langroid/langroid

What it solves

Langroid is a lightweight and extensible Python framework designed to simplify the development of LLM-powered applications. It addresses the complexity of building multi-agent systems where multiple AI agents must collaborate, exchange messages, and solve problems together using a principled approach inspired by the Actor Framework.

How it works

Developers set up Agents and equip them with optional components such as an LLM, a vector store, and specific tools or functions. These agents are then assigned to Tasks, which manage the interaction loops. Agents can collaborate by exchanging messages, and the framework supports a wide range of LLMs (including local and remote models), multimodal inputs (PDFs, images), and integration with MCP (Model Context Protocol) servers to extend agent capabilities via tool adapters.

Who it’s for

It is intended for developers and researchers building production-ready LLM applications, particularly those requiring multi-agent orchestration, structured information extraction, or RAG (Retrieval-Augmented Generation) systems.

Highlights

• Multi-Agent Orchestration: Inspired by the Actor Framework for collaborative problem solving.
• Broad LLM Support: Works with practically any LLM, including local models (e.g., Mistral) and remote APIs (e.g., OpenAI, Gemini, DeepSeek).
• Extensible Tooling: Supports function-calling, MCP tool adapters, and a variety of PDF parsers and vector databases (e.g., Qdrant, Milvus, Pinecone).
• Advanced RAG Capabilities: Includes DocChatAgent with Reciprocal Rank Fusion (RRF) for improved retrieval and structure-aware Markdown chunking.
• Developer Experience: Focuses on a simplified setup and flexible abstractions for Agents and Tasks.