Skip to content

Executions

Open it: Executions · under Runtimes & runs in the console menu.

Every agent turn and every flow run is an execution: a run of a flow, kept with each of its steps. Items lists runs newest first, each with its flow, what started it, how long ago, how long it took and how it ended (completed, failed, running, paused, cancelled). It shows the runs of the project picked in the top bar that you may see: a project’s runs to its members, and your own runs of platform flows.

Executions: runs, newest first

A run's Overview: its status, what ran, who started it and when

TabWhat it shows
OverviewIts status and duration, how many steps were recorded and how often it resumed; why it failed, if it did, with the error each step gave; What ran: the flow, Started by, when it was queued, started and finished, and its run id. Run again starts it with the same input
StepsEach step it ran, with its time. Pick one for its input, output and the context after it
TimelineWhere the time went: waiting for a worker, running, paused, and each step, on one axis
JournalThe durable record of each step, which is what lets a run resume after a restart
Resume pointsWhere it can resume from
Input & outputWhat it was started with, and what it gave
What started itShown as
A triggerTrigger <name>, with Open; a schedule shows Schedule of trigger <name>
A turn of an agentA turn of agent <id>, for the person it acted for, with Open
An app or a person through the APIapi, and who

A run's Steps: each step with its time

Step paths show where each step sits in the flow: /load_agent/try/agent_row is the agent_row step inside the try of load_agent; /turn/sub/… are the steps of a subflow turn runs; for[0] is the first pass of a loop.

A run's Timeline: queued, running, and each step on one axis

The trigger test on the Triggers page started run 01a1df10-…: the help-desk agent’s turn, Started by Trigger refund-requests. It completed in 8.98 s over 24 steps; nearly all of it was /turn, the model call. Its Input & output tab shows the mapped message and the agent’s answer.