docling 2.38.1__tar.gz → 2.40.0__tar.gz

{docling-2.38.1 → docling-2.40.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling
-Version: 2.38.1
+Version: 2.40.0
 Summary: SDK and CLI for parsing PDF, DOCX, HTML, and more, to a unified document representation for powering downstream workflows such as gen AI applications.
 Author-email: Christoph Auer <cau@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Maxim Lysak <mly@zurich.ibm.com>, Nikos Livathinos <nli@zurich.ibm.com>, Ahmed Nassar <ahn@zurich.ibm.com>, Panos Vagenas <pva@zurich.ibm.com>, Peter Staar <taa@zurich.ibm.com>
 License-Expression: MIT
@@ -26,9 +26,9 @@ Requires-Python: <4.0,>=3.9
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: pydantic<3.0.0,>=2.0.0
-Requires-Dist: docling-core[chunking]<3.0.0,>=2.29.0
-Requires-Dist: docling-ibm-models<4.0.0,>=3.4.4
+Requires-Dist: docling-core[chunking]<3.0.0,>=2.39.0
 Requires-Dist: docling-parse<5.0.0,>=4.0.0
+Requires-Dist: docling-ibm-models<4,>=3.6.0
 Requires-Dist: filetype<2.0.0,>=1.2.0
 Requires-Dist: pypdfium2<5.0.0,>=4.30.0
 Requires-Dist: pydantic-settings<3.0.0,>=2.3.0
@@ -57,7 +57,7 @@ Requires-Dist: ocrmac<2.0.0,>=1.0.0; sys_platform == "darwin" and extra == "ocrm
 Provides-Extra: vlm
 Requires-Dist: transformers<5.0.0,>=4.46.0; extra == "vlm"
 Requires-Dist: accelerate<2.0.0,>=1.2.1; extra == "vlm"
-Requires-Dist: mlx-vlm>=0.1.22; (python_version >= "3.10" and sys_platform == "darwin" and platform_machine == "arm64") and extra == "vlm"
+Requires-Dist: mlx-vlm<0.2,>=0.1.22; (python_version >= "3.10" and sys_platform == "darwin" and platform_machine == "arm64") and extra == "vlm"
 Provides-Extra: rapidocr
 Requires-Dist: rapidocr-onnxruntime<2.0.0,>=1.4.0; python_version < "3.13" and extra == "rapidocr"
 Requires-Dist: onnxruntime<2.0.0,>=1.7.0; extra == "rapidocr"

{docling-2.38.1 → docling-2.40.0}/docling/backend/docling_parse_v4_backend.py

@@ -187,7 +187,17 @@ class DoclingParseV4DocumentBackend(PdfDocumentBackend):
 
     def unload(self):
         super().unload()
-        self.dp_doc.unload()
-        with pypdfium2_lock:
-            self._pdoc.close()
-        self._pdoc = None
+        # Unload docling-parse document first
+        if self.dp_doc is not None:
+            self.dp_doc.unload()
+            self.dp_doc = None
+
+        # Then close pypdfium2 document with proper locking
+        if self._pdoc is not None:
+            with pypdfium2_lock:
+                try:
+                    self._pdoc.close()
+                except Exception:
+                    # Ignore cleanup errors
+                    pass
+            self._pdoc = None

{docling-2.38.1 → docling-2.40.0}/docling/backend/html_backend.py

@@ -17,6 +17,7 @@ from docling_core.types.doc import (
     TableData,
 )
 from docling_core.types.doc.document import ContentLayer
+from pydantic import BaseModel
 from typing_extensions import override
 
 from docling.backend.abstract_backend import DeclarativeDocumentBackend
@@ -48,6 +49,11 @@ TAGS_FOR_NODE_ITEMS: Final = [
 ]
 
 
+class _Context(BaseModel):
+    list_ordered_flag_by_ref: dict[str, bool] = {}
+    list_start_by_ref: dict[str, int] = {}
+
+
 class HTMLDocumentBackend(DeclarativeDocumentBackend):
     @override
     def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]):
@@ -59,6 +65,7 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
         self.max_levels = 10
         self.level = 0
         self.parents: dict[int, Optional[Union[DocItem, GroupItem]]] = {}
+        self.ctx = _Context()
         for i in range(self.max_levels):
             self.parents[i] = None
 
@@ -121,6 +128,7 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
             self.content_layer = (
                 ContentLayer.BODY if headers is None else ContentLayer.FURNITURE
             )
+            self.ctx = _Context()  # reset context
             self.walk(content, doc)
         else:
             raise RuntimeError(
@@ -294,28 +302,25 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
     def handle_list(self, element: Tag, doc: DoclingDocument) -> None:
         """Handles list tags (ul, ol) and their list items."""
 
