AgentX

# AgentX - Intelligent Dialogue System Platform [](https://opensource.org/licenses/MIT) AgentX is an intelligent Agent building platform based on Large Language Models (LLM) and Multi-Capability Platforms (MCP). It aims to simplify the process of creating Agents, allowing users to build personalized intelligent Agents through natural language and tool integration without complex process nodes or drag-and-drop operations. ## 🔗 Related Links ### 📦 Sub-repositories - 🛡️ **High Availability Gateway**: [API-Premium-Gateway](https://github.com/lucky-aeon/API-Premium-Gateway) - High availability component for models - 🌐 **MCP Gateway**: [mcp-gateway](https://github.com/lucky-aeon/mcp-gateway) - Unified management of MCP services - 🏪 **MCP Community**: [agent-mcp-community](https://github.com/lucky-aeon/agent-mcp-community) - Open-source community for MCP Server ### 📚 Learning Resources - 🎥 **Project Tutorial**: [Bilibili Video Tutorial](https://www.bilibili.com/video/BV1qaTWzPERJ/?spm_id_from=333.1387.homepage.video_card.click) - 📖 **Detailed Teaching**: [Qiao Ya Community - code.xhyovo.cn](https://code.xhyovo.cn/) - 🎯 **Project Demo**: [Online PPT Introduction](https://needless-comparison.surge.sh) ## ⏳ Features - [x] Agent Management (Create/Publish) - [x] LLM Context Management (Sliding Window, Summarization Algorithm) - [x] Agent Strategy (MCP) - [x] Large Model Service Providers - [x] Users - [x] Tool Marketplace - [x] MCP Server Community - [x] MCP Gateway - [x] Pre-configured Tools - [x] Agent Scheduled Tasks - [ ] Agent OpenAPI - [ ] High Availability Components for Models - [ ] RAG - [ ] Billing - [ ] Multi-Agent - [ ] Agent Monitoring ## 🚀 How to Install and Start ### 🛠️ Environment Preparation * **Node.js & npm**: It is recommended to use the LTS version. * **Java Development Kit (JDK)**: JDK 17 or higher. * **Docker & Docker Compose**: For deploying databases and other dependent services. ### 💻 Local Startup #### 1\. Clone the Repository ```bash git clone https://github.com/your-username/AgentX.git # Replace with the actual repository address cd AgentX ``` #### 2\. Start the Database (PostgreSQL) Navigate to the `script` directory and execute the startup script. This script will use Docker Compose to start a PostgreSQL container and initialize the database. ```bash cd script chmod +x setup_with_compose.sh ./setup_with_compose.sh ``` Once successfully started, you will see the connection information for PostgreSQL: ``` 🎉 PostgreSQL container has been successfully started! Container Name: agentx-postgres Connection Information: Host: localhost Port: 5432 User: postgres Password: postgres Database: agentx Connection URL: jdbc:postgresql://localhost:5432/agentx You can connect to the database using the following command: docker exec -it agentx-postgres psql -U postgres -d agentx ✅ Database initialization completed! ``` #### 3\. Start the Backend Service (AgentX Java Application) Return to the project root directory, navigate to the `AgentX` directory, and build and run the backend service using Maven or Gradle (if used). ```bash cd ../AgentX # If it's a Maven project, typically: ./mvnw clean install ./mvnw spring-boot:run # Or run according to the actual jar file path # java -jar target/AgentX-0.0.1-SNAPSHOT.jar # Replace with the actual jar file name ``` Once the backend service is started, it will typically listen on port `8080`. #### 4\. Start the Frontend Service (AgentX-Frontend-Plus) Return to the project root directory, navigate to the `agentx-frontend-plus` directory, install dependencies, and start the frontend service. ```bash cd ../agentx-frontend-plus npm install --legacy-peer-deps npm run dev ``` Once the frontend service is started, it will typically listen on port `3000`. ### ⚙️ Common Docker Compose Commands In the `script` directory: * **Start all services**: `./setup_with_compose.sh` (recommended for the first run or when reinitializing the database) * **Start/Restart services (without initializing the database)**: `docker-compose up -d` * **Stop all services**: `docker-compose down` * **Check service status**: `docker ps` * **View database logs**: `docker logs agentx-postgres` ## Feature Introduction ### Agent Management Users can create Agents using LLM + "Plugins", where plugins refer to tools, knowledge bases, etc. Anything that serves the LLM is called a "plugin". Agents created by users can be published for others to use. When using an Agent, the choice of model is determined by the user. ### Token Context Management Although langchain4j provides built-in Token context handling, the system also offers a Token-based sliding window and summarization algorithm. ### Agent Strategy Agent = LLM + Tools The project implements Agents using LLM + MCP, facilitated by langchain4j's MCP. In the future, we will develop Agent strategies through self-testing and research. ### Users Registration and login are available via GitHub and Email. ### Tool Marketplace Users can upload their own tools or use official tools, which can be utilized by Agents. ### MCP Server Community Tools uploaded by users will be synchronized to https://github.com/lucky-aeon/agent-mcp-community after passing `manual review`. ### MCP Gateway A self-developed MCP gateway is used to unify the management of all MCP Servers: https://github.com/lucky-aeon/mcp-gateway ## Contact Us We are committed to building an active developer community and welcome all forms of communication and collaboration! ### 📱 Private WeChat For technical questions or business cooperation, you can add the developer's WeChat: <img src="docs/images/wechat.jpg" alt="Private WeChat" width="200"/> ### 👥 WeChat Group Join our technical discussion group and engage with more developers: <img src="docs/images/group.jpg" alt="WeChat Group" width="200"/> ### 📢 WeChat Official Account Follow our official account for the latest technical updates and product news: <img src="docs/images/微信公众号.jpg" alt="Official WeChat Account" width="200"/> --- **If the QR code is expired or cannot be scanned, please contact me via Private WeChat.**