Promethist Watchdog

Email alerts

Configure the relay in Settings, then use Send test email before trusting it. Three details are not obvious from the dialog:

It is a plain SMTP relay — no Microsoft 365 OAuth, no mailbox to set up. With authentication turned off, it works against an IP-allowlisted relay without credentials.
When authentication is on, the password goes into the Windows Credential Manager, never the configuration file. Therefore, it is omitted when you export your configuration and must be set again on the next machine.
Alerts are throttled per app and per event, so an app crash-looping at 3am cannot bury another app’s alert or your inbox.

The events worth an email are chosen in Settings → Email alerts → Advanced → Email on these events. Four of them are about the whole machine rather than one app, and all four are included by default:

reboot: The optional daily PC reboot is about to happen.
emergency_shutdown: The building lost power (Shutting down when the power fails). Sent before anything is stopped, so it leaves while the UPS still has battery.
network_down: The internet has been unreachable for several checks in a row.
network_restored: It came back, with how long it was gone.

A network_down alert has to travel over the connection that has just failed, so usually it does not arrive. That is why network_restored exists and carries the outage’s length: it is the one that reaches you, and it tells you what you needed to know. Nothing is lost either way — every check is written to a file regardless. See Proving the network was the problem.

Two more are about a web address Watchdog checks (Watching a server by its web address), and are also included by default:

url_down: The address has failed several checks in a row (two by default).
url_restored: It is answering again, with how long it was gone.