VibeCommit 🚀

Offline AI agent for Git workflows—voice-dictated commits, smart summaries, and agentic checks. Built in Swift for macOS devs who vibe code without cloud BS.

Inspired by agentic engineering: No more stalled prompts or manual loops. Just dictate "Commit vibe: fixed auth bug," and it generates summaries with local gpt-oss-20b, verifies diffs, and pushes if ready.

✨ Features

• 📝 Auto-generate commit messages from recent commits (last 10) using local AI (gpt-oss-20b).
• 🔍 Vibe checks: Validates summary length and optionally auto-commits.
• 🎤 Voice dictation integration with macOS native Speech framework (transcribe vibes hands-free).
• ⚙️ Extensible CLI with flags for mock mode (testing without AI) and auto-commit.
• Zero config: Runs in any Git repo.

🚀 Quick Start

1. Clone: git clone https://github.com/ClayWarren/VibeCommit.git
2. Build: swift build
3. Run examples (in a Git repo):
   • Basic summary: swift run VibeCommit --summarize
   • With voice input: swift run VibeCommit --summarize --voice
   • Mock mode for testing: swift run VibeCommit --summarize --voice --mock
   • Auto-commit: swift run VibeCommit --summarize --autoCommit (prompts for confirmation)

Setup Notes

• AI Features: Install gpt-oss-20b via Hugging Face for real summaries (skip with --mock). Requires Python 3+ with transformers library: pip install transformers. The model will auto-download on first use.
• Voice Permissions (macOS): On first run with --voice, grant microphone and speech recognition access via System Settings > Privacy & Security. If no prompt appears, reset with tccutil reset Microphone and tccutil reset SpeechRecognition in Terminal.
• Dependencies: Swift 6.2+, Git, Python for AI bridge.

📅 Roadmap

• Phase 1: Core CLI with Git hooks and basic AI summaries. ✅
• Phase 2: Voice input and agent completions. ✅
• Future: Integrations (visual diffs, MCP for Claude fallback if local fails), full Git push automation.

Contributions welcome! See Issues for "good first issue" tags.

MIT © 2025 Clay Warren