Gia — Mindful 20-20-20

A mindful Chrome extension powered by Gemini Nano to help your eyes breathe.

🌿 Inspiration

Turning an optometrist’s advice into a UX project. Building tech that helps people take care of themselves.

Around the same time I was planning my entry for the Google Built-In AI Challenge, my optometrist told me I needed to take more screen breaks to manage eye strain and headaches.
The timing couldn’t have been better — Gemini Nano had everything I needed to build it locally and keep the experience private, responsive, and kind.

So I built Gia, a mindful focus companion that helps you follow the (20\text{-}20\text{-}20) rule during deep work sessions.

Two tones, same goal:

Mindful — calm, grounding reminders
Goofy — slightly unhinged humor and bad jokes

One helps you relax; the other helps you laugh — both remind you to take a break. Sometimes the best UX nudge is a really bad joke.

Rooted in cognitive psychology and ocular health best practices, Gia uses short, intentional interruptions to help you notice how long you’ve been staring. It also encourages longer reset breaks after extended screen time — turning healthy habits into part of your daily grind.

💡 What It Does

Gia helps users build healthier screen habits by encouraging short, intentional breaks and longer reset moments throughout the day.

After about 20–25 minutes of screen time, Gia gently reminds you to look away for 20 seconds following the (20\text{-}20\text{-}20) rule.
After a user-selected interval, Gia suggests a longer “reset” break — a few minutes to breathe, stretch, or reflect — to help maintain steady energy and prevent fatigue.

By turning science-based guidance into friendly nudges, Gia helps people care for their focus and their eyes — without breaking their flow.

🧩 About This Project

Gia was built using Chrome’s Built-in AI to create a fully local, privacy-first experience powered by Gemini Nano.

The Prompt API classifies short voice commands like “start,” “snooze,” and “dismiss,” enabling natural, hands-free control during work or study sessions.
The Translator API adds multilingual accessibility, allowing users to experience both the Mindful and Goofy tones in their preferred language — all processed privately and on-device.

By combining these APIs with thoughtful UX and Material Design principles, Gia turns the science-backed (20\text{-}20\text{-}20) rule into a simple, human-centered habit.

🧠 How It Works

Technical Implementation

Gia is a Manifest V3 Chrome Extension powered by Gemini Nano LLM through Chrome’s Built-in AI.
All AI processing occurs on-device for full privacy and offline functionality.

APIs Used

Prompt API — classify short voice intents (“start,” “snooze,” “dismiss”) locally
Translator API — multilingual reminders and UI strings, processed on-device

Core Chrome APIs

Alarms, Storage, TTS, Offscreen — timers, local data, and audio cues

Frontend: HTML, CSS, JavaScript
Design System: Material Design (Figma)
AI Engine: Gemini Nano (on-device)
Version Control: GitHub (public submission)

🔊 Multimodal Functionality

Voice Input — say “snooze” or “dismiss”; Prompt API handles local intent recognition
Visual + Textual Feedback — soft cards adapt tone (Mindful or Goofy)
Optional Audio Output — short voice lines or tone-matched sound cues, all on-device

🛠️ How We Built It

UI in Figma using Material Design for accessibility, clarity, consistency
Manifest V3 setup with alarms, storage, offscreen for background timing
Prompt API for local voice intent; Translator API for multilingual support
Tone profiles for Mindful vs. Goofy copy and timing
Testing & iteration to keep reminders helpful, not disruptive

🏆 Accomplishments

Local-first, privacy-preserving AI extension
Turned the (20\text{-}20\text{-}20) rule into a practical, adaptive habit
Integrated short and long break cycles that feel natural

🧩 Challenges

Offscreen scripting for voice + TTS without blocking UI
Tuning on-device latency and cadence so reminders don’t feel intrusive
Balancing humor and mindfulness in tone design

📚 What We Learned

Small changes in tone, timing, and context transform how people respond to breaks — and privacy can coexist with personalization when AI runs locally.