Assuming you're fixed on not disconnecting: I would recommend using ollama (or install llamacpp if you're on Windows) to bring along a modestly capable LLM you

Assuming you're fixed on not disconnecting: I would recommend using ollama (or install llamacpp if you're on Windows) to bring along a modestly capable LLM you can run locally such a CodeLlama or Minstrel. They won't be as capable as ChatGPT, but a lot of knowledge gets compressed into these relatively small models.