This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
TrailMate AI is an outdoor micro-adventure generator powered by locally running Gemma 3 4B.
It transforms simple preferences, such as available time, activity type, difficulty, and environment, into outdoor missions with checkpoints and a grounding exercise.
The goal is simple: use AI to give people a reason to put their phones away and engage with the world around them.
I tested TrailMate AI in a real park and completed all three generated checkpoints: Observe the Canopy, Listen to the Wind, and Feel the Ground. The app recorded 3/3 checkpoints completed and awarded 125 points.
Demo
Project demonstration:
The demo shows TrailMate AI generating an outdoor mission using locally running Gemma and presenting its checkpoints.
Code
GitHub repository: https://github.com/Anasdarzi009/trailmate-ai
TrailMate AI is open source under the MIT License.
How I Built It
TrailMate AI uses Gemma 3 4B, an open-weight language model, through Ollama for local inference.
The application is built with:
- AI model: Gemma 3 4B
- Local inference: Ollama
- Backend: Python, FastAPI, and Pydantic
- Frontend: React, Vite, and Tailwind CSS
- Mission history and points: Browser localStorage
The React frontend sends mission requests to the FastAPI backend, which communicates with the local Ollama runtime. The generated mission is then returned to the interface.
Once the model is installed, mission generation does not require a hosted AI API. Initial setup and model downloads require internet access.
Why Does Open Innovation Matter?
Open-weight AI made it possible to build TrailMate AI around local inference instead of depending on a hosted AI API for every mission.
This gives developers greater control over where inference happens and provides a foundation for privacy-conscious, locally powered applications.
It also makes the project easier for other developers to explore, modify, and extend using open-source tools.
For this project, open innovation supports the central idea: AI can help people engage with the real world without making constant online interaction a requirement.
My Agent Session
This section is optional. Add a DevRelay session link if you have one to share; otherwise, remove this section.
Prize Categories
Best Use of Gemma — Gemma 3 4B is the core model used to generate outdoor missions, running locally through Ollama.
What's Next?
Potential improvements include richer mission personalisation, additional outdoor activities, and more ways to record and track adventures.
These are future ideas, not features claimed as already implemented.
Built for Hacktoberfest 2026 Week 1: Touch Grass 🌿
Top comments (0)