docling 2.39.0__tar.gz → 2.66.0__tar.gz

{docling-2.39.0 → docling-2.66.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling
-Version: 2.39.0
+Version: 2.66.0
 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
@@ -22,34 +22,40 @@ Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
 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,>=2.39.0
-Requires-Dist: docling-ibm-models<4.0.0,>=3.4.4
-Requires-Dist: docling-parse<5.0.0,>=4.0.0
+Requires-Dist: docling-core[chunking]<3.0.0,>=2.50.1
+Requires-Dist: docling-parse<5.0.0,>=4.7.0
+Requires-Dist: docling-ibm-models<4,>=3.9.1
 Requires-Dist: filetype<2.0.0,>=1.2.0
-Requires-Dist: pypdfium2<5.0.0,>=4.30.0
+Requires-Dist: pypdfium2!=4.30.1,<5.0.0,>=4.30.0
 Requires-Dist: pydantic-settings<3.0.0,>=2.3.0
 Requires-Dist: huggingface_hub<1,>=0.23
 Requires-Dist: requests<3.0.0,>=2.32.2
-Requires-Dist: easyocr<2.0,>=1.7
+Requires-Dist: ocrmac<2.0.0,>=1.0.0; sys_platform == "darwin"
+Requires-Dist: rapidocr<4.0.0,>=3.3
 Requires-Dist: certifi>=2024.7.4
 Requires-Dist: rtree<2.0.0,>=1.3.0
-Requires-Dist: typer<0.17.0,>=0.12.5
+Requires-Dist: typer<0.20.0,>=0.12.5
 Requires-Dist: python-docx<2.0.0,>=1.1.2
 Requires-Dist: python-pptx<2.0.0,>=1.0.2
 Requires-Dist: beautifulsoup4<5.0.0,>=4.12.3
 Requires-Dist: pandas<3.0.0,>=2.1.4
 Requires-Dist: marko<3.0.0,>=2.1.2
 Requires-Dist: openpyxl<4.0.0,>=3.1.5
-Requires-Dist: lxml<6.0.0,>=4.0.0
+Requires-Dist: lxml<7.0.0,>=4.0.0
 Requires-Dist: pillow<12.0.0,>=10.0.0
 Requires-Dist: tqdm<5.0.0,>=4.65.0
 Requires-Dist: pluggy<2.0.0,>=1.0.0
 Requires-Dist: pylatexenc<3.0,>=2.10
 Requires-Dist: scipy<2.0.0,>=1.6.0
+Requires-Dist: accelerate<2,>=1.0.0
+Requires-Dist: polyfactory>=2.22.2
+Provides-Extra: easyocr
+Requires-Dist: easyocr<2.0,>=1.7; extra == "easyocr"
 Provides-Extra: tesserocr
 Requires-Dist: tesserocr<3.0.0,>=2.7.1; extra == "tesserocr"
 Provides-Extra: ocrmac
@@ -57,12 +63,15 @@ 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<1.0.0,>=0.3.0; (python_version >= "3.10" and python_version < "3.14" and sys_platform == "darwin" and platform_machine == "arm64") and extra == "vlm"
+Requires-Dist: vllm<1.0.0,>=0.10.0; (python_version >= "3.10" and python_version < "3.14" and sys_platform == "linux" and platform_machine == "x86_64") and extra == "vlm"
+Requires-Dist: qwen-vl-utils>=0.0.11; 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"
+Requires-Dist: rapidocr<4.0.0,>=3.3; extra == "rapidocr"
+Requires-Dist: onnxruntime<2.0.0,>=1.7.0; python_version < "3.14" and extra == "rapidocr"
 Provides-Extra: asr
-Requires-Dist: openai-whisper>=20240930; extra == "asr"
+Requires-Dist: mlx-whisper>=0.4.3; (python_version >= "3.10" and python_version < "3.14" and sys_platform == "darwin" and platform_machine == "arm64") and extra == "asr"
+Requires-Dist: openai-whisper>=20250625; python_version < "3.14" and extra == "asr"
 Dynamic: license-file
 
