natural-pdf 0.1.11__py3-none-any.whl → 0.1.12__py3-none-any.whl

natural_pdf/elements/collections.py

@@ -20,10 +20,10 @@ from typing import (
 )
 
 from pdfplumber.utils.geometry import objects_to_bbox
-from PIL import Image, ImageDraw, ImageFont
 
 # New Imports
 from pdfplumber.utils.text import TEXTMAP_KWARGS, WORD_EXTRACTOR_KWARGS, chars_to_textmap
+from PIL import Image, ImageDraw, ImageFont
 from tqdm.auto import tqdm
 
 from natural_pdf.classification.manager import ClassificationManager
@@ -46,6 +46,7 @@ except ImportError:
 
 try:
     from natural_pdf.exporters.searchable_pdf import create_searchable_pdf
+
     pass
 except ImportError:
     create_searchable_pdf = None
@@ -61,7 +62,7 @@ logger = logging.getLogger(__name__)
 
 if TYPE_CHECKING:
     from natural_pdf.core.page import Page
-    from natural_pdf.core.pdf import PDF # ---> ADDED PDF type hint
+    from natural_pdf.core.pdf import PDF  # ---> ADDED PDF type hint
     from natural_pdf.elements.region import Region
 
 T = TypeVar("T")
@@ -840,6 +841,7 @@ class ElementCollection(
         labels: bool = True,  # Use 'labels' consistent with service
         legend_position: str = "right",
         render_ocr: bool = False,
+        width: Optional[int] = None,  # Add width parameter
     ) -> Optional["Image.Image"]:
         """
         Generates a temporary preview image highlighting elements in this collection
@@ -862,6 +864,7 @@ class ElementCollection(
             labels: Whether to include a legend for the temporary highlights.
             legend_position: Position of the legend ('right', 'left', 'top', 'bottom').
             render_ocr: Whether to render OCR text.
+            width: Optional width for the output image in pixels.
 
         Returns:
             PIL Image object of the temporary preview, or None if rendering fails or
@@ -922,6 +925,7 @@ class ElementCollection(
                 page_index=page.index,
                 temporary_highlights=highlight_data_list,
                 scale=scale,
+                width=width,  # Pass the width parameter
                 labels=labels,  # Use 'labels'
                 legend_position=legend_position,
                 render_ocr=render_ocr,
@@ -1159,10 +1163,96 @@ class ElementCollection(
 
         Args:
             selector: CSS-like selector string
+            contains: How to determine if elements are inside: 'all' (fully inside),
+                      'any' (any overlap), or 'center' (center point inside).
+                      (default: "all")
             apply_exclusions: Whether to exclude elements in exclusion regions
         """
         return self.apply(lambda element: element.find(selector, **kwargs))
 
