📈 Observability

A thousand identical stack traces are one problem — this makes them one node.

The watcher reads the deployment's logs, fingerprints each error, and folds recurrences into a single incident carrying its count and its first and last sighting. What you triage is a list of distinct problems rather than a scrolling wall.

What ships

• Log ingest for the cluster's own log shipper, gated by a shared secret.
• Incident nodes with fingerprint, occurrence count, and a resolution state.
• The triage surface — open incidents, newest recurrence first.

Contents