LangChain-Chatchat

LangChain-Chatchat is an open-source RAG and Agent application with 38,000+ stars on GitHub that lets you chat with your own documents using local, open-weight language models — fully offline.

Overview

Originally released as Langchain-ChatGLM, Chatchat was one of the first projects to package the now-standard local knowledge-base recipe: load your files, split and embed the text, store it in a vector database such as FAISS or Milvus, retrieve the most relevant passages and let a local LLM answer with that context. Everything can run on open models — GLM, Qwen, Llama and others, served through Xinference or Ollama — so documents never leave your network, while OpenAI’s API is supported if you prefer it. It exposes the system two ways: a FastAPI service for programmatic access and a Streamlit WebUI for interactive use. The project has been especially influential among Chinese-speaking developers building private knowledge assistants, and it remains maintained, with the 0.3.x line current and updates into late 2025.

Key Features

• Local knowledge-base QA – the complete load, split, embed, retrieve, answer pipeline out of the box.
• Fully offline capable – run end to end on open models with no external API calls.
• Many model backends – GLM, Qwen, Llama and more via Xinference or Ollama, plus OpenAI API support.
• Vector store choice – FAISS, Milvus and other LangChain-supported stores.
• Two interfaces – FastAPI endpoints for integration and a Streamlit WebUI for people.
• Agent features – tool-using agent workflows on top of the knowledge base.
• Flexible deployment – pip install, source install or Docker images.

💳 Pricing

LangChain-Chatchat is free and open source under the Apache License 2.0. Running it with local models costs nothing beyond your hardware; using a hosted model API is optional.

Who Is It For

Teams and individuals — especially in Chinese-language environments — who want a private, document-grounded chatbot over internal files without sending data to a third-party service.