Cette page est affichée en anglais. Une traduction française est en cours.
ScriptsMar 31, 2026·2 min de lecture

RAGFlow — Deep Document Understanding RAG Engine

Open-source RAG engine with deep document understanding. Parses complex PDFs, tables, images. Agent-powered Q&A with citations. Multi-model. 77K+ stars.

Introduction

RAGFlow is a leading open-source RAG engine that combines deep document understanding with agent capabilities. Unlike basic RAG that chunks text blindly, RAGFlow uses vision models to understand document layouts — tables, images, charts, multi-column text, and complex formatting. Provides accurate, cited answers with full source traceability. 77,000+ GitHub stars, Apache 2.0.

Best for: Enterprise document Q&A, knowledge bases with complex PDFs, legal/financial document analysis Works with: OpenAI, Anthropic, Ollama, Azure, any OpenAI-compatible API


Key Features

Deep Document Parsing

Vision-based layout analysis that understands:

  • Complex tables (merged cells, nested headers)
  • Embedded images and charts
  • Multi-column layouts
  • Headers, footers, page numbers
  • Mathematical formulas

Accurate Citations

Every answer includes exact source references — page number, paragraph, and highlighted text.

Agent Workflows

Built-in agent for multi-step reasoning, web search augmentation, and tool use.

Template-Based Chunking

Document-type-aware chunking strategies — different templates for papers, resumes, contracts, manuals.

Multi-Model Support

OpenAI, Anthropic, Google, Ollama, Azure, DeepSeek, Zhipu, and custom endpoints.

Knowledge Base Management

Web UI for uploading, organizing, and managing document collections across teams.


FAQ

Q: What is RAGFlow? A: An open-source RAG engine with deep document understanding. Parses complex PDFs with tables, images, and charts for accurate Q&A with citations. 77K+ GitHub stars.

Q: How is RAGFlow different from basic RAG? A: Basic RAG splits text into chunks by character count. RAGFlow uses vision models to understand document structure — tables remain as tables, images are captioned, and layouts are preserved.


🙏

Source et remerciements

Created by InfiniFlow. Licensed under Apache 2.0. infiniflow/ragflow — 77,000+ GitHub stars

Discussion

Connectez-vous pour rejoindre la discussion.
Aucun commentaire pour l'instant. Soyez le premier à partager votre avis.

Actifs similaires