code-agent

# code-agent: code reading and terminal execution agent based on mcp ## Project Introduction This project is an experimental project based on Large Language Model (LLM), Model Context Protocol (MCP), and Retrieval-Augmented Generation (RAG). It demonstrates how to build an AI assistant system that can interact with external tools and utilize retrieval-augmented generation techniques. Reference project: https://github.com/StrayDragon/exp-llm-mcp-rag ```mermaid graph TD %% 核心节点 U[用户] G[Gradio界面] S[状态管理] A[Agent代理] L[LLM模型] M[MCP客户端] Log[日志工具] %% 主流程 U -->|输入查询/点击发送| G G -->|获取状态| S S -->|返回状态数据| G G -->|初始化/调用| A A -->|需要模型处理| L L -->|生成响应/工具调用| A A -->|需要工具执行| M M -->|返回工具结果| A A -->|整理结果| G G -->|流式展示回复| U %% 辅助流程 U -->|点击清空| G G -->|重置历史| S U -->|点击重置| G G -->|销毁Agent| S A -->|记录操作| Log M -->|记录调用| Log ``` ### Core Features - Large language model calls based on OpenAI API - LLM interaction with external tools via MCP (Model Context Protocol) - Uses a third-party MCP server: Desktop Commander - Supports file system operations, web content retrieval, and terminal command interaction ## Quick Start ### Environment Preparation 1. Make sure Python 3.12 or higher is installed 2. Clone this repository 3. In `.env` and fill in the necessary configuration information: - `OPENAI_API_KEY`: OpenAI API key - `OPENAI_BASE_URL`: OpenAI API base URL, be sure to keep '/v1' at the end (defaults to 'https://api.openai.com/v1') - `DEFAULT_MODEL_NAME`: (Optional) Default model name (defaults to "gpt-4o-mini") - `EMBEDDING_KEY`: (Optional) Embedding model API key (defaults to $OPENAI_API_KEY) - `EMBEDDING_BASE_URL`: (Optional) Embedding model API base URL, such as Silicon-based flow API or OpenAI-compatible API (defaults to $OPENAI_BASE_URL) - `USE_CN_MIRROR`: (Optional) Whether to use the Chinese mirror, set any value (such as '1') to true (defaults to false) - `PROXY_URL`: (Optional) Proxy URL (such as "http(s)://xxx") for `fetch` (mcp-tool) to go through the proxy ### Install Node.js (including npm, npx, etc.) Go to the official website to install the latest version https://nodejs.org/en/blog/release/v20.19.3 ### Install the scoop package management tool ```shell # 安装 scoop Set-ExecutionPolicy RemoteSigned -Scope CurrentUser -Force Invoke-Expression (Invoke-RestMethod -Uri https://get.scoop.sh) ``` ### Install the uv package management tool ```bash # 安装 uv scoop install main/uv ``` ### Install dependencies ```bash # 使用 uv 安装依赖 uv sync uv pip install numpy uv pip install opencv-python uv pip install matplotlib uv pip install pillow ``` ### Activate the virtual environment ```bash # 使用 uv 安装依赖 .venv/Scripts/activate ``` ### Install desktop-commander ```bash # 使用 npx 安装依赖 npx -y --verbose @wonderwhy-er/desktop-commander@latest setup --ignore-scripts ``` ### Run the example This project uses the `just` command tool to run different examples: ```bash # 查看可用命令 just help ``` ### Start chatting ```bash # chat_openai启动 just chat_openai ``` ### Start agent ```bash # agent启动 just agent ``` ## Project Structure - `src/augmented/`: Main source code directory - `agent.py`: Agent implementation, responsible for coordinating LLM and tools - `chat_openai.py`: OpenAI API client encapsulation - `mcp_client.py`: MCP client implementation - `mcp_tools.py`: MCP tool definition - `utils/`: Tool functions - `info.py`: Project information and configuration - `pretty.py`: Unified log output system - `justfile`: Task execution configuration file