PDF extraction gives different results in offline mode

[lekslekha]

Working setup : Python PDF extraction, with models downloaded offline & path set via artifacts_path. Works perfectly & extracts 35 sections from PDF document.
Same setup deploy on Azure function app (it cant connect to public network), works but extracts only one section from PDF.
Why ?

Code :

ap = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(file))), "docling_models")
pipeline_options = PdfPipelineOptions(artifacts_path = ap)

def extract_text_from_bytes(file_bytes: bytes, filename: str = "document") -> str:
"""
Extract text from byte object using Docling.

Args:
    file_bytes: Byte content of the document
    filename: Optional filename for context (helps with format detection)

Returns:
    Extracted text in markdown format
"""
file_like_object = io.BytesIO(file_bytes)

source = DocumentStream(name = filename, stream=file_like_object)
converter = DocumentConverter(format_options={InputFormat.PDF: PdfFormatOption(pipeline_options=pipeline_options)})
result = converter.convert(source)
md = result.document.export_to_markdown()
#logger.info(f"Text extracted from {filename}: {md}")
return md