Shai - AI 驱动的 Shell 命令生成器

English | 中文

中文版

📖 项目简介

Shai 是一个 Shell 命令生成工具，利用大型语言模型（LLM）根据自然语言描述自动生成相应的 Shell 命令，提供安全的风险评估和交互式确认机制。

✨ 核心特性

• 🤖 智能命令生成 - 基于自然语言描述生成准确的 Shell 命令
• 🔒 安全评估 - 自动评估命令风险等级（安全/警告/危险）
• ⚙️ 灵活配置 - 支持多样模型配置
• 🎨 交互式界面 - 彩色输出和加载动画，提升用户体验

🚀 快速开始

安装方法

uv tool install shai-python

基本使用

命令行模式：

shai 列出当前目录下的所有 Python 文件

交互式模式：

shai
>> 查找大于 10MB 的文件

配置编辑：

shai config

⚙️ 配置说明

首次运行配置

首次运行时会自动创建配置文件，你需要配置 API 密钥：

配置文件位置

• 配置文件: ~/.config/shai/config.yaml

配置示例

# ~/.config/shai/config.yaml

# 调用模型
default_model: "deepseek" # Point to ID in models

# 提供商配置
providers:
  deepseek:
    api_key: "${DEEPSEEK_API_KEY}" # refer to environment variable
    #base_url: "https://api.deepseek.com"  # 使用缺省
  智普:
    api_key: "sk-..." # 建议使用环境变量，比如 ${GLM_API_KEY}
    base_url: "https://open.bigmodel.cn/api/paas/v4"
  火山方舟:
    api_key: "${VOLCANO_API_KEY}"
    base_url: "https://ark.cn-beijing.volces.com/api/v3"
  openai:
    api_key: "${OPENAI_API_KEY}" 
  google:
    api_key: "${GOOGLE_API_KEY}" 
  anthropic:
    api_key: "${ANTHROPIC_API_KEY}"
  xai:
    api_key: "${XAI_API_KEY}" 
  openrouter:
    api_key: "${OPENROUTER_API_KEY}" 

# 模型配置
models:
  - id: "deepseek-chat"
    provider: "deepseek"
    alias: "deepseek"

  - id: "glm-4.7-flash"
    provider: "智普"

  - id: "doubao-seed-1-6-flash-250828"
    provider: "火山方舟"
    alias: "doubao-flash"

🔧 功能详解

命令生成流程

1. 输入描述 - 用户输入自然语言任务描述
2. AI 分析 - 模型分析任务并生成合适的命令
3. 风险评估 - 自动评估命令的风险等级
4. 用户确认 - 根据风险等级提示用户确认执行
5. 命令执行 - 用户确认后执行生成的命令

风险等级

• 🟢 安全 (Safe) - 无害命令，可直接执行
• 🟡 警告 (Caution) - 可能有副作用，需要确认
• 🔴 危险 (Danger) - 高风险命令，需要明确确认

🐛 故障排除

常见问题

Q: API 密钥配置错误 A: 运行 shai config 重新配置，或设置环境变量

Q: 命令执行失败 A: 检查网络连接和 API 配额，确认配置正确

Q: 权限错误 A: 配置文件会自动创建在当前用户目录，无需特殊权限

📄 许可证

本项目采用 MIT 许可证 - 查看 LICENSE 文件了解详情。

---

安全提示: 使用 AI 生成的命令时请谨慎，特别是涉及系统修改或数据删除的操作。建议在非生产环境中先测试命令效果。

English Version

📖 Project Introduction

Shai uses large language models (LLMs) to generate shell command based on natural language descriptions. It provides safety risk assessment, interactive confirmation mechanisms.

✨ Core Features

• 🤖 Intelligent Command Generation - Generate accurate shell commands from natural language descriptions
• 🔒 Safety Assessment - Automatic risk level evaluation (Safe/Caution/Danger)
• ⚙️ Flexible Configuration - Various models configuration
• 🎨 Interactive Interface - Colored output and loading animations for better user experience

🚀 Quick Start

Installation

uv tool install shai-python

Basic Usage

Command Line Mode:

shai list all Python files in current directory

Interactive Mode:

shai
>> find files larger than 10MB

Configuration Editing:

shai config

⚙️ Configuration

First Run Setup

Configuration file is automatically created on first run:

Configuration File Location

• Config File: ~/.config/shai/config.yaml

Configuration Example

# ~/.config/shai/config.yaml

# Default settings
default_model: "deepseek" # Point to ID or alias in models

# Providers
providers:
  deepseek:
    api_key: "${DEEPSEEK_API_KEY}" # refer to environment variable
    #base_url: "https://api.deepseek.com"  # use default base_url
  openai:
    api_key: "sk..."  # recommended NOT to use plain text API KEY.
  google:
    api_key: "${GOOGLE_API_KEY}" 
  anthropic:
    api_key: "${ANTHROPIC_API_KEY}"
  xai:
    api_key: "${XAI_API_KEY}" 
  openrouter:
    api_key: "${OPENROUTER_API_KEY}" 

# Models
models:
  - id: "deepseek-chat"
    provider: "deepseek"
    alias: "deepseek"

🔧 Features

Command Generation Process

1. Input Description - User provides natural language task description
2. AI Analysis - Model analyzes task and generates appropriate command
3. Risk Assessment - Automatic risk level evaluation
4. User Confirmation - Prompt user based on risk level
5. Command Execution - Execute generated command after confirmation

Risk Levels

• 🟢 Safe - Harmless commands, can be executed directly
• 🟡 Caution - Commands with potential side effects, require confirmation
• 🔴 Danger - High-risk commands, require explicit confirmation

🐛 Troubleshooting

Common Issues

Q: API Key Configuration Error A: Run shai config to reconfigure, or set environment variables

Q: Command Execution Failed A: Check network connection and API quota, verify configuration

Q: Module Not Found Error A: Ensure installing shai-python package, not shai

Q: Permission Error A: Configuration files are created in user directory, no special permissions needed

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Security Notice: Use AI-generated commands with caution, especially those involving system modifications or data deletion. Test command effects in non-production environments first.