feat: Images?

[totalitarian]

Will it be possible to extract images like biff does my drawing a highlight around the image?

[sabidib]

This could be possible with a biff specific extractor!

There's a guide here on how to write an extractor.

The extractor would import biff and then run it on the downloaded documents.

As for the output, right now remarking focuses on text based highlights. There may be room for image based highlights.

There are a few options for that:

• Stick with text based highlights and output a gcs path/filesystem path for the images extracted from biff.
• Add a ImageHighlight model and have that as a possible return value from extractors. The imageHighlight model would have a field indicating the location of the image.
• Creating a whole new kind of extractor: ImageHighlightExtractor which is similar to the HighlightExtractorbut with some image specific outputs... This would be the cleanest, but adds a bit too much abstraction.

The first 2 options might be interesting to see in a possible solution, which could be refactored into the third option if we have more image based extractors suggested.

I'm not planning on working on that immediately, but will gladly advise on any PRs.