docling-core 2.31.1__py3-none-any.whl → 2.32.0__py3-none-any.whl

docling_core/transforms/chunker/hybrid_chunker.py

@@ -9,6 +9,7 @@ from functools import cached_property
 from typing import Any, Iterable, Iterator, Optional, Union
 
 from pydantic import BaseModel, ConfigDict, Field, computed_field, model_validator
+from transformers import PreTrainedTokenizerBase
 
 from docling_core.transforms.chunker.hierarchical_chunker import (
     ChunkingSerializerProvider,
@@ -70,23 +71,36 @@ class HybridChunker(BaseChunker):
     @model_validator(mode="before")
     @classmethod
     def _patch(cls, data: Any) -> Any:
-        if isinstance(data, dict) and (tokenizer := data.get("tokenizer")):
+        if isinstance(data, dict):
+            tokenizer = data.get("tokenizer")
             max_tokens = data.get("max_tokens")
-            if isinstance(tokenizer, BaseTokenizer):
-                pass
-            else:
+            if not isinstance(tokenizer, BaseTokenizer) and (
+                # some legacy param passed:
+                tokenizer is not None
+                or max_tokens is not None
+            ):
                 from docling_core.transforms.chunker.tokenizer.huggingface import (
                     HuggingFaceTokenizer,
                 )
 
+                warnings.warn(
+                    "Deprecated initialization parameter types for HybridChunker. "
+                    "For updated usage check out "
+                    "https://docling-project.github.io/docling/examples/hybrid_chunking/",
+                    DeprecationWarning,
+                )
+
                 if isinstance(tokenizer, str):
                     data["tokenizer"] = HuggingFaceTokenizer.from_pretrained(
                         model_name=tokenizer,
                         max_tokens=max_tokens,
                     )
-                else:
-                    # migrate previous HF-based tokenizers
-                    kwargs = {"tokenizer": tokenizer}
+                elif tokenizer is None or isinstance(
+                    tokenizer, PreTrainedTokenizerBase
+                ):
+                    kwargs = {
+                        "tokenizer": tokenizer or _get_default_tokenizer().tokenizer
+                    }
                     if max_tokens is not None:
                         kwargs["max_tokens"] = max_tokens
                     data["tokenizer"] = HuggingFaceTokenizer(**kwargs)
@@ -141,7 +155,6 @@ class HybridChunker(BaseChunker):
         meta = DocMeta(
             doc_items=doc_items,
             headings=doc_chunk.meta.headings,
-            captions=doc_chunk.meta.captions,
             origin=doc_chunk.meta.origin,
         )
         window_text = (
@@ -220,7 +233,9 @@ class HybridChunker(BaseChunker):
             )
             if available_length <= 0:
                 warnings.warn(
-                    f"Headers and captions for this chunk are longer than the total amount of size for the chunk, chunk will be ignored: {doc_chunk.text=}"  # noqa
+                    "Headers and captions for this chunk are longer than the total "
+                    "amount of size for the chunk, chunk will be ignored: "
+                    f"{doc_chunk.text=}"
                 )
                 return []
             text = doc_chunk.text
@@ -235,10 +250,10 @@ class HybridChunker(BaseChunker):
         num_chunks = len(chunks)
         while window_end < num_chunks:
             chunk = chunks[window_end]
-            headings_and_captions = (chunk.meta.headings, chunk.meta.captions)
+            headings = chunk.meta.headings
             ready_to_append = False
             if window_start == window_end:
-                current_headings_and_captions = headings_and_captions
+                current_headings = headings
                 window_end += 1
                 first_chunk_of_window = chunk
             else:
@@ -249,13 +264,12 @@ class HybridChunker(BaseChunker):
                     text=self.delim.join([chk.text for chk in chks]),
                     meta=DocMeta(
                         doc_items=doc_items,
-                        headings=current_headings_and_captions[0],
-                        captions=current_headings_and_captions[1],
+                        headings=current_headings,
                         origin=chunk.meta.origin,
                     ),
                 )
                 if (
-                    headings_and_captions == current_headings_and_captions
+                    headings == current_headings
                     and self._count_chunk_tokens(doc_chunk=candidate) <= self.max_tokens
                 ):
                     # there is room to include the new chunk so add it to the window and

docling_core/transforms/chunker/tokenizer/huggingface.py

@@ -1,10 +1,11 @@
 """HuggingFace tokenization."""
 
