natural-pdf 0.1.7__py3-none-any.whl → 0.1.8__py3-none-any.whl

natural_pdf/core/pdf.py

@@ -1,11 +1,13 @@
-import copy  # Add import for deepcopy
+import copy
 import logging
 import os
 import re
 import tempfile
 import urllib.request
-from pathlib import Path  # Added Path
-from typing import (  # Added Iterable and TYPE_CHECKING
+import time
+import threading
+from pathlib import Path
+from typing import (
     TYPE_CHECKING,
     Any,
     Callable,
@@ -17,29 +19,33 @@ from typing import (  # Added Iterable and TYPE_CHECKING
     Type,
     Union,
 )
-from pathlib import Path
-
+from natural_pdf.utils.tqdm_utils import get_tqdm
 
 import pdfplumber
 from PIL import Image
 
-from natural_pdf.analyzers.layout.layout_manager import (  # Import the new LayoutManager
-    LayoutManager,
-)
-from natural_pdf.core.highlighting_service import HighlightingService  # <-- Import the new service
+from natural_pdf.analyzers.layout.layout_manager import LayoutManager
+from natural_pdf.core.highlighting_service import HighlightingService
 from natural_pdf.core.page import Page
 from natural_pdf.elements.collections import ElementCollection
 from natural_pdf.elements.region import Region
 from natural_pdf.ocr import OCRManager, OCROptions
 from natural_pdf.selectors.parser import parse_selector
 
-# Import the flag directly - this should always work
+from natural_pdf.classification.manager import ClassificationManager
+from natural_pdf.classification.manager import ClassificationError
+from natural_pdf.classification.results import ClassificationResult
+from natural_pdf.extraction.manager import StructuredDataManager
+
+from natural_pdf.utils.locks import pdf_render_lock
+from natural_pdf.elements.base import Element
+from natural_pdf.classification.mixin import ClassificationMixin
+from natural_pdf.extraction.mixin import ExtractionMixin
 
-# --- Add Search Service Imports (needed for new methods) ---
 try:
-    from typing import Any as TypingAny  # Import Any if not already
+    from typing import Any as TypingAny
 
-    from natural_pdf.search import TextSearchOptions  # Keep for ask default
+    from natural_pdf.search import TextSearchOptions
     from natural_pdf.search import (
         BaseSearchOptions,
         SearchOptions,
@@ -47,25 +53,24 @@ try:
         get_search_service,
     )
 except ImportError:
-    # Define dummies if needed for type hints within the class
     SearchServiceProtocol = object
     SearchOptions, TextSearchOptions, BaseSearchOptions = object, object, object
     TypingAny = object
 
-    # Dummy factory needed for default arg in methods
     def get_search_service(**kwargs) -> SearchServiceProtocol:
         raise ImportError(
             "Search dependencies are not installed. Install with: pip install natural-pdf[search]"
         )
 
-
-# --- End Search Service Imports ---
-
-# Set up logger early
 logger = logging.getLogger("natural_pdf.core.pdf")
+tqdm = get_tqdm()
 
+DEFAULT_MANAGERS = {
+    "classification": ClassificationManager,
+    "structured_data": StructuredDataManager,
+}
 
-class PDF:
+class PDF(ExtractionMixin):
     """
     Enhanced PDF wrapper built on top of pdfplumber.
 
@@ -86,35 +91,23 @@ class PDF:
         Args:
             path_or_url: Path to the PDF file or a URL to a PDF
             reading_order: Whether to use natural reading order
-            font_attrs: Font attributes to consider when grouping characters into words.
-                       Default: ['fontname', 'size'] (Group by font name and size)
-                       None: Only consider spatial relationships
-                       List: Custom attributes to consider (e.g., ['fontname', 'size', 'color'])
-            keep_spaces: Whether to include spaces in word elements (default: True).
-                       True: Spaces are part of words, better for multi-word searching
-                       False: Break text at spaces, each word is separate (legacy behavior)
+            font_attrs: Font attributes for grouping characters into words
+            keep_spaces: Whether to include spaces in word elements
         """
-        # Check if the input is a URL
         is_url = path_or_url.startswith("http://") or path_or_url.startswith("https://")
 
-        # Initialize path-related attributes
         self._original_path = path_or_url
         self._temp_file = None
-        self._resolved_path = None  # Store the actual path used by pdfplumber
+        self._resolved_path = None
 
         if is_url:
             logger.info(f"Downloading PDF from URL: {path_or_url}")
             try:
-                # Create a temporary file to store the downloaded PDF
                 self._temp_file = tempfile.NamedTemporaryFile(suffix=".pdf", delete=False)
-
-                # Download the PDF
                 with urllib.request.urlopen(path_or_url) as response:
                     self._temp_file.write(response.read())
                     self._temp_file.flush()
                     self._temp_file.close()
-
-                # Use the temporary file path
                 self._resolved_path = self._temp_file.name
                 logger.info(f"PDF downloaded to temporary file: {self._resolved_path}")
             except Exception as e:
@@ -126,7 +119,6 @@ class PDF:
                 logger.error(f"Failed to download PDF from URL: {e}")
                 raise ValueError(f"Failed to download PDF from URL: {e}")
         else:
-            # Use the provided path directly
             self._resolved_path = path_or_url
 
         logger.info(f"Initializing PDF from {self._resolved_path}")
@@ -137,42 +129,68 @@ class PDF:
         try:
             self._pdf = pdfplumber.open(self._resolved_path)
         except Exception as e:
-            logger.error(
-                f"Failed to open PDF with pdfplumber: {self._resolved_path}. Error: {e}",
-                exc_info=True,
-            )
-            # Clean up temp file if creation failed
+            logger.error(f"Failed to open PDF: {e}", exc_info=True)
             self.close()
             raise IOError(f"Failed to open PDF file/URL: {path_or_url}") from e
 
-        self._path = self._resolved_path  # Keep original path too?
-        self.path = self._resolved_path  # Public attribute for the resolved path
-        self.source_path = self._original_path  # Public attribute for the user-provided path/URL
+        self._path = self._resolved_path
+        self.path = self._resolved_path
+        self.source_path = self._original_path
 
         self._reading_order = reading_order
         self._config = {"keep_spaces": keep_spaces}
+        self._font_attrs = font_attrs
 
-        self._font_attrs = font_attrs  # Store the font attribute configuration
-
-        # Initialize Managers and Services (conditionally available)
         self._ocr_manager = OCRManager() if OCRManager else None
         self._layout_manager = LayoutManager() if LayoutManager else None
         self.highlighter = HighlightingService(self)
+        self._classification_manager_instance = ClassificationManager()
+        self._manager_registry = {}
 
-        # Initialize pages last, passing necessary refs
         self._pages = [
             Page(p, parent=self, index=i, font_attrs=font_attrs)
             for i, p in enumerate(self._pdf.pages)
         ]
 
-        # Other state
         self._element_cache = {}
-        self._exclusions = []  # List to store exclusion functions/regions
-        self._regions = []  # List to store region functions/definitions
+        self._exclusions = []
+        self._regions = []
 
-        logger.info("Initialized HighlightingService.")
         logger.info(f"PDF '{self.source_path}' initialized with {len(self._pages)} pages.")
 