-        if element.name == "ul":
-            # create a list group
-            self.parents[self.level + 1] = doc.add_group(
-                parent=self.parents[self.level],
-                name="list",
-                label=GroupLabel.LIST,
-                content_layer=self.content_layer,
-            )
-        elif element.name == "ol":
+        start: Optional[int] = None
+        if is_ordered := element.name == "ol":
             start_attr = element.get("start")
-            start: int = (
-                int(start_attr)
-                if isinstance(start_attr, str) and start_attr.isnumeric()
-                else 1
-            )
-            # create a list group
-            self.parents[self.level + 1] = doc.add_group(
-                parent=self.parents[self.level],
-                name="ordered list" + (f" start {start}" if start != 1 else ""),
-                label=GroupLabel.ORDERED_LIST,
-                content_layer=self.content_layer,
-            )
+            if isinstance(start_attr, str) and start_attr.isnumeric():
+                start = int(start_attr)
+            name = "ordered list" + (f" start {start}" if start is not None else "")
+        else:
+            name = "list"
+        # create a list group
+        list_group = doc.add_list_group(
+            name=name,
+            parent=self.parents[self.level],
+            content_layer=self.content_layer,
+        )
+        self.parents[self.level + 1] = list_group
+        self.ctx.list_ordered_flag_by_ref[list_group.self_ref] = is_ordered
+        if is_ordered and start is not None:
+            self.ctx.list_start_by_ref[list_group.self_ref] = start
+
         self.level += 1
 
         self.walk(element, doc)
@@ -331,16 +336,11 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
         if parent is None:
             _log.debug(f"list-item has no parent in DoclingDocument: {element}")
             return
-        parent_label: str = parent.label
-        index_in_list = len(parent.children) + 1
-        if (
-            parent_label == GroupLabel.ORDERED_LIST
-            and isinstance(parent, GroupItem)
-            and parent.name
-        ):
-            start_in_list: str = parent.name.split(" ")[-1]
-            start: int = int(start_in_list) if start_in_list.isnumeric() else 1
-            index_in_list += start - 1
+        enumerated = self.ctx.list_ordered_flag_by_ref.get(parent.self_ref, False)
+        if enumerated and (start := self.ctx.list_start_by_ref.get(parent.self_ref)):
+            marker = f"{start + len(parent.children)}."
+        else:
+            marker = ""
 
         if nested_list:
             # Text in list item can be hidden within hierarchy, hence
@@ -350,12 +350,6 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
             text = text.replace("\n", "").replace("\r", "")
             text = " ".join(text.split()).strip()
 
-            marker = ""
-            enumerated = False
-            if parent_label == GroupLabel.ORDERED_LIST:
-                marker = str(index_in_list)
-                enumerated = True
-
             if len(text) > 0:
                 # create a list-item
                 self.parents[self.level + 1] = doc.add_list_item(
@@ -375,11 +369,6 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
         elif element.text.strip():
             text = element.text.strip()
 
-            marker = ""
-            enumerated = False
-            if parent_label == GroupLabel.ORDERED_LIST:
-                marker = f"{index_in_list!s}."
-                enumerated = True
             doc.add_list_item(
                 text=text,
                 enumerated=enumerated,

{docling-2.38.1 → docling-2.40.0}/docling/backend/md_backend.py

@@ -14,13 +14,12 @@ from docling_core.types.doc import (
     DocItemLabel,
     DoclingDocument,
     DocumentOrigin,
-    GroupLabel,
     NodeItem,
     TableCell,
     TableData,
     TextItem,
 )
-from docling_core.types.doc.document import Formatting, OrderedList, UnorderedList
+from docling_core.types.doc.document import Formatting
 from marko import Markdown
 from pydantic import AnyUrl, BaseModel, Field, TypeAdapter
 from typing_extensions import Annotated
@@ -51,6 +50,7 @@ class _HeadingCreationPayload(BaseModel):
 
 class _ListItemCreationPayload(BaseModel):
     kind: Literal["list_item"] = "list_item"
+    enumerated: bool
 