-import sys
+import json
 from os import PathLike
 from typing import Optional, Union
 
-from pydantic import ConfigDict, PositiveInt, TypeAdapter, model_validator
+from huggingface_hub import hf_hub_download
+from pydantic import ConfigDict, model_validator
 from typing_extensions import Self
 
 from docling_core.transforms.chunker.tokenizer.base import BaseTokenizer
@@ -28,16 +29,23 @@ class HuggingFaceTokenizer(BaseTokenizer):
 
     @model_validator(mode="after")
     def _patch(self) -> Self:
-        if hasattr(self.tokenizer, "model_max_length"):
-            model_max_tokens: PositiveInt = TypeAdapter(PositiveInt).validate_python(
-                self.tokenizer.model_max_length
-            )
-            user_max_tokens = self.max_tokens or sys.maxsize
-            self.max_tokens = min(model_max_tokens, user_max_tokens)
-        elif self.max_tokens is None:
-            raise ValueError(
-                "max_tokens must be defined as model does not define model_max_length"
-            )
+        if self.max_tokens is None:
+            try:
+                # try to use SentenceTransformers-specific config as that seems to be
+                # reliable (whenever available)
+                config_name = "sentence_bert_config.json"
+                config_path = hf_hub_download(
+                    repo_id=self.tokenizer.name_or_path,
+                    filename=config_name,
+                )
+                with open(config_path) as f:
+                    data = json.load(f)
+                self.max_tokens = int(data["max_seq_length"])
+            except Exception as e:
+                raise RuntimeError(
+                    "max_tokens could not be determined automatically; please set "
+                    "explicitly."
+                ) from e
         return self
 
     def count_tokens(self, text: str):

docling_core/transforms/serializer/common.py

@@ -11,7 +11,7 @@ from functools import cached_property
 from pathlib import Path
 from typing import Any, Iterable, Optional, Tuple, Union
 
-from pydantic import AnyUrl, BaseModel, NonNegativeInt, computed_field
+from pydantic import AnyUrl, BaseModel, ConfigDict, NonNegativeInt, computed_field
 from typing_extensions import Self, override
 
 from docling_core.transforms.serializer.base import (
@@ -39,7 +39,11 @@ from docling_core.types.doc.document import (
     KeyValueItem,
     NodeItem,
     OrderedList,
+    PictureClassificationData,
+    PictureDataType,
+    PictureDescriptionData,
     PictureItem,
+    PictureMoleculeData,
     TableItem,
     TextItem,
     UnorderedList,
@@ -118,6 +122,23 @@ def _iterate_items(
         yield item
 
 
+def _get_picture_annotation_text(annotation: PictureDataType) -> Optional[str]:
+    result = None
+    if isinstance(annotation, PictureClassificationData):
+        predicted_class = (
+            annotation.predicted_classes[0].class_name
+            if annotation.predicted_classes
+            else None
+        )
+        if predicted_class is not None:
+            result = predicted_class.replace("_", " ")
+    elif isinstance(annotation, PictureDescriptionData):
+        result = annotation.text
+    elif isinstance(annotation, PictureMoleculeData):
+        result = annotation.smi
+    return result
+
+
 def create_ser_result(
     *,
     text: str = "",
@@ -176,11 +197,7 @@ class CommonParams(BaseModel):
 class DocSerializer(BaseModel, BaseDocSerializer):
     """Class for document serializers."""
 
-    class Config:
-        """Pydantic config."""
-
-        arbitrary_types_allowed = True
-        extra = "forbid"
+    model_config = ConfigDict(arbitrary_types_allowed=True, extra="forbid")
 
     doc: DoclingDocument
 

docling_core/transforms/serializer/html.py

@@ -35,6 +35,7 @@ from docling_core.transforms.serializer.base import (
 from docling_core.transforms.serializer.common import (
     CommonParams,
     DocSerializer,
+    _get_picture_annotation_text,
     create_ser_result,
 )
 from docling_core.transforms.serializer.html_styles import (
@@ -110,6 +111,8 @@ class HTMLParams(CommonParams):
     # Enable charts to be printed into HTML as tables
     enable_chart_tables: bool = True
 