+    @overload
+    def find_all(
+        self,
+        *,
+        text: str,
+        contains: str = "all",
+        apply_exclusions: bool = True,
+        regex: bool = False,
+        case: bool = True,
+        **kwargs,
+    ) -> "ElementCollection": ...
+
+    @overload
+    def find_all(
+        self,
+        selector: str,
+        *,
+        contains: str = "all",
+        apply_exclusions: bool = True,
+        regex: bool = False,
+        case: bool = True,
+        **kwargs,
+    ) -> "ElementCollection": ...
+
+    def find_all(
+        self,
+        selector: Optional[str] = None,
+        *,
+        text: Optional[str] = None,
+        contains: str = "all",
+        apply_exclusions: bool = True,
+        regex: bool = False,
+        case: bool = True,
+        **kwargs,
+    ) -> "ElementCollection":
+        """
+        Find all elements within each element of this collection matching the selector OR text,
+        and return a flattened collection of all found sub-elements.
+
+        Provide EITHER `selector` OR `text`, but not both.
+
+        Args:
+            selector: CSS-like selector string.
+            text: Text content to search for (equivalent to 'text:contains(...)').
+            contains: How to determine if elements are inside: 'all' (fully inside),
+                     'any' (any overlap), or 'center' (center point inside).
+                     (default: "all")
+            apply_exclusions: Whether to apply exclusion regions (default: True).
+            regex: Whether to use regex for text search (`selector` or `text`) (default: False).
+            case: Whether to do case-sensitive text search (`selector` or `text`) (default: True).
+            **kwargs: Additional parameters for element filtering.
+
+        Returns:
+            A new ElementCollection containing all matching sub-elements from all elements
+            in this collection.
+        """
+        if selector is None and text is None:
+            raise ValueError("Either 'selector' or 'text' must be provided to find_all.")
+        if selector is not None and text is not None:
+            raise ValueError("Provide either 'selector' or 'text' to find_all, not both.")
+
+        all_found_elements: List[Element] = []
+        for element in self._elements:
+            if hasattr(element, "find_all") and callable(element.find_all):
+                # Element.find_all returns an ElementCollection
+                found_in_element: "ElementCollection" = element.find_all(
+                    selector=selector,
+                    text=text,
+                    contains=contains,
+                    apply_exclusions=apply_exclusions,
+                    regex=regex,
+                    case=case,
+                    **kwargs,
+                )
+                if found_in_element and found_in_element.elements:
+                    all_found_elements.extend(found_in_element.elements)
+            # else:
+            # Elements in the collection are expected to support find_all.
+            # If an element type doesn't, an AttributeError will naturally occur,
+            # or a more specific check/handling could be added here if needed.
+
+        return ElementCollection(all_found_elements)
+
     def extract_each_text(self, **kwargs) -> List[str]:
         """
         Extract text from each element in this region.
@@ -1633,6 +1723,7 @@ class PageCollection(Generic[P], ApplyMixin):
         self,
         *,
         text: str,
+        contains: str = "all",
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1644,6 +1735,7 @@ class PageCollection(Generic[P], ApplyMixin):
         self,
         selector: str,
         *,
+        contains: str = "all",
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1655,6 +1747,7 @@ class PageCollection(Generic[P], ApplyMixin):
         selector: Optional[str] = None,
         *,
         text: Optional[str] = None,
+        contains: str = "all",
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1668,6 +1761,9 @@ class PageCollection(Generic[P], ApplyMixin):
         Args:
             selector: CSS-like selector string.
             text: Text content to search for (equivalent to 'text:contains(...)').
+            contains: How to determine if elements are inside: 'all' (fully inside),
+                     'any' (any overlap), or 'center' (center point inside).
+                     (default: "all")
             apply_exclusions: Whether to exclude elements in exclusion regions (default: True).
             regex: Whether to use regex for text search (`selector` or `text`) (default: False).
             case: Whether to do case-sensitive text search (`selector` or `text`) (default: True).
@@ -1681,6 +1777,7 @@ class PageCollection(Generic[P], ApplyMixin):
             element = page.find(
                 selector=selector,
                 text=text,
+                contains=contains,
                 apply_exclusions=apply_exclusions,
                 regex=regex,
                 case=case,
@@ -1695,6 +1792,7 @@ class PageCollection(Generic[P], ApplyMixin):
         self,
         *,
         text: str,
+        contains: str = "all",
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1706,6 +1804,7 @@ class PageCollection(Generic[P], ApplyMixin):
         self,
         selector: str,
         *,
+        contains: str = "all",
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1717,6 +1816,7 @@ class PageCollection(Generic[P], ApplyMixin):
         selector: Optional[str] = None,
         *,
         text: Optional[str] = None,
+        contains: str = "all",
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1730,6 +1830,9 @@ class PageCollection(Generic[P], ApplyMixin):
         Args:
             selector: CSS-like selector string.
             text: Text content to search for (equivalent to 'text:contains(...)').
+            contains: How to determine if elements are inside: 'all' (fully inside),
+                     'any' (any overlap), or 'center' (center point inside).
+                     (default: "all")
             apply_exclusions: Whether to exclude elements in exclusion regions (default: True).
             regex: Whether to use regex for text search (`selector` or `text`) (default: False).
             case: Whether to do case-sensitive text search (`selector` or `text`) (default: True).
@@ -1744,6 +1847,7 @@ class PageCollection(Generic[P], ApplyMixin):
             elements = page.find_all(
                 selector=selector,
                 text=text,
+                contains=contains,
                 apply_exclusions=apply_exclusions,
                 regex=regex,
                 case=case,
@@ -2314,8 +2418,10 @@ class PageCollection(Generic[P], ApplyMixin):
         try:
             from PIL import Image, ImageDraw, ImageFont
         except ImportError:
-             logger.error("Pillow library not found, required for to_image(). Install with 'pip install Pillow'")
-             return None
+            logger.error(
+                "Pillow library not found, required for to_image(). Install with 'pip install Pillow'"
+            )
+            return None
 
         if not self.pages:
             logger.warning("Cannot generate image for empty PageCollection")
@@ -2334,27 +2440,34 @@ class PageCollection(Generic[P], ApplyMixin):
                 try:
                     font = ImageFont.load_default(16)
                 except IOError:
-                     logger.warning("Default font not found. Labels cannot be added.")
-                     add_labels = False # Disable if no font
+                    logger.warning("Default font not found. Labels cannot be added.")
+                    add_labels = False  # Disable if no font
 