 
 _CreationPayload = Annotated[
@@ -187,15 +187,13 @@ class MarkdownDocumentBackend(DeclarativeDocumentBackend):
         doc: DoclingDocument,
         parent_item: Optional[NodeItem],
         text: str,
+        enumerated: bool,
         formatting: Optional[Formatting] = None,
         hyperlink: Optional[Union[AnyUrl, Path]] = None,
     ):
-        if not isinstance(parent_item, (OrderedList, UnorderedList)):
-            _log.warning("ListItem would have not had a list parent, adding one.")
-            parent_item = doc.add_unordered_list(parent=parent_item)
         item = doc.add_list_item(
             text=text,
-            enumerated=(isinstance(parent_item, OrderedList)),
+            enumerated=enumerated,
             parent=parent_item,
             formatting=formatting,
             hyperlink=hyperlink,
@@ -238,6 +236,7 @@ class MarkdownDocumentBackend(DeclarativeDocumentBackend):
         creation_stack: list[
             _CreationPayload
         ],  # stack for lazy item creation triggered deep in marko's AST (on RawText)
+        list_ordered_flag_by_ref: dict[str, bool],
         parent_item: Optional[NodeItem] = None,
         formatting: Optional[Formatting] = None,
         hyperlink: Optional[Union[AnyUrl, Path]] = None,
@@ -275,10 +274,8 @@ class MarkdownDocumentBackend(DeclarativeDocumentBackend):
             self._close_table(doc)
             _log.debug(f" - List {'ordered' if element.ordered else 'unordered'}")
             if has_non_empty_list_items:
-                label = GroupLabel.ORDERED_LIST if element.ordered else GroupLabel.LIST
-                parent_item = doc.add_group(
-                    label=label, name="list", parent=parent_item
-                )
+                parent_item = doc.add_list_group(name="list", parent=parent_item)
+                list_ordered_flag_by_ref[parent_item.self_ref] = element.ordered
 
         elif (
             isinstance(element, marko.block.ListItem)
@@ -289,16 +286,22 @@ class MarkdownDocumentBackend(DeclarativeDocumentBackend):
             self._close_table(doc)
             _log.debug(" - List item")
 
+            enumerated = (
+                list_ordered_flag_by_ref.get(parent_item.self_ref, False)
+                if parent_item
+                else False
+            )
             if len(child.children) > 1:  # inline group will be created further down
                 parent_item = self._create_list_item(
                     doc=doc,
                     parent_item=parent_item,
                     text="",
+                    enumerated=enumerated,
                     formatting=formatting,
                     hyperlink=hyperlink,
                 )
             else:
-                creation_stack.append(_ListItemCreationPayload())
+                creation_stack.append(_ListItemCreationPayload(enumerated=enumerated))
 
         elif isinstance(element, marko.inline.Image):
             self._close_table(doc)
@@ -335,7 +338,7 @@ class MarkdownDocumentBackend(DeclarativeDocumentBackend):
             _log.debug(f" - Paragraph (raw text): {element.children}")
             snippet_text = element.children.strip()
             # Detect start of the table:
-            if "|" in snippet_text:
+            if "|" in snippet_text or self.in_table:
                 # most likely part of the markdown table
                 self.in_table = True
                 if len(self.md_table_buffer) > 0:
@@ -349,10 +352,18 @@ class MarkdownDocumentBackend(DeclarativeDocumentBackend):
                     while len(creation_stack) > 0:
                         to_create = creation_stack.pop()
                         if isinstance(to_create, _ListItemCreationPayload):
+                            enumerated = (
+                                list_ordered_flag_by_ref.get(
+                                    parent_item.self_ref, False
+                                )
+                                if parent_item
+                                else False
+                            )
                             parent_item = self._create_list_item(
                                 doc=doc,
                                 parent_item=parent_item,
                                 text=snippet_text,
+                                enumerated=enumerated,
                                 formatting=formatting,
                                 hyperlink=hyperlink,
                             )
@@ -453,6 +464,7 @@ class MarkdownDocumentBackend(DeclarativeDocumentBackend):
                     doc=doc,
                     visited=visited,
                     creation_stack=creation_stack,
+                    list_ordered_flag_by_ref=list_ordered_flag_by_ref,
                     parent_item=parent_item,
                     formatting=formatting,
                     hyperlink=hyperlink,
@@ -497,6 +509,7 @@ class MarkdownDocumentBackend(DeclarativeDocumentBackend):
                 parent_item=None,
                 visited=set(),
                 creation_stack=[],
+                list_ordered_flag_by_ref={},
             )
             self._close_table(doc=doc)  # handle any last hanging table
 

