed.status == "failed"
+    assert "RuntimeError" in (failed.error or "")
+
+    outbox.replay(failed.message_id)
+    sent = outbox.dispatch(failed.message_id, fail_once)
+    sent_again = outbox.dispatch(sent.message_id, fail_once)
+
+    assert sent.status == "sent"
+    assert sent_again == sent
+    assert received_keys == [pending.idempotency_key, pending.idempotency_key]
+
+
+def test_outbox_rejects_tampered_state_invariants(tmp_path: Path) -> None:
+    outbox = Outbox(tmp_path)
+    message = outbox.enqueue(
+        idempotency_key="run-001:tamper",
+        topic="audit.write",
+        payload={},
+        run_id="run-001",
+    )
+    path = outbox.message_path(message.message_id)
+    raw = json.loads(path.read_text(encoding="utf-8"))
+    raw["status"] = "sent"
+    raw["sent_at"] = None
+    path.write_text(json.dumps(raw), encoding="utf-8")
+
+    with pytest.raises(ValueError, match="invalid outbox message"):
+        outbox.get(message.message_id)
+
+
+def test_outbox_files_are_structured_and_do_not_store_callable_results(tmp_path: Path) -> None:
+    outbox = Outbox(tmp_path)
+    message = outbox.enqueue(
+        idempotency_key="run-001:audit",
+        topic="audit.write",
+        payload={"value": 1},
+        run_id="run-001",
+    )
+
+    raw = json.loads(outbox.message_path(message.message_id).read_text(encoding="utf-8"))
+
+    assert raw["schema_version"] == 1
+    assert raw["status"] == "pending"
+    assert raw["payload"] == {"value": 1}
+    assert set(raw) == {
+        "schema_version",
+        "message_id",
+        "idempotency_key",
+        "topic",
+        "payload",
+        "run_id",
+        "status",
+        "attempts",
+        "created_at",
+        "updated_at",
+        "sent_at",
+        "error",
+    }
diff --git a/tests/test_product_commerce_module.py b/tests/test_product_commerce_module.py
new file mode 100644
index 0000000..96c35bc
--- /dev/null
+++ b/tests/test_product_commerce_module.py
@@ -0,0 +1,42 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+from gyxx_flow.adapters.native import DeferredModuleCommandStep
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.modules.product_commerce import (
+    PRODUCT_MANUAL_WORKFLOW_IDS,
+    PRODUCT_RESOURCE,
+    PRODUCT_SCHEDULED_WORKFLOW_IDS,
+    PRODUCT_TIMEOUT_SECONDS,
+    PRODUCT_WORKFLOW_IDS,
+    ProductCommerceModule,
+)
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+
+
+def test_product_module_registers_all_available_entries() -> None:
+    module = ProductCommerceModule.from_catalog(
+        WorkflowCatalog.load(PROJECT_ROOT / "config")
+    )
+
+    assert tuple(item.workflow_id for item in module.workflow_definitions()) == PRODUCT_WORKFLOW_IDS
+    assert len(PRODUCT_SCHEDULED_WORKFLOW_IDS) == 7
+    assert len(PRODUCT_MANUAL_WORKFLOW_IDS) == 3
+
+
+def test_product_workflows_use_project_owned_module_commands() -> None:
+    module = ProductCommerceModule.from_catalog(
+        WorkflowCatalog.load(PROJECT_ROOT / "config")
+    )
+
+    for definition in module.workflow_definitions():
+        step = definition.steps[0]
+        assert step.step_id == "module_execute"
+        assert isinstance(step.action, DeferredModuleCommandStep)
+        assert step.timeout_seconds == PRODUCT_TIMEOUT_SECONDS == 6 * 60 * 60
+        assert step.max_attempts == 1
+        assert step.resources == (PRODUCT_RESOURCE,) == ("module:product_commerce",)
+        assert step.production_sink is True
+
diff --git a/tests/test_product_runtime_data_boundaries.py b/tests/test_product_runtime_data_boundaries.py
new file mode 100644
index 0000000..f18fe6e
--- /dev/null
+++ b/tests/test_product_runtime_data_boundaries.py
@@ -0,0 +1,148 @@
+from __future__ import annotations
+
+import importlib.util
+import sys
+from pathlib import Path
+
+import pytest
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+RUNTIME_ROOT = (
+    PROJECT_ROOT
+    / "src"
+    / "gyxx_flow"
+    / "modules"
+    / "product_commerce"
+    / "runtime"
+)
+
+
+def _load_runtime_paths(monkeypatch: pytest.MonkeyPatch, data_root: Path):
+    monkeypatch.setenv("GYXX_DATA_ROOT", str(data_root))
+    monkeypatch.setenv("GYXX_MODULE_ROOT", str(RUNTIME_ROOT))
+    module_path = RUNTIME_ROOT / "runtime_paths.py"
+    name = "_gyxx_product_runtime_paths_boundaries"
+    spec = importlib.util.spec_from_file_location(name, module_path)
+    assert spec and spec.loader
+    module = importlib.util.module_from_spec(spec)
+    sys.modules[name] = module
+    try:
+        spec.loader.exec_module(module)
+    finally:
+        sys.modules.pop(name, None)
+    return module
+
+
+def test_managed_data_path_rejects_writes_outside_data_home(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    paths = _load_runtime_paths(monkeypatch, tmp_path)
+
+    assert paths.managed_data_path("data/raw/product_commerce/run.json") == (
+        tmp_path / "data" / "raw" / "product_commerce" / "run.json"
+    )
+    assert paths.managed_data_path(paths.TMP_ROOT / "batch.json") == (
+        tmp_path / "tmp" / "product_commerce" / "batch.json"
+    )
+    with pytest.raises(ValueError, match="outside GYXX_DATA_ROOT"):
+        paths.managed_data_path(tmp_path.parent / "escape.json")
+    with pytest.raises(ValueError, match="outside GYXX_DATA_ROOT"):
+        paths.managed_data_path("../escape.json")
+
+
+def test_artifact_relative_path_is_relative_to_data_home(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    paths = _load_runtime_paths(monkeypatch, tmp_path)
+
+    artifact = paths.RAW_DATA_ROOT / "jd" / "record.json"
+    assert paths.artifact_relative_path(artifact) == (
+        "data/raw/product_commerce/jd/record.json"
+    )
+    with pytest.raises(ValueError, match="outside GYXX_DATA_ROOT"):
+        paths.artifact_relative_path(RUNTIME_ROOT / "record.json")
+
+
+@pytest.mark.parametrize(
+    ("relative", "forbidden"),
+    [
+        ("collect_retry_utils.py", 'Path("data/logs")'),
+        ("rebuild_market_rank_documents.py", 'PROJECT_ROOT / "data'),
+        ("scripts/insert_main_image_records.py", 'PROJECT_ROOT / "_tmp_'),
+        ("jd_product_data_collector.py", 'os.path.join(PROJECT_ROOT, "debug"'),
+        ("vendors/jd-data-flow/jd_data_collector.py", "os.getcwd()"),
+        (
+            "vendors/jd-data-flow/jd_peer_product_data_collector.py",
+            "os.path.dirname(os.path.abspath(__file__)), \"data\"",
+        ),
+    ],
+)
+def test_runtime_writers_do_not_fall_back_to_cwd_or_source_tree(
+    relative: str, forbidden: str
+) -> None:
+    text = (RUNTIME_ROOT / relative).read_text(encoding="utf-8-sig")
+    assert forbidden not in text
+
+
+@pytest.mark.parametrize(
+    "relative",
+    [
+        "collect_jd_market_rank.py",
+        "collect_sycm_market_rank.py",
+        "import_product_reviews.py",
+        "scripts/insert_jd_main_image_records.py",
+        "scripts/insert_main_image_records.py",
+    ],
+)
+def test_runtime_artifact_references_do_not_assume_source_tree_parent(
+    relative: str,
+) -> None:
+    text = (RUNTIME_ROOT / relative).read_text(encoding="utf-8-sig")
+    assert ".relative_to(PROJECT_ROOT)" not in text
+
+
+def test_checkpoints_and_operational_json_use_state_and_log_roots() -> None:
+    retry_text = (RUNTIME_ROOT / "collect_retry_utils.py").read_text(
+        encoding="utf-8-sig"
+    )
+    assert "LOG_ROOT / \"failures\"" in retry_text
+
+    for relative in (
+        "collect_dy_market_rank.py",
+        "collect_jd_market_rank.py",
+        "collect_sycm_market_rank.py",
+    ):
+        text = (RUNTIME_ROOT / relative).read_text(encoding="utf-8-sig")
+        assert "STATE_ROOT" in text
+        assert "CHECKPOINT_PATH = STATE_ROOT" in text
+
+    decline = (RUNTIME_ROOT / "check_nine_day_decline.py").read_text(
+        encoding="utf-8-sig"
+    )
+    insert = (RUNTIME_ROOT / "insert_bitable_records.py").read_text(
+        encoding="utf-8-sig"
+    )
+    assert "LOG_PATH = LOG_ROOT" in decline
+    assert "failures_path = LOG_ROOT" in insert
+
+
+@pytest.mark.parametrize(
+    "relative",
+    [
+        "aggregate_daily_final.py",
+        "collect_erp_yesterday_metrics.py",
+        "check_nine_day_decline.py",
+        "jd_main_image_collector.py",
+        "jd_self_inventory_sales_collector.py",
+        "jd_product_data_collector.py",
+        "taobao_dmp_item_crowd_insight_screenshots.py",
+        "taobao_wanxiang_ai_creative_report.py",
+        "scripts/insert_jd_main_image_records.py",
+        "scripts/insert_main_image_records.py",
+        "vendors/jd-data-flow/jd_data_collector.py",
+        "vendors/jd-data-flow/jd_product_data_collector.py",
+    ],
+)
+def test_cli_output_overrides_use_the_data_home_guard(relative: str) -> None:
+    text = (RUNTIME_ROOT / relative).read_text(encoding="utf-8-sig")
+    assert "managed_data_path" in text
diff --git a/tests/test_product_source_migration.py b/tests/test_product_source_migration.py
new file mode 100644
index 0000000..605a32e
--- /dev/null
+++ b/tests/test_product_source_migration.py
@@ -0,0 +1,429 @@
+from __future__ import annotations
+
+import ast
+import hashlib
+import importlib.util
+import json
+import re
+from pathlib import Path
+
+from gyxx_flow.security import scan_repository
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+RUNTIME_ROOT = (
+    PROJECT_ROOT
+    / "src"
+    / "gyxx_flow"
+    / "modules"
+    / "product_commerce"
+    / "runtime"
+)
+MANIFEST_PATH = (
+    PROJECT_ROOT / "config" / "source-manifests" / "product_commerce.json"
+)
+
+PRODUCTION_PYTHON = {
+    "aggregate_daily_final.py",
+    "analyze_style_with_hermes.py",
+    "backfill_poseidon_sales.py",
+    "backfill_collect.py",
+    "backfill_one_day.py",
+    "check_nine_day_decline.py",
+    "collect_dy_market_rank.py",
+    "collect_dy_persona_to_bitable.py",
+    "collect_erp_yesterday_metrics.py",
+    "collect_jd_market_rank.py",
+    "collect_jd_persona_to_bitable.py",
+    "collect_persona_to_bitable.py",
+    "collect_retry_utils.py",
+    "collect_sycm_market_rank.py",
+    "config/__init__.py",
+    "config/style_config_loader.py",
+    "db/__init__.py",
+    "db/sync_dim_style.py",
+    "db/sync_sku_master.py",
+    "dy_audience_profile_collect.py",
+    "dy_product_scraping.py",
+    "erp_login_product_analysis.py",
+    "erp_metric_overrides.py",
+    "export_bitable_records.py",
+    "feishu_doc_native.py",
+    "import_product_daily.py",
+    "import_product_reviews.py",
+    "insert_bitable_records.py",
+    "jd_main_image_collector.py",
+    "jd_product_data_collector.py",
+    "jd_self_inventory_sales_collector.py",
+    "lark_cli_runtime.py",
+    "main_image_db.py",
+    "main_image_paths.py",
+    "market_rank_hermes_notification.py",
+    "market_rank_report_archive.py",
+    "orchestrate_daily_collection.py",
+    "orchestrate_market_rank_collection.py",
+    "orchestrate_review_collection.py",
+    "reapply_erp_override.py",
+    "rebuild_market_rank_documents.py",
+    "run_alerts_with_retry.py",
+    "run_daily_persona.py",
+    "run_weekly_jd_main_image.py",
+    "run_weekly_main_image.py",
+    "scripts/insert_jd_main_image_records.py",
+    "scripts/insert_main_image_records.py",
+    "taobao_dmp_item_crowd_insight_screenshots.py",
+    "taobao_sycm_collect.py",
+    "taobao_sycm_collect_backfill.py",
+    "taobao_sycm_products.py",
+    "taobao_wanxiang_ai_creative_report.py",
+    "upload_video_to_guanghe.py",
+    "vendors/dy-data-flow/adaptive_selectors.py",
+    "vendors/dy-data-flow/dy_store_competitor_store_scraping.py",
+    "vendors/dy-data-flow/dynamic_session_src.py",
+    "vendors/jd-data-flow/collection_progress.py",
+    "vendors/jd-data-flow/config.py",
+    "vendors/jd-data-flow/jd_data_collector.py",
+    "vendors/jd-data-flow/jd_peer_product_data_collector.py",
+    "vendors/jd-data-flow/jd_product_data_collector.py",
+    "vendors/jd-data-flow/state.py",
+    "weekly_aggregate.py",
+}
+
+RESOURCES = {
+    ".env.example",
+    "bitable_main_image_map.json",
+    "bitable_style_map.json",
+    "config/auto-flow-config.example.json",
+    "db/schema.sql",
+    "styles_input.json",
+}
+
+LEGACY_TESTS = {
+    "test_backfill_poseidon_sales.py",
+    "test_db_config.py",
+    "test_dy_market_rank.py",
+    "test_dy_session_reuse.py",
+    "test_erp_metric_overrides.py",
+    "test_erp_no_data_freshness.py",
+    "test_erp_slow_skip_codes.py",
+    "test_feishu_doc_native.py",
+    "test_guanghe_metadata.py",
+    "test_guanghe_store_routing.py",
+    "test_import_product_daily.py",
+    "test_jd_enter_shangzhi.py",
+    "test_jd_market_rank.py",
+    "test_main_image_concurrency.py",
+    "test_market_rank_hermes_notification.py",
+    "test_market_rank_report_archive.py",
+    "test_market_rank_workflow.py",
+    "test_persona_launcher.py",
+    "test_style_analysis_orchestration.py",
+    "test_style_config_loader_erp_source.py",
+    "test_sycm_market_rank.py",
+    "test_tm_persona_recovery.py",
+    "test_wanxiang_report_template.py",
+}
+
+LAUNCHER_REFERENCES = {
+    "run_alerts.bat",
+    "run_daily_collect.bat",
+    "run_daily_import.bat",
+    "run_daily_import.ps1",
+    "run_daily_persona.bat",
+    "run_style_analysis_3d.bat",
+    "run_weekly_jd_main_image.bat",
+    "run_weekly_main_image.bat",
+    "run_weekly_market_rank.bat",
+}
+
+INTENTIONALLY_EXCLUDED = {
+    "debug_canvax_dump.py",
+    "debug_canvax_dump2.py",
+    "debug_erp_filter.py",
+    "debug_jd_main_image_snapshot.py",
+    "inspect_dy_comment_filters.py",
+    "jd_collect_test.py",
+}
+
+SUBPROCESS_TARGETS = {
+    "aggregate_daily_final.py",
+    "analyze_style_with_hermes.py",
+    "check_nine_day_decline.py",
+    "collect_dy_market_rank.py",
+    "collect_dy_persona_to_bitable.py",
+    "collect_erp_yesterday_metrics.py",
+    "collect_jd_market_rank.py",
+    "collect_jd_persona_to_bitable.py",
+    "collect_persona_to_bitable.py",
+    "collect_sycm_market_rank.py",
+    "dy_audience_profile_collect.py",
+    "dy_product_scraping.py",
+    "export_bitable_records.py",
+    "import_product_reviews.py",
+    "insert_bitable_records.py",
+    "jd_main_image_collector.py",
+    "jd_product_data_collector.py",
+    "orchestrate_daily_collection.py",
+    "scripts/insert_jd_main_image_records.py",
+    "scripts/insert_main_image_records.py",
+    "taobao_dmp_item_crowd_insight_screenshots.py",
+    "taobao_sycm_collect.py",
+    "taobao_sycm_collect_backfill.py",
+    "taobao_sycm_products.py",
+    "taobao_wanxiang_ai_creative_report.py",
+}
+
+
+def _sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as handle:
+        for block in iter(lambda: handle.read(1024 * 1024), b""):
+            digest.update(block)
+    return digest.hexdigest()
+
+
+def _load_runtime_paths():
+    path = RUNTIME_ROOT / "runtime_paths.py"
+    spec = importlib.util.spec_from_file_location("product_runtime_paths_test", path)
+    assert spec is not None and spec.loader is not None
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module
+
+
+def test_product_runtime_contains_the_complete_reviewed_source_snapshot() -> None:
+    assert len(PRODUCTION_PYTHON) == 63
+    assert all((RUNTIME_ROOT / relative).is_file() for relative in PRODUCTION_PYTHON)
+    assert all((RUNTIME_ROOT / relative).is_file() for relative in RESOURCES)
+    assert all(
+        (RUNTIME_ROOT / "runtime_tests" / name).is_file() for name in LEGACY_TESTS
+    )
+    assert all(
+        (RUNTIME_ROOT / "launchers_reference" / name).is_file()
+        for name in LAUNCHER_REFERENCES
+    )
+
+
+def test_product_source_manifest_is_complete_and_verifiable() -> None:
+    manifest = json.loads(MANIFEST_PATH.read_text(encoding="utf-8"))
+    assert set(manifest) == {
+        "schema_version",
+        "module",
+        "snapshot",
+        "files",
+        "intentionally_excluded",
+        "transformation_rules",
+    }
+    assert manifest["schema_version"] == 1
+    assert manifest["module"] == "product_commerce"
+    assert manifest["snapshot"] == "current-filesystem"
+    audit = next(
+        rule["metadata"]
+        for rule in manifest["transformation_rules"]
+        if rule["id"] == "migration-audit"
+    )
+    assert audit["source_project"] == "product-collector-analyze-flow"
+    assert audit["snapshot_kind"] == "working_tree"
+    security_review = audit["security_review"]
+    assert security_review["reviewed_source_python_findings"] == 15
+    assert security_review["migrated_findings"] == 14
+    assert security_review["excluded_findings"] == 1
+    assert security_review["target_findings"] == 0
+    assert len(security_review["decisions"]) == 15
+    assert {
+        item["source_relative_path"]
+        for item in manifest["intentionally_excluded"]
+    } == INTENTIONALLY_EXCLUDED
+
+    expected_targets = {
+        *(
+            f"src/gyxx_flow/modules/product_commerce/runtime/{path}"
+            for path in PRODUCTION_PYTHON | RESOURCES
+        ),
+        *(
+            f"src/gyxx_flow/modules/product_commerce/runtime/runtime_tests/{name}"
+            for name in LEGACY_TESTS
+        ),
+        *(
+            f"src/gyxx_flow/modules/product_commerce/runtime/launchers_reference/{name}"
+            for name in LAUNCHER_REFERENCES
+        ),
+    }
+    entries = manifest["files"]
+    assert {item["target_relative_path"] for item in entries} == expected_targets
+    assert len(entries) == len(expected_targets)
+
+    for item in entries:
+        target = PROJECT_ROOT / item["target_relative_path"]
+        assert target.is_file()
+        assert item["category"] in {
+            "production_source",
+            "source_resource",
+            "runtime_resource",
+            "regression_test",
+            "launcher_provenance",
+        }
+        assert isinstance(item["transformed"], bool)
+        assert len(item["source_sha256"]) == 64
+        assert item["target_sha256"] == _sha256(target)
+        if not item["transformed"]:
+            assert item["source_sha256"] == item["target_sha256"]
+
+    generated = audit["generated_files"]
+    assert {item["target_relative_path"] for item in generated} == {
+        "src/gyxx_flow/modules/product_commerce/runtime/runtime_paths.py"
+    }
+    for item in generated:
+        target = PROJECT_ROOT / item["target_relative_path"]
+        assert item["target_sha256"] == _sha256(target)
+    assert manifest["transformation_rules"]
+
+
+def test_migrated_product_snapshot_has_no_plaintext_credentials_or_old_roots() -> None:
+    assert scan_repository(RUNTIME_ROOT) == []
+
+    forbidden = ("E:\\auto-flow", "D:\\product-collector-analyze-flow")
+    for path in RUNTIME_ROOT.rglob("*"):
+        if not path.is_file() or path.suffix.casefold() in {".pyc", ".pyo"}:
+            continue
+        try:
+            text = path.read_text(encoding="utf-8-sig")
+        except (OSError, UnicodeDecodeError):
+            continue
+        assert not any(value.casefold() in text.casefold() for value in forbidden), path
+
+
+def test_product_runtime_paths_are_portable_and_layered(tmp_path, monkeypatch) -> None:
+    monkeypatch.setenv("GYXX_MODULE_ROOT", str(RUNTIME_ROOT))
+    monkeypatch.setenv("GYXX_DATA_ROOT", str(tmp_path))
+    paths = _load_runtime_paths()
+
+    assert paths.MODULE_ROOT == RUNTIME_ROOT.resolve()
+    assert paths.DATA_HOME == tmp_path.resolve()
+    assert paths.RAW_DATA_ROOT == tmp_path / "data" / "raw" / "product_commerce"
+    assert paths.NORMALIZED_DATA_ROOT == (
+        tmp_path / "data" / "normalized" / "product_commerce"
+    )
+    assert paths.CURATED_DATA_ROOT == (
+        tmp_path / "data" / "curated" / "product_commerce"
+    )
+    assert paths.EXPORTS_DATA_ROOT == (
+        tmp_path / "data" / "exports" / "product_commerce"
+    )
+    assert paths.STATE_ROOT == tmp_path / "state" / "product_commerce"
+    assert paths.PROFILE_ROOT == (
+        tmp_path / "state" / "product_commerce" / "browser-profiles"
+    )
+    assert paths.LOG_ROOT == tmp_path / "logs" / "product_commerce"
+    assert paths.TMP_ROOT == tmp_path / "tmp" / "product_commerce"
+
+    assert paths.vendor_source_root("jd") == (
+        RUNTIME_ROOT / "vendors" / "jd-data-flow"
+    ).resolve()
+    assert paths.vendor_source_root("dy") == (
+        RUNTIME_ROOT / "vendors" / "dy-data-flow"
+    ).resolve()
+    for relative in SUBPROCESS_TARGETS:
+        assert paths.runtime_script(relative).is_file()
+
+
+def test_product_runtime_has_no_machine_or_legacy_path_escape_hatches() -> None:
+    forbidden = (
+        "D:\\yingxiaoyunying",
+        "D:\\shop-data-flow",
+        "D:\\product-collector-analyze-flow",
+        "E:\\auto-flow",
+        "D:\\jd-data-flow",
+        "E:\\jd-data-flow",
+        "D:\\dy-data-flow",
+        "E:\\dy-data-flow",
+        "JD_DATA_FLOW_ROOT",
+        "DY_DATA_FLOW_ROOT",
+        "AUTO_FLOW_CONFIG",
+        "C:\\ChromeDebug",
+        "C:\\Users\\Administrator",
+        'PROJECT_ROOT / "data"',
+        'ROOT / "data"',
+        'os.path.join(PROJECT_ROOT, "data"',
+    )
+    for path in RUNTIME_ROOT.rglob("*"):
+        if not path.is_file() or path.suffix.casefold() in {".pyc", ".pyo"}:
+            continue
+        try:
+            text = path.read_text(encoding="utf-8-sig")
+        except (OSError, UnicodeDecodeError):
+            continue
+        assert not any(value.casefold() in text.casefold() for value in forbidden), path
+
+
+def test_product_runtime_has_no_hardcoded_long_hex_keys() -> None:
+    hardcoded_key = re.compile(
+        r"(?is)(?:api[_-]?key|secret|token).{0,160}?[\"'][0-9a-f]{32,}[\"']"
+    )
+    for path in RUNTIME_ROOT.rglob("*.py"):
+        text = path.read_text(encoding="utf-8-sig")
+        assert hardcoded_key.search(text) is None, path
+
+    analyzer = (RUNTIME_ROOT / "analyze_style_with_hermes.py").read_text(
+        encoding="utf-8-sig"
+    )
+    assert 'os.getenv("HERMES_API_KEY")' in analyzer
+    assert "HERMES_API_KEY is required" in analyzer
+
+
+def test_product_core_runtime_has_no_known_f821_regressions() -> None:
+    erp_tree = ast.parse(
+        (RUNTIME_ROOT / "collect_erp_yesterday_metrics.py").read_text(
+            encoding="utf-8-sig"
+        )
+    )
+    assert any(
+        isinstance(node, ast.FunctionDef) and node.name == "_ensure_iframe_alive"
+        for node in erp_tree.body
+    )
+
+    class ScopeNames(ast.NodeVisitor):
+        def __init__(self, root: ast.FunctionDef) -> None:
+            self.root = root
+            self.loaded: set[str] = set()
+            self.stored: set[str] = set()
+
+        def visit_FunctionDef(self, node: ast.FunctionDef) -> None:
+            if node is self.root:
+                self.generic_visit(node)
+
+        def visit_Name(self, node: ast.Name) -> None:
+            if isinstance(node.ctx, ast.Load):
+                self.loaded.add(node.id)
+            elif isinstance(node.ctx, ast.Store):
+                self.stored.add(node.id)
+
+    jd_paths = (
+        RUNTIME_ROOT / "jd_product_data_collector.py",
+        RUNTIME_ROOT / "vendors" / "jd-data-flow" / "jd_product_data_collector.py",
+    )
+    for path in jd_paths:
+        tree = ast.parse(path.read_text(encoding="utf-8-sig"))
+        implementations = [
+            node
+            for node in tree.body
+            if isinstance(node, ast.FunctionDef)
+            and node.name == "collect_each_spu_detail"
+        ]
+        assert implementations
+        for implementation in implementations:
+            names = ScopeNames(implementation)
+            names.visit(implementation)
+            assert "week_range" not in names.loaded or "week_range" in names.stored
+
+
+def test_dynamic_session_snapshot_is_classified_as_source_resource() -> None:
+    manifest = json.loads(MANIFEST_PATH.read_text(encoding="utf-8"))
+    item = next(
+        item
+        for item in manifest["files"]
+        if item["source_relative_path"]
+        == "vendors/dy-data-flow/dynamic_session_src.py"
+    )
+    assert item["category"] == "source_resource"
+    assert item["lint_policy"] == "not_a_standalone_module"
+    assert item["reason"]
diff --git a/tests/test_relocated_project.py b/tests/test_relocated_project.py
new file mode 100644
index 0000000..84ac221
--- /dev/null
+++ b/tests/test_relocated_project.py
@@ -0,0 +1,81 @@
+from __future__ import annotations
+
+import json
+import os
+import shutil
+import subprocess
+import sys
+from pathlib import Path
+
+import pytest
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+
+
+def _run_relocated(
+    relocated: Path, data_root: Path, *args: str
+) -> subprocess.CompletedProcess[str]:
+    env = os.environ.copy()
+    for name in tuple(env):
+        if name.startswith("GYXX_LEGACY_"):
+            env.pop(name)
+    env["PYTHONPATH"] = str(relocated / "src")
+    env["GYXX_DATA_ROOT"] = str(data_root)
+    return subprocess.run(
+        [sys.executable, "-m", "gyxx_flow", *args],
+        cwd=relocated,
+        env=env,
+        text=True,
+        encoding="utf-8",
+        capture_output=True,
+        check=False,
+        timeout=60,
+    )
+
+
+@pytest.fixture
+def relocated_project(tmp_path: Path) -> Path:
+    relocated = tmp_path / "moved-gyxx-flow"
+    shutil.copytree(PROJECT_ROOT / "src", relocated / "src")
+    shutil.copytree(PROJECT_ROOT / "config", relocated / "config")
+    return relocated
+
+
+def test_catalog_and_scripts_work_after_project_and_data_roots_move(
+    relocated_project: Path, tmp_path: Path
+) -> None:
+    data_root = tmp_path / "independent-data"
+
+    workflows = _run_relocated(relocated_project, data_root, "list", "--json")
+    scripts = _run_relocated(
+        relocated_project, data_root, "scripts", "list", "--json"
+    )
+
+    assert workflows.returncode == 0, workflows.stderr
+    assert scripts.returncode == 0, scripts.stderr
+    assert len(json.loads(workflows.stdout)) == 29
+    assert len(json.loads(scripts.stdout)) >= 100
+
+
+@pytest.mark.parametrize(
+    "workflow_id",
+    (
+        "content.metrics.daily",
+        "product.daily",
+        "shop.metrics.weekly",
+        "supply.purchase_confirmation.daily",
+    ),
+)
+def test_each_relocated_module_dry_runs_without_old_projects(
+    relocated_project: Path, tmp_path: Path, workflow_id: str
+) -> None:
+    result = _run_relocated(
+        relocated_project,
+        tmp_path / "independent-data",
+        "run",
+        workflow_id,
+        "--date",
+        "2026-07-27",
+    )
+
+    assert result.returncode == 0, result.stderr
diff --git a/tests/test_runtime_integrations.py b/tests/test_runtime_integrations.py
new file mode 100644
index 0000000..ab0598c
--- /dev/null
+++ b/tests/test_runtime_integrations.py
@@ -0,0 +1,252 @@
+from __future__ import annotations
+
+import json
+import sys
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.adapters import (
+    BrowserCookieStore,
+    RuntimeIntegrationCatalog,
+    RuntimeIntegrationError,
+    RuntimeServicePolicy,
+)
+from gyxx_flow.adapters.bootstrap import _rewrite_browser_arguments
+from gyxx_flow.adapters.native import ModuleCommandAdapter, ModuleSourceRoots
+from gyxx_flow.catalog import WorkflowEntry
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.script_catalog import ScriptCatalog, ScriptEntry
+
+
+def _catalog(tmp_path: Path) -> ScriptCatalog:
+    root = tmp_path / "runtime"
+    root.mkdir()
+    first = root / "first.py"
+    second = root / "nested" / "second.py"
+    second.parent.mkdir()
+    first.write_text("print('first')\n", encoding="utf-8")
+    second.write_text("print('second')\n", encoding="utf-8")
+    return ScriptCatalog(
+        (
+            ScriptEntry("demo:first.py", "demo", "first.py", "python", first),
+            ScriptEntry(
+                "demo:nested/second.py",
+                "demo",
+                "nested/second.py",
+                "python",
+                second,
+            ),
+        )
+    )
+
+
+def _write_config(path: Path) -> None:
+    path.write_text(
+        json.dumps(
+            {
+                "schema_version": 1,
+                "cdp_host": "127.0.0.1",
+                "scripts": {
+                    "demo:first.py": 22001,
+                    "demo:nested/second.py": 22002,
+                },
+                "services": {
+                    "feishu": "legacy",
+                    "postgres": "cloud",
+                    "hermes": "local",
+                    "hermes_url": "http://127.0.0.1:8642/v1/chat/completions",
+                },
+            }
+        ),
+        encoding="utf-8",
+    )
+
+
+def test_project_registry_covers_every_script_with_unique_stable_port() -> None:
+    project_root = Path(__file__).resolve().parents[1]
+    scripts = ScriptCatalog.discover_default()
+    first = RuntimeIntegrationCatalog.load(
+        project_root / "config" / "runtime-bindings.json",
+        scripts=scripts,
+        data_root=project_root / "var",
+    )
+    second = RuntimeIntegrationCatalog.load(
+        project_root / "config" / "runtime-bindings.json",
+        scripts=scripts,
+        data_root=project_root / "var",
+    )
+
+    assert len(scripts.script_ids) == 131
+    bindings = [first.binding_for(script_id) for script_id in scripts.script_ids]
+    assert len({binding.cdp_port for binding in bindings}) == 131
+    assert all(22000 <= binding.cdp_port <= 22999 for binding in bindings)
+    assert all(binding.cdp_url.startswith("http://127.0.0.1:") for binding in bindings)
+    assert bindings == [second.binding_for(item.script_id) for item in bindings]
+
+
+def test_bindings_isolate_browser_state_and_relocate_with_data_root(tmp_path: Path) -> None:
+    scripts = _catalog(tmp_path)
+    config = tmp_path / "bindings.json"
+    _write_config(config)
+    data_root = tmp_path / "portable-data"
+    catalog = RuntimeIntegrationCatalog.load(config, scripts=scripts, data_root=data_root)
+
+    first = catalog.binding_for("demo:first.py")
+    second = catalog.binding_for("demo:nested/second.py")
+    assert first.profile_dir != second.profile_dir
+    assert first.cookie_file != second.cookie_file
+    assert first.storage_state_file != second.storage_state_file
+    for path in (
+        first.profile_dir,
+        first.cookie_file,
+        first.storage_state_file,
+        second.profile_dir,
+    ):
+        assert path.is_relative_to(data_root.resolve())
+    assert not first.profile_dir.exists()
+
+
+def test_parent_and_nested_child_receive_different_runtime_environment(tmp_path: Path) -> None:
+    scripts = _catalog(tmp_path)
+    config = tmp_path / "bindings.json"
+    _write_config(config)
+    catalog = RuntimeIntegrationCatalog.load(config, scripts=scripts, data_root=tmp_path / "data")
+
+    parent = catalog.environment_for("demo:first.py", {"KEEP": "yes"})
+    child = catalog.environment_for("demo:nested/second.py", parent)
+
+    assert child["KEEP"] == "yes"
+    assert parent["GYXX_SCRIPT_ID"] == "demo:first.py"
+    assert child["GYXX_SCRIPT_ID"] == "demo:nested/second.py"
+    assert parent["GYXX_BROWSER_CDP_PORT"] != child["GYXX_BROWSER_CDP_PORT"]
+    assert parent["GYXX_BROWSER_PROFILE_DIR"] != child["GYXX_BROWSER_PROFILE_DIR"]
+    assert parent["GYXX_BROWSER_COOKIE_FILE"] != child["GYXX_BROWSER_COOKIE_FILE"]
+
+
+def test_module_command_adapter_injects_entry_script_binding(tmp_path: Path) -> None:
+    scripts = _catalog(tmp_path)
+    config = tmp_path / "bindings.json"
+    _write_config(config)
+    catalog = RuntimeIntegrationCatalog.load(config, scripts=scripts, data_root=tmp_path / "data")
+    runtime_root = tmp_path / "runtime"
+    adapter = ModuleCommandAdapter(
+        ModuleSourceRoots({"demo": runtime_root}),
+        base_env={"KEEP": "yes"},
+        project_root=tmp_path,
+        data_root=tmp_path / "data",
+        integration_catalog=catalog,
+    )
+    entry = WorkflowEntry("demo.workflow", "demo", "manual", "first.py")
+    context = RunContext.create("demo.workflow", "2026-07-27")
+
+    command = adapter.build(entry, context=context)
+
+    assert command.env["KEEP"] == "yes"
+    assert command.env["GYXX_SCRIPT_ID"] == "demo:first.py"
+    assert command.env["GYXX_BROWSER_CDP_PORT"] == "22001"
+
+
+def test_cookie_store_round_trip_and_storage_state_are_atomic(tmp_path: Path) -> None:
+    store = BrowserCookieStore(
+        cookie_file=tmp_path / "cookies" / "cookies.json",
+        storage_state_file=tmp_path / "cookies" / "storage_state.json",
+    )
+    cookies = [{"name": "session", "value": "secret-sentinel", "domain": ".example.test", "path": "/"}]
+    state = {"cookies": cookies, "origins": []}
+
+    assert store.load_cookies() == []
+    assert store.load_storage_state() is None
+    store.save_cookies(cookies)
+    store.save_storage_state(state)
+
+    assert store.load_cookies() == cookies
+    assert store.load_storage_state() == state
+    assert not list((tmp_path / "cookies").glob("*.tmp"))
+    assert "secret-sentinel" not in repr(store)
+
+
+def test_service_policy_preserves_original_feishu_and_cloud_db_and_local_hermes() -> None:
+    policy = RuntimeServicePolicy(
+        hermes_url="http://127.0.0.1:8642/v1/chat/completions"
+    )
+    original = {
+        "LARK_PROFILE": "original-profile",
+        "PG_HOST": "cloud-db.example.test",
+        "PG_PASSWORD": "placeholder",
+        "DB_HOST": "",
+        "CUSTOM": "keep",
+    }
+
+    result = policy.apply(original)
+
+    assert result["LARK_PROFILE"] == "original-profile"
+    assert result["PG_HOST"] == "cloud-db.example.test"
+    assert result["PG_PASSWORD"] == "placeholder"
+    assert result["DB_HOST"] == "cloud-db.example.test"
+    assert result["AUTOFLOW_PG_HOST"] == "cloud-db.example.test"
+    assert result["GYXX_FEISHU_MODE"] == "legacy"
+    assert result["GYXX_POSTGRES_MODE"] == "cloud"
+    assert result["GYXX_HERMES_MODE"] == "local"
+    assert result["HERMES_ANALYZER_URL"].startswith("http://127.0.0.1:")
+    assert "secret-sentinel" not in repr(policy)
+
+
+@pytest.mark.parametrize(
+    ("environment", "message"),
+    [
+        ({"PG_HOST": "127.0.0.1"}, "cloud"),
+        ({"DB_HOST": "localhost"}, "cloud"),
+        ({"AUTOFLOW_PG_HOST": "::1"}, "cloud"),
+        ({"HERMES_ANALYZER_URL": "https://remote.example.test/v1"}, "local"),
+        ({"ANALYZER_API_SERVER_URL": "http://10.0.0.8:8642/v1"}, "local"),
+    ],
+)
+def test_service_policy_rejects_local_database_or_remote_hermes(
+    environment: dict[str, str], message: str
+) -> None:
+    with pytest.raises(RuntimeIntegrationError, match=message):
+        RuntimeServicePolicy().apply(environment)
+
+
+def test_registry_rejects_missing_duplicate_or_unknown_script_allocations(tmp_path: Path) -> None:
+    scripts = _catalog(tmp_path)
+    config = tmp_path / "bindings.json"
+    _write_config(config)
+    payload = json.loads(config.read_text(encoding="utf-8"))
+    payload["scripts"]["demo:nested/second.py"] = 22001
+    config.write_text(json.dumps(payload), encoding="utf-8")
+    with pytest.raises(RuntimeIntegrationError, match="unique"):
+        RuntimeIntegrationCatalog.load(config, scripts=scripts, data_root=tmp_path / "data")
+
+    payload["scripts"].pop("demo:nested/second.py")
+    payload["scripts"]["demo:unknown.py"] = 22003
+    config.write_text(json.dumps(payload), encoding="utf-8")
+    with pytest.raises(RuntimeIntegrationError, match="coverage"):
+        RuntimeIntegrationCatalog.load(config, scripts=scripts, data_root=tmp_path / "data")
+
+
+def test_nested_script_browser_arguments_are_rebound(monkeypatch, tmp_path: Path) -> None:
+    scripts = _catalog(tmp_path)
+    config = tmp_path / "bindings.json"
+    _write_config(config)
+    binding = RuntimeIntegrationCatalog.load(
+        config, scripts=scripts, data_root=tmp_path / "data"
+    ).binding_for("demo:nested/second.py")
+    monkeypatch.setattr(
+        "sys.argv",
+        [
+            "second.py",
+            "--user-data-dir",
+            "parent-profile",
+            "--cdp-url=http://127.0.0.1:22001",
+            "--cdp-port",
+            "22001",
+        ],
+    )
+
+    _rewrite_browser_arguments(binding)
+
+    assert sys.argv[2] == str(binding.profile_dir)
+    assert sys.argv[3] == f"--cdp-url={binding.cdp_url}"
+    assert sys.argv[5] == str(binding.cdp_port)
diff --git a/tests/test_runtime_output_boundaries.py b/tests/test_runtime_output_boundaries.py
new file mode 100644
index 0000000..9232b13
--- /dev/null
+++ b/tests/test_runtime_output_boundaries.py
@@ -0,0 +1,47 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+import pytest
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+MODULE_ROOT = PROJECT_ROOT / "src" / "gyxx_flow" / "modules"
+
+
+@pytest.mark.parametrize(
+    ("module", "forbidden"),
+    [
+        ("content_marketing", 'Path("data") / "v2_results"'),
+        ("product_commerce", 'Path("data/logs")'),
+        ("product_commerce", 'SCRIPT_DIR / "data"'),
+        ("product_commerce", 'PROJECT_ROOT / "debug"'),
+        ("product_commerce", "relative_to(PROJECT_ROOT)"),
+        ("shop_intelligence", 'os.path.join(os.getcwd(), "debug"'),
+        ("supply_chain", 'PACKAGE_ROOT / "shared-data"'),
+        ("supply_chain", "~/Downloads/停产产品.xlsx"),
+    ],
+)
+def test_migrated_runtimes_have_no_known_mutable_output_bypasses(
+    module: str, forbidden: str
+) -> None:
+    offenders = []
+    for path in (MODULE_ROOT / module / "runtime").rglob("*.py"):
+        source = path.read_text(encoding="utf-8-sig")
+        if forbidden in source:
+            offenders.append(path.relative_to(PROJECT_ROOT).as_posix())
+
+    assert offenders == [], f"{forbidden!r} bypasses the canonical data boundary"
+
+
+def test_runtime_path_modules_do_not_use_source_local_data_fallbacks() -> None:
+    path_files = [
+        MODULE_ROOT / "content_marketing" / "runtime" / "runtime_paths.py",
+        MODULE_ROOT / "product_commerce" / "runtime" / "runtime_paths.py",
+        MODULE_ROOT / "shop_intelligence" / "runtime" / "paths.py",
+        MODULE_ROOT / "supply_chain" / "runtime" / "paths.py",
+    ]
+    forbidden = ('MODULE_ROOT / ".runtime-data"', 'module_root / "var"', 'PROJECT_ROOT / "data"')
+
+    for path in path_files:
+        source = path.read_text(encoding="utf-8-sig")
+        assert not any(pattern in source for pattern in forbidden), path
diff --git a/tests/test_scheduler.py b/tests/test_scheduler.py
new file mode 100644
index 0000000..a2d2965
--- /dev/null
+++ b/tests/test_scheduler.py
@@ -0,0 +1,147 @@
+from __future__ import annotations
+
+import json
+import subprocess
+import xml.etree.ElementTree as ET
+from datetime import date
+from pathlib import Path
+
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.scheduler import (
+    CurrentScheduledTask,
+    PowerShellCurrentTaskProvider,
+    SchedulerConfig,
+    build_schedule_plan,
+    detect_schedule_drift,
+    write_schedule_plan_bundle,
+)
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+NS = {"t": "http://schemas.microsoft.com/windows/2004/02/mit/task"}
+
+
+def test_windows_provider_reads_managed_tasks_without_mutation(monkeypatch) -> None:
+    captured: dict[str, object] = {}
+
+    def fake_run(command, **kwargs):
+        captured["command"] = command
+        captured["kwargs"] = kwargs
+        return subprocess.CompletedProcess(
+            command,
+            0,
+            stdout=json.dumps(
+                [{"full_name": r"\GYXX\product.daily", "xml": "<Task />"}]
+            ),
+            stderr="",
+        )
+
+    monkeypatch.setattr(subprocess, "run", fake_run)
+    tasks = PowerShellCurrentTaskProvider().current_tasks("\\GYXX\\")
+
+    assert tasks == [CurrentScheduledTask(r"\GYXX\product.daily", "<Task />")]
+    assert captured["command"][0] == "powershell.exe"
+    assert "Register-ScheduledTask" not in captured["command"][-1]
+    assert captured["kwargs"]["shell"] is False
+
+
+def _plan(tmp_path: Path):
+    catalog = WorkflowCatalog.load(PROJECT_ROOT / "config")
+    return build_schedule_plan(
+        catalog,
+        SchedulerConfig(
+            python_executable=Path(r"C:\Portable\Python312\python.exe"),
+            project_root=tmp_path / "portable-project",
+        ),
+        start_date=date(2026, 7, 27),
+    )
+
+
+def test_plan_generates_21_inert_portable_windows_task_definitions(tmp_path: Path) -> None:
+    plan = _plan(tmp_path)
+
+    assert plan.timezone == "Asia/Shanghai"
+    assert len(plan.tasks) == 21
+    assert len({task.full_name for task in plan.tasks}) == 21
+    assert all(task.full_name.startswith("\\GYXX\\") for task in plan.tasks)
+    assert all("Register-ScheduledTask" not in task.xml for task in plan.tasks)
+
+    sample = next(task for task in plan.tasks if task.workflow_id == "product.daily")
+    root = ET.fromstring(sample.xml)
+    assert root.findtext(".//t:Exec/t:Command", namespaces=NS) == str(
+        plan.config.python_executable
+    )
+    assert root.findtext(".//t:Exec/t:Arguments", namespaces=NS) == (
+        "-m gyxx_flow run product.daily --scheduled"
+    )
+    assert root.findtext(".//t:Exec/t:WorkingDirectory", namespaces=NS) == str(
+        plan.config.project_root
+    )
+    assert root.findtext(".//t:MultipleInstancesPolicy", namespaces=NS) == "IgnoreNew"
+
+
+def test_plan_preserves_daily_weekly_monthly_and_interval_semantics(tmp_path: Path) -> None:
+    plan = _plan(tmp_path)
+    by_id = {task.workflow_id: ET.fromstring(task.xml) for task in plan.tasks}
+
+    assert by_id["product.daily"].findtext(
+        ".//t:ScheduleByDay/t:DaysInterval", namespaces=NS
+    ) == "1"
+    weekly = by_id["content.comments.weekly"]
+    assert weekly.find(".//t:ScheduleByWeek/t:DaysOfWeek/t:Sunday", NS) is not None
+    monthly = by_id["content.summary.monthly"]
+    assert monthly.findtext(
+        ".//t:ScheduleByMonth/t:DaysOfMonth/t:Day", namespaces=NS
+    ) == "1"
+    interval = by_id["product.style_analysis.interval"]
+    assert interval.findtext(
+        ".//t:ScheduleByDay/t:DaysInterval", namespaces=NS
+    ) == "3"
+    assert interval.findtext(".//t:StartBoundary", namespaces=NS).startswith(
+        "2026-07-25T11:00:00"
+    )
+
+
+def test_bundle_write_is_a_plan_only_and_never_applies_tasks(tmp_path: Path) -> None:
+    plan = _plan(tmp_path)
+    destination = tmp_path / "schedule-plan"
+
+    written = write_schedule_plan_bundle(destination, plan)
+
+    assert len(list((destination / "tasks").glob("*.xml"))) == 21
+    assert written == destination.resolve()
+    payload = json.loads((destination / "plan.json").read_text(encoding="utf-8"))
+    assert payload["apply_required"] is True
+    assert payload["applied"] is False
+    install = (destination / "install.ps1").read_text(encoding="utf-8")
+    assert "Register-ScheduledTask" in install
+    assert "China Standard Time" in install
+    assert "-WhatIf" in install
+    assert "[string]$WorkflowId" in install
+    assert "Apply requires exactly one safe -WorkflowId" in install
+    assert "ForEach-Object" not in install
+
+
+def test_drift_reports_missing_changed_and_managed_extra_without_values(
+    tmp_path: Path,
+) -> None:
+    plan = _plan(tmp_path)
+    first, second, *rest = plan.tasks
+    changed_xml = second.xml.replace("--scheduled", "--changed")
+    current = [
+        CurrentScheduledTask(first.full_name, first.xml),
+        CurrentScheduledTask(second.full_name, changed_xml),
+        *[CurrentScheduledTask(task.full_name, task.xml) for task in rest],
+        CurrentScheduledTask(r"\GYXX\obsolete.workflow", first.xml),
+        CurrentScheduledTask(r"\Unrelated\keep", "secret-value-must-not-appear"),
+    ]
+    current = [item for item in current if item.full_name != rest[-1].full_name]
+
+    report = detect_schedule_drift(plan, current)
+
+    assert report.is_clean is False
+    assert [(item.full_name, item.status) for item in report.items] == [
+        (second.full_name, "changed"),
+        (rest[-1].full_name, "missing"),
+        (r"\GYXX\obsolete.workflow", "extra"),
+    ]
+    assert "secret-value-must-not-appear" not in json.dumps(report.as_dict())
diff --git a/tests/test_scheduler_cli.py b/tests/test_scheduler_cli.py
new file mode 100644
index 0000000..aaf5ee2
--- /dev/null
+++ b/tests/test_scheduler_cli.py
@@ -0,0 +1,85 @@
+from __future__ import annotations
+
+import io
+import json
+from datetime import date
+from pathlib import Path
+
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.cli import main
+from gyxx_flow.core.config import Settings
+from gyxx_flow.scheduler import CurrentScheduledTask
+from gyxx_flow.workflow import WorkflowDefinition
+from gyxx_flow.workflow.registry import WorkflowRegistry
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+
+
+class FakeCurrentTaskProvider:
+    def __init__(self, tasks: list[CurrentScheduledTask]) -> None:
+        self.tasks = tasks
+        self.requested_path: str | None = None
+
+    def current_tasks(self, task_path: str) -> list[CurrentScheduledTask]:
+        self.requested_path = task_path
+        return self.tasks
+
+
+def test_schedule_plan_cli_writes_inert_bundle_and_drift_report(tmp_path: Path) -> None:
+    provider = FakeCurrentTaskProvider([])
+    output = io.StringIO()
+    destination = tmp_path / "generated-plan"
+    settings = Settings(project_root=PROJECT_ROOT, data_root=tmp_path / "data")
+
+    exit_code = main(
+        [
+            "schedule",
+            "plan",
+            "--output",
+            str(destination),
+            "--start-date",
+            "2026-07-27",
+            "--python-executable",
+            str(PROJECT_ROOT / ".venv" / "Scripts" / "python.exe"),
+        ],
+        settings=settings,
+        current_task_provider=provider,
+        stdout=output,
+        stderr=io.StringIO(),
+    )
+
+    assert exit_code == 0
+    payload = json.loads(output.getvalue())
+    assert payload["desired_count"] == 21
+    assert payload["applied"] is False
+    assert payload["drift_count"] == 21
+    assert provider.requested_path == "\\GYXX\\"
+    assert (destination / "install.ps1").is_file()
+    assert (destination / "drift.json").is_file()
+    assert json.loads((destination / "plan.json").read_text(encoding="utf-8"))[
+        "applied"
+    ] is False
+
+
+def test_scheduled_run_uses_shanghai_business_date_and_executes_registered_workflow(
+    monkeypatch, tmp_path: Path
+) -> None:
+    catalog = WorkflowCatalog.load(PROJECT_ROOT / "config")
+    registry = WorkflowRegistry(catalog)
+    registry.register(WorkflowDefinition("product.daily", ()))
+    monkeypatch.setattr("gyxx_flow.cli._today_shanghai", lambda: date(2026, 7, 27))
+    output = io.StringIO()
+
+    exit_code = main(
+        ["run", "product.daily", "--scheduled"],
+        registry=registry,
+        settings=Settings(project_root=PROJECT_ROOT, data_root=tmp_path),
+        stdout=output,
+        stderr=io.StringIO(),
+    )
+
+    assert exit_code == 0
+    payload = json.loads(output.getvalue())
+    assert payload["business_date"] == "2026-07-27"
+    assert payload["dry_run"] is False
+    assert payload["status"] == "success"
diff --git a/tests/test_script_catalog.py b/tests/test_script_catalog.py
new file mode 100644
index 0000000..5d933d4
--- /dev/null
+++ b/tests/test_script_catalog.py
@@ -0,0 +1,93 @@
+from __future__ import annotations
+
+import io
+import json
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.cli import EXIT_SUCCESS, main
+from gyxx_flow.core.config import Settings
+from gyxx_flow.script_catalog import ScriptCatalog, ScriptCatalogError
+
+
+def test_script_catalog_discovers_python_main_and_launchers(tmp_path: Path) -> None:
+    runtime = tmp_path / "content_marketing" / "runtime"
+    (runtime / "tools").mkdir(parents=True)
+    (runtime / "tools" / "run_job.py").write_text(
+        'if __name__ == "__main__":\n    raise SystemExit(0)\n', encoding="utf-8"
+    )
+    (runtime / "library.py").write_text("def helper():\n    return 1\n", encoding="utf-8")
+    (runtime / "daily.bat").write_text("@echo off\n", encoding="utf-8")
+    (runtime / "maint.ps1").write_text("Write-Output ok\n", encoding="utf-8")
+    (runtime / "runtime_tests").mkdir()
+    (runtime / "runtime_tests" / "test_old.py").write_text(
+        'if __name__ == "__main__":\n    pass\n', encoding="utf-8"
+    )
+
+    catalog = ScriptCatalog.discover({"content_marketing": runtime})
+
+    assert catalog.script_ids == (
+        "content_marketing:daily.bat",
+        "content_marketing:maint.ps1",
+        "content_marketing:tools/run_job.py",
+    )
+    assert catalog.get("content_marketing:tools/run_job.py").entry == "tools/run_job.py"
+
+
+def test_default_script_catalog_discovers_every_project_owned_runtime() -> None:
+    catalog = ScriptCatalog.discover_default()
+
+    modules = {item.module for item in catalog.scripts}
+    assert modules == {
+        "content_marketing",
+        "product_commerce",
+        "shop_intelligence",
+        "supply_chain",
+    }
+    assert len(catalog.scripts) > 21
+
+
+@pytest.mark.parametrize("script_id", ("missing", "content_marketing:../run.py"))
+def test_script_catalog_rejects_unknown_or_unsafe_ids(
+    tmp_path: Path, script_id: str
+) -> None:
+    runtime = tmp_path / "content_marketing" / "runtime"
+    runtime.mkdir(parents=True)
+    catalog = ScriptCatalog.discover({"content_marketing": runtime})
+
+    with pytest.raises(ScriptCatalogError):
+        catalog.get(script_id)
+
+
+def test_cli_lists_and_dry_runs_project_owned_scripts(tmp_path: Path) -> None:
+    output = io.StringIO()
+    settings = Settings(project_root=Path(__file__).parents[1], data_root=tmp_path)
+
+    assert main(["scripts", "list", "--json"], settings=settings, stdout=output) == 0
+    rows = json.loads(output.getvalue())
+    assert len(rows) > 21
+    assert {row["module"] for row in rows} == {
+        "content_marketing",
+        "product_commerce",
+        "shop_intelligence",
+        "supply_chain",
+    }
+
+    output = io.StringIO()
+    exit_code = main(
+        [
+            "scripts",
+            "run",
+            "content_marketing:run_all.py",
+            "--date",
+            "2026-07-27",
+        ],
+        settings=settings,
+        stdout=output,
+    )
+    payload = json.loads(output.getvalue())
+    assert exit_code == EXIT_SUCCESS
+    assert payload["dry_run"] is True
+    assert payload["script_id"] == "content_marketing:run_all.py"
+    assert payload["steps"] == {"module_script": "skipped"}
diff --git a/tests/test_security_scanner.py b/tests/test_security_scanner.py
new file mode 100644
index 0000000..31087e8
--- /dev/null
+++ b/tests/test_security_scanner.py
@@ -0,0 +1,133 @@
+from __future__ import annotations
+
+from dataclasses import asdict
+
+import pytest
+
+
+def test_scan_repository_reports_only_location_and_rule(tmp_path):
+    from gyxx_flow.security.scanner import scan_repository
+
+    first_value = "not-a-real-" + "password-7f31"
+    second_value = "not-a-real-" + "url-secret-91ac"
+    private_key_header = "-----BEGIN " + "PRIVATE KEY-----"
+    source = tmp_path / "src" / "settings.py"
+    source.parent.mkdir()
+    source.write_text(
+        "DEBUG = True\n"
+        f'DATABASE_PASSWORD = "{first_value}"\n'
+        f'endpoint = "https://service-user:{second_value}@example.invalid/api"\n'
+        f'key = "{private_key_header}"\n',
+        encoding="utf-8",
+    )
+
+    findings = scan_repository(tmp_path)
+
+    assert [asdict(finding) for finding in findings] == [
+        {"file": "src/settings.py", "line": 2, "rule": "plaintext-credential"},
+        {"file": "src/settings.py", "line": 3, "rule": "credential-in-url"},
+        {"file": "src/settings.py", "line": 4, "rule": "private-key-material"},
+    ]
+    rendered = repr(findings)
+    assert first_value not in rendered
+    assert second_value not in rendered
+    assert set(asdict(findings[0])) == {"file", "line", "rule"}
+
+
+@pytest.mark.parametrize("directory", ["var", ".git", ".venv", ".learnings"])
+def test_scan_repository_excludes_runtime_and_tool_directories(tmp_path, directory):
+    from gyxx_flow.security.scanner import scan_repository
+
+    value = "not-a-real-" + "excluded-secret-18d2"
+    source = tmp_path / directory / "nested" / "settings.py"
+    source.parent.mkdir(parents=True)
+    source.write_text(f'API_KEY = "{value}"\n', encoding="utf-8")
+
+    assert scan_repository(tmp_path) == []
+
+
+def test_scan_repository_ignores_empty_and_externalized_values(tmp_path):
+    from gyxx_flow.security.scanner import scan_repository
+
+    password_name = "DATABASE_" + "PASSWORD"
+    api_key_name = "API_" + "KEY"
+    client_secret_name = "CLIENT_" + "SECRET"
+    access_token_name = "ACCESS_" + "TOKEN"
+    source = tmp_path / "config" / "settings.env"
+    source.parent.mkdir()
+    source.write_text(
+        f"{password_name}=\n"
+        f"{api_key_name}=${{GYXX_API_KEY}}\n"
+        f"{client_secret_name}=$GYXX_CLIENT_SECRET\n"
+        f"{access_token_name}=os.getenv('GYXX_ACCESS_TOKEN')\n",
+        encoding="utf-8",
+    )
+
+    assert scan_repository(tmp_path) == []
+
+
+def test_scan_repository_detects_common_application_secret_names(tmp_path):
+    from gyxx_flow.security.scanner import scan_repository
+
+    app_secret_name = "FEISHU_APP_" + "SECRET"
+    secret_key_name = "DJANGO_" + "SECRET_KEY"
+    first_value = "not-a-real-" + "application-secret-20f9"
+    second_value = "not-a-real-" + "signing-secret-52bb"
+    source = tmp_path / "config" / "application.env"
+    source.parent.mkdir()
+    source.write_text(
+        f"{app_secret_name}={first_value}\n"
+        f"{secret_key_name}={second_value}\n",
+        encoding="utf-8",
+    )
+
+    assert [asdict(finding) for finding in scan_repository(tmp_path)] == [
+        {"file": "config/application.env", "line": 1, "rule": "plaintext-credential"},
+        {"file": "config/application.env", "line": 2, "rule": "plaintext-credential"},
+    ]
+
+
+def test_scan_repository_ignores_declarations_references_and_comparisons(tmp_path):
+    from gyxx_flow.security.scanner import scan_repository
+
+    password_name = "DATABASE_" + "PASSWORD"
+    api_key_name = "API_" + "KEY"
+    client_secret_name = "CLIENT_" + "SECRET"
+    source = tmp_path / "src" / "model.py"
+    source.parent.mkdir()
+    source.write_text(
+        f"{password_name}: str\n"
+        f"{api_key_name}: SecretStr\n"
+        f"{client_secret_name} = settings.client_secret\n"
+        f"if {password_name} == candidate:\n    pass\n",
+        encoding="utf-8",
+    )
+
+    assert scan_repository(tmp_path) == []
+
+
+def test_scan_repository_rejects_a_non_directory_root(tmp_path):
+    from gyxx_flow.security.scanner import scan_repository
+
+    target = tmp_path / "settings.py"
+    target.write_text("DEBUG = True\n", encoding="utf-8")
+
+    with pytest.raises(NotADirectoryError):
+        scan_repository(target)
+
+
+def test_scan_repository_detects_multiline_environment_secret_default(tmp_path):
+    from gyxx_flow.security.scanner import scan_repository
+
+    source = tmp_path / "collector.py"
+    source.write_text(
+        'API_KEY = os.getenv(\n    "SERVICE_API_KEY",\n'
+        '    "0123456789abcdef0123456789abcdef",\n)\n',
+        encoding="utf-8",
+    )
+
+    findings = scan_repository(tmp_path)
+
+    assert [(item.line, item.rule) for item in findings] == [
+        (1, "hardcoded-long-hex-credential")
+    ]
diff --git a/tests/test_shadow_comparison.py b/tests/test_shadow_comparison.py
new file mode 100644
index 0000000..c960f5f
--- /dev/null
+++ b/tests/test_shadow_comparison.py
@@ -0,0 +1,270 @@
+from __future__ import annotations
+
+import csv
+import json
+from datetime import datetime, timezone
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.migration.shadow import (
+    ComparisonSpec,
+    MetricTolerance,
+    compare_structured_files,
+    write_comparison_report,
+)
+
+FIXED_TIME = datetime(2026, 7, 27, 8, 30, tzinfo=timezone.utc)
+
+
+def _write_jsonl(path: Path, rows: list[dict[str, object]]) -> None:
+    path.write_text(
+        "".join(json.dumps(row, ensure_ascii=False) + "\n" for row in rows),
+        encoding="utf-8",
+    )
+
+
+def _write_csv(path: Path, rows: list[dict[str, object]]) -> None:
+    with path.open("w", encoding="utf-8", newline="") as handle:
+        writer = csv.DictWriter(handle, fieldnames=list(rows[0]))
+        writer.writeheader()
+        writer.writerows(rows)
+
+
+def test_compare_jsonl_to_csv_reports_key_metric_and_error_differences_without_rows(
+    tmp_path: Path,
+) -> None:
+    baseline = tmp_path / "baseline.jsonl"
+    candidate = tmp_path / "candidate.csv"
+    _write_jsonl(
+        baseline,
+        [
+            {
+                "shop_id": "shop-secret-1",
+                "sku": "sku-secret-a",
+                "category": "tops",
+                "revenue": 100.0,
+                "errors": ["upstream token exposed"],
+                "customer_email": "private@example.test",
+            },
+            {
+                "shop_id": "shop-secret-1",
+                "sku": "sku-secret-b",
+                "category": "shoes",
+                "revenue": 20.0,
+                "errors": [],
+                "customer_email": "hidden@example.test",
+            },
+        ],
+    )
+    _write_csv(
+        candidate,
+        [
+            {
+                "shop_id": "shop-secret-1",
+                "sku": "sku-secret-a",
+                "category": "tops",
+                "revenue": "101.5",
+                "errors": "replacement failure",
+                "customer_email": "private@example.test",
+            },
+            {
+                "shop_id": "shop-secret-2",
+                "sku": "sku-secret-c",
+                "category": "accessories",
+                "revenue": "9.0",
+                "errors": "",
+                "customer_email": "do-not-report@example.test",
+            },
+        ],
+    )
+    spec = ComparisonSpec(
+        primary_key=("shop_id", "sku"),
+        metrics={"revenue": MetricTolerance(absolute=1.0, relative=0.0)},
+        error_fields=("errors",),
+        safe_fields=("category",),
+        hash_secret="test-only-hmac-key",
+    )
+
+    report = compare_structured_files(
+        baseline, candidate, spec=spec, generated_at=FIXED_TIME
+    )
+    payload = report.to_dict()
+
+    assert payload["schema_version"] == 1
+    assert payload["status"] == "mismatch"
+    assert payload["generated_at"] == "2026-07-27T08:30:00+00:00"
+    assert payload["sources"] == {
+        "baseline_format": "jsonl",
+        "candidate_format": "csv",
+    }
+    assert payload["summary"] == {
+        "baseline_rows": 2,
+        "candidate_rows": 2,
+        "common_keys": 1,
+        "missing_keys": 1,
+        "extra_keys": 1,
+        "metric_mismatches": 1,
+        "error_set_differences": 2,
+    }
+    assert payload["keys"]["missing"][0]["safe_fields"] == {"category": "shoes"}
+    assert payload["keys"]["extra"][0]["safe_fields"] == {
+        "category": "accessories"
+    }
+    assert len(payload["keys"]["missing"][0]["key_hash"]) == 64
+    assert payload["metrics"]["revenue"]["mismatch_count"] == 1
+    assert payload["metrics"]["revenue"]["max_absolute_delta"] == 1.5
+    assert len(payload["errors"]["only_baseline_hashes"][0]) == 64
+    assert len(payload["errors"]["only_candidate_hashes"][0]) == 64
+
+    rendered = json.dumps(payload, ensure_ascii=False, sort_keys=True)
+    for sensitive in (
+        "shop-secret-1",
+        "shop-secret-2",
+        "sku-secret-a",
+        "sku-secret-b",
+        "sku-secret-c",
+        "upstream token exposed",
+        "replacement failure",
+        "private@example.test",
+        "hidden@example.test",
+        "do-not-report@example.test",
+        "test-only-hmac-key",
+    ):
+        assert sensitive not in rendered
+
+
+def test_metric_values_within_absolute_or_relative_tolerance_match(tmp_path: Path) -> None:
+    baseline = tmp_path / "baseline.json"
+    candidate = tmp_path / "candidate.jsonl"
+    baseline.write_text(
+        json.dumps(
+            {
+                "records": [
+                    {"id": "one", "amount": 100.0},
+                    {"id": "two", "amount": 0.0},
+                ]
+            }
+        ),
+        encoding="utf-8",
+    )
+    _write_jsonl(
+        candidate,
+        [
+            {"id": "one", "amount": 100.5},
+            {"id": "two", "amount": 0.01},
+        ],
+    )
+    spec = ComparisonSpec(
+        primary_key=("id",),
+        metrics={
+            "amount": MetricTolerance(absolute=0.01, relative=0.01),
+        },
+    )
+
+    payload = compare_structured_files(baseline, candidate, spec=spec).to_dict()
+
+    assert payload["status"] == "match"
+    assert payload["metrics"]["amount"]["mismatch_count"] == 0
+    assert payload["summary"]["metric_mismatches"] == 0
+
+
+def test_comparison_rejects_duplicate_or_missing_primary_keys(tmp_path: Path) -> None:
+    baseline = tmp_path / "baseline.jsonl"
+    candidate = tmp_path / "candidate.jsonl"
+    _write_jsonl(baseline, [{"id": "same"}, {"id": "same"}])
+    _write_jsonl(candidate, [{"id": "same"}])
+
+    with pytest.raises(ValueError, match="duplicate primary key"):
+        compare_structured_files(
+            baseline, candidate, spec=ComparisonSpec(primary_key=("id",))
+        )
+
+    _write_jsonl(baseline, [{"name": "missing"}])
+    with pytest.raises(ValueError, match="missing primary key field"):
+        compare_structured_files(
+            baseline, candidate, spec=ComparisonSpec(primary_key=("id",))
+        )
+
+
+def test_comparison_rejects_unsupported_or_malformed_inputs(tmp_path: Path) -> None:
+    bad_extension = tmp_path / "rows.txt"
+    candidate = tmp_path / "candidate.jsonl"
+    bad_extension.write_text("{}\n", encoding="utf-8")
+    _write_jsonl(candidate, [{"id": "1", "amount": 1}])
+    spec = ComparisonSpec(
+        primary_key=("id",), metrics={"amount": MetricTolerance()}
+    )
+
+    with pytest.raises(ValueError, match="unsupported structured data format"):
+        compare_structured_files(bad_extension, candidate, spec=spec)
+
+    malformed = tmp_path / "malformed.jsonl"
+    malformed.write_text('[1, 2, 3]\n', encoding="utf-8")
+    with pytest.raises(ValueError, match="record must be a JSON object"):
+        compare_structured_files(malformed, candidate, spec=spec)
+
+    too_large = tmp_path / "too-large.jsonl"
+    _write_jsonl(too_large, [{"id": "1", "amount": "1e9999"}])
+    with pytest.raises(ValueError, match="exceeds report range"):
+        compare_structured_files(too_large, candidate, spec=spec)
+
+
+def test_difference_samples_are_deterministic_and_bounded(tmp_path: Path) -> None:
+    baseline = tmp_path / "baseline.jsonl"
+    candidate = tmp_path / "candidate.jsonl"
+    _write_jsonl(baseline, [{"id": f"old-{index}"} for index in range(5)])
+    _write_jsonl(candidate, [{"id": f"new-{index}"} for index in range(5)])
+    spec = ComparisonSpec(primary_key=("id",), difference_limit=2)
+
+    payload = compare_structured_files(baseline, candidate, spec=spec).to_dict()
+
+    assert payload["summary"]["missing_keys"] == 5
+    assert payload["keys"]["missing_total"] == 5
+    assert payload["keys"]["missing_truncated"] is True
+    assert len(payload["keys"]["missing"]) == 2
+    assert payload["keys"]["missing"] == sorted(
+        payload["keys"]["missing"], key=lambda item: item["key_hash"]
+    )
+
+
+def test_write_comparison_report_is_atomic_and_machine_readable(tmp_path: Path) -> None:
+    baseline = tmp_path / "baseline.jsonl"
+    candidate = tmp_path / "candidate.jsonl"
+    destination = tmp_path / "reports" / "comparison.json"
+    _write_jsonl(baseline, [{"id": "same"}])
+    _write_jsonl(candidate, [{"id": "same"}])
+    report = compare_structured_files(
+        baseline,
+        candidate,
+        spec=ComparisonSpec(primary_key=("id",)),
+        generated_at=FIXED_TIME,
+    )
+
+    written = write_comparison_report(destination, report)
+
+    assert written == destination
+    assert json.loads(destination.read_text(encoding="utf-8")) == report.to_dict()
+    assert not list(destination.parent.glob(".*.tmp"))
+
+
+@pytest.mark.parametrize(
+    "kwargs, message",
+    [
+        ({"primary_key": ()}, "primary_key"),
+        ({"primary_key": ("id", "id")}, "primary_key"),
+        ({"primary_key": ("id",), "difference_limit": 0}, "difference_limit"),
+    ],
+)
+def test_comparison_spec_validates_declarations(
+    kwargs: dict[str, object], message: str
+) -> None:
+    with pytest.raises(ValueError, match=message):
+        ComparisonSpec(**kwargs)  # type: ignore[arg-type]
+
+
+def test_metric_tolerance_rejects_negative_or_non_finite_values() -> None:
+    with pytest.raises(ValueError, match="absolute"):
+        MetricTolerance(absolute=-0.1)
+    with pytest.raises(ValueError, match="relative"):
+        MetricTolerance(relative=float("inf"))
diff --git a/tests/test_shared_adapters.py b/tests/test_shared_adapters.py
new file mode 100644
index 0000000..a47cb0d
--- /dev/null
+++ b/tests/test_shared_adapters.py
@@ -0,0 +1,93 @@
+from __future__ import annotations
+
+import subprocess
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.adapters import (
+    BrowserProfileManager,
+    FeishuOutboxAdapter,
+    HermesCommandAdapter,
+    PostgresOutboxAdapter,
+)
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.core.locks import ResourceBusyError
+from gyxx_flow.ops import Outbox
+
+
+def test_feishu_and_postgres_writes_share_idempotent_outbox_boundary(tmp_path: Path) -> None:
+    outbox = Outbox(tmp_path)
+    feishu = FeishuOutboxAdapter(outbox)
+    postgres = PostgresOutboxAdapter(outbox)
+
+    first = feishu.upsert(
+        run_id="run-001",
+        operation_key="weekly-report",
+        artifact_id="artifact-report-001",
+    )
+    duplicate = feishu.upsert(
+        run_id="run-001",
+        operation_key="weekly-report",
+        artifact_id="artifact-report-001",
+    )
+    db_message = postgres.upsert(
+        run_id="run-001",
+        operation_key="shop-metrics",
+        artifact_id="artifact-metrics-001",
+    )
+
+    assert duplicate == first
+    assert first.topic == "feishu.upsert"
+    assert db_message.topic == "postgres.upsert"
+    assert first.payload == {"artifact_id": "artifact-report-001"}
+    assert db_message.payload == {"artifact_id": "artifact-metrics-001"}
+    assert first.idempotency_key != db_message.idempotency_key
+
+
+def test_hermes_adapter_builds_explicit_command_and_dry_run_starts_no_process(
+    monkeypatch, tmp_path: Path
+) -> None:
+    prompt = tmp_path / "prompt.md"
+    prompt.write_text("analyze artifact-001", encoding="utf-8")
+    adapter = HermesCommandAdapter(executable="hermes", base_env={"PROFILE": "analysis"})
+    command = adapter.build(prompt_file=prompt, cwd=tmp_path)
+
+    assert command.argv == ("hermes", "-z", "analyze artifact-001")
+    assert command.cwd == tmp_path.resolve()
+    assert command.env == {"PROFILE": "analysis"}
+
+    monkeypatch.setattr(
+        subprocess,
+        "run",
+        lambda *args, **kwargs: pytest.fail("dry-run must not start Hermes"),
+    )
+    context = RunContext.create("product.style", "2026-07-27")
+    outcome = command.execute(context=context, timeout_seconds=60, dry_run=True)
+    assert outcome.skipped is True
+
+
+def test_browser_profiles_are_portable_and_exclusively_locked(tmp_path: Path) -> None:
+    manager = BrowserProfileManager(tmp_path)
+
+    with manager.acquire("shop-compass", owner="run-001") as first:
+        assert first.profile_path == (
+            tmp_path / "state" / "browser_profiles" / "shop-compass"
+        ).resolve()
+        assert first.profile_path.is_dir()
+        with pytest.raises(ResourceBusyError):
+            with manager.acquire(
+                "shop-compass", owner="run-002", timeout_seconds=0
+            ):
+                pass
+
+    with manager.acquire("shop-compass", owner="run-002", timeout_seconds=0) as second:
+        assert second.profile_path == first.profile_path
+
+
+@pytest.mark.parametrize("unsafe", ["../escape", "a/b", "a\\b", "", ".."])