         # Render individual page images
         page_images = []
         for page in pages_to_render:
             try:
                 # Assume page.to_image returns a PIL Image or None
-                img = page.to_image(width=page_width, include_highlights=True) # Render with highlights for visual context
+                img = page.to_image(
+                    width=page_width, include_highlights=True
+                )  # Render with highlights for visual context
                 if img is None:
-                     logger.warning(f"Failed to generate image for page {page.number}. Skipping.")
-                     continue
+                    logger.warning(f"Failed to generate image for page {page.number}. Skipping.")
+                    continue
             except Exception as img_err:
-                 logger.error(f"Error generating image for page {page.number}: {img_err}", exc_info=True)
-                 continue
-
+                logger.error(
+                    f"Error generating image for page {page.number}: {img_err}", exc_info=True
+                )
+                continue
 
             # Add page number label
             if add_labels and font:
                 draw = ImageDraw.Draw(img)
-                pdf_name = Path(page.pdf.path).stem if hasattr(page, "pdf") and page.pdf and hasattr(page.pdf, "path") else ""
+                pdf_name = (
+                    Path(page.pdf.path).stem
+                    if hasattr(page, "pdf") and page.pdf and hasattr(page.pdf, "path")
+                    else ""
+                )
                 label_text = f"p{page.number}"
                 if pdf_name:
                     label_text += f" - {pdf_name}"
@@ -2364,43 +2477,65 @@ class PageCollection(Generic[P], ApplyMixin):
                     # Placeholder logic - adjust based on how classification results are stored
                     category = None
                     confidence = None
-                    if hasattr(page, 'analyses') and page.analyses and 'classification' in page.analyses:
-                        result = page.analyses['classification']
+                    if (
+                        hasattr(page, "analyses")
+                        and page.analyses
+                        and "classification" in page.analyses
+                    ):
+                        result = page.analyses["classification"]
                         # Adapt based on actual structure of classification result
-                        category = getattr(result, 'label', None) or result.get('label', None) if isinstance(result, dict) else None
-                        confidence = getattr(result, 'score', None) or result.get('score', None) if isinstance(result, dict) else None
+                        category = (
+                            getattr(result, "label", None) or result.get("label", None)
+                            if isinstance(result, dict)
+                            else None
+                        )
+                        confidence = (
+                            getattr(result, "score", None) or result.get("score", None)
+                            if isinstance(result, dict)
+                            else None
+                        )
 
                     if category is not None and confidence is not None:
-                         try:
-                            category_str = f"{category} ({confidence:.2f})" # Format confidence
+                        try:
+                            category_str = f"{category} ({confidence:.2f})"  # Format confidence
                             label_text += f"\\n{category_str}"
-                         except (TypeError, ValueError): pass # Ignore formatting errors
-
+                        except (TypeError, ValueError):
+                            pass  # Ignore formatting errors
 
                 # Calculate bounding box for multi-line text and draw background/text
                 try:
                     # Using textbbox for potentially better accuracy with specific fonts
                     # Note: textbbox needs Pillow 8+
-                    bbox = draw.textbbox((5, 5), label_text, font=font, spacing=2) # Use textbbox if available
-                    bg_rect = (max(0, bbox[0] - 2), max(0, bbox[1] - 2),
-                               min(img.width, bbox[2] + 2), min(img.height, bbox[3] + 2))
+                    bbox = draw.textbbox(
+                        (5, 5), label_text, font=font, spacing=2
+                    )  # Use textbbox if available
+                    bg_rect = (
+                        max(0, bbox[0] - 2),
+                        max(0, bbox[1] - 2),
+                        min(img.width, bbox[2] + 2),
+                        min(img.height, bbox[3] + 2),
+                    )
 