+    include_annotations: bool = True
+
 
 class HTMLTextSerializer(BaseModel, BaseTextSerializer):
     """HTML-specific text item serializer."""
@@ -943,18 +946,46 @@ class HTMLDocSerializer(DocSerializer):
         params = self.params.merge_with_patch(patch=kwargs)
         results: list[SerializationResult] = []
         text_res = ""
+        excluded_refs = self.get_excluded_refs(**kwargs)
+
         if DocItemLabel.CAPTION in params.labels:
-            results = [
-                create_ser_result(text=it.text, span_source=it)
-                for cap in item.captions
-                if isinstance(it := cap.resolve(self.doc), TextItem)
-                and it.self_ref not in self.get_excluded_refs(**kwargs)
-            ]
-            text_res = params.caption_delim.join([r.text for r in results])
-            if text_res:
-                text_dir = get_text_direction(text_res)
-                dir_str = f' dir="{text_dir}"' if text_dir == "rtl" else ""
-                text_res = f"<{tag}{dir_str}>{html.escape(text_res)}</{tag}>"
+            for cap in item.captions:
+                if (
+                    isinstance(it := cap.resolve(self.doc), TextItem)
+                    and it.self_ref not in excluded_refs
+                ):
+                    text_cap = it.text
+                    text_dir = get_text_direction(text_cap)
+                    dir_str = f' dir="{text_dir}"' if text_dir == "rtl" else ""
+                    cap_ser_res = create_ser_result(
+                        text=(
+                            f'<div class="caption"{dir_str}>'
+                            f"{html.escape(text_cap)}"
+                            f"</div>"
+                        ),
+                        span_source=it,
+                    )
+                    results.append(cap_ser_res)
+
+        if params.include_annotations and item.self_ref not in excluded_refs:
+            if isinstance(item, PictureItem):
+                for ann in item.annotations:
+                    if ann_text := _get_picture_annotation_text(annotation=ann):
+                        text_dir = get_text_direction(ann_text)
+                        dir_str = f' dir="{text_dir}"' if text_dir == "rtl" else ""
+                        ann_ser_res = create_ser_result(
+                            text=(
+                                f'<div data-annotation-kind="{ann.kind}"{dir_str}>'
+                                f"{html.escape(ann_text)}"
+                                f"</div>"
+                            ),
+                            span_source=item,
+                        )
+                        results.append(ann_ser_res)
+
+        text_res = params.caption_delim.join([r.text for r in results])
+        if text_res:
+            text_res = f"<{tag}>{text_res}</{tag}>"
         return create_ser_result(text=text_res, span_source=results)
 
     def _generate_head(self) -> str:

docling_core/transforms/serializer/markdown.py

@@ -29,6 +29,7 @@ from docling_core.transforms.serializer.base import (
 from docling_core.transforms.serializer.common import (
     CommonParams,
     DocSerializer,
+    _get_picture_annotation_text,
     _PageBreakSerResult,
     create_ser_result,
 )
@@ -69,6 +70,8 @@ class MarkdownParams(CommonParams):
     page_break_placeholder: Optional[str] = None  # e.g. "<!-- page break -->"
     escape_underscores: bool = True
     escape_html: bool = True
+    include_annotations: bool = True
+    mark_annotations: bool = False
 
 
 class MarkdownTextSerializer(BaseModel, BaseTextSerializer):
@@ -210,6 +213,24 @@ class MarkdownPictureSerializer(BasePictureSerializer):
             res_parts.append(cap_res)
 
         if item.self_ref not in doc_serializer.get_excluded_refs(**kwargs):
+            if params.include_annotations:
+
+                for ann in item.annotations:
+                    if ann_text := _get_picture_annotation_text(annotation=ann):
+                        ann_ser_res = create_ser_result(
+                            text=(
+                                (
+                                    f'<!--<annotation kind="{ann.kind}">-->'
+                                    f"{ann_text}"
+                                    f"<!--<annotation/>-->"
+                                )
+                                if params.mark_annotations
+                                else ann_text
+                            ),
+                            span_source=item,
+                        )
+                        res_parts.append(ann_ser_res)
+
             img_res = self._serialize_image_part(
                 item=item,
                 doc=doc,