{docling-2.38.1 → docling-2.40.0}/docling/backend/msexcel_backend.py

@@ -337,10 +337,17 @@ class MsExcelDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentBacken
         # Collect the data within the bounds
         data = []
         visited_cells: set[tuple[int, int]] = set()
-        for ri in range(start_row, max_row + 1):
-            for rj in range(start_col, max_col + 1):
-                cell = sheet.cell(row=ri + 1, column=rj + 1)  # 1-based indexing
-
+        for ri, row in enumerate(
+            sheet.iter_rows(
+                min_row=start_row + 1,  # start_row is 0-based but iter_rows is 1-based
+                max_row=max_row + 1,
+                min_col=start_col + 1,
+                max_col=max_col + 1,
+                values_only=False,
+            ),
+            start_row,
+        ):
+            for rj, cell in enumerate(row, start_col):
                 # Check if the cell belongs to a merged range
                 row_span = 1
                 col_span = 1
@@ -397,10 +404,16 @@ class MsExcelDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentBacken
         """
         max_row: int = start_row
 
-        while max_row < sheet.max_row - 1:
-            # Get the cell value or check if it is part of a merged cell
-            cell = sheet.cell(row=max_row + 2, column=start_col + 1)
-
+        for ri, (cell,) in enumerate(
+            sheet.iter_rows(
+                min_row=start_row + 2,
+                max_row=sheet.max_row,
+                min_col=start_col + 1,
+                max_col=start_col + 1,
+                values_only=False,
+            ),
+            start_row + 1,
+        ):
             # Check if the cell is part of a merged range
             merged_range = next(
                 (mr for mr in sheet.merged_cells.ranges if cell.coordinate in mr),
@@ -414,7 +427,7 @@ class MsExcelDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentBacken
             if merged_range:
                 max_row = max(max_row, merged_range.max_row - 1)
             else:
-                max_row += 1
+                max_row = ri
 
         return max_row
 
@@ -433,10 +446,16 @@ class MsExcelDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentBacken
         """
         max_col: int = start_col
 
-        while max_col < sheet.max_column - 1:
-            # Get the cell value or check if it is part of a merged cell
-            cell = sheet.cell(row=start_row + 1, column=max_col + 2)
-
+        for rj, (cell,) in enumerate(
+            sheet.iter_cols(
+                min_row=start_row + 1,
+                max_row=start_row + 1,
+                min_col=start_col + 2,
+                max_col=sheet.max_column,
+                values_only=False,
+            ),
+            start_col + 1,
+        ):
             # Check if the cell is part of a merged range
             merged_range = next(
                 (mr for mr in sheet.merged_cells.ranges if cell.coordinate in mr),
@@ -450,7 +469,7 @@ class MsExcelDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentBacken
             if merged_range:
                 max_col = max(max_col, merged_range.max_col - 1)
             else:
-                max_col += 1
+                max_col = rj
 
         return max_col
 

{docling-2.38.1 → docling-2.40.0}/docling/backend/mspowerpoint_backend.py

@@ -121,7 +121,9 @@ class MsPowerpointDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentB
 
         return prov
 
-    def handle_text_elements(self, shape, parent_slide, slide_ind, doc, slide_size):
+    def handle_text_elements(
+        self, shape, parent_slide, slide_ind, doc: DoclingDocument, slide_size
+    ):
         is_list_group_created = False
         enum_list_item_value = 0
         new_list = None
@@ -165,10 +167,7 @@ class MsPowerpointDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentB
                 enumerated = bullet_type == "Numbered"
 
                 if not is_list_group_created:
-                    new_list = doc.add_group(
-                        label=GroupLabel.ORDERED_LIST
-                        if enumerated
-                        else GroupLabel.LIST,
+                    new_list = doc.add_list_group(
                         name="list",
                         parent=parent_slide,
                     )

