Skip to main content

Runs and resources

Studio presents the same generic record model as graphenectl: a run owns a tree of resources, every record has state and history, and available operations come from its kind and the caller's permissions.

Start and follow a run

Open a pipeline, choose the active revision, and fill the generated typed form. Studio obtains this form from the revision schema; it does not keep a separate copy of pipeline parameters. Submit the run, then use:

  • trace for execution progress and dependencies;
  • owned tree for everything created under the run;
  • rerun to submit the same pipeline inputs again;
  • cancel to request cancellation of an active run.

Cancellation is asynchronous. Keep following the run until it reaches a terminal outcome; the worker still needs to finalize the owned resource tree.

Inspect a record

Selecting a run, resource, agent, artifact, or system record opens the generic inspector. Depending on its kind and permissions, it provides:

ViewUse
StateCurrent spec, outputs, phase, owner, and labels
EventsOrdered lifecycle and reconciliation history
LogsFiltered record logs and live tail
MetricsCharts from the configured metrics query backend
TracesSpans from the configured trace query backend
DownloadArtifact or other downloadable record payload
CommandsKind-defined operational commands

Transfer and delete are ownership operations, not UI shortcuts. Transfer moves the selected record subtree to another owner. Delete asks Graphene to finalize that subtree. Both remain subject to server authorization and resource-kind rules.

Topology and terminal

Topology combines ownership with explicitly declared data and control flow; it does not infer hidden network dependencies. Terminal opens an interactive session on a connected agent. Terminal access is privileged: use a least-privilege token, verify the selected installation and agent, and close the session when the task is complete.

If a view is empty, distinguish “the record emitted no data” from “the query backend is not configured.” Use the troubleshooting map and observability model to identify the layer.