Hey, I have trouble setting up the Custom LLM server response. Can someone share the response structure that the vapi client expects from the server?