 <p align="center">
@@ -88,6 +97,8 @@ Dynamic: license-file
 [![License MIT](https://img.shields.io/github/license/docling-project/docling)](https://opensource.org/licenses/MIT)
 [![PyPI Downloads](https://static.pepy.tech/badge/docling/month)](https://pepy.tech/projects/docling)
 [![Docling Actor](https://apify.com/actor-badge?actor=vancura/docling?fpr=docling)](https://apify.com/vancura/docling)
+[![Chat with Dosu](https://dosu.dev/dosu-chat-badge.svg)](https://app.dosu.dev/097760a8-135e-4789-8234-90c8837d7f1c/ask?utm_source=github)
+[![Discord](https://img.shields.io/discord/1399788921306746971?color=6A7EC2&logo=discord&logoColor=ffffff)](https://docling.ai/discord)
 [![OpenSSF Best Practices](https://www.bestpractices.dev/projects/10101/badge)](https://www.bestpractices.dev/projects/10101)
 [![LF AI & Data](https://img.shields.io/badge/LF%20AI%20%26%20Data-003778?logo=linuxfoundation&logoColor=fff&color=0094ff&labelColor=003778)](https://lfaidata.foundation/projects/)
 
@@ -95,17 +106,24 @@ Docling simplifies document processing, parsing diverse formats — including ad
 
 ## Features
 
-* 🗂️  Parsing of [multiple document formats][supported_formats] incl. PDF, DOCX, PPTX, XLSX, HTML, WAV, MP3, images (PNG, TIFF, JPEG, ...), and more
+* 🗂️ Parsing of [multiple document formats][supported_formats] incl. PDF, DOCX, PPTX, XLSX, HTML, WAV, MP3, VTT, images (PNG, TIFF, JPEG, ...), and more
 * 📑 Advanced PDF understanding incl. page layout, reading order, table structure, code, formulas, image classification, and more
 * 🧬 Unified, expressive [DoclingDocument][docling_document] representation format
-* ↪️  Various [export formats][supported_formats] and options, including Markdown, HTML, [DocTags](https://arxiv.org/abs/2503.11576) and lossless JSON
+* ↪️ Various [export formats][supported_formats] and options, including Markdown, HTML, [DocTags](https://arxiv.org/abs/2503.11576) and lossless JSON
 * 🔒 Local execution capabilities for sensitive data and air-gapped environments
 * 🤖 Plug-and-play [integrations][integrations] incl. LangChain, LlamaIndex, Crew AI & Haystack for agentic AI
 * 🔍 Extensive OCR support for scanned PDFs and images
-* 👓 Support of several Visual Language Models ([SmolDocling](https://huggingface.co/ds4sd/SmolDocling-256M-preview))
-* 🎙️  Support for Audio with Automatic Speech Recognition (ASR) models
+* 👓 Support of several Visual Language Models ([GraniteDocling](https://huggingface.co/ibm-granite/granite-docling-258M))
+* 🎙️ Audio support with Automatic Speech Recognition (ASR) models
+* 🔌 Connect to any agent using the [MCP server](https://docling-project.github.io/docling/usage/mcp/)
 * 💻 Simple and convenient CLI
 
+### What's new
+* 📤 Structured [information extraction][extraction] \[🧪 beta\]
+* 📑 New layout model (**Heron**) by default, for faster PDF parsing
+* 🔌 [MCP server](https://docling-project.github.io/docling/usage/mcp/) for agentic applications
+* 💬 Parsing of Web Video Text Tracks (WebVTT) files
+
 ### Coming soon
 
 * 📝 Metadata extraction, including title, authors, references & language
@@ -136,7 +154,7 @@ result = converter.convert(source)
 print(result.document.export_to_markdown())  # output: "## Docling Technical Report[...]"
 ```
 
-More [advanced usage options](https://docling-project.github.io/docling/usage/) are available in
+More [advanced usage options](https://docling-project.github.io/docling/usage/advanced_options/) are available in
 the docs.
 
 ## CLI
@@ -147,9 +165,9 @@ Docling has a built-in CLI to run conversions.
 docling https://arxiv.org/pdf/2206.01062
 ```
 
-You can also use 🥚[SmolDocling](https://huggingface.co/ds4sd/SmolDocling-256M-preview) and other VLMs via Docling CLI:
+You can also use 🥚[GraniteDocling](https://huggingface.co/ibm-granite/granite-docling-258M) and other VLMs via Docling CLI:
 ```bash
-docling --pipeline vlm --vlm-model smoldocling https://arxiv.org/pdf/2206.01062
+docling --pipeline vlm --vlm-model granite_docling https://arxiv.org/pdf/2206.01062
 ```
 This will use MLX acceleration on supported Apple Silicon hardware.
 
