该项目使用模型上下文协议（MCP）为GitHub仓库提供语义搜索功能。它支持针对索引的GitHub仓库进行检索增强生成（RAG）查询，支持公共和私有仓库。

组成部分

• vscode-extension：VSCode 扩展，用于与 Copilot 聊天窗口集成为 MCP。
• 工作流程：后端MCP服务器，使用Cloudflare基础设施处理索引和搜索查询。
• 配置包：用于部署的配置和打包工具。

快速启动

MCP 配置

1. 在你的IDE中配置MCP服务器（详见工作流程README中的设置说明）。
2. 通过提供 @owner 和 repository 参数，使用语义搜索工具。
3. 如果仓库尚未被索引，请稍后再回来查看，因为初始索引需要时间。

特色

• 跨 GitHub 仓库的语义代码搜索
• 支持私有仓库
• 与VSCode及其他MCP兼容工具的集成
• 基于Cloudflare的后端支持扩展性

有关详细的安装、部署和使用说明，请参阅每个组件目录中的 README 文件。

GitHub Semantic Search

This project provides semantic search capabilities for GitHub repositories using the Model Context Protocol (MCP). It enables retrieval-augmented generation (RAG) queries against indexed GitHub repositories, supporting both public and private repos.

Components

• vscode-extension: VSCode extension for integration with Copilot Chat Window as MCP.
• workflow: Backend MCP server that handles indexing and search queries using Cloudflare infrastructure.
• config-packs: Configuration and packaging utilities for deployment.

Quick Start

MCP Configuration

{
  "mcpServers": {
    "github-semantic-search-server": {
      "type": "streamable-http",
      "url": "https://github-search.lokeel.com/mcp",
      "headers": {
        "GITHUB_TOKEN": ""
      }
    }
  }
}

1. Configure the MCP server in your IDE (see workflow README for setup instructions).
2. Use the semantic search tool by providing @owner and repository parameters.
3. If a repository hasn’t been indexed yet, check back later as initial indexing takes time.

Features

• Semantic code search across GitHub repositories
• Support for private repositories
• Integration with VSCode and other MCP-compatible tools
• Cloudflare-powered backend for scalability

For detailed installation, deployment, and usage instructions, refer to the README files in each component directory.