Alerts and weekly digest
Get notified when average health drops, danger or failure rates rise, or call volume changes, and receive a weekly summary for every workspace.
Dashboards are for looking; alerts are for not having to. VoiceLens evaluates alert rules every minute against a window of recent calls and notifies you by email and webhook when a metric crosses a threshold, then again when it recovers. A weekly digest summarises each workspace every Monday.
Alert rules
Each rule watches one metric over a window and compares it to a threshold.
| Field | Options |
|---|---|
| Metric | AverageHealthScore, DangerCallRate (share of calls with at least one danger fault), FailedCallRate (share of calls whose processing failed), CallVolume |
| Comparison | Below or Above |
| Threshold | The value to compare against |
| Window | The last N calls, or the last N hours |
| Channel | Optional. Scope the rule to one channel; otherwise the whole workspace. |
| Minimum sample size | The rule stays silent until the window contains at least this many calls, so a single bad call at 3 a.m. does not page anyone. |
| Notify | Up to 10 email addresses and/or one or more webhook endpoints |
| Notify on recovery | On by default. Sends a resolution when the condition clears. |
| Cooldown | Minimum minutes between two triggers of the same rule. 60 by default, up to 7 days. |
A workspace can have up to 25 rules. Creating and editing them requires the Admin role.
Examples
- Bot regression:
AverageHealthScoreBelow70 over the last 50 calls on the bot's channel, minimum sample 20. - Compliance drift:
DangerCallRateAbove0.10 over the last 24 hours, workspace-wide. - Pipeline trouble:
FailedCallRateAbove0.05 over the last 6 hours. - Silence:
CallVolumeBelow10 over the last 4 hours on a channel that should be busy; usually means the integration stopped sending.
What a notification contains
Emails and the alert.triggered webhook event include the rule name, metric, comparison and threshold, the observed value, the sample size, the window, the channel (if scoped), the time, a link to the dashboard, and up to a handful of sample calls from the window so the recipient can start reviewing immediately. When the condition clears and recovery notifications are enabled, an alert.resolved event and email follow.
Webhook payload shapes are documented in Webhooks.
Weekly digest
Every Monday at 07:00 UTC, each active member of a workspace receives one email covering the previous seven days compared with the week before:
- Calls received, completed and failed, and how many contained a danger fault
- Average health score and its change
- The five most frequent faults
- The five busiest channels with their health
- Alerts triggered during the week
Weeks with no calls are skipped rather than sent empty. The digest is toggled per workspace under workspace settings (Owner or Admin).
Threshold webhooks without a rule
Independently of alert rules, a webhook endpoint can carry its own health-score threshold. Every completed call that scores strictly below it, or that contains at least one danger fault, produces a threshold.breached event for that endpoint alone. This is the right tool for routing individual calls to a QA queue; alert rules are the right tool for trends.
Recommendations
- Scope bot alerts to the bot's channel and set the minimum sample size to at least 20; bots produce bursts.
- Keep one workspace-wide
FailedCallRaterule; it is the earliest sign of a broken integration or an expired token. - Point alerts at a shared inbox or a webhook that opens a ticket, not at one person's email.