+def test_browser_profile_id_rejects_path_escape(tmp_path: Path, unsafe: str) -> None:
+    with pytest.raises(ValueError, match="profile"):
+        with BrowserProfileManager(tmp_path).acquire(unsafe, owner="run-001"):
+            pass
+
diff --git a/tests/test_shop_intelligence_module.py b/tests/test_shop_intelligence_module.py
new file mode 100644
index 0000000..a30246b
--- /dev/null
+++ b/tests/test_shop_intelligence_module.py
@@ -0,0 +1,38 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+from gyxx_flow.adapters.native import DeferredModuleCommandStep
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.modules.shop_intelligence import (
+    COLLECTION_TIMEOUT_SECONDS,
+    SHOP_RESOURCE,
+    SHOP_WORKFLOW_IDS,
+    ShopIntelligenceModule,
+)
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+
+
+def test_shop_module_registers_both_weekly_workflows() -> None:
+    module = ShopIntelligenceModule.from_catalog(
+        WorkflowCatalog.load(PROJECT_ROOT / "config")
+    )
+
+    assert tuple(item.workflow_id for item in module.workflow_definitions()) == SHOP_WORKFLOW_IDS
+
+
+def test_shop_workflows_use_project_owned_module_commands() -> None:
+    module = ShopIntelligenceModule.from_catalog(
+        WorkflowCatalog.load(PROJECT_ROOT / "config")
+    )
+
+    for definition in module.workflow_definitions():
+        step = definition.steps[0]
+        assert step.step_id == "module_collect"
+        assert isinstance(step.action, DeferredModuleCommandStep)
+        assert step.timeout_seconds == COLLECTION_TIMEOUT_SECONDS == 4 * 60 * 60
+        assert step.max_attempts == 1
+        assert step.resources == (SHOP_RESOURCE,) == ("module:shop_intelligence",)
+        assert step.production_sink is True
+
diff --git a/tests/test_shop_supply_data_routing.py b/tests/test_shop_supply_data_routing.py
new file mode 100644
index 0000000..9685274
--- /dev/null
+++ b/tests/test_shop_supply_data_routing.py
@@ -0,0 +1,240 @@
+from __future__ import annotations
+
+import importlib.util
+import sys
+from datetime import datetime, timezone
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.adapters.native import ModuleCommandAdapter, ModuleSourceRoots
+from gyxx_flow.catalog import WorkflowEntry
+from gyxx_flow.core.context import RunContext
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+MODULES_ROOT = PROJECT_ROOT / "src" / "gyxx_flow" / "modules"
+
+
+def _read(relative: str) -> str:
+    return (PROJECT_ROOT / relative).read_text(encoding="utf-8-sig")
+
+
+def _load_shop_paths(monkeypatch: pytest.MonkeyPatch, data_root: Path):
+    monkeypatch.setenv("GYXX_DATA_ROOT", str(data_root))
+    path = MODULES_ROOT / "shop_intelligence" / "runtime" / "paths.py"
+    name = "_gyxx_test_shop_routing_paths"
+    spec = importlib.util.spec_from_file_location(name, path)
+    assert spec and spec.loader
+    module = importlib.util.module_from_spec(spec)
+    sys.modules[name] = module
+    try:
+        spec.loader.exec_module(module)
+    finally:
+        sys.modules.pop(name, None)
+    return module
+
+
+def test_shop_output_resolver_confines_collectors_to_raw_root(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    paths = _load_shop_paths(monkeypatch, tmp_path / "portable")
+    default = paths.RAW_ROOT / "京东" / "店铺数据"
+
+    assert paths.resolve_raw_output(None, default=default) == default
+    assert paths.resolve_raw_output("京东/竞店数据", default=default) == (
+        paths.RAW_ROOT / "京东" / "竞店数据"
+    )
+    assert paths.resolve_raw_output(
+        str(paths.RAW_ROOT / "京东" / "店铺数据"), default=default
+    ) == (paths.RAW_ROOT / "京东" / "店铺数据")
+
+    with pytest.raises(ValueError, match="RAW_ROOT"):
+        paths.resolve_raw_output(tmp_path / "outside", default=default)
+    with pytest.raises(ValueError, match="RAW_ROOT"):
+        paths.resolve_raw_output("../outside", default=default)
+
+
+def test_shop_debug_and_cli_outputs_use_canonical_roots() -> None:
+    jd_shop = _read(
+        "src/gyxx_flow/modules/shop_intelligence/runtime/collectors/jd_data_collector.py"
+    )
+    jd_peer = _read(
+        "src/gyxx_flow/modules/shop_intelligence/runtime/collectors/"
+        "jd_peer_store_data_collector.py"
+    )
+
+    assert "os.getcwd()" not in jd_shop
+    assert 'TMP_ROOT / "debug" / "jd_date_picker"' in jd_shop
+    assert "resolve_raw_output(args.output" in jd_shop
+    assert "resolve_raw_output(args.output" in jd_peer
+
+
+def test_supply_python_outputs_use_state_raw_tmp_and_export_layers() -> None:
+    feishu = _read(
+        "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/feishu_sheets.py"
+    )
+    workflow = _read(
+        "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/mcp_workflow.py"
+    )
+    monitor = _read(
+        "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/monitor.py"
+    )
+    replenishment = _read(
+        "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/"
+        "ProductReplenishment.py"
+    )
+    confirmation = _read(
+        "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/"
+        "PurchaseConfirmation.py"
+    )
+    batch = _read(
+        "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/"
+        "batch_process.py"
+    )
+    bitable = _read(
+        "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/"
+        "insert_replenishment_bitable.py"
+    )
+
+    assert 'STATE_ROOT / "cache" / "discontinued_skus_feishu.json"' in feishu
+    assert 'STATE_ROOT / "locks"' in workflow
+    assert "STATUS_FILE = STATE_ROOT / \"status.json\"" in monitor
+    assert "PROCESSING_FILE = STATE_ROOT / \"processing_status.json\"" in monitor
+
+    assert "REPLENISHMENT_RAW_ROOT" in replenishment
+    assert "REPLENISHMENT_WORK_ROOT" in replenishment
+    assert "REPLENISHMENT_EXPORT_ROOT" in replenishment
+    assert 'output_dir="data"' not in replenishment
+    assert "os.listdir(raw_dir)" not in replenishment
+    assert "PURCHASE_CONFIRMATION_RAW_ROOT" in confirmation
+    assert "PURCHASE_CONFIRMATION_EXPORT_ROOT" in confirmation
+    assert "Downloads" not in replenishment
+    assert "Downloads" not in batch
+    assert 'glob("run_id=*/pending_insert.json")' in bitable
+
+
+def test_supply_raw_outputs_are_run_scoped_and_never_cleaned_in_place() -> None:
+    workflow = _read(
+        "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/mcp_workflow.py"
+    )
+    for script in (
+        "collect_confirmation.ps1",
+        "collect_purchase_order_update.ps1",
+        "collect_replenishment.ps1",
+    ):
+        source = _read(
+            "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/" + script
+        )
+        assert '"run_id=$workflowId"' in source
+        assert (
+            "Get-ChildItem -Path $sharedDir -File -ErrorAction SilentlyContinue | Remove-Item"
+            not in source
+        )
+        assert (
+            "Get-ChildItem -Path $sharedReplenishmentDir -File -ErrorAction SilentlyContinue | Remove-Item"
+            not in source
+        )
+        assert (
+            "Get-ChildItem -Path $sharedAlertDir -File -ErrorAction SilentlyContinue | Remove-Item"
+            not in source
+        )
+        assert "Remove-Item -LiteralPath $chromeProfileDir" not in source
+
+    clear_body = workflow.split("def _clear_shared_outputs", 1)[1].split(
+        "def _matching_shared_outputs", 1
+    )[0]
+    cleanup_body = workflow.split("def _cleanup_workflow_artifacts", 1)[1].split(
+        "def _verify_bitable_insert_in_pg", 1
+    )[0]
+    assert ".unlink(" not in clear_body
+    assert "SHARED_DIR" not in cleanup_body
+    assert ".rglob(pattern)" in workflow
+    assert 'collector_dir / "data"' not in workflow
+    assert "def _publish_recent_script_outputs" not in workflow
+
+
+def test_purchase_order_trigger_uses_state_task_and_never_cleans_raw_or_source() -> None:
+    trigger = _read(
+        "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/"
+        "trigger_purchase_order_update.py"
+    )
+    collector = _read(
+        "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/"
+        "collect_purchase_order_update.ps1"
+    )
+
+    assert "TASK_FILE = STATE_ROOT" in trigger
+    assert "cleanup_old_artifacts" not in trigger
+    assert ".unlink(" not in trigger
+    assert '$taskFile = Join-Path $stateRoot "tasks\\purchase-order-update\\task.json"' in collector
+
+
+@pytest.mark.parametrize(
+    "script",
+    [
+        "collect_confirmation.ps1",
+        "collect_purchase_order_update.ps1",
+        "collect_replenishment.ps1",
+    ],
+)
+def test_supply_powershell_prefers_injected_module_roots(script: str) -> None:
+    source = _read(
+        "src/gyxx_flow/modules/supply_chain/runtime/orchestrator/scripts/" + script
+    )
+
+    for variable in (
+        "GYXX_SUPPLY_RAW_ROOT",
+        "GYXX_SUPPLY_STATE_ROOT",
+        "GYXX_SUPPLY_EXPORT_ROOT",
+        "GYXX_SUPPLY_WORK_ROOT",
+    ):
+        assert f"$env:{variable}" in source
+    assert 'Join-Path $projectRoot "var"' in source
+    assert 'Join-Path $dataRoot "raw\\supply_chain"' not in source
+    assert 'Join-Path $dataRoot "exports\\supply_chain"' not in source
+
+    if script == "collect_confirmation.ps1":
+        assert "$collectionStartedAt = Get-Date" in source
+        assert 'Get-ChildItem -Path $dataDir -Include "*.csv", "*.xlsx"' not in source
+
+
+def test_native_adapter_injects_supply_module_roots(tmp_path: Path) -> None:
+    runtime = tmp_path / "runtime"
+    runtime.mkdir()
+    (runtime / "collect.ps1").write_text("Write-Output ok", encoding="utf-8")
+    adapter = ModuleCommandAdapter(
+        ModuleSourceRoots({"supply_chain": runtime}),
+        base_env={},
+        project_root=tmp_path,
+        data_root=tmp_path / "portable-data",
+    )
+    entry = WorkflowEntry(
+        workflow_id="supply.collect",
+        module="supply_chain",
+        trigger="manual",
+        entry="collect.ps1",
+        args=(),
+        source_project="supply",
+    )
+    context = RunContext.create(
+        "supply.collect",
+        "2026-07-27",
+        now=datetime(2026, 7, 27, tzinfo=timezone.utc),
+        random_suffix="routing1",
+    )
+
+    command = adapter.build(entry, context=context)
+
+    module_root = tmp_path / "portable-data"
+    assert command.env["GYXX_SUPPLY_RAW_ROOT"] == str(
+        (module_root / "data" / "raw" / "supply_chain").resolve()
+    )
+    assert command.env["GYXX_SUPPLY_STATE_ROOT"] == str(
+        (module_root / "state" / "supply_chain").resolve()
+    )
+    assert command.env["GYXX_SUPPLY_EXPORT_ROOT"] == str(
+        (module_root / "data" / "exports" / "supply_chain").resolve()
+    )
+    assert command.env["GYXX_SUPPLY_WORK_ROOT"] == str(
+        (module_root / "tmp" / "supply_chain").resolve()
+    )
diff --git a/tests/test_shop_supply_source_snapshot.py b/tests/test_shop_supply_source_snapshot.py
new file mode 100644
index 0000000..9dcee4e
--- /dev/null
+++ b/tests/test_shop_supply_source_snapshot.py
@@ -0,0 +1,201 @@
+from __future__ import annotations
+
+import hashlib
+import json
+import re
+from pathlib import Path, PurePosixPath
+
+import pytest
+
+from gyxx_flow.security.scanner import scan_repository
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+MANIFEST_ROOT = PROJECT_ROOT / "config" / "source-manifests"
+SHA256 = re.compile(r"^[0-9a-f]{64}$")
+OLD_ROOTS = (
+    "d:\\shop-data-flow",
+    "d:/shop-data-flow",
+    "e:\\auto-flow",
+    "e:/auto-flow",
+)
+NON_PORTABLE_RUNTIME_PATTERNS = (
+    re.compile(r"e:[\\/]+auto-flow", re.IGNORECASE),
+    re.compile(r"d:[\\/]+erp", re.IGNORECASE),
+    re.compile(r"path\s*\(\s*['\"]d:[\\/]+['\"]\s*\)\.iterdir", re.IGNORECASE),
+    re.compile(r"sys\.path\.insert\s*\(", re.IGNORECASE),
+)
+EXPECTED = {
+    "shop_intelligence": {
+        "minimum_files": 22,
+        "required": {
+            "adaptive_selectors.py",
+            "collectors/jd_data_collector.py",
+            "collectors/jd_peer_store_data_collector.py",
+            "collectors/dy_store_competitor_store_scraping.py",
+            "collectors/taobao_sycm.py",
+            "runners/run_shop.py",
+            "runners/run_peer_store.py",
+            "writers/shop_base_writer.py",
+            "writers/peer_store_writer.py",
+            "db/db.py",
+            "db/schema.sql",
+            "scripts/setup_scheduler.ps1",
+        },
+        "classifications": {"scheduled", "internal", "maintenance"},
+    },
+    "supply_chain": {
+        "minimum_files": 37,
+        "required": {
+            "run.py",
+            "orchestrator/mcp_workflow.py",
+            "orchestrator/monitor.py",
+            "orchestrator/pg_writer.py",
+            "orchestrator/scripts/PurchaseConfirmation.py",
+            "orchestrator/scripts/ProductReplenishment.py",
+            "orchestrator/scripts/PurchaseOrderUpdate.py",
+            "orchestrator/scripts/batch_process.py",
+            "orchestrator/scripts/trigger_purchase_order_update.py",
+            "orchestrator/scripts/insert_replenishment_bitable.py",
+            "orchestrator/scripts/send_card_notification.py",
+            "orchestrator/sql/backfill_history.py",
+            "skills/collector/SKILL.md",
+            "skills/analyzer/SKILL.md",
+        },
+        "classifications": {
+            "scheduled",
+            "manual",
+            "event",
+            "maintenance",
+            "internal",
+        },
+    },
+}
+
+
+def _sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as handle:
+        for block in iter(lambda: handle.read(1024 * 1024), b""):
+            digest.update(block)
+    return digest.hexdigest()
+
+
+@pytest.mark.parametrize("module", sorted(EXPECTED))
+def test_source_manifest_covers_runtime_sources_and_entrypoints(module: str) -> None:
+    expected = EXPECTED[module]
+    manifest_path = MANIFEST_ROOT / f"{module}.json"
+    manifest = json.loads(manifest_path.read_text(encoding="utf-8"))
+
+    assert manifest["schema_version"] == 1
+    assert manifest["module"] == module
+    files = manifest["files"]
+    assert len(files) >= expected["minimum_files"]
+    assert expected["required"] <= {
+        item["source_relative_path"] for item in files
+    }
+
+    target_paths = [item["target_relative_path"] for item in files]
+    assert len(target_paths) == len(set(target_paths))
+    for item in files:
+        target = PurePosixPath(item["target_relative_path"])
+        assert not target.is_absolute()
+        assert ".." not in target.parts
+        assert target.parts[:5] == (
+            "src",
+            "gyxx_flow",
+            "modules",
+            module,
+            "runtime",
+        )
+        assert SHA256.fullmatch(item["source_sha256"])
+        assert SHA256.fullmatch(item["target_sha256"])
+        assert item["transformed"] is (
+            item["source_sha256"] != item["target_sha256"]
+        )
+
+    entrypoints = manifest["entrypoints"]
+    assert {item["classification"] for item in entrypoints} == expected[
+        "classifications"
+    ]
+    assert all(item["source_relative_path"] for item in entrypoints)
+    assert all(item["target_relative_path"] for item in entrypoints)
+
+
+@pytest.mark.parametrize("module", sorted(EXPECTED))
+def test_runtime_files_match_manifest_and_are_decoupled(module: str) -> None:
+    manifest_path = MANIFEST_ROOT / f"{module}.json"
+    manifest = json.loads(manifest_path.read_text(encoding="utf-8"))
+    runtime_root = (
+        PROJECT_ROOT / "src" / "gyxx_flow" / "modules" / module / "runtime"
+    )
+
+    for item in manifest["files"]:
+        target = PROJECT_ROOT.joinpath(
+            *PurePosixPath(item["target_relative_path"]).parts
+        )
+        assert target.is_file(), item["target_relative_path"]
+        assert _sha256(target) == item["target_sha256"]
+
+    for path in runtime_root.rglob("*"):
+        if not path.is_file() or b"\x00" in path.read_bytes():
+            continue
+        text = path.read_text(encoding="utf-8-sig", errors="strict").casefold()
+        assert not any(root in text for root in OLD_ROOTS), path
+
+    manifest_text = manifest_path.read_text(encoding="utf-8").casefold()
+    assert not any(root in manifest_text for root in OLD_ROOTS)
+    assert scan_repository(runtime_root) == []
+
+
+@pytest.mark.parametrize("module", sorted(EXPECTED))
+def test_source_manifest_explicitly_excludes_secrets_and_runtime_state(
+    module: str,
+) -> None:
+    manifest = json.loads(
+        (MANIFEST_ROOT / f"{module}.json").read_text(encoding="utf-8")
+    )
+    excluded = {item["pattern"] for item in manifest["intentionally_excluded"]}
+
+    assert {
+        ".git/**",
+        ".venv/**",
+        "**/__pycache__/**",
+        ".env",
+        "**/*cookies*",
+        "**/*profile*/**",
+        "**/data/**",
+        "**/logs/**",
+    } <= excluded
+
+
+@pytest.mark.parametrize("module", sorted(EXPECTED))
+def test_runtime_has_no_legacy_absolute_paths_or_sys_path_bootstrap(
+    module: str,
+) -> None:
+    runtime_root = (
+        PROJECT_ROOT / "src" / "gyxx_flow" / "modules" / module / "runtime"
+    )
+    for path in runtime_root.rglob("*"):
+        if not path.is_file() or b"\x00" in path.read_bytes():
+            continue
+        text = path.read_text(encoding="utf-8-sig", errors="strict")
+        assert not any(
+            pattern.search(text) for pattern in NON_PORTABLE_RUNTIME_PATTERNS
+        ), path
+
+
+def test_supply_backfill_reads_only_from_portable_shared_data_root() -> None:
+    backfill = (
+        PROJECT_ROOT
+        / "src"
+        / "gyxx_flow"
+        / "modules"
+        / "supply_chain"
+        / "runtime"
+        / "orchestrator"
+        / "sql"
+        / "backfill_history.py"
+    ).read_text(encoding="utf-8-sig")
+
+    assert "from gyxx_flow.modules.supply_chain.runtime.paths import SHARED_ROOT" in backfill
+    assert "ROOT = SHARED_ROOT" in backfill
diff --git a/tests/test_source_migration_acceptance.py b/tests/test_source_migration_acceptance.py
new file mode 100644
index 0000000..f5aa266
--- /dev/null
+++ b/tests/test_source_migration_acceptance.py
@@ -0,0 +1,62 @@
+from __future__ import annotations
+
+import hashlib
+import json
+from pathlib import Path, PurePosixPath
+
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.script_catalog import ScriptCatalog
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+MODULES = (
+    "content_marketing",
+    "product_commerce",
+    "shop_intelligence",
+    "supply_chain",
+)
+
+
+def _sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as stream:
+        for chunk in iter(lambda: stream.read(1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def test_all_source_manifests_use_one_schema_and_verify_target_hashes() -> None:
+    for module in MODULES:
+        path = PROJECT_ROOT / "config" / "source-manifests" / f"{module}.json"
+        manifest = json.loads(path.read_text(encoding="utf-8-sig"))
+        assert manifest["schema_version"] == 1
+        assert manifest["module"] == module
+        assert isinstance(manifest["snapshot"], str) and manifest["snapshot"]
+        assert manifest["files"]
+        assert isinstance(manifest["intentionally_excluded"], list)
+        for item in manifest["files"]:
+            target = PROJECT_ROOT.joinpath(
+                *PurePosixPath(item["target_relative_path"]).parts
+            )
+            assert target.is_file(), target
+            assert _sha256(target) == item["target_sha256"], target
+
+
+def test_every_workflow_and_manual_script_resolves_inside_new_project() -> None:
+    catalog = WorkflowCatalog.load(PROJECT_ROOT / "config")
+    for workflow in catalog.workflows:
+        if workflow.trigger == "unavailable":
+            continue
+        target = (
+            PROJECT_ROOT
+            / "src"
+            / "gyxx_flow"
+            / "modules"
+            / workflow.module
+            / "runtime"
+            / workflow.entry
+        ).resolve(strict=True)
+        assert target.is_relative_to(PROJECT_ROOT)
+
+    scripts = ScriptCatalog.discover_default()
+    assert len(scripts.scripts) >= 100
+    assert all(script.path.is_relative_to(PROJECT_ROOT) for script in scripts.scripts)
diff --git a/tests/test_supply_chain_module.py b/tests/test_supply_chain_module.py
new file mode 100644
index 0000000..60e3789
--- /dev/null
+++ b/tests/test_supply_chain_module.py
@@ -0,0 +1,42 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+from gyxx_flow.adapters.native import DeferredModuleCommandStep
+from gyxx_flow.catalog import WorkflowCatalog
+from gyxx_flow.modules.supply_chain import (
+    MANUAL_SUPPLY_WORKFLOW_IDS,
+    SCHEDULED_SUPPLY_WORKFLOW_IDS,
+    SUPPLY_RESOURCE,
+    SUPPLY_TIMEOUT_SECONDS,
+    SUPPLY_WORKFLOW_IDS,
+    SupplyChainModule,
+)
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+
+
+def test_supply_module_registers_scheduled_and_manual_workflows() -> None:
+    module = SupplyChainModule.from_catalog(
+        WorkflowCatalog.load(PROJECT_ROOT / "config")
+    )
+
+    assert tuple(item.workflow_id for item in module.workflow_definitions()) == SUPPLY_WORKFLOW_IDS
+    assert len(SCHEDULED_SUPPLY_WORKFLOW_IDS) == 3
+    assert len(MANUAL_SUPPLY_WORKFLOW_IDS) == 1
+
+
+def test_supply_workflows_use_project_owned_module_commands() -> None:
+    module = SupplyChainModule.from_catalog(
+        WorkflowCatalog.load(PROJECT_ROOT / "config")
+    )
+
+    for definition in module.workflow_definitions():
+        step = definition.steps[0]
+        assert step.step_id == "module_supply"
+        assert isinstance(step.action, DeferredModuleCommandStep)
+        assert step.timeout_seconds == SUPPLY_TIMEOUT_SECONDS == 2 * 60 * 60
+        assert step.max_attempts == 1
+        assert step.resources == (SUPPLY_RESOURCE,) == ("module:supply_chain",)
+        assert step.production_sink is True
+
diff --git a/tests/test_unified_module_data_paths.py b/tests/test_unified_module_data_paths.py
new file mode 100644
index 0000000..62e36b2
--- /dev/null
+++ b/tests/test_unified_module_data_paths.py
@@ -0,0 +1,144 @@
+from __future__ import annotations
+
+import importlib.util
+import sys
+from pathlib import Path
+
+import pytest
+
+from gyxx_flow.core.layout import DataLayout
+
+PROJECT_ROOT = Path(__file__).resolve().parents[1]
+RUNTIME_ROOT = PROJECT_ROOT / "src" / "gyxx_flow" / "modules"
+
+
+def _load_path_module(module: str, filename: str, monkeypatch, data_root: Path):
+    monkeypatch.setenv("GYXX_DATA_ROOT", str(data_root))
+    runtime_root = RUNTIME_ROOT / module / "runtime"
+    monkeypatch.setenv("GYXX_MODULE_ROOT", str(runtime_root))
+    path = runtime_root / filename
+    name = f"_gyxx_test_paths_{module}"
+    spec = importlib.util.spec_from_file_location(name, path)
+    assert spec and spec.loader
+    loaded = importlib.util.module_from_spec(spec)
+    sys.modules[name] = loaded
+    try:
+        spec.loader.exec_module(loaded)
+    finally:
+        sys.modules.pop(name, None)
+    return loaded
+
+
+@pytest.mark.parametrize(
+    ("module", "filename", "names"),
+    [
+        (
+            "content_marketing",
+            "runtime_paths.py",
+            ("raw_root", "normalized_root", "curated_root", "exports_root"),
+        ),
+        (
+            "product_commerce",
+            "runtime_paths.py",
+            (
+                "RAW_DATA_ROOT",
+                "NORMALIZED_DATA_ROOT",
+                "CURATED_DATA_ROOT",
+                "EXPORTS_DATA_ROOT",
+            ),
+        ),
+        (
+            "shop_intelligence",
+            "paths.py",
+            ("RAW_ROOT", "NORMALIZED_ROOT", "CURATED_ROOT", "EXPORTS_ROOT"),
+        ),
+        (
+            "supply_chain",
+            "paths.py",
+            ("SHARED_ROOT", "NORMALIZED_ROOT", "CURATED_ROOT", "BACKUP_ROOT"),
+        ),
+    ],
+)
+def test_all_module_path_boundaries_use_the_canonical_data_tree(
+    tmp_path: Path,
+    monkeypatch,
+    module: str,
+    filename: str,
+    names: tuple[str, str, str, str],
+) -> None:
+    data_root = tmp_path / "relocatable-data"
+    loaded = _load_path_module(module, filename, monkeypatch, data_root)
+
+    layers = ("raw", "normalized", "curated", "exports")
+    for name, layer in zip(names, layers, strict=True):
+        assert getattr(loaded, name) == data_root / "data" / layer / module
+    assert not data_root.exists(), "path resolution must not mutate the filesystem"
+
+
+def test_module_data_paths_cover_collected_and_exported_file_types(tmp_path: Path) -> None:
+    paths = DataLayout(tmp_path).for_module("content_marketing")
+
+    assert paths.raw_path("api", "response.json") == (
+        tmp_path / "data" / "raw" / "content_marketing" / "api" / "response.json"
+    )
+    assert paths.raw_path("downloads", "source.xlsx") == (
+        tmp_path / "data" / "raw" / "content_marketing" / "downloads" / "source.xlsx"
+    )
+    assert paths.raw_path("downloads", "source.xls") == (
+        tmp_path / "data" / "raw" / "content_marketing" / "downloads" / "source.xls"
+    )
+    assert paths.raw_path("downloads", "source.csv") == (
+        tmp_path / "data" / "raw" / "content_marketing" / "downloads" / "source.csv"
+    )
+    assert paths.export_path("reports", "weekly.md") == (
+        tmp_path / "data" / "exports" / "content_marketing" / "reports" / "weekly.md"
+    )
+    assert list(tmp_path.iterdir()) == []
+
+
+@pytest.mark.parametrize("unsafe", ["", "..", "a/b", "a\\b", "C:\\data"])
+def test_module_data_paths_reject_unsafe_module_names(tmp_path: Path, unsafe: str) -> None:
+    with pytest.raises(ValueError, match="path segment"):
+        DataLayout(tmp_path).for_module(unsafe)
+
+
+@pytest.mark.parametrize("unsafe", [("..", "escape.json"), ("C:\\", "escape.json")])
+def test_module_data_paths_reject_paths_outside_the_layer(
+    tmp_path: Path, unsafe: tuple[str, str]
+) -> None:
+    with pytest.raises(ValueError, match="escapes module data root"):
+        DataLayout(tmp_path).for_module("product_commerce").raw_path(*unsafe)
+
+
+@pytest.mark.parametrize("configured", ["", "   "])
+@pytest.mark.parametrize(
+    ("module", "filename", "root_name"),
+    [
+        ("content_marketing", "runtime_paths.py", "data_root"),
+        ("product_commerce", "runtime_paths.py", "DATA_HOME"),
+        ("shop_intelligence", "paths.py", "DATA_ROOT"),
+        ("supply_chain", "paths.py", "DATA_ROOT"),
+    ],
+)
+def test_blank_data_root_uses_the_project_var_default(
+    monkeypatch,
+    configured: str,
+    module: str,
+    filename: str,
+    root_name: str,
+) -> None:
+    monkeypatch.setenv("GYXX_DATA_ROOT", configured)
+    runtime_root = RUNTIME_ROOT / module / "runtime"
+    monkeypatch.setenv("GYXX_MODULE_ROOT", str(runtime_root))
+    path = runtime_root / filename
+    name = f"_gyxx_test_blank_paths_{module}"
+    spec = importlib.util.spec_from_file_location(name, path)
+    assert spec and spec.loader
+    loaded = importlib.util.module_from_spec(spec)
+    sys.modules[name] = loaded
+    try:
+        spec.loader.exec_module(loaded)
+    finally:
+        sys.modules.pop(name, None)
+
+    assert getattr(loaded, root_name) == PROJECT_ROOT / "var"
diff --git a/tests/test_workflow_engine.py b/tests/test_workflow_engine.py
new file mode 100644
index 0000000..82a86e8
--- /dev/null
+++ b/tests/test_workflow_engine.py
@@ -0,0 +1,365 @@
+from __future__ import annotations
+
+import json
+import subprocess
+from collections import deque
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+
+import pytest
+
+from gyxx_flow.core.context import RunContext
+from gyxx_flow.core.layout import DataLayout
+from gyxx_flow.core.locks import LockManager
+from gyxx_flow.core.records import RunJournal
+from gyxx_flow.ops.effects import EffectLedger
+from gyxx_flow.workflow.engine import WorkflowEngine
+from gyxx_flow.workflow.model import (
+    StepDefinition,
+    WorkflowDefinition,
+    WorkflowValidationError,
+)
+from gyxx_flow.workflow.steps import CommandStep, StepExecution
+
+
+def _context(*, shadow: bool = False) -> RunContext:
+    return RunContext.create(
+        "shop.weekly",
+        "2026-07-27",
+        shadow=shadow,
+        now=datetime(2026, 7, 27, 4, 0, tzinfo=timezone.utc),
+        random_suffix="abc123",
+    )
+
+
+class FakeStep:
+    def __init__(self, *outcomes: StepExecution) -> None:
+        self.outcomes = deque(outcomes or (StepExecution(exit_code=0),))
+        self.calls: list[dict[str, Any]] = []
+
+    def execute(
+        self,
+        *,
+        context: RunContext,
+        timeout_seconds: float | None,
+        dry_run: bool,
+    ) -> StepExecution:
+        self.calls.append(
+            {
+                "context": context,
+                "timeout_seconds": timeout_seconds,
+                "dry_run": dry_run,
+            }
+        )
+        return self.outcomes.popleft()
+
+
+def test_workflow_validates_relations_and_uses_stable_topological_order() -> None:
+    action = FakeStep()
+    workflow = WorkflowDefinition(
+        workflow_id="shop.weekly",
+        steps=(
+            StepDefinition("publish", action, depends_on=("collect",)),
+            StepDefinition("audit", action),
+            StepDefinition("collect", action),
+        ),
+    )
+
+    assert [step.step_id for step in workflow.ordered_steps()] == [
+        "audit",
+        "collect",
+        "publish",
+    ]
+
+    with pytest.raises(WorkflowValidationError, match="duplicate step id"):
+        WorkflowDefinition(
+            "shop.weekly",
+            (StepDefinition("collect", action), StepDefinition("collect", action)),
+        )
+
+    with pytest.raises(WorkflowValidationError, match="unknown dependency"):
+        WorkflowDefinition(
+            "shop.weekly",
+            (StepDefinition("publish", action, depends_on=("missing",)),),
+        )
+
+    with pytest.raises(WorkflowValidationError, match="cycle"):
+        WorkflowDefinition(
+            "shop.weekly",
+            (
+                StepDefinition("first", action, depends_on=("second",)),
+                StepDefinition("second", action, depends_on=("first",)),
+            ),
+        )
+
+
+@pytest.mark.parametrize(
+    ("kwargs", "message"),
+    [
+        ({"max_attempts": 0}, "max_attempts"),
+        ({"timeout_seconds": 0}, "timeout_seconds"),
+        ({"retry_delay_seconds": -1}, "retry_delay_seconds"),
+        ({"depends_on": ("one", "one")}, "duplicate dependency"),
+    ],
+)
+def test_step_definition_validates_retry_timeout_and_dependencies(
+    kwargs: dict[str, object], message: str
+) -> None:
+    with pytest.raises(WorkflowValidationError, match=message):
+        StepDefinition("collect", FakeStep(), **kwargs)
+
+
+def test_command_step_uses_argv_explicit_cwd_env_and_timeout(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
+) -> None:
+    captured: dict[str, object] = {}
+
+    def fake_run(argv: list[str], **kwargs: object) -> subprocess.CompletedProcess[str]:
+        captured["argv"] = argv
+        captured.update(kwargs)
+        return subprocess.CompletedProcess(argv, 0, stdout="ok", stderr="")
+
+    monkeypatch.setattr(subprocess, "run", fake_run)
+    step = CommandStep(
+        argv=("C:/Python312/python.exe", "worker.py", "--date", "2026-07-27"),
+        cwd=tmp_path,
+        env={"MODE": "test"},
+    )
+
+    outcome = step.execute(context=_context(), timeout_seconds=9.5, dry_run=False)
+
+    assert outcome.exit_code == 0
+    assert outcome.stdout == "ok"
+    assert captured["argv"] == [
+        "C:/Python312/python.exe",
+        "worker.py",
+        "--date",
+        "2026-07-27",
+    ]
+    assert captured["cwd"] == tmp_path
+    assert captured["env"] == {"MODE": "test"}
+    assert captured["timeout"] == 9.5
+    assert captured["shell"] is False
+
+
+def test_command_step_dry_run_never_starts_a_process(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
+) -> None:
+    def fail_if_called(*args: object, **kwargs: object) -> None:
+        raise AssertionError("subprocess must not be started")
+
+    monkeypatch.setattr(subprocess, "run", fail_if_called)
+    step = CommandStep(argv=("tool.exe", "--write"), cwd=tmp_path, env={})
+
+    outcome = step.execute(context=_context(), timeout_seconds=None, dry_run=True)
+
+    assert outcome.skipped is True
+    assert outcome.reason == "dry-run"
+    assert outcome.exit_code == 0
+
+
+def test_command_step_reports_timeout_as_a_failed_exit(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
+) -> None:
+    def timeout(*args: object, **kwargs: object) -> None:
+        raise subprocess.TimeoutExpired(cmd=["tool.exe"], timeout=2)
+
+    monkeypatch.setattr(subprocess, "run", timeout)
+    step = CommandStep(argv=("tool.exe",), cwd=tmp_path, env={})
+
+    outcome = step.execute(context=_context(), timeout_seconds=2, dry_run=False)
+
+    assert outcome.exit_code == 124
+    assert "timed out" in (outcome.error or "")
+
+
+def test_engine_retries_records_every_attempt_and_aggregates_failures(tmp_path: Path) -> None:
+    flaky = FakeStep(
+        StepExecution(exit_code=7, error="temporary"),
+        StepExecution(exit_code=0),
+    )
+    optional = FakeStep(StepExecution(exit_code=5, error="optional failed"))
+    required = FakeStep(StepExecution(exit_code=9, error="required failed"))
+    workflow = WorkflowDefinition(
+        "shop.weekly",
+        (
+            StepDefinition(
+                "flaky",
+                flaky,
+                timeout_seconds=4,
+                max_attempts=2,
+                retry_delay_seconds=0,
+                resources=("browser:shop",),
+            ),
+            StepDefinition("optional", optional, critical=False),
+            StepDefinition("required", required),
+        ),
+    )
+    context = _context()
+    layout = DataLayout(tmp_path)
+    journal = RunJournal.create(layout, context)
+
+    result = WorkflowEngine(LockManager(tmp_path / "locks")).execute(
+        workflow,
+        context=context,
+        journal=journal,
+    )
+
+    assert result.status == "failed"
+    assert result.exit_code == 1
+    assert result.steps["flaky"].status == "success"
+    assert result.steps["flaky"].final_exit_code == 0
+    assert [attempt.exit_code for attempt in result.steps["flaky"].attempts] == [7, 0]
+    assert result.steps["optional"].status == "failed"
+    assert result.steps["required"].status == "failed"
+    assert len(flaky.calls) == 2
+    assert all(call["timeout_seconds"] == 4 for call in flaky.calls)
+
+    payload = json.loads(journal.path.read_text(encoding="utf-8"))
+    assert payload["status"] == "failed"
+    assert payload["steps"]["flaky.attempt-1"]["exit_code"] == 7
+    assert payload["steps"]["flaky.attempt-2"]["exit_code"] == 0
+    assert payload["steps"]["required.attempt-1"]["exit_code"] == 9
+    assert payload["trace"]["inputs"] == [
+        "step:flaky",
+        "step:optional",
+        "step:required",
+    ]
+    assert "step:flaky:attempt:1:failed:exit:7" in payload["trace"]["outputs"]
+    assert "step:flaky:attempt:2:success:exit:0" in payload["trace"]["outputs"]
+    assert list((tmp_path / "locks").glob("*.lock")) == []
+
+
+def test_noncritical_failure_does_not_fail_the_workflow(tmp_path: Path) -> None:
+    workflow = WorkflowDefinition(
+        "shop.weekly",
+        (
+            StepDefinition(
+                "optional",
+                FakeStep(StepExecution(exit_code=2, error="not available")),
+                critical=False,
+            ),
+            StepDefinition("required", FakeStep()),
+        ),
+    )
+    context = _context()
+    journal = RunJournal.create(DataLayout(tmp_path), context)
+
+    result = WorkflowEngine(LockManager(tmp_path / "locks")).execute(
+        workflow, context=context, journal=journal
+    )
+
+    assert result.status == "success"
+    assert result.exit_code == 0
+    assert result.warnings == ("optional",)
+
+
+def test_engine_shadow_mode_skips_production_sinks_and_official_notifications(
+    tmp_path: Path,
+) -> None:
+    safe = FakeStep()
+    sink = FakeStep()
+    notify = FakeStep()
+    workflow = WorkflowDefinition(
+        "shop.weekly",
+        (
+            StepDefinition("safe", safe),
+            StepDefinition("publish", sink, production_sink=True, depends_on=("safe",)),
+            StepDefinition(
+                "notify",
+                notify,
+                official_notification=True,
+                depends_on=("publish",),
+            ),
+        ),
+    )
+    context = _context(shadow=True)
+    journal = RunJournal.create(DataLayout(tmp_path), context)
+
+    result = WorkflowEngine(LockManager(tmp_path / "locks")).execute(
+        workflow, context=context, journal=journal
+    )
+
+    assert result.status == "success"
+    assert result.steps["safe"].status == "success"
+    assert result.steps["publish"].status == "skipped"
+    assert result.steps["publish"].reason == "shadow-policy"
+    assert result.steps["notify"].status == "skipped"
+    assert sink.calls == []
+    assert notify.calls == []
+
+    payload = json.loads(journal.path.read_text(encoding="utf-8"))
+    assert payload["steps"]["publish.attempt-1"]["status"] == "skipped"
+    assert payload["steps"]["notify.attempt-1"]["status"] == "skipped"
+    assert payload["trace"]["external_writes"] == []
+
+
+def test_successful_sink_records_declared_external_write_in_trace(tmp_path: Path) -> None:
+    workflow = WorkflowDefinition(
+        "shop.weekly",
+        (StepDefinition("publish", FakeStep(), production_sink=True),),
+    )
+    context = _context()
+    journal = RunJournal.create(DataLayout(tmp_path), context)
+
+    result = WorkflowEngine(LockManager(tmp_path / "locks")).execute(
+        workflow, context=context, journal=journal
+    )
+
+    assert result.status == "success"
+    payload = json.loads(journal.path.read_text(encoding="utf-8"))
+    assert payload["trace"]["external_writes"] == ["step:publish:production_sink"]
+
+
+def test_effect_ledger_skips_a_successful_production_sink_replay(tmp_path: Path) -> None:
+    action = FakeStep()
+    workflow = WorkflowDefinition(
+        "shop.weekly",
+        (StepDefinition("publish", action, production_sink=True),),
+    )
+    ledger = EffectLedger(tmp_path)
+    engine = WorkflowEngine(
+        LockManager(tmp_path / "locks"), effect_ledger=ledger
+    )
+    first_context = _context()
+    first = engine.execute(
+        workflow,
+        context=first_context,
+        journal=RunJournal.create(DataLayout(tmp_path), first_context),
+    )
+    second_context = RunContext.create(
+        "shop.weekly", "2026-07-27", random_suffix="replay2"
+    )
+    second = engine.execute(
+        workflow,
+        context=second_context,
+        journal=RunJournal.create(DataLayout(tmp_path), second_context),
+    )
+
+    assert first.status == "success"
+    assert second.status == "success"
+    assert second.steps["publish"].reason == "idempotency-replay"
+    assert len(action.calls) == 1
+
+
+def test_failed_dependency_skips_downstream_critical_step(tmp_path: Path) -> None:
+    downstream = FakeStep()
+    workflow = WorkflowDefinition(
+        "shop.weekly",
+        (
+            StepDefinition("collect", FakeStep(StepExecution(exit_code=3, error="bad input"))),
+            StepDefinition("publish", downstream, depends_on=("collect",)),
+        ),
+    )
+    context = _context()
+    journal = RunJournal.create(DataLayout(tmp_path), context)
+
+    result = WorkflowEngine(LockManager(tmp_path / "locks")).execute(
+        workflow, context=context, journal=journal
+    )
+
+    assert result.status == "failed"
+    assert result.steps["publish"].status == "skipped"
+    assert result.steps["publish"].reason == "dependency-failed"
+    assert downstream.calls == []
diff --git a/uv.lock b/uv.lock
new file mode 100644
index 0000000..3ff853b
--- /dev/null
+++ b/uv.lock
@@ -0,0 +1,2872 @@
+version = 1
+revision = 3
+requires-python = ">=3.12"
+resolution-markers = [
+    "python_full_version >= '3.14' and sys_platform == 'win32'",
+    "python_full_version >= '3.14' and sys_platform == 'emscripten'",
+    "python_full_version >= '3.14' and sys_platform != 'emscripten' and sys_platform != 'win32'",
+    "python_full_version < '3.14' and sys_platform == 'win32'",
+    "python_full_version < '3.14' and sys_platform == 'emscripten'",
+    "python_full_version < '3.14' and sys_platform != 'emscripten' and sys_platform != 'win32'",
+]
+
+[[package]]
+name = "annotated-types"
+version = "0.8.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/5f/56/a8120250d128bed162cd73c76d45f6ef9991f3e068f62a8ee060afa3104a/annotated_types-0.8.0.tar.gz", hash = "sha256:13b2beaad985e05e2d6407ee4c4f35590b11f8d693a258a561055cac8f64cab7", size = 15893, upload-time = "2026-07-23T20:16:13.995Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/99/91/8acff4f5e50511b911bbccb72b8628a49c68ce14148cd9f6431094859a90/annotated_types-0.8.0-py3-none-any.whl", hash = "sha256:f072f4d804ea359e4eaf198b1af7a8b0943881a87f31bb764f8bf219bb9419e0", size = 13427, upload-time = "2026-07-23T20:16:12.938Z" },
+]
+
+[[package]]
+name = "anthropic"
+version = "0.120.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anyio" },
+    { name = "distro" },
+    { name = "docstring-parser" },
+    { name = "httpx" },
+    { name = "jiter" },
+    { name = "pydantic" },
+    { name = "sniffio" },
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/92/5c/3331da4fc009d448008a50c78d86cc929e8c937cd1442245ce3f80561c4e/anthropic-0.120.0.tar.gz", hash = "sha256:6ba6007dc9b00365b20f6101a6618f5196ac1ceef81512e4b5cc0e7436d4975d", size = 1008042, upload-time = "2026-07-24T16:32:52.384Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/1b/8a/8522bdf809e1f95f0d9c936540987a3f6afba01d2921a2bf488dedf836a8/anthropic-0.120.0-py3-none-any.whl", hash = "sha256:591bd531563ec7b63a1e138f5c11f14cb94edda99623b349c2ce2ece8e08b8a5", size = 1022602, upload-time = "2026-07-24T16:32:50.506Z" },
+]
+
+[[package]]
+name = "anyio"
+version = "4.14.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "idna" },
+    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/61/cc/a381afa6efea9f496eff839d4a6a1aed3bfafc7b3ab4b0d1b243a12573dd/anyio-4.14.2.tar.gz", hash = "sha256:cfa139f3ed1a23ee8f88a145ddb5ac7605b8bbfd8592baacd7ce3d8bb4313c7f", size = 260176, upload-time = "2026-07-12T20:29:07.082Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/da/35/f2287558c17e29fafc8ef3daf819bb9834061cfa43bff8014f7df7f63bdc/anyio-4.14.2-py3-none-any.whl", hash = "sha256:9f505dda5ac9f0c8309b5e8bd445a8c2bf7246f3ce950121e45ea15bc41d1494", size = 125813, upload-time = "2026-07-12T20:29:05.763Z" },
+]
+
+[[package]]
+name = "apify-fingerprint-datapoints"
+version = "0.13.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/f9/f1/b74f95767581372ab849c8b13e384b62f60d034584892c60c4a3442d9312/apify_fingerprint_datapoints-0.13.0.tar.gz", hash = "sha256:263141c19e9bc90a821e6b4e2b845925f17e0b8fbd53a897fc71546bd50df7f1", size = 934827, upload-time = "2026-05-04T09:08:45.036Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/fc/58/8402442bf6af5a3a8068fe5431c42ea4f73c1eb18f621f9bf7c5de80caf5/apify_fingerprint_datapoints-0.13.0-py3-none-any.whl", hash = "sha256:0213d42297be19e8035202b41fb2e840a1e5d79874c99c882a5027a7d0b1a0eb", size = 761652, upload-time = "2026-05-04T09:08:43.347Z" },
+]
+
+[[package]]
+name = "asttokens"
+version = "3.0.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/25/1e/faf0f247f6f881b98fc4d6d07e14085cb89d13665084e6d6ac1dc2c03d0b/asttokens-3.0.2.tar.gz", hash = "sha256:3ecdbd8f2cc195f53ccada3a613538bb5f9ef6f6869129f13e03c30a677b8fe2", size = 63136, upload-time = "2026-07-12T03:31:49.084Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/d4/2b/04b8a15f3a1c77bc79ddf5c73875327f34b4fa75982df2b76e45e402d364/asttokens-3.0.2-py3-none-any.whl", hash = "sha256:9da13157f5b28becde0bd374fc677dcd3c290614264eff096f167c469cd9f933", size = 28702, upload-time = "2026-07-12T03:31:47.542Z" },
+]
+
+[[package]]
+name = "attrs"
+version = "26.1.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/9a/8e/82a0fe20a541c03148528be8cac2408564a6c9a0cc7e9171802bc1d26985/attrs-26.1.0.tar.gz", hash = "sha256:d03ceb89cb322a8fd706d4fb91940737b6642aa36998fe130a9bc96c985eff32", size = 952055, upload-time = "2026-03-19T14:22:25.026Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/64/b4/17d4b0b2a2dc85a6df63d1157e028ed19f90d4cd97c36717afef2bc2f395/attrs-26.1.0-py3-none-any.whl", hash = "sha256:c647aa4a12dfbad9333ca4e71fe62ddc36f4e63b2d260a37a8b83d2f043ac309", size = 67548, upload-time = "2026-03-19T14:22:23.645Z" },
+]
+
+[[package]]
+name = "beautifulsoup4"
+version = "4.15.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "soupsieve" },
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/43/65/318323f98dbee45d42dff61d8f047181bc6f2268a9068cfad035a46be5af/beautifulsoup4-4.15.0.tar.gz", hash = "sha256:288e3ca7d54b06f2ac191970bc275c1939cb46d450b255bf6718b04aa37ab4f7", size = 632571, upload-time = "2026-06-07T16:44:20.453Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/88/c6/92fcd42f1ba33e1184263f25bfabf3d27c383410470f169e4b8163bf9c17/beautifulsoup4-4.15.0-py3-none-any.whl", hash = "sha256:d6f88de62e1d4e38ecb1077eb9724cd0eff29d2a08ca16a401e9b9e93f117cf9", size = 109924, upload-time = "2026-06-07T16:44:21.566Z" },
+]
+
+[[package]]
+name = "browserforge"
+version = "1.2.4"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "apify-fingerprint-datapoints" },
+    { name = "click" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/78/6f/8975af88d203efd70cc69477ebac702babef38201d04621c9583f2508f25/browserforge-1.2.4.tar.gz", hash = "sha256:05686473793769856ebd3528c69071f5be0e511260993e8b2ba839863711a0c4", size = 36700, upload-time = "2026-02-03T02:52:09.721Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/dd/35/ce962f738ae28ffce6293e7607b129075633e6bb185a5ab87e49246eedc2/browserforge-1.2.4-py3-none-any.whl", hash = "sha256:fb1c14e62ac09de221dcfc73074200269f697596c642cb200ceaab1127a17542", size = 37890, upload-time = "2026-02-03T02:52:08.745Z" },
+]
+
+[[package]]
+name = "certifi"
+version = "2026.7.22"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/a3/c2/24167ea9858356b47a87a50d39908bfdb72ceeefe0041586e704e5376b3a/certifi-2026.7.22.tar.gz", hash = "sha256:741e2c3b351ddf169a738da9f2c048608ff7f2c5cc02f1ebc6b118bb090d5d55", size = 138112, upload-time = "2026-07-22T03:35:12.644Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/0b/a7/71ac2cff56fec219ed242bb11b8efb69fcc4bec75db06fb7bfe35de520e6/certifi-2026.7.22-py3-none-any.whl", hash = "sha256:62f22742b58a1a33014a2b6b706588a8d7e2a88ae7bd1a6ebe8c992928483775", size = 136983, upload-time = "2026-07-22T03:35:11.276Z" },
+]
+
+[[package]]
+name = "cffi"
+version = "2.1.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "pycparser", marker = "implementation_name != 'PyPy'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/57/5f/ff100cae70ebe9d8df1c01a00e510e45d9adb5c1fdda84791b199141de97/cffi-2.1.0.tar.gz", hash = "sha256:efc1cdd798b1aaf39b4610bba7aad28c9bea9b910f25c784ccf9ec1fa719d1f9", size = 531036, upload-time = "2026-07-06T21:34:30.382Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/1e/85/990925db5df586ec90beb97529c853497e7f85ba0234830447faf41c3057/cffi-2.1.0-cp312-cp312-macosx_10_15_x86_64.whl", hash = "sha256:df2b82571a1b30f58a87bf4e5a9e78d2b1eff6c6ce8fd3aa3757221f93f0863f", size = 184829, upload-time = "2026-07-06T21:32:44.324Z" },
+    { url = "https://files.pythonhosted.org/packages/4b/92/e7bb136ad6b5352603732cf907ef862ca103f20f2031c1735a46300c20c9/cffi-2.1.0-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:78474632761faa0fb96f30b1c928c84ebcf68713cbb80d15bab09dfe61640fde", size = 184728, upload-time = "2026-07-06T21:32:45.683Z" },
+    { url = "https://files.pythonhosted.org/packages/c3/c0/d1ec30ffb370f748f2fb54425972bfef9871e0132e82fb589c46b6676049/cffi-2.1.0-cp312-cp312-manylinux1_i686.manylinux2014_i686.manylinux_2_17_i686.manylinux_2_5_i686.whl", hash = "sha256:5972433ad71a9e46516584ef60a0fda12d9dc459938d1539c3ddecf9bdc1368d", size = 214815, upload-time = "2026-07-06T21:32:48.557Z" },
+    { url = "https://files.pythonhosted.org/packages/1b/dc/5620cf930688be01f2d673804291de757a934c90b946dbdc3d84130c2ea4/cffi-2.1.0-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:b6422532152adf4e59b110cb2808cee7a033800952f5c036b4af047ee43199e7", size = 222429, upload-time = "2026-07-06T21:32:49.848Z" },
+    { url = "https://files.pythonhosted.org/packages/4b/a4/77b53abbf7a1e0beb9637edbef2a94d15f9c822f591e85d439ffd91519a6/cffi-2.1.0-cp312-cp312-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:46b1c8db8f6122420f32d02fffb924c2fe9bc772d228c7c711748fff56aabb2b", size = 210315, upload-time = "2026-07-06T21:32:51.221Z" },
+    { url = "https://files.pythonhosted.org/packages/58/0c/f528df19cc94b675087324d4760d9e6d5bfae97d6217aa4fac43de4f5fcc/cffi-2.1.0-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:d9fafc5aa2e2a39aaf7f8cc0c1f044a9b07fca12e558dca53a3cc5c654ad67a7", size = 208859, upload-time = "2026-07-06T21:32:52.512Z" },
+    { url = "https://files.pythonhosted.org/packages/62/f2/c9522a81c32132799a1972c39f5c5f8b4c8b9f00488a23feaa6c06f07741/cffi-2.1.0-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:1e9f50d192a3e525b15a75ab5114e442d83d657b7ec29182a991bc9a88fd3a66", size = 221844, upload-time = "2026-07-06T21:32:53.704Z" },
+    { url = "https://files.pythonhosted.org/packages/6e/28/bd53988b9833e8f8ad539d26f4c07a6b3f6bcb1e9e02e7ca038250b3428d/cffi-2.1.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:98fff996e983a36d3aa2eca83af40c5821202e7e6f32d13ae94e3d2286f10cfe", size = 225287, upload-time = "2026-07-06T21:32:54.907Z" },
+    { url = "https://files.pythonhosted.org/packages/79/99/0d0fd37f055224085f42bbb2c022d002e17dde4a97972822327b07d84101/cffi-2.1.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:379de10ce1ba048b1448599d1b37b24caee16309d1ac98d3982fc997f768700b", size = 223681, upload-time = "2026-07-06T21:32:56.329Z" },
+    { url = "https://files.pythonhosted.org/packages/b0/80/c138990aa2a70b1a269f6e06348729836d733d6f970867943f61d367f8cc/cffi-2.1.0-cp312-cp312-win32.whl", hash = "sha256:9b8f0f26ca4e7513c534d351eca551947d053fac438f2a04ac96d882909b0d3a", size = 175269, upload-time = "2026-07-06T21:32:57.777Z" },
+    { url = "https://files.pythonhosted.org/packages/a8/eb/f636456ff21a83fc13c032b58cc5dde061691546ac79efa284b2989b7982/cffi-2.1.0-cp312-cp312-win_amd64.whl", hash = "sha256:c97f080ea627e2863524c5af3836e2270b5f5dfff1f104392b959f8df0c5d384", size = 185881, upload-time = "2026-07-06T21:32:59.253Z" },
+    { url = "https://files.pythonhosted.org/packages/dd/2c/400ea43e721727dca8a65c4521390e9196757caba4a45643acb2b63271b8/cffi-2.1.0-cp312-cp312-win_arm64.whl", hash = "sha256:6d194185eabd279f1c05ebe3504265ddfc5ad2b58d0714f7db9f01da592e9eb6", size = 180088, upload-time = "2026-07-06T21:33:02.278Z" },
+    { url = "https://files.pythonhosted.org/packages/96/88/a996879e2eeccb815f6e3a5967b12a308257412acec882039d386bd2aa7b/cffi-2.1.0-cp313-cp313-ios_13_0_arm64_iphoneos.whl", hash = "sha256:10537b1df4967ca26d21e5072d7d54188354483b91dc75058968d3f0cf13fbda", size = 194331, upload-time = "2026-07-06T21:33:03.697Z" },
+    { url = "https://files.pythonhosted.org/packages/58/85/7ae00d5c8dd6266f4e944c3db630f3c5c9a98b61d469c714d848b1d8138a/cffi-2.1.0-cp313-cp313-ios_13_0_arm64_iphonesimulator.whl", hash = "sha256:a95b05f9baf29b91171b3a8bd2020b028835243e7b0ff6bb23e2a3c228518b1b", size = 196966, upload-time = "2026-07-06T21:33:05.353Z" },
+    { url = "https://files.pythonhosted.org/packages/8c/e9/45c3a76ad8d43ad9261f4c95436da61128d3ca545d72b9612c0ab5be0b1c/cffi-2.1.0-cp313-cp313-macosx_10_15_x86_64.whl", hash = "sha256:15faec4adfff450819f3aee0e2e02c812de6edb88203aa58807955db2003472a", size = 184795, upload-time = "2026-07-06T21:33:06.699Z" },
+    { url = "https://files.pythonhosted.org/packages/84/4c/82f132cb4418ee6d953d982b19191e87e2a6372c8a4ce36e50b69d6ade4a/cffi-2.1.0-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:716ff8ec22f20b4d988b12884086bcef0fc99737043e503f7a3935a6be99b1ea", size = 184746, upload-time = "2026-07-06T21:33:08.071Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/1c/4ed5a0e5bdca6cbc275556de3328dd1b76fd0c11cc13c88fe66d1d8715f2/cffi-2.1.0-cp313-cp313-manylinux1_i686.manylinux2014_i686.manylinux_2_17_i686.manylinux_2_5_i686.whl", hash = "sha256:63960549e4f8dc41e31accb97b975abaecfc44c03e396c093a6436763c2ea7db", size = 214747, upload-time = "2026-07-06T21:33:09.671Z" },
+    { url = "https://files.pythonhosted.org/packages/3a/a6/e879bb68cc23a2bc9ba8f4b7d8019f0c2694bad2ab6c4a3701d429439f58/cffi-2.1.0-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:ff067a8d8d880e7809e4ac88eb009bb848870115317b306666502ccad30b147f", size = 222392, upload-time = "2026-07-06T21:33:10.896Z" },
+    { url = "https://files.pythonhosted.org/packages/88/f6/01890cfd63c08f8eb96a8319b0443690197d240a8bd6346048cf7bde9190/cffi-2.1.0-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:3b926723c13eba9f81d2ef3820d63aeceec3b2d4639906047bf675cb8a7a500d", size = 210285, upload-time = "2026-07-06T21:33:12.251Z" },
+    { url = "https://files.pythonhosted.org/packages/a6/cf/2b684132056f438567b61e19d690dd31cd0921ace051e0a458be6074369e/cffi-2.1.0-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:47ff3a8bfd8cb9da1af7524b965127095055654c177fcfc7578debcb015eecd0", size = 208801, upload-time = "2026-07-06T21:33:13.617Z" },
+    { url = "https://files.pythonhosted.org/packages/6f/08/f2e7d62c460faae0926f2d6e423694aa409ced3bc1fe2927a0a6e5f05416/cffi-2.1.0-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:799416bae98336e400981ff6e532d67d5c709cfb30afb79865a1315f94b0e224", size = 221808, upload-time = "2026-07-06T21:33:15.466Z" },
+    { url = "https://files.pythonhosted.org/packages/38/37/04f54b8e63a02f3d908332c9effbf8c366167c6f733ed8a3d4f79b7e2a1e/cffi-2.1.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:961be50688f7fba2fa65f63712d3b9b341a22311f5253460ce933f52f0de1c8c", size = 225241, upload-time = "2026-07-06T21:33:16.869Z" },
+    { url = "https://files.pythonhosted.org/packages/a9/d6/c72eecca433cd3e681c65ed313ab4835d9d4a379704d0f628a6a05f51c2e/cffi-2.1.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:bf5c6cf48238b0eb4c086978c492ad1cbc22373fc5b2d7353b3a598ce6db887a", size = 223588, upload-time = "2026-07-06T21:33:18.239Z" },
+    { url = "https://files.pythonhosted.org/packages/c6/4b/e706f67279140f92939da3475ad610df18bfd52d50f14953a8e5fede71d5/cffi-2.1.0-cp313-cp313-win32.whl", hash = "sha256:db3eb7d46527159a878ec3460e9d40615bc25ba337d477db681aea6e4f05c5d2", size = 175248, upload-time = "2026-07-06T21:33:19.799Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/47/59eb7975cb0e4ef0afa764ea945b29a5bb4537a9f771cb7d6c8a5dd74c95/cffi-2.1.0-cp313-cp313-win_amd64.whl", hash = "sha256:8e74a6135550c4748af665b1b1118b6aab33b1fc6a16f9aff630af107c3b4512", size = 185717, upload-time = "2026-07-06T21:33:21.47Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/af/34fee85c48f8d94efc8597bc09470c9dd274c145f1c12e0fbc6ab6d38d74/cffi-2.1.0-cp313-cp313-win_arm64.whl", hash = "sha256:2282cd5e38aa8accd03e99d1256af8411c84cdbee6a89d841b563fdbd1f3e50f", size = 180114, upload-time = "2026-07-06T21:33:22.515Z" },
+    { url = "https://files.pythonhosted.org/packages/d8/f0/81478e482afa03f6d18dc8f2afb5edc45b3080853b634b5ed91961be0998/cffi-2.1.0-cp314-cp314-ios_13_0_arm64_iphoneos.whl", hash = "sha256:d2117334c3af3bdcb9a88522b844a2bdb5efdc4f71c6c822df55486ae1c3347a", size = 194142, upload-time = "2026-07-06T21:33:23.657Z" },
+    { url = "https://files.pythonhosted.org/packages/7d/95/8de304305cd9204974b0ca051b86d307cafca13aa575a0ef1b44d92c0d8c/cffi-2.1.0-cp314-cp314-ios_13_0_arm64_iphonesimulator.whl", hash = "sha256:702c436735fbe99d59ada02a1f65cfc0d31c0ee8b7290912f8fbc5cd1e4b16c3", size = 196819, upload-time = "2026-07-06T21:33:25.007Z" },
+    { url = "https://files.pythonhosted.org/packages/20/71/7c8372d30e42415602ed9f268f7cfd66f1b855fed881ecd168bcb45dbc0b/cffi-2.1.0-cp314-cp314-macosx_10_15_x86_64.whl", hash = "sha256:1ff3456eab0d889592d1936d6125bbfbc7ae4d3354a700f8bd80450a66445d4d", size = 184965, upload-time = "2026-07-06T21:33:26.605Z" },
+    { url = "https://files.pythonhosted.org/packages/d6/5c/584e626835f0375c928176c04137c96927165cb8733cdb3150ec04e5ee5e/cffi-2.1.0-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:c4165821e131d6d4ca444347c2b694e2311bcfa3fe5a861cc72968f28867beac", size = 184952, upload-time = "2026-07-06T21:33:27.823Z" },
+    { url = "https://files.pythonhosted.org/packages/2e/d2/065fcae1c73979fac8e054462478d0ff8a29c40cdc2ed7ea5676a061df53/cffi-2.1.0-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:276f20fffd7b396e12516ba8edf9509210ac248cbbc5acbc39cd512f9f59ebe6", size = 222353, upload-time = "2026-07-06T21:33:29.178Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/a5/e8bbb1ce5b3ac2f53ad6a10bde44318a5a8d99d4f4a000d44a6e39aeb3e4/cffi-2.1.0-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:7d5980a3433d4b71a5e120f9dd551403d7824e31e2e67124fe2769c404c06913", size = 210051, upload-time = "2026-07-06T21:33:30.534Z" },
+    { url = "https://files.pythonhosted.org/packages/28/ed/c127d3ac36e899c965e3361357c3befacd6578c03f40125183e41c3b219e/cffi-2.1.0-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:6ca4919c6e4f89aa99c42510b42cf54596892c00b3f9077f6bdd1505e24b9c8d", size = 208630, upload-time = "2026-07-06T21:33:31.753Z" },
+    { url = "https://files.pythonhosted.org/packages/cc/d7/97d3136f81db489ec8d1d67748c110d6c994268fd7528014aa9f2b085e4e/cffi-2.1.0-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:d53d10f7da99ae46f7373b9150393e9c5eab9b224909982b43832668de4779f5", size = 221593, upload-time = "2026-07-06T21:33:33.044Z" },
+    { url = "https://files.pythonhosted.org/packages/d3/27/93195977168ee63aed233a1a0993a2178798654d1f4bddcdd321d6fd3b21/cffi-2.1.0-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:c351efb95e832a853a29361675f33a7ce53de1a109cd73fd47af0712213aa4ce", size = 225146, upload-time = "2026-07-06T21:33:34.224Z" },
+    { url = "https://files.pythonhosted.org/packages/b3/c1/6dbd291ee2ae5a50a034aa057207081f545923bbf15dad4511e985aafff5/cffi-2.1.0-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:dbf7c7a88e2bac086f06d14577332760bdeecc42bdec8ac4077f6260557d9326", size = 223240, upload-time = "2026-07-06T21:33:35.57Z" },
+    { url = "https://files.pythonhosted.org/packages/0f/6f/ade5ce9863a57992a6ea3d0d10d7e29b8749fc127204b3d493d667b2815f/cffi-2.1.0-cp314-cp314-win32.whl", hash = "sha256:1854b724d00f6654c742097d5387569021be12d3a0f770eae1df8f8acfcc6acd", size = 177723, upload-time = "2026-07-06T21:33:51.626Z" },
+    { url = "https://files.pythonhosted.org/packages/41/de/92b9eeed4ae4a21d6fd9b2a2c8505cbed573299902ea73981cc13f7ff62c/cffi-2.1.0-cp314-cp314-win_amd64.whl", hash = "sha256:1b96bfe2c4bd825681b7d311ad6d9b7280a091f43e8f63da5729638083cd3bfb", size = 187937, upload-time = "2026-07-06T21:33:53.403Z" },
+    { url = "https://files.pythonhosted.org/packages/2e/1a/cc6ae6c2913a03aab8898eee57963cf1035b8df5872ed8b9115fcc7e2be8/cffi-2.1.0-cp314-cp314-win_arm64.whl", hash = "sha256:7d28dff1db6764108bc30788d85d61c876beff416d9a49cb9dd7c5a9f34f5804", size = 183001, upload-time = "2026-07-06T21:33:54.74Z" },
+    { url = "https://files.pythonhosted.org/packages/14/f0/134c00ce0779ec86dea2aa1aac69339c2741a8045072676763512363a2ea/cffi-2.1.0-cp314-cp314t-macosx_10_15_x86_64.whl", hash = "sha256:7ea6b3e2c4250ff1de21c630fe72d0f63eb95c2c32ffbf64a358cf4a8836d714", size = 188538, upload-time = "2026-07-06T21:33:36.792Z" },
+    { url = "https://files.pythonhosted.org/packages/50/d8/3b86aba791cb610d24e8a3e1b2cd529e71fa15096b04e4d4e360049d4a4c/cffi-2.1.0-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:6af371f3767faeffc6ac1ef57cdfd25844403e9d3f476c5537caee499de96376", size = 188230, upload-time = "2026-07-06T21:33:38.011Z" },
+    { url = "https://files.pythonhosted.org/packages/14/d0/117dcd9209255ad8571fbc8c92ef32593a1d294dcec91ddc4e4db50606f2/cffi-2.1.0-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:eb4e8997a49aa2c08a3e43c9045d224448b8941d88e7ac163c7d383e560cbf98", size = 223899, upload-time = "2026-07-06T21:33:39.514Z" },
+    { url = "https://files.pythonhosted.org/packages/b6/3d/f20f8b886b254e3ad10e15cd4186d3aed49f3e6a35ab37aab9f8f25f7c03/cffi-2.1.0-cp314-cp314t-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:bf01d8c84cbea96b944c73b22182e6c7c432b3475632b8111dbfdc95ddad6e13", size = 211652, upload-time = "2026-07-06T21:33:40.851Z" },
+    { url = "https://files.pythonhosted.org/packages/28/3b/fad54de07260b93ddeef4b96d0131d57ea900675df1d410ae1deee52d7a6/cffi-2.1.0-cp314-cp314t-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:33eb1ad83ebe8f313e0df035c406227d55a79456704a863fad9842136af5ad7d", size = 210755, upload-time = "2026-07-06T21:33:42.183Z" },
+    { url = "https://files.pythonhosted.org/packages/cc/82/3d5c705acb7abbba9bbd7d79b8e62e0f25b6120eb7ae6ac49f1b721722fe/cffi-2.1.0-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:ac0f1a2d0cfa7eea3f2aaf006ab6e70e8feeb16b75d65b7e5939982ca2f11056", size = 223933, upload-time = "2026-07-06T21:33:43.603Z" },
+    { url = "https://files.pythonhosted.org/packages/6c/d0/47e338384ab6b1004241002fa616301020cea4fc95f283506565d252f276/cffi-2.1.0-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:c16914df9fb7f500e440e6875fa23ff5e0b31db01fa9c06af98d59a91f0dc2e4", size = 226749, upload-time = "2026-07-06T21:33:45.046Z" },
+    { url = "https://files.pythonhosted.org/packages/70/25/65bd5b58ea4bfdfc15cde02cb5365f89ef8ab8b2adfb8fe5c4bd4233382f/cffi-2.1.0-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:5ecbd0499275d57506d397eebe1981cee87b47fcd9ef5c22cab7ed7644a39a94", size = 225703, upload-time = "2026-07-06T21:33:46.374Z" },
+    { url = "https://files.pythonhosted.org/packages/dc/78/aa01ac599a8a4322533d45a1f9bc93b338276d2d59dabbe7c6d92a775c81/cffi-2.1.0-cp314-cp314t-win32.whl", hash = "sha256:7d034dcffa09e9a46c93fa3a3be402096cb5354ac6e41ab8e5cc9cd8b642ad76", size = 182857, upload-time = "2026-07-06T21:33:47.696Z" },
+    { url = "https://files.pythonhosted.org/packages/b9/26/d00496b22de4d4228f32dde94ad996f350c8aad676d63bcca0743c8dea4d/cffi-2.1.0-cp314-cp314t-win_amd64.whl", hash = "sha256:0582a58f3051372229ca8e7f5f589f9e5632678208d8636fea3676711fdf7fe5", size = 194065, upload-time = "2026-07-06T21:33:48.953Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/dd/0c7dbf815a579ff005008a2d815a55d6bb047c349eef536d9dc53d3f0a8d/cffi-2.1.0-cp314-cp314t-win_arm64.whl", hash = "sha256:510aeeeac94811b138077451da1fb18b308a5feab47dd2b603af55804155e1c8", size = 186404, upload-time = "2026-07-06T21:33:50.309Z" },
+    { url = "https://files.pythonhosted.org/packages/55/c7/8c8c50cb11c6750051daf12164098a9a6f027ac4356967fd4d800a07f242/cffi-2.1.0-cp315-cp315-ios_13_0_arm64_iphoneos.whl", hash = "sha256:2e9dabb9abcb7ad15938c7196ad5c1718a4e6d33cc79b4c0209bdb64c4a54a5c", size = 194121, upload-time = "2026-07-06T21:33:56.109Z" },
+    { url = "https://files.pythonhosted.org/packages/99/e2/67680bf19a6b60d2bb7ff83baefa2a4c3d2d7dc0f3277034b802e1fc504c/cffi-2.1.0-cp315-cp315-ios_13_0_arm64_iphonesimulator.whl", hash = "sha256:37f525a7e7e50c017fdebe58b787be310ad59357ae43a053943a6e1a6c526001", size = 196820, upload-time = "2026-07-06T21:33:57.288Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/da/4bbe583a3b3a5c8c60892124fe17f3fa3656523faf0d3484eae90f091853/cffi-2.1.0-cp315-cp315-macosx_10_15_x86_64.whl", hash = "sha256:95f2954c2c9473d892eca6e0409f3568b37ab62a8eedb122461f73cc273476e3", size = 184936, upload-time = "2026-07-06T21:33:58.765Z" },
+    { url = "https://files.pythonhosted.org/packages/e5/4b/1f4c36ab273980d7aa75bb126ea4f8971f24a96108acad3a0a084028c57b/cffi-2.1.0-cp315-cp315-macosx_11_0_arm64.whl", hash = "sha256:cdf2448aab5f661c9315308ec8b93f4e8a1a67a3c733f8631067a2b67d5913dc", size = 185045, upload-time = "2026-07-06T21:34:00.085Z" },
+    { url = "https://files.pythonhosted.org/packages/ef/c3/ad299dc38f3583f8d916b299f028af418a9ec98bc695fcbebeae7420691c/cffi-2.1.0-cp315-cp315-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:90bec57cf82089383bd06a605b3eb8daebf7e5a668520beaf6e327a83a947699", size = 222342, upload-time = "2026-07-06T21:34:01.814Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/d8/df4543cc087245044ed02ef3ad8e0a26619d0075ac7a77a12dc81177851b/cffi-2.1.0-cp315-cp315-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:6274dcb2d15cef48daa73ed1be5a40d501d74dccd0cd6db364776d12cb6ba022", size = 210073, upload-time = "2026-07-06T21:34:03.255Z" },
+    { url = "https://files.pythonhosted.org/packages/2c/0e/fac738d73728c6cea2a88a2883dca54892496cbba88a1dc1f2909cb8a6f5/cffi-2.1.0-cp315-cp315-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:2b71d409cccee78310ab5dec549aed052aaea483346e282c7b02362596e01bb0", size = 208551, upload-time = "2026-07-06T21:34:04.433Z" },
+    { url = "https://files.pythonhosted.org/packages/e6/3f/0b04a700dd64f465c93020253a793a82c9b4dff9961f48facd0df945d9b8/cffi-2.1.0-cp315-cp315-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:7d3538f9c0e50670f4deb93dbb696576e60590369cae2faf7de681e597a8a1f1", size = 221649, upload-time = "2026-07-06T21:34:06.157Z" },
+    { url = "https://files.pythonhosted.org/packages/5d/7c/b7379a5704c79eda57ce075869ba70a0368d1c850f803b3c0d078d39dcaf/cffi-2.1.0-cp315-cp315-musllinux_1_2_aarch64.whl", hash = "sha256:8f9ec95b8a043d3dfbc74d9abc6f7baf524dd27a8dc160b0a32ff9cdab650c28", size = 225203, upload-time = "2026-07-06T21:34:07.489Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/02/d5e6c43ea85c41bda2a184a3418f195fe7cf602967a8d2b94e085b83deef/cffi-2.1.0-cp315-cp315-musllinux_1_2_x86_64.whl", hash = "sha256:af5e2915d41fe6c961694d7bfdc8562942638200f3ce2765dfb8b745cf997629", size = 223263, upload-time = "2026-07-06T21:34:08.712Z" },
+    { url = "https://files.pythonhosted.org/packages/2c/d8/772b8259bf75749adffb1c546828978381fb516f60cf701f6c83daf60c85/cffi-2.1.0-cp315-cp315-win32.whl", hash = "sha256:0a42c688d19fca6e095a53c6a6e2295a5b050a8b289f109adab02a9e61a25de6", size = 177696, upload-time = "2026-07-06T21:34:26.355Z" },
+    { url = "https://files.pythonhosted.org/packages/2f/dd/afa2191fc6d57fedd26e5844a2fe2fcc0bbfa00961bbaa5a41e4921e7cca/cffi-2.1.0-cp315-cp315-win_amd64.whl", hash = "sha256:bccbbb5ee76a61f9d99b5bf3846a51d7fca4b6a732fe46f89295610edaf41853", size = 187914, upload-time = "2026-07-06T21:34:27.58Z" },
+    { url = "https://files.pythonhosted.org/packages/05/ef/6cd4f8c671517162379dc79cfae5aea9106bc38abb89628d5c16adf6a838/cffi-2.1.0-cp315-cp315-win_arm64.whl", hash = "sha256:8d35c139744adb3e727cd51b1a18324bbe44b8bd41bf8322bca4d41289f48eda", size = 183004, upload-time = "2026-07-06T21:34:28.905Z" },
+    { url = "https://files.pythonhosted.org/packages/11/b6/12fc55092817a5faa26fb8c40c7f9d662e11a46ee248c137aafc42517d92/cffi-2.1.0-cp315-cp315t-macosx_10_15_x86_64.whl", hash = "sha256:f9912624a0c0b834b7520d7769b3644453aabc0a7e1c839da7359f050750e9bc", size = 188378, upload-time = "2026-07-06T21:34:09.926Z" },
+    { url = "https://files.pythonhosted.org/packages/8d/2e/cdac88979f295fde5daa69622c7d2111e56e7ceb94f211357fbe452339e4/cffi-2.1.0-cp315-cp315t-macosx_11_0_arm64.whl", hash = "sha256:df92f2aba50eb4d96718b68ef76f2e57a57b54f2fa62333496d16c6d585a85ca", size = 188319, upload-time = "2026-07-06T21:34:11.101Z" },
+    { url = "https://files.pythonhosted.org/packages/e0/27/1d0b408497e41a74795af122d7b603c418c5fed0171450f899afd04e594f/cffi-2.1.0-cp315-cp315t-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:0520e1f4c35f44e209cbbb421b67eec42e6a157f59444dfb6058874ff3610e5d", size = 223904, upload-time = "2026-07-06T21:34:12.606Z" },
+    { url = "https://files.pythonhosted.org/packages/8b/31/e115c985105dd7ffb32444505f18ceb874bb42d992af05d5dced7ecf1980/cffi-2.1.0-cp315-cp315t-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:3681e031db29958a7502f5c0c9d6bbc4c36cb20f7b104086fa642d1799631ff8", size = 211554, upload-time = "2026-07-06T21:34:13.987Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/67/9e6e09409336d9e515c58367e7cfcf4f89df06ad25252675595a58eb59d5/cffi-2.1.0-cp315-cp315t-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:762f99479dcb369f60ab9017ad4ab97a36a1dd7c1ee5a3b15db0f4b8659120cd", size = 210795, upload-time = "2026-07-06T21:34:15.972Z" },
+    { url = "https://files.pythonhosted.org/packages/19/e5/d3cc82a4a0be7902af279c04181ad038449c096734464a5ae1de3e1401bd/cffi-2.1.0-cp315-cp315t-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:0611e7ebf90573a535ebdc33ae9da222d037853983e13359f580fab781ca017f", size = 223843, upload-time = "2026-07-06T21:34:17.509Z" },
+    { url = "https://files.pythonhosted.org/packages/b9/65/b434abc97ce7cecc2c640fde160507c0ecc7e21544b483ba3325d2e2ea17/cffi-2.1.0-cp315-cp315t-musllinux_1_2_aarch64.whl", hash = "sha256:86cf8755a791f72c85dc287128cc62d4f24d392e3f1e15837245623f4a33cccc", size = 226773, upload-time = "2026-07-06T21:34:19.05Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/9f/d4dc66ca651eb1145a133314cda721abf13cfac3d28c4a0402263ae6ad75/cffi-2.1.0-cp315-cp315t-musllinux_1_2_x86_64.whl", hash = "sha256:ba00f661f8ba35d075c937174e27c2c421cec3942fd2e0ea3e66996757c0fdd9", size = 225719, upload-time = "2026-07-06T21:34:20.576Z" },
+    { url = "https://files.pythonhosted.org/packages/68/5a/e536c528bc8057496c360c0978559a2dc45653f89dd6151078aa7d8fca1a/cffi-2.1.0-cp315-cp315t-win32.whl", hash = "sha256:cb96698e3c7413d906ce83f8ffd245ec1bd94707541f299d0ce4d6b0193e982b", size = 182760, upload-time = "2026-07-06T21:34:22.059Z" },
+    { url = "https://files.pythonhosted.org/packages/d3/0b/0ffe8b82d3875bced5fa1e7986a7a46b748262a40ab7f60b475eb9fb1bb3/cffi-2.1.0-cp315-cp315t-win_amd64.whl", hash = "sha256:f146d154428a2523f9cc7936c02353c2459b8f6cf07d3cd1ee1c0a611109c5d5", size = 193769, upload-time = "2026-07-06T21:34:23.589Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/17/1073b53b68c9b5ca6914adf5f8bf55aacc2d3be102418c90700160ea8605/cffi-2.1.0-cp315-cp315t-win_arm64.whl", hash = "sha256:cbb7640ce37159548d2147b5b8c241f962143d4c71231431820783f4dc78f210", size = 186405, upload-time = "2026-07-06T21:34:24.857Z" },
+]
+
+[[package]]
+name = "charset-normalizer"
+version = "3.4.9"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/bd/2a/23f34ec9d04624958e137efdc394888716353190e75f25dd22c7a2c7a8aa/charset_normalizer-3.4.9.tar.gz", hash = "sha256:673611bbd43f0810bec0b0f028ddeaaa501190339cac411f347ac76917c3ae7b", size = 152439, upload-time = "2026-07-07T14:34:58.454Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/70/4a/ecbd131485c07fcdfad54e28946d513e3da22ef3b4bd854dcafae54ec739/charset_normalizer-3.4.9-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:45b0cc4e3556cd875e09102988d1ab8356c998b596c9fced84547c8138b487a0", size = 319300, upload-time = "2026-07-07T14:33:15.666Z" },
+    { url = "https://files.pythonhosted.org/packages/ec/96/5d9364e3342d69f3a045e1777bc47c85c383e6e9466d561b33fdb419d1f9/charset_normalizer-3.4.9-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:9b2aff1c7b3884512b9512c3eaadd9bab39fb45042ffaaa1dd08ff2b9f8109d9", size = 215802, upload-time = "2026-07-07T14:33:17.031Z" },
+    { url = "https://files.pythonhosted.org/packages/4b/4c/5361f9aa7f2cb58d94f2ab831b3d493f69efb1d239654b4744e3c09527cb/charset_normalizer-3.4.9-cp312-cp312-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:9104ed0bd76a429d46f9ec0dbc9b08ad1d2dcdf2b00a5a0daa1c145329b35b44", size = 237171, upload-time = "2026-07-07T14:33:18.576Z" },
+    { url = "https://files.pythonhosted.org/packages/50/78/ce342ca4ff30b2eb49fe6d9578df85974f90c67d294113e94efdd9664cbd/charset_normalizer-3.4.9-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:7b86a2b16095d250c6f58b3d9b2eee6f4147754344f3dab0922f7c9bf7d226c9", size = 233075, upload-time = "2026-07-07T14:33:20.084Z" },
+    { url = "https://files.pythonhosted.org/packages/01/c4/4fa4c8b3097a11f3c5f09a35b72ed6855fb1d332469504962ab7bafcc702/charset_normalizer-3.4.9-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:5e226f6218febc71f6c1fc2fafb91c226f75bdc1d8fb12d66823716e891608fd", size = 224256, upload-time = "2026-07-07T14:33:21.747Z" },
+    { url = "https://files.pythonhosted.org/packages/87/3a/ad914516df7e358a81aae018caa5e0470ba827fa6d763b1d2e87d920a5f6/charset_normalizer-3.4.9-cp312-cp312-manylinux_2_31_armv7l.whl", hash = "sha256:90c44bc373b7687f6948b693cceaea1348ae0975d7474746559494468e3c1d84", size = 208784, upload-time = "2026-07-07T14:33:23.313Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/74/3c12f9755717dfe5c5c87da63f35d765fa0c00382ec26bf23f7fae34f2ba/charset_normalizer-3.4.9-cp312-cp312-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:9cdef90ae47919cae358d8ab15797a800ed41da7aba5d72419fb510729e2ed4b", size = 219928, upload-time = "2026-07-07T14:33:24.814Z" },
+    { url = "https://files.pythonhosted.org/packages/33/9a/895095b83e7907abd6d3d99aad3a38ad0d9686cc186cb0c94c24320fe63e/charset_normalizer-3.4.9-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:60f44ade2cf573dad7a277e6f8ca9a51a21dda572b13bd7d8539bb3cd5dbedde", size = 218489, upload-time = "2026-07-07T14:33:26.42Z" },
+    { url = "https://files.pythonhosted.org/packages/a1/34/ef5c05f412f42520d7709b7d3784d19640839eb7366ded1755511585429f/charset_normalizer-3.4.9-cp312-cp312-musllinux_1_2_armv7l.whl", hash = "sha256:a1786910334ed46ab1dd73222f2cd1e05c2c3bb39f6dddb4f8b36fc382058a39", size = 210267, upload-time = "2026-07-07T14:33:27.952Z" },
+    { url = "https://files.pythonhosted.org/packages/83/dc/9b29fa4412b318bf3bfea985c35d67eb55e04b59a7c3f2237168b0e0be6f/charset_normalizer-3.4.9-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:03d07803992c6c7bbc976327f34b18b6160327fc81cb82c9d504720ac0be3b62", size = 226030, upload-time = "2026-07-07T14:33:29.397Z" },
+    { url = "https://files.pythonhosted.org/packages/0e/42/6dbc00b8cd16011691203e33570fa42ed5746599a2e878112d16eab403a3/charset_normalizer-3.4.9-cp312-cp312-win32.whl", hash = "sha256:78841cccf1af7b40f6f716338d50c0902dbe88d9f800b3c973b7a9a0a693a642", size = 151185, upload-time = "2026-07-07T14:33:30.781Z" },
+    { url = "https://files.pythonhosted.org/packages/80/cc/f920afd1a23c58ccd53c1d36085a71893a4737ff5e66e0371efab6809850/charset_normalizer-3.4.9-cp312-cp312-win_amd64.whl", hash = "sha256:4b3dac63058cc36820b0dd072f89898604e2d39686fe05321729d00d8ac185a0", size = 162557, upload-time = "2026-07-07T14:33:32.176Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/e6/0386d43a261ff4e4b30c5857af7df877254b46bec7b9d1b74b6bf969a90b/charset_normalizer-3.4.9-cp312-cp312-win_arm64.whl", hash = "sha256:78fa18e436a1a0e58dbd7e02fc4473f3f32cceb12df9dfca542d075961c307d2", size = 152665, upload-time = "2026-07-07T14:33:33.711Z" },
+    { url = "https://files.pythonhosted.org/packages/b2/06/97ec2aeae780b31d742b6352218b43841a6871e2564578ca522dce4a45c3/charset_normalizer-3.4.9-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:440eede837960000d74978f0eba527be106b5b9aee0daf779d395276ed0b0614", size = 317688, upload-time = "2026-07-07T14:33:35.408Z" },
+    { url = "https://files.pythonhosted.org/packages/d0/39/8ff066c672434225f8d25f8b739f992af250944392173dcc88362681c9bf/charset_normalizer-3.4.9-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:21e764fd1e70b6a3e205a0e46f3051701f98a8cb3fad66eeb80e48bb502f8698", size = 214982, upload-time = "2026-07-07T14:33:36.996Z" },
+    { url = "https://files.pythonhosted.org/packages/92/8f/3a47a3667c83c2df9483d91644c6c107de3bf8874aa1793da9d3012eb986/charset_normalizer-3.4.9-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:e4fd89cc178bced6ad29cb3e6dd4aa63fa5017c3524dbd0b25998fb64a87cc8b", size = 236460, upload-time = "2026-07-07T14:33:38.536Z" },
+    { url = "https://files.pythonhosted.org/packages/f1/60/b22cdbee7e4013dab8b0d7647fc6181120fbbbc8f7025c226d15bd5a47fc/charset_normalizer-3.4.9-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:bd47ba7fc3ca94896759ea0109775132d3e7ab921fbf54038e1bab2e46c313c9", size = 232003, upload-time = "2026-07-07T14:33:40.059Z" },
+    { url = "https://files.pythonhosted.org/packages/ea/f8/72eb13dcabe7257035cea8aefd922caad2f110d252bf9f67c4c2ca763aee/charset_normalizer-3.4.9-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:84fd18bcc17526fc2b3c1af7d2b9217d32c9c04448c16ec693b9b4f1985c3d33", size = 223149, upload-time = "2026-07-07T14:33:41.631Z" },
+    { url = "https://files.pythonhosted.org/packages/b0/3e/faee8f9de92b14ee1198e9163252bb15efee7301b31256a3b6d9ebfdd0dd/charset_normalizer-3.4.9-cp313-cp313-manylinux_2_31_armv7l.whl", hash = "sha256:5b10cd92fc5c498b35a8635df6d5a100207f88b63a4dc1de7ef9a548e1e2cd63", size = 207901, upload-time = "2026-07-07T14:33:43.209Z" },
