InternLM/agentlego
AgentLego provides a library of versatile tool APIs for extending and enhancing LLM-based agents with multimodal capabilities.

AgentLego is an open-source library of tool APIs designed to extend large language model agents. It offers tools for visual perception, image generation and editing, speech processing, and visual-language reasoning. The library provides a flexible interface allowing users to define custom tools with arbitrary arguments and outputs. It integrates with existing agent frameworks including LangChain, Transformers Agents, and Lagent, and supports remote tool serving for models requiring GPU environments.