gemini-antigravity-cli

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Gemini Antigravity CLI

Skill by ara.so — Devtools Skills collection.

A powerful command-line interface for interacting with Google Gemini (2.5 Pro/Flash, 3.5 Flash) and Antigravity AI models. Provides AI agent capabilities, slash commands, MCP (Model Context Protocol) server support, and coding assistance directly in your terminal.

由ara.so开发的Skill工具——属于Devtools Skills集合。

这是一款功能强大的命令行界面工具，可与Google Gemini（2.5 Pro/Flash、3.5 Flash）和Antigravity AI模型交互。它在终端中直接提供AI Agent能力、斜杠命令、MCP（Model Context Protocol）服务器支持以及编码辅助功能。

What It Does

功能介绍

Gemini Antigravity CLI enables developers to:

Interact with Google Gemini AI models from the terminal
Use advanced AI coding agent features for code generation and assistance
Execute slash commands for quick workflows
Integrate with MCP servers and plugins
Maintain context-aware conversations
Switch between different Gemini/Antigravity models on the fly

Gemini Antigravity CLI帮助开发者实现：

在终端中与Google Gemini AI模型交互
使用高级AI编码Agent功能生成代码并获得编码协助
执行斜杠命令以快速完成工作流
与MCP服务器及插件集成
维持上下文感知的对话
即时切换不同的Gemini/Antigravity模型

Installation

安装方式

Global NPM Installation (Recommended)

全局NPM安装（推荐）

bash

npm install -g gemini-antigravity-cli

bash

npm install -g gemini-antigravity-cli

Binary Download

二进制文件下载

Download the pre-built executable from GitHub Releases:

bash

undefined

从GitHub Releases下载预构建的可执行文件：

bash

undefined

Download antigravity-cli.zip

下载antigravity-cli.zip

wget https://github.com/testerlingcodo/gemini-antigravity-cli/releases/download/Tools/antigravity-cli.zip unzip antigravity-cli.zip chmod +x antigravity # Unix-like systems

undefined

wget https://github.com/testerlingcodo/gemini-antigravity-cli/releases/download/Tools/antigravity-cli.zip unzip antigravity-cli.zip chmod +x antigravity # 类Unix系统

undefined

Requirements

系统要求

Node.js 16+ (for npm installation)
Google API Key with Gemini access
Internet connection for API calls

Node.js 16+（适用于NPM安装方式）
拥有Gemini访问权限的Google API密钥
用于API调用的网络连接

Configuration

配置说明

Setting Up API Key

设置API密钥

Set your Google Gemini API key as an environment variable:

bash

undefined

将Google Gemini API密钥设置为环境变量：

bash

undefined

Linux/macOS

export GOOGLE_API_KEY="your_api_key_here"

Windows PowerShell

$env:GOOGLE_API_KEY="your_api_key_here"

Windows CMD

set GOOGLE_API_KEY=your_api_key_here


Add to your shell profile for persistence:
```bash

set GOOGLE_API_KEY=your_api_key_here


将密钥添加到shell配置文件中以永久生效：
```bash

~/.bashrc or ~/.zshrc

~/.bashrc 或 ~/.zshrc

export GOOGLE_API_KEY="your_api_key_here"

undefined

export GOOGLE_API_KEY="your_api_key_here"

undefined

Configuration File

配置文件

The CLI typically stores configuration in:

Linux/macOS:

~/.config/gemini-antigravity-cli/config.json

Windows:

%APPDATA%\gemini-antigravity-cli\config.json

Example configuration:

json

{
  "model": "gemini-2.5-flash",
  "temperature": 0.7,
  "maxTokens": 8192,
  "systemPrompt": "You are a helpful coding assistant.",
  "mcpServers": [],
  "plugins": []
}

CLI通常将配置存储在以下位置：

Linux/macOS：

~/.config/gemini-antigravity-cli/config.json

Windows：

%APPDATA%\gemini-antigravity-cli\config.json

示例配置：

json

{
  "model": "gemini-2.5-flash",
  "temperature": 0.7,
  "maxTokens": 8192,
  "systemPrompt": "You are a helpful coding assistant.",
  "mcpServers": [],
  "plugins": []
}