+    { url = "https://files.pythonhosted.org/packages/3a/25/45f30093ae27dd7b92a793b61882a38685f993700113ca36e0c9c14965e1/charset_normalizer-3.4.9-cp313-cp313-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:a4fbdde9dd4a9ce5fd52c2b3a347bb50cc89483ef783f1cb00d408c13f7a96c0", size = 219176, upload-time = "2026-07-07T14:33:44.725Z" },
+    { url = "https://files.pythonhosted.org/packages/48/18/c8f397329c35e32f6a837e488986f4ae03bd2abebc453b48714991630c2f/charset_normalizer-3.4.9-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:416c229f77e5ea25b3dfd4b582f8d73d7e43c22320302b9ab128a2d3a0b38efe", size = 217356, upload-time = "2026-07-07T14:33:46.192Z" },
+    { url = "https://files.pythonhosted.org/packages/86/7e/5ce0bba863470fd1902d5e5843968951bddf38abe4742fc97116ef4598b3/charset_normalizer-3.4.9-cp313-cp313-musllinux_1_2_armv7l.whl", hash = "sha256:75286256590a6320cf106a0d28970d3560aad9ee09aa7b34fb40524792436d35", size = 209614, upload-time = "2026-07-07T14:33:47.705Z" },
+    { url = "https://files.pythonhosted.org/packages/6c/ef/2473d3c4d869155be4af1191111d59c4d5c4e0173026f7e85b176e23bf65/charset_normalizer-3.4.9-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:69b157c5d3292bcd443faca052f3096f637f1e074b98212a933c074ae23dc3b8", size = 224991, upload-time = "2026-07-07T14:33:49.238Z" },
+    { url = "https://files.pythonhosted.org/packages/d0/a3/53ddae3db108a088156aa8ddfafd411ebbc1340f48c5573f697b27f69a39/charset_normalizer-3.4.9-cp313-cp313-win32.whl", hash = "sha256:51307f5c71007673a2bf8232ad973483d281e74cb99c8c5a990af1eefa6277d9", size = 150622, upload-time = "2026-07-07T14:33:50.711Z" },
+    { url = "https://files.pythonhosted.org/packages/e8/ef/6953a77c7cf2c2ff9998e6f575ab3e380119f100223381565a4f94c1f836/charset_normalizer-3.4.9-cp313-cp313-win_amd64.whl", hash = "sha256:fe2c7201c642b7c308f1675355ad7ff7b66acfe3541625efe5a3ad38f29d6115", size = 161947, upload-time = "2026-07-07T14:33:52.197Z" },
+    { url = "https://files.pythonhosted.org/packages/6e/fb/d560d1d1555debbfe7849d9cac6145c1b537709d79576bf22557ed803b82/charset_normalizer-3.4.9-cp313-cp313-win_arm64.whl", hash = "sha256:611057cc5d5c0afc743ba8be6bd828c17e0aaa8643f9d0a9b9bb7dea80eb8012", size = 152594, upload-time = "2026-07-07T14:33:53.486Z" },
+    { url = "https://files.pythonhosted.org/packages/7e/8d/496817fa0944239ecae662dd57ea765cfeaec6a735f9f025d4b7b72e7143/charset_normalizer-3.4.9-cp314-cp314-macosx_10_15_universal2.whl", hash = "sha256:0327fcd59a935777d83410750c50600ee9571af2846f71ce40f25b13da1ef380", size = 317253, upload-time = "2026-07-07T14:33:54.994Z" },
+    { url = "https://files.pythonhosted.org/packages/2b/f9/ef4a69ea338ad3c0deceea0f5f7d2380ae8b52132b06d652cb0d2cd86706/charset_normalizer-3.4.9-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:8a79d9f4d8001473a30c163556b3c3bfebec837495a412dde78b51672f6134f9", size = 215898, upload-time = "2026-07-07T14:33:56.334Z" },
+    { url = "https://files.pythonhosted.org/packages/8c/e7/5ddfd76fc061eb52de219658a4aa431cbacadf0a0219c8854f00da50d289/charset_normalizer-3.4.9-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:33bdcc2a32c0a0e861f60841a512c8acc658c87c2ac59d89e3a46dacf7d866e4", size = 236718, upload-time = "2026-07-07T14:33:57.9Z" },
+    { url = "https://files.pythonhosted.org/packages/49/ba/768fa3f36048d81c477a0ce61f813bc1454d80917ccfe550abd9f44f5e24/charset_normalizer-3.4.9-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:f840ed6d8ecba8255df8c42b87fadeda98ddfc6eeec05e2dc66e26d46dd6f58a", size = 232519, upload-time = "2026-07-07T14:33:59.811Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/c4/b3e049d2aa3766180c78507110543d9d50894cc97f57de543f1be521dcdc/charset_normalizer-3.4.9-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:c25fe15c70c59eb7c5ce8c06a1f3fa1da0ecc5ea1e7a5922c40fd2fa9b0d5046", size = 223143, upload-time = "2026-07-07T14:34:01.517Z" },
+    { url = "https://files.pythonhosted.org/packages/19/79/55c32d06d76ae4feafe053f061f3e3ab70bcf19f4007797ce8c3efda7830/charset_normalizer-3.4.9-cp314-cp314-manylinux_2_31_armv7l.whl", hash = "sha256:f7fb7d750cfa0a070d2c24e831fd3481019a60dd317ea2b39acbcebc08b6ed81", size = 206742, upload-time = "2026-07-07T14:34:03.04Z" },
+    { url = "https://files.pythonhosted.org/packages/10/e0/47c079dd82d217c807479cd59ffd30af56307ea31c108b75758970459ad3/charset_normalizer-3.4.9-cp314-cp314-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:4d1c96a7a18b9690a4d46df09e3e3382406ae3213727cd1019ebade1c4a81917", size = 219191, upload-time = "2026-07-07T14:34:04.657Z" },
+    { url = "https://files.pythonhosted.org/packages/42/ab/b9bc2e77d6b44a7e46ef62ec5cac1c9a6ba7b9135a5d560f002696ec9995/charset_normalizer-3.4.9-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:a4cfde78a9f2880208d16a93b795726a3017d5977e08d1e162a7a31322479c41", size = 218328, upload-time = "2026-07-07T14:34:06.115Z" },
+    { url = "https://files.pythonhosted.org/packages/f1/78/c9c71d599f5aa2d42bcdd35cbbd46d7f535351a57e40ff7d8e5a7e219401/charset_normalizer-3.4.9-cp314-cp314-musllinux_1_2_armv7l.whl", hash = "sha256:d4d6fcde76f94f5cb9e43e9e9a61f16dacefd228cbbf6f1a09bd9b219a92f1a1", size = 207406, upload-time = "2026-07-07T14:34:07.554Z" },
+    { url = "https://files.pythonhosted.org/packages/f6/39/c914445c321a845097ce4f6ac7de9a18228a77b766272125a1ce00d851eb/charset_normalizer-3.4.9-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:898f0e9068ca27d37f8e83a5b962821df851532e6c4a7d615c1c033f9da6eedf", size = 225157, upload-time = "2026-07-07T14:34:09.061Z" },
+    { url = "https://files.pythonhosted.org/packages/9b/f2/c0d4b8508565a36bc5c624e88ed297f5b0b1095011034d7f5b83a69908b5/charset_normalizer-3.4.9-cp314-cp314-win32.whl", hash = "sha256:c1c948747b03be832dceed96ca815cef7360de9aa19d37c730f8e3f6101aca48", size = 151095, upload-time = "2026-07-07T14:34:10.901Z" },
+    { url = "https://files.pythonhosted.org/packages/49/fd/a1d26144398c67486422a72bf5812cda22cb4ccfcd95a290fb41ceb4b8e2/charset_normalizer-3.4.9-cp314-cp314-win_amd64.whl", hash = "sha256:16b65ea0f2465b6fb52aa22de5eca612aa964ddfec00a912e26f4656cbef890b", size = 162796, upload-time = "2026-07-07T14:34:12.47Z" },
+    { url = "https://files.pythonhosted.org/packages/20/95/d75e82f8ce9fd323ebf059c16c9aadefb22a1ecde13b7840b35835e4886c/charset_normalizer-3.4.9-cp314-cp314-win_arm64.whl", hash = "sha256:40a126142a56b2dfc0aacbad1de8310cbf60da7656db0e6b16eebd48e3e93519", size = 153334, upload-time = "2026-07-07T14:34:14.044Z" },
+    { url = "https://files.pythonhosted.org/packages/00/5e/17398df3a139985ba9d11ed072531986f408c8fca952835ef1ab1820c02b/charset_normalizer-3.4.9-cp314-cp314t-macosx_10_15_universal2.whl", hash = "sha256:609b3ba8fcc0fb5ab7af00719d0fb6ad0cb518e48e7712d12fd68f1327951198", size = 338848, upload-time = "2026-07-07T14:34:15.688Z" },
+    { url = "https://files.pythonhosted.org/packages/cd/91/7253a32e86b7e1d1239b1b36ba6dd0f021a21107ab33054b53119cc083b9/charset_normalizer-3.4.9-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:51447e9aa2684679af07ca5021c3db526e0284347ebf4ffcec1154c3350cfe32", size = 223022, upload-time = "2026-07-07T14:34:17.248Z" },
+    { url = "https://files.pythonhosted.org/packages/cb/32/2e64bd2be10e89c61e57ebe6a93fd98ae88eb7ebe414b5121f22c96c69eb/charset_normalizer-3.4.9-cp314-cp314t-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:cc1b0fff8ead343dae06305f954eb8468ba0ec1a97881f42489d198e4ce3c632", size = 241590, upload-time = "2026-07-07T14:34:18.813Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/ef/d96ec496cfea0c21db43b0ad03891308b02388d054cc902cf0e5a1ad6a88/charset_normalizer-3.4.9-cp314-cp314t-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:fa36ec09ef71d158186bc79e359ff5fdd6e7996fe8ab638f00d6b93139ba4fcf", size = 239584, upload-time = "2026-07-07T14:34:20.52Z" },
+    { url = "https://files.pythonhosted.org/packages/d4/ce/9af95f7876194bd7a14e3dfe4a4de2e0bff02666a3910d72beafd06cc297/charset_normalizer-3.4.9-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:df115d4d83168fdf2cae48ef1ff6d1cb4c466364e30861b37121de0f3bf1b990", size = 230224, upload-time = "2026-07-07T14:34:22.189Z" },
+    { url = "https://files.pythonhosted.org/packages/52/94/af74dde74a3996bd959c350709bfe50e297823d70a8c1cbd54b838880863/charset_normalizer-3.4.9-cp314-cp314t-manylinux_2_31_armv7l.whl", hash = "sha256:f86c6358749bd4fda175388691e3ba8c46e24c5347d0afd20f9b7edfc9faf07d", size = 212667, upload-time = "2026-07-07T14:34:23.857Z" },
+    { url = "https://files.pythonhosted.org/packages/ee/f0/f1c4fe746c395922961b5916ed1d7d6e7d4c84851d19ed43cc89980ec953/charset_normalizer-3.4.9-cp314-cp314t-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:32286a2c8d167e897177b673176c1e3e00d4057caf5d2b64eef9a3666b03018e", size = 227179, upload-time = "2026-07-07T14:34:25.586Z" },
+    { url = "https://files.pythonhosted.org/packages/e4/56/6c745619ac397e8871e2bcd3cea1eec86b877488f33888b3aef5c3ed506e/charset_normalizer-3.4.9-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:83aed2c10721ddd90f68140685391b50811a880af20654c59af6b6c66c40513c", size = 225372, upload-time = "2026-07-07T14:34:27.212Z" },
+    { url = "https://files.pythonhosted.org/packages/78/ad/98aae8630ac71f16711968e38a5acfecce41b778bf2f0312851020f565a8/charset_normalizer-3.4.9-cp314-cp314t-musllinux_1_2_armv7l.whl", hash = "sha256:cd6c3d4b783c556fa00bf540854e42f135e2f256abd29669fcd0da0f2dec79c2", size = 215222, upload-time = "2026-07-07T14:34:28.774Z" },
+    { url = "https://files.pythonhosted.org/packages/f7/40/9593d54209765207a7f11073c06494c1721e4ca4a0a426c597679bf7f91e/charset_normalizer-3.4.9-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:ee2f2a527e3c1a6e6411eb4209642e138b544a2d72fe5d0d76daf77b24063534", size = 231958, upload-time = "2026-07-07T14:34:30.345Z" },
+    { url = "https://files.pythonhosted.org/packages/b1/27/693ee5e8a18191eb38647360c51cd505013e2bd3b366aa43fd5344c21e3c/charset_normalizer-3.4.9-cp314-cp314t-win32.whl", hash = "sha256:0d861473f743244d349b50f850d10eb87aeb22bbdcc8e64f79273c94af5a8226", size = 155580, upload-time = "2026-07-07T14:34:31.884Z" },
+    { url = "https://files.pythonhosted.org/packages/80/3f/bd97d3d9c613013d07cb7733d299385b41df37f0471310f5a73dc359f0b8/charset_normalizer-3.4.9-cp314-cp314t-win_amd64.whl", hash = "sha256:9b8e0f3107e2200b76f6054de99016eac3ee6762713587b36baaa7e4bd2ae177", size = 167620, upload-time = "2026-07-07T14:34:33.438Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/c6/eee9dca4439b1061f76373f06ea855678cc4a64c1c3c90b50e479edbb8eb/charset_normalizer-3.4.9-cp314-cp314t-win_arm64.whl", hash = "sha256:19ac87f93086ce37b86e098888555c4b4bc48102279bae3350098c0ed664b501", size = 158037, upload-time = "2026-07-07T14:34:35.018Z" },
+    { url = "https://files.pythonhosted.org/packages/98/2b/f97f1c193fb855c345d678f5077d6926034db0722df74c8f057020e05a25/charset_normalizer-3.4.9-py3-none-any.whl", hash = "sha256:68e5f26a1ad57ded6d1cfb85331d1c1a195314756471d97758c48498bb4dcdf5", size = 64538, upload-time = "2026-07-07T14:34:56.993Z" },
+]
+
+[[package]]
+name = "click"
+version = "8.4.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "colorama", marker = "sys_platform == 'win32'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/76/d4/81420972a676e8ffea40450d8c8c92943e7218a78fe9b64359836cc9876b/click-8.4.2.tar.gz", hash = "sha256:9a6cea6e60b17ebe0a44c5cc636d94f09bd66142c1cd7d8b4cd731c4917a15f6", size = 338000, upload-time = "2026-06-24T17:45:15.148Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/fb/e2/79c688af8b210d232694e31e59da9f6ec747bae31c3f5946e4e9b98860d5/click-8.4.2-py3-none-any.whl", hash = "sha256:e6f9f66136c816745b9d65817da91d61d957fb16e02e4dcd0552553c5a197b76", size = 119243, upload-time = "2026-06-24T17:45:13.73Z" },
+]
+
+[[package]]
+name = "colorama"
+version = "0.4.6"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/d8/53/6f443c9a4a8358a93a6792e2acffb9d9d5cb0a5cfd8802644b7b1c9a02e4/colorama-0.4.6.tar.gz", hash = "sha256:08695f5cb7ed6e0531a20572697297273c47b8cae5a63ffc6d6ed5c201be6e44", size = 27697, upload-time = "2022-10-25T02:36:22.414Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/d1/d6/3965ed04c63042e047cb6a3e6ed1a63a35087b6a609aa3a15ed8ac56c221/colorama-0.4.6-py2.py3-none-any.whl", hash = "sha256:4f1d9991f5acc0ca119f9d443620b77f9d6b33703e51011c16baf57afb285fc6", size = 25335, upload-time = "2022-10-25T02:36:20.889Z" },
+]
+
+[[package]]
+name = "cryptography"
+version = "49.0.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "cffi", marker = "platform_python_implementation != 'PyPy'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/1f/99/d1c90d6041656cc6ee229dc99cd67fd0cd5aec3c5f7d72fffc27cc750054/cryptography-49.0.0.tar.gz", hash = "sha256:f89660a348f4f78a92366240a61404e337586ef7f5909a2fef59ca88ef505493", size = 854345, upload-time = "2026-06-12T20:02:30.512Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/9b/22/adf66990e63584a68dfb50c24f48a125c07b1699899381c8151e63ed458c/cryptography-49.0.0-cp311-abi3-macosx_11_0_arm64.whl", hash = "sha256:966fe0e9c67490071f14c0d2b1cb2dfb3023c5ce39457343931415f08382f2db", size = 4032100, upload-time = "2026-06-12T20:02:32.143Z" },
+    { url = "https://files.pythonhosted.org/packages/09/41/3797cfaf69cae04a13ee78ebd83f0678d9c02b4779d21ce24445326f1a69/cryptography-49.0.0-cp311-abi3-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:36d1709f992593689b45bda411498d62c6e365f2ca00b84657d4dadd24de16db", size = 4692978, upload-time = "2026-06-12T20:01:21.305Z" },
+    { url = "https://files.pythonhosted.org/packages/e6/8b/43011f7ebe515a8aa20d61f290a326cd890c2e738e16e59eaff8d9c3a412/cryptography-49.0.0-cp311-abi3-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:0e959b578856a3924bc0cbb710fc12c387b9412a951389f3ca61704a9e25f325", size = 4716422, upload-time = "2026-06-12T20:01:48.566Z" },
+    { url = "https://files.pythonhosted.org/packages/4a/91/01ce7303a4579e6d3a6abef01bd322848e9ea7a219adcabc5048b9033571/cryptography-49.0.0-cp311-abi3-manylinux_2_28_aarch64.whl", hash = "sha256:53ecee2e23f7169b6117e99fc8a944e5e50f79e69758a83b52a00cb98ab2b2d2", size = 4700503, upload-time = "2026-06-12T20:02:47.091Z" },
+    { url = "https://files.pythonhosted.org/packages/62/99/a2c95cf8293f07491e9e27c20cc4dcd18176d944e674679adeb1d0173fd6/cryptography-49.0.0-cp311-abi3-manylinux_2_28_ppc64le.whl", hash = "sha256:2eda353d8a27bcbcaa4cbed18994a74ab4d19a2ca897db188ea269ab9b71419b", size = 5309779, upload-time = "2026-06-12T20:02:08.987Z" },
+    { url = "https://files.pythonhosted.org/packages/20/2c/0622f20ff02b2ef32558733443805dc82fd4c275be01b2d19d14676f3a1b/cryptography-49.0.0-cp311-abi3-manylinux_2_28_x86_64.whl", hash = "sha256:2afe9051da7ae7bd5905da5a949280c7d2bb75682e188f650a9d0f2756b834c6", size = 4749683, upload-time = "2026-06-12T20:02:03.335Z" },
+    { url = "https://files.pythonhosted.org/packages/a3/5b/c5246635d5fd3b64e0d45ae10e99fd32fe9676a79915ccfe5a61ba9af1a5/cryptography-49.0.0-cp311-abi3-manylinux_2_31_armv7l.whl", hash = "sha256:0b82e28ee398a386f0807bba7884d30f25218855690f45115831bcce5d90822c", size = 4337874, upload-time = "2026-06-12T20:02:54.323Z" },
+    { url = "https://files.pythonhosted.org/packages/6d/88/05563c7fe2e914e87d1a536d06fe83e66b4e1d95cb593e05aea375531da8/cryptography-49.0.0-cp311-abi3-manylinux_2_34_aarch64.whl", hash = "sha256:ccac2bfebc306b862133e3bb71f3f6ee8bb525240089b2d952e4144b3a6d5da7", size = 4700283, upload-time = "2026-06-12T20:01:34.822Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/b6/d7696e4e890d6ae1469935164c9e5215c557671cb78d6e3f458ccceaa632/cryptography-49.0.0-cp311-abi3-manylinux_2_34_ppc64le.whl", hash = "sha256:d0527ce944105f257f605a827d6ebead966c752038b6e8656abb9c5edee6fc68", size = 5265844, upload-time = "2026-06-12T20:01:24.09Z" },
+    { url = "https://files.pythonhosted.org/packages/a9/3c/f3ad17eecc1a57b0ba236dc01f90e783c51f4a2f35f64777cc4f47a184b2/cryptography-49.0.0-cp311-abi3-manylinux_2_34_x86_64.whl", hash = "sha256:cbc77da8c523d5abd028635ba850a6966fcee2c82e2bf65a41d1d8afe0f98be9", size = 4749290, upload-time = "2026-06-12T20:01:30.848Z" },
+    { url = "https://files.pythonhosted.org/packages/4f/01/339573cf1023163a400b0b5d16f6d507de413b9f60be6fd1b77feeaf6737/cryptography-49.0.0-cp311-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:b87e65d263b3e5d3bb92a57e2a6638e2f31110fa7aa890c7b2dbba42248d0a3f", size = 4834612, upload-time = "2026-06-12T20:01:29.246Z" },
+    { url = "https://files.pythonhosted.org/packages/71/fd/577302e213a1be9468f92d1afef66fcf1ef83d516819d9992ca547f592bd/cryptography-49.0.0-cp311-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:66ec79c3904820572d7e987abdf304281f141d37ad9a489b8e97066e7b9b6459", size = 4980804, upload-time = "2026-06-12T20:01:42.853Z" },
+    { url = "https://files.pythonhosted.org/packages/1f/09/f42b1d190c5ba75f72062a387f8030d1d75f6ab035788f1d9c4b01de6525/cryptography-49.0.0-cp311-abi3-win_amd64.whl", hash = "sha256:e5dfc1e64de5677cec922ffa8da89c546d0415bf6efdf081842e5d44c84e1f0e", size = 3810026, upload-time = "2026-06-12T20:02:39.262Z" },
+    { url = "https://files.pythonhosted.org/packages/ec/9e/db72b3ae7fc9cfad53e630e56c6ae83b9b6ff0bf3718ffb8012d20b3aabf/cryptography-49.0.0-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:73a205dce83953d131a4aa1e0fd917a2fd1c5b1eef251e9d7152efefcbf5caf7", size = 4013892, upload-time = "2026-06-12T20:02:10.735Z" },
+    { url = "https://files.pythonhosted.org/packages/86/12/c48a424f38db03027be9f7ed5c7dc5de9933dbee992865f98b13727a009d/cryptography-49.0.0-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:196ecd6a36e4e9aa10270393bb98d8df88fccee0bf1e5128b91ae4eb4375896d", size = 4678835, upload-time = "2026-06-12T20:02:48.743Z" },
+    { url = "https://files.pythonhosted.org/packages/68/28/8a3ad4653662c93fc44dc4e5d8fd374c25c42e07b34bbfbadf49cf57a5a8/cryptography-49.0.0-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:7abcee80084cda3f7691f3eb1ce480d8df49cec637b429aa35986c1de71738aa", size = 4697239, upload-time = "2026-06-12T20:02:56.03Z" },
+    { url = "https://files.pythonhosted.org/packages/a8/b2/2193fc74f81aee4f9b62733133b73b5176718932ed8f2e4b03fa040480a6/cryptography-49.0.0-cp314-cp314t-manylinux_2_28_aarch64.whl", hash = "sha256:4ae387c9cb68ea569ca17e490d66d8142b81c3cc814bf179974b7d146e490bbb", size = 4685593, upload-time = "2026-06-12T20:02:50.666Z" },
+    { url = "https://files.pythonhosted.org/packages/47/f1/1d3eaa243bfc5de4a187b22aa8c048b3e4980bfbe830ac46e6bac2e66947/cryptography-49.0.0-cp314-cp314t-manylinux_2_28_ppc64le.whl", hash = "sha256:f37d847238971164fdbc68ade6f6574aecc9c0af714190e2083429ff68f4ce9d", size = 5289961, upload-time = "2026-06-12T20:01:46.468Z" },
+    { url = "https://files.pythonhosted.org/packages/58/39/2d51306721330c486495853eda1c567880ff036de15a14c4b74f399934af/cryptography-49.0.0-cp314-cp314t-manylinux_2_28_x86_64.whl", hash = "sha256:c2bc30226390d60ea19d9f82b19db005fe0452154a23c1c410c12ea801e43561", size = 4731145, upload-time = "2026-06-12T20:02:16.832Z" },
+    { url = "https://files.pythonhosted.org/packages/17/50/983e838c7fd0d87fd8c969bcdd328edaf5f756e38df5281637424c155873/cryptography-49.0.0-cp314-cp314t-manylinux_2_31_armv7l.whl", hash = "sha256:07cab27cc7b7e0fd28e5e26bb9eeedde5c135c868b46de4a27845abe94af6122", size = 4321719, upload-time = "2026-06-12T20:02:52.611Z" },
+    { url = "https://files.pythonhosted.org/packages/a7/f5/8f571d7e27c55bce9f76f026143bcb1e040a4233149ecca0bea5fa5dd5f7/cryptography-49.0.0-cp314-cp314t-manylinux_2_34_aarch64.whl", hash = "sha256:b20133d204d2bb56ba047642199603876c872026ca53e79c35b83772ab2cc505", size = 4685209, upload-time = "2026-06-12T20:02:07.282Z" },
+    { url = "https://files.pythonhosted.org/packages/e7/84/0e27016a6fc5a0886f797018b26aa42f40c09a82332bff77822a451deaaa/cryptography-49.0.0-cp314-cp314t-manylinux_2_34_ppc64le.whl", hash = "sha256:b970c6da94d5bb18629db453d14f2a1300f6bf59b61e9b82377931ef95504866", size = 5246285, upload-time = "2026-06-12T20:01:32.439Z" },
+    { url = "https://files.pythonhosted.org/packages/11/2d/5e1fb307cb5931881516b464c98774b3f2c36b5d4bb9a2830253cf553cad/cryptography-49.0.0-cp314-cp314t-manylinux_2_34_x86_64.whl", hash = "sha256:d8ecde755e2e91bf773fc94e8c9d730cd7f2007004cb492263a794ec3899a1c8", size = 4730441, upload-time = "2026-06-12T20:02:01.469Z" },
+    { url = "https://files.pythonhosted.org/packages/e4/c0/bff5a02ee731d207d6a1ed51732549d8c53d2bc8da1d10ec6f2844201d68/cryptography-49.0.0-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:e3fb64c420688e5319ae25113a354015abbd8dffbfbc41781a1ea66fc7622ac3", size = 4815869, upload-time = "2026-06-12T20:01:36.574Z" },
+    { url = "https://files.pythonhosted.org/packages/b9/26/814681d14248d95d73d5c3eea0c39a94eb8302df966f670a2c60de90974b/cryptography-49.0.0-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:32703d93296f5c1f4b53349ad3a250c2cae0fdecd3a3dd5d47e616d8d616af27", size = 4960948, upload-time = "2026-06-12T20:02:18.688Z" },
+    { url = "https://files.pythonhosted.org/packages/4c/fe/93ecac273d3738939d023612ad12cca9a3740a5345d69fda04134c43fd96/cryptography-49.0.0-cp314-cp314t-win_amd64.whl", hash = "sha256:33cd0565932807baddb67b96dbee92f2c374b5c89dee09fd74079aeb8c8dba61", size = 3799153, upload-time = "2026-06-12T20:01:39.059Z" },
+    { url = "https://files.pythonhosted.org/packages/19/2a/5bb823f5bedcf80718cea7fbc95ec5515cca3769633c4b01a32be7f30e7c/cryptography-49.0.0-cp39-abi3-macosx_11_0_arm64.whl", hash = "sha256:ec5e529fb80935c94fe7b729f9972b50e351a0e6b50aa294fd5cabb109fcc29a", size = 4025947, upload-time = "2026-06-12T20:01:25.745Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/df/40577043ca124e17012f408ddddaeb213b856336ac82ddb3bc915f39e29f/cryptography-49.0.0-cp39-abi3-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:f78ff2c9ed8dc2d036b0f4d640e22522213d047c1b14e61205a7e55c80a494d4", size = 4692429, upload-time = "2026-06-12T20:01:53.628Z" },
+    { url = "https://files.pythonhosted.org/packages/2c/99/2d13299eb3dd27b02dcfaafcc91d6b5cb3329f7cbd6d8f51921acd566c1a/cryptography-49.0.0-cp39-abi3-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:35b151772baff2c74cba7fa290ceaff4c3b11c0c881eb93eb5dbc05a7cfbba18", size = 4700968, upload-time = "2026-06-12T20:02:45.383Z" },
+    { url = "https://files.pythonhosted.org/packages/a5/4d/9c0cd02f95e2602dd5e563da149ee0830abef3537be8b34dc56281ebe27a/cryptography-49.0.0-cp39-abi3-manylinux_2_28_aarch64.whl", hash = "sha256:0f21641cf4b30fca7aee061ced0ec7ad7b073518088b7c9969a297c0ae796c69", size = 4697758, upload-time = "2026-06-12T20:01:41.13Z" },
+    { url = "https://files.pythonhosted.org/packages/24/01/186c825898477d77e2324d5360fefe622ff1d8d1963ec0554e2cada8ec77/cryptography-49.0.0-cp39-abi3-manylinux_2_28_ppc64le.whl", hash = "sha256:9e82dcc8e56052715fb18b2429e3bca4823b1629136a2084fc45a9a5cecb9b64", size = 5298863, upload-time = "2026-06-12T20:02:24.579Z" },
+    { url = "https://files.pythonhosted.org/packages/b8/7b/62cbbab75d0659865bf0273790031544a0b16c8072d258f9428dcd8190dc/cryptography-49.0.0-cp39-abi3-manylinux_2_28_x86_64.whl", hash = "sha256:6f2debedf9ca60cf1d5bd466475638af5130f89965605cd818484d19987d3a21", size = 4735983, upload-time = "2026-06-12T20:01:50.14Z" },
+    { url = "https://files.pythonhosted.org/packages/6c/72/3e798c064bc39e471008075d0f9bc9daf77a80879c092e4a8e170c585ed4/cryptography-49.0.0-cp39-abi3-manylinux_2_31_armv7l.whl", hash = "sha256:8c25ceb16df5b9435f3f6a9829204985b0e0cbee3b48aacd432c7d2c850b44d9", size = 4334173, upload-time = "2026-06-12T20:01:44.743Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/ee/6fca21d1ac73e06f8bef71940abfd4d2f6472b4bca284d770f32bd4086f6/cryptography-49.0.0-cp39-abi3-manylinux_2_34_aarch64.whl", hash = "sha256:28d8b15e6275f12c8a207dc309dfa957903c927d08d0cc937ee3f63f200693cc", size = 4697298, upload-time = "2026-06-12T20:02:20.918Z" },
+    { url = "https://files.pythonhosted.org/packages/67/d0/a5fcd3515f0bae49a7b6d0413cc1bdccdcc1fc0047037a0d480642cdc5d6/cryptography-49.0.0-cp39-abi3-manylinux_2_34_ppc64le.whl", hash = "sha256:6fc361c34fb6aac015ce19435876635e5c6d21db31998b0920f675f131e043b8", size = 5254338, upload-time = "2026-06-12T20:02:22.737Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/84/84fe36f19caf857d61cb7fc9c63035a47ffabd84ea12d1d393148efa3615/cryptography-49.0.0-cp39-abi3-manylinux_2_34_x86_64.whl", hash = "sha256:2400ef9c9e2299a25614eb1dea3db54a69b1349efd043bfac9c67630d136df36", size = 4735650, upload-time = "2026-06-12T20:02:41.389Z" },
+    { url = "https://files.pythonhosted.org/packages/6c/a0/db537264e234f7273a73ec020873d6d6b39dfd8a53db78b550ca8320440e/cryptography-49.0.0-cp39-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:67e1d20ad9ef3a563c59ef22e7a8a0b8210bd26604369ea4a30a7c66aefe504e", size = 4834820, upload-time = "2026-06-12T20:01:51.847Z" },
+    { url = "https://files.pythonhosted.org/packages/93/77/8df9eb486495979bccecd1062e2eaf435250e84437040295b57d09048b0b/cryptography-49.0.0-cp39-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:42b0684e0e40cf26122427802486f6d93aea593612603a94fbf260c7eb1e9c1b", size = 4967968, upload-time = "2026-06-12T20:02:12.524Z" },
+    { url = "https://files.pythonhosted.org/packages/c2/e6/f60198ea8d9dfa15fff9ed4ca02ce362f6eadd9ba757dcc50634c4257b63/cryptography-49.0.0-cp39-abi3-win_amd64.whl", hash = "sha256:026ac7423e6fa66872d3bf889be5974507da3944f866f704fa200eadacd00001", size = 3785547, upload-time = "2026-06-12T20:02:26.847Z" },
+]
+
+[[package]]
+name = "cssselect"
+version = "1.4.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/ec/2e/cdfd8b01c37cbf4f9482eefd455853a3cf9c995029a46acd31dfaa9c1dd6/cssselect-1.4.0.tar.gz", hash = "sha256:fdaf0a1425e17dfe8c5cf66191d211b357cf7872ae8afc4c6762ddd8ac47fc92", size = 40589, upload-time = "2026-01-29T07:00:26.701Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/20/0c/7bb51e3acfafd16c48875bf3db03607674df16f5b6ef8d056586af7e2b8b/cssselect-1.4.0-py3-none-any.whl", hash = "sha256:c0ec5c0191c8ee39fcc8afc1540331d8b55b0183478c50e9c8a79d44dbceb1d8", size = 18540, upload-time = "2026-01-29T07:00:24.994Z" },
+]
+
+[[package]]
+name = "curl-cffi"
+version = "0.15.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "certifi" },
+    { name = "cffi" },
+    { name = "rich" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/48/5b/89fcfebd3e5e85134147ac99e9f2b2271165fd4d71984fc65da5f17819b7/curl_cffi-0.15.0.tar.gz", hash = "sha256:ea0c67652bf6893d34ee0f82c944f37e488f6147e9421bef1771cc6545b02ded", size = 196437, upload-time = "2026-04-03T11:12:31.525Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/5e/42/54ddd442c795f30ce5dd4e49f87ce77505958d3777cd96a91567a3975d2a/curl_cffi-0.15.0-cp310-abi3-macosx_10_9_x86_64.whl", hash = "sha256:bda66404010e9ed743b1b83c20c86f24fe21a9a6873e17479d6e67e29d8ded28", size = 2795267, upload-time = "2026-04-03T11:11:46.48Z" },
+    { url = "https://files.pythonhosted.org/packages/83/2d/3915e238579b3c5a92cead5c79130c3b8d20caaba7616cc4d894650e1d6b/curl_cffi-0.15.0-cp310-abi3-macosx_11_0_arm64.whl", hash = "sha256:a25620d9bf989c9c029a7d1642999c4c265abb0bad811deb2f77b0b5b2b12e5b", size = 2573544, upload-time = "2026-04-03T11:11:47.951Z" },
+    { url = "https://files.pythonhosted.org/packages/2a/b3/9d2f1057749a1b07ba1989db3c1503ce8bed998310bae9aea2c43aa64f20/curl_cffi-0.15.0-cp310-abi3-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:582e570aa2586b96ed47cf4a17586b9a3c462cbe43f780487c3dc245c6ef1527", size = 10515369, upload-time = "2026-04-03T11:11:50.126Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/1d/6d10dded5ce3fd8157e558ebd97d09e551b77a62cdc1c31e93d0a633cee5/curl_cffi-0.15.0-cp310-abi3-manylinux2014_i686.manylinux_2_17_i686.whl", hash = "sha256:838e48212447d9c81364b04707a5c861daf08f8320f9ecb3406a8919d1d5c3b3", size = 10160045, upload-time = "2026-04-03T11:11:52.664Z" },
+    { url = "https://files.pythonhosted.org/packages/5c/12/c70b835487ace3b9ba1502631912e3440082b8ae3a162f60b59cb0b6444d/curl_cffi-0.15.0-cp310-abi3-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:2b6c847d86283b07ae69bb72c82eb8a59242277142aa35b89850f89e792a02fc", size = 11090433, upload-time = "2026-04-03T11:11:55.049Z" },
+    { url = "https://files.pythonhosted.org/packages/ea/0d/78edcc4f71934225db99df68197a107386d59080742fc7bf6bb4d007924f/curl_cffi-0.15.0-cp310-abi3-manylinux_2_28_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:9e5e69eee735f659287e2c84444319d68a1fa68dd37abf228943a4074864283a", size = 10479178, upload-time = "2026-04-03T11:11:57.685Z" },
+    { url = "https://files.pythonhosted.org/packages/5b/84/1e101c1acb1ea2f0b4992f5c3024f596d8e21db0d53540b9d583f673c4e7/curl_cffi-0.15.0-cp310-abi3-manylinux_2_34_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:aa1323950224db24f4c510d010b3affa02196ca853fb424191fa917a513d3f4b", size = 10317051, upload-time = "2026-04-03T11:12:00.295Z" },
+    { url = "https://files.pythonhosted.org/packages/28/42/8ef236b22a6c23d096c85a1dc507efe37bfdfc7a2f8a4b34efb590197369/curl_cffi-0.15.0-cp310-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:41f80170ba844009273b2660da1964ec31e99e5719d16b3422ada87177e32e13", size = 11299660, upload-time = "2026-04-03T11:12:02.791Z" },
+    { url = "https://files.pythonhosted.org/packages/1d/01/56aeb055d962da87a1be0d74c6c644e251c7e88129b5471dc44ac724e678/curl_cffi-0.15.0-cp310-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:1977e1e12cfb5c11352cbb74acef1bed24eb7d226dab61ca57c168c21acd4d61", size = 11945049, upload-time = "2026-04-03T11:12:05.912Z" },
+    { url = "https://files.pythonhosted.org/packages/d8/8c/2abf99a38d6340d66cf0557e0c750ef3f8883dfc5d450087e01c85861343/curl_cffi-0.15.0-cp310-abi3-win_amd64.whl", hash = "sha256:5a0c1896a0d5a5ac1eb89cd24b008d2b718dd1df6fd2f75451b59ca66e49e572", size = 1661649, upload-time = "2026-04-03T11:12:07.948Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/39/dfd54f2240d3a9b96d77bacc62b97813b35e2aa8ecf5cd5013c683f1ba96/curl_cffi-0.15.0-cp310-abi3-win_arm64.whl", hash = "sha256:a6d57f8389273a3a1f94370473c74897467bcc36af0a17336989780c507fa43d", size = 1410741, upload-time = "2026-04-03T11:12:10.073Z" },
+    { url = "https://files.pythonhosted.org/packages/19/6a/c24df8a4fc22fa84070dcd94abeba43c15e08cc09e35869565c0bad196fd/curl_cffi-0.15.0-cp313-abi3-android_24_arm64_v8a.whl", hash = "sha256:4682dc38d4336e0eb0b185374db90a760efde63cbea994b4e63f3521d44c4c92", size = 7190427, upload-time = "2026-04-03T11:12:12.142Z" },
+    { url = "https://files.pythonhosted.org/packages/11/56/132225cb3491d07cc6adcce5fe395e059bde87c68cff1ef87a31c88c7819/curl_cffi-0.15.0-cp314-cp314t-macosx_10_15_x86_64.whl", hash = "sha256:967ad7355bd8e9586f8c2d02eaa99953747549e7ea4a9b25cd53353e6b67fe6d", size = 2795723, upload-time = "2026-04-03T11:12:13.668Z" },
+    { url = "https://files.pythonhosted.org/packages/07/8f/f4f83cd303bef7e8f1749512e5dd157e7e5d08b0a36c8211f9640a2757bf/curl_cffi-0.15.0-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:7e63539d0d839d0a8c5eacf86229bc68c57803547f35e0db7ee0986328b478c3", size = 2573739, upload-time = "2026-04-03T11:12:15.08Z" },
+    { url = "https://files.pythonhosted.org/packages/e8/5c/643d65c7fc9acd742876aa55c2d7823c438cb7665810acd2e66c9976c4d9/curl_cffi-0.15.0-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:08c799b89740b9bc49c09fbc3d5907f13ac1f845ca52620507ef9466d4639dd5", size = 10521046, upload-time = "2026-04-03T11:12:17.034Z" },
+    { url = "https://files.pythonhosted.org/packages/7f/0b/9b8037113c93f4c5323096163471fa7c35c7676c3f608eeaf1287cd99d58/curl_cffi-0.15.0-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:7b7a92767a888ee90147e18964b396d8435ff42737030d6fb00824ffd6094805", size = 11096115, upload-time = "2026-04-03T11:12:19.694Z" },
+    { url = "https://files.pythonhosted.org/packages/5f/96/fff2fcbd924ef4042e0d67379f751a8a4e3186a91e75e35a4cf218b306ee/curl_cffi-0.15.0-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:829cc357061ecb99cc2d406301f609a039e05665322f5c025ec67c38b0dc49ce", size = 11305346, upload-time = "2026-04-03T11:12:22.151Z" },
+    { url = "https://files.pythonhosted.org/packages/53/1b/304b253a45ab28691c8c5e8cca1e6cbb9cf8e46dfceae4648dd536f75e73/curl_cffi-0.15.0-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:408d6f14e346841cd889c2e0962832bb235ba3b6749ebf609f347f747da5e60f", size = 11949834, upload-time = "2026-04-03T11:12:24.986Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/ff/4723d92f08259c707a974aba27a08d0a822b9555e35ca581bf18d055a364/curl_cffi-0.15.0-cp314-cp314t-win_amd64.whl", hash = "sha256:b624c7ce087bfda967a013ed0a64702a525444e5b6e97d23534d567ccc6525aa", size = 1702771, upload-time = "2026-04-03T11:12:28.201Z" },
+    { url = "https://files.pythonhosted.org/packages/59/8c/36bbe06d66fa2b765e4a07199f643a59a9cd1a754207a96335402a9520f4/curl_cffi-0.15.0-cp314-cp314t-win_arm64.whl", hash = "sha256:0b6c0543b993996670e9e4b78e305a2d60809d5681903ffb5568e21a387434d3", size = 1466312, upload-time = "2026-04-03T11:12:30.054Z" },
+]
+
+[[package]]
+name = "decorator"
+version = "5.3.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/60/8b/32f9823da46cde7df2087faa08cd98d01b908f8dcab982cdba9c84e85355/decorator-5.3.1.tar.gz", hash = "sha256:4cbcdd55a6efadb9dbea26b858f4fb3264567b52d69ca0d25b721b553f60ea82", size = 58084, upload-time = "2026-05-18T06:03:28.057Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/05/7f/798705f5296a58ca505d600456748d1be48078eac8a7050d8a98bc9edb89/decorator-5.3.1-py3-none-any.whl", hash = "sha256:f47fe6fdbd2edd623ecfe36875d37aba411624e2670dd395dddae1358689bb3c", size = 10365, upload-time = "2026-05-18T06:03:26.517Z" },
+]
+
+[[package]]
+name = "distro"
+version = "1.9.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/fc/f8/98eea607f65de6527f8a2e8885fc8015d3e6f5775df186e443e0964a11c3/distro-1.9.0.tar.gz", hash = "sha256:2fa77c6fd8940f116ee1d6b94a2f90b13b5ea8d019b98bc8bafdcabcdd9bdbed", size = 60722, upload-time = "2023-12-24T09:54:32.31Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/12/b3/231ffd4ab1fc9d679809f356cebee130ac7daa00d6d6f3206dd4fd137e9e/distro-1.9.0-py3-none-any.whl", hash = "sha256:7bffd925d65168f85027d8da9af6bddab658135b840670a223589bc0c8ef02b2", size = 20277, upload-time = "2023-12-24T09:54:30.421Z" },
+]
+
+[[package]]
+name = "docstring-parser"
+version = "0.18.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/e0/4d/f332313098c1de1b2d2ff91cf2674415cc7cddab2ca1b01ae29774bd5fdf/docstring_parser-0.18.0.tar.gz", hash = "sha256:292510982205c12b1248696f44959db3cdd1740237a968ea1e2e7a900eeb2015", size = 29341, upload-time = "2026-04-14T04:09:19.867Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a7/5f/ed01f9a3cdffbd5a008556fc7b2a08ddb1cc6ace7effa7340604b1d16699/docstring_parser-0.18.0-py3-none-any.whl", hash = "sha256:b3fcbed555c47d8479be0796ef7e19c2670d428d72e96da63f3a40122860374b", size = 22484, upload-time = "2026-04-14T04:09:18.638Z" },
+]
+
+[[package]]
+name = "et-xmlfile"
+version = "2.0.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/d3/38/af70d7ab1ae9d4da450eeec1fa3918940a5fafb9055e934af8d6eb0c2313/et_xmlfile-2.0.0.tar.gz", hash = "sha256:dab3f4764309081ce75662649be815c4c9081e88f0837825f90fd28317d4da54", size = 17234, upload-time = "2024-10-25T17:25:40.039Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/c1/8b/5fe2cc11fee489817272089c4203e679c63b570a5aaeb18d852ae3cbba6a/et_xmlfile-2.0.0-py3-none-any.whl", hash = "sha256:7a91720bc756843502c3b7504c77b8fe44217c85c537d85037f0f536151b2caa", size = 18059, upload-time = "2024-10-25T17:25:39.051Z" },
+]
+
+[[package]]
+name = "executing"
+version = "2.2.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/cc/28/c14e053b6762b1044f34a13aab6859bbf40456d37d23aa286ac24cfd9a5d/executing-2.2.1.tar.gz", hash = "sha256:3632cc370565f6648cc328b32435bd120a1e4ebb20c77e3fdde9a13cd1e533c4", size = 1129488, upload-time = "2025-09-01T09:48:10.866Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/c1/ea/53f2148663b321f21b5a606bd5f191517cf40b7072c0497d3c92c4a13b1e/executing-2.2.1-py2.py3-none-any.whl", hash = "sha256:760643d3452b4d777d295bb167ccc74c64a81df23fb5e08eff250c425a4b2017", size = 28317, upload-time = "2025-09-01T09:48:08.5Z" },
+]
+
+[[package]]
+name = "greenlet"
+version = "3.5.4"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/a3/74/b13368064b09053253555d3f2839cc2684d22d5aed0d2ccffbf7a6736558/greenlet-3.5.4.tar.gz", hash = "sha256:0232ae1de90a8e07867bb127d7a6ba2301e859145489f25cda8a6096dabe1d20", size = 206538, upload-time = "2026-07-22T12:47:14.468Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/f3/04/81bd731d6d1e3a469d9a4c36f5eb069bcf0cbb2d5d342c9fec22245b91fc/greenlet-3.5.4-cp312-cp312-macosx_11_0_universal2.whl", hash = "sha256:3d66250e8b09f182ede05490998c818b5961f7a3640332d44c4927caec7bbfe4", size = 295909, upload-time = "2026-07-22T11:38:09.261Z" },
+    { url = "https://files.pythonhosted.org/packages/cc/dd/f5f22903a6ae70f5ea328ed0beaec92ad903f0e3b7d2845133b354abc4b8/greenlet-3.5.4-cp312-cp312-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:c90e930c9c192e5b3ee9fb8bcd920ea3926155e2e3ded39fc697323addecee17", size = 612011, upload-time = "2026-07-22T12:26:40.69Z" },
+    { url = "https://files.pythonhosted.org/packages/8e/10/92a4a88d12b915d74ea5b6d288e4afefda4771647caa34442c156f7a454f/greenlet-3.5.4-cp312-cp312-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:791fdfeeb9c6e0c7b10fa151bf110d2a6974866f13dcb5b1c7efae698245893a", size = 624299, upload-time = "2026-07-22T12:29:02.089Z" },
+    { url = "https://files.pythonhosted.org/packages/6c/f9/03e26be3487c5238e81f2b84714959a86ea8515a869828cf41f4fc54b34e/greenlet-3.5.4-cp312-cp312-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:b7c895310363f310361e0fe2072af85269d2a2a285cd04c0c59e79a5e3670dcf", size = 629603, upload-time = "2026-07-22T12:43:43.456Z" },
+    { url = "https://files.pythonhosted.org/packages/50/6d/0b14bb9db2989f32cd9fe7f76afedea01ee8bee3f87c07e69f24adfe7e63/greenlet-3.5.4-cp312-cp312-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:f88193799d43dbf8c8a806d6405c9c52fe2af40bf75072a606357b33cc336c7f", size = 621541, upload-time = "2026-07-22T11:51:09.464Z" },
+    { url = "https://files.pythonhosted.org/packages/57/6b/7c55ca72ef80d57c16c4a55210f82582622462dc4485799a30f4ec6f3372/greenlet-3.5.4-cp312-cp312-manylinux_2_39_riscv64.whl", hash = "sha256:13b980043cb1b3134e81ea469da1250ddcc6bfe6d245bbaa59168d9cdc8f228f", size = 432554, upload-time = "2026-07-22T12:39:51.379Z" },
+    { url = "https://files.pythonhosted.org/packages/48/3d/25e9a2d9eb6b2e8b7ca4e80a3a26cb887cce6c8e0a87c921164f11bc5574/greenlet-3.5.4-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:b7a5f095767c4493afcd06067f2bb3b8716e3f3f9e92b99c88e7e99f885b3d4d", size = 1581444, upload-time = "2026-07-22T12:25:03.818Z" },
+    { url = "https://files.pythonhosted.org/packages/b9/96/4c9bf2e2c408dcc0556edce69efa9f802e82223573c53240136a086821f1/greenlet-3.5.4-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:42afdc1ab5f66da8c586c32af9224a74a706b4f0ea0dc3a4188a0860a09c65c9", size = 1645842, upload-time = "2026-07-22T11:51:12.295Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/41/303ecb26a3a56122c0f4d4073ee078881847bd6b6f463ae0ec57ec20223b/greenlet-3.5.4-cp312-cp312-win_amd64.whl", hash = "sha256:60149df8f462d1b230038e6590c23c3b4768bb5d6c022b3b6e82532b34b0b8a3", size = 247169, upload-time = "2026-07-22T11:38:19.893Z" },
+    { url = "https://files.pythonhosted.org/packages/a4/e3/ef56864b4c35fcb3eb3b41b869f6cc46f4cd3f5e2c68e74acde8ac433951/greenlet-3.5.4-cp312-cp312-win_arm64.whl", hash = "sha256:77d6ce04fed0d9aeed42e0f37923cc43eba9b027bdd9c34546bb4ccd143d0fe0", size = 245565, upload-time = "2026-07-22T11:38:27.061Z" },
+    { url = "https://files.pythonhosted.org/packages/c0/9a/e51225dcd58713f16ccbdcc501a8da21098ea14515b7870f1f94459e5ff5/greenlet-3.5.4-cp313-cp313-macosx_11_0_universal2.whl", hash = "sha256:24e61b88cb7e1b1d794b32a10cc346ac779681d6d74ff137a3e0a444d2bf1f02", size = 294831, upload-time = "2026-07-22T11:38:53.389Z" },
+    { url = "https://files.pythonhosted.org/packages/9f/ea/de50a50fadf979713ab18b46f22ad5ff5f2dcfc637a3ebdecf669801e1a5/greenlet-3.5.4-cp313-cp313-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:870d730fec833f5a06906a32596cc099b9161594642a92a520b7a88911c95356", size = 614619, upload-time = "2026-07-22T12:26:42.282Z" },
+    { url = "https://files.pythonhosted.org/packages/db/c7/2aae27fea41205b8650294c301f042a2a4bb6155eea48c995b890a92f2c1/greenlet-3.5.4-cp313-cp313-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:ec5ff0d1878df6af3bf9b638a5a92a7d5693291de77c91bff10fa48519c604ef", size = 627021, upload-time = "2026-07-22T12:29:03.445Z" },
+    { url = "https://files.pythonhosted.org/packages/1b/80/fb4d4788bbc8e54761f1fc88533af9523a6e86299fa113d6e8a8503ed9fc/greenlet-3.5.4-cp313-cp313-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:07bd44616608d873d06735b63ef1a88191d6ca57c8d291d6559c71bc14c0893c", size = 632845, upload-time = "2026-07-22T12:43:45.19Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/56/79fd826f9ccaae0b84e1b4ef68dabba5e105bb044ffcd448a0b782fcba9a/greenlet-3.5.4-cp313-cp313-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:d84d993f6e575c950d91a23c1345d18fe1a4310d447bf630849d7809196b52f0", size = 624002, upload-time = "2026-07-22T11:51:11.391Z" },
+    { url = "https://files.pythonhosted.org/packages/42/e3/6086fa578ebb72772722cdc4bcd628459814b42e0c2db1e3cbd6552b3271/greenlet-3.5.4-cp313-cp313-manylinux_2_39_riscv64.whl", hash = "sha256:3529a8a933582ad19e224792cac7372489526576b75b4c124e8e4f29948f4861", size = 435053, upload-time = "2026-07-22T12:39:52.715Z" },
+    { url = "https://files.pythonhosted.org/packages/0a/1a/27319f97e731298513dcba1a2e91b63e9d8811d9de22130f960b129b1bf1/greenlet-3.5.4-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:58023945f421093de5e6fa108c0985a8659d43f49e0216da25099369a121bcbd", size = 1581533, upload-time = "2026-07-22T12:25:05.322Z" },
+    { url = "https://files.pythonhosted.org/packages/b1/6d/24240bf562e9786dd2799ee0a4a4dadb4ded22510f41b20245099159ac8c/greenlet-3.5.4-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:bae2728e1897aa8df8cb1af38cd48b3a743aefe29372de7b8b7a9f532501e69f", size = 1645781, upload-time = "2026-07-22T11:51:14.805Z" },
+    { url = "https://files.pythonhosted.org/packages/c1/5a/442ab1a9ef7ca6bf7210e5397a95972206a91a31033a03c8900866a10039/greenlet-3.5.4-cp313-cp313-win_amd64.whl", hash = "sha256:ca5726c0b08ca35ae873557266a78b2c3f3b2b7d7401aa5ff886c2045dd0111c", size = 247133, upload-time = "2026-07-22T11:39:20.661Z" },
+    { url = "https://files.pythonhosted.org/packages/3e/e6/9160210222386b1a378ff94db846b9508ca24a121cf684991561fdb69280/greenlet-3.5.4-cp313-cp313-win_arm64.whl", hash = "sha256:7c1303791d603080cac6fc3b34df51c3b75b723739c282c8029e48a0d241672f", size = 245500, upload-time = "2026-07-22T11:40:22.185Z" },
+    { url = "https://files.pythonhosted.org/packages/a5/a7/6ab1d4f9cd548d15ab90da29947f2076100130bb179b0bde59f795a459e3/greenlet-3.5.4-cp314-cp314-macosx_11_0_universal2.whl", hash = "sha256:7e8afa5eac028f8140ceafe5ceec66e6aa127ddcb21452d2a564dcd2900b5f22", size = 295410, upload-time = "2026-07-22T11:40:35.747Z" },
+    { url = "https://files.pythonhosted.org/packages/cd/7a/422f63b4715cbc0b24385305407adf38b48f6bb68b3e6b04090e994d0f5a/greenlet-3.5.4-cp314-cp314-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:73b37afe369021423ea53dd3123e04bffa7e93ac64429b9f50835b2e4fcae7cf", size = 661286, upload-time = "2026-07-22T12:26:43.8Z" },
+    { url = "https://files.pythonhosted.org/packages/d0/31/5a1cac663bf5582190c5a714ef81364f03cde232227f39748f8ae4c11da5/greenlet-3.5.4-cp314-cp314-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:3ef964f56dfcb6f9bbef2a190d9126795eac408716aeae47b5e7c73c32aafca9", size = 673517, upload-time = "2026-07-22T12:29:04.815Z" },
+    { url = "https://files.pythonhosted.org/packages/9c/bf/250c2921c7b585dde12f5239e313ca2dcbc464d161ecca36e4e6ef21762d/greenlet-3.5.4-cp314-cp314-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:cef589bc65fae02d10bca2ac341191c5b33acc2967892ebf4fcbd10eabb7a74c", size = 677968, upload-time = "2026-07-22T12:43:46.788Z" },
+    { url = "https://files.pythonhosted.org/packages/15/4a/2a82a1e3f8aaca020853ac8d12211280ca2b231aa08ea39f636f1060c319/greenlet-3.5.4-cp314-cp314-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:9c53ff01a5c53a40f2c16820ebc56d7c61a77f5fbe009dadd96292d5682f80f8", size = 670917, upload-time = "2026-07-22T11:51:13.589Z" },
+    { url = "https://files.pythonhosted.org/packages/18/40/10bfcf6513558d82f7b95dd728001c63bd388259fe27d3e30ae01f103430/greenlet-3.5.4-cp314-cp314-manylinux_2_39_riscv64.whl", hash = "sha256:dfc41ae893d9ceaf22c824f2153a88b30651b20e8758c2cd9ac143f23640563c", size = 480643, upload-time = "2026-07-22T12:39:54.149Z" },
+    { url = "https://files.pythonhosted.org/packages/68/b0/e379a152b17bfdfa95795af4049e37c0fd1b4d81f020d426db104ed07c77/greenlet-3.5.4-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:ecca4d80d55a01ad6b23b33262662956149fbb7b2c6be2910f1705921958cbf3", size = 1628478, upload-time = "2026-07-22T12:25:06.678Z" },
+    { url = "https://files.pythonhosted.org/packages/5e/43/bffdfa64f7317f954c5c1230b5dd5922676ce198689a68c1ac1ed4b1b1a5/greenlet-3.5.4-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:2ffbc533e0eaf8e80d8471411646ab88fe58f641d508c0b02b24494479f4d9ec", size = 1692021, upload-time = "2026-07-22T11:51:17.008Z" },
+    { url = "https://files.pythonhosted.org/packages/d0/11/f799f9637e2c6e9b0b716015e339040598b058cf7654dfc0d67468b177ed/greenlet-3.5.4-cp314-cp314-win_amd64.whl", hash = "sha256:305f69e6c4523d7f6979ed001cff4e5853c063e5da04880296603aa0227e544c", size = 248031, upload-time = "2026-07-22T11:40:11.007Z" },
+    { url = "https://files.pythonhosted.org/packages/05/75/625bcdd74d5e6b2dca1ecba3c3ac77bcf8a026c21a649a46cef23e421f97/greenlet-3.5.4-cp314-cp314-win_arm64.whl", hash = "sha256:f260930bbbbcf9caee661211235a5111c86dfe5832fdf6ae4570da1e0995320f", size = 246892, upload-time = "2026-07-22T11:40:27.357Z" },
+    { url = "https://files.pythonhosted.org/packages/ec/69/35c62ed49c320cb4d98e14698ccca5467d3bfe683984172be9cb564d9ce3/greenlet-3.5.4-cp314-cp314t-macosx_11_0_universal2.whl", hash = "sha256:41ddab54e4b238f4a6c323f39b4e59e176affd5a94d461a9fb7583dac74240a3", size = 305571, upload-time = "2026-07-22T11:40:31.659Z" },
+    { url = "https://files.pythonhosted.org/packages/5b/6c/64d60216b3640dcb0b62d913dd9e0d80030c09115bb2e4ba70c95d10ca45/greenlet-3.5.4-cp314-cp314t-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:b3dabe3e2809013052c68bdf0b7fa5f5f2859c43a80803131ad61af9cabd7867", size = 672568, upload-time = "2026-07-22T12:26:45.298Z" },
+    { url = "https://files.pythonhosted.org/packages/5c/de/ba3ab0a96292e53039530333b0d2ae18d9e508f3a325cd7bf15f8172944c/greenlet-3.5.4-cp314-cp314t-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:27d3f00718634d4520a3a150154ac5da36f257869d41321953375b90bfbbc72c", size = 680076, upload-time = "2026-07-22T12:29:06.125Z" },
+    { url = "https://files.pythonhosted.org/packages/ae/db/24a10af12bf8e639cec46c38b9ce1a282543ba42ff4fb0b31a970f1ab603/greenlet-3.5.4-cp314-cp314t-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:39169a11d87a6a263afda3e9a27d1df16d0f919d40a4837cc73986c9884c0dd8", size = 681690, upload-time = "2026-07-22T12:43:48.109Z" },
+    { url = "https://files.pythonhosted.org/packages/a5/be/aeada79083c6f1c15f45d77a332f9c441af263ee298e3eb17522cd337d22/greenlet-3.5.4-cp314-cp314t-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:cbd60b5763c6543c1827e48faaf14ea9bfbad245f52b1a4d76a2a2d8884c6c66", size = 676733, upload-time = "2026-07-22T11:51:16.027Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/60/44a2eca7b9fd71ae0fae7ff184da1cd3169d176652b97aa1cffcbb0ef961/greenlet-3.5.4-cp314-cp314t-manylinux_2_39_riscv64.whl", hash = "sha256:bd3d1145f603b2db19feb9078c2e6855eb7c67e15580c010ed815cee519b86fd", size = 510263, upload-time = "2026-07-22T12:39:55.678Z" },
+    { url = "https://files.pythonhosted.org/packages/e9/10/2392fc3a98948652ef5fd1e7275c04f861dd13f74b78a2b4309f4ee4d090/greenlet-3.5.4-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:f00f910f0e7b35416c63b23ad78b769aeccfc1775f712b43c4ee525624a2eef7", size = 1637327, upload-time = "2026-07-22T12:25:07.879Z" },
+    { url = "https://files.pythonhosted.org/packages/55/c6/e7237a3dfa1f205ed0d9ea1e46d70bd2811b32d516266399fb59d28ab90a/greenlet-3.5.4-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:91c26423753b92caf41ab3f98fd547d7374d4d9fc2d85be041886c1579d9255e", size = 1697493, upload-time = "2026-07-22T11:51:19.214Z" },
+    { url = "https://files.pythonhosted.org/packages/55/e3/4ba8154ba2a3d43729e499f72471b4b5c993f3826d3e24da81d5f06d6572/greenlet-3.5.4-cp314-cp314t-win_amd64.whl", hash = "sha256:ee032b91fd8ec29ec6c4cea2b8c561b178435134bd0752c7334b94e9c736c132", size = 251637, upload-time = "2026-07-22T11:40:37.44Z" },
+    { url = "https://files.pythonhosted.org/packages/90/03/e3f96dfc100261a29545ddc8270cafe58f9195b6651466910e820910de77/greenlet-3.5.4-cp315-cp315-macosx_11_0_universal2.whl", hash = "sha256:178111881dd7a6c946471fda85485ec796e1043c2b939f694b096e2ecf986809", size = 296076, upload-time = "2026-07-22T11:39:38.364Z" },
+    { url = "https://files.pythonhosted.org/packages/a4/3d/da52d208e5c977bce8667e784729e584e38b5785f4c1ec0f4c836e9a1c42/greenlet-3.5.4-cp315-cp315-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:d92df08dd65fede97fc37aad36c2e9dcda3b31c467f8e0c2c096456cb818e927", size = 666870, upload-time = "2026-07-22T12:26:46.691Z" },
+    { url = "https://files.pythonhosted.org/packages/2d/8a/7e6dee25cb8a8cf9b362c8e597cc269593378bd916f16c736c059a52e85a/greenlet-3.5.4-cp315-cp315-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:99e8f8c4ebc4fd80aa26c1280ae9ad43a0976e786349703a181cf0bae60413e5", size = 677678, upload-time = "2026-07-22T12:29:07.508Z" },
+    { url = "https://files.pythonhosted.org/packages/51/a7/dafc7415d430b0a43a16396eb49ecb3b62fd720877fb259cc4dcfaf5f31e/greenlet-3.5.4-cp315-cp315-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:1f17e362d78e37559e0506c5a7d066bdd45073c36a0127a543e8a0df27242ff3", size = 681428, upload-time = "2026-07-22T12:43:49.623Z" },
+    { url = "https://files.pythonhosted.org/packages/6c/21/5a38699fa45de749e3857d93b8f07e4c20489e77c2d35d915a2e1c456606/greenlet-3.5.4-cp315-cp315-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:394de08dad5ffcb1f50c2159d93e398d9d2da3ed437645eaa54771fa720db9f0", size = 676067, upload-time = "2026-07-22T11:51:18.163Z" },
+    { url = "https://files.pythonhosted.org/packages/2e/d9/6298f3432de301d4718766cf934bd73c418c73f81fbb77247319364b0d96/greenlet-3.5.4-cp315-cp315-manylinux_2_39_riscv64.whl", hash = "sha256:cd320d998cbaa032932830448e39abf3c6a12901295e386e8114db926e10cffb", size = 487446, upload-time = "2026-07-22T12:39:57.044Z" },
+    { url = "https://files.pythonhosted.org/packages/5e/ba/863116ab8ff1ca7a729e327800268939d182db47aa433db70e216e7d9194/greenlet-3.5.4-cp315-cp315-musllinux_1_2_aarch64.whl", hash = "sha256:c883d61f2282d72c767a14936641b3efcbde9d82f1080712aaea0b1d3126cb88", size = 1633489, upload-time = "2026-07-22T12:25:09.605Z" },
+    { url = "https://files.pythonhosted.org/packages/fa/06/7466ced82818d6132462d7f26b3f83c66ea15d2b193a6c0088d558ed7d95/greenlet-3.5.4-cp315-cp315-musllinux_1_2_x86_64.whl", hash = "sha256:2a924f15d17957e252a810acefcb5942f5ca712298e8b6fcaed9a307d357522c", size = 1696584, upload-time = "2026-07-22T11:51:21.304Z" },
+    { url = "https://files.pythonhosted.org/packages/f9/4d/55b638489260065de9ffce606c8b5d04507bef705de4b212a0c3d6a1a0df/greenlet-3.5.4-cp315-cp315-win_amd64.whl", hash = "sha256:ed17e5f3420360d5b459de8462efb52060399a5326a613d4cde31cef63ef95da", size = 248297, upload-time = "2026-07-22T11:42:04.055Z" },
+    { url = "https://files.pythonhosted.org/packages/bb/08/9dd4ae635da93d41dc268bc34bd62a9d711ed8b8825c5d22ac910c7d6e6d/greenlet-3.5.4-cp315-cp315-win_arm64.whl", hash = "sha256:f908898d6fa484ce4b6f447ce70ea99b52c503fee419e53cf74d60a16bc9e667", size = 247423, upload-time = "2026-07-22T11:44:00.764Z" },
+    { url = "https://files.pythonhosted.org/packages/19/66/7c87ed9cdbf1d49c2c6cd1c7b9dd4d16c33b24235ca03972293a1876b30c/greenlet-3.5.4-cp315-cp315t-macosx_11_0_universal2.whl", hash = "sha256:1833637f17d5e7472548a48575c394fe39f1b1890d676d162d86593610f44d8c", size = 306487, upload-time = "2026-07-22T11:41:25.118Z" },
+    { url = "https://files.pythonhosted.org/packages/5b/05/0a4201e7c0054866eefc05da234f236dd4c950d0fbf9ca0517141f01b269/greenlet-3.5.4-cp315-cp315t-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:12cda9122e03341f1cb6b8207a19d7a9d375e52f1b4e9243918375f40fd7b4b9", size = 676479, upload-time = "2026-07-22T12:26:48.129Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/c0/4b6b8c5a3aec70f0649cd89662d120fdd6421e2bdc8e3b15c3ab5ec568d8/greenlet-3.5.4-cp315-cp315t-manylinux_2_24_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:d83ae0e32d14957ab7170785a20f582635c8474deab1bfbb552b17e769a6ce25", size = 684321, upload-time = "2026-07-22T12:29:08.925Z" },
+    { url = "https://files.pythonhosted.org/packages/88/15/0b167aeea95285b0e654ddce651922f666c089363c2ec528ca8b9a9ba74f/greenlet-3.5.4-cp315-cp315t-manylinux_2_24_s390x.manylinux_2_28_s390x.whl", hash = "sha256:123aa379c962ed5fe90a880327e0c3066124ac64ec99e12a238be9fd8eb3db3d", size = 685995, upload-time = "2026-07-22T12:43:50.993Z" },
+    { url = "https://files.pythonhosted.org/packages/24/c9/b49c31c9a972eee91e260445770e922244a7efc542697f51a012ec046d0f/greenlet-3.5.4-cp315-cp315t-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:9f1467de1bb767f75db0aa34c195e3a496d8d1278c796e70c24ce205d3e99cde", size = 681293, upload-time = "2026-07-22T11:51:20.43Z" },
+    { url = "https://files.pythonhosted.org/packages/de/90/c023ec337f32ff505be7db759c80d98f0532bb94d0c6fa13645efe9bee2e/greenlet-3.5.4-cp315-cp315t-manylinux_2_39_riscv64.whl", hash = "sha256:adf2244d7f69409925a8f22ed22cc5f93cdfe5c9dc87ff3476be2c2aaae61a05", size = 516928, upload-time = "2026-07-22T12:39:58.359Z" },
+    { url = "https://files.pythonhosted.org/packages/95/6f/7f2d4653770500eee667866016d42d7a68e3d3462f80df6b8e3fcd48a0eb/greenlet-3.5.4-cp315-cp315t-musllinux_1_2_aarch64.whl", hash = "sha256:0fa53040b78b578120eecdc0265e3f1051487cc425d11a2b7c761daadf4feaa8", size = 1642474, upload-time = "2026-07-22T12:25:10.819Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/d8/8cba31036a4caae448087ba5d150660ab03b4a0f54d9150f6495a3be7262/greenlet-3.5.4-cp315-cp315t-musllinux_1_2_x86_64.whl", hash = "sha256:60e0bc961d367df506660e9ac0177a76bc6d81305300704b0977d1634f76efe2", size = 1701012, upload-time = "2026-07-22T11:51:23.17Z" },
+    { url = "https://files.pythonhosted.org/packages/e3/cd/3f77a4cce3bae631b08eb52f53a82a976669600337e21dfdba811cb50267/greenlet-3.5.4-cp315-cp315t-win_amd64.whl", hash = "sha256:f680e549edb3eaf21eea4e7fe101e15ec180c74b7879ab46adc080f22d4015d2", size = 251977, upload-time = "2026-07-22T11:41:38.125Z" },
+    { url = "https://files.pythonhosted.org/packages/93/e8/65e8707d00fe2a49bf12f609a9b2b39ba6dd23c2810eacad877c4fc94bfe/greenlet-3.5.4-cp315-cp315t-win_arm64.whl", hash = "sha256:08fc36de8442d5c3e95b044550dbea9bf144d31ec0cc58e36fb241cb6ef6a994", size = 250538, upload-time = "2026-07-22T11:40:17.985Z" },
+]
+
+[[package]]
+name = "gyxx-flow"
+version = "0.1.0"
+source = { editable = "." }
+dependencies = [
+    { name = "browserforge" },
+    { name = "curl-cffi" },
+    { name = "httpx" },
+    { name = "imageio-ffmpeg" },
+    { name = "langchain" },
+    { name = "langchain-anthropic" },
+    { name = "langchain-core" },
+    { name = "langchain-openai" },
+    { name = "langgraph" },
+    { name = "lark-oapi" },
+    { name = "lxml" },
+    { name = "msgspec" },
+    { name = "openpyxl" },
+    { name = "ormsgpack" },
+    { name = "pandas" },
+    { name = "patchright" },
+    { name = "pillow" },
+    { name = "playwright" },
+    { name = "psutil" },
+    { name = "psycopg", extra = ["binary", "pool"] },
+    { name = "psycopg2-binary" },
+    { name = "python-dotenv" },
+    { name = "requests" },
+    { name = "scrapling", extra = ["all"] },
+    { name = "selenium" },
+    { name = "xlrd" },
+]
+
+[package.optional-dependencies]
+test = [
+    { name = "pytest" },
+]
+
+[package.metadata]
+requires-dist = [
+    { name = "browserforge", specifier = ">=1.2.4" },
+    { name = "curl-cffi", specifier = ">=0.15.0" },
+    { name = "httpx", specifier = ">=0.27" },
+    { name = "imageio-ffmpeg", specifier = ">=0.6.0" },
+    { name = "langchain", specifier = ">=0.3.0" },
+    { name = "langchain-anthropic", specifier = ">=0.3.0" },
+    { name = "langchain-core", specifier = ">=0.3.0" },
+    { name = "langchain-openai", specifier = ">=0.2.0" },
+    { name = "langgraph", specifier = ">=0.2.0" },
+    { name = "lark-oapi", specifier = ">=1.0.0" },
+    { name = "lxml", specifier = ">=5.0" },
+    { name = "msgspec", specifier = ">=0.21.1" },
+    { name = "openpyxl", specifier = ">=3.1.5" },
+    { name = "ormsgpack", specifier = ">=1.2.0" },
+    { name = "pandas", specifier = ">=3.0.3" },
+    { name = "patchright", specifier = ">=1.60.1" },
+    { name = "pillow", specifier = ">=10" },
+    { name = "playwright", specifier = ">=1.60.0" },
+    { name = "psutil", specifier = ">=5.9" },
+    { name = "psycopg", extras = ["binary", "pool"], specifier = ">=3.1" },
+    { name = "psycopg2-binary", specifier = ">=2.9.12" },
+    { name = "pytest", marker = "extra == 'test'", specifier = ">=8.0" },
+    { name = "python-dotenv", specifier = ">=1.0.0" },
+    { name = "requests", specifier = ">=2.31.0" },
+    { name = "scrapling", extras = ["all"], specifier = ">=0.4.9" },
+    { name = "selenium", specifier = ">=4.20" },
+    { name = "xlrd", specifier = ">=2.0.1" },
+]
+provides-extras = ["test"]
+
+[[package]]
+name = "h11"
+version = "0.16.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/01/ee/02a2c011bdab74c6fb3c75474d40b3052059d95df7e73351460c8588d963/h11-0.16.0.tar.gz", hash = "sha256:4e35b956cf45792e4caa5885e69fba00bdbc6ffafbfa020300e549b208ee5ff1", size = 101250, upload-time = "2025-04-24T03:35:25.427Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/04/4b/29cac41a4d98d144bf5f6d33995617b185d14b22401f75ca86f384e87ff1/h11-0.16.0-py3-none-any.whl", hash = "sha256:63cf8bbe7522de3bf65932fda1d9c2772064ffb3dae62d55932da54b31cb6c86", size = 37515, upload-time = "2025-04-24T03:35:24.344Z" },
+]
+
+[[package]]
+name = "httpcore"
+version = "1.0.9"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "certifi" },
+    { name = "h11" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/06/94/82699a10bca87a5556c9c59b5963f2d039dbd239f25bc2a63907a05a14cb/httpcore-1.0.9.tar.gz", hash = "sha256:6e34463af53fd2ab5d807f399a9b45ea31c3dfa2276f15a2c3f00afff6e176e8", size = 85484, upload-time = "2025-04-24T22:06:22.219Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/7e/f5/f66802a942d491edb555dd61e3a9961140fd64c90bce1eafd741609d334d/httpcore-1.0.9-py3-none-any.whl", hash = "sha256:2d400746a40668fc9dec9810239072b40b4484b640a8c38fd654a024c7a1bf55", size = 78784, upload-time = "2025-04-24T22:06:20.566Z" },
+]
+
+[[package]]
+name = "httpx"
+version = "0.28.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anyio" },
+    { name = "certifi" },
+    { name = "httpcore" },
+    { name = "idna" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b1/df/48c586a5fe32a0f01324ee087459e112ebb7224f646c0b5023f5e79e9956/httpx-0.28.1.tar.gz", hash = "sha256:75e98c5f16b0f35b567856f597f06ff2270a374470a5c2392242528e3e3e42fc", size = 141406, upload-time = "2024-12-06T15:37:23.222Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/2a/39/e50c7c3a983047577ee07d2a9e53faf5a69493943ec3f6a384bdc792deb2/httpx-0.28.1-py3-none-any.whl", hash = "sha256:d909fcccc110f8c7faf814ca82a9a4d816bc5a6dbfea25d6591d6985b8ba59ad", size = 73517, upload-time = "2024-12-06T15:37:21.509Z" },
+]
+
+[[package]]
+name = "httpx-sse"
+version = "0.4.3"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/0f/4c/751061ffa58615a32c31b2d82e8482be8dd4a89154f003147acee90f2be9/httpx_sse-0.4.3.tar.gz", hash = "sha256:9b1ed0127459a66014aec3c56bebd93da3c1bc8bb6618c8082039a44889a755d", size = 15943, upload-time = "2025-10-10T21:48:22.271Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/d2/fd/6668e5aec43ab844de6fc74927e155a3b37bf40d7c3790e49fc0406b6578/httpx_sse-0.4.3-py3-none-any.whl", hash = "sha256:0ac1c9fe3c0afad2e0ebb25a934a59f4c7823b60792691f779fad2c5568830fc", size = 8960, upload-time = "2025-10-10T21:48:21.158Z" },
+]
+
+[[package]]
+name = "idna"
+version = "3.18"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/cd/63/9496c57188a2ee585e0f1db071d75089a11e98aa86eb99d9d7618fc1edce/idna-3.18.tar.gz", hash = "sha256:ffb385a7e039654cef1ab9ef32c6fafe283c0c0467bba1d9029738ce4a14a848", size = 196711, upload-time = "2026-06-02T14:34:07.794Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/1e/5e/d4e9f1a599fb8e573b7b87160658329fbf28d19eac2718f51fc3def3aa5a/idna-3.18-py3-none-any.whl", hash = "sha256:7f952cbe720b688055e3f87de14f5c3e5fdaa8bc3928985c4077ca689de849a2", size = 65455, upload-time = "2026-06-02T14:34:06.319Z" },
+]
+
+[[package]]
+name = "imageio-ffmpeg"
+version = "0.6.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/44/bd/c3343c721f2a1b0c9fc71c1aebf1966a3b7f08c2eea8ed5437a2865611d6/imageio_ffmpeg-0.6.0.tar.gz", hash = "sha256:e2556bed8e005564a9f925bb7afa4002d82770d6b08825078b7697ab88ba1755", size = 25210, upload-time = "2025-01-16T21:34:32.747Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/da/58/87ef68ac83f4c7690961bce288fd8e382bc5f1513860fc7f90a9c1c1c6bf/imageio_ffmpeg-0.6.0-py3-none-macosx_10_9_intel.macosx_10_9_x86_64.whl", hash = "sha256:9d2baaf867088508d4a3458e61eeb30e945c4ad8016025545f66c4b5aaef0a61", size = 24932969, upload-time = "2025-01-16T21:34:20.464Z" },
+    { url = "https://files.pythonhosted.org/packages/40/5c/f3d8a657d362cc93b81aab8feda487317da5b5d31c0e1fdfd5e986e55d17/imageio_ffmpeg-0.6.0-py3-none-macosx_11_0_arm64.whl", hash = "sha256:b1ae3173414b5fc5f538a726c4e48ea97edc0d2cdc11f103afee655c463fa742", size = 21113891, upload-time = "2025-01-16T21:34:00.277Z" },
+    { url = "https://files.pythonhosted.org/packages/33/e7/1925bfbc563c39c1d2e82501d8372734a5c725e53ac3b31b4c2d081e895b/imageio_ffmpeg-0.6.0-py3-none-manylinux2014_aarch64.whl", hash = "sha256:1d47bebd83d2c5fc770720d211855f208af8a596c82d17730aa51e815cdee6dc", size = 25632706, upload-time = "2025-01-16T21:33:53.475Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/2d/43c8522a2038e9d0e7dbdf3a61195ecc31ca576fb1527a528c877e87d973/imageio_ffmpeg-0.6.0-py3-none-manylinux2014_x86_64.whl", hash = "sha256:c7e46fcec401dd990405049d2e2f475e2b397779df2519b544b8aab515195282", size = 29498237, upload-time = "2025-01-16T21:34:13.726Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/13/59da54728351883c3c1d9fca1710ab8eee82c7beba585df8f25ca925f08f/imageio_ffmpeg-0.6.0-py3-none-win32.whl", hash = "sha256:196faa79366b4a82f95c0f4053191d2013f4714a715780f0ad2a68ff37483cc2", size = 19652251, upload-time = "2025-01-16T21:34:06.812Z" },
+    { url = "https://files.pythonhosted.org/packages/2c/c6/fa760e12a2483469e2bf5058c5faff664acf66cadb4df2ad6205b016a73d/imageio_ffmpeg-0.6.0-py3-none-win_amd64.whl", hash = "sha256:02fa47c83703c37df6bfe4896aab339013f62bf02c5ebf2dce6da56af04ffc0a", size = 31246824, upload-time = "2025-01-16T21:34:28.6Z" },
+]
+
+[[package]]
+name = "iniconfig"
+version = "2.3.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/72/34/14ca021ce8e5dfedc35312d08ba8bf51fdd999c576889fc2c24cb97f4f10/iniconfig-2.3.0.tar.gz", hash = "sha256:c76315c77db068650d49c5b56314774a7804df16fee4402c1f19d6d15d8c4730", size = 20503, upload-time = "2025-10-18T21:55:43.219Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/cb/b1/3846dd7f199d53cb17f49cba7e651e9ce294d8497c8c150530ed11865bb8/iniconfig-2.3.0-py3-none-any.whl", hash = "sha256:f631c04d2c48c52b84d0d0549c99ff3859c98df65b3101406327ecc7d53fbf12", size = 7484, upload-time = "2025-10-18T21:55:41.639Z" },
+]
+
+[[package]]
+name = "ipython"
+version = "9.15.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "colorama", marker = "sys_platform == 'win32'" },
+    { name = "decorator" },
+    { name = "ipython-pygments-lexers" },
+    { name = "jedi" },
+    { name = "matplotlib-inline" },
+    { name = "pexpect", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "prompt-toolkit" },
+    { name = "psutil", marker = "sys_platform != 'cygwin' and sys_platform != 'emscripten'" },
+    { name = "pygments" },
+    { name = "stack-data" },
+    { name = "traitlets" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/53/59/165d3b4d75cc34add3122c4417ecb229085140ac573103c223cd01dde96f/ipython-9.15.0.tar.gz", hash = "sha256:da2819ce2aa83135257df830660b1176d986c3d2876db24df01974fa955b2756", size = 4442580, upload-time = "2026-06-26T11:03:35.913Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/40/3a/948263ca3b9d65bb2b1b0c521b3a49fad5d59ada58724bd87d2bd5ff3f36/ipython-9.15.0-py3-none-any.whl", hash = "sha256:515ad9c3cdf0c932a5a9f6245419e8aba706b7bd03c3e1d3a1c83d9351d6aa6e", size = 630895, upload-time = "2026-06-26T11:03:33.809Z" },
+]
+
+[[package]]
+name = "ipython-pygments-lexers"
+version = "1.1.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "pygments" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/ef/4c/5dd1d8af08107f88c7f741ead7a40854b8ac24ddf9ae850afbcf698aa552/ipython_pygments_lexers-1.1.1.tar.gz", hash = "sha256:09c0138009e56b6854f9535736f4171d855c8c08a563a0dcd8022f78355c7e81", size = 8393, upload-time = "2025-01-17T11:24:34.505Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/d9/33/1f075bf72b0b747cb3288d011319aaf64083cf2efef8354174e3ed4540e2/ipython_pygments_lexers-1.1.1-py3-none-any.whl", hash = "sha256:a9462224a505ade19a605f71f8fa63c2048833ce50abc86768a0d81d876dc81c", size = 8074, upload-time = "2025-01-17T11:24:33.271Z" },
+]
+
+[[package]]
+name = "jedi"
+version = "0.20.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "parso" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/46/b7/a3635f6a2d7cf5b5dd98064fc1d5fbbafcb25477bcea204a3a92145d158b/jedi-0.20.0.tar.gz", hash = "sha256:c3f4ccbd276696f4b19c54618d4fb18f9fc24b0aef02acf704b23f487daa1011", size = 3119416, upload-time = "2026-05-01T23:38:47.814Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/9a/93/242e2eab5fe682ffcb8b0084bde703a41d51e17ee0f3a31ff0d9d813620a/jedi-0.20.0-py2.py3-none-any.whl", hash = "sha256:7bdd9c2634f56713299976f4cbd59cb3fa92165cc5e05ea811fb253480728b67", size = 4884812, upload-time = "2026-05-01T23:38:43.919Z" },
+]
+
+[[package]]
+name = "jiter"
+version = "0.16.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/1d/1f/10936e16d8860c70698a1aa939a46aa0224813b782bce4e000e637da0b2d/jiter-0.16.0.tar.gz", hash = "sha256:7b24c3492c5f4f84a37946ad9cf504910cf6a782d6a4e0689b6673c5894b4a1c", size = 176431, upload-time = "2026-06-29T13:05:13.657Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/83/2b/52ace16ed031354f0539749a49e4bf33797d82bea5137910835fa4b09793/jiter-0.16.0-cp312-cp312-macosx_10_12_x86_64.whl", hash = "sha256:67c3bc1760f8c99d805dcab4e644027142a53b1d5d861f18780ebdbd5d40b72a", size = 306943, upload-time = "2026-06-29T13:03:14.035Z" },
+    { url = "https://files.pythonhosted.org/packages/94/2e/34957c2c1b661c252ba9bcc60ae0bddc27e0f7202c6073326a13c5390eec/jiter-0.16.0-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:5af7780e4a26bd7d0d989592bf9ef12ebf806b74ab709223ecca37c749872ea9", size = 307779, upload-time = "2026-06-29T13:03:15.418Z" },
+    { url = "https://files.pythonhosted.org/packages/88/6c/59bd309cab4460c54cf1079f3eb7fe7af6a4c895c5c957a53378693bad2b/jiter-0.16.0-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d5bf78d0e05e45cfdd66558893938d59afe3d1b1a824a202039b20e607d25a72", size = 335826, upload-time = "2026-06-29T13:03:17.11Z" },
+    { url = "https://files.pythonhosted.org/packages/3b/8c/f5ef7b65f0df47afa16596969defb281ebb86e96df346d62be6fd853d620/jiter-0.16.0-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:f4444a83f946605990c98f625cdd3d2725bfb818158760c5748c653170a20e0e", size = 362573, upload-time = "2026-06-29T13:03:18.781Z" },
+    { url = "https://files.pythonhosted.org/packages/2b/0b/ace4354da061ee38844a0c27dc2c21eecd27aea119e8da324bea987522d0/jiter-0.16.0-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:3a23f0e4f957e1be65752d2dfac9a5a06b1917af8dc85deb639c3b9d02e31290", size = 457979, upload-time = "2026-06-29T13:03:20.293Z" },
+    { url = "https://files.pythonhosted.org/packages/55/40/c0253d3772eb9dcd8e6606ee9b2d53ec8e5b814589c47f140aa585f21eaa/jiter-0.16.0-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:c22a488f7b9218e245a0025a9ba6b100e2e54700831cf4cf16833a27fba3ad01", size = 372302, upload-time = "2026-06-29T13:03:21.739Z" },
+    { url = "https://files.pythonhosted.org/packages/a8/d2/4839422241aa12860ce597b20068727094ba0bc480723c74924ca5bad483/jiter-0.16.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:46add52f4ad47a08bfb1219f3e673da972191489a33016edefdb5ea55bfa8c48", size = 343805, upload-time = "2026-06-29T13:03:23.384Z" },
+    { url = "https://files.pythonhosted.org/packages/e2/59/e196888a05befdda7dbe299b722d56f2f6eec65402bc34c0a3306d595feb/jiter-0.16.0-cp312-cp312-manylinux_2_31_riscv64.whl", hash = "sha256:9c8a956fd72c2cf1e730d01ea080341f13aa0a97a4a33b51abebe725b7ae9ca9", size = 351107, upload-time = "2026-06-29T13:03:24.815Z" },
+    { url = "https://files.pythonhosted.org/packages/ec/74/4cd9e0fca65232136400354b630fbfcd2de634e22ccbb96567725981b548/jiter-0.16.0-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:561926e0573ffe4a32498420a76d64b16c513e1ab413b9d28158a8764ac701e5", size = 388441, upload-time = "2026-06-29T13:03:26.266Z" },
+    { url = "https://files.pythonhosted.org/packages/d9/8c/554691e48bc711299c0a293dd8a6179e24b2d66a54dc295421fcf64569c0/jiter-0.16.0-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:44d019fa8cdaf89bf29c71b39e3712143fdd0ac76725c6ef954f9957a5ea8730", size = 516354, upload-time = "2026-06-29T13:03:28.02Z" },
+    { url = "https://files.pythonhosted.org/packages/a4/cb/01e9d69dc2cc6759d4f91e230b34489c4fdb2518992650633f9e20bece89/jiter-0.16.0-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:0df91907609837f33341b8e6fe73b95991fdaa57caf1a0fbd343dffe826f386f", size = 547880, upload-time = "2026-06-29T13:03:29.534Z" },
+    { url = "https://files.pythonhosted.org/packages/79/70/2953195f1c6ad00f49fa67e13df7e60acb3dd4f387101bc15abccddd905e/jiter-0.16.0-cp312-cp312-win32.whl", hash = "sha256:51d7b836acb0108d7c77df1742332cac2a1fa04a74d6dacec46e7091f0e91274", size = 203473, upload-time = "2026-06-29T13:03:31.025Z" },
+    { url = "https://files.pythonhosted.org/packages/2d/05/2909a8b10699a4d560f8c502b6b2c5f3991b682b1922c1eedda242b225bd/jiter-0.16.0-cp312-cp312-win_amd64.whl", hash = "sha256:1878349266f8ee36ecb1375cc5ba2f115f35fd9f0a1a4119e725e379126647f7", size = 196905, upload-time = "2026-06-29T13:03:32.472Z" },
+    { url = "https://files.pythonhosted.org/packages/e9/a9/6b82bb1c8d7790d602489b967b982a909e5d092875a6c2ade96444c8dfc5/jiter-0.16.0-cp312-cp312-win_arm64.whl", hash = "sha256:2ed5738ae4af18271a51a528b8811b0cbfa4a1858de9d83359e4169855d6a331", size = 190618, upload-time = "2026-06-29T13:03:34.672Z" },
+    { url = "https://files.pythonhosted.org/packages/91/c0/555fc60473d30d66894ba825e63615e3be7524fac23858356afa7a38906c/jiter-0.16.0-cp313-cp313-macosx_10_12_x86_64.whl", hash = "sha256:41977aa5654023948c2dae2a81cbf9c43343954bef1cd59a154dd15a4d84c195", size = 306203, upload-time = "2026-06-29T13:03:36.243Z" },
+    { url = "https://files.pythonhosted.org/packages/d0/2b/c3eaf16f5d7c9bad66ea32f40a95bd169b29a91217fcc7f081375157e99c/jiter-0.16.0-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:d28bb3c26762358dadf3e5bf0bccd29ae987d65e6988d2e6f49829c76b003c09", size = 306489, upload-time = "2026-06-29T13:03:37.846Z" },
+    { url = "https://files.pythonhosted.org/packages/96/3f/02fdfc6705cad96127d883af5c34e4867f554f29ec7705ec1a46156400a9/jiter-0.16.0-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:0542a7189c26920778658fc8fcf2af8bae05bae9924577f71804acef37996536", size = 335453, upload-time = "2026-06-29T13:03:39.221Z" },
+    { url = "https://files.pythonhosted.org/packages/b2/a6/e4bda5920d4b0d7c5dfb7174ce4a6b2e4d3e11c9162c452ef0eab4cdbdbd/jiter-0.16.0-cp313-cp313-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:8fb8de1e23a0cb2a7f53c335049c7b72b6db41aa6227cdcc0972a1de5cb39450", size = 361625, upload-time = "2026-06-29T13:03:40.597Z" },
+    { url = "https://files.pythonhosted.org/packages/b7/97/4e6b59b2c6e55cbb3e183595f81ad65dcfb21c915fee5e19e335df21bc55/jiter-0.16.0-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:b72d0b2990ca754a9102779ac98d8597b7cb31678958562214a007f909eab78e", size = 456958, upload-time = "2026-06-29T13:03:42.074Z" },
+    { url = "https://files.pythonhosted.org/packages/15/e0/97e9557686d2f94f4b93786eccb7eed28e9228ad132ea8237f44727314a7/jiter-0.16.0-cp313-cp313-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:d5f91b1c27fc22a57993d5a5cb8a627cb8ed4b10502716fac1ffbfe1d19d84e8", size = 372017, upload-time = "2026-06-29T13:03:43.658Z" },
+    { url = "https://files.pythonhosted.org/packages/0f/94/db768b6938e0df35c86beeba3dfbbb025c9ee5c19e1aa271f2396e50864d/jiter-0.16.0-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:c682bea068a90b764577bdb78a60a4c1d1606daf9cd4c893832a37c7cc9d9026", size = 343320, upload-time = "2026-06-29T13:03:45.226Z" },
+    { url = "https://files.pythonhosted.org/packages/c1/d6/5a59d938244a30735fe62d9433fd325f9021ea29d89780ea4596ea93bc89/jiter-0.16.0-cp313-cp313-manylinux_2_31_riscv64.whl", hash = "sha256:8d031aabecc4f1b6276adfb42e3aabb77c89d468bf616600e8d3a11328929053", size = 350520, upload-time = "2026-06-29T13:03:46.671Z" },
+    { url = "https://files.pythonhosted.org/packages/67/f8/c4a857f49c9af125f6bbcac7e3eee7f7978ed89682833062e2dbf62576b1/jiter-0.16.0-cp313-cp313-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:eab2cd170150e70153de16896a1774e3a1dca80154c56b54d7a812c479a7165e", size = 387550, upload-time = "2026-06-29T13:03:48.361Z" },
+    { url = "https://files.pythonhosted.org/packages/8b/d6/5fbc2f7d6b67b754caa61a993a2e626e815dec47ffc2f9e35f01adfebec7/jiter-0.16.0-cp313-cp313-musllinux_1_1_aarch64.whl", hash = "sha256:6edb63a46e65a82c26800a868e49b2cac30dd5a4218b88d74bc2c848c8ad60bb", size = 515424, upload-time = "2026-06-29T13:03:49.881Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/54/284f0164b64a5fed915fea6ba7e9ba9b3d8d37c67d59cf2e3bb99d45cdfe/jiter-0.16.0-cp313-cp313-musllinux_1_1_x86_64.whl", hash = "sha256:659039cc50b5addcc35fcc87ae2c1833b7c0a8e5326ef631a75e4478447bcf84", size = 546981, upload-time = "2026-06-29T13:03:51.363Z" },
+    { url = "https://files.pythonhosted.org/packages/13/c5/2a467585a576594384e1d2c43e1224deaafc085f24e243529cf98beef8e1/jiter-0.16.0-cp313-cp313-win32.whl", hash = "sha256:c9c53be232c2e206ef9cdbad81a48bfa74c3d3f08bcf8124630a8a748aad993e", size = 202853, upload-time = "2026-06-29T13:03:53.015Z" },
+    { url = "https://files.pythonhosted.org/packages/88/6a/de61d04b9eec69c71719968d2f716532a3bc121170c44a39e14979c6be81/jiter-0.16.0-cp313-cp313-win_amd64.whl", hash = "sha256:baad945ed47f163ad833314f8e3288c396118934f94e7bbb9e243ce4b341a4fd", size = 196160, upload-time = "2026-06-29T13:03:54.447Z" },
+    { url = "https://files.pythonhosted.org/packages/19/4b/b390ed59bafb3f31d008d1218578f10327714484b334439947f7e5b11e7f/jiter-0.16.0-cp313-cp313-win_arm64.whl", hash = "sha256:3c1fd2dbe1b0af19e987f03fe66c5f5bd105a2229c1aff4ab14890b24f41d21a", size = 189862, upload-time = "2026-06-29T13:03:55.754Z" },
+    { url = "https://files.pythonhosted.org/packages/a7/89/bc4f1b57d5da938fd344a466396541e586d161320d70bffd929aaafcd8f4/jiter-0.16.0-cp314-cp314-macosx_10_12_x86_64.whl", hash = "sha256:b2c61484666ad42726029af0c00ef4541f0f3b5cdc550221f56c2343208018ee", size = 308239, upload-time = "2026-06-29T13:03:57.205Z" },
+    { url = "https://files.pythonhosted.org/packages/65/7a/c415453e5213001bf3b411ff65dec3d303b0e76a4a2cfea9768cd4960994/jiter-0.16.0-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:63efadc657488f45db1c676d81e704cac2abf3fdb892def1faea61db053127e2", size = 308928, upload-time = "2026-06-29T13:03:58.643Z" },
+    { url = "https://files.pythonhosted.org/packages/11/fc/1f4fb7ebf9a724c7741994f4aae18fba1e2f3133df14521a79194952c34a/jiter-0.16.0-cp314-cp314-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:cf0d73f50e7b6935677854f6e8e31d499ca7064dd24734f703e060f5b237d883", size = 336998, upload-time = "2026-06-29T13:04:00.071Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/8d/72cadaac05ccfa7cc3a0a2232862e6c72443ca40cf300ba8b57f9f18b69b/jiter-0.16.0-cp314-cp314-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:bf3ea07d9bc8e7d03a9fbc051295462e6dbc295b894fd72457c3136e3e43d898", size = 362112, upload-time = "2026-06-29T13:04:01.52Z" },
+    { url = "https://files.pythonhosted.org/packages/58/4a/c4b0d5f651fda90a24ffce9f8d56cde462a2e09d31ae3de3c68cef34c04e/jiter-0.16.0-cp314-cp314-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:26798522707abb47d767db536e4148ceac1b14446bf028ee85e579a2e043cfe5", size = 459807, upload-time = "2026-06-29T13:04:03.214Z" },
