docling-core 2.48.4__tar.gz → 2.50.0__tar.gz

{docling_core-2.48.4 → docling_core-2.50.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling-core
-Version: 2.48.4
+Version: 2.50.0
 Summary: A python library to define and validate data types in Docling.
 Author-email: Cesar Berrospi Ramis <ceb@zurich.ibm.com>, Panos Vagenas <pva@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Christoph Auer <cau@zurich.ibm.com>, Peter Staar <taa@zurich.ibm.com>
 Maintainer-email: Panos Vagenas <pva@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Christoph Auer <cau@zurich.ibm.com>, Peter Staar <taa@zurich.ibm.com>, Cesar Berrospi Ramis <ceb@zurich.ibm.com>
@@ -15,12 +15,17 @@ Classifier: Intended Audience :: Developers
 Classifier: Intended Audience :: Science/Research
 Classifier: Natural Language :: English
 Classifier: Operating System :: OS Independent
-Classifier: Programming Language :: Python :: 3
 Classifier: Topic :: Database
 Classifier: Topic :: Scientific/Engineering :: Information Analysis
 Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Classifier: Typing :: Typed
 Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.9
+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
@@ -29,7 +34,7 @@ Requires-Dist: pydantic!=2.10.0,!=2.10.1,!=2.10.2,<3.0.0,>=2.6.0
 Requires-Dist: jsonref<2.0.0,>=1.1.0
 Requires-Dist: tabulate<0.10.0,>=0.9.0
 Requires-Dist: pandas<3.0.0,>=2.1.4
-Requires-Dist: pillow<12.0.0,>=10.0.0
+Requires-Dist: pillow<13.0.0,>=10.0.0
 Requires-Dist: pyyaml<7.0.0,>=5.1
 Requires-Dist: typing-extensions<5.0.0,>=4.12.2
 Requires-Dist: typer<0.20.0,>=0.12.5
@@ -39,7 +44,7 @@ Requires-Dist: semchunk<3.0.0,>=2.2.0; extra == "chunking"
 Requires-Dist: transformers<5.0.0,>=4.34.0; extra == "chunking"
 Provides-Extra: chunking-openai
 Requires-Dist: semchunk; extra == "chunking-openai"
-Requires-Dist: tiktoken<0.10.0,>=0.9.0; extra == "chunking-openai"
+Requires-Dist: tiktoken<0.13.0,>=0.9.0; extra == "chunking-openai"
 Dynamic: license-file
 
 # Docling Core

{docling_core-2.48.4 → docling_core-2.50.0}/docling_core/cli/view.py

@@ -39,9 +39,17 @@ def view(
         typer.Argument(
             ...,
             metavar="source",
-            help="Docling JSON file to view.",
+            help="Docling JSON or YAML file to view.",
         ),
     ],
+    split_view: Annotated[
+        bool,
+        typer.Option(
+            "--split-view",
+            "-s",
+            help="Split view of the document.",
+        ),
+    ] = False,
     version: Annotated[
         Optional[bool],
         typer.Option(
@@ -52,11 +60,19 @@ def view(
         ),
     ] = None,
 ):
-    """Display a Docling JSON file on the default browser."""
+    """Display a DoclingDocument file on the default browser."""
     path = resolve_source_to_path(source=source)
-    doc = DoclingDocument.load_from_json(filename=path)
-    target_path = Path(tempfile.mkdtemp()) / "out.html"
-    html_output = doc.export_to_html(image_mode=ImageRefMode.EMBEDDED)
+    if path.suffix == ".json":
+        doc = DoclingDocument.load_from_json(filename=path)
+    elif path.suffix in [".yaml", ".yml"]:
+        doc = DoclingDocument.load_from_yaml(filename=path)
+    else:
+        raise ValueError(f"Unsupported file type: {path.suffix}")
+    target_path = Path(tempfile.mkdtemp()) / f"{path.stem}.html"
+    html_output = doc.export_to_html(
+        image_mode=ImageRefMode.EMBEDDED,
+        split_page_view=split_view,
+    )
     with open(target_path, "w", encoding="utf-8") as f:
         f.write(html_output)
     webbrowser.open(url=f"file://{target_path.absolute().resolve()}")

{docling_core-2.48.4 → docling_core-2.50.0}/docling_core/transforms/serializer/base.py

@@ -9,6 +9,7 @@ from pathlib import Path
 from typing import Any, Optional, Union
 
 from pydantic import AnyUrl, BaseModel
+from typing_extensions import deprecated
 
 from docling_core.types.doc.document import (
     DocItem,
@@ -258,6 +259,7 @@ class BaseDocSerializer(ABC):
         """Serialize the item's captions."""
         ...
 
