Clusters overview

A cluster is a group of signals the Failure Agent has confirmed and grouped by cause. A single signal is one data point; a cluster is a pattern worth fixing. Experiments in Simulation also start by picking a cluster. If you have no clusters yet, confirm signals in Signals. Confirming a signal seeds the first cluster.

​

The list view

Open Clusters in the sidebar to see the workspace’s clusters as a table.

• Pattern the cluster’s summary title, generated by Nora from a representative signal
• Flow which Flow the failures came from
• Signals how many signals are grouped
• Cases how many cases have accumulated (see Cases)
• Severity
• Status Open, Improving, or Resolved

Filter from the top of the page.

• Search clusters
• Date range: All time, Last 7 days, Last 14 days, Last 30 days, Last 90 days
• Agent: All agents or a specific agent
• Status chips: All, Open, Improving, Resolved (each chip shows a count)

​

Cluster detail

Click a row to open the cluster. The header shows the title along with a category tag (for example tool failure or ungrounded claim), the status, and the case count, source-signal count, and Flow. The body is the case table (see Cases). You can also arrive from the signal side: View parent cluster in a signal’s detail opens the cluster that signal belongs to.

​

Managing folders

Clusters are managed like folders.

• New folder at the top right of the list creates an empty cluster folder.
• The ⋯ menu on the detail page offers Rename, Merge into…, and Delete folder.
• Deleting a folder does not delete its cases; they move to Unclustered.

​

Starting an experiment from a cluster

Run simulation at the top right of the detail page starts a new experiment with this cluster already selected (see New experiment).

​

Learn more