Notification / Alerting Agent

The backend agent that watches error rates on long-running batch jobs as they happen and sends multi-channel alerts the instant a threshold is crossed or a critical task completes.

What Is the Notification / Alerting Agent?

The Notification / Alerting Agent is one of TotalApp's Core Agents. It watches error rates and progress on long-running batch jobs — a large reconciliation run, a regulatory reporting batch — as they happen, and sends multi-channel alerts (email, in-app terminal, canvas) the instant a threshold is crossed or a critical task completes.

A batch job can fail quietly in the middle of the night, and by the time anyone checks a dashboard the window to react has already closed. The Notification / Alerting Agent exists so nothing fails silently — the right people know the moment it matters, on whichever channel they're actually watching.

In one sentence

Configure a threshold for a job type — the agent watches error rate and progress in real time and fires a multi-channel alert with enough context to act, the instant that threshold is crossed or the job completes.

How It Works — Watch, Detect, Alert

1. Threshold Monitoring → 2. Breach Detection → 3. Multi-Channel Alert

Phase	What happens
1. Threshold Monitoring	Monitors error rate and progress in real time, not just at job completion. Thresholds are configurable per job type — a reporting batch and a data import have different acceptable error rates.
2. Breach Detection	Distinguishes a transient blip from a genuine threshold breach, so a momentary spike doesn't trigger unnecessary alert noise.
3. Multi-Channel Alert	Sends through email, in-app terminal, and canvas notifications from one alert, routed to the right recipients based on job ownership and on-call configuration.

Success alerts too, not just failures

The agent also sends completion alerts for critical tasks — knowing when a critical batch finished successfully can matter just as much as knowing when one failed, especially for time-sensitive processes like regulatory reporting.

Key Capabilities

Real-Time Threshold Watch

Tracks error rates and progress live, not just at job completion — catching a problem while there's still time to react.

Email, Terminal & Canvas

One alert reaches every relevant channel — nothing depends on someone checking one specific screen at the right moment.

Smart Recipient Routing

Alerts route based on job ownership and on-call configuration, not a static distribution list that goes stale.

Completion Alerts Too

Notifies on successful completion of critical tasks, not just failures — closing the loop on time-sensitive jobs.

Output — What You Get Back

Field	Meaning
`alert_type`	Threshold breach, stalled step, or successful completion of a critical task.
`channels`	Which channels the alert was sent through — email, in-app terminal, canvas, or a combination.
`recipients`	Who received the alert, resolved from job ownership and on-call configuration.
`context`	What failed (or succeeded), how many records were affected, and where to look next — enough detail to act immediately.

Not tied to one specific integration

The agent works across any long-running batch job configured to report progress and error rate — it is not hard-wired to a single reporting pipeline or job type.

Frequently Asked Questions

Does every job type share the same alert threshold?

No — thresholds are configurable per job type, since a reporting batch and a bulk data import don't have the same acceptable error rate or urgency profile.

Will a brief error spike trigger a false alarm?

The agent is designed to distinguish a transient blip from a genuine threshold breach, reducing alert fatigue from momentary noise that resolves on its own.

Who receives an alert?

Recipients are resolved from job ownership and on-call configuration, not a static list — so alerts reach whoever is actually responsible at the time, rather than a fixed set of names that may be outdated.

Does this only fire on failures?

No — it also fires completion alerts for critical tasks, so successful completion of a time-sensitive job is communicated just as reliably as a failure would be.