LangDB

judgeval is an open-source tool designed for agent behavior monitoring. It facilitates tracking and judging agent behavior in both online and offline environments. The tool enables users to set up alerts and analyze agent behaviors and topic patterns at scale. judgeval supports agent post-training and monitoring.

power-pwn is an open-source security toolset designed for offensive and defensive security measures. It facilitates the discovery, reconnaissance, and ethical assessment of AI agents. The tool helps security professionals and AI developers identify and address potential vulnerabilities in AI systems. It is available on GitHub.

adversarial-attacks-pytorch is a PyTorch library that provides adversarial attacks for generating adversarial examples. It offers a PyTorch-like interface to simplify the implementation of these attacks. The library is designed to help PyTorch users test and improve the robustness of their models against adversarial threats.

Guardrails is an open-source toolkit for adding programmable guardrails to LLM-based conversational systems. It allows developers to easily implement safety measures and control the behavior of LLMs. The tool helps ensure that conversational AI systems are safe, reliable, and aligned with desired policies.

AIX360 is an open-source toolkit for interpretability and explainability of data and machine learning models. It helps users understand how AI models make decisions. The toolkit supports various methods for explaining AI models and datasets. It is available as a Python package.

ai.robots.txt is a curated list of AI agent and robot crawlers to block from accessing websites. It helps webmasters manage AI-related traffic. The list is open-source and encourages community contributions for identifying and implementing crawler blocking rules. It aids in controlling website access and resource usage.