                     # Draw semi-transparent background
-                    overlay = Image.new('RGBA', img.size, (255, 255, 255, 0))
+                    overlay = Image.new("RGBA", img.size, (255, 255, 255, 0))
                     draw_overlay = ImageDraw.Draw(overlay)
-                    draw_overlay.rectangle(bg_rect, fill=(255, 255, 255, 180)) # White with alpha
-                    img = Image.alpha_composite(img.convert('RGBA'), overlay).convert('RGB')
-                    draw = ImageDraw.Draw(img) # Recreate draw object
+                    draw_overlay.rectangle(bg_rect, fill=(255, 255, 255, 180))  # White with alpha
+                    img = Image.alpha_composite(img.convert("RGBA"), overlay).convert("RGB")
+                    draw = ImageDraw.Draw(img)  # Recreate draw object
 
                     # Draw the potentially multi-line text
                     draw.multiline_text((5, 5), label_text, fill=(0, 0, 0), font=font, spacing=2)
-                except AttributeError: # Fallback for older Pillow without textbbox
+                except AttributeError:  # Fallback for older Pillow without textbbox
                     # Approximate size and draw
                     # This might not be perfectly aligned
-                     draw.rectangle((2, 2, 150, 40), fill=(255, 255, 255, 180)) # Simple fixed background
-                     draw.multiline_text((5, 5), label_text, fill=(0, 0, 0), font=font, spacing=2)
+                    draw.rectangle(
+                        (2, 2, 150, 40), fill=(255, 255, 255, 180)
+                    )  # Simple fixed background
+                    draw.multiline_text((5, 5), label_text, fill=(0, 0, 0), font=font, spacing=2)
                 except Exception as draw_err:
-                     logger.error(f"Error drawing label on page {page.number}: {draw_err}", exc_info=True)
+                    logger.error(
+                        f"Error drawing label on page {page.number}: {draw_err}", exc_info=True
+                    )
 
             page_images.append(img)
 
@@ -2408,7 +2543,6 @@ class PageCollection(Generic[P], ApplyMixin):
             logger.warning("No page images were successfully rendered for the grid.")
             return None
 
-
         # Calculate grid dimensions if not provided
         num_images = len(page_images)
         if not rows and not cols:
@@ -2418,24 +2552,23 @@ class PageCollection(Generic[P], ApplyMixin):
             cols = (num_images + rows - 1) // rows
         elif cols and not rows:
             rows = (num_images + cols - 1) // cols
-        cols = max(1, cols if cols else 1) # Ensure at least 1
+        cols = max(1, cols if cols else 1)  # Ensure at least 1
         rows = max(1, rows if rows else 1)
 
-
         # Get maximum dimensions for consistent grid cells
         max_width = max(img.width for img in page_images) if page_images else 1
         max_height = max(img.height for img in page_images) if page_images else 1
 
-
         # Create grid image
         grid_width = cols * max_width + (cols + 1) * spacing
         grid_height = rows * max_height + (rows + 1) * spacing
-        grid_img = Image.new("RGB", (grid_width, grid_height), (220, 220, 220)) # Lighter gray background
-
+        grid_img = Image.new(
+            "RGB", (grid_width, grid_height), (220, 220, 220)
+        )  # Lighter gray background
 
         # Place images in grid
         for i, img in enumerate(page_images):
-            if i >= rows * cols: # Ensure we don't exceed grid capacity
+            if i >= rows * cols:  # Ensure we don't exceed grid capacity
                 break
 
             row = i // cols
@@ -2484,8 +2617,8 @@ class PageCollection(Generic[P], ApplyMixin):
         if not self.pages:
             raise ValueError("Cannot save an empty PageCollection.")
 
-        if not (ocr ^ original): # XOR: exactly one must be true
-             raise ValueError("Exactly one of 'ocr' or 'original' must be True.")
+        if not (ocr ^ original):  # XOR: exactly one must be true
+            raise ValueError("Exactly one of 'ocr' or 'original' must be True.")
 
         output_path_obj = Path(output_path)
         output_path_str = str(output_path_obj)
@@ -2494,18 +2627,29 @@ class PageCollection(Generic[P], ApplyMixin):
             if create_searchable_pdf is None:
                 raise ImportError(
                     "Saving with ocr=True requires 'pikepdf' and 'Pillow'. "
-                    "Install with: pip install \\\"natural-pdf[ocr-export]\\\"" # Escaped quotes
+                    'Install with: pip install \\"natural-pdf[ocr-export]\\"'  # Escaped quotes
                 )
 
