Overview
Error Feed reads your traces, groups the problems it finds into issues, and points at the layer to fix
Errors in an AI system rarely show up as one clean failure. They show up as a pattern: the same kind of mistake repeating across dozens of requests, buried in traces you’d otherwise have to read one by one to catch.
What is Error Feed?
Error Feed reads a sample of the traces in an Observe project, decides for itself what went wrong in each one, and groups the traces that went wrong the same way into a single issue you work like a ticket. Nothing has to mark a trace as failed first: the scan is what finds the problem. Each issue carries a severity, a status, an assignee, and an optional link to a Linear ticket, and points at the fix layer, the part of your system the fix actually belongs in.
Before you start
Error Feed requires an Enterprise or Cloud license.
Scanning ships off because a project’s sampling rate starts at 0. Nothing is scanned, and no issues appear, until you raise it.
Start here
Turn on Error Feed
Raise the sampling rate on a project and get your first issues
Understanding Error Feed
The object model behind an issue, from a single finding to the fix layer
Severity & Status
The two independent axes every issue carries, and how they change
Triage issues
Filter, scan, and work a feed down with bulk actions
Investigate an issue
Read the evidence behind one issue before you touch the fix
Run a root cause analysis
Get a written root cause and a proposed fix for one issue
Questions & Discussion