TODO Streaming responses from openai server, send messages with chunks of response Parse user's audio messages Handle errors Handle /start