Alarm thresholds and notifications
A threshold defines when a call statistic requires attention. Set a threshold value for the statistic that you want to monitor, and Live Hub raises a Threshold Alarm when the statistic crosses that value.
Each metric supports two alert levels: Minor alert and Major alert. You can configure either level independently or configure both levels for the same metric.
Live Hub evaluates each alarm over a 5-minute window, and raises an alert when the aggregated value crosses a configured threshold.
Alarms created before release 2.20 use Average over the 5-minute window. To alert on the worst call instead, edit the alarm and select Max.
Configure alarm thresholds
The Thresholds tab lists a set of default alarms, and lets you add your own across Calls, LLM, Tools, Bot latency, Quality, Billing, and, where your account has post-call-analysis metrics defined, Insights.
To view the alarm thresholds:
-
In the Navigation pane, select Alarms, and then select the Thresholds tab.
-
Read the table.
Column What it holds Enable/disable toggle Whether the alarm is active. Alarm name The name you gave the alarm, or its default name. Metric The statistic it watches. Threshold Exceeds threshold or Drops below. Value(s) The value the alarm watches, for a metric with a fixed set of values, such as Voice quality. Entity The bot, connection, or number the alarm applies to, or All. Minor The value that raises a minor alert. Major The value that raises a major alert. -
Click a toggle to activate or deactivate an alarm, or click the filters icon to narrow the table. The number of active alarms appears above the table, for example 2/2 Active.
Default alarm thresholds
A new account starts with the following 7 alarms. 'Failed calls (%)' and 'Poor quality calls (%)' are active by default. The others stay inactive until you turn them on, and you can delete any of them. Each default alarm applies to All entities.
These defaults apply to standalone accounts and subaccounts. A parent account has a single default alarm, 'Remaining funds ($)'.
| Alarm name | Metric | Value(s) | Threshold | Minor | Major |
|---|---|---|---|---|---|
| Average call duration (sec) | Call duration | – | Drops below | 10 s | 5 s |
| Calls rejected by limit | Calls rejected by limit | – | Exceeds | Off | Off |
| Failed calls (%) | Call status | failure | Exceeds | 5% | 10% |
| Maximum call attempts per second | Max call attempts per second | – | Exceeds | 2 calls/s | 5 calls/s |
| Maximum concurrent calls | Max concurrent calls | – | Exceeds | 3 calls | 5 calls |
| Poor quality calls (%) | Call min quality | poor | Exceeds | 5% | 10% |
| Remaining funds ($) | Remaining funds | – | Drops below | $10 | $5 |
The billing alarm depends on the account's payment type. A prepaid account gets 'Remaining funds ($)'. A postpaid account gets 'Monthly cost ($)' instead, which raises an alert when the month's cost exceeds $100 (Minor) or $1000 (Major). A subaccount's billing alarm alerts only the subaccount, not its parent account or the parent's other subaccounts.
'Calls rejected by limit' has no levels set. To use the 'Calls rejected by limit' alarm, edit it and set a Minor value, a Major value, or both.
Add or edit a threshold alarm
To add a threshold alarm:
-
Click Add threshold alarm, or, to edit an existing alarm, click the Edit icon on the alarm's row.
-
In the 'Alarm name' field, enter a name unique in the account.
-
In the 'Metric' field, click Select. Browse the categories LLM, Tools, Bot latency, Calls, Quality, and Billing, or search by name, description, or category. Click the metric to watch, and then click Select.
-
From the 'Entity' drop-down list, select All, or a specific entity. LLM, Tools, and Insights only offer AI Agent bots.
-
From the 'Threshold' drop-down list, select Exceeds threshold or Drops below.
-
Depending on the metric, set one more field:
- A metric with a fixed set of values, such as Call status, adds a 'Value(s)' field. Select the value to watch, for example failure.
- A bot error metric adds a 'Value(s)' field listing the ways a call can end, for example busy. Select every value that counts toward the alarm. Only the values you select are counted. This applies to bots only.
-
A numeric metric, such as Call duration, adds an 'Aggregation' field. Select Average, Sum, Min, or Max. The options depend on the metric.
Aggregation Use it for Average The typical value over the 5-minute window. Occasional errors can be hidden by normal calls. Sum The total over the window, which works like a rate, for example the number of LLM failures in 5 minutes. Max The worst single value in the window, for example the longest call delay. Min The lowest single value in the window.
-
Select Minor, Major, or both, and enter a value for each. At least one is required.
-
Click Create, or Update if you are editing.
Insights alarms only appear when your account has post-call-analysis metrics defined. If a metric used by an alarm is later deleted from post-call analysis, the alarm stays in the table marked as no longer available, and you can only delete it.
Setting a threshold requires the livehub/alarms/thresholds/update role, and viewing
the tab requires livehub/alarms/thresholds/read.
Configure alarm notification
Configure email notifications to receive updates when alarms are raised or cleared.
Live Hub sends notification emails approximately every five minutes. Each email includes only the alarm changes that occurred since the previous notification.
To enable alarm notifications:
-
In the Navigation pane, select Settings.
-
Select the Advanced tab.
-
From the 'Forward alarms via email' drop-down list, select Enabled.
-
Click Update.