After asking to the AI, the response should be delivered as a real-time stream rather than all at once