+        self._initialize_managers()
+        self._initialize_highlighter()
+
+    def _initialize_managers(self):
+        """Initialize manager instances based on DEFAULT_MANAGERS."""
+        self._managers = {}
+        for key, manager_class in DEFAULT_MANAGERS.items():
+            try:
+                self._managers[key] = manager_class()
+                logger.debug(f"Initialized manager for key '{key}': {manager_class.__name__}")
+            except Exception as e:
+                logger.error(f"Failed to initialize manager {manager_class.__name__}: {e}")
+                self._managers[key] = None
+
+    def get_manager(self, key: str) -> Any:
+        """Retrieve a manager instance by its key."""
+        if key not in self._managers:
+            raise KeyError(f"No manager registered for key '{key}'. Available: {list(self._managers.keys())}")
+        
+        manager_instance = self._managers.get(key)
+        
+        if manager_instance is None:
+             manager_class = DEFAULT_MANAGERS.get(key)
+             if manager_class:
+                  raise RuntimeError(f"Manager '{key}' ({manager_class.__name__}) failed to initialize previously.")
+             else:
+                  raise RuntimeError(f"Manager '{key}' failed to initialize (class not found).")
+
+        return manager_instance
+
+    def _initialize_highlighter(self):
+        pass
+
     @property
     def metadata(self) -> Dict[str, Any]:
         """Access metadata as a dictionary."""
@@ -183,7 +201,6 @@ class PDF:
         """Access pages as a PageCollection object."""
         from natural_pdf.elements.collections import PageCollection
 
-        # Ensure _pages is initialized
         if not hasattr(self, "_pages"):
             raise AttributeError("PDF pages not yet initialized.")
         return PageCollection(self._pages)
@@ -195,12 +212,10 @@ class PDF:
         Returns:
             Self for method chaining
         """
-        # Ensure _pages is initialized
         if not hasattr(self, "_pages"):
             raise AttributeError("PDF pages not yet initialized.")
 
         self._exclusions = []
-        # Also clear from pages
         for page in self._pages:
             page.clear_exclusions()
         return self
@@ -212,99 +227,75 @@ class PDF:
         Add an exclusion function to the PDF. Text from these regions will be excluded from extraction.
 
         Args:
-            exclusion_func: A function that takes a Page and returns a Region to exclude, or None.
+            exclusion_func: A function that takes a Page and returns a Region to exclude, or None
             label: Optional label for this exclusion
 
         Returns:
             Self for method chaining
         """
-        # Ensure _pages is initialized
         if not hasattr(self, "_pages"):
             raise AttributeError("PDF pages not yet initialized.")
 
-        # Store exclusion with its label at PDF level
         exclusion_data = (exclusion_func, label)
         self._exclusions.append(exclusion_data)
 
-        # Apply this exclusion to all pages
         for page in self._pages:
-            # We pass the original function, Page.add_exclusion handles calling it
             page.add_exclusion(exclusion_func, label=label)
 
         return self
 
     def apply_ocr(
         self,
-        pages: Optional[Union[Iterable[int], range, slice]] = None,
         engine: Optional[str] = None,
-        # --- Common OCR Parameters (Direct Arguments) ---
         languages: Optional[List[str]] = None,
-        min_confidence: Optional[float] = None,  # Min confidence threshold
+        min_confidence: Optional[float] = None,
         device: Optional[str] = None,
-        resolution: Optional[int] = None,  # DPI for rendering before OCR
-        apply_exclusions: bool = True,  # New parameter
+        resolution: Optional[int] = None,
+        apply_exclusions: bool = True,
         detect_only: bool = False,
-        # --- Engine-Specific Options --- Use 'options=' for this
-        options: Optional[Any] = None,  # e.g., EasyOCROptions(...), PaddleOCROptions(...), or dict
-        # **kwargs: Optional[Dict[str, Any]] = None # Allow potential extra args?
+        replace: bool = True,
+        options: Optional[Any] = None,
+        pages: Optional[Union[Iterable[int], range, slice]] = None,
     ) -> "PDF":
         """
-        Applies OCR to specified pages (or all pages) of the PDF using batch processing.
-
-        This method renders the specified pages to images, sends them as a batch
-        to the OCRManager, and adds the resulting TextElements to each respective page.
+        Applies OCR to specified pages of the PDF using batch processing.
 
         Args:
-            pages: An iterable of 0-based page indices (list, range, tuple),
-                   a slice object, or None to process all pages.
-            engine: Name of the OCR engine (e.g., 'easyocr', 'paddleocr', 'surya').
-                    Uses manager's default ('easyocr') if None.
-            languages: List of language codes (e.g., ['en', 'fr'], ['en', 'ch_sim']).
-                       **Must be codes understood by the specific selected engine.**
-                       No mapping is performed. Overrides manager/engine default.
-            min_confidence: Minimum confidence threshold for detected text (0.0 to 1.0).
-                            Overrides manager/engine default.
-            device: Device to run OCR on (e.g., 'cpu', 'cuda', 'mps').
-                    Overrides manager/engine default.
-            resolution: DPI resolution to render page images before OCR (e.g., 150, 300).
-                        Affects input quality for OCR. Defaults to 150 if not set.
-            apply_exclusions: If True (default), render page image for OCR with
-                              excluded areas masked (whited out). If False, OCR
-                              the raw page image without masking exclusions.
-            detect_only: If True, only detect text bounding boxes, don't perform OCR.
-            options: An engine-specific options object (e.g., EasyOCROptions) or dict
-                     containing parameters specific to the chosen engine.
+            engine: Name of the OCR engine
+            languages: List of language codes
+            min_confidence: Minimum confidence threshold 
+            device: Device to run OCR on
+            resolution: DPI resolution for page images
+            apply_exclusions: Whether to mask excluded areas
+            detect_only: If True, only detect text boxes
+            replace: Whether to replace existing OCR elements
+            options: Engine-specific options
+            pages: Page indices to process or None for all pages
 
