activity-core/src/activity_core/workflows.py

"""Temporal workflow definitions for activity-core.

Two workflows are registered here:
  - RunActivityWorkflow  → orchestrator-tq
  - TaskExecutorWorkflow → task-execution-tq

Workflow IDs follow the conventions in docs/conventions.md:
  RunActivityWorkflow:  activity-{activity_id}:{trigger_key}
  TaskExecutorWorkflow: task-{run_id}:{task_type}:{index}
"""

from __future__ import annotations

import uuid
from datetime import timedelta

from temporalio import workflow
from temporalio.common import RetryPolicy, SearchAttributeKey, TypedSearchAttributes, SearchAttributePair

with workflow.unsafe.imports_passed_through():
    from activity_core.activities import (
        load_activity_definition,
        log_run,
        persist_task_instance,
        resolve_context,
    )
    from activity_core.template_engine import evaluate_templates
    from activity_core.schedule_manager import SCHEDULED_TRIGGER_KEY

# T32: Custom search attributes for Temporal visibility (must be registered in Temporal first).
# Registration: temporal operator search-attribute create --name ActivityId --type Keyword
_ACTIVITY_ID_KEY = SearchAttributeKey.for_keyword("ActivityId")
_ACTIVITY_NAME_KEY = SearchAttributeKey.for_keyword("ActivityName")

_RETRY_POLICY = RetryPolicy(
    initial_interval=timedelta(seconds=1),
    backoff_coefficient=2.0,
    maximum_interval=timedelta(minutes=5),
    maximum_attempts=10,
)

_ACTIVITY_TIMEOUT = timedelta(minutes=5)
_TASK_QUEUE = "task-execution-tq"


@workflow.defn
class RunActivityWorkflow:
    """Durable orchestration workflow.

    Sequence:
      1. load_activity_definition(activity_id)  → defn dict
      2. resolve_context(defn.context_sources)  → context snapshot
      3. evaluate_templates(templates, context) → task specs (pure, no activity)
      4. log_run(...)                           → run_id
      5. start_child_workflow per task spec     (fire-and-forget, detached)
    """

    @workflow.run
    async def run(
        self,
        activity_id: str,
        trigger_key: str,
        scheduled_for: str | None = None,
    ) -> dict:
        """
        Args:
            activity_id:    UUID of the ActivityDefinition row.
            trigger_key:    ISO-8601 datetime (cron) or event_id (event trigger).
                            Used as the idempotency key component.
            scheduled_for:  ISO-8601 string of the nominal scheduled time (cron only).

        Returns:
            {"run_id": str, "tasks_spawned": int}
        """
        # ── 1. Load definition ────────────────────────────────────────────────
        defn: dict = await workflow.execute_activity(
            load_activity_definition,
            activity_id,
            start_to_close_timeout=_ACTIVITY_TIMEOUT,
            retry_policy=_RETRY_POLICY,
        )

        # T32: Tag this workflow execution with activity metadata so runs are
        # filterable in the Temporal UI (requires ActivityId + ActivityName to be
        # registered as custom search attributes — see docs/runbook.md).
        workflow.upsert_search_attributes(
            TypedSearchAttributes([
                SearchAttributePair(_ACTIVITY_ID_KEY, activity_id),
                SearchAttributePair(_ACTIVITY_NAME_KEY, defn.get("name", "")),
            ])
        )

        # ── 2. Resolve context ────────────────────────────────────────────────
        context_snapshot: dict = await workflow.execute_activity(
            resolve_context,
            defn["context_sources"],
            start_to_close_timeout=_ACTIVITY_TIMEOUT,
            retry_policy=_RETRY_POLICY,
        )

        # ── 3. Evaluate templates (pure — no activity) ────────────────────────
        task_specs: list[dict] = evaluate_templates(
            defn["task_templates"], context_snapshot
        )

        # ── 4. Log the run ────────────────────────────────────────────────────
        # run_id is derived deterministically so log_run retries are idempotent.
        # For schedule-fired runs the trigger_key is the sentinel "scheduled";
        # each fire has a unique workflow_id (embeds ${firstScheduledTime}), so
        # we use the workflow_id as the dedup key instead.
        if trigger_key == SCHEDULED_TRIGGER_KEY:
            dedup_source = workflow.info().workflow_id
        else:
            dedup_source = f"{activity_id}:{trigger_key}"
        run_id = str(uuid.uuid5(uuid.NAMESPACE_URL, dedup_source))
        await workflow.execute_activity(
            log_run,
            {
                "run_id": run_id,
                "activity_id": activity_id,
                "scheduled_for": scheduled_for,
                "context_snapshot": context_snapshot,
                "tasks_spawned": len(task_specs),
                "version_used": defn["version"],
            },
            start_to_close_timeout=_ACTIVITY_TIMEOUT,
            retry_policy=_RETRY_POLICY,
        )

        # ── 5. Spawn task executor children (fire-and-forget) ─────────────────
        for index, spec in enumerate(task_specs):
            child_id = f"task-{run_id}:{spec['task_type']}:{index}"
            await workflow.start_child_workflow(
                TaskExecutorWorkflow,
                args=[run_id, spec["task_type"], spec["params"]],
                id=child_id,
                task_queue=_TASK_QUEUE,
                parent_close_policy=workflow.ParentClosePolicy.ABANDON,
            )

        return {"run_id": run_id, "tasks_spawned": len(task_specs)}


@workflow.defn
class TaskExecutorWorkflow:
    """Child workflow that executes one concrete task instance.

    Stub behaviour: persists a task_instances row with status=done and
    returns immediately. Real task execution logic replaces this in a
    later workstream.

    task_id is derived deterministically from the workflow's own ID so
    persist_task_instance retries remain idempotent.
    """

    @workflow.run
    async def run(self, run_id: str, task_type: str, params: dict) -> dict:
        # Derive a stable task_id from this workflow's own ID.
        task_id = str(
            uuid.uuid5(uuid.NAMESPACE_URL, workflow.info().workflow_id)
        )

        workflow.logger.info(
            "TaskExecutorWorkflow started",
            extra={"run_id": run_id, "task_type": task_type, "task_id": task_id},
        )

        await workflow.execute_activity(
            persist_task_instance,
            {
                "id": task_id,
                "run_id": run_id,
                "type": task_type,
                "params": params,
                "status": "done",
            },
            task_queue=_TASK_QUEUE,
            start_to_close_timeout=_ACTIVITY_TIMEOUT,
            retry_policy=_RETRY_POLICY,
        )

        return {"task_id": task_id, "status": "done"}