+    @deprecated("Use serialize_meta() instead.")
     @abstractmethod
     def serialize_annotations(
         self,
@@ -267,6 +269,15 @@ class BaseDocSerializer(ABC):
         """Serialize the item's annotations."""
         ...
 
+    @abstractmethod
+    def serialize_meta(
+        self,
+        item: NodeItem,
+        **kwargs: Any,
+    ) -> SerializationResult:
+        """Serialize the item's meta."""
+        ...
+
     @abstractmethod
     def get_excluded_refs(self, **kwargs: Any) -> set[str]:
         """Get references to excluded items."""
@@ -287,6 +298,26 @@ class BaseSerializerProvider(ABC):
         ...
 
 
+class BaseMetaSerializer(ABC):
+    """Base class for meta serializers."""
+
+    @abstractmethod
+    def serialize(
+        self,
+        *,
+        item: NodeItem,
+        doc: DoclingDocument,
+        **kwargs: Any,
+    ) -> SerializationResult:
+        """Serializes the meta of the passed item."""
+        ...
+
+    def _humanize_text(self, text: str, title: bool = False) -> str:
+        tmp = text.replace("__", "_").replace("_", " ")
+        return tmp.title() if title else tmp.capitalize()
+
+
+@deprecated("Use BaseMetaSerializer() instead.")
 class BaseAnnotationSerializer(ABC):
     """Base class for annotation serializers."""
 

{docling_core-2.48.4 → docling_core-2.50.0}/docling_core/transforms/serializer/common.py

@@ -4,6 +4,7 @@
 #
 
 """Define base classes for serialization."""
+import logging
 import re
 import sys
 from abc import abstractmethod
@@ -11,7 +12,14 @@ from functools import cached_property
 from pathlib import Path
 from typing import Any, Iterable, Optional, Tuple, Union
 
-from pydantic import AnyUrl, BaseModel, ConfigDict, NonNegativeInt, computed_field
+from pydantic import (
+    AnyUrl,
+    BaseModel,
+    ConfigDict,
+    Field,
+    NonNegativeInt,
+    computed_field,
+)
 from typing_extensions import Self, override
 
 from docling_core.transforms.serializer.base import (
@@ -22,6 +30,7 @@ from docling_core.transforms.serializer.base import (
     BaseInlineSerializer,
     BaseKeyValueSerializer,
     BaseListSerializer,
+    BaseMetaSerializer,
     BasePictureSerializer,
     BaseTableSerializer,
     BaseTextSerializer,
@@ -56,6 +65,9 @@ _DEFAULT_LABELS = DOCUMENT_TOKENS_EXPORT_LABELS
 _DEFAULT_LAYERS = {cl for cl in ContentLayer}
 
 
+_logger = logging.getLogger(__name__)
+
+
 class _PageBreakNode(NodeItem):
     """Page break node."""
 
@@ -76,11 +88,11 @@ def _iterate_items(
     traverse_pictures: bool = False,
     add_page_breaks: bool = False,
     visited: Optional[set[str]] = None,
-):
+) -> Iterable[Tuple[NodeItem, int]]:
     my_visited: set[str] = visited if visited is not None else set()
     prev_page_nr: Optional[int] = None
     page_break_i = 0
-    for item, _ in doc.iterate_items(
+    for item, lvl in doc.iterate_items(
         root=node,
         with_groups=True,
         included_content_layers=layers,
@@ -93,7 +105,7 @@ def _iterate_items(
             ):
                 # if group starts with new page, yield page break before group node
                 my_visited.add(item.self_ref)
-                for it in _iterate_items(
+                for it, _ in _iterate_items(
                     doc=doc,
                     layers=layers,
                     node=item,
@@ -108,7 +120,7 @@ def _iterate_items(
                                 self_ref=f"#/pb/{page_break_i}",
                                 prev_page=prev_page_nr,
                                 next_page=page_no,
-                            )
+                            ), lvl
                         break
             elif isinstance(item, DocItem) and item.prov:
                 page_no = item.prov[0].page_no
@@ -118,10 +130,10 @@ def _iterate_items(
                             self_ref=f"#/pb/{page_break_i}",
                             prev_page=prev_page_nr,
                             next_page=page_no,
-                        )
+                        ), lvl
                         page_break_i += 1
                     prev_page_nr = page_no
-        yield item
+        yield item, lvl
 
 
 def _get_annotation_text(
@@ -188,9 +200,22 @@ class CommonParams(BaseModel):
     start_idx: NonNegativeInt = 0
     stop_idx: NonNegativeInt = sys.maxsize
 
