TouchGrass AI

#devchallenge #hf26challenge

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

What I Built

TouchGrass AI 🌿 — A Local-First Outdoor Activity Companion

TouchGrass AI is a local-first, open-weight AI companion that turns screen time into small, practical outdoor missions. Users select their mood, available time, energy level, and preferred activities to receive personalized suggestions for mindful walks, nature exploration, bird listening, gardening, and gentle movement.

The goal is simple: make the screen the shortest part of the experience. Get a plan, put the phone away, and reconnect with the world outside.

The application includes a built-in fallback planner, so users can still receive activity suggestions without downloading an AI model. It is designed without user accounts, analytics, or third-party cloud AI dependencies.

Demo

Local demo: Run the application on your machine using the setup instructions in the repository's README.

To run it, install the dependencies with npm install, start the server using npm start, and visit http://localhost:3000.

Code

GitHub Repository — TouchGrass AI

How I Built It

I built TouchGrass AI using Node.js and Express for the backend, HTML, CSS, and JavaScript for the frontend, and Ollama to support local inference with the open-weight llama3.2:3b model.

The application collects a user's activity preferences and constructs a prompt that asks the local model to generate a realistic outdoor mission, including its duration, difficulty, step-by-step instructions, and an alternative for when going outside is not practical.

The backend communicates with Ollama through its local API. If the model is unavailable, the application falls back to its built-in activity planner, keeping the core experience accessible without a cloud AI service.

I also designed the interface to be responsive and straightforward, with a focus on reducing unnecessary screen interaction rather than encouraging users to spend more time inside an application.

Why Does Open Innovation Matter?

Open innovation is central to TouchGrass AI because the information people share about their mood, routines, and personal preferences should not automatically need to leave their devices.

Using an open-weight model with local inference makes it possible to generate personalized activity suggestions without depending on a paid, proprietary AI API. It reduces dependence on external services, gives users more control over their data, and allows developers to inspect and modify the prompts, experiment with different models, and customize the application's behavior.

The built-in fallback also ensures that the project remains useful when a model has not been installed or is unavailable.

Most importantly, open-source AI makes this experience more accessible to people who want to experiment, learn, and build without recurring inference costs. Instead of locking the core functionality behind a closed service, TouchGrass AI keeps the implementation open and adaptable.

For this project, the value of AI is not keeping people engaged with technology for longer. It is using technology to help people step away from it.

My Agent Session

This project does not currently include a published DevRelay agent session. I may add one later to document the development process and demonstrate how the local AI integration works.