         Returns:
-            Self for method chaining.
-
-        Raises:
-            ValueError: If page indices are invalid.
-            TypeError: If 'options' is not compatible with the engine.
-            RuntimeError: If the OCRManager or selected engine is not available.
+            Self for method chaining
         """
         if not self._ocr_manager:
             logger.error("OCRManager not available. Cannot apply OCR.")
-            # Or raise RuntimeError("OCRManager not initialized.")
             return self
 
-        # --- Determine Target Pages (unchanged) ---
-        target_pages: List[Page] = []
+        thread_id = threading.current_thread().name
+        logger.debug(f"[{thread_id}] PDF.apply_ocr starting for {self.path}")
+        
+        target_pages = []
         if pages is None:
             target_pages = self._pages
         elif isinstance(pages, slice):
             target_pages = self._pages[pages]
-        elif hasattr(pages, "__iter__"):  # Check if it's iterable (list, range, tuple, etc.)
+        elif hasattr(pages, "__iter__"):
             try:
                 target_pages = [self._pages[i] for i in pages]
             except IndexError:
                 raise ValueError("Invalid page index provided in 'pages' iterable.")
             except TypeError:
-                raise TypeError(
-                    "'pages' must be None, a slice, or an iterable of page indices (int)."
-                )
+                raise TypeError("'pages' must be None, a slice, or an iterable of page indices.")
         else:
-            raise TypeError("'pages' must be None, a slice, or an iterable of page indices (int).")
+            raise TypeError("'pages' must be None, a slice, or an iterable of page indices.")
 
         if not target_pages:
             logger.warning("No pages selected for OCR processing.")
@@ -312,26 +303,20 @@ class PDF:
 
         page_numbers = [p.number for p in target_pages]
         logger.info(f"Applying batch OCR to pages: {page_numbers}...")
-        # --- Determine Rendering Resolution ---
-        # Priority: 1. direct `resolution` arg, 2. PDF config, 3. default 150
-        final_resolution = resolution  # Use direct arg if provided
-        if final_resolution is None:
-            final_resolution = getattr(self, "_config", {}).get("resolution", 150)
-
-        logger.debug(f"Using OCR image rendering resolution: {final_resolution} DPI")
-
-        # --- Render Images for Batch ---
-        images_pil: List[Image.Image] = []
-        page_image_map: List[Tuple[Page, Image.Image]] = []  # Store page and its image
-        logger.info(
-            f"Rendering {len(target_pages)} pages to images at {final_resolution} DPI (apply_exclusions={apply_exclusions})..."
-        )
-        failed_page_num = "unknown"  # Keep track of potentially failing page
+        
+        final_resolution = resolution or getattr(self, "_config", {}).get("resolution", 150)
+        logger.debug(f"Using OCR image resolution: {final_resolution} DPI")
+
+        images_pil = []
+        page_image_map = []
+        logger.info(f"[{thread_id}] Rendering {len(target_pages)} pages...")
+        failed_page_num = "unknown"
+        render_start_time = time.monotonic()
+        
         try:
-            for i, page in enumerate(target_pages):
-                failed_page_num = page.number  # Update current page number in case of error
+            for i, page in enumerate(tqdm(target_pages, desc="Rendering pages", leave=False)):
+                failed_page_num = page.number
                 logger.debug(f"  Rendering page {page.number} (index {page.index})...")
-                # Use the determined final_resolution and apply exclusions if requested
                 to_image_kwargs = {
                     "resolution": final_resolution,
                     "include_highlights": False,
@@ -340,68 +325,64 @@ class PDF:
                 img = page.to_image(**to_image_kwargs)
                 if img is None:
                     logger.error(f"  Failed to render page {page.number} to image.")
-                    # Decide how to handle: skip page, raise error? For now, skip.
-                    continue  # Skip this page if rendering failed
+                    continue
                 images_pil.append(img)
-                page_image_map.append((page, img))  # Store pair
+                page_image_map.append((page, img))
         except Exception as e:
-            logger.error(f"Failed to render one or more pages for batch OCR: {e}", exc_info=True)
+            logger.error(f"Failed to render pages for batch OCR: {e}")
             raise RuntimeError(f"Failed to render page {failed_page_num} for OCR.") from e
+            
+        render_end_time = time.monotonic()
+        logger.debug(f"[{thread_id}] Finished rendering {len(images_pil)} images (Duration: {render_end_time - render_start_time:.2f}s)")
 
         if not images_pil or not page_image_map:
             logger.error("No images were successfully rendered for batch OCR.")
             return self
 
-        # --- Prepare Arguments for Manager ---
-        # Pass common args directly, engine-specific via options
         manager_args = {
             "images": images_pil,
             "engine": engine,
             "languages": languages,
-            "min_confidence": min_confidence,  # Use the renamed parameter
+            "min_confidence": min_confidence,
             "device": device,
             "options": options,
             "detect_only": detect_only,
-            # Note: resolution is used for rendering, not passed to OCR manager directly
         }
-        # Filter out None values so manager can use its defaults
         manager_args = {k: v for k, v in manager_args.items() if v is not None}
 
-        # --- Call OCR Manager for Batch Processing ---
-        logger.info(
-            f"Calling OCR Manager with args: { {k:v for k,v in manager_args.items() if k!='images'} } ..."
-        )
+        ocr_call_args = {k:v for k,v in manager_args.items() if k!='images'}
+        logger.info(f"[{thread_id}] Calling OCR Manager with args: {ocr_call_args}...")
+        ocr_start_time = time.monotonic()
+        
         try:
-            # Manager's apply_ocr signature needs to accept common args directly
             batch_results = self._ocr_manager.apply_ocr(**manager_args)
 
             if not isinstance(batch_results, list) or len(batch_results) != len(images_pil):
-                logger.error(
-                    f"OCR Manager returned unexpected result format or length for batch processing. "
-                    f"Expected list of length {len(images_pil)}, got {type(batch_results)} "
-                    f"with length {len(batch_results) if isinstance(batch_results, list) else 'N/A'}."
-                )
+                logger.error(f"OCR Manager returned unexpected result format or length.")
                 return self
 
             logger.info("OCR Manager batch processing complete.")
-
         except Exception as e:
-            logger.error(f"Batch OCR processing failed: {e}", exc_info=True)
+            logger.error(f"Batch OCR processing failed: {e}")
             return self
+            
+        ocr_end_time = time.monotonic()
+        logger.debug(f"[{thread_id}] OCR processing finished (Duration: {ocr_end_time - ocr_start_time:.2f}s)")
 
