headquarter/openspec/changes/notification-center/apply-pr2.md

# PR-2 Apply Report: Backend Integration for Notification Center

## Status: COMPLETE

All 5 tasks for PR-2 (NC-PR2-001 through NC-PR2-005) have been implemented, tested, and validated.

## What Was Implemented

### NC-PR2-001: Wire lifecycle_hooks.py to NotificationService

**File:** `apps/api/src/services/lifecycle_hooks.py`

- Imported `notification_service` singleton from `src.services.notification_service`
- Added `_derive_title(event_type)` helper mapping lifecycle events to human-readable titles:
  - `instance.created` → "Container created"
  - `instance.started` → "Container started"
  - `instance.stopped` → "Container stopped"
  - `instance.restarted` → "Container restarted"
  - `instance.deleted` → "Container deleted"
  - `instance.error` → "Container error"
- After `event_bus.publish(...)`, calls `notification_service.create_notification(...)` with:
  - `user_id = instance.owner_id`
  - `category = "instance"`
  - `severity = "error"` for `instance.error`, `"info"` for all others
  - `source_type = "tool_instances"`, `source_id = instance.id`
- Wrapped in `try/except`; logs failure with `correlation_id` and continues
- Event bus publish and audit row insert are unaffected by notification failure

### NC-PR2-002: Wire health_monitor.py to NotificationService

**File:** `apps/api/src/services/health_monitor.py`

- Imported `notification_service` singleton
- After `self._event_bus.publish(event_type, payload)`, calls `notification_service.create_notification(...)` with:
  - `user_id = instance.owner_id`
  - `category = "instance"` for `new_status == "error"`
  - `category = "health"` for `instance.health_changed`
  - `severity` mapped:
    - `"error"` for crash
    - `"warning"` for unhealthy
    - `"info"` for recovery (running)
  - `title` mapped:
    - "Container failed" for error
    - "Container unhealthy" for unhealthy
    - "Container recovered" for running
- Wrapped in `try/except`; logs failure with `correlation_id` and continues
- Original event bus publish and health check insert are unaffected

### NC-PR2-003: Extend UserConfig schema for notification preferences

**File:** `apps/api/src/api/user_config.py`

- Added `notification_mute_categories: list[str] | None = None` to `UserConfigResponse`
- Added `notification_toast_level: str | None = None` to `UserConfigResponse`
- Added the same fields to `UserConfigUpdate`
- Existing config keys are unaffected; new fields are optional with `None` defaults

### NC-PR2-004: Event producer integration tests (RED)

**File:** `apps/api/tests/integration/test_notifications_lifecycle.py` *(new)*

6 integration tests covering:

1. `test_lifecycle_event_creates_notification` — lifecycle hook `instance.started` creates `severity="info"` notification for owner
2. `test_health_monitor_error_creates_notification` — simulated crash creates `severity="error"` notification for owner
3. `test_notification_failure_does_not_block_event_pipeline` — mocked `create_notification` raising `RuntimeError`; event still published, no exception escapes
4. `test_notification_ownership_matches_instance_owner` — notification `user_id` equals `instance.owner_id`, not the API caller
5. `test_lifecycle_error_creates_error_notification` — `instance.error` maps to `severity="error"`, title="Container error"
6. `test_health_monitor_unhealthy_creates_warning_notification` — tunnel failure creating `severity="warning"`, category="health"

### NC-PR2-005: Verify producer tests and clean up (GREEN / REFACTOR)

- All 6 new integration tests pass
- 13 unit tests for `NotificationService` pass (no regressions)
- 10 integration tests for notifications API pass (no regressions)
- 6 existing health monitor unit tests pass (no regressions)
- 6 existing event integration tests pass (no regressions)
- `ruff check` passes on all modified files

