feat: complete production workflow migration

2026-08-06 14:29:57 +08:00
parent 7f215e79c4
commit 8df5266abb
448 changed files with 56937 additions and 14619 deletions
@@ -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.
@@ -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%
@@ -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%
@@ -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%
@@ -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
@@ -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%
@@ -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%
@@ -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%
@@ -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%
@@ -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%