* hard-coded openai models * ensure index if disable ingest with langflow is active * update backend to not update embedding model when flag is disabled * initialize index on startup when feature flag is enabled * put config.yaml on docker compose