Skip to content
repository radar
RAG & Memory

📊RAG-Anything

ACTIVE BREAKOUT

HKUDS/RAG-Anything · homepage ↗

"RAG-Anything: All-in-One RAG Framework"

⭐ Very popular: 23k stars, gaining about 147 a week

View on GitHub ↗

repo profile

vintage 2025 1 year old
delivery library
language Python
license MIT

momentum

total stars 23k
stars added last week +147
weekly downloads 7k /wk · PyPI
commits / week 4 accelerating · +325% vs prior mo
issues closed 35% ~16d to close, median
contributors 60 (+5 in 15d)
release cadence monthly
last activity 2d ago

durability

backing foundation-backed HKU Data Intelligence Lab
openness permissive
bus factor 3 concentrated
top-author share 24% 6 mo

bus factor = how many people it takes to cover more than half the commits (6 months). 1 is a solo project; higher means the work is spread across a team. top-author share is the single busiest author's slice of those commits.

since we covered it

monthly average + 1k/mo (+20%/mo) · + 16k total since PR#18

why it's a big deal

  • Lets teams run RAG over mixed-content documents, PDFs, Office files, and images containing text, tables, equations, and charts, without stitching together separate pipelines.
  • Targets developers whose source material is not plain text, handling scanned tables, LaTeX equations, and image captions that text-only RAG systems drop.
  • Ships as a pip-installable Python package, so existing LightRAG users can extend to multimodal retrieval without rebuilding their stack.

under the hood

  • Runs a five-stage pipeline in Python: parsing, content understanding, multimodal analysis, knowledge graph construction, and modality-aware retrieval, built on top of LightRAG.
  • Parses documents with MinerU by default, with Docling and PaddleOCR as alternatives, and routes images, tables, and equations to dedicated analyzers including vision language models for captioning.
  • Retrieval combines vector similarity with knowledge graph traversal and supports text, VLM-enhanced, and multimodal query modes, with direct content-list insertion to skip parsing for pre-processed input.

Radar summary, generated from the project's public sources

star history

PR#18 · 7k23k now Jun 2025Aug 2026
  1. PR#18 7k 2025-10-01
  2. now 23k + 16k since first covered

curve is sampled from GitHub's star history, plus our own daily readings since we covered it; the dashed stretch is before we first covered it, the solid line since. figures at coverage are the numbers we printed then (approx.), current count is live.

understory

Better known than its recent output, coasting a little on attention.

-20 understory score output 51 · clout 71
Aug 2025 Jul 2026
  • output, commits & releases
  • clout, star velocity

output = commits & releases; clout = star velocity, both 0 to 100 monthly indices; the gap where output runs above clout is the understory. The understory →

covered in

  • PR#18 2025-10-01 below the radar

    All-in-One Multimodal RAG System

similar projects

compare these →
  • 📑 PageIndex

    📑 PageIndex: Document Index for Vectorless, Reasoning-based RAG

    35k ACTIVE
  • 🔮 STORM

    An LLM-powered knowledge curation system that researches a topic and generates a full-length report with citations.

    31k ACTIVE
  • ⚙️ llmware

    Unified framework for building enterprise RAG pipelines with small, specialized models

    15k ACTIVE

comments

Sign in with GitHub to add your blip on RAG-Anything.

loading comments…