Unable to parse PDFs: unknown type in init_ws

[wwwslinger]

With docling-parse version 3.0.0, I receive the attached exception when attempting to convert the attached PDF and many others like it. I don't have this error with prior versions. This means I can't use docling 2.10.0+. Is there a workaround?

Here is the code:

from docling.datamodel.base_models import InputFormat
from docling.document_converter import DocumentConverter, PdfFormatOption
from docling.datamodel.pipeline_options import PdfPipelineOptions, TableFormerMode

def get_docling_converter(method='fast'):
    pipeline_options = PdfPipelineOptions(do_table_structure=True, generate_picture_images=True)
    if method == 'accurate':
        pipeline_options.table_structure_options.mode = TableFormerMode.ACCURATE
        
    elif method == 'predicted':
        pipeline_options.table_structure_options.mode = TableFormerMode.ACCURATE
        pipeline_options.table_structure_options.do_cell_matching = False  # uses text cells predicted from table structure model
        
    return DocumentConverter(
            format_options={
                InputFormat.PDF: PdfFormatOption(pipeline_options=pipeline_options)
            }
        )

converter = ct.get_docling_converter('predicted')
result = converter.convert('Prot_001.pdf')

Versions:
Python 3.11.10
docling==2.10.0
docling-core==2.9.0
docling-ibm-models==2.0.8
docling-parse==3.0.0

[cau-git]

Thanks for providing this example, we will look into it. Meanwhile, you can use docling in the latest version if you choose the DoclingParseDocumentBackend (v1), which is still available as it was before. We only changed the default backend.

[PeterStaar-IBM]

Fixed in PR: #101

[wwwslinger]

@PeterStaar-IBM thanks, I'll give it a try!
How do you get the view with the boxed elements? That would also be helpful for my task.

[PeterStaar-IBM]

@wwwslinger If you clone the package and install, you can,

python ./docling_parse/visualize.py -i <pdf-file>  -p 1  -l info -c word --interactive --log-text