+    { url = "https://files.pythonhosted.org/packages/80/58/ef77879ea9aa56b50824edc5a445e226422c7a8d211f3fd2a56bcb9493cf/jiter-0.16.0-cp314-cp314-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:bc837c1b9631be10abfe0191537fe8009838204cec7e44827401ace390ddb567", size = 373181, upload-time = "2026-06-29T13:04:04.629Z" },
+    { url = "https://files.pythonhosted.org/packages/49/2e/ffbc3f254e4d8a66da3062c624a7df4b7c2b2cf9e1fe43cf394b3e104041/jiter-0.16.0-cp314-cp314-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:49060fd70737fad59d33ba9dcc0d83247dc9e77187de26053a19c16c9f32bd69", size = 344927, upload-time = "2026-06-29T13:04:06.067Z" },
+    { url = "https://files.pythonhosted.org/packages/9a/f6/0be5dc6d64a89f80aa8fec984f94dedb2973e251edcae55841d60786d578/jiter-0.16.0-cp314-cp314-manylinux_2_31_riscv64.whl", hash = "sha256:adbb8edeadd431bc4477879d5d371ece7cb1334486584e0f252656dd7ffada29", size = 352754, upload-time = "2026-06-29T13:04:07.477Z" },
+    { url = "https://files.pythonhosted.org/packages/da/6e/7d31243b3b91cd261dd19e9d3557fc3251a80883d3d8049c86174e7ab7af/jiter-0.16.0-cp314-cp314-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:31aaee5b80f672c1dc21272bcfb9cbdcfc1ea04ff50f00ed5af500b80c44fa93", size = 390553, upload-time = "2026-06-29T13:04:08.92Z" },
+    { url = "https://files.pythonhosted.org/packages/25/33/51ae371fde3c88897520f62b4d5f8b27ad7103e2bb10812ff52195609853/jiter-0.16.0-cp314-cp314-musllinux_1_1_aarch64.whl", hash = "sha256:6722bcef4ffc86c835574b1b2fac6b33b9fb4a889c781e67950e891591f3c55a", size = 516900, upload-time = "2026-06-29T13:04:10.407Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/45/6449b3d123ea439ba79507c657288f461d55049e7bcbdc2cf8eb8210f491/jiter-0.16.0-cp314-cp314-musllinux_1_1_x86_64.whl", hash = "sha256:5ab4f50ff971b611d656554ea10b75f80097392c827bc32923c6eeb6386c8b00", size = 548754, upload-time = "2026-06-29T13:04:12.046Z" },
+    { url = "https://files.pythonhosted.org/packages/9b/e7/fd2fb11ae3e2649333da3aa170d04d7b3000bbdc3b270f6513382fdf4e04/jiter-0.16.0-cp314-cp314-pyemscripten_2026_0_wasm32.whl", hash = "sha256:710cc51d4ebdcd3c1f70b232c1db1ea1344a075770422bbd4bede5708335acbe", size = 122381, upload-time = "2026-06-29T13:04:13.413Z" },
+    { url = "https://files.pythonhosted.org/packages/26/80/f0b147a62c315a164ed2168908286ca302310824c218d3aae52b06c0c9a9/jiter-0.16.0-cp314-cp314-win32.whl", hash = "sha256:57b37fc887a32d44798e4d8ebfa7c9683ff3da1d5bf38f08d1bb3573ccb39106", size = 204578, upload-time = "2026-06-29T13:04:14.813Z" },
+    { url = "https://files.pythonhosted.org/packages/5e/e6/4758a14304b4523a6f5adb2419340086aa3593bd4327c2b25b5948a90548/jiter-0.16.0-cp314-cp314-win_amd64.whl", hash = "sha256:cbd18dd5e2df96b580487b5745adf57ef64ad89ba2d9662fc3c19386acce7db8", size = 198154, upload-time = "2026-06-29T13:04:16.272Z" },
+    { url = "https://files.pythonhosted.org/packages/26/be/41fa54a2e7ea41d6c99f1dc5b1f0fd4cb474680304b5d268dd518e81da3a/jiter-0.16.0-cp314-cp314-win_arm64.whl", hash = "sha256:a32d2027a9fa67f109ff245a3252ece3ccc32cc56703e1deab6cc846a59e0585", size = 191458, upload-time = "2026-06-29T13:04:17.707Z" },
+    { url = "https://files.pythonhosted.org/packages/81/6b/59127338b86d9fe4d99418f5a15118bea778103ee0fe9d9dd7e0af174e95/jiter-0.16.0-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:2577196f4474ef3fc4779a088a23b0897bbf86f9ea3679c372d45b8383b43207", size = 316739, upload-time = "2026-06-29T13:04:19.663Z" },
+    { url = "https://files.pythonhosted.org/packages/2d/95/49461034d5388196d3dabf98748935f017b7785d8f3f5349f834bcc4ed0d/jiter-0.16.0-cp314-cp314t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:616e89e008a93c01104161c75b4988e58716b01d62307ebfe161e52a56d2a818", size = 340911, upload-time = "2026-06-29T13:04:21.257Z" },
+    { url = "https://files.pythonhosted.org/packages/cd/97/a4369f2fb82cb3dda13b98622f31249b2e014b223fe64ee534413ad72294/jiter-0.16.0-cp314-cp314t-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:0e2e9efbe042210df657bade597f66d6d75723e3d8f45a12ea6d8167ff8bbce3", size = 361747, upload-time = "2026-06-29T13:04:22.677Z" },
+    { url = "https://files.pythonhosted.org/packages/28/51/49b6ed456261646e1906016a6760367a28aacd3c24805e4e5fe64116c1db/jiter-0.16.0-cp314-cp314t-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:3f4d9e473a5ce7d27fef8b848df4dc16e283893d3f53b4a585e72c9595f3c284", size = 460225, upload-time = "2026-06-29T13:04:24.441Z" },
+    { url = "https://files.pythonhosted.org/packages/33/b5/5689aff4f66c5b60be63106e591dbfcba2190df97d2c9c7cf052361ddb98/jiter-0.16.0-cp314-cp314t-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:8d30a4a1c87713060c8d1cc59a7b6c8fb6b8ef0a6900368014c76c87922a2929", size = 373169, upload-time = "2026-06-29T13:04:25.884Z" },
+    { url = "https://files.pythonhosted.org/packages/a2/96/3ae1b85ee0d6d6cab254fb7f8da018272b932bbf2d69b07e98aa2a96c746/jiter-0.16.0-cp314-cp314t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:bae96332410f866e5900d809298b1ed82735932986c672495f9701daacd80620", size = 350332, upload-time = "2026-06-29T13:04:27.302Z" },
+    { url = "https://files.pythonhosted.org/packages/15/32/c99d7bafd78986556c95bf60ce84c6cc98786eac56066c12d7f828bb6747/jiter-0.16.0-cp314-cp314t-manylinux_2_31_riscv64.whl", hash = "sha256:da3d7ec75dc83bb18bca888b5edfae0656a26849056c59e05a7728badd17e7af", size = 353377, upload-time = "2026-06-29T13:04:28.731Z" },
+    { url = "https://files.pythonhosted.org/packages/0e/4b/f99a8e571287c3dec766bcc18528bbe8e8fb5365522ab5e6d64c93e87066/jiter-0.16.0-cp314-cp314t-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:ee6162b77d49a9939229df666dfa8af3e656b6701b54c4c84966d740e189264e", size = 387746, upload-time = "2026-06-29T13:04:30.319Z" },
+    { url = "https://files.pythonhosted.org/packages/75/69/c78a5b3f71040e34eb5917df26fb7ae9a2174cad1ccbf277512507c53a6e/jiter-0.16.0-cp314-cp314t-musllinux_1_1_aarch64.whl", hash = "sha256:63ffdbdae7d4499f4cda14eadc12ddcabef0fc0c081191bdc2247489cb698077", size = 517292, upload-time = "2026-06-29T13:04:31.709Z" },
+    { url = "https://files.pythonhosted.org/packages/c2/f7/095b38eda4c70d03651c403f29a5590f16d12ddc5d544aac9f9cddf72277/jiter-0.16.0-cp314-cp314t-musllinux_1_1_x86_64.whl", hash = "sha256:a111256a7193bea0759267b10385e5870949c239ed7b6ddbaaf57573edb38734", size = 549259, upload-time = "2026-06-29T13:04:33.721Z" },
+    { url = "https://files.pythonhosted.org/packages/2e/c5/6a0207d90e5f656d95af98ebd0934f382d37674416f215aeda2ff8063e51/jiter-0.16.0-cp314-cp314t-win32.whl", hash = "sha256:de5ba8763e56b793561f43bed197c9ea55776daa5e9a6b91eed68a909bc9cdbf", size = 206523, upload-time = "2026-06-29T13:04:35.068Z" },
+    { url = "https://files.pythonhosted.org/packages/a5/31/c757d5f30a8980fd945ce7b98be10be9e4ff59c7c42f5fd86804c2e87db8/jiter-0.16.0-cp314-cp314t-win_amd64.whl", hash = "sha256:b8a3f9a6008048fe9def7bf465180564a6e458047d2ce499149cfbe73c3ae9db", size = 200366, upload-time = "2026-06-29T13:04:36.61Z" },
+    { url = "https://files.pythonhosted.org/packages/7c/a2/d88de6d313d734a544a7901353ad5db67cb38dcfcd91713b7979dafc345d/jiter-0.16.0-cp314-cp314t-win_arm64.whl", hash = "sha256:0fa25b09b13075c46f5bc174f2690525a925a4fc2f7c82969a2bbabff22386ce", size = 190516, upload-time = "2026-06-29T13:04:38.004Z" },
+    { url = "https://files.pythonhosted.org/packages/98/ab/664fd8c4be028b2bedd3d2ff08769c4ede23d0dbc87a77c62384a0515b5d/jiter-0.16.0-graalpy312-graalpy250_312_native-macosx_10_12_x86_64.whl", hash = "sha256:f17d61a28b4b3e0e3e2ba98490c70501403b4d196f78732439160e7fd3678127", size = 303106, upload-time = "2026-06-29T13:05:07.118Z" },
+    { url = "https://files.pythonhosted.org/packages/1a/07/421f1d5b65493a76e16027b848aba6a7d28073ae75944fa4289cc914d39f/jiter-0.16.0-graalpy312-graalpy250_312_native-macosx_11_0_arm64.whl", hash = "sha256:96e38eea538c8ddf853a35727c7be0741c76c13f04148ac5c116222f50ece3b3", size = 304658, upload-time = "2026-06-29T13:05:08.708Z" },
+    { url = "https://files.pythonhosted.org/packages/0a/db/bba1155f01a01c3c37a89425d571da751bbedf5c54247b831a04cb971798/jiter-0.16.0-graalpy312-graalpy250_312_native-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d284fb8d94d5855d60c44fefcab4bf966f1da6fada73992b01f6f0c9bc0c6702", size = 339719, upload-time = "2026-06-29T13:05:10.41Z" },
+    { url = "https://files.pythonhosted.org/packages/78/f7/18a1afcd64f35314b68c1f23afcd9994d0bc13e65cc77517afff4e83986d/jiter-0.16.0-graalpy312-graalpy250_312_native-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:64d613743df53199b1aa256a7d328340da6d7078aac7705a7db9d7a791e9cfd2", size = 343885, upload-time = "2026-06-29T13:05:12.087Z" },
+]
+
+[[package]]
+name = "jsonpatch"
+version = "1.33"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "jsonpointer" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/42/78/18813351fe5d63acad16aec57f94ec2b70a09e53ca98145589e185423873/jsonpatch-1.33.tar.gz", hash = "sha256:9fcd4009c41e6d12348b4a0ff2563ba56a2923a7dfee731d004e212e1ee5030c", size = 21699, upload-time = "2023-06-26T12:07:29.144Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/73/07/02e16ed01e04a374e644b575638ec7987ae846d25ad97bcc9945a3ee4b0e/jsonpatch-1.33-py2.py3-none-any.whl", hash = "sha256:0ae28c0cd062bbd8b8ecc26d7d164fbbea9652a1a3693f3b956c1eae5145dade", size = 12898, upload-time = "2023-06-16T21:01:28.466Z" },
+]
+
+[[package]]
+name = "jsonpointer"
+version = "3.1.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/18/c7/af399a2e7a67fd18d63c40c5e62d3af4e67b836a2107468b6a5ea24c4304/jsonpointer-3.1.1.tar.gz", hash = "sha256:0b801c7db33a904024f6004d526dcc53bbb8a4a0f4e32bfd10beadf60adf1900", size = 9068, upload-time = "2026-03-23T22:32:32.458Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/9e/6a/a83720e953b1682d2d109d3c2dbb0bc9bf28cc1cbc205be4ef4be5da709d/jsonpointer-3.1.1-py3-none-any.whl", hash = "sha256:8ff8b95779d071ba472cf5bc913028df06031797532f08a7d5b602d8b2a488ca", size = 7659, upload-time = "2026-03-23T22:32:31.568Z" },
+]
+
+[[package]]
+name = "jsonschema"
+version = "4.26.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "attrs" },
+    { name = "jsonschema-specifications" },
+    { name = "referencing" },
+    { name = "rpds-py" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b3/fc/e067678238fa451312d4c62bf6e6cf5ec56375422aee02f9cb5f909b3047/jsonschema-4.26.0.tar.gz", hash = "sha256:0c26707e2efad8aa1bfc5b7ce170f3fccc2e4918ff85989ba9ffa9facb2be326", size = 366583, upload-time = "2026-01-07T13:41:07.246Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/69/90/f63fb5873511e014207a475e2bb4e8b2e570d655b00ac19a9a0ca0a385ee/jsonschema-4.26.0-py3-none-any.whl", hash = "sha256:d489f15263b8d200f8387e64b4c3a75f06629559fb73deb8fdfb525f2dab50ce", size = 90630, upload-time = "2026-01-07T13:41:05.306Z" },
+]
+
+[[package]]
+name = "jsonschema-specifications"
+version = "2025.9.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "referencing" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/19/74/a633ee74eb36c44aa6d1095e7cc5569bebf04342ee146178e2d36600708b/jsonschema_specifications-2025.9.1.tar.gz", hash = "sha256:b540987f239e745613c7a9176f3edb72b832a4ac465cf02712288397832b5e8d", size = 32855, upload-time = "2025-09-08T01:34:59.186Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/41/45/1a4ed80516f02155c51f51e8cedb3c1902296743db0bbc66608a0db2814f/jsonschema_specifications-2025.9.1-py3-none-any.whl", hash = "sha256:98802fee3a11ee76ecaca44429fda8a41bff98b00a0f2838151b113f210cc6fe", size = 18437, upload-time = "2025-09-08T01:34:57.871Z" },
+]
+
+[[package]]
+name = "langchain"
+version = "1.3.14"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "langchain-core" },
+    { name = "langgraph" },
+    { name = "pydantic" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/29/68/a6dbad9c22df4087a0f9e79ddd46226c442b30128bfeee538d5889492a73/langchain-1.3.14.tar.gz", hash = "sha256:1b6696c72ba3bbbce54d745e0180742c9f6ece8bbc59ed5a46c3e20b9a435929", size = 645181, upload-time = "2026-07-16T13:28:18.29Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a9/ec/0f942e78a621f8e3162ff1ed24284f469aaf51fb4607ee5831c626f2b2bc/langchain-1.3.14-py3-none-any.whl", hash = "sha256:4d10dbe91005952cddd56d0dc77aa108964da6bae90ab20063653957e901f782", size = 139560, upload-time = "2026-07-16T13:28:16.498Z" },
+]
+
+[[package]]
+name = "langchain-anthropic"
+version = "1.5.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anthropic" },
+    { name = "langchain-core" },
+    { name = "pydantic" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/46/be/07f180f98cee1826c31fd194bd676a43c0e99cdc746966eb61c0e30044a6/langchain_anthropic-1.5.2.tar.gz", hash = "sha256:11cda4fac1129f229022a5c7a9be47156dea02ecf6e666acbb050943eb2df613", size = 712706, upload-time = "2026-07-24T19:08:09.637Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/7d/1f/554cdfcacec55ea00bbefa5d25c80bd3870f0509b532c89e26fffd25f56b/langchain_anthropic-1.5.2-py3-none-any.whl", hash = "sha256:49f94c4daccde31fd06bee70a1a0d00a0b9a91dda10fe1115495e63c9c2f90da", size = 53898, upload-time = "2026-07-24T19:08:08.517Z" },
+]
+
+[[package]]
+name = "langchain-core"
+version = "1.5.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "jsonpatch" },
+    { name = "langchain-protocol" },
+    { name = "langsmith" },
+    { name = "packaging" },
+    { name = "pydantic" },
+    { name = "pyyaml" },
+    { name = "tenacity" },
+    { name = "typing-extensions" },
+    { name = "uuid-utils" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/a7/fc/84e23e8adff5adcd7792273ad610c4972ec534658a52698fb8db6defa87a/langchain_core-1.5.1.tar.gz", hash = "sha256:b0df382704c6403c1e0c9603415bce09290455d4aeeb38f350d15f78ca597483", size = 972219, upload-time = "2026-07-23T20:13:22.282Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/55/1e/1eb8833dc59b9c1f49da89a4a1ad7510440c9eb57fb539ea2203294a1acd/langchain_core-1.5.1-py3-none-any.whl", hash = "sha256:c5ec8f51dd05124f950c9afd0fd8bb3f7be4e405eeb868d481b2f8ff652cb9d2", size = 561634, upload-time = "2026-07-23T20:13:20.717Z" },
+]
+
+[[package]]
+name = "langchain-openai"
+version = "1.4.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "langchain-core" },
+    { name = "openai" },
+    { name = "tiktoken" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/bf/1b/a83bf6cae4632363cef0b6f2ee1b4f62c8a5ebcf22cd8ef24430a736c2a8/langchain_openai-1.4.1.tar.gz", hash = "sha256:6d16be615d997db80294731b8e768783f1fb8e0313668e64acd50cd68acbad20", size = 3262416, upload-time = "2026-07-23T20:31:13.053Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/fe/1c/8b604dc8be2735c8ae5c655e520066231057d1301958c20c776e62bd00fb/langchain_openai-1.4.1-py3-none-any.whl", hash = "sha256:8528bb34cc78fdfd2d895573c7917f9441cbb82db5f18ae0e6b3b75d95bdefb3", size = 122067, upload-time = "2026-07-23T20:31:11.809Z" },
+]
+
+[[package]]
+name = "langchain-protocol"
+version = "0.0.18"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/d2/59/b5959aea96faa9146e2e49a7a22882b3528c62efafe9a6a95beab30c2305/langchain_protocol-0.0.18.tar.gz", hash = "sha256:ec3e11782f1ed0c9db38e5a9ed01b0e7a0d3fba406faa8aef6594b73c56a63e6", size = 6150, upload-time = "2026-06-18T17:08:26.959Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/99/2e/d82db9eec13ad0f72e7aaad5c4bc730ab111934fdc83c85523206eb9b0a0/langchain_protocol-0.0.18-py3-none-any.whl", hash = "sha256:70b53a86fbf9cedc863555effe44da192ab02d556ddbf2cf95b8873adcf41b5a", size = 7221, upload-time = "2026-06-18T17:08:25.996Z" },
+]
+
+[[package]]
+name = "langgraph"
+version = "1.2.9"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "langchain-core" },
+    { name = "langgraph-checkpoint" },
+    { name = "langgraph-prebuilt" },
+    { name = "langgraph-sdk" },
+    { name = "pydantic" },
+    { name = "xxhash" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/41/4b/0d1130e26b41a99dcc88353bbe7162a1f255c4db746bd94024268e6af27b/langgraph-1.2.9.tar.gz", hash = "sha256:385f87bc1802c35af7e0aa479278ecba8582d103515eb48256cb2ddcd42d0bd4", size = 722869, upload-time = "2026-07-10T01:30:14.985Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/41/16/0b8dc48823f1326f3e0c8012a3c07a40da6f194299e2ec080df236287baf/langgraph-1.2.9-py3-none-any.whl", hash = "sha256:c2d98ad94333937922ba04148641c1da2bfe45b5b8e55d7b6dcb0bb2df809e76", size = 247473, upload-time = "2026-07-10T01:30:13.733Z" },
+]
+
+[[package]]
+name = "langgraph-checkpoint"
+version = "4.1.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "langchain-core" },
+    { name = "ormsgpack" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/83/47/886af6f886f0bff2273164a45f008694e48a96ff3cd25ff0228f2aa9480e/langgraph_checkpoint-4.1.1.tar.gz", hash = "sha256:6c2bdb530c91f91d7d9c1bd100925d0fc4f498d418c17f3587d1526279482a25", size = 184020, upload-time = "2026-05-22T16:57:38.503Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/bd/b4/71425e3e38be92611300b9cc5e46a5bf98ab23f5ea8a75b73d02a2f1413c/langgraph_checkpoint-4.1.1-py3-none-any.whl", hash = "sha256:25d29144b082827218e7bc3f1e9b0566a4bb007895cd6cc26f66a8428739f56e", size = 56212, upload-time = "2026-05-22T16:57:37.203Z" },
+]
+
+[[package]]
+name = "langgraph-prebuilt"
+version = "1.1.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "langchain-core" },
+    { name = "langgraph-checkpoint" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/29/66/ed9b93f56bc17ef22d551892f0ac2b225a97fe0fcf23a511b857f70d590b/langgraph_prebuilt-1.1.0.tar.gz", hash = "sha256:3c579cf6eed2d17f9c157c2d0fcaddcd8688524e7022d3b22b37a3bf4589d528", size = 178833, upload-time = "2026-05-12T03:37:49.332Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e9/43/3fe1a700b8490ed02679cdbbc8c915eb23a092faf496c9c1118abcd10be3/langgraph_prebuilt-1.1.0-py3-none-any.whl", hash = "sha256:51e311747d755b751d5c6b39b0c1446124d3a7643d2515017e6714b323508fc9", size = 41043, upload-time = "2026-05-12T03:37:48.007Z" },
+]
+
+[[package]]
+name = "langgraph-sdk"
+version = "0.4.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "httpx" },
+    { name = "langchain-core" },
+    { name = "langchain-protocol" },
+    { name = "orjson" },
+    { name = "websockets" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b4/2b/bd8ac26d4e97f6df88ef05ce5b6a38945a3903e1025d926f4752aa88aa97/langgraph_sdk-0.4.2.tar.gz", hash = "sha256:b88f0f5f6328ac0680d6790614a905b2bcfa257f2276dba4e38f0e86db0aa738", size = 348327, upload-time = "2026-06-01T17:51:19.856Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a0/05/aac507337cceae773c2cc9ab91eb6301963af7aeeb55b4217a00e15aff17/langgraph_sdk-0.4.2-py3-none-any.whl", hash = "sha256:75fa5096c1177ce39c847096a8fe3745ffd480ddb412995f836e9f5f884c43dd", size = 160521, upload-time = "2026-06-01T17:51:18.849Z" },
+]
+
+[[package]]
+name = "langsmith"
+version = "0.10.10"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anyio" },
+    { name = "distro" },
+    { name = "httpx" },
+    { name = "orjson", marker = "platform_python_implementation != 'PyPy'" },
+    { name = "packaging" },
+    { name = "pydantic" },
+    { name = "requests" },
+    { name = "requests-toolbelt" },
+    { name = "sniffio" },
+    { name = "typing-extensions" },
+    { name = "uuid-utils" },
+    { name = "websockets" },
+    { name = "xxhash" },
+    { name = "zstandard" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/10/38/c709ff119172e490a8e8bccd10deeda8da239f15f11521009a58c7b904ac/langsmith-0.10.10.tar.gz", hash = "sha256:e0e175e9dd8de6d96dbb55244482e20590a2691ec7c5e087afc1f302e33d98fe", size = 4739726, upload-time = "2026-07-23T04:18:56.246Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/3e/ec/0b2348bea502e07cb3c2a6e5109423debd82aef9fcef6f807c159ceb5297/langsmith-0.10.10-py3-none-any.whl", hash = "sha256:eb5e9da635f5853fc657cddd1c27f40a8e8b2f00c38051555d608c8e57eb605d", size = 675232, upload-time = "2026-07-23T04:18:54.399Z" },
+]
+
+[[package]]
+name = "lark-oapi"
+version = "1.7.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "httpx" },
+    { name = "pycryptodome" },
+    { name = "requests" },
+    { name = "requests-toolbelt" },
+    { name = "websockets" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/57/d0/0d4abc5e6dce888d71683d84a8ceb6fa9b798c2766e0f070eaacbdca5770/lark_oapi-1.7.1.tar.gz", hash = "sha256:57dc4616010cb9355dba11be3cf9b571b1a088b13abc22baf233c7e3527dd396", size = 2282040, upload-time = "2026-07-08T06:21:40.23Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a5/32/3650ba9230ade1245fa8d6021c9fe8febd27c9b5fdec78ffa465dbbc962e/lark_oapi-1.7.1-py3-none-any.whl", hash = "sha256:16945576a8c783a10e50911c00e421c1212fde3a76b2c2a6a4879b8606c1af8f", size = 7807753, upload-time = "2026-07-08T06:21:30.818Z" },
+]
+
+[[package]]
+name = "lxml"
+version = "6.1.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/05/3b/aab6728cae887456f409b4d75e8a01856e4f04bd510de38052a47768b680/lxml-6.1.1.tar.gz", hash = "sha256:ba96ae44888e0185281e937633a743ea90d5a196c6000f82565ebb0580012d40", size = 4197430, upload-time = "2026-05-18T19:19:06.424Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/6a/6e/c4add832b6fc1e887125b96f880d7b9b70aae5248718e046b1704bcac4b9/lxml-6.1.1-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:104c09bda8d2a562824c0e319d0768ce26a779b7601e0931d33b09b53c392ef7", size = 8570821, upload-time = "2026-05-18T19:17:42.068Z" },
+    { url = "https://files.pythonhosted.org/packages/22/00/ff3009c88e65de8011630acf8ab5a09cb2becd2aaf47fba2f3449f6224e9/lxml-6.1.1-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:25c6997a9a534e016695a0ba06b2f07945de682731ff01065b6d5a4474179da1", size = 4624252, upload-time = "2026-05-18T19:17:47.897Z" },
+    { url = "https://files.pythonhosted.org/packages/42/95/bb63f0fd62e554fe078e1fb3c8fe9083c14ddc7ad7fa178d10e57e071ac7/lxml-6.1.1-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:c921ba5c51e4e9f63b8b00267d06566e1f63407408a0496da2d1d0bfc819c7fc", size = 4930746, upload-time = "2026-05-18T19:18:29.637Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/99/0013e8d9b5960f4f041cf0b73e2f80c23eb5205b1f7bfb20203243651359/lxml-6.1.1-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:54a7f95e4de5fb94e2f9f4b9055c6ba33bf3d628fd77a1d647c5923caa2cdcdc", size = 5093723, upload-time = "2026-05-18T19:18:34.168Z" },
+    { url = "https://files.pythonhosted.org/packages/29/91/317b332636bfc7bddcff828d41b3307f50043f4b237e40849c333d80fa1a/lxml-6.1.1-cp312-cp312-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:96f2ec43df44b1f76249ee0a615334f9b5b060e1c8bd90e706dad2d14d02f383", size = 5005557, upload-time = "2026-05-18T19:18:39.798Z" },
+    { url = "https://files.pythonhosted.org/packages/42/2f/cc9bf06afe70f9c9093ae60855d9759da9db601ec4080f7473319666ffd7/lxml-6.1.1-cp312-cp312-manylinux_2_26_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:70ef8a7e102a1508f8121aae5b0867abd663f72c14f0a9c937e6554cb4587b7b", size = 5631036, upload-time = "2026-05-18T19:18:44.858Z" },
+    { url = "https://files.pythonhosted.org/packages/08/f6/af32e23e563971ffb0fb86be52bc5be5c2c118858ffc119bf6a9039b173d/lxml-6.1.1-cp312-cp312-manylinux_2_26_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:ebe6af670449830d6d9b752c256a983291c766a1365ba5d5460048f9e33a7818", size = 5240367, upload-time = "2026-05-18T19:18:49.217Z" },
+    { url = "https://files.pythonhosted.org/packages/78/83/8555d40948b09ce86f1bd0c68a7ac31d07b1929f92cc1b074006c97ef2d2/lxml-6.1.1-cp312-cp312-manylinux_2_28_i686.whl", hash = "sha256:27acc820660aaffa4f7c087f29120e12980f7779d56d8492d263170111284740", size = 5350171, upload-time = "2026-05-18T19:18:52.779Z" },
+    { url = "https://files.pythonhosted.org/packages/63/75/5d92da93729b7bad783689e6496049fa40927b45bec7bf183c981de3ca70/lxml-6.1.1-cp312-cp312-manylinux_2_31_armv7l.whl", hash = "sha256:1db753c9115ec7100d073b744d17e25e88a8f90f5c39b2f5dd878149af59671f", size = 4694874, upload-time = "2026-05-18T19:18:55.139Z" },
+    { url = "https://files.pythonhosted.org/packages/c5/b5/3aad415a9a25b822e783f15deeb4dffccf5113030f1afa2222dd929313d9/lxml-6.1.1-cp312-cp312-manylinux_2_38_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:c4f469aebd783bb741c2ecb2a681008fd26bfe5c16a9a72ed5467f834e810df2", size = 5244492, upload-time = "2026-05-18T19:19:01.28Z" },
+    { url = "https://files.pythonhosted.org/packages/f1/a1/5fcf7eb9904b80086aa47dcf0027de07b1bb990afad2e6823144c368ae04/lxml-6.1.1-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:766b010012d59470072c1816b5b6c69f1d243e5db36ea5968e94accf430a4635", size = 5048232, upload-time = "2026-05-18T19:18:12.67Z" },
+    { url = "https://files.pythonhosted.org/packages/77/74/1f601b63c7a69fcdf10fa9b148c81da8442204194f6c55509cc485c786b9/lxml-6.1.1-cp312-cp312-musllinux_1_2_armv7l.whl", hash = "sha256:b8d812c6011c08b8111a15e54dd990b8923692d80adf35488bee34026c35accf", size = 4777023, upload-time = "2026-05-18T19:18:15.928Z" },
+    { url = "https://files.pythonhosted.org/packages/a2/b9/7a78f51aec95b1bf780d78e12705a9f6533284f8693dc5c0e6724fa53d3f/lxml-6.1.1-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:fe0306bd29505a9177aac19f1877174b0e7422c222a59f70b2cd41633448c3dc", size = 5645773, upload-time = "2026-05-18T19:18:23.223Z" },
+    { url = "https://files.pythonhosted.org/packages/a5/6e/98a7b7ad54e4e74fa1f20fff776913980619d0ebe5558232d7da6580bdd8/lxml-6.1.1-cp312-cp312-musllinux_1_2_riscv64.whl", hash = "sha256:5ba186ad207446c65d3bb3d3e0412b032b1d9f595e59861e2354798c5703d955", size = 5233088, upload-time = "2026-05-18T19:18:31.433Z" },
+    { url = "https://files.pythonhosted.org/packages/65/d1/bc0ed2427bf609f2ee10da303a6a226f9c8bce94f945dc29a32ce55de6e4/lxml-6.1.1-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:aa366a1e55b8ebfe8ca8ddc3cfe75c8ebade181aeb0f661d0cb05986b647f72a", size = 5260995, upload-time = "2026-05-18T19:18:37.091Z" },
+    { url = "https://files.pythonhosted.org/packages/69/8b/6772e1a4b513fc50a8d931f19edde0e13ae6918510a1e13ff67864f3e5ed/lxml-6.1.1-cp312-cp312-win32.whl", hash = "sha256:126c93f7f56f0eda92f6d8c619edc463a4f23d9252f1c9d0405a76f25fa9f11a", size = 3596382, upload-time = "2026-05-18T19:17:18.37Z" },
+    { url = "https://files.pythonhosted.org/packages/1b/89/45198e9624762af2dfd2cb8782598477ceb29f6e59caab560388ae1f4ec1/lxml-6.1.1-cp312-cp312-win_amd64.whl", hash = "sha256:26e6eda8d38c1fcab1090dd196ee87cbd13788e531937610e2589085de074e77", size = 3997255, upload-time = "2026-05-18T19:17:56.781Z" },
+    { url = "https://files.pythonhosted.org/packages/90/a9/7a54b6834088d9ae528a7b780584ba6a39a9457b0ac330479f20ffbc9449/lxml-6.1.1-cp312-cp312-win_arm64.whl", hash = "sha256:6540377fbd53fe1b629172288c464fb18db11ce1fa7dc15891da10aa9dcc3e7f", size = 3659610, upload-time = "2026-05-19T19:22:50.843Z" },
+    { url = "https://files.pythonhosted.org/packages/a5/eb/7e6f37c5584ccbb2ff267f56fd0339016938c1c8684cfefab9b33ffc2f36/lxml-6.1.1-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:68a9198d0fc122d14bb76837de9aa80cf84caed990b5b237f532ed87d3706736", size = 8559780, upload-time = "2026-05-18T19:17:57.661Z" },
+    { url = "https://files.pythonhosted.org/packages/a1/36/587c2521cf23a2cd6c9c22108aa7528f683a1f195ed7ccd23a4b1786ad36/lxml-6.1.1-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:7d47866cb32fb503450b6edc9df355d10dc49836af2e89901bd6ac6b0896d9d9", size = 4618006, upload-time = "2026-05-18T19:18:04.452Z" },
+    { url = "https://files.pythonhosted.org/packages/6e/ca/ab7bfe2bf4c972af5e7878262845ead3a24a929a9b04bc11c7c1ece6c82a/lxml-6.1.1-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:eb7c9811bfaa8b1ed5ed319f5d370dfbcaa59d52ea64be2a5a85e18195930354", size = 4924139, upload-time = "2026-05-18T19:19:04.873Z" },
+    { url = "https://files.pythonhosted.org/packages/6b/55/a0c72851dfee5ecc689f949723a73dea457758912542cb955b108eaf0d8f/lxml-6.1.1-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:762ff394d5bd56da0cf034a23dcce4e13923f15321a2adfa2ac00201dc6d3fca", size = 5082329, upload-time = "2026-05-18T19:19:09.728Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/b6/0608f7d61a3b96cc67e5648a3d906e31a5082093e10e7be65b3886289938/lxml-6.1.1-cp313-cp313-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:a088f287f7d8275a33c07f2cac6c50b9319309a0200a39e7e75d80c707723099", size = 4993564, upload-time = "2026-05-18T19:19:13.608Z" },
+    { url = "https://files.pythonhosted.org/packages/4c/66/ae227524b066d29d55bf0b453d93d2d793c40218657d643dcbbca13b8faf/lxml-6.1.1-cp313-cp313-manylinux_2_26_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:e902da4b04e6b52e5893900d4b8ab46068f75f3561f01bf1080957f9fd932ed6", size = 5613467, upload-time = "2026-05-18T19:19:16.228Z" },
+    { url = "https://files.pythonhosted.org/packages/a6/76/dbe4a00b50385e40194231dcfe5a12c059de7cf90e89c83407d2b085b719/lxml-6.1.1-cp313-cp313-manylinux_2_26_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:1d4962d4c66bf830a7e59ed6cfc17d148149898a3aefa8ec6e59763e6e3ed085", size = 5228304, upload-time = "2026-05-18T19:19:19.354Z" },
+    { url = "https://files.pythonhosted.org/packages/1c/01/00b1b8442ed2041793336868ba0b9ea4b13d7da7c085c6404c207a63bf79/lxml-6.1.1-cp313-cp313-manylinux_2_28_i686.whl", hash = "sha256:581d4c8ae690a6609e64862dd6b7c2489635c2d13907fc2b20f2bc200ff1d21e", size = 5341607, upload-time = "2026-05-18T19:19:22.297Z" },
+    { url = "https://files.pythonhosted.org/packages/63/36/1ad29931e9a4638bb707869f01d423a6c815f82152138d1a40dfcfde2b95/lxml-6.1.1-cp313-cp313-manylinux_2_31_armv7l.whl", hash = "sha256:876e1ff5930ed8bf295ec5ef9a8155e9b6b1876bbf1deed8b3a8069311875a8f", size = 4700168, upload-time = "2026-05-18T19:19:25.133Z" },
+    { url = "https://files.pythonhosted.org/packages/3c/d1/a9536cecf9be18a0dc72d32bead283a2332d1ffebd2dd3ac70ce444686e5/lxml-6.1.1-cp313-cp313-manylinux_2_38_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:9eb9b5a968f6e0f6d640092a567e14529ff8cea2e29d00da6f78a79fa49f013c", size = 5232487, upload-time = "2026-05-18T19:19:28.603Z" },
+    { url = "https://files.pythonhosted.org/packages/0e/77/b4fb1e03bf5d130e879214d3100092e386418807fb74dd0adc4b0a48f351/lxml-6.1.1-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:aa49e06d94aba782c6a02eecb7e507969e7e7a41b267f1b359bb35585f295d5b", size = 5044231, upload-time = "2026-05-18T19:18:42.246Z" },
+    { url = "https://files.pythonhosted.org/packages/26/4c/d00daeeb0a5530c4028a9232aa1b93db3ef4ed2158c116ea73c79a9765b3/lxml-6.1.1-cp313-cp313-musllinux_1_2_armv7l.whl", hash = "sha256:70cdfd80589d59e43e18005dd7244e8895e93db8ab6a620b7e23df5445a4e3d2", size = 4769450, upload-time = "2026-05-18T19:18:48.013Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/6a/715a3a8d156ce42f29cf014706f5410c2ff3b02267774110fc23266409fe/lxml-6.1.1-cp313-cp313-musllinux_1_2_ppc64le.whl", hash = "sha256:aad9aa39483ed8ec44d6d2e59e5b98a0d80676ef0d92f44bfc374836111f62f5", size = 5635874, upload-time = "2026-05-18T19:18:51.914Z" },
+    { url = "https://files.pythonhosted.org/packages/45/37/0544bc21dde2a88f3a17b504e6fc79c0e01d25a33c2f6079724e9e72b9c7/lxml-6.1.1-cp313-cp313-musllinux_1_2_riscv64.whl", hash = "sha256:d49514be2f28d895c38cf9d2b72d7b9a07d00314519f456c0b50b53cfcf4c785", size = 5223987, upload-time = "2026-05-18T19:18:59.715Z" },
+    { url = "https://files.pythonhosted.org/packages/4d/f8/f6a5e8185bcb28c2befae3d31f8e3df3b811cb0f47746517a81279fcafe1/lxml-6.1.1-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:47402e62c52ff5988c1e8c6c63177f5708bccf48e366dea4e3dcf1e645e04947", size = 5250276, upload-time = "2026-05-18T19:19:03.834Z" },
+    { url = "https://files.pythonhosted.org/packages/c7/f2/1a2b9f1b7a49d45495369be7ef9ad05b262930f2eab3e3145706fca8083f/lxml-6.1.1-cp313-cp313-win32.whl", hash = "sha256:3483644525531e1d5762b0c44a8e18b6efba321b6dcf8a8952de10b037618bca", size = 3596903, upload-time = "2026-05-18T19:17:29.863Z" },
+    { url = "https://files.pythonhosted.org/packages/e6/99/f4ffb024f238eec2131aaa09f3278fb6129cf892741bf68e1fc1afb8c100/lxml-6.1.1-cp313-cp313-win_amd64.whl", hash = "sha256:a10bd2fd62e8ce916ececb342f348f190724a098c1faa056fdfb2a22ad5e8660", size = 3995869, upload-time = "2026-05-18T19:18:02.596Z" },
+    { url = "https://files.pythonhosted.org/packages/d1/53/70eb8c5c6037f27448f1e3c54ebede9545a801ae63f0a7254afca4fe8e45/lxml-6.1.1-cp313-cp313-win_arm64.whl", hash = "sha256:424aa57aca0897eb922aef34395bd1289b3b6f04e6bae20ea123c0c7e333cffc", size = 3658490, upload-time = "2026-05-19T19:22:53.846Z" },
+    { url = "https://files.pythonhosted.org/packages/13/e2/2e325795566de01d0d7c3bb57d3c370616b2d07b01214e84eec5d3b10963/lxml-6.1.1-cp314-cp314-macosx_10_15_universal2.whl", hash = "sha256:19b7ab10b210b0b3ad7985d9ac4eb66ab09a90b20fe6e2f7ba55d01a234345d0", size = 8577146, upload-time = "2026-05-18T19:18:17.765Z" },
+    { url = "https://files.pythonhosted.org/packages/93/cf/5630b5e4be7d2e6bee8efe83865c925221103cf0221303b104ce134b01e2/lxml-6.1.1-cp314-cp314-macosx_10_15_x86_64.whl", hash = "sha256:c08e5c694306507275f2290073350c4f32e383db15213b2c69e7ff39c1193840", size = 4623866, upload-time = "2026-05-18T19:18:30.669Z" },
+    { url = "https://files.pythonhosted.org/packages/d2/51/3904907c063451cf8d4a5c9fe0cad95fa1f4ec57f4e3884fa0731bd7a305/lxml-6.1.1-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:74a9717fd0d82effef5c2854f0d917231d5324b5a3eb7275c43ac9fa32f97a14", size = 4950022, upload-time = "2026-05-18T19:19:31.958Z" },
+    { url = "https://files.pythonhosted.org/packages/94/cd/9c7611a51c37a2830928405817cc5d56a97f64fab83cc3f628748b135749/lxml-6.1.1-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:efe0374196335f93b53269acd811b944f2e6bdc88e8894f214bd636455484909", size = 5086695, upload-time = "2026-05-18T19:19:34.764Z" },
+    { url = "https://files.pythonhosted.org/packages/da/d6/24e3b5906abb0b674ff2ae195bc3ce59708df2bcd17cf17703b2d7dd643a/lxml-6.1.1-cp314-cp314-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:ac931cdc9442c1763b8a8f6cd62c0c938737eafc5be75eff88df55fc73bc0d00", size = 5031642, upload-time = "2026-05-18T19:19:37.771Z" },
+    { url = "https://files.pythonhosted.org/packages/2d/db/6ec54f99019838bff54785c51da07f189eb4676861c5f2730962b0d8d665/lxml-6.1.1-cp314-cp314-manylinux_2_26_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:aee395f5d0927f947758b4ec119fd5fc8ec71f07a1c5c52077b30b04c0fa6955", size = 5647338, upload-time = "2026-05-18T19:19:40.553Z" },
+    { url = "https://files.pythonhosted.org/packages/42/3d/ef4dcfffd22d27a61805d8ed9f7fb888495bc6aa88648fa07c1eaa5586b6/lxml-6.1.1-cp314-cp314-manylinux_2_26_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:9395002973c827b3ed67db77e6ec09f092919a587022174554096a269378fb13", size = 5239528, upload-time = "2026-05-18T19:19:43.657Z" },
+    { url = "https://files.pythonhosted.org/packages/62/bb/37fb3f0dff146bdcfa78eec47879273820b2a0bf350ec236ce14bd0b1c26/lxml-6.1.1-cp314-cp314-manylinux_2_28_i686.whl", hash = "sha256:73bc2086f141224ebddb7fc5c6a36ca58b31b94b561e1dfe8e073e3270fad1e7", size = 5350730, upload-time = "2026-05-18T19:19:46.307Z" },
+    { url = "https://files.pythonhosted.org/packages/90/42/43253f168388df4fae1f38c01df36ddb9bee39e2048167b54cdcbae85ea3/lxml-6.1.1-cp314-cp314-manylinux_2_31_armv7l.whl", hash = "sha256:3779def59032b81e44a5f70096ef6bf2082f8d901937dca354474ba09782e245", size = 4697530, upload-time = "2026-05-18T19:19:49.889Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/a8/c5a8504f81bbdfc8e7094c2c850cdb4ed6777fc4d5ddd9e5ab819f3b0d54/lxml-6.1.1-cp314-cp314-manylinux_2_38_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:86c89b9d55ebf820ad7c90bc533410f0d098054f293351f10603c0c46ff598f5", size = 5250670, upload-time = "2026-05-18T19:19:53.199Z" },
+    { url = "https://files.pythonhosted.org/packages/77/b7/c7e76ab18744d75e21f320ebf9ff9d1ceae2b54dd431ea5a64caf26c9672/lxml-6.1.1-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:19607c6bbff2a44cf3fe8250abccd20942d3462473e0a721d01d379ed017e462", size = 5084485, upload-time = "2026-05-18T19:19:08.422Z" },
+    { url = "https://files.pythonhosted.org/packages/31/31/b35c53f8ef7b7c31cacd23d3638652fff7bcd1deb6eedb709ab43b685908/lxml-6.1.1-cp314-cp314-musllinux_1_2_armv7l.whl", hash = "sha256:c6ed5141a5c7507cf3ee76bd363b0d6f801e3321adc35b5d825a23115faa5465", size = 4737635, upload-time = "2026-05-18T19:19:12.321Z" },
+    { url = "https://files.pythonhosted.org/packages/d9/06/31f23c813a7fe8e0cb1b175e915b08c9bf4e86d225b210feadbdbe519667/lxml-6.1.1-cp314-cp314-musllinux_1_2_ppc64le.whl", hash = "sha256:62aeb7e85b5d60320b9d77eef2e773994e2c0ce10121b277e0a19804e1654a5a", size = 5670681, upload-time = "2026-05-18T19:19:15.001Z" },
+    { url = "https://files.pythonhosted.org/packages/1a/bc/ce619bccc89b1fd9ad8a8e1330ee3f3beff9f2ff95b712d7bbcdd6e22fc3/lxml-6.1.1-cp314-cp314-musllinux_1_2_riscv64.whl", hash = "sha256:b1b963fd8f5caa68e99dfae060d54de1fe9cba899b8718b44a00cdca53c3e590", size = 5238229, upload-time = "2026-05-18T19:19:18.131Z" },
+    { url = "https://files.pythonhosted.org/packages/2f/5d/b329acbbedc0b619ebc2be6cf7ee9ed07e80892c88d4dfd612c33805789a/lxml-6.1.1-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:63876be28efefa04a1df615b46770e82042cce445cfdce55160522f57b231ccb", size = 5264191, upload-time = "2026-05-18T19:19:21.118Z" },
+    { url = "https://files.pythonhosted.org/packages/d6/85/be36fb1425b30db3c3f9df75fe86343ebffb79e6320bd7f588e25bfeac39/lxml-6.1.1-cp314-cp314-win32.whl", hash = "sha256:7f7a92e8583f06b1fd49d01158143b8461cfcd135dcb10ec807270a3051bd603", size = 3657202, upload-time = "2026-05-18T19:17:39.509Z" },
+    { url = "https://files.pythonhosted.org/packages/b8/ce/3cf9a827342269f54d405a6202397de63f07c69cbd6ce7d183a3f0cba1e9/lxml-6.1.1-cp314-cp314-win_amd64.whl", hash = "sha256:b2d444f2e66624d68e9c6b211e28a76e22fff5fcabcfff4deac18b529b7d4137", size = 4064497, upload-time = "2026-05-18T19:18:14.662Z" },
+    { url = "https://files.pythonhosted.org/packages/d9/3e/1a957bde8f0760039e627f94699f82caa782c9d838d86c3d28245ee67212/lxml-6.1.1-cp314-cp314-win_arm64.whl", hash = "sha256:3fd9728a2735fda14f4e8235830c86b539e9661e849665bf926d3f867943b4bf", size = 3741991, upload-time = "2026-05-19T19:22:59.111Z" },
+    { url = "https://files.pythonhosted.org/packages/78/b2/00ed55b3a2efa4658fb795c38d1090ec9b3e8a6c3683d4441fa517f09c3b/lxml-6.1.1-cp314-cp314t-macosx_10_15_universal2.whl", hash = "sha256:787b2496d0dbe8cd180984e8d29e3a6f76e7ea34db781cb3bd55e4ba1ef8b4ee", size = 8827545, upload-time = "2026-05-18T19:18:41.193Z" },
+    { url = "https://files.pythonhosted.org/packages/c0/73/74573db19baa618d5f266f2407898b087ff6927115b00b71e5fc1b700847/lxml-6.1.1-cp314-cp314t-macosx_10_15_x86_64.whl", hash = "sha256:2c8daa471358dc2d6fcf02165e80ec68f77871a286df95bc5cc3816153b0fd2c", size = 4735736, upload-time = "2026-05-18T19:18:46.761Z" },
+    { url = "https://files.pythonhosted.org/packages/16/02/6f7061f4f95f51e545d48e87647c54791d204a4e881be4156e7a26ba5338/lxml-6.1.1-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:acd7d70b64c0aae0c7922cca83d288a16f5f6da523637697872253415269baef", size = 4970291, upload-time = "2026-05-18T19:19:56.215Z" },
+    { url = "https://files.pythonhosted.org/packages/b0/02/55fc057d8283427dea7d6edb102e7a840239c77a64a983d92f62a304c0e9/lxml-6.1.1-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:4f0dd2f01f9f8a89f565d000e03abcf0a13d692a346c8d22f628d49af098777a", size = 5102822, upload-time = "2026-05-18T19:19:59.223Z" },
+    { url = "https://files.pythonhosted.org/packages/e4/48/8e1cf78d89d66850121d9255a2a24414c98f775da93b90cf976956c24b14/lxml-6.1.1-cp314-cp314t-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:0b7e8a14c8634bf6f7a568634cb395305a6d964aeb5b7ee32248094bed3a7e2c", size = 5027923, upload-time = "2026-05-18T19:20:01.549Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/00/0632a0647612c8af24d26997b3b961397daa9d5b2581444805933629a4cb/lxml-6.1.1-cp314-cp314t-manylinux_2_26_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:86281fbdd6a8162756f8d603f37e3435bfa38043adb79c6dc6a2dfee065e7525", size = 5595843, upload-time = "2026-05-18T19:20:03.93Z" },
+    { url = "https://files.pythonhosted.org/packages/bc/86/ab008a7dc360711b66858d61c80a5979a70a09f2aa2b05d9698df80b803d/lxml-6.1.1-cp314-cp314t-manylinux_2_26_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:c5d7152ec39ca7c402d8fb9bad86140a15b9503bd0c54484e3f1bbe3dd37ceca", size = 5224515, upload-time = "2026-05-18T19:20:06.381Z" },
+    { url = "https://files.pythonhosted.org/packages/75/c6/2702ff375e728e34f56d9a45339a9cf7e4427e917f542225242d63a05afa/lxml-6.1.1-cp314-cp314t-manylinux_2_28_i686.whl", hash = "sha256:88d8cb75b9d82858497a5393e3c63cfbf03035225e4b35a49ed7ccb151e4dc0e", size = 5312511, upload-time = "2026-05-18T19:20:09.308Z" },
+    { url = "https://files.pythonhosted.org/packages/b7/57/a5807c98f87a86f10ef9ffab35516df7c0f0c4b6d5d33e9f608ab9c04a31/lxml-6.1.1-cp314-cp314t-manylinux_2_31_armv7l.whl", hash = "sha256:f64ec5397ea6a41fc1b4af0380d79b44a755b5531dcaccd9940fb260dca93038", size = 4639206, upload-time = "2026-05-18T19:20:11.704Z" },
+    { url = "https://files.pythonhosted.org/packages/1f/e1/8a0a2c35734812395f4da4eaf33748a7e5705bfb2a58b128da764339d5ec/lxml-6.1.1-cp314-cp314t-manylinux_2_38_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:d34bbf07dbc7ca5970671b1512e928991fb5e9d95365636c9b2d8b4f53af405e", size = 5232404, upload-time = "2026-05-18T19:20:14.064Z" },
+    { url = "https://files.pythonhosted.org/packages/c2/e2/0e6a4dd5ad84d01d99aa7bae7cfefd4a760a0e0f8176818241de17d9b6c0/lxml-6.1.1-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:17e0e18d4ad8adbd0399291bc44845b69d9dd68439a3cdebdf35ff902ec05072", size = 5083769, upload-time = "2026-05-18T19:19:23.758Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/7e/161f33d463f6ffc1c7679104b65086dea120080d49dde4d238f015aaee2f/lxml-6.1.1-cp314-cp314t-musllinux_1_2_armv7l.whl", hash = "sha256:3ab541146f1f6968c462d6c2ac495148e8cdba2f8347700b2141b6ec5a75bf52", size = 4758936, upload-time = "2026-05-18T19:19:27.256Z" },
+    { url = "https://files.pythonhosted.org/packages/f1/fb/2369825e3f6ca99305bf9f7b7085fda91c8b0922a89e54d900974aa3ef85/lxml-6.1.1-cp314-cp314t-musllinux_1_2_ppc64le.whl", hash = "sha256:2a0217714657e023ef4293500f65aa20fce6164c8fd6b08fa5bd4a859fb14b9b", size = 5620296, upload-time = "2026-05-18T19:19:29.993Z" },
+    { url = "https://files.pythonhosted.org/packages/30/90/d61e383146f74c5ab683947ea14dc7b82778838ab9b95ea73a23b60d0191/lxml-6.1.1-cp314-cp314t-musllinux_1_2_riscv64.whl", hash = "sha256:05a82eb6e1530a64f26225b55cbd178113bd0b5af1c2b625f25e5296742c26d2", size = 5228598, upload-time = "2026-05-18T19:19:33.523Z" },
+    { url = "https://files.pythonhosted.org/packages/76/2d/2dafd8149e94b05bb070690efd5bb2680720681e03ff03fc57d2b70a1105/lxml-6.1.1-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:9e36f163528fc50cbef305f02a5fd66d404edf7049cdaff211dbc2cba5a7013e", size = 5247845, upload-time = "2026-05-18T19:19:36.649Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/68/b30e913340c380ddac9580c6e6230991fc37240ec4f64704833e4f3e2769/lxml-6.1.1-cp314-cp314t-win32.whl", hash = "sha256:649dda677cf3bd6ac9ae14007ba0c824ded8ce5808b53fc7431d9140399118c1", size = 3897345, upload-time = "2026-05-18T19:17:33.562Z" },
+    { url = "https://files.pythonhosted.org/packages/3c/4e/9eb2af5335545f9fbcd7af57bcf87c6025d31eaa31b14ec184a6c8675328/lxml-6.1.1-cp314-cp314t-win_amd64.whl", hash = "sha256:793033d6c5cdf33a573f910d9bea14ef8f5771820411d118da8e1182edb53d5e", size = 4393350, upload-time = "2026-05-18T19:18:10.076Z" },
+    { url = "https://files.pythonhosted.org/packages/7f/2c/0f1e93c636720e8a3eb59af2bfda99d98b55891e1c53bc30c2e0e865f01b/lxml-6.1.1-cp314-cp314t-win_arm64.whl", hash = "sha256:58bb955caba94e467d2a96da17660d2d704e0675894cba21ab8a775b8621fd1c", size = 3817223, upload-time = "2026-05-19T19:22:56.823Z" },
+]
+
+[[package]]
+name = "markdown-it-py"
+version = "4.2.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "mdurl" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/06/ff/7841249c247aa650a76b9ee4bbaeae59370dc8bfd2f6c01f3630c35eb134/markdown_it_py-4.2.0.tar.gz", hash = "sha256:04a21681d6fbb623de53f6f364d352309d4094dd4194040a10fd51833e418d49", size = 82454, upload-time = "2026-05-07T12:08:28.36Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/b3/81/4da04ced5a082363ecfa159c010d200ecbd959ae410c10c0264a38cac0f5/markdown_it_py-4.2.0-py3-none-any.whl", hash = "sha256:9f7ebbcd14fe59494226453aed97c1070d83f8d24b6fc3a3bcf9a38092641c4a", size = 91687, upload-time = "2026-05-07T12:08:27.182Z" },
+]
+
+[[package]]
+name = "markdownify"
+version = "1.2.3"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "beautifulsoup4" },
+    { name = "six" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/92/ab/d1297139c0e2ceb151ae564c8c4f57ac0155d8f1f8b4cbd5d6523c82ea36/markdownify-1.2.3.tar.gz", hash = "sha256:1a176f05522c8a2cb1dd3ab9d307dcdadbed5c26ae717855bfc42b3b6d38d937", size = 18852, upload-time = "2026-06-30T20:27:39.06Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/04/10/fa543d484e8b1199243fe20eedd02cc5af050edebce98a7293a5773df592/markdownify-1.2.3-py3-none-any.whl", hash = "sha256:a189a0bedfd14009030fde5f85bb6f77c56897cb839b5c25315dd7d4e3e290ba", size = 15732, upload-time = "2026-06-30T20:27:38.094Z" },
+]
+
+[[package]]
+name = "matplotlib-inline"
+version = "0.2.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "traitlets" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/bd/c0/9f7c9a46090390368a4d7bcb76bb87a4a36c421e4c0792cdb53486ffac7a/matplotlib_inline-0.2.2.tar.gz", hash = "sha256:72f3fe8fce36b70d4a5b612f899090cd0401deddc4ea90e1572b9f4bfb058c79", size = 8150, upload-time = "2026-05-08T17:33:33.49Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/41/09/5b161152e2d90f7b87f781c2e1267494aef9c32498df793f73ad0a0a494a/matplotlib_inline-0.2.2-py3-none-any.whl", hash = "sha256:3c821cf1c209f59fb2d2d64abbf5b23b67bcb2210d663f9918dd851c6da1fcf6", size = 9534, upload-time = "2026-05-08T17:33:32.055Z" },
+]
+
+[[package]]
+name = "mcp"
+version = "1.28.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anyio" },
+    { name = "httpx" },
+    { name = "httpx-sse" },
+    { name = "jsonschema" },
+    { name = "pydantic" },
+    { name = "pydantic-settings" },
+    { name = "pyjwt", extra = ["crypto"] },
+    { name = "python-multipart" },
+    { name = "pywin32", marker = "sys_platform == 'win32'" },
+    { name = "sse-starlette" },
+    { name = "starlette" },
+    { name = "typing-extensions" },
+    { name = "typing-inspection" },
+    { name = "uvicorn", marker = "sys_platform != 'emscripten'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/6e/77/9450b8f251a13affb6281997d0523c4615f8a8b35d0b21ff30db3a5aac9d/mcp-1.28.1.tar.gz", hash = "sha256:d51e36a5f5644faea4f85ea649bfffa6bc6c26770d42798ad6a3de3d2ba69683", size = 638501, upload-time = "2026-06-26T12:57:29.093Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e2/5e/d118fce19f87a2e7d8101c35c8ae0ec289098a4df0ff244cec23e415aca0/mcp-1.28.1-py3-none-any.whl", hash = "sha256:2726bca5e7193f61c5dde8b12500a6de2d9acf6d1a1c0be9e8c2e706437991df", size = 222620, upload-time = "2026-06-26T12:57:27.218Z" },
+]
+
+[[package]]
+name = "mdurl"
+version = "0.1.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/d6/54/cfe61301667036ec958cb99bd3efefba235e65cdeb9c84d24a8293ba1d90/mdurl-0.1.2.tar.gz", hash = "sha256:bb413d29f5eea38f31dd4754dd7377d4465116fb207585f97bf925588687c1ba", size = 8729, upload-time = "2022-08-14T12:40:10.846Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/b3/38/89ba8ad64ae25be8de66a6d463314cf1eb366222074cfda9ee839c56a4b4/mdurl-0.1.2-py3-none-any.whl", hash = "sha256:84008a41e51615a49fc9966191ff91509e3c40b939176e643fd50a5c2196b8f8", size = 9979, upload-time = "2022-08-14T12:40:09.779Z" },
+]
+
+[[package]]
+name = "msgspec"
+version = "0.21.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/e3/60/f79b9b013a16fa3a58350c9295ddc6789f2e335f36ea61ed10a21b215364/msgspec-0.21.1.tar.gz", hash = "sha256:2313508e394b0d208f8f56892ca9b2799e2561329de9763b19619595a6c0f72c", size = 319193, upload-time = "2026-04-12T21:44:50.394Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/6e/cf/317224852c00248c620a9bcf4b26e2e4ab8afd752f18d2a6ef73ebd423b6/msgspec-0.21.1-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:d4248cf0b6129b7d230eacd493c17cc2d4f3989f3bb7f633a928a85b7dcfa251", size = 196188, upload-time = "2026-04-12T21:44:07.181Z" },
+    { url = "https://files.pythonhosted.org/packages/6d/81/074612945c0666078f7366f40000013de9f6ba687491d450df699bceebc9/msgspec-0.21.1-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:5102c7e9b3acff82178449b85006d96310e690291bb1ea0142f1b24bcb8aabcb", size = 188473, upload-time = "2026-04-12T21:44:08.736Z" },
+    { url = "https://files.pythonhosted.org/packages/8a/37/655101799590bcc5fddb2bd3fe0e6194e816c2d1da7c361725f5eb89a910/msgspec-0.21.1-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:846758412e9518252b2ac9bffd6f0e54d9ff614f5f9488df7749f81ff5c80920", size = 218871, upload-time = "2026-04-12T21:44:09.917Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/d1/d4cd9fe89c7d400d7a18f86ccc94daa3f0927f53558846fcb60791dce5d6/msgspec-0.21.1-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:21995e74b5c598c2e004110ad66ec7f1b8c20bf2bcf3b2de8fd9a3094422d3ff", size = 225025, upload-time = "2026-04-12T21:44:11.191Z" },
+    { url = "https://files.pythonhosted.org/packages/24/bf/e20549e602b9edccadeeff98760345a416f9cce846a657e8b18e3396b212/msgspec-0.21.1-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:6129f0cca52992e898fd5344187f7c8127b63d810b2fd73e36fca73b4c6475ee", size = 222672, upload-time = "2026-04-12T21:44:12.481Z" },
+    { url = "https://files.pythonhosted.org/packages/b4/68/04d7a8f0f786545cf9b8c280c57aa6befb5977af6e884b8b54191cbe44b3/msgspec-0.21.1-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:ef3ec2296248d1f8b9231acb051b6d471dfde8f21819e86c9adaaa9f42918521", size = 227303, upload-time = "2026-04-12T21:44:13.709Z" },
+    { url = "https://files.pythonhosted.org/packages/cc/4d/619866af2840875be408047bf9e70ceafbae6ab50660de7134ed1b25eb86/msgspec-0.21.1-cp312-cp312-win_amd64.whl", hash = "sha256:d4ab834a054c6f0cbeef6df9e7e1b33d5f1bc7b86dea1d2fd7cad003873e783d", size = 190017, upload-time = "2026-04-12T21:44:14.977Z" },
+    { url = "https://files.pythonhosted.org/packages/5e/2e/a8f9eca8fd00e097d7a9e99ba8a4685db994494448e3d4f0b7f6e9a3c0f7/msgspec-0.21.1-cp312-cp312-win_arm64.whl", hash = "sha256:628aaa35c74950a8c59da330d7e98917e1c7188f983745782027748ee4ca573e", size = 175345, upload-time = "2026-04-12T21:44:16.431Z" },
+    { url = "https://files.pythonhosted.org/packages/7e/74/f11ede02839b19ff459f88e3145df5d711626ca84da4e23520cebf819367/msgspec-0.21.1-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:764173717a01743f007e9f74520ed281f24672c604514f7d76c1c3a10e8edb66", size = 196176, upload-time = "2026-04-12T21:44:17.613Z" },
+    { url = "https://files.pythonhosted.org/packages/bb/40/4476c1bd341418a046c4955aff632ec769315d1e3cb94e6acf86d461f9ed/msgspec-0.21.1-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:344c7cd0eaed1fb81d7959f99100ef71ec9b536881a376f11b9a6c4803365697", size = 188524, upload-time = "2026-04-12T21:44:18.815Z" },
+    { url = "https://files.pythonhosted.org/packages/ca/d9/9e9d7d7e5061b47540d03d640fab9b3965ba7ae49c1b2154861c8f007518/msgspec-0.21.1-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:48943e278b3854c2f89f955ddc6f9f430d3f0784b16e47d10604ee0463cd21f5", size = 218880, upload-time = "2026-04-12T21:44:20.028Z" },
+    { url = "https://files.pythonhosted.org/packages/74/66/2bb344f34abb4b57e60c7c9c761994e0417b9718ec1460bf00c296f2a7ea/msgspec-0.21.1-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:a9aa659ebb0101b1cbc31461212b87e341d961f0ab0772aaf068a99e001ec4aa", size = 225050, upload-time = "2026-04-12T21:44:21.577Z" },
+    { url = "https://files.pythonhosted.org/packages/1a/84/7c1e412f76092277bf760cef12b7979d03314d259ab5b5cafde5d0c1722d/msgspec-0.21.1-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:f7b27d1a8ead2b6f5b0c4f2d07b8be1ccfcc041c8a0e704781edebe3ae13c484", size = 222713, upload-time = "2026-04-12T21:44:22.83Z" },
+    { url = "https://files.pythonhosted.org/packages/4e/27/0bba04b2b4ef05f3d068429410bc71d2cea925f1596a8f41152cccd5edb8/msgspec-0.21.1-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:38fe93e86b61328fe544cb7fd871fad5a27c8734bfda90f65e5dbe288ae50f61", size = 227259, upload-time = "2026-04-12T21:44:24.11Z" },
+    { url = "https://files.pythonhosted.org/packages/b0/2d/09574b0eea02fed2c2c1383dbaae2c7f79dc16dcd6487a886000afb5d7c4/msgspec-0.21.1-cp313-cp313-win_amd64.whl", hash = "sha256:8bc666331c35fcce05a7cd2d6221adbe0f6058f8e750711413d22793c080ac6a", size = 189857, upload-time = "2026-04-12T21:44:25.359Z" },
+    { url = "https://files.pythonhosted.org/packages/46/34/105b1576ad182879914f0c821f17ee1d13abb165cb060448f96fe2aff078/msgspec-0.21.1-cp313-cp313-win_arm64.whl", hash = "sha256:42bb1241e0750c1a4346f2aa84db26c5ffd99a4eb3a954927d9f149ff2f42898", size = 175403, upload-time = "2026-04-12T21:44:26.608Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/ad/86954e987d1d6a5c579e2c2e7832b65e0fff194179fdac4f581536086024/msgspec-0.21.1-cp314-cp314-macosx_10_15_x86_64.whl", hash = "sha256:fab48eb45fdbfbdb2c0edfec00ffc53b6b6085beefc6b50b61e01659f9f8757f", size = 196261, upload-time = "2026-04-12T21:44:27.807Z" },
+    { url = "https://files.pythonhosted.org/packages/d1/a1/c5e46c3e42b866199365e35d11dddfd1fbd8bba4fdb3c52f965b1607ce94/msgspec-0.21.1-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:3cb779ea0c35bc807ff941d415875c1f69ca0be91a2e907ab99a171811d86a9a", size = 188729, upload-time = "2026-04-12T21:44:28.99Z" },
+    { url = "https://files.pythonhosted.org/packages/85/7d/1e29a319d678d6cb962ae5bdf32a6858ebdf38f73bc654c0e9c742a0c2c8/msgspec-0.21.1-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:68604db36b3b4dd9bf160e436e12798a4738848144cea1aca1cb984011eb160f", size = 219866, upload-time = "2026-04-12T21:44:31.104Z" },
+    { url = "https://files.pythonhosted.org/packages/25/1f/cca084ca2572810fff12ea9dbdcbe39eac048f40daf4a9077b49fcbe8cee/msgspec-0.21.1-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:3d6b9dc50948eaf65df54d2fd0ff66e6d8c32f116037209ee861810eb9b676cb", size = 224993, upload-time = "2026-04-12T21:44:32.649Z" },
+    { url = "https://files.pythonhosted.org/packages/71/94/d2120fc9d419a89a3a7c13e5b7078798c4b392a96a02a6e2b3ce43a8766c/msgspec-0.21.1-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:52c5e21930942302394429c5a582ce7e6b62c7f983b3760834c2ce107e0dd6df", size = 223535, upload-time = "2026-04-12T21:44:33.839Z" },
+    { url = "https://files.pythonhosted.org/packages/75/17/42418b66a3ad972a89bab73dd78b79cc6282bb488a25e73c853cee7443b9/msgspec-0.21.1-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:abbb39d65681fa24ed394e01af3d59d869068324f900c61d06062b7fb9980f2f", size = 227222, upload-time = "2026-04-12T21:44:35.093Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/33/265c894268cca88ff67b144ca2b4c522fc8b9a6f1966a3640c70516e78e1/msgspec-0.21.1-cp314-cp314-win_amd64.whl", hash = "sha256:5666b1b560b97b6ec2eb3fca8a502298ebac56e13bbca1f88523538ce83d01ea", size = 193810, upload-time = "2026-04-12T21:44:36.612Z" },
+    { url = "https://files.pythonhosted.org/packages/3b/8f/a6d35f25bf1fc63c492fdd88fdce01ba0875ead48c2b91f90f33653b4131/msgspec-0.21.1-cp314-cp314-win_arm64.whl", hash = "sha256:d8b8578e4c83b14ceea4cef0d0b747e31d9330fe4b03b2b2ad4063866a178f93", size = 179125, upload-time = "2026-04-12T21:44:38.198Z" },
+    { url = "https://files.pythonhosted.org/packages/c6/39/74839641e64b99d87da55af0fc472854d42b46e2183b9e2a67fe1bb2a512/msgspec-0.21.1-cp314-cp314t-macosx_10_15_x86_64.whl", hash = "sha256:15f523d51c00ebad412213bfe9f06f0a50ec2b93e0c19e824a2d267cabb48ea2", size = 200171, upload-time = "2026-04-12T21:44:39.414Z" },
+    { url = "https://files.pythonhosted.org/packages/70/9b/ce0cca6d2d87fcd4b6ff97600790494e64f26a2c55d61507cd2755c16193/msgspec-0.21.1-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:4e47390360583ba3d5c6cb44cf0a9f61b0a06a899d3c2c00627cedebb2e2884b", size = 192879, upload-time = "2026-04-12T21:44:40.882Z" },
+    { url = "https://files.pythonhosted.org/packages/a7/08/673a7bb05e5702dc787ddd3011195b509f9867927970da59052211929987/msgspec-0.21.1-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:f60800e6299b798142dc40b0644da77ceac5ea0568be58228417eae14135c847", size = 226281, upload-time = "2026-04-12T21:44:42.181Z" },
+    { url = "https://files.pythonhosted.org/packages/7d/45/86508cf57283e9070b3c447e3ab25b792a7a0855a3ea4e0c6d111ac34c97/msgspec-0.21.1-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:5f8e9dfcd98419cf7568808470c4317a3fb30bef0e3715b568730a2b272a20d7", size = 229863, upload-time = "2026-04-12T21:44:43.442Z" },
+    { url = "https://files.pythonhosted.org/packages/2c/62/e7c9367cd08d590559faacd711edbae36840342843e669440363f33c7d36/msgspec-0.21.1-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:92d89dfad13bd1ea640dc3e37e724ed380da1030b272bdf5ecafb983c3ad7c75", size = 230445, upload-time = "2026-04-12T21:44:44.806Z" },
+    { url = "https://files.pythonhosted.org/packages/42/b4/c0f54632103846b658a10930025f4de41c8724b5e4805a5f3b395586cb7e/msgspec-0.21.1-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:0d03867786e5d7ba25d666df4b11320c27170f4aeafcb8e3a8b0a50a4fb742ca", size = 231822, upload-time = "2026-04-12T21:44:46.343Z" },
+    { url = "https://files.pythonhosted.org/packages/ea/1d/0d85cc79d0ccf5508e9c846cc66552a6a16bf92abd1dbd8362617f7b35cd/msgspec-0.21.1-cp314-cp314t-win_amd64.whl", hash = "sha256:740fbf1c9d59992ca3537d6fbe9ebbf9eaf726a65fbf31448e0ecbc710697a63", size = 206650, upload-time = "2026-04-12T21:44:47.601Z" },
+    { url = "https://files.pythonhosted.org/packages/90/91/56c5d560f20e6c20e9e4f55bd0e458f7f162aa689ee350346c04c48eac0b/msgspec-0.21.1-cp314-cp314t-win_arm64.whl", hash = "sha256:0d2cc73df6058d811a126ac3a8ad63a4dfa210c82f9cf5a004802eaf4712de90", size = 183149, upload-time = "2026-04-12T21:44:48.833Z" },
+]
+
+[[package]]
+name = "numpy"
+version = "2.5.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/22/fd/89965aa4ac08c74998539fcbf24fa3540f3e15237fbeb6bcf9c908f4aade/numpy-2.5.1.tar.gz", hash = "sha256:a48a113e6afea91f5608793bafa7ef2ad481fefbda87ec5069f483de61cb9fa3", size = 20755553, upload-time = "2026-07-04T17:08:00.933Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/62/7b/14687aa674250e5e546f616f486b0d56d3631cd5b2415739141ce40bdcea/numpy-2.5.1-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:2c889b56fe48b1018f764b0eec8df59ab654e9148aa91faa12596043500de277", size = 16801574, upload-time = "2026-07-04T17:06:12.423Z" },
+    { url = "https://files.pythonhosted.org/packages/e1/19/cc5bb2a3f2913d27d6dbb2c78d25921fabaedc6741d4a5a615a11f3c5bf3/numpy-2.5.1-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:ab451b59c5643c570974c43aef780703ef1d3b4965d2be07afd530615a9358d1", size = 11772250, upload-time = "2026-07-04T17:06:15.726Z" },
+    { url = "https://files.pythonhosted.org/packages/42/77/fdf34a71dd30f54979b18603bee915e0aaf825b07afe79acd60b04b691e2/numpy-2.5.1-cp312-cp312-macosx_14_0_arm64.whl", hash = "sha256:78798bd5b9ad744056af8efa90e3b9ddaa53272a0848a483084a1cc0a13b2dc0", size = 5331516, upload-time = "2026-07-04T17:06:17.913Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/e2/eb7efa015b4cce41e2517bf182a7fce0d7d5b9d9ed76a29bfa0f4fe4505c/numpy-2.5.1-cp312-cp312-macosx_14_0_x86_64.whl", hash = "sha256:2ae0ca40bcb22d6ba59c1dfd5446f49940b0f2d821fde133f10dda11f816b84e", size = 6664863, upload-time = "2026-07-04T17:06:20.02Z" },
+    { url = "https://files.pythonhosted.org/packages/a9/4b/a2b32dd94ee9ffbeecb28152240042a3949db33b1c834d44090b80e1b3b8/numpy-2.5.1-cp312-cp312-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:61ac47e772e6b8ea489e1d2f441a34c5c3ac17327e7ce294cbdf535795ad4e75", size = 15167977, upload-time = "2026-07-04T17:06:21.621Z" },
+    { url = "https://files.pythonhosted.org/packages/b8/a9/6e73d68500f80773f65f0654ea932019d6694329a0eb0ed0533de38df376/numpy-2.5.1-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:59fda5e192b570217ec2580c96f00e9a7e12ef6866a900eb089b62c1a32545ca", size = 16672469, upload-time = "2026-07-04T17:06:24.064Z" },
+    { url = "https://files.pythonhosted.org/packages/24/7d/ad3e59015135f5261c95fd4cafeff159c955febd83a99a1d9250c4233815/numpy-2.5.1-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:f7119ebff1a9829e9f431a4f9d28e703023bb6b9fe7c8f724467dbfc27c94ab3", size = 16527531, upload-time = "2026-07-04T17:06:26.69Z" },
+    { url = "https://files.pythonhosted.org/packages/83/d0/a39b2fbcde9cb17a1dac678f254b33a6336298af9df338824c685425d5e8/numpy-2.5.1-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:e824c2acf8862052246be5a44c15da1777940c60d010dd2aab897824d9c430f9", size = 18431940, upload-time = "2026-07-04T17:06:29.521Z" },
+    { url = "https://files.pythonhosted.org/packages/04/12/cff070947791c1ed425ff76413189adbdc2fbe215eba7ce7fa454a03c7f8/numpy-2.5.1-cp312-cp312-win32.whl", hash = "sha256:08d60c810432eb83360958dea0999ac4cfb94531ea8efcbf0b7f277c2068aeb2", size = 6066764, upload-time = "2026-07-04T17:06:32.571Z" },
+    { url = "https://files.pythonhosted.org/packages/65/66/53f31807a48a750f9d748da273bc3fcedd12b27ff1f3e373bfec55ef2dc0/numpy-2.5.1-cp312-cp312-win_amd64.whl", hash = "sha256:f7d60026c0bdb1380e83bfa7a0419c4577ee4b9a08880afcb6dadeb74c649fa2", size = 12430966, upload-time = "2026-07-04T17:06:34.926Z" },
+    { url = "https://files.pythonhosted.org/packages/2b/2a/d1a88066b1c14186f5d3c0d18c94f17b064511982bab0578d49ee9d43c29/numpy-2.5.1-cp312-cp312-win_arm64.whl", hash = "sha256:17a25e09640602e10bc8de0e6fa2b3fd68eedd84ba6d7842dc8f32f9ab87bd0b", size = 10350488, upload-time = "2026-07-04T17:06:37.785Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/07/ec2a3f0c91761581d4b7104a740791800025983f9a4dc4e73f91a99aeac4/numpy-2.5.1-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:0bfebd8695f9863592fe744be833a258120b14a9f39da255e8aa8fade2c0ddd1", size = 16796419, upload-time = "2026-07-04T17:06:40.37Z" },
+    { url = "https://files.pythonhosted.org/packages/ab/ab/ddb499fc4f8780354395face5b65c7fd107bcd6e1d667a5f07d046956f6f/numpy-2.5.1-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:30b44a6b53a7ae63c54c089a8726e5563ed302716c5b7ccc85afade40b0e7ff6", size = 11765832, upload-time = "2026-07-04T17:06:42.768Z" },
+    { url = "https://files.pythonhosted.org/packages/88/b3/3c28c558a09fc72100c646dac6d2fce8e834c471b0edca01a29996706117/numpy-2.5.1-cp313-cp313-macosx_14_0_arm64.whl", hash = "sha256:6165343f81b56ef8f514f396989e529b61d9dc709b99421b07e9f3e698e2287d", size = 5325143, upload-time = "2026-07-04T17:06:45.466Z" },
+    { url = "https://files.pythonhosted.org/packages/5e/0e/ce19b985bb15c596f4f05954e76cccc77c845083b3b8f938a6c68e523128/numpy-2.5.1-cp313-cp313-macosx_14_0_x86_64.whl", hash = "sha256:4939237038ada79308dda3204ac6462df056b5672b2e25db1149cf873668b3e1", size = 6659749, upload-time = "2026-07-04T17:06:47.288Z" },
+    { url = "https://files.pythonhosted.org/packages/2e/20/1ee6614d64332a1bba6411f38e68cb79eec1b2459e20a623777c5c5492a2/numpy-2.5.1-cp313-cp313-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:1c6759f538fb912fc46de0a6b1758ccf7b57bc7c7ebebc23974fdac3de8db0cd", size = 15164716, upload-time = "2026-07-04T17:06:49.494Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/a7/2bcd3fdbb87804755c35b729bf8709d62025c5f4cfd7d5b2415997097515/numpy-2.5.1-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:9726558e8db4a5bf7929a70ae50f63abda4daf0efe810e3bfbab95976f75fc1a", size = 16661440, upload-time = "2026-07-04T17:06:52.061Z" },
+    { url = "https://files.pythonhosted.org/packages/fc/d7/a41e3310c886fe457d36e670bbf24fae411aca8a7b6ad92a32afd924077c/numpy-2.5.1-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:3935f3b419b244a02732676fa5317a9193cc596a4c0646db07e5b421229ac9f7", size = 16526305, upload-time = "2026-07-04T17:06:54.605Z" },
+    { url = "https://files.pythonhosted.org/packages/53/75/4333a9a707c1edd3a4e1a0c58eca52c0f31e55089fa80db02b5565b24df7/numpy-2.5.1-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:dc932a65ded7ce9013d120845a2514dcccb1a67bfc8deb8d37633762951904a6", size = 18423008, upload-time = "2026-07-04T17:06:57.54Z" },
+    { url = "https://files.pythonhosted.org/packages/ee/90/e314a32b1c11a2ffe818ddad3a57b50b4b6e1b6c487192eb50cdef0415d0/numpy-2.5.1-cp313-cp313-win32.whl", hash = "sha256:4b4ff1608417eb7a59da7b967bbb798cacfe071d2caf526a24281cd562072ed9", size = 6063885, upload-time = "2026-07-04T17:07:00.14Z" },
+    { url = "https://files.pythonhosted.org/packages/10/70/800b3fca480af32df9e8ea9f3d4a0c8feb4b32d7f195d174eabbda4829ad/numpy-2.5.1-cp313-cp313-win_amd64.whl", hash = "sha256:6c3fe51bc6a16453d452997053454f309e8e0ed7b42d6b361ce4ac8c32913d74", size = 12425674, upload-time = "2026-07-04T17:07:02.387Z" },
+    { url = "https://files.pythonhosted.org/packages/8b/0b/196350c122f50f6ca56846f2d71efd5e0d24b7b2e07355e019b2e2c7a11e/numpy-2.5.1-cp313-cp313-win_arm64.whl", hash = "sha256:f7feb014281029e628ba2d5a007407443b06e418b6fe451d1e2adcbc8eba0107", size = 10350256, upload-time = "2026-07-04T17:07:04.878Z" },
+    { url = "https://files.pythonhosted.org/packages/db/f4/731b6085a83faf6ca843394cbd5e217280c214399f7e8b21b9f552af0ae2/numpy-2.5.1-cp314-cp314-macosx_10_15_x86_64.whl", hash = "sha256:7c786fe9a5bbe360022e584c5a34cf6b54265c71bd7ec8ac3d8fec38968071f8", size = 16795063, upload-time = "2026-07-04T17:07:07.374Z" },
+    { url = "https://files.pythonhosted.org/packages/bf/64/0e215f2048dd11a55bb989ed41b3585ef57452404e638d703a211a3e4157/numpy-2.5.1-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:32985c896d897419ef8da6917872d80b78ad0ea26d85b23245c7366ffde76d75", size = 11776652, upload-time = "2026-07-04T17:07:09.907Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/59/2b844c7a6e9deff69b404a66221e1542937734f65d5e6e39411876053862/numpy-2.5.1-cp314-cp314-macosx_14_0_arm64.whl", hash = "sha256:efd736408cc97c79b9e6917338dfc8f06013b2274f992e96b1d9a81a71e2a2c2", size = 5335944, upload-time = "2026-07-04T17:07:12.227Z" },
+    { url = "https://files.pythonhosted.org/packages/86/51/9bf7cb2cabcebc9e017e4ec7e6322b378317a542c08b4cb68479c1efc716/numpy-2.5.1-cp314-cp314-macosx_14_0_x86_64.whl", hash = "sha256:ab84dc6b074fa881cae55bea94cc4f68e285181ba7f32497bf7dee6b1496165b", size = 6656266, upload-time = "2026-07-04T17:07:14.368Z" },
+    { url = "https://files.pythonhosted.org/packages/83/3e/fb7615b211b82a32f44d5180a6d421b61f84d4fadd578b48ba4ac34e189f/numpy-2.5.1-cp314-cp314-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:caf3e317d33d60c37986b452613f4ab51246d0691350c03d0cb4a898627f4a95", size = 15179720, upload-time = "2026-07-04T17:07:16.272Z" },
+    { url = "https://files.pythonhosted.org/packages/41/5f/0f992cb24560673496c5d68de61913b57166ce530ffda07c1f280e0cc464/numpy-2.5.1-cp314-cp314-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:54ad769f17bc2d833b620851989f62054fb9ab93c969d9e1dc3c8e3d56beea21", size = 16664835, upload-time = "2026-07-04T17:07:19.021Z" },
+    { url = "https://files.pythonhosted.org/packages/a2/2f/97d6475ee91afe2587797d09446f9d3e475ad4cb681662d824809327b75a/numpy-2.5.1-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:c12afb53450fa976d4c681c50a7423729a4c51c0465ed9f32b8a9cabbc472373", size = 16539135, upload-time = "2026-07-04T17:07:22.015Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/5b/4db81e4ba0be7e2776b1de68c82aa862c7f8ec27e1b4927d4ae075e20678/numpy-2.5.1-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:e8c11c405efc5ff6816d5983c96cdfa215bab3428961243af3ff59b228490438", size = 18426684, upload-time = "2026-07-04T17:07:24.941Z" },
+    { url = "https://files.pythonhosted.org/packages/1f/64/c0ba2d90724d450279a7df8f32057241070250a26a7e2b5337d77347f481/numpy-2.5.1-cp314-cp314-win32.whl", hash = "sha256:f2479a47f8d5932d1718168a681ad6e536a9df484c83cfcf9de365e164537ace", size = 6116103, upload-time = "2026-07-04T17:07:27.622Z" },
+    { url = "https://files.pythonhosted.org/packages/c1/1a/837f9ed7405adcd7a40538792eb169eddd8fa5630c16a1ef49dae71a30f4/numpy-2.5.1-cp314-cp314-win_amd64.whl", hash = "sha256:24d0eb82c0541d3415a33425db64ae439dffccd7b4dbcb30e7c35120205c506a", size = 12562177, upload-time = "2026-07-04T17:07:29.887Z" },
+    { url = "https://files.pythonhosted.org/packages/22/ed/49707938b6dd0a78a9178dd93227dc89e4c11af47f5c798d70366e8d0483/numpy-2.5.1-cp314-cp314-win_arm64.whl", hash = "sha256:5a4c988b38d261deeeaad9954e3deb091ad905c94e8bb6708654ef1d97f286b0", size = 10627739, upload-time = "2026-07-04T17:07:32.568Z" },
+    { url = "https://files.pythonhosted.org/packages/a6/c7/bb4b882cfe7f299cbc8b66e42e7dd78cf9d14e40f9469fc5e3db7e15b3bd/numpy-2.5.1-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:a33276be12fa045805f477f22482088b66bb758ffbe89a9d21457de863a32e22", size = 11894709, upload-time = "2026-07-04T17:07:34.941Z" },
+    { url = "https://files.pythonhosted.org/packages/40/3f/5af7f4a7f6224aef48017aa82bb6174c7a659d724be0c75017b7e64a55b4/numpy-2.5.1-cp314-cp314t-macosx_14_0_arm64.whl", hash = "sha256:f089d7b00756190aacf1f5d34bdf38c3c430ac82b4f868f8cede73380460fce7", size = 5453810, upload-time = "2026-07-04T17:07:37.495Z" },
+    { url = "https://files.pythonhosted.org/packages/20/c9/3474309bc94d634d3f9c3eddf03250ecb8c22cd948ef16fef69a77cc5d7b/numpy-2.5.1-cp314-cp314t-macosx_14_0_x86_64.whl", hash = "sha256:09e9bfd8d2cf479c7d174804fb3811c53a8e9f20a37444008606b57d6b7a826d", size = 6761189, upload-time = "2026-07-04T17:07:39.563Z" },
+    { url = "https://files.pythonhosted.org/packages/90/8a/558ae39fdd55d7e7f7fef9a84a6e964ac6b23edbd2a07e52bb084500507d/numpy-2.5.1-cp314-cp314t-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:e68d8dd1e7eba712948f2053a29ec86917bc70ba1358df869d9f06649ef9cf09", size = 15225039, upload-time = "2026-07-04T17:07:41.682Z" },
+    { url = "https://files.pythonhosted.org/packages/63/27/ca7392b2d030277bdf0273e7d23255b3ee57d57a7c170a6f4fb3981e1e5d/numpy-2.5.1-cp314-cp314t-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:99d5095fa265a0c4152e7bb12759e14381ef5496152f1ce58f44bdf55c44beb4", size = 16701306, upload-time = "2026-07-04T17:07:44.611Z" },
+    { url = "https://files.pythonhosted.org/packages/02/42/03d53ae7996c44d4374a8262e9dc41671fd56cbb98f7d47ef85cf5da4c6b/numpy-2.5.1-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:ab87a91b3cc3382b8956095bd8f95e00cf679bb81554339be1a2ba404a1473c1", size = 16589955, upload-time = "2026-07-04T17:07:47.694Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/15/6c1784ae469640e65db111e9a34b3d0f14d91e8a38b9ce34810ced370dbb/numpy-2.5.1-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:224ca51130ef7da85bea2191625181cb4f337f9cb64b471f10c1a12aa8b60077", size = 18464252, upload-time = "2026-07-04T17:07:50.684Z" },
+    { url = "https://files.pythonhosted.org/packages/94/a8/f98e50356cf167df656c526c2dfeec2d7dde182f2a3da4b458a5938e2776/numpy-2.5.1-cp314-cp314t-win32.whl", hash = "sha256:6eab239876581b2b3c5a242281b6007bbdbcd1c7085d7709bb57c5929b11e6bf", size = 6263298, upload-time = "2026-07-04T17:07:53.445Z" },
+    { url = "https://files.pythonhosted.org/packages/72/ac/96ae880cdecad0b3275d9359fcec72667b49a4863c9f12942e43679dda02/numpy-2.5.1-cp314-cp314t-win_amd64.whl", hash = "sha256:83ce9c80d5b521b0d77ddcbe5447c218d247929b6cc056ca5351342accfff0af", size = 12748623, upload-time = "2026-07-04T17:07:55.384Z" },
+    { url = "https://files.pythonhosted.org/packages/a1/5a/4d2b1601df3602dba7a14f3348ba9bfe94a18adb428e693df6154c293831/numpy-2.5.1-cp314-cp314t-win_arm64.whl", hash = "sha256:5a6db61f9aaa57e369905c67d852045d3c4f7126405b29d09b19dec118e9c9cb", size = 10697674, upload-time = "2026-07-04T17:07:58.506Z" },
+]
+
+[[package]]
+name = "openai"
+version = "2.48.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anyio" },
+    { name = "distro" },
+    { name = "httpx" },
+    { name = "jiter" },
+    { name = "pydantic" },
+    { name = "sniffio" },
+    { name = "tqdm" },
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/2a/ae/d4d1835488c0350424009dac5095b9a3e173bee12fd2e421ee27e2142c42/openai-2.48.0.tar.gz", hash = "sha256:231b1e7661dda14574986c2f71451e9d584b7fe69e0ee6480e12ed090b48fc16", size = 1093427, upload-time = "2026-07-23T20:15:50.402Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/b5/2a/dcb891114e303c4379d4a498f10222e33eee540bcef4e1e493bd0af2b242/openai-2.48.0-py3-none-any.whl", hash = "sha256:c98df30aaaf93c51979f64d3e7c5b76464f8be0173368266229eb8fe6bd30f2c", size = 1648520, upload-time = "2026-07-23T20:15:48.052Z" },
+]
+
+[[package]]
+name = "openpyxl"
+version = "3.1.5"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "et-xmlfile" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/3d/f9/88d94a75de065ea32619465d2f77b29a0469500e99012523b91cc4141cd1/openpyxl-3.1.5.tar.gz", hash = "sha256:cf0e3cf56142039133628b5acffe8ef0c12bc902d2aadd3e0fe5878dc08d1050", size = 186464, upload-time = "2024-06-28T14:03:44.161Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/c0/da/977ded879c29cbd04de313843e76868e6e13408a94ed6b987245dc7c8506/openpyxl-3.1.5-py2.py3-none-any.whl", hash = "sha256:5282c12b107bffeef825f4617dc029afaf41d0ea60823bbb665ef3079dc79de2", size = 250910, upload-time = "2024-06-28T14:03:41.161Z" },
+]
+
+[[package]]
+name = "orjson"
+version = "3.11.9"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/7e/0c/964746fcafbd16f8ff53219ad9f6b412b34f345c75f384ad434ceaadb538/orjson-3.11.9.tar.gz", hash = "sha256:4fef17e1f8722c11587a6ef18e35902450221da0028e65dbaaa543619e68e48f", size = 5599163, upload-time = "2026-05-06T15:11:08.309Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/16/6d/11867a3ffa3a3608d84a4de51ef4dd0896d6b5cc9132fbe1daf593e677bc/orjson-3.11.9-cp312-cp312-macosx_10_15_x86_64.macosx_11_0_arm64.macosx_10_15_universal2.whl", hash = "sha256:9ef6fe90aadef185c7b128859f40beb24720b4ecea95379fc9000931179c3a49", size = 228515, upload-time = "2026-05-06T15:09:57.265Z" },
+    { url = "https://files.pythonhosted.org/packages/24/75/05912954c8b288f34fcf5cd4b9b071cb4f6e77b9961e175e56ebb258089f/orjson-3.11.9-cp312-cp312-macosx_15_0_arm64.whl", hash = "sha256:e5c9b8f28e726e97d97696c826bc7bea5d71cecd63576dba92924a32c1961291", size = 128409, upload-time = "2026-05-06T15:09:59.063Z" },
+    { url = "https://files.pythonhosted.org/packages/ab/86/1c3a47df3bc8191ea9ac51603bbb872a95167a364320c269f2557911f406/orjson-3.11.9-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:26a473dbb4162108b27901492546f83c76fdcea3d0eadff00ae7a07e18dcce09", size = 132106, upload-time = "2026-05-06T15:10:00.798Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/cf/b33b5f3e695ae7d63feef9d915c37cc3b8f465493dcd4f8e0b4c697a2366/orjson-3.11.9-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:011382e2a60fda9d46f1cdee31068cfc52ffe952b587d683ec0463002802a0f4", size = 127864, upload-time = "2026-05-06T15:10:02.15Z" },
+    { url = "https://files.pythonhosted.org/packages/31/6a/6cf69385a58208024fcb8c014e2141b8ce838aba6492b589f8acfff97fab/orjson-3.11.9-cp312-cp312-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:c2d3dc759490128c5c1711a53eeaa8ee1d437fd0038ffd2b6008abf46db3f882", size = 135213, upload-time = "2026-05-06T15:10:03.515Z" },
+    { url = "https://files.pythonhosted.org/packages/e8/f8/0b1bd3e8f2efcdd376af5c8cfd79eaf13f018080c0089c80ebd724e3c7fb/orjson-3.11.9-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:d8ea516b3726d190e1b4297e6f4e7a8650347ae053868a18163b4dd3641d1fff", size = 145994, upload-time = "2026-05-06T15:10:05.083Z" },
+    { url = "https://files.pythonhosted.org/packages/f3/59/dab79f61044c529d2c81aecdc589b1f833a1c8dec11ba3b1c2498a02ca7e/orjson-3.11.9-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:380cdce7ba24989af81d0a7013d0aaec5d0e2a21734c0e2681b1bc4f141957fe", size = 132744, upload-time = "2026-05-06T15:10:06.853Z" },
+    { url = "https://files.pythonhosted.org/packages/0e/a4/82b7a2fe5d8a67a59ed831b24d59a3d46ea7d207b66e1602d376541d94a6/orjson-3.11.9-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:be4fa4f0af7fa18951f7ab3fc2148e223af211bf03f59e1c6034ec3f97f21d61", size = 134014, upload-time = "2026-05-06T15:10:08.213Z" },
+    { url = "https://files.pythonhosted.org/packages/50/c7/375e83a76851b73b2e39f3bcf0e5a19e2b89bad13e5bca97d0b293d27f24/orjson-3.11.9-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:a8f5f8bc7ce7d59f08d9f99fa510c06496164a24cb5f3d34537dbd9ca30132e2", size = 141509, upload-time = "2026-05-06T15:10:09.595Z" },
+    { url = "https://files.pythonhosted.org/packages/7f/7c/49d5d82a3d3097f641f094f552131f1e2723b0b8cb0fa2874ab65ecfffa6/orjson-3.11.9-cp312-cp312-musllinux_1_2_armv7l.whl", hash = "sha256:4d7fde5501b944f83b3e665e1b31343ff6e154b15560a16b7130ea1e594a4206", size = 415127, upload-time = "2026-05-06T15:10:11.049Z" },
+    { url = "https://files.pythonhosted.org/packages/3a/dc/7446c538590d55f455647e5f3c61fc33f7108714e7afcffa6a2a033f8350/orjson-3.11.9-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:cde1a448023ba7d5bb4c01c5afb48894380b5e4956e0627266526587ef4e535f", size = 148025, upload-time = "2026-05-06T15:10:12.842Z" },