+    include_non_meta: bool = True
+
     include_formatting: bool = True
     include_hyperlinks: bool = True
     caption_delim: str = " "
+    use_legacy_annotations: bool = Field(
+        default=False, description="Use legacy annotation serialization."
+    )
+    allowed_meta_names: Optional[set[str]] = Field(
+        default=None,
+        description="Meta name to allow; None means all meta names are allowed.",
+    )
+    blocked_meta_names: set[str] = Field(
+        default_factory=set,
+        description="Meta name to block; takes precedence over allowed_meta_names.",
+    )
 
     def merge_with_patch(self, patch: dict[str, Any]) -> Self:
         """Create an instance by merging the provided patch dict on top of self."""
@@ -215,6 +240,7 @@ class DocSerializer(BaseModel, BaseDocSerializer):
     list_serializer: BaseListSerializer
     inline_serializer: BaseInlineSerializer
 
+    meta_serializer: Optional[BaseMetaSerializer] = None
     annotation_serializer: BaseAnnotationSerializer
 
     params: CommonParams = CommonParams()
@@ -245,7 +271,7 @@ class DocSerializer(BaseModel, BaseDocSerializer):
         if refs is None:
             refs = {
                 item.self_ref
-                for ix, item in enumerate(
+                for ix, (item, _) in enumerate(
                     _iterate_items(
                         doc=self.doc,
                         traverse_pictures=True,
@@ -301,103 +327,130 @@ class DocSerializer(BaseModel, BaseDocSerializer):
     ) -> SerializationResult:
         """Serialize a given node."""
         my_visited: set[str] = visited if visited is not None else set()
+        parts: list[SerializationResult] = []
+        delim: str = kwargs.get("delim", "\n")
+        my_params = self.params.model_copy(update=kwargs)
         my_kwargs = {**self.params.model_dump(), **kwargs}
         empty_res = create_ser_result()
-        if item is None or item == self.doc.body:
-            if self.doc.body.self_ref not in my_visited:
-                my_visited.add(self.doc.body.self_ref)
-                return self._serialize_body(**my_kwargs)
+
+        my_item = item or self.doc.body
+
+        if my_item == self.doc.body:
+            if my_item.meta and not my_params.use_legacy_annotations:
+                meta_part = self.serialize_meta(item=my_item, **my_kwargs)
+                if meta_part.text:
+                    parts.append(meta_part)
+
+            if my_item.self_ref not in my_visited:
+                my_visited.add(my_item.self_ref)
+                part = self._serialize_body(**my_kwargs)
+                if part.text:
+                    parts.append(part)
+                return create_ser_result(
+                    text=delim.join([p.text for p in parts if p.text]),
+                    span_source=parts,
+                )
             else:
                 return empty_res
 
