DoclingDocument to PDF Conversion

[JTZ18]

Description

Docling’s ability to convert PDFs into structured formats like Markdown and JSON is incredibly useful! I’m curious if there’s an existing or planned feature for the reverse process: exporting a DoclingDocument back to PDF.

Use Case

For a document translation pipeline, it would be helpful to:

1.	Import a PDF (e.g., in Chinese) into a DoclingDocument.
2.	Extract and process text content with an LLM for translation (e.g., Chinese to English).
3.	Replace the original text in DoclingDocument with the translated content.
4.	Export the updated DoclingDocument back to PDF to complete the translated document.

Questions

•	Is such a feature currently supported or on the roadmap?
•	Do you foresee any challenges in implementing a DoclingDocument-to-PDF export feature?

Thanks for considering this question, any insights would be appreciated

[PeterStaar-IBM]

@JTZ18 I think this is indeed something that will come as an add-on. We need to carefully design it, but it might be possible.

[mailtoshwetha09]

@JTZ18 I am doing similar use case where i am importing document in japanese. Hierarchical chunker is extracting as single text as below instead of word. Can you please let me know how did manage to append as word and call LLM.
Cell(id=37, text='金', bbox=BoundingBox(l=433.7615661621094, t=202.373291015625, r=441.44830322265625, b=209.35009765625, coord_origin=<CoordOrigin.TOPLEFT: 'TOPLEFT'>)), Cell(id=38, text='純', bbox=BoundingBox(l=298.4635314941406, t=311.658203125, r=441.2472229003906, b=319.3939208984375, coord_origin=<CoordOrigin.TOPLEFT: 'TOPLEFT'>)), Cell(id=39, text='資 ', bbox=BoundingBox(l=298.56146240234375, t=325.3382568359375, r=305.8238525390625, b=332.8536071777344, coord_origin=<CoordOrigin.TOPLEFT: 'TOPLEFT'>)), Cell(id=40, text='本 ', bbox=BoundingBox(l=366.1123352050781, t=325.55859375, r=373.8316955566406, b=333.0412902832031, coord_origin=<CoordOrigin.TOPLEFT: 'TOPLEFT'>)), Cell(id=41, text='金', bbox=BoundingBox(l=433.7615661621094, t=325.4932861328125, r=441.44830322265625, b=332.4700927734375, coord_origin=<CoordOrigin.TOPLEFT: 'TOPLEFT'>))