Anyone using Custom LLM and it works for them? Because im just piping through OpenAI response and it still doesnt work.