perf: client httpx riusato per gli embedding (lazy, chiuso a shutdown)

- _get_http(): creazione lazy, keep-alive riusato tra le chiamate
- lifespan shutdown: aclose del client
- verificato sul server: status + search OK (istanza di test)

gateway/main.py

@@ -82,6 +82,10 @@ async def lifespan(_app: FastAPI):
             await cleanup_task
         except asyncio.CancelledError:
             pass
+        global _http
+        if _http is not None:
+            await _http.aclose()
+            _http = None
 
 
 app = FastAPI(title="Memory Gateway", version="2.5.0", lifespan=lifespan)
@@ -184,16 +188,25 @@ def _parse_ts(value: Optional[str]) -> Optional[float]:
 
 
 # ---------------------------------------------------------------------------
-# Embedding via Ollama (BGE-M3)
+# Embedding via Ollama (BGE-M3) — client httpx riusato (creato lazy, chiuso a shutdown)
 # ---------------------------------------------------------------------------
+_http: Optional[httpx.AsyncClient] = None
+
+
+def _get_http() -> httpx.AsyncClient:
+    global _http
+    if _http is None:
+        _http = httpx.AsyncClient(timeout=30)
+    return _http
+
+
 async def embed(text: str) -> list[float]:
-    async with httpx.AsyncClient(timeout=30) as client:
-        r = await client.post(
-            f"{OLLAMA_URL}/api/embed",
-            json={"model": EMBED_MODEL, "input": text},
-        )
-        r.raise_for_status()
-        return r.json()["embeddings"][0]
+    r = await _get_http().post(
+        f"{OLLAMA_URL}/api/embed",
+        json={"model": EMBED_MODEL, "input": text},
+    )
+    r.raise_for_status()
+    return r.json()["embeddings"][0]
 
 
 # ---------------------------------------------------------------------------