refac: owner-bind ephemeral web-search RAG collections (#26706)

The web-search-* namespace was the one collection namespace filter_accessible_collections admitted unconditionally for any non-admin user, on both read and write, unlike file-*, user-memory-* and knowledge bases which are owner-scoped. process_web_search now mints these ephemeral per-query collections as web-search-{user.id}-<hash>, and the access helper only admits web-search-{requester.id}-* names, so a web-search collection is readable and writable only by the user who created it (admins keep their bypass). The collections hold transient public web-search results and their names are non-enumerable query hashes, so there was no demonstrated cross-user access path; this removes the namespace exception so the per-user scoping the other namespaces enforce also covers web-search.

Co-authored-by: rexpository <rexpository@users.noreply.github.com>

backend/open_webui/retrieval/utils.py

@@ -1267,7 +1267,7 @@ async def filter_accessible_collections(
       - any name with characters outside [A-Za-z0-9_-] → rejected
       - file-*          → validated via has_access_to_file
       - user-memory-*   → must match user's own memory collection
-      - web-search-*    → ephemeral per-query collections, always allowed
+      - web-search-*    → ephemeral per-query collections, owner-bound to web-search-{user.id}-*
       - knowledge-bases → always denied (system meta-collection)
       - everything else → if the name matches a knowledge base, validated
                           via Knowledges.check_access_by_user_id; if no
@@ -1302,10 +1302,10 @@ async def filter_accessible_collections(
             if name == f'user-memory-{user.id}':
                 validated.add(name)
         elif name.startswith('web-search-'):
-            # Ephemeral collections created by process_web_search — safe
-            # to allow because they contain only transient web-search
-            # results scoped to the requesting user's session.
-            validated.add(name)
+            # Ephemeral per-query collections, owner-bound: process_web_search mints
+            # them as web-search-{user.id}-<hash>, so only the creator may read/write.
+            if name.startswith(f'web-search-{user.id}-'):
+                validated.add(name)
         else:
             # May be a knowledge-base ID or a legacy/ephemeral collection.
             # If it IS a KB, enforce access control.  If no such KB

backend/open_webui/routers/retrieval.py

@@ -2657,7 +2657,8 @@ async def process_web_search(request: Request, form_data: SearchForm, user=Depen
             }
         else:
             # Create a single collection for all documents
-            collection_name = f'web-search-{calculate_sha256_string("-".join(form_data.queries))}'[:63]
+            # Bind the ephemeral collection to its owner so filter_accessible_collections can scope it per-user.
+            collection_name = f'web-search-{user.id}-{calculate_sha256_string("-".join(form_data.queries))}'[:63]
 
             try:
                 await run_in_threadpool(