Plus I like to write tests for my code, and figuring out how to serve a mock HTTP response to each different LLM client library in turn is a whole lot more work than just using httpx and pytest-httpx – which I already know

→ View original post on X — @simonw