Key Commands

核心命令

Starting the CLI

启动CLI

bash

undefined

bash

undefined

Start with default settings

使用默认设置启动

antigravity

Or use the alias

或使用别名

gemini

undefined

gemini

undefined

Slash Commands

斜杠命令

Once inside the CLI, use these slash commands:

进入CLI后，可使用以下斜杠命令：

/help

/help

Display all available commands:

/help

显示所有可用命令：

/help

/model

/model

Switch between AI models:

/model gemini-2.5-pro
/model gemini-2.5-flash
/model gemini-3.5-flash
/model antigravity

切换AI模型：

/model gemini-2.5-pro
/model gemini-2.5-flash
/model gemini-3.5-flash
/model antigravity

/clear

/clear

Clear conversation history:

/clear

清除对话历史：

/clear

/context

/context

Add file or directory context:

/context ./src/index.ts
/context ./src/

添加文件或目录上下文：

/context ./src/index.ts
/context ./src/

/save

/save

Save conversation to file:

/save conversation.md
/save output.txt

将对话保存到文件：

/save conversation.md
/save output.txt

/load

/load

Load previous conversation:

/load conversation.md

加载之前的对话：

/load conversation.md

/config

/config

View or modify configuration:

/config
/config model gemini-2.5-flash
/config temperature 0.9

查看或修改配置：

/config
/config model gemini-2.5-flash
/config temperature 0.9

/exit

/exit

Exit the CLI:

/exit

退出CLI：

/exit

Usage Patterns

使用场景

Basic Code Assistance

基础编码辅助

Start the CLI and ask coding questions directly:

typescript

// User input in CLI:
Create a TypeScript function to fetch data from an API with retry logic

// The AI will generate:
async function fetchWithRetry<T>(
  url: string,
  maxRetries: number = 3,
  delayMs: number = 1000
): Promise<T> {
  let lastError: Error;
  
  for (let attempt = 0; attempt < maxRetries; attempt++) {
    try {
      const response = await fetch(url);
      if (!response.ok) {
        throw new Error(`HTTP ${response.status}: ${response.statusText}`);
      }
      return await response.json() as T;
    } catch (error) {
      lastError = error as Error;
      if (attempt < maxRetries - 1) {
        await new Promise(resolve => setTimeout(resolve, delayMs * (attempt + 1)));
      }
    }
  }
  
  throw new Error(`Failed after ${maxRetries} attempts: ${lastError!.message}`);
}

启动CLI后直接询问编码问题：

typescript

// CLI中的用户输入：
创建一个带有重试逻辑的TypeScript API数据获取函数

// AI将生成：
async function fetchWithRetry<T>(
  url: string,
  maxRetries: number = 3,
  delayMs: number = 1000
): Promise<T> {
  let lastError: Error;
  
  for (let attempt = 0; attempt < maxRetries; attempt++) {
    try {
      const response = await fetch(url);
      if (!response.ok) {
        throw new Error(`HTTP ${response.status}: ${response.statusText}`);
      }
      return await response.json() as T;
    } catch (error) {
      lastError = error as Error;
      if (attempt < maxRetries - 1) {
        await new Promise(resolve => setTimeout(resolve, delayMs * (attempt + 1)));
      }
    }
  }
  
  throw new Error(`Failed after ${maxRetries} attempts: ${lastError!.message}`);
}

Context-Aware Coding

上下文感知编码

Load project files for context-aware assistance:

bash

undefined

加载项目文件以获得上下文感知的编码协助：

bash

undefined

In the CLI:

在CLI中：

/context ./src/api/client.ts

Then ask:

然后询问：

Refactor this API client to use async/await instead of promises

undefined

将这个API客户端重构为使用async/await而非Promise链式调用

undefined

Multi-File Analysis

多文件分析

bash

/context ./src/
/context ./tests/

bash

/context ./src/
/context ./tests/

Then:

