WIP feat: accept binary PDF instead of just path to PDF

While the library expects a filepath to be given to the
PdfConverter, the underlying pdfium.PdfDocument constructor
supports bytes and other types. Adding 'bytes' as an accepted
type for the Document Pydantic model allows binary PDFs to be
parsed by the library.

marker/processors/debug.py

@@ -49,7 +49,8 @@ class DebugProcessor(BaseProcessor):
 
     def __call__(self, document: Document):
         # Remove extension from doc name
-        doc_base = os.path.basename(document.filepath).rsplit(".", 1)[0]
+        filepath = document.filepath if isinstance(document.filepath, str) else ""
+        doc_base = os.path.basename(filepath).rsplit(".", 1)[0]
         self.debug_folder = os.path.join(self.debug_data_folder, doc_base)
         if any([self.debug_layout_images, self.debug_pdf_images, self.debug_json]):
             os.makedirs(self.debug_folder, exist_ok=True)

marker/providers/__init__.py

@@ -19,7 +19,7 @@ def raw_text(self):
 ProviderPageLines = Dict[int, List[ProviderOutput]]
 
 class BaseProvider:
-    def __init__(self, filepath: str, config: Optional[BaseModel | dict] = None):
+    def __init__(self, filepath: str | bytes, config: Optional[BaseModel | dict] = None):
         assign_config(self, config)
         self.filepath = filepath
 

marker/providers/pdf.py

@@ -26,8 +26,8 @@ class PdfProvider(BaseProvider):
     ocr_newline_threshold: float = .6
     ocr_alphanum_threshold: float = .3
 
-    def __init__(self, filepath: str, config=None):
-        super().__init__(filepath, config)
+    def __init__(self, file: str | bytes, config=None):
+        super().__init__(file, config)
 
         self.doc: pdfium.PdfDocument = pdfium.PdfDocument(self.filepath)
         self.page_lines: ProviderPageLines = {i: [] for i in range(len(self.doc))}

marker/schema/document.py

@@ -23,7 +23,7 @@ class TocItem(BaseModel):
 
 
 class Document(BaseModel):
-    filepath: str
+    filepath: str | bytes
     pages: List[PageGroup]
     block_type: BlockTypes = BlockTypes.Document
     table_of_contents: List[TocItem] | None = None