Telemetry v2

monitoring_lab/docker-compose.yml

@@ -0,0 +1,20 @@
+services:
+  prometheus:
+    image: prom/prometheus:latest
+    container_name: prometheus
+    ports:
+      - "9090:9090"
+    volumes:
+      - ./prometheus.yml:/etc/prometheus/prometheus.yml
+    extra_hosts:
+      - "host.docker.internal:host-gateway"
+
+  grafana:
+    image: grafana/grafana:latest
+    container_name: grafana
+    ports:
+      - "3000:3000"
+    environment:
+      - GF_SECURITY_ADMIN_PASSWORD=admin
+    depends_on:
+      - prometheus

monitoring_lab/prometheus.yml

@@ -0,0 +1,8 @@
+lobal:
+  scrape_interval: 5s # How often to fetch metrics
+
+scrape_configs:
+  - job_name: 'MLPA_local'
+    metrics_path: '/metrics'
+    static_configs:
+      - targets: ['host.docker.internal:8080']

src/mlpa/core/auth/authorize.py

@@ -4,6 +4,8 @@
 
 from mlpa.core.classes import AuthorizedChatRequest, ChatRequest, ServiceType
 from mlpa.core.config import env
+from mlpa.core.logger import logger
+from mlpa.core.prometheus_metrics import metrics
 from mlpa.core.routers.appattest import app_attest_auth
 from mlpa.core.routers.fxa import fxa_auth
 from mlpa.core.utils import parse_app_attest_jwt
@@ -16,13 +18,19 @@ async def authorize_request(
     use_app_attest: Annotated[bool | None, Header()] = None,
     use_qa_certificates: Annotated[bool | None, Header()] = None,
 ) -> AuthorizedChatRequest:
+    metrics.auth_request_count_total.inc()
     if not authorization:
+        metrics.auth_error_count_total.labels(error=f"MalformedRequest")
         raise HTTPException(status_code=401, detail="Missing authorization header")
     if use_app_attest:
         assertionAuth = parse_app_attest_jwt(authorization, "assert")
         data = await app_attest_auth(assertionAuth, chat_request, use_qa_certificates)
         if data:
             if data.get("error"):
+                metrics.auth_error_count_total.labels(error=f"AppattestUnauthorized")
+                logger.error(
+                    "Unauthorized", extra={"error": data["error"], "type": "appattest"}
+                )
                 raise HTTPException(status_code=401, detail=data["error"])
             return AuthorizedChatRequest(
                 user=f"{assertionAuth.key_id_b64}:{service_type.value}",  # "user" is key_id_b64 from app attest
@@ -33,11 +41,17 @@ async def authorize_request(
         fxa_user_id = fxa_auth(authorization)
         if fxa_user_id:
             if fxa_user_id.get("error"):
+                metrics.auth_error_count_total.labels(error=f"FxAUnauthorized")
+                logger.error(
+                    "Unauthorized", extra={"error": fxa_user_id["error"], "type": "FxA"}
+                )
                 raise HTTPException(status_code=401, detail=fxa_user_id["error"])
             return AuthorizedChatRequest(
                 user=f"{fxa_user_id['user']}:{service_type.value}",
                 **chat_request.model_dump(exclude_unset=True),
             )
+    metrics.auth_error_count_total.labels(error=f"Cancelled")
+    logger.error("Auth Cancelled", extra={"error": "Cancelled", "type": "None"})
     raise HTTPException(
         status_code=401, detail="Please authenticate with App Attest or FxA."
     )