             # Check for non-OCR vector elements (provide a warning)
             has_vector_elements = False
             for page in self.pages:
                 # Simplified check for common vector types or non-OCR chars/words
-                if (hasattr(page, 'rects') and page.rects or
-                    hasattr(page, 'lines') and page.lines or
-                    hasattr(page, 'curves') and page.curves or
-                    (hasattr(page, 'chars') and any(getattr(el, 'source', None) != 'ocr' for el in page.chars)) or
-                    (hasattr(page, 'words') and any(getattr(el, 'source', None) != 'ocr' for el in page.words))):
+                if (
+                    hasattr(page, "rects")
+                    and page.rects
+                    or hasattr(page, "lines")
+                    and page.lines
+                    or hasattr(page, "curves")
+                    and page.curves
+                    or (
+                        hasattr(page, "chars")
+                        and any(getattr(el, "source", None) != "ocr" for el in page.chars)
+                    )
+                    or (
+                        hasattr(page, "words")
+                        and any(getattr(el, "source", None) != "ocr" for el in page.words)
+                    )
+                ):
                     has_vector_elements = True
                     break
             if has_vector_elements:
@@ -2532,22 +2676,22 @@ class PageCollection(Generic[P], ApplyMixin):
             if create_original_pdf is None:
                 raise ImportError(
                     "Saving with original=True requires 'pikepdf'. "
-                    "Install with: pip install \\\"natural-pdf[ocr-export]\\\"" # Escaped quotes
+                    'Install with: pip install \\"natural-pdf[ocr-export]\\"'  # Escaped quotes
                 )
 
             # Check for OCR elements (provide a warning) - keep this check here
             has_ocr_elements = False
             for page in self.pages:
-                 # Use find_all which returns a collection; check if it's non-empty
-                 if hasattr(page, 'find_all'):
-                     ocr_text_elements = page.find_all("text[source=ocr]")
-                     if ocr_text_elements: # Check truthiness of collection
-                         has_ocr_elements = True
-                         break
-                 elif hasattr(page, 'words'): # Fallback check if find_all isn't present?
-                     if any(getattr(el, 'source', None) == 'ocr' for el in page.words):
-                          has_ocr_elements = True
-                          break
+                # Use find_all which returns a collection; check if it's non-empty
+                if hasattr(page, "find_all"):
+                    ocr_text_elements = page.find_all("text[source=ocr]")
+                    if ocr_text_elements:  # Check truthiness of collection
+                        has_ocr_elements = True
+                        break
+                elif hasattr(page, "words"):  # Fallback check if find_all isn't present?
+                    if any(getattr(el, "source", None) == "ocr" for el in page.words):
+                        has_ocr_elements = True
+                        break
 
             if has_ocr_elements:
                 logger.warning(
@@ -2565,5 +2709,5 @@ class PageCollection(Generic[P], ApplyMixin):
             except Exception as e:
                 # Error logging is handled within create_original_pdf
                 # Re-raise the exception caught from the exporter
-                raise e # Keep the original exception type (ValueError, RuntimeError, etc.)
+                raise e  # Keep the original exception type (ValueError, RuntimeError, etc.)
             # <--- END MODIFIED

natural_pdf/elements/region.py

@@ -772,6 +772,7 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
         # Add a default color for standalone show
         color: Optional[Union[Tuple, str]] = "blue",
         label: Optional[str] = None,
+        width: Optional[int] = None,  # Add width parameter
     ) -> "Image.Image":
         """
         Show the page with just this region highlighted temporarily.
@@ -782,6 +783,7 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
             legend_position: Position of the legend
             color: Color to highlight this region (default: blue)
             label: Optional label for this region in the legend
+            width: Optional width for the output image in pixels
 
