System Administration

Errors

Admins only: This tab is for site administrators.

The Errors tab collects errors (and optionally warnings) logged by every part of Kenseme: the web app and both MCP hosts. Identical errors are grouped by a fingerprint so a flood of the same failure reads as one row with a count. From any error you can see who hit it and where, jump to the full trace, and ask an AI agent to suggest a cause and fix.

In the app: Settings › System Administration › Errors tab (or go straight to /admin/errors)

[screenshot]
The Errors tab in Grouped view, with the filter row and a list of fingerprints sorted by Last seen.

Choose a view

The switch at the top right picks the view:

  • Grouped — one row per fingerprint: Last seen, Exception type, Sample message, Count, how many Users and Orgs were affected, the Last page, and the Host. Start here.
  • All events — one row per occurrence: Time, Level, Host, Exception type, Message, Page, User, Org, and Correlation.

Filter

Filter Options
Time range Last hour, Last 24 hours, Last 7 days, Last 30 days.
Level Errors or Warnings and above. Only in All events; Grouped shows errors.
Host All hosts, Web, Mcp (the schema MCP host), or McpData (the ontology data MCP host).
Organization id One organization’s numeric id, or blank for any.
Search Text in the message or exception type.

Click Refresh to reload, or turn on Auto to refresh on a timer. When you’ve drilled into one fingerprint, a Filtered to fingerprint chip appears; click Clear to go back to everything.

Read an error

Click a row to open the error’s detail. It’s organized into sections:

  • Who — the user’s name, email, and user id, and the organization.
  • Where — the page, host, app version, component, and source context.
  • What — exception type, message, when it occurred, and the stack trace.
  • Correlation — the correlation id that ties this error to the request’s other log lines and traces.
  • Occurrences — how often this fingerprint has happened. Show all events switches the tab to All events filtered to this fingerprint.

Two actions sit at the bottom of the detail:

  • Search in App Insights opens the matching query in Azure Application Insights.
  • Copy as Claude Code prompt copies a ready-made prompt describing the error, for pasting into a coding agent.

Copy and Copy JSON copy individual values or the whole record.

Ask an agent for remediation

The Agent remediation section asks an AI agent to explain the error from its trace and the code it points at.

  1. Click Ask agent for remediation. The request runs as a background job named “Error remediation”, which you can watch on System Administration’s Background Jobs tab. The section updates by itself when the run finishes.
  2. When it finishes, the section shows the agent’s Probable cause, a Suggested fix, the Affected locations (file, line, and why), a Risk note, and its Confidence.
  3. Click Copy fix as Claude Code prompt to hand the fix to a coding agent.

If the run fails, click Retry. Asking again for the same fingerprint while a run is already in progress reuses that run rather than starting a second one. Earlier runs are listed with when they were Requested and their Status.

Note: The agent’s answer is a starting point. Check the suggested fix against the code before acting on it.