RAG-Anything

About

RAG-Anything is an all-in-one multimodal RAG framework developed by HKUDS, built upon LightRAG. It addresses the limitations of traditional RAG systems in processing diverse content like text, images, tables, and equations. This comprehensive system offers end-to-end multimodal document processing, enabling seamless retrieval and querying across all content modalities through intelligent parsing, multimodal content understanding, and knowledge graph construction, eliminating the need for multiple specialized tools, including VLM-Enhanced Query capabilities.

Features

• End-to-End Multimodal Processing Pipeline
• Universal Document & Content Type Support
• Multimodal Knowledge Graph Construction
• Adaptive Content Decomposition & Specialized Analyzers
• Hybrid Intelligent Retrieval & VLM-Enhanced Query

Supported Platforms