Michele Dolfi

In Docling this would not be an export mode but a choice of which reading order model/module to use. At the moment there is only one reading order model but...

Looking at the input PDF I think what you see is Docling setting all content before the first header as furniture. Since furnitures as excluded (by default) from the markdown...

```py from docling_core.types.doc import ContentLayer document.export_to_markdown(included_content_layers={ContentLayer.BODY, ContentLayer.FURNITURE}) ```

It turned out all the anonymous invites cannot be made without expiration, so the current flow has to make an API call to Discord to get the current invitation link...

@b-g-d The problem with your proposed solution is that Docling should be able to load the models from an artifacts directory which is not the HF cache dir.

Please note that proper GPU support is not yet released, but it is coming very soon. To understand your timings, let us propose another experiment. Please add the following in...

For 2 I guess you are referring to this book? Can you share the extract, or provide the exact page range you tested?

You might be interested in trying out Data-Prep-Kit which will orchestrate a distributed batch conversion using Docling, and it has already support for s3 storage: https://ds4sd.github.io/docling/integrations/data_prep_kit/ Connectors directly in Docling...