This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
I built Fieldquest, a small web app that turns the time, energy, and kind of activity you’re in the mood for into a tiny outdoor quest.
Choose whether you want to notice things, move a little, or make something; set aside 5, 15, 20, or 45 minutes; and Fieldquest gives you a few simple steps to try. The suggestions are designed to start near wherever you are, need no special gear, and leave the screen behind once you head out.
Fieldquest is for anyone who wants a low-pressure nudge outside—whether that means a walk around the block, a moment on a balcony, or a closer look at a nearby patch of green. No streaks, scores, or account required. Small counts.
Demo
file:///C:/Users/Pranjal%20kumar/Pictures/Screenshots/index.html
The demo mode works without an AI model. To try locally generated quests, run Ollama with the gemma3:1b model on your computer.
Code
Fieldquest
A little outside, your way. Fieldquest turns your available time, energy, and mood into a short outdoor micro-adventure. The goal is to make the screen the shortest part of the experience.
Run it
Open index.html in a browser for the built-in demo mode. No account, API key, install, or network connection is required for the demo.
To generate quests with an open-weight model, run Ollama, download Gemma 3 1B, and serve this folder on localhost:
ollama pull gemma3:1b
ollama serve
python -m http.server 8000
Open http://localhost:8000. If the browser cannot reach Ollama, allow local browser access by setting OLLAMA_ORIGINS=* for the Ollama process, then restart it. The app calls Ollama's local chat endpoint directly; there is no app server and no cloud model API.
Why open-source AI?
The model can run on the user's own machine. Their starting point, energy, and preferences stay in their browser and…
Built with plain HTML, CSS, and JavaScript. No framework or third-party JavaScript dependencies.
How I Built It
Fieldquest is a static, responsive website. Its controls let you choose a starting place, time, energy level, and activity style.
For AI-generated quests, the site sends those choices to a local Ollama instance running Gemma 3 1B. The model returns a short title, description, and set of steps. If Ollama or the model isn’t available, Fieldquest uses built-in quests instead, so the core experience still works offline.
I kept the suggestions low-pressure and practical: no required equipment, no trespassing, no picking plants, and no assumptions about local weather or wildlife.
Why Does Open Innovation Matter?
The local model makes the experience more private and adaptable. A person can run quest generation on their own machine, keep their choices on-device, and choose which model to use. There’s no hosted AI account or closed API required for the local setup.
Open models also make the project easier to experiment with: the prompt and model can be changed, and the built-in demo mode remains available when local inference isn’t set up. That makes the AI an option people can control, while keeping the app useful either way.
My Agent Session
[Add a DevRelay session link if you create one. This section is optional.]
Prize Categories
- Best Use of Gemma
Top comments (0)