ghostshell-ai
v1.0.1
Published
AI-powered real-time interview copilot. Captures audio, transcribes with Whisper, generates answers with Gemma AI -- all in an invisible overlay.
Maintainers
Readme
GhostShell AI
Real-time AI interview copilot for Windows. Captures audio, transcribes with Whisper, generates answers with Gemma AI -- all in an invisible overlay.
Install
npm install -g ghostshell-aiUsage
ghostshell # Launch the app
ghostshell --setup # Re-configure API keys
ghostshell --persona resume.txt # Use custom resume
ghostshell --version
ghostshell --helpFirst Run
On first launch, you'll be asked to choose:
- Use default keys -- works immediately, shared free-tier quota
- Use your own keys -- browser opens to get free API keys (2 minutes)
Keys are saved locally in ~/.ghostshell/config.json and never uploaded anywhere.
How It Works
- Press REC to capture audio (interviewer or your mic)
- Press STOP -- audio is transcribed by Whisper in <1 second
- AI generates an answer in ~2 seconds using your resume context
- Answer appears in the transparent overlay
The overlay is invisible to screen sharing (Windows Display Affinity).
Commands
| Key | Action |
|-----|--------|
| Ctrl+Shift+G | Toggle click-through mode |
| REC / STOP | Toggle recording |
| SCAN | Screenshot analysis (MCQ, coding, diagrams) |
Requirements
- Windows 10/11
- Node.js 16+
API Keys (Free)
| Service | What For | Get Key | |---------|----------|---------| | Google AI Studio | Gemma 3 AI model | aistudio.google.com/apikey | | Groq | Whisper transcription | console.groq.com |
Both are completely free with generous limits.
License
MIT
