Last released Nov 12, 2026
A self-hosted eval harness for RAG and LLM systems: known-answer questions, a yes/no judge, error bars, and a static report.