I hard-killed my hiking safety app mid-trip. The alert still went out.

Description and use

TrailWatch uses a local Gemma model, Temporal durable workflows, and ntfy push notifications. Even after hard‑killing the worker, timers survive and an alert is sent 10 s after restart.

Problem

Hikers forget to notify others of delays, and emergency contacts may not act promptly if the hiker is silent.

Outcome

Even after hard‑killing the local worker, the alert was sent only 10 s after the worker restarted, proving durable timers survive process death.

Workflow

• 1. Hiker enters trip description. 2. Gemma parses it into a plan. 3. Temporal workflow starts durable timers for check‑in, grace, and re‑alert. 4. Hiker taps "I'm OK" or "SOS" via phone or ntfy lock‑screen. 5. If no check‑in, Temporal nudges, then alerts contact with template and Gemma brief. 6. Contact receives all‑clear when hiker taps "I'm OK" after return.

Evidence

If you miss a check‑in, TrailWatch nudges you, waits a grace period, then alerts your emergency contact.

Source

Kill the worker, restart the web app, reboot the Temporal server mid-hike: the timer still fires.

Source

The alert went out about 10 seconds after the worker came back.

Source