Giskard

About

Giskard is an open-source Python library designed to enhance transparency and accountability in AI model testing and evaluation. It provides a comprehensive framework for detecting performance, bias, and security issues in AI applications, supporting models ranging from traditional machine learning to Large Language Models (LLMs). Giskard's key features include automated vulnerability detection, which identifies issues such as hallucinations, prompt injection, and discrimination. It also offers tools like the Retrieval Augmented Generation Evaluation Toolkit (RAGET) for assessing RAG applications. Giskard integrates seamlessly with CI/CD workflows, enabling continuous testing and monitoring of models. It allows for customized test suites and advanced scan configurations, focusing on specific aspects of a model. The library promotes fairness, robustness, and explainability in AI systems, providing insights into model predictions to enhance transparency and trust. By automating testing processes, Giskard helps save time and reduce AI risks, making it a valuable tool for ensuring the quality and reliability of AI models across various industries.