Search documentation

Search for a topic, a control, or an event type.

    Docs / Workflows / Runs

    Runs

    A run is the record of one event’s path through a workflow. Test runs let you choose what happens for real. Live runs do everything their jobs are set up to do.

    A run starts when an event reaches the first job of a workflow. Each job that runs can publish an output event, and the jobs listening for that event continue the same run. One run therefore covers every job that a single starting event reached. For each job, the run records the input event, status, reported activity, output event, duration and cost.

    The Logs panel in the builder shows a workflow’s runs, and the Runs tab in Activity lists runs across workflows. Logs and Activity explain how to read them.

    Kinds of run

    The Logs panel labels each run with its type.

    Triggered run
    A live workflow started by an incoming event that matched a job’s trigger and passed its Filters. An incoming event that does not pass them starts no run and leaves no record.
    Manual run
    A live run started by a person, such as Run workflow on the live version with every permission on.
    Test run
    Run workflow on a draft or paused workflow, or with any permission turned off. It follows the permissions described below.
    Partial test run
    One job started on its own with Run job.
    Re-run
    A recorded job started again from its original input event.

    Run workflow and Run job

    Run workflow in the builder’s run toolbar starts a run of the version open on the canvas. A run needs an input event. OAIZ selects the input of the first job’s most recent run when it still matches the trigger, and otherwise the latest stored event that matches. Run workflow stays disabled until an input event exists, unless the workflow starts from a Schedule.

    The arrow beside Run workflow, labelled Configure run workflow, opens Permissions and Change event. Change event opens Select Test Event, which lists stored events filtered by the workflow trigger. Show all lists every event.

    Run job, the play button beside a job on the canvas, runs that one job with its own Permissions and Change event. It always runs as a test and starts with only Read on. With Publish Events off, only that job runs. With Publish Events on, its output event can start the jobs after it. A later job can run on its own once the jobs before it have finished with a matching output event.

    During a run, the toolbar shows Running workflow and the elapsed time. Stop watching stops showing progress. Jobs that are already running continue, and jobs that have not started do not start. A run that publishes events cannot be stopped from the builder. Its close button only hides the progress, and Open in Runs in the button’s tooltip shows the run in Activity.

    Permissions

    Permissions decide what a run does for real. Each card lists the integration actions and jobs it affects in this workflow. An entry marked May be used belongs to an Agent or Code job, which calls that tool only when its model or code decides to.

    Read
    Integration actions that read from a connected service, such as listing records. With Read off, they return example results instead of calling the service.
    Write
    Integration actions that change a service, Webhook requests, database writes and knowledge document changes, including the Agent and Code tools that write to a database or document. With Write off, an integration write returns example results and a webhook sends no request. A database write runs and is rolled back without row events, and a knowledge job saves nothing.
    Publish Events
    Each job’s output event goes onto the event queue, where other jobs and workflows listening for it can start. It also covers Agent and Code jobs that create events. With it off, output events stay inside the run, and Delay and Wait for Event continue at once instead of waiting.

    Model calls and image generation run on every run, test runs included. Always runs in the menu lists those jobs and the Agent or Code tools that act regardless of permissions. Agent and Code jobs always run too. The permissions limit only the integration, database, knowledge and event tools they call.

    Human Approval has no permission. In a test run it is auto-approved and nobody is asked. In a live run it waits for a decision in Approvals, unless its own auto-approval conditions match.

    A test starts with Read on and Write and Publish Events off. On the live version, all three start on. The choices apply to runs started from the builder and are not saved with the workflow. Each action reference describes that action’s test result.

    Test runs and live runs

    Run workflow runs for real only on the live version with Read, Write and Publish Events all on. That run is a Manual run of the live workflow, so approvals wait for a person and delays wait for their time. Turning any permission off, or running a draft or paused workflow, makes it a test run. A test never makes a draft live. Go live does that, as Workflows explains.

    Live runs also start on their own. While a workflow is live, each incoming event that a job listens for and that passes the job’s Filters starts a Triggered run, as Events and Jobs describe.

    Run and job statuses

    The canvas marks each job as a run reaches it. Logs and the Runs tab in Activity show the recorded status with the same labels.

    Job statuses
    StatusMeaning
    In queueWaiting for an earlier job in the run.
    RunningIn progress. The canvas shows Testing during a test run.
    SucceededThe action finished.
    Filters not metThe input did not pass the job’s Filters, so its action did not run.
    Event mismatchShown on the canvas when the input event does not match the job’s event source.
    FailedThe action returned an error.
    Timed outThe action ran past its time limit.
    RejectedA person rejected the Human Approval request.
    Budget exceededA spend limit stopped the job.
    Waiting for approvalHuman Approval is waiting for a decision.
    Waiting for delayDelay is holding the run until its time.
    Waiting for eventWait for Event is holding the run until a matching event or its deadline.
    Did not runThe run ended or was stopped before this job started. In a saved live run, a job whose Filters were not met also shows Did not run.

    A run shows Failed when any of its jobs failed. Otherwise a rejected, timed out, running or waiting job sets the run’s status, in that order. A run shows Succeeded when every job that ran succeeded, even when Filters stopped some of its paths, so a successful run does not mean every job ran.

    Re-running a recorded job

    Re-run step in Log Details starts a recorded job again. To re-run several, select up to 20 steps in Log Details or the Runs tab and choose Re-run selected steps. A confirmation dialog comes first.

    A re-run uses the job’s original input event with the job’s current live configuration. The configuration recorded with the original run is not restored. A re-run has Read, Write and Publish Events on, so it sends real messages and writes, and its output event can start the jobs after it. Each re-run starts a new run marked Re-run. A step whose input event was not recorded, or is no longer available, cannot be re-run.

    Notify on failure

    Notify on failure sends an email when a job fails or times out. The workflow’s setting is a switch in Edit Name & Details in the workflow’s menu. Each job also has a Notify on failure field in its details.

    Auto
    Follows the workflow’s switch.
    Enabled
    Emails for this job even when the workflow’s switch is off.
    Disabled
    Never emails for this job, even when the workflow’s switch is on.

    The email names the job, the workflow, the error and the time, and links to the job’s Log Details. It goes to the person who manages the job’s account, normally the person who created the job. A failed job in a test run sends it too.