Exercise 01 — Streaming Chatbot

Timebox: 30-45 minutes

Goal

Turn chatbot.py into a multi-turn streaming chatbot that keeps history and behaves cleanly under transient failures.

Work in

• chatbot.py

Tasks

1. Replace the mock stream with a real provider call or keep the mock and improve the architecture first.
2. Keep conversation history across turns.
3. Allow a system prompt.
4. Print token or usage information after each response.
5. Retry on rate-limit or transient failures.

Done when

• The CLI supports multiple turns
• Responses stream instead of arriving all at once
• History is preserved across turns
• Failures retry before surfacing