Update PDF to Markdown converter API with NVIDIA L4 support

pdf_converter/convert_pdf_to_md.py

@@ -1,6 +1,8 @@
 import marker
 import os
 import sys
+from marker.config.parser import ConfigParser
+from marker.models import create_model_dict
 
 def convert_pdf(pdf_input_path, output_md_path=None):
     """
@@ -20,11 +22,27 @@ def convert_pdf(pdf_input_path, output_md_path=None):
     print(f"Starting conversion of '{pdf_input_path}'...")
 
     try:
-        # Use marker.Marker() for newer versions of marker-pdf
-        md_converter = marker.Marker()
+        # Create configuration
+        config_parser = ConfigParser({})
+        
+        # Load models
+        models = create_model_dict()
+        
+        # Get converter class and create converter
+        converter_cls = config_parser.get_converter_cls()
+        converter = converter_cls(
+            config=config_parser.generate_config_dict(),
+            artifact_dict=models,
+            processor_list=config_parser.get_processors(),
+            renderer=config_parser.get_renderer(),
+            llm_service=config_parser.get_llm_service()
+        )
         
         # Convert the PDF to markdown using marker
-        markdown_text = md_converter.convert_file(pdf_input_path)
+        result = converter(pdf_input_path)
+        
+        # The converter returns a dictionary with the markdown content
+        markdown_text = result.get('markdown', '')
         
         # If output path is provided, save the markdown
         if output_md_path:
@@ -40,4 +58,5 @@ def convert_pdf(pdf_input_path, output_md_path=None):
 
     except Exception as e:
         print(f"An error occurred during conversion: {e}", file=sys.stderr)
+        print(f"Error details: {str(type(e))}", file=sys.stderr)
         raise