Tag: testing

Build an LLM Eval Harness That Runs in CI

Forty cases, a scorer, a baseline and a GitHub Action. How to catch AI regressions that unit tests miss, without paying for an eval platform.