M22: implement operational observability

backend/app/api/routers/observability.py

@@ -0,0 +1,40 @@
+from __future__ import annotations
+
+import hmac
+
+from fastapi import APIRouter, Header, HTTPException
+from fastapi.responses import Response
+from prometheus_client import CONTENT_TYPE_LATEST, generate_latest
+from sqlalchemy import func, select
+
+from app.core.config import get_settings
+from app.core.db import SessionLocal
+from app.core.observability import OUTBOX_EVENTS
+from app.models.outbox import DEMO_SCENARIO_ERROR_CODE, OutboxEvent
+
+router = APIRouter(tags=["observability"])
+settings = get_settings()
+
+
+def _refresh_database_metrics() -> None:
+    with SessionLocal() as db:
+        rows = db.execute(
+            select(
+                OutboxEvent.delivery_status,
+                (OutboxEvent.last_error_code == DEMO_SCENARIO_ERROR_CODE).label("demo"),
+                func.count(),
+            ).group_by(OutboxEvent.delivery_status, "demo")
+        ).all()
+    OUTBOX_EVENTS.clear()
+    for status, is_demo, count in rows:
+        OUTBOX_EVENTS.labels(str(status), "synthetic" if is_demo else "operational").set(count)
+
+
+@router.get("/metrics", include_in_schema=False)
+def metrics(authorization: str | None = Header(default=None)) -> Response:
+    if settings.metrics_bearer_token:
+        supplied = authorization.removeprefix("Bearer ") if authorization else ""
+        if not hmac.compare_digest(supplied, settings.metrics_bearer_token):
+            raise HTTPException(status_code=401, detail="Metrics token required")
+    _refresh_database_metrics()
+    return Response(content=generate_latest(), media_type=CONTENT_TYPE_LATEST)