This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
EcoQuest is an AI-powered outdoor habit-building web app designed to help people turn screen time into real-world exploration.
Instead of simply telling people to "spend less time on their phones," EcoQuest gives them something specific to do outside.
Users choose:
EcoQuest then uses Google Gemma 4 to generate a personalized outdoor quest.
For example, a user might choose:
20 minutes · Park · Curious · Easy
and receive a quest such as The Tree Detective, with simple observation and exploration steps they can complete outdoors.
The experience then continues with:
Choose → Generate → Explore → Complete → Earn XP → Build a Streak → Journal
The app tracks outdoor minutes, completed quests, XP, nature streaks, and personal field notes locally in the browser.
EcoQuest is built around a simple idea:
The goal isn't to spend more time on another app. It's to spend less time on screens and more time in the real world.
Live Demo: https://ecoquest-txcy.onrender.com/
The live application demonstrates the complete experience:
GitHub: https://github.com/pnkj006/EcoQuest
The project is built as a small, focused MVP with the AI generation layer separated from the frontend.
EcoQuest is built with:
The core architecture is:
React Frontend
↓
/api/generate-quest
↓
Node.js Server
↓
Gemma 4
↓
Structured Quest JSON
↓
Personalized Outdoor Quest
The user preferences are sent to a server-side API endpoint. The server asks Gemma 4 to generate a structured quest containing the title, duration, environment, difficulty, description, steps, and optional bonus challenge.
I used Gemma 4 (gemma-4-26b-a4b-it) because the quest-generation task is short, structured, and highly adaptable to user preferences.
The model is instructed to create realistic and safe outdoor activities without requiring special equipment. It also avoids dangerous activities, unsafe road crossings, approaching wildlife, disturbing plants or animals, and entering restricted/private areas.
The rest of the application is intentionally lightweight. User progress is stored locally using localStorage, so the MVP does not require authentication or a database.
For EcoQuest, the AI model isn't just an additional feature. It is the component responsible for turning a user's mood, available time, and surroundings into an actionable outdoor experience.
Using an open-weight model such as Gemma gives the project a path beyond a single closed AI provider.
Because Gemma is an open-weight model, the quest-generation layer can potentially be adapted, fine-tuned, evaluated, or eventually self-hosted as the project grows.
That matters for a project like EcoQuest because the goal is not simply to generate text. The long-term goal is to create an AI system that understands different environments and creates useful, safe, context-aware experiences while keeping the underlying technology more open and adaptable.
For this MVP, Gemma made it possible to experiment with that idea without building a large recommendation system or manually creating hundreds of quests.
I built EcoQuest with an AI-assisted development workflow and used Antigravity throughout the project for implementation, testing, debugging, and iteration.
The development process included:
EcoQuest is entering the Best Use of Gemma category.
Gemma 4 is used as the core AI engine for personalized outdoor quest generation. The model directly transforms the user's selected time, environment, mood, and difficulty into structured quests that drive the main product experience.
EcoQuest is currently an MVP, but there are several directions I would like to explore:
For now, the most important thing is simple:
Close the laptop. Go outside. Complete a quest.