Prior art
BehavTest stands on ideas from Promptfoo, DeepEval, Inspect AI, and Ragas, and on the pass@k / pass^k reliability framing from τ-bench. If you need a hosted platform, deep RAG metrics today, or production observability, those tools are excellent. Sourced, side-by-side comparisons: BehavTest vs Promptfoo, DeepEval, Ragas and LangSmith. BehavTest's bet is a small, vendor-neutral, statistically honest behavioral regression testing tool you can run anywhere.