"A calmer, safer, and more explainable way to use Git — powered by voice and AI."
Gitty is a desktop application that transforms the Git experience into a voice-driven, anxiety-free workflow.
Built with Electron, React, and TypeScript, it combines Speech Recognition, a Groq LLM, and a real PTY terminal to help developers perform Git operations using natural language — with full transparency and zero risk.
Simply speak your intent:
“Create a new branch for the login feature.”
…and the app:
- Transcribes your speech.
- Uses an LLM (via Groq API) to interpret intent.
- Takes a live snapshot of your repository (branch, upstream, changes, etc.).
- Displays the exact Git command and a plain-English explanation.
- Waits for your confirmation — you stay in control.
No destructive actions.
No blind execution.
No Git anxiety.
Even experienced developers feel a twinge of stress before running high-impact Git commands.
Syntax errors, missing upstreams, or accidental resets can have serious consequences in shared codebases.
The Voice-Controlled Git Coach helps professionals:
- Understand what each command will do before running it.
- Trust that actions are safe and explainable.
- Learn the rationale behind commands over time.
- Reduce cognitive load under time pressure.
Alex Chen — Mid-level software engineer, comfortable with Git but often juggling multiple branches and deadlines.
He knows Git inside out, but that small voice of doubt — “Am I about to break something?” — never quite goes away.
This app quiets that voice by showing him exactly what will happen before anything runs.
| Feature | Description |
|---|---|
| 🎙️ Voice-to-Git | Speak natural commands like “commit my changes” or “undo the last commit.” |
| 🧠 Context-Aware AI | Uses live repo snapshots (branch, upstream, ahead/behind, uncommitted changes) to interpret intent accurately. |
| 💬 Plain-English Explanations | Every command is accompanied by a short, human-readable explanation of its impact. |
| ✅ Pre-flight Confirmation | The app never executes automatically — you always see the command first. |
| 🧱 Real Terminal Core | Built on child_processes and xterm.js; runs genuine shell commands, not simulations. |
| 🧭 Minimalist UI | A clean terminal interface with only essential overlays: Idle / Listening / Parsing / Ready. |
User Speaks
↓
Vosk (Speech-to-Text)
↓
Repo Snapshot Collector (branch, upstream, status, ahead/behind)
↓
Groq LLM (Intent + JSON Response)
↓
Command Validation (One line explanation)
↓
Confirmation Overlay (Command + Explanation)
↓
User Confirms → executes command → Output streams to xterm.js
| Layer | Technology |
|---|---|
| Desktop Shell | Electron |
| Frontend | React + Vite + TypeScript |
| Terminal | xterm.js |
| AI Engine | Groq LLM API (Llama 3.3 70B) |
| Speech Recognition | Vosk |
| Validation | Zod + jsonrepair |
| Styling | Tailwind CSS |
| Packaging | electron-builder |
- Node.js: ≥ 22.12 (or 20.19 LTS)
- npm
- Groq API Key (obtain from console.groq.com)
- Only Possible on MacOS :(
git clone [email protected]:pauravhp/git-helper.git
cd git-helper
npm installnpm run transpile:electronnpm run dist:mac # macOSCreate a .env file in the project root:
GROQ_API_KEY=your_api_key_herenpm run devThe packaged app will appear in /dist.
- Human confirmation required: Every command requires user consent.
- Explain before execute: Transparency first, automation second.
- Add fine-grained permission levels for org/team settings.
- Integrate with VS Code as an extension.
- Offline LLM fallback (e.g., Mistral 7B quantized).
- Multi-language STT support.
- Advanced “scenario explanations” for educational use.
Paurav Hosur Param Abhay Anoop C Vivek Reddy Daniaal Tahir Mahmood
Git should feel empowering — not intimidating.
The Voice-Controlled Gitty brings clarity, confidence, and calm to one of the most essential tools in software development.
“See before you run.”
Understand Git. Reduce anxiety. Stay in control.
MIT © 2025 Voice-Controlled Git Coach Team