feat-request: Document Schema filepath property data type

[eddiedozier]

Hi, I'm currently facing a pydantic error, https://errors.pydantic.dev/2.11/v/string_unicode, when I pass bytes to the TableConverter filepath argument. After tracing the source of the error to /schema/document.py , I was able to resolve it by converting the filepath type of str to Union[str | bytes].

class Document(BaseModel):
    filepath: Union[str | bytes]

The use case for this change is utilizing Marker in a GPU serverless function that will take a PDF file upload from users that will go downstream to the serverless function where Marker will process the file.

[ZeyBal]

I ran into the same issue, but the change you proposed did not solved it.

I find it strange that PDFConverter accepts bytes, while TableConverter does not.
Is there a specific reason why the __call__ function in PDFConverter accepts bytes, but the one in TableConverter doesn’t?

# converters/pdf.py
def __call__(self, filepath: str | io.BytesIO):
    with self.filepath_to_str(filepath) as temp_path:  # filepath_to_str creates a temporary file if filepath is in bytes
        document = self.build_document(temp_path)
        self.page_count = len(document.pages)
        renderer = self.resolve_dependencies(self.renderer)
        rendered = renderer(document)
    return rendered

# converters/table.py
def __call__(self, filepath: str):
    document = self.build_document(filepath)
    self.page_count = len(document.pages)
    renderer = self.resolve_dependencies(self.renderer)
    return renderer(document)