## TDD Cycle Evidence

| Cycle | Task | Test File | RED | GREEN | Evidence |
|-------|------|-----------|-----|-------|----------|
| 1 | NC-PR2-004 (producer flow) | `tests/integration/test_notifications_lifecycle.py` | 4 tests written against unwired producers | Wired `lifecycle_hooks.py` and `health_monitor.py` | 4 passed |
| 2 | NC-PR2-004 (severity mapping) | `tests/integration/test_notifications_lifecycle.py` | Added error + warning severity tests | Already green from implementation | 6 passed |
| 3 | NC-PR2-003 (UserConfig schema) | `src/api/user_config.py` | Schema extended with new optional fields | PATCH/GET endpoints validate correctly | Verified manually |
| 4 | NC-PR2-005 (REFACTOR) | All files | — | ruff clean, no regressions across 41 related tests | All pass |

## Changed Files

1. `apps/api/src/services/lifecycle_hooks.py` — Wired `NotificationService` after event bus publish
2. `apps/api/src/services/health_monitor.py` — Wired `NotificationService` after state change event publish
3. `apps/api/src/api/user_config.py` — Added `notification_mute_categories` and `notification_toast_level` to Pydantic schemas
4. `apps/api/tests/integration/test_notifications_lifecycle.py` *(new)* — 6 integration tests for event-to-notification flow

## Test Commands & Exit Codes

```bash
# New integration tests for event producers (6 tests)
cd apps/api && python -m pytest tests/integration/test_notifications_lifecycle.py -v
# Exit: 0 — 6 passed

# NotificationService unit tests (no regressions)
cd apps/api && python -m pytest tests/unit/test_notification_service.py -v
# Exit: 0 — 13 passed

# Notifications API integration tests (no regressions)
cd apps/api && python -m pytest tests/integration/test_notifications_api.py -v
# Exit: 0 — 10 passed

# Health monitor unit tests (no regressions)
cd apps/api && python -m pytest tests/unit/test_health_monitor.py -v
# Exit: 0 — 6 passed

# Event integration tests (no regressions)
cd apps/api && python -m pytest tests/integration/test_events.py -v
# Exit: 0 — 6 passed

# Combined relevant test suite
cd apps/api && python -m pytest \
  tests/unit/test_notification_service.py \
  tests/integration/test_notifications_api.py \
  tests/integration/test_notifications_lifecycle.py \
  tests/unit/test_health_monitor.py \
  tests/integration/test_events.py \
  -v
# Exit: 0 — 41 passed

# Ruff linting on all modified files
cd apps/api && python -m ruff check \
  src/services/lifecycle_hooks.py \
  src/services/health_monitor.py \
  src/api/user_config.py \
  tests/integration/test_notifications_lifecycle.py
# Exit: 0 — All checks passed
```

## Deviations from Design

None. All mappings and behaviors match the design spec (section 1.3) and task requirements exactly.

## Surprises / Decisions

1. **Health monitor `test_health_monitor_unhealthy_creates_warning_notification` required `public_url`:** The health monitor only checks tunnel health when `instance.public_url` is truthy. Without setting it on the test fixture instance, `_derive_status` returned `"running"` instead of `"unhealthy"`, which created an `"info"` notification. Fixed by setting `test_instance.public_url` in the test before calling `_check_instance`.

2. **Patch target for failure test:** The `test_notification_failure_does_not_block_event_pipeline` patches `src.services.lifecycle_hooks.notification_service.create_notification`. This only works because `lifecycle_hooks.py` imports `notification_service` at module level, making the attribute resolvable by `unittest.mock.patch`.

3. **No schema migration needed for UserConfig:** Preferences are stored in the existing JSON `config` blob, consistent with the existing pattern (theme, editor, git identity). No Alembic migration required.

## Risks

- **None:** All changes are additive. Event producers use `try/except` so notification failures cannot block the event pipeline. No existing test regressions introduced.

## PR Boundary

This PR covers PR-2 only (NC-PR2-001 through NC-PR2-005). PR-3 (frontend core) and PR-4 (toast coordination) are out of scope.