TrailRelay: Offline-First Trail Hazard Reporting with Gemma and Temporal

TrailRelay is an offline-first trail hazard reporting system that allows users to report issues such as fallen trees, flooding, and damaged trails through a web interface, even when they have no internet connection. The system stores reports locally in the browser using IndexedDB, ensuring availability when connectivity is unavailable. When the backend is reachable, pending reports are synchronized for processing.

The core of TrailRelay lies in its combination of local AI, durable workflow orchestration, and GitHub integration. The locally running Gemma 3 1B model classifies the hazard descriptions without requiring a hosted AI API, while Temporal orchestrates the report-processing workflow, manages activity execution, and retries failed operations. GitHub Issues provides a way to track processed reports.

The project demonstrates how local AI and durable workflows can work together to create a more resilient system for outdoor safety. By running Gemma locally, the application gains more control over AI inference and enables local experimentation without depending on a paid inference service. Temporal, on the other hand, ensures that workflow activities are resilient to failures through orchestration and retries.

Browser-based storage preserves reports during connectivity issues, ensuring they can be processed once internet access is restored.

The use of open technologies in TrailRelay allows for inspection, local running, and adaptation, making it easier to understand and modify the system. This open innovation approach enables users to build and adapt the application using tools that meet their specific needs. By combining Google's open-weight Gemma model with Temporal workflow orchestration and GitHub tracking, TrailRelay addresses the challenge of reporting trail hazards in remote areas, contributing to safer outdoor exploration without the need for constant internet connectivity.