Welcome to Ragflow’s documentation!

Contents:

• About RAGFlow: Named Among GitHub’s Fastest-Growing Open Source Projects
  • The Rise of Retrieval-Augmented Generation in Production
  • Why RAGFlow Resonates in the AI Era
  • A Project in Active Development
  • Conclusion
• Security Concerns
  • considerations:
  • sandbox:
  • elasticsearch:
• RAGFlow System Architecture
  • Overview
  • Core Infrastructure Containers
  • RAGFlow Application Containers
  • High-Level Architecture Diagram
  • Data & Execution Flow
• From RAG to Context - A 2025 Year-End Review of RAG
  • Can RAG Still Be Improved?
• example:
  • Why Infiniflow RAGFlow Uses a Reranker, an Embedding Model, and a Chat Model
• Synergy of the Three Models
• Why vLLM is Used to Serve the Reranker Model
  • Key Reasons for Using vLLM to Serve the Reranker
  • Serving the Reranker Locally with vLLM
  • RAGFlow Integration
• Serving vLLM Reranker Using Docker (CPU-Only)
  • Docker Compose Configuration (CPU Mode)
  • Key Components Explained
  • Why the Model Must Be Pre-Downloaded Locally
  • Why CPU-Only (No GPU)?
  • Start the Service
  • Verify Availability
  • Integration with RAGFlow
  • Benefits of This CPU + Docker Setup
• Integrating vLLM with RAGFlow via Docker Network
  • Why Network Configuration is Required
  • Step-by-Step: Configure Docker Network
  • Architecture Diagram
  • Verification
  • Benefits of This Setup
  • Troubleshooting Tips
  • Summary
• Batch Processing and Metadata Management in Infiniflow RAGFlow
  • API Base URL
  • Authentication
  • Step 1: Retrieve Dataset and Document IDs
  • Step 2: Add Metadata to a Document (via PUT)
  • Use Case: Batch Metadata Enrichment
  • Other Batch-Capable Endpoints
  • Best Practices
  • See also:
  • Summary
• How the Knowledge Graph in Infiniflow/RAGFlow Works
  • Overview
  • Construction Process
  • Query and Retrieval Process
  • Key Features and Limitations
• Running Llama 3.1 with llama.cpp
  • 1. Model Format: GGUF
  • 2. Compile llama.cpp for Intel i7 (CPU-only)
  • 3. Run the Model with Web Interface
  • 4. Compile llama.cpp with NVIDIA GPU Support (CUDA)
  • Summary
• Running Multiple Models on llama.cpp Using Docker
  • Example docker-compose.yml
  • Key Configuration Notes
  • Usage
• Deploying LLMs in Hybrid Cloud: Why llama.cpp Wins for Us
  • 1. Current Setup: Ollama in Testing
  • 2. Production Requirements: Hybrid Cloud & Multi-User Access
  • 3. Evaluation: vLLM vs llama.cpp
  • 4. Why llama.cpp Is Our Production Choice
  • 5. Migration Path: From Ollama → llama.cpp
  • Summary
• How InfiniFlow RAGFlow Uses gVisor
  • gVisor Integration
  • How RAGFlow Uses gVisor Sandboxes
  • Security Advantages in Practice
  • Enabling/Disabling the Sandbox
  • Summary
• RAGFlow GPU vs CPU: Full Explanation (2025 Edition)
  • Why Does RAGFlow Still Need a GPU Even When Using Ollama?
  • Complete RAGFlow Pipeline (with GPU usage marked)
  • What DeepDoc Actually Does (and Why GPU Makes It 5–20× Faster)
  • Real-World Performance Numbers
  • When Do You Actually Need ragflow-gpu?
  • Recommended Setup in 2025 (Best of Both Worlds)
  • Monitoring & Verification
  • Conclusion
• Upgrade to latest release :
• upload document
  • response
• Graphrag
  • reponse (json)
  • graphrag
  • So slow, need to see progress
  • Stuck ???? what now????
• Chat
• Why Infinity is a Good Alternative in RAGFlow
  • Tailored for RAG/LLM Workloads
  • Performance & Efficiency
  • Integration in RAGFlow
  • Fully Open Source & Vendor-Neutral
  • Drawbacks
• MinerU and Its Use in RAGFlow
  • Introduction to MinerU
  • MinerU in RAGFlow
  • Comparison with Existing PDF Ingestion Tools
  • Summary of MinerU Advantages
  • Activating MinerU in RagFlow
• What is Agent context engine?
  • Beyond the hype: The reality of today’s “intelligent” Agents
  • Deconstructing the Agent Context Engine
  • Why we need a dedicated engine? The case for a unified substrate
  • RAGFlow: A resolute march toward the context engine of Agents
• Using SearXNG with RAGFlow
  • Introduction to SearXNG in RAGFlow Agents
  • Benefits of Using SearXNG with RAGFlow
  • Benefits for Intranet Page Browsing
  • Deployment in Docker Containers
  • Conclusion
• homelab
  • problem older hardware :
  • software :
  • problem reranking:
  • possible solution:

Indices and tables

• Index

• Module Index

• Search Page