⚡ Zerqon — Real-Time AI Meeting Assistant

Listen. Understand. Suggest. Summarise.

Zerqon is a Windows desktop app that sits quietly in your system tray and assists you during meetings. It detects when you join a call, transcribes speech in real time, suggests what to say next, and automatically summarises the meeting when it ends.

---

Screenshots

---

Features

• Auto-detects meetings — Opens automatically when Zoom, Teams, Google Meet, Discord, Slack Huddle, and more are detected
• Real-time transcription — Speech recognition runs 100% locally using Whisper (no audio leaves your machine)
• AI suggestions — Powered by Google Gemini, suggests what to say next as the conversation develops
• Meeting summaries — Automatically generates structured summaries (key points, decisions, action items) when the call ends
• Headphone support — Works with any Windows audio input: built-in mic, wired headset, USB mic, Bluetooth
• System tray app — Runs silently in the background, zero interference with your workflow
• One-click installer — No technical knowledge needed

---

Supported Meeting Apps

App	Detection Method
Zoom	Process + Browser
Microsoft Teams	Process + Browser
Google Meet	Browser window title
Discord	Process
Slack Huddle	Process
Cisco Webex	Process
Skype	Process
GoTo Meeting	Process
RingCentral	Process
Whereby, Jitsi	Browser window title

---

Installation

For end users (no technical knowledge needed)

1. Go to the Releases page
2. Download Zerqon.Setup.exe
3. Double-click it and follow the setup wizard
4. You will need a free Gemini API key

For developers (build from source)

Prerequisites — you need these installed:

• Python 3.11 — tick "Add to PATH" during install
• Node.js 20 LTS
• Git

Clone and build:

# Clone the repo
git clone https://github.com/YOUR_USERNAME/zerqon.git
cd zerqon

# Create your .env file with your Gemini API key
echo GEMINI_API_KEY=AIza...your_key_here > .env

# Run the one-click setup (installs all dependencies and builds the app)
INSTALL.bat

The setup script handles everything else automatically — Python packages, Node modules, PyInstaller bundling, and Electron packaging.

For development (without building the full installer):

# Terminal 1 — Python backend
python -m venv venv
venv\Scripts\activate
pip install -r backend/requirements.txt
python -m backend.main

# Terminal 2 — Electron frontend
cd frontend
npm install
npm start

---

How It Works

Microphone / Headset
        │
        ▼
  sounddevice (3s chunks)
        │
        ▼
  faster-whisper (local STT, ~0.5s)
        │
        ▼
  Rolling 300-word context window
        │
        ▼
  Google Gemini API (~0.8s)
        │
        ▼
  WebSocket → Electron overlay
        │
        ▼
  Suggestion shown on screen

Total latency: ~1.5–2.5 seconds end to end.

Speech recognition (Whisper) runs entirely on your CPU — no audio is ever sent to any server. Only the text transcript is sent to Gemini for suggestion generation.

---

Project Structure

zerqon/
├── backend/
│   ├── audio/capture.py          # Mic capture via sounddevice
│   ├── transcriber/whisper_stt.py # Local STT via faster-whisper
│   ├── llm/handler.py            # Gemini suggestions + summaries
│   ├── server/ws_server.py       # WebSocket bridge to Electron
│   ├── context_manager.py        # Rolling context window
│   ├── meeting_detector.py       # Detects meeting apps + browser calls
│   ├── meeting_summarizer.py     # Generates + saves meeting summaries
│   ├── config.py                 # All settings (reads from .env)
│   └── main.py                   # Entry point
├── frontend/
│   ├── main.js                   # Electron main process
│   └── windows/
│       ├── setup.html            # First-run wizard
│       ├── overlay.html          # Live suggestion overlay
│       ├── prompt.html           # Meeting detected popup
│       ├── summary.html          # Meeting summary viewer
│       └── settings.html         # Settings window
├── installer/
│   └── custom.nsh                # NSIS installer hooks
├── zerqon-backend.spec           # PyInstaller config
├── INSTALL.bat                   # One-click setup script
├── setup.ps1                     # PowerShell automation
└── .env.example                  # Template for your API key

---

Configuration

Copy .env.example to .env and fill in your values:

GEMINI_API_KEY=AIza...your_key_here
GEMINI_MODEL=gemini-1.5-flash
WHISPER_MODEL=base
WHISPER_LANGUAGE=en
SUGGESTION_TRIGGER_WORDS=20

Whisper model sizes:

Model	Speed	Accuracy	VRAM
tiny	Fastest	Lower	~1GB
base	Fast (recommended)	Good	~1GB
small	Medium	Better	~2GB
medium	Slow	High	~5GB

---

Tech Stack

Component	Technology
Speech recognition	faster-whisper (local)
AI suggestions	Google Gemini (free tier)
Audio capture	sounddevice
Frontend	Electron
IPC bridge	WebSocket (localhost)
Meeting detection	psutil + pygetwindow
Packaging	PyInstaller + electron-builder

---

Privacy

• Audio never leaves your machine. Whisper runs 100% locally on your CPU.
• Only text is sent to Gemini. The transcript (not audio) is sent to Google's API for suggestion generation.
• Your API key is stored locally in a .env file on your PC. It is never transmitted anywhere except directly to Google's API.
• Meeting summaries are saved to Documents\Zerqon\Summaries\ on your local machine only.

---

Contributing

Pull requests are welcome. For major changes, please open an issue first.

1. Fork the repo
2. Create your feature branch: git checkout -b feature/my-feature
3. Commit your changes: git commit -m 'Add my feature'
4. Push to the branch: git push origin feature/my-feature
5. Open a Pull Request

---

Support the project

If Zerqon has saved you from an awkward silence, consider buying me a coffee:

---

License

MIT License — free to use, modify, and distribute.