KarmaMessageAI

Generate perfect Git commit messages using OpenAI, local models through Ollama, or your Claude Code subscription, following the Karma / Conventional Commits specification.

!Version !VSCode !License

Features

• ✨ One-click generation — button in the Source Control panel header
• Cancellable — stop a slow generation from the progress notification
• Smart diff — skips lockfiles, dist/ and minified files, and shares the size budget fairly between files
• Editable preview — review and modify the message before applying (Ctrl+Enter to confirm)
• Karma convention — enforces feat, fix, docs, refactor, and all standard types
• Message history — side panel with the last 50 generated messages (click to re-apply)
• Claude subscription support — use the Claude Code CLI with your existing login, no API key
• Secure API key storage — stored in VS Code's encrypted SecretStorage (OS Keychain)
• Configurable model — choose between gpt-4o-mini, gpt-4o, gpt-4-turbo, and gpt-3.5-turbo

Karma Convention

Messages follow this format:

<type>(<scope>): <subject>

<body>

<footer>

Examples:

feat(auth): add JWT refresh token rotation

Refresh tokens now rotate on each use to prevent token theft.

Closes [#482](https://github.com/danielboggianosa/vscode-karma-message-ai/issues/482)

fix(api): handle null response from payment gateway

chore(deps): update dependencies to latest versions

Requirements

• VS Code ^1.85.0
• Git installed and available in your PATH
• One of: an OpenAI API Key, Ollama with a downloaded local model, or the Claude Code CLI logged in with your Claude subscription.

Installation

Option A — Install from VSIX (recommended)

1. Build the package:

   npm install
   node esbuild.js --production
   npx vsce package --no-dependencies
   

2. In VS Code: Ctrl+Shift+P → Extensions: Install from VSIX...
3. Select the generated karma-message-ai-1.1.0.vsix file
4. Reload VS Code

Option B — Development mode (for testing)

1. Open the Commit Message VsCode folder in VS Code
2. Open src/extension.ts
3. Press F5 — a new Extension Development Host window will open
4. In that window, open any Git repository and start using the extension

Setup

1. Choose a provider

OpenAI remains the default. For a local option without an API key or per-request API fees:

1. Install and start Ollama.

2. Download the default model: ollama pull qwen2.5-coder:3b (about 1.9 GB).

3. Add to VS Code settings:

   "commitAI.provider": "ollama",
   "commitAI.ollama.model": "qwen2.5-coder:3b"
   

Keep Ollama running; use ollama serve if the service is not already active. After downloading the model, local generation works offline. Speed and quality depend on your hardware and selected model. Change commitAI.ollama.model to another model you have downloaded. Local models use your computer's RAM and processing resources.

The default server is http://localhost:11434. Configure commitAI.ollama.baseUrl if needed and commitAI.ollama.timeout (default 120 seconds) for slow model loading. With Remote SSH, WSL or containers, localhost refers to the machine running the extension host. The server must be reachable from there.

See Ollama authentication: local API requests need no key; cloud models are a separate service and may require sign-in.

Claude Code setup (use your Claude subscription)

1. Install the Claude Code CLI and log in once by running claude in a terminal.

2. Add to VS Code settings:

   "commitAI.provider": "claude-code",
   "commitAI.claudeCode.model": "haiku"
   

The extension runs claude -p with no tools and no saved session, from a temporary directory so project CLAUDE.md files are not included. Usage counts against your Claude subscription. ANTHROPIC_API_KEY is removed from the CLI's environment so requests are not billed to the API. If VS Code cannot find the CLI (it may not inherit your shell PATH), set commitAI.claudeCode.path to the absolute path, e.g. /Users/you/.local/bin/claude.

OpenAI setup

Set commitAI.provider to openai and configure your API key:

Open the Command Palette (Ctrl+Shift+P) and run:

Commit AI: Set OpenAI API Key

Paste your key (sk-...). It is stored securely using VS Code's built-in SecretStorage — it never touches disk in plain text.

2. Stage your changes

git add .
# or stage specific files in the Source Control panel

3. Generate the commit message

Click the ✨ button in the Source Control panel header.

The message appears in an editable panel — review it, make any tweaks, then click Apply to Commit or press Ctrl+Enter.

Usage

Extension Settings

Configure via File → Preferences → Settings and search for Commit AI:

Project Structure

src/
├── extension.ts          # Entry point — registers commands and views
├── gitService.ts         # Reads staged diff, writes to SCM input box
├── openaiService.ts      # Calls OpenAI, Ollama or Claude Code; manages API key
├── commitPreviewPanel.ts # Webview panel for editing the generated message
└── historyProvider.ts    # TreeView provider for the history sidebar

Development

# Install dependencies
npm install

# Build (development)
node esbuild.js

# Build (production / minified)
node esbuild.js --production

# Watch mode
node esbuild.js --watch

# Run tests
npm test

# Type-check only
npx tsc --noEmit

# Package as .vsix
npx vsce package --no-dependencies

Privacy

• With OpenAI selected, your git diff is sent to OpenAI's API.
• With Claude Code selected, your diff is sent to Anthropic through the Claude Code CLI under your subscription; see Anthropic's privacy policy.
• With Ollama selected, your diff is sent only to the configured Ollama server. A local server with a local model keeps generation on that machine; remote servers or cloud models have different privacy implications.
• The API key is stored locally in VS Code's SecretStorage (OS Keychain).
• No data is stored or sent anywhere else.
• Review OpenAI's privacy policy for details on how your data is handled.

License

MIT © Daniel Boggiano