Alert rules and silences

Create alert rules that watch hospital measurements such as approval waits, overdue bills, walk-in waits and low stock, and use silences to pause alerts for a planned window.

An alert rule watches one measurement and raises an alert when it crosses a threshold for long enough. A silence pauses alerts for a time window, for example during planned maintenance. This page is for administrators. Staff who act on alerts should read Alerts feed.

Open Alert Rules

Go to SettingsPlatform ConfigurationAlert Rules. The screen has two tabs, Rules and Silences. Changing alert rules and silences needs the Configure permission on Settings. Without it you see "Alerting configuration is admin only", and you can still use the Alerts page.

What you can watch

Pick one of these in the Metric list. The description under the field explains each one.

Metric	What it measures	Values	Can be grouped by
`hos.appointment.pending_approval_age_max`	Age of the oldest appointment still pending approval, per centre	Time in seconds	`centre_uuid`
`hos.bill.overdue_count`	Issued or partially paid bills past their due date, per centre	Count	`centre_uuid`
`hos.walk_in.queue_wait_max`	Longest current wait of a waiting walk-in token, per centre and queue type	Time in seconds	`centre_uuid`, `queue_type`
`hos.inventory.low_stock_margin`	Quantity on hand minus the store reorder point, per store and item. Negative means below the reorder point.	Quantity	`centre_uuid`, `store_uuid`, `item_uuid`
`hos.whatsapp.delivery_failure_rate`	Share of WhatsApp notifications that failed recently, across the organisation	Ratio from 0 to 1 (0.2 is shown as 20%)	Nothing

Create a rule

Start a new rule

Administrator

Settings > Platform Configuration > Alert Rules

On the Rules tab, click New rule (or Create the first rule if there are none).

Expected result: The New rule dialog opens.

Name it

Enter a Name staff will recognise, for example "Pending approvals older than 30 min". Enter a Rule key: lowercase words separated by single hyphens, up to 64 characters, for example pending-approvals-stale. The rule key cannot be changed later.

Choose what to watch

Pick the Metric, the Aggregation, the Condition and the Threshold (see the field reference below).

Choose the timing

Set Lookback window (s), Must hold for (s) and Cooldown (s). The helper text under each field shows the time in minutes or hours.

Choose severity and grouping

Pick the Severity and, if the metric allows it, Group alerts by.

Add escalation if needed

Under Escalation, click Add escalation level for each level and enter Level 1 after (s), Level 2 after (s) and so on.

Create

Leave Rule is enabled ticked and click Create rule.

Expected result: "Rule created. It is evaluated on the next check."

If you see "That rule key is already in use.", choose another key.

Field reference

Field	What it does	Default
Aggregation	How the values in the lookback window are combined: Latest, Sum, Average, Minimum, Maximum or Count.	Latest
Condition	is above (>), is at or above (≥), is below (<), is at or below (≤), equals (=), does not equal (≠).	is above (>)
Threshold	The value to compare with, in the metric's unit. For times, enter seconds (1800 for 30 minutes).	0
Lookback window (s)	How far back values are looked at for the aggregation.	300
Must hold for (s)	How long the condition must stay true before an alert opens. 0 means it fires immediately. Unless Aggregation is Latest, the lookback window must be at least this long.	0
Cooldown (s)	After an alert from this rule is resolved, how long before the same alert can open again. 0 means it can re-fire straight away.	0
Severity	Info, Warning or Critical. Shown on the Alerts page and available to notification overrides.	Warning
Group alerts by	Split the rule into one alert per centre, store, item or queue type. Up to 5 keys. Without grouping you get one alert for the whole organisation.	None
Only evaluate samples matching	Label filters: limit the rule to one centre, store, item or queue type. Click Add label filter, pick a Label and enter its Value.	None
Escalation	Up to 5 levels. Each level sends an Alert Escalated notification if the alert is still firing that many seconds after it started. At most 24 hours.	None
Rule is enabled	Disabled rules are not checked.	On

Group by place so the right people are told

Grouping decides what each alert is about. A low-stock rule grouped by store_uuid and item_uuid gives one alert per item in each store, and lets notification rules send it to that store's staff. A walk-in wait rule grouped by centre_uuid gives one alert per centre, so each centre's alert can be sent to that centre's staff.

