Ask HN: Best file format for AI reports output?

[azkalam]

I want to use LLMs to construct reports. These reports will be of varying quality so I need to do a few rounds of QA (LLMs and human reviewers). So I need a format that is good for both feeding into a model and rendering as HTML.

Markdown makes sense but if I add rich media like images it won't be a self contained file.

Markdown is versatile enough to contain tables, images, headings etc. It also has low token overhead.

How can Markdown handle page breaks?

PDF is self contained but I want a tighter format that is quick to pull text and images from.

HTML is interesting, but again I want a restricted feature set - no JavaScript etc.

I am currently thinking about using zips containing Markdown files and image assets.

What are people reaching for in this space?