@@ -216,3 +234,4 @@ The project was started by the AI for knowledge team at IBM Research Zurich.
 [supported_formats]: https://docling-project.github.io/docling/usage/supported_formats/
 [docling_document]: https://docling-project.github.io/docling/concepts/docling_document/
 [integrations]: https://docling-project.github.io/docling/integrations/
+[extraction]: https://docling-project.github.io/docling/examples/extraction/

{docling-2.39.0 → docling-2.66.0}/README.md

@@ -21,6 +21,8 @@
 [![License MIT](https://img.shields.io/github/license/docling-project/docling)](https://opensource.org/licenses/MIT)
 [![PyPI Downloads](https://static.pepy.tech/badge/docling/month)](https://pepy.tech/projects/docling)
 [![Docling Actor](https://apify.com/actor-badge?actor=vancura/docling?fpr=docling)](https://apify.com/vancura/docling)
+[![Chat with Dosu](https://dosu.dev/dosu-chat-badge.svg)](https://app.dosu.dev/097760a8-135e-4789-8234-90c8837d7f1c/ask?utm_source=github)
+[![Discord](https://img.shields.io/discord/1399788921306746971?color=6A7EC2&logo=discord&logoColor=ffffff)](https://docling.ai/discord)
 [![OpenSSF Best Practices](https://www.bestpractices.dev/projects/10101/badge)](https://www.bestpractices.dev/projects/10101)
 [![LF AI & Data](https://img.shields.io/badge/LF%20AI%20%26%20Data-003778?logo=linuxfoundation&logoColor=fff&color=0094ff&labelColor=003778)](https://lfaidata.foundation/projects/)
 
@@ -28,17 +30,24 @@ Docling simplifies document processing, parsing diverse formats — including ad
 
 ## Features
 
-* 🗂️  Parsing of [multiple document formats][supported_formats] incl. PDF, DOCX, PPTX, XLSX, HTML, WAV, MP3, images (PNG, TIFF, JPEG, ...), and more
+* 🗂️ Parsing of [multiple document formats][supported_formats] incl. PDF, DOCX, PPTX, XLSX, HTML, WAV, MP3, VTT, images (PNG, TIFF, JPEG, ...), and more
 * 📑 Advanced PDF understanding incl. page layout, reading order, table structure, code, formulas, image classification, and more
 * 🧬 Unified, expressive [DoclingDocument][docling_document] representation format
-* ↪️  Various [export formats][supported_formats] and options, including Markdown, HTML, [DocTags](https://arxiv.org/abs/2503.11576) and lossless JSON
+* ↪️ Various [export formats][supported_formats] and options, including Markdown, HTML, [DocTags](https://arxiv.org/abs/2503.11576) and lossless JSON
 * 🔒 Local execution capabilities for sensitive data and air-gapped environments
 * 🤖 Plug-and-play [integrations][integrations] incl. LangChain, LlamaIndex, Crew AI & Haystack for agentic AI
 * 🔍 Extensive OCR support for scanned PDFs and images
-* 👓 Support of several Visual Language Models ([SmolDocling](https://huggingface.co/ds4sd/SmolDocling-256M-preview))
-* 🎙️  Support for Audio with Automatic Speech Recognition (ASR) models
+* 👓 Support of several Visual Language Models ([GraniteDocling](https://huggingface.co/ibm-granite/granite-docling-258M))
+* 🎙️ Audio support with Automatic Speech Recognition (ASR) models
+* 🔌 Connect to any agent using the [MCP server](https://docling-project.github.io/docling/usage/mcp/)
 * 💻 Simple and convenient CLI
 
+### What's new
+* 📤 Structured [information extraction][extraction] \[🧪 beta\]
+* 📑 New layout model (**Heron**) by default, for faster PDF parsing
+* 🔌 [MCP server](https://docling-project.github.io/docling/usage/mcp/) for agentic applications
+* 💬 Parsing of Web Video Text Tracks (WebVTT) files
+
 ### Coming soon
 
