LeetWatcher

A macOS menubar app that helps you stay focused during LeetCode sessions by tracking when you look away from the screen or switch to other apps.

Why

Solving LeetCode problems effectively means staying focused without reaching for AI helpers or other distractions. LeetWatcher uses an honor-system approach: it doesn't block anything, but tracks your focus so you can reflect on your session afterward.

How It Works

• Window monitoring: Detects when you switch away from a browser with LeetCode open
• Gaze detection: Uses your webcam + MediaPipe to detect when you're looking away from the screen
• Session summary: After each session, shows your focus percentage and prompts reflection on your approach

Violations are recorded after 5 seconds of looking away or switching windows, giving you time for natural glances.

Requirements

• macOS (uses Quartz for window detection)
• Python 3.13+
• Webcam access (for gaze detection)

Permissions

LeetWatcher requires these macOS permissions to function properly:

Permission	Why	How to Grant
Camera	Gaze detection via webcam	System Settings → Privacy & Security → Camera → Enable for Terminal/Python
Screen Recording	Reading browser window titles to detect LeetCode	System Settings → Privacy & Security → Screen Recording → Enable for Terminal/Python
Automation	Getting browser URL via AppleScript	Prompted automatically when first accessing browser URL

Granting Permissions

1. Camera: You'll be prompted on first run. Click "OK" to allow.

2. Screen Recording:

   • Go to System Settings → Privacy & Security → Screen Recording
   • Enable the app running LeetWatcher (usually "Terminal" or "Python")
   • Restart the app after granting

3. Automation (for URL capture):

   • When you start a session with a browser open, macOS will ask to allow control
   • Click "OK" to let LeetWatcher read the current URL

Note: Without Screen Recording permission, LeetWatcher can still detect app switches but cannot see window titles. It will assume any browser is on LeetCode to avoid false violations.

Installation

# Clone and install
git clone <repo-url>
cd LeetWatcher

# Download the MediaPipe model
curl -L -o leetwatcher/models/face_landmarker.task \
  "https://storage.googleapis.com/mediapipe-models/face_landmarker/face_landmarker/float16/1/face_landmarker.task"

# Install dependencies
uv sync

Usage

# Run the menubar app
uv run leetwatcher

# Preview gaze detection (calibration)
uv run python -m leetwatcher.preview

In the menubar:

• Start Session: Begin tracking focus
• End Session: Stop and see your summary
• Gaze Preview: Test what movements trigger violations
• View History: See past sessions

Gaze Preview

The preview window shows:

• GREEN: Focused (looking at screen)
• RED: Looking away (would trigger violation after 5s)
• ORANGE: No face detected

Use this to calibrate your setup and understand what counts as "looking away."