Local AI Git Commit Generator

Generate meaningful, conventional git commit messages using any local Ollama AI model — completely offline, no API keys needed.

Features

• AI-Powered Commit Messages — Generates Conventional Commits format messages by analyzing your diffs
• Any Ollama Model — Works with any model installed on your Ollama instance (qwen2.5-coder, llama3, codellama, mistral, deepseek-coder, etc.)
• Per-File Messages — Each changed file gets its own tailored commit message
• Editable Messages — Click on any generated message to modify it before committing
• Batch Commit — Select multiple files with checkboxes and commit them sequentially
• Stage / Unstage — Stage or unstage files directly from the panel
• Background Generation — Messages are generated in the background as you code
• Project Context Aware — Scans your project structure to generate more accurate messages
• Fully Offline — All AI processing happens locally via Ollama. No data leaves your machine.

Prerequisites

1. Ollama installed and running locally

   # macOS / Linux
   curl -fsSL https://ollama.ai/install.sh | sh
   
   # Then start the server
   ollama serve
   

2. Pull a model (any model works)

   # Recommended for code tasks
   ollama pull qwen2.5-coder:14b
   
   # Or use any other model
   ollama pull llama3
   ollama pull codellama
   ollama pull mistral
   ollama pull deepseek-coder:6.7b
   

Installation

From VS Code Marketplace

1. Open VS Code
2. Go to Extensions (Cmd+Shift+X / Ctrl+Shift+X)
3. Search for "Local AI Git Commit Generator"
4. Click Install

From VSIX File

1. Download the .vsix file from the Releases page
2. In VS Code: Cmd+Shift+P → "Extensions: Install from VSIX..."
3. Select the downloaded file

Usage

Open the Panel

• Command Palette: Cmd+Shift+P → Local AI Git Commit: Generate Commit Messages
• Keyboard Shortcut: Cmd+Shift+G Cmd+Shift+M (Mac) / Ctrl+Shift+G Ctrl+Shift+M (Windows/Linux)

Workflow

1. Make changes to your code as usual
2. Open the panel — it detects all changed files automatically
3. Review messages — AI generates a commit message for each file
4. Edit if needed — click on any message to modify it
5. Stage & Commit — use the per-file buttons, or:
6. Batch commit — check multiple files → click "Commit Selected"

Per-File Actions

Batch Commit

1. Select files using checkboxes (or "Select All")
2. Click "Commit Selected (N)"
3. Files are committed one at a time, sequentially — each commit is verified before the next starts
4. A live progress panel shows success/failure per file

Configuration

Open VS Code Settings (Cmd+,) and search for "Local AI Commit":

Example: Using a Different Model

// settings.json
{
  "localAiCommit.model": "llama3",
  "localAiCommit.temperature": 0.1
}

Example: Remote Ollama Server

{
  "localAiCommit.ollamaUrl": "http://192.168.1.100:11434"
}

Commit Message Format

All generated messages follow Conventional Commits:

type(scope): description

Allowed Types

Rules

• Scope is optional and always lowercase
• Description is lowercase, no trailing period
• Max header length: 100 characters

Examples

feat(auth): add biometric login support
fix(api): handle null response from server
refactor(home): simplify feed list rendering
chore: update dependencies to latest versions
test(utils): add unit tests for date formatter

Troubleshooting

"Ollama is not running"

# Start the Ollama server
ollama serve

"Model not found"

# List installed models
ollama list

# Pull the model configured in settings
ollama pull qwen2.5-coder:14b

Messages are generic or wrong

• Make sure diffs are not too large (the extension truncates at 12KB per file)
• Try a larger/better model: qwen2.5-coder:14b or deepseek-coder:33b
• Lower the temperature to 0.1 for more predictable output

Extension not activating

• Ensure you have a workspace folder open (not just a single file)
• The folder must be a git repository (git init if not)

Building from Source

git clone <repo-url>
cd local-ai-git-commit-generator
npm install
npm run build

# Package as .vsix
npx @vscode/vsce package

License

MIT

Setup

npm install
npm run build

Running the Extension

1. Open this folder in VS Code
2. Press F5 (or Run → Start Debugging)
3. A new VS Code window (Extension Development Host) opens
4. Open any git repository in that window
5. Open the Command Palette (Cmd+Shift+P) and run: AI Commit: Generate Commit Message

Usage

• The panel shows all staged + unstaged changed files
• Each file gets an AI-generated commit message (streaming)
• Edit any message manually by clicking the text
• Regenerate per file with the ↻ button
• Toggle between Per-file and Combined commit modes
• Click Commit when ready

Keyboard Shortcut

Cmd+Shift+G Cmd+Shift+M (macOS) / Ctrl+Shift+G Ctrl+Shift+M (Windows/Linux)

Architecture

src/
├── extension.ts          # Entry point, command registration
├── types.ts              # Shared TypeScript types
├── gitService.ts         # Git operations (simple-git)
├── ollamaService.ts      # Ollama API with streaming + concurrency
├── panelProvider.ts       # Webview panel lifecycle & message handling
└── webview/
    └── webviewContent.ts  # Full HTML/CSS/JS for the webview UI