 * 📝 Metadata extraction, including title, authors, references & language
@@ -69,7 +78,7 @@ result = converter.convert(source)
 print(result.document.export_to_markdown())  # output: "## Docling Technical Report[...]"
 ```
 
-More [advanced usage options](https://docling-project.github.io/docling/usage/) are available in
+More [advanced usage options](https://docling-project.github.io/docling/usage/advanced_options/) are available in
 the docs.
 
 ## CLI
@@ -80,9 +89,9 @@ Docling has a built-in CLI to run conversions.
 docling https://arxiv.org/pdf/2206.01062
 ```
 
-You can also use 🥚[SmolDocling](https://huggingface.co/ds4sd/SmolDocling-256M-preview) and other VLMs via Docling CLI:
+You can also use 🥚[GraniteDocling](https://huggingface.co/ibm-granite/granite-docling-258M) and other VLMs via Docling CLI:
 ```bash
-docling --pipeline vlm --vlm-model smoldocling https://arxiv.org/pdf/2206.01062
+docling --pipeline vlm --vlm-model granite_docling https://arxiv.org/pdf/2206.01062
 ```
 This will use MLX acceleration on supported Apple Silicon hardware.
 
@@ -149,3 +158,4 @@ The project was started by the AI for knowledge team at IBM Research Zurich.
 [supported_formats]: https://docling-project.github.io/docling/usage/supported_formats/
 [docling_document]: https://docling-project.github.io/docling/concepts/docling_document/
 [integrations]: https://docling-project.github.io/docling/integrations/
+[extraction]: https://docling-project.github.io/docling/examples/extraction/

{docling-2.39.0 → docling-2.66.0}/docling/backend/abstract_backend.py

@@ -1,10 +1,16 @@
 from abc import ABC, abstractmethod
 from io import BytesIO
 from pathlib import Path
-from typing import TYPE_CHECKING, Set, Union
+from typing import TYPE_CHECKING, Union
 
 from docling_core.types.doc import DoclingDocument
 
+from docling.datamodel.backend_options import (
+    BackendOptions,
+    BaseBackendOptions,
+    DeclarativeBackendOptions,
+)
+
 if TYPE_CHECKING:
     from docling.datamodel.base_models import InputFormat
     from docling.datamodel.document import InputDocument
@@ -12,11 +18,17 @@ if TYPE_CHECKING:
 
 class AbstractDocumentBackend(ABC):
     @abstractmethod
-    def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]):
+    def __init__(
+        self,
+        in_doc: "InputDocument",
+        path_or_stream: Union[BytesIO, Path],
+        options: BaseBackendOptions = BaseBackendOptions(),
+    ):
         self.file = in_doc.file
         self.path_or_stream = path_or_stream
         self.document_hash = in_doc.document_hash
         self.input_format = in_doc.format
+        self.options = options
 
     @abstractmethod
     def is_valid(self) -> bool:
@@ -35,7 +47,7 @@ class AbstractDocumentBackend(ABC):
 
     @classmethod
     @abstractmethod
-    def supported_formats(cls) -> Set["InputFormat"]:
+    def supported_formats(cls) -> set["InputFormat"]:
         pass
 
 
@@ -58,6 +70,15 @@ class DeclarativeDocumentBackend(AbstractDocumentBackend):
     straight without a recognition pipeline.
     """
 
+    @abstractmethod
+    def __init__(
+        self,
+        in_doc: "InputDocument",
+        path_or_stream: Union[BytesIO, Path],
+        options: BackendOptions = DeclarativeBackendOptions(),
+    ) -> None:
+        super().__init__(in_doc, path_or_stream, options)
+
     @abstractmethod
     def convert(self) -> DoclingDocument:
         pass

{docling-2.39.0 → docling-2.66.0}/docling/backend/asciidoc_backend.py

@@ -2,7 +2,7 @@ import logging
 import re
 from io import BytesIO
 from pathlib import Path
-from typing import Final, Set, Union
+from typing import Final, Union
 
 from docling_core.types.doc import (
     DocItemLabel,
@@ -27,7 +27,7 @@ DEFAULT_IMAGE_HEIGHT: Final = 128
 
 
 class AsciiDocBackend(DeclarativeDocumentBackend):
-    def __init__(self, in_doc: InputDocument, path_or_stream: Union[BytesIO, Path]):
+    def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]):
         super().__init__(in_doc, path_or_stream)
 
         self.path_or_stream = path_or_stream
@@ -58,7 +58,7 @@ class AsciiDocBackend(DeclarativeDocumentBackend):
         return
 
