Skip to main content

What It Tells You

An alert is one sentence about one field: the window that has opened or closed, the risk that has crossed a threshold, or the sensor that has gone quiet, and what to do about it. It names the reading or the modelled value behind it, so you can check the number before you act. A frost alert says which night and how cold. A nitrogen alert says the window is open and why. A leaching alert says how much rain is coming.

How It Works

Every model writes an advisory row when its state changes and sends a push notification and an email. Alert text is rendered from templates. A language model never writes it. A state that persists across runs raises a single alert. The alert links to the field, where the reading and the model behind it are shown. A custom trigger is announced in words: the field, the reading and its unit, and the threshold it crossed. It fires once on the crossing, not on every reading in breach, and re-arms once the reading has come back past the threshold by a margin. A run of identical alerts folds into one card in the notification centre.
Cartoon of a phone with a frost alert bubble that shows a temperature chip and points at a probe in the field.

A frost alert names the night and the temperature, and points at the probe behind it.