RAG in Action: Retrieval-Augmented Generation System Development
A Chinese-language course repository implementing a complete DeepSeek-based retrieval-augmented generation system, with modules for chunking, embedding, vector storage, retrieval and evaluation.
rag-in-action is the code repository for a Chinese-language course on building a retrieval-augmented generation system end to end, built around DeepSeek. It is organized into modules covering a basic RAG implementation, data loading, document chunking strategies, text embedding, vector database operations with Milvus and Chroma, pre-retrieval query expansion, index optimization, post-retrieval reranking and filtering, answer generation, system evaluation with RAGAS and TruLens, and advanced techniques such as graph RAG and multi-agent RAG.
It supports both LangChain and LlamaIndex, with separate GPU and CPU requirements files for Ubuntu, macOS and Windows through WSL2, and is accompanied by a published book covering the same material.