🌿 TouchGrass AI: Turning Open-Source AI into a Real-World Adventure — Less Screen, More Green

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

🌿 TouchGrass AI — Less Screen, More Green

What I Built

TouchGrass AI is an AI-powered outdoor activity companion that encourages people to take a break from their screens, step outside, and reconnect with nature.

In a world where we spend hours scrolling, studying, coding, and working on our devices, it is easy to forget the simple joy of going outside. I wanted to build something that uses AI not to keep people online for longer, but to help them spend more time in the real world.

TouchGrass AI turns the simple question “What should I do outside?” into personalized outdoor activity suggestions.

🌱 Key Features

• AI-Powered Activity Suggestions: Get outdoor activity ideas based on your mood, available time, and surroundings.
• Mood-Based Recommendations: Find suitable activities when you're feeling stressed, low on energy, curious, or social.
• Flexible Time Options: Choose activities that fit the amount of time you have available.
• Daily Nature Challenges: Try small challenges that encourage you to explore your surroundings and develop healthier habits.
• Nature Journal: Record your outdoor experiences and reflect on the moments you spend away from your screen.
• Progress Tracking: Keep track of your outdoor activities and build a consistent nature-friendly routine.
• Fallback Activity Planner: If the AI model cannot load, the application can use predefined JavaScript recommendations to keep the experience useful.

🎯 Who Is It For?

TouchGrass AI is designed for students, developers, remote workers, and anyone who spends a lot of time indoors or on digital devices.

You don't need to plan a hike or travel somewhere special. Sometimes, a short walk, sitting in a park, observing birds, or simply getting some fresh air is enough to get started.

The goal is simple: use technology to help people reconnect with the world beyond their screens.

Demo

🌐 Try the live website:
https://kamya5907-sketch.github.io/-TouchGrass-AI-/

Explore the interface, select your mood, choose your available time and surroundings, and discover ideas for spending more time outdoors.

Code

💻 GitHub Repository:
https://github.com/kamya5907-sketch/-TouchGrass-AI-

The project is built with three separate core files:

• index.html — Defines the structure and content of the website.
• style.css — Provides the visual design, nature-inspired colors, responsive layout, and styling.
• script.js — Handles interactions, activity recommendations, AI integration, journal entries, and progress tracking.

The project is designed to be accessible through a web browser without requiring a complex installation process.

How I Built It

I built TouchGrass AI using HTML, CSS, JavaScript, and Transformers.js, with an open-weight language model as the intended core of the AI experience.

🛠️ Technology Stack

• HTML5: Structures the website, including activity-planning controls, recommendation areas, challenges, and journaling features.
• CSS3: Creates a calming, nature-inspired visual experience with green tones, clean layouts, and responsive styling.
• JavaScript: Connects the interface to the recommendation logic and manages user interactions, challenges, and saved progress.
• Transformers.js: Provides the browser-side integration for running supported machine-learning models.
• FLAN-T5-small: The open-weight model configured for the AI recommendation experience.

🤖 How the AI Fits In

The idea behind TouchGrass AI is to make recommendations more relevant to a person's current situation instead of displaying the same generic list of activities to everyone.

The application collects simple preferences, such as mood, available time, and surroundings, and uses them to guide outdoor activity recommendations.

The configured model is Xenova/flan-t5-small, accessed through Transformers.js. The application also includes fallback recommendations so that users can still receive activity ideas if the model fails to load.

The model's availability and successful inference depend on browser compatibility, network access for the initial model download, and the runtime environment. The fallback planner is separate from the AI-generated recommendation path.

🌍 Why This Approach?

I wanted to explore how open-weight AI can be integrated into a practical, accessible web application using technologies that are familiar to beginner developers.

Instead of building a complicated backend, I focused on a lightweight frontend experience that people can open directly in their browsers.

This project also helped me explore how AI, user-centered design, and everyday behavior change can work together to solve a small but meaningful problem.

Why Does Open Innovation Matter?

Open innovation matters because AI should be more than a collection of closed services that developers can only access through paid APIs.

Using an open-weight model such as FLAN-T5-small gives developers an opportunity to explore model integration, understand the inference process, and experiment with AI without making a proprietary hosted API the only possible foundation.

For a project like TouchGrass AI, this creates several possibilities:

• Accessibility for learners: Students and independent developers can experiment with AI using familiar web technologies.
• More control: Developers can investigate model behavior, adapt prompts, change models, and improve the application without depending entirely on one closed provider.
• Experimentation: Open models make it easier to prototype new ideas and explore different ways of generating useful recommendations.
• Potential for local inference: With compatible models and runtimes, browser-based inference can reduce the need to send every interaction to a remote inference service.
• Community collaboration: Other developers can inspect the code, report issues, suggest improvements, and build on the project.

For this project, open innovation also reflects the message behind the product itself. Technology should empower people rather than capture all their attention.

I wanted to experiment with AI in a way that encourages users to close their laptops, walk outside, and experience something beyond a screen.

The most meaningful use of AI is not always keeping people engaged with technology. Sometimes, it is helping them know when to step away from it. 🌿

My Agent Session

I do not have a DevRelay agent-session link to share for this submission. This section can be updated if a session is recorded and made available.

Prize Categories

My submission is focused on the Open-Source AI theme, using an open-weight model and Transformers.js as part of the project's AI implementation.

If the challenge provides specific partner prize categories, I will select only those for which the project meets the published eligibility requirements.

---

Built with curiosity, JavaScript, and a little love for the outdoors. 🌱

Project: TouchGrass AI
Developer: Kamya Varshney
Live Demo: https://kamya5907-sketch.github.io/-TouchGrass-AI-/
Source Code: https://github.com/kamya5907-sketch/-TouchGrass-AI-

Thanks for participating! 🌍