+    { url = "https://files.pythonhosted.org/packages/df/e5/4d2d8af06f788329b4f78f8cc3679bb395392fcaa1e4d8d3c33e85308fa4/orjson-3.11.9-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:71e63adb0e1f1ed5d9e168f50a91ceb93ae6420731d222dc7da5c69409aa47aa", size = 136943, upload-time = "2026-05-06T15:10:14.405Z" },
+    { url = "https://files.pythonhosted.org/packages/06/69/850264ccf6d80f6b174620d30a87f65c9b1490aba33fe6b62798e618cad3/orjson-3.11.9-cp312-cp312-win32.whl", hash = "sha256:2d057a602cdd19a0ad680417527c45b6961a095081c0f46fe0e03e304aac6470", size = 131606, upload-time = "2026-05-06T15:10:15.791Z" },
+    { url = "https://files.pythonhosted.org/packages/b9/d5/973a43fc9c55e20f2051e9830997649f669be0cb3ca52192087c0143f118/orjson-3.11.9-cp312-cp312-win_amd64.whl", hash = "sha256:59e403b1cc5a676da8eaf31f6254801b7341b3e29efa85f92b48d272637e77be", size = 127101, upload-time = "2026-05-06T15:10:17.129Z" },
+    { url = "https://files.pythonhosted.org/packages/fe/ae/495470f0e4a18f73fa10b7f6b84b464ec4cc5291c4e0c7c2a6c400bef006/orjson-3.11.9-cp312-cp312-win_arm64.whl", hash = "sha256:9af678d6488357948f1f84c6cd1c1d397c014e1ae2f98ae082a44eb48f602624", size = 126736, upload-time = "2026-05-06T15:10:18.645Z" },
+    { url = "https://files.pythonhosted.org/packages/32/33/93fcc25907235c344ae73122f8a4e01d2d393ef062b4af7d2e2487a32c37/orjson-3.11.9-cp313-cp313-macosx_10_15_x86_64.macosx_11_0_arm64.macosx_10_15_universal2.whl", hash = "sha256:4bab1b2d6141fe7b32ae71dac905666ece4f94936efbfb13d55bb7739a3a6021", size = 228458, upload-time = "2026-05-06T15:10:20.079Z" },
+    { url = "https://files.pythonhosted.org/packages/8f/27/b1e6dadb3c080313c03fdd8067b85e6a0460c7d8d6a1c3984ef77b904e4d/orjson-3.11.9-cp313-cp313-macosx_15_0_arm64.whl", hash = "sha256:844417969855fc7a41be124aafe83dc424592a7f77cd4501900c67307122b92c", size = 128368, upload-time = "2026-05-06T15:10:21.549Z" },
+    { url = "https://files.pythonhosted.org/packages/21/0f/c9ede0bf052f6b4051e64a7d4fa91b725cccf8321a6a786e86eb03519f00/orjson-3.11.9-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:ffe02797b5e9f3a9d8292ddcd289b474ad13e81ad83cd1891a240811f1d2cb81", size = 132070, upload-time = "2026-05-06T15:10:23.371Z" },
+    { url = "https://files.pythonhosted.org/packages/fd/26/d398e28048dc18205bbe812f2c88cb9b40313db2470778e25964796458fe/orjson-3.11.9-cp313-cp313-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:0e4eed3b200023042814d2fc8a5d2e880f13b52e1ed2485e83da4f3962f7dc1a", size = 127892, upload-time = "2026-05-06T15:10:24.714Z" },
+    { url = "https://files.pythonhosted.org/packages/66/60/52b0054c4c700d5aa7fc5b7ca96917400d8f061307778578e67a10e25852/orjson-3.11.9-cp313-cp313-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:8aff7da9952a5ad1cef8e68017724d96c7b9a66e99e91d6252e1b133d67a7b10", size = 135217, upload-time = "2026-05-06T15:10:26.084Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/97/1e3dc2b2a28b7b2528f403d2fc1d79ec5f39af3bc143ab65d3ec26426385/orjson-3.11.9-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:4d4e98d6f3b8afed8bc8cd9718ec0cdf46661826beefb53fe8eafb37f2bf0362", size = 145980, upload-time = "2026-05-06T15:10:28.062Z" },
+    { url = "https://files.pythonhosted.org/packages/fc/39/31fbfe7850f2de32dee7e7e5c09f26d403ab01e440ac96001c6b01ad3c99/orjson-3.11.9-cp313-cp313-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:3a81d52442a7c99b3662333235b3adf96a1715864658b35bb797212be7bddb97", size = 132738, upload-time = "2026-05-06T15:10:29.727Z" },
+    { url = "https://files.pythonhosted.org/packages/a1/08/dca0082dd2a194acb93e5457e73455388e2e2ca464a2672449a9ddbb679d/orjson-3.11.9-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:4e39364e726a8fff737309aff059ff67d8a8c8d5b677be7bb49a8b3e84b7e218", size = 134033, upload-time = "2026-05-06T15:10:31.152Z" },
+    { url = "https://files.pythonhosted.org/packages/11/d4/5bdb0626801230139987385554c5d4c42255218ac906525bf4347f22cd95/orjson-3.11.9-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:4fd66214623f1b17501df9f0543bef0b833979ab5b6ded1e1d123222866aa8c9", size = 141492, upload-time = "2026-05-06T15:10:32.641Z" },
+    { url = "https://files.pythonhosted.org/packages/fa/88/a21fb53b3ede6703aede6dce4710ed4111e5b201cfa6bbff5e544f9d47d7/orjson-3.11.9-cp313-cp313-musllinux_1_2_armv7l.whl", hash = "sha256:8ecc30f10465fa1e0ce13fd01d9e22c316e5053a719a8d915d4545a09a5ff677", size = 415087, upload-time = "2026-05-06T15:10:34.438Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/57/1b30daf70f0d8180e9a73cefbfbdd99e4bf19eb020466502b01fba7e0e50/orjson-3.11.9-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:97db4c94a7db398a5bd636273324f0b3fd58b350bbbac8bb380ceb825a9b40f4", size = 148031, upload-time = "2026-05-06T15:10:36.358Z" },
+    { url = "https://files.pythonhosted.org/packages/04/83/45fbb6d962e260807f99441db9613cee868ceda4baceda59b3720a563f97/orjson-3.11.9-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:9f78cf8fec5bd627f4082b8dfeac7871b43d7f3274904492a43dab39f18a19a0", size = 136915, upload-time = "2026-05-06T15:10:38.013Z" },
+    { url = "https://files.pythonhosted.org/packages/5f/cc/2d10025f9056d376e4127ec05a5808b218d46f035fdc08178a5411b34250/orjson-3.11.9-cp313-cp313-win32.whl", hash = "sha256:d4087e5c0209a0a8efe4de3303c234b9c44d1174161dcd851e8eea07c7560b32", size = 131613, upload-time = "2026-05-06T15:10:39.569Z" },
+    { url = "https://files.pythonhosted.org/packages/67/bd/2775ff28bfe883b9aa1ff348300542eb2ef1ee18d8ae0e3a49846817a865/orjson-3.11.9-cp313-cp313-win_amd64.whl", hash = "sha256:051b102c93b4f634e89f3866b07b9a9a98915ada541f4ec30f177067b2694979", size = 127086, upload-time = "2026-05-06T15:10:41.262Z" },
+    { url = "https://files.pythonhosted.org/packages/91/2b/d26799e580939e32a7da9a39531bc9e58e15ca32ffaa6a8cb3e9bb0d22cd/orjson-3.11.9-cp313-cp313-win_arm64.whl", hash = "sha256:cce9127885941bd28f080cecf1f1d288336b7e0d812c345b08be88b572796254", size = 126696, upload-time = "2026-05-06T15:10:42.651Z" },
+    { url = "https://files.pythonhosted.org/packages/8e/eb/5da01e356015aee6ecfa1187ced87aef51364e306f5e695dd52719bf0e78/orjson-3.11.9-cp314-cp314-macosx_10_15_x86_64.macosx_11_0_arm64.macosx_10_15_universal2.whl", hash = "sha256:b6ef1979adc4bc243523f1a2ba91418030a8e29b0a99cbe7e0e2d6807d4dce6e", size = 228465, upload-time = "2026-05-06T15:10:44.097Z" },
+    { url = "https://files.pythonhosted.org/packages/64/62/3e0e0c14c957133bcd855395c62b55ed4e3b0af23ffea11b032cb1dcbdb1/orjson-3.11.9-cp314-cp314-macosx_15_0_arm64.whl", hash = "sha256:f36b7f32c7c0db4a719f1fc5824db4a9c6f8bd1a354debb91faf26ebf3a4c71e", size = 128364, upload-time = "2026-05-06T15:10:45.839Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/5a/07d8aa117211a8ed7630bda80c8c0b14d04e0f8dcf99bcf49656e4a710eb/orjson-3.11.9-cp314-cp314-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:08f4d8ebb44925c794e535b2bebc507cebf32209df81de22ae285fb0d8d66de0", size = 132063, upload-time = "2026-05-06T15:10:47.267Z" },
+    { url = "https://files.pythonhosted.org/packages/d6/ec/4acaf21483e18aa945be74a474c74b434f284b549f275a0a39b9f98956e9/orjson-3.11.9-cp314-cp314-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:6cc7923789694fd58f001cbcac7e47abc13af4d560ebbfcf3b41a8b1a0748124", size = 122356, upload-time = "2026-05-06T15:10:48.765Z" },
+    { url = "https://files.pythonhosted.org/packages/13/d8/5f0555e7638801323b7a75850f92e7dfa891bc84fe27a1ba4449170d1200/orjson-3.11.9-cp314-cp314-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:ea5c46eb2d3af39e806b986f4b09d5c2706a1f5afde3cbf7544ce6616127173c", size = 129592, upload-time = "2026-05-06T15:10:50.13Z" },
+    { url = "https://files.pythonhosted.org/packages/b6/30/ed9860412a3603ceb3c5955bfd72d28b9d0e7ba6ed81add14f83d7114236/orjson-3.11.9-cp314-cp314-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:f5d89a2ed90731df3be64bab0aa44f78bff39fdc9d71c291f4a8023aa46425b7", size = 140491, upload-time = "2026-05-06T15:10:51.582Z" },
+    { url = "https://files.pythonhosted.org/packages/d0/17/adc514dea7ac7c505527febf884934b815d34f0c7b8693c1a8b39c5c4a57/orjson-3.11.9-cp314-cp314-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:25e4aed0312d292c09f61af25bba34e0b2c88546041472b09088c39a4d828af1", size = 127309, upload-time = "2026-05-06T15:10:53.329Z" },
+    { url = "https://files.pythonhosted.org/packages/76/3e/c0b690253f0b82d86e99949af13533363acfb5432ecb5d53dd5b3bce9c34/orjson-3.11.9-cp314-cp314-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:aaea64f3f467d22e70eeed68bdccb3bc4f83f650446c4a03c59f2cba28a108db", size = 134030, upload-time = "2026-05-06T15:10:54.988Z" },
+    { url = "https://files.pythonhosted.org/packages/c1/7a/bc82a0bb25e9faaf92dc4d9ef002732efc09737706af83e346788641d4a7/orjson-3.11.9-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:a028425d1b440c5d92a6be1e1a020739dfe67ea87d96c6dbe828c1b30041728b", size = 141482, upload-time = "2026-05-06T15:10:56.663Z" },
+    { url = "https://files.pythonhosted.org/packages/01/55/e69188b939f77d5d32a9833745ace31ea5ccae3ab613a1ec185d3cd2c4fb/orjson-3.11.9-cp314-cp314-musllinux_1_2_armv7l.whl", hash = "sha256:5b192c6cf397e4455b11523c5cf2b18ed084c1bbd61b6c0926344d2129481972", size = 415178, upload-time = "2026-05-06T15:10:58.446Z" },
+    { url = "https://files.pythonhosted.org/packages/2e/1a/b8a5a7ac527e80b9cb11d51e3f6689b709279183264b9ec5c7bc680bb8b5/orjson-3.11.9-cp314-cp314-musllinux_1_2_i686.whl", hash = "sha256:ea407d4ccf5891d667d045fecae97a7a1e5e87b3b97f97ae1803c2e741130be0", size = 148089, upload-time = "2026-05-06T15:11:00.441Z" },
+    { url = "https://files.pythonhosted.org/packages/97/4e/00503f64204bf859b37213a63927028f30fb6268cd8677fb0a5ad48155e1/orjson-3.11.9-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:5f63aaf97afd9f6dec5b1a68e1b8da12bfccb4cb9a9a65c3e0b6c847849e7586", size = 136921, upload-time = "2026-05-06T15:11:02.176Z" },
+    { url = "https://files.pythonhosted.org/packages/0d/ba/a23b82a0a8d0ed7bed4e5f5035aae751cad4ff6a1e8d2ecd14d8860f5929/orjson-3.11.9-cp314-cp314-win32.whl", hash = "sha256:e30ab17845bb9fa54ccf67fa4f9f5282652d54faa6d17452f47d0f369d038673", size = 131638, upload-time = "2026-05-06T15:11:03.696Z" },
+    { url = "https://files.pythonhosted.org/packages/f3/c3/0c6798456bade745c75c452342dabacce5798196483e77e643be1f53877d/orjson-3.11.9-cp314-cp314-win_amd64.whl", hash = "sha256:32ef5f4283a3be81913947d19608eacb7c6608026851123790cd9cc8982af34b", size = 127078, upload-time = "2026-05-06T15:11:05.123Z" },
+    { url = "https://files.pythonhosted.org/packages/16/21/5a3f1e8913103b703a436a5664238e5b965ec392b555fe68943ea3691e6b/orjson-3.11.9-cp314-cp314-win_arm64.whl", hash = "sha256:eebdbdeef0094e4f5aefa20dcd4eb2368ab5e7a3b4edea27f1e7b2892e009cf9", size = 126687, upload-time = "2026-05-06T15:11:06.602Z" },
+]
+
+[[package]]
+name = "ormsgpack"
+version = "1.12.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/12/0c/f1761e21486942ab9bb6feaebc610fa074f7c5e496e6962dea5873348077/ormsgpack-1.12.2.tar.gz", hash = "sha256:944a2233640273bee67521795a73cf1e959538e0dfb7ac635505010455e53b33", size = 39031, upload-time = "2026-01-18T20:55:28.023Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/4c/36/16c4b1921c308a92cef3bf6663226ae283395aa0ff6e154f925c32e91ff5/ormsgpack-1.12.2-cp312-cp312-macosx_10_12_x86_64.macosx_11_0_arm64.macosx_10_12_universal2.whl", hash = "sha256:7a29d09b64b9694b588ff2f80e9826bdceb3a2b91523c5beae1fab27d5c940e7", size = 378618, upload-time = "2026-01-18T20:55:50.835Z" },
+    { url = "https://files.pythonhosted.org/packages/c0/68/468de634079615abf66ed13bb5c34ff71da237213f29294363beeeca5306/ormsgpack-1.12.2-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:0b39e629fd2e1c5b2f46f99778450b59454d1f901bc507963168985e79f09c5d", size = 203186, upload-time = "2026-01-18T20:56:11.163Z" },
+    { url = "https://files.pythonhosted.org/packages/73/a9/d756e01961442688b7939bacd87ce13bfad7d26ce24f910f6028178b2cc8/ormsgpack-1.12.2-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:958dcb270d30a7cb633a45ee62b9444433fa571a752d2ca484efdac07480876e", size = 210738, upload-time = "2026-01-18T20:56:09.181Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/ba/795b1036888542c9113269a3f5690ab53dd2258c6fb17676ac4bd44fcf94/ormsgpack-1.12.2-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:58d379d72b6c5e964851c77cfedfb386e474adee4fd39791c2c5d9efb53505cc", size = 212569, upload-time = "2026-01-18T20:56:06.135Z" },
+    { url = "https://files.pythonhosted.org/packages/6c/aa/bff73c57497b9e0cba8837c7e4bcab584b1a6dbc91a5dd5526784a5030c8/ormsgpack-1.12.2-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:8463a3fc5f09832e67bdb0e2fda6d518dc4281b133166146a67f54c08496442e", size = 387166, upload-time = "2026-01-18T20:55:36.738Z" },
+    { url = "https://files.pythonhosted.org/packages/d3/cf/f8283cba44bcb7b14f97b6274d449db276b3a86589bdb363169b51bc12de/ormsgpack-1.12.2-cp312-cp312-musllinux_1_2_armv7l.whl", hash = "sha256:eddffb77eff0bad4e67547d67a130604e7e2dfbb7b0cde0796045be4090f35c6", size = 482498, upload-time = "2026-01-18T20:55:29.626Z" },
+    { url = "https://files.pythonhosted.org/packages/05/be/71e37b852d723dfcbe952ad04178c030df60d6b78eba26bfd14c9a40575e/ormsgpack-1.12.2-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:fcd55e5f6ba0dbce624942adf9f152062135f991a0126064889f68eb850de0dd", size = 425518, upload-time = "2026-01-18T20:55:49.556Z" },
+    { url = "https://files.pythonhosted.org/packages/7a/0c/9803aa883d18c7ef197213cd2cbf73ba76472a11fe100fb7dab2884edf48/ormsgpack-1.12.2-cp312-cp312-win_amd64.whl", hash = "sha256:d024b40828f1dde5654faebd0d824f9cc29ad46891f626272dd5bfd7af2333a4", size = 117462, upload-time = "2026-01-18T20:55:47.726Z" },
+    { url = "https://files.pythonhosted.org/packages/c8/9e/029e898298b2cc662f10d7a15652a53e3b525b1e7f07e21fef8536a09bb8/ormsgpack-1.12.2-cp312-cp312-win_arm64.whl", hash = "sha256:da538c542bac7d1c8f3f2a937863dba36f013108ce63e55745941dda4b75dbb6", size = 111559, upload-time = "2026-01-18T20:55:54.273Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/29/bb0eba3288c0449efbb013e9c6f58aea79cf5cb9ee1921f8865f04c1a9d7/ormsgpack-1.12.2-cp313-cp313-macosx_10_12_x86_64.macosx_11_0_arm64.macosx_10_12_universal2.whl", hash = "sha256:5ea60cb5f210b1cfbad8c002948d73447508e629ec375acb82910e3efa8ff355", size = 378661, upload-time = "2026-01-18T20:55:57.765Z" },
+    { url = "https://files.pythonhosted.org/packages/6e/31/5efa31346affdac489acade2926989e019e8ca98129658a183e3add7af5e/ormsgpack-1.12.2-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:f3601f19afdbea273ed70b06495e5794606a8b690a568d6c996a90d7255e51c1", size = 203194, upload-time = "2026-01-18T20:56:08.252Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/56/d0087278beef833187e0167f8527235ebe6f6ffc2a143e9de12a98b1ce87/ormsgpack-1.12.2-cp313-cp313-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:29a9f17a3dac6054c0dce7925e0f4995c727f7c41859adf9b5572180f640d172", size = 210778, upload-time = "2026-01-18T20:55:17.694Z" },
+    { url = "https://files.pythonhosted.org/packages/1c/a2/072343e1413d9443e5a252a8eb591c2d5b1bffbe5e7bfc78c069361b92eb/ormsgpack-1.12.2-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:39c1bd2092880e413902910388be8715f70b9f15f20779d44e673033a6146f2d", size = 212592, upload-time = "2026-01-18T20:55:32.747Z" },
+    { url = "https://files.pythonhosted.org/packages/a2/8b/a0da3b98a91d41187a63b02dda14267eefc2a74fcb43cc2701066cf1510e/ormsgpack-1.12.2-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:50b7249244382209877deedeee838aef1542f3d0fc28b8fe71ca9d7e1896a0d7", size = 387164, upload-time = "2026-01-18T20:55:40.853Z" },
+    { url = "https://files.pythonhosted.org/packages/19/bb/6d226bc4cf9fc20d8eb1d976d027a3f7c3491e8f08289a2e76abe96a65f3/ormsgpack-1.12.2-cp313-cp313-musllinux_1_2_armv7l.whl", hash = "sha256:5af04800d844451cf102a59c74a841324868d3f1625c296a06cc655c542a6685", size = 482516, upload-time = "2026-01-18T20:55:42.033Z" },
+    { url = "https://files.pythonhosted.org/packages/fb/f1/bb2c7223398543dedb3dbf8bb93aaa737b387de61c5feaad6f908841b782/ormsgpack-1.12.2-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:cec70477d4371cd524534cd16472d8b9cc187e0e3043a8790545a9a9b296c258", size = 425539, upload-time = "2026-01-18T20:55:24.727Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/e8/0fb45f57a2ada1fed374f7494c8cd55e2f88ccd0ab0a669aa3468716bf5f/ormsgpack-1.12.2-cp313-cp313-win_amd64.whl", hash = "sha256:21f4276caca5c03a818041d637e4019bc84f9d6ca8baa5ea03e5cc8bf56140e9", size = 117459, upload-time = "2026-01-18T20:55:56.876Z" },
+    { url = "https://files.pythonhosted.org/packages/7a/d4/0cfeea1e960d550a131001a7f38a5132c7ae3ebde4c82af1f364ccc5d904/ormsgpack-1.12.2-cp313-cp313-win_arm64.whl", hash = "sha256:baca4b6773d20a82e36d6fd25f341064244f9f86a13dead95dd7d7f996f51709", size = 111577, upload-time = "2026-01-18T20:55:43.605Z" },
+    { url = "https://files.pythonhosted.org/packages/94/16/24d18851334be09c25e87f74307c84950f18c324a4d3c0b41dabdbf19c29/ormsgpack-1.12.2-cp314-cp314-macosx_10_12_x86_64.macosx_11_0_arm64.macosx_10_12_universal2.whl", hash = "sha256:bc68dd5915f4acf66ff2010ee47c8906dc1cf07399b16f4089f8c71733f6e36c", size = 378717, upload-time = "2026-01-18T20:55:26.164Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/a2/88b9b56f83adae8032ac6a6fa7f080c65b3baf9b6b64fd3d37bd202991d4/ormsgpack-1.12.2-cp314-cp314-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:46d084427b4132553940070ad95107266656cb646ea9da4975f85cb1a6676553", size = 203183, upload-time = "2026-01-18T20:55:18.815Z" },
+    { url = "https://files.pythonhosted.org/packages/a9/80/43e4555963bf602e5bdc79cbc8debd8b6d5456c00d2504df9775e74b450b/ormsgpack-1.12.2-cp314-cp314-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:c010da16235806cf1d7bc4c96bf286bfa91c686853395a299b3ddb49499a3e13", size = 210814, upload-time = "2026-01-18T20:55:33.973Z" },
+    { url = "https://files.pythonhosted.org/packages/78/e1/7cfbf28de8bca6efe7e525b329c31277d1b64ce08dcba723971c241a9d60/ormsgpack-1.12.2-cp314-cp314-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:18867233df592c997154ff942a6503df274b5ac1765215bceba7a231bea2745d", size = 212634, upload-time = "2026-01-18T20:55:28.634Z" },
+    { url = "https://files.pythonhosted.org/packages/95/f8/30ae5716e88d792a4e879debee195653c26ddd3964c968594ddef0a3cc7e/ormsgpack-1.12.2-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:b009049086ddc6b8f80c76b3955df1aa22a5fbd7673c525cd63bf91f23122ede", size = 387139, upload-time = "2026-01-18T20:56:02.013Z" },
+    { url = "https://files.pythonhosted.org/packages/dc/81/aee5b18a3e3a0e52f718b37ab4b8af6fae0d9d6a65103036a90c2a8ffb5d/ormsgpack-1.12.2-cp314-cp314-musllinux_1_2_armv7l.whl", hash = "sha256:1dcc17d92b6390d4f18f937cf0b99054824a7815818012ddca925d6e01c2e49e", size = 482578, upload-time = "2026-01-18T20:55:35.117Z" },
+    { url = "https://files.pythonhosted.org/packages/bd/17/71c9ba472d5d45f7546317f467a5fc941929cd68fb32796ca3d13dcbaec2/ormsgpack-1.12.2-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:f04b5e896d510b07c0ad733d7fce2d44b260c5e6c402d272128f8941984e4285", size = 425539, upload-time = "2026-01-18T20:56:04.009Z" },
+    { url = "https://files.pythonhosted.org/packages/2e/a6/ac99cd7fe77e822fed5250ff4b86fa66dd4238937dd178d2299f10b69816/ormsgpack-1.12.2-cp314-cp314-win_amd64.whl", hash = "sha256:ae3aba7eed4ca7cb79fd3436eddd29140f17ea254b91604aa1eb19bfcedb990f", size = 117493, upload-time = "2026-01-18T20:56:07.343Z" },
+    { url = "https://files.pythonhosted.org/packages/3a/67/339872846a1ae4592535385a1c1f93614138566d7af094200c9c3b45d1e5/ormsgpack-1.12.2-cp314-cp314-win_arm64.whl", hash = "sha256:118576ea6006893aea811b17429bfc561b4778fad393f5f538c84af70b01260c", size = 111579, upload-time = "2026-01-18T20:55:21.161Z" },
+    { url = "https://files.pythonhosted.org/packages/49/c2/6feb972dc87285ad381749d3882d8aecbde9f6ecf908dd717d33d66df095/ormsgpack-1.12.2-cp314-cp314t-macosx_10_12_x86_64.macosx_11_0_arm64.macosx_10_12_universal2.whl", hash = "sha256:7121b3d355d3858781dc40dafe25a32ff8a8242b9d80c692fd548a4b1f7fd3c8", size = 378721, upload-time = "2026-01-18T20:55:52.12Z" },
+    { url = "https://files.pythonhosted.org/packages/a3/9a/900a6b9b413e0f8a471cf07830f9cf65939af039a362204b36bd5b581d8b/ormsgpack-1.12.2-cp314-cp314t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:4ee766d2e78251b7a63daf1cddfac36a73562d3ddef68cacfb41b2af64698033", size = 203170, upload-time = "2026-01-18T20:55:44.469Z" },
+    { url = "https://files.pythonhosted.org/packages/87/4c/27a95466354606b256f24fad464d7c97ab62bce6cc529dd4673e1179b8fb/ormsgpack-1.12.2-cp314-cp314t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:292410a7d23de9b40444636b9b8f1e4e4b814af7f1ef476e44887e52a123f09d", size = 212816, upload-time = "2026-01-18T20:55:23.501Z" },
+    { url = "https://files.pythonhosted.org/packages/73/cd/29cee6007bddf7a834e6cd6f536754c0535fcb939d384f0f37a38b1cddb8/ormsgpack-1.12.2-cp314-cp314t-win_amd64.whl", hash = "sha256:837dd316584485b72ef451d08dd3e96c4a11d12e4963aedb40e08f89685d8ec2", size = 117232, upload-time = "2026-01-18T20:55:45.448Z" },
+]
+
+[[package]]
+name = "outcome"
+version = "1.3.0.post0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "attrs" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/98/df/77698abfac98571e65ffeb0c1fba8ffd692ab8458d617a0eed7d9a8d38f2/outcome-1.3.0.post0.tar.gz", hash = "sha256:9dcf02e65f2971b80047b377468e72a268e15c0af3cf1238e6ff14f7f91143b8", size = 21060, upload-time = "2023-10-26T04:26:04.361Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/55/8b/5ab7257531a5d830fc8000c476e63c935488d74609b50f9384a643ec0a62/outcome-1.3.0.post0-py2.py3-none-any.whl", hash = "sha256:e771c5ce06d1415e356078d3bdd68523f284b4ce5419828922b6871e65eda82b", size = 10692, upload-time = "2023-10-26T04:26:02.532Z" },
+]
+
+[[package]]
+name = "packaging"
+version = "26.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/d7/f1/e7a6dd94a8d4a5626c03e4e99c87f241ba9e350cd9e6d75123f992427270/packaging-26.2.tar.gz", hash = "sha256:ff452ff5a3e828ce110190feff1178bb1f2ea2281fa2075aadb987c2fb221661", size = 228134, upload-time = "2026-04-24T20:15:23.917Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/df/b2/87e62e8c3e2f4b32e5fe99e0b86d576da1312593b39f47d8ceef365e95ed/packaging-26.2-py3-none-any.whl", hash = "sha256:5fc45236b9446107ff2415ce77c807cee2862cb6fac22b8a73826d0693b0980e", size = 100195, upload-time = "2026-04-24T20:15:22.081Z" },
+]
+
+[[package]]
+name = "pandas"
+version = "3.0.5"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "numpy" },
+    { name = "python-dateutil" },
+    { name = "tzdata", marker = "sys_platform == 'emscripten' or sys_platform == 'win32'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/be/4f/5f3422a2afec5ffc46308b79e53291365a93748b498ac2e58bead0197916/pandas-3.0.5.tar.gz", hash = "sha256:dca3734d6ab7c906e6730f0788b0a1dbb9f2467731f9711f77995c8e9d62d712", size = 4658219, upload-time = "2026-07-22T22:19:28.819Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/1c/54/1dc810ea558d1320b597aa140a514f2fdf1d2ea09c38cf556f13ea712ec9/pandas-3.0.5-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:fa290c16964d4963fbfbc358928239cf3bd755b20e988ce944877def2f44471d", size = 10411717, upload-time = "2026-07-22T22:18:08.307Z" },
+    { url = "https://files.pythonhosted.org/packages/68/56/fbe81c09195924d8b7b8d4461a20458fe80a6a5ed6b24f0314da684277e1/pandas-3.0.5-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:c2e26bb46934b8a2ca0c3de1d3d606fc5f6746584791b2db264d58cf370e08dc", size = 9957095, upload-time = "2026-07-22T22:18:10.6Z" },
+    { url = "https://files.pythonhosted.org/packages/e0/51/fac252f4a913ed5eabf3c11b880a9e8d5a6c10f0b2129d0462212d238b4d/pandas-3.0.5-cp312-cp312-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:73fa87b08a7ef706f8aafda39ddaccf2a99047bea62d8c88a0361bcafb2237bc", size = 10485458, upload-time = "2026-07-22T22:18:12.834Z" },
+    { url = "https://files.pythonhosted.org/packages/12/98/e976540c1addf70442be7842a18cf70884a964abbf69442504f4d2939989/pandas-3.0.5-cp312-cp312-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:d373ce03ffd84010ed9839fa73672a9c8256990532e158440c0085db7d914b34", size = 10998091, upload-time = "2026-07-22T22:18:15.209Z" },
+    { url = "https://files.pythonhosted.org/packages/a4/8c/1f29b5be8d3fc47dd7567eb167fabba2085879b31e0287ce7cba6d3d2ff4/pandas-3.0.5-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:2a29c53d85ea98c5e792c59ef82ee9fbe6ca902c0d0adb6b23f45ef894cd7bf6", size = 11499501, upload-time = "2026-07-22T22:18:17.689Z" },
+    { url = "https://files.pythonhosted.org/packages/9d/e2/bd9c98ad2df7b38bde002adde4cdf353519da51881634323b126c55997f9/pandas-3.0.5-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:a5ad3b02ed6bc7d7ae9b70804b2c6aa31827489d150f8e623ce82491b82085d7", size = 12060559, upload-time = "2026-07-22T22:18:20.147Z" },
+    { url = "https://files.pythonhosted.org/packages/f3/9a/ffbd852d58bd74a617fe2f8ee6a58a96982271ce41cf981eab22190b4a4b/pandas-3.0.5-cp312-cp312-pyemscripten_2024_0_wasm32.whl", hash = "sha256:b2acb4650527eec6822c3dadb2b771277b65e7dae7a267d4bccf65fd1bb3fbce", size = 7197652, upload-time = "2026-07-22T22:18:22.502Z" },
+    { url = "https://files.pythonhosted.org/packages/70/b5/d2d3e9ae73362ba4229651b0ee1455cf78073a1ce585f6ff693782ce263e/pandas-3.0.5-cp312-cp312-win_amd64.whl", hash = "sha256:80a611068e8a3ac23f7398c6c14eb46dc974e5cc9997f653e2dcfd1da74edd41", size = 9831691, upload-time = "2026-07-22T22:18:24.534Z" },
+    { url = "https://files.pythonhosted.org/packages/52/51/dea1e89d6a6796b9c43f85a09b484ee03edb8a4c4842e73e200a8c11301c/pandas-3.0.5-cp312-cp312-win_arm64.whl", hash = "sha256:25ff585b972a18ef1fe9ffa3ac6544d9950508aa76832e5147640b6022821e49", size = 9105796, upload-time = "2026-07-22T22:18:27.064Z" },
+    { url = "https://files.pythonhosted.org/packages/bf/09/7b95c4a0025227d6f118c4039b423412ac6a982db02864166185d812fbc7/pandas-3.0.5-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:c1c05a767fe8e5b4fe9e1c29806829c582052eaedb9120a3da83ba3f69e24a5b", size = 10385742, upload-time = "2026-07-22T22:18:29.346Z" },
+    { url = "https://files.pythonhosted.org/packages/8d/0c/dc78fd8c4da477b4b5e8ad37295af352190d21ef63a9ee1bc071753074cc/pandas-3.0.5-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:b86765f268b56f7e665b93bce9d5df69dee7f99e595cf8fb839483ab315942a3", size = 9932067, upload-time = "2026-07-22T22:18:31.833Z" },
+    { url = "https://files.pythonhosted.org/packages/3e/71/3592c055cf44df9808550f9368ceda80ff2b224d355ef73fe251dcda1802/pandas-3.0.5-cp313-cp313-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:c597ecf5616b5c420372c1d4d4c00dbbfba7398bea857dcc984347e1ea48417b", size = 10466756, upload-time = "2026-07-22T22:18:34.195Z" },
+    { url = "https://files.pythonhosted.org/packages/e3/70/4363150359f95b4cb4bcbb34ca23572bb5495749a621a8f3d5a1ddfd293c/pandas-3.0.5-cp313-cp313-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:4b11c36e218331d0387cbe3a0a5f75162357a1d92d57b2b08a336ff94b19b2be", size = 10938525, upload-time = "2026-07-22T22:18:36.81Z" },
+    { url = "https://files.pythonhosted.org/packages/f7/d0/317e7a0c67c0e69fa905a0161409397a7dc2d46ff611f6ca4803352c042b/pandas-3.0.5-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:cf52e1f61d229496da17dc7ab54acdee627357e7008fd4fecba3d0ba2937fa58", size = 11489303, upload-time = "2026-07-22T22:18:39.287Z" },
+    { url = "https://files.pythonhosted.org/packages/f1/8d/36dade89b49e4f9d5cbdbe863772581f98c0c6d78fc39ad4c557f6f2e17e/pandas-3.0.5-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:db172144bb56422bd157812f3b021eacc255451470b31e2c633c349490a1cfee", size = 11989004, upload-time = "2026-07-22T22:18:42.208Z" },
+    { url = "https://files.pythonhosted.org/packages/9c/ba/18c4ec8a746e177da05a9e7a7963781d8ea195780724f854601b6ebd6b78/pandas-3.0.5-cp313-cp313-win_amd64.whl", hash = "sha256:0d298e951f23016ce4699951d044ae6418dbc91bf68cefca0f77666fcbb4e5c6", size = 9826896, upload-time = "2026-07-22T22:18:44.539Z" },
+    { url = "https://files.pythonhosted.org/packages/de/ec/28a57266b753799a87b8bc79e7887ac6fd981b8c6d2978a0b7e7b6bd708c/pandas-3.0.5-cp313-cp313-win_arm64.whl", hash = "sha256:66266d3442a5e8b3c90274c2b8b230bee42dd1c286bc822cc2f9f2c7e12b883e", size = 9094790, upload-time = "2026-07-22T22:18:47.468Z" },
+    { url = "https://files.pythonhosted.org/packages/51/2f/cf6aae281264f4463f0875bcbb15fd2bb6d291cc535187dad1732475e4a9/pandas-3.0.5-cp314-cp314-macosx_10_15_x86_64.whl", hash = "sha256:2f264fc46911cc8131a7322a16199bbf8e353d27c10bb211f5bd0c814324dc36", size = 10390034, upload-time = "2026-07-22T22:18:49.818Z" },
+    { url = "https://files.pythonhosted.org/packages/06/ec/5189518c7a7659c4bdcc6b1eb32c46c6f3c86b0661ffd84143d1112c7732/pandas-3.0.5-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:53730687fcd161883b24e10411c06d6a4c0f2275d2faf3bb2bc25deb4ba8007c", size = 9980065, upload-time = "2026-07-22T22:18:52.249Z" },
+    { url = "https://files.pythonhosted.org/packages/ea/f1/598503ce8d7e3c35601e0747ba288c7864baae66380725bc12f13f884dfe/pandas-3.0.5-cp314-cp314-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:960d3ebcf249f75206899fcd2c6de53f736b7265759ced0d3e559df0b8b709b0", size = 10545532, upload-time = "2026-07-22T22:18:54.813Z" },
+    { url = "https://files.pythonhosted.org/packages/fa/de/ceae2adf7034e07e9910299fe412e1819c4f0dd520700a888bcb03625448/pandas-3.0.5-cp314-cp314-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:9e94c2c5ca43bd3ca32bf64d32308887b65e5f9bfd8023ea52755107a999f93b", size = 10963120, upload-time = "2026-07-22T22:18:57.42Z" },
+    { url = "https://files.pythonhosted.org/packages/66/25/86e0f4451874eb79e688deeebe3c451fec4557f8952005818d800ee8ac7e/pandas-3.0.5-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:e819dd5f62966b481a8cb649d3299ebd886a1ea91ed5a99bf7ce77c98d18ab94", size = 11563178, upload-time = "2026-07-22T22:18:59.729Z" },
+    { url = "https://files.pythonhosted.org/packages/f3/45/8643daa3b4147e433adfcccefdd0380d3aad79d86b15d8999730fe1944d5/pandas-3.0.5-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:3c5ed2e7c06e91d340dfd091d7934f9bc82e4a36b95f647f090b9d1c9ac649da", size = 12028708, upload-time = "2026-07-22T22:19:02.164Z" },
+    { url = "https://files.pythonhosted.org/packages/96/58/ad979ae617615576e8aafd569c9d4b62f1191d896e38f51d66ba06f3b89a/pandas-3.0.5-cp314-cp314-win_amd64.whl", hash = "sha256:cd8f7c6dc98527058ee6264219343f5392240a6f1bfa654fc5d79023020d0c92", size = 9951806, upload-time = "2026-07-22T22:19:04.596Z" },
+    { url = "https://files.pythonhosted.org/packages/69/32/7ac03886b304049a9d2625ee88f59af760d8a93bd30ed9239bce7b9869a8/pandas-3.0.5-cp314-cp314-win_arm64.whl", hash = "sha256:5183427f5a8156d480f30333777bc978be93650a49a7c01db26adffe95b31e85", size = 9238297, upload-time = "2026-07-22T22:19:06.836Z" },
+    { url = "https://files.pythonhosted.org/packages/be/ed/1d1f2ee5547d5167face2376d11c8b2a4c7bfff5a416ee7a9046891fab1e/pandas-3.0.5-cp314-cp314t-macosx_10_15_x86_64.whl", hash = "sha256:303da736987d481074ca720ada325f8bd80c64ebc2d45ed79b29df3aaa4a26ca", size = 10849690, upload-time = "2026-07-22T22:19:09.391Z" },
+    { url = "https://files.pythonhosted.org/packages/57/55/17e17152e98fbb0c4b1e562bc65387a2f20a80db0f4a86bf8d3a0e4248d4/pandas-3.0.5-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:3b2801bbb049d0136f6c213eae02b5fca969384fc2064dd728d8620552aa49da", size = 10509945, upload-time = "2026-07-22T22:19:11.773Z" },
+    { url = "https://files.pythonhosted.org/packages/88/90/817d44dbf83facf9556f33576d9af0a241981e7bb5c00606c0bcb5df8dda/pandas-3.0.5-cp314-cp314t-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:cce3a9d11d2b1f82c69a27ec1f4948a170e2c403c4bbfa8cca62e3fdebe2ef3a", size = 10392197, upload-time = "2026-07-22T22:19:14.024Z" },
+    { url = "https://files.pythonhosted.org/packages/f1/da/889f00c0a6f5aa1545add70abbf01502dff87ab577adb855bd631c54d2f2/pandas-3.0.5-cp314-cp314t-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:ef01af4d8dc6cd2c8d6c7736f149574ef93fe043811eeb5e445f2647154b5040", size = 10862726, upload-time = "2026-07-22T22:19:16.351Z" },
+    { url = "https://files.pythonhosted.org/packages/bc/98/f1e934fb3c98fce859c6147c6785816c7b5b9ab7821115c5d8c4de9842b9/pandas-3.0.5-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:e2759e890db96dfcffdbd9b86c3c2cb6afaf58def482820317e06163ec1066cd", size = 11414864, upload-time = "2026-07-22T22:19:18.981Z" },
+    { url = "https://files.pythonhosted.org/packages/fe/be/d448af7d657d82e1888dd8551f79c6d6fb161080b5b9752d84d910ec2319/pandas-3.0.5-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:b58b1b39d46a5862e3fb18f50d1a201398619d16a0f9f73f57eea5583cf0e63c", size = 11925105, upload-time = "2026-07-22T22:19:21.515Z" },
+    { url = "https://files.pythonhosted.org/packages/29/c1/ccb4238212c8c4f496c584f3044d94e0c030ed8e1d68999db46c91c2242f/pandas-3.0.5-cp314-cp314t-win_amd64.whl", hash = "sha256:1c10461f6eeb35d8f05b6184c65c8b9991663b66c46b1d559b682cb34ae7c6ea", size = 10387612, upload-time = "2026-07-22T22:19:24.257Z" },
+    { url = "https://files.pythonhosted.org/packages/d2/cf/6a51b2c38980e04c279fd2fa908a1b0982064e860444acfca4ec2e2c8359/pandas-3.0.5-cp314-cp314t-win_arm64.whl", hash = "sha256:3c5015fd1730fbf883647e88068176c839c102cea883ba1769a6f4593bfc1f8c", size = 9509776, upload-time = "2026-07-22T22:19:26.694Z" },
+]
+
+[[package]]
+name = "parso"
+version = "0.8.7"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/30/4b/90c937815137d43ce71ba043cd3566221e9df6b9c805f24b5d138c9d40a7/parso-0.8.7.tar.gz", hash = "sha256:eaaac4c9fdd5e9e8852dc778d2d7405897ec510f2a298071453e5e3a07914bb1", size = 401824, upload-time = "2026-05-01T23:13:02.138Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/99/5d/8268b644392ee874ee82a635cd0df1773de230bde356c38de28e298392cc/parso-0.8.7-py2.py3-none-any.whl", hash = "sha256:a8926eb2a1b915486941fdbd31e86a4baf88fe8c210f25f2f35ecec5b574ca1c", size = 107025, upload-time = "2026-05-01T23:12:58.867Z" },
+]
+
+[[package]]
+name = "patchright"
+version = "1.61.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "greenlet" },
+    { name = "pyee" },
+]
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/79/42/a76e2ea17dd4feccbf5ef6df017fb2e67f4b72a6f559f59d9aad1197f83a/patchright-1.61.2-py3-none-macosx_10_13_x86_64.whl", hash = "sha256:43ae10cd2a3e46b222179a120ab39ba88f801f3a42aa41cab45fac9928f2103d", size = 43727572, upload-time = "2026-07-05T21:17:52.382Z" },
+    { url = "https://files.pythonhosted.org/packages/57/e7/1ff89853c9aeaf26cd20559f401375ec71a40e4041cf7ace03d41a741f06/patchright-1.61.2-py3-none-macosx_11_0_arm64.whl", hash = "sha256:a3da24d7890631255c7bef467b94ffe7f7bfa36f96fc72c10d5787d3cb396873", size = 42512921, upload-time = "2026-07-05T21:17:55.657Z" },
+    { url = "https://files.pythonhosted.org/packages/e4/d0/d730536a7fc43c3aaf2781518595ca71ffa1fc2f24e580c0d0cf04e00eca/patchright-1.61.2-py3-none-macosx_11_0_universal2.whl", hash = "sha256:edef852aa7d28791fa2d3f7ce135edfea2ce07a2bc5e0a74fb2dfe269af62223", size = 43727573, upload-time = "2026-07-05T21:17:59.121Z" },
+    { url = "https://files.pythonhosted.org/packages/a2/b5/c76dcda275cb0d9651321e343f3268609ae24cef196bff56751754ffba16/patchright-1.61.2-py3-none-manylinux1_x86_64.whl", hash = "sha256:545bdf2c0bb5f9ad78ab315e91c8c2990cc6a702e7f18650b71fe1071049b5c1", size = 47729338, upload-time = "2026-07-05T21:18:02.608Z" },
+    { url = "https://files.pythonhosted.org/packages/19/ef/5de6feaedf47a6ba8efb9e8c4f8a42bad29cc59c35bbd088b1729a36b271/patchright-1.61.2-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:1824da30cf6560749bc1bb4a7f002fec325d5b709090dd7c8c1b643899331cee", size = 47426969, upload-time = "2026-07-05T21:18:06.88Z" },
+    { url = "https://files.pythonhosted.org/packages/6c/6e/ce2c5d03a4c785ac8f3e0180bbf83ef85149cf41fdfa41c09bd8cbcfd945/patchright-1.61.2-py3-none-win32.whl", hash = "sha256:a5aaf4cec2bd38714f2447f84bcdfe7dcb592a16e9616bf6d2d872066d1c78af", size = 38155048, upload-time = "2026-07-05T21:18:09.938Z" },
+    { url = "https://files.pythonhosted.org/packages/46/c8/fceb57b9d32ed078f53fc6a757c6e4b54cb117f59412938252a3359105e5/patchright-1.61.2-py3-none-win_amd64.whl", hash = "sha256:506bca9b72e609fd8fee35e2a3ad26e5d92e25337002f5c2cdad53d5cacb0b67", size = 38155052, upload-time = "2026-07-05T21:18:13.051Z" },
+    { url = "https://files.pythonhosted.org/packages/33/96/67a791c6b7cd76424b44d356beaf414ab97c3aef2d42fad4641e7905b41d/patchright-1.61.2-py3-none-win_arm64.whl", hash = "sha256:3dee1e2daada9f98653763831f688d39c924e0cf0eeb988e8047b703766cee73", size = 34268624, upload-time = "2026-07-05T21:18:15.966Z" },
+]
+
+[[package]]
+name = "pexpect"
+version = "4.9.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "ptyprocess", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/42/92/cc564bf6381ff43ce1f4d06852fc19a2f11d180f23dc32d9588bee2f149d/pexpect-4.9.0.tar.gz", hash = "sha256:ee7d41123f3c9911050ea2c2dac107568dc43b2d3b0c7557a33212c398ead30f", size = 166450, upload-time = "2023-11-25T09:07:26.339Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/9e/c3/059298687310d527a58bb01f3b1965787ee3b40dce76752eda8b44e9a2c5/pexpect-4.9.0-py2.py3-none-any.whl", hash = "sha256:7236d1e080e4936be2dc3e326cec0af72acf9212a7e1d060210e70a47e253523", size = 63772, upload-time = "2023-11-25T06:56:14.81Z" },
+]
+
+[[package]]
+name = "pillow"
+version = "12.3.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/1c/3d/bb7fca845737cf9d7dbde16ed1843984665ff2e0a518f5db43e77ec540b9/pillow-12.3.0.tar.gz", hash = "sha256:3b8182a766685eaa002637e28b4ec8d6b18819a0c71f579bf0dbaa5830297cce", size = 47025035, upload-time = "2026-07-01T11:56:38.965Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/37/bf/fb3ebff8ddcb76aac5a01389251bbbb9519922a9b520d8247c1ca864a25d/pillow-12.3.0-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:ba09209fbe443b4acccebe845d8a138b89a8f4fbaeedd44953490b5315d5e965", size = 5345969, upload-time = "2026-07-01T11:54:06.397Z" },
+    { url = "https://files.pythonhosted.org/packages/d8/66/9a386a92561f402389a4fc70c18838bf6d35eb5eb5c6850b4b2dc64f5048/pillow-12.3.0-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:ffd0c5368496f41b0944be820fcb7a838aa6e623d250b01acf2643939c3f99d7", size = 4780323, upload-time = "2026-07-01T11:54:09.351Z" },
+    { url = "https://files.pythonhosted.org/packages/25/27/ac8f99618ffd3dde21db0f4d4b1d2ab00c0880595bfd17df103f7f39fd0c/pillow-12.3.0-cp312-cp312-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:d9c7f76c0673154f044e9d78c8655fb4213f6ca31a836df48b40fe5d187717b9", size = 6266838, upload-time = "2026-07-01T11:54:11.71Z" },
+    { url = "https://files.pythonhosted.org/packages/84/21/a35af28dcc61f37ed850a2d64c65c701321dfbf25085e469d5559360cbbf/pillow-12.3.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:78cb2c6865a35ab8ff8b75fd122f6033b92a62c82801110e48ddd6c936a45d91", size = 6940830, upload-time = "2026-07-01T11:54:13.732Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/51/8b08617af3ad95e33ce6d7dd2c99ed6c8298f7fb131636303956be022e25/pillow-12.3.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:e491916b378fba47242221bb9ead245211b70d504f495d105d17b14a24b4907c", size = 6344383, upload-time = "2026-07-01T11:54:15.756Z" },
+    { url = "https://files.pythonhosted.org/packages/1d/72/cf78ac9780bb93c28328f408973845a309d4d145041665f734572ced1b52/pillow-12.3.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:0dd2064cbc55aaec028ef5fbb60fa47bb6c3e7918e07ff17935284b227a9d2df", size = 7052934, upload-time = "2026-07-01T11:54:17.721Z" },
+    { url = "https://files.pythonhosted.org/packages/20/20/25e0f4dc178a6bc0696793720055519a0de89e7661dae886992decbd2f81/pillow-12.3.0-cp312-cp312-win32.whl", hash = "sha256:dbce0b29841537a2fa4a214c2bbf14de3587c9680caa9b4e217568472490b28f", size = 6472684, upload-time = "2026-07-01T11:54:19.839Z" },
+    { url = "https://files.pythonhosted.org/packages/45/89/da2f7971a317f83d807fdd4065c0af40208e59e692cc43d315a71a0e96d1/pillow-12.3.0-cp312-cp312-win_amd64.whl", hash = "sha256:a2b55dd6b2a4c4b7d87ffa56bdb33fdc5fdb9a462173861a7bc097f17d91cb09", size = 7227137, upload-time = "2026-07-01T11:54:22.025Z" },
+    { url = "https://files.pythonhosted.org/packages/de/47/4845a0a6c0dbf1db8456bd9fc791f13c5ced7ced20606d08a0aacfd25b49/pillow-12.3.0-cp312-cp312-win_arm64.whl", hash = "sha256:331b624368d4f1d069149002f25f44bc61c8919ce8ddb3c45bdad8f6e2d89510", size = 2568267, upload-time = "2026-07-01T11:54:24.051Z" },
+    { url = "https://files.pythonhosted.org/packages/9d/ac/31fb64e1e7efb5a4b50cd3d92049ba89ac6e4d8d3bb6a74e15048ca3353e/pillow-12.3.0-cp313-cp313-ios_13_0_arm64_iphoneos.whl", hash = "sha256:21900ce7ba264168cd50defae43cd75d25c833ad4ad6e73ffc5596d12e25ac89", size = 4161684, upload-time = "2026-07-01T11:54:25.934Z" },
+    { url = "https://files.pythonhosted.org/packages/87/b4/9805e23d2b4d77842b468513841fda254ee42f0289d25088340e4ff46e2d/pillow-12.3.0-cp313-cp313-ios_13_0_arm64_iphonesimulator.whl", hash = "sha256:4e8c2a84d977f50b9daed6eeaf3baef67d00d5d74d932288f02cb94518ee3ace", size = 4255487, upload-time = "2026-07-01T11:54:27.935Z" },
+    { url = "https://files.pythonhosted.org/packages/df/39/ecf519435a200c693fe053a6ee4d835b41cf963a4dfc2551c4e637cb2a71/pillow-12.3.0-cp313-cp313-ios_13_0_x86_64_iphonesimulator.whl", hash = "sha256:ae26d61dfa7a47befdc7572b521024e8745f3d809bd95ca9505a7bba9ef849ec", size = 3696433, upload-time = "2026-07-01T11:54:29.813Z" },
+    { url = "https://files.pythonhosted.org/packages/42/92/2fc3ffad878ae8dd5469ec1bc8eb83b71f48e13efdf68f02709003982a32/pillow-12.3.0-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:7a743ff716f746fc19a9557f60dab1600d4613255f8a7aeb3cdde4db7eb15a66", size = 5345889, upload-time = "2026-07-01T11:54:31.97Z" },
+    { url = "https://files.pythonhosted.org/packages/10/76/8803c13605b763d33d156c4678fc77f8443389c0c51c8aef707bb02015f4/pillow-12.3.0-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:d69141514cc30b774ceea5e3ed3a6635c8d8a96edf664689b890f4089111fb35", size = 4780109, upload-time = "2026-07-01T11:54:34.026Z" },
+    { url = "https://files.pythonhosted.org/packages/1f/01/e18aff37cb0b4aac47ac90f016d347a49aca667ef97f190b06ac2aabc928/pillow-12.3.0-cp313-cp313-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:f7401aebd7f581d7f83a439d87d474999317ee099218e5ad25d125290990ba65", size = 6263736, upload-time = "2026-07-01T11:54:36.131Z" },
+    { url = "https://files.pythonhosted.org/packages/f7/62/de5bdd77d935331f4f802edc11e4d82950f642caad6cb2f949837b8560e2/pillow-12.3.0-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:0847a763afefb695bc912d7c131e7e0632d4edc1d8698f58ddabec8e46b8b6d3", size = 6937129, upload-time = "2026-07-01T11:54:38.216Z" },
+    { url = "https://files.pythonhosted.org/packages/70/4d/105627a13300c5e0df1d174230b32fd1273062c96f7745fd552b945d1e1d/pillow-12.3.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:571b9fcb07b97ef3a492028fb3d2dc0993ca23a06138b0315286566d29ef718a", size = 6339562, upload-time = "2026-07-01T11:54:40.354Z" },
+    { url = "https://files.pythonhosted.org/packages/6b/1d/f13de01a553988ab895ba1c722e06cf3144d4f57656fd5b81b6d881f1179/pillow-12.3.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:756c768d0c9c2955feb7a56c37ea24aea2e369f8d36a88da270b6a9f19e62b5e", size = 7049439, upload-time = "2026-07-01T11:54:42.489Z" },
+    { url = "https://files.pythonhosted.org/packages/c9/f9/066794cca041b969964f779ee5fa66a9498bbf34248ac39c5d7954e4198f/pillow-12.3.0-cp313-cp313-win32.whl", hash = "sha256:a876864214e136f0eb367788dbd7df045f4806801518e2cfe9e13229cfe06d8f", size = 6473287, upload-time = "2026-07-01T11:54:44.9Z" },
+    { url = "https://files.pythonhosted.org/packages/a6/9b/7a58e61d62be561da3a356fe2384d4059a6345fc130e23ef1c36a5b81d24/pillow-12.3.0-cp313-cp313-win_amd64.whl", hash = "sha256:1cca606cd25738df4ed873d5ad46bbdb3d83b5cbca291f6b4ff13a4df6b0bbe8", size = 7239691, upload-time = "2026-07-01T11:54:47.141Z" },
+    { url = "https://files.pythonhosted.org/packages/aa/b0/c4ed4f0ef8f8fa5ee8351537db6650bb8189f7e118842978dd6589065692/pillow-12.3.0-cp313-cp313-win_arm64.whl", hash = "sha256:b629de27fda84b42cde7edef0d85f13b958b47f6e9bbcbba9b673c562a89bd8b", size = 2568185, upload-time = "2026-07-01T11:54:49.137Z" },
+    { url = "https://files.pythonhosted.org/packages/dc/01/001f65b68192f0228cc1dbbc8d2530ab5d58b61037ba0587f946fea607cd/pillow-12.3.0-cp314-cp314-ios_13_0_arm64_iphoneos.whl", hash = "sha256:9cf95fe4d0f84c82d282745d9bb08ad9f926efa00be4697e767b814ce40d4330", size = 4161736, upload-time = "2026-07-01T11:54:51.156Z" },
+    { url = "https://files.pythonhosted.org/packages/1a/d2/0219746d0fd16fc8a84498e79452375be3797d3ce4044596ce565164b84f/pillow-12.3.0-cp314-cp314-ios_13_0_arm64_iphonesimulator.whl", hash = "sha256:8728f216dcdb6e6d555cf971cb34076139ad74b31fc2c14da4fafc741c5f6217", size = 4255435, upload-time = "2026-07-01T11:54:53.414Z" },
+    { url = "https://files.pythonhosted.org/packages/c8/02/8d0bc62ef0302318c46ff2a512822d2610e81c7aa46c9b3abe6cbaca5ad0/pillow-12.3.0-cp314-cp314-ios_13_0_x86_64_iphonesimulator.whl", hash = "sha256:a45650e8ce7fafffd731db8550230db6b0d306d181a90b67d3e6bca2f1990930", size = 3696262, upload-time = "2026-07-01T11:54:55.739Z" },
+    { url = "https://files.pythonhosted.org/packages/85/e2/73c77d218410b14f5f2d565e8a998d5317b7b9c75368d29985139f7a46f0/pillow-12.3.0-cp314-cp314-macosx_10_15_x86_64.whl", hash = "sha256:ba54cfebe86920a559a7c4d6b9050791c20513650a1952ebe3368c7dc70306f8", size = 5350344, upload-time = "2026-07-01T11:54:57.657Z" },
+    { url = "https://files.pythonhosted.org/packages/c7/da/32c752228ae345f489e3a42499d817b6c3996da7e8a3bc7a04fc806b243b/pillow-12.3.0-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:e158cb00350dc278f3b91551101aa7d12415a66ebf2c91d8d5ac14e56ddd3ad0", size = 4780131, upload-time = "2026-07-01T11:54:59.713Z" },
+    { url = "https://files.pythonhosted.org/packages/b1/9d/8b2c807dbef61a5197c047afe99823787eb66f63daf9fb2432f91d6f0462/pillow-12.3.0-cp314-cp314-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:e9aeb04d6aef139de265b29683e119b638208f88cf73cdd1658aa07221165321", size = 6263757, upload-time = "2026-07-01T11:55:01.778Z" },
+    { url = "https://files.pythonhosted.org/packages/5c/44/c85361f65dbe00eea8576ee467c768d25129989efb76e94f205e9ca9bb46/pillow-12.3.0-cp314-cp314-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:251bf95b67017e27b13d82f5b326234ca62d70f9cf4c2b9032de2358a3b12c7b", size = 6936962, upload-time = "2026-07-01T11:55:03.93Z" },
+    { url = "https://files.pythonhosted.org/packages/18/7e/e483414b35800b86b6f08dbbc7803fb5cd52c4d6f897f47d53ea2c7e6f65/pillow-12.3.0-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:fe3cca2e4e8a592be0f269a1ca4835c25199d9f3ce815c8491048f785b0a0198", size = 6339171, upload-time = "2026-07-01T11:55:05.989Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/f4/68c491844841ede6bed70189546b3ee9731cf9f2cbad396faff5e1ccba45/pillow-12.3.0-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:23aceaa007d6172b02c277f0cd359c79492bbb14f7072b4ede9fbcaf20648130", size = 7048116, upload-time = "2026-07-01T11:55:08.131Z" },
+    { url = "https://files.pythonhosted.org/packages/a3/34/77f3f793fed8efc7d243f21b33c5a3f0d1c97ee70346d3db855587e155ff/pillow-12.3.0-cp314-cp314-win32.whl", hash = "sha256:af8d94b0db561cf68b88a267c5c44b49e134f525d0dc2cb7ed413a66bc23559a", size = 6467209, upload-time = "2026-07-01T11:55:10.408Z" },
+    { url = "https://files.pythonhosted.org/packages/f1/e0/492879f69d94f91f60fc8cd05ba03650e9520afebb2fb7aa12777d7c7f38/pillow-12.3.0-cp314-cp314-win_amd64.whl", hash = "sha256:fdafc9cce40277e0f7a0feabce0ee50dd2fa1800f3b38015e51296b5e814048d", size = 7237707, upload-time = "2026-07-01T11:55:12.745Z" },
+    { url = "https://files.pythonhosted.org/packages/c9/ac/6b11f2875f1c2ac040d84e1bbf9cf22a88038f901ca1037898b280b38365/pillow-12.3.0-cp314-cp314-win_arm64.whl", hash = "sha256:e91206ee562682b51b98ef4b26a6ef48fd84e15fd4c4bc5ec768eb641d206838", size = 2565995, upload-time = "2026-07-01T11:55:14.736Z" },
+    { url = "https://files.pythonhosted.org/packages/52/69/c2208e56af9bfc1913afb24020297a691eb1d4ef688474c8a04913f65e04/pillow-12.3.0-cp314-cp314t-macosx_10_15_x86_64.whl", hash = "sha256:164b31cd1a0490ab6efae01aa5df49da7061be0af1b30e035b6e9a1bfe34ee6e", size = 5352503, upload-time = "2026-07-01T11:55:17.076Z" },
+    { url = "https://files.pythonhosted.org/packages/07/70/e5686d753e898a45d778ff1718dba8516ead6ab6b95d85fc8c4b70650cf2/pillow-12.3.0-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:5afb51d599ea772b8365ae807ae557f18bccfe46ab261fd1c2a9ed700fc6eb17", size = 4782956, upload-time = "2026-07-01T11:55:19.448Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/37/25c6692f06927ee973ff18c8d9ee98ad0b4d84ee67a09610c2dd1447958e/pillow-12.3.0-cp314-cp314t-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:3edce1d53195db527e0191f84b71d02022de0540bf43a16ed734ed7537b07385", size = 6322855, upload-time = "2026-07-01T11:55:21.613Z" },
+    { url = "https://files.pythonhosted.org/packages/cc/91/420637fcb8f1bc11029e403b4538e6694744428d8246118e45719f944556/pillow-12.3.0-cp314-cp314t-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:bf16ba1b4d0b6b7c8e534936632270cf70eb00dbe09005bc345b2677b726855c", size = 6989642, upload-time = "2026-07-01T11:55:24.006Z" },
+    { url = "https://files.pythonhosted.org/packages/10/08/b94d7811281ccf0d143a1cf768d1c49e1e54af63e7b708ab2ee3eb87face/pillow-12.3.0-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:24870b09b224f7ae3c39ed07d10e819d06f8720bc551847b1d623832b5b0e28d", size = 6391281, upload-time = "2026-07-01T11:55:26.252Z" },
+    { url = "https://files.pythonhosted.org/packages/d2/87/24233f785f55474dc02ce3e739c5528a77e3a862e9333d1dd7a25cc31f70/pillow-12.3.0-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:30f2aa603c41533cc25c05acd0da21636e84a315768feb631c937177db558931", size = 7096716, upload-time = "2026-07-01T11:55:28.318Z" },
+    { url = "https://files.pythonhosted.org/packages/23/26/fcb2f6e37175b04f53570b59937867e2b80ee1685e744023153028fc14f9/pillow-12.3.0-cp314-cp314t-win32.whl", hash = "sha256:4b0a7fe987b14c31ebda6083f74f22b561fd3739bc0ac51e019622e3d72668c7", size = 6474125, upload-time = "2026-07-01T11:55:30.956Z" },
+    { url = "https://files.pythonhosted.org/packages/90/de/3634abee5f1c9e13c56787b7d5517b0ba8d6de51700b95578cf338349c9f/pillow-12.3.0-cp314-cp314t-win_amd64.whl", hash = "sha256:962864dc93511324d51ddbb5b9f8731bf71675b93ca612a07441896f4688fb8c", size = 7242939, upload-time = "2026-07-01T11:55:34.044Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/2a/fd13f8eb24de5714a6eb444a3d67e2842c6c576e159a43793adf23051351/pillow-12.3.0-cp314-cp314t-win_arm64.whl", hash = "sha256:0740a512dc522224c77d9aa5a8d70d8b7d73fb91f2c21125d8d025d3b8990e45", size = 2567506, upload-time = "2026-07-01T11:55:35.988Z" },
+    { url = "https://files.pythonhosted.org/packages/5d/dc/8fdce34ec725a33c81c6ba122b904d6b9024e50ea9ac7bede62fab54506c/pillow-12.3.0-cp315-cp315-ios_13_0_arm64_iphoneos.whl", hash = "sha256:0feb2e9d6ad6c9e3c06effe9d00f3f1e618a6643273576b016f591e9315a7139", size = 4162063, upload-time = "2026-07-01T11:55:37.941Z" },
+    { url = "https://files.pythonhosted.org/packages/76/66/2044b9a63d3b84ff048228dfcb7cd9bf0df983e8470971bf7d4c57b693de/pillow-12.3.0-cp315-cp315-ios_13_0_arm64_iphonesimulator.whl", hash = "sha256:9e881fca225083806662a5c43d627d215f258ff43c890f831966c7d7ba9c7402", size = 4255549, upload-time = "2026-07-01T11:55:40.022Z" },
+    { url = "https://files.pythonhosted.org/packages/52/7e/1f67e6f4ece6b582ee4b539decbcc9f848dc245a93ed8cd7338bafef72f1/pillow-12.3.0-cp315-cp315-ios_13_0_x86_64_iphonesimulator.whl", hash = "sha256:4998562bf62a445225f22e07c896bb04b35b1b1f2eb6d760584c9c51d7a5f78c", size = 3696331, upload-time = "2026-07-01T11:55:41.98Z" },
+    { url = "https://files.pythonhosted.org/packages/12/40/d306fc2c8e4d45d7f175c77edca7063be7b86fe7fe6e68f4353bf71d808c/pillow-12.3.0-cp315-cp315-macosx_10_15_x86_64.whl", hash = "sha256:dc624f6bc473dacdf7ef7eb8678d0d08edf15cd94fad6ae5c7d6cc67a4e4902f", size = 5350370, upload-time = "2026-07-01T11:55:44.028Z" },
+    { url = "https://files.pythonhosted.org/packages/dd/44/668fb1437e8ce420f62d6106eb66e44a5971602a4d794615bdf79315d82d/pillow-12.3.0-cp315-cp315-macosx_11_0_arm64.whl", hash = "sha256:71d6097b330eea8fd15097780c8e89cb1a8ce7838669f48c5bacd6f663dd4701", size = 4780147, upload-time = "2026-07-01T11:55:46.073Z" },
+    { url = "https://files.pythonhosted.org/packages/0c/08/93fa2e70e30a2d81547e481b6ee2bb9522117221fb1e0ce4b5df70967677/pillow-12.3.0-cp315-cp315-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:28ce87c5ab450a9dd970b52e5aca5fe63ed432d18a2eaddd1979a00a1ba24ace", size = 6273659, upload-time = "2026-07-01T11:55:48.264Z" },
+    { url = "https://files.pythonhosted.org/packages/f8/6d/043e96ff814fc31a33077e4cba86082167db520c93632afdf2042febbb0c/pillow-12.3.0-cp315-cp315-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:6b02afb9b97f65fbca5f31db6a2a3ba21aa93030225f150fa3f249717e938fb4", size = 6947439, upload-time = "2026-07-01T11:55:50.503Z" },
+    { url = "https://files.pythonhosted.org/packages/af/92/ba71d2ee2ac0edf3fa33bd9d5ee9ee080da70b1766f3ca3934f9938ddac9/pillow-12.3.0-cp315-cp315-musllinux_1_2_aarch64.whl", hash = "sha256:1182d52bc2d5e5d7d0949503aa7e36d12f42205dc287e4883f407b1988820d39", size = 6353577, upload-time = "2026-07-01T11:55:52.697Z" },
+    { url = "https://files.pythonhosted.org/packages/0f/ce/e63064e2122923ff687c8ad792d0d736a7b3920a56a46982e81a7fdd25d6/pillow-12.3.0-cp315-cp315-musllinux_1_2_x86_64.whl", hash = "sha256:e795b7eb908249c4e43c7c99fac7c2c75dab0c43566e37db472a355f63693d71", size = 7060394, upload-time = "2026-07-01T11:55:55.149Z" },
+    { url = "https://files.pythonhosted.org/packages/54/76/a09cc3ccc8d773a7283d34c38bec1708f9e3cc932093cbc4c5e71ac4060b/pillow-12.3.0-cp315-cp315-win32.whl", hash = "sha256:57b3d78c95ba9059768b10e28b813002261d3f3dfc55cc48b0c988f625175827", size = 6467375, upload-time = "2026-07-01T11:55:57.769Z" },
+    { url = "https://files.pythonhosted.org/packages/3e/03/1846c49ba3b1d5550392a4bbd06d6fb4578e1cd91a803198b5c90f5f7d53/pillow-12.3.0-cp315-cp315-win_amd64.whl", hash = "sha256:fa4ecea169a355be7a3ade2c783e2ed12f0e40d2c5621cda8b3297faf7fbb9f5", size = 7237048, upload-time = "2026-07-01T11:55:59.975Z" },
+    { url = "https://files.pythonhosted.org/packages/fb/bb/89f35dcc79610423f9f195504d7def7f0d1416a711541b42867e25fe3412/pillow-12.3.0-cp315-cp315-win_arm64.whl", hash = "sha256:877c3f311ff35410f690861c4409e7ccbf0cd2f878e50628a28e5a0bb689e658", size = 2566006, upload-time = "2026-07-01T11:56:02.143Z" },
+    { url = "https://files.pythonhosted.org/packages/30/88/707027ba09942dfa2c28759b5c222d769290a41c6d20ea60ec250801941f/pillow-12.3.0-cp315-cp315t-macosx_10_15_x86_64.whl", hash = "sha256:e9871b1ffbfa9656b60aeee92ed5136a5742696006fa322b29ea3d8da0ecc9cf", size = 5352509, upload-time = "2026-07-01T11:56:04.2Z" },
+    { url = "https://files.pythonhosted.org/packages/b0/6d/00352fa25332c2569cd387851f568cc5a4b75a9adbfb37ac4fbce4c02eec/pillow-12.3.0-cp315-cp315t-macosx_11_0_arm64.whl", hash = "sha256:53aa02d20d10c3d814d536aa4e5ac9b84ca0ff5a88377963b085ad6822f93e64", size = 4783167, upload-time = "2026-07-01T11:56:06.631Z" },
+    { url = "https://files.pythonhosted.org/packages/13/4f/9e049dfa21af7c22427275720e2490267ba8138120add5c4c574deb69782/pillow-12.3.0-cp315-cp315t-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:446c34dcc4324b084a53b705127dc15717b22c5e140ae0a3c38349d4efec071e", size = 6329237, upload-time = "2026-07-01T11:56:08.868Z" },
+    { url = "https://files.pythonhosted.org/packages/36/16/cf6eeaae8d0fce8dd390a33437cf68c5d5bd73834a2bc6e2f14efda0ab45/pillow-12.3.0-cp315-cp315t-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:cf1845d02ad822a369a49f2bb9345b1614744267682e7a03527dc3bf6eea1777", size = 6997047, upload-time = "2026-07-01T11:56:11.379Z" },
+    { url = "https://files.pythonhosted.org/packages/1e/69/dbf769bdd55f48bf5733cac28edc6364ffaa072ec9ba336266e4fe66be55/pillow-12.3.0-cp315-cp315t-musllinux_1_2_aarch64.whl", hash = "sha256:186941b6aef820ad110fb01fb06eb925374dc3a21b17e37ec9a53b250c6fe2d1", size = 6400440, upload-time = "2026-07-01T11:56:13.908Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/e1/ffc9cfc2eea0d178da8018e18e959301ad9d6bc9f3edb7181e748a474b97/pillow-12.3.0-cp315-cp315t-musllinux_1_2_x86_64.whl", hash = "sha256:f13c32a3abd6079a66d9526e18dad9b6d280384d49d7c54040cd57b6424041d9", size = 7105895, upload-time = "2026-07-01T11:56:16.575Z" },
+    { url = "https://files.pythonhosted.org/packages/18/f0/a5595c1e8c3ae44b9828cb2f0fa8155e5095ef04d6327b8f61cf44a3df85/pillow-12.3.0-cp315-cp315t-win32.whl", hash = "sha256:1657923d2d45afb66526e5b933e5b3052e6bdea196c90d3abb2424e18c77dae8", size = 6474384, upload-time = "2026-07-01T11:56:18.855Z" },
+    { url = "https://files.pythonhosted.org/packages/e4/04/62bcd9f844984c5938d3b05264a61d797a29d3e0812341a8204af70bbdee/pillow-12.3.0-cp315-cp315t-win_amd64.whl", hash = "sha256:8cd2f7bdda092d99c9fc2fb7391354f306d01443d22785d0cbfafa2e2c8bb418", size = 7243537, upload-time = "2026-07-01T11:56:21.214Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/68/1f3066acedf37673694a7141381d8f811ae97f30d34413d236abe7d489f1/pillow-12.3.0-cp315-cp315t-win_arm64.whl", hash = "sha256:06ff022112bc9cbf83b60f8e028d94ad87b60621706487e65f673de61610ab59", size = 2567491, upload-time = "2026-07-01T11:56:23.506Z" },
+]
+
+[[package]]
+name = "playwright"
+version = "1.61.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "greenlet" },
+    { name = "pyee" },
+]
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/44/ee/31e4e0db36588b817a10b299a0285082545fde7d36543c2abe498bb3d61a/playwright-1.61.0-py3-none-macosx_10_13_x86_64.whl", hash = "sha256:ff138c3a604f69911e9d42fd036e55c2a171e5616edf04c1e7f60a2a285540b0", size = 43421877, upload-time = "2026-06-29T10:32:48.428Z" },
+    { url = "https://files.pythonhosted.org/packages/42/35/71395dd3ecc798965be4a3ef8c443217d4abca168e7cb34536304f9489e6/playwright-1.61.0-py3-none-macosx_11_0_arm64.whl", hash = "sha256:009588c2a7e499bc5a8b425b61fa65490968bbda9cd69e0cf2cff10f8304659a", size = 42205016, upload-time = "2026-06-29T10:32:52.104Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/44/323164cf5cd1647bdefce76ffce27651aadb959d089b48f53ea40918276e/playwright-1.61.0-py3-none-macosx_11_0_universal2.whl", hash = "sha256:9f7de4536088d12037c13a52b7ea34b59270b78926bb56935070597ffac6b1af", size = 43421884, upload-time = "2026-06-29T10:32:55.773Z" },
+    { url = "https://files.pythonhosted.org/packages/ab/f8/a35bf179e4ba2522c1893635094a64e407572547bd61528820fc0abc87fe/playwright-1.61.0-py3-none-manylinux1_x86_64.whl", hash = "sha256:54f3b39f6eab832e33458c1dd7da0b5682aedab3b09ae731b5c59fa12fd2024e", size = 47421381, upload-time = "2026-06-29T10:32:59.903Z" },
+    { url = "https://files.pythonhosted.org/packages/b7/eb/e3f922348ec17c315f98c463f72faa1181a1c3de0bfe31a8d2edf6561723/playwright-1.61.0-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:93454322ade8c11d5d6c211bfd91bdfb9ffb4810e3e026371bcbc4bec1b7ee4c", size = 47120545, upload-time = "2026-06-29T10:33:03.574Z" },
+    { url = "https://files.pythonhosted.org/packages/c2/a6/5be4e52b40a9c0c8a073e7c5b0785c05cf5a9ea8f8a7b5b260e32d970342/playwright-1.61.0-py3-none-win32.whl", hash = "sha256:372d55a6f1248fa1dd47599686980cb8fb5bbe6fcda59eab793eb657c11d8a9b", size = 37844841, upload-time = "2026-06-29T10:33:07.361Z" },
+    { url = "https://files.pythonhosted.org/packages/6c/fd/2b78036e5fbe9d5f5645bbe08a1eac7160c51243c0093963edbcf67c35d9/playwright-1.61.0-py3-none-win_amd64.whl", hash = "sha256:35c6cc4589a5d00964a59d7b3e59641e0aac0c02f15479a7af77d20f6bc79597", size = 37844846, upload-time = "2026-06-29T10:33:10.637Z" },
+    { url = "https://files.pythonhosted.org/packages/27/0d/1b0f3c4ee4eb0514bc805b5c2f9a223e5b6de4f11a926f5235d51d0fc81b/playwright-1.61.0-py3-none-win_arm64.whl", hash = "sha256:e9fcbffcf557a8620fdedd92491eb59a32d18e23d6f3b4f6214b952be324fe51", size = 33955127, upload-time = "2026-06-29T10:33:14.008Z" },
+]
+
+[[package]]
+name = "pluggy"
+version = "1.6.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/f9/e2/3e91f31a7d2b083fe6ef3fa267035b518369d9511ffab804f839851d2779/pluggy-1.6.0.tar.gz", hash = "sha256:7dcc130b76258d33b90f61b658791dede3486c3e6bfb003ee5c9bfb396dd22f3", size = 69412, upload-time = "2025-05-15T12:30:07.975Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/54/20/4d324d65cc6d9205fabedc306948156824eb9f0ee1633355a8f7ec5c66bf/pluggy-1.6.0-py3-none-any.whl", hash = "sha256:e920276dd6813095e9377c0bc5566d94c932c33b27a3e3945d8389c374dd4746", size = 20538, upload-time = "2025-05-15T12:30:06.134Z" },
+]
+
+[[package]]
+name = "prompt-toolkit"
+version = "3.0.53"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "wcwidth" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/7d/ea/39b988c938f75cb75d7045b5c69f8bfed47ee2152c8837fb403de29d6fb8/prompt_toolkit-3.0.53.tar.gz", hash = "sha256:9ec8a0ad96d5c56148b3f914aa79c1564c3fde5d2e6b876e7bc327e353cf8fa6", size = 435492, upload-time = "2026-07-26T20:56:14.758Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/54/6f/84908cad2d6aa5144abcf7b42709fe4fdb459bc640ec7ac5786e7693dabc/prompt_toolkit-3.0.53-py3-none-any.whl", hash = "sha256:01c0891d7f9237d5e339f7d3e42cdae80b7534abb1c7c0e3352efba6231492f2", size = 392288, upload-time = "2026-07-26T20:56:12.512Z" },
+]
+
+[[package]]
+name = "protego"
+version = "0.6.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/7d/1b/6b0ee60bb1561843bfc62f5c7c3cb1ef6147b47e829a5fd6b7fcd2752471/protego-0.6.2.tar.gz", hash = "sha256:88ff004544ce44e61269cc6f8735f7837d12e09bb77619fc94fbb26b72e5d137", size = 3137854, upload-time = "2026-06-25T09:33:16.361Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/61/10/cbd3c06603bb8b3e0e871df24ee793a6e3b53d8c4cc8763f1b8b936649aa/protego-0.6.2-py3-none-any.whl", hash = "sha256:714de21d82527c9be900066c3211b266985dd6a19b6e70c57e033fc1a589f3ff", size = 10296, upload-time = "2026-06-25T09:33:14.771Z" },
+]
+
+[[package]]
+name = "psutil"
+version = "7.2.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/aa/c6/d1ddf4abb55e93cebc4f2ed8b5d6dbad109ecb8d63748dd2b20ab5e57ebe/psutil-7.2.2.tar.gz", hash = "sha256:0746f5f8d406af344fd547f1c8daa5f5c33dbc293bb8d6a16d80b4bb88f59372", size = 493740, upload-time = "2026-01-28T18:14:54.428Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/51/08/510cbdb69c25a96f4ae523f733cdc963ae654904e8db864c07585ef99875/psutil-7.2.2-cp313-cp313t-macosx_10_13_x86_64.whl", hash = "sha256:2edccc433cbfa046b980b0df0171cd25bcaeb3a68fe9022db0979e7aa74a826b", size = 130595, upload-time = "2026-01-28T18:14:57.293Z" },
+    { url = "https://files.pythonhosted.org/packages/d6/f5/97baea3fe7a5a9af7436301f85490905379b1c6f2dd51fe3ecf24b4c5fbf/psutil-7.2.2-cp313-cp313t-macosx_11_0_arm64.whl", hash = "sha256:e78c8603dcd9a04c7364f1a3e670cea95d51ee865e4efb3556a3a63adef958ea", size = 131082, upload-time = "2026-01-28T18:14:59.732Z" },
+    { url = "https://files.pythonhosted.org/packages/37/d6/246513fbf9fa174af531f28412297dd05241d97a75911ac8febefa1a53c6/psutil-7.2.2-cp313-cp313t-manylinux2010_x86_64.manylinux_2_12_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:1a571f2330c966c62aeda00dd24620425d4b0cc86881c89861fbc04549e5dc63", size = 181476, upload-time = "2026-01-28T18:15:01.884Z" },
+    { url = "https://files.pythonhosted.org/packages/b8/b5/9182c9af3836cca61696dabe4fd1304e17bc56cb62f17439e1154f225dd3/psutil-7.2.2-cp313-cp313t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:917e891983ca3c1887b4ef36447b1e0873e70c933afc831c6b6da078ba474312", size = 184062, upload-time = "2026-01-28T18:15:04.436Z" },
+    { url = "https://files.pythonhosted.org/packages/16/ba/0756dca669f5a9300d0cbcbfae9a4c30e446dfc7440ffe43ded5724bfd93/psutil-7.2.2-cp313-cp313t-win_amd64.whl", hash = "sha256:ab486563df44c17f5173621c7b198955bd6b613fb87c71c161f827d3fb149a9b", size = 139893, upload-time = "2026-01-28T18:15:06.378Z" },
+    { url = "https://files.pythonhosted.org/packages/1c/61/8fa0e26f33623b49949346de05ec1ddaad02ed8ba64af45f40a147dbfa97/psutil-7.2.2-cp313-cp313t-win_arm64.whl", hash = "sha256:ae0aefdd8796a7737eccea863f80f81e468a1e4cf14d926bd9b6f5f2d5f90ca9", size = 135589, upload-time = "2026-01-28T18:15:08.03Z" },
+    { url = "https://files.pythonhosted.org/packages/81/69/ef179ab5ca24f32acc1dac0c247fd6a13b501fd5534dbae0e05a1c48b66d/psutil-7.2.2-cp314-cp314t-macosx_10_15_x86_64.whl", hash = "sha256:eed63d3b4d62449571547b60578c5b2c4bcccc5387148db46e0c2313dad0ee00", size = 130664, upload-time = "2026-01-28T18:15:09.469Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/64/665248b557a236d3fa9efc378d60d95ef56dd0a490c2cd37dafc7660d4a9/psutil-7.2.2-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:7b6d09433a10592ce39b13d7be5a54fbac1d1228ed29abc880fb23df7cb694c9", size = 131087, upload-time = "2026-01-28T18:15:11.724Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/2e/e6782744700d6759ebce3043dcfa661fb61e2fb752b91cdeae9af12c2178/psutil-7.2.2-cp314-cp314t-manylinux2010_x86_64.manylinux_2_12_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:1fa4ecf83bcdf6e6c8f4449aff98eefb5d0604bf88cb883d7da3d8d2d909546a", size = 182383, upload-time = "2026-01-28T18:15:13.445Z" },
+    { url = "https://files.pythonhosted.org/packages/57/49/0a41cefd10cb7505cdc04dab3eacf24c0c2cb158a998b8c7b1d27ee2c1f5/psutil-7.2.2-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:e452c464a02e7dc7822a05d25db4cde564444a67e58539a00f929c51eddda0cf", size = 185210, upload-time = "2026-01-28T18:15:16.002Z" },
+    { url = "https://files.pythonhosted.org/packages/dd/2c/ff9bfb544f283ba5f83ba725a3c5fec6d6b10b8f27ac1dc641c473dc390d/psutil-7.2.2-cp314-cp314t-win_amd64.whl", hash = "sha256:c7663d4e37f13e884d13994247449e9f8f574bc4655d509c3b95e9ec9e2b9dc1", size = 141228, upload-time = "2026-01-28T18:15:18.385Z" },
+    { url = "https://files.pythonhosted.org/packages/f2/fc/f8d9c31db14fcec13748d373e668bc3bed94d9077dbc17fb0eebc073233c/psutil-7.2.2-cp314-cp314t-win_arm64.whl", hash = "sha256:11fe5a4f613759764e79c65cf11ebdf26e33d6dd34336f8a337aa2996d71c841", size = 136284, upload-time = "2026-01-28T18:15:19.912Z" },
+    { url = "https://files.pythonhosted.org/packages/e7/36/5ee6e05c9bd427237b11b3937ad82bb8ad2752d72c6969314590dd0c2f6e/psutil-7.2.2-cp36-abi3-macosx_10_9_x86_64.whl", hash = "sha256:ed0cace939114f62738d808fdcecd4c869222507e266e574799e9c0faa17d486", size = 129090, upload-time = "2026-01-28T18:15:22.168Z" },
+    { url = "https://files.pythonhosted.org/packages/80/c4/f5af4c1ca8c1eeb2e92ccca14ce8effdeec651d5ab6053c589b074eda6e1/psutil-7.2.2-cp36-abi3-macosx_11_0_arm64.whl", hash = "sha256:1a7b04c10f32cc88ab39cbf606e117fd74721c831c98a27dc04578deb0c16979", size = 129859, upload-time = "2026-01-28T18:15:23.795Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/70/5d8df3b09e25bce090399cf48e452d25c935ab72dad19406c77f4e828045/psutil-7.2.2-cp36-abi3-manylinux2010_x86_64.manylinux_2_12_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:076a2d2f923fd4821644f5ba89f059523da90dc9014e85f8e45a5774ca5bc6f9", size = 155560, upload-time = "2026-01-28T18:15:25.976Z" },
+    { url = "https://files.pythonhosted.org/packages/63/65/37648c0c158dc222aba51c089eb3bdfa238e621674dc42d48706e639204f/psutil-7.2.2-cp36-abi3-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:b0726cecd84f9474419d67252add4ac0cd9811b04d61123054b9fb6f57df6e9e", size = 156997, upload-time = "2026-01-28T18:15:27.794Z" },
+    { url = "https://files.pythonhosted.org/packages/8e/13/125093eadae863ce03c6ffdbae9929430d116a246ef69866dad94da3bfbc/psutil-7.2.2-cp36-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:fd04ef36b4a6d599bbdb225dd1d3f51e00105f6d48a28f006da7f9822f2606d8", size = 148972, upload-time = "2026-01-28T18:15:29.342Z" },
+    { url = "https://files.pythonhosted.org/packages/04/78/0acd37ca84ce3ddffaa92ef0f571e073faa6d8ff1f0559ab1272188ea2be/psutil-7.2.2-cp36-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:b58fabe35e80b264a4e3bb23e6b96f9e45a3df7fb7eed419ac0e5947c61e47cc", size = 148266, upload-time = "2026-01-28T18:15:31.597Z" },
+    { url = "https://files.pythonhosted.org/packages/b4/90/e2159492b5426be0c1fef7acba807a03511f97c5f86b3caeda6ad92351a7/psutil-7.2.2-cp37-abi3-win_amd64.whl", hash = "sha256:eb7e81434c8d223ec4a219b5fc1c47d0417b12be7ea866e24fb5ad6e84b3d988", size = 137737, upload-time = "2026-01-28T18:15:33.849Z" },
+    { url = "https://files.pythonhosted.org/packages/8c/c7/7bb2e321574b10df20cbde462a94e2b71d05f9bbda251ef27d104668306a/psutil-7.2.2-cp37-abi3-win_arm64.whl", hash = "sha256:8c233660f575a5a89e6d4cb65d9f938126312bca76d8fe087b947b3a1aaac9ee", size = 134617, upload-time = "2026-01-28T18:15:36.514Z" },
+]
+
+[[package]]
+name = "psycopg"
+version = "3.3.4"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+    { name = "tzdata", marker = "sys_platform == 'win32'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/db/2f/cb91e5502ec9de1de6f1b76cfbf69531932725361168bb06963620c77e2e/psycopg-3.3.4.tar.gz", hash = "sha256:e21207764952cff81b6b8bdacad9a3939f2793367fdac2987b3aac36a651b5bc", size = 165799, upload-time = "2026-05-01T23:31:55.179Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/5c/e0/7b3dee031daae7743609ce3c746565d4a3ed7c2c186479eb48e34e838c64/psycopg-3.3.4-py3-none-any.whl", hash = "sha256:b6bbc25ccf05c8fad3b061d9db2ef0909a555171b84b07f29458a447253d679a", size = 213001, upload-time = "2026-05-01T23:20:50.816Z" },
+]
+
+[package.optional-dependencies]
+binary = [
+    { name = "psycopg-binary", marker = "implementation_name != 'pypy'" },
+]
+pool = [
+    { name = "psycopg-pool" },
+]
+
+[[package]]
+name = "psycopg-binary"
+version = "3.3.4"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/95/7d/03818e13ba7f36de93573c93ee3482006d3dfa8b0f8d28df511bad0a1a92/psycopg_binary-3.3.4-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:5ab28a2a7649df3b72e6b674b4c190e448e8e77cf496a65bd846472048de2089", size = 4591122, upload-time = "2026-05-01T23:27:56.162Z" },
+    { url = "https://files.pythonhosted.org/packages/a5/b9/11b341edf8d54e2694726b273fe9652b254d989f4f63e3ac6816ad6b55f4/psycopg_binary-3.3.4-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:6402a9d8146cf4b3974ded3fd28a971e83dc6a0333eb7822524a3aa20b546578", size = 4669943, upload-time = "2026-05-01T23:28:04.522Z" },
+    { url = "https://files.pythonhosted.org/packages/8b/18/4665bacd65e7865b4372fcd8abb8b9186ada4b0025f8c2ca691b364a556c/psycopg_binary-3.3.4-cp312-cp312-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:580ae30a5f95ccd90008ec697d3ed6a4a2047a516407ad904283fa42086936e9", size = 5469697, upload-time = "2026-05-01T23:28:11.337Z" },
+    { url = "https://files.pythonhosted.org/packages/7c/b1/b83136c6e510593d9b0c759ba5384337bc4ad82d19fda675adc4b2703c84/psycopg_binary-3.3.4-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:e7510c37550f91a187e3660a8cc50d4b760f8c3b8b2f89ebc5698cd2c7f2c85d", size = 5152995, upload-time = "2026-05-01T23:28:20.529Z" },
+    { url = "https://files.pythonhosted.org/packages/67/8d/a9821e2a648afe6091989929982a3b0f00b2631a859cb81379728f08fb75/psycopg_binary-3.3.4-cp312-cp312-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:77df19583501ea288eaf15ac0fe7ad01e6d8091a91d5c41df5c718f307d8e31b", size = 6738180, upload-time = "2026-05-01T23:28:30.654Z" },
+    { url = "https://files.pythonhosted.org/packages/7e/58/2e349e8d23905dc2317b80ac65f48fb6f821a4777a4e994a60da91c4850f/psycopg_binary-3.3.4-cp312-cp312-manylinux_2_38_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:018fbed325936da502feb546642c982dcc4b9ffdea32dfef78dbf3b7f7ad4070", size = 4978828, upload-time = "2026-05-01T23:28:37.277Z" },
+    { url = "https://files.pythonhosted.org/packages/45/48/57b00d03b4721878326122a1f1e6b0a90b85bcaec56b5b2f8ea6cfa45235/psycopg_binary-3.3.4-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:17a21953a9e5ff3a16dab692625a3676e2f101db5e40072f39dbee2250194d68", size = 4509757, upload-time = "2026-05-01T23:28:43.078Z" },
+    { url = "https://files.pythonhosted.org/packages/25/37/33b47d8c007df69aec500df5889767c4d313748e8e9e27a2fef8a6dabcee/psycopg_binary-3.3.4-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:eb05ee1c2b817d27c537333224c9e83c7afb86fe7296ba970990068baf819b16", size = 4190546, upload-time = "2026-05-01T23:28:50.016Z" },
+    { url = "https://files.pythonhosted.org/packages/ca/c6/32b0835dbc2122617902b649d76a91c1e75406e76bf3d595b0c3bb5ffad6/psycopg_binary-3.3.4-cp312-cp312-musllinux_1_2_riscv64.whl", hash = "sha256:773d573e11f437ce0bdb95b7c18dc58390494f96d43f8b45b9760436114f7652", size = 3926197, upload-time = "2026-05-01T23:28:55.55Z" },
+    { url = "https://files.pythonhosted.org/packages/cd/68/d190ef0c0c5b16ded07831dabc8ddd412f4cdab07ec6e30ed38d9bda0e1f/psycopg_binary-3.3.4-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:71e55ccbdfae79a2ed9c6369c3008a3025817ff9d7e27b32a2d84e2a4267e66e", size = 4236627, upload-time = "2026-05-01T23:29:05.336Z" },
+    { url = "https://files.pythonhosted.org/packages/25/8f/81dcbc2e8454b74d14881275ea45f00791052dac531a9fa8be1730d1685b/psycopg_binary-3.3.4-cp312-cp312-win_amd64.whl", hash = "sha256:494ca54901be8cf9eb7e02c25b731f2317c378efa44f43e8f9bd0e1184ae7be4", size = 3560782, upload-time = "2026-05-01T23:29:11.967Z" },
+    { url = "https://files.pythonhosted.org/packages/09/43/13e9c406fbbf354580476e248a16b64802a376873ebe6339e30bb655572d/psycopg_binary-3.3.4-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:fbd1d4ed566895ad2d3bf4ddfd8bae90026930ddf29df3b9d91d32c8c47866a7", size = 4590377, upload-time = "2026-05-01T23:29:18.782Z" },
+    { url = "https://files.pythonhosted.org/packages/22/be/2923cd7c3683e7afdecf4f10796a18de02f5c5ddc0969aa2ad0a8cdd3bbd/psycopg_binary-3.3.4-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:75a9067e236f9b9ae3535b66fe99bddb33d39c0de10112e49b9ab11eee53dc31", size = 4669023, upload-time = "2026-05-01T23:29:25.884Z" },
+    { url = "https://files.pythonhosted.org/packages/96/a0/2c913d6fe13d6a8bd13597d36739bf47af063ad9399e402cfecab16f3c1e/psycopg_binary-3.3.4-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:b56b603ebcea8aa10b46228b8410ba7f13e7c2ee54389d4d9be0927fd8ce2a70", size = 5467423, upload-time = "2026-05-01T23:29:33.416Z" },
+    { url = "https://files.pythonhosted.org/packages/e7/38/205d10bc1ad0df4a21c5c51659126bd3ea0ef98fcad1e852f78c249bb9c3/psycopg_binary-3.3.4-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:c677c4ad433cb7150c8cd304a0769ae3bcfbe5ea0676eb53faa7b1443b16d0d3", size = 5151137, upload-time = "2026-05-01T23:29:42.013Z" },
+    { url = "https://files.pythonhosted.org/packages/36/fc/f0381ddcd45eff3bb70dbca6823a996048d7f507b2ec3fc92c6fabc0fe87/psycopg_binary-3.3.4-cp313-cp313-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:26df2717e59c0473e4465a97dfb1b7afebaa479277870fd5784d1436470db47c", size = 6736671, upload-time = "2026-05-01T23:29:51.626Z" },
+    { url = "https://files.pythonhosted.org/packages/95/40/fa545ae152c24327651e5624e4902121e808270be36c10b12e9939be09bc/psycopg_binary-3.3.4-cp313-cp313-manylinux_2_38_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:1dc1f79fd16bb1f3f4421417a514607539f17804d95c7ed617265369d1981cae", size = 4979601, upload-time = "2026-05-01T23:29:56.961Z" },
+    { url = "https://files.pythonhosted.org/packages/86/e4/2f8a47ee97f90cd2b933d0463081d35631ff419de2b8c984a5f369857de0/psycopg_binary-3.3.4-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:136f199a407b5348b9b857c504aff60c77622a28482e7195839ce1b51238c4cc", size = 4510513, upload-time = "2026-05-01T23:30:07.243Z" },
+    { url = "https://files.pythonhosted.org/packages/0e/0e/94e842ff4a7f98ed162580ca2e8b8864b28c1e0350f2443f8ee47f821167/psycopg_binary-3.3.4-cp313-cp313-musllinux_1_2_ppc64le.whl", hash = "sha256:b6f5a29e9c775b9f12a1a717aa7a2c80f9e1db6f27ba44a5b59c80ac61d2ffcf", size = 4187243, upload-time = "2026-05-01T23:30:15.352Z" },
+    { url = "https://files.pythonhosted.org/packages/d0/83/fc6c174b672e29b7de996ea77b6cbddf46c891751c3355f6974292baa6b4/psycopg_binary-3.3.4-cp313-cp313-musllinux_1_2_riscv64.whl", hash = "sha256:ee17a2cf4943cde261adfad1bbc5bf38d6b3776d7afff74c7cabcbeaeb08c260", size = 3927347, upload-time = "2026-05-01T23:30:21.186Z" },
+    { url = "https://files.pythonhosted.org/packages/e9/65/768364d4a97a15b1a7f47ba52688c1686f22941d8332a8398cefc468e25f/psycopg_binary-3.3.4-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:5c4ab71be17bdca30cb34c34c4e1496e2f5d6f20c199c12bad226070b22ef9bf", size = 4236393, upload-time = "2026-05-01T23:30:26.211Z" },
+    { url = "https://files.pythonhosted.org/packages/bd/3b/218efbc9e645becd80cdf651acda05f85cfe546b7a9c0458c7cbc8fe1f74/psycopg_binary-3.3.4-cp313-cp313-win_amd64.whl", hash = "sha256:dbfdb9b6cc79f31104a7b162a2b921b765fcc62af6c00540a167a8de47e4ed38", size = 3564592, upload-time = "2026-05-01T23:30:31.764Z" },