docling_core/types/doc/document.py

@@ -11,6 +11,7 @@ import os
 import re
 import sys
 import typing
+import warnings
 from enum import Enum
 from io import BytesIO
 from pathlib import Path
@@ -2924,6 +2925,7 @@ class DoclingDocument(BaseModel):
         page_no: Optional[int] = None,
         included_content_layers: Optional[set[ContentLayer]] = None,
         page_break_placeholder: Optional[str] = None,
+        include_annotations: bool = True,
     ):
         """Save to markdown."""
         if isinstance(filename, str):
@@ -2951,6 +2953,7 @@ class DoclingDocument(BaseModel):
             page_no=page_no,
             included_content_layers=included_content_layers,
             page_break_placeholder=page_break_placeholder,
+            include_annotations=include_annotations,
         )
 
         with open(filename, "w", encoding="utf-8") as fw:
@@ -2972,6 +2975,8 @@ class DoclingDocument(BaseModel):
         page_no: Optional[int] = None,
         included_content_layers: Optional[set[ContentLayer]] = None,
         page_break_placeholder: Optional[str] = None,  # e.g. "<!-- page break -->",
+        include_annotations: bool = True,
+        mark_annotations: bool = False,
     ) -> str:
         r"""Serialize to Markdown.
 
@@ -2991,9 +2996,9 @@ class DoclingDocument(BaseModel):
         :type labels: Optional[set[DocItemLabel]] = None
         :param strict_text: Deprecated.
         :type strict_text: bool = False
-        :param escaping_underscores: bool: Whether to escape underscores in the
+        :param escape_underscores: bool: Whether to escape underscores in the
             text content of the document. (Default value = True).
-        :type escaping_underscores: bool = True
+        :type escape_underscores: bool = True
         :param image_placeholder: The placeholder to include to position
             images in the markdown. (Default value = "\<!-- image --\>").
         :type image_placeholder: str = "<!-- image -->"
@@ -3009,6 +3014,12 @@ class DoclingDocument(BaseModel):
         :param page_break_placeholder: The placeholder to include for marking page
             breaks. None means no page break placeholder will be used.
         :type page_break_placeholder: Optional[str] = None
+        :param include_annotations: bool: Whether to include annotations in the export.
+            (Default value = True).
+        :type include_annotations: bool = True
+        :param mark_annotations: bool: Whether to mark annotations in the export; only
+            relevant if include_annotations is True. (Default value = False).
+        :type mark_annotations: bool = False
         :returns: The exported Markdown representation.
         :rtype: str
         """
@@ -3038,6 +3049,8 @@ class DoclingDocument(BaseModel):
                 indent=indent,
                 wrap_width=text_width if text_width > 0 else None,
                 page_break_placeholder=page_break_placeholder,
+                include_annotations=include_annotations,
+                mark_annotations=mark_annotations,
             ),
         )
         ser_res = serializer.serialize()
@@ -3087,6 +3100,7 @@ class DoclingDocument(BaseModel):
         html_head: str = "null",  # should be deprecated
         included_content_layers: Optional[set[ContentLayer]] = None,
         split_page_view: bool = False,
+        include_annotations: bool = True,
     ):
         """Save to HTML."""
         if isinstance(filename, str):
@@ -3112,6 +3126,7 @@ class DoclingDocument(BaseModel):
             html_head=html_head,
             included_content_layers=included_content_layers,
             split_page_view=split_page_view,
+            include_annotations=include_annotations,
         )
 
         with open(filename, "w", encoding="utf-8") as fw:
@@ -3164,6 +3179,7 @@ class DoclingDocument(BaseModel):
         html_head: str = "null",  # should be deprecated ...
         included_content_layers: Optional[set[ContentLayer]] = None,
         split_page_view: bool = False,
+        include_annotations: bool = True,
     ) -> str:
         r"""Serialize to HTML."""
         from docling_core.transforms.serializer.html import (
@@ -3195,6 +3211,7 @@ class DoclingDocument(BaseModel):
             html_head=html_head,
             html_lang=html_lang,
             output_style=output_style,
+            include_annotations=include_annotations,
         )
 
         if html_head == "null":
@@ -4109,7 +4126,10 @@ class DoclingDocument(BaseModel):
     @classmethod
     def validate_document(cls, d: "DoclingDocument"):
         """validate_document."""
-        if not d.validate_tree(d.body) or not d.validate_tree(d.furniture):
-            raise ValueError("Document hierachy is inconsistent.")
+        with warnings.catch_warnings():
+            # ignore warning from deprecated furniture
+            warnings.filterwarnings("ignore", category=DeprecationWarning)
+            if not d.validate_tree(d.body) or not d.validate_tree(d.furniture):
+                raise ValueError("Document hierachy is inconsistent.")
 
