This is a project for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass.
What I Built
TouchGrass AI is an outdoor adventure companion that turns a little free time into a practical reason to step outside.
Choose your available time, interests, energy level, and preferred environment to get outdoor mission ideas. Each mission includes preparation notes, step-by-step instructions, things to observe, and safety considerations.
When it’s time to go, Adventure Mode shows one instruction at a time and saves progress locally. Afterward, you can record a reflection in your journal and see completed adventures in your impact dashboard.
The app also includes a curated library of 24 missions. These templates keep discovery useful when Gemini is unavailable, but they are not a local AI model.
Demo
Code
View the TouchGrass AI repository on GitHub
How I Built It
I built the app with Next.js App Router, TypeScript, Tailwind CSS, Zod, and the official Google Gen AI SDK.
Adventure recommendations, nature explanations, and garden plans use the hosted Gemini API through server-side routes. The API key stays in a server environment variable. Zod validates requests and generated adventure data.
Saved missions, active mission progress, and journal entries are stored in the browser. The offline adventure library uses deterministic filtering; it does not run AI inference locally.
Why Does Open Innovation Matter?
This version uses hosted Gemini and does not use an open-weight model or an open-source AI agent framework. The offline templates are curated content, not AI inference, so I can’t claim that open-source AI powers the project.
The local-first parts do show one useful privacy choice: saved missions, progress, and journal notes stay in the browser instead of being sent to an AI service. Gemini requests need an internet connection.
My Agent Session
No public agent-session link is included.
Prize Categories
None.
Top comments (0)