     @classmethod
-    def supported_formats(cls) -> Set[InputFormat]:
+    def supported_formats(cls) -> set[InputFormat]:
         return {InputFormat.ASCIIDOC}
 
     def convert(self) -> DoclingDocument:
@@ -78,7 +78,7 @@ class AsciiDocBackend(DeclarativeDocumentBackend):
 
         return doc
 
-    def _parse(self, doc: DoclingDocument):  # noqa: C901
+    def _parse(self, doc: DoclingDocument):
         """
         Main function that orchestrates the parsing by yielding components:
         title, section headers, text, lists, and tables.

{docling-2.39.0 → docling-2.66.0}/docling/backend/docling_parse_v4_backend.py

@@ -12,6 +12,7 @@ from PIL import Image
 from pypdfium2 import PdfPage
 
 from docling.backend.pdf_backend import PdfDocumentBackend, PdfPageBackend
+from docling.datamodel.backend_options import PdfBackendOptions
 from docling.datamodel.base_models import Size
 from docling.utils.locks import pypdfium2_lock
 
@@ -22,15 +23,64 @@ _log = logging.getLogger(__name__)
 
 
 class DoclingParseV4PageBackend(PdfPageBackend):
-    def __init__(self, parsed_page: SegmentedPdfPage, page_obj: PdfPage):
+    def __init__(
+        self,
+        *,
+        dp_doc: PdfDocument,
+        page_obj: PdfPage,
+        page_no: int,
+        create_words: bool = True,
+        create_textlines: bool = True,
+        keep_chars: bool = False,
+        keep_lines: bool = False,
+        keep_images: bool = True,
+    ):
         self._ppage = page_obj
-        self._dpage = parsed_page
-        self.valid = parsed_page is not None
+        self._dp_doc = dp_doc
+        self._page_no = page_no
+
+        self._create_words = create_words
+        self._create_textlines = create_textlines
+
+        self._keep_chars = keep_chars
+        self._keep_lines = keep_lines
+        self._keep_images = keep_images
+
+        self._dpage: Optional[SegmentedPdfPage] = None
+        self._unloaded = False
+        self.valid = (self._ppage is not None) and (self._dp_doc is not None)
+
+    def _ensure_parsed(self) -> None:
+        if self._dpage is not None:
+            return
+
+        seg_page = self._dp_doc.get_page(
+            self._page_no + 1,
+            keep_chars=self._keep_chars,
+            keep_lines=self._keep_lines,
+            keep_bitmaps=self._keep_images,
+            create_words=self._create_words,
+            create_textlines=self._create_textlines,
+            enforce_same_font=True,
+        )
+
+        # In Docling, all TextCell instances are expected with top-left origin.
+        [
+            tc.to_top_left_origin(seg_page.dimension.height)
+            for tc in seg_page.textline_cells
+        ]
+        [tc.to_top_left_origin(seg_page.dimension.height) for tc in seg_page.char_cells]
+        [tc.to_top_left_origin(seg_page.dimension.height) for tc in seg_page.word_cells]
+
+        self._dpage = seg_page
 
     def is_valid(self) -> bool:
         return self.valid
 
     def get_text_in_rect(self, bbox: BoundingBox) -> str:
+        self._ensure_parsed()
+        assert self._dpage is not None
+
         # Find intersecting cells on the page
         text_piece = ""
         page_size = self.get_size()
@@ -56,12 +106,19 @@ class DoclingParseV4PageBackend(PdfPageBackend):
         return text_piece
 
     def get_segmented_page(self) -> Optional[SegmentedPdfPage]:
+        self._ensure_parsed()
         return self._dpage
 
     def get_text_cells(self) -> Iterable[TextCell]:
+        self._ensure_parsed()
+        assert self._dpage is not None
+
         return self._dpage.textline_cells
 
     def get_bitmap_rects(self, scale: float = 1) -> Iterable[BoundingBox]:
+        self._ensure_parsed()
+        assert self._dpage is not None
+
         AREA_THRESHOLD = 0  # 32 * 32
 
         images = self._dpage.bitmap_resources
@@ -123,18 +180,33 @@ class DoclingParseV4PageBackend(PdfPageBackend):
         # )
 
