RuoYi AI is a Java-based MCP server that integrates enterprise-grade AI capabilities through Spring AI, Langchain4j, and a modular AI tooling ecosystem. It provides an extensible platform for building AI-assisted workflows, knowledge graphs, and dynamic tool orchestration within an enterprise context. With real-time communication via WebSocket/SSE and a structured tool ecosystem, you can connect multiple AI models and platforms (such as FastGPT, Coze, and DIFY) and orchestrate them using knowledge graphs and AI processes. Typical use involves starting the Spring Boot application and using the unified chat interface or admin backend to configure AI tools, knowledge bases, and workflows, enabling a seamless, RAG-enabled assistant for business tasks.

After the server is up, you can interact with the AI through the web client or API endpoints. The MCP integration exposes capabilities for tool invocation, workflow orchestration, and intelligent routing based on the knowledge graph. You can customize model providers, vector stores, and local deployment options to suit data privacy and performance requirements, making it suitable for on-premises or private cloud deployments in enterprise environments.

Prerequisites:

• Java 17+ (JDK) installed
• Maven or Gradle build system (depending on project configuration)
• Optional: MySQL 8.x or other supported databases, Redis, and a vector store (Milvus/Weaviate/Qdrant) if using RAG features

Installation steps:

1. Clone the repository: git clone https://github.com/ageerle/ruoyi-ai.git cd ruoyi-ai

2. Build the project (choose the appropriate build tool used by the project, e.g., Maven): mvn clean package -DskipTests

   or if using Gradle:

   ./gradlew bootJar

3. Prepare configuration:

   • Ensure a database is available and accessible (update application.properties or application.yml with DB credentials).
   • Configure any required environment variables (JAVA_HOME, SPRING_PROFILES_ACTIVE, etc.).
   • If using local vector stores or external AI providers, ensure their endpoints/credentials are configured.

4. Run the application: java -jar target/ruoyi-ai-<version>.jar

   or via your IDE if using a run configuration

5. Verify startup:

   • Check logs for successful startup messages (Spring context initialized).
   • Access management and user interfaces at the configured URLs (e.g., http://localhost:8080/admin and http://localhost:8080/).

6. Optional: run with Docker (if Dockerfile is provided by the project): docker build -t ruoyi-ai:latest . docker run -p 8080:8080 -e JAVA_HOME=/path/to/java ruoyi-ai:latest

Notes and tips:

• If you plan to use private knowledge bases and local vector stores, ensure the vector database is properly configured and accessible from the MCP server.
• When deploying in production, consider enabling SSL termination, configuring a reverse proxy, and setting up proper authentication/authorization (JWT or Sa-Token).
• The MCP integration supports multiple AI platforms; when adding providers (OpenAI, Azure, local models), store API keys securely (e.g., in environment variables or secret management).
• For debugging RAG or knowledge-graph related issues, consult the troubleshooting guides in the docs and check logs under the logs directory for detailed error messages.
• If deploying with Docker, you may need to adjust memory limits and JVM options to accommodate the model load and vector store operations.