Interview Assistant

AI-powered interview assistant for Linux (Ubuntu) with screen-share avoidance capabilities. Captures interviewer audio, transcribes questions in real-time, and provides AI-generated answers via Claude.

Features

• Interview Mode: Real-time audio capture and transcription of interviewer questions
• AI Answers: Claude-powered responses with code examples and complexity analysis
• Screen-Share Avoidance: Multiple stealth modes to stay invisible during screen sharing
• Multiple Interview Types: DSA, System Design, and Behavioral modes with specialized prompts
• Glassmorphism UI: Beautiful, modern dark theme with semi-transparent panels
• History: SQLite-backed Q&A history with search and export
• System Tray: Quick access controls from system tray

Requirements

• Ubuntu 22.04+ (or other Linux with GTK4)
• Python 3.10+
• PipeWire or PulseAudio
• Anthropic API key

Installation

1. Install System Dependencies

sudo apt update
sudo apt install -y \
    libgtk-4-dev \
    gir1.2-gtk-4.0 \
    libadwaita-1-dev \
    gir1.2-adw-1 \
    libgtksourceview-5-dev \
    gir1.2-gtksource-5 \
    libpulse-dev \
    pipewire \
    pipewire-pulse \
    ffmpeg \
    python3-pip \
    python3-venv

2. Create Virtual Environment

python3 -m venv venv
source venv/bin/activate

3. Install Python Dependencies

pip install -e .
# Or using requirements.txt
pip install -r requirements.txt

4. Download Whisper Model

# The model will download automatically on first run
# Or pre-download with:
python -c "from faster_whisper import WhisperModel; WhisperModel('base')"

5. Set Up AI Backend

The app supports two AI backends:

Option A: Ollama (Recommended - Free & Offline)

1. Install Ollama:

   curl -fsSL https://ollama.com/install.sh | sh

2. Start Ollama:

   ollama serve

3. Pull a model (choose one):

   # Fast, good for coding (recommended)
   ollama pull llama3.1:8b
   
   # Best for coding interviews
   ollama pull deepseek-coder-v2:16b
   
   # Lightweight alternative
   ollama pull mistral:7b

4. The app uses Ollama by default - just run it!

Option B: Claude API (Cloud, Paid)

1. Create account at https://console.anthropic.com
2. Add billing and generate API key
3. Configure in the app:
   • Open Settings > AI
   • Select "Claude (Cloud, Paid)" as backend
   • Enter your API key

Usage

Run the Application

# From project directory
python -m interview_assistant

# Or if installed
interview-assistant

Interview Mode

1. Select Interview Type: Choose DSA, System Design, or Behavioral from the dropdown
2. Start Recording: Click the record button or press Ctrl+Alt+R
3. Speak Questions: The app captures system audio and transcribes questions
4. View Answers: AI-generated answers appear with syntax highlighting

Stealth Modes

Configure in Settings > Stealth:

• Normal: Standard window (visible in screen share)
• Overlay: X11 window tricks to bypass some capture tools
• Secondary Monitor: Shows only on secondary display
• Hotkey Popup: Brief popup that auto-hides

Keyboard Shortcuts

Shortcut	Action
Ctrl+Alt+R	Toggle recording
Ctrl+Alt+I	Toggle window visibility
Ctrl+Q	Quit application

Project Structure

interview_assistant/
├── core/           # Config, events, session management
├── audio/          # Audio capture and VAD
├── transcription/  # Whisper transcription
├── ai/             # Claude API integration
├── ui/             # GTK4 UI components
├── stealth/        # Screen-share avoidance
├── storage/        # SQLite database
├── services/       # System tray, shortcuts
└── resources/      # CSS themes, icons

Configuration

Configuration is stored in ~/.config/interview-assistant/config.toml

[ai]
api_key = "your-api-key"
model = "claude-sonnet-4-5-20250929"
max_tokens = 4096

[transcription]
model_size = "base"
language = "en"

[stealth]
mode = "normal"
auto_hide_timeout = 5000
opacity = 0.95

Limitations

Screen-Share Avoidance

There is no guaranteed method to hide windows from all screen capture tools. Effectiveness depends on:

• Screen capture method (X11 vs XDG Portal)
• Desktop environment and compositor
• Screen sharing mode (entire screen vs window)

Recommendation: Secondary monitor mode is most reliable if you share only your primary display.

Audio Capture

System audio capture requires PipeWire or PulseAudio monitor sources. Ensure your audio is routed through the default output device.

Development

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Format code
black interview_assistant/
ruff check interview_assistant/

License

MIT License - See LICENSE file