Dev Tools · 2h ago
Apte merges unit tests and LLM evals in one async Python framework
Apte is an async-first Python test framework that runs unit tests and LLM evals in the same engine with shared fixtures. It uses typed, inspectable dependencies via Annotated[..., Use(...)] instead of pytest's name-based fixture resolution. The framework aims to eliminate the tooling split between code testing and model evaluation.
Meridian48 take
The pitch is sensible—unifying test and eval tooling reduces friction—but adoption depends on whether the typed-fixture ergonomics outweigh the ecosystem gravity of pytest.
python-testingllm-evals