livekit/agents ? reverse-engineered prompt

I want to build a real-time voice AI bot in Python. It should be able to listen to a user speaking, figure out what they said, process it with an LLM like one from OpenAI, and then speak a response back. The whole thing should feel like a natural, real-time conversation.

It's important that the agent can tell when a person has finished their thought so it doesn't interrupt them. I also want it to be flexible, so I can easily swap in different models for speech-to-text, the AI brain, and the text-to-speech voice. Let's give it the ability to use tools, maybe a simple function to look up the weather. It should also be able to connect to phone calls. Please use the LiveKit agents framework for this; feel free to check their latest docs online to make sure you get the setup right.