然后询问：

Find all the test cases that are missing error handling

undefined

找出所有缺少错误处理的测试用例

undefined

Code Review Workflow

代码审查工作流

bash

/context ./pull-request.diff

bash

/context ./pull-request.diff

Ask:

询问：

Review this code for potential bugs, security issues, and best practices

undefined

审查这段代码，找出潜在的bug、安全问题及不符合最佳实践的地方

undefined

Integration Examples

集成示例

Using in Scripts

在脚本中使用

You can pipe input to the CLI for automated workflows:

bash

undefined

可通过管道将输入传递给CLI以实现自动化工作流：

bash

undefined

Send a single prompt

发送单个提示词

echo "Generate a README.md template for a TypeScript project" | antigravity

Process file content

处理文件内容

cat ./src/index.ts | antigravity "Explain what this code does"

undefined

cat ./src/index.ts | antigravity "Explain what this code does"

undefined

Batch Processing

批量处理

bash

undefined

bash

undefined

Create a script for batch processing

创建批量处理脚本

#!/bin/bash for file in ./src/*.ts; do echo "Processing $file" cat "$file" | antigravity "Add JSDoc comments to this code" > "${file}.documented" done

undefined

#!/bin/bash for file in ./src/*.ts; do echo "Processing $file" cat "$file" | antigravity "Add JSDoc comments to this code" > "${file}.documented" done

undefined

CI/CD Integration

CI/CD集成

yaml

undefined

yaml

undefined

.github/workflows/code-review.yml

name: AI Code Review on: [pull_request]

jobs: review: runs-on: ubuntu-latest steps: - uses: actions/checkout@v3 - name: Install CLI run: npm install -g gemini-antigravity-cli - name: Run Review env: GOOGLE_API_KEY: ${{ secrets.GOOGLE_API_KEY }} run: | git diff origin/main...HEAD | antigravity "Review this code change"

undefined

name: AI Code Review on: [pull_request]

undefined

MCP Server Integration

MCP服务器集成

Connecting to MCP Servers

连接MCP服务器

Configure MCP servers in your config file:

json

{
  "mcpServers": [
    {
      "name": "filesystem",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/workspace"]
    },
    {
      "name": "github",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
      }
    }
  ]
}

在配置文件中配置MCP服务器：

json

{
  "mcpServers": [
    {
      "name": "filesystem",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/workspace"]
    },
    {
      "name": "github",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
      }
    }
  ]
}

Using MCP Tools

使用MCP工具

bash

undefined

bash

undefined

Once MCP servers are configured, use them in prompts:

配置MCP服务器后，可在提示词中使用：

Use the filesystem server to read all package.json files in the workspace

使用filesystem服务器读取工作区中所有的package.json文件

Or:

或者：

Search GitHub for similar implementations using the github server

undefined

使用github服务器搜索GitHub上的类似实现

undefined

Common Patterns

常见使用模式

Project Setup Assistant

项目搭建助手

typescript

// User: Set up a new Express.js API project with TypeScript

// CLI generates project structure:
mkdir my-api && cd my-api
npm init -y
npm install express
npm install -D typescript @types/express @types/node ts-node

// Creates tsconfig.json:
{
  "compilerOptions": {
    "target": "ES2020",
    "module": "commonjs",
    "outDir": "./dist",
    "strict": true,
    "esModuleInterop": true
  }
}

// Creates src/index.ts:
import express from 'express';

const app = express();
const port = process.env.PORT || 3000;

app.use(express.json());

app.get('/health', (req, res) => {
  res.json({ status: 'ok' });
});

app.listen(port, () => {
  console.log(`Server running on port ${port}`);
});

typescript

// 用户：搭建一个新的Express.js API项目并使用TypeScript

// CLI生成项目结构：
mkdir my-api && cd my-api
npm init -y
npm install express
npm install -D typescript @types/express @types/node ts-node

// 创建tsconfig.json：
{
  "compilerOptions": {
    "target": "ES2020",
    "module": "commonjs",
    "outDir": "./dist",
    "strict": true,
    "esModuleInterop": true
  }
}

