Anyone working with custom LLM? 1. I added my endpoint url to vapi assistant with custom llm 2. I trigger outboud call to my own number, i pick up, I hear initial message defined in vapi assistant. 3. I anwer with some generic message (like hellow how are you" 4. I get response to my endpoint with my message transcribed 5. My llm generates answer as simple text which I want to send back for VAPI to TTS to me in the call. What's the structure that I send that response so I hear my llm's answer on the phone? In https://docs.vapi.ai/customization/custom-llm/using-your-server it says:

5. Receive and Format Response: The response from OpenAI, containing the generated text, is received and formatted according to Vapi’s expected structure.

What is that structure? I cant find it.