     def unload(self):
+        if not self._unloaded and self._dp_doc is not None:
+            self._dp_doc.unload_pages((self._page_no + 1, self._page_no + 2))
+            self._unloaded = True
+
         self._ppage = None
         self._dpage = None
+        self._dp_doc = None
 
 
 class DoclingParseV4DocumentBackend(PdfDocumentBackend):
-    def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]):
-        super().__init__(in_doc, path_or_stream)
-
+    def __init__(
+        self,
+        in_doc: "InputDocument",
+        path_or_stream: Union[BytesIO, Path],
+        options: PdfBackendOptions = PdfBackendOptions(),
+    ):
+        super().__init__(in_doc, path_or_stream, options)
+
+        password = (
+            self.options.password.get_secret_value() if self.options.password else None
+        )
         with pypdfium2_lock:
-            self._pdoc = pdfium.PdfDocument(self.path_or_stream)
+            self._pdoc = pdfium.PdfDocument(self.path_or_stream, password=password)
         self.parser = DoclingPdfParser(loglevel="fatal")
-        self.dp_doc: PdfDocument = self.parser.load(path_or_stream=self.path_or_stream)
+        self.dp_doc: PdfDocument = self.parser.load(
+            path_or_stream=self.path_or_stream, password=password
+        )
         success = self.dp_doc is not None
 
         if not success:
@@ -157,37 +229,32 @@ class DoclingParseV4DocumentBackend(PdfDocumentBackend):
         self, page_no: int, create_words: bool = True, create_textlines: bool = True
     ) -> DoclingParseV4PageBackend:
         with pypdfium2_lock:
-            seg_page = self.dp_doc.get_page(
-                page_no + 1,
-                create_words=create_words,
-                create_textlines=create_textlines,
-            )
-
-            # In Docling, all TextCell instances are expected with top-left origin.
-            [
-                tc.to_top_left_origin(seg_page.dimension.height)
-                for tc in seg_page.textline_cells
-            ]
-            [
-                tc.to_top_left_origin(seg_page.dimension.height)
-                for tc in seg_page.char_cells
-            ]
-            [
-                tc.to_top_left_origin(seg_page.dimension.height)
-                for tc in seg_page.word_cells
-            ]
-
-            return DoclingParseV4PageBackend(
-                seg_page,
-                self._pdoc[page_no],
-            )
+            ppage = self._pdoc[page_no]
+
+        return DoclingParseV4PageBackend(
+            dp_doc=self.dp_doc,
+            page_obj=ppage,
+            page_no=page_no,
+            create_words=create_words,
+            create_textlines=create_textlines,
+        )
 
     def is_valid(self) -> bool:
         return self.page_count() > 0
 
     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-2.66.0/docling/backend/docx/drawingml/utils.py

