Self-Improving Agentic RAG Released as OSS, Updating Its Own Procedures via Multi-Dimensional Evaluation

Developer Fareed Khan has published a reference implementation of a "self-improving Agentic RAG" on GitHub that combines a team of specialist agents with multi-dimensional evaluation to raise its own performance. Targeting the medical and healthcare domain, it demonstrates an end-to-end loop in which agents score their outputs across several axes, diagnose weaknesses, and rewrite the operating procedures themselves.**