-        # --- Distribute Results and Add Elements to Pages (unchanged) ---
         logger.info("Adding OCR results to respective pages...")
         total_elements_added = 0
+        
         for i, (page, img) in enumerate(page_image_map):
             results_for_page = batch_results[i]
             if not isinstance(results_for_page, list):
-                logger.warning(
-                    f"Skipping results for page {page.number}: Expected list, got {type(results_for_page)}"
-                )
+                logger.warning(f"Skipping results for page {page.number}: Expected list, got {type(results_for_page)}")
                 continue
 
             logger.debug(f"  Processing {len(results_for_page)} results for page {page.number}...")
             try:
+                if manager_args.get("replace", True) and hasattr(page, "_element_mgr"):
+                    page._element_mgr.remove_ocr_elements()
+                
                 img_scale_x = page.width / img.width if img.width > 0 else 1
                 img_scale_y = page.height / img.height if img.height > 0 else 1
                 elements = page._element_mgr.create_text_elements_from_ocr(
@@ -414,53 +395,39 @@ class PDF:
                 else:
                     logger.debug(f"  No valid TextElements created for page {page.number}.")
             except Exception as e:
-                logger.error(
-                    f"  Error adding OCR elements to page {page.number}: {e}", exc_info=True
-                )
+                logger.error(f"  Error adding OCR elements to page {page.number}: {e}")
 
-        logger.info(
-            f"Finished adding OCR results. Total elements added across {len(target_pages)} pages: {total_elements_added}"
-        )
+        logger.info(f"Finished adding OCR results. Total elements added: {total_elements_added}")
         return self
 
     def add_region(
         self, region_func: Callable[["Page"], Optional[Region]], name: str = None
     ) -> "PDF":
         """
-        Add a region function to the PDF. This creates regions on all pages using the provided function.
+        Add a region function to the PDF.
 
         Args:
-            region_func: A function that takes a Page and returns a Region, or None.
+            region_func: A function that takes a Page and returns a Region, or None
             name: Optional name for the region
 
         Returns:
             Self for method chaining
         """
-        # Ensure _pages is initialized
         if not hasattr(self, "_pages"):
             raise AttributeError("PDF pages not yet initialized.")
 
-        # Store region with its name at PDF level
         region_data = (region_func, name)
         self._regions.append(region_data)
 
-        # Apply this region to all pages
         for page in self._pages:
             try:
-                # Call the function to get the region for this specific page
                 region_instance = region_func(page)
                 if region_instance and isinstance(region_instance, Region):
-                    # If a valid region is returned, add it to the page
                     page.add_region(region_instance, name=name, source="named")
                 elif region_instance is not None:
-                    logger.warning(
-                        f"Region function did not return a valid Region object for page {page.number}. Got: {type(region_instance)}"
-                    )
+                    logger.warning(f"Region function did not return a valid Region for page {page.number}")
             except Exception as e:
-                logger.error(
-                    f"Error executing or adding region function for page {page.number}: {e}",
-                    exc_info=True,
-                )
+                logger.error(f"Error adding region for page {page.number}: {e}")
 
         return self
 
@@ -471,22 +438,19 @@ class PDF:
         Find the first element matching the selector.
 
         Args:
-            selector: CSS-like selector string (e.g., 'text:contains("Annual Report")')
-            apply_exclusions: Whether to exclude elements in exclusion regions (default: True)
-            regex: Whether to use regex for text search in :contains (default: False)
-            case: Whether to do case-sensitive text search (default: True)
+            selector: CSS-like selector string
+            apply_exclusions: Whether to exclude elements in exclusion regions
+            regex: Whether to use regex for text search
+            case: Whether to do case-sensitive text search
             **kwargs: Additional filter parameters
 
         Returns:
             Element object or None if not found
         """
-        # Ensure _pages is initialized
         if not hasattr(self, "_pages"):
             raise AttributeError("PDF pages not yet initialized.")
 
         selector_obj = parse_selector(selector)
-
-        # Pass regex and case flags to selector function
         kwargs["regex"] = regex
         kwargs["case"] = case
 
@@ -502,22 +466,19 @@ class PDF:
         Find all elements matching the selector.
 
         Args:
-            selector: CSS-like selector string (e.g., 'text[color=(1,0,0)]')
-            apply_exclusions: Whether to exclude elements in exclusion regions (default: True)
-            regex: Whether to use regex for text search in :contains (default: False)
-            case: Whether to do case-sensitive text search (default: True)
+            selector: CSS-like selector string
+            apply_exclusions: Whether to exclude elements in exclusion regions
+            regex: Whether to use regex for text search
+            case: Whether to do case-sensitive text search
             **kwargs: Additional filter parameters
 
         Returns:
             ElementCollection with matching elements
         """
-        # Ensure _pages is initialized
         if not hasattr(self, "_pages"):
             raise AttributeError("PDF pages not yet initialized.")
 
         selector_obj = parse_selector(selector)
-
-        # Pass regex and case flags to selector function
         kwargs["regex"] = regex
         kwargs["case"] = case
 
@@ -534,8 +495,8 @@ class PDF:
 
         Args:
             selector_obj: Parsed selector dictionary
-            apply_exclusions: Whether to exclude elements in exclusion regions (default: True)
-            first_only: If True, stop searching after the first match is found.
+            apply_exclusions: Whether to exclude elements in exclusion regions
+            first_only: If True, stop searching after the first match is found
             **kwargs: Additional filter parameters
 
         Returns:
@@ -543,57 +504,45 @@ class PDF:
         """
         from natural_pdf.elements.collections import ElementCollection
 
-        # Determine page range to search
         page_indices = kwargs.get("pages", range(len(self._pages)))
         if isinstance(page_indices, int):
             page_indices = [page_indices]
         elif isinstance(page_indices, slice):
             page_indices = range(*page_indices.indices(len(self._pages)))
 
-        # Check for cross-page pseudo-classes (currently not supported)
         for pseudo in selector_obj.get("pseudo_classes", []):
             if pseudo.get("name") in ("spans", "continues"):
                 logger.warning("Cross-page selectors ('spans', 'continues') are not yet supported.")
                 return ElementCollection([])
 
-        # Regular case: collect elements from each page
         all_elements = []
         for page_idx in page_indices:
             if 0 <= page_idx < len(self._pages):
                 page = self._pages[page_idx]
-                # Pass first_only down to page._apply_selector
                 page_elements_collection = page._apply_selector(
                     selector_obj, apply_exclusions=apply_exclusions, first_only=first_only, **kwargs
                 )
                 if page_elements_collection:
                     page_elements = page_elements_collection.elements
                     all_elements.extend(page_elements)
-                    # If we only need the first match overall, and we found one on this page, stop
                     if first_only and page_elements:
-                        break  # Stop iterating through pages
+                        break
             else:
                 logger.warning(f"Page index {page_idx} out of range (0-{len(self._pages)-1}).")
 
