MCP Project Overview

MCP Server for creating code project file architecture and summary using the Model Context Protocol (MCP).

Description

This is a Node.js application built with TypeScript that implements an MCP server to help analyze and summarize code project architecture. It uses the @modelcontextprotocol/sdk to provide resources and tools for project analysis.

Features

• Project Structure Resource: Get a structured view of a project's file hierarchy
• File Analysis Tool: Analyze the content and purpose of individual files
• Project Summary Tool: Generate a comprehensive summary of a project including its structure and package information
• LangChain Integration: Utilize LangChain for advanced language model capabilities
• Chroma Vector Database: Store and retrieve vector embeddings for semantic search

Prerequisites

• Node.js (v18 or higher)
• npm (v6 or higher)
• OpenAI API key (for LangChain and vector embeddings)
• Chroma DB server (optional, defaults to http://localhost:8000)

Installation

1. Clone the repository:

   git clone https://github.com/Pengguancheng/mcp-project-overview.git
   cd mcp-project-overview
   

2. Install dependencies:

   npm install
   

MCP Server 配置

要在您的編輯器或開發環境中整合 MCP Server，請使用以下配置：

{
  "mcpServers": {
    "chroma": {
      "command": "node",
      "args": [
        "dist/index.js"
      ],
      "env": {
        "OPENAI_API_KEY": "your-openai-api-key-here",
        "PROJECT_NAME": "your-project-name"
      }
    }
  }
}

環境變量說明

• OPENAI_API_KEY: 您的 OpenAI API 密鑰，用於生成嵌入和項目摘要
• PROJECT_NAME: 項目名稱，用作 Chroma 集合名稱
• CHROMA_URL: (可選) Chroma 數據庫服務器 URL，默認為 http://localhost:8000

MCP 使用範例

添加代碼文檔到向量數據庫

// 添加類文檔
await mcpClient.callTool("vector-add", {
  type: "class",
  name: "UserRepository",
  namespace: "app.repositories",
  text: "負責用戶數據的 CRUD 操作，包含創建、讀取、更新和刪除用戶的方法。",
  projectName: "my-project",
  filePath: "src/repositories/UserRepository.ts",
  summary: "用戶數據訪問層",
  references: ["src/models/User.ts", "typeorm"]
});

語義搜索代碼實體

// 自然語言查詢
const results = await mcpClient.callTool("vector-search", {
  query: "用戶認證相關的代碼",
  projectName: "my-project",
  limit: 5
});

// 按類型過濾
const classResults = await mcpClient.callTool("vector-search", {
  query: "數據庫操作",
  type: "class",
  projectName: "my-project"
});

// 按引用關係過濾
const refResults = await mcpClient.callTool("vector-search", {
  query: "用戶管理",
  references: ["src/models/User.ts"],
  projectName: "my-project"
});

生成項目概覽

await mcpClient.callTool("summary-code", {
  projectDir: "/path/to/your/project",
  targetDir: "src",
  outputFile: "docs/overview.md",
  summaryType: "overview"
});

清除向量數據庫

await mcpClient.callTool("vector-clear", {
  projectName: "my-project"
});

使用資源查詢

// 使用資源進行自然語言查詢
const resourceResult = await mcpClient.getResource("chroma://query/視頻處理相關的代碼");