mini-eval

mini-eval is a small, dependency light library for evaluating systems built on top of large language models. It gives you a simple way to define test cases, run them against your prompts or agents, and score the results so you can see whether a change made things better or worse.

The goal is to keep evaluation close to the code. You write evals as plain functions, run them locally or in CI, and get readable output without setting up a heavy framework.

You can find the source, examples, and installation instructions on GitHub.