Apps/Focus Guardian

Focus Guardian

Live

Your robot accountability partner

A productivity body-double app that uses Reachy Mini as an accountability partner. The robot watches you work, notices when you get distracted, and provides gentle encouragement through expressions and movements. Based on the body-doubling technique used for ADHD focus.

Try on HuggingFace →View Source

PythonGradioMediaPipeOpenCVReachy SDK

Demo Coming Soon

We're working on recording a demo. In the meantime, try running it yourself!

Features

Attention Tracking

Uses head pose detection to notice when you look away from your screen. Reachy responds with curious or encouraging expressions.

Pomodoro Sessions

Built-in focus timer with customizable work/break intervals. Reachy celebrates completed sessions with antenna wiggles.

Expressive Feedback

A library of robot expressions — encouraging nods, playful tilts, celebratory antenna dances — that respond to your focus state.

Session Analytics

Track your focus patterns over time. See when you're most productive and what breaks your concentration.

How It Works

Start a Focus Session

Launch the app and set your work duration. Reachy settles into 'focus mode' — attentive but calm.

Work While Watched

The camera tracks your head pose. As long as you're focused, Reachy stays supportively still with occasional encouraging movements.

Get Gentle Nudges

Look away too long? Reachy notices and gives you a gentle reminder — a head tilt, an antenna waggle, nothing aggressive.

Celebrate Completion

Finish your session and Reachy celebrates with you. Take your break knowing you earned it.

Getting Started

Prerequisites

•Reachy Mini Lite (physical robot or simulation)
•Python 3.10+
•Webcam for head pose detection
•Reachy daemon running on port 8000

# Clone the repo
git clone https://github.com/BioInfo/reachy.git
cd reachy/apps/focus-guardian

# Install dependencies
pip install -r requirements.txt

# Run with simulation
python app.py --simulation

# Run with physical robot
python app.py

The Build Story

claude code

Claude Contributions

Expression System Design

Designed the mapping between focus states and robot expressions. Created a library of 12 distinct expressions with smooth transitions.

"Help me design expressions that feel supportive, not judgmental"

Head Pose Integration

Integrated MediaPipe face mesh for real-time head pose estimation. Handles edge cases like partial face visibility.

What I Learned

→Body-doubling works even with robots — the sense of 'being watched' helps focus
→Expressions need to be subtle; too much movement becomes distracting
→Head pose detection is surprisingly sensitive to lighting conditions