Build & Connect

Adding a node

Open the editor and click the + Node button in the canvas toolbar, or drag a node from the left drawer onto the canvas. Pick the node type that matches what you're modeling:

• Dataset — Anything that produces data (a CSV, an image folder, a streaming source).
• Model — A trained or trainable model (transformer, classifier, embedder).
• Logic — Transformations, filters, joins, splits.
• Custom — Your own Python node when nothing else fits.

Every node has a default label and a slug. Click the label to rename it; the slug stays stable so connections survive renames.

Wiring ports

Hover a node and you'll see input ports on the left and output ports on the right. To create a connection:

1. Click and drag from an output port.
2. Release on a compatible input port on another node.

Ports are color-coded by semantic type — image, tensor, float, string, label. Two ports connect cleanly when their types match.

Editing node config

Click any node to open the Inspector on the right. The Inspector has two tabs:

• Config — Name, label, parameters, custom key/value pairs.
• Run Data — Results from the last run, populated after you execute the graph.

Press Escape or click anywhere on the canvas to dismiss the Inspector. A thin tab on the right edge of the canvas re-opens it for the last selected node.

Selecting and moving

• Click a node to select it.
• Shift-click to add to the selection.
• Drag on empty canvas to draw a marquee.
• Cmd/Ctrl+A selects everything.
• Delete removes selected nodes (and all edges touching them).

Tidy up

Once your graph gets dense, hit the Tidy button in the canvas toolbar to auto-layout nodes into a clean topological flow. Tidy preserves selection and connections; it only repositions.

Saving and undo

Every mutation pushes onto a per-variant history stack. Use Cmd/Ctrl+Z to undo and Cmd/Ctrl+Shift+Z to redo. Your graph saves to local storage automatically; you don't need to hit save.