Remove apps/alerts/tasks.py

Problem

apps/alerts/tasks.py implements a legacy Celery-based pipeline that duplicates what apps/orchestration/ does properly. It violates the project’s boundary rule (“Only apps.orchestration advances the pipeline”) by running all 4 stages from inside apps.alerts, bypassing PipelineOrchestrator entirely.

Consequences of the current state:

• No PipelineRun / StageExecution records — runs through this path are invisible
• No trace_id / run_id correlation
• Retry logic reimplemented ad-hoc via Celery decorators instead of the orchestrator’s state machine
• notify_channels task hardcodes its own driver lookup, severity mapping, and message formatting

Current test coverage: 12%.

Decision

Delete apps/alerts/tasks.py and route the async webhook path through apps.orchestration.tasks.run_pipeline_task.

Changes

1. apps/alerts/views.py — Replace apps.alerts.tasks.orchestrate_event import with apps.orchestration.tasks.run_pipeline_task. Adapt .delay() call to match run_pipeline_task’s signature (payload, source, etc.). Keep the 202 async response shape.

2. apps/alerts/tasks.py — Delete entirely.

3. apps/alerts/_tests/test_tasks.py — Delete entirely.

4. apps/alerts/_tests/views/test_webhook.py — Update mocks referencing apps.alerts.tasks.orchestrate_event to reference apps.orchestration.tasks.run_pipeline_task.

5. docs/Architecture.md — Update references to old task names if they point to apps.alerts.tasks.

What stays the same

• Async/sync split in views.py (Celery available → 202; Celery down → sync fallback)
• ENABLE_CELERY_ORCHESTRATION env var
• CELERY_TASK_ALWAYS_EAGER check
• Sync fallback calls AlertOrchestrator.process_webhook() directly