-        my_visited.add(item.self_ref)
-
-        ########
-        # groups
-        ########
-        if isinstance(item, ListGroup):
-            part = self.list_serializer.serialize(
-                item=item,
-                doc_serializer=self,
-                doc=self.doc,
-                list_level=list_level,
-                is_inline_scope=is_inline_scope,
-                visited=my_visited,
-                **my_kwargs,
-            )
-        elif isinstance(item, InlineGroup):
-            part = self.inline_serializer.serialize(
-                item=item,
-                doc_serializer=self,
-                doc=self.doc,
-                list_level=list_level,
-                visited=my_visited,
-                **my_kwargs,
-            )
-        ###########
-        # doc items
-        ###########
-        elif isinstance(item, TextItem):
-            if item.self_ref in self._captions_of_some_item:
-                # those captions will be handled by the floating item holding them
-                return empty_res
-            else:
-                part = (
-                    self.text_serializer.serialize(
-                        item=item,
-                        doc_serializer=self,
-                        doc=self.doc,
-                        is_inline_scope=is_inline_scope,
-                        visited=my_visited,
-                        **my_kwargs,
+        my_visited.add(my_item.self_ref)
+
+        if my_item.meta and not my_params.use_legacy_annotations:
+            meta_part = self.serialize_meta(item=my_item, **my_kwargs)
+            if meta_part.text:
+                parts.append(meta_part)
+
+        if my_params.include_non_meta:
+            ########
+            # groups
+            ########
+            if isinstance(my_item, ListGroup):
+                part = self.list_serializer.serialize(
+                    item=my_item,
+                    doc_serializer=self,
+                    doc=self.doc,
+                    list_level=list_level,
+                    is_inline_scope=is_inline_scope,
+                    visited=my_visited,
+                    **my_kwargs,
+                )
+            elif isinstance(my_item, InlineGroup):
+                part = self.inline_serializer.serialize(
+                    item=my_item,
+                    doc_serializer=self,
+                    doc=self.doc,
+                    list_level=list_level,
+                    visited=my_visited,
+                    **my_kwargs,
+                )
+            ###########
+            # doc items
+            ###########
+            elif isinstance(my_item, TextItem):
+                if my_item.self_ref in self._captions_of_some_item:
+                    # those captions will be handled by the floating item holding them
+                    return empty_res
+                else:
+                    part = (
+                        self.text_serializer.serialize(
+                            item=my_item,
+                            doc_serializer=self,
+                            doc=self.doc,
+                            is_inline_scope=is_inline_scope,
+                            visited=my_visited,
+                            **my_kwargs,
+                        )
+                        if my_item.self_ref not in self.get_excluded_refs(**kwargs)
+                        else empty_res
                     )
-                    if item.self_ref not in self.get_excluded_refs(**kwargs)
-                    else empty_res
+            elif isinstance(my_item, TableItem):
+                part = self.table_serializer.serialize(
+                    item=my_item,
+                    doc_serializer=self,
+                    doc=self.doc,
+                    visited=my_visited,
+                    **my_kwargs,
                 )
-        elif isinstance(item, TableItem):
-            part = self.table_serializer.serialize(
-                item=item,
-                doc_serializer=self,
-                doc=self.doc,
-                visited=my_visited,
-                **my_kwargs,
-            )
-        elif isinstance(item, PictureItem):
-            part = self.picture_serializer.serialize(
-                item=item,
-                doc_serializer=self,
-                doc=self.doc,
-                visited=my_visited,
-                **my_kwargs,
-            )
-        elif isinstance(item, KeyValueItem):
-            part = self.key_value_serializer.serialize(
-                item=item,
-                doc_serializer=self,
-                doc=self.doc,
-                **my_kwargs,
-            )
-        elif isinstance(item, FormItem):
-            part = self.form_serializer.serialize(
-                item=item,
-                doc_serializer=self,
-                doc=self.doc,
-                **my_kwargs,
-            )
-        elif isinstance(item, _PageBreakNode):
-            part = _PageBreakSerResult(
-                text=self._create_page_break(node=item),
-                node=item,
-            )
-        else:
-            part = self.fallback_serializer.serialize(
-                item=item,
-                doc_serializer=self,
-                doc=self.doc,
-                visited=my_visited,
-                **my_kwargs,
-            )
-        return part
+            elif isinstance(my_item, PictureItem):
+                part = self.picture_serializer.serialize(
+                    item=my_item,
+                    doc_serializer=self,
+                    doc=self.doc,
+                    visited=my_visited,
+                    **my_kwargs,
+                )
+            elif isinstance(my_item, KeyValueItem):
+                part = self.key_value_serializer.serialize(
+                    item=my_item,
+                    doc_serializer=self,
+                    doc=self.doc,
+                    **my_kwargs,
+                )
+            elif isinstance(my_item, FormItem):
+                part = self.form_serializer.serialize(
+                    item=my_item,
+                    doc_serializer=self,
+                    doc=self.doc,
+                    **my_kwargs,
+                )
+            elif isinstance(my_item, _PageBreakNode):
+                part = _PageBreakSerResult(
+                    text=self._create_page_break(node=my_item),
+                    node=my_item,
+                )
+            else:
+                part = self.fallback_serializer.serialize(
+                    item=my_item,
+                    doc_serializer=self,
+                    doc=self.doc,
+                    visited=my_visited,
+                    **my_kwargs,
+                )
+            parts.append(part)
+
+        return create_ser_result(
+            text=delim.join([p.text for p in parts if p.text]), span_source=parts
+        )
 
