Plexiglass

Plexiglass is an innovative toolbox designed to test vulnerabilities and safeguard Language Learning Models (LLMs), developed by SafeLlama. This Python-based security toolbox operates with a focus on the rigorous testing and enhancement of security measures for LLMs.

Key Features

• Security-focused LLM Toolbox: Plexiglass is crafted as a security toolbox specifically for testing and safeguarding LLMs, making it an essential tool for developers working with LLMs.

• Versatile Testing Modes: It offers two primary modes of operation: a command line interface (CLI) or a standalone python module. This versatility allows developers to choose the mode that best suits their needs.

Inspiration and Compatibility

• The development of Plexiglass is motivated by the increasing importance of security in machine learning and AI. It aligns with the current trend of creating robust and secure AI systems, especially in the realm of language learning models.

• Plexiglass uses litellm under the hood, suggesting compatibility with a range of LLMs and providing a versatile tool for developers and researchers in the field.

Community Involvement

• SafeLlama emphasizes the importance of community involvement in Plexiglass’s development. They are actively seeking contributors to further enrich the toolbox and extend its capabilities.

Install

• Build and Install

Usage

• Quick Start Guide

Further Reading

• Code of Conduct