ααLoop docs
On this page

Rendering graphs from Python

A Linnet graph can carry ordinary Python objects and render directly in a notebook. This guide uses a document-processing workflow to explore the default drawing, layout selection, and styles computed from application data. Begin with the Python quickstart for installation and basic graph queries.

The notebook below loads automatically when it comes into view. Python, Linnet, and the Typst renderer run in your browser through WebAssembly. The first visit downloads their dependencies; no local Python installation or notebook server is needed.

Try the rendering API

Choose a layout and enable the custom line theme or half-edge IDs. Each change rebuilds the example graph and redraws it. The cells are editable, so you can change the workflow or its selectors and run the affected cell with its play button.

This interactive notebook requires JavaScript and this documentation build’s notebook assets. The rendering notebook source is also available to run locally.

What to explore

  • Switch between Force and Stable layered. Both render the same graph; force layout tunes geometric positions, while the layered layout organizes the workflow in its chosen direction.
  • Enable Custom Python line theme. Edge selectors inspect each channel’s protocol to choose its color, label, and pattern. Endpoint selectors place direction marks using source/sink flow.
  • Enable Half-edge IDs to see how the incoming request and outgoing publication differ from internal edges. The notebook reports the node, edge, and half-edge counts below the drawings.
  • Switch the node store and inspect the payload checks. Node, edge, and endpoint dataclass instances remain attached by identity, including their shared runtime object.

The rendering boundary

Returning a Graph from a cell invokes its SVG representation. graph.to_svg() explicitly produces the same kind of drawing; the notebook shows both forms. The graph’s RenderConfig combines layout options, drawing defaults, and Python selectors. Only topology and the selectors’ typed drawing results pass to Typst; application payloads stay in Python.

The optional theme demonstrates that domain conventions belong in those selectors. It does not change the workflow records or require them to be serializable. For the complete types and method signatures, use the Python API reference.

To experiment with a DOT-backed physics example, continue to GammaLoop DOT input. For a continuously updating view of the force solver itself, use the live DOT playground.