Has anyone managed to use custom LLM with Responses API (not chat completions endpoint)? There are no tutorials or example code out there I could find, but Responses API would be way better to use, coz it offers session management and you can directly use file_search (for RAG).