Skip to content

products / developer tool

Git Commit Analyzer

A local AI Git plugin. It reads your staged diff and writes a Conventional Commits message with llama.cpp and a GGUF model, so your code never leaves your machine.
View on GitHubDownload a release

  • Local AI

    GGUF models run on your machine

  • Conventional Commits

    type(scope): subject

  • Interactive

    Use, edit, or cancel

  • Private

    No cloud calls, no telemetry

install

Installed in two commands

Homebrew is the recommended route on macOS, with prebuilt binaries for Apple Silicon and Intel. No Rust toolchain needed.

Homebrew (recommended)
brew tap zh30/tap
brew install git-ca
One line install script
bash -c "$(curl -fsSL https://sh.zhanghe.dev/install-git-ca.sh)"

System requirements

Git
Version 2.30 or later
Model
GGUF, downloaded on first run
macOS
Homebrew (Intel and Apple Silicon)
Windows
Prebuilt binary from GitHub Releases
Linux
Prebuilt releases are temporarily unavailable

usage

Stage, run, confirm

  1. 01

    Stage your changes

    git add <files>
  2. 02

    Run the analyzer

    git ca
  3. 03

    Confirm the message

    Use, edit, or cancel the suggestion

Configuration

git ca model
Select a model
git ca language
Output language (EN / ZH)
git ca doctor
Diagnostics

features

What it does for you

  • Reads the change

    Analyzes staged changes and writes a message that fits them, using a local GGUF model.

  • Consistent history

    Produces type(scope): subject messages for a clean, maintainable log.

  • Interactive mode

    Use, edit, or cancel every suggestion. You stay in control.

  • Fully local

    Inference runs through llama.cpp on your machine. No external API calls.

privacy

Your code stays on your machine

  • Local processing. All inference runs through llama.cpp with GGUF models.

  • No cloud. Code and commit messages never leave your device.

  • Open source. The whole codebase is public and auditable on GitHub.

  • No telemetry. The CLI collects no usage data or analytics.

related

Related workflows

official

Official links

Use these canonical links for citations, installation, source review, privacy checks, and support.

WebsiteGitHubGitHub ReleasesSupport

faq

Frequently asked questions

No. Git Commit Analyzer runs entirely on your local machine using llama.cpp. Your code and commit messages never leave your device, ensuring 100% privacy and security.

No. Once installed and the model is downloaded, the tool works completely offline. You can generate commit messages even without an internet connection.

The tool uses GGUF models optimized for local inference. It automatically downloads a recommended model, but you can configure it to use other compatible models.

Yes. You can configure the output language to Chinese or other languages supported by the underlying model using the configuration commands.