Alerts
Set up agent alerts
Merchant Portal → Notifications → Agent alerts.
"Get an email the moment the agent stops or needs a hand — separate from the scheduled report, and to whoever is actually at the pharmacy."
1. Add recipients first
Under Alert recipients, add the addresses that should be told when the agent needs help — the people physically able to open QS1 or press Start. Until you do, the field reads "No alert recipients yet — add whoever should be told when the agent needs help."
Note: This is a separate list from Report recipients. Enabling alerts with an empty list is refused: "Add at least one alert recipient before enabling alerts."
2. Turn alerts on
Flip the toggle at the top of the card from Disabled to Enabled.
3. Choose which events email you
Under Alert me when the agent…:
| Checkbox | Default | Hint shown in the portal |
|---|---|---|
| Stopped with an error | on | "The agent hit a fault it couldn't recover from and stopped." |
| Stuck, needs attention | on | "Still running but waiting on someone — QS1 isn't open, or the cloud is unreachable. Nothing processes until a person acts." |
| Went offline | on | "No contact while its schedule says it should be running — a crash, or the machine is off. The only signal the agent can't send itself." |
| Finished a run | off | "Informational: the run drained the queue or someone pressed Stop. Fires on every normal run." |
Important: Leave Finished a run off unless you specifically want a heartbeat email. It fires on every normal run, and a mailbox full of everything-is-fine mail is how the genuine alerts stop being read.
4. Tune the two dials
Don't repeat within — the cooldown. Options are 15, 30, 60, 120, 240, 480 minutes and 24 hours; the default is 60 minutes.
"One email per device per problem type in this window."
Per device and per problem type, so an ongoing stuck agent never masks a separate error on the same machine.
Offline after — how much silence counts as offline. Options are 5, 10, 20, 30, 60, 120, 240 minutes; the default is 20 minutes. Only counted inside the run schedule, so an agent that is off-shift is not offline. The field is disabled unless Went offline is ticked.
Note: Offline detection compares silence against the Schedules page. A location with no schedule rows is treated as always expected, so alerting still works. It is most reliable at catching a machine that is off or crashed between runs; do not rely on it as a guaranteed dead-man's switch during a single long continuous run.
5. Prove it works
Click Send test alert:
"Sends a clearly-labelled sample alert. Does not affect real alerts."
The test is deliberately excluded from the alert history and does not consume the cooldown, so a setup check can never swallow a real alert minutes later. Confirm it arrives — including that it is not filtered as spam — before you rely on alerts.
What the emails look like
Subject: [PharmAutomate] <what happened> — <location>
The location comes first because a pharmacy group needs to know which location before anything else. The headline is one of:
- Agent stopped with an error
- Agent needs attention
- Agent offline
- Agent finished its run
The body carries the location, the device, the agent's own description of the problem, the time in your organisation's timezone, and the one action that clears it.
Note: Link click-tracking is deliberately switched off for these emails, so nothing rewrites the addresses in them. That is intentional and should not be re-enabled.