The Workflow Editor

The workflow editor is the screen where you open a workflow, look at its graph, fill in its inputs, run it, and watch the run. You reach it from the Execute Workflow tile on the Home screen. Opening it that way loads a world into the simulator first, so you may see a full-screen overlay reading Loading world… before the editor appears.

This page describes what is on the screen. For how to design a workflow, pick node types, and bind inputs, see Authoring a workflow. For custom input canvases, see Creating a canvas.

Layout

The screen is a top bar above three side-by-side panels. Every divider between the panels is draggable, so you can widen the panel you are working in. The right panel can be dragged all the way closed.

The top bar

The bar is titled Workflow Editor. Reading left to right:

Workflow

A searchable dropdown under the label Workflow. Closed, it reads Select workflow...; open, it has a search box reading Search workflows... and shows No workflows found. when nothing matches. The first entry is always Create New Workflow. Choosing it swaps the dropdown for a text box (Enter workflow name...) with a tick button to confirm and a cross to cancel. Enter confirms, Escape cancels.

Next to the dropdown is a refresh icon button that re-reads the workflow list from the server.

The save control

There is no permanent Save button. A save icon button appears beside the workflow dropdown only when the workflow has unsaved changes, and its tooltip reads Save changes. When you save, a coloured banner appears directly under the top bar with the result. A success banner clears itself after a few seconds. A warning or an error stays until you dismiss it with its close button.

In a read-only sample project the save control is disabled, so edits stay local to your browser tab.

World and Load

A second searchable dropdown under the label World (Select world..., Search worlds..., No worlds available.), followed by a Load button. Picking a world in the dropdown does not change the scene. Load is what pushes the chosen world into the viewport, and it is disabled while a run is in progress or while another load is running.

Status badge

A single badge showing the editor's execution state. It reads one of READY, RUNNING, STEPPING, PAUSED, ROLLING BACK, COMPLETED, FAILED, CANCELLED.

Execution world

A dropdown offering Simulation and Real Hardware.

Reset

In the cloud, the next control is a Reset button, labelled Reset arms to watching position. It moves the simulated arms back to their default pose. It is disabled while a run is in progress.

On a local install this position holds a pipette reconnect icon button instead, labelled Reconnect pipette. That button is not present in the cloud.

Run, Pause, Resume

One slot in the bar changes with the run state.

• Idle: a Run button. It opens a dialog titled Name this run with a text box (e.g. calibration test 3) and Cancel and Start buttons. The name is what you will see later in the resume picker. Leaving it blank uses an auto-generated id. Run is disabled while the graph, the world, or the object list is still loading, and while a custom input canvas is attached but not yet confirmed.
• Running or stepping: a Pause button.
• Paused: one or more resume buttons. If the paused step offered no choices, you get a single green Resume. If it did, you get one button per choice, labelled with the text the step supplied.

Resume run

A separate dropdown button labelled Resume run, next to the run slot. Opening it lists past runs that did not finish, under the heading Resume an incomplete run; each row shows the run's name, its status, and when it started. While the list loads it shows Loading…, and if there is nothing to offer it shows No resumable runs. Picking a row opens a confirmation dialog before anything happens, because resuming resets the scene and reseeds the inputs. The button is disabled while a run is active.

This is not the same control as the green Resume above, which continues the run that is paused right now.

Stop

A red Stop button, always visible and disabled when nothing is running.

World View

The left panel, headed World View, with an information tooltip reading Real-time 3D preview of the world state during workflow execution. It shows the scene as the run moves it. At the right of the header is a gizmo toggle, whose tooltip reads Hide gizmos or Show gizmos depending on its current state.

The viewport itself has its own controls and its own quirks. See The Viser 3D viewport.

Graph Editor

The middle panel, headed Graph Editor, holds the graph. Nodes are boxes, edges are the arrows between them.

You can click a node or an edge to select it (its details then appear in the Edit tab), drag a node to move it, drag from a node's handle to another node to create an edge, and press Delete with a node or edge selected to remove it. Dropping a node type from the palette onto the graph places it where you dropped it. Editing is switched off while a run is in progress.

A spinner appears next to the panel heading while the graph is loading.

The progress strip

Once a run starts, a strip appears below the graph:

• A header row with elapsed time, the run status, and a count of completed nodes against the total number of nodes in the graph.
• A highlighted row naming the step currently executing.
• A scrolling list of finished steps, newest at the top, each with a tick, the clock time it finished, its label, and how long it took.
• A red row with the error text if the run failed.

The right panel

Four tabs across the top: Schema, Inputs, Edit, Log. Two of them carry a dot indicator: Inputs shows an amber dot when the workflow has inputs that are not all filled in, and Log shows a green dot when there are log entries.

