Real-Time Interview Copilot

An AI-powered desktop application that provides real-time interview copilot. The app listens to live audio, transcribes speech in real-time using OpenAI Whisper, detects questions, and generates professional answers using OpenAI GPT-4.

Features

• 🎤 Real-time Speech Recognition - Captures microphone audio and transcribes in real-time using OpenAI Whisper
• 🤖 AI-Powered Answers - Generates professional interview answers using OpenAI GPT-4
• 🔒 Screen Share Safe - Window is hidden from screen sharing (Zoom, Teams, Google Meet)
• ⚡ Low Latency - Sub-2 second response time from question detection to answer generation
• 🎯 Smart Question Detection - Uses linguistic patterns and pause detection to identify questions
• 📌 Always on Top - Keeps the copilot visible during interviews
• 🌙 Dark Theme - Easy on the eyes during long interview sessions

Tech Stack

• Electron - Cross-platform desktop app framework
• React 19 - UI library with hooks
• TypeScript - Type-safe JavaScript
• Tailwind CSS - Utility-first CSS framework
• Zustand - Lightweight state management
• OpenAI SDK - GPT-4o answer generation

Prerequisites

Before running this application, you'll need:

1. Node.js (v18 or higher)
2. pnpm package manager
3. OpenAI API Key - Get one at platform.openai.com

Installation

1. Clone the repository:

git clone <repository-url>
cd interview-copilot

2. Install dependencies:

pnpm install

Development

Run the app in development mode with hot reload:

pnpm dev

Building

Build for Windows:

pnpm build:win

Build for other platforms:

pnpm build:mac    # macOS
pnpm build:linux  # Linux

Usage

1. Launch the app - The settings modal will open if API keys aren't configured
2. Configure API keys - Enter your OpenAI API key in Settings
3. Click "Start" - Begin listening for interview questions
4. Speak or let the interviewer speak - The app transcribes audio in real-time
5. Wait for question detection - When a pause is detected after a question, an answer is generated
6. View suggested answers - Professional answers appear in the bottom panel
7. Click "Stop" when done

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                        Electron Main Process                     │
│  ┌─────────────────┐  ┌─────────────────┐  ┌─────────────────┐  │
│  │  Audio Capture  │  │  OpenAI Whisper │  │  OpenAI LLM     │  │
│  │  (System Audio) │  │  (WebSocket)    │  │  (Answer Gen)   │  │
│  └─────────────────┘  └─────────────────┘  └─────────────────┘  │
└─────────────────────────────────────────────────────────────────┘
                              │ IPC
┌─────────────────────────────────────────────────────────────────┐
│                      Electron Renderer (React)                   │
│  ┌─────────────────┐  ┌─────────────────┐  ┌─────────────────┐  │
│  │  Mic Capture    │  │  Transcript     │  │  Answer         │  │
│  │  (Web Audio)    │  │  Display        │  │  Display        │  │
│  └─────────────────┘  └─────────────────┘  └─────────────────┘  │
└─────────────────────────────────────────────────────────────────┘

Project Structure

src/
├── main/
│   ├── index.ts                    # Main process entry
│   ├── ipc/
│   │   └── handlers.ts             # IPC handlers
│   └── services/
│       ├── whisperService.ts      # Speech-to-text service
│       ├── openaiService.ts        # Answer generation service
│       ├── questionDetector.ts     # Question detection logic
│       └── settingsManager.ts      # Settings persistence
├── preload/
│   ├── index.ts                    # Preload scripts
│   └── index.d.ts                  # Type definitions
└── renderer/src/
    ├── App.tsx                     # Main React component
    ├── components/
    │   ├── Header.tsx              # Title bar with controls
    │   ├── StatusBar.tsx           # Recording status
    │   ├── TranscriptPanel.tsx     # Live transcription display
    │   ├── AnswerPanel.tsx         # Generated answers display
    │   └── SettingsModal.tsx       # Settings configuration
    ├── hooks/
    │   ├── useAudioCapture.ts      # Audio capture hook
    │   └── useInterview.ts         # Main interview logic
    ├── store/
    │   └── interviewStore.ts       # Zustand state management
    └── services/
        └── audioCapture.ts         # Web Audio implementation

Configuration Options

Setting	Description	Default
OpenAI API Key	Key for Whisper speech-to-text service	Required
OpenAI API Key	Key for answer generation	Required
OpenAI Model	GPT model to use	gpt-4o-mini
Pause Threshold	Silence duration before processing (ms)	1500
Window Opacity	Window transparency	100%
Always on Top	Keep window above others	Enabled

Security Features

• Content Protection - Window is hidden from screen capture software
• Secure Storage - API keys are encrypted using Electron's safeStorage API in local storage
• Context Isolation - Renderer process is isolated from main process
• CSP Headers - Content Security Policy restricts resource loading

Troubleshooting

Microphone not working

• Ensure microphone permissions are granted in Windows Settings
• Try selecting a different audio input device

No transcription appearing

• Check that your OpenAI API key is valid
• Ensure you have an active internet connection

Answers not generating

• Verify your OpenAI API key has sufficient credits
• Check the console for error messages

Window visible in screen share

• The content protection feature works on Windows 10/11
• Some older screen sharing methods may still capture the window

License

MIT License - See LICENSE file for details

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

Interview Copilot