-        # Create a combined collection
         combined = ElementCollection(all_elements)
 
-        # Sort in document order if requested and not first_only (already sorted by page)
         if not first_only and kwargs.get("document_order", True):
-            # Check if elements have page, top, x0 before sorting
             if all(
                 hasattr(el, "page") and hasattr(el, "top") and hasattr(el, "x0")
                 for el in combined.elements
             ):
                 combined.sort(key=lambda el: (el.page.index, el.top, el.x0))
             else:
-                # Elements might be Regions without inherent sorting order yet
-                # Attempt sorting by page index if possible
                 try:
                     combined.sort(key=lambda el: el.page.index)
                 except AttributeError:
-                    logger.warning(
-                        "Cannot sort elements in document order: Missing required attributes (e.g., page)."
-                    )
+                    logger.warning("Cannot sort elements in document order: Missing required attributes.")
 
         return combined
 
@@ -610,24 +559,21 @@ class PDF:
 
         Args:
             selector: Optional selector to filter elements
-            preserve_whitespace: Whether to keep blank characters (default: True)
-            use_exclusions: Whether to apply exclusion regions (default: True)
-            debug_exclusions: Whether to output detailed debugging for exclusions (default: False)
+            preserve_whitespace: Whether to keep blank characters
+            use_exclusions: Whether to apply exclusion regions
+            debug_exclusions: Whether to output detailed debugging for exclusions
             **kwargs: Additional extraction parameters
 
         Returns:
             Extracted text as string
         """
-        # Ensure _pages is initialized
         if not hasattr(self, "_pages"):
             raise AttributeError("PDF pages not yet initialized.")
 
-        # If selector is provided, find elements first
         if selector:
             elements = self.find_all(selector, apply_exclusions=use_exclusions, **kwargs)
             return elements.extract_text(preserve_whitespace=preserve_whitespace, **kwargs)
 
-        # Otherwise extract from all pages
         if debug_exclusions:
             print(f"PDF: Extracting text with exclusions from {len(self.pages)} pages")
             print(f"PDF: Found {len(self._exclusions)} document-level exclusions")
@@ -648,25 +594,6 @@ class PDF:
 
         return "\n".join(texts)
 
-    def extract(self, selector: str, preserve_whitespace=True, **kwargs) -> str:
-        """
-        Shorthand for finding elements and extracting their text.
-
-        Args:
-            selector: CSS-like selector string
-            preserve_whitespace: Whether to keep blank characters (default: True)
-            **kwargs: Additional extraction parameters
-
-        Returns:
-            Extracted text from matching elements
-        """
-        # Ensure _pages is initialized
-        if not hasattr(self, "_pages"):
-            raise AttributeError("PDF pages not yet initialized.")
-        return self.extract_text(
-            selector, preserve_whitespace=preserve_whitespace, use_exclusions=True, **kwargs
-        )  # apply_exclusions is handled by find_all in extract_text
-
     def extract_tables(
         self, selector: Optional[str] = None, merge_across_pages: bool = False, **kwargs
     ) -> List[Any]:
@@ -681,54 +608,43 @@ class PDF:
         Returns:
             List of extracted tables
         """
-        # Ensure _pages is initialized
         if not hasattr(self, "_pages"):
             raise AttributeError("PDF pages not yet initialized.")
-        # TODO: Implement table extraction
+            
         logger.warning("PDF.extract_tables is not fully implemented yet.")
         all_tables = []
+        
         for page in self.pages:
-            # Assuming page.extract_tables(**kwargs) exists or is added
             if hasattr(page, "extract_tables"):
                 all_tables.extend(page.extract_tables(**kwargs))
             else:
                 logger.debug(f"Page {page.number} does not have extract_tables method.")
-        # Placeholder filtering
+                
         if selector:
             logger.warning("Filtering extracted tables by selector is not implemented.")
-            # Would need to parse selector and filter the list `all_tables`
-        # Placeholder merging
+            
         if merge_across_pages:
             logger.warning("Merging tables across pages is not implemented.")
-            # Would need logic to detect and merge related tables
+            
         return all_tables
 
-    # --- New Method: save_searchable ---
     def save_searchable(self, output_path: Union[str, "Path"], dpi: int = 300, **kwargs):
         """
         Saves the PDF with an OCR text layer, making content searchable.
 
         Requires optional dependencies. Install with: pip install "natural-pdf[ocr-save]"
 
-        Note: OCR must have been applied to the pages beforehand
-              (e.g., using pdf.apply_ocr()).
-
         Args:
-            output_path: Path to save the searchable PDF.
-            dpi: Resolution for rendering and OCR overlay (default 300).
-            **kwargs: Additional keyword arguments passed to the exporter.
+            output_path: Path to save the searchable PDF
+            dpi: Resolution for rendering and OCR overlay
+            **kwargs: Additional keyword arguments passed to the exporter
         """
-        # Import moved here, assuming it's always available now
         from natural_pdf.exporters.searchable_pdf import create_searchable_pdf
 
-        # Convert pathlib.Path to string if necessary
         output_path_str = str(output_path)
-
         create_searchable_pdf(self, output_path_str, dpi=dpi, **kwargs)
         logger.info(f"Searchable PDF saved to: {output_path_str}")
 
-    # --- End New Method ---
-
     def ask(
         self,
         question: str,
@@ -750,27 +666,21 @@ class PDF:
             **kwargs: Additional parameters passed to the QA engine
 
         Returns:
-            A dictionary containing the answer, confidence, and other metadata.
-            Result will have an 'answer' key containing the answer text.
+            A dictionary containing the answer, confidence, and other metadata
         """
         from natural_pdf.qa import get_qa_engine
 
-        # Initialize or get QA engine
         qa_engine = get_qa_engine() if model is None else get_qa_engine(model_name=model)
 
-        # Determine which pages to query
         if pages is None:
             target_pages = list(range(len(self.pages)))
         elif isinstance(pages, int):
-            # Single page
             target_pages = [pages]
         elif isinstance(pages, (list, range)):
-            # List or range of pages
             target_pages = pages
         else:
             raise ValueError(f"Invalid pages parameter: {pages}")
 
-        # Actually query each page and gather results
         results = []
         for page_idx in target_pages:
             if 0 <= page_idx < len(self.pages):
@@ -779,211 +689,148 @@ class PDF:
                     page=page, question=question, min_confidence=min_confidence, **kwargs
                 )
 
-                # Add to results if it found an answer
                 if page_result and page_result.get("found", False):
                     results.append(page_result)
 
