HKUDS/LightRAG

[EMNLP2025] LightRAG: Simple and Fast Retrieval-Augmented Generation

What it solves

LightRAG is a lightweight, graph-based Retrieval-Augmented Generation (RAG) framework designed to be a more efficient alternative to Microsoft GraphRAG. It solves the problems of heavy computational overhead, slow response times, and high costs associated with incremental updates in large-scale graph indexing and retrieval, while overcoming the fragmented context limitations of traditional chunk-based vector RAG.

How it works

LightRAG uses a dual-layer architecture that manages both knowledge graphs (KGs) and vector embeddings. This allows the system to perform dual-level retrieval, integrating detailed facts and abstract concepts simultaneously. It avoids inefficient multi-hop reasoning or community reports for complex queries, which reduces the number of LLM calls and lowers latency. For data ingestion, it supports multiple parsing engines (MinerU, Docling, Native) and four chunking strategies (Fixed, Recursive, Vector, and Paragraph semantic) to maintain document structure.

Who it’s for

It is designed for developers and organizations needing high-quality RAG in vertical domains (such as legal or financial) that require global comprehension, logical reasoning, and the ability to handle complex, cross-document queries across large-scale datasets.

Highlights

  • Dual-Layer Architecture: Combines knowledge graphs and vector embeddings for comprehensive and diverse retrieval.
  • Efficient Updates: Supports incremental updates and selective document deletion with automatic KG regeneration.
  • Multimodal Support: Integrates multimodal processing for images, formulas, and tables, connecting them to the body text via the knowledge graph.
  • Flexible Storage: Supports various production-ready backends including PostgreSQL, MongoDB, Neo4J, and OpenSearch.
  • Advanced Parsing: Includes a native engine that corrects section headings in Word documents and aligns chunk boundaries with semantic boundaries.

Related

  • Project
    FalkorDB/GraphRAG-SDK

    GraphRAG‑SDK is a Python SDK that builds a knowledge graph in FalkorDB from your documents and then answers questions by traversing that graph. It adds provenance edges, multi‑hop retrieval, and optional abstention to dramatically cut down LLM hallucinations, and it ships with benchmarks showing it outperforms standard vector‑RAG. Install via pip, run a FalkorDB container, ingest texts (or PDFs/markdown), call `finalize()`, and query with `completion()`. The library supports schemas, incremental updates, custom providers, and is Apache‑2.0 licensed.

  • Project
    raphaelmansuy/edgequake

    A high-performance Graph-RAG framework written in Rust that transforms documents into knowledge graphs to enable complex multi-hop reasoning and thematic retrieval.

  • Project
    hhy-huang/HiRAG

    HiRAG is a Retrieval-Augmented Generation framework that uses hierarchical knowledge to improve the quality and comprehensiveness of LLM responses compared to standard RAG.

  • Project
    Alibaba-NLP/VRAG

    A multi-turn multimodal RAG framework that uses agentic reinforcement learning to enable VLMs to reason and retrieve information across text, images, and videos.

  • Project
    HKUDS/RAG-Anything

    RAG‑Anything is a Python library that builds a multimodal Retrieval‑Augmented Generation system. It parses PDFs, Office files, images, tables, and equations using MinerU, creates a cross‑modal knowledge graph, and retrieves relevant pieces with a hybrid vector‑graph search before passing them to an LLM for answer generation. Install via `pip install raganything`, configure with `RAGAnythingConfig`, ingest documents, and query in natural language.