Skip to content
Orvanta Docs

Workflow editor

The workflow editor is the node-based canvas where you wire individual steps into a workflow. Drag in nodes, connect them with edges to set the execution order, configure each node’s inputs, and run the whole thing (or just one node) to inspect results as they come in.

The shared canvas

Section titled “The shared canvas”

Historically, DAG Flow (this page) and BPMN Flow each had their own visual language for rendering a graph. As of 6.0.0 they share one canvas: DAG Flow renders and edits on the same canvas primitives as BPMN — shapes, icons, orthogonal edge routing, and a left-to-right or top-down layout toggle — rather than the older node vocabulary. There’s no artifact-format change behind this: a DAG Flow is still an OpenFlow JSON/YAML document, not BPMN XML; only the on-screen drawing changed.

On the shared canvas, control flow reads closer to a BPMN diagram than the old bracket-pair layout:

  • Branch one / branch all render as an exclusive (✕) or parallel (+) gateway, splitting into one lane per arm and rejoining at a merge gateway. Arms keep their declared order; an empty arm still reserves its own lane rather than being skipped.
  • For loop / while loop render as a single container node — a bordered box holding the loop body — instead of a separate start/end node pair.
  • A failure handler runs in its own lane, offset from the main flow so it doesn’t crowd the happy path.
  • A preprocessor appears inline right after Start, labelled to make clear it only runs when the flow starts from an external trigger.
  • An AI agent step renders as a hexagon with its tools listed underneath, connected by dashed association lines rather than solid execution edges — the same convention BPMN association lines use for “related to, not executed in sequence.”

Editing on this canvas works the way an n8n or Kestra user would expect: click a + on any edge or empty slot to open the Steps panel in the primary side bar (the same place the BPMN editor lists its shapes) and pick a card to insert it at that position, or at the end of the main lane when no + is pending (an embedded editor has no side bar and keeps a floating picker), drag a step to reorder it, copy and paste a step (or a whole branch/loop subtree — internal results.<id> references are rewritten automatically so a pasted copy doesn’t silently point back at the original), and collapse a group of steps to a single band.

Selecting a step opens a Step tab in the bottom panel at half the viewport height, so the step’s editor sits under the graph instead of covering it. The Runs tab lists editor test runs (flowpreview jobs) as well as deployed runs. The status bar shows a badge for the editor in use: DAG, BPMN, or a script’s runtime.

The run viewer

Section titled “The run viewer”

The run viewer uses the same canvas. Approve and cancel buttons sit on the suspended node itself, with the same resume request as the run page’s approval panel (a step with a resume form opens that form in a popover). A Graph / Timeline toggle swaps the graph for the Gantt timeline. Double-click a step to open a read-only view over the graph with this run’s input, the step definition, and its output and logs, with previous and next buttons in flow order. Assets are drawn as data-object nodes with arrowed associations, and an Assets toggle hides them.

This is the editor, not a preview. Since 6.0.0 the shared canvas is the only DAG canvas, in the flow editor and the run viewer; the classic node canvas and its toggle were removed. No screenshots yet — the canvas is still moving release to release.

Section titled “Related”