LightRAG review

Q: How do I install LightRAG?

Run uv tool install "lightrag-hku[api]". The lightrag-hku package needs Python 3.10 or newer. The install section above has the other options from the project's README.

LightRAG is an open-source Python framework for retrieval-augmented generation that combines knowledge graphs with vector embeddings for indexing and retrieval.

Updated Nov 1, 2026. Review written Oct 22, 2026 from the project's own sources; usage data refreshed every Monday.

TL;DR

LightRAG is a lightweight knowledge-graph RAG framework from the Data Intelligence Lab at HKU, described in an EMNLP 2025 paper. It is for developers building RAG over large or changing document sets. LightRAG uses dual-level retrieval over graphs and vectors and supports incremental updates and document deletion without full rebuilds.

40.5kGitHub stars
+166new stars, week of Oct 25
35.6kdownloads a week, +7% vs 4-week avg
#26in RAG, data and memory this week

LightRAG at a glance

Category: RAG, data and memory
Maintained by: ✨Data Intelligence Lab@HKU✨ (organization, @HKUDS on GitHub)
License: MIT
Language: Python, TypeScript, Shell
Repository: github.com/HKUDS/LightRAG
Website: arxiv.org/abs/2410.05779
PyPI package: lightrag-hku 1.5.7
Latest release: v1.5.7 on GitHub, Sep 2, 2026
First published: Oct 2, 2024

What LightRAG does

LightRAG builds a knowledge graph and vector embeddings from documents, then retrieves both detailed facts and abstract concepts through dual-level retrieval.
LightRAG supports incremental updates and document deletion, using the indexing LLM cache to rebuild affected entities and relationships.
LightRAG parses documents with MinerU, Docling, or a Native engine that handles images, tables, and formulas in Word and Markdown files.
LightRAG offers four text chunking strategies: Fixed-length, Recursive character, Vector semantic, and Paragraph semantic.
LightRAG includes a WebUI to insert, query, and visualize knowledge, plus a query-only /workspace entry for end users.
LightRAG supports citations, reranking, RAGAS evaluation, and Langfuse tracing.

How it works

LightRAG runs as a Python server, lightrag-server, configured through a .env file with your LLM and embedding settings. You can set separate LLMs for extraction, querying, keywords, and vision. The server provides a WebUI, an API, and Ollama-compatible /api/* routes. LightRAG stores data in backends such as PostgreSQL, MongoDB, Neo4J, or OpenSearch, and can run under Docker Compose.

How to install and start using LightRAG

With uv

uv tool install "lightrag-hku[api]"

With pip

pip install "lightrag-hku[api]"

With Docker Compose

docker compose up

Requirements

Python 3.10 or later
Access to an LLM and an embedding model, configured in .env
The system libcairo library to rasterize SVG images in Markdown documents
spaCy language models, but only if you enable smart_heading for docx files

First steps

Copy the env file with cp env.example .env, or run make env-base to use the interactive setup wizard.
Add your LLM and embedding configuration to the .env file.
Configure authentication with LIGHTRAG_API_KEY or AUTH_ACCOUNTS and TOKEN_SECRET before exposing the server on a network.
Start the server with lightrag-server, then insert and query documents through the WebUI.

Who LightRAG is for

Developers who need graph-based RAG over cross-document or domain material, such as legal or financial text
Teams with document collections that change and need incremental updates or deletions
Teams running open-source LLMs locally or in air-gapped environments
Projects that want one PostgreSQL database for all four storage types

Limitations to know

LightRAG binds to all interfaces by default, and every endpoint is public unless you configure authentication.
LightRAG leaves the Ollama-compatible /api/* routes open by default unless you set WHITELIST_PATHS.
LightRAG skips SVG images in Markdown documents if libcairo is missing, because pip and uv cannot install it.
LightRAG needs LLM calls during indexing as well as querying, though the project says it reduces how many it makes.

Is LightRAG growing? Usage data

In the week of Oct 25–31, 2026, LightRAG ranked #237 on the Devtoolwatch Board and #26 in RAG, data and memory, with a momentum score of 39 out of 100. Its main package had 35.6k downloads that week, +7% against its own 4-week average.

New GitHub stars per week for the last 12 weeks, from GitHub's stargazer history (today's stargazers by the week they starred).

Show the weekly numbers

New GitHub stars per week for LightRAG
Week of	New stars
Aug 9, 2026	246
Aug 16, 2026	235
Aug 23, 2026	176
Aug 30, 2026	178
Sep 6, 2026	190
Sep 13, 2026	175
Sep 20, 2026	117
Sep 27, 2026	109
Oct 4, 2026	115
Oct 11, 2026	158
Oct 18, 2026	144
Oct 25, 2026	166

On Hacker News

Successful room temperature ambient-pressure magnetic levitation of LK-99, 2408 points and 1199 comments (Aug 4, 2023)
The first room-temperature ambient-pressure superconductor?, 1690 points and 877 comments (Jul 25, 2023)
DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL, 1351 points and 1056 comments (Jan 25, 2025)
ArXiv now offers papers in HTML format, 1204 points and 312 comments (Dec 21, 2023)
The Era of 1-bit LLMs: ternary parameters for cost-effective computing, 1040 points and 447 comments (Feb 28, 2024)

The Board ranks tools by four signals each week: new stars, new stars for the repo's size, added downloads and download growth beyond the market. How the ranking works.

