🎯 Interview Copilot

Real-time AI-powered interview assistant with transparent overlay, speech-to-text, and intelligent response generation using Azure GPT-4o.

✨ Features

• 🎤 Real-time Speech-to-Text: Captures and transcribes interview questions using Deepgram Nova-2
• 🤖 AI-Powered Responses: Generates intelligent answers using Azure GPT-4o with streaming
• 📸 Screen Analysis: Analyzes coding problem screenshots with GPT-4o Vision
• 👁️ Transparent Overlay: Always-on-top window that doesn't interfere with your interview
• ⚡ Ultra-Low Latency: < 4 seconds from speech to answer
• 🎮 Global Shortcuts: Control without switching focus

🏗️ Architecture

Backend (Python + FastAPI)

• WebSocket server for bi-directional streaming
• Deepgram integration for speech-to-text
• Azure OpenAI GPT-4o for response generation
• Smart routing for coding vs. general questions

Frontend (Electron + React + TypeScript)

• Desktop overlay application
• System audio capture
• Screen capture functionality
• Real-time UI updates with streaming responses

🚀 Quick Start

Prerequisites

• Python 3.10+
• Node.js 18+
• Azure OpenAI API key with GPT-4o access
• Deepgram API key

1. Setup Backend

cd backend
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

Create backend/.env:

AZURE_OPENAI_API_KEY=your_azure_key
AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
AZURE_OPENAI_DEPLOYMENT_NAME=gpt-4o
DEEPGRAM_API_KEY=your_deepgram_key

Run server:

python run.py

2. Setup Frontend

cd frontend
npm install
npm run dev

📖 Documentation

• ⚡ Quick Start Guide - Get started in 5 minutes!
• Complete Setup Guide - Detailed installation and configuration
• Project Specification - Technical architecture and design
• Backend Documentation - API and WebSocket events
• Frontend Documentation - Component structure and services

🎮 Usage

Global Shortcuts

• Ctrl+Shift+L - Toggle audio listening
• Ctrl+Shift+S - Capture screen for coding problems
• Ctrl+Shift+H - Hide/Show overlay

Workflow

1. Start the backend server
2. Launch the frontend app
3. Start listening to capture interview questions
4. Get instant AI responses displayed in the overlay
5. Capture screens for coding problem analysis

🔧 Tech Stack

Component	Technology	Purpose
Frontend	Electron + React + TypeScript	Desktop overlay UI
Backend	FastAPI + Python	AI orchestration
Speech-to-Text	Deepgram Nova-2	Real-time transcription
LLM	Azure GPT-4o	Response generation
Vision	GPT-4o Vision	Screenshot analysis
Communication	WebSocket (Socket.IO)	Streaming data
State Management	Zustand	Frontend state

📊 Performance Targets

• End-to-End Latency: < 4 seconds (Speech → Answer)
• STT Latency: ~300ms (Deepgram Nova-2)
• LLM Streaming: Real-time token delivery
• Audio Sample Rate: 16kHz Linear16 PCM

🔒 Security

• API keys stored in environment variables
• No sensitive data in frontend code
• Secure WebSocket communication
• User consent required for audio/screen capture

📁 Project Structure

Ai_dialer_response/
├── backend/              # Python FastAPI backend
│   ├── services/        # Deepgram & Azure OpenAI services
│   ├── main.py          # FastAPI app with WebSocket
│   ├── config.py        # Configuration management
│   └── requirements.txt # Python dependencies
├── frontend/            # Electron + React frontend
│   ├── electron/        # Electron main process
│   ├── src/
│   │   ├── components/  # React components
│   │   ├── services/    # WebSocket, Audio, Screen capture
│   │   ├── hooks/       # Custom React hooks
│   │   └── store/       # Zustand state management
│   └── package.json     # Node dependencies
├── SETUP_GUIDE.md       # Detailed setup instructions
└── README.md            # This file

🤝 Contributing

This project is built for interview assistance. Contributions are welcome!

📄 License

See LICENSE file for details.

⚠️ Disclaimer

This tool is designed for educational purposes and interview practice. Always ensure you have permission to use AI assistance during actual interviews. Use responsibly and ethically.