Run statuses

Every run carries one status:

  • pending: the run is queued and has not started.
  • running: nodes are executing.
  • paused: a node failed under Pause on first failure. Resume continues it.
  • awaiting approval: the run waits for approval under Needs your decision in Automations; see Run policy.
  • completed: every node finished.
  • failed: a node failed, or the run timed out.
  • cancelled: stopped by Stop or Force cancel in the editor, or because its deployment was paused or stopped.
  • skipped: the run started and found nothing to do.

Skipped is not a failure. A run skips when each failed node had nothing to work on, such as a missing symbol, an empty watchlist or an app turned off in the project. If any other node fails, the run is failed.

Run History

Run History, the clock button in the toolbar, opens a side panel of this workflow's runs, newest first. Each row shows the status, any error, the duration, when it finished, and a word for what started the run:

  • manual: Run in the editor, or chat.
  • schedule, webhook or event (a signal).
  • batch or episode: a dashboard's Run watchlist or Run all. Each is an ordinary run of main; see Run workflows and batch runs.
  • sub-workflow: a Run workflow node.

  1. Expand a run

    Click its row. It opens with the Run id and Nodes: each node with its status, duration and any error.

  2. Click View on canvas

    The nodes take on that run's state and the Console opens with its outputs.

Run and node timeouts

  1. Open the settings

    Click the gear at the right of the toolbar (Settings). Workflow Settings opens.

  2. Set the budgets

    Under Wall-clock timeouts, Run timeout (minutes) caps the whole run (default 60), and Node timeout (minutes) caps any one node (default 10). Each takes 1 to 1440.

  3. Choose what a failure does

    Pick an option under On node failure.

  4. Click Save

    The toast reads Timeouts saved.

A run copies these settings when it starts, so a change never touches a run in flight. A run that passes its budget fails, and so does a node that passes its own; see A run timed out.

After a node fails, On node failure decides the rest:

  • Continue (current behavior): nodes that depend on the failed one are skipped; independent branches keep running.
  • Halt on first failure: everything still to run is marked skipped and the run ends as failed.
  • Pause on first failure: the run pauses, and a banner reads Run paused at and the node's id. Resume reruns the failed node from the version the run started with and reuses the outputs of the nodes before it.

Versions

The row under the toolbar names the version the canvas holds:

  • v3 with a Main badge: the main version, or with a Make main button when it is not main.
  • Edited: changes no version has. Save version… sits beside it.
  • No versions: nothing has been saved as a version yet.

  1. Click Save version…

    Unsaved edits are saved first.

  2. Say what changed

    Type a note in What changed?, up to 200 characters, or leave it blank.

  3. Click Save

    The toast reads Saved v4.

Identical content never gets a new number: if the canvas already matches a version, the toast reads No changes since v3. Moving nodes or autosave makes no version. A backtest of code that has no version gives it one, so every backtest records the version it tested. A Program must be built before the workflow can be saved as a version.

Versions opens the list, newest first, with each version's note and who saved it and when; a version an agent saved reads Agent. Each row's More menu holds Restore and Clone.

Main

Until a workflow has a main, the next version you save becomes main on its own, with two exceptions:

  • An enabled schedule, webhook or signal cannot run that version, because it needs an input that automation does not supply. Main waits, and the Saved v1 toast says why: Morning scan can't run v1.
  • The workflow is a Rules or Program strategy. Main starts at the first version whose backtest, run by a person, completes.

A version an agent saves, or an agent's backtest, never starts main.

After that, main moves only by Make main: from the bar, the Versions list or the strategy's page in Backtesting. Its toast offers Undo when there was a main before. Make main refuses for the same two reasons, and main stays where it was: Morning scan can't run v4. or Backtest v4 first.

What a run executes depends on what started it:

  • Run in the editor, or chat: the canvas.
  • A schedule, a signal monitor, an incoming webhook, a dashboard (a batch included), another workflow or an API key: main.
  • A deployment: the version it was deployed with; see Deploy a strategy.

A workflow with no main runs the canvas everywhere except in a deployment. Once it has one, an edit on the canvas changes nothing automated until you save a version and make it main.

Restore and clone

Restore copies a version's graph onto the canvas, and main stays where it is.

Clone makes a new workflow from the version, named after this one with copy added, and opens it. It starts with no versions and no automations.