Skip to content

DocsUsing the workspace

Read an incident

What each part of the incident view means, and what to do with it.

Open Issues and select one. The page is arranged in the order you should read it: what happened, what it means, and what you can do.

The parts of the page

SectionWhat it tells you
SummaryA plain-language description of the failure
Root causeWhat the model believes started it, and why
Risk level and scoreHow urgent this is — calculated by rules, not by the model alone
Business impactWho is affected and how, in non-technical terms
Affected areasThe parts of the system implicated
Developer notesDetail written for whoever picks up the work
Suggested fixesOne or more proposed changes, each with an explanation
Event detailThe raw stack trace and request context, with sensitive fields removed

How much to trust it

The analysis is a starting point, not a verdict. Two habits keep it useful:

  1. Read the root cause against the event detail below it. If the trace does not support the story, the story is wrong.
  2. Treat the risk score as the priority signal and the narrative as the explanation. They are produced separately on purpose.

Closing the loop

Every incident can be moved to one of three states:

StateUse it when
AcknowledgedSomeone has seen it and owns it
ResolvedThe cause is fixed and you expect it to stop
IgnoredIt is known, accepted, and should stop asking for attention
Ignoring an issue does not stop it being collected — it stops it competing for attention. The history stays.