{docling-2.38.1 → docling-2.40.0}/docling/backend/msword_backend.py

@@ -10,11 +10,12 @@ from docling_core.types.doc import (
     DocumentOrigin,
     GroupLabel,
     ImageRef,
+    ListGroup,
     NodeItem,
     TableCell,
     TableData,
 )
-from docling_core.types.doc.document import Formatting, OrderedList, UnorderedList
+from docling_core.types.doc.document import Formatting
 from docx import Document
 from docx.document import Document as DocxDocument
 from docx.oxml.table import CT_Tc
@@ -688,7 +689,7 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         paragraph_elements: list,
     ) -> Optional[NodeItem]:
         return (
-            doc.add_group(label=GroupLabel.INLINE, parent=prev_parent)
+            doc.add_inline_group(parent=prev_parent)
             if len(paragraph_elements) > 1
             else prev_parent
         )
@@ -781,9 +782,7 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
             else:
                 # Inline equation
                 level = self._get_level()
-                inline_equation = doc.add_group(
-                    label=GroupLabel.INLINE, parent=self.parents[level - 1]
-                )
+                inline_equation = doc.add_inline_group(parent=self.parents[level - 1])
                 text_tmp = text
                 for eq in equations:
                     if len(text_tmp) == 0:
@@ -931,18 +930,22 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         level: int,
     ) -> None:
         # This should not happen by construction
-        if not isinstance(self.parents[level], (OrderedList, UnorderedList)):
+        if not isinstance(self.parents[level], ListGroup):
+            return
+        if not elements:
             return
+
         if len(elements) == 1:
             text, format, hyperlink = elements[0]
-            doc.add_list_item(
-                marker=marker,
-                enumerated=enumerated,
-                parent=self.parents[level],
-                text=text,
-                formatting=format,
-                hyperlink=hyperlink,
-            )
+            if text:
+                doc.add_list_item(
+                    marker=marker,
+                    enumerated=enumerated,
+                    parent=self.parents[level],
+                    text=text,
+                    formatting=format,
+                    hyperlink=hyperlink,
+                )
         else:
             new_item = doc.add_list_item(
                 marker=marker,
@@ -950,15 +953,16 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
                 parent=self.parents[level],
                 text="",
             )
-            new_parent = doc.add_group(label=GroupLabel.INLINE, parent=new_item)
+            new_parent = doc.add_inline_group(parent=new_item)
             for text, format, hyperlink in elements:
-                doc.add_text(
-                    label=DocItemLabel.TEXT,
-                    parent=new_parent,
-                    text=text,
-                    formatting=format,
-                    hyperlink=hyperlink,
-                )
+                if text:
+                    doc.add_text(
+                        label=DocItemLabel.TEXT,
+                        parent=new_parent,
+                        text=text,
+                        formatting=format,
+                        hyperlink=hyperlink,
+                    )
 
     def _add_list_item(
         self,
@@ -979,8 +983,8 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         if self._prev_numid() is None:  # Open new list
             self.level_at_new_list = level
 
-            self.parents[level] = doc.add_group(
-                label=GroupLabel.LIST, name="list", parent=self.parents[level - 1]
+            self.parents[level] = doc.add_list_group(
+                name="list", parent=self.parents[level - 1]
             )
 
             # Set marker and enumerated arguments if this is an enumeration element.
@@ -1001,19 +1005,10 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
                 self.level_at_new_list + prev_indent + 1,
                 self.level_at_new_list + ilevel + 1,
             ):
-                # Determine if this is an unordered list or an ordered list.
-                # Set GroupLabel.ORDERED_LIST when it fits.
                 self.listIter = 0
-                if is_numbered:
-                    self.parents[i] = doc.add_group(
-                        label=GroupLabel.ORDERED_LIST,
-                        name="list",
-                        parent=self.parents[i - 1],
-                    )
-                else:
-                    self.parents[i] = doc.add_group(
-                        label=GroupLabel.LIST, name="list", parent=self.parents[i - 1]
-                    )
+                self.parents[i] = doc.add_list_group(
+                    name="list", parent=self.parents[i - 1]
+                )
 
