All projects

groundcheck

A specialized evaluation framework built to measure the reliability, accuracy, and traceability of Retrieval-Augmented Generation systems. Groundcheck evaluates retrieval performance using metrics such as Recall@K, Mean Reciprocal Rank, and Precision@K, while independently verifying numerical accuracy, citation correctness, document revisions, and unsupported claims. Unlike evaluation systems that depend on another language model to judge responses, Groundcheck uses deterministic programmatic checks to produce reproducible results. Its benchmark includes negative-control questions to measure whether an AI system can recognize when the available documents do not contain an answer. Built in Python with 179 automated tests and a minimal dependency footprint, it integrates directly into CI pipelines and blocks deployments when evaluation metrics regress.

Python · Evals · RAG · CI gate · CLI