CampusWalk

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

CampusWalk 🌿 — Touch Grass, One Quest at a Time!

CampusWalk turns ordinary outdoor time into meaningful mini-adventures.

It is an AI-powered outdoor quest generator that helps students explore their campus, connect with nature, and take refreshing breaks from screens.

Features

• 🤖 AI-generated quests based on available time, activity, and difficulty
• 🌱 Nature, walking, exploration, and photography challenges
• 🏆 XP, achievements, and progress tracking
• 📔 An adventure journal to save completed quests and reflections

Built With

React, Vite, JavaScript, Transformers.js, and an open-weight language model running in the browser.

Our goal: Make stepping outside feel like an adventure, one quest at a time!

Demo

https://campus-walk-lk2mnd9rs-riya-rathis-projects.vercel.app/

Code

https://github.com/riyarathi2211/CampusWalk/tree/main/src

How I Built It

CampusWalk is an AI-powered outdoor quest generator designed to help students step away from their screens and explore the world around them.

I built the frontend using React, Vite, and JavaScript, with a clean, nature-inspired interface. Users can choose an activity—Nature, Walking, Explore, or Photography—along with their available time and preferred difficulty.

The AI component uses Transformers.js with the open-weight onnx-community/LFM2.5-350M-ONNX model. It runs inference directly in the browser using WebGPU, where supported, to generate personalized outdoor missions without relying on a hosted LLM API for each generation.

I also built an achievement system with XP and unlockable badges, along with an Adventure Journal that stores completed quests and reflections locally on the user's device.

Why Does Open Innovation Matter?

Open innovation made it possible to build an AI-powered experience without depending entirely on a proprietary, closed-model API.

Using an open-weight model with browser-based inference gives CampusWalk the potential to reduce dependence on external inference services, improve user privacy, and make experimentation more accessible. It also lets developers inspect the model's capabilities, explore different deployment options, and adapt the experience without being locked into a single hosted AI provider.

For a project focused on encouraging people to engage with the real world, keeping the AI experience lightweight and accessible is an important part of the vision.

My Agent Session

CampusWalk uses browser-based model inference rather than a multi-step autonomous coding or tool-using agent. I don't have a DevRelay session link to share, so I'm leaving this section out.