# BreakTrail: Less Scrolling, More Outside 🌿

This story, derived from the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass submission, introduces BreakTrail, an innovative tool designed to encourage outdoor breaks for busy individuals. The creator, a developer, aimed to combat the habit of excessive screen time by providing a simple solution to plan quick outdoor activities.

BreakTrail allows users to schedule brief outdoor breaks based on various factors like available time, starting location, desired mood (such as Quiet, Nature, Movement, or Discovery), and budget. What sets BreakTrail apart from other recommendations tools is its practical approach. Instead of presenting a list of options, it offers a clear plan that includes the location, activity, and rationale behind the suggestion.

Key features of BreakTrail include:

- Mood selection for a personalized experience

- AI-assisted activity selection from a curated list of locations

- Ability to save plans and export them as JSON files or visual cards

- An outdoor timer with pause and resume options

- Offline functionality for recording feelings and observations post-break

- Cached access to saved plans and timer data without an internet connection

The current prototype showcases a Jaipur location catalogue, although it does not offer live navigation, real-time weather updates, or verified opening hours. The demo website can be accessed at https://breaktrail-nu.vercel.app/.

To build BreakTrail, the developer utilized a variety of technologies and tools, including React, TypeScript, Vite, Tailwind CSS for the frontend, Node.js, Express, SQLite, and Zod for the backend. For local AI inference, they integrated Qwen3 via Ollama, using the qwen3:0.6b model. The planning process involved filtering catalog entries based on user criteria, with the AI model selecting a place and suggesting an activity and explanation. The backend ensured that these selections remained grounded in the available catalogue.

To enhance accessibility, the project includes an OpenRouter adapter for free hosted inference, enabling the backend to request AI output without requiring visitors to install Ollama. This setup allows the deployed backend to request AI output through different free models, ensuring flexibility in the hosted version. The Progressive Web App (PWA) functionality caches the interface for offline access, while browser storage manages saved plans and reflections. A queue system ensures that the offline data syncs with the backend once connectivity is restored.

For public deployment, private browser tokens were implemented to isolate each visitor's backend records, keeping API keys securely on the server. The developer also employed an AI coding assistant to assist in inspecting the project, preparing deployment changes, enhancing visitor data isolation, and integrating the hosted AI adapter.

This collaboration helped troubleshoot deployment issues, such as backend tracking as a nested Git repository, adjusting deployment URLs, configuring CORS settings, addressing device authentication, and validating AI responses.

The significance of open innovation in this context is evident. By using open-weight models, users can potentially run a locally-hosted version of BreakTrail, eliminating reliance on a hosted inference service. This approach allows developers to inspect the integration, modify prompts, experiment with different model sizes, and run inference on their own hardware.

Such openness not only benefits the project but also encourages others to adapt the catalogue to their own city, improve accessibility features, or develop better outdoor activities. For the creator, open innovation represents a means to democratize the tool, enabling a broader community to understand, modify, and enhance the application's capabilities.