KUBERNETES / INCIDENT WORKSPACE

Investigate the symptom.
Show the evidence.

Turn Pod events and status output into a reviewable investigation brief, with source passages from official Kubernetes documentation.

Browser-based analysis · pinned source links · exportable incident brief

01 Describe the symptom02 Inspect the evidence03 Keep the brief

Describe the incident

Use sanitized diagnostic output. Recognized credential patterns are redacted locally, but detection is incomplete. Remove sensitive data before pasting.

Synthetic examples

This public deployment has no model endpoint configured. The evidence workflow is live; model examples, when available, are labeled recorded. How this workflow works ↗

Loading pinned runbooks…

YOUR INVESTIGATION

Start with one useful clue.

Choose a synthetic example or describe your incident. We’ll match the symptom to a pinned runbook and keep the supporting passages visible.

EXAMPLE SIGNALImagePullBackOff

Observed status → matching runbook → source evidence → investigation brief

The analysis uses the text you provide. Your cluster stays under your control.

02 / A VISIBLE DECISION PATH

Input → signals → evidence

Known status strings expand the retrieval query. The same pinned passages remain inspectable, whether or not a model is configured.

Model outputs must pass checks

The optional answer path requires valid JSON, supplied source IDs, matching quotes, and supported read-only command forms. Failed or truncated responses are rejected.

Measure before claiming a win

A frozen protocol defines a future comparison of excerpt-only retrieval and base/SFT served artifacts. Model inference is deferred. Completion, evidence validity, phrase coverage and abstention are separate measurements.