refactor: split gateway and pi-qmem extension into modules

- gateway: separate config, models, state, audit, guardrail, embeddings,
  store, metrics, cleanup and routes; keep main.py as FastAPI bootstrap
- extension: split client/config, six tools, config command and rules;
  preserve jiti entrypoint and registrations
- Dockerfile copies the complete gateway module set
- tests: update monkeypatch boundaries for modular config/state

gateway/metrics.py

@@ -0,0 +1,70 @@
+"""Metriche in-memory e push Prometheus/VictoriaMetrics."""
+from __future__ import annotations
+
+import time
+from typing import Any
+
+from config import VM_PUSH_INTERVAL, VM_PUSH_URL, _metrics, log
+
+
+def record_request(endpoint: str, duration: float, status_code: int) -> None:
+    _metrics["requests"][endpoint] += 1
+    _metrics["duration_sum"][endpoint] += duration
+    _metrics["duration_count"][endpoint] += 1
+    if status_code >= 400:
+        _metrics["errors"][(endpoint, status_code)] += 1
+
+
+def record_search(hits: int) -> None:
+    _metrics["search_queries"] += 1
+    _metrics["search_hits"] += hits
+
+
+def snapshot(qdrant: Any, collection: str) -> dict:
+    try:
+        points = qdrant.get_collection(collection).points_count
+    except Exception:  # noqa: BLE001
+        points = None
+    return {
+        "requests": dict(_metrics["requests"]),
+        "avg_duration_ms": {
+            endpoint: round(_metrics["duration_sum"][endpoint] / _metrics["duration_count"][endpoint] * 1000, 2)
+            for endpoint in _metrics["duration_count"]
+        },
+        "errors": {f"{endpoint}:{status}": count for (endpoint, status), count in _metrics["errors"].items()},
+        "search_queries": _metrics["search_queries"],
+        "search_hits": _metrics["search_hits"],
+        "points": points,
+    }
+
+
+def prometheus_lines(qdrant: Any, collection: str) -> list[str]:
+    lines: list[str] = []
+    for endpoint, count in _metrics["requests"].items():
+        lines.append(f'qmem_requests_total{{endpoint="{endpoint}"}} {count}')
+    for endpoint, total in _metrics["duration_sum"].items():
+        count = _metrics["duration_count"][endpoint]
+        lines.append(f'qmem_request_duration_seconds_sum{{endpoint="{endpoint}"}} {total:.6f}')
+        lines.append(f'qmem_request_duration_seconds_count{{endpoint="{endpoint}"}} {count}')
+    for (endpoint, status), count in _metrics["errors"].items():
+        lines.append(f'qmem_errors_total{{endpoint="{endpoint}",status="{status}"}} {count}')
+    lines.append(f"qmem_search_queries_total {_metrics['search_queries']}")
+    lines.append(f"qmem_search_hits_total {_metrics['search_hits']}")
+    try:
+        lines.append(f"qmem_points {qdrant.get_collection(collection).points_count}")
+    except Exception:  # noqa: BLE001
+        pass
+    return lines
+
+
+async def push_loop(qdrant: Any, collection: str, get_http) -> None:
+    while True:
+        try:
+            now_ms = int(time.time() * 1000)
+            body = "\n".join(f"{line} {now_ms}" for line in prometheus_lines(qdrant, collection)) + "\n"
+            response = await get_http().post(VM_PUSH_URL, content=body, headers={"Content-Type": "text/plain"})
+            if response.status_code >= 300:
+                log.warning("metrics push: HTTP %s", response.status_code)
+        except Exception as exc:  # noqa: BLE001
+            log.warning("metrics push error: %s", exc)
+        await __import__("asyncio").sleep(VM_PUSH_INTERVAL)