AWS introduces open-source unified Knowledge Graph RAG for complex, relational queries

AWS has launched `unified-kg-rag-on-aws`, an open-source framework for knowledge-graph Retrieval-augmented generation (RAG) that combines Microsoft's GraphRAG and HKUDS's LightRAG methodologies. Built on Amazon Bedrock, Neptune, and OpenSearch Service, it enables AI models to reason over entity relationships, addressing the limitations of vector-only RAG for complex, multi-hop, and cross-document queries. This solution allows per-query methodology selection and supports incremental indexing and multilingual processing, beneficial for developers building advanced RAG systems for structured and relational data on AWS. It is particularly useful for use cases like compliance analysis where answers span multiple, interconnected documents.