feat(gateway): stadio rerank con catena di fallback resiliente (frigate→brain)

- gateway/rerank.py: catena da RERANK_CHAIN (JSON, per-nodo key+timeout),
  cooldown 60s sui nodi falliti, score sigmoide [0,1], degrada con grazia
  all'ordine di fusione se tutti i nodi sono giù
- routes: /v1/memories:search applica il rerank post-fusione (fetch esteso a
  RERANK_CANDIDATES), risposta con rerank{used,backend,took_ms}, flag
  per-query rerank=false; /v1/version espone lo stato rerank
- store: search() accetta limit esteso; models: SearchIn.rerank
- metrics: qmem_rerank_calls_total + durata per backend
- test: 10 nuovi (fallback, cooldown, degradazione, integrazione) — 46 pass

gateway/config.py

@@ -23,7 +23,7 @@ GUARDRAIL_BLOCK_THRESHOLD = float(os.environ.get("GUARDRAIL_BLOCK_THRESHOLD", "0
 GUARDRAIL_WARN_THRESHOLD = float(os.environ.get("GUARDRAIL_WARN_THRESHOLD", "0.70"))
 GUARDRAIL_VERSION = "similarity-v1"
 GIT_COMMIT = os.environ.get("GIT_COMMIT", "unknown").strip()
-GATEWAY_VERSION = os.environ.get("GATEWAY_VERSION", "2.8.0").strip()
+GATEWAY_VERSION = os.environ.get("GATEWAY_VERSION", "2.9.0").strip()
 
 VM_PUSH_URL = os.environ.get("VM_PUSH_URL", "http://host.docker.internal:8428/api/v1/import/prometheus")
 VM_PUSH_INTERVAL = int(os.environ.get("VM_PUSH_INTERVAL", "30"))
@@ -33,6 +33,15 @@ logging.basicConfig(level=logging.INFO, format="%(asctime)s %(levelname)s %(mess
 log = logging.getLogger("memory-gateway")
 
 SPARSE_VECTOR_NAME = "bm25"
+
+# Re-ranking: catena di fallback resiliente (frigate → brain locale).
+# Il default nel codice è OFF; il deploy imposta RERANK_ENABLED=true e la catena.
+RERANK_ENABLED = os.environ.get("RERANK_ENABLED", "false").lower() == "true"
+RERANK_MODEL = os.environ.get("RERANK_MODEL", "bge-reranker-v2-m3")
+RERANK_CANDIDATES = int(os.environ.get("RERANK_CANDIDATES", "16"))
+RERANK_TIMEOUT_MS = int(os.environ.get("RERANK_TIMEOUT_MS", "10000"))
+RERANK_RETRY_COOLDOWN_S = int(os.environ.get("RERANK_RETRY_COOLDOWN_S", "60"))
+RERANK_CHAIN = os.environ.get("RERANK_CHAIN", "")
 _metrics: dict[str, Any] = {
     "requests": Counter(),
     "duration_sum": Counter(),
@@ -40,4 +49,6 @@ _metrics: dict[str, Any] = {
     "errors": Counter(),
     "search_queries": 0,
     "search_hits": 0,
+    "rerank_calls": Counter(),
+    "rerank_duration_sum": Counter(),
 }