fix: Preserve arrays in metadata + refresh OAuth token before sync

1. process_metadata (vector/utils.py):
   - Lists/arrays now preserved as-is (Pinecone supports native arrays)
   - Only datetime and dict types converted to strings
   - Enables Pinecone queries like: filter: {labels: {$in: ['SENT']}}
   - Fixed iteration bug (was modifying dict while iterating)

2. _sync_user_periodic (gmail_auto_sync.py):
   - Now uses OAuth manager to get/refresh token before sync
   - Prevents 'OAuth token expired' errors during attachment fetching
   - Falls back to stored token if refresh fails

backend/open_webui/retrieval/vector/utils.py

@@ -13,16 +13,42 @@ def filter_metadata(metadata: dict[str, any]) -> dict[str, any]:
 def process_metadata(
     metadata: dict[str, any],
 ) -> dict[str, any]:
-    for key, value in metadata.items():
-        # Remove large fields
-        if key in KEYS_TO_EXCLUDE:
-            del metadata[key]
+    """
+    Process metadata for vector storage.
 
-        # Convert non-serializable fields to strings
-        if (
-            isinstance(value, datetime)
-            or isinstance(value, list)
-            or isinstance(value, dict)
-        ):
+    - Removes large excluded fields
+    - Converts non-JSON-serializable types (datetime, dict) to strings
+    - Preserves lists/arrays (Pinecone supports array filtering with $in)
+    """
+    keys_to_delete = []
+
+    for key, value in metadata.items():
+        # Mark large fields for deletion
+        if key in KEYS_TO_EXCLUDE:
+            keys_to_delete.append(key)
+            continue
+
+        # Convert datetime to ISO string
+        if isinstance(value, datetime):
+            metadata[key] = value.isoformat()
+        # Convert dict to string (nested objects not supported in most vector DBs)
+        elif isinstance(value, dict):
             metadata[key] = str(value)
+        # KEEP lists as-is - Pinecone supports arrays for $in filtering
+        # e.g., filter: {"labels": {"$in": ["SENT"]}}
+        elif isinstance(value, list):
+            # Ensure list items are serializable (strings, numbers, booleans)
+            metadata[key] = [
+                (
+                    str(item)
+                    if not isinstance(item, (str, int, float, bool, type(None)))
+                    else item
+                )
+                for item in value
+            ]
+
+    # Delete excluded keys (can't modify dict while iterating)
+    for key in keys_to_delete:
+        del metadata[key]
+
     return metadata

backend/open_webui/utils/gmail_auto_sync.py

@@ -1238,10 +1238,20 @@ async def _sync_user_periodic(user_id: str) -> bool:
             logger.debug(f"⏭️  No Google OAuth session for user {user_id}")
             return False
 
-        # Use OAuth token from session
-        # Note: Token refresh happens automatically on next OAuth login
-        # If token is expired, Gmail API will fail and we'll retry on next cycle
-        oauth_token = oauth_session.token
+        # Get refreshed OAuth token using OAuth manager (auto-refreshes if expired)
+        # This is critical for long-running syncs where token may expire mid-process
+        try:
+            from open_webui.main import app
+
+            oauth_token = await app.state.oauth_manager.get_oauth_token(
+                user_id=user_id,
+                session_id=oauth_session.id,
+            )
+        except Exception as e:
+            logger.warning(
+                f"⚠️  Failed to get/refresh OAuth token for user {user_id}: {e}"
+            )
+            oauth_token = oauth_session.token  # Fallback to stored token
 
         # Validation: Check OAuth token exists and has access_token
         if not oauth_token or not oauth_token.get("access_token"):