Update app.py

app.py

@@ -15,17 +15,16 @@ app = FastAPI(title="Free MarkItDown Service")
 md = MarkItDown()
 
 IMAGE_EXTENSIONS = {".png", ".jpg", ".jpeg", ".bmp", ".tiff", ".webp"}
+LEGACY_OFFICE_EXTENSIONS = {".doc", ".xls", ".ppt", ".rtf"}
 
 def run_free_ocr(file_path: str) -> str:
-    """Fallback OCR pipeline using Tesseract (100% free)."""
+    """Fallback OCR pipeline using Tesseract."""
     file_ext = os.path.splitext(file_path)[1].lower()
     
-    # Image OCR
     if file_ext in IMAGE_EXTENSIONS:
         img = Image.open(file_path)
         return pytesseract.image_to_string(img)
     
-    # PDF OCR (page by page)
     elif file_ext == ".pdf":
         pages = convert_from_path(file_path)
         extracted_text = []
@@ -37,14 +36,26 @@ def run_free_ocr(file_path: str) -> str:
     
     return ""
 
+def convert_legacy_office_file(file_path: str, ext: str) -> str:
+    """Converts legacy .doc, .xls, .ppt to modern XML format via LibreOffice."""
+    out_dir = "/tmp"
+    target_ext = "docx" if ext == ".doc" else "xlsx" if ext == ".xls" else "pptx"
+    
+    subprocess.run(
+        ["libreoffice", "--headless", "--convert-to", target_ext, file_path, "--outdir", out_dir],
+        check=True
+    )
+    
+    base_name = os.path.splitext(os.path.basename(file_path))[0]
+    return os.path.join(out_dir, f"{base_name}.{target_ext}")
+
 def process_input(source_path_or_url: str, original_filename: str = None) -> str:
-    """Unified document converter with explicit fallback handling."""
     if not source_path_or_url or not source_path_or_url.strip():
         return "Please supply a valid file or URL."
 
     target = source_path_or_url.strip()
 
-    # If it's a URL, let MarkItDown fetch and parse it directly
+    # If HTTP/HTTPS URL
     if target.startswith("http://") or target.startswith("https://"):
         try:
             res = md.convert(target)
@@ -52,51 +63,41 @@ def process_input(source_path_or_url: str, original_filename: str = None) -> str
         except Exception as e:
             return f"URL Conversion Error: {str(e)}"
 
-    # Identify true file extension
-    ext = ""
-    if original_filename:
-        ext = os.path.splitext(original_filename)[1].lower()
-    else:
-        ext = os.path.splitext(target)[1].lower()
-
-    # Ensure the file path on disk has the correct file extension
+    # 1. Detect extension from the original filename (e.g., 'document.doc')
+    ext = os.path.splitext(original_filename or target)[1].lower()
+    
+    # 2. FORCE the temp file on disk to have the real extension so MarkItDown knows what it is!
     working_path = target
-    temp_renamed = None
+    temp_with_ext = None
     if ext and not target.lower().endswith(ext):
-        temp_renamed = f"{target}{ext}"
-        shutil.copyfile(target, temp_renamed)
-        working_path = temp_renamed
+        temp_with_ext = f"{target}{ext}"
+        shutil.copyfile(target, temp_with_ext)
+        working_path = temp_with_ext
 
-    converted_docx = None
+    temp_converted_office = None
 
     try:
-        # Route 1: Directly handle pure image files using local Tesseract OCR
+        # Route Images directly to Tesseract
         if ext in IMAGE_EXTENSIONS:
             return run_free_ocr(working_path)
 
-        # Route 2: Convert legacy binary .doc files using LibreOffice
-        if ext == ".doc":
-            out_dir = "/tmp"
-            subprocess.run(
-                ["libreoffice", "--headless", "--convert-to", "docx", working_path, "--outdir", out_dir],
-                check=True
-            )
-            converted_docx = os.path.join(out_dir, os.path.splitext(os.path.basename(working_path))[0] + ".docx")
-            working_path = converted_docx
-
-        # Route 3: Standard MarkItDown extraction
+        # Route Legacy Office files (.doc) to LibreOffice first
+        if ext in LEGACY_OFFICE_EXTENSIONS:
+            temp_converted_office = convert_legacy_office_file(working_path, ext)
+            working_path = temp_converted_office
+
+        # Standard MarkItDown extraction
         result = md.convert(working_path)
         text_content = result.text_content if result else ""
 
-        # Fallback to Tesseract OCR if MarkItDown extracts no text
-        if not text_content or not text_content.strip():
+        # Fallback to Tesseract if extraction yielded blank text
+        if not text_content.strip():
             text_content = run_free_ocr(working_path)
 
         return text_content if text_content.strip() else "Unable to extract text from document."
 
-    except (FileConversionException, Exception) as e:
-        # If MarkItDown fails to parse the file format, run Tesseract OCR as final safety net
-        print(f"MarkItDown failed on '{working_path}'. Attempting Tesseract OCR fallback...")
+    except Exception as e:
+        print(f"MarkItDown failed on '{working_path}'. Executing Tesseract OCR fallback...")
         try:
             ocr_text = run_free_ocr(working_path)
             if ocr_text and ocr_text.strip():
@@ -106,17 +107,17 @@ def process_input(source_path_or_url: str, original_filename: str = None) -> str
 
         print("--- CONVERSION ERROR ---")
         traceback.print_exc()
-        return f"Conversion Error: {str(e)}"
+        return f"Conversion Error: File type '{ext}' could not be processed."
 
     finally:
-        # Clean up temporary renamed or converted files
-        if temp_renamed and os.path.exists(temp_renamed):
-            os.remove(temp_renamed)
-        if converted_docx and os.path.exists(converted_docx):
-            os.remove(converted_docx)
+        # Cleanup temporary files
+        if temp_with_ext and os.path.exists(temp_with_ext):
+            os.remove(temp_with_ext)
+        if temp_converted_office and os.path.exists(temp_converted_office):
+            os.remove(temp_converted_office)
 
 # -------------------------------------------------------------
-# 1. FastAPI Endpoint (For TypeScript backend calls)
+# FastAPI Endpoint
 # -------------------------------------------------------------
 @app.post("/convert")
 async def convert_file_api(file: UploadFile = None, url: str = None):
@@ -144,11 +145,11 @@ async def convert_file_api(file: UploadFile = None, url: str = None):
             os.remove(temp_path)
 
 # -------------------------------------------------------------
-# 2. Gradio Web Interface (For browser uploads)
+# Gradio Handler
 # -------------------------------------------------------------
 def gradio_handler(file_obj, url_input):
     if file_obj is not None:
-        # Pass file_obj.name (disk path) along with original filename metadata
+        # Extract the real filename (e.g. "my_doc.pdf" instead of "/tmp/gradio/xyz123")
         orig_name = getattr(file_obj, "orig_name", file_obj.name)
         return process_input(file_obj.name, original_filename=orig_name)
     elif url_input and url_input.strip():
@@ -157,7 +158,6 @@ def gradio_handler(file_obj, url_input):
 
 with gr.Blocks(title="Free MarkItDown Tester") as demo:
     gr.Markdown("# 📄 Free Document & URL Converter")
-    gr.Markdown("Upload a document (`.pdf`, `.doc`, `.docx`, `.xlsx`, `.pptx`, `.png`, `.jpg`) or enter a public URL.")
     
     with gr.Row():
         with gr.Column():