cmt - AI-Powered Git Commit Message Generator

cmt is a command-line tool that generates meaningful git commit messages using AI models. It analyzes your staged changes and generates a well-formatted, descriptive commit message following conventional commit standards.

Features

• 🤖 Supports multiple AI models:
  • Anthropic's Claude 3.5 Sonnet (default, temperature 0.3)
  • OpenAI's GPT-4 Optimized (temperature 1.0)
• 📝 Follows conventional commit format (type: subject)
• 📊 Optional diff statistics
• 💡 Contextual hints to guide message generation
• ⚙️ Configurable AI model and parameters
• 🔑 Supports environment variables for API keys

Installation

Using Install Script (Recommended)

The easiest way to install cmt is using our install script:

curl -sSL https://raw.githubusercontent.com/clifton/cmt/main/scripts/install.sh | sh

This will automatically download and install the latest version for your platform.

Installing from crates.io

You can also install cmt directly from crates.io:

cargo install cmt

Installing from source

Alternatively, you can build from source:

# Clone the repository
git clone https://github.com/clifton/cmt.git
cd cmt

# Build and install
cargo install --path .

Configuration

Set your API key(s) either:

1. As environment variables:

   # For Claude (default)
   export ANTHROPIC_API_KEY='your-api-key'
   
   # For OpenAI (optional)
   export OPENAI_API_KEY='your-api-key'
   

2. Or in a .env file in your project directory:

   ANTHROPIC_API_KEY=your-api-key
   OPENAI_API_KEY=your-api-key
   

Usage

Basic Usage

# Stage your changes first
git add .

# Generate a commit message using Claude (default)
cmt

# Generate a commit message using OpenAI
cmt --openai

# Use the generated message directly with git
git commit -F <(cmt --message-only)

Command-line Options

Usage: cmt [OPTIONS]

Options:
  -m, --message-only               Only output the generated commit message, without formatting
  -s, --show-diff                  Show the diff of staged changes
      --model <MODEL>              Use a specific AI model (defaults to claude-3-5-sonnet-latest or gpt-4o depending on provider)
      --openai                     Use OpenAI instead of Claude (which is default)
      --anthropic                  Use Anthropic instead of OpenAI (which is default)
  -t, --temperature <TEMPERATURE>  Adjust the creativity of the generated message (0.0 to 2.0)
      --hint <HINT>                Add a hint to guide the AI in generating the commit message
  -h, --help                       Print help
  -V, --version                    Print version

Examples

# Show diff statistics along with the message
cmt --show-diff

# Use OpenAI with a custom temperature
cmt --openai --temperature 0.8

# Provide a hint for context
cmt --hint "This fixes the login timeout issue"

# Combine multiple options
cmt --openai --model gpt-4 --hint "Update dependencies for security" --show-diff

# Use with git commit directly
git commit -F <(cmt --message-only --hint "Refactor to improve performance")

How It Works

1. When you run cmt, it analyzes your staged git changes
2. The changes are sent to the selected AI model (Claude or OpenAI) along with:
   • A system prompt that guides the model to generate conventional commits
   • Your optional hint for additional context
   • The staged changes as the user prompt
3. The AI generates a commit message following the conventional commit format
4. The message is displayed (with optional diff statistics) or output directly for use with git

Commit Message Format

The generated commit messages follow the conventional commit format:

type: subject

- Detailed change 1
- Detailed change 2

Where type is one of:

• feat: New features
• fix: Bug fixes
• docs: Documentation changes
• style: Code style changes (formatting, etc.)
• refactor: Code refactoring
• test: Adding or modifying tests
• chore: Maintenance tasks

Development

Building from source

# Clone the repository
git clone https://github.com/yourusername/cmt.git
cd cmt

# Build
cargo build

# Run tests
cargo test

# Run in development
cargo run

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

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