-        # Sort results by confidence
         results.sort(key=lambda x: x.get("confidence", 0), reverse=True)
 
-        # Return the best result, or a default result if none found
         if results:
             return results[0]
         else:
-            # Return a structure indicating no answer found
             return {
                 "answer": None,
                 "confidence": 0.0,
                 "found": False,
-                "page_num": None,  # Or maybe the pages searched?
+                "page_num": None,
                 "source_elements": [],
             }
 
     def search_within_index(
         self,
         query: Union[str, Path, Image.Image, Region],
-        search_service: SearchServiceProtocol,  # Now required
+        search_service: SearchServiceProtocol,
         options: Optional[SearchOptions] = None,
     ) -> List[Dict[str, Any]]:
         """
-        Finds relevant documents specifically originating from THIS PDF document
-        within a search index managed by the provided SearchService.
-
-        This method uses a pre-configured SearchService instance and adds
-        a filter to the search query to scope results only to pages from
-        this specific PDF object (based on its resolved path).
+        Finds relevant documents from this PDF within a search index.
 
         Args:
-            query: The search query (text, image path, PIL Image, Region).
-            search_service: A pre-configured SearchService instance pointing to the
-                            index where this PDF's content (or related content)
-                            is expected to be found.
-            options: Optional SearchOptions to configure the query (top_k, filters, etc.).
-                     Any existing filters in `options` will be combined with the
-                     PDF-scoping filter using an 'AND' condition.
+            query: The search query (text, image path, PIL Image, Region)
+            search_service: A pre-configured SearchService instance
+            options: Optional SearchOptions to configure the query
 
         Returns:
-            A list of result dictionaries, sorted by relevance, containing only
-            results originating from this PDF's pages.
+            A list of result dictionaries, sorted by relevance
 
         Raises:
-            ImportError: If search dependencies are not installed.
-            ValueError: If search_service is None.
-            TypeError: If search_service does not conform to the protocol.
-            FileNotFoundError: If the collection managed by the service does not exist.
-            RuntimeError: For other search failures.
+            ImportError: If search dependencies are not installed
+            ValueError: If search_service is None
+            TypeError: If search_service does not conform to the protocol
+            FileNotFoundError: If the collection managed by the service does not exist
+            RuntimeError: For other search failures
         """
         if not search_service:
             raise ValueError("A configured SearchServiceProtocol instance must be provided.")
-        # Optional stricter check:
-        # if not isinstance(search_service, SearchServiceProtocol):
-        #     raise TypeError("Provided search_service does not conform to SearchServiceProtocol.")
 
-        # Get collection name from service for logging
         collection_name = getattr(search_service, "collection_name", "<Unknown Collection>")
-        logger.info(
-            f"Searching within index '{collection_name}' (via provided service) for content from PDF '{self.path}'. Query type: {type(query).__name__}."
-        )
+        logger.info(f"Searching within index '{collection_name}' for content from PDF '{self.path}'")
+
+        service = search_service
 
-        # --- 1. Get Search Service Instance --- (REMOVED - provided directly)
-        # service: SearchServiceProtocol
-        # if search_service:
-        #     service = search_service
-        # else:
-        #     logger.debug(f"Getting SearchService instance via factory (persist={persist}, collection={collection_name})...")
-        #     factory_args = {**kwargs, 'collection_name': collection_name, 'persist': persist}
-        #     # TODO: Pass embedding model from options/pdf config if needed?
-        #     service = get_search_service(**factory_args)
-        service = search_service  # Use validated provided service
-
-        # --- 2. Prepare Query and Options ---
         query_input = query
-        # Resolve options (use default TextSearch if none provided)
         effective_options = copy.deepcopy(options) if options is not None else TextSearchOptions()
 
-        # Handle Region query - extract text for now
         if isinstance(query, Region):
             logger.debug("Query is a Region object. Extracting text.")
             if not isinstance(effective_options, TextSearchOptions):
-                logger.warning(
-                    "Querying with Region image requires MultiModalSearchOptions (Not fully implemented). Falling back to text extraction."
-                )
+                logger.warning("Querying with Region image requires MultiModalSearchOptions. Falling back to text extraction.")
             query_input = query.extract_text()
             if not query_input or query_input.isspace():
                 logger.error("Region has no extractable text for query.")
                 return []
 
-        # --- 3. Add Filter to Scope Search to THIS PDF ---
-        # Assume metadata field 'pdf_path' stores the resolved path used during indexing
+        # Add filter to scope search to THIS PDF
         pdf_scope_filter = {
-            "field": "pdf_path",  # Or potentially "source_path" depending on indexing metadata
+            "field": "pdf_path",
             "operator": "eq",
-            "value": self.path,  # Use the resolved path of this PDF instance
+            "value": self.path,
         }
         logger.debug(f"Applying filter to scope search to PDF: {pdf_scope_filter}")
 
         # Combine with existing filters in options (if any)
         if effective_options.filters:
-            logger.debug(
-                f"Combining PDF scope filter with existing filters: {effective_options.filters}"
-            )
-            # Assume filters are compatible with the underlying search service
-            # If existing filters aren't already in an AND block, wrap them
-            if (
-                isinstance(effective_options.filters, dict)
-                and effective_options.filters.get("operator") == "AND"
-            ):
-                # Already an AND block, just append the condition
+            logger.debug(f"Combining PDF scope filter with existing filters")
+            if isinstance(effective_options.filters, dict) and effective_options.filters.get("operator") == "AND":
                 effective_options.filters["conditions"].append(pdf_scope_filter)
             elif isinstance(effective_options.filters, list):
-                # Assume list represents implicit AND conditions
                 effective_options.filters = {
                     "operator": "AND",
                     "conditions": effective_options.filters + [pdf_scope_filter],
                 }
-            elif isinstance(effective_options.filters, dict):  # Single filter dict
+            elif isinstance(effective_options.filters, dict):
                 effective_options.filters = {
                     "operator": "AND",
                     "conditions": [effective_options.filters, pdf_scope_filter],
                 }
             else:
-                logger.warning(
-                    f"Unsupported format for existing filters: {type(effective_options.filters)}. Overwriting with PDF scope filter."
-                )
+                logger.warning(f"Unsupported format for existing filters. Overwriting with PDF scope filter.")
                 effective_options.filters = pdf_scope_filter
         else:
             effective_options.filters = pdf_scope_filter
 
         logger.debug(f"Final filters for service search: {effective_options.filters}")
 
