Skip to main content
A cluster is a grouping of similar findings, observations Bento pulled from your runs and bundled because they look alike. In the dashboard they’re the scatter view under Monitoring: each dot is a finding, and dots that sit close together form a cluster. The clusters group reads that view for the whole workspace. Where an issue is one tracked problem you work and close, a cluster is the looser, zoomed-out picture: the shapes in your monitoring data before anyone has named them. Each cluster in the workspace view carries a label, a description, a member_count, a trend, and a few sample_quotes. The view returns only active clusters, the ones worth looking at. Each member is a Finding, extracted from a Trajectory. A Signal is what turns a recurring cluster into a detector, and has its own clusters via signals list-clusters <signal-id>.

​
The commands

Both are read-only and workspace-scoped. Neither takes query flags or a body; the only options are the shared --workspace and --output.

​
Read the scatter view

See the clusters, with labels, counts, and a few quotes each:
Pull the names and sizes for an overview:
Get the plot points behind the picture. Each is a finding with its 2D coordinates, kind, severity, and quote; centers gives the labeled middle of each cluster. Unclustered points carry a null cluster_id:

​
See also

Signals

The detectors a recurring cluster becomes.

Findings

The individual observations that make up each cluster.

Trajectories

The analyzed runs those findings come from.