fix(backend): set search_path for vector type visibility in hybrid search

- Add SET LOCAL search_path TO platform, public; to queries using vector types
- This ensures the vector type is found while keeping operators working
- Fixes hybrid search on databases using platform schema

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

autogpt_platform/backend/backend/api/features/store/embeddings.py

@@ -108,8 +108,10 @@ async def store_embedding(
         embedding_str = "[" + ",".join(str(x) for x in embedding) + "]"
 
         # Upsert the embedding
+        # Set search_path to include public for vector type visibility
         await client.execute_raw(
             """
+            SET LOCAL search_path TO platform, public;
             INSERT INTO platform."StoreListingEmbedding" (
                 "id", "storeListingVersionId", "embedding",
                 "searchableText", "contentHash", "createdAt", "updatedAt"

autogpt_platform/backend/backend/api/features/store/hybrid_search.py

@@ -147,7 +147,9 @@ async def hybrid_search(
         # Build hybrid search query with weighted scoring
         # The semantic score is (1 - cosine_distance), normalized to [0,1]
         # The lexical score is ts_rank_cd, normalized by max value
+        # Set search_path to include public for vector type visibility
         sql_query = f"""
+            SET LOCAL search_path TO platform, public;
             WITH search_scores AS (
                 SELECT
                     sa.*,
@@ -224,6 +226,7 @@ async def hybrid_search(
 
         # Count query - must also filter by min_score
         count_query = f"""
+            SET LOCAL search_path TO platform, public;
             WITH search_scores AS (
                 SELECT
                     sa.slug,