AutoCommit

AutoCommit generates git commit messages from your staged changes using AI. It currently supports OpenAI, Ollama 2, Mistral, Groq and Google AI models.

Features

• AI-driven commit messages based on diffs
• Interactive CLI: commit, copy, regenerate, add instructions, or exit
• Editable instruction template
• Shell completion scripts for Bash, Zsh, Fish, PowerShell

Installation

Go

Ensure you have go installed and run:

go install github.com/christian-gama/autocommit/v2@latest

Manual

git clone https://github.com/christian-gama/autocommit.git
cd autocommit
make build
make install # Linux/macOS

Enabling Shell Completion (Optional)

Enable shell completion by following the autocommit completion --help instructions. If you need help to set up for a specific shell you can run autocommit completion [shell] --help. Available shells are:

• bash
• zsh
• fish
• powershell

Example for Zsh (macOS):

echo "autoload -U compinit; compinit" >> ~/.zshrc
autocommit completion zsh > $(brew --prefix)/share/zsh/site-functions/_autocommit

Configuration

Run autocommit configure to choose a provider. Set your credential (such as API Key) and then select your preferred model. Your settings are stored locally.

Supported Providers

• OpenAI

  • Requires an API key from OpenAI's platform
  • Supported models:
    • GPT-4o
    • GPT-4.1
    • GPT-4.1-mini
    • GPT-4.1-nano
    • o1
    • o1-mini
    • o3
    • o3-mini
    • o4-mini

• Ollama

  • Requires Ollama to be installed and running locally
  • No API key needed, but:
    1. The Ollama service must be active (run ollama serve before invoking)
    2. You must have the models you want to use already downloaded locally (e.g., run ollama pull llama4 to download llama4 before selecting it in the autocommit configuration interface)
  • Supported models:
    • gemma:1b, gemma:4b, gemma:12b, gemma:27b
    • qwen3:0.6b, qwen3:1.7b, qwen3:4b, qwen3:8b, qwen3:14b, qwen3:30b, qwen3:32b, qwen3:235b
    • deepseek-r1:1.5b, deepseek-r1:7b, deepseek-r1:8b, deepseek-r1:14b, deepseek-r1:32b, deepseek-r1:70b, deepseek-r1:671b
    • llama4
    • llama3.3

• Mistral

  • Requires an API key from Mistral AI Platform
  • Supported models:
    • mistral-large-latest
    • mistral-medium-latest
    • mistral-small-latest

• Google AI

  • Requires a Google AI API key from Google AI Studio
  • Supported models:
    • gemini-2.0-flash
    • gemini-2.5-pro-exp-03-25
    • gemini-2.5-pro-preview-05-06
    • gemini-2.5-flash-preview-04-17

• Groq

  • Requires a Groq API key from Groq
  • Supported models:
    • gemma2-9b-it
    • llama-3.3-70b-versatile
    • llama-3.1-8b-instant
    • llama3-70b-8192
    • llama3-8b-8192

You can change provider settings anytime by running autocommit configure again.

Usage

1. Stage your changes: git add .
2. Run AutoCommit: autocommit OR (optional) autocommit --instruction "Add additional instructions to improve the commit message here"
3. Follow the prompts to:
   • Commit with the generated message
   • Copy the message to clipboard
   • Regenerate with new instructions
   • Add a custom instruction
   • Exit without committing

Instructions Template

• Open and edit the AI prompt template: autocommit instructions
• Restore default template: autocommit instructions restore

If the template is missing, AutoCommit recreates it automatically.

Quality of Commit Messages

Message quality depends on the model and context size. Large diffs dilute focus, and older or smaller models often miss instructions or skip key details. Reasoning models tend to follow prompts more faithfully and produce clearer, more accurate messages, but tend to cost more. For best results, keep your commits focused, minimize diff size, and use a reasoning-capable model.

License

MIT License See LICENSE for details.