Awesome Infra for AI › Workflow Orchestration for AI

chatchat-space/LangGraph-Chatchat

⭐ 194 Python repository created 2024-09-30

LangGraph-Chatchat is an open-source project focused on developing a Retrieval-Augmented Generation (RAG) and Agent application. It leverages the LangGraph framework to implement local knowledge-based Q&A functionalities, specifically optimized for Chinese scenarios and open-source models, enabling offline deployment. The project integrates with popular open-source LLMs like ChatGLM, GLM-4-Chat, Qwen2-Instruct, and Llama3, as well as embedding models and vector databases. It supports frameworks such as Xinference and Ollama for model integration and uses Streamlit for its WebUI. The core principle involves loading and processing documents into a knowledge base, vectorizing text, performing similarity searches for user queries, and using the retrieved context to enhance LLM responses. Unlike its predecessor, LangChain-Chatchat, this project emphasizes LangGraph for building entire conversation flows, including both agentic and RAG components, necessitating LLMs with function-calling capabilities. It offers a more customizable and observable agent development experience through LangGraph's state machine capabilities and potential integration with LangSmith. The project also centralizes historical message management in `sqlite` or `PostgreSQL`, improving developer convenience and maintainability. It provides registrars for graphs and tools, along with demo implementations to guide developers in creating custom graph-based agents.

https://github.com/chatchat-space/LangGraph-Chatchat

agentagentsfastapilangchainlangchain-agentlangchain-agentslangchain-pythonlanggraphlanggraph-pythonllmllm-opsRAGknowledge baseopen-source LLMvector databasefunction calling