         Returns:
             PIL Image of the page with only this region highlighted
@@ -812,6 +814,7 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
             page_index=self._page.index,
             temporary_highlights=[temp_highlight_data],
             scale=scale,
+            width=width,  # Pass the width parameter
             labels=labels,
             legend_position=legend_position,
         )
@@ -1333,6 +1336,7 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
         self,
         *,
         text: str,
+        contains: str = "all",
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1344,6 +1348,7 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
         self,
         selector: str,
         *,
+        contains: str = "all",
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1355,6 +1360,7 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
         selector: Optional[str] = None,  # Now optional
         *,
         text: Optional[str] = None,  # New text parameter
+        contains: str = "all",  # New parameter for containment behavior
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1368,6 +1374,9 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
         Args:
             selector: CSS-like selector string.
             text: Text content to search for (equivalent to 'text:contains(...)').
+            contains: How to determine if elements are inside: 'all' (fully inside),
+                     'any' (any overlap), or 'center' (center point inside).
+                     (default: "all")
             apply_exclusions: Whether to exclude elements in exclusion regions (default: True).
             regex: Whether to use regex for text search (`selector` or `text`) (default: False).
             case: Whether to do case-sensitive text search (`selector` or `text`) (default: True).
@@ -1380,6 +1389,7 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
         elements = self.find_all(
             selector=selector,
             text=text,
+            contains=contains,
             apply_exclusions=apply_exclusions,
             regex=regex,
             case=case,
@@ -1392,6 +1402,7 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
         self,
         *,
         text: str,
+        contains: str = "all",
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1403,6 +1414,7 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
         self,
         selector: str,
         *,
+        contains: str = "all",
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1414,6 +1426,7 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
         selector: Optional[str] = None,  # Now optional
         *,
         text: Optional[str] = None,  # New text parameter
+        contains: str = "all",  # New parameter to control inside/overlap behavior
         apply_exclusions: bool = True,
         regex: bool = False,
         case: bool = True,
@@ -1427,6 +1440,9 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
         Args:
             selector: CSS-like selector string.
             text: Text content to search for (equivalent to 'text:contains(...)').
+            contains: How to determine if elements are inside: 'all' (fully inside),
+                     'any' (any overlap), or 'center' (center point inside).
+                     (default: "all")
             apply_exclusions: Whether to exclude elements in exclusion regions (default: True).
             regex: Whether to use regex for text search (`selector` or `text`) (default: False).
             case: Whether to do case-sensitive text search (`selector` or `text`) (default: True).
@@ -1442,6 +1458,10 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
         if selector is None and text is None:
             raise ValueError("Provide either 'selector' or 'text'.")
 
+        # Validate contains parameter
+        if contains not in ["all", "any", "center"]:
+            raise ValueError(f"Invalid contains value: {contains}. Must be 'all', 'any', or 'center'")
+
         # Construct selector if 'text' is provided
         effective_selector = ""
         if text is not None:
@@ -1481,22 +1501,34 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
             # Let the page handle its exclusion logic if needed
             potential_elements = self.page.find_all(
                 selector=effective_selector,
-                apply_exclusions=False,  # Apply exclusions LATER based on region bbox
+                apply_exclusions=apply_exclusions,
                 regex=regex,
                 case=case,
                 **kwargs,
             )
 
-            # Filter these elements to those strictly within the region's bounds
+            # Filter these elements based on the specified containment method
             region_bbox = self.bbox
-            matching_elements = [
-                el
-                for el in potential_elements
-                if el.x0 >= region_bbox[0]
-                and el.top >= region_bbox[1]
-                and el.x1 <= region_bbox[2]
-                and el.bottom <= region_bbox[3]
-            ]
+            matching_elements = []
+            
+            if contains == "all":  # Fully inside (strict)
+                matching_elements = [
+                    el for el in potential_elements
+                    if el.x0 >= region_bbox[0]
+                    and el.top >= region_bbox[1]
+                    and el.x1 <= region_bbox[2]
+                    and el.bottom <= region_bbox[3]
+                ]
+            elif contains == "any":  # Any overlap
+                matching_elements = [
+                    el for el in potential_elements
+                    if self.intersects(el)
+                ]
+            elif contains == "center":  # Center point inside
+                matching_elements = [
+                    el for el in potential_elements
+                    if self.is_element_center_inside(el)
+                ]
 
             return ElementCollection(matching_elements)
 
@@ -1988,7 +2020,7 @@ class Region(DirectionalMixin, ClassificationMixin, ExtractionMixin):
             from natural_pdf.qa.document_qa import get_qa_engine
         except ImportError:
             logger.error(
-                "Question answering requires optional dependencies. Install with `pip install natural-pdf[qa]`"
+                "Question answering requires optional dependencies. Install with `pip install natural-pdf[core-ml]`"
             )
             return {
                 "answer": None,