graphrag reimplementation from scratch
knowledge graphs · retrieval · llm pipelines
A from-scratch reimplementation of the GraphRAG retrieval architecture, built to understand the system deeply.
GraphRAG (Graph Retrieval-Augmented Generation) extends standard RAG by building a knowledge graph from the source corpus and using community detection to answer questions that require synthesising across many documents.
The reimplementation focused initially on rebuilding the original graphrag capabilities implemented by Microsoft Research, and then iterating on the design by adding multimodal (specifically image) support.
Available on GitHub.