TRAINYOURAGENT

What is up, what is degraded, and what happened.

Component-level status for the phone leg, the reasoning chain, integrations, and the customer portal, plus a history of past incidents with real root causes. Degraded is reported as degraded rather than as operational-with-a-note, because the note is what customers never read.

Components tracked separately

How incidents are communicated

Affected customers hear directly, not just from this page, and they hear before the incident is resolved rather than after. Updates carry what we know, what we do not yet know, and the next update time. A written post-mortem follows within a week for anything customer-visible.

Why upstream outages still count

When a model provider degrades, the fallback chain absorbs it and the incident is still logged here, because a caller experiencing a slower agent does not care whose infrastructure caused it. Reporting only our own faults would make this page uselessly clean.