Raised on LibreChat-AI#16246: LibreChat-AI#16246 (comment)
conversationList.maxEndpointFilters (librechat.yaml, default 50, added in LibreChat-AI#16245) bounds how many endpoints a list request may name, and parseConversationListFilters answers 400 past it. The chat list Endpoint facet in LibreChat-AI#16246 offers every configured endpoint as an unbounded multi-select and the client never learns the limit, so a deployment that sets the limit below the number of endpoints it serves lets a user build a selection the server refuses; the Chats section then shows its load-error state until the facet is reset.
With the default of 50 this needs a deliberately low setting. The fix spans both sides: publish the limit where the client reads config (startup config), then stop the facet at the limit (disable further endpoints and say why) rather than submitting an over-limit request.
Raised on LibreChat-AI#16246: LibreChat-AI#16246 (comment)
conversationList.maxEndpointFilters(librechat.yaml, default 50, added in LibreChat-AI#16245) bounds how many endpoints a list request may name, andparseConversationListFiltersanswers 400 past it. The chat list Endpoint facet in LibreChat-AI#16246 offers every configured endpoint as an unbounded multi-select and the client never learns the limit, so a deployment that sets the limit below the number of endpoints it serves lets a user build a selection the server refuses; the Chats section then shows its load-error state until the facet is reset.With the default of 50 this needs a deliberately low setting. The fix spans both sides: publish the limit where the client reads config (startup config), then stop the facet at the limit (disable further endpoints and say why) rather than submitting an over-limit request.