Facing mysterious latency with your Python-based AI agents under load? Seeing extremely large time-to-first-tokens for new requests? Before blaming your LLM provider, look into the framework you're using, or your own code. Maybe the event loop is getting blocked.

I built `pyleak` to catch these event loop blocks automatically: https://github.com/deepankarm/pyleak

Use it as a pytest plugin:

    import pytest

    @pytest.mark.no_leak
    async def test_my_agent():
       ...