         return d

{docling_core-2.31.1.dist-info → docling_core-2.32.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: docling-core
-Version: 2.31.1
+Version: 2.32.0
 Summary: A python library to define and validate data types in Docling.
 Home-page: https://github.com/docling-project
 License: MIT

{docling_core-2.31.1.dist-info → docling_core-2.32.0.dist-info}/RECORD

@@ -20,18 +20,18 @@ docling_core/transforms/__init__.py,sha256=P81y_oqkiTN4Ld5crh1gQ6BbHqqR6C6nBt9AC
 docling_core/transforms/chunker/__init__.py,sha256=YdizSKXLmmK9eyYBsarHWr8Mx_AoA0PT7c0absibZMk,306
 docling_core/transforms/chunker/base.py,sha256=kJaRrGQynglG9wpy0IaAYTf4MKheWH5BAPzx4LE9yIg,2824
 docling_core/transforms/chunker/hierarchical_chunker.py,sha256=7Fpwwsn2BoiR12KGPrn8fU1uuhqBLp85MRLMF0aIsL8,8281
-docling_core/transforms/chunker/hybrid_chunker.py,sha256=67Whij6zSPZbVQA-fToyBtTfLtDK6BdnZ-Mhlz0p8ZQ,11886
+docling_core/transforms/chunker/hybrid_chunker.py,sha256=i4Yskms48XRUAVhec8pTGDP1dbrTEgc1pNh5fNXqfKQ,12317
 docling_core/transforms/chunker/tokenizer/__init__.py,sha256=-bhXOTpoI7SYk7vn47z8Ek-RZFjJk4TfZawxsFuNHnE,34
 docling_core/transforms/chunker/tokenizer/base.py,sha256=2gOBQPYJYC0iWXOgMG3DiNP7xEBtii7DYcib0iECq5o,575
-docling_core/transforms/chunker/tokenizer/huggingface.py,sha256=JQ-D3b5vTPQbvu4HaMfYqFzSBLbV_HnmoBGv7d6Kqn4,2220
+docling_core/transforms/chunker/tokenizer/huggingface.py,sha256=aZ_RNQIzcNkAHGHZw3SBCoqJHM2Ihb65eiM29O9BR6o,2506
 docling_core/transforms/chunker/tokenizer/openai.py,sha256=zt2kwcC-r8MafeEG0CESab8E4RIC9aaFXxxnxOGyTMA,918
 docling_core/transforms/serializer/__init__.py,sha256=CECQlMoCDUxkg4RAUdC3itA3I3qFhKhe2HcYghN6_xw,105
 docling_core/transforms/serializer/base.py,sha256=9bgpWA0oMmZNRc3yIuZVnu5bJ1glClBsswtVF1vYwMI,6046
-docling_core/transforms/serializer/common.py,sha256=TC1EwHIp9PYcI8jeTKeavUAPtounYmS0V1bfS_wDKm0,17427
+docling_core/transforms/serializer/common.py,sha256=mkajw0QRL--WgVL42Vlp2e2PuUQVh79D6EKP4_3YKy0,18112
 docling_core/transforms/serializer/doctags.py,sha256=mEmRWVuebcG5pZcR1_HX146cyUk0_FjaLQtMXSgh9hs,17870
-docling_core/transforms/serializer/html.py,sha256=Xq9CU5qZTDdwstizYqWNL_TFNDs9NHK_6JvvZk0TP98,34571
+docling_core/transforms/serializer/html.py,sha256=_HN1WFKH_WJkxtZrmvm1a6-UDxsEGt_ChWdUysS1qjY,35843
 docling_core/transforms/serializer/html_styles.py,sha256=-jBwS4EU7yfKoz0GSoxhwx90OmIKieO6TwPw57IuxcA,4692
-docling_core/transforms/serializer/markdown.py,sha256=4thokWJIaF3dvpchjp-Y7NTSzUuXwTmfNey4MQj-c5I,17873
