commitwiz 🤖

AI-powered git commit message generator. Stage your changes, run commitwiz, get perfect Conventional Commits — every time.

$ git add -A && commitwiz

🤖 Analyzing diff... (3 files changed, +47 -12)

  feat(auth): add JWT refresh token rotation

  Implement automatic refresh token rotation on each access token
  renewal. Expired refresh tokens are now invalidated server-side
  to prevent replay attacks.

? [C]ommit  [E]dit  [R]egenerate  [Q]uit: C
✅ Committed!

Why?

• Consistent history — No more fix stuff, wip, or asdf. Every commit follows Conventional Commits format.
• Context-aware — Reads your actual diff, not just file names. Understands what changed and why.
• Fast — Uses gpt-4o-mini by default. Most commits generate in under 2 seconds.
• Works with local models — Ollama, LM Studio, or any OpenAI-compatible API. Your code never has to leave your machine.
• Interactive — Review before committing. Edit, regenerate, or bail out.

Install

npm install -g commitwiz

Or run without installing:

npx commitwiz

Quick Start

# 1. Set your API key (or use a local model — see below)
export OPENAI_API_KEY=sk-...

# 2. Stage changes
git add -A

# 3. Generate and commit
commitwiz

Usage

# Interactive mode (default) — review, edit, regenerate, or quit
commitwiz

# Auto-commit without confirmation
commitwiz --auto

# Use a specific model
commitwiz --model gpt-4o

# Use a local model via Ollama
OPENAI_BASE_URL=http://localhost:11434/v1 commitwiz --model llama3

Features

Feature	Details
🤖 AI-powered	Uses OpenAI (or any compatible API) to analyze your staged diff
📝 Conventional Commits	Always generates properly formatted type(scope): description messages
⚡ Fast	gpt-4o-mini by default — sub-2-second generation
🔄 Interactive	Commit, edit, regenerate, or quit — you're always in control
🏠 Local models	Works with Ollama, LM Studio, or any OpenAI-compatible endpoint
🔒 Private	Use local models and your code never leaves your machine

Configuration

OpenAI (default)

export OPENAI_API_KEY=sk-...

Ollama (local, free)

# Install Ollama and pull a model
ollama pull llama3

# Point commitwiz at Ollama
export OPENAI_BASE_URL=http://localhost:11434/v1
commitwiz --model llama3

LM Studio (local, free)

# Start LM Studio's local server, then:
export OPENAI_BASE_URL=http://localhost:1234/v1
export OPENAI_API_KEY=lm-studio
commitwiz

Any OpenAI-compatible API

export OPENAI_BASE_URL=https://your-api.example.com/v1
export OPENAI_API_KEY=your-key
commitwiz --model your-model

Recipes

Git hook (auto-generate on every commit)

# .git/hooks/prepare-commit-msg
#!/bin/sh
if [ -z "$2" ]; then
  MSG=$(commitwiz --auto --stdout 2>/dev/null)
  [ -n "$MSG" ] && echo "$MSG" > "$1"
fi

Team workflow with commitlint

# Install commitlint for validation
npm install -D @commitlint/cli @commitlint/config-conventional

# commitwiz generates → commitlint validates
commitwiz --auto && npx commitlint --edit

CI: Validate commit messages in pull requests

# .github/workflows/commitlint.yml
name: Lint Commits
on: [pull_request]
jobs:
  commitlint:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with: { fetch-depth: 0 }
      - run: npx commitlint --from ${{ github.event.pull_request.base.sha }}

Quick alias

# Add to ~/.bashrc or ~/.zshrc
alias gc='git add -A && commitwiz'

Batch commits (monorepo)

# Commit each package separately
for dir in packages/*/; do
  cd "$dir"
  git add -A
  commitwiz --auto
  cd ../..
done

How It Works

1. Reads your staged diff via git diff --cached
2. Sends the diff to your configured LLM with a system prompt enforcing Conventional Commits format
3. Presents the message for review (or auto-commits with --auto)
4. Commits using git commit -m "..." on confirmation

The prompt ensures:

• Correct type(scope): description format
• Imperative mood ("add" not "added")
• Concise subject line (≤72 chars)
• Optional body for complex changes

Comparison

Tool	AI-Powered	Local Models	Interactive	Conventional Commits	Zero Config
commitwiz	✅	✅	✅	✅	✅
aicommits	✅	❌	❌	❌	✅
cz-cli	❌	N/A	✅	✅	❌
commitlint	❌	N/A	❌	✅ (validation only)	❌
Manual	❌	N/A	N/A	Sometimes	✅

Use Cases

• Solo developers — Stop agonizing over commit wording. Let AI draft, you approve.
• Teams — Enforce consistent Conventional Commits without training everyone on the spec.
• Open source — Clean commit history makes changelogs and releases trivial.
• Monorepos — Scope-aware messages help track which package changed.
• CI/CD — Conventional Commits enable automatic versioning with semantic-release.

Related Tools

• env-lint-cli — Validate .env files before deploying
• dep-size — Check npm dependency sizes before installing
• license-maker — Generate LICENSE files from the CLI
• npm-name-check — Check npm package name availability
• kill-port-cli — Kill processes on any port

License

MIT