Closed
Description
Issue description
When using VertexAiRagMemoryService, memory entries retrieved via search_memory are not filtered by user_id, only by session_id. This can result in memory being shared or leaked between different users if session IDs are reused or not globally unique.
In my application, I stored information under a session for user_id = "uid1". Later, when querying from a session with user_id = "uid2" and a different session ID, the same information was still retrieved, suggesting memory is not correctly scoped by user.
Expected behavior:
Memory entries should be filtered, preferably on the server side, by at least the user_id to ensure proper isolation between users.
Possible fix:
Uncomment this line.
if not context.source_display_name.startswith(f"{app_name}.{user_id}."):
continue