cortex/context.py

@@ -163,21 +163,18 @@ async def _search_neomem(
         List of memory objects with full structure, or empty list on failure
     """
     try:
-        client = NeoMemClient(base_url=NEOMEM_API)
+        # NeoMemClient reads NEOMEM_API from environment, no base_url parameter
+        client = NeoMemClient()
         results = await client.search(
             query=query,
             user_id=user_id,
-            limit=limit
+            limit=limit,
+            threshold=RELEVANCE_THRESHOLD
         )
 
-        # Filter by relevance threshold
-        filtered = [
-            r for r in results
-            if r.get("score", 0.0) >= RELEVANCE_THRESHOLD
-        ]
-
-        logger.info(f"NeoMem search returned {len(filtered)}/{len(results)} relevant results")
-        return filtered
+        # Results are already filtered by threshold in NeoMemClient.search()
+        logger.info(f"NeoMem search returned {len(results)} relevant results")
+        return results
 
     except Exception as e:
         logger.warning(f"NeoMem search failed: {e}")