Schema

Headed Input Schema, with the subtitle Define inputs required before execution and an Add Input button. This tab defines what a run will ask for, not the values themselves.

Each row has Name (e.g., Destination), a Type dropdown, Description (optional) (Brief description), an Array checkbox, and a delete icon. A duplicate name is flagged in red with Duplicate name - must be unique. For types that support a nested shape, a button reads Define Schema, or Edit Schema (n fields) once one exists, with the note Schema defined beside it. Empty, the tab reads No inputs defined yet. and Add inputs to parameterize this workflow. The footer counts how many inputs are defined and flags duplicates.

The whole tab is disabled while a run is in progress.

Inputs

This is where you supply values before a run.

By default it shows the generated form, headed Workflow Inputs, with a refresh icon button that re-reads the objects in the world. Under the heading is a completion line reading All inputs set in green, or the number of inputs still remaining in amber. Each input is a card with its name, a badge for its type, an array badge if it takes a list, its description, and a control matched to its type. Filled cards turn green. If a workflow carries older-style object bindings, they appear below a divider headed Legacy Object Bindings, each with its display name, its alias, and a dropdown reading Select world object..., marked Bound once set. The footer counts the world objects available to bind against. With nothing to fill in, the tab reads No inputs defined. and Ready for execution.

If the workflow has a custom canvas attached, the canvas replaces that form. While it loads you see Loading canvas…. If it cannot run, an amber card reads Canvas unavailable with the reason and a Use standard form instead button that switches back to the generated form for this session. A workflow with a canvas will not start until you have confirmed your setup inside the canvas.

The canvas authoring view is reached from the Edit tab, opens over this tab, and closes with its Done button.

Edit

What this tab shows depends on what is selected in the graph.

With an edge selected: a card headed Edge Connection with the edge's condition as a badge, a From: and To: row naming the two nodes, and a Delete Edge button.

With a node selected: a Node Label text box, a badge reading the node's type followed by the word node, the node's id, and a Description box. Below that is a Delete Node button. It is disabled for the entry and exit nodes, with the note Start and End nodes cannot be deleted.

Then a section that depends on the node's type:

• Skill nodes get a searchable Skill dropdown (Select skill..., Search skills..., No skills available.) with the chosen skill's description underneath, then Skill Parameters: one control per parameter, each with a source picker so a parameter can take a fixed value (Set manually) or come from a workflow input. A skill with no parameters shows This skill has no configurable parameters.
• Loop nodes get a Loop Type dropdown offering Count (Fixed iterations), Collection (Iterate over items), and Conditional (While condition), plus the field that type needs: Iterations, Collection Key, or Condition.
• Conditional nodes get an Expression box for the branching condition.

With nothing selected: the tab becomes the build panel.

• Edit Graph, a line of guidance reading Click a node in the graph to edit its parameters, or add new nodes below.
• Add Nodes, holding the node palette.
• Canvas, with a single button reading Add canvas or Edit canvas, and a line stating whether a canvas is already attached.
• Tips, three reminders: Drag from node handles to create connections, Select a node or edge and press Delete to remove it, Drag nodes to reposition them on the canvas.

The node palette

The palette lists node types, not individual skills. It offers exactly three: Start (Graph entry point), Skill (Execute a robotic skill), and End (Graph exit point). Its footer reads Click to add or drag to position, which is what it does: clicking adds the node, dragging drops it where you release.

Adding a skill step is therefore two moves. Add a Skill node from the palette, then select it and choose which skill it runs from the Skill dropdown described above. Loop and conditional nodes are not in this palette; the editor can display and edit them, but does not create them here.

Log

Headed Run Log, with a Download .txt button that is disabled until a run exists.

Below the header is a row of filter chips, all, notes, steps, events, and a search box (Search…). Entries are grouped by step: a step header shows its time, a tick when finished (with its duration) or starting… while it is still going, and collapses to hide the entries underneath it. Filtering or searching flattens the grouping.

Scrolling away from the newest entry raises a Jump to latest button in the corner.

Empty, the tab reads Start a run to see the log., or Log will populate as the run progresses. once a run is underway. If your filter matches nothing it reads No entries match your filter.

At the bottom is a note composer with an Add note button. Its box reads Add a note… (Enter to send, Shift+Enter for newline) during a run, and Start a run to add notes otherwise. Enter sends, Shift and Enter together add a line break.

Dialogs raised during a run

If a step asks you a question, a dialog appears over the editor with the question, any images the step captured, and one button per answer. Its wording comes from the step, not from the app. You cannot dismiss it with Escape, by clicking outside, or with a close button. Answering is the only way out.