feat: complete production workflow migration

src/gyxx_flow/workflow/model.py

@@ -3,9 +3,10 @@
 from __future__ import annotations
 
 import heapq
+import math
 import re
 from dataclasses import dataclass
-from typing import TYPE_CHECKING, Protocol
+from typing import TYPE_CHECKING, Literal, Protocol
 
 if TYPE_CHECKING:
     from gyxx_flow.core.context import RunContext
@@ -15,6 +16,7 @@ if TYPE_CHECKING:
 
 _ID = re.compile(r"^[a-z][a-z0-9]*(?:[._-][a-z0-9]+)*$")
 _RESOURCE = re.compile(r"^[A-Za-z0-9][A-Za-z0-9:._-]*$")
+ReplayPolicy = Literal["guarded", "idempotent"]
 
 
 class WorkflowValidationError(ValueError):
@@ -47,6 +49,8 @@ class StepDefinition:
     resources: tuple[str, ...] = ()
     production_sink: bool = False
     official_notification: bool = False
+    replay_policy: ReplayPolicy = "guarded"
+    run_after_failure: bool = False
 
     def __post_init__(self) -> None:
         object.__setattr__(self, "depends_on", tuple(self.depends_on))
@@ -65,7 +69,10 @@ class StepDefinition:
             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
+            isinstance(self.timeout_seconds, bool)
+            or not isinstance(self.timeout_seconds, (int, float))
+            or not math.isfinite(self.timeout_seconds)
+            or self.timeout_seconds <= 0
         ):
             raise WorkflowValidationError(
                 f"timeout_seconds must be positive for step {self.step_id!r}"
@@ -89,6 +96,20 @@ class StepDefinition:
                 raise WorkflowValidationError(
                     f"invalid resource for step {self.step_id!r}: {resource!r}"
                 )
+        if not isinstance(self.run_after_failure, bool):
+            raise WorkflowValidationError(
+                f"run_after_failure must be boolean for step {self.step_id!r}"
+            )
+        if self.replay_policy not in {"guarded", "idempotent"}:
+            raise WorkflowValidationError(
+                f"invalid replay_policy for step {self.step_id!r}: "
+                f"{self.replay_policy!r}"
+            )
+        if self.replay_policy == "idempotent" and not self.production_sink:
+            raise WorkflowValidationError(
+                f"idempotent replay_policy requires production_sink for step "
+                f"{self.step_id!r}"
+            )
 
 
 @dataclass(frozen=True, slots=True)