giskard-oss

Giskard is an open-source Python library designed for testing and evaluating agentic systems, featuring a modular architecture and an async-first approach. Its primary use case involves dynamic, multi-turn testing of AI agents, enhancing capabilities for vulnerability scanning and RAG evaluation through its tools like `giskard-checks` and `giskard-scan`. Notably, Giskard v3 introduces improved efficiency by eliminating heavy dependencies and focuses on the evaluation of various agent types, including LLMs and multi-step pipelines.