git commit-at

AI-powered git commit message generator — local, private, no API key needed.

git commit-at analyzes your staged changes and suggests 3 conventional commit messages using a locally-running AI model (Ollama). Pick one, customize it, and commit — all from the terminal.

Why git commit-at?

git commit-at Manual commits Cloud AI tools
Private ✅ 100% local ✅ ❌ sends code to cloud
No API key ✅ ✅ ❌ requires subscription
Conventional commits ✅ auto ❌ manual discipline ⚠️ sometimes
Branch visualizer ✅ built-in ❌ ❌
Commit history ✅ tracked ❌ ❌

Install

npm install -g git commit-at

Requires: Docker (all AI/backend services run in containers — no extra setup).

How to use

# 1. Stage your changes
git add .

# 2. Run the tool
git commit-at

# 3. First time: register/login at http://localhost:7860
# 4. Pick a suggested commit message
# 5. Confirm — done!

On first run, Docker starts Ollama, Redis, and the Gradio UI automatically.

Features

  • 🤖 3 AI suggestions per run, tailored to your actual diff
  • 📋 Conventional commits — feat:, fix:, refactor:, chore:, etc.
  • 🎫 Ticket integration — prefix messages with JIRA/Linear ticket numbers
  • 🌿 Branch visualizer — live git branch DAG in the web UI
  • 📊 Commit history — track everything you've committed across repos
  • 🔐 Session persistence — stay logged in between runs
  • 🐳 Zero local dependencies — everything runs in Docker

Numbers

(TBH — to be updated with real metrics)

Metric Value
Average suggestion time ~5–10 s
First-run setup time ~30 s
AI model size ~1 GB (qwen2.5-coder:1.5b)
Runs entirely offline after setup ✅

Links

Install now

View on npm · GitHub

Quick Start

# Install
npm install -g git commit-at

# Go to any git repo
cd your-project
git add .

# Generate commit message
git commit-at

The first run starts all Docker services and opens the login UI at http://localhost:7860. After login, you'll see 3 commit message suggestions based on your staged diff.

How it works

git commit-at (CLI)
      │
      ├── Checks Docker services
      │     ├── Ollama  (AI model, port 11434)
      │     ├── Redis   (session + cache, port 6379)
      │     └── Gradio  (web UI, port 7860)
      │
      ├── Waits for login at http://localhost:7860
      │
      └── Runs commit flow
            ├── Reads git diff of staged files
            ├── Sends diff to Ollama (qwen2.5-coder:1.5b)
            ├── Streams 3 commit suggestions
            ├── Prompts: ticket number? format?
            ├── You pick a message
            └── git commit -m "..."