This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass.
What if AI could help us spend less time staring at screens and more time exploring the world outside?
That's the idea behind TrailMate AI — an AI-powered outdoor adventure companion that turns outdoor activities into quests, rewards progress, and encourages people to reconnect with nature. 🌎
🌱 What I Built
TrailMate AI is designed for students, explorers, nature lovers, and anyone who wants to build healthier habits by spending more time outdoors.
Instead of making technology another reason to stay online, I wanted to explore how AI could encourage real-world experiences.
✨ Key Features
- 🥾 Outdoor Quests: Turn outdoor activities into engaging challenges.
- ⭐ XP and Progress Tracking: Earn experience points as you complete quests.
- 🏆 Badges and Achievements: Celebrate milestones and stay motivated.
- 📓 Adventure Journal: Record experiences and reflections from your adventures.
- 🤖 AI Trail Guide: Get AI-generated guidance to support your outdoor journey.
- 💾 Progress Persistence: Keep track of quest progress and journal entries between visits.
The goal is simple: make going outside feel rewarding, engaging, and fun!
🚀 Demo
Live Demo: https://trail-mate-ai-three.vercel.app
The project is currently available as source code on GitHub.
💻 Code
GitHub Repository: TrailMate AI
Feel free to explore the code, learn from the implementation, and share suggestions or improvements.
🛠️ How I Built It
I built TrailMate AI using modern web technologies and an open-weight AI model.
Tech stack:
- Next.js — Application structure and API routes.
- JavaScript and React — Interactive components and application state.
- Tailwind CSS — Responsive styling and UI design.
- Ollama + Phi-3 — Local language-model inference for the AI trail guide.
- Browser localStorage — Persistence for quest progress and journal entries.
Ollama allows the Phi-3 model to run locally, making it possible to experiment with AI-powered features without relying entirely on a proprietary hosted AI API.
The application combines an interactive web experience with AI guidance and gamification. Quests, XP, achievements, and journaling are intended to make outdoor exploration feel like a continuing adventure rather than a one-time activity.
🧠 Why Does Open Innovation Matter?
Open innovation makes AI experimentation more accessible to developers.
Using an open-weight model through Ollama gives developers an opportunity to experiment with local inference, understand how language models integrate into applications, and build experiences with greater control over their development environment.
For TrailMate AI, this approach helped me explore an important question:
Can we use AI not just to keep people engaged with technology, but to encourage them to engage with the real world?
Open-source tools and publicly shared code also create opportunities for other developers to learn, contribute, and build on existing ideas.
I believe technology should help us live better — not simply spend more time online.
🏆 Prize Categories
This project is submitted to the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass.
Partner prize categories: None claimed at this time.
🌍 What's Next?
I'd love to keep improving TrailMate AI by making outdoor quests more engaging, refining the AI guide, and exploring more ways to encourage people to step away from their screens.
This is an opportunity to experiment with AI in a different way: using technology as a bridge to the outdoors.
💬 Your Turn!
If you could create one outdoor quest for yourself, what would it be?
A morning walk? A hike? Exploring a nearby park? Or simply watching the sunset without checking your phone?
Share your idea in the comments. Let's inspire each other to touch grass! 🌿
Build with AI. Explore the real world. Scroll less.
Top comments (4)
tr.ee/dev-to
how does the edge function handle cold‑start latency for embedding inference, and is there a fallback if the model service hits rate limits?
Thanks for asking, Alex! 😊 Currently, I'm using Ollama with Phi-3 for AI responses. Embedding inference and rate-limit fallback aren't implemented yet, but they're good ideas for future improvements!
phi-3 on ollama is fine for prototyping, but without embedding inference you'll hit latency at scale, consider a lightweight vector store or on‑device index to avoid fallback thrashing