HKUDS/RAG-Anything

Project overview

RAG-Anything is a Python library that processes and queries multimodal document content, combining vector similarity search and graph traversal to handle text, images, tables, and equations within a single framework.

Project type

RAG · Vector Database · Document Processing

Use cases

Documents & Office · Knowledge Q&A

Deployment

Refer to project documentation

License

MIT

Best for

• Researchers, developers, and enterprise teams who need to process documents containing diverse multimodal content and retrieve answers using a combined vector and graph approach.

Key capabilities

• Provides a workflow from document ingestion and parsing through multimodal query answering.
• Processes PDFs, Office documents, images, and other supported file formats.
• Analyzes images, tables, mathematical equations, and heterogeneous content with dedicated processors.
• Combines vector similarity search and graph traversal for multimodal content retrieval.
• Accepts externally pre-parsed content lists instead of requiring document parsing.

Limitations and risks

• Processing Office documents requires a separate LibreOffice installation.
• The documented end-to-end example requires an OpenAI-compatible API key and base URL. It configures gpt-4o-mini as the LLM model, gpt-4o for image processing, and text-embedding-3-large for embeddings.
• GPU requirements, operating-system requirements, database requirements, data boundaries, cost dependencies, and telemetry are not documented.

Getting started

• Basic installation requires running a single pip command, though full document-format support may require optional Python extras and a separate LibreOffice installation. Alternatively, users can install uv, clone the repository, and synchronize dependencies.

Evidence and sources

• README: from raganything import RAGAnything, RAGAnythingConfig
• README: Modern documents increasingly contain diverse multimodal content—text, images, tables, equations, charts, and multimedia—that traditional text-focused RAG systems cannot effective…
• README: As a unified solution, RAG-Anything **eliminates the need for multiple specialized tools**. It provides **seamless processing and querying across all content modalities** within a…
• Release: v1.3.1
• README: - **🔄 End-to-End Multimodal Pipeline** - Complete workflow from document ingestion and parsing to intelligent multimodal query answering

Reviewed at: 2026-07-17T11:21:55Z