+    { url = "https://files.pythonhosted.org/packages/48/a6/828c9185701dab71b234c2a76c38a08b098ebfec5020716b4e93807492b5/psycopg_binary-3.3.4-cp314-cp314-macosx_10_15_x86_64.whl", hash = "sha256:28b7398fdd19db3232c884fb24550bdfe951221f510e195e233299e4c9b78f97", size = 4607292, upload-time = "2026-05-01T23:30:38.962Z" },
+    { url = "https://files.pythonhosted.org/packages/92/58/5b40dbc9d839045c9dae956960e4fb6d20bcabe6c59a2aa34fc3a371913f/psycopg_binary-3.3.4-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:1fbaa292a3c8bb61b45df1ad3da1908ccee7cb889db9425e3557d9e34e2a4829", size = 4687023, upload-time = "2026-05-01T23:30:47.227Z" },
+    { url = "https://files.pythonhosted.org/packages/85/a9/793f0ac107a9003b48441d0d1f9f616d96e0f37458dd8dc12528ceff55fb/psycopg_binary-3.3.4-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:94596f9e7633ee3f6440711d43bb70aa31cc0a46a900ab8b4201a366ace5c9e7", size = 5486985, upload-time = "2026-05-01T23:30:55.517Z" },
+    { url = "https://files.pythonhosted.org/packages/8f/26/42e8533497e2592334f68ec529cf5f840f7fa4e99575a4bb61aa184dbfbf/psycopg_binary-3.3.4-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:8c0056529e68dbe9184cd4019a1f3d8f3a4ead2f6fc7a5afcf27d3314edd1277", size = 5168745, upload-time = "2026-05-01T23:31:01.904Z" },
+    { url = "https://files.pythonhosted.org/packages/15/af/b7151776cc08d5935d45c833ec818a9beb417cf7c08239af1aafbdae78ee/psycopg_binary-3.3.4-cp314-cp314-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:2c09aad7051326e7603c14e50636db9c01f78272dc54b3accff03d46370461e6", size = 6761486, upload-time = "2026-05-01T23:31:14.511Z" },
+    { url = "https://files.pythonhosted.org/packages/d0/ed/c92533b9124712d592cbf1cd6c76da933a2e0acea81dfe1fbe7e735f0cff/psycopg_binary-3.3.4-cp314-cp314-manylinux_2_38_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:514404ed543efd620c85602b747df2a23cf1241b4067199e1a66f2d2757aaa41", size = 4997427, upload-time = "2026-05-01T23:31:20.901Z" },
+    { url = "https://files.pythonhosted.org/packages/a2/23/ccadfd0de416aa188356daa199453af24087b042e296088706d190ae0295/psycopg_binary-3.3.4-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:46893c26858be12cc49ca4226ed6a60b4bfccadd946b3bebb783a60b38788228", size = 4533549, upload-time = "2026-05-01T23:31:26.204Z" },
+    { url = "https://files.pythonhosted.org/packages/fd/a0/c8f43cee36386f7bc891ab41a9d31ea07cf9826038e732da79f26b1e5f34/psycopg_binary-3.3.4-cp314-cp314-musllinux_1_2_ppc64le.whl", hash = "sha256:df1d567fc430f6df15c9fcf67d87685fc49bdb325adc0db5af1adfb2f44eb5c9", size = 4210256, upload-time = "2026-05-01T23:31:33.884Z" },
+    { url = "https://files.pythonhosted.org/packages/4e/2c/c1547871be3790676e8868b38655496422f94f0978dfb66b74bdba2f1676/psycopg_binary-3.3.4-cp314-cp314-musllinux_1_2_riscv64.whl", hash = "sha256:6b9016b1714da4dd5ecaaa75b82098aa5a0b87854ce9b092e21c27c4ae23e014", size = 3946204, upload-time = "2026-05-01T23:31:39.626Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/b1/f6670f00fa7ea601584623f6c11602ab92117d83eaff885e0210f6de7418/psycopg_binary-3.3.4-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:47c656a8a7ba6eb0cff1801a4caaa9c8bdc12d03080e273aff1c8ac39971a77e", size = 4255811, upload-time = "2026-05-01T23:31:44.986Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/e6/5fff07a70d1f945ed90ae131c3bd76cab32beff7c58c6db15ad5820b6d1f/psycopg_binary-3.3.4-cp314-cp314-win_amd64.whl", hash = "sha256:c37e024c07308cd06cf3ec51bfd0e7f6157585a4d84d1bce4a7f5f7913719bf8", size = 3666849, upload-time = "2026-05-01T23:31:51.165Z" },
+]
+
+[[package]]
+name = "psycopg-pool"
+version = "3.3.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/90/82/7a23d26039827ecd4ebe93905651029ddd307c5182ad59296dfb6f67b528/psycopg_pool-3.3.1.tar.gz", hash = "sha256:b10b10b7a175d5cc1592147dc5b7eec8a9e0834eb3ed2c4a92c858e2f51eb63c", size = 31661, upload-time = "2026-05-01T23:31:59.809Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/37/ed/89c2c620af0e1660354cd8aabf9f5b21f911597ce22acb37c805d6c86bc8/psycopg_pool-3.3.1-py3-none-any.whl", hash = "sha256:2af5b432941c4c9ad5c87b3fa410aec910ec8f7c122855897983a06c45f2e4b5", size = 40023, upload-time = "2026-05-01T23:31:53.136Z" },
+]
+
+[[package]]
+name = "psycopg2-binary"
+version = "2.9.12"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/2a/60/a3624f79acea344c16fbef3a94d28b89a8042ddfb8f3e4ca83f538671409/psycopg2_binary-2.9.12.tar.gz", hash = "sha256:5ac9444edc768c02a6b6a591f070b8aae28ff3a99be57560ac996001580f294c", size = 379686, upload-time = "2026-04-21T09:40:34.304Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e2/9f/ef4ef3c8e15083df90ca35265cfd1a081a2f0cc07bb229c6314c6af817f4/psycopg2_binary-2.9.12-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:5cdc05117180c5fa9c40eea8ea559ce64d73824c39d928b7da9fb5f6a9392433", size = 3712459, upload-time = "2026-04-20T23:34:30.549Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/01/3dd14e46ba48c1e1a6ec58ee599fa1b5efa00c246d5046cd903d0eeb1af1/psycopg2_binary-2.9.12-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:d3227a3bc228c10d21011a99245edca923e4e8bf461857e869a507d9a41fe9f6", size = 3822936, upload-time = "2026-04-20T23:34:32.77Z" },
+    { url = "https://files.pythonhosted.org/packages/a6/f7/0640e4901119d8a9f7a1784b927f494e2198e213ceb593753d1f2c8b1b30/psycopg2_binary-2.9.12-cp312-cp312-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:995ce929eede89db6254b50827e2b7fd61e50d11f0b116b29fffe4a2e53c4580", size = 4578676, upload-time = "2026-04-20T23:34:35.18Z" },
+    { url = "https://files.pythonhosted.org/packages/b0/55/44df3965b5f297c50cc0b1b594a31c67d6127a9d133045b8a66611b14dfb/psycopg2_binary-2.9.12-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:9fe06d93e72f1c048e731a2e3e7854a5bfaa58fc736068df90b352cefe66f03f", size = 4274917, upload-time = "2026-04-20T23:34:37.982Z" },
+    { url = "https://files.pythonhosted.org/packages/b0/4b/74535248b1eac0c9336862e8617c765ac94dac76f9e25d7c4a79588c8907/psycopg2_binary-2.9.12-cp312-cp312-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:40e7b28b63aaf737cb3a1edc3a9bbc9a9f4ad3dcb7152e8c1130e4050eddcb7d", size = 5894843, upload-time = "2026-04-20T23:34:40.856Z" },
+    { url = "https://files.pythonhosted.org/packages/f2/ba/f1bf8d2ae71868ad800b661099086ee52bc0f8d9f05be1acd8ebb06757cc/psycopg2_binary-2.9.12-cp312-cp312-manylinux_2_38_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:89d19a9f7899e8eb0656a2b3a08e0da04c720a06db6e0033eab5928aabe60fa9", size = 4110556, upload-time = "2026-04-20T23:34:44.016Z" },
+    { url = "https://files.pythonhosted.org/packages/45/46/c15706c338403b7c420bcc0c2905aad116cc064545686d8bf85f1999ea00/psycopg2_binary-2.9.12-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:612b965daee295ae2da8f8218ce1d274645dc76ef3f1abf6a0a94fd57eff876d", size = 3655714, upload-time = "2026-04-20T23:34:46.233Z" },
+    { url = "https://files.pythonhosted.org/packages/b3/7c/a2d5dc09b64a4564db242a0fe418fde7d33f6f8259dd2c5b9d7def00fb5a/psycopg2_binary-2.9.12-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:b9a339b79d37c1b45f3235265f07cdeb0cb5ad7acd2ac7720a5920989c17c24e", size = 3301154, upload-time = "2026-04-20T23:34:49.528Z" },
+    { url = "https://files.pythonhosted.org/packages/c0/e8/cc8c9a4ce71461f9ec548d38cadc41dc184b34c73e6455450775a9334ccd/psycopg2_binary-2.9.12-cp312-cp312-musllinux_1_2_riscv64.whl", hash = "sha256:3471336e1acfd9c7fe507b8bad5af9317b6a89294f9eb37bd9a030bb7bebcdc6", size = 3048882, upload-time = "2026-04-20T23:34:51.86Z" },
+    { url = "https://files.pythonhosted.org/packages/19/6a/31e2296bc0787c5ab75d3d118e40b239db8151b5192b90b77c72bc9256e9/psycopg2_binary-2.9.12-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:7af18183109e23502c8b2ae7f6926c0882766f35b5175a4cd737ad825e4d7a1b", size = 3351298, upload-time = "2026-04-20T23:34:54.124Z" },
+    { url = "https://files.pythonhosted.org/packages/5f/a8/75f4e3e11203b590150abed2cf7794b9c9c9f7eceddae955191138b44dde/psycopg2_binary-2.9.12-cp312-cp312-win_amd64.whl", hash = "sha256:398fcd4db988c7d7d3713e2b8e18939776fd3fb447052daae4f24fa39daede4c", size = 2757230, upload-time = "2026-04-20T23:34:56.242Z" },
+    { url = "https://files.pythonhosted.org/packages/91/bb/4608c96f970f6e0c56572e87027ef4404f709382a3503e9934526d7ba051/psycopg2_binary-2.9.12-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:7c729a73c7b1b84de3582f73cdd27d905121dc2c531f3d9a3c32a3011033b965", size = 3712419, upload-time = "2026-04-20T23:34:58.754Z" },
+    { url = "https://files.pythonhosted.org/packages/5e/af/48f76af9d50d61cf390f8cd657b503168b089e2e9298e48465d029fcc713/psycopg2_binary-2.9.12-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:4413d0caef93c5cf50b96863df4c2efe8c269bf2267df353225595e7e15e8df7", size = 3822990, upload-time = "2026-04-20T23:35:00.821Z" },
+    { url = "https://files.pythonhosted.org/packages/7a/df/aba0f99397cd811d32e06fc0cc781f1f3ce98bc0e729cb423925085d781a/psycopg2_binary-2.9.12-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:4dfcf8e45ebb0c663be34a3442f65e17311f3367089cd4e5e3a3e8e62c978777", size = 4578696, upload-time = "2026-04-20T23:35:03.409Z" },
+    { url = "https://files.pythonhosted.org/packages/95/9c/eaa74021ac4e4d5c2f83d82fc6615a63f4fe6c94dc4e94c3990427053f67/psycopg2_binary-2.9.12-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:c41321a14dd74aceb6a9a643b9253a334521babfa763fa873e33d89cfa122fb5", size = 4274982, upload-time = "2026-04-20T23:35:05.583Z" },
+    { url = "https://files.pythonhosted.org/packages/35/ed/c25deff98bd26187ba48b3b250a3ffc3037c46c5b89362534a15d200e0db/psycopg2_binary-2.9.12-cp313-cp313-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:83946ba43979ebfdc99a3cd0ee775c89f221df026984ba19d46133d8d75d3cd9", size = 5894867, upload-time = "2026-04-20T23:35:07.902Z" },
+    { url = "https://files.pythonhosted.org/packages/9a/81/8d0e21ca77373c6c9589e5c4528f6e8f0c08c62cafc76fb0bddb7a2cee22/psycopg2_binary-2.9.12-cp313-cp313-manylinux_2_38_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:411e85815652d13560fbe731878daa5d92378c4995a22302071890ec3397d019", size = 4110578, upload-time = "2026-04-20T23:35:10.149Z" },
+    { url = "https://files.pythonhosted.org/packages/00/fc/f481e2435bd8f742d0123309174aae4165160ad3ef17c1b99c3622c241d2/psycopg2_binary-2.9.12-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:1c8ad4c08e00f7679559eaed7aff1edfffc60c086b976f93972f686384a95e2c", size = 3655816, upload-time = "2026-04-20T23:35:12.56Z" },
+    { url = "https://files.pythonhosted.org/packages/53/79/b9f46466bdbe9f239c96cde8be33c1aace4842f06013b47b730dc9759187/psycopg2_binary-2.9.12-cp313-cp313-musllinux_1_2_ppc64le.whl", hash = "sha256:00814e40fa23c2b37ef0a1e3c749d89982c73a9cb5046137f0752a22d432e82f", size = 3301307, upload-time = "2026-04-20T23:35:15.029Z" },
+    { url = "https://files.pythonhosted.org/packages/3f/19/7dc003b32fe35024df89b658104f7c8538a8b2dcbde7a4e746ce929742e7/psycopg2_binary-2.9.12-cp313-cp313-musllinux_1_2_riscv64.whl", hash = "sha256:98062447aebc20ed20add1f547a364fd0ef8933640d5372ff1873f8deb9b61be", size = 3048968, upload-time = "2026-04-20T23:35:16.757Z" },
+    { url = "https://files.pythonhosted.org/packages/91/58/2dbd7db5c604d45f4950d988506aae672a14126ec22998ced5021cbb76bb/psycopg2_binary-2.9.12-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:66a7685d7e548f10fb4ce32fb01a7b7f4aa702134de92a292c7bd9e0d3dbd290", size = 3351369, upload-time = "2026-04-20T23:35:18.933Z" },
+    { url = "https://files.pythonhosted.org/packages/42/ee/dee8dcaad07f735824de3d6563bc67119fa6c28257b17977a8d624f02fab/psycopg2_binary-2.9.12-cp313-cp313-win_amd64.whl", hash = "sha256:b6937f5fe4e180aeee87de907a2fa982ded6f7f15d7218f78a083e4e1d68f2a0", size = 2757347, upload-time = "2026-04-20T23:35:21.283Z" },
+    { url = "https://files.pythonhosted.org/packages/13/1b/708c0dca874acfad6d65314271859899a79007686f3a1f74e82a2ed4b645/psycopg2_binary-2.9.12-cp314-cp314-macosx_10_15_x86_64.whl", hash = "sha256:6f3b3de8a74ef8db215f22edffb19e32dc6fa41340456de7ec99efdc8a7b3ec2", size = 3712428, upload-time = "2026-04-20T23:35:23.453Z" },
+    { url = "https://files.pythonhosted.org/packages/d6/39/ddbea9d4b4de6aca9431b6ed253f530f8a02d3b8f9bcfd0dbfe2b3de6fe4/psycopg2_binary-2.9.12-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:1006fb62f0f0bc5ce256a832356c6262e91be43f5e4eb15b5eaf38079464caf2", size = 3823184, upload-time = "2026-04-20T23:35:25.92Z" },
+    { url = "https://files.pythonhosted.org/packages/bf/a0/bc2fef74b106fa345567122a0659e6d94512ed7dc0131ec44c9e5aba3725/psycopg2_binary-2.9.12-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:840066105706cd2eb29b9a1c2329620056582a4bf3e8169dec5c447042d0869f", size = 4579157, upload-time = "2026-04-20T23:35:28.542Z" },
+    { url = "https://files.pythonhosted.org/packages/57/d7/d4e3b2005d3de607ca4fbb0e8742e248056e52184a6b94ebda3c1c2c329b/psycopg2_binary-2.9.12-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:863f5d12241ebe1c76a72a04c2113b6dc905f90b9cef0e9be0efd994affd9354", size = 4274970, upload-time = "2026-04-20T23:35:30.418Z" },
+    { url = "https://files.pythonhosted.org/packages/2e/42/c9853f8db3967fe08bcde11f53d53b85d351750cae726ce001cb68afa9c1/psycopg2_binary-2.9.12-cp314-cp314-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:a99eaab34a9010f1a086b126de467466620a750634d114d20455f3a824aae033", size = 5895175, upload-time = "2026-04-20T23:35:33.584Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/fd/b82b5601a97630308bef079f545ffec481bbbc795c2ba5ec416a01d03f60/psycopg2_binary-2.9.12-cp314-cp314-manylinux_2_38_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:ffdd7dc5463ccd61845ac37b7012d0f35a1548df9febe14f8dd549be4a0bc81e", size = 4110658, upload-time = "2026-04-20T23:35:35.638Z" },
+    { url = "https://files.pythonhosted.org/packages/62/8c/32ca69b0389ef25dd22937bf9e8fbe2ce27aea20b05ded48c4ce4cb42475/psycopg2_binary-2.9.12-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:54a0dfecab1b48731f934e06139dfe11e24219fb6d0ceb32177cf0375f14c7b5", size = 3656251, upload-time = "2026-04-20T23:35:37.854Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/29/96992a2b59e3b9d730fcf9612d0a387305025dc867a9fc490a9e496e074e/psycopg2_binary-2.9.12-cp314-cp314-musllinux_1_2_ppc64le.whl", hash = "sha256:96937c9c5d891f772430f418a7a8b4691a90c3e6b93cf72b5bd7cad8cbca32a5", size = 3301810, upload-time = "2026-04-20T23:35:39.927Z" },
+    { url = "https://files.pythonhosted.org/packages/56/ad/44b06659949b243ae10112cd3b20a197f9bf3e81d5651379b9eb889bfaad/psycopg2_binary-2.9.12-cp314-cp314-musllinux_1_2_riscv64.whl", hash = "sha256:77b348775efd4cdab410ec6609d81ccecd1139c90265fa583a7255c8064bc03d", size = 3048977, upload-time = "2026-04-20T23:35:41.806Z" },
+    { url = "https://files.pythonhosted.org/packages/1d/f2/10a1bcebadb6aa55e280e1f58975c36a7b560ea525184c7aa4064c466633/psycopg2_binary-2.9.12-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:527e6342b3e44c2f0544f6b8e927d60de7f163f5723b8f1dfa7d2a84298738cd", size = 3351466, upload-time = "2026-04-20T23:35:43.993Z" },
+    { url = "https://files.pythonhosted.org/packages/20/be/b732c8418ffa5bcfda002890f5dc4c869fc17db66ff11f53b17cfe44afc0/psycopg2_binary-2.9.12-cp314-cp314-win_amd64.whl", hash = "sha256:f12ae41fcafadb39b2785e64a40f9db05d6de2ac114077457e0e7c597f3af980", size = 2848762, upload-time = "2026-04-20T23:35:46.421Z" },
+]
+
+[[package]]
+name = "ptyprocess"
+version = "0.7.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/20/e5/16ff212c1e452235a90aeb09066144d0c5a6a8c0834397e03f5224495c4e/ptyprocess-0.7.0.tar.gz", hash = "sha256:5c5d0a3b48ceee0b48485e0c26037c0acd7d29765ca3fbb5cb3831d347423220", size = 70762, upload-time = "2020-12-28T15:15:30.155Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/22/a6/858897256d0deac81a172289110f31629fc4cee19b6f01283303e18c8db3/ptyprocess-0.7.0-py2.py3-none-any.whl", hash = "sha256:4b41f3967fce3af57cc7e94b888626c18bf37a083e3651ca8feeb66d492fef35", size = 13993, upload-time = "2020-12-28T15:15:28.35Z" },
+]
+
+[[package]]
+name = "pure-eval"
+version = "0.2.3"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/cd/05/0a34433a064256a578f1783a10da6df098ceaa4a57bbeaa96a6c0352786b/pure_eval-0.2.3.tar.gz", hash = "sha256:5f4e983f40564c576c7c8635ae88db5956bb2229d7e9237d03b3c0b0190eaf42", size = 19752, upload-time = "2024-07-21T12:58:21.801Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/8e/37/efad0257dc6e593a18957422533ff0f87ede7c9c6ea010a2177d738fb82f/pure_eval-0.2.3-py3-none-any.whl", hash = "sha256:1db8e35b67b3d218d818ae653e27f06c3aa420901fa7b081ca98cbedc874e0d0", size = 11842, upload-time = "2024-07-21T12:58:20.04Z" },
+]
+
+[[package]]
+name = "pycparser"
+version = "3.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/1b/7d/92392ff7815c21062bea51aa7b87d45576f649f16458d78b7cf94b9ab2e6/pycparser-3.0.tar.gz", hash = "sha256:600f49d217304a5902ac3c37e1281c9fe94e4d0489de643a9504c5cdfdfc6b29", size = 103492, upload-time = "2026-01-21T14:26:51.89Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/0c/c3/44f3fbbfa403ea2a7c779186dc20772604442dde72947e7d01069cbe98e3/pycparser-3.0-py3-none-any.whl", hash = "sha256:b727414169a36b7d524c1c3e31839a521725078d7b2ff038656844266160a992", size = 48172, upload-time = "2026-01-21T14:26:50.693Z" },
+]
+
+[[package]]
+name = "pycryptodome"
+version = "3.23.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/8e/a6/8452177684d5e906854776276ddd34eca30d1b1e15aa1ee9cefc289a33f5/pycryptodome-3.23.0.tar.gz", hash = "sha256:447700a657182d60338bab09fdb27518f8856aecd80ae4c6bdddb67ff5da44ef", size = 4921276, upload-time = "2025-05-17T17:21:45.242Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/04/5d/bdb09489b63cd34a976cc9e2a8d938114f7a53a74d3dd4f125ffa49dce82/pycryptodome-3.23.0-cp313-cp313t-macosx_10_13_universal2.whl", hash = "sha256:0011f7f00cdb74879142011f95133274741778abba114ceca229adbf8e62c3e4", size = 2495152, upload-time = "2025-05-17T17:20:20.833Z" },
+    { url = "https://files.pythonhosted.org/packages/a7/ce/7840250ed4cc0039c433cd41715536f926d6e86ce84e904068eb3244b6a6/pycryptodome-3.23.0-cp313-cp313t-macosx_10_13_x86_64.whl", hash = "sha256:90460fc9e088ce095f9ee8356722d4f10f86e5be06e2354230a9880b9c549aae", size = 1639348, upload-time = "2025-05-17T17:20:23.171Z" },
+    { url = "https://files.pythonhosted.org/packages/ee/f0/991da24c55c1f688d6a3b5a11940567353f74590734ee4a64294834ae472/pycryptodome-3.23.0-cp313-cp313t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:4764e64b269fc83b00f682c47443c2e6e85b18273712b98aa43bcb77f8570477", size = 2184033, upload-time = "2025-05-17T17:20:25.424Z" },
+    { url = "https://files.pythonhosted.org/packages/54/16/0e11882deddf00f68b68dd4e8e442ddc30641f31afeb2bc25588124ac8de/pycryptodome-3.23.0-cp313-cp313t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:eb8f24adb74984aa0e5d07a2368ad95276cf38051fe2dc6605cbcf482e04f2a7", size = 2270142, upload-time = "2025-05-17T17:20:27.808Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/fc/4347fea23a3f95ffb931f383ff28b3f7b1fe868739182cb76718c0da86a1/pycryptodome-3.23.0-cp313-cp313t-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:d97618c9c6684a97ef7637ba43bdf6663a2e2e77efe0f863cce97a76af396446", size = 2309384, upload-time = "2025-05-17T17:20:30.765Z" },
+    { url = "https://files.pythonhosted.org/packages/6e/d9/c5261780b69ce66d8cfab25d2797bd6e82ba0241804694cd48be41add5eb/pycryptodome-3.23.0-cp313-cp313t-musllinux_1_2_aarch64.whl", hash = "sha256:9a53a4fe5cb075075d515797d6ce2f56772ea7e6a1e5e4b96cf78a14bac3d265", size = 2183237, upload-time = "2025-05-17T17:20:33.736Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/6f/3af2ffedd5cfa08c631f89452c6648c4d779e7772dfc388c77c920ca6bbf/pycryptodome-3.23.0-cp313-cp313t-musllinux_1_2_i686.whl", hash = "sha256:763d1d74f56f031788e5d307029caef067febf890cd1f8bf61183ae142f1a77b", size = 2343898, upload-time = "2025-05-17T17:20:36.086Z" },
+    { url = "https://files.pythonhosted.org/packages/9a/dc/9060d807039ee5de6e2f260f72f3d70ac213993a804f5e67e0a73a56dd2f/pycryptodome-3.23.0-cp313-cp313t-musllinux_1_2_x86_64.whl", hash = "sha256:954af0e2bd7cea83ce72243b14e4fb518b18f0c1649b576d114973e2073b273d", size = 2269197, upload-time = "2025-05-17T17:20:38.414Z" },
+    { url = "https://files.pythonhosted.org/packages/f9/34/e6c8ca177cb29dcc4967fef73f5de445912f93bd0343c9c33c8e5bf8cde8/pycryptodome-3.23.0-cp313-cp313t-win32.whl", hash = "sha256:257bb3572c63ad8ba40b89f6fc9d63a2a628e9f9708d31ee26560925ebe0210a", size = 1768600, upload-time = "2025-05-17T17:20:40.688Z" },
+    { url = "https://files.pythonhosted.org/packages/e4/1d/89756b8d7ff623ad0160f4539da571d1f594d21ee6d68be130a6eccb39a4/pycryptodome-3.23.0-cp313-cp313t-win_amd64.whl", hash = "sha256:6501790c5b62a29fcb227bd6b62012181d886a767ce9ed03b303d1f22eb5c625", size = 1799740, upload-time = "2025-05-17T17:20:42.413Z" },
+    { url = "https://files.pythonhosted.org/packages/5d/61/35a64f0feaea9fd07f0d91209e7be91726eb48c0f1bfc6720647194071e4/pycryptodome-3.23.0-cp313-cp313t-win_arm64.whl", hash = "sha256:9a77627a330ab23ca43b48b130e202582e91cc69619947840ea4d2d1be21eb39", size = 1703685, upload-time = "2025-05-17T17:20:44.388Z" },
+    { url = "https://files.pythonhosted.org/packages/db/6c/a1f71542c969912bb0e106f64f60a56cc1f0fabecf9396f45accbe63fa68/pycryptodome-3.23.0-cp37-abi3-macosx_10_9_universal2.whl", hash = "sha256:187058ab80b3281b1de11c2e6842a357a1f71b42cb1e15bce373f3d238135c27", size = 2495627, upload-time = "2025-05-17T17:20:47.139Z" },
+    { url = "https://files.pythonhosted.org/packages/6e/4e/a066527e079fc5002390c8acdd3aca431e6ea0a50ffd7201551175b47323/pycryptodome-3.23.0-cp37-abi3-macosx_10_9_x86_64.whl", hash = "sha256:cfb5cd445280c5b0a4e6187a7ce8de5a07b5f3f897f235caa11f1f435f182843", size = 1640362, upload-time = "2025-05-17T17:20:50.392Z" },
+    { url = "https://files.pythonhosted.org/packages/50/52/adaf4c8c100a8c49d2bd058e5b551f73dfd8cb89eb4911e25a0c469b6b4e/pycryptodome-3.23.0-cp37-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:67bd81fcbe34f43ad9422ee8fd4843c8e7198dd88dd3d40e6de42ee65fbe1490", size = 2182625, upload-time = "2025-05-17T17:20:52.866Z" },
+    { url = "https://files.pythonhosted.org/packages/5f/e9/a09476d436d0ff1402ac3867d933c61805ec2326c6ea557aeeac3825604e/pycryptodome-3.23.0-cp37-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:c8987bd3307a39bc03df5c8e0e3d8be0c4c3518b7f044b0f4c15d1aa78f52575", size = 2268954, upload-time = "2025-05-17T17:20:55.027Z" },
+    { url = "https://files.pythonhosted.org/packages/f9/c5/ffe6474e0c551d54cab931918127c46d70cab8f114e0c2b5a3c071c2f484/pycryptodome-3.23.0-cp37-abi3-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:aa0698f65e5b570426fc31b8162ed4603b0c2841cbb9088e2b01641e3065915b", size = 2308534, upload-time = "2025-05-17T17:20:57.279Z" },
+    { url = "https://files.pythonhosted.org/packages/18/28/e199677fc15ecf43010f2463fde4c1a53015d1fe95fb03bca2890836603a/pycryptodome-3.23.0-cp37-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:53ecbafc2b55353edcebd64bf5da94a2a2cdf5090a6915bcca6eca6cc452585a", size = 2181853, upload-time = "2025-05-17T17:20:59.322Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/ea/4fdb09f2165ce1365c9eaefef36625583371ee514db58dc9b65d3a255c4c/pycryptodome-3.23.0-cp37-abi3-musllinux_1_2_i686.whl", hash = "sha256:156df9667ad9f2ad26255926524e1c136d6664b741547deb0a86a9acf5ea631f", size = 2342465, upload-time = "2025-05-17T17:21:03.83Z" },
+    { url = "https://files.pythonhosted.org/packages/22/82/6edc3fc42fe9284aead511394bac167693fb2b0e0395b28b8bedaa07ef04/pycryptodome-3.23.0-cp37-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:dea827b4d55ee390dc89b2afe5927d4308a8b538ae91d9c6f7a5090f397af1aa", size = 2267414, upload-time = "2025-05-17T17:21:06.72Z" },
+    { url = "https://files.pythonhosted.org/packages/59/fe/aae679b64363eb78326c7fdc9d06ec3de18bac68be4b612fc1fe8902693c/pycryptodome-3.23.0-cp37-abi3-win32.whl", hash = "sha256:507dbead45474b62b2bbe318eb1c4c8ee641077532067fec9c1aa82c31f84886", size = 1768484, upload-time = "2025-05-17T17:21:08.535Z" },
+    { url = "https://files.pythonhosted.org/packages/54/2f/e97a1b8294db0daaa87012c24a7bb714147c7ade7656973fd6c736b484ff/pycryptodome-3.23.0-cp37-abi3-win_amd64.whl", hash = "sha256:c75b52aacc6c0c260f204cbdd834f76edc9fb0d8e0da9fbf8352ef58202564e2", size = 1799636, upload-time = "2025-05-17T17:21:10.393Z" },
+    { url = "https://files.pythonhosted.org/packages/18/3d/f9441a0d798bf2b1e645adc3265e55706aead1255ccdad3856dbdcffec14/pycryptodome-3.23.0-cp37-abi3-win_arm64.whl", hash = "sha256:11eeeb6917903876f134b56ba11abe95c0b0fd5e3330def218083c7d98bbcb3c", size = 1703675, upload-time = "2025-05-17T17:21:13.146Z" },
+]
+
+[[package]]
+name = "pydantic"
+version = "2.13.4"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "annotated-types" },
+    { name = "pydantic-core" },
+    { name = "typing-extensions" },
+    { name = "typing-inspection" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/18/a5/b60d21ac674192f8ab0ba4e9fd860690f9b4a6e51ca5df118733b487d8d6/pydantic-2.13.4.tar.gz", hash = "sha256:c40756b57adaa8b1efeeced5c196f3f3b7c435f90e84ea7f443901bec8099ef6", size = 844775, upload-time = "2026-05-06T13:43:05.343Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/fd/7b/122376b1fd3c62c1ed9dc80c931ace4844b3c55407b6fb2d199377c9736f/pydantic-2.13.4-py3-none-any.whl", hash = "sha256:45a282cde31d808236fd7ea9d919b128653c8b38b393d1c4ab335c62924d9aba", size = 472262, upload-time = "2026-05-06T13:43:02.641Z" },
+]
+
+[[package]]
+name = "pydantic-core"
+version = "2.46.4"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/9d/56/921726b776ace8d8f5db44c4ef961006580d91dc52b803c489fafd1aa249/pydantic_core-2.46.4.tar.gz", hash = "sha256:62f875393d7f270851f20523dd2e29f082bcc82292d66db2b64ea71f64b6e1c1", size = 471464, upload-time = "2026-05-06T13:37:06.98Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/ce/8c/af022f0af448d7747c5154288d46b5f2bc5f17366eaa0e23e9aa04d59f3b/pydantic_core-2.46.4-cp312-cp312-macosx_10_12_x86_64.whl", hash = "sha256:3245406455a5d98187ec35530fd772b1d799b26667980872c8d4614991e2c4a2", size = 2106158, upload-time = "2026-05-06T13:38:57.215Z" },
+    { url = "https://files.pythonhosted.org/packages/19/95/6195171e385007300f0f5574592e467c568becce2d937a0b6804f218bc49/pydantic_core-2.46.4-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:962ccbab7b642487b1d8b7df90ef677e03134cf1fd8880bf698649b22a69371f", size = 1951724, upload-time = "2026-05-06T13:37:02.697Z" },
+    { url = "https://files.pythonhosted.org/packages/8e/bc/f47d1ff9cbb1620e1b5b697eef06010035735f07820180e74178226b27b3/pydantic_core-2.46.4-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:8233f2947cf85404441fd7e0085f53b10c93e0ee78611099b5c7237e36aacbf7", size = 1975742, upload-time = "2026-05-06T13:37:09.448Z" },
+    { url = "https://files.pythonhosted.org/packages/5b/11/9b9a5b0306345664a2da6410877af6e8082481b5884b3ddd78d47c6013ce/pydantic_core-2.46.4-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:3a233125ac121aa3ffba9a2b59edfc4a985a76092dc8279586ab4b71390875e7", size = 2052418, upload-time = "2026-05-06T13:37:38.234Z" },
+    { url = "https://files.pythonhosted.org/packages/f1/b7/a65fec226f5d78fc39f4a13c4cc0c768c22b113438f60c14adc9d2865038/pydantic_core-2.46.4-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:5b712b53160b79a5850310b912a5ef8e57e56947c8ad690c227f5c9d7e561712", size = 2232274, upload-time = "2026-05-06T13:38:27.753Z" },
+    { url = "https://files.pythonhosted.org/packages/68/f0/92039db98b907ef49269a8271f67db9cb78ae2fc68062ef7e4e77adb5f61/pydantic_core-2.46.4-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:9401557acd873c3a7f3eb9383edef8ac4968f9510e340f4808d427e75667e7b4", size = 2309940, upload-time = "2026-05-06T13:38:05.353Z" },
+    { url = "https://files.pythonhosted.org/packages/5f/97/2aab507d3d00ca626e8e57c1eac6a79e4e5fbcc63eb99733ff55d1717f65/pydantic_core-2.46.4-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:926c9541b14b12b1681dca8a0b75feb510b06c6341b70a8e500c2fdcff837cce", size = 2094516, upload-time = "2026-05-06T13:39:10.577Z" },
+    { url = "https://files.pythonhosted.org/packages/22/37/a8aca44d40d737dde2bc05b3c6c07dff0de07ce6f82e9f3167aeaf4d5dea/pydantic_core-2.46.4-cp312-cp312-manylinux_2_31_riscv64.whl", hash = "sha256:56cb4851bcaf3d117eddcef4fe66afd750a50274b0da8e22be256d10e5611987", size = 2136854, upload-time = "2026-05-06T13:40:22.59Z" },
+    { url = "https://files.pythonhosted.org/packages/24/99/fcef1b79238c06a8cbec70819ac722ba76e02bc8ada9b0fd66eba40da01b/pydantic_core-2.46.4-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:c68fcd102d71ea85c5b2dfac3f4f8476eff42a9e078fd5faefff6d145063536b", size = 2180306, upload-time = "2026-05-06T13:40:10.666Z" },
+    { url = "https://files.pythonhosted.org/packages/ae/6c/fc44000918855b42779d007ae63b0532794739027b2f417321cddbc44f6a/pydantic_core-2.46.4-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:b2f69dec1725e79a012d920df1707de5caf7ed5e08f3be4435e25803efc47458", size = 2190044, upload-time = "2026-05-06T13:40:43.231Z" },
+    { url = "https://files.pythonhosted.org/packages/6b/65/d9cadc9f1920d7a127ad2edba16c1db7916e59719285cd6c94600b0080ba/pydantic_core-2.46.4-cp312-cp312-musllinux_1_1_armv7l.whl", hash = "sha256:8d0820e8192167f80d88d64038e609c31452eeca865b4e1d9950a27a4609b00b", size = 2329133, upload-time = "2026-05-06T13:39:57.365Z" },
+    { url = "https://files.pythonhosted.org/packages/d0/cf/c873d91679f3a30bcf5e7ac280ce5573483e72295307685120d0d5ad3416/pydantic_core-2.46.4-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:fbdb89b3e1c94a30cc5edfce477c6e6a5dc4d8f84665b455c27582f211a1c72c", size = 2374464, upload-time = "2026-05-06T13:38:06.976Z" },
+    { url = "https://files.pythonhosted.org/packages/47/bd/6f2fc8188f31bf10590f1e98e7b306336161fac930a8c514cd7bd828c7dc/pydantic_core-2.46.4-cp312-cp312-win32.whl", hash = "sha256:9aa768456404a8bf48a4406685ac2bec8e72b62c69313734fa3b73cf33b3a894", size = 1974823, upload-time = "2026-05-06T13:40:47.985Z" },
+    { url = "https://files.pythonhosted.org/packages/40/8c/985c1d41ea1107c2534abd9870e4ed5c8e7669b5c308297835c001e7a1c4/pydantic_core-2.46.4-cp312-cp312-win_amd64.whl", hash = "sha256:e9c26f834c65f5752f3f06cb08cb86a913ceb7274d0db6e267808a708b46bc89", size = 2072919, upload-time = "2026-05-06T13:39:21.153Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/ba/f463d006e0c47373ca7ec5e1a261c59dc01ef4d62b2657af925fb0deee3a/pydantic_core-2.46.4-cp312-cp312-win_arm64.whl", hash = "sha256:4fc73cb559bdb54b1134a706a2802a4cddd27a0633f5abb7e53056268751ac6a", size = 2027604, upload-time = "2026-05-06T13:39:03.753Z" },
+    { url = "https://files.pythonhosted.org/packages/51/a2/5d30b469c5267a17b39dec53208222f76a8d351dfac4af661888c5aee77d/pydantic_core-2.46.4-cp313-cp313-macosx_10_12_x86_64.whl", hash = "sha256:5d5902252db0d3cedf8d4a1bc68f70eeb430f7e4c7104c8c476753519b423008", size = 2106306, upload-time = "2026-05-06T13:37:48.029Z" },
+    { url = "https://files.pythonhosted.org/packages/c1/81/4fa520eaffa8bd7d1525e644cd6d39e7d60b1592bc5b516693c7340b50f1/pydantic_core-2.46.4-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:c94f0688e7b8d0a67abf40e57a7eaaecd17cc9586706a31b76c031f63df052b4", size = 1951906, upload-time = "2026-05-06T13:37:17.012Z" },
+    { url = "https://files.pythonhosted.org/packages/03/d5/fd02da45b659668b05923b17ba3a0100a0a3d5541e3bd8fcc4ecb711309e/pydantic_core-2.46.4-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:f027324c56cd5406ca49c124b0db10e56c69064fec039acc571c29020cc87c76", size = 1976802, upload-time = "2026-05-06T13:37:35.113Z" },
+    { url = "https://files.pythonhosted.org/packages/21/f2/95727e1368be3d3ed485eaab7adbd7dda408f33f7a36e8b48e0144002b91/pydantic_core-2.46.4-cp313-cp313-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:e739fee756ba1010f8bcccb534252e85a35fe45ae92c295a06059ce58b74ccd3", size = 2052446, upload-time = "2026-05-06T13:37:12.313Z" },
+    { url = "https://files.pythonhosted.org/packages/9c/86/5d99feea3f77c7234b8718075b23db11532773c1a0dbd9b9490215dc2eeb/pydantic_core-2.46.4-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:9d56801be94b86a9da183e5f3766e6310752b99ff647e38b09a9500d88e46e76", size = 2232757, upload-time = "2026-05-06T13:39:01.149Z" },
+    { url = "https://files.pythonhosted.org/packages/d2/3a/508ac615935ef7588cf6d9e9b91309fdc2da751af865e02a9098de88258c/pydantic_core-2.46.4-cp313-cp313-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:2412e734dcb48da14d4e4006b82b46b74f2518b8a26ee7e58c6844a6cd6d03c4", size = 2309275, upload-time = "2026-05-06T13:37:41.406Z" },
+    { url = "https://files.pythonhosted.org/packages/07/f8/41db9de19d7987d6b04715a02b3b40aea467000275d9d758ffaa31af7d50/pydantic_core-2.46.4-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:9551187363ffc0de2a00b2e47c25aeaeb1020b69b668762966df15fc5659dd5a", size = 2094467, upload-time = "2026-05-06T13:39:18.847Z" },
+    { url = "https://files.pythonhosted.org/packages/2c/e2/f35033184cb11d0052daf4416e8e10a502ea2ac006fc4f459aee872727d1/pydantic_core-2.46.4-cp313-cp313-manylinux_2_31_riscv64.whl", hash = "sha256:0186750b482eefa11d7f435892b09c5c606193ef3375bcf94aa00ae6bfb66262", size = 2134417, upload-time = "2026-05-06T13:40:17.944Z" },
+    { url = "https://files.pythonhosted.org/packages/7e/7b/6ceeb1cc90e193862f444ebe373d8fdf613f0a82572dde03fb10734c6c71/pydantic_core-2.46.4-cp313-cp313-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:5855698a4856556d86e8e6cd8434bc3ac0314ee8e12089ae0e143f64c6256e4e", size = 2179782, upload-time = "2026-05-06T13:40:32.618Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/f2/c8d7773ede6af08036423a00ae0ceffce266c3c52a096c435d68c896083f/pydantic_core-2.46.4-cp313-cp313-musllinux_1_1_aarch64.whl", hash = "sha256:cbaf13819775b7f769bf4a1f066cb6df7a28d4480081a589828ef190226881cd", size = 2188782, upload-time = "2026-05-06T13:36:51.018Z" },
+    { url = "https://files.pythonhosted.org/packages/59/31/0c864784e31f09f05cdd87606f08923b9c9e7f6e51dd27f20f62f975ce9f/pydantic_core-2.46.4-cp313-cp313-musllinux_1_1_armv7l.whl", hash = "sha256:633147d34cf4550417f12e2b1a0383973bdf5cdfde212cb09e9a581cf10820be", size = 2328334, upload-time = "2026-05-06T13:40:37.764Z" },
+    { url = "https://files.pythonhosted.org/packages/c2/eb/4f6c8a41efa30baa755590f4141abf3a8c370fab610915733e74134a7270/pydantic_core-2.46.4-cp313-cp313-musllinux_1_1_x86_64.whl", hash = "sha256:82cf5301172168103724d49a1444d3378cb20cdee30b116a1bd6031236298a5d", size = 2372986, upload-time = "2026-05-06T13:39:34.152Z" },
+    { url = "https://files.pythonhosted.org/packages/5b/24/b375a480d53113860c299764bfe9f349a3dc9108b3adc0d7f0d786492ebf/pydantic_core-2.46.4-cp313-cp313-win32.whl", hash = "sha256:9fa8ae11da9e2b3126c6426f147e0fba88d96d65921799bb30c6abd1cb2c97fb", size = 1973693, upload-time = "2026-05-06T13:37:55.072Z" },
+    { url = "https://files.pythonhosted.org/packages/7e/e8/cff247591966f2d22ec8c003cd7587e27b7ba7b81ab2fb888e3ab75dc285/pydantic_core-2.46.4-cp313-cp313-win_amd64.whl", hash = "sha256:6b3ace8194b0e5204818c92802dcdca7fc6d88aabbb799d7c795540d9cd6d292", size = 2071819, upload-time = "2026-05-06T13:38:49.139Z" },
+    { url = "https://files.pythonhosted.org/packages/c6/1a/f4aee670d5670e9e148e0c82c7db98d780be566c6e6a97ee8035528ca0b3/pydantic_core-2.46.4-cp313-cp313-win_arm64.whl", hash = "sha256:184c081504d17f1c1066e430e117142b2c77d9448a97f7b65c6ac9fd9aee238d", size = 2027411, upload-time = "2026-05-06T13:40:45.796Z" },
+    { url = "https://files.pythonhosted.org/packages/8d/74/228a26ddad29c6672b805d9fd78e8d251cd04004fa7eed0e622096cd0250/pydantic_core-2.46.4-cp314-cp314-macosx_10_12_x86_64.whl", hash = "sha256:428e04521a40150c85216fc8b85e8d39fece235a9cf5e383761238c7fa9b96fb", size = 2102079, upload-time = "2026-05-06T13:38:41.019Z" },
+    { url = "https://files.pythonhosted.org/packages/ad/1f/8970b150a4b4365623ae00fc88603491f763c627311ae8031e3111356d6e/pydantic_core-2.46.4-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:23ace664830ee0bfe014a0c7bc248b1f7f25ed7ad103852c317624a1083af462", size = 1952179, upload-time = "2026-05-06T13:36:59.812Z" },
+    { url = "https://files.pythonhosted.org/packages/95/30/5211a831ae054928054b2f79731661087a2bc5c01e825c672b3a4a8f1b3e/pydantic_core-2.46.4-cp314-cp314-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:ce5c1d2a8b27468f433ca974829c44060b8097eedc39933e3c206a90ee49c4a9", size = 1978926, upload-time = "2026-05-06T13:37:39.933Z" },
+    { url = "https://files.pythonhosted.org/packages/57/e9/689668733b1eb67adeef047db3c2e8788fcf65a7fd9c9e2b46b7744fe245/pydantic_core-2.46.4-cp314-cp314-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:7283d57845ecf5a163403eb0702dfc220cc4fbdd18919cb5ccea4f95ee1cdab4", size = 2046785, upload-time = "2026-05-06T13:38:01.995Z" },
+    { url = "https://files.pythonhosted.org/packages/60/d9/6715260422ff50a2109878fd24d948a6c3446bb2664f34ee78cd972b3acd/pydantic_core-2.46.4-cp314-cp314-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:8daafc69c93ee8a0204506a3b6b30f586ef54028f52aeeeb5c4cfc5184fd5914", size = 2228733, upload-time = "2026-05-06T13:40:50.371Z" },
+    { url = "https://files.pythonhosted.org/packages/18/ae/fdb2f64316afca925640f8e70bb1a564b0ec2721c1389e25b8eb4bf9a299/pydantic_core-2.46.4-cp314-cp314-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:cd2213145bcc2ba85884d0ac63d222fece9209678f77b9b4d76f054c561adb28", size = 2307534, upload-time = "2026-05-06T13:37:21.531Z" },
+    { url = "https://files.pythonhosted.org/packages/89/1d/8eff589b45bb8190a9d12c49cfad0f176a5cbd1534908a6b5125e2886239/pydantic_core-2.46.4-cp314-cp314-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:7a5f930472650a82629163023e630d160863fce524c616f4e5186e5de9d9a49b", size = 2099732, upload-time = "2026-05-06T13:39:31.942Z" },
+    { url = "https://files.pythonhosted.org/packages/06/d5/ee5a3366637fee41dee51a1fc91562dcf12ddbc68fda34e6b253da2324bb/pydantic_core-2.46.4-cp314-cp314-manylinux_2_31_riscv64.whl", hash = "sha256:c1b3f518abeca3aa13c712fd202306e145abf59a18b094a6bafb2d2bbf59192c", size = 2129627, upload-time = "2026-05-06T13:37:25.033Z" },
+    { url = "https://files.pythonhosted.org/packages/94/33/2414be571d2c6a6c4d08be21f9292b6d3fdb08949a97b6dfe985017821db/pydantic_core-2.46.4-cp314-cp314-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:1a7dd0b3ee80d90150e3495a3a13ac34dbcbfd4f012996a6a1d8900e91b5c0fb", size = 2179141, upload-time = "2026-05-06T13:37:14.046Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/79/7daa95be995be0eecc4cf75064cb33f9bbbfe3fe0158caf2f0d4a996a5c7/pydantic_core-2.46.4-cp314-cp314-musllinux_1_1_aarch64.whl", hash = "sha256:3fb702cd90b0446a3a1c5e470bfa0dd23c0233b676a9099ddcc964fa6ca13898", size = 2184325, upload-time = "2026-05-06T13:36:53.615Z" },
+    { url = "https://files.pythonhosted.org/packages/9f/cb/d0a382f5c0de8a222dc61c65348e0ce831b1f68e0a018450d31c2cace3a5/pydantic_core-2.46.4-cp314-cp314-musllinux_1_1_armv7l.whl", hash = "sha256:b8458003118a712e66286df6a707db01c52c0f52f7db8e4a38f0da1d3b94fc4e", size = 2323990, upload-time = "2026-05-06T13:40:29.971Z" },
+    { url = "https://files.pythonhosted.org/packages/05/db/d9ba624cc4a5aced1598e88c04fdbd8310c8a69b9d38b9a3d39ce3a61ed7/pydantic_core-2.46.4-cp314-cp314-musllinux_1_1_x86_64.whl", hash = "sha256:372429a130e469c9cd698925ce5fc50940b7a1336b0d82038e63d5bbc4edc519", size = 2369978, upload-time = "2026-05-06T13:37:23.027Z" },
+    { url = "https://files.pythonhosted.org/packages/f2/20/d15df15ba918c423461905802bfd2981c3af0bfa0e40d05e13edbfa48bc3/pydantic_core-2.46.4-cp314-cp314-win32.whl", hash = "sha256:85bb3611ff1802f3ee7fdd7dbff26b56f343fb432d57a4728fdd49b6ef35e2f4", size = 1966354, upload-time = "2026-05-06T13:38:03.499Z" },
+    { url = "https://files.pythonhosted.org/packages/fc/b6/6b8de4c0a7d7ab3004c439c80c5c1e0a3e8d78bbae19379b01960383d9e5/pydantic_core-2.46.4-cp314-cp314-win_amd64.whl", hash = "sha256:811ff8e9c313ab425368bcbb36e5c4ebd7108c2bbf4e4089cfbb0b01eff63fac", size = 2072238, upload-time = "2026-05-06T13:39:40.807Z" },
+    { url = "https://files.pythonhosted.org/packages/32/36/51eb763beec1f4cf59b1db243a7dcc39cbb41230f050a09b9d69faaf0a48/pydantic_core-2.46.4-cp314-cp314-win_arm64.whl", hash = "sha256:bfec22eab3c8cc2ceec0248aec886624116dc079afa027ecc8ad4a7e62010f8a", size = 2018251, upload-time = "2026-05-06T13:37:26.72Z" },
+    { url = "https://files.pythonhosted.org/packages/e8/91/855af51d625b23aa987116a19e231d2aaef9c4a415273ddc189b79a45fee/pydantic_core-2.46.4-cp314-cp314t-macosx_10_12_x86_64.whl", hash = "sha256:af8244b2bef6aaad6d92cda81372de7f8c8d36c9f0c3ea36e827c60e7d9467a0", size = 2099593, upload-time = "2026-05-06T13:39:47.682Z" },
+    { url = "https://files.pythonhosted.org/packages/fb/1b/8784a54c65edb5f49f0a14d6977cf1b209bba85a4c77445b255c2de58ab3/pydantic_core-2.46.4-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:5a4330cdbc57162e4b3aa303f588ba752257694c9c9be3e7ebb11b4aca659b5d", size = 1935226, upload-time = "2026-05-06T13:40:40.428Z" },
+    { url = "https://files.pythonhosted.org/packages/e8/e7/1955d28d1afc56dd4b3ad7cc0cf39df1b9852964cf16e5d13912756d6d6b/pydantic_core-2.46.4-cp314-cp314t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:29c61fc04a3d840155ff08e475a04809278972fe6aef51e2720554e96367e34b", size = 1974605, upload-time = "2026-05-06T13:37:32.029Z" },
+    { url = "https://files.pythonhosted.org/packages/93/e2/3fedbf0ba7a22850e6e9fd78117f1c0f10f950182344d8a6c535d468fdd8/pydantic_core-2.46.4-cp314-cp314t-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:c50f2528cf200c5eed56faf3f4e22fcd5f38c157a8b78576e6ba3168ec35f000", size = 2030777, upload-time = "2026-05-06T13:38:55.239Z" },
+    { url = "https://files.pythonhosted.org/packages/f8/61/46be275fcaaba0b4f5b9669dd852267ce1ff616592dccf7a7845588df091/pydantic_core-2.46.4-cp314-cp314t-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:0cbe8b01f948de4286c74cdd6c667aceb38f5c1e26f0693b3983d9d74887c65e", size = 2236641, upload-time = "2026-05-06T13:37:08.096Z" },
+    { url = "https://files.pythonhosted.org/packages/60/db/12e93e46a8bac9988be3c016860f83293daea8c716c029c9ace279036f2f/pydantic_core-2.46.4-cp314-cp314t-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:617d7e2ca7dcb8c5cf6bcb8c59b8832c94b36196bbf1cbd1bfb56ed341905edd", size = 2286404, upload-time = "2026-05-06T13:40:20.221Z" },
+    { url = "https://files.pythonhosted.org/packages/e2/4a/4d8b19008f38d31c53b8219cfedc2e3d5de5fe99d90076b7e767de29274f/pydantic_core-2.46.4-cp314-cp314t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:7027560ee92211647d0d34e3f7cd6f50da56399d26a9c8ad0da286d3869a53f3", size = 2109219, upload-time = "2026-05-06T13:38:12.153Z" },
+    { url = "https://files.pythonhosted.org/packages/88/70/3cbc40978fefb7bb09c6708d40d4ad1a5d70fd7213c3d17f971de868ec1f/pydantic_core-2.46.4-cp314-cp314t-manylinux_2_31_riscv64.whl", hash = "sha256:f99626688942fb746e545232e7726926f3be91b5975f8b55327665fafda991c7", size = 2110594, upload-time = "2026-05-06T13:40:02.971Z" },
+    { url = "https://files.pythonhosted.org/packages/9d/20/b8d36736216e29491125531685b2f9e61aa5b4b2599893f8268551da3338/pydantic_core-2.46.4-cp314-cp314t-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:fc3e9034a63de20e15e8ade85358bc6efc614008cab72898b4b4952bea0509ff", size = 2159542, upload-time = "2026-05-06T13:39:27.506Z" },
+    { url = "https://files.pythonhosted.org/packages/1d/a2/367df868eb584dacf6bf82a389272406d7178e301c4ac82545ab98bc2dd9/pydantic_core-2.46.4-cp314-cp314t-musllinux_1_1_aarch64.whl", hash = "sha256:97e7cf2be5c77b7d1a9713a05605d49460d02c6078d38d8bef3cbe323c548424", size = 2168146, upload-time = "2026-05-06T13:38:31.93Z" },
+    { url = "https://files.pythonhosted.org/packages/c1/b8/4460f77f7e201893f649a29ab355dddd3beee8a97bcb1a320db414f9a06e/pydantic_core-2.46.4-cp314-cp314t-musllinux_1_1_armv7l.whl", hash = "sha256:3bf92c5d0e00fefaab325a4d27828fe6b6e2a21848686b5b60d2d9eeb09d76c6", size = 2306309, upload-time = "2026-05-06T13:37:44.717Z" },
+    { url = "https://files.pythonhosted.org/packages/64/c4/be2639293acd87dc8ddbcec41a73cee9b2ebf996fe6d892a1a74e88ad3f7/pydantic_core-2.46.4-cp314-cp314t-musllinux_1_1_x86_64.whl", hash = "sha256:3ecbc122d18468d06ca279dc26a8c2e2d5acb10943bb35e36ae92096dc3b5565", size = 2369736, upload-time = "2026-05-06T13:37:05.645Z" },
+    { url = "https://files.pythonhosted.org/packages/30/a6/9f9f380dbb301f67023bf8f707aaa75daadf84f7152d95c410fd7e81d994/pydantic_core-2.46.4-cp314-cp314t-win32.whl", hash = "sha256:e846ae7835bf0703ae43f534ab79a867146dadd59dc9ca5c8b53d5c8f7c9ef02", size = 1955575, upload-time = "2026-05-06T13:38:51.116Z" },
+    { url = "https://files.pythonhosted.org/packages/40/1f/f1eb9eb350e795d1af8586289746f5c5677d16043040d63710e22abc43c9/pydantic_core-2.46.4-cp314-cp314t-win_amd64.whl", hash = "sha256:2108ba5c1c1eca18030634489dc544844144ee36357f2f9f780b93e7ddbb44b5", size = 2051624, upload-time = "2026-05-06T13:38:21.672Z" },
+    { url = "https://files.pythonhosted.org/packages/f6/d2/42dd53d0a85c27606f316d3aa5d2869c4e8470a5ed6dec30e4a1abe19192/pydantic_core-2.46.4-cp314-cp314t-win_arm64.whl", hash = "sha256:4fcbe087dbc2068af7eda3aa87634eba216dbda64d1ae73c8684b621d33f6596", size = 2017325, upload-time = "2026-05-06T13:40:52.723Z" },
+    { url = "https://files.pythonhosted.org/packages/9d/1d/8987ad40f65ae1432753072f214fb5c74fe47ffbd0698bb9cbbb585664f8/pydantic_core-2.46.4-graalpy312-graalpy250_312_native-macosx_10_12_x86_64.whl", hash = "sha256:1d8ba486450b14f3b1d63bc521d410ec7565e52f887b9fb671791886436a42f7", size = 2095527, upload-time = "2026-05-06T13:39:52.283Z" },
+    { url = "https://files.pythonhosted.org/packages/64/d3/84c282a7eee1d3ac4c0377546ef5a1ea436ce26840d9ac3b7ed54a377507/pydantic_core-2.46.4-graalpy312-graalpy250_312_native-macosx_11_0_arm64.whl", hash = "sha256:3009f12e4e90b7f88b4f9adb1b0c4a3d58fe7820f3238c190047209d148026df", size = 1936024, upload-time = "2026-05-06T13:40:15.671Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/ca/eac61596cdeb4d7e174d3dc0bd8a6238f14f75f97a24e7b7db4c7e7340a0/pydantic_core-2.46.4-graalpy312-graalpy250_312_native-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:ad785e92e6dc634c21555edc8bd6b64957ab844541bcb96a1366c202951ae526", size = 1990696, upload-time = "2026-05-06T13:38:34.717Z" },
+    { url = "https://files.pythonhosted.org/packages/fa/c3/7c8b240552251faf6b3a957db200fcfbbcec36763c050428b601e0c9b83b/pydantic_core-2.46.4-graalpy312-graalpy250_312_native-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:00c603d540afdd6b80eb39f078f33ebd46211f02f33e34a32d9f053bba711de0", size = 2147590, upload-time = "2026-05-06T13:39:29.883Z" },
+]
+
+[[package]]
+name = "pydantic-settings"
+version = "2.14.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "pydantic" },
+    { name = "python-dotenv" },
+    { name = "typing-inspection" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/5c/b5/8f48e906c3e0205276e8bd8cb7512217a87b2685304d64be27cad5b3019f/pydantic_settings-2.14.2.tar.gz", hash = "sha256:c19dd64b19097f1de80184f0cc7b0272a13ae6e170cbf240a3e27e381ed14a5f", size = 237700, upload-time = "2026-06-19T13:44:56.324Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/77/c1/6e422f34e569cf8e18df68d1939c81c099d2b61e4f7d9621c8a77560799c/pydantic_settings-2.14.2-py3-none-any.whl", hash = "sha256:a20c97b37910b6550d5ea50fbcc2d4187defe58cd57070b73863d069419c9440", size = 61715, upload-time = "2026-06-19T13:44:55.02Z" },
+]
+
+[[package]]
+name = "pyee"
+version = "13.0.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/8b/04/e7c1fe4dc78a6fdbfd6c337b1c3732ff543b8a397683ab38378447baa331/pyee-13.0.1.tar.gz", hash = "sha256:0b931f7c14535667ed4c7e0d531716368715e860b988770fc7eb8578d1f67fc8", size = 31655, upload-time = "2026-02-14T21:12:28.044Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a0/c4/b4d4827c93ef43c01f599ef31453ccc1c132b353284fc6c87d535c233129/pyee-13.0.1-py3-none-any.whl", hash = "sha256:af2f8fede4171ef667dfded53f96e2ed0d6e6bd7ee3bb46437f77e3b57689228", size = 15659, upload-time = "2026-02-14T21:12:26.263Z" },
+]
+
+[[package]]
+name = "pygments"
+version = "2.20.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/c3/b2/bc9c9196916376152d655522fdcebac55e66de6603a76a02bca1b6414f6c/pygments-2.20.0.tar.gz", hash = "sha256:6757cd03768053ff99f3039c1a36d6c0aa0b263438fcab17520b30a303a82b5f", size = 4955991, upload-time = "2026-03-29T13:29:33.898Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/f4/7e/a72dd26f3b0f4f2bf1dd8923c85f7ceb43172af56d63c7383eb62b332364/pygments-2.20.0-py3-none-any.whl", hash = "sha256:81a9e26dd42fd28a23a2d169d86d7ac03b46e2f8b59ed4698fb4785f946d0176", size = 1231151, upload-time = "2026-03-29T13:29:30.038Z" },
+]
+
+[[package]]
+name = "pyjwt"
+version = "2.13.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/3b/81/58d0ac84e1ef3a3843791d6954d94c0b33d526c75eeb1efbce9d0a4c4077/pyjwt-2.13.0.tar.gz", hash = "sha256:41571c89ca91598c79e8ef18a2d07367d4810fbbd6f637794879baf1b7703423", size = 107515, upload-time = "2026-05-21T19:54:36.618Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a3/5e/ecf12fdb62546d64385c158514e9b2b671f7832108ef2ecd2020ce0af2d1/pyjwt-2.13.0-py3-none-any.whl", hash = "sha256:66adcc2aff09b3f1bbd95fc1e1577df8ac8723c978552fd43304c8a290ac5728", size = 31274, upload-time = "2026-05-21T19:54:35.362Z" },
+]
+
+[package.optional-dependencies]
+crypto = [
+    { name = "cryptography" },
+]
+
+[[package]]
+name = "pysocks"
+version = "1.7.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/bd/11/293dd436aea955d45fc4e8a35b6ae7270f5b8e00b53cf6c024c83b657a11/PySocks-1.7.1.tar.gz", hash = "sha256:3f8804571ebe159c380ac6de37643bb4685970655d3bba243530d6558b799aa0", size = 284429, upload-time = "2019-09-20T02:07:35.714Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/8d/59/b4572118e098ac8e46e399a1dd0f2d85403ce8bbaad9ec79373ed6badaf9/PySocks-1.7.1-py3-none-any.whl", hash = "sha256:2725bd0a9925919b9b51739eea5f9e2bae91e83288108a9ad338b2e3a4435ee5", size = 16725, upload-time = "2019-09-20T02:06:22.938Z" },
+]
+
+[[package]]
+name = "pytest"
+version = "9.1.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "colorama", marker = "sys_platform == 'win32'" },
+    { name = "iniconfig" },
+    { name = "packaging" },
+    { name = "pluggy" },
+    { name = "pygments" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/e4/47/b9efed96c114afcfa3c9d3fe98a76a1d14c74a9e266d397cf6eb64be5e01/pytest-9.1.1.tar.gz", hash = "sha256:1088fbde8f2b49d95a549a195707afa7a76a3ce9bcadc26b6d71f0ffda5fe313", size = 1636369, upload-time = "2026-06-19T10:58:32.857Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/24/25/1de2678b631f5a49215c6c96fff41ba892b0a34df68d6d80292b1b48aa7f/pytest-9.1.1-py3-none-any.whl", hash = "sha256:37a86b45efb9a47a61a36449063e8e18d0cab3161329fc099eb21783169c4f0c", size = 386536, upload-time = "2026-06-19T10:58:31.347Z" },
+]
+
+[[package]]
+name = "python-dateutil"
+version = "2.9.0.post0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "six" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/66/c0/0c8b6ad9f17a802ee498c46e004a0eb49bc148f2fd230864601a86dcf6db/python-dateutil-2.9.0.post0.tar.gz", hash = "sha256:37dd54208da7e1cd875388217d5e00ebd4179249f90fb72437e91a35459a0ad3", size = 342432, upload-time = "2024-03-01T18:36:20.211Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/ec/57/56b9bcc3c9c6a792fcbaf139543cee77261f3651ca9da0c93f5c1221264b/python_dateutil-2.9.0.post0-py2.py3-none-any.whl", hash = "sha256:a8b2bc7bffae282281c8140a97d3aa9c14da0b136dfe83f850eea9a5f7470427", size = 229892, upload-time = "2024-03-01T18:36:18.57Z" },
+]
+
+[[package]]
+name = "python-dotenv"
+version = "1.2.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/82/ed/0301aeeac3e5353ef3d94b6ec08bbcabd04a72018415dcb29e588514bba8/python_dotenv-1.2.2.tar.gz", hash = "sha256:2c371a91fbd7ba082c2c1dc1f8bf89ca22564a087c2c287cd9b662adde799cf3", size = 50135, upload-time = "2026-03-01T16:00:26.196Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/0b/d7/1959b9648791274998a9c3526f6d0ec8fd2233e4d4acce81bbae76b44b2a/python_dotenv-1.2.2-py3-none-any.whl", hash = "sha256:1d8214789a24de455a8b8bd8ae6fe3c6b69a5e3d64aa8a8e5d68e694bbcb285a", size = 22101, upload-time = "2026-03-01T16:00:25.09Z" },
+]
+
+[[package]]
+name = "python-multipart"
+version = "0.0.32"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/5b/42/55c32bb9b12693c092ad250a0e82edb5b31ddeda6eb772de5f308b3804ad/python_multipart-0.0.32.tar.gz", hash = "sha256:be54b7f3fa167bb83e4fcd936b887b708f4e57fe75911c02aebf53efaf8d938e", size = 46881, upload-time = "2026-06-04T16:18:58.647Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e1/04/e8135ebd1ad02c56ec633277529b2602ff99ff634be76cdba5744cf554fd/python_multipart-0.0.32-py3-none-any.whl", hash = "sha256:ff6d3f776f16878c894e52e107296ffc890e913c611b1a4ec6c44e2821fe2e23", size = 30042, upload-time = "2026-06-04T16:18:57.319Z" },
+]
+
+[[package]]
+name = "pywin32"
+version = "312"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/83/ff/32aa7d2ed0ab12b323aaa64f9b75e6ad4f8fd09f9ccfc28c79414d46838d/pywin32-312-cp312-cp312-win32.whl", hash = "sha256:dab4f65ac9c4e48400a2a0530c46c3c579cd5905ecd11b80692373915269208b", size = 6371877, upload-time = "2026-06-04T07:49:28.836Z" },
+    { url = "https://files.pythonhosted.org/packages/03/d9/77040d3b43df3f3be32ea289433d660d2727f5ba327bc73be835127d9d60/pywin32-312-cp312-cp312-win_amd64.whl", hash = "sha256:b457f6d628a47e8a7346ce22acb7e1a46a4a78b52e1d17e1af56871bd19a93bc", size = 6914841, upload-time = "2026-06-04T07:49:31.85Z" },
+    { url = "https://files.pythonhosted.org/packages/e3/cc/7b1ec671775756020a0ee7f4feeaf3c568f0ab86bd3900088cf986937a92/pywin32-312-cp312-cp312-win_arm64.whl", hash = "sha256:6017c58e12f6809fbb0555b75df144c2922a9ffd18e4b9b5afa863b6c1a9d950", size = 6727901, upload-time = "2026-06-04T07:49:34.244Z" },
+    { url = "https://files.pythonhosted.org/packages/2d/41/12fbfd7f36ed2146d8bc9de96c2741296bf0d490b98508496cff322e274c/pywin32-312-cp313-cp313-win32.whl", hash = "sha256:7a27df850933d16a8eabfbaeb73d52b273e2da667f80d70b01a89d1f6828d02c", size = 6370184, upload-time = "2026-06-04T07:49:36.253Z" },
+    { url = "https://files.pythonhosted.org/packages/ba/db/36a78e3403099d31d9746d13fdcde5accc43c1155f375a34d15983a479a7/pywin32-312-cp313-cp313-win_amd64.whl", hash = "sha256:c53e878d15a1c44788082bfe712a905433473aa38f86375b7cf8b45e3acbaaf9", size = 6914298, upload-time = "2026-06-04T07:49:38.876Z" },
+    { url = "https://files.pythonhosted.org/packages/84/37/c1697194092b76de9ed47ca124323f02c57ffc8a45c06f88a3d5acaf01eb/pywin32-312-cp313-cp313-win_arm64.whl", hash = "sha256:59aba5d5940842075343a5ddc6b11f1cdf0d1567fe745290359dfbcc7c2eb831", size = 6727640, upload-time = "2026-06-04T07:49:41.083Z" },
+    { url = "https://files.pythonhosted.org/packages/fc/2b/1f3cded5822fd49c02f40544cbb5f58c7cfd6b1694869fd476cb6170ee97/pywin32-312-cp314-cp314-win32.whl", hash = "sha256:a77a90fbb6881238d2ca9c6fd797b25817f3768fe78d214a90137ff055a75f5b", size = 6468928, upload-time = "2026-06-04T07:49:43.188Z" },
+    { url = "https://files.pythonhosted.org/packages/21/82/3bf86d2e2808902013132e1ce905a7da0da53790f3836c64bf44d55e24f3/pywin32-312-cp314-cp314-win_amd64.whl", hash = "sha256:a4dd3a848290ef724347b19f301045831d8e802fa4464f491b98b1e0a081432e", size = 7024157, upload-time = "2026-06-04T07:49:45.34Z" },
+    { url = "https://files.pythonhosted.org/packages/a4/0e/73f6d6800b4f27655abd9e9f6aaeaefcddb2b946e4674efa2bab184a7f7b/pywin32-312-cp314-cp314-win_arm64.whl", hash = "sha256:9fce94568364e0155e6dfb781ac5d95903be8baf28670632beab1b523f300daa", size = 6839598, upload-time = "2026-06-04T07:49:47.613Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/61/caa39686032d2ebdd04ff0ab5cbe163126c0066d98e00c9018646e42393b/pywin32-312-cp315-cp315-win32.whl", hash = "sha256:5c1fbe4a937a73ae9297384a3da38518cbc694c68ad8a809b2e19acd350f03ed", size = 6471159, upload-time = "2026-06-04T07:49:50.035Z" },
+    { url = "https://files.pythonhosted.org/packages/0f/cd/7e1de64a4a6f69c04214169657ccab0d93a670ea50e35eb8f489d7378249/pywin32-312-cp315-cp315-win_amd64.whl", hash = "sha256:c2f03a0f73f804a13c2735b99392b0cd426bb4f2c4d0178e5ac966a0f21618d5", size = 7025293, upload-time = "2026-06-04T07:49:54.857Z" },
+    { url = "https://files.pythonhosted.org/packages/23/ed/4532e9388e65fa16b46776ef47ad631a64eda1631884488af707666350ed/pywin32-312-cp315-cp315-win_arm64.whl", hash = "sha256:a8597d28f267b39074aef51fa593530082b39cbe5a074226096857b1fed2dfb9", size = 6840337, upload-time = "2026-06-04T07:49:57.531Z" },
+]
+
+[[package]]
+name = "pyyaml"
+version = "6.0.3"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/05/8e/961c0007c59b8dd7729d542c61a4d537767a59645b82a0b521206e1e25c2/pyyaml-6.0.3.tar.gz", hash = "sha256:d76623373421df22fb4cf8817020cbb7ef15c725b9d5e45f17e189bfc384190f", size = 130960, upload-time = "2025-09-25T21:33:16.546Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/d1/33/422b98d2195232ca1826284a76852ad5a86fe23e31b009c9886b2d0fb8b2/pyyaml-6.0.3-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:7f047e29dcae44602496db43be01ad42fc6f1cc0d8cd6c83d342306c32270196", size = 182063, upload-time = "2025-09-25T21:32:11.445Z" },
+    { url = "https://files.pythonhosted.org/packages/89/a0/6cf41a19a1f2f3feab0e9c0b74134aa2ce6849093d5517a0c550fe37a648/pyyaml-6.0.3-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:fc09d0aa354569bc501d4e787133afc08552722d3ab34836a80547331bb5d4a0", size = 173973, upload-time = "2025-09-25T21:32:12.492Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/23/7a778b6bd0b9a8039df8b1b1d80e2e2ad78aa04171592c8a5c43a56a6af4/pyyaml-6.0.3-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:9149cad251584d5fb4981be1ecde53a1ca46c891a79788c0df828d2f166bda28", size = 775116, upload-time = "2025-09-25T21:32:13.652Z" },
+    { url = "https://files.pythonhosted.org/packages/65/30/d7353c338e12baef4ecc1b09e877c1970bd3382789c159b4f89d6a70dc09/pyyaml-6.0.3-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:5fdec68f91a0c6739b380c83b951e2c72ac0197ace422360e6d5a959d8d97b2c", size = 844011, upload-time = "2025-09-25T21:32:15.21Z" },
+    { url = "https://files.pythonhosted.org/packages/8b/9d/b3589d3877982d4f2329302ef98a8026e7f4443c765c46cfecc8858c6b4b/pyyaml-6.0.3-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:ba1cc08a7ccde2d2ec775841541641e4548226580ab850948cbfda66a1befcdc", size = 807870, upload-time = "2025-09-25T21:32:16.431Z" },
+    { url = "https://files.pythonhosted.org/packages/05/c0/b3be26a015601b822b97d9149ff8cb5ead58c66f981e04fedf4e762f4bd4/pyyaml-6.0.3-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:8dc52c23056b9ddd46818a57b78404882310fb473d63f17b07d5c40421e47f8e", size = 761089, upload-time = "2025-09-25T21:32:17.56Z" },
+    { url = "https://files.pythonhosted.org/packages/be/8e/98435a21d1d4b46590d5459a22d88128103f8da4c2d4cb8f14f2a96504e1/pyyaml-6.0.3-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:41715c910c881bc081f1e8872880d3c650acf13dfa8214bad49ed4cede7c34ea", size = 790181, upload-time = "2025-09-25T21:32:18.834Z" },
+    { url = "https://files.pythonhosted.org/packages/74/93/7baea19427dcfbe1e5a372d81473250b379f04b1bd3c4c5ff825e2327202/pyyaml-6.0.3-cp312-cp312-win32.whl", hash = "sha256:96b533f0e99f6579b3d4d4995707cf36df9100d67e0c8303a0c55b27b5f99bc5", size = 137658, upload-time = "2025-09-25T21:32:20.209Z" },
+    { url = "https://files.pythonhosted.org/packages/86/bf/899e81e4cce32febab4fb42bb97dcdf66bc135272882d1987881a4b519e9/pyyaml-6.0.3-cp312-cp312-win_amd64.whl", hash = "sha256:5fcd34e47f6e0b794d17de1b4ff496c00986e1c83f7ab2fb8fcfe9616ff7477b", size = 154003, upload-time = "2025-09-25T21:32:21.167Z" },
+    { url = "https://files.pythonhosted.org/packages/1a/08/67bd04656199bbb51dbed1439b7f27601dfb576fb864099c7ef0c3e55531/pyyaml-6.0.3-cp312-cp312-win_arm64.whl", hash = "sha256:64386e5e707d03a7e172c0701abfb7e10f0fb753ee1d773128192742712a98fd", size = 140344, upload-time = "2025-09-25T21:32:22.617Z" },
+    { url = "https://files.pythonhosted.org/packages/d1/11/0fd08f8192109f7169db964b5707a2f1e8b745d4e239b784a5a1dd80d1db/pyyaml-6.0.3-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:8da9669d359f02c0b91ccc01cac4a67f16afec0dac22c2ad09f46bee0697eba8", size = 181669, upload-time = "2025-09-25T21:32:23.673Z" },
+    { url = "https://files.pythonhosted.org/packages/b1/16/95309993f1d3748cd644e02e38b75d50cbc0d9561d21f390a76242ce073f/pyyaml-6.0.3-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:2283a07e2c21a2aa78d9c4442724ec1eb15f5e42a723b99cb3d822d48f5f7ad1", size = 173252, upload-time = "2025-09-25T21:32:25.149Z" },
+    { url = "https://files.pythonhosted.org/packages/50/31/b20f376d3f810b9b2371e72ef5adb33879b25edb7a6d072cb7ca0c486398/pyyaml-6.0.3-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:ee2922902c45ae8ccada2c5b501ab86c36525b883eff4255313a253a3160861c", size = 767081, upload-time = "2025-09-25T21:32:26.575Z" },
+    { url = "https://files.pythonhosted.org/packages/49/1e/a55ca81e949270d5d4432fbbd19dfea5321eda7c41a849d443dc92fd1ff7/pyyaml-6.0.3-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:a33284e20b78bd4a18c8c2282d549d10bc8408a2a7ff57653c0cf0b9be0afce5", size = 841159, upload-time = "2025-09-25T21:32:27.727Z" },
+    { url = "https://files.pythonhosted.org/packages/74/27/e5b8f34d02d9995b80abcef563ea1f8b56d20134d8f4e5e81733b1feceb2/pyyaml-6.0.3-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:0f29edc409a6392443abf94b9cf89ce99889a1dd5376d94316ae5145dfedd5d6", size = 801626, upload-time = "2025-09-25T21:32:28.878Z" },
+    { url = "https://files.pythonhosted.org/packages/f9/11/ba845c23988798f40e52ba45f34849aa8a1f2d4af4b798588010792ebad6/pyyaml-6.0.3-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:f7057c9a337546edc7973c0d3ba84ddcdf0daa14533c2065749c9075001090e6", size = 753613, upload-time = "2025-09-25T21:32:30.178Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/e0/7966e1a7bfc0a45bf0a7fb6b98ea03fc9b8d84fa7f2229e9659680b69ee3/pyyaml-6.0.3-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:eda16858a3cab07b80edaf74336ece1f986ba330fdb8ee0d6c0d68fe82bc96be", size = 794115, upload-time = "2025-09-25T21:32:31.353Z" },
+    { url = "https://files.pythonhosted.org/packages/de/94/980b50a6531b3019e45ddeada0626d45fa85cbe22300844a7983285bed3b/pyyaml-6.0.3-cp313-cp313-win32.whl", hash = "sha256:d0eae10f8159e8fdad514efdc92d74fd8d682c933a6dd088030f3834bc8e6b26", size = 137427, upload-time = "2025-09-25T21:32:32.58Z" },
+    { url = "https://files.pythonhosted.org/packages/97/c9/39d5b874e8b28845e4ec2202b5da735d0199dbe5b8fb85f91398814a9a46/pyyaml-6.0.3-cp313-cp313-win_amd64.whl", hash = "sha256:79005a0d97d5ddabfeeea4cf676af11e647e41d81c9a7722a193022accdb6b7c", size = 154090, upload-time = "2025-09-25T21:32:33.659Z" },
+    { url = "https://files.pythonhosted.org/packages/73/e8/2bdf3ca2090f68bb3d75b44da7bbc71843b19c9f2b9cb9b0f4ab7a5a4329/pyyaml-6.0.3-cp313-cp313-win_arm64.whl", hash = "sha256:5498cd1645aa724a7c71c8f378eb29ebe23da2fc0d7a08071d89469bf1d2defb", size = 140246, upload-time = "2025-09-25T21:32:34.663Z" },
+    { url = "https://files.pythonhosted.org/packages/9d/8c/f4bd7f6465179953d3ac9bc44ac1a8a3e6122cf8ada906b4f96c60172d43/pyyaml-6.0.3-cp314-cp314-macosx_10_13_x86_64.whl", hash = "sha256:8d1fab6bb153a416f9aeb4b8763bc0f22a5586065f86f7664fc23339fc1c1fac", size = 181814, upload-time = "2025-09-25T21:32:35.712Z" },
+    { url = "https://files.pythonhosted.org/packages/bd/9c/4d95bb87eb2063d20db7b60faa3840c1b18025517ae857371c4dd55a6b3a/pyyaml-6.0.3-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:34d5fcd24b8445fadc33f9cf348c1047101756fd760b4dacb5c3e99755703310", size = 173809, upload-time = "2025-09-25T21:32:36.789Z" },
+    { url = "https://files.pythonhosted.org/packages/92/b5/47e807c2623074914e29dabd16cbbdd4bf5e9b2db9f8090fa64411fc5382/pyyaml-6.0.3-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:501a031947e3a9025ed4405a168e6ef5ae3126c59f90ce0cd6f2bfc477be31b7", size = 766454, upload-time = "2025-09-25T21:32:37.966Z" },
+    { url = "https://files.pythonhosted.org/packages/02/9e/e5e9b168be58564121efb3de6859c452fccde0ab093d8438905899a3a483/pyyaml-6.0.3-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:b3bc83488de33889877a0f2543ade9f70c67d66d9ebb4ac959502e12de895788", size = 836355, upload-time = "2025-09-25T21:32:39.178Z" },
+    { url = "https://files.pythonhosted.org/packages/88/f9/16491d7ed2a919954993e48aa941b200f38040928474c9e85ea9e64222c3/pyyaml-6.0.3-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:c458b6d084f9b935061bc36216e8a69a7e293a2f1e68bf956dcd9e6cbcd143f5", size = 794175, upload-time = "2025-09-25T21:32:40.865Z" },
+    { url = "https://files.pythonhosted.org/packages/dd/3f/5989debef34dc6397317802b527dbbafb2b4760878a53d4166579111411e/pyyaml-6.0.3-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:7c6610def4f163542a622a73fb39f534f8c101d690126992300bf3207eab9764", size = 755228, upload-time = "2025-09-25T21:32:42.084Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/ce/af88a49043cd2e265be63d083fc75b27b6ed062f5f9fd6cdc223ad62f03e/pyyaml-6.0.3-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:5190d403f121660ce8d1d2c1bb2ef1bd05b5f68533fc5c2ea899bd15f4399b35", size = 789194, upload-time = "2025-09-25T21:32:43.362Z" },
+    { url = "https://files.pythonhosted.org/packages/23/20/bb6982b26a40bb43951265ba29d4c246ef0ff59c9fdcdf0ed04e0687de4d/pyyaml-6.0.3-cp314-cp314-win_amd64.whl", hash = "sha256:4a2e8cebe2ff6ab7d1050ecd59c25d4c8bd7e6f400f5f82b96557ac0abafd0ac", size = 156429, upload-time = "2025-09-25T21:32:57.844Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/f4/a4541072bb9422c8a883ab55255f918fa378ecf083f5b85e87fc2b4eda1b/pyyaml-6.0.3-cp314-cp314-win_arm64.whl", hash = "sha256:93dda82c9c22deb0a405ea4dc5f2d0cda384168e466364dec6255b293923b2f3", size = 143912, upload-time = "2025-09-25T21:32:59.247Z" },
+    { url = "https://files.pythonhosted.org/packages/7c/f9/07dd09ae774e4616edf6cda684ee78f97777bdd15847253637a6f052a62f/pyyaml-6.0.3-cp314-cp314t-macosx_10_13_x86_64.whl", hash = "sha256:02893d100e99e03eda1c8fd5c441d8c60103fd175728e23e431db1b589cf5ab3", size = 189108, upload-time = "2025-09-25T21:32:44.377Z" },
+    { url = "https://files.pythonhosted.org/packages/4e/78/8d08c9fb7ce09ad8c38ad533c1191cf27f7ae1effe5bb9400a46d9437fcf/pyyaml-6.0.3-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:c1ff362665ae507275af2853520967820d9124984e0f7466736aea23d8611fba", size = 183641, upload-time = "2025-09-25T21:32:45.407Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/5b/3babb19104a46945cf816d047db2788bcaf8c94527a805610b0289a01c6b/pyyaml-6.0.3-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:6adc77889b628398debc7b65c073bcb99c4a0237b248cacaf3fe8a557563ef6c", size = 831901, upload-time = "2025-09-25T21:32:48.83Z" },
+    { url = "https://files.pythonhosted.org/packages/8b/cc/dff0684d8dc44da4d22a13f35f073d558c268780ce3c6ba1b87055bb0b87/pyyaml-6.0.3-cp314-cp314t-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:a80cb027f6b349846a3bf6d73b5e95e782175e52f22108cfa17876aaeff93702", size = 861132, upload-time = "2025-09-25T21:32:50.149Z" },
+    { url = "https://files.pythonhosted.org/packages/b1/5e/f77dc6b9036943e285ba76b49e118d9ea929885becb0a29ba8a7c75e29fe/pyyaml-6.0.3-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:00c4bdeba853cc34e7dd471f16b4114f4162dc03e6b7afcc2128711f0eca823c", size = 839261, upload-time = "2025-09-25T21:32:51.808Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/88/a9db1376aa2a228197c58b37302f284b5617f56a5d959fd1763fb1675ce6/pyyaml-6.0.3-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:66e1674c3ef6f541c35191caae2d429b967b99e02040f5ba928632d9a7f0f065", size = 805272, upload-time = "2025-09-25T21:32:52.941Z" },
+    { url = "https://files.pythonhosted.org/packages/da/92/1446574745d74df0c92e6aa4a7b0b3130706a4142b2d1a5869f2eaa423c6/pyyaml-6.0.3-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:16249ee61e95f858e83976573de0f5b2893b3677ba71c9dd36b9cf8be9ac6d65", size = 829923, upload-time = "2025-09-25T21:32:54.537Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/7a/1c7270340330e575b92f397352af856a8c06f230aa3e76f86b39d01b416a/pyyaml-6.0.3-cp314-cp314t-win_amd64.whl", hash = "sha256:4ad1906908f2f5ae4e5a8ddfce73c320c2a1429ec52eafd27138b7f1cbe341c9", size = 174062, upload-time = "2025-09-25T21:32:55.767Z" },
+    { url = "https://files.pythonhosted.org/packages/f1/12/de94a39c2ef588c7e6455cfbe7343d3b2dc9d6b6b2f40c4c6565744c873d/pyyaml-6.0.3-cp314-cp314t-win_arm64.whl", hash = "sha256:ebc55a14a21cb14062aa4162f906cd962b28e2e9ea38f9b4391244cd8de4ae0b", size = 149341, upload-time = "2025-09-25T21:32:56.828Z" },
+]
+
+[[package]]
+name = "referencing"
+version = "0.37.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "attrs" },
+    { name = "rpds-py" },
+    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/22/f5/df4e9027acead3ecc63e50fe1e36aca1523e1719559c499951bb4b53188f/referencing-0.37.0.tar.gz", hash = "sha256:44aefc3142c5b842538163acb373e24cce6632bd54bdb01b21ad5863489f50d8", size = 78036, upload-time = "2025-10-13T15:30:48.871Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/2c/58/ca301544e1fa93ed4f80d724bf5b194f6e4b945841c5bfd555878eea9fcb/referencing-0.37.0-py3-none-any.whl", hash = "sha256:381329a9f99628c9069361716891d34ad94af76e461dcb0335825aecc7692231", size = 26766, upload-time = "2025-10-13T15:30:47.625Z" },
+]
+
+[[package]]
+name = "regex"
+version = "2026.7.19"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/20/98/04b13f1ddfb63158025291c02e03eb42fbb7acb51d091d541050eb4e35e8/regex-2026.7.19.tar.gz", hash = "sha256:7e77b324909c1617cbb4c668677e2c6ae13f44d7c1de0d4f15f2e3c10f3315b5", size = 416440, upload-time = "2026-07-19T00:19:48.923Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/3b/b9/d11d7e501ac8fd7d617684423ebb9561e0b998481c1e4cbc0cb212c5d74a/regex-2026.7.19-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:2cc3460cedf7579948486eab03bc9ad7089df4d7281c0f47f4afe03e8d13f02d", size = 496778, upload-time = "2026-07-19T00:17:05.677Z" },
+    { url = "https://files.pythonhosted.org/packages/3f/a9/a5ab6f312f24318019170dc485d5421fe4f89e43a98640da50d95a8a7041/regex-2026.7.19-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:0e9554c8785eac5cffe6300f69a91f58ba72bc88a5f8d661235ad7c6aa5b8ccd", size = 297122, upload-time = "2026-07-19T00:17:07.59Z" },
+    { url = "https://files.pythonhosted.org/packages/b3/63/4cab4d7f2d384a144d420b763d97674cb70619c878ea6fcd7640d0e62143/regex-2026.7.19-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:d7da47a0f248977f08e2cb659ff3c17ddc13a4d39b3a7baa0a81bf5b415430f6", size = 292009, upload-time = "2026-07-19T00:17:09.648Z" },
+    { url = "https://files.pythonhosted.org/packages/22/85/102a81b218298957d4ea7d2f084fae537a71add9d6ff93c8e67284c5f45e/regex-2026.7.19-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:93db40c8de0815baab96a06e08a984bac71f989d13bab789e382158c5d426797", size = 796708, upload-time = "2026-07-19T00:17:11.542Z" },
+    { url = "https://files.pythonhosted.org/packages/78/b5/dc136af5629938a037cd2b304c12240e132ec92f38be8ff9cc89af2a1f2d/regex-2026.7.19-cp312-cp312-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:66bd62c59a5427746e8c44becae1d9b99d22fb13f30f492083dfb9ad7c45cc18", size = 865651, upload-time = "2026-07-19T00:17:13.312Z" },
+    { url = "https://files.pythonhosted.org/packages/e0/75/67402ae3cd9c8c988a4c805d15ee3eef015e7ca4cb112cf3e640fc1f4153/regex-2026.7.19-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:1649eb39fcc9ea80c4d2f110fde2b8ab2aef3877b98f02ab9b14e961f418c511", size = 911756, upload-time = "2026-07-19T00:17:15.015Z" },
+    { url = "https://files.pythonhosted.org/packages/2a/8e/096d00c7c480ef2ff4265349b14e2261d4ab787ba1f74e2e80d1c58079c3/regex-2026.7.19-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:9dce8ec9695f531a1b8a6f314fd4b393adcccf2ea861db480cdf97a301d01a68", size = 801798, upload-time = "2026-07-19T00:17:17.208Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/41/e7ecac6edb5722417f85cc67eaf386322fbe8acf6918ec2fdc37c20dd9d0/regex-2026.7.19-cp312-cp312-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:3080a7fd38ef049bd489e01c970c97dd84ff446a885b0f1f6b26d9b1ad13ce11", size = 776933, upload-time = "2026-07-19T00:17:19.347Z" },
+    { url = "https://files.pythonhosted.org/packages/6f/69/03c9b3f058d66403e0ca2c938696e81d51cd4c6d47ec5265f02f96948d9a/regex-2026.7.19-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:1d793a7988e04fcb1e2e135567443d82173225d657419ec09414a9b5a145b986", size = 784338, upload-time = "2026-07-19T00:17:21.057Z" },
+    { url = "https://files.pythonhosted.org/packages/f6/f7/b38ab3d43f284afbb618fcd15d0e77eb786ae461ce1f6bc7494619ddc0f2/regex-2026.7.19-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:e8b0abe7d870f53ca5143895fef7d1041a0c831a140d3dc2c760dd7ba25d4a8b", size = 860452, upload-time = "2026-07-19T00:17:23.119Z" },
+    { url = "https://files.pythonhosted.org/packages/15/5c/ff60ef0571121714f3cf9920bc183071e384a10b556d042e0fdb06cc07a5/regex-2026.7.19-cp312-cp312-musllinux_1_2_riscv64.whl", hash = "sha256:4e5413bd5f13d3a4e3539ca98f70f75e7fca92518dd7f117f030ebedd10b60cb", size = 765958, upload-time = "2026-07-19T00:17:24.81Z" },
+    { url = "https://files.pythonhosted.org/packages/aa/0f/bd34021162c0ab47f9a315bd56cd5642e920c8e5668a75ef6c6a6fca590d/regex-2026.7.19-cp312-cp312-musllinux_1_2_s390x.whl", hash = "sha256:73b133a9e6fb512858e7f065e96f1180aa46646bc74a83aea62f1d314f3dd035", size = 851765, upload-time = "2026-07-19T00:17:26.993Z" },
+    { url = "https://files.pythonhosted.org/packages/2a/20/a2ca43edade0595cccfdc98636739f536d9e26898e7dbddc2b9e98898953/regex-2026.7.19-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:dbe6493fbd27321b1d1f2dd4f5c7e5bd4d8b1d7cab7f32fd67db3d0b2ed8248a", size = 789714, upload-time = "2026-07-19T00:17:28.699Z" },
+    { url = "https://files.pythonhosted.org/packages/5d/47/e02db4015d424fc83c00ea0ac8c5e5ec14397943de9abf909d5ce3a25931/regex-2026.7.19-cp312-cp312-win32.whl", hash = "sha256:ddd67571c10869f65a5d7dde536d1e066e306cc90de57d7de4d5f34802428bb5", size = 267157, upload-time = "2026-07-19T00:17:31.051Z" },
+    { url = "https://files.pythonhosted.org/packages/08/8e/c780c131f79b42ed22d1bd7da4096c2c35f813e835acd02ef0f018bd892c/regex-2026.7.19-cp312-cp312-win_amd64.whl", hash = "sha256:e30d40268a28d54ce0437031750497004c22602b8e3ab891f759b795a003b312", size = 277777, upload-time = "2026-07-19T00:17:32.848Z" },
+    { url = "https://files.pythonhosted.org/packages/3e/4c/e4d7e086449bdf379d89774bf1f89dc4a41943f3c5a6125a03905b34b5fb/regex-2026.7.19-cp312-cp312-win_arm64.whl", hash = "sha256:de9208bb427130c82a5dbfd104f92c8876fc9559278c880b3002755bbbe9c83d", size = 277136, upload-time = "2026-07-19T00:17:34.803Z" },
+    { url = "https://files.pythonhosted.org/packages/5d/3d/84165e4299ff76f3a40fe1f2abf939e976f693383a08d2beea6af62bd2c1/regex-2026.7.19-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:f035d9dc1d25eff9d361456572231c7d27b5ccd473ca7dc0adfce732bd006d40", size = 496552, upload-time = "2026-07-19T00:17:36.808Z" },
+    { url = "https://files.pythonhosted.org/packages/02/a2/a65293e6e4cf28eb7ee1be5335a5386c40d6742e9f47fafc8fec785e16c7/regex-2026.7.19-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:c42572142ed0b9d5d261ba727157c426510da78e20828b66bbb855098b8a4e38", size = 296983, upload-time = "2026-07-19T00:17:38.816Z" },
+    { url = "https://files.pythonhosted.org/packages/95/47/2d0564e93d87bc48618360ddca232a2ca612bbdf53ce8465d45ca5ce14ee/regex-2026.7.19-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:40b34dd88658e4fedd2fddbf0275ac970d00614b731357f425722a3ed1983d11", size = 291832, upload-time = "2026-07-19T00:17:40.726Z" },
+    { url = "https://files.pythonhosted.org/packages/07/cd/42dfbabff3dfc9603c501c0e2e2c5adbb09d127b267bf5348de0af338c15/regex-2026.7.19-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:0c41c63992bf1874cebb6e7f56fd7d3c007924659a604ae3d90e427d40d4fd13", size = 796775, upload-time = "2026-07-19T00:17:42.382Z" },
+    { url = "https://files.pythonhosted.org/packages/df/5d/f6a4839f2b934e3eed5973fd07f5929ee97d4c98939fb275ea23c274ee16/regex-2026.7.19-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:1d3372064506b94dd2c67c845f2db8062e9e9ba84d04e33cb96d7d33c11fe1ae", size = 865687, upload-time = "2026-07-19T00:17:44.185Z" },
+    { url = "https://files.pythonhosted.org/packages/14/b0/b47d6c36049bc59806a50bd4c86ced70bbe058d787f80281b1d7a9b0e024/regex-2026.7.19-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:fce7760bf283405b2c7999cab3da4e72f7deca6396013115e3f7a955db9760da", size = 911962, upload-time = "2026-07-19T00:17:46.442Z" },
+    { url = "https://files.pythonhosted.org/packages/2a/be/ff61f28f9273658cfe23acbbac5217221f6519960ed401e61dfdab12bc35/regex-2026.7.19-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:c0d702548d89d572b2929879bc883bb7a4c4709efafe4512cadee56c55c9bd15", size = 801817, upload-time = "2026-07-19T00:17:48.25Z" },