@@ -0,0 +1,131 @@
+import os
+import shutil
+import subprocess
+from pathlib import Path
+from tempfile import mkdtemp
+from typing import Callable, Optional
+
+import pypdfium2
+from docx.document import Document
+from PIL import Image, ImageChops
+
+
+def get_libreoffice_cmd(raise_if_unavailable: bool = False) -> Optional[str]:
+    """Return the libreoffice cmd and optionally test it."""
+
+    libreoffice_cmd = (
+        shutil.which("libreoffice")
+        or shutil.which("soffice")
+        or (
+            "/Applications/LibreOffice.app/Contents/MacOS/soffice"
+            if os.path.isfile("/Applications/LibreOffice.app/Contents/MacOS/soffice")
+            else None
+        )
+    )
+
+    if raise_if_unavailable:
+        if libreoffice_cmd is None:
+            raise RuntimeError("Libreoffice not found")
+
+        # The following test will raise if the libreoffice_cmd cannot be used
+        subprocess.run(
+            [
+                libreoffice_cmd,
+                "-h",
+            ],
+            stdout=subprocess.DEVNULL,
+            stderr=subprocess.DEVNULL,
+            check=True,
+        )
+
+    return libreoffice_cmd
+
+
+def get_docx_to_pdf_converter() -> Optional[Callable]:
+    """
+    Detects the best available DOCX to PDF tool and returns a conversion function.
+    The returned function accepts (input_path, output_path).
+    Returns None if no tool is available.
+    """
+
+    # Try LibreOffice
+    libreoffice_cmd = get_libreoffice_cmd()
+
+    if libreoffice_cmd:
+
+        def convert_with_libreoffice(input_path, output_path):
+            subprocess.run(
+                [
+                    libreoffice_cmd,
+                    "--headless",
+                    "--convert-to",
+                    "pdf",
+                    "--outdir",
+                    os.path.dirname(output_path),
+                    input_path,
+                ],
+                stdout=subprocess.DEVNULL,
+                stderr=subprocess.DEVNULL,
+                check=True,
+            )
+
+            expected_output = os.path.join(
+                os.path.dirname(output_path),
+                os.path.splitext(os.path.basename(input_path))[0] + ".pdf",
+            )
+            if expected_output != output_path:
+                os.rename(expected_output, output_path)
+
+        return convert_with_libreoffice
+
+    ## Space for other DOCX to PDF converters if available
+
+    # No tools found
+    return None
+
+
+def crop_whitespace(image: Image.Image, bg_color=None, padding=0) -> Image.Image:
+    if bg_color is None:
+        bg_color = image.getpixel((0, 0))
+
+    bg = Image.new(image.mode, image.size, bg_color)
+    diff = ImageChops.difference(image, bg)
+    bbox = diff.getbbox()
+
+    if bbox:
+        left, upper, right, lower = bbox
+        left = max(0, left - padding)
+        upper = max(0, upper - padding)
+        right = min(image.width, right + padding)
+        lower = min(image.height, lower + padding)
+        return image.crop((left, upper, right, lower))
+    else:
+        return image
+
+
+def get_pil_from_dml_docx(
+    docx: Document, converter: Optional[Callable]
+) -> Optional[Image.Image]:
+    if converter is None:
+        return None
+
+    temp_dir = Path(mkdtemp())
+    temp_docx = Path(temp_dir / "drawing_only.docx")
+    temp_pdf = Path(temp_dir / "drawing_only.pdf")
+
+    # 1) Save docx temporarily
+    docx.save(str(temp_docx))
+
+    # 2) Export to PDF
+    converter(temp_docx, temp_pdf)
+
+    # 3) Load PDF as PNG
+    pdf = pypdfium2.PdfDocument(temp_pdf)
+    page = pdf[0]
+    image = crop_whitespace(page.render(scale=2).to_pil())
+    page.close()
+    pdf.close()
+
+    shutil.rmtree(temp_dir, ignore_errors=True)
+
+    return image

{docling-2.39.0 → docling-2.66.0}/docling/backend/docx/latex/latex_dict.py

@@ -65,6 +65,11 @@ CHR_BO = {
     "\u2210": "\\coprod",
     "\u2211": "\\sum",
     "\u222b": "\\int",
+    "\u222c": "\\iint",
+    "\u222d": "\\iiint",
+    "\u222e": "\\oint",
+    "\u222f": "\\oiint",
+    "\u2230": "\\oiiint",
     "\u22c0": "\\bigwedge",
     "\u22c1": "\\bigvee",
     "\u22c2": "\\bigcap",

{docling-2.39.0 → docling-2.66.0}/docling/backend/docx/latex/omml.py

@@ -260,7 +260,15 @@ class oMath2Latex(Tag2Method):
         the fraction object
         """
         c_dict = self.process_children_dict(elm)
-        pr = c_dict["fPr"]
+        pr = c_dict.get("fPr")
+        if pr is None:
+            # Handle missing fPr element gracefully
+            _log.debug("Missing fPr element in fraction, using default formatting")
+            latex_s = F_DEFAULT
+            return latex_s.format(
+                num=c_dict.get("num"),
+                den=c_dict.get("den"),
+            )
         latex_s = get_val(pr.type, default=F_DEFAULT, store=F)
         return pr.text + latex_s.format(num=c_dict.get("num"), den=c_dict.get("den"))
 
@@ -373,7 +381,8 @@ class oMath2Latex(Tag2Method):
         bo = ""
         for stag, t, e in self.process_children_list(elm):
             if stag == "naryPr":
-                bo = get_val(t.chr, store=CHR_BO)
+                # if <m:naryPr> contains no <m:chr>, the n-ary represents an integral
+                bo = get_val(t.chr, default="\\int", store=CHR_BO)
             else:
                 res.append(t)
         return bo + BLANK.join(res)