+docling_core/transforms/serializer/markdown.py,sha256=ussKqIptiKPTCRNjy3edjap4DOsy52no-FLSeAyv9S0,18759
 docling_core/transforms/visualizer/__init__.py,sha256=gUfF25yiJ_KO46ZIUNqZQOZGy2PLx6gnnr6AZYxKHXI,35
 docling_core/transforms/visualizer/base.py,sha256=aEF7b3rHq6DVdX8zDYEPoq55BHDYe4Hh_97lBdcW4lY,555
 docling_core/transforms/visualizer/layout_visualizer.py,sha256=ulXxWGIl69-HMKDPFk_XKgNCgQeDNc969PVt_X0-drA,7823
@@ -40,7 +40,7 @@ docling_core/types/__init__.py,sha256=MVRSgsk5focwGyAplh_TRR3dEecIXpd98g_u3zZ5HX
 docling_core/types/base.py,sha256=PusJskRVL19y-hq0BgXr5e8--QEqSqLnFNJ8UbOqW88,8318
 docling_core/types/doc/__init__.py,sha256=bysJn2iwjAHwThSWDPXEdVUUij7p_ax12_nx2_0CMdg,653
 docling_core/types/doc/base.py,sha256=sM3IyFXzVh2WT8IGh5nejXYh8sf39yBh8TBSlHeJ9CI,12611
-docling_core/types/doc/document.py,sha256=eboNYL-QVnDNnw3vL7PPVdPosfs5oNfsrVofxmdBDHw,140884
+docling_core/types/doc/document.py,sha256=wxPxTOh3pfZr33rGPgnrFSY6b70C5Fe20tqqgYRUxrI,141930
 docling_core/types/doc/labels.py,sha256=vp4h3e7AmBvezRmgrfuPehjAHTZOufphErLB4ENhdME,7171
 docling_core/types/doc/page.py,sha256=1JMPwglaTITBvg959L_pcWPb-fXoDYGh-e_tGZMzVMQ,41060
 docling_core/types/doc/tokens.py,sha256=z22l9J81_sg9CYMvOuLmPuLsNT7h_s7wao2UT89DvI8,9278
@@ -73,8 +73,8 @@ docling_core/utils/generate_jsonschema.py,sha256=uNX1O5XnjyB5nA66XqZXTt3YbGuR2ty
 docling_core/utils/legacy.py,sha256=DrI3QGoL755ZCIoKHF74-pTWm8R0zfFo2C2vB5dT2aY,24463
 docling_core/utils/validate.py,sha256=aQ11UbFyl8iD_N7yTTZmm_VVeXz8KcCyn3GLXgkfYRM,2049
 docling_core/utils/validators.py,sha256=azcrndLzhNkTWnbFSu9shJ5D3j_znnLrIFA5R8hzmGU,2798
-docling_core-2.31.1.dist-info/LICENSE,sha256=2M9-6EoQ1sxFztTOkXGAtwUDJvnWaAHdB9BYWVwGkIw,1087
-docling_core-2.31.1.dist-info/METADATA,sha256=O13NvxzbHR0wUzP_3yQbOFqSI63LUvCzWvXEkiqQePY,5976
-docling_core-2.31.1.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-docling_core-2.31.1.dist-info/entry_points.txt,sha256=oClcdb2L2RKx4jdqUykY16Kum_f0_whwWhGzIodyidc,216
-docling_core-2.31.1.dist-info/RECORD,,
+docling_core-2.32.0.dist-info/LICENSE,sha256=2M9-6EoQ1sxFztTOkXGAtwUDJvnWaAHdB9BYWVwGkIw,1087
+docling_core-2.32.0.dist-info/METADATA,sha256=3ZtmT43WNDIQQd4I9H-Yf1Z954rod4Y5ZDrvTNPsRBQ,5976
+docling_core-2.32.0.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+docling_core-2.32.0.dist-info/entry_points.txt,sha256=oClcdb2L2RKx4jdqUykY16Kum_f0_whwWhGzIodyidc,216
+docling_core-2.32.0.dist-info/RECORD,,