-        # --- 4. Call SearchService ---
         try:
-            # Call the service's search method (no collection_name needed)
             results = service.search(
                 query=query_input,
                 options=effective_options,
             )
-            logger.info(
-                f"SearchService returned {len(results)} results scoped to PDF '{self.path}' within collection '{collection_name}'."
-            )
+            logger.info(f"SearchService returned {len(results)} results from PDF '{self.path}'")
             return results
         except FileNotFoundError as fnf:
-            logger.error(
-                f"Search failed: Collection '{collection_name}' not found by service. Error: {fnf}"
-            )
-            raise  # Re-raise specific error
+            logger.error(f"Search failed: Collection not found. Error: {fnf}")
+            raise
         except Exception as e:
-            logger.error(
-                f"SearchService search failed for PDF '{self.path}' in collection '{collection_name}': {e}",
-                exc_info=True,
-            )
-            raise RuntimeError(
-                f"Search within index failed for PDF '{self.path}'. See logs for details."
-            ) from e
+            logger.error(f"SearchService search failed: {e}")
+            raise RuntimeError(f"Search within index failed. See logs for details.") from e
 
     def export_ocr_correction_task(self, output_zip_path: str, **kwargs):
         """
-        Exports OCR results from this PDF into a correction task package (zip file).
+        Exports OCR results from this PDF into a correction task package.
 
         Args:
-            output_zip_path: The path to save the output zip file.
+            output_zip_path: The path to save the output zip file
             **kwargs: Additional arguments passed to create_correction_task_package
-                      (e.g., image_render_scale, overwrite).
         """
         try:
             from natural_pdf.utils.packaging import create_correction_task_package
-
             create_correction_task_package(source=self, output_zip_path=output_zip_path, **kwargs)
         except ImportError:
-            logger.error(
-                "Failed to import 'create_correction_task_package'. Packaging utility might be missing."
-            )
-            # Or raise
+            logger.error("Failed to import 'create_correction_task_package'. Packaging utility might be missing.")
         except Exception as e:
-            logger.error(f"Failed to export correction task for {self.path}: {e}", exc_info=True)
-            raise  # Re-raise the exception from the utility function
+            logger.error(f"Failed to export correction task: {e}")
+            raise
 
     def correct_ocr(
         self,
         correction_callback: Callable[[Any], Optional[str]],
         pages: Optional[Union[Iterable[int], range, slice]] = None,
-    ) -> "PDF":  # Return self for chaining
+        max_workers: Optional[int] = None,
+        progress_callback: Optional[Callable[[], None]] = None,
+    ) -> "PDF":
         """
-        Applies corrections to OCR-generated text elements using a callback function,
-        delegating the core work to the `Page.correct_ocr` method.
+        Applies corrections to OCR text elements using a callback function.
 
         Args:
-            correction_callback: A function that accepts a single argument (an element
-                                object) and returns `Optional[str]`. It returns the
-                                corrected text string if an update is needed, otherwise None.
+            correction_callback: Function that takes an element and returns corrected text or None
             pages: Optional page indices/slice to limit the scope of correction
-                (default: all pages).
+            max_workers: Maximum number of threads to use for parallel execution
+            progress_callback: Optional callback function for progress updates
 
         Returns:
-            Self for method chaining.
+            Self for method chaining
         """
-        # Determine target pages
-        target_page_indices: List[int] = []
+        target_page_indices = []
         if pages is None:
             target_page_indices = list(range(len(self._pages)))
         elif isinstance(pages, slice):
@@ -991,56 +838,49 @@ class PDF:
         elif hasattr(pages, "__iter__"):
             try:
                 target_page_indices = [int(i) for i in pages]
-                # Validate indices
                 for idx in target_page_indices:
                     if not (0 <= idx < len(self._pages)):
                         raise IndexError(f"Page index {idx} out of range (0-{len(self._pages)-1}).")
             except (IndexError, TypeError, ValueError) as e:
-                raise ValueError(
-                    f"Invalid page index or type provided in 'pages': {pages}. Error: {e}"
-                ) from e
+                raise ValueError(f"Invalid page index in 'pages': {pages}. Error: {e}") from e
         else:
-            raise TypeError("'pages' must be None, a slice, or an iterable of page indices (int).")
+            raise TypeError("'pages' must be None, a slice, or an iterable of page indices.")
 
         if not target_page_indices:
             logger.warning("No pages selected for OCR correction.")
             return self
 
-        logger.info(
-            f"Starting OCR correction process via Page delegation for pages: {target_page_indices}"
-        )
+        logger.info(f"Starting OCR correction for pages: {target_page_indices}")
 
-        # Iterate through target pages and call their correct_ocr method
         for page_idx in target_page_indices:
             page = self._pages[page_idx]
             try:
-                page.correct_ocr(correction_callback=correction_callback)
+                page.correct_ocr(
+                    correction_callback=correction_callback,
+                    max_workers=max_workers,
+                    progress_callback=progress_callback,
+                )
             except Exception as e:
-                logger.error(f"Error during correct_ocr on page {page_idx}: {e}", exc_info=True)
-                # Optionally re-raise or just log and continue
+                logger.error(f"Error during correct_ocr on page {page_idx}: {e}")
 
-        logger.info(f"OCR correction process finished for requested pages.")
+        logger.info("OCR correction process finished.")
         return self
 
     def __len__(self) -> int:
         """Return the number of pages in the PDF."""
-        # Ensure _pages is initialized
         if not hasattr(self, "_pages"):
-            # Return 0 or raise error if not fully initialized? Let's return 0.
             return 0
         return len(self._pages)
 
-    def __getitem__(self, key) -> Union[Page, "PageCollection"]:  # Return PageCollection for slice
+    def __getitem__(self, key) -> Union[Page, "PageCollection"]:
         """Access pages by index or slice."""
-        # Check if self._pages has been initialized
         if not hasattr(self, "_pages"):
             raise AttributeError("PDF pages not initialized yet.")
+            
         if isinstance(key, slice):
-            # Return a PageCollection slice
             from natural_pdf.elements.collections import PageCollection
-
             return PageCollection(self._pages[key])
-        # Check index bounds before accessing
+            
         if isinstance(key, int):
             if 0 <= key < len(self._pages):
                 return self._pages[key]
@@ -1054,13 +894,12 @@ class PDF:
         if hasattr(self, "_pdf") and self._pdf is not None:
             try:
                 self._pdf.close()
-                logger.debug(f"Closed underlying pdfplumber PDF object for {self.source_path}")
+                logger.debug(f"Closed pdfplumber PDF object for {self.source_path}")
             except Exception as e:
                 logger.warning(f"Error closing pdfplumber object: {e}")
             finally:
                 self._pdf = None
 
