JavaScript imposition - using LLMs (AI) to write imposition scripts

As JavaScript imposition configurations use JavaScript, and AI is quite good at writing JavaScript, isn't it possible to do away with manually writing imposition configurations, and use a good AI engine instead?

Yes, and no.

The world of AI moves quite quickly, so it is entirely possible that the advice in this help article is outdated by the time you read it. Use this article as guidelines, but think how you can take advantage of newer AI features (or ask your AI).

This article was written in June 2026.

JavaScript versus pdfToolbox JavaScript

pdfToolbox uses standard JavaScript language in its imposition configuration, but it extends the JavaScript object model with new objects and convenience methods. This is an important point if you want to use an AI engine to help you, because any answers that work for "generic" JavaScript, are likely not to work for pdfToolbox imposition JavaScript.

Your AI engine needs to at least understand the custom objects and methods added by pdfToolbox, and the terminology used in the documentation and JavaScript.

It is your responsibility to teach your AI engine about these custom extensions.

Hallucinations

The biggest problem here is not that you need to get your AI to write something for you. It will happily do so while telling you that you have great ideas. The biggest problem is stopping your AI engine from hallucinating, and inventing something just to make you happy.

Providing a proper framework to stop this is the most critical part of making this work.

Projects

The concept of projects as embedded (at the time of writing) in ChatGPT and Claude, is quite essential to making this work. A project allows you to setup a version of the truth that is continuously checked as content is being generated. Through experimentation, it became clear that a project needs to have:

  • The relevant parts of the pdfToolbox documentation file,
  • Guidance files with rules, canonical patters the AI should favor, style guide descriptions for the type of JavaScript you want to generate and so on,
  • Examples of good imposition code for the AI to use as further reference materials.

The download at the end of this article contains all of this material, ready to drop into a new project, or (preferably) update to match your requirements and preferences before using it with your AI engine.

Prompting

The second important part is how prompts are used to make the AI do the correct thing. In the downloadable material, two prompts are included:

  • A project prompt: this is meant to prime the project. Setting up the correct expectations for the AI engine, and (very importantly) figure out whether the AI still has unresolved questions before working on something concrete.
  • A chat prompt: this is meant as the initial prompt for any concrete project where you want the AI to write imposition script code.

Downloadable material

Download and unzip the following archive:

attachment AI Project material.zip

Follow this procedure with the contents:

  • In a newly created AI project, add the files from "Framework", "Sources" and "Examples" to the files section of the project. It may take a few seconds for the AI engine to process all files.
  • Start a new chat and copy the contents of the "Project prompt" file into the chat. Look at the questions your AI engine comes up with and answer them. It is worthwhile to repeat "Do you have any more questions right now" before moving on to a concrete topic.
  • Once all questions are settled, start a new chat (in the same project obviously) and complete it with your actual question.

Lean back and relax while the wonders of AI come up with an answer for your problem!