How escalation works

Escalation levels count from when the alert starts firing. At each level, if the alert is still Firing, LeapHealthOS sends the Alert Escalated notification with the level number. Acknowledging, silencing or resolving the alert stops the remaining levels.

Who receives escalations is set on the Alert Escalated event in SettingsPlatform ConfigurationNotification Rules. To send higher levels to more senior staff, add an override that checks escalation_level. See Notification events and recipient strategies.

Example rules

Goal	Metric	Aggregation and condition	Threshold	Timing	Group by
Approvals waiting more than 30 minutes	`hos.appointment.pending_approval_age_max`	Latest, is above (>)	1800	Must hold for 0	`centre_uuid`
Any overdue bills	`hos.bill.overdue_count`	Latest, is above (>)	0	Must hold for 0, Cooldown 3600	`centre_uuid`
Walk-in wait over 20 minutes for 5 minutes	`hos.walk_in.queue_wait_max`	Latest, is above (>)	1200	Must hold for 300	`centre_uuid`, `queue_type`
Stock at or below reorder point	`hos.inventory.low_stock_margin`	Latest, is at or below (≤)	0	Must hold for 0	`store_uuid`, `item_uuid`
WhatsApp failures above 20%	`hos.whatsapp.delivery_failure_rate`	Latest, is above (>)	0.2	Must hold for 600	None

For the low-stock example, also see Batches, expiry and low-stock alerts.

Manage rules

Each rule card shows the name, metric, severity, Enabled or Disabled, the condition in words and how alerts are grouped. Filter the list by Metric, Severity and State (Enabled and disabled, Enabled, Disabled).

Action	Effect
Edit	Change anything except the rule key. Click Save changes.
Disable	Asks "Disable this rule?". Every open alert from the rule is resolved. You can enable it again at any time.
Enable	The rule is checked again from the next check, about a minute later. Nothing re-opens immediately.
Delete	Asks "Delete this rule?". The rule stops raising alerts and leaves the list. Its open alerts are resolved; past alerts stay readable on the Alerts page.

Silences

A silence holds back matching alerts for a time window: a maintenance slot, a known outage, a centre that is closed for the day. While a silence is active, matching alerts that open start as Silenced and nobody is notified.

Create a silence

Start a new silence

Administrator

Settings > Platform Configuration > Alert Rules

Open the Silences tab and click New silence.

Choose what it covers

Pick a Rule (or leave "Any rule"), a Severity (or "Any severity"), and optionally add Label and Value pairs with Add label to narrow it further.

Choose the window

Set Starts at (leave empty to start now) and Ends at (one hour from now by default). Add a Reason such as "Planned maintenance".

Create it

Click Create silence.

Expected result: "Silence created. Alerts opening inside the window stay quiet."

A silence with no rule, severity or labels silences everything

If you leave all the matchers empty, LeapHealthOS asks "Silence every alert?". Confirming suppresses every alert in the organisation until the silence ends, and nobody is notified about anything.

A new silence does not change alerts that are already firing. To pause an alert that is already open, use Silence on the alert itself on the Alerts page.

End a silence

The Silences tab lists silences with what they match, their window and Active or Expired. Use State to show "Active only" or "Active and expired". Click End now to stop an active silence early and confirm "End this silence now?".

Result: "Silence ended. Held alerts return to firing on the next check." Alerts it was holding return to firing about a minute later if their condition still holds, and their notifications go out then. The silence stays in the list as history.

alertsalert rulessilencesescalationoperationsadmin