// 创建src/index.ts：
import express from 'express';

const app = express();
const port = process.env.PORT || 3000;

app.use(express.json());

app.get('/health', (req, res) => {
  res.json({ status: 'ok' });
});

app.listen(port, () => {
  console.log(`Server running on port ${port}`);
});

Debugging Assistant

调试助手

bash

/context ./src/buggy-code.ts

bash

/context ./src/buggy-code.ts

Ask:

询问：

This code throws "Cannot read property 'length' of undefined". Find and fix the bug.

undefined

这段代码抛出"Cannot read property 'length' of undefined"错误。找出并修复这个bug。

undefined

Documentation Generator

文档生成器

bash

/context ./src/utils/

bash

/context ./src/utils/

Generate docs:

生成文档：

Create comprehensive API documentation for all exported functions in markdown format

undefined

为所有导出函数创建全面的Markdown格式API文档

undefined

Test Generation

测试用例生成

bash

/context ./src/user-service.ts

bash

/context ./src/user-service.ts

Generate tests:

生成测试用例：

Create Jest unit tests for this service with 100% code coverage

undefined

为这个服务创建Jest单元测试，实现100%代码覆盖率

undefined

Troubleshooting

故障排除

API Key Issues

API密钥问题

Problem: "Invalid API key" error

Solution:

bash

undefined

问题： "Invalid API key"错误

解决方案：

bash

undefined

Verify API key is set

验证API密钥是否已设置

echo $GOOGLE_API_KEY # Linux/macOS echo %GOOGLE_API_KEY% # Windows

Re-export with correct key

使用正确的密钥重新导出

export GOOGLE_API_KEY="your_valid_key"

Test with curl

使用curl测试

curl "https://generativelanguage.googleapis.com/v1/models?key=$GOOGLE_API_KEY"

undefined

curl "https://generativelanguage.googleapis.com/v1/models?key=$GOOGLE_API_KEY"

undefined

Connection Errors

连接错误

Problem: Network timeout or connection refused

Solution:

bash

undefined

问题： 网络超时或连接被拒绝

解决方案：

bash

undefined

Check internet connectivity

检查网络连通性

ping google.com

Verify proxy settings if behind corporate firewall

如果处于企业防火墙后，验证代理设置

export HTTP_PROXY="http://proxy.company.com:8080" export HTTPS_PROXY="http://proxy.company.com:8080"

Test API endpoint

测试API端点

curl -I https://generativelanguage.googleapis.com

undefined

curl -I https://generativelanguage.googleapis.com

undefined

Model Not Available

模型不可用

Problem: "Model not found" error

Solution:

bash

undefined

问题： "Model not found"错误

解决方案：

bash

undefined

List available models

列出可用模型

/model

Use a supported model:

使用支持的模型：

/model gemini-2.5-flash

Check API access for specific models in Google AI Studio

在Google AI Studio中检查特定模型的API访问权限

undefined

undefined

Rate Limiting

速率限制

Problem: "Rate limit exceeded" or 429 errors

Solution:

Wait before retrying (rate limits reset after a period)
Use
```
gemini-2.5-flash
```
for higher rate limits
Implement exponential backoff in scripts:

bash

#!/bin/bash
attempt=0
max_attempts=5

while [ $attempt -lt $max_attempts ]; do
  if echo "$prompt" | antigravity; then
    break
  fi
  attempt=$((attempt + 1))
  sleep $((2 ** attempt))
done

问题： "Rate limit exceeded"或429错误

解决方案：

等待一段时间后重试（速率限制会在一段时间后重置）
使用
```
gemini-2.5-flash
```
以获得更高的速率限制
在脚本中实现指数退避：

bash

#!/bin/bash
attempt=0
max_attempts=5

while [ $attempt -lt $max_attempts ]; do
  if echo "$prompt" | antigravity; then
    break
  fi
  attempt=$((attempt + 1))
  sleep $((2 ** attempt))
done

