Update app.py

app.py

@@ -1,85 +1,76 @@
 import gradio as gr
 from paddleocr import PaddleOCR
 import os
+import tempfile
+from pdf2image import convert_from_path
 
-# --- Configuration ---
-# Initialize PaddleOCR globally for efficiency. 
-# Setting 'use_angle_cls=False' and 'use_text_cls=False' speeds up inference,
-# making it better suited for the default CPU tier on Hugging Face Spaces.
-# We are using the English model (en) for general document parsing.
-# PaddleOCR natively supports taking a PDF file path as input using pymupdf, 
-# which it installs as a dependency.
-
-# Note on 'use_gpu': Hugging Face Spaces typically default to CPU. 
-# If you deploy on a paid GPU Space, set this to True.
+# --- Global Initialization (Fixes Applied Here) ---
+# 1. Deprecation fix: 'use_angle_cls' changed to 'use_textline_orientation'
+# 2. Initialization error fix: Removed 'use_gpu=False' as it caused an 'Unknown argument' error.
 try:
-    ocr = PaddleOCR(lang='en', use_angle_cls=False, use_gpu=False)
+    print("Initializing PaddleOCR (lang='en', use_textline_orientation=False)...")
+    ocr = PaddleOCR(lang='en', use_textline_orientation=False)
 except Exception as e:
-    # Fallback initialization in case of deployment issues
-    print(f"Error initializing PaddleOCR: {e}. Attempting default initialization.")
-    ocr = PaddleOCR()
+    print(f"Error initializing PaddleOCR: {e}. Falling back to default initialization.")
+    ocr = PaddleOCR(lang='en')
 
 
-def process_pdf_for_ocr(pdf_file_path):
+def process_pdf_for_ocr(pdf_file_obj):
     """
-    Takes a PDF file path, runs PaddleOCR on it, and returns the extracted text.
-    Only the first page is processed for demonstration and speed on a free tier.
+    Processes the uploaded PDF file object to perform OCR and extract text.
+    The Gradio file object provides the file path automatically.
     """
-    if pdf_file_path is None:
-        return "Please upload a PDF file to analyze."
+    if pdf_file_obj is None:
+        return "Please upload a PDF file."
 
+    pdf_file_path = pdf_file_obj.name
     print(f"Processing PDF: {pdf_file_path}")
-    
-    # Run OCR inference. PaddleOCR intelligently handles PDF input.
-    # It will extract text from the first page by default.
-    result = ocr.ocr(pdf_file_path, cls=False, det=True, rec=True)
 
-    # --- Post-processing: Format the results into clean text ---
-    
-    extracted_text = []
-    
-    # PaddleOCR result format: list of pages -> list of detection results
-    # Each detection result is: [bounding_box, (text, confidence)]
-    
-    # Check if result is not None and has content
-    if result and result[0] is not None:
-        # Assuming single-page processing for simplicity. `result[0]` is the first page.
-        for line in result[0]:
-            # The text is the first element of the tuple inside the list (line[1][0])
-            text = line[1][0]
-            extracted_text.append(text)
-            
-    if not extracted_text:
-        return "OCR analysis complete, but no readable text was detected on the first page."
-    
-    # Join all detected lines into a single, clean block of text
-    return "\n".join(extracted_text)
+    try:
+        # --- Function Call Fix Applied Here (Former Line 35) ---
+        # The 'cls=False' argument was removed because it is no longer supported
+        # and caused the TypeError in PaddleOCR's internal 'predict' function.
+        result = ocr.ocr(pdf_file_path, det=True, rec=True)
+        
+        extracted_text_lines = []
+        
+        # Parse the OCR result
+        for page_result in result:
+            if page_result is None:
+                continue
+            for line in page_result:
+                # line format: [bounding_box, (text, confidence)]
+                if isinstance(line, list) and len(line) == 2 and isinstance(line[1], tuple):
+                    extracted_text_lines.append(line[1][0])
 
+        if not extracted_text_lines:
+            return "OCR completed, but no text was found in the document."
 
-# --- Gradio Interface Setup ---
+        return "\n".join(extracted_text_lines)
 
-# Create a custom Gradio component for PDF upload
-pdf_input = gr.File(
-    label="Upload PDF Document (Only the first page is processed)",
-    file_types=[".pdf"],
-)
+    except Exception as e:
+        # Use a custom message box instead of alert/confirm
+        error_message = f"An error occurred during OCR processing: {e}"
+        print(error_message)
+        return error_message
 
-# Create a text output box
-text_output = gr.Textbox(
-    label="Extracted Text Results",
-    lines=20,
-    placeholder="The text extracted from the PDF will appear here."
-)
+# --- Gradio Interface Setup ---
 
-# Define the Gradio Interface
 iface = gr.Interface(
     fn=process_pdf_for_ocr,
-    inputs=pdf_input,
-    outputs=text_output,
-    title="PDF OCR Parser using PaddleOCR",
-    description="Upload a PDF file, and this app will use the powerful PaddleOCR system (PP-OCRv3) to extract the text from the document (first page only for quick demo)."
+    inputs=gr.File(
+        label="Upload PDF for Multi-Page OCR", 
+        file_types=['.pdf'], 
+        type="filepath" # Ensure we get the path for PaddleOCR
+    ),
+    outputs=gr.Textbox(
+        label="Extracted Text Content", 
+        placeholder="Upload a PDF and click Submit to see the OCR results here."
+    ),
+    title="Fixed PaddleOCR PDF Processor",
+    description="This application uses PaddleOCR to extract text from a multi-page PDF file. The previously encountered errors regarding unknown arguments and unsupported keywords have been fixed."
 )
 
-# Launch the app
 if __name__ == "__main__":
-    iface.launch()
+    # Launch configuration based on your original logs
+    iface.launch(server_name="0.0.0.0", server_port=7860)