Hiveflow
Visual workflow orchestration for AI agents and automation
RAGFlow
RAG engine with deep document understanding and agents
RAGFlow is an open-source RAG engine combining retrieval-augmented generation with agent capabilities, offering deep document understanding, template-based chunking, and automated workflows for production AI systems.
Overview
What is RAGFlow?
RAGFlow is a Retrieval-Augmented Generation (RAG) engine that fuses advanced RAG techniques with agent capabilities to create a superior context layer for large language models. It offers a streamlined RAG workflow adaptable to enterprises of any scale, powered by a converged context engine and pre-built agent templates.
Core Capabilities
The platform excels at deep document understanding, extracting knowledge from unstructured data with complicated formats including Word, slides, Excel, images, scanned copies, structured data, and web pages. Its template-based chunking approach provides intelligent, explainable text segmentation with multiple template options. RAGFlow minimizes hallucinations through grounded citations, visualization of text chunking for human intervention, and traceable references.
Deployment & Integration
RAGFlow requires Docker and runs on systems with at least 4 CPU cores, 16 GB RAM, and 50 GB disk space. The platform supports both CPU and GPU acceleration for embedding and document processing tasks. It offers configurable LLMs and embedding models, multiple recall paired with fused re-ranking, and intuitive APIs for seamless business integration. Recent updates include support for GPT-5, agentic workflows, MCP, cross-language queries, and multi-modal image understanding.
Highlights
Deep document understanding extracts knowledge from complex unstructured formats with unlimited token capacity
Template-based chunking with visualization enables intelligent, explainable text segmentation and human intervention
Grounded citations with traceable references reduce hallucinations and support verifiable answers
Automated RAG workflow with configurable LLMs, embedding models, and fused re-ranking for enterprise scale
Pros
- Handles heterogeneous data sources including documents, images, scanned copies, and web pages
- Visual chunking interface allows human oversight and quality control
- Combines RAG with agent capabilities and supports agentic workflows with MCP
- Docker-based deployment simplifies installation and scaling for production environments
Considerations
- Requires substantial resources: minimum 16 GB RAM and 50 GB disk space
- Docker images built for x86 platforms only; ARM64 users must build custom images
- Slim Docker images exclude embedding models, requiring separate configuration
- gVisor dependency required for code executor sandbox feature adds complexity
Managed products teams compare with
When teams consider RAGFlow, these hosted platforms usually appear on the same shortlist.
Looking for a hosted option? These are the services engineering teams benchmark against before choosing open source.
Fit guide
Great for
- Enterprises needing production-ready RAG systems with deep document understanding
- Teams processing complex documents with mixed formats, tables, and images
- Organizations requiring traceable citations and explainable AI outputs
- Developers building agentic workflows with retrieval-augmented generation
Not ideal when
- Resource-constrained environments with less than 16 GB RAM or limited disk space
- ARM64 platforms without Docker build expertise or custom image requirements
- Projects requiring lightweight, serverless deployment without Docker infrastructure
- Simple question-answering use cases without complex document processing needs
How teams use it
Enterprise Document Intelligence
Extract structured knowledge from mixed-format corporate documents, scanned files, and presentations with grounded citations for compliance and audit trails
Multi-Modal Research Assistant
Process PDFs with embedded images using multi-modal models, combine with internet search for deep research capabilities across unlimited document tokens
Text-to-SQL Analytics
Transform natural language queries into SQL statements through RAG, enabling business users to query structured databases without technical expertise
Cross-Language Knowledge Retrieval
Query documents in multiple languages with automatic translation and context preservation, supporting global teams and multilingual content repositories
Tech snapshot
Python49%
TypeScript49%
Less1%
Shell1%
HTML1%
CSS1%
Frequently asked questions
What are the minimum system requirements to run RAGFlow?
RAGFlow requires at least 4 CPU cores, 16 GB RAM, 50 GB disk space, Docker 24.0.0+, and Docker Compose v2.26.1+. GPU acceleration is optional for embedding and document processing tasks.
What is the difference between full and slim Docker images?
Full images (~9 GB) include pre-built embedding models for immediate use. Slim images (~2 GB) exclude embedding models, requiring separate configuration but offering faster downloads and smaller footprint.
Does RAGFlow support ARM64 platforms like Apple Silicon?
Docker images are built for x86 platforms only. ARM64 users must build custom Docker images following the provided guide in the repository documentation.
How does RAGFlow reduce hallucinations in AI responses?
RAGFlow provides visualization of text chunking for human intervention, displays key references with traceable citations, and uses grounded retrieval to ensure answers are backed by source documents.
What document formats does RAGFlow support?
RAGFlow supports Word, PowerPoint, Excel, TXT, images, scanned copies, structured data, web pages, PDFs, and DOCX files with embedded images processed through multi-modal models.
Project at a glance
Active- Stars
- 71,867
- Watchers
- 71,867
- Forks
- 7,907
LicenseApache-2.0
Repo age2 years old
Last commit2 hours ago
Primary languagePython
Last synced 49 minutes ago