HKUDS/RAG-Anything

Health v2 · maintenance only

Adoption not part of the score

23081 stars · 2673 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

RAG-Anything is an all-in-one Python framework for multimodal Retrieval-Augmented Generation, built on LightRAG. It treats text, images, tables, and equations in documents as interconnected knowledge entities and uses dual-graph construction with cross-modal hybrid retrieval to answer queries over heterogeneous content.

Use cases

• build a RAG pipeline over multimodal documents with images and tables
• query PDFs containing figures, tables, and equations with an LLM
• index long documents for retrieval-augmented question answering
• extract and search knowledge across text and visual content
• add multimodal retrieval to a chatbot or knowledge base
• run graph-based RAG over research papers or reports

When to choose

• your documents mix text with images, tables, and math expressions
• you want a ready-made multimodal RAG framework based on LightRAG
• traditional text-only RAG fails on long or visually rich documents

When to avoid

• you only need simple text-only RAG with minimal dependencies
• you need a fully managed hosted RAG service rather than a Python framework
• your project is not Python-based

Facets

framework · maturity active

rag llm-inference parser search-engine machine-learning large-language-models artificial-intelligence pdf python cross-platform multimodal-rag knowledge-graph document-parsing lightrag dual-graph-retrieval retrieval-augmented-generation natural-language-processing

Member repositories

For agents

markdown · JSON · MCP: product_card(name="HKUDS/RAG-Anything")