ksachdeva/langchain-graphrag

Implements the GraphRAG paper with LangChain integration, enabling modular swapping of LLM providers (OpenAI, Azure, Ollama), embeddings, and vector stores without core logic changes. Builds knowledge graphs through text unit and entity extraction, then performs hierarchical retrieval via local and global search using context selection and prompt-based summarization. Designed for experimental extensibility—nearly all components can be replaced with custom implementations matching required interfaces.

164 stars.