Tenants

The device ID is how this device identifies itself for the rest of its life and cannot be changed afterwards, so use the name the site uses. The credential is shown once, on the next screen, and is not recoverable — if it is lost the device gets a new one, which means visiting it.

Held for stops a spike being an incident: the condition must hold this long before an alarm is raised. Clears at is a separate threshold to come back below, so a value sitting on the line does not raise and clear all night. Leave them empty and the rule fires on the first reading and clears on the first reading back.

Priority is lowest-first, and the list below is the order. A critical alarm tries every provider in turn until one accepts; anything less serious stops after two, because unbounded retry on a warning costs money without making anyone safer. A provider that refuses is stepped over for a minute, and a provider that accepts is believed — sending again because a gateway was slow means two messages, two charges, and one alarm.

The credential is written to the server’s secret store, never to the database. It is not in any backup, not in any response, and this console cannot read it back — when you edit a provider later you will see that a credential is set and a box to replace it, never the value itself. The record keeps only the name it is stored under.

{{to}}, {{text}}, {{from}}, {{validity}}, {{callback}} and {{accountId}} are substituted, each escaped for the format you chose — so an alarm message containing an ampersand cannot turn into two fields. {{accountId}} also works in the path.

Validity tells the carrier to stop trying after this long. An alarm delivered an hour late is worse than one that never arrives: it reports a situation that has moved on, and somebody acts on it.

The callback is stored and not yet used. Nothing in the platform reads delivery receipts today, so a message this gateway accepts is reported as accepted and not as delivered. Setting this now means the address is already right when that lands.

SMS is charged per message. Email is not. The two limits bound what a single situation can cost: a rate limit across the whole group, and a cap on how many messages one incident may produce.

Neither limit can silence a critical alarm. Over the limit, a critical notification is still sent, the overspend is recorded against the tenant, and it is reported. A cap reached at 02:00 must not be why nobody heard about a bearing seizing at 02:40. Lower severities are capped for real, because a warning at 03:00 is a cost decision rather than a safety one.