+    { url = "https://files.pythonhosted.org/packages/c3/bb/8b4f7f26b333f9f79e1b453613c39bb4776f51d38ae66dd0ba31d6b354ca/regex-2026.7.19-cp313-cp313-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:d446c6ac40bb6e05025ccee55b84d80fe9bf8e93010ffc4bb9484f13d498835f", size = 776908, upload-time = "2026-07-19T00:17:50.183Z" },
+    { url = "https://files.pythonhosted.org/packages/09/13/610110fc5921d380516d03c26b652555f08aa0d23ea78a771231873c3638/regex-2026.7.19-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:4c3501bfa814ab07b5580741f9bf78dfdfe146a04057f82df9e2402d2a975939", size = 784426, upload-time = "2026-07-19T00:17:52.454Z" },
+    { url = "https://files.pythonhosted.org/packages/ca/f5/1ef9e2a83a5947c57ebff0b377cb5727c3d5ec1992317a320d035cd0dbb6/regex-2026.7.19-cp313-cp313-musllinux_1_2_ppc64le.whl", hash = "sha256:c4585c3e64b4f9e583b4d2683f18f5d5d872b3d71dcf24594b74ecc23602fa96", size = 860600, upload-time = "2026-07-19T00:17:54.229Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/02/073af33a3ec149241d11c80acea91e722aa0adbf05addd50f251c4fe89c3/regex-2026.7.19-cp313-cp313-musllinux_1_2_riscv64.whl", hash = "sha256:571fde9741eb0ccde23dd4e0c1d50fbae910e901fa7e629faf39b2dda740d220", size = 765950, upload-time = "2026-07-19T00:17:56.041Z" },
+    { url = "https://files.pythonhosted.org/packages/81/a9/d1e9f819dc394a568ef370cd56cf25394e957a2235f8370f23b576e5a475/regex-2026.7.19-cp313-cp313-musllinux_1_2_s390x.whl", hash = "sha256:15b364b9b98d6d2fe1a85034c23a3180ff913f46caddc3895f6fd65186255ccc", size = 851794, upload-time = "2026-07-19T00:17:57.897Z" },
+    { url = "https://files.pythonhosted.org/packages/03/3a/8ae83eda7579feacdf984e71fb9e70635fb6f832eeddca58427ec4fca926/regex-2026.7.19-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:ffd8893ccc1c2fce6e0d6ca402d716fe1b29db70c7132609a05955e31b2aa8f2", size = 789845, upload-time = "2026-07-19T00:17:59.97Z" },
+    { url = "https://files.pythonhosted.org/packages/4b/23/c195cbfe5a75fdec64d8f6554fd15237b837919d2c61bdc141d7c807b08b/regex-2026.7.19-cp313-cp313-win32.whl", hash = "sha256:f0fa4fa9c3632d708742baf2282f2055c11d888a790362670a403cbf48a2c404", size = 267135, upload-time = "2026-07-19T00:18:01.958Z" },
+    { url = "https://files.pythonhosted.org/packages/b2/80/a11de8404b7272b70acb45c1c05987cce60b45d5693da2e176f0e390d564/regex-2026.7.19-cp313-cp313-win_amd64.whl", hash = "sha256:d51ffd3427640fa2da6ade574ceba932f210ad095f65fcc450a2b0a0d454868e", size = 277747, upload-time = "2026-07-19T00:18:04.121Z" },
+    { url = "https://files.pythonhosted.org/packages/d1/29/0f5c8eff1b4f1f3d83276d365fccecf666afcc7d947420943bf394d07adb/regex-2026.7.19-cp313-cp313-win_arm64.whl", hash = "sha256:c670fe7be5b6020b76bc6e8d2196074657e1327595bca93a389e1a76ab130ad8", size = 277129, upload-time = "2026-07-19T00:18:05.821Z" },
+    { url = "https://files.pythonhosted.org/packages/dc/4c/44b74742052cedda40f9ae469532a037112f7311a36669a891fba8984bb0/regex-2026.7.19-cp313-cp313t-macosx_10_13_universal2.whl", hash = "sha256:db47b561c9afd884baa1f96f797c9ca369872c4b65912bc691cfa99e68340af2", size = 501134, upload-time = "2026-07-19T00:18:07.567Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/45/bbd038b5e39ee5613a5a689290145b40058cc152c41de9cc23639d2b9734/regex-2026.7.19-cp313-cp313t-macosx_10_13_x86_64.whl", hash = "sha256:65dcd28d3eba2ab7c2fd906485cc301392b47cc2234790d27d4e4814e02cdfda", size = 299418, upload-time = "2026-07-19T00:18:09.38Z" },
+    { url = "https://files.pythonhosted.org/packages/65/38/c5bde94b4cedfd5850d64c3f08222d8e1600e84f6ee71d9b44b4b8163f74/regex-2026.7.19-cp313-cp313t-macosx_11_0_arm64.whl", hash = "sha256:f2e7f8e2ab6c2922be02c7ec45185aa5bd771e2e57b95455ee343a44d8130dff", size = 294486, upload-time = "2026-07-19T00:18:11.188Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/6a/2f5e107cb26c960b781967178899daf2787a7ab151844ed3c01d6fc95474/regex-2026.7.19-cp313-cp313t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:fe31f28c94402043161876a258a9c6f757cb485905c7614ce8d6cd40e6b7bdc1", size = 811643, upload-time = "2026-07-19T00:18:12.975Z" },
+    { url = "https://files.pythonhosted.org/packages/37/d4/a2f963406d7d73a62eed84ba05a258afb6cad1b21aa4517443ce40506b78/regex-2026.7.19-cp313-cp313t-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:f8f6fa298bb4f7f58a33334406218ba74716e68feddf5e4e54cd5d8082705abf", size = 871081, upload-time = "2026-07-19T00:18:14.733Z" },
+    { url = "https://files.pythonhosted.org/packages/45/a3/44be546340bedb15f13063f5e7fe16793ea4d9ea2e805d09bd174ac27724/regex-2026.7.19-cp313-cp313t-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:cc1b2440423a851fad781309dd87843868f4f66a6bcd1ddb9225cf4ec2c84732", size = 917372, upload-time = "2026-07-19T00:18:16.724Z" },
+    { url = "https://files.pythonhosted.org/packages/f8/f6/e0870b0fd2a40dba0074e4b76e514b21313d37946c9248453e34ec43923e/regex-2026.7.19-cp313-cp313t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:8ac59a0900474a52b7c04af8196affc22bd9842acb0950df12f7b813e983609a", size = 816089, upload-time = "2026-07-19T00:18:18.617Z" },
+    { url = "https://files.pythonhosted.org/packages/ae/27/957e8e22690ad6634572b39b71f130a6105f4d0718bb16849eac00fff147/regex-2026.7.19-cp313-cp313t-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:4896db1f4ce0576765b8272aa922df324e0f5b9bb2c3d03044ff32a7234a9aba", size = 785206, upload-time = "2026-07-19T00:18:20.464Z" },
+    { url = "https://files.pythonhosted.org/packages/76/a4/186e410941e731037c01166069ab86da9f65e8f8110c18009ccf4bd623ee/regex-2026.7.19-cp313-cp313t-musllinux_1_2_aarch64.whl", hash = "sha256:4e6883a021db30511d9fb8cfb0f222ce1f2c369f7d4d8b0448f449a93ba0bdfc", size = 800431, upload-time = "2026-07-19T00:18:22.716Z" },
+    { url = "https://files.pythonhosted.org/packages/73/9f/e4e10e023d291d64a33e246610b724493bf1ce98e0e59c9b7c837e5acfb7/regex-2026.7.19-cp313-cp313t-musllinux_1_2_ppc64le.whl", hash = "sha256:09523a592938aa9f587fb74467c63ff0cf88fc3df14c82ab0f0517dcf76aaa62", size = 864906, upload-time = "2026-07-19T00:18:24.772Z" },
+    { url = "https://files.pythonhosted.org/packages/24/57/ccb20b6be5f1f52a053d1ba2a8f7a077edb9d918248b8490d7506c6832b3/regex-2026.7.19-cp313-cp313t-musllinux_1_2_riscv64.whl", hash = "sha256:1ebac3474b8589fce2f9b225b650afd61448f7c73a5d0255a10cc6366471aed1", size = 773559, upload-time = "2026-07-19T00:18:27.008Z" },
+    { url = "https://files.pythonhosted.org/packages/a3/82/f3b263cf8fad927dc102891da8502e718b7ff9d19af7a2a07c03865d7188/regex-2026.7.19-cp313-cp313t-musllinux_1_2_s390x.whl", hash = "sha256:4a0530bb1b8c1c985e7e2122e2b4d3aedd8a3c21c6bfddae6767c4405668b56e", size = 857739, upload-time = "2026-07-19T00:18:29.107Z" },
+    { url = "https://files.pythonhosted.org/packages/47/2e/1687bd1b6c2aed5e672ccf845fc11557821fe7366d921b50889ea5ce57bf/regex-2026.7.19-cp313-cp313t-musllinux_1_2_x86_64.whl", hash = "sha256:2ef7eeb108c47ce7bcc9513e51bcb1bf57e8f483d52fce68a8642e3527141ae0", size = 804522, upload-time = "2026-07-19T00:18:31.362Z" },
+    { url = "https://files.pythonhosted.org/packages/76/7c/cc4e7655181b2d9235b704f2c5e19d8eff002bbc437bae59baee0e381aca/regex-2026.7.19-cp313-cp313t-win32.whl", hash = "sha256:64b6ca7391a1395c2638dd5c7456d67bea44fc6c5e8e92c5dc8aa6a8f23292b4", size = 269141, upload-time = "2026-07-19T00:18:33.479Z" },
+    { url = "https://files.pythonhosted.org/packages/bb/14/961b4c7b05a2391c32dbc85e27773076671ef8f97f36cec70fe414734c02/regex-2026.7.19-cp313-cp313t-win_amd64.whl", hash = "sha256:f04b9f56b0e0614c0126be12c2c2d9f8850c1e57af302bd0a63bed379d4af974", size = 280036, upload-time = "2026-07-19T00:18:35.419Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/67/795644550d788ddbb6dc458c95895f8009978ea6d6ea76b005eb3f45e8c9/regex-2026.7.19-cp313-cp313t-win_arm64.whl", hash = "sha256:fcee38cd8e5089d6d4f048ba1233b3ad76e5954f545382180889112ff5cb712d", size = 279394, upload-time = "2026-07-19T00:18:37.454Z" },
+    { url = "https://files.pythonhosted.org/packages/d2/25/0c4c452f8ef3efe456745b2f33195f5904b573fb4c2ff3f0cb9ec188461e/regex-2026.7.19-cp314-cp314-macosx_10_13_universal2.whl", hash = "sha256:a81758ed242b861b72e778ba34d41366441a2e10b16b472784c88da2dea7e2dd", size = 496750, upload-time = "2026-07-19T00:18:39.633Z" },
+    { url = "https://files.pythonhosted.org/packages/24/9e/b70ca6c1704f6c7cd32a9e143c86cc5968d10981eca284bad670c245ea7d/regex-2026.7.19-cp314-cp314-macosx_10_13_x86_64.whl", hash = "sha256:4aa5435cdb3eb6f55fe98a171b05e3fbcd95fadaa4aa32acf62afd9b0cfdbcac", size = 297093, upload-time = "2026-07-19T00:18:41.583Z" },
+    { url = "https://files.pythonhosted.org/packages/87/74/0b692da2520d51fbff19c88b83d97e4c702909dd02386c585998b7e2dbed/regex-2026.7.19-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:60be8693a1dadc210bbcbc0db3e26da5f7d01d1d5a3da594e99b4fa42df404f5", size = 292043, upload-time = "2026-07-19T00:18:43.347Z" },
+    { url = "https://files.pythonhosted.org/packages/e3/a7/1d478e614016045a33feae57446215f9fd65b665a5ceb2f891fb3183bc52/regex-2026.7.19-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:d19662dbedbe783d323196312d38f5ba53cf56296378252171985da6899887d3", size = 797214, upload-time = "2026-07-19T00:18:45.362Z" },
+    { url = "https://files.pythonhosted.org/packages/aa/ae/11b9c9411d92c30e3d2db32df5a31133e4a99a8fc397a604fd08f6c4bffb/regex-2026.7.19-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:d15df07081d91b76ff20d43f94592ee110330152d617b730fdbe5ef9fb680053", size = 866433, upload-time = "2026-07-19T00:18:47.315Z" },
+    { url = "https://files.pythonhosted.org/packages/b1/62/2b2efc4992f91d6d204b24c647c9f9412e85379d92b7c0ab9fdae622327e/regex-2026.7.19-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:56ad4d9f77df871a99e25c37091052a02528ec0eb059de928ee33956b854b45b", size = 911360, upload-time = "2026-07-19T00:18:49.588Z" },
+    { url = "https://files.pythonhosted.org/packages/14/71/986ceea9aa3da548bf1357cad89b63915ec6d21ec957c8113b29ece567df/regex-2026.7.19-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:7322ec6cc9fba9d49ab888bb82d67ac5625627aa168f0165139b17018df3fb8a", size = 801275, upload-time = "2026-07-19T00:18:51.767Z" },
+    { url = "https://files.pythonhosted.org/packages/15/be/ce9d9534b2cda96eab32c548261224b9b4e220a4126f098f60f42ae7b4cd/regex-2026.7.19-cp314-cp314-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:9c7472192ebfad53a6be7c4a8bfb2d64b81c0e93a1fc8c57e1dd0b638297b5d1", size = 777131, upload-time = "2026-07-19T00:18:54.053Z" },
+    { url = "https://files.pythonhosted.org/packages/61/2b/58b5c710f2c3929515a25f3a1ca0dad0dcd4518d4fff3cf23bc7adb8dcd2/regex-2026.7.19-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:c10b82c2634df08dfb13b1f04e38fe310d086ee092f4f69c0c8da234251e556e", size = 785020, upload-time = "2026-07-19T00:18:56.579Z" },
+    { url = "https://files.pythonhosted.org/packages/84/03/5fe091935b74f15fe0f97998c215cae418d1c0413f6258c7d4d2e83aa37f/regex-2026.7.19-cp314-cp314-musllinux_1_2_ppc64le.whl", hash = "sha256:17ed5692f6acc4183e98331101a5f9e4f64d72fe58b753da4d444a2c77d05b12", size = 861263, upload-time = "2026-07-19T00:18:58.64Z" },
+    { url = "https://files.pythonhosted.org/packages/d8/fa/d60bf82e10841eef62a9e32aac401468f05fddfbcb2942e342b1ba3d2433/regex-2026.7.19-cp314-cp314-musllinux_1_2_riscv64.whl", hash = "sha256:22a992de9a0d91bda927bf02b94351d737a0302905432c88a53de7c4b9ce62e2", size = 766199, upload-time = "2026-07-19T00:19:00.705Z" },
+    { url = "https://files.pythonhosted.org/packages/bf/5d/11e64d151b0662b81d6bf644c74dc118d461df85bdf2577fadbbf751788a/regex-2026.7.19-cp314-cp314-musllinux_1_2_s390x.whl", hash = "sha256:618a0aed532be87294c4477b0481f3aa0f1520f4014a4374dd4cf789b4cd2c97", size = 851317, upload-time = "2026-07-19T00:19:03.015Z" },
+    { url = "https://files.pythonhosted.org/packages/7c/34/532efb87488d90807bae6a443d357ee5e2728a478c597619c8aaa17cc0bd/regex-2026.7.19-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:2ce9e679f776649746729b6c86382da519ef649c8e34cc41df0d2e5e0f6c36d4", size = 789557, upload-time = "2026-07-19T00:19:05.338Z" },
+    { url = "https://files.pythonhosted.org/packages/d6/90/3a8d5ca977171ec3ae21a71207d2228b2663bde14d7f7ef0e6363ecf9290/regex-2026.7.19-cp314-cp314-win32.whl", hash = "sha256:73f272fba87b8ccfe70a137d02a54af386f6d27aa509fbffdd978f5947aae1aa", size = 272531, upload-time = "2026-07-19T00:19:07.487Z" },
+    { url = "https://files.pythonhosted.org/packages/96/e1/8862885e70409de70e8c005f57fb2e7be8d9ef0317250d60f4c9660a300d/regex-2026.7.19-cp314-cp314-win_amd64.whl", hash = "sha256:d721e53758b2cca74990185eb0671dd466d7a388a1a45d0c6f4c13cef41a68ac", size = 280831, upload-time = "2026-07-19T00:19:09.46Z" },
+    { url = "https://files.pythonhosted.org/packages/08/82/2693e53e29f9104d9de95d37ce4dd826bd32d5f9c0085d3aa6ac042675c4/regex-2026.7.19-cp314-cp314-win_arm64.whl", hash = "sha256:65fa6cb38ed5e9c3637e68e544f598b39c3b86b808ed0627a67b68320384b459", size = 281099, upload-time = "2026-07-19T00:19:11.398Z" },
+    { url = "https://files.pythonhosted.org/packages/92/b7/9a01aa16461a18cde9d7b9c3ab21e501db2ce33725f53014342b91df2b0a/regex-2026.7.19-cp314-cp314t-macosx_10_13_universal2.whl", hash = "sha256:5a2721c8720e2cb3c209925dfb9200199b4b07361c9e01d321719404b21458b3", size = 501121, upload-time = "2026-07-19T00:19:13.425Z" },
+    { url = "https://files.pythonhosted.org/packages/f3/5e/bbaeca815dc9191c424c94a4fdc5c87c75748a64a6271821212ebdd4e1a3/regex-2026.7.19-cp314-cp314t-macosx_10_13_x86_64.whl", hash = "sha256:199535629f25caf89698039af3d1ad5fcae7f933e2112c73f1cdf49165c99518", size = 299415, upload-time = "2026-07-19T00:19:15.43Z" },
+    { url = "https://files.pythonhosted.org/packages/cd/d6/0dd1a321afaab95eb7ff44aa0f637301786f1dc71c6b797b9ed236ed8890/regex-2026.7.19-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:9b60d7814174f059e5de4ab98271cc5ba9259cfea55273a81544dceea32dc8d9", size = 294483, upload-time = "2026-07-19T00:19:17.879Z" },
+    { url = "https://files.pythonhosted.org/packages/92/5f/40bacf91d0904f812e13bbbab3864604c463eced8afdc54aeaa50492ea95/regex-2026.7.19-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:dbece16025afda5e3031af0c4059207e61dcf73ef13af844964f57f387d1c435", size = 811833, upload-time = "2026-07-19T00:19:20.102Z" },
+    { url = "https://files.pythonhosted.org/packages/94/7c/4902744261f775aeede8b5627314b38482da29cf49a57b66a6fb753246c5/regex-2026.7.19-cp314-cp314t-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:d24ecb4f5e009ea0bd275ee37ad9953b32005e2e5e60f8bbae16da0dbbf0d3a0", size = 871270, upload-time = "2026-07-19T00:19:22.365Z" },
+    { url = "https://files.pythonhosted.org/packages/16/70/6980c9be6bf21c0a60ed3e0aea39cf419ecf3b08d1d9947bc56e196ef186/regex-2026.7.19-cp314-cp314t-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:8cae6fd77a5b72dae505084b1a2ee0360139faf72fedbab667cd7cc65aae7a6a", size = 917534, upload-time = "2026-07-19T00:19:24.529Z" },
+    { url = "https://files.pythonhosted.org/packages/52/92/8b2bd872782ce8c42691e39acb38eb8efe014e5ddb78ad7d943d6f197ce9/regex-2026.7.19-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:9724e6cb5e478cd7d8cabf027826178739cb18cf0e117d0e32814d479fa02276", size = 816135, upload-time = "2026-07-19T00:19:26.919Z" },
+    { url = "https://files.pythonhosted.org/packages/de/2d/33a602f657bdc4041f17d79f92ab18261d255d91a06117a6e29df023e5e2/regex-2026.7.19-cp314-cp314t-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:572fc57b0009c735ee56c175ea021b637a15551a312f56734277f923d6fd0f6c", size = 785492, upload-time = "2026-07-19T00:19:29.192Z" },
+    { url = "https://files.pythonhosted.org/packages/9e/36/0987cf4cb271680064a70d24a475873775a151d0b7058698a006cb0cae4a/regex-2026.7.19-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:20568e182eb82d39a6bf7cff3fd58566f14c75c6f74b2c8c96537eecf9010e3a", size = 800658, upload-time = "2026-07-19T00:19:31.392Z" },
+    { url = "https://files.pythonhosted.org/packages/a8/24/c14f31c135e1ba55fa4f9a58ca98d0842512bf6188230763c31c8f449e3b/regex-2026.7.19-cp314-cp314t-musllinux_1_2_ppc64le.whl", hash = "sha256:1d58561843f0ff7dc78b4c28b5e2dc388f3eff94ebc8a232a3adba961fc00009", size = 865073, upload-time = "2026-07-19T00:19:33.485Z" },
+    { url = "https://files.pythonhosted.org/packages/14/85/181a12211f22469f24d2de1ebddfe397d2396e2c29013b9a58134a91069a/regex-2026.7.19-cp314-cp314t-musllinux_1_2_riscv64.whl", hash = "sha256:61bb1bd45520aacd56dd80943bd34991fb5350afdd1f36f2282230fd5154a218", size = 773684, upload-time = "2026-07-19T00:19:35.599Z" },
+    { url = "https://files.pythonhosted.org/packages/23/58/bd1a0c1a62251366f8d21f41b1ea3c76994962071b8b6ea42f72d505c0f0/regex-2026.7.19-cp314-cp314t-musllinux_1_2_s390x.whl", hash = "sha256:cd3584591ea4429026cdb931b054342c2bcf189b44ff367f8d5c15bc092a2966", size = 857769, upload-time = "2026-07-19T00:19:37.738Z" },
+    { url = "https://files.pythonhosted.org/packages/e4/4f/f7e2dad6756b2fe1fe75dd90a628c3b45f249d39f948dd90cd2476325417/regex-2026.7.19-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:5cc26a66e212fa5d6c6170c3a40d99d888db3020c6fdab1523250d4341382e44", size = 804546, upload-time = "2026-07-19T00:19:40.229Z" },
+    { url = "https://files.pythonhosted.org/packages/2b/d7/01d31d5bdb09bc026fab77f59a371fdf8f9b292e4810546c56182ca70498/regex-2026.7.19-cp314-cp314t-win32.whl", hash = "sha256:2c4e61e2e1be56f63ec3cc618aa9e0de81ef6f43d177205451840022e24f5b78", size = 274526, upload-time = "2026-07-19T00:19:42.398Z" },
+    { url = "https://files.pythonhosted.org/packages/52/0e/cea4ce73bc0a8247a0748228ae6669984c7e1f8134b6fa66e59c0572e0ea/regex-2026.7.19-cp314-cp314t-win_amd64.whl", hash = "sha256:c639ea314df70a7b2811e8020448c75af8c9445f5a60f8a4ced81c306a9380c2", size = 283763, upload-time = "2026-07-19T00:19:44.644Z" },
+    { url = "https://files.pythonhosted.org/packages/6f/b6/26e41975febae63b7a6e3e02f32cff6cff2e4f10d19c929082f56aebf7c6/regex-2026.7.19-cp314-cp314t-win_arm64.whl", hash = "sha256:9a15e785f244f3e07847b984ce8773fc3da10a9f3c131cc49a4c5b4d672b4547", size = 283451, upload-time = "2026-07-19T00:19:46.639Z" },
+]
+
+[[package]]
+name = "requests"
+version = "2.34.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "certifi" },
+    { name = "charset-normalizer" },
+    { name = "idna" },
+    { name = "urllib3" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/ac/c3/e2a2b89f2d3e2179abd6d00ebd70bff6273f37fb3e0cc209f48b39d00cbf/requests-2.34.2.tar.gz", hash = "sha256:f288924cae4e29463698d6d60bc6a4da69c89185ad1e0bcc4104f584e960b9ed", size = 142856, upload-time = "2026-05-14T19:25:27.735Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a0/f4/c67b0b3f1b9245e8d266f0f112c500d50e5b4e83cb6f3b71b6528104182a/requests-2.34.2-py3-none-any.whl", hash = "sha256:2a0d60c172f83ac6ab31e4554906c0f3b3588d37b5cb939b1c061f4907e278e0", size = 73075, upload-time = "2026-05-14T19:25:26.443Z" },
+]
+
+[[package]]
+name = "requests-toolbelt"
+version = "1.0.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "requests" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/f3/61/d7545dafb7ac2230c70d38d31cbfe4cc64f7144dc41f6e4e4b78ecd9f5bb/requests-toolbelt-1.0.0.tar.gz", hash = "sha256:7681a0a3d047012b5bdc0ee37d7f8f07ebe76ab08caeccfc3921ce23c88d5bc6", size = 206888, upload-time = "2023-05-01T04:11:33.229Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/3f/51/d4db610ef29373b879047326cbf6fa98b6c1969d6f6dc423279de2b1be2c/requests_toolbelt-1.0.0-py2.py3-none-any.whl", hash = "sha256:cccfdd665f0a24fcf4726e690f65639d272bb0637b9b92dfd91a5568ccf6bd06", size = 54481, upload-time = "2023-05-01T04:11:28.427Z" },
+]
+
+[[package]]
+name = "rich"
+version = "15.0.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "markdown-it-py" },
+    { name = "pygments" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/c0/8f/0722ca900cc807c13a6a0c696dacf35430f72e0ec571c4275d2371fca3e9/rich-15.0.0.tar.gz", hash = "sha256:edd07a4824c6b40189fb7ac9bc4c52536e9780fbbfbddf6f1e2502c31b068c36", size = 230680, upload-time = "2026-04-12T08:24:00.75Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/82/3b/64d4899d73f91ba49a8c18a8ff3f0ea8f1c1d75481760df8c68ef5235bf5/rich-15.0.0-py3-none-any.whl", hash = "sha256:33bd4ef74232fb73fe9279a257718407f169c09b78a87ad3d296f548e27de0bb", size = 310654, upload-time = "2026-04-12T08:24:02.83Z" },
+]
+
+[[package]]
+name = "rpds-py"
+version = "2026.6.3"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/aa/2a/9618a122aeb2a169a28b03889a2995fe297588964333d4a7d67bdf46e147/rpds_py-2026.6.3.tar.gz", hash = "sha256:1cebd1337c242e4ec2293e541f712b2da849b29f48f0c293684b71c0632625d4", size = 64051, upload-time = "2026-06-30T07:17:53.009Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/5c/be/2e8974163072e7bab7df1a5acd54c4498e75e35d6d18b864d3a9d5dadc92/rpds_py-2026.6.3-cp312-cp312-macosx_10_12_x86_64.whl", hash = "sha256:a0811d33247c3d6128a3001d763f2aa056bb3425204335400ac54f89eec3a0d0", size = 343691, upload-time = "2026-06-30T07:15:14.96Z" },
+    { url = "https://files.pythonhosted.org/packages/a4/73/319dfa745dd668efe89309141ded489126461fcecd2b8f3a3cda185129b6/rpds_py-2026.6.3-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:538949e262e46caa31ac01bdb3c1e8f642622922cacbabbae6a8445d9dc33eaf", size = 338542, upload-time = "2026-06-30T07:15:16.267Z" },
+    { url = "https://files.pythonhosted.org/packages/21/63/4239893be1c4d09b709b1a8f6be4188f0870084ff547f46606b8a75f1b03/rpds_py-2026.6.3-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:55927d532399c2c646100ff7feb48eaa940ad70f42cd68e1328f3ded9f81ca24", size = 368180, upload-time = "2026-06-30T07:15:17.62Z" },
+    { url = "https://files.pythonhosted.org/packages/1c/ca/9c5de382225234ceb37b1844ebdb140db12b2a278bb9efe2fcd19f6c82ce/rpds_py-2026.6.3-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:f56f1695bc5c0871cbc33dc0130fcf503aab0c57dcc5a6700a4f49eba4f2652e", size = 375067, upload-time = "2026-06-30T07:15:18.952Z" },
+    { url = "https://files.pythonhosted.org/packages/87/dc/863f69d1bf04ade34b7fe0d59b9fdf6f0135fe2d7cbca74f1d665589559d/rpds_py-2026.6.3-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:270b293dae9058fc9fcedab50f13cebf46fb8ed1d1d54e0521a9da5d6b211975", size = 490509, upload-time = "2026-06-30T07:15:20.434Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/ef/eac16a12048b45ec7c7fa94f2be3438a5f26bf9cc8580b18a1cfd609b7f6/rpds_py-2026.6.3-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:127565fead0a10943b282957bd5447804ff3160ad79f2ad2635e6d249e380680", size = 382754, upload-time = "2026-06-30T07:15:21.831Z" },
+    { url = "https://files.pythonhosted.org/packages/04/8f/d2f3f532616be4d06c316ef119683e832bd3d41e112bf3a88f4151c95b17/rpds_py-2026.6.3-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ecabd69db66de867690f9797f2f8fa27ba501bbc24540cbdbdc649cd15888ba6", size = 366189, upload-time = "2026-06-30T07:15:23.371Z" },
+    { url = "https://files.pythonhosted.org/packages/e3/29/41a7b0e98a4b44cd676ab7598419623373eb43b20be68c084935c1a8cf88/rpds_py-2026.6.3-cp312-cp312-manylinux_2_31_riscv64.whl", hash = "sha256:58eadac9cd119677b60e1cf8ac4052f35949d71b8a9e5556efccbe82533cf22a", size = 377750, upload-time = "2026-06-30T07:15:24.659Z" },
+    { url = "https://files.pythonhosted.org/packages/2e/05/ecda0bec46f9a1565090bcdc941d023f6a25aff85fda28f89f8d19878152/rpds_py-2026.6.3-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:7491ee23305ac3eb59e492b6945881f5cd77a6f731061a3f25b77fd40f9e99a4", size = 395576, upload-time = "2026-06-30T07:15:25.987Z" },
+    { url = "https://files.pythonhosted.org/packages/68/a8/6ed52f03ee6cb854ce78785cc9a9a672eb880e83fd7224d471f667d151f1/rpds_py-2026.6.3-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:2c99f7e8ccb3dd6e3e4bfeac657a7b208c9bac8075f4b078c02d7404c34107fa", size = 543807, upload-time = "2026-06-30T07:15:27.356Z" },
+    { url = "https://files.pythonhosted.org/packages/8f/d6/156c0d3eea27ba09b92562ba2364ba124c0a061b199e17eac637cd25a5e2/rpds_py-2026.6.3-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:62698275682bf121181861295c9181e789030a2d516071f5b8f3c23c170cd0fc", size = 611187, upload-time = "2026-06-30T07:15:28.931Z" },
+    { url = "https://files.pythonhosted.org/packages/f1/31/774212ed989c62f7f310220089f9b0a3fb8f40f5443d1727abd5d9f52bc9/rpds_py-2026.6.3-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:a214c993455f99a89aaeadc9b21241900037adc9d97203e374d75513c5911822", size = 573030, upload-time = "2026-06-30T07:15:30.553Z" },
+    { url = "https://files.pythonhosted.org/packages/c9/50/22f73127a41f1ce4f87fe39aadfb9a126345801c274aa93ae88456249327/rpds_py-2026.6.3-cp312-cp312-win32.whl", hash = "sha256:501f9f04a588d6a09179368c57071301445191767c64e4b52a6aa9871f1ef5ed", size = 202185, upload-time = "2026-06-30T07:15:32.027Z" },
+    { url = "https://files.pythonhosted.org/packages/04/3a/f0ee4d4dde9d3b69dedf1b5f74e7a40017046d55052d173e418c6a94f960/rpds_py-2026.6.3-cp312-cp312-win_amd64.whl", hash = "sha256:2c958bf94822e9290a40aaf2a822d4bc5c88099093e3948ad6c571eca9272e5f", size = 220394, upload-time = "2026-06-30T07:15:33.359Z" },
+    { url = "https://files.pythonhosted.org/packages/f3/83/3382fe37f809b59f02aac04dbc4e765b480b46ee0227ed516e3bdc4d3dfc/rpds_py-2026.6.3-cp312-cp312-win_arm64.whl", hash = "sha256:22bffe6042b9bcb0822bcd1955ec00e245daf17b4344e4ed8e9551b976b63e96", size = 215753, upload-time = "2026-06-30T07:15:34.778Z" },
+    { url = "https://files.pythonhosted.org/packages/a4/9e/b818ee580026ec578138e961027a68820c40afeb1ec8f6819b54fb99e196/rpds_py-2026.6.3-cp313-cp313-macosx_10_12_x86_64.whl", hash = "sha256:3cfe765c1da0072636ca06628261e0ea05688e160d5c8a03e0217c3854037223", size = 343012, upload-time = "2026-06-30T07:15:36.005Z" },
+    { url = "https://files.pythonhosted.org/packages/f3/6b/686d9dc4359a8f163cfbbf89ee0b4e586431de22fe8248edb63a8cf50d49/rpds_py-2026.6.3-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:f4d78253f6996be4901669ad25319f842f740eccf4d58e3c7f3dd39e6dde1d8f", size = 338203, upload-time = "2026-06-30T07:15:37.462Z" },
+    { url = "https://files.pythonhosted.org/packages/9e/9b/069aa329940f8207615e091f5eedbbd40e1e15eac68a0790fd05ccdf796c/rpds_py-2026.6.3-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:54f45a148e28767bf343d33a684693c70e451c6f4c0e9904709a723fafbdfc1f", size = 367984, upload-time = "2026-06-30T07:15:39.008Z" },
+    { url = "https://files.pythonhosted.org/packages/14/db/34c203e4becff3703e4d3bc121842c00b8689197f398161203a880052f4e/rpds_py-2026.6.3-cp313-cp313-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:842e7b070435622248c7a2c44ae53fa1440e073cc3023bc919fed570884097a7", size = 374815, upload-time = "2026-06-30T07:15:40.253Z" },
+    { url = "https://files.pythonhosted.org/packages/ee/7d/8071067d2cc453d916ad836e828c943f575e8a44612537759002a1e07381/rpds_py-2026.6.3-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:8020133a74bd81b4572dd8e4be028a6b1ebcd70e6726edc3918008c08bee6ee6", size = 490545, upload-time = "2026-06-30T07:15:41.729Z" },
+    { url = "https://files.pythonhosted.org/packages/a3/42/da06c5aa8f0484ff07f270787434204d9f4535e2f8c3b51ed402267e63c3/rpds_py-2026.6.3-cp313-cp313-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:cdc7e35386f3847df728fbcb5e887e2d79c19e2fa1eba9e51b6621d23e3243af", size = 382828, upload-time = "2026-06-30T07:15:43.327Z" },
+    { url = "https://files.pythonhosted.org/packages/57/d7/fe978efc2ae50abe48eb7464668ea99f53c010c60aeebb7b35ad27f23661/rpds_py-2026.6.3-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:acac386b453c2516111b50985d60ce46e7fadb5ea71ae7b25f4c946935bf27cf", size = 365678, upload-time = "2026-06-30T07:15:44.992Z" },
+    { url = "https://files.pythonhosted.org/packages/69/9d/1d8922e1990b2a6eb532b6ff53d3e73d2b3bbffc84116c75826bee73dfc6/rpds_py-2026.6.3-cp313-cp313-manylinux_2_31_riscv64.whl", hash = "sha256:425560c6fa0415f27261727bb20bd097568485e5eb0c121f1949417d1c516885", size = 377811, upload-time = "2026-06-30T07:15:46.523Z" },
+    { url = "https://files.pythonhosted.org/packages/b1/3d/198dceafb4fb034a6a47347e1b0735d34e0bd4a50be4e898d408ee66cb14/rpds_py-2026.6.3-cp313-cp313-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:a550fb4950a06dde3beb4721f5ad4b25bf4513784665b0a8522c792e2bd822a4", size = 395382, upload-time = "2026-06-30T07:15:47.955Z" },
+    { url = "https://files.pythonhosted.org/packages/1f/f1/13968e49655d40b6b19d8b9140296bbc6f1d86b3f0f6c346cf9f1adddf4b/rpds_py-2026.6.3-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:4f4bca01b63096f606e095734dd56e74e175f94cfbf24ff3d63281cec61f7bb7", size = 543832, upload-time = "2026-06-30T07:15:49.33Z" },
+    { url = "https://files.pythonhosted.org/packages/ac/ab/289bcb1b90bd3e40a2900c561fa0e2087345ecbb094f0b870f2345142b7c/rpds_py-2026.6.3-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:ccffae9a092a00deb7efd545fe5e2c33c33b88e7c054337e9a74c179347d0b7d", size = 611011, upload-time = "2026-06-30T07:15:50.847Z" },
+    { url = "https://files.pythonhosted.org/packages/1e/16/5043105e679436ccfbc8e5e0dd2d663ed18a8b8113515fd06a5e5d77c83e/rpds_py-2026.6.3-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:1cf01971c4f2c5553b772a542e4aaf191789cd331bc2cd4ff0e6e65ba49e1e97", size = 572431, upload-time = "2026-06-30T07:15:52.394Z" },
+    { url = "https://files.pythonhosted.org/packages/85/ed/adab103321c0a6565d5ae1c2998349bc3ee175b82ccc5ae8fc04cc413075/rpds_py-2026.6.3-cp313-cp313-win32.whl", hash = "sha256:8c3d1e9c15b9d51ca0391e13da1a25a0a4df3c58a37c9dc368e0736cf7f69df0", size = 201710, upload-time = "2026-06-30T07:15:53.894Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/ed/a03b09668e74e5dabbf2e211f6468e1820c0552f7b0500082da31841bf7b/rpds_py-2026.6.3-cp313-cp313-win_amd64.whl", hash = "sha256:9250a9a0a6fd4648b3f868da8d91a4c52b5811a62df58e753d50ae4454a36f80", size = 219454, upload-time = "2026-06-30T07:15:55.25Z" },
+    { url = "https://files.pythonhosted.org/packages/27/17/b8642c12930b71bc2b25831f6708ccf0f75abcd11883932ec9ce54ba3a78/rpds_py-2026.6.3-cp313-cp313-win_arm64.whl", hash = "sha256:900a67df3fd1660b035a4761c4ce73c382ea6b35f90f9863c36c6fd8bf8b09bb", size = 215063, upload-time = "2026-06-30T07:15:56.573Z" },
+    { url = "https://files.pythonhosted.org/packages/b6/36/7fbe9dcdaf857fb3f63c2a2284b62492d95f5e8334e947e5fb6e7f68c9be/rpds_py-2026.6.3-cp314-cp314-macosx_10_12_x86_64.whl", hash = "sha256:931908d9fc855d8f74783377822be318edb6dcb19e47169dc038f9a1bf60b06e", size = 344510, upload-time = "2026-06-30T07:15:57.921Z" },
+    { url = "https://files.pythonhosted.org/packages/ba/54/f785cc3d3f60839ca57a5af4927a9f347b07b2799c373fc20f7949f87c7e/rpds_py-2026.6.3-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:d7469697dce35be237db177d42e2a2ee26e6dcc5fc052078a6fefabd288c6edd", size = 339495, upload-time = "2026-06-30T07:15:59.238Z" },
+    { url = "https://files.pythonhosted.org/packages/63/ef/d4cdaf309e6b095b43597103cf8c0b951d6cca2acce68c474f75ec12e0c7/rpds_py-2026.6.3-cp314-cp314-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:bcfbcf66006befb9fd2aeaa9e01feaf881b4dc330a02ba07d2322b1c11be7b5d", size = 369454, upload-time = "2026-06-30T07:16:01.021Z" },
+    { url = "https://files.pythonhosted.org/packages/96/4a/9559a68b7ee15db09d7981212e8c2e219d2a1d6d4faa0391d813c3496a36/rpds_py-2026.6.3-cp314-cp314-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:847927daf4cffbd4e90e42bc890069897101edd015f956cb8721b3473372edda", size = 374583, upload-time = "2026-06-30T07:16:02.287Z" },
+    { url = "https://files.pythonhosted.org/packages/ef/75/8964aa7d2c6e8ac43eba8eb6e6b0fdda1f46d39f2fc3e6aa9f2cb17f485d/rpds_py-2026.6.3-cp314-cp314-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:aca6c1ef08a82bfe327cc156da694660f599923e2e6665b6d81c9c2d0ac9ffc8", size = 492919, upload-time = "2026-06-30T07:16:03.723Z" },
+    { url = "https://files.pythonhosted.org/packages/8f/97/6908094ac804115e65aedfd90f1b5fee4eebebd3f6c4cfc5419939267565/rpds_py-2026.6.3-cp314-cp314-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:ae50181a047c871561212bb97f7932a2d45fb53e947bd9b57ebad85b529cbc53", size = 383725, upload-time = "2026-06-30T07:16:05.305Z" },
+    { url = "https://files.pythonhosted.org/packages/d1/9c/0d1fdc2e7aba23e290d603bc494e97bd205bae262ce33c6b32a69768ed5e/rpds_py-2026.6.3-cp314-cp314-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:dc319e5a1de4b6913aac94bf6a2f9e847371e0a140a43dd4991db1a09bc2d504", size = 367255, upload-time = "2026-06-30T07:16:07.086Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/fe/f0209ca4a9ed074bc8acb44dfd0e81c3122e94c9689f5645b7973a866719/rpds_py-2026.6.3-cp314-cp314-manylinux_2_31_riscv64.whl", hash = "sha256:e4316bf32babbed84e691e352faf967ce2f0f024174a8643c37c94a1080374fc", size = 379060, upload-time = "2026-06-30T07:16:08.525Z" },
+    { url = "https://files.pythonhosted.org/packages/c6/8d/f1cc54c616b9d8897de8738aac148d20afca93f68187475fe194d09a71b9/rpds_py-2026.6.3-cp314-cp314-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:8c6e5a2f750cc71c3e3b11d71661f21d6f9bc6cebc6564b1466417a1ec03ec77", size = 395960, upload-time = "2026-06-30T07:16:09.989Z" },
+    { url = "https://files.pythonhosted.org/packages/fb/04/aafff00f73aeca2945f734f1d483c64ab8f472d0864ab02377fd8e89c3b2/rpds_py-2026.6.3-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:4470ce197d4090875cf6affbf1f853338387428df97c4fb7b7106317b8214698", size = 545356, upload-time = "2026-06-30T07:16:11.816Z" },
+    { url = "https://files.pythonhosted.org/packages/fd/cc/e229663b9e4ddac5a4acbe9085dd80a71af2a5d356b8b39d6bff233f24b0/rpds_py-2026.6.3-cp314-cp314-musllinux_1_2_i686.whl", hash = "sha256:ea964164cc9afa72d4d9b23cc28dafae93693c0a53e0b42acbff15b22c3f9ddd", size = 612319, upload-time = "2026-06-30T07:16:13.586Z" },
+    { url = "https://files.pythonhosted.org/packages/e3/7a/8a0e6d3e6cd066af108b71b43122c3fe158dd9eb86acac626593a2582eb1/rpds_py-2026.6.3-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:639c8929aa0afe81be836b04de888460d6bed38b9c54cfc18da8f6bfabf5af5d", size = 573508, upload-time = "2026-06-30T07:16:15.23Z" },
+    { url = "https://files.pythonhosted.org/packages/87/03/2a69ab618a789cf6cf85c86bb844c62d090e700ab1a2aa676b3741b6c516/rpds_py-2026.6.3-cp314-cp314-win32.whl", hash = "sha256:882076c00c0a608b131187055ddc5ae29f2e7eaf870d6168980420d58528a5c8", size = 202504, upload-time = "2026-06-30T07:16:16.893Z" },
+    { url = "https://files.pythonhosted.org/packages/85/62/a3892ba945f4e24c78f352e5de3c7620d8479f73f211406a97263d13c7d2/rpds_py-2026.6.3-cp314-cp314-win_amd64.whl", hash = "sha256:0be972be84cfcaf46c8c6edf690ca0f154ac17babf1f6a955a51579b34ad2dc5", size = 220380, upload-time = "2026-06-30T07:16:18.108Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/e7/c2bd44dc831931815ad11ebb5f430b5a0a4d3caa9de837107876c30c3432/rpds_py-2026.6.3-cp314-cp314-win_arm64.whl", hash = "sha256:2a9c6f195058cb45335e8cc3802745c603d716eb96bc9625950c1aac71c0c703", size = 215976, upload-time = "2026-06-30T07:16:19.654Z" },
+    { url = "https://files.pythonhosted.org/packages/79/9c/fff7b74bce9a091ec9a012a03f9ff5f69364eaf9451060dfc4486da2ffdd/rpds_py-2026.6.3-cp314-cp314t-macosx_10_12_x86_64.whl", hash = "sha256:f90938e92afda60266da758ee7d363447f7f0138c9559f9e1811629580582d90", size = 346840, upload-time = "2026-06-30T07:16:21.268Z" },
+    { url = "https://files.pythonhosted.org/packages/e9/44/77bcb1168b33704908295533d27f10eb811e9e3e193e8993dc99572211d3/rpds_py-2026.6.3-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:ec829541c45bca16e61c7ae50c20501f213605beb75d1aba91a6ee37fbbb56a4", size = 340282, upload-time = "2026-06-30T07:16:22.875Z" },
+    { url = "https://files.pythonhosted.org/packages/87/3c/7a9081c7c9e645b39efe19e4ffbeccd80add246327cd9b888aecffd72317/rpds_py-2026.6.3-cp314-cp314t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:afd70d95892096cdb26f15a00c45907b17817577aa8d1c76b2dcc2788391f9e9", size = 370403, upload-time = "2026-06-30T07:16:24.415Z" },
+    { url = "https://files.pythonhosted.org/packages/f7/69/af47021eb7dad6ff3396cb001c08f0f3c4d06c20253f75be6421a59fe6b7/rpds_py-2026.6.3-cp314-cp314t-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:29dfa0533a5d4c94d4dfa1b694fcb56c9c63aad8330ffdd816fd225d0a7a162f", size = 376055, upload-time = "2026-06-30T07:16:26.111Z" },
+    { url = "https://files.pythonhosted.org/packages/81/fc/a3bcf517084396a6dd258c592567a3c011ba4557f2fde23dceaf26e74f2e/rpds_py-2026.6.3-cp314-cp314t-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:af05d726809bff6b141be124d4c7ce998f9c9c7f30edb1f46c07aa103d540b41", size = 494419, upload-time = "2026-06-30T07:16:27.596Z" },
+    { url = "https://files.pythonhosted.org/packages/c9/eb/13d529d1788135425c7bf207f8463458ca5d92e43f3f701365b83e9dffc1/rpds_py-2026.6.3-cp314-cp314t-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:9826217f048f620d9a712672818bf231442c1b35d96b227a07eabd11b4bb6945", size = 384848, upload-time = "2026-06-30T07:16:29.183Z" },
+    { url = "https://files.pythonhosted.org/packages/8e/f4/b7ac49f30013aba8f7b9566b1dd07e81de95e708c1374b7bacc5b9bc5c9c/rpds_py-2026.6.3-cp314-cp314t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:536bceea4fa4acf7e1c61da2b5786304367c816c8895be71b8f537c480b0ea1f", size = 371369, upload-time = "2026-06-30T07:16:30.912Z" },
+    { url = "https://files.pythonhosted.org/packages/31/86/6260bafa622f788b07ddec0e52d810305c8b9b0b8c27f58a2ab04bf62b4f/rpds_py-2026.6.3-cp314-cp314t-manylinux_2_31_riscv64.whl", hash = "sha256:bc0011654b91cc4fb2ae701bec0a0ba1e552c0714247fa7af6c59e0ccfa3a4e1", size = 379673, upload-time = "2026-06-30T07:16:32.486Z" },
+    { url = "https://files.pythonhosted.org/packages/19/c3/03f1ee79a047b48daeca157c89a18509cde22b6b951d642b9b0af1be660a/rpds_py-2026.6.3-cp314-cp314t-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:539d75de9e0d536c84ff18dfeb805398e58227001ce09231a26a08b9aed1ee0e", size = 397500, upload-time = "2026-06-30T07:16:34.471Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/95/8ed0cd8c377dca12aea498f119fe639fc474d1461545c39d2b5872eb1c0f/rpds_py-2026.6.3-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:166cf54d9f44fc6ceb53c7860258dde44a81406646de79f8ed3234fca3b6e538", size = 545978, upload-time = "2026-06-30T07:16:36.45Z" },
+    { url = "https://files.pythonhosted.org/packages/d3/f2/0eb57f0eaa83f8fc152a7e03de968ab77e1f00732bebc892b190c6eebde7/rpds_py-2026.6.3-cp314-cp314t-musllinux_1_2_i686.whl", hash = "sha256:d34c20167764fbcf927194d532dd7e0c56772f0a5f943fa5ef9e9afbba8fb9db", size = 613350, upload-time = "2026-06-30T07:16:38.213Z" },
+    { url = "https://files.pythonhosted.org/packages/5b/de/e0674bdbc3ef7634989b3f854c3f34bc1f587d36e5bfdc5c378d57034619/rpds_py-2026.6.3-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:ea7bb13b7c9a29791f87a0387ba7d3ad3a6d783d827e4d3f27b40a0ff44495e2", size = 576486, upload-time = "2026-06-30T07:16:39.797Z" },
+    { url = "https://files.pythonhosted.org/packages/f2/f6/21101359743cd136ada781e8210a85769578422ba460672eea0e29739200/rpds_py-2026.6.3-cp314-cp314t-win32.whl", hash = "sha256:6de4744d05bd1aa1be4ed7ea1189e3979196808008113bbbf899a460966b925e", size = 201068, upload-time = "2026-06-30T07:16:41.316Z" },
+    { url = "https://files.pythonhosted.org/packages/a6/b2/9574d4d44f7760c2aa32d92a0a4f41698e33f5b204a0bf5c9758f52c79d5/rpds_py-2026.6.3-cp314-cp314t-win_amd64.whl", hash = "sha256:c7b9a2f8f4d8e90af72571d3d495deebdd7e3c75451f5b41719aee166e940fc2", size = 220600, upload-time = "2026-06-30T07:16:43.091Z" },
+    { url = "https://files.pythonhosted.org/packages/08/ae/f23a2697e6ee6340a578b0f136be6483657bef0c6f9497b752bb5c0964bb/rpds_py-2026.6.3-cp315-cp315-macosx_10_12_x86_64.whl", hash = "sha256:e059c5dde6452b44424bd1834557556c226b57781dee1227af23518459722b13", size = 344726, upload-time = "2026-06-30T07:16:44.5Z" },
+    { url = "https://files.pythonhosted.org/packages/c3/63/e7b3a1a5358dd32c930a1062d8e15b67fd6e8922e81df9e91706d66ee5c8/rpds_py-2026.6.3-cp315-cp315-macosx_11_0_arm64.whl", hash = "sha256:2f7c26fbc5acd2522b95d4177fe4710ffd8e9b20529e703ffbf8db4d93903f05", size = 339587, upload-time = "2026-06-30T07:16:46.255Z" },
+    { url = "https://files.pythonhosted.org/packages/ec/64/10a85681916ca55fffb91b0a211f84e34297c109243484dd6394660a8a7c/rpds_py-2026.6.3-cp315-cp315-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:a3086b538543802f84c843911242db20447de00d8752dd0efc936dbcf02218ba", size = 369585, upload-time = "2026-06-30T07:16:48.101Z" },
+    { url = "https://files.pythonhosted.org/packages/76/c2/baf95c7c38823e12ba34407c5f5767a89e5cf2233895e56f608167ae9493/rpds_py-2026.6.3-cp315-cp315-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:8f2e5c5ee828d42cb11760761c0af6507927bec42d0ad5458f97c9203b054617", size = 375479, upload-time = "2026-06-30T07:16:49.93Z" },
+    { url = "https://files.pythonhosted.org/packages/6a/94/0aad06c72d65101e11d33528d438cda99a39ce0da99466e156158f2541d3/rpds_py-2026.6.3-cp315-cp315-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:ed0c1e5d10cdc7135537988c74a0188da68e2f3c30813ba3744ab1e42e0480f9", size = 492418, upload-time = "2026-06-30T07:16:51.641Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/17/de3f5a479a1f056535d7489819639d8cd591ea6281d700390b43b1abd745/rpds_py-2026.6.3-cp315-cp315-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:8c2642a7603ec0b16ed77da4555db3b4b472341904873788327c0b0d7b95f1bb", size = 384123, upload-time = "2026-06-30T07:16:53.622Z" },
+    { url = "https://files.pythonhosted.org/packages/46/7d/bf09bd1b145bb2671c03e1e6d1ab8651858d90d8c7dfeadd85a37a934fd8/rpds_py-2026.6.3-cp315-cp315-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:8e4320744c1ffdd95a603def63344bfab2d33edeab301c5007e7de9f9f5b3885", size = 367351, upload-time = "2026-06-30T07:16:55.241Z" },
+    { url = "https://files.pythonhosted.org/packages/a3/ea/1bb734f314b8be319149ddee80b18bd41372bdcfbdf88d28131c0cd37719/rpds_py-2026.6.3-cp315-cp315-manylinux_2_31_riscv64.whl", hash = "sha256:a9f4645593036b81bbdb36b9c8e0ea0d1c3fee968c4d59db0344c14087ef143a", size = 378827, upload-time = "2026-06-30T07:16:56.841Z" },
+    { url = "https://files.pythonhosted.org/packages/4b/93/d9611e5b25e26df9a3649813ed66193ace9347a7c7fc4ab7cf70e94851c0/rpds_py-2026.6.3-cp315-cp315-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:e55d236be29255554da47abe5c577637db7c24a02b8b46f0ca9524c855801868", size = 395966, upload-time = "2026-06-30T07:16:58.557Z" },
+    { url = "https://files.pythonhosted.org/packages/c3/cb/99d77e16e5534ae1d90629bbe419ba6ee170833a6a85e3aa1cc41726fbbc/rpds_py-2026.6.3-cp315-cp315-musllinux_1_2_aarch64.whl", hash = "sha256:24e9c5386e16669b674a69c156c8eeefcb578f3b3397b713b08e6d60f3c7b187", size = 545680, upload-time = "2026-06-30T07:17:00.164Z" },
+    { url = "https://files.pythonhosted.org/packages/59/15/11a29755f790cef7a2f755e8e14f4f0c33f39489e1893a632a2eee59672b/rpds_py-2026.6.3-cp315-cp315-musllinux_1_2_i686.whl", hash = "sha256:c60924535c75f1566b6eb75b5c31a48a43fef04fa2d0d201acbad8a9969c6107", size = 611853, upload-time = "2026-06-30T07:17:01.962Z" },
+    { url = "https://files.pythonhosted.org/packages/68/86/0c27547e21644da938fb530f7e1a8148dd24d02db07e7a5f2567a17ce710/rpds_py-2026.6.3-cp315-cp315-musllinux_1_2_x86_64.whl", hash = "sha256:38a2fea2787428f811719ceb9114cb78964a3138838320c29ac39526c79c16ba", size = 573715, upload-time = "2026-06-30T07:17:03.693Z" },
+    { url = "https://files.pythonhosted.org/packages/29/71/4d8fcf700931815594bce892255bbd973b94efaf0fc1932b0590df18d886/rpds_py-2026.6.3-cp315-cp315-win32.whl", hash = "sha256:d483fe17f01ad64b7bf7cc38fcefff1ca9fb83f8c2b2542b68f97ffe0611b369", size = 202864, upload-time = "2026-06-30T07:17:05.746Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/62/b577562de0edbb55b2be85ce5fd09c33e386b9b13eee09833af4240fd5c4/rpds_py-2026.6.3-cp315-cp315-win_amd64.whl", hash = "sha256:67e3a721ffc5d8d2210d3671872298c4a84e4b8035cfe42ffd7cde35d772b146", size = 220430, upload-time = "2026-06-30T07:17:07.471Z" },
+    { url = "https://files.pythonhosted.org/packages/c8/95/d6d0b2509825141eef60669a5739eec88dbc6a48053d6c92993a5704defe/rpds_py-2026.6.3-cp315-cp315-win_arm64.whl", hash = "sha256:6e84adbcf4bf841aed8116a8264b9f50b4cb3e7bd89b516122e616ac56ca269e", size = 215877, upload-time = "2026-06-30T07:17:09.008Z" },
+    { url = "https://files.pythonhosted.org/packages/b7/bf/f3ea278f0afd615c1d0f19cb69043a41526e2bb600c2b536eb192218eb27/rpds_py-2026.6.3-cp315-cp315t-macosx_10_12_x86_64.whl", hash = "sha256:ae6dd8f10bd17aad820876d24caec9efdafd80a318d16c0a48edb5e136902c6b", size = 346933, upload-time = "2026-06-30T07:17:10.762Z" },
+    { url = "https://files.pythonhosted.org/packages/9d/29/9907bdf1c5346763cf10b7f6852aad86652168c259def904cbe0082c5864/rpds_py-2026.6.3-cp315-cp315t-macosx_11_0_arm64.whl", hash = "sha256:bdbd97738551fca3917c1bd7188bec1920bb520104f28e7e1007f9ceb17b7690", size = 340274, upload-time = "2026-06-30T07:17:12.266Z" },
+    { url = "https://files.pythonhosted.org/packages/6f/2c/8e03767b5778ef25cebf74a7a91a2c3806f8eced4c92cb7406bbe060756d/rpds_py-2026.6.3-cp315-cp315t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:8b95977e7211527ab0ba576e286d023389fbeeb32a6b7b771665d333c60e5342", size = 370763, upload-time = "2026-06-30T07:17:14.107Z" },
+    { url = "https://files.pythonhosted.org/packages/2e/e1/df2a7e1ba2efd796af26194250b8d42c821b46592311595162af9ef0528d/rpds_py-2026.6.3-cp315-cp315t-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:d15fde0e6fb0d88a60d221204873743e5d9f0b7d29165e62cd86d0413ad74ba6", size = 376467, upload-time = "2026-06-30T07:17:15.76Z" },
+    { url = "https://files.pythonhosted.org/packages/6b/de/8a0814d1946af29cb068fb259aa8622f856df1d0bab58429448726b537f5/rpds_py-2026.6.3-cp315-cp315t-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:a136d453475ac0fcbda502ef1e6504bd28d6d904700915d278deeab0d00fe140", size = 496689, upload-time = "2026-06-30T07:17:17.308Z" },
+    { url = "https://files.pythonhosted.org/packages/df/f3/f19e0c852ba13694f5a79f3b719331051573cb5693feacf8a88ffffc3a71/rpds_py-2026.6.3-cp315-cp315t-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:f826877d462181e5eb1c26a0026b8d0cab05d99844ecb6d8bf3627a2ca0c0442", size = 385340, upload-time = "2026-06-30T07:17:18.928Z" },
+    { url = "https://files.pythonhosted.org/packages/e2/ae/7ec3a9d2d4351f99e37bcb06b6b6f954512646bfdbf9742e1de727865daf/rpds_py-2026.6.3-cp315-cp315t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:79486287de1730dbaff3dbd124d0ca4d2ef7f9d29bf2544f1f93c09b5bcbbd12", size = 372179, upload-time = "2026-06-30T07:17:20.539Z" },
+    { url = "https://files.pythonhosted.org/packages/d3/ac/9cee911dff2aaa9a5a8354f6610bf2e6a616de9197c5fff4f54f82585f1e/rpds_py-2026.6.3-cp315-cp315t-manylinux_2_31_riscv64.whl", hash = "sha256:808345f53cb952433ca2816f1604ff3515608a81784954f38d4452acfe8e61d5", size = 379993, upload-time = "2026-06-30T07:17:22.212Z" },
+    { url = "https://files.pythonhosted.org/packages/83/6b/7c2a07ba88d1e9a936612f7a5d067467ed03d971d5a06f7d309dff044a7e/rpds_py-2026.6.3-cp315-cp315t-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:1967debc37f64f2c4dc90a7f563aec558b471966e12adcac4e1c4240496b6ebf", size = 398909, upload-time = "2026-06-30T07:17:23.66Z" },
+    { url = "https://files.pythonhosted.org/packages/97/0b/776ffcb66783637b0031f6d58d6fb55913c8b5abf00aeecd46bf933fb477/rpds_py-2026.6.3-cp315-cp315t-musllinux_1_2_aarch64.whl", hash = "sha256:f0840b5b17057f7fd918b76183a4b5a0635f43e14eb2ce60dce1d4ee4707ea00", size = 546584, upload-time = "2026-06-30T07:17:25.264Z" },
+    { url = "https://files.pythonhosted.org/packages/55/33/ba3bc04d7092bd553c9b2b195624992d2cc4f3de1f380b7b93cbee67bd79/rpds_py-2026.6.3-cp315-cp315t-musllinux_1_2_i686.whl", hash = "sha256:faa679d19a6696fd54259ad321251ad77a13e70e03dd834daa762a44fb6196ef", size = 614357, upload-time = "2026-06-30T07:17:26.888Z" },
+    { url = "https://files.pythonhosted.org/packages/8b/71/14edf065f04630b1a8472f7653cad03f6c478bcf95ea0e6aed55451e33ea/rpds_py-2026.6.3-cp315-cp315t-musllinux_1_2_x86_64.whl", hash = "sha256:23a439f31ccbeff1574e24889128821d1f7917470e830cf6544dced1c662262a", size = 576533, upload-time = "2026-06-30T07:17:28.546Z" },
+    { url = "https://files.pythonhosted.org/packages/ba/76/65002b08596c389105720a8c0d22298b8dc25a4baf89b2ce431343c8b1de/rpds_py-2026.6.3-cp315-cp315t-win32.whl", hash = "sha256:913ca42ccad3f8cc6e292b587ae8ae49c8c823e5dce51a736252fc7c7cdfa577", size = 201204, upload-time = "2026-06-30T07:17:30.193Z" },
+    { url = "https://files.pythonhosted.org/packages/8c/97/d855d6b3c322d1f27e26f5241c42016b56cf01377ea8ed348285f54652f0/rpds_py-2026.6.3-cp315-cp315t-win_amd64.whl", hash = "sha256:ae3d4fe8c0b9213624fdce7279d70e3b148b682ca20719ebd193a23ebfa47324", size = 220719, upload-time = "2026-06-30T07:17:31.788Z" },
+]
+
+[[package]]
+name = "scrapling"
+version = "0.4.12"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "cssselect" },
+    { name = "lxml" },
+    { name = "orjson" },
+    { name = "tld" },
+    { name = "typing-extensions" },
+    { name = "w3lib" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/12/a1/f1dc696450c65bd2b12e0b2b5c88de9d38ddec3f13f584c1859e12725f2b/scrapling-0.4.12.tar.gz", hash = "sha256:c6f06d0ea54208d430d47402c4e66760280718dc5b116fa99985beb7b9a517f4", size = 167523, upload-time = "2026-07-26T20:24:03.506Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/7d/64/845a2c5e4dbda0403e0662d553b94c84bd08a60f9ded5b04abcb9994e43e/scrapling-0.4.12-py3-none-any.whl", hash = "sha256:bff83185a226f14c1e50e4408a2d24c93bc6ffe4b83a904d2f11ad1ad3823a91", size = 169304, upload-time = "2026-07-26T20:24:01.958Z" },
+]
+
+[package.optional-dependencies]
+all = [
+    { name = "anyio" },
+    { name = "apify-fingerprint-datapoints" },
+    { name = "browserforge" },
+    { name = "click" },
+    { name = "curl-cffi" },
+    { name = "ipython" },
+    { name = "markdownify" },
+    { name = "mcp" },
+    { name = "msgspec" },
+    { name = "patchright" },
+    { name = "playwright" },
+    { name = "protego" },
+]
+
+[[package]]
+name = "selenium"
+version = "4.46.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "certifi" },
+    { name = "trio" },
+    { name = "trio-websocket" },
+    { name = "typing-extensions" },
+    { name = "urllib3", extra = ["socks"] },
+    { name = "websocket-client" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/a0/53/053df98ef0c38535a74ea732217eb2dbfddc7ff5442a47a4b315ba577f85/selenium-4.46.0.tar.gz", hash = "sha256:54f7e1a4df5f7508ac8c38ce2ea584db1b27083dd79962b22524219219df5cbe", size = 1006000, upload-time = "2026-07-11T00:38:41.11Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/30/f4/365c723a37c7e9ba6a6a19bf0c9e407137703f56a72c6201d0e4cb1432cb/selenium-4.46.0-py3-none-any.whl", hash = "sha256:f03e864770a21ab32009961c9d015cb2c6275eba45c926c272e02d90af423aad", size = 9428587, upload-time = "2026-07-11T00:38:38.84Z" },
+]
+
+[[package]]
+name = "six"
+version = "1.17.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/94/e7/b2c673351809dca68a0e064b6af791aa332cf192da575fd474ed7d6f16a2/six-1.17.0.tar.gz", hash = "sha256:ff70335d468e7eb6ec65b95b99d3a2836546063f63acc5171de367e834932a81", size = 34031, upload-time = "2024-12-04T17:35:28.174Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/b7/ce/149a00dd41f10bc29e5921b496af8b574d8413afcd5e30dfa0ed46c2cc5e/six-1.17.0-py2.py3-none-any.whl", hash = "sha256:4721f391ed90541fddacab5acf947aa0d3dc7d27b2e1e8eda2be8970586c3274", size = 11050, upload-time = "2024-12-04T17:35:26.475Z" },
+]
+
+[[package]]
+name = "sniffio"
+version = "1.3.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/a2/87/a6771e1546d97e7e041b6ae58d80074f81b7d5121207425c964ddf5cfdbd/sniffio-1.3.1.tar.gz", hash = "sha256:f4324edc670a0f49750a81b895f35c3adb843cca46f0530f79fc1babb23789dc", size = 20372, upload-time = "2024-02-25T23:20:04.057Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e9/44/75a9c9421471a6c4805dbf2356f7c181a29c1879239abab1ea2cc8f38b40/sniffio-1.3.1-py3-none-any.whl", hash = "sha256:2f6da418d1f1e0fddd844478f41680e794e6051915791a034ff65e5f100525a2", size = 10235, upload-time = "2024-02-25T23:20:01.196Z" },
+]
+
+[[package]]
+name = "sortedcontainers"
+version = "2.4.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/e8/c4/ba2f8066cceb6f23394729afe52f3bf7adec04bf9ed2c820b39e19299111/sortedcontainers-2.4.0.tar.gz", hash = "sha256:25caa5a06cc30b6b83d11423433f65d1f9d76c4c6a0c90e3379eaa43b9bfdb88", size = 30594, upload-time = "2021-05-16T22:03:42.897Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/32/46/9cb0e58b2deb7f82b84065f37f3bffeb12413f947f9388e4cac22c4621ce/sortedcontainers-2.4.0-py2.py3-none-any.whl", hash = "sha256:a163dcaede0f1c021485e957a39245190e74249897e2ae4b2aa38595db237ee0", size = 29575, upload-time = "2021-05-16T22:03:41.177Z" },
+]
+
+[[package]]
+name = "soupsieve"
+version = "2.9.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/d9/38/e12680bbe6b4f8f3d17adcaf38d26850aa756c85cf4a80e79fc12a018fe8/soupsieve-2.9.1.tar.gz", hash = "sha256:c33e6605bbc71dd628b00c632d58ae607c22bade247e52553928f83bbb75b4ba", size = 122261, upload-time = "2026-07-21T16:57:17.452Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/0f/2c/437fe806897c2d6cfdc3ee43a18da8bf8e568530a4ae9bac781541ca9896/soupsieve-2.9.1-py3-none-any.whl", hash = "sha256:4f4477399246b7a0c720a88ca2454b11cd6bb9ae4c9d170140786e916776c14c", size = 37404, upload-time = "2026-07-21T16:57:16.421Z" },
+]
+
+[[package]]
+name = "sse-starlette"
+version = "3.4.6"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anyio" },
+    { name = "starlette" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/6c/10/a34c656829ffc1c4b22ef36d70d9ebb6b99c020e2aeb17cee5485099f028/sse_starlette-3.4.6.tar.gz", hash = "sha256:725f8a1bd6d26ae1b2c9610c0ef5065dfdd496f3988d28adcf8c4b49dc25c627", size = 32542, upload-time = "2026-07-20T14:16:32.201Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/49/36/e10c1d1b7ca881d2625db2ec28508578499187bb1c389952c398474e1834/sse_starlette-3.4.6-py3-none-any.whl", hash = "sha256:56217ab4c9a9f9c5db7b21e08732d3e7c2b807f45231ad23de0551a24c4a41f6", size = 16516, upload-time = "2026-07-20T14:16:30.978Z" },
+]
+
+[[package]]
+name = "stack-data"
+version = "0.6.3"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "asttokens" },
+    { name = "executing" },
+    { name = "pure-eval" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/28/e3/55dcc2cfbc3ca9c29519eb6884dd1415ecb53b0e934862d3559ddcb7e20b/stack_data-0.6.3.tar.gz", hash = "sha256:836a778de4fec4dcd1dcd89ed8abff8a221f58308462e1c4aa2a3cf30148f0b9", size = 44707, upload-time = "2023-09-30T13:58:05.479Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/f1/7b/ce1eafaf1a76852e2ec9b22edecf1daa58175c090266e9f6c64afcd81d91/stack_data-0.6.3-py3-none-any.whl", hash = "sha256:d5558e0c25a4cb0853cddad3d77da9891a08cb85dd9f9f91b9f8cd66e511e695", size = 24521, upload-time = "2023-09-30T13:58:03.53Z" },
+]
+
+[[package]]
+name = "starlette"
+version = "1.3.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anyio" },
+    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/eb/e3/7c1dc7381d9f8ab7d854328ebfa884e62cb3f3d8549ddfd37c7814f42afa/starlette-1.3.1.tar.gz", hash = "sha256:05d0213193f2fbaae60e2ecb593b4add4262ad4e46536b54abe36f11a71724e0", size = 2703240, upload-time = "2026-06-12T09:23:11.602Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/ec/bb/2799cc2ede3ed41131f8975621e7213dfc7ef4acbbaadfa440f32500c370/starlette-1.3.1-py3-none-any.whl", hash = "sha256:c7372aae11c3c3f26a42df7bd626cec2f47d03483d261d369516a615a53714c6", size = 73632, upload-time = "2026-06-12T09:23:10.017Z" },
+]
+
+[[package]]
+name = "tenacity"
+version = "9.1.4"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/47/c6/ee486fd809e357697ee8a44d3d69222b344920433d3b6666ccd9b374630c/tenacity-9.1.4.tar.gz", hash = "sha256:adb31d4c263f2bd041081ab33b498309a57c77f9acf2db65aadf0898179cf93a", size = 49413, upload-time = "2026-02-07T10:45:33.841Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/d7/c1/eb8f9debc45d3b7918a32ab756658a0904732f75e555402972246b0b8e71/tenacity-9.1.4-py3-none-any.whl", hash = "sha256:6095a360c919085f28c6527de529e76a06ad89b23659fa881ae0649b867a9d55", size = 28926, upload-time = "2026-02-07T10:45:32.24Z" },
+]
+
+[[package]]
+name = "tiktoken"
+version = "0.13.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "regex" },
+    { name = "requests" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/e4/e5/5f3cb2159769d0f4324c0e9e87f9de3c4b1cd45848a96b2eb3566ad5ca77/tiktoken-0.13.0.tar.gz", hash = "sha256:c9435714c3a84c2319499de9a300c0e604449dd0799ff246458b3bb6a7f433c1", size = 38986, upload-time = "2026-05-15T04:51:27.153Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/85/8e/144bde4e01df66b34bb865557c7cd754ed08b036217ebd79c9db5e9048a9/tiktoken-0.13.0-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:32ac870a806cfb260a02d0cb70426aef02e038297f8ad50df5040bb5af360791", size = 1034888, upload-time = "2026-05-15T04:50:31.579Z" },
+    { url = "https://files.pythonhosted.org/packages/36/18/d4ac9d20956cdebca04841316660ed584c2fecdc2b81722a28bc7ad3b1e4/tiktoken-0.13.0-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:4d9980f11429ed2d737c463bb1fb78cf330caa026adf002f714aced7849a687b", size = 982970, upload-time = "2026-05-15T04:50:32.961Z" },
+    { url = "https://files.pythonhosted.org/packages/74/ed/6bb8d05b9f731f749fee5c6f5ca63e981143c826a5985877330507bd13b7/tiktoken-0.13.0-cp312-cp312-manylinux_2_28_aarch64.whl", hash = "sha256:3f277ebea5edd7b8bf03c6f9431e1d67d517530115572b2dc1d465326e8f88c7", size = 1115741, upload-time = "2026-05-15T04:50:34.475Z" },
+    { url = "https://files.pythonhosted.org/packages/34/de/2ca96b07a82d972b74fe4b46de055b79c904e45c7eab699354a0bfa697dc/tiktoken-0.13.0-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:a116178fa7e1b4065bff05214360373a65cac22f965be7b3f73d00a0dbfe7649", size = 1136523, upload-time = "2026-05-15T04:50:35.782Z" },
+    { url = "https://files.pythonhosted.org/packages/ee/dc/9dafec002c2d4424378563cf4cf5c7fb93631d2a55013c8b87554ee4012c/tiktoken-0.13.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:2c397ddda233208345b01bd30f2fca79ff730e55731d0108a603f9bc57f6af3b", size = 1181954, upload-time = "2026-05-15T04:50:36.99Z" },
+    { url = "https://files.pythonhosted.org/packages/a1/d0/1f8578c45b2f24759b46f0b50d31878c63c73e6bf0f2227e10ec5c5408dc/tiktoken-0.13.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:95097e4f89b06403976e498abf61a0ee73a7497e73fb599cb211d8197a054d91", size = 1240069, upload-time = "2026-05-15T04:50:38.221Z" },
+    { url = "https://files.pythonhosted.org/packages/aa/90/28d7f154888610aa9237e541986beb62b479df29d193a5a0617dbb1514d0/tiktoken-0.13.0-cp312-cp312-win_amd64.whl", hash = "sha256:8f2d16e7a7c783ad81f36e457d046d1f1c8af70b22aec8a13238efe531977c41", size = 874748, upload-time = "2026-05-15T04:50:39.587Z" },
+    { url = "https://files.pythonhosted.org/packages/9c/83/b096c859c2a47c11731bf2f5885f4028b809dfe2396582883eed9cae372f/tiktoken-0.13.0-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:5df5d1507bd245f1ccad4a074698240021239e455eb0bb4ced4e3d7181872154", size = 1034228, upload-time = "2026-05-15T04:50:40.988Z" },
+    { url = "https://files.pythonhosted.org/packages/53/61/c68e123b6d753e3fc2751e9b18e732c9d8bf1e1926762e736eee935d931c/tiktoken-0.13.0-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:8fe806a50664e83a6ffd56cbd1e4f5dcc6cd32a3e7538f70dc38b1a271384545", size = 982978, upload-time = "2026-05-15T04:50:42.195Z" },
+    { url = "https://files.pythonhosted.org/packages/ef/8b/96cc178cc584e65d363134500f297790b06cd48cdeb1e8fcf7bbe60f4715/tiktoken-0.13.0-cp313-cp313-manylinux_2_28_aarch64.whl", hash = "sha256:125bc05005e747f993a83dc67934249932d6e4209854452cd4c0b1d53fba3ba2", size = 1116355, upload-time = "2026-05-15T04:50:43.564Z" },
+    { url = "https://files.pythonhosted.org/packages/86/f5/bab735d2c72ea55404b295d02d092644eb5f7cc6205e34d35eb9abfb9ab2/tiktoken-0.13.0-cp313-cp313-manylinux_2_28_x86_64.whl", hash = "sha256:5e6358911cab4adee6712da27d65573496a4f68cf8a2b5fca6a4ad10fc5748cf", size = 1135772, upload-time = "2026-05-15T04:50:44.782Z" },
+    { url = "https://files.pythonhosted.org/packages/4e/b9/6de04ebdf904edfaad87788011b3735087a0c9ea671b9027e1e4e965e8c8/tiktoken-0.13.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:975cbd78d085d75d26b59660e262736dcaed1e35f8f142cd6291025c01d25486", size = 1182415, upload-time = "2026-05-15T04:50:46.422Z" },
+    { url = "https://files.pythonhosted.org/packages/0d/9c/470a05f3b1caf038f44880e334d47ab674e0c80d514c66b375d14d5afa10/tiktoken-0.13.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:75ab9bc99fa020a4c283424590ecd7f3afd70c1c281cb3fa3192a6c3af9f9615", size = 1239879, upload-time = "2026-05-15T04:50:48.052Z" },
+    { url = "https://files.pythonhosted.org/packages/42/a6/c1936d16055436cb32e6c6128d68629622e00f4768562f55653752d34768/tiktoken-0.13.0-cp313-cp313-win_amd64.whl", hash = "sha256:6b1615f0ff71953d19729ceb18865429c185b0a23c5353f1bbca34a394bf60f7", size = 874829, upload-time = "2026-05-15T04:50:49.202Z" },
+    { url = "https://files.pythonhosted.org/packages/d6/07/acb5992c3772b5a36284f742cfb7a5895aa4471d1848ac31464ad50d7fdf/tiktoken-0.13.0-cp313-cp313t-macosx_10_13_x86_64.whl", hash = "sha256:6eb4a5bfbc6426938026b1a334e898ac53541360d62d8c689870160cc80abd67", size = 1033600, upload-time = "2026-05-15T04:50:50.4Z" },
+    { url = "https://files.pythonhosted.org/packages/14/e9/742e9aec30f59b9f161f7ff7cd072e02ea836c9e1c0854a8076dfcd40d5c/tiktoken-0.13.0-cp313-cp313t-macosx_11_0_arm64.whl", hash = "sha256:43cee3e5400573b2046fbf092cc7a5bc30164f9e4c95ce20714da929df48737a", size = 982516, upload-time = "2026-05-15T04:50:52.03Z" },
+    { url = "https://files.pythonhosted.org/packages/72/74/ca1541b053e7648254d2e4b42a253e1bb4359f2c91a0a8d49228c794e1a0/tiktoken-0.13.0-cp313-cp313t-manylinux_2_28_aarch64.whl", hash = "sha256:7de52e3f566d19b3b11bd37eea552c6c305ad74081f736882bd44d148ed4c48d", size = 1115518, upload-time = "2026-05-15T04:50:53.543Z" },
+    { url = "https://files.pythonhosted.org/packages/46/e3/93825eaf5a4a504795b787e5d5dea07fbeb3dabf97aa7b450be8bde59c89/tiktoken-0.13.0-cp313-cp313t-manylinux_2_28_x86_64.whl", hash = "sha256:51384448aa508e4df84c0f7c1dc3211c7f7b8096325660ee5fc82f3e11b381ce", size = 1136867, upload-time = "2026-05-15T04:50:55.191Z" },
+    { url = "https://files.pythonhosted.org/packages/8c/46/002b68de6827091d5ae90b048f326e8aad8d953520950e5ce1508879414f/tiktoken-0.13.0-cp313-cp313t-musllinux_1_2_aarch64.whl", hash = "sha256:e28157350f7ebf35008dd8e9e0fdb621f976e4230c881099c85e8cf07eaa50e2", size = 1181826, upload-time = "2026-05-15T04:50:56.296Z" },
+    { url = "https://files.pythonhosted.org/packages/db/c6/d393e3185a276505182f7abd93fe714f3c444a2be9180798fa052347504e/tiktoken-0.13.0-cp313-cp313t-musllinux_1_2_x86_64.whl", hash = "sha256:165cf1820ea4a354985c2490a5205d4cc74661c934aca79dd0368232fff94e0f", size = 1239489, upload-time = "2026-05-15T04:50:57.918Z" },
+    { url = "https://files.pythonhosted.org/packages/b7/4d/bc07d1f1635d4897a202acc0ae11c2886eaa7325c359ba4741b47bf8e225/tiktoken-0.13.0-cp313-cp313t-win_amd64.whl", hash = "sha256:6c43a675ca14f6f2749ba7f12075d37456015a24b859f2517b9beb4ef30807ec", size = 873820, upload-time = "2026-05-15T04:50:59.528Z" },
+    { url = "https://files.pythonhosted.org/packages/8c/93/0dd6adca026a616c3a92974566b43381eea4b475ce1f36c062b8271a9ac5/tiktoken-0.13.0-cp314-cp314-macosx_10_13_x86_64.whl", hash = "sha256:eaaaef47c2406277181d2086484c317bf7fc433e2d5d03ff94f56b0dcec87471", size = 1034977, upload-time = "2026-05-15T04:51:00.957Z" },
+    { url = "https://files.pythonhosted.org/packages/d9/77/5ec6e6bc5b30bed6d93f7f2162d8f6b32437b3ba27cb527cfe004f6109c9/tiktoken-0.13.0-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:ca8b310bd93b3772cb1b7922d915446864860f562bdfe4825c63a0aed3fb28cd", size = 983635, upload-time = "2026-05-15T04:51:02.629Z" },
+    { url = "https://files.pythonhosted.org/packages/94/b0/c8ae9aff00d625c50659b4513e707a0462c4bf5d4d6cc1b802103225c02e/tiktoken-0.13.0-cp314-cp314-manylinux_2_28_aarch64.whl", hash = "sha256:32e0c12305105002c047b3bb1070b0dd9a73b0cb3b2856a8972b810e7a4f5881", size = 1116036, upload-time = "2026-05-15T04:51:04.082Z" },
+    { url = "https://files.pythonhosted.org/packages/1b/ac/6a5dddd1d0a6018ecb389bd0353e6b4a515eb4d2286611bd0ace1937b9e1/tiktoken-0.13.0-cp314-cp314-manylinux_2_28_x86_64.whl", hash = "sha256:5ba5fd62507a932d1241346179e3b39bc7bf7408f03c272652d93b3bedf5db24", size = 1135544, upload-time = "2026-05-15T04:51:05.229Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/b8/585032b4384b2f7dcdaddcb52865c83a701a420d09e3c2b4a2be1c450c57/tiktoken-0.13.0-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:d108bc2d470fc53c8ecd24f2c0fd2b5f98c33e87cdb6aa2e9b8c5dced703d273", size = 1182217, upload-time = "2026-05-15T04:51:06.517Z" },
+    { url = "https://files.pythonhosted.org/packages/cd/b6/993ff1ded3958215fd341a847b8e5ffeb5de473f435296870d314fc91ac4/tiktoken-0.13.0-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:cb99cb5127449f58d0a2d5f5ccfb390d8dbdfd919c221246caaee29d8725ed51", size = 1239404, upload-time = "2026-05-15T04:51:07.843Z" },
+    { url = "https://files.pythonhosted.org/packages/dd/3d/fef7e06e3b33e7538db0ced734cf9fe23b6832d2ac4990c119c377aec55e/tiktoken-0.13.0-cp314-cp314-win_amd64.whl", hash = "sha256:115c4f26ffa11caac8b54eea35c2ad38c612c20a48d35dd15d70a02ac6f51f58", size = 918686, upload-time = "2026-05-15T04:51:08.925Z" },
+    { url = "https://files.pythonhosted.org/packages/c1/82/a7fc44582bc32ab00de988a2299bf77c077f59068b233109e34b7d6ca7e6/tiktoken-0.13.0-cp314-cp314t-macosx_10_13_x86_64.whl", hash = "sha256:472527e9132952f2fbf77cd290658bacf003d4d5a3fabc18e5fbd407cbae4d9b", size = 1034454, upload-time = "2026-05-15T04:51:10.035Z" },
+    { url = "https://files.pythonhosted.org/packages/37/d0/24d8a890c14f432a05cea669c17bebeaa99f96a7c79523b590f564246411/tiktoken-0.13.0-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:4e2f67d27c9626cdd25fe33d9313c5cdb3d8d82da646b68d6eb8e7e9c20e6448", size = 982976, upload-time = "2026-05-15T04:51:11.23Z" },
+    { url = "https://files.pythonhosted.org/packages/49/b7/2ab43f62788a9266187a9bfc1d3af99ad83e5eaa25fbef168a69cd5ad14f/tiktoken-0.13.0-cp314-cp314t-manylinux_2_28_aarch64.whl", hash = "sha256:2b920b35805cd64585a37c3dc7ce65fba4d2d36016be01e1d7942482ca29093a", size = 1115526, upload-time = "2026-05-15T04:51:12.608Z" },
+    { url = "https://files.pythonhosted.org/packages/64/39/1494321ed323ce7a14d88e3cd6cb9058625977df1c6961ddc492bd10a9f3/tiktoken-0.13.0-cp314-cp314t-manylinux_2_28_x86_64.whl", hash = "sha256:493af3aa28a4aaf2e3d2600a2ee717252c9bf5ab38fff94eb5a02db5ab77e5ad", size = 1136466, upload-time = "2026-05-15T04:51:13.926Z" },
+    { url = "https://files.pythonhosted.org/packages/96/d9/dfd086aa2d918c563a140720e0ce296cada1634efd2783d5cf51e05f984e/tiktoken-0.13.0-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:6644c9c2b5cf3916f5a3641d7d12fdb3f006a7b3d9ff6acdaec44e29ab1ff91e", size = 1181863, upload-time = "2026-05-15T04:51:15.025Z" },
+    { url = "https://files.pythonhosted.org/packages/2f/68/a18b4f307086954fdae32714cb4f85562e34f9d34ab206e61f1816aa6018/tiktoken-0.13.0-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:5cb65b60b9408563676d874a3a4ee573370066f0dc4e29d84e82e989c6517424", size = 1239218, upload-time = "2026-05-15T04:51:16.103Z" },
+    { url = "https://files.pythonhosted.org/packages/16/5b/f2aa703a4fc5d2dff73460a7d46cc2f3f44aa0f3dd8eeb20d2a0ecf68862/tiktoken-0.13.0-cp314-cp314t-win_amd64.whl", hash = "sha256:85b78cc3a2c3d48723ca751fa981f1fedccd54194ca0471b957364353a898b07", size = 918110, upload-time = "2026-05-15T04:51:17.237Z" },
+]
+
+[[package]]
+name = "tld"
+version = "0.13.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/5c/5d/76b4383ac4e5b5e254e50c09807b3e13820bed6d6c11cd540264988d6802/tld-0.13.2.tar.gz", hash = "sha256:d983fa92b9d717400742fca844e29d5e18271079c7bcfabf66d01b39b4a14345", size = 467175, upload-time = "2026-03-06T23:50:34.498Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/9e/90/39a85a4b63c84213e78b3c17d22e1bf45328acf8ebb33ef93be30d0a3911/tld-0.13.2-py2.py3-none-any.whl", hash = "sha256:9b8fdbdb880e7ba65b216a4937f2c94c49a7226723783d5838fc958ac76f4e0c", size = 296743, upload-time = "2026-03-06T23:50:32.465Z" },
+]
+
+[[package]]
+name = "tqdm"
+version = "4.69.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "colorama", marker = "sys_platform == 'win32'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/dd/84/da0e5038228fa34dfd77c5026b173ed035d2a3ba31f1077590c013de2bff/tqdm-4.69.1.tar.gz", hash = "sha256:2be21080a0ce17e902c2f1baeb6a74bf551b67bbdfa4bc0109fad471d0b4cb0d", size = 793046, upload-time = "2026-07-24T14:22:02.08Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/01/50/5817619a0fca56aff06383dbfde7ae017b3ca383915b3f1e4713164273cf/tqdm-4.69.1-py3-none-any.whl", hash = "sha256:0a654b96f7a2660cceb615b56f307ec2bef96c515409014a429a561981ab52b4", size = 675452, upload-time = "2026-07-24T14:22:00.048Z" },
+]
+
+[[package]]
+name = "traitlets"
+version = "5.15.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/57/a9/a2584b8313b89f94869ddb3c4074617a691de1812a614d2d50e32ca5a7a6/traitlets-5.15.1.tar.gz", hash = "sha256:7b1c07854fe25acb39e009bae49f11b79ff6cbb2f27999104e9110e7a6b53722", size = 163344, upload-time = "2026-06-03T12:26:06.181Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/96/8d/1080ee4c231f361b6ce4470d556c8c435b67c7e0753aaa641497ee92f88b/traitlets-5.15.1-py3-none-any.whl", hash = "sha256:770a53705f84b81ac107e83a1b3328ff2dae16094d8fc3cfc004e4b22dfd8e92", size = 85858, upload-time = "2026-06-03T12:26:04.395Z" },
+]
+
+[[package]]
+name = "trio"
+version = "0.33.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "attrs" },
+    { name = "cffi", marker = "implementation_name != 'pypy' and os_name == 'nt'" },
+    { name = "idna" },
+    { name = "outcome" },
+    { name = "sniffio" },
+    { name = "sortedcontainers" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/52/b6/c744031c6f89b18b3f5f4f7338603ab381d740a7f45938c4607b2302481f/trio-0.33.0.tar.gz", hash = "sha256:a29b92b73f09d4b48ed249acd91073281a7f1063f09caba5dc70465b5c7aa970", size = 605109, upload-time = "2026-02-14T18:40:55.386Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/1c/93/dab25dc87ac48da0fe0f6419e07d0bfd98799bed4e05e7b9e0f85a1a4b4b/trio-0.33.0-py3-none-any.whl", hash = "sha256:3bd5d87f781d9b0192d592aef28691f8951d6c2e41b7e1da4c25cde6c180ae9b", size = 510294, upload-time = "2026-02-14T18:40:53.313Z" },
+]
+
+[[package]]
+name = "trio-websocket"
+version = "0.12.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "outcome" },
+    { name = "trio" },
+    { name = "wsproto" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/d1/3c/8b4358e81f2f2cfe71b66a267f023a91db20a817b9425dd964873796980a/trio_websocket-0.12.2.tar.gz", hash = "sha256:22c72c436f3d1e264d0910a3951934798dcc5b00ae56fc4ee079d46c7cf20fae", size = 33549, upload-time = "2025-02-25T05:16:58.947Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/c7/19/eb640a397bba49ba49ef9dbe2e7e5c04202ba045b6ce2ec36e9cadc51e04/trio_websocket-0.12.2-py3-none-any.whl", hash = "sha256:df605665f1db533f4a386c94525870851096a223adcb97f72a07e8b4beba45b6", size = 21221, upload-time = "2025-02-25T05:16:57.545Z" },
+]
+
+[[package]]
+name = "typing-extensions"
+version = "4.16.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/f6/cc/6253133b5bb138fc3306cebfbda2c520f545d36b5be2c7255cc528bb45d6/typing_extensions-4.16.0.tar.gz", hash = "sha256:dc983d19a509c94dba722ee6abd33940f7c05a89e243c47e907eb4db6f1a43e5", size = 113555, upload-time = "2026-07-02T08:40:05.92Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/49/d3/b8441a820a491ddfc024b0b0cf0393375b75ea13866d9c66727e54c2fc80/typing_extensions-4.16.0-py3-none-any.whl", hash = "sha256:481caa481374e813c1b176ada14e97f1f67a4539ce9cfeb3f350d78d6370c2e8", size = 45571, upload-time = "2026-07-02T08:40:04.659Z" },
+]
+
+[[package]]
+name = "typing-inspection"
+version = "0.4.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/55/e3/70399cb7dd41c10ac53367ae42139cf4b1ca5f36bb3dc6c9d33acdb43655/typing_inspection-0.4.2.tar.gz", hash = "sha256:ba561c48a67c5958007083d386c3295464928b01faa735ab8547c5692e87f464", size = 75949, upload-time = "2025-10-01T02:14:41.687Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/dc/9b/47798a6c91d8bdb567fe2698fe81e0c6b7cb7ef4d13da4114b41d239f65d/typing_inspection-0.4.2-py3-none-any.whl", hash = "sha256:4ed1cacbdc298c220f1bd249ed5287caa16f34d44ef4e9c3d0cbad5b521545e7", size = 14611, upload-time = "2025-10-01T02:14:40.154Z" },
+]
+
+[[package]]
+name = "tzdata"
+version = "2026.3"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/92/ff/5a28bdfd8c3ebec42564ac7d0e54ca3db65044a9314a97f9564fa7a1e926/tzdata-2026.3.tar.gz", hash = "sha256:4a1518b8993086a7982523e071643f3c0e5f213e75b21318e78bcabfff9d1415", size = 198674, upload-time = "2026-07-10T08:50:37.887Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e5/6d/b53b99a9f2766d095985947a5782f1702cabb129a34f7a802d7197af832f/tzdata-2026.3-py2.py3-none-any.whl", hash = "sha256:dc096730c87af6cab1b171c9d532be840741ff5d459015e7f6947bd7d7e54931", size = 348168, upload-time = "2026-07-10T08:50:36.46Z" },
+]
+
+[[package]]
+name = "urllib3"
+version = "2.7.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/53/0c/06f8b233b8fd13b9e5ee11424ef85419ba0d8ba0b3138bf360be2ff56953/urllib3-2.7.0.tar.gz", hash = "sha256:231e0ec3b63ceb14667c67be60f2f2c40a518cb38b03af60abc813da26505f4c", size = 433602, upload-time = "2026-05-07T16:13:18.596Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/7f/3e/5db95bcf282c52709639744ca2a8b149baccf648e39c8cc87553df9eae0c/urllib3-2.7.0-py3-none-any.whl", hash = "sha256:9fb4c81ebbb1ce9531cce37674bbc6f1360472bc18ca9a553ede278ef7276897", size = 131087, upload-time = "2026-05-07T16:13:17.151Z" },
+]
+
+[package.optional-dependencies]
+socks = [
+    { name = "pysocks" },
+]
+
+[[package]]
+name = "uuid-utils"
+version = "0.17.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/e7/91/63938e0e7e7876658e5e40178e7c0735b53527886fe11797a11699c55edd/uuid_utils-0.17.0.tar.gz", hash = "sha256:abb5667a36119019b3fa320c4d10c21ebccfcc87c8a739e6a0056cee7f48dde2", size = 43220, upload-time = "2026-07-09T13:49:58.433Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/20/80/a7e685968e3cec99d6fe2fb25d0f5726310e1bba356da68c13dfd8b7d140/uuid_utils-0.17.0-cp312-cp312-macosx_10_12_x86_64.macosx_11_0_arm64.macosx_10_12_universal2.whl", hash = "sha256:9205068badf453d2f0821fd5d340389b4679992d7ff79d4f3e5608996dd1b287", size = 556403, upload-time = "2026-07-09T13:48:27.022Z" },
+    { url = "https://files.pythonhosted.org/packages/56/47/3102d93bcb7b0bfe6bede63ff8f221a7f91348e10a37f682773be27c56d9/uuid_utils-0.17.0-cp312-cp312-macosx_10_12_x86_64.whl", hash = "sha256:0fcca4e838af9ac9243b3358d7c14afa4dca286a87781124c272d6c4cad9c968", size = 285608, upload-time = "2026-07-09T13:48:28.769Z" },
+    { url = "https://files.pythonhosted.org/packages/55/fb/d59695f0f8db065b93c63316eaafa05a22d75a0486978a33736c52c646d5/uuid_utils-0.17.0-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:0f3729e839209f3457d0d8b6a35a376fdf65577a5aecaf4cc3587d3305759ba6", size = 319926, upload-time = "2026-07-09T13:48:29.965Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/03/62fabcd1e990e07a0e220e8d552af45bc16f107fa8e55c2014a706bb1a1e/uuid_utils-0.17.0-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:3dac0ad0cd9a2818d1775215365a4e8c2f8ada215529dd26f3f8cceeb67a6988", size = 327172, upload-time = "2026-07-09T13:48:31.187Z" },
+    { url = "https://files.pythonhosted.org/packages/d9/37/a5081391338b459e2f8d8b12581f00f8caa6317fab510e0e85c18c59e938/uuid_utils-0.17.0-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:e671b2322ef09106ecb1ca0f4c398b134d5e2c1f80d7a4f3336847a3072c0e94", size = 439075, upload-time = "2026-07-09T13:48:32.295Z" },
+    { url = "https://files.pythonhosted.org/packages/59/30/91795bd01e17a13661280d4899fbf38fb05e3f38e873f9aaec106ec30aa0/uuid_utils-0.17.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:8eb3e5caca8d3a6f72ea4cce024583f989f6f2e9186f98800213fff0176e8bcc", size = 320247, upload-time = "2026-07-09T13:48:33.64Z" },
+    { url = "https://files.pythonhosted.org/packages/e5/11/09102b78303e4eb62069d6d88ef9fd661dc523e8f429e1fd67eaa78a6f44/uuid_utils-0.17.0-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:8b72c2002202038666bf647f9a790906214c7c11cd0d6efef77b7d07bef3034a", size = 344738, upload-time = "2026-07-09T13:48:34.786Z" },
+    { url = "https://files.pythonhosted.org/packages/74/f9/be95bad6954b60328878c3800258f01a6accd24fd75112d13f023462d53f/uuid_utils-0.17.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:4e2ac1c0b56f2c91b6f158e29ed96b1503223fe8aa6e79b1be1dc55bd8a5131c", size = 496845, upload-time = "2026-07-09T13:48:36.057Z" },
