Phase 5: Notifications — WebSocket, APScheduler, AI tool, health review

Backend:
- Notification model + Alembic migration
- Notification service: CRUD, mark read, unread count, pending scheduled
- WebSocket manager singleton for real-time push
- WebSocket endpoint /ws/notifications with JWT auth via query param
- APScheduler integration: periodic notification sender (every 60s),
  daily proactive health review job (8 AM)
- AI tool: schedule_notification (immediate or scheduled)
- Health review worker: analyzes user memory via Claude, creates
  ai_generated notifications with WebSocket push

Frontend:
- Notification API client + Zustand store
- WebSocket hook with auto-reconnect (exponential backoff)
- Notification bell in header with unread count badge + dropdown
- Notifications page with type badges, mark read, mark all read
- WebSocket initialized in AppLayout for app-wide real-time updates
- Enabled notifications nav in sidebar
- English + Russian translations

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

backend/app/schemas/notification.py

@@ -0,0 +1,30 @@
+import uuid
+from datetime import datetime
+
+from pydantic import BaseModel, Field
+
+
+class NotificationResponse(BaseModel):
+    id: uuid.UUID
+    user_id: uuid.UUID
+    title: str
+    body: str
+    type: str
+    channel: str
+    status: str
+    scheduled_at: datetime | None
+    sent_at: datetime | None
+    read_at: datetime | None
+    metadata: dict | None = Field(None, alias="metadata_")
+    created_at: datetime
+
+    model_config = {"from_attributes": True, "populate_by_name": True}
+
+
+class NotificationListResponse(BaseModel):
+    notifications: list[NotificationResponse]
+    unread_count: int
+
+
+class UnreadCountResponse(BaseModel):
+    count: int