feat(metrics): enable repair tracking and Prometheus lifecycle for observability

app/routers/nl2sql.py

@@ -108,7 +108,7 @@ def _cache_gc(now: float) -> None:
     for k, (ts, _) in list(_CACHE.items()):
         if now - ts > _CACHE_TTL:
             _CACHE.pop(k, None)
-    # size eviction (ساده)
+    # size eviction
     while len(_CACHE) > _CACHE_MAX:
         _CACHE.pop(next(iter(_CACHE)), None)
 

docker-compose.yml

@@ -1,40 +1,31 @@
-version: "3.9"
-
 services:
-  postgres:
-    image: postgres:16
-    container_name: nl2sql_pg
-    environment:
-      POSTGRES_USER: postgres
-      POSTGRES_PASSWORD: postgres
-      POSTGRES_DB: demo
-    volumes:
-      - pgdata:/var/lib/postgresql/data
-      - ./infra/migrate.sql:/docker-entrypoint-initdb.d/00_init.sql:ro
-    ports:
-      - "5432:5432"
-    healthcheck:
-      test: ["CMD-SHELL", "pg_isready -U postgres -d demo"]
-      interval: 5s
-      timeout: 3s
-      retries: 10
-
   api:
-    build:
-      context: .
-      dockerfile: Dockerfile
-    container_name: nl2sql_api
-    depends_on:
-      postgres:
-        condition: service_healthy
     environment:
-      DB_MODE: postgres
       POSTGRES_DSN: dbname=demo user=postgres password=postgres host=postgres port=5432
-      OPENAI_MODEL_ID: gpt-4o-mini
+      OPENAI_MODEL: gpt-4o-mini
       OPENAI_API_KEY: ${OPENAI_API_KEY}
     ports:
       - "8000:8000"
     command: ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "8000", "--proxy-headers"]
 
+  prometheus:
+    image: prom/prometheus:latest
+    container_name: prometheus
+    restart: unless-stopped
+    ports:
+      - "9090:9090"
+    volumes:
+      # 📘 Prometheus config
+      - ./provisioning/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml:ro
+      # 📘 Rules (recording + alert rules)
+      - ./provisioning/prometheus/rules.yml:/etc/prometheus/rules.yml:ro
+      # 📘 Persistent TSDB storage
+      - prometheus_data:/prometheus
+    command:
+      - "--config.file=/etc/prometheus/prometheus.yml"
+      - "--storage.tsdb.path=/prometheus"
+      - "--web.enable-lifecycle"
+      - "--web.enable-admin-api"
+
 volumes:
-  pgdata:
+  prometheus_data:

nl2sql/metrics.py

@@ -55,7 +55,7 @@ verifier_failures_total = Counter(
 repair_attempts_total = Counter(
     "repair_attempts_total",
     "Number of repair loop attempts",
-    ["outcome"],  # success | failed
+    ["outcome"],  # attempt | success | failed
     registry=REGISTRY,
 )
 

nl2sql/pipeline.py

@@ -305,6 +305,7 @@ class Pipeline:
             if not verified:
                 for _attempt in range(2):
                     # repair
+                    repair_attempts_total.labels(outcome="attempt").inc()
                     t0 = time.perf_counter()
                     r_fix = self._safe_stage(
                         self.repair.run,

prometheus/rules.yml

@@ -23,11 +23,8 @@ groups:
     # repair success rate (0..1)
     - record: nl2sql:repair_success_rate
       expr: |
-        (
-          sum(rate(repair_attempts_total{outcome="success"}[5m]))
-        )
-        /
-        clamp_min(sum(rate(repair_attempts_total[5m])), 1)
+          (sum(increase(repair_attempts_total{outcome="success"}[30m]))) /
+          clamp_min(sum(increase(repair_attempts_total[30m])), 1)
 
     # cache hit ratio (0..1)
     - record: nl2sql:cache_hit_ratio