             # TODO: Set marker and enumerated arguments if this is an enumeration element.
             self.listIter += 1

{docling-2.38.1 → docling-2.40.0}/docling/datamodel/pipeline_options.py

@@ -1,4 +1,5 @@
 import logging
+from datetime import datetime
 from enum import Enum
 from pathlib import Path
 from typing import Any, ClassVar, Dict, List, Literal, Optional, Union
@@ -265,6 +266,12 @@ class VlmPipelineOptions(PaginatedPipelineOptions):
     )
 
 
+class LayoutOptions(BaseModel):
+    """Options for layout processing."""
+
+    create_orphan_clusters: bool = True  # Whether to create clusters for orphaned cells
+
+
 class AsrPipelineOptions(PipelineOptions):
     asr_options: Union[InlineAsrOptions] = asr_model_specs.WHISPER_TINY
     artifacts_path: Optional[Union[Path, str]] = None
@@ -289,6 +296,7 @@ class PdfPipelineOptions(PaginatedPipelineOptions):
     picture_description_options: PictureDescriptionBaseOptions = (
         smolvlm_picture_description
     )
+    layout_options: LayoutOptions = LayoutOptions()
 
     images_scale: float = 1.0
     generate_page_images: bool = False

{docling-2.38.1 → docling-2.40.0}/docling/models/base_ocr_model.py

@@ -3,14 +3,13 @@ import logging
 from abc import abstractmethod
 from collections.abc import Iterable
 from pathlib import Path
-from typing import List, Optional, Type
+from typing import TYPE_CHECKING, List, Optional, Type
 
 import numpy as np
 from docling_core.types.doc import BoundingBox, CoordOrigin
 from docling_core.types.doc.page import TextCell
 from PIL import Image, ImageDraw
 from rtree import index
-from scipy.ndimage import binary_dilation, find_objects, label
 
 from docling.datamodel.accelerator_options import AcceleratorOptions
 from docling.datamodel.base_models import Page
@@ -31,11 +30,16 @@ class BaseOcrModel(BasePageModel, BaseModelWithOptions):
         options: OcrOptions,
         accelerator_options: AcceleratorOptions,
     ):
+        # Make sure any delay/error from import occurs on ocr model init and not first use
+        from scipy.ndimage import binary_dilation, find_objects, label
+
         self.enabled = enabled
         self.options = options
 
     # Computes the optimum amount and coordinates of rectangles to OCR on a given page
     def get_ocr_rects(self, page: Page) -> List[BoundingBox]:
+        from scipy.ndimage import binary_dilation, find_objects, label
+
         BITMAP_COVERAGE_TRESHOLD = 0.75
         assert page.size is not None
 

{docling-2.38.1 → docling-2.40.0}/docling/models/layout_model.py

@@ -7,12 +7,12 @@ from typing import Optional
 
 import numpy as np
 from docling_core.types.doc import DocItemLabel
-from docling_ibm_models.layoutmodel.layout_predictor import LayoutPredictor
 from PIL import Image
 
 from docling.datamodel.accelerator_options import AcceleratorOptions
 from docling.datamodel.base_models import BoundingBox, Cluster, LayoutPrediction, Page
 from docling.datamodel.document import ConversionResult
+from docling.datamodel.pipeline_options import LayoutOptions
 from docling.datamodel.settings import settings
 from docling.models.base_model import BasePageModel
 from docling.models.utils.hf_model_download import download_hf_model
@@ -49,8 +49,15 @@ class LayoutModel(BasePageModel):
     CONTAINER_LABELS = [DocItemLabel.FORM, DocItemLabel.KEY_VALUE_REGION]
 
     def __init__(
-        self, artifacts_path: Optional[Path], accelerator_options: AcceleratorOptions
+        self,
+        artifacts_path: Optional[Path],
+        accelerator_options: AcceleratorOptions,
+        options: LayoutOptions,
     ):
+        from docling_ibm_models.layoutmodel.layout_predictor import LayoutPredictor
+
+        self.options = options
+
         device = decide_device(accelerator_options.device)
 
         if artifacts_path is None:
@@ -176,7 +183,7 @@ class LayoutModel(BasePageModel):
                     # Apply postprocessing
 
                     processed_clusters, processed_cells = LayoutPostprocessor(
-                        page, clusters
+                        page, clusters, self.options
                     ).postprocess()
                     # Note: LayoutPostprocessor updates page.cells and page.parsed_page internally