Skip to main content

Interactive Figure Design

A plugin that helps an AI assistant make clear explanatory figures, then faithful interactive versions

Author

Vagish Hemmige

Written for version 0.1.5.

An early version. It comes from the maintainer’s own figures, such as the TransplantID workflow; it has not yet been used by other people.

What it is

Interactive Figure Design is a plugin for ChatGPT and Claude. A plugin is a set of instructions the assistant reads when a task calls for them, so it works the same way every time and in every chat. This one is for explanatory figures: workflow diagrams, study designs, system overviews and similar pictures that explain how something works.

With the plugin installed, you ask in plain words (“Draw a figure explaining how our weekly literature search works”) and the assistant follows the plugin’s steps.

It’s built for explaining processes, not plotting data: a plot of real numbers is better made by the analysis code that produces it.

What it does differently

Without instructions, an assistant tends to draw first and ask later, which is why explanatory figures often take many rounds. The plugin changes the order:

  • Content before drawing. The assistant writes out, in plain language, what each view of the figure will explain, and you confirm it before anything is drawn.
  • Static before interactive. Each view is first a static SVG you can look at and approve. Interactivity comes only after that, and only if you ask.
  • Checked before shown. The assistant inspects the rendered picture, not just its code, and gets an independent design review for substantial figures when it can.
  • Your style, not a generic one. The default look is the information-rich style of the maintainer’s TransplantID figure: numbered pastel phases, strong headings, meaningful small illustrations and clearly labelled arrows.

How it works shows the six stages.

Where to start

Before you rely on it

The plugin guides an assistant; it does not check your facts. Review what the figure says as you would a colleague’s draft before it goes into a paper, a talk or a website. What’s been tested lists what has and has not been checked.

The source, including every instruction the assistant reads, is on GitHub. Original text and code are MIT licensed; the bundled reference figure is not covered (see the house style).