+    { url = "https://files.pythonhosted.org/packages/2d/02/8a19a34e0530d987488a068a71576a236f5c8c746630b870b57f71eb24ef/uuid_utils-0.17.0-cp312-cp312-musllinux_1_2_armv7l.whl", hash = "sha256:6c142bd0cb4dba31c10babe00d59f7ef6460f0ef55eaa9c1a9da270684af996a", size = 603233, upload-time = "2026-07-09T13:48:37.512Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/a8/b1abab36ff73b0248d82179816467f6d39a2e80fd64329a895ca94f3508e/uuid_utils-0.17.0-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:e252db239eb41c32248e096e0d170bce5896a4fd3405556362bc3dd83d912206", size = 561401, upload-time = "2026-07-09T13:48:38.977Z" },
+    { url = "https://files.pythonhosted.org/packages/61/91/70e7b528b351cc03a9ca43e6116371cdde31bb12bcead7ca2ca1367366cc/uuid_utils-0.17.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:237722b6581bb5b4eb4cefbcbe5c6e2980a440aabe781fbe50ebf1cb71eee4cc", size = 525314, upload-time = "2026-07-09T13:48:40.599Z" },
+    { url = "https://files.pythonhosted.org/packages/d6/f6/9167e90cf9937d6558f92d022ff3024a69d938a514d9c8faa4080f73b001/uuid_utils-0.17.0-cp312-cp312-win32.whl", hash = "sha256:46a73cacdf512f473a81f65dbf84186e08cfe6e9118fa582b6c6b33a8288a30d", size = 166831, upload-time = "2026-07-09T13:48:41.862Z" },
+    { url = "https://files.pythonhosted.org/packages/5c/7d/0b889654d9ee3413f810cf4685e241285f650d98a4103ac9f3c6bcc95f29/uuid_utils-0.17.0-cp312-cp312-win_amd64.whl", hash = "sha256:e59b60a0a4cb7541480e02090d37dc2df3b72df4c2e776fff64ce3a4e3dd4637", size = 172944, upload-time = "2026-07-09T13:48:42.992Z" },
+    { url = "https://files.pythonhosted.org/packages/be/35/8c6e1bf65e4d400352885dadc656ad6d0af96e89231e3f04686bc2197128/uuid_utils-0.17.0-cp312-cp312-win_arm64.whl", hash = "sha256:d561a4c5747a1e6c7fa7c49a0292e78b4e8c456332caa084fc7abad8de828652", size = 172459, upload-time = "2026-07-09T13:48:44.271Z" },
+    { url = "https://files.pythonhosted.org/packages/d2/dd/614fb9912157ac0128e6050859ccf06d9f13df9a944a803e8f80f6157e38/uuid_utils-0.17.0-cp313-cp313-macosx_10_12_x86_64.macosx_11_0_arm64.macosx_10_12_universal2.whl", hash = "sha256:d11a7bc1e02da8984d32e6de9e0826c6edac00eac17de270f372bf32f9a0af63", size = 557259, upload-time = "2026-07-09T13:48:45.664Z" },
+    { url = "https://files.pythonhosted.org/packages/3e/11/d072711704de3d21bec08b6c2f36a215200ca1d5e01a390ea1ac434080a0/uuid_utils-0.17.0-cp313-cp313-macosx_10_12_x86_64.whl", hash = "sha256:7a49f47ac26df3e431c56b825c1bae8e6d3d591fdbb7438c227cc9845a7e3d73", size = 286271, upload-time = "2026-07-09T13:48:47.018Z" },
+    { url = "https://files.pythonhosted.org/packages/18/6d/8a63e5eb2d5a6ba69a6c2036e305075bd6f5a022e7ea25fc6ce0eb7c51d2/uuid_utils-0.17.0-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:32df1944808877702ceea398c103881c09a679bb672a215e01c2a84231266bf9", size = 320025, upload-time = "2026-07-09T13:48:48.208Z" },
+    { url = "https://files.pythonhosted.org/packages/f7/2d/bdc2caf9719d9090d7c46043242ae6136cba4f7a7ee384992ab905ad9aa1/uuid_utils-0.17.0-cp313-cp313-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:98c88d3edd08e7245562e9815996dbc6f0bd4745e1c76462f24af5ae4e187dd1", size = 327931, upload-time = "2026-07-09T13:48:49.673Z" },
+    { url = "https://files.pythonhosted.org/packages/b6/33/9219d09d51ead282b578b2a4e0a515c2cce3ec52076cada8bfb7e35727d5/uuid_utils-0.17.0-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:5a4370089c8b2e42f1db51d76408c7fa8eaa2934bf854d17983d16179c07c098", size = 438537, upload-time = "2026-07-09T13:48:50.842Z" },
+    { url = "https://files.pythonhosted.org/packages/d8/79/e8e0f8b3955f2081c116157119d87659937893242eb834aa170da04d660b/uuid_utils-0.17.0-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:09a55b7a5ae764985cb46467496a1787678d0a1400356157a080ad95b1a36869", size = 320656, upload-time = "2026-07-09T13:48:52.164Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/5e/d1ceddc430ff04b6e21704b2030d4438074a2f478b265dab43da957791c1/uuid_utils-0.17.0-cp313-cp313-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:56aa6488b931246fae11924e4bd0e2b32677e63945eecb71c29e3c2ca0dc3131", size = 345310, upload-time = "2026-07-09T13:48:54.076Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/62/89438e12f389a843e626b7e37691319a057b3d6b80914609106891faadda/uuid_utils-0.17.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:309a35f12d99dde19032bc2259cda6431c85eeac0879134dc777cc3087d7e1cb", size = 496771, upload-time = "2026-07-09T13:48:55.365Z" },
+    { url = "https://files.pythonhosted.org/packages/87/d2/eedcd99f522d60e238ead03844f0d51743ba84d33044959e230b756bf212/uuid_utils-0.17.0-cp313-cp313-musllinux_1_2_armv7l.whl", hash = "sha256:21c79b61ff750abcf057163dd764ccb6196cde7a26cda1b31b45cd97769e03b3", size = 603631, upload-time = "2026-07-09T13:48:56.746Z" },
+    { url = "https://files.pythonhosted.org/packages/0e/a8/bb1b38aaddd7243b6e562c6694f499bf094800918316192fd8cb2cdc2620/uuid_utils-0.17.0-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:4134353bfe3026ddab8e886002dc52bc5a0ab04611aabb0eaae23c32e6e57f64", size = 562008, upload-time = "2026-07-09T13:48:58.241Z" },
+    { url = "https://files.pythonhosted.org/packages/b4/77/5f7ed930dc105e293845c09e4d5bd84076318a12f45a46783e1af64906d7/uuid_utils-0.17.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:7c89359affecebe2e39e6a116d069b363c936511a9572b308402489a26957d89", size = 525527, upload-time = "2026-07-09T13:48:59.784Z" },
+    { url = "https://files.pythonhosted.org/packages/fd/25/1b55697adf6811a6f92cff6340e6b03e31fd6bc51066a5c10698c29b3679/uuid_utils-0.17.0-cp313-cp313-pyemscripten_2025_0_wasm32.whl", hash = "sha256:6a019a31bc4db89a0903a3e4f6b218571f3a6ff0ad4b3d3fe1c8f91a05ff6e3e", size = 97965, upload-time = "2026-07-09T13:49:01.217Z" },
+    { url = "https://files.pythonhosted.org/packages/26/bf/cd729343de4684230be8a966bad7bfc2cf10ce3e643b1189a8b5370dbe35/uuid_utils-0.17.0-cp313-cp313-win32.whl", hash = "sha256:b3131a82d0c7611f0aa480a6d36929e001a3f54ba0fc029a8118a5863cce513c", size = 167316, upload-time = "2026-07-09T13:49:02.354Z" },
+    { url = "https://files.pythonhosted.org/packages/76/f0/e602ae0a1b139a7826e5189b93d91902564def06d5006324fd2faf82c8fc/uuid_utils-0.17.0-cp313-cp313-win_amd64.whl", hash = "sha256:9e311f908d2f842fca4c7dcebc4f10306b8089b204ef04cf6704b4332c9ff6ff", size = 173630, upload-time = "2026-07-09T13:49:03.529Z" },
+    { url = "https://files.pythonhosted.org/packages/1a/52/024ebece265b387154115dc4f1d9727174ef82623069f4bec8b7ed7e73f7/uuid_utils-0.17.0-cp313-cp313-win_arm64.whl", hash = "sha256:c351737e2e65497c7200ab4ffb8af97e9f48be6488309abdd265fe08d66ee92f", size = 173214, upload-time = "2026-07-09T13:49:04.836Z" },
+    { url = "https://files.pythonhosted.org/packages/56/44/e2fd3fdf356e1b55d2acf1b956b4f3f29ffb215a99c387eba04b1c5fba66/uuid_utils-0.17.0-cp314-cp314-macosx_10_12_x86_64.macosx_11_0_arm64.macosx_10_12_universal2.whl", hash = "sha256:673d89cc434cc9b97a0b4cf61272f6fca70a81f64eb0afbface2a0d9f77f06cd", size = 562232, upload-time = "2026-07-09T13:49:06.201Z" },
+    { url = "https://files.pythonhosted.org/packages/19/28/65e0980d668a6d44e699f59d1acf43d6b5d4893592c115ce7c680bb4dfa1/uuid_utils-0.17.0-cp314-cp314-macosx_10_12_x86_64.whl", hash = "sha256:387cf7437c94ddec08651a0f1081381299c7075bc48a6251d8922bf39973378a", size = 287858, upload-time = "2026-07-09T13:49:07.45Z" },
+    { url = "https://files.pythonhosted.org/packages/8f/8d/5e97bcebc90fb6a10f98af3dc1ba552e04183aba59e2edc0b9cf486dd998/uuid_utils-0.17.0-cp314-cp314-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:220b52746d99e11964badac3c0869016e0c24bafb70a7dd5c2c072a6be3da9cc", size = 321587, upload-time = "2026-07-09T13:49:09.489Z" },
+    { url = "https://files.pythonhosted.org/packages/8b/d7/88b2a2370cc3d455ba0515fb6f5c8f7ac0c0f55a86801b6e56a432f22c17/uuid_utils-0.17.0-cp314-cp314-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:0ab4a66e7a035ad6625cfc1fbdb34f5c2d25a80ae1ef4bfee458ea2036333c6d", size = 328964, upload-time = "2026-07-09T13:49:11.292Z" },
+    { url = "https://files.pythonhosted.org/packages/bd/0f/181c5da673953dfc0958cb4fb3a4984a9098673ddb05cac68e994bc8511b/uuid_utils-0.17.0-cp314-cp314-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:5641071337eb11d61a001ea08793bf72216f3241f0a433ed2764804b2a3e3cc7", size = 442909, upload-time = "2026-07-09T13:49:12.644Z" },
+    { url = "https://files.pythonhosted.org/packages/ec/38/5c5e665af542884a8fd3c61725c38453239e13940326b5b70f3ef8881a97/uuid_utils-0.17.0-cp314-cp314-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:9082e709014946b1f6e96ae6ecd93652efca2d2a6a3ab67dbe151c8b4bf193a4", size = 323076, upload-time = "2026-07-09T13:49:13.897Z" },
+    { url = "https://files.pythonhosted.org/packages/f5/35/7de97de18cbf226c2a4f2104ad15e56ca4491717c81c0b71795c0c585b4e/uuid_utils-0.17.0-cp314-cp314-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:1fd6f0e8a162dc0e9255b6aebe3cd175e76c33202f1bf39da9e6294b93db0099", size = 347360, upload-time = "2026-07-09T13:49:15.237Z" },
+    { url = "https://files.pythonhosted.org/packages/26/a1/9915d5dd59fdd1957ded5d188c0ea0b9db5a1d84d42c8d8828a7b83b366e/uuid_utils-0.17.0-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:d63010803d7c368963bbe6f7ec379593e76dd581d7db0f29118d88713c9e0354", size = 499267, upload-time = "2026-07-09T13:49:16.774Z" },
+    { url = "https://files.pythonhosted.org/packages/c0/05/88108405262ec850cea0f95733445d6873e5772af3292baabd9ef8457740/uuid_utils-0.17.0-cp314-cp314-musllinux_1_2_armv7l.whl", hash = "sha256:a46bedc273b6f58f11dee816ff74999625ef8d007890f411b7a4975bf1c89330", size = 604940, upload-time = "2026-07-09T13:49:18.147Z" },
+    { url = "https://files.pythonhosted.org/packages/89/d5/6dbcd300de47cc443cff2656cd5327a385751213dcb2101cfee7388170b2/uuid_utils-0.17.0-cp314-cp314-musllinux_1_2_i686.whl", hash = "sha256:405233a5f625b3d995648f4647fa6befa4567cf3f74e1f6b9837e16f7310f0e0", size = 564172, upload-time = "2026-07-09T13:49:19.593Z" },
+    { url = "https://files.pythonhosted.org/packages/ab/94/e8057f2288a415fba8a978bca4b589f5cb6b91a028a5dc07a1775938b33f/uuid_utils-0.17.0-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:b6c5d2d71e1f17329150ad9427d27f4a3f29a01792e7ecdc64a98ac5368fc4d5", size = 528533, upload-time = "2026-07-09T13:49:21.075Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/6b/31713148c77e48e62f51aa042a98a54a8be0396912ea5130f83f52ae722d/uuid_utils-0.17.0-cp314-cp314-pyemscripten_2026_0_wasm32.whl", hash = "sha256:f7e9b8728ba07a3cb2f29d5aa1a266c2664eb8ef0fd43afa34627c92f7fac8f0", size = 99197, upload-time = "2026-07-09T13:49:22.351Z" },
+    { url = "https://files.pythonhosted.org/packages/f3/f3/ca6f6ac5428312df8ed632f6dd9f9e6aba23090471fcdeae53eab027e8b3/uuid_utils-0.17.0-cp314-cp314-win32.whl", hash = "sha256:58838921e377791ef22c64cc92141bfae030f43651ff9272f0f28a208a9e6a5a", size = 169540, upload-time = "2026-07-09T13:49:23.563Z" },
+    { url = "https://files.pythonhosted.org/packages/c6/cd/7ede0db66411fa09817d79b680f7454ea9bee2d374e1922e4efd065760a3/uuid_utils-0.17.0-cp314-cp314-win_amd64.whl", hash = "sha256:42275ebd0e8e74e32cdbfb8bd88fc99576567d51d54a508020611fd8f4f463a0", size = 175984, upload-time = "2026-07-09T13:49:24.703Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/81/533b5f80cd4918c0693f4e1b7b90ceb1caa45f4266ae8b528135d7ecca5d/uuid_utils-0.17.0-cp314-cp314-win_arm64.whl", hash = "sha256:b5d11cccba076a32321ef1380dea956821f0b51794ef59df64e58fb1cd543aae", size = 174749, upload-time = "2026-07-09T13:49:25.886Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/13/f400ac39d06fd8be5b099c09e41bb975205926722a3e8d53348817cb7ff9/uuid_utils-0.17.0-cp314-cp314t-macosx_10_12_x86_64.macosx_11_0_arm64.macosx_10_12_universal2.whl", hash = "sha256:fae8b282f0cb22a5de222999f7723f4e5ec04f6fcdf4aaef879b5b36625ae2b0", size = 562610, upload-time = "2026-07-09T13:49:27.374Z" },
+    { url = "https://files.pythonhosted.org/packages/03/8c/c71c8312304c56f6d0bcba87cd402fa79bec35d18ffc8c41954196ca68e5/uuid_utils-0.17.0-cp314-cp314t-macosx_10_12_x86_64.whl", hash = "sha256:967955620df45e6cffe2e9950cb9903cb455649396f896b26b04363a91a5054b", size = 289473, upload-time = "2026-07-09T13:49:28.989Z" },
+    { url = "https://files.pythonhosted.org/packages/bb/cd/522117e2e5184ca1d4f0f85ee833e9e21bd8c6b99eff8a4d1a8e5a194e33/uuid_utils-0.17.0-cp314-cp314t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:375cde148430d60a4a07c03abaa0774c4fddfdd90de99b4ba02f24088bc9d750", size = 321600, upload-time = "2026-07-09T13:49:30.4Z" },
+    { url = "https://files.pythonhosted.org/packages/6d/f4/0d81f9bd346fc717bc561c08fa6457e0328966eb76e536b938fe77d56459/uuid_utils-0.17.0-cp314-cp314t-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:975c17da26c5b9d46c336b03c52a057ac28378d6f9d98b58d32a038589bb3912", size = 329569, upload-time = "2026-07-09T13:49:31.732Z" },
+    { url = "https://files.pythonhosted.org/packages/5e/41/26e1363f36a94c9e8ec2dd21d5f63088d3e7c723adbb12dcc8fdc77be417/uuid_utils-0.17.0-cp314-cp314t-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:3150d836290c88f1d26eb59c4db280d87417dd3bfaadd2889c77416c8f0ff6fa", size = 442051, upload-time = "2026-07-09T13:49:33.024Z" },
+    { url = "https://files.pythonhosted.org/packages/2b/a7/2c1ed1b34d7df7fdcc11c28fd26d94d44843b37d9af2435ff9fd8abdbc08/uuid_utils-0.17.0-cp314-cp314t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:9472a8de37faf8bd216c628e0e68c8f6bef730d3ba0a5060f3b0fa460c992ac2", size = 324372, upload-time = "2026-07-09T13:49:34.554Z" },
+    { url = "https://files.pythonhosted.org/packages/78/bf/328d3c6bb22c496944a1b3b732207d71aa6964eb604e5e3b9dcb91ed0a00/uuid_utils-0.17.0-cp314-cp314t-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:d27c531edb8d1f38ca2eddaa1fa24913a460aeb721f2efd4ef42a124ce94e354", size = 348548, upload-time = "2026-07-09T13:49:35.898Z" },
+    { url = "https://files.pythonhosted.org/packages/3e/76/a07de5cb7b90582fdbbc830fd19be129cbbb9897cfe239fef469d7bd2d09/uuid_utils-0.17.0-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:5670c52a438e21483ce715776144914a4e2a2a5c62d9dee15f8a3e90cf128ae6", size = 498985, upload-time = "2026-07-09T13:49:37.142Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/62/9966e46ae34fcec6b06119631fb3c09705ea78835035ce3a82d3348eb61a/uuid_utils-0.17.0-cp314-cp314t-musllinux_1_2_armv7l.whl", hash = "sha256:6f29689a76fe7a49cbd629a794d0ec1eab48814e323a00a146a741b0195bde68", size = 605183, upload-time = "2026-07-09T13:49:38.648Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/4e/bb962ba0fe31e903b199f22cf4c1a6cba35a8987aef526d287277ab8ca8b/uuid_utils-0.17.0-cp314-cp314t-musllinux_1_2_i686.whl", hash = "sha256:4441600447d340ae103a353f01dbcd22ff680e5ee1a22988efe8d7b791d8fdb3", size = 565412, upload-time = "2026-07-09T13:49:40.115Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/9e/122adfeeeae8a84ccfd43bce627b104d12a2180a93bffd2c0e1b54dad7a6/uuid_utils-0.17.0-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:e7b04935a79c03c41ad08d0a5f390aac968bfb561f1268897bc5b0f077971efd", size = 529885, upload-time = "2026-07-09T13:49:41.513Z" },
+    { url = "https://files.pythonhosted.org/packages/b3/4f/257304dded339dc35fc9bf35722ac68fd4fdb930f255b8f7bccdf74ebba9/uuid_utils-0.17.0-cp314-cp314t-win32.whl", hash = "sha256:239d8a281fe10bae33205b5d43185834d556b18434e0a113b5dc1dfb2fd97e91", size = 169472, upload-time = "2026-07-09T13:49:42.871Z" },
+    { url = "https://files.pythonhosted.org/packages/35/c8/e78c06db7e9ce317ce7b8759ff2058333eac75caa8c22b75f0059589c9be/uuid_utils-0.17.0-cp314-cp314t-win_amd64.whl", hash = "sha256:e288a06cbbbcd01b44386e767985c9e21d2ad9bf59829aa7058d9a2a494804ab", size = 176271, upload-time = "2026-07-09T13:49:44.105Z" },
+    { url = "https://files.pythonhosted.org/packages/a7/11/bd1c70e1ad3301163cebe66c8d26de26e6814d52f642a849448bd2833626/uuid_utils-0.17.0-cp314-cp314t-win_arm64.whl", hash = "sha256:1776a80d16369999b21627028cc5dbce819be83e1e079fdd7a51b587d2916db9", size = 175004, upload-time = "2026-07-09T13:49:45.591Z" },
+]
+
+[[package]]
+name = "uvicorn"
+version = "0.51.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "click", marker = "sys_platform != 'emscripten'" },
+    { name = "h11", marker = "sys_platform != 'emscripten'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/a2/65/b7c6c443ccc58678c91e1e973bbe2a878591538655d6e1d47f24ba1c51f3/uvicorn-0.51.0.tar.gz", hash = "sha256:f6f4b69b657c312f516dd2d268ab9ae6f254b11e4bac504f37b2ab58b24dd0b0", size = 94412, upload-time = "2026-07-08T10:59:05.962Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/45/ec/dbb7e5a6b91f86bfb9eb7d2988a2730907b6a729875b949c7f022e8b88fa/uvicorn-0.51.0-py3-none-any.whl", hash = "sha256:5d38af6cd620f2ae3849fb44fd4879e0890aa1febe8d47eb355fb45d93fe6a5b", size = 73219, upload-time = "2026-07-08T10:59:04.44Z" },
+]
+
+[[package]]
+name = "w3lib"
+version = "2.4.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/c0/91/b2eb59c2cf243de5de1e91c963655df78c015509f51297685a8c86a27b8c/w3lib-2.4.1.tar.gz", hash = "sha256:8dd69ee39ff6398d708c793abc779c334a69bac7cee1cdf71736c669ed6be864", size = 48494, upload-time = "2026-03-20T09:50:27.477Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/66/c3/f8b216cbd742e5b84c40f045204c764ccb7524d2aeab021054ec69446b0a/w3lib-2.4.1-py3-none-any.whl", hash = "sha256:40930132907e68de906a5b89331ab8c8ff4f01bd35b5539ef7896017d814138d", size = 21695, upload-time = "2026-03-20T09:50:26.187Z" },
+]
+
+[[package]]
+name = "wcwidth"
+version = "0.8.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/34/74/c6428f875774288bec1396f5bfcbc2d925700a4dad61727fd5f2b12f249d/wcwidth-0.8.2.tar.gz", hash = "sha256:91fbef97204b96a3d4d421609b80340b760cf33e26da123ff243d76b1fda8dda", size = 1466253, upload-time = "2026-06-29T18:11:11.601Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/96/42/3e5985a0a7e57de470b320c6d6a1a67c844f6737a587f3d44dd13d1819e7/wcwidth-0.8.2-py3-none-any.whl", hash = "sha256:d63947694a0539a1d51e01eda7caf800c291020e6cdd7e28ad7b14dd33ad4f85", size = 323166, upload-time = "2026-06-29T18:11:09.888Z" },
+]
+
+[[package]]
+name = "websocket-client"
+version = "1.9.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/2c/41/aa4bf9664e4cda14c3b39865b12251e8e7d239f4cd0e3cc1b6c2ccde25c1/websocket_client-1.9.0.tar.gz", hash = "sha256:9e813624b6eb619999a97dc7958469217c3176312b3a16a4bd1bc7e08a46ec98", size = 70576, upload-time = "2025-10-07T21:16:36.495Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/34/db/b10e48aa8fff7407e67470363eac595018441cf32d5e1001567a7aeba5d2/websocket_client-1.9.0-py3-none-any.whl", hash = "sha256:af248a825037ef591efbf6ed20cc5faa03d3b47b9e5a2230a529eeee1c1fc3ef", size = 82616, upload-time = "2025-10-07T21:16:34.951Z" },
+]
+
+[[package]]
+name = "websockets"
+version = "15.0.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/21/e6/26d09fab466b7ca9c7737474c52be4f76a40301b08362eb2dbc19dcc16c1/websockets-15.0.1.tar.gz", hash = "sha256:82544de02076bafba038ce055ee6412d68da13ab47f0c60cab827346de828dee", size = 177016, upload-time = "2025-03-05T20:03:41.606Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/51/6b/4545a0d843594f5d0771e86463606a3988b5a09ca5123136f8a76580dd63/websockets-15.0.1-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:3e90baa811a5d73f3ca0bcbf32064d663ed81318ab225ee4f427ad4e26e5aff3", size = 175437, upload-time = "2025-03-05T20:02:16.706Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/71/809a0f5f6a06522af902e0f2ea2757f71ead94610010cf570ab5c98e99ed/websockets-15.0.1-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:592f1a9fe869c778694f0aa806ba0374e97648ab57936f092fd9d87f8bc03665", size = 173096, upload-time = "2025-03-05T20:02:18.832Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/69/1a681dd6f02180916f116894181eab8b2e25b31e484c5d0eae637ec01f7c/websockets-15.0.1-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:0701bc3cfcb9164d04a14b149fd74be7347a530ad3bbf15ab2c678a2cd3dd9a2", size = 173332, upload-time = "2025-03-05T20:02:20.187Z" },
+    { url = "https://files.pythonhosted.org/packages/a6/02/0073b3952f5bce97eafbb35757f8d0d54812b6174ed8dd952aa08429bcc3/websockets-15.0.1-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:e8b56bdcdb4505c8078cb6c7157d9811a85790f2f2b3632c7d1462ab5783d215", size = 183152, upload-time = "2025-03-05T20:02:22.286Z" },
+    { url = "https://files.pythonhosted.org/packages/74/45/c205c8480eafd114b428284840da0b1be9ffd0e4f87338dc95dc6ff961a1/websockets-15.0.1-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:0af68c55afbd5f07986df82831c7bff04846928ea8d1fd7f30052638788bc9b5", size = 182096, upload-time = "2025-03-05T20:02:24.368Z" },
+    { url = "https://files.pythonhosted.org/packages/14/8f/aa61f528fba38578ec553c145857a181384c72b98156f858ca5c8e82d9d3/websockets-15.0.1-cp312-cp312-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:64dee438fed052b52e4f98f76c5790513235efaa1ef7f3f2192c392cd7c91b65", size = 182523, upload-time = "2025-03-05T20:02:25.669Z" },
+    { url = "https://files.pythonhosted.org/packages/ec/6d/0267396610add5bc0d0d3e77f546d4cd287200804fe02323797de77dbce9/websockets-15.0.1-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:d5f6b181bb38171a8ad1d6aa58a67a6aa9d4b38d0f8c5f496b9e42561dfc62fe", size = 182790, upload-time = "2025-03-05T20:02:26.99Z" },
+    { url = "https://files.pythonhosted.org/packages/02/05/c68c5adbf679cf610ae2f74a9b871ae84564462955d991178f95a1ddb7dd/websockets-15.0.1-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:5d54b09eba2bada6011aea5375542a157637b91029687eb4fdb2dab11059c1b4", size = 182165, upload-time = "2025-03-05T20:02:30.291Z" },
+    { url = "https://files.pythonhosted.org/packages/29/93/bb672df7b2f5faac89761cb5fa34f5cec45a4026c383a4b5761c6cea5c16/websockets-15.0.1-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:3be571a8b5afed347da347bfcf27ba12b069d9d7f42cb8c7028b5e98bbb12597", size = 182160, upload-time = "2025-03-05T20:02:31.634Z" },
+    { url = "https://files.pythonhosted.org/packages/ff/83/de1f7709376dc3ca9b7eeb4b9a07b4526b14876b6d372a4dc62312bebee0/websockets-15.0.1-cp312-cp312-win32.whl", hash = "sha256:c338ffa0520bdb12fbc527265235639fb76e7bc7faafbb93f6ba80d9c06578a9", size = 176395, upload-time = "2025-03-05T20:02:33.017Z" },
+    { url = "https://files.pythonhosted.org/packages/7d/71/abf2ebc3bbfa40f391ce1428c7168fb20582d0ff57019b69ea20fa698043/websockets-15.0.1-cp312-cp312-win_amd64.whl", hash = "sha256:fcd5cf9e305d7b8338754470cf69cf81f420459dbae8a3b40cee57417f4614a7", size = 176841, upload-time = "2025-03-05T20:02:34.498Z" },
+    { url = "https://files.pythonhosted.org/packages/cb/9f/51f0cf64471a9d2b4d0fc6c534f323b664e7095640c34562f5182e5a7195/websockets-15.0.1-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:ee443ef070bb3b6ed74514f5efaa37a252af57c90eb33b956d35c8e9c10a1931", size = 175440, upload-time = "2025-03-05T20:02:36.695Z" },
+    { url = "https://files.pythonhosted.org/packages/8a/05/aa116ec9943c718905997412c5989f7ed671bc0188ee2ba89520e8765d7b/websockets-15.0.1-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:5a939de6b7b4e18ca683218320fc67ea886038265fd1ed30173f5ce3f8e85675", size = 173098, upload-time = "2025-03-05T20:02:37.985Z" },
+    { url = "https://files.pythonhosted.org/packages/ff/0b/33cef55ff24f2d92924923c99926dcce78e7bd922d649467f0eda8368923/websockets-15.0.1-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:746ee8dba912cd6fc889a8147168991d50ed70447bf18bcda7039f7d2e3d9151", size = 173329, upload-time = "2025-03-05T20:02:39.298Z" },
+    { url = "https://files.pythonhosted.org/packages/31/1d/063b25dcc01faa8fada1469bdf769de3768b7044eac9d41f734fd7b6ad6d/websockets-15.0.1-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:595b6c3969023ecf9041b2936ac3827e4623bfa3ccf007575f04c5a6aa318c22", size = 183111, upload-time = "2025-03-05T20:02:40.595Z" },
+    { url = "https://files.pythonhosted.org/packages/93/53/9a87ee494a51bf63e4ec9241c1ccc4f7c2f45fff85d5bde2ff74fcb68b9e/websockets-15.0.1-cp313-cp313-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:3c714d2fc58b5ca3e285461a4cc0c9a66bd0e24c5da9911e30158286c9b5be7f", size = 182054, upload-time = "2025-03-05T20:02:41.926Z" },
+    { url = "https://files.pythonhosted.org/packages/ff/b2/83a6ddf56cdcbad4e3d841fcc55d6ba7d19aeb89c50f24dd7e859ec0805f/websockets-15.0.1-cp313-cp313-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:0f3c1e2ab208db911594ae5b4f79addeb3501604a165019dd221c0bdcabe4db8", size = 182496, upload-time = "2025-03-05T20:02:43.304Z" },
+    { url = "https://files.pythonhosted.org/packages/98/41/e7038944ed0abf34c45aa4635ba28136f06052e08fc2168520bb8b25149f/websockets-15.0.1-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:229cf1d3ca6c1804400b0a9790dc66528e08a6a1feec0d5040e8b9eb14422375", size = 182829, upload-time = "2025-03-05T20:02:48.812Z" },
+    { url = "https://files.pythonhosted.org/packages/e0/17/de15b6158680c7623c6ef0db361da965ab25d813ae54fcfeae2e5b9ef910/websockets-15.0.1-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:756c56e867a90fb00177d530dca4b097dd753cde348448a1012ed6c5131f8b7d", size = 182217, upload-time = "2025-03-05T20:02:50.14Z" },
+    { url = "https://files.pythonhosted.org/packages/33/2b/1f168cb6041853eef0362fb9554c3824367c5560cbdaad89ac40f8c2edfc/websockets-15.0.1-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:558d023b3df0bffe50a04e710bc87742de35060580a293c2a984299ed83bc4e4", size = 182195, upload-time = "2025-03-05T20:02:51.561Z" },
+    { url = "https://files.pythonhosted.org/packages/86/eb/20b6cdf273913d0ad05a6a14aed4b9a85591c18a987a3d47f20fa13dcc47/websockets-15.0.1-cp313-cp313-win32.whl", hash = "sha256:ba9e56e8ceeeedb2e080147ba85ffcd5cd0711b89576b83784d8605a7df455fa", size = 176393, upload-time = "2025-03-05T20:02:53.814Z" },
+    { url = "https://files.pythonhosted.org/packages/1b/6c/c65773d6cab416a64d191d6ee8a8b1c68a09970ea6909d16965d26bfed1e/websockets-15.0.1-cp313-cp313-win_amd64.whl", hash = "sha256:e09473f095a819042ecb2ab9465aee615bd9c2028e4ef7d933600a8401c79561", size = 176837, upload-time = "2025-03-05T20:02:55.237Z" },
+    { url = "https://files.pythonhosted.org/packages/fa/a8/5b41e0da817d64113292ab1f8247140aac61cbf6cfd085d6a0fa77f4984f/websockets-15.0.1-py3-none-any.whl", hash = "sha256:f7a866fbc1e97b5c617ee4116daaa09b722101d4a3c170c787450ba409f9736f", size = 169743, upload-time = "2025-03-05T20:03:39.41Z" },
+]
+
+[[package]]
+name = "wsproto"
+version = "1.3.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "h11" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/c7/79/12135bdf8b9c9367b8701c2c19a14c913c120b882d50b014ca0d38083c2c/wsproto-1.3.2.tar.gz", hash = "sha256:b86885dcf294e15204919950f666e06ffc6c7c114ca900b060d6e16293528294", size = 50116, upload-time = "2025-11-20T18:18:01.871Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a4/f5/10b68b7b1544245097b2a1b8238f66f2fc6dcaeb24ba5d917f52bd2eed4f/wsproto-1.3.2-py3-none-any.whl", hash = "sha256:61eea322cdf56e8cc904bd3ad7573359a242ba65688716b0710a5eb12beab584", size = 24405, upload-time = "2025-11-20T18:18:00.454Z" },
+]
+
+[[package]]
+name = "xlrd"
+version = "2.0.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/07/5a/377161c2d3538d1990d7af382c79f3b2372e880b65de21b01b1a2b78691e/xlrd-2.0.2.tar.gz", hash = "sha256:08b5e25de58f21ce71dc7db3b3b8106c1fa776f3024c54e45b45b374e89234c9", size = 100167, upload-time = "2025-06-14T08:46:39.039Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/1a/62/c8d562e7766786ba6587d09c5a8ba9f718ed3fa8af7f4553e8f91c36f302/xlrd-2.0.2-py2.py3-none-any.whl", hash = "sha256:ea762c3d29f4cca48d82df517b6d89fbce4db3107f9d78713e48cd321d5c9aa9", size = 96555, upload-time = "2025-06-14T08:46:37.766Z" },
+]
+
+[[package]]
+name = "xxhash"
+version = "3.8.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/8e/63/71aa56b151a1b28770037a61bd4e461c2619cfc8866a4fcaf1548605e325/xxhash-3.8.1.tar.gz", hash = "sha256:b0de4bf3aa66363552d52c6a89003c479911f12098cd48a53d44a0f7a25f7c46", size = 86223, upload-time = "2026-07-06T10:49:58.937Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/42/91/f65c34a7aa7b4e7cf4854f8e6ef3f7ee32ceac41d4f008da0780db0612f6/xxhash-3.8.1-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:e6e49370822c1f4d8d90e678b06dbcb08b51a026a7c4b55479e7d467f2e813bc", size = 34680, upload-time = "2026-07-06T10:44:40.932Z" },
+    { url = "https://files.pythonhosted.org/packages/57/04/b10a245a4c09a9cfa88f8e9ae755029413ad1ac17047f9a61906e5ae0799/xxhash-3.8.1-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:220d68130f83f7cc86d6edfdeab176adc73d7200bf3a8ec10c629e8cf605c215", size = 32397, upload-time = "2026-07-06T10:44:42.196Z" },
+    { url = "https://files.pythonhosted.org/packages/3a/75/45ab795b5945b6388583bd75202106af505537935566c15a1577797a0e08/xxhash-3.8.1-cp312-cp312-manylinux1_i686.manylinux_2_28_i686.manylinux_2_5_i686.whl", hash = "sha256:4d365ee1892c1fa803536f8c6ce21d24b29c9718ec75eb856095c07830f8c478", size = 220549, upload-time = "2026-07-06T10:44:43.603Z" },
+    { url = "https://files.pythonhosted.org/packages/13/44/5ba2bd0a14ddf4193fc7d8ec29625f659f22c06d60b28f04bf46305d8330/xxhash-3.8.1-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:852bfe059720632e2f16a6a4745e41d20937b2bf2a42a401e2412046bb6971cc", size = 241186, upload-time = "2026-07-06T10:44:45.534Z" },
+    { url = "https://files.pythonhosted.org/packages/23/32/c4147def4d1e4538b906f82731e0ba23424377fc50a7cddd03cd284c8f63/xxhash-3.8.1-cp312-cp312-manylinux2014_armv7l.manylinux_2_17_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:2f8c25a7061d952de589bd0ea0eaadee32378ff83dd6a677b267f9cd86f401f8", size = 264852, upload-time = "2026-07-06T10:44:47.199Z" },
+    { url = "https://files.pythonhosted.org/packages/6c/bd/71ed14f4f0318bb7fd7b2ec51999413487fa8da8d41208e84d50d1ef0f98/xxhash-3.8.1-cp312-cp312-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:868a8dcaff1a84ba78038e1cef14fc88ccf84d9b4d12ea604696e0693296aa56", size = 242663, upload-time = "2026-07-06T10:44:48.846Z" },
+    { url = "https://files.pythonhosted.org/packages/91/09/70af22c565a8473b3f2ae73f88e7721af281bc4a575236dbd1970c9f76f6/xxhash-3.8.1-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:6536d8677d2fff7e64cd0b98b976df9de7aee0e69590044c2af5f51b76b7a170", size = 473510, upload-time = "2026-07-06T10:44:50.695Z" },
+    { url = "https://files.pythonhosted.org/packages/18/96/34db781c8f0cf99c544ca1f2bc2e5bf55426e1eb4ca6de8ea5da56a9f352/xxhash-3.8.1-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:82c0cedd280eab2e8291270e6c04894dbc096f8159a39dcf1807429f026ca3cc", size = 220469, upload-time = "2026-07-06T10:44:52.422Z" },
+    { url = "https://files.pythonhosted.org/packages/93/5f/9a184f615fa5a4dce30c01534f62946ce5a11ce40f73785cbd356ccabaa9/xxhash-3.8.1-cp312-cp312-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:daa86e4b68221d38e669bb236ba112d0335353829fb627c82e5909e4bbe8694c", size = 310290, upload-time = "2026-07-06T10:44:54.142Z" },
+    { url = "https://files.pythonhosted.org/packages/a9/dc/9b9a9789011ee153723a5eb9e7dd7fcbae2ba9b3fe7a729249ca7c252056/xxhash-3.8.1-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:2bc7113e6f2b6b3922dd61796ca9f36af09da3773898e7003038dc992fc83b8d", size = 238173, upload-time = "2026-07-06T10:44:55.693Z" },
+    { url = "https://files.pythonhosted.org/packages/ec/4d/71c6005ada9dcb608a4e1902e8475ecadb5f3fbfa04e1e244d276a2d0c43/xxhash-3.8.1-cp312-cp312-musllinux_1_2_armv7l.whl", hash = "sha256:5eed32dad81d6ba8e62dc7b9ffa0500199385d7810a8dd9d4eafaceb8c6e20bb", size = 269026, upload-time = "2026-07-06T10:44:57.424Z" },
+    { url = "https://files.pythonhosted.org/packages/2f/87/d6c036ba25dfbd9c8633be5aa86fc9474bbb9e2c68212a841d090abe7344/xxhash-3.8.1-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:83697b0ea1f10e7f5d8b26a4906fa851393c61546c63839643a2b7fe2d868061", size = 224970, upload-time = "2026-07-06T10:44:59.085Z" },
+    { url = "https://files.pythonhosted.org/packages/48/62/4c1f035a41c5752aa05e195b6c904c07b94fe9061a16de61e72a6e6b135f/xxhash-3.8.1-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:36fc69160465ae75c6ec4ac9f781bb2aa16ae7ff869e73c26fee85fbb11b9887", size = 240820, upload-time = "2026-07-06T10:45:00.746Z" },
+    { url = "https://files.pythonhosted.org/packages/da/14/d39d565069b87e86d21a2af2a31d04db79249d25aa8d5b62959056a89857/xxhash-3.8.1-cp312-cp312-musllinux_1_2_riscv64.whl", hash = "sha256:445e0f5a31f2f3546ae0895d4811e159518cdc9d824c11419898d40cfadb677e", size = 300619, upload-time = "2026-07-06T10:45:02.716Z" },
+    { url = "https://files.pythonhosted.org/packages/13/22/75467acc887edc8cf71c97ab1708feb3df7a88bda589b9f399765c6387d2/xxhash-3.8.1-cp312-cp312-musllinux_1_2_s390x.whl", hash = "sha256:dfe0580fbfd5e4af87d0cc52d2044f155d55ebd8c8a93568758a2ea7d8e15975", size = 443267, upload-time = "2026-07-06T10:45:04.653Z" },
+    { url = "https://files.pythonhosted.org/packages/a4/b6/1da3baa5fa6ef705e3425fddd382be7dfc4dfba2686df90a20f16e9c7b1b/xxhash-3.8.1-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:095e1323fa108be1292c54c86da3ef3c7a7dc015b105a52133973bc07a6ad11a", size = 217338, upload-time = "2026-07-06T10:45:06.304Z" },
+    { url = "https://files.pythonhosted.org/packages/78/dd/b5295a9f97484e7a1c2b283a742ca45e3104991c55a1ef670dde161829ba/xxhash-3.8.1-cp312-cp312-win32.whl", hash = "sha256:bf28f55e427e0483acb1f666bd0d869b6d5e5a716680c216ad7befe3d4cfba2e", size = 31970, upload-time = "2026-07-06T10:45:07.823Z" },
+    { url = "https://files.pythonhosted.org/packages/ec/31/3fa0b807d7e21515cd975e7fe5c039d52ac3e9401a96d6ad68dae6305215/xxhash-3.8.1-cp312-cp312-win_amd64.whl", hash = "sha256:2256e80e4960ee282f63428adb349cb7f8bd8efe4db770d88eb815f4b9860724", size = 32741, upload-time = "2026-07-06T10:45:09.42Z" },
+    { url = "https://files.pythonhosted.org/packages/b8/05/86feada74e239600e6875aa507afb40482a89b92700aa74a92da83bdcb77/xxhash-3.8.1-cp312-cp312-win_arm64.whl", hash = "sha256:9df56e6df96a60590935e22373041cccc91fd55858763dcffb55bf63b3a2b396", size = 29234, upload-time = "2026-07-06T10:45:10.809Z" },
+    { url = "https://files.pythonhosted.org/packages/6b/8c/446bb782cd0d27007a917b5569a08dd73219c3e8d6e459014db104b27bdb/xxhash-3.8.1-cp313-cp313-android_21_arm64_v8a.whl", hash = "sha256:3c682fcd96eb4bf64be32a4d95f96107e1588005831bd8a741b324fdda01b913", size = 38562, upload-time = "2026-07-06T10:45:12.425Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/ec/c0c45627eaa6be7a5d6117423adf8f7a15b17ee74b4b17072cca5959a225/xxhash-3.8.1-cp313-cp313-android_21_x86_64.whl", hash = "sha256:036a024d8b9c01f70782e09ed98d532e76fd23f950ae7154bd950fe94e90ebec", size = 36656, upload-time = "2026-07-06T10:45:13.932Z" },
+    { url = "https://files.pythonhosted.org/packages/f6/94/8324c04cc7597154caaeba6c094e01fbd2e7601d01e7a13eea9f5420e77b/xxhash-3.8.1-cp313-cp313-ios_13_0_arm64_iphoneos.whl", hash = "sha256:d6a5c0bce213b23b0166fe0d35bcbbe23ce4b968f257cc7eb6fd57cb8e1e6297", size = 31169, upload-time = "2026-07-06T10:45:15.687Z" },
+    { url = "https://files.pythonhosted.org/packages/40/a4/beb6bb26e1184e126dbe7a5682330214ef54dcfbf882078aa9f4b5428d42/xxhash-3.8.1-cp313-cp313-ios_13_0_arm64_iphonesimulator.whl", hash = "sha256:5177aa44eddaa97c6ef0cc00c6d540edb64d51781d2f8fb941612ec61a92c9ed", size = 32177, upload-time = "2026-07-06T10:45:17.035Z" },
+    { url = "https://files.pythonhosted.org/packages/56/0f/fc4c92a5a528f839b34b6419b2e53c8597f2a629d5a1f5d721f65bfa1fd6/xxhash-3.8.1-cp313-cp313-ios_13_0_x86_64_iphonesimulator.whl", hash = "sha256:7801b7223db017b9c0c9ccf37e44524edb35a1544a1c032add22c061c6af0276", size = 34642, upload-time = "2026-07-06T10:45:18.39Z" },
+    { url = "https://files.pythonhosted.org/packages/d4/58/edbfb141d4000767ac6a9694f8ac0763e2c2e983e65c9e31620ba56e2667/xxhash-3.8.1-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:9e80238259655bf69d7bcd08226a970d7f42605f3157786bfa76dd13472d7fa0", size = 34684, upload-time = "2026-07-06T10:45:20.033Z" },
+    { url = "https://files.pythonhosted.org/packages/07/3f/5072f1f0f5714186f0ac2a0b5a4929ce30d4b845e94886b6c01b6ebda0be/xxhash-3.8.1-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:bcab50a389cc04d87f90092af78a6adba2ab3deca63175a3344ca83514045315", size = 32401, upload-time = "2026-07-06T10:45:21.414Z" },
+    { url = "https://files.pythonhosted.org/packages/49/c7/802ea2f9c2ed59219934d6d65c470d502b1788043eae277a52af8658bda6/xxhash-3.8.1-cp313-cp313-manylinux1_i686.manylinux_2_28_i686.manylinux_2_5_i686.whl", hash = "sha256:a2489d3a776fa380cb8e71f54c7fda268a9baf3de9b1395093fd280f95735907", size = 220617, upload-time = "2026-07-06T10:45:23.234Z" },
+    { url = "https://files.pythonhosted.org/packages/99/a8/e10488efd31fcb13fcd6acbc6e788f10c6f8e3a0cc4ae3eb89dc19c55a12/xxhash-3.8.1-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:32ab1e5432690276e71192be7401b55f96db2d0eedea5d44eb1f164505669cc0", size = 241295, upload-time = "2026-07-06T10:45:25.364Z" },
+    { url = "https://files.pythonhosted.org/packages/18/cc/14180b17d44892a631f8ae7323c30bfbb1328efc8209e528a480293528ac/xxhash-3.8.1-cp313-cp313-manylinux2014_armv7l.manylinux_2_17_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:b30e01a0b97a4bc3f519a4d7a82da3dc53251fb0de5eeea8660dcd4ff094c0c2", size = 264688, upload-time = "2026-07-06T10:45:27.09Z" },
+    { url = "https://files.pythonhosted.org/packages/a9/72/a14019d0c5f6c41ee407a503036ae32787c91325ca218a96a9b5627be651/xxhash-3.8.1-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:1f44275ddb0978b67a58a951501903f04d49335a91f7681c9ce122ecb8ccb329", size = 242740, upload-time = "2026-07-06T10:45:28.753Z" },
+    { url = "https://files.pythonhosted.org/packages/68/08/92550e556c6fcfcb96c6a336945eb53a431ed43120ed749636debb16c5cf/xxhash-3.8.1-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:e3b87cbd974512c0c5fc7b469c36b2cdc9ee6d76e4ec78bccb2c7184611c49b0", size = 473599, upload-time = "2026-07-06T10:45:30.524Z" },
+    { url = "https://files.pythonhosted.org/packages/29/83/e361d3c1acd1b21e1d489616de6fa4aaf843365d8179f612e3743eac20a9/xxhash-3.8.1-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:98ee81b4b7f3023c9cb04a78cc67610baffcb5812d92f2096cb5a5efc6f19437", size = 220559, upload-time = "2026-07-06T10:45:32.979Z" },
+    { url = "https://files.pythonhosted.org/packages/05/01/006a4243c2c2a6831827f9999f6d1c23feeef100eb023c1f886022a00bf3/xxhash-3.8.1-cp313-cp313-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:2666f059a1588a99267e33605365ed89cea92f424b3522806a9f4bd8ad2e3d62", size = 310383, upload-time = "2026-07-06T10:45:35.875Z" },
+    { url = "https://files.pythonhosted.org/packages/d8/20/af388e8bf9f9a0f89eeef7d2a1935d176ee1c20bc6adeda05035879379cf/xxhash-3.8.1-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:b0093cf7eeb91b84776e8742113afa4bdf47533d36cf719179aaaf1f56f6f8bf", size = 238228, upload-time = "2026-07-06T10:45:38.02Z" },
+    { url = "https://files.pythonhosted.org/packages/63/6b/4666579a87eebd1744663c404297355fa0658617b015cedfa58810ee7036/xxhash-3.8.1-cp313-cp313-musllinux_1_2_armv7l.whl", hash = "sha256:3a800912a2e5e975d4128969d645c4a2a80aa886ccd6c9b1c6f44529e327e8cf", size = 269137, upload-time = "2026-07-06T10:45:39.954Z" },
+    { url = "https://files.pythonhosted.org/packages/de/d3/e963a8a46f900a137d91b02144d8ea07a8f812971b138204a3b2f8b8e55c/xxhash-3.8.1-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:0fe37f72a207223d22a4eddc3149d4298993385aa9daef25c039246ca5a309f3", size = 225068, upload-time = "2026-07-06T10:45:41.718Z" },
+    { url = "https://files.pythonhosted.org/packages/aa/80/9d181dbcde4b0fe48375f48833a5832d4b8cd2b349b15110c92ee472d874/xxhash-3.8.1-cp313-cp313-musllinux_1_2_ppc64le.whl", hash = "sha256:5db43f249b4be9f99ef4b967863f37094fb40e67effafb78ba4f0356b6396104", size = 240874, upload-time = "2026-07-06T10:45:43.414Z" },
+    { url = "https://files.pythonhosted.org/packages/39/15/ce3ab5a1cd27ead25a5196e55a7284220f6ad6e316da494ffd900b2b600f/xxhash-3.8.1-cp313-cp313-musllinux_1_2_riscv64.whl", hash = "sha256:c4ed42965c2cd9081f011be22f69d0e65d3b6165fe7734072fd0c232840bbd4e", size = 300702, upload-time = "2026-07-06T10:45:45.135Z" },
+    { url = "https://files.pythonhosted.org/packages/96/c0/2281a8ab5f2a62dbf57a23c58a01ccc1d98abf40f71193c8a81f59e759b5/xxhash-3.8.1-cp313-cp313-musllinux_1_2_s390x.whl", hash = "sha256:3557bec8fcb11738a8920eeb68974bc76b75262f6947998d3147954ce0a4b893", size = 443351, upload-time = "2026-07-06T10:45:47.188Z" },
+    { url = "https://files.pythonhosted.org/packages/81/2e/071a58c1a53a52d4f7a3aa0987be0c396dffd40da8204805fe1b130a81f4/xxhash-3.8.1-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:00de40f3b42240db23a82a5c682b55d7263d84a26a953240c1aee463409660e3", size = 217396, upload-time = "2026-07-06T10:45:48.925Z" },
+    { url = "https://files.pythonhosted.org/packages/68/44/36ab58134badd9d3433fc7b53c4ca8d113d8e807782885628640f8297a4d/xxhash-3.8.1-cp313-cp313-win32.whl", hash = "sha256:b5196cc2574cfec572a5f3fb7cfa5ade27305ae3d06516a082132441aff4c83a", size = 31974, upload-time = "2026-07-06T10:45:50.591Z" },
+    { url = "https://files.pythonhosted.org/packages/96/2a/2a0b84798448e766f7b89ceed073cb0cb5a43fc9ebbacbdea74a38de18e3/xxhash-3.8.1-cp313-cp313-win_amd64.whl", hash = "sha256:538f5f865df6cd8c32dd63158a0e5b4f5dd08d732a7da8b7228a5a0776c8ce55", size = 32739, upload-time = "2026-07-06T10:45:52.221Z" },
+    { url = "https://files.pythonhosted.org/packages/d4/60/bb51dbf7c363ff88a7cbd50b7959718219577ef44d7cf255929ffc4a2194/xxhash-3.8.1-cp313-cp313-win_arm64.whl", hash = "sha256:a6617f30641ba0d8baa1635fbefb1dffc5165ec36d26921bd5cee13497cd937a", size = 29239, upload-time = "2026-07-06T10:45:53.714Z" },
+    { url = "https://files.pythonhosted.org/packages/56/d3/827ca123c2ee5443a6aaed3c5dd199237dc2f010e2bebd7ec09ef36f3a5f/xxhash-3.8.1-cp313-cp313t-macosx_10_13_x86_64.whl", hash = "sha256:bfcd82852c62a60e314670a9602de354c4460f8adad916e2e42a20860c7870bc", size = 34964, upload-time = "2026-07-06T10:45:55.535Z" },
+    { url = "https://files.pythonhosted.org/packages/05/67/67ae2a3ccdeb8b8ef025d35aee9edd1d26c3abe5051d47da9286232afbf8/xxhash-3.8.1-cp313-cp313t-macosx_11_0_arm64.whl", hash = "sha256:08ea2081f5e88615fec8622a9f87fbe21b8ea58d88cfc02163ca11026ee62a92", size = 32697, upload-time = "2026-07-06T10:45:57.288Z" },
+    { url = "https://files.pythonhosted.org/packages/38/5a/3d3994346e1f45493679cb5c1ffc2bf454e410e9d1e8a662d253becee91e/xxhash-3.8.1-cp313-cp313t-manylinux1_i686.manylinux_2_28_i686.manylinux_2_5_i686.whl", hash = "sha256:2e32855b6f9e5b18f449e59d45e3d5778bdeb660632ef2693cca267a11246c75", size = 225954, upload-time = "2026-07-06T10:45:58.897Z" },
+    { url = "https://files.pythonhosted.org/packages/3f/2c/53169270309b7cd8e05504e07fe123bac053b89d00ac63617faacf0a2ec0/xxhash-3.8.1-cp313-cp313t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:a6e088bd7870775624256a0d84c2a6714afd223b2eeb56b0ca58398e52a32fda", size = 249776, upload-time = "2026-07-06T10:46:00.977Z" },
+    { url = "https://files.pythonhosted.org/packages/70/e0/5c551d8d592f944506f7c5185e210255c15e672a3c6008c156a1bd9b775e/xxhash-3.8.1-cp313-cp313t-manylinux2014_armv7l.manylinux_2_17_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:72eb5ae575cc7ae2b23f6f8064a8b10f638c7149819ae9cc6d20ebd4d37a1629", size = 274776, upload-time = "2026-07-06T10:46:02.869Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/2a/d3a762270cee2d7bcd0e25e28c623e5f3f5c0dc637b66e3e47dd5b0bb3f0/xxhash-3.8.1-cp313-cp313t-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:d0b48cdf690a64cedf7258c3dc9506cc41fc86edd7739c40e3098952265dc068", size = 252056, upload-time = "2026-07-06T10:46:04.688Z" },
+    { url = "https://files.pythonhosted.org/packages/c1/8f/b78e4373b2cb6d1c42af60ea2d7e9146ad0710b239ac7f706d5d31d5bb98/xxhash-3.8.1-cp313-cp313t-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:fb9e256a357dfcede7818c6d34e70db2d6b664394803d1de4b6984d2de76c0f1", size = 482108, upload-time = "2026-07-06T10:46:06.498Z" },
+    { url = "https://files.pythonhosted.org/packages/e6/0d/642d923336ea61a15f8ce64fc7e078729e6e06c3a026e517fa79b2c23b7a/xxhash-3.8.1-cp313-cp313t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:51f71a6e2ad071e70c937e41fcb6c19f82c3f9f49831eba850ed4a106ffbb647", size = 226739, upload-time = "2026-07-06T10:46:08.598Z" },
+    { url = "https://files.pythonhosted.org/packages/a6/0a/a37d6da6427d45a8d23e3ee3a0ca9c9d4a90364849c6637fe2963a755f9b/xxhash-3.8.1-cp313-cp313t-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:e4a6443968c4e8dc69967e12776776a5952c119cc1bd94168ad1c5ad667c2be1", size = 319658, upload-time = "2026-07-06T10:46:10.504Z" },
+    { url = "https://files.pythonhosted.org/packages/4a/51/ebbd40da8a3f1bc53b4b7a9a87f8e28bd95c5f21bc14b8a57860cf367d1b/xxhash-3.8.1-cp313-cp313t-musllinux_1_2_aarch64.whl", hash = "sha256:714503083a1f2065c9ad15340dd49ac8a8e948a505a705ffa1750cb951519113", size = 246059, upload-time = "2026-07-06T10:46:12.634Z" },
+    { url = "https://files.pythonhosted.org/packages/24/4c/d9014030147e1f0bb26e7da47aa240dd9ec61c763c573e558111d869f8e1/xxhash-3.8.1-cp313-cp313t-musllinux_1_2_armv7l.whl", hash = "sha256:77f74e45a1e5574bbbf80181c8027b3a4c65c2248fffbd557bd596fff13102f9", size = 275535, upload-time = "2026-07-06T10:46:14.614Z" },
+    { url = "https://files.pythonhosted.org/packages/84/86/caee2db41fadcd5a25aa4323213f9afec5a8586d4e419241e3d659362bd7/xxhash-3.8.1-cp313-cp313t-musllinux_1_2_i686.whl", hash = "sha256:4e0e1b0fb0259c1b75d1251ac0bb4d7ab675d36f7a6bf4ba6aa630dae94f9ffa", size = 231292, upload-time = "2026-07-06T10:46:16.452Z" },
+    { url = "https://files.pythonhosted.org/packages/0b/60/f52f08bcdc904c4514ea5c25caa19e9f3214144434a6ff96dc82dc1cbddd/xxhash-3.8.1-cp313-cp313t-musllinux_1_2_ppc64le.whl", hash = "sha256:10e4393ec33633c2f05ad01869e546ad080b1a18f2650503731f153774608b31", size = 250490, upload-time = "2026-07-06T10:46:18.318Z" },
+    { url = "https://files.pythonhosted.org/packages/24/a0/94dc7ae310838f250669c6ad7168e6d6fca17d49dac1053f06dc232c4a56/xxhash-3.8.1-cp313-cp313t-musllinux_1_2_riscv64.whl", hash = "sha256:b3ba794c3d885803db6c3116686923f1ec13bc86e621e169a375282b63ea1cc6", size = 309861, upload-time = "2026-07-06T10:46:20.503Z" },
+    { url = "https://files.pythonhosted.org/packages/8b/f9/adeead7d0eb28cdfc2832544ea639ffbc6749ccde47a8e228d667459182e/xxhash-3.8.1-cp313-cp313t-musllinux_1_2_s390x.whl", hash = "sha256:57189a69c0891e4818853feaa521c972d22c880a001453addea015f48e3c3398", size = 448739, upload-time = "2026-07-06T10:46:22.79Z" },
+    { url = "https://files.pythonhosted.org/packages/04/a4/22ec0e07db57d901c9298ae98aa3cf2be45bafded6f07c13131e85b89032/xxhash-3.8.1-cp313-cp313t-musllinux_1_2_x86_64.whl", hash = "sha256:d59e71153fe9ff85648d00e18649b07e9b22c797291abb7e27274fa06df8b838", size = 223657, upload-time = "2026-07-06T10:46:24.831Z" },
+    { url = "https://files.pythonhosted.org/packages/94/32/8a9531f37b59e5a013003db7cb7414baf4ce7e0e1268e0d5947cd3d6a2df/xxhash-3.8.1-cp313-cp313t-win32.whl", hash = "sha256:5b96f0024e9840f449bd91b2d005c921a4b666055a0d1b6492463799f32aae22", size = 32377, upload-time = "2026-07-06T10:46:26.86Z" },
+    { url = "https://files.pythonhosted.org/packages/e7/ab/2ca45fd7f671de5f81fc297ef1c95080b40c86ec6be0cc6034b8f7707ac8/xxhash-3.8.1-cp313-cp313t-win_amd64.whl", hash = "sha256:37d5a56c36dcc0b9a87b814cd992598d33863ff683749de6c86081f278d5e629", size = 33274, upload-time = "2026-07-06T10:46:28.39Z" },
+    { url = "https://files.pythonhosted.org/packages/5a/54/20d7163463ddb6438b73a427d1655a77a502cf9b9b0c3ada3599629d9c0a/xxhash-3.8.1-cp313-cp313t-win_arm64.whl", hash = "sha256:6696c8752aded28ff3b16f33ef28ce28fb5d209b80c206746f943199fcf5fd65", size = 29375, upload-time = "2026-07-06T10:46:29.962Z" },
+    { url = "https://files.pythonhosted.org/packages/c2/8b/df2ba04f22a6cd6b39f96a6577329a8471a55c90ef8d8e2f7c102363613f/xxhash-3.8.1-cp314-cp314-android_24_arm64_v8a.whl", hash = "sha256:9db455cb649dcfe4504d6d68a6d83a7315a99a3ca59871dc3ff840671f99adba", size = 38430, upload-time = "2026-07-06T10:46:31.496Z" },
+    { url = "https://files.pythonhosted.org/packages/b2/4f/6a059e8ad3ca8deedc91dfe335b211204900895152212c03ebbe721de68b/xxhash-3.8.1-cp314-cp314-android_24_x86_64.whl", hash = "sha256:affb37f152e55b5e4494bb9d0107f7bb08515c6704fbed82d9f61214d74adc17", size = 36558, upload-time = "2026-07-06T10:46:33.078Z" },
+    { url = "https://files.pythonhosted.org/packages/cb/95/40be178205acce092ae418feb20ac737b32a02c7b864926ed0717354c9f8/xxhash-3.8.1-cp314-cp314-ios_13_0_arm64_iphoneos.whl", hash = "sha256:460261045936975193bfd20549a0de1cd52a33b405cbb972f0d80940c42266cd", size = 31181, upload-time = "2026-07-06T10:46:34.793Z" },
+    { url = "https://files.pythonhosted.org/packages/3f/89/2da4dbf051bafa156c0e3f12012db2b0ac3b84ff37ca1f021f6bfffcdfbb/xxhash-3.8.1-cp314-cp314-ios_13_0_arm64_iphonesimulator.whl", hash = "sha256:38c887aedb696ef8bca19983206d270848558cfae4a91afa6a2fb05dde58ffc5", size = 32192, upload-time = "2026-07-06T10:46:36.393Z" },
+    { url = "https://files.pythonhosted.org/packages/7c/4e/e000bbae3566bc8e0be771a8a0f294aa99075e3f0bc4ef43922ebffdebc8/xxhash-3.8.1-cp314-cp314-ios_13_0_x86_64_iphonesimulator.whl", hash = "sha256:594131ce1aad18db3689781f806db1b065cdaa04f4df36b4c038d2013aefd0bf", size = 34691, upload-time = "2026-07-06T10:46:38.1Z" },
+    { url = "https://files.pythonhosted.org/packages/b4/4a/ea954aacc7d1c8711880ac2b55da94429a9b4296b151c4fc0966549ca1ee/xxhash-3.8.1-cp314-cp314-macosx_10_15_x86_64.whl", hash = "sha256:78c794b643d214f1522e7a288bcf5a2de120d26cd170516749a4009dc92722c9", size = 34807, upload-time = "2026-07-06T10:46:39.647Z" },
+    { url = "https://files.pythonhosted.org/packages/ca/29/df598e738ff37558ac627264deb2e560902d9bf7f46d3bd5175c9eee593e/xxhash-3.8.1-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:af0c9fedc4a2c24e8664953882fe8185f3790b8338c9c700f76f5ad660817711", size = 32410, upload-time = "2026-07-06T10:46:41.359Z" },
+    { url = "https://files.pythonhosted.org/packages/59/9c/81ab40e7d33ada0b3df5d1bc884894d15dbf4f805cd645b685e4606bb8e0/xxhash-3.8.1-cp314-cp314-manylinux1_i686.manylinux_2_28_i686.manylinux_2_5_i686.whl", hash = "sha256:115772daeb71b2f3b9381177017f53e6cf3f3439c840737fdabd21aba6e54920", size = 220564, upload-time = "2026-07-06T10:46:43.463Z" },
+    { url = "https://files.pythonhosted.org/packages/fd/6f/62ae6f5c8606320a0e2a41c2dc8c6d91cc5d63d0f84dd9582e9543779dd8/xxhash-3.8.1-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:000435984a0469b0f822fe76f35bddea0f96a4d6521b3339a60a6428cdee1edc", size = 241462, upload-time = "2026-07-06T10:46:45.509Z" },
+    { url = "https://files.pythonhosted.org/packages/15/a1/9c3a0ec6cb524396f551eddd102a76690a795494eb9784fc67542b0daa37/xxhash-3.8.1-cp314-cp314-manylinux2014_armv7l.manylinux_2_17_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:2f1c68394818e0595569c2ff3cbc1e6d5a36a434e796f5c526b987b80c8a8c62", size = 264491, upload-time = "2026-07-06T10:46:47.655Z" },
+    { url = "https://files.pythonhosted.org/packages/64/f2/700a4674e4308eb59d2fdb973977e82eae231bea5044753fee5c9eec0e0c/xxhash-3.8.1-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:46b39976d008e2a845758650f0ff7136bca004f40da0c8798bd37ac37860154f", size = 242905, upload-time = "2026-07-06T10:46:49.857Z" },
+    { url = "https://files.pythonhosted.org/packages/f3/8a/72d9874375c8d4cbc64a8cd1d659d5695a8765c3db82efa82dc5bd9f14d0/xxhash-3.8.1-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:d5006c65ec507a333479e76e00e2c368781f16c24ededa764763956b32a0e93e", size = 473873, upload-time = "2026-07-06T10:46:51.953Z" },
+    { url = "https://files.pythonhosted.org/packages/03/f0/6db07590ed7e0a77f186ef0bcea8d52553bf1ba57833e09467a2411f0f2d/xxhash-3.8.1-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:c31a2649bcf1fe97cf11c79848d761df33ac46b3896942d31b640557b486ff6b", size = 220765, upload-time = "2026-07-06T10:46:55.41Z" },
+    { url = "https://files.pythonhosted.org/packages/8f/10/00d12d8b8beabbf49a8bbc626fb9f40445145a8887eb41a6acfb69149ac4/xxhash-3.8.1-cp314-cp314-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:8f759eed402448c2bdbb492e4fba1f20668ffe29688605ea61f0f67f9e4e386d", size = 310478, upload-time = "2026-07-06T10:46:57.729Z" },
+    { url = "https://files.pythonhosted.org/packages/1f/f9/12a82394eefb0f185d15a7f7b9f627c61c475a72dd83718436a5b84b42ac/xxhash-3.8.1-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:7b5f97ecfede10d5b2870383620e2d25c8561e217c7bf9081073802b54248d2b", size = 238393, upload-time = "2026-07-06T10:46:59.87Z" },
+    { url = "https://files.pythonhosted.org/packages/20/f3/53f963e320b9ce678337aa7273f39ce692ded8b99e3d22a866ec722159ab/xxhash-3.8.1-cp314-cp314-musllinux_1_2_armv7l.whl", hash = "sha256:1da930bbcac3e8fbe2191850e2abb57977a99348c12c4b385e1058ac1b0a9ecc", size = 268704, upload-time = "2026-07-06T10:47:01.806Z" },
+    { url = "https://files.pythonhosted.org/packages/0a/50/5b5badbd87c82d9f9b5f58ac74a3f29ef08f6fc387b324b8fd482450b862/xxhash-3.8.1-cp314-cp314-musllinux_1_2_i686.whl", hash = "sha256:747476436f6891b9773374ce8d48edcc8b12cb5b61b67c6fb6289633747d088f", size = 225015, upload-time = "2026-07-06T10:47:03.784Z" },
+    { url = "https://files.pythonhosted.org/packages/30/93/3ca68265afe7b4e69435e08a7b6a1d9d0f2a071e889da1f8041ed00fe878/xxhash-3.8.1-cp314-cp314-musllinux_1_2_ppc64le.whl", hash = "sha256:4ef09bbc2519a93cd0f95f2ceb5f7b85919dffea643278e02362bf40e3c4bed1", size = 240951, upload-time = "2026-07-06T10:47:05.816Z" },
+    { url = "https://files.pythonhosted.org/packages/dd/a6/27e19670c40f46b5e76e11f2f4713d21054804568425d870670e757172ad/xxhash-3.8.1-cp314-cp314-musllinux_1_2_riscv64.whl", hash = "sha256:a5eed9d41995a83f3332b4e3396abb7f433cac584222bd7e305b606d8353861e", size = 300751, upload-time = "2026-07-06T10:47:07.95Z" },
+    { url = "https://files.pythonhosted.org/packages/bc/fb/b33e27689959fe7ed2ae0b830af41560d65213943983afa9db3a8d481bce/xxhash-3.8.1-cp314-cp314-musllinux_1_2_s390x.whl", hash = "sha256:53f3ed9118397074ff63a79b66b7fec1c84c782eecde35c5bc94e420a971c231", size = 443480, upload-time = "2026-07-06T10:47:10Z" },
+    { url = "https://files.pythonhosted.org/packages/26/60/0e0d973be5fe280753ef02fbc89349492ad6e903bf1dcb870b668f94b662/xxhash-3.8.1-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:d247b34bf433c92b41689318fd25d246313cab2275a6a47e2efac178b80d6efe", size = 217657, upload-time = "2026-07-06T10:47:12.196Z" },
+    { url = "https://files.pythonhosted.org/packages/ad/68/c9e3ecef4a9a417d464cb5bd200aa12f73192dee677901b9e08e0ad0d1bb/xxhash-3.8.1-cp314-cp314-win32.whl", hash = "sha256:d58ce8b6cfa9c4d2f230557f69caf7c06369e318015d0b19485095bc2c5963ab", size = 32690, upload-time = "2026-07-06T10:47:14.204Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/99/e9e44588c0b62837bbec5ba7927816de0afa03406b1a0b6c7a7e1d1a30a0/xxhash-3.8.1-cp314-cp314-win_amd64.whl", hash = "sha256:6cee733fe4ccb1737e0997135283c82341e5cfa9cf214b165f9087fb663aaf4f", size = 33460, upload-time = "2026-07-06T10:47:16.021Z" },
+    { url = "https://files.pythonhosted.org/packages/45/2b/64f36d86380b3657ad9031967ab814f3ef31307174650853f69c18932ebc/xxhash-3.8.1-cp314-cp314-win_arm64.whl", hash = "sha256:58346024d47e84f7d8b3e7f5d6faa1d58acbbe49a8771497872059f58c1d8ea5", size = 30092, upload-time = "2026-07-06T10:47:17.81Z" },
+    { url = "https://files.pythonhosted.org/packages/92/cb/18b64bff88c58a0ca209dc533e63cf02d7ae5aa6b1b9a9fd14e81b5dbd60/xxhash-3.8.1-cp314-cp314t-macosx_10_15_x86_64.whl", hash = "sha256:01cab782f8a0a05ecad2c63d7ef10f7ab475f660e0d6419d069418c14d88de7c", size = 35024, upload-time = "2026-07-06T10:47:19.821Z" },
+    { url = "https://files.pythonhosted.org/packages/af/1d/72d8a70520e5dcddb472ea0486d299da3240745a10658290cd7b5690ede2/xxhash-3.8.1-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:717b12fdc51819833704e85e6926d76981ffa3f780ef92e33ebb8b26d46bb230", size = 32697, upload-time = "2026-07-06T10:47:21.649Z" },
+    { url = "https://files.pythonhosted.org/packages/9c/b8/e041f555903c56db3d0a731b3d72a6575d75e0ed868b1bd2e5176111ca44/xxhash-3.8.1-cp314-cp314t-manylinux1_i686.manylinux_2_28_i686.manylinux_2_5_i686.whl", hash = "sha256:ec55d80e9b8a519d742669e0b49e8ce9e6747be42bf3c138158b6543a9c8e489", size = 226044, upload-time = "2026-07-06T10:47:23.612Z" },
+    { url = "https://files.pythonhosted.org/packages/3a/7e/5cdcf06bf6ec4b5d2ac073feb23432ec1d603fd438864cbd2c09c7cb45e1/xxhash-3.8.1-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:98d8ac1129b4dd39098cffed94d1284aceb61c3aa396757ccc736ac392e4cee5", size = 249899, upload-time = "2026-07-06T10:47:25.812Z" },
+    { url = "https://files.pythonhosted.org/packages/c0/c0/eb7e059cb5e1dba11fd30d2fdf882f56e5a417a3eaa43669d43623767f45/xxhash-3.8.1-cp314-cp314t-manylinux2014_armv7l.manylinux_2_17_armv7l.manylinux_2_31_armv7l.whl", hash = "sha256:3bc0fa90830df1e1277f33cc6e55de9990b83c0319fd8c7412866cfde38b025e", size = 274892, upload-time = "2026-07-06T10:47:27.931Z" },
+    { url = "https://files.pythonhosted.org/packages/66/74/a600aaf7cd39957fd1510adeedb1749c1e7eb82bd632a1153d9c664c3135/xxhash-3.8.1-cp314-cp314t-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:c73b6f652f0745425aa6378319c331293b5341756262e9408ed3d45f183375e6", size = 252243, upload-time = "2026-07-06T10:47:30.288Z" },
+    { url = "https://files.pythonhosted.org/packages/ad/04/78d88fa75a6763e5d09bf1b947a392a27988903381b219006f92f3c68fc8/xxhash-3.8.1-cp314-cp314t-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:f6114692261eff4266386cdec0f7d87eee24e317ab397c218b7ae6a76b4c6339", size = 482191, upload-time = "2026-07-06T10:47:32.45Z" },
+    { url = "https://files.pythonhosted.org/packages/7f/06/07a8aea1108d682de8791ce608cdf367d75ff4e7e57cd3c154bdc6f47b23/xxhash-3.8.1-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:4df57c0b161ec1b3ed0526a67b0db0914b557e86ee8aae51887aec941b261542", size = 226877, upload-time = "2026-07-06T10:47:34.705Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/b5/86bade5618a524d2c06c4041aa2fe8e5749ce16e88afba60d67c1684a21f/xxhash-3.8.1-cp314-cp314t-manylinux_2_31_riscv64.manylinux_2_39_riscv64.whl", hash = "sha256:9043877a917be88ccf230aa5667c1bd059bce80f4c2727e4defa1b29b7f48b08", size = 319794, upload-time = "2026-07-06T10:47:37.08Z" },
+    { url = "https://files.pythonhosted.org/packages/23/69/9b1a2b89b1621bb740fbcb7beb512f60f99480c1bdc680c0c90e1f56ff75/xxhash-3.8.1-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:559e3cabe522231909f9de98ef06929edbd53782046bd21aae0c72db6f2a0775", size = 246202, upload-time = "2026-07-06T10:47:39.676Z" },
+    { url = "https://files.pythonhosted.org/packages/08/ea/662ed6cb49f1d34078b6a3a3e0f3d29ff93fd7b5a03c0bc9ecfd9b2159c3/xxhash-3.8.1-cp314-cp314t-musllinux_1_2_armv7l.whl", hash = "sha256:264710bd335016f303763ce1275c6486df30bb57c2245c91b224c983d7ac39b8", size = 275628, upload-time = "2026-07-06T10:47:41.99Z" },
+    { url = "https://files.pythonhosted.org/packages/13/f5/49fc9e4c6728a5a3bd8fe639199d2fa67609b3a84f938aff6e8568dd3e4f/xxhash-3.8.1-cp314-cp314t-musllinux_1_2_i686.whl", hash = "sha256:e14800b9b10bb39d7a60ad4a310e403164d7b8988a27ae933d4e40618a44088e", size = 231390, upload-time = "2026-07-06T10:47:44.233Z" },
+    { url = "https://files.pythonhosted.org/packages/64/9d/3acaf8f599c0e0b30e910a3a11ba32929da53c86dc73c7c55fe6a010b4e9/xxhash-3.8.1-cp314-cp314t-musllinux_1_2_ppc64le.whl", hash = "sha256:ea6a3e734b0fd41b82784a400be946821900daebe610c050a5e0760838a34f99", size = 250600, upload-time = "2026-07-06T10:47:47.611Z" },
+    { url = "https://files.pythonhosted.org/packages/23/64/8acab4c5ec60dbe664b5b9858fd44c2413b07e535b09556a0a5022e78aa6/xxhash-3.8.1-cp314-cp314t-musllinux_1_2_riscv64.whl", hash = "sha256:cf399fac542a1c7a4734a435b93df2c55e858c7d31abf6c1bdf46f9ae67fbfd0", size = 310032, upload-time = "2026-07-06T10:47:49.88Z" },
+    { url = "https://files.pythonhosted.org/packages/56/47/a0288d7329b1fe63e2734a32d19d444a96ae2b4810f545bc61e561224917/xxhash-3.8.1-cp314-cp314t-musllinux_1_2_s390x.whl", hash = "sha256:44c89d915a75c11d2547eaee9098fcd80398987c4bff2974a0497a925bf92c07", size = 448882, upload-time = "2026-07-06T10:47:52.631Z" },
+    { url = "https://files.pythonhosted.org/packages/01/e7/3071dfd3beb5c38204ce1cf56bf7749fce08de900fa92714b81d1d8ca1f2/xxhash-3.8.1-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:358650d5bda9c635da699c53adf4e8134af492ecc79c960f917eebf088bb6799", size = 223728, upload-time = "2026-07-06T10:47:55.093Z" },
+    { url = "https://files.pythonhosted.org/packages/12/11/b99949f0ba2b07e9f9ffe83b9c86faa685f9080725dc21a916a607313be5/xxhash-3.8.1-cp314-cp314t-win32.whl", hash = "sha256:c240939e963653054fc7e4a17c382829cda4aa88a7daf0af841715dbded1b497", size = 33150, upload-time = "2026-07-06T10:47:57.274Z" },
+    { url = "https://files.pythonhosted.org/packages/54/1c/09703eb341f8416e74e58d6c6732d4b5c46de59c942363203cb237cc95b0/xxhash-3.8.1-cp314-cp314t-win_amd64.whl", hash = "sha256:7258ee276e8772599bc19e14b36f6260306e21b637190cd7cb489a2449d48684", size = 34005, upload-time = "2026-07-06T10:47:59.434Z" },
+    { url = "https://files.pythonhosted.org/packages/d6/f9/6ed7251bb6a8af10ac73b1821c60583d2826e5b2064e45a979c935287c98/xxhash-3.8.1-cp314-cp314t-win_arm64.whl", hash = "sha256:8f454166c2ffed45636c8d501741e649851ba2f346c4eb73a64c07ac00428f20", size = 30239, upload-time = "2026-07-06T10:48:01.874Z" },
+]
+
+[[package]]
+name = "zstandard"
+version = "0.25.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/fd/aa/3e0508d5a5dd96529cdc5a97011299056e14c6505b678fd58938792794b1/zstandard-0.25.0.tar.gz", hash = "sha256:7713e1179d162cf5c7906da876ec2ccb9c3a9dcbdffef0cc7f70c3667a205f0b", size = 711513, upload-time = "2025-09-14T22:15:54.002Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/82/fc/f26eb6ef91ae723a03e16eddb198abcfce2bc5a42e224d44cc8b6765e57e/zstandard-0.25.0-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:7b3c3a3ab9daa3eed242d6ecceead93aebbb8f5f84318d82cee643e019c4b73b", size = 795738, upload-time = "2025-09-14T22:16:56.237Z" },
+    { url = "https://files.pythonhosted.org/packages/aa/1c/d920d64b22f8dd028a8b90e2d756e431a5d86194caa78e3819c7bf53b4b3/zstandard-0.25.0-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:913cbd31a400febff93b564a23e17c3ed2d56c064006f54efec210d586171c00", size = 640436, upload-time = "2025-09-14T22:16:57.774Z" },
+    { url = "https://files.pythonhosted.org/packages/53/6c/288c3f0bd9fcfe9ca41e2c2fbfd17b2097f6af57b62a81161941f09afa76/zstandard-0.25.0-cp312-cp312-manylinux2010_i686.manylinux2014_i686.manylinux_2_12_i686.manylinux_2_17_i686.whl", hash = "sha256:011d388c76b11a0c165374ce660ce2c8efa8e5d87f34996aa80f9c0816698b64", size = 5343019, upload-time = "2025-09-14T22:16:59.302Z" },
+    { url = "https://files.pythonhosted.org/packages/1e/15/efef5a2f204a64bdb5571e6161d49f7ef0fffdbca953a615efbec045f60f/zstandard-0.25.0-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:6dffecc361d079bb48d7caef5d673c88c8988d3d33fb74ab95b7ee6da42652ea", size = 5063012, upload-time = "2025-09-14T22:17:01.156Z" },
+    { url = "https://files.pythonhosted.org/packages/b7/37/a6ce629ffdb43959e92e87ebdaeebb5ac81c944b6a75c9c47e300f85abdf/zstandard-0.25.0-cp312-cp312-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:7149623bba7fdf7e7f24312953bcf73cae103db8cae49f8154dd1eadc8a29ecb", size = 5394148, upload-time = "2025-09-14T22:17:03.091Z" },
+    { url = "https://files.pythonhosted.org/packages/e3/79/2bf870b3abeb5c070fe2d670a5a8d1057a8270f125ef7676d29ea900f496/zstandard-0.25.0-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:6a573a35693e03cf1d67799fd01b50ff578515a8aeadd4595d2a7fa9f3ec002a", size = 5451652, upload-time = "2025-09-14T22:17:04.979Z" },
+    { url = "https://files.pythonhosted.org/packages/53/60/7be26e610767316c028a2cbedb9a3beabdbe33e2182c373f71a1c0b88f36/zstandard-0.25.0-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:5a56ba0db2d244117ed744dfa8f6f5b366e14148e00de44723413b2f3938a902", size = 5546993, upload-time = "2025-09-14T22:17:06.781Z" },
+    { url = "https://files.pythonhosted.org/packages/85/c7/3483ad9ff0662623f3648479b0380d2de5510abf00990468c286c6b04017/zstandard-0.25.0-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:10ef2a79ab8e2974e2075fb984e5b9806c64134810fac21576f0668e7ea19f8f", size = 5046806, upload-time = "2025-09-14T22:17:08.415Z" },
+    { url = "https://files.pythonhosted.org/packages/08/b3/206883dd25b8d1591a1caa44b54c2aad84badccf2f1de9e2d60a446f9a25/zstandard-0.25.0-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:aaf21ba8fb76d102b696781bddaa0954b782536446083ae3fdaa6f16b25a1c4b", size = 5576659, upload-time = "2025-09-14T22:17:10.164Z" },
+    { url = "https://files.pythonhosted.org/packages/9d/31/76c0779101453e6c117b0ff22565865c54f48f8bd807df2b00c2c404b8e0/zstandard-0.25.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:1869da9571d5e94a85a5e8d57e4e8807b175c9e4a6294e3b66fa4efb074d90f6", size = 4953933, upload-time = "2025-09-14T22:17:11.857Z" },
+    { url = "https://files.pythonhosted.org/packages/18/e1/97680c664a1bf9a247a280a053d98e251424af51f1b196c6d52f117c9720/zstandard-0.25.0-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:809c5bcb2c67cd0ed81e9229d227d4ca28f82d0f778fc5fea624a9def3963f91", size = 5268008, upload-time = "2025-09-14T22:17:13.627Z" },
+    { url = "https://files.pythonhosted.org/packages/1e/73/316e4010de585ac798e154e88fd81bb16afc5c5cb1a72eeb16dd37e8024a/zstandard-0.25.0-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:f27662e4f7dbf9f9c12391cb37b4c4c3cb90ffbd3b1fb9284dadbbb8935fa708", size = 5433517, upload-time = "2025-09-14T22:17:16.103Z" },
+    { url = "https://files.pythonhosted.org/packages/5b/60/dd0f8cfa8129c5a0ce3ea6b7f70be5b33d2618013a161e1ff26c2b39787c/zstandard-0.25.0-cp312-cp312-musllinux_1_2_s390x.whl", hash = "sha256:99c0c846e6e61718715a3c9437ccc625de26593fea60189567f0118dc9db7512", size = 5814292, upload-time = "2025-09-14T22:17:17.827Z" },
+    { url = "https://files.pythonhosted.org/packages/fc/5f/75aafd4b9d11b5407b641b8e41a57864097663699f23e9ad4dbb91dc6bfe/zstandard-0.25.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:474d2596a2dbc241a556e965fb76002c1ce655445e4e3bf38e5477d413165ffa", size = 5360237, upload-time = "2025-09-14T22:17:19.954Z" },
+    { url = "https://files.pythonhosted.org/packages/ff/8d/0309daffea4fcac7981021dbf21cdb2e3427a9e76bafbcdbdf5392ff99a4/zstandard-0.25.0-cp312-cp312-win32.whl", hash = "sha256:23ebc8f17a03133b4426bcc04aabd68f8236eb78c3760f12783385171b0fd8bd", size = 436922, upload-time = "2025-09-14T22:17:24.398Z" },
+    { url = "https://files.pythonhosted.org/packages/79/3b/fa54d9015f945330510cb5d0b0501e8253c127cca7ebe8ba46a965df18c5/zstandard-0.25.0-cp312-cp312-win_amd64.whl", hash = "sha256:ffef5a74088f1e09947aecf91011136665152e0b4b359c42be3373897fb39b01", size = 506276, upload-time = "2025-09-14T22:17:21.429Z" },
+    { url = "https://files.pythonhosted.org/packages/ea/6b/8b51697e5319b1f9ac71087b0af9a40d8a6288ff8025c36486e0c12abcc4/zstandard-0.25.0-cp312-cp312-win_arm64.whl", hash = "sha256:181eb40e0b6a29b3cd2849f825e0fa34397f649170673d385f3598ae17cca2e9", size = 462679, upload-time = "2025-09-14T22:17:23.147Z" },
+    { url = "https://files.pythonhosted.org/packages/35/0b/8df9c4ad06af91d39e94fa96cc010a24ac4ef1378d3efab9223cc8593d40/zstandard-0.25.0-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:ec996f12524f88e151c339688c3897194821d7f03081ab35d31d1e12ec975e94", size = 795735, upload-time = "2025-09-14T22:17:26.042Z" },
+    { url = "https://files.pythonhosted.org/packages/3f/06/9ae96a3e5dcfd119377ba33d4c42a7d89da1efabd5cb3e366b156c45ff4d/zstandard-0.25.0-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:a1a4ae2dec3993a32247995bdfe367fc3266da832d82f8438c8570f989753de1", size = 640440, upload-time = "2025-09-14T22:17:27.366Z" },
+    { url = "https://files.pythonhosted.org/packages/d9/14/933d27204c2bd404229c69f445862454dcc101cd69ef8c6068f15aaec12c/zstandard-0.25.0-cp313-cp313-manylinux2010_i686.manylinux2014_i686.manylinux_2_12_i686.manylinux_2_17_i686.whl", hash = "sha256:e96594a5537722fdfb79951672a2a63aec5ebfb823e7560586f7484819f2a08f", size = 5343070, upload-time = "2025-09-14T22:17:28.896Z" },
+    { url = "https://files.pythonhosted.org/packages/6d/db/ddb11011826ed7db9d0e485d13df79b58586bfdec56e5c84a928a9a78c1c/zstandard-0.25.0-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:bfc4e20784722098822e3eee42b8e576b379ed72cca4a7cb856ae733e62192ea", size = 5063001, upload-time = "2025-09-14T22:17:31.044Z" },
+    { url = "https://files.pythonhosted.org/packages/db/00/87466ea3f99599d02a5238498b87bf84a6348290c19571051839ca943777/zstandard-0.25.0-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:457ed498fc58cdc12fc48f7950e02740d4f7ae9493dd4ab2168a47c93c31298e", size = 5394120, upload-time = "2025-09-14T22:17:32.711Z" },
+    { url = "https://files.pythonhosted.org/packages/2b/95/fc5531d9c618a679a20ff6c29e2b3ef1d1f4ad66c5e161ae6ff847d102a9/zstandard-0.25.0-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:fd7a5004eb1980d3cefe26b2685bcb0b17989901a70a1040d1ac86f1d898c551", size = 5451230, upload-time = "2025-09-14T22:17:34.41Z" },
+    { url = "https://files.pythonhosted.org/packages/63/4b/e3678b4e776db00f9f7b2fe58e547e8928ef32727d7a1ff01dea010f3f13/zstandard-0.25.0-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:8e735494da3db08694d26480f1493ad2cf86e99bdd53e8e9771b2752a5c0246a", size = 5547173, upload-time = "2025-09-14T22:17:36.084Z" },
+    { url = "https://files.pythonhosted.org/packages/4e/d5/ba05ed95c6b8ec30bd468dfeab20589f2cf709b5c940483e31d991f2ca58/zstandard-0.25.0-cp313-cp313-musllinux_1_1_aarch64.whl", hash = "sha256:3a39c94ad7866160a4a46d772e43311a743c316942037671beb264e395bdd611", size = 5046736, upload-time = "2025-09-14T22:17:37.891Z" },
+    { url = "https://files.pythonhosted.org/packages/50/d5/870aa06b3a76c73eced65c044b92286a3c4e00554005ff51962deef28e28/zstandard-0.25.0-cp313-cp313-musllinux_1_1_x86_64.whl", hash = "sha256:172de1f06947577d3a3005416977cce6168f2261284c02080e7ad0185faeced3", size = 5576368, upload-time = "2025-09-14T22:17:40.206Z" },
+    { url = "https://files.pythonhosted.org/packages/5d/35/398dc2ffc89d304d59bc12f0fdd931b4ce455bddf7038a0a67733a25f550/zstandard-0.25.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:3c83b0188c852a47cd13ef3bf9209fb0a77fa5374958b8c53aaa699398c6bd7b", size = 4954022, upload-time = "2025-09-14T22:17:41.879Z" },
+    { url = "https://files.pythonhosted.org/packages/9a/5c/36ba1e5507d56d2213202ec2b05e8541734af5f2ce378c5d1ceaf4d88dc4/zstandard-0.25.0-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:1673b7199bbe763365b81a4f3252b8e80f44c9e323fc42940dc8843bfeaf9851", size = 5267889, upload-time = "2025-09-14T22:17:43.577Z" },
+    { url = "https://files.pythonhosted.org/packages/70/e8/2ec6b6fb7358b2ec0113ae202647ca7c0e9d15b61c005ae5225ad0995df5/zstandard-0.25.0-cp313-cp313-musllinux_1_2_ppc64le.whl", hash = "sha256:0be7622c37c183406f3dbf0cba104118eb16a4ea7359eeb5752f0794882fc250", size = 5433952, upload-time = "2025-09-14T22:17:45.271Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/01/b5f4d4dbc59ef193e870495c6f1275f5b2928e01ff5a81fecb22a06e22fb/zstandard-0.25.0-cp313-cp313-musllinux_1_2_s390x.whl", hash = "sha256:5f5e4c2a23ca271c218ac025bd7d635597048b366d6f31f420aaeb715239fc98", size = 5814054, upload-time = "2025-09-14T22:17:47.08Z" },
+    { url = "https://files.pythonhosted.org/packages/b2/e5/fbd822d5c6f427cf158316d012c5a12f233473c2f9c5fe5ab1ae5d21f3d8/zstandard-0.25.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:4f187a0bb61b35119d1926aee039524d1f93aaf38a9916b8c4b78ac8514a0aaf", size = 5360113, upload-time = "2025-09-14T22:17:48.893Z" },
+    { url = "https://files.pythonhosted.org/packages/8e/e0/69a553d2047f9a2c7347caa225bb3a63b6d7704ad74610cb7823baa08ed7/zstandard-0.25.0-cp313-cp313-win32.whl", hash = "sha256:7030defa83eef3e51ff26f0b7bfb229f0204b66fe18e04359ce3474ac33cbc09", size = 436936, upload-time = "2025-09-14T22:17:52.658Z" },
+    { url = "https://files.pythonhosted.org/packages/d9/82/b9c06c870f3bd8767c201f1edbdf9e8dc34be5b0fbc5682c4f80fe948475/zstandard-0.25.0-cp313-cp313-win_amd64.whl", hash = "sha256:1f830a0dac88719af0ae43b8b2d6aef487d437036468ef3c2ea59c51f9d55fd5", size = 506232, upload-time = "2025-09-14T22:17:50.402Z" },
+    { url = "https://files.pythonhosted.org/packages/d4/57/60c3c01243bb81d381c9916e2a6d9e149ab8627c0c7d7abb2d73384b3c0c/zstandard-0.25.0-cp313-cp313-win_arm64.whl", hash = "sha256:85304a43f4d513f5464ceb938aa02c1e78c2943b29f44a750b48b25ac999a049", size = 462671, upload-time = "2025-09-14T22:17:51.533Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/5c/f8923b595b55fe49e30612987ad8bf053aef555c14f05bb659dd5dbe3e8a/zstandard-0.25.0-cp314-cp314-macosx_10_13_x86_64.whl", hash = "sha256:e29f0cf06974c899b2c188ef7f783607dbef36da4c242eb6c82dcd8b512855e3", size = 795887, upload-time = "2025-09-14T22:17:54.198Z" },
+    { url = "https://files.pythonhosted.org/packages/8d/09/d0a2a14fc3439c5f874042dca72a79c70a532090b7ba0003be73fee37ae2/zstandard-0.25.0-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:05df5136bc5a011f33cd25bc9f506e7426c0c9b3f9954f056831ce68f3b6689f", size = 640658, upload-time = "2025-09-14T22:17:55.423Z" },
+    { url = "https://files.pythonhosted.org/packages/5d/7c/8b6b71b1ddd517f68ffb55e10834388d4f793c49c6b83effaaa05785b0b4/zstandard-0.25.0-cp314-cp314-manylinux2010_i686.manylinux_2_12_i686.manylinux_2_28_i686.whl", hash = "sha256:f604efd28f239cc21b3adb53eb061e2a205dc164be408e553b41ba2ffe0ca15c", size = 5379849, upload-time = "2025-09-14T22:17:57.372Z" },
+    { url = "https://files.pythonhosted.org/packages/a4/86/a48e56320d0a17189ab7a42645387334fba2200e904ee47fc5a26c1fd8ca/zstandard-0.25.0-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:223415140608d0f0da010499eaa8ccdb9af210a543fac54bce15babbcfc78439", size = 5058095, upload-time = "2025-09-14T22:17:59.498Z" },
+    { url = "https://files.pythonhosted.org/packages/f8/ad/eb659984ee2c0a779f9d06dbfe45e2dc39d99ff40a319895df2d3d9a48e5/zstandard-0.25.0-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:2e54296a283f3ab5a26fc9b8b5d4978ea0532f37b231644f367aa588930aa043", size = 5551751, upload-time = "2025-09-14T22:18:01.618Z" },
+    { url = "https://files.pythonhosted.org/packages/61/b3/b637faea43677eb7bd42ab204dfb7053bd5c4582bfe6b1baefa80ac0c47b/zstandard-0.25.0-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:ca54090275939dc8ec5dea2d2afb400e0f83444b2fc24e07df7fdef677110859", size = 6364818, upload-time = "2025-09-14T22:18:03.769Z" },
+    { url = "https://files.pythonhosted.org/packages/31/dc/cc50210e11e465c975462439a492516a73300ab8caa8f5e0902544fd748b/zstandard-0.25.0-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:e09bb6252b6476d8d56100e8147b803befa9a12cea144bbe629dd508800d1ad0", size = 5560402, upload-time = "2025-09-14T22:18:05.954Z" },
+    { url = "https://files.pythonhosted.org/packages/c9/ae/56523ae9c142f0c08efd5e868a6da613ae76614eca1305259c3bf6a0ed43/zstandard-0.25.0-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:a9ec8c642d1ec73287ae3e726792dd86c96f5681eb8df274a757bf62b750eae7", size = 4955108, upload-time = "2025-09-14T22:18:07.68Z" },
+    { url = "https://files.pythonhosted.org/packages/98/cf/c899f2d6df0840d5e384cf4c4121458c72802e8bda19691f3b16619f51e9/zstandard-0.25.0-cp314-cp314-musllinux_1_2_i686.whl", hash = "sha256:a4089a10e598eae6393756b036e0f419e8c1d60f44a831520f9af41c14216cf2", size = 5269248, upload-time = "2025-09-14T22:18:09.753Z" },
+    { url = "https://files.pythonhosted.org/packages/1b/c0/59e912a531d91e1c192d3085fc0f6fb2852753c301a812d856d857ea03c6/zstandard-0.25.0-cp314-cp314-musllinux_1_2_ppc64le.whl", hash = "sha256:f67e8f1a324a900e75b5e28ffb152bcac9fbed1cc7b43f99cd90f395c4375344", size = 5430330, upload-time = "2025-09-14T22:18:11.966Z" },
+    { url = "https://files.pythonhosted.org/packages/a0/1d/7e31db1240de2df22a58e2ea9a93fc6e38cc29353e660c0272b6735d6669/zstandard-0.25.0-cp314-cp314-musllinux_1_2_s390x.whl", hash = "sha256:9654dbc012d8b06fc3d19cc825af3f7bf8ae242226df5f83936cb39f5fdc846c", size = 5811123, upload-time = "2025-09-14T22:18:13.907Z" },
+    { url = "https://files.pythonhosted.org/packages/f6/49/fac46df5ad353d50535e118d6983069df68ca5908d4d65b8c466150a4ff1/zstandard-0.25.0-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:4203ce3b31aec23012d3a4cf4a2ed64d12fea5269c49aed5e4c3611b938e4088", size = 5359591, upload-time = "2025-09-14T22:18:16.465Z" },
+    { url = "https://files.pythonhosted.org/packages/c2/38/f249a2050ad1eea0bb364046153942e34abba95dd5520af199aed86fbb49/zstandard-0.25.0-cp314-cp314-win32.whl", hash = "sha256:da469dc041701583e34de852d8634703550348d5822e66a0c827d39b05365b12", size = 444513, upload-time = "2025-09-14T22:18:20.61Z" },
+    { url = "https://files.pythonhosted.org/packages/3a/43/241f9615bcf8ba8903b3f0432da069e857fc4fd1783bd26183db53c4804b/zstandard-0.25.0-cp314-cp314-win_amd64.whl", hash = "sha256:c19bcdd826e95671065f8692b5a4aa95c52dc7a02a4c5a0cac46deb879a017a2", size = 516118, upload-time = "2025-09-14T22:18:17.849Z" },
+    { url = "https://files.pythonhosted.org/packages/f0/ef/da163ce2450ed4febf6467d77ccb4cd52c4c30ab45624bad26ca0a27260c/zstandard-0.25.0-cp314-cp314-win_arm64.whl", hash = "sha256:d7541afd73985c630bafcd6338d2518ae96060075f9463d7dc14cfb33514383d", size = 476940, upload-time = "2025-09-14T22:18:19.088Z" },
+]