datawhalechina/llm-universe

Covers unified API wrappers for major domestic and international LLM providers (GPT, Baidu Wenxin, iFlytek Spark, Zhipu GLM) alongside LangChain integration, enabling consistent multi-model invocation without API-specific implementation details. Teaches RAG architecture through a practical personal knowledge base assistant project, combining document loading/chunking, vector database construction with embedding APIs, and Streamlit deployment—all executable on standard hardware without GPU requirements. Structured in three progressive tracks: foundational LLM application development, advanced RAG optimization techniques (hybrid retrieval, prompt engineering, fine-tuning), and open-source project case studies.

12,159 stars.