Context Size Errors

上下文长度超限

Problem: "Context length exceeded" error

Solution:

bash

undefined

问题： "Context length exceeded"错误

解决方案：

bash

undefined

Reduce context size

减小上下文大小

/clear # Clear previous conversation

/clear # 清除之前的对话

Use smaller file selections

使用更小的文件选择范围

/context ./src/specific-file.ts # Instead of entire directory

/context ./src/specific-file.ts # 而非整个目录

Summarize large files first

先总结大文件内容

cat large-file.ts | head -n 100 | antigravity "Summarize this code"

undefined

cat large-file.ts | head -n 100 | antigravity "Summarize this code"

undefined

Installation Issues

安装问题

Problem: npm install fails

Solution:

bash

undefined

问题： npm安装失败

解决方案：

bash

undefined

Clear npm cache

清除npm缓存

npm cache clean --force

Try with explicit registry

使用明确的镜像源尝试安装

npm install -g gemini-antigravity-cli --registry=https://registry.npmjs.org

Or use npx without global install

或使用npx无需全局安装

npx gemini-antigravity-cli

undefined

npx gemini-antigravity-cli

undefined

Binary Execution Issues

二进制文件执行问题

Problem: Permission denied on Unix systems

Solution:

bash

chmod +x antigravity
./antigravity

Problem: Windows SmartScreen warning

Solution:

Click "More info" → "Run anyway"
Or build from source if concerned about security

问题： Unix系统上出现权限拒绝

解决方案：

bash

chmod +x antigravity
./antigravity

问题： Windows SmartScreen警告

解决方案：

点击"更多信息" → "仍要运行"
若担心安全问题，可从源代码构建

Advanced Configuration

高级配置

Custom System Prompts

自定义系统提示词

Modify behavior with custom system prompts:

json

{
  "systemPrompt": "You are an expert TypeScript developer specializing in React and Node.js. Provide concise, production-ready code with error handling. Always use modern ES6+ syntax."
}

通过自定义系统提示词修改AI行为：

json

{
  "systemPrompt": "You are an expert TypeScript developer specializing in React and Node.js. Provide concise, production-ready code with error handling. Always use modern ES6+ syntax."
}

Temperature and Token Settings

温度与Token设置

Control response creativity and length:

json

{
  "temperature": 0.2,  // Lower = more deterministic (0.0-1.0)
  "maxTokens": 4096,   // Maximum response length
  "topP": 0.95,        // Nucleus sampling threshold
  "topK": 40           // Top-K sampling parameter
}

控制响应的创造性和长度：

json

{
  "temperature": 0.2,  // 数值越低，输出越确定（范围0.0-1.0）
  "maxTokens": 4096,   // 最大响应长度
  "topP": 0.95,        // 核采样阈值
  "topK": 40           // Top-K采样参数
}

Plugin Configuration

插件配置

Enable community plugins (if supported):

json

{
  "plugins": [
    "git-integration",
    "code-formatter",
    "security-scanner"
  ]
}

启用社区插件（若支持）：

json

{
  "plugins": [
    "git-integration",
    "code-formatter",
    "security-scanner"
  ]
}

Best Practices

最佳实践

Always set API key as environment variable — never hardcode in scripts
Use
/clear
between unrelated tasks — prevents context confusion
Provide specific context with
/context
— improves accuracy
Start with
gemini-2.5-flash
— faster and cheaper for most tasks
Save important conversations — use
```
/save
```
for future reference
Use lower temperature (0.2-0.4) for code generation — more consistent output
Higher temperature (0.7-0.9) for creative tasks — like naming or brainstorming

始终将API密钥设置为环境变量——切勿在脚本中硬编码
在不相关任务之间使用
/clear
——避免上下文混淆
使用
/context
提供特定上下文——提高AI响应准确性
优先使用
gemini-2.5-flash
——大多数任务下速度更快、成本更低
保存重要对话——使用
```
/save
```
以便日后参考
代码生成时使用较低温度（0.2-0.4）——输出更一致
创意任务使用较高温度（0.7-0.9）——比如命名或头脑风暴