-        # Clean up temporary file if it exists
         if hasattr(self, "_temp_file") and self._temp_file is not None:
             temp_file_path = None
             try:
@@ -1070,7 +909,7 @@ class PDF:
                         os.unlink(temp_file_path)
                         logger.debug(f"Removed temporary PDF file: {temp_file_path}")
             except Exception as e:
-                logger.warning(f"Failed to clean up temporary PDF file '{temp_file_path}': {e}")
+                logger.warning(f"Failed to clean up temporary file '{temp_file_path}': {e}")
             finally:
                 self._temp_file = None
 
@@ -1082,6 +921,176 @@ class PDF:
         """Context manager exit."""
         self.close()
 
-    # --- Indexable Protocol Methods --- Needed for search/sync
     def get_id(self) -> str:
+        """Get unique identifier for this PDF."""
         return self.path
+
+    # --- Classification Methods --- #
+
+    def classify_pages(
+        self,
+        categories: List[str],
+        model: Optional[str] = None,
+        pages: Optional[Union[Iterable[int], range, slice]] = None,
+        analysis_key: str = "classification",
+        using: Optional[str] = None,
+        **kwargs,
+    ) -> "PDF":
+        """
+        Classifies specified pages of the PDF.
+
+        Args:
+            categories: List of category names
+            model: Model identifier ('text', 'vision', or specific HF ID)
+            pages: Page indices, slice, or None for all pages
+            analysis_key: Key to store results in page's analyses dict
+            using: Processing mode ('text' or 'vision')
+            **kwargs: Additional arguments for the ClassificationManager
+
+        Returns:
+            Self for method chaining
+        """
+        if not categories:
+            raise ValueError("Categories list cannot be empty.")
+
+        try:
+            manager = self.get_manager('classification')
+        except (ValueError, RuntimeError) as e:
+            raise ClassificationError(f"Cannot get ClassificationManager: {e}") from e
+
+        if not manager or not manager.is_available():
+            try:
+                from natural_pdf.classification.manager import _CLASSIFICATION_AVAILABLE
+                if not _CLASSIFICATION_AVAILABLE:
+                    raise ImportError("Classification dependencies missing.")
+            except ImportError:
+                raise ImportError(
+                    "Classification dependencies missing. "
+                    "Install with: pip install \"natural-pdf[classification]\""
+                )
+            raise ClassificationError("ClassificationManager not available.")
+
+        target_pages = []
+        if pages is None:
+            target_pages = self._pages
+        elif isinstance(pages, slice):
+            target_pages = self._pages[pages]
+        elif hasattr(pages, "__iter__"):
+            try:
+                target_pages = [self._pages[i] for i in pages]
+            except IndexError:
+                raise ValueError("Invalid page index provided.")
+            except TypeError:
+                raise TypeError("'pages' must be None, a slice, or an iterable of page indices.")
+        else:
+            raise TypeError("'pages' must be None, a slice, or an iterable of page indices.")
+
+        if not target_pages:
+            logger.warning("No pages selected for classification.")
+            return self
+
+        inferred_using = manager.infer_using(model if model else manager.DEFAULT_TEXT_MODEL, using)
+        logger.info(f"Classifying {len(target_pages)} pages using model '{model or '(default)'}' (mode: {inferred_using})")
+
+        page_contents = []
+        pages_to_classify = []
+        logger.debug(f"Gathering content for {len(target_pages)} pages...")
+        
+        for page in target_pages:
+            try:
+                content = page._get_classification_content(model_type=inferred_using, **kwargs)
+                page_contents.append(content)
+                pages_to_classify.append(page)
+            except ValueError as e:
+                logger.warning(f"Skipping page {page.number}: Cannot get content - {e}")
+            except Exception as e:
+                logger.warning(f"Skipping page {page.number}: Error getting content - {e}")
+
+        if not page_contents:
+            logger.warning("No content could be gathered for batch classification.")
+            return self
+            
+        logger.debug(f"Gathered content for {len(pages_to_classify)} pages.")
+
+        try:
+            batch_results = manager.classify_batch(
+                item_contents=page_contents,
+                categories=categories,
+                model_id=model,
+                using=inferred_using,
+                **kwargs,
+            )
+        except Exception as e:
+            logger.error(f"Batch classification failed: {e}")
+            raise ClassificationError(f"Batch classification failed: {e}") from e
+
+        if len(batch_results) != len(pages_to_classify):
+            logger.error(f"Mismatch between number of results ({len(batch_results)}) and pages ({len(pages_to_classify)})")
+            return self
+
+        logger.debug(f"Distributing {len(batch_results)} results to pages under key '{analysis_key}'...")
+        for page, result_obj in zip(pages_to_classify, batch_results):
+            try:
+                if not hasattr(page, 'analyses') or page.analyses is None:
+                    page.analyses = {}
+                page.analyses[analysis_key] = result_obj
+            except Exception as e:
+                logger.warning(f"Failed to store classification results for page {page.number}: {e}")
+
+        logger.info(f"Finished classifying PDF pages.")
+        return self
+
+    # --- End Classification Methods --- #
+
+    # --- Extraction Support --- #
+    def _get_extraction_content(self, using: str = 'text', **kwargs) -> Any:
+        """
+        Retrieves the content for the entire PDF.
+
+        Args:
+            using: 'text' or 'vision'
+            **kwargs: Additional arguments passed to extract_text or page.to_image
+
+        Returns:
+            str: Extracted text if using='text'
+            List[PIL.Image.Image]: List of page images if using='vision'
+            None: If content cannot be retrieved
+        """
+        if using == 'text':
+            try:
+                layout = kwargs.pop('layout', True)
+                return self.extract_text(layout=layout, **kwargs)
+            except Exception as e:
+                logger.error(f"Error extracting text from PDF: {e}")
+                return None
+        elif using == 'vision':
+            page_images = []
+            logger.info(f"Rendering {len(self.pages)} pages to images...")
+            
+            resolution = kwargs.pop('resolution', 72)
+            include_highlights = kwargs.pop('include_highlights', False)
+            labels = kwargs.pop('labels', False)
+            
+            try:
+                for page in tqdm(self.pages, desc="Rendering Pages"):
+                    img = page.to_image(
+                        resolution=resolution,
+                        include_highlights=include_highlights,
+                        labels=labels,
+                        **kwargs
+                    )
+                    if img:
+                        page_images.append(img)
+                    else:
+                        logger.warning(f"Failed to render page {page.number}, skipping.")
+                if not page_images:
+                    logger.error("Failed to render any pages.")
+                    return None
+                return page_images
+            except Exception as e:
+                logger.error(f"Error rendering pages: {e}")
+                return None
+        else:
+            logger.error(f"Unsupported value for 'using': {using}")
+            return None
+    # --- End Extraction Support --- #