A powerful CLI tool that leverages OpenAI's GPT models to generate high-quality, conventional commit messages from your staged changes.
- 🤖 Uses OpenAI's GPT models to analyze your staged changes
- 📝 Generates conventional commit messages that follow best practices
- 🎯 Interactive selection from multiple commit message suggestions
- ✏️ Edit messages directly or request AI revisions
- 🧠 Advanced reasoning mode for enhanced AI interactions
- 🔍 Comprehensive debugging capabilities with file or stdout logging
- ⚡ Streaming responses for real-time feedback
- 🔄 Auto-update checks to keep you on the latest version
- 🎨 Beautiful terminal UI with color-coded output
- ⚙️ Configurable settings via YAML config file
cargo install turbocommit
Pro tip: Add an alias to your shell configuration for quicker access:
# Add to your .bashrc, .zshrc, etc.
alias tc='turbocommit'
- Stage your changes:
git add . # or stage specific files
- Generate commit messages:
turbocommit # or 'tc' if you set up the alias
After generating commit messages, you can:
- Select your preferred message from multiple suggestions
- Edit the message directly before committing
- Request AI revisions with additional context or requirements
- Commit the message once you're satisfied
-n <number>
- Number of commit message suggestions to generate-t <temperature>
- Temperature for GPT model (0.0 to 2.0) (no effect in reasoning mode)-f <frequency_penalty>
- Frequency penalty (-2.0 to 2.0)-m <model>
- Specify the GPT model to use-r, --enable-reasoning
- Enable support for models with reasoning capabilities (like o-series)--reasoning-effort <level>
- Set reasoning effort for supported models (low/medium/high, default: medium)-d, --debug
- Show basic debug info in console--debug-file <path>
- Write detailed debug logs to file (use '-' for stdout)--auto-commit
- Automatically commit with the generated message--amend
- Amend the last commit with the generated message (useful with Git hooks)--api-key <key>
- Provide API key directly--api-endpoint <url>
- Custom API endpoint URL-p, --print-once
- Disable streaming output
When using models that support reasoning capabilities (like OpenAI's o-series), this mode enables their built-in reasoning features. These models are specifically designed to analyze code changes and generate commit messages with their own reasoning process.
Example usage:
turbocommit -r -m o3-mini -n 1 # Enable reasoning mode with default effort
turbocommit -r --reasoning-effort high -m o3-mini -n 1 # Specify reasoning effort
Debug output helps troubleshoot API interactions:
turbocommit -d # Basic info to console
turbocommit --debug-file debug.log # Detailed logs to file
turbocommit --debug-file - # Detailed logs to stdout
The debug logs include:
- Request details (model, tokens, parameters)
- API responses and errors
- Timing information
- Full request/response JSON (in file mode)
Different models have different capabilities and limitations:
- Support reasoning mode
- Do not support temperature/frequency parameters
- May not support multiple choices (
-n
) - Optimized for specific tasks
- Support all parameters
- Multiple choices available
- Temperature and frequency tuning
- Standard reasoning capabilities
For more options, run:
turbocommit --help
turboCommit creates a config file at ~/.turbocommit.yaml
on first run. You can customize:
- Default model
- API endpoint
- Temperature and frequency penalty
- Number of suggestions
- System message prompt
- Auto-update checks
- Reasoning mode defaults
- And more!
Example configuration:
model: "gpt-4"
default_temperature: 1.0
default_frequency_penalty: 0.0
default_number_of_choices: 3
enable_reasoning: true
reasoning_effort: "medium"
disable_print_as_stream: false
disable_auto_update_check: false
You can maintain multiple configuration files for different use cases (e.g., different providers or environments) and specify which one to use with the -c
or --config
option:
# Use a local config file
turbocommit -c ./local-config.yaml
# Use a different provider's config
turbocommit -c ~/.turbocommit-azure.yaml
# Use the default config
turbocommit # uses ~/.turbocommit.yaml
Each config file follows the same format as shown above. This allows you to easily switch between different configurations without modifying the default config file.
Contributions are welcome! Feel free to open issues and pull requests.
Licensed under MIT - see the LICENSE file for details.
The --amend
option allows you to change the commit message of your last commit. This is useful when:
- You want to improve the message of your last commit
- You want to fix a typo in your commit message
- You want to add more context to your commit message
Usage:
# First, make sure you have no staged changes
git status # Should show no staged changes
# Then use --amend to improve the last commit's message
turbocommit --amend # This will analyze the last commit's changes and suggest a new message
Important Notes:
- When using
--amend
, you must not have any staged changes - The tool will analyze only the changes from your last commit
- If you want to include new changes in the amended commit:
- Either commit them first normally, then amend that commit
- Or use
git commit --amend
manually to include them
You can also combine this with auto-commit for a quick message update:
turbocommit --amend --auto-commit # Automatically amend with the first generated message
If your project uses Git hooks (e.g., linters, formatters), here's how to use turboCommit effectively:
- Stage and commit your changes normally:
git add .
turbocommit
-
If hooks fail:
- Fix the issues reported by hooks
- Stage the fixed files (
git add .
) - Commit again
-
If you want to improve the commit message after all hooks pass:
# Make sure you have no staged changes
git status
# Then improve the message
turbocommit --amend # This will analyze the commit and suggest a better message
This workflow ensures that:
- Code quality checks run before the commit
- You can improve the commit message after all checks pass
- The final commit message is high-quality and descriptive