Actually, it’s probably the LLM because if it’s saying the first message, then the text to speech must be working. Might also be the transcription not capturing what the user is saying so that the LLM doesn’t respond.