LightRAG alternatives

Other RAG, data and memory tools on the Board in the week of Oct 25–31, 2026, by momentum:

RAG, data and memory ranked by momentum, the week of Oct 25–31, 2026
Rank	Tool	Stars	Downloads/wk
#1	XERJXERJ is a local-first search engine and MCP server for AI agents, packaged as one Rust binary with an Elasticsearch-compatible API.	17.9k	–
#2	MemvidMemvid is a single-file memory layer for AI agents that packages data, embeddings, search indexes, and metadata into one portable .mv2 file.	17.3k	–
#3	LEANNLEANN is a Python vector database and RAG library that recomputes embeddings on demand to keep indexes small on personal devices.	13.7k	–
#4	GraphifyGraphify is an AI coding assistant skill that turns a codebase, with its docs, PDFs, images and videos, into a queryable knowledge graph.	140k	480.7k
#5	MarkItDownMarkItDown is a lightweight Python utility from Microsoft that converts PDFs, Office documents, images, audio and other files to…	193.2k	4.7M
#6	FirecrawlFirecrawl is an API and open source project to search, scrape, crawl, and interact with web pages, returning LLM-ready markdown or JSON…	197.4k	2.9M

All RAG, data and memory tools, ranked

Recent LightRAG releases

v1.5.7, Sep 2, 2026
v1.5.7rc2, Aug 19, 2026
v1.5.7rc1 (pre-release), Aug 17, 2026
v1.5.6, Aug 6, 2026
v1.5.5, Jul 31, 2026

LightRAG FAQ

Is LightRAG free and open source?

Yes. LightRAG's source code is on GitHub (HKUDS/LightRAG) under the MIT license, so you can use it for free.

How do I install LightRAG?

Run uv tool install "lightrag-hku[api]". The lightrag-hku package needs Python 3.10 or newer. The install section above has the other options from the project's README.

Is LightRAG still maintained?

Yes. The HKUDS/LightRAG repository was last updated on Oct 20, 2026, and the latest release, v1.5.7 on GitHub, came out on Sep 2, 2026.

How popular is LightRAG?

LightRAG has 40.5k GitHub stars as of Nov 1, 2026. It gained 166 new stars in the week of Oct 25–31, 2026, against a 12-week average of 167 a week. Its main package had 35.6k downloads that week, +7% against its own 4-week average. That made it #26 among RAG, data and memory tools on the Devtoolwatch Board.

LightRAG vs GraphRAG: how do they compare?

LightRAG and GraphRAG are both tracked by Devtoolwatch as RAG, data and memory tools. LightRAG has 40.5k GitHub stars and GraphRAG has 36.5k. In the week of Oct 25–31, 2026, LightRAG's main package had 35.6k downloads and GraphRAG's had 11.1k. GraphRAG is a Microsoft Research Python pipeline that extracts knowledge graphs from unstructured text to improve LLM question answering over private data. GraphRAG review.

LightRAG vs Graphify: how do they compare?

LightRAG and Graphify are both tracked by Devtoolwatch as RAG, data and memory tools. LightRAG has 40.5k GitHub stars and Graphify has 140k. In the week of Oct 25–31, 2026, LightRAG's main package had 35.6k downloads and Graphify's had 480.7k. Graphify is an AI coding assistant skill that turns a codebase, with its docs, PDFs, images and videos, into a queryable knowledge graph. Graphify review.

What are the best alternatives to LightRAG?

Other RAG, data and memory tools on the Devtoolwatch Board in the week of Oct 25–31, 2026, by momentum: XERJ (#1), Memvid (#2), LEANN (#3), Graphify (#4), MarkItDown (#5). The alternatives table above has their numbers.

What is LightRAG used for?

LightRAG is used for retrieval-augmented generation over document collections. It indexes documents into a knowledge graph plus vector embeddings so an LLM can answer complex, cross-document queries with citations.

How does LightRAG work?

LightRAG adds graph structures to text indexing and retrieval. It extracts entities and relationships with an LLM and stores them alongside vectors. A dual-level retrieval system then combines low-level facts with high-level concepts. An incremental update algorithm adds new data without rebuilding the index.

Is LightRAG a GraphRAG?

Yes, LightRAG is a graph-based RAG framework. The project describes it as an efficient alternative to Microsoft GraphRAG that does not rely on community reports or multi-hop reasoning for complex queries.

Is LightRAG slow?

LightRAG is designed to reduce LLM calls during indexing and querying to lower latency. The project also removed processing bottlenecks for large datasets. Its v1.5.7 release candidates added graph-first ingestion with deferred vector indexing.

Can LightRAG process PDFs and images?

Yes. LightRAG parses multimodal content through MinerU or Docling services, which come from its merged RAG-Anything integration. Its Native engine parses images, tables, and formulas in Word and Markdown documents.

Sources

LightRAG README badge

Maintainers can show the tool's weekly Board rank in their README:

[![Devtoolwatch](https://index.devtoolwatch.com/badge/lightrag.svg)](https://index.devtoolwatch.com/tools/lightrag/)

How this page is made: Devtoolwatch writes each review from the project's README, releases, package metadata and website, with an AI model, then checks every claim against the source it cites and drops what the sources don't support. The numbers come from Devtoolwatch's daily GitHub snapshots, GitHub's stargazer history, and the npm and PyPI download APIs. Spotted an error? Email [email protected].