alg-mcp

## Getting Started **Algorithmic Learning Graybeard - Empowering Intelligent Querying with RAG + MCP Dual-Engine Architecture.** *Built upon the foundation of Large Language Models (LLM), leveraging MCP Server's capabilities in data fusion and semantic indexing for databases, this project constructs a dual-engine architecture that synergizes RAG and MCP. It achieves precise conversion from natural language to structured queries through intelligent syntax parsing and query mapping, forming a complete technical loop. This can be used to assist enterprises/teams in technical research and development, product design, and other supporting tasks.* > **🌟 If this project is helpful to you, please give it a star 🌟** ## System Composition and Documentation [English](README_en.md) <br> > 👉 Code Address: [github](https://github.com/chenlinyang/alg-mcp) or [gitee](https://gitee.com/chenlinyang/alg-mcp) If you have thousands of data directories and billions of data facts, and you need to perform data retrieval or analysis, you can extend and optimize this project. Architecture Reference: - [Claude MCP Protocol](https://www.claudemcp.com/specification) - [MCP SDK Introduction](https://modelcontextprotocol.io/introduction) - Overall Architecture<br> <img src="./images/GeneralArchitecture.png" width = "800" alt="Overall Architecture" /> - Life Cycle<br> <img src="./images/LifeCycle.png" width = "600" alt="Life Cycle" /> - Operation Steps<br> <img src="./images/Operation.png" width = "600" alt="Operation Steps" /> ## Features * MCP Server * MCP Client * Chat Demo Address: <br> http://localhost:4000/ - Weather Query > External interface free docking  - Database Operation > One sentence to complete grouping, ranking, sampling and other SQL operations  ## Supported Models * Zhipu-AI * Also supports access to other models: <br> Supports DeepSeek, ChatGPT 3.5, Tongyi Qianwen, Wenxin Yiyan, Ollama, etc. ## Technology Stack This repository is for backend services Technology Stack: * jdk 17 * springboot 3.4.4 * [langchain4j 1.0.0-beta3](https://github.com/langchain4j/langchain4j) * **mcp-server** utilizes custom annotations to implement the MCP protocol. Completed interfaces include: <br> initialize, tools/call, tools/list, notifications/initialized, prompts/list, prompts/get, etc. * **mcp-client** uses RAG custom QueryTransformer to read MCP server and implement NPL2SQL functionality ## How to Run ### mcp-server Configuration **a. Configure Database Connection** * Database (MySQL) ```yaml spring: datasource: type: com.zaxxer.hikari.HikariDataSource url: jdbc:mysql://<ip>:<port>/<database_name>?characterEncoding=utf8&serverTimezone=GMT%2B8&useSSL=false driver-class-name: com.mysql.cj.jdbc.Driver username: <username> password: <password> ``` * Supports PostgreSQL, Oracle, SQLServer, H2 and other relational databases<br> Configuration reference: [spring-boot-data-jdbc](https://docs.spring.io/spring-boot/docs/current/reference/html/data.html#data.sql.jdbc) **b. Configure Third-Party Weather API** * Apply for QWeather API-KEY and API-HOST<br> [Application Address](https://id.qweather.com/#/login) ```yaml weather: api: host: <api-host> api-key: <api-key> ``` ### mcp-client Configuration **a. Configure mcp-server Connection** * SSE Endpoint ```yaml mcp: server: sse-url: http://localhost:3002/sse ``` **b. Configure Large Model** * Large model api-key and model configuration ```yaml zhipu-ai: api-key: <api-key> model: <model> ```