RuoYi AI is an MIT-licensed, Java-based enterprise AI development framework for building AI agents, aimed at developers and platform teams that need multi-provider LLM management, private knowledge bases, visual workflow orchestration and multi-agent coordination in a single self-hostable stack.

What it is

RuoYi AI is a full-stack AI assistant platform built around a Spring Boot 3.5.8 and Langchain4j backend, a Vue 3 frontend (Vben Admin and element-plus-x), MySQL 8.0 and Redis for storage, and a choice of Milvus, Weaviate or Qdrant as the vector database. It bundles five concerns into one system: model management, knowledge management, tool management, visual workflow orchestration and multi-agent execution. Document processing covers PDF, Word and Excel parsing with intelligent image analysis, and the communication layer uses WebSocket plus SSE streaming. It lives in the Java and Spring Boot ecosystem, which is unusual for this class of tool, and it is compatible with mainstream Agent Skill standards.

The concrete problem it solves is the assembly tax of building an enterprise AI assistant from separate parts. Instead of stitching together a model gateway, a standalone RAG pipeline, a workflow engine and hand-written orchestration code, teams configure all of that behind one admin panel. Model management reaches DeepSeek, Zhipu, MIMO, Bailian and OpenAI directly, and can also front Coze, DIFY, FastGPT or RAGFlow as upstream platforms, so an existing retrieval or agent service does not have to be thrown away. The distributed module set — backend, user frontend, admin panel, drama, copilot and a uniapp mini-app — replaces the collection of unrelated repositories that such a platform usually grows into.

Key capabilities

• Unified model management across DeepSeek, Zhipu, MIMO, Bailian and OpenAI, with multi-modal understanding and integration with Coze, DIFY, FastGPT and RAGFlow.
• Knowledge management combining local RAG with Milvus, Weaviate or Qdrant vector databases, plus document parsing for PDF, Word and Excel with intelligent image analysis.
• Tool management over the MCP protocol, with a Skills capability and an extensible tool ecosystem.
• Visual workflow designer with drag-and-drop node orchestration and SSE streaming execution; documented nodes include model calls, email sending and manual review.
• Multi-agent framework based on Langchain4j with Supervisor mode orchestration and support for multiple decision models, combinable with tools and skills.
• Security and observability layers: Sa-Token plus JWT dual-layer authentication, comprehensive logging, performance monitoring and service health checks.

Who uses it and how

• Enterprise teams building private AI assistants over internal document corpora, keeping data in their own MySQL 8.0, Redis and vector database deployment.
• Teams already standardised on Java and Spring Boot, who want the backend on Spring Boot 3.5.8 with Langchain4j and the frontend on Vue 3, Vben Admin and element-plus-x rather than a Python stack.
• Workflow builders who need human-in-the-loop steps, using the manual review node alongside the email sending node for notifications.
• Organisations that route between Chinese model providers such as DeepSeek, Zhipu, MIMO and Bailian and OpenAI through one configuration surface.
• Multi-surface deployments that need more than a chat window, using the separate repositories for backend, user frontend, admin panel, drama, copilot and the uniapp mini-app.

Getting started

The README documents Docker deployment in two variants, the first being a one-click start method. A hosted alternative exists for evaluation: an admin panel at http://129.226.199.247:25666 and a user frontend at http://129.226.199.247:25137, both reachable with the default admin / admin123 account, with documentation at https://doc.ruoyiai.chat.

How it compares

The facts do not include a list of paid products that RuoYi AI replaces, and they name no direct peer in this registry, so on that axis it stands alone here. Coze, DIFY, FastGPT and RAGFlow appear only as platforms RuoYi AI integrates with through model management, not as alternatives it is positioned against. Its distinguishing traits against anything comparable are the MIT licence, the Java and Spring Boot implementation and full self-hosting.

When to use it — and when not to

A self-hoster must operate MySQL 8.0, Redis and one of the supported vector databases, and should expect to run several front-end modules rather than a single binary. Teams that do not want to manage a JVM service, a relational database and a vector store, or that need a Python-native stack, are better served elsewhere. The facts also describe a separate commercial edition at https://web.ruoyiai.chat with WeChat QR code login, so groups outside China should confirm which capabilities live only there before committing to the open-source track.

# Clone the v3.1.0 release
git clone --depth 1 --branch v3.1.0 https://github.com/ageerle/ruoyi-ai.git
cd ruoyi-ai

# Pin the image version. Public GHCR images do not require docker login.
cp docs/docker/ruoyi-ai/.env.example docs/docker/ruoyi-ai/.env
sed -i 's/^RUIYI_VERSION=.*/RUIYI_VERSION=v3.1.0/' docs/docker/ruoyi-ai/.env

# Pull pre-built images from GHCR and start all services
docker compose --env-file docs/docker/ruoyi-ai/.env \
  -f docs/docker/ruoyi-ai/docker-compose-all.yaml pull
docker compose --env-file docs/docker/ruoyi-ai/.env \
  -f docs/docker/ruoyi-ai/docker-compose-all.yaml up -d

# Check service status
docker compose --env-file docs/docker/ruoyi-ai/.env \
  -f docs/docker/ruoyi-ai/docker-compose-all.yaml ps

# Access services (replace SERVER_IP with the server address)
# Admin Panel: http://SERVER_IP:25666 (admin / admin123)
# User Frontend: http://SERVER_IP:25137
# Backend API: http://SERVER_IP:26039

# Clone the v3.1.0 release
git clone --depth 1 --branch v3.1.0 https://github.com/ageerle/ruoyi-ai.git
cd ruoyi-ai

# Create the environment configuration file
Copy-Item docs\docker\ruoyi-ai\.env.example docs\docker\ruoyi-ai\.env

# Pin the image version
(Get-Content docs\docker\ruoyi-ai\.env) `
  -replace '^RUIYI_VERSION=.*', 'RUIYI_VERSION=v3.1.0' |
  Set-Content docs\docker\ruoyi-ai\.env

# Check the environment configuration
Get-Content docs\docker\ruoyi-ai\.env

# Pull pre-built images from GHCR and start all services
docker compose --env-file docs\docker\ruoyi-ai\.env `
  -f docs\docker\ruoyi-ai\docker-compose-all.yaml pull
docker compose --env-file docs\docker\ruoyi-ai\.env `
  -f docs\docker\ruoyi-ai\docker-compose-all.yaml up -d

# Check service status
docker compose --env-file docs\docker\ruoyi-ai\.env `
  -f docs\docker\ruoyi-ai\docker-compose-all.yaml ps

# Enter backend project directory
cd ruoyi-ai

# Start backend service (build from source)
docker-compose up -d --build

# Wait for backend service to start
docker-compose logs -f backend

# Enter admin panel project directory
cd ruoyi-admin

# Build and start admin panel
docker-compose up -d --build

# Access admin panel
# URL: http://localhost:5666

# Enter user frontend project directory
cd ruoyi-web

# Build and start user frontend
docker-compose up -d --build

# Access user frontend
# URL: http://localhost:5137