I'm talking about this : https://docs.vapi.ai/customization/custom-llm/using-your-server where in the code they are saying to format the response according to vapi structure