Documentation
Node graph
How a graph works
A graph is a set of nodes connected by wires. Geometry flows from a node’s output (bottom) into the inputs of the next node (top). Each node does one job and returns new geometry; nothing is modified in place, so every step stays editable. Results are cached: when you change a parameter, only that node and what is downstream of it re-cook.
Adding nodes
- Right-click empty graph space, press Tab with the mouse over the graph, or click + Node. The node menu opens at the cursor.
- The menu has three columns: categories, the nodes in the selected category, and the docs of the highlighted node.
- Type to search across all 164 nodes. Use ↑ ↓ to move, ← → to switch between categories and nodes, Enter to add, Esc (or Tab) to close.
- The node is placed where you opened the menu and named after its type (
box1,box2…). If a node is selected and the new node has inputs, it is wired to the selected node; if the selected node held the display flag, the new node takes it over.
Right-click only opens the menu on a click: if you drag with the right button, it pans instead.
Connecting
- Drag from a node’s output port to another node’s input port. An input holds one wire; connecting a new one replaces the old.
- Nodes with several inputs (for example
copy_to_points: geometry, points) show one port per input.mergetakes up to 64 inputs. - Click a wire to select it and press Del to disconnect it.
Display flag
The viewport shows the display node of the current graph. Double-click a node, select it and press R or Enter, or click Display in the parameter panel. To inspect an upstream step, just move the display flag there and back. A good habit (and what the AI does) is ending each graph with a null named OUT_something that holds the display flag.
Bypass
Press B on a selected node (or click Bypass) to pass its first input through unchanged. Use it to compare with and without a step, or to switch off something expensive (erosion, sculpt_detail) while you work.
Select, move, delete
- Click a node to select it; its parameters appear on the right. Click empty space to deselect.
- Drag a node to move it. Hold Ctrl while dragging to snap to a 10-unit grid.
- Del (or Backspace) deletes the selected node; nodes that were fed by it are reconnected to its first input, so the chain stays intact. Undo with Ctrl+Z.
Layout and navigation
- L or Layout: automatic layout of the whole graph.
- H, F (over the graph) or Home: frame the whole network.
- Drag empty space (left, middle or right button) to pan; the mouse wheel zooms.
Names and comments
Rename a node by editing the name field at the top of its parameter panel. Write a Comment at the bottom of the panel; it is shown under the node. Meaningful names (ground_noise, tower_body, OUT_city) make graphs readable for you and for the AI, and expressions refer to nodes by name.
Errors and warnings
A node that fails is marked in the graph; its parameter panel shows the error in red (warnings in yellow). Downstream nodes cannot cook until it is fixed. The Log tab collects all messages.
Subnets and loops (reusable assets)
A project can hold several graphs. Any graph can be used as a reusable asset:
- Inside the asset graph, start with
subnet_inputnodes and end with anoutputnode. subnetruns that graph as a single node; its inputs feed thesubnet_inputnodes.for_eachruns the graph once per piece of its input (pieces are the prim attributename).object_mergepulls the output of a node in another graph;graph_refbrings in another graph’s output with a transform (see Projects & graphs).
The Building + subnet example uses a building generator as an asset over many lot footprints with for_each.
Categories
Nodes are grouped by category in the node menu: