in practice, you want the LLM to have the entire body of text prior to responding. what you should do i begin streaming the response from the LLM and send that to your speech to text processor if you want to improve voice speed.