Open-source testing framework for ML models — detect bias, hallucinations, and vulnerabilities before deployment.