     # making some assumptions about the kwargs it can pass
     @override
@@ -416,7 +469,7 @@ class DocSerializer(BaseModel, BaseDocSerializer):
         my_visited: set[str] = visited if visited is not None else set()
         params = self.params.merge_with_patch(patch=kwargs)
 
-        for node in _iterate_items(
+        for node, lvl in _iterate_items(
             node=item,
             doc=self.doc,
             layers=params.layers,
@@ -426,15 +479,17 @@ class DocSerializer(BaseModel, BaseDocSerializer):
                 continue
             else:
                 my_visited.add(node.self_ref)
+
             part = self.serialize(
                 item=node,
                 list_level=list_level,
                 is_inline_scope=is_inline_scope,
                 visited=my_visited,
-                **kwargs,
+                **(dict(level=lvl) | kwargs),
             )
             if part.text:
                 parts.append(part)
+
         return parts
 
     @override
@@ -528,6 +583,31 @@ class DocSerializer(BaseModel, BaseDocSerializer):
             text_res = ""
         return create_ser_result(text=text_res, span_source=results)
 
+    @override
+    def serialize_meta(
+        self,
+        item: NodeItem,
+        **kwargs: Any,
+    ) -> SerializationResult:
+        """Serialize the item's meta."""
+        if self.meta_serializer:
+            if item.self_ref not in self.get_excluded_refs(**kwargs):
+                return self.meta_serializer.serialize(
+                    item=item,
+                    doc=self.doc,
+                    **(self.params.model_dump() | kwargs),
+                )
+            else:
+                return create_ser_result(
+                    text="", span_source=item if isinstance(item, DocItem) else []
+                )
+        else:
+            _logger.warning("No meta serializer found.")
+            return create_ser_result(
+                text="", span_source=item if isinstance(item, DocItem) else []
+            )
+
+    # TODO deprecate
     @override
     def serialize_annotations(
         self,

{docling_core-2.48.4 → docling_core-2.50.0}/docling_core/transforms/serializer/doctags.py

@@ -44,6 +44,7 @@ from docling_core.types.doc.document import (
     PictureTabularChartData,
     ProvenanceItem,
     SectionHeaderItem,
+    TableData,
     TableItem,
     TextItem,
 )
@@ -233,13 +234,22 @@ class DocTagsPictureSerializer(BasePictureSerializer):
                     ysize=params.ysize,
                 )
 
-            classifications = [
-                ann
-                for ann in item.annotations
-                if isinstance(ann, PictureClassificationData)
-            ]
-            if len(classifications) > 0:
+            # handle classification data
+            predicted_class: Optional[str] = None
+            if item.meta and item.meta.classification:
+                predicted_class = (
+                    item.meta.classification.get_main_prediction().class_name
+                )
+            elif (
+                classifications := [
+                    ann
+                    for ann in item.annotations
+                    if isinstance(ann, PictureClassificationData)
+                ]
+            ) and classifications[0].predicted_classes:
                 predicted_class = classifications[0].predicted_classes[0].class_name
+            if predicted_class:
+                body += DocumentToken.get_picture_classification_token(predicted_class)
                 if predicted_class in [
                     PictureClassificationLabel.PIE_CHART,
                     PictureClassificationLabel.BAR_CHART,
@@ -250,26 +260,31 @@ class DocTagsPictureSerializer(BasePictureSerializer):
                     PictureClassificationLabel.HEATMAP,
                 ]:
                     is_chart = True
-                body += DocumentToken.get_picture_classification_token(predicted_class)
 
-            smiles_annotations = [
+            # handle molecule data
+            smi: Optional[str] = None
+            if item.meta and item.meta.molecule:
+                smi = item.meta.molecule.smi
+            elif smiles_annotations := [
                 ann for ann in item.annotations if isinstance(ann, PictureMoleculeData)
-            ]
-            if len(smiles_annotations) > 0:
-                body += _wrap(
-                    text=smiles_annotations[0].smi, wrap_tag=DocumentToken.SMILES.value
-                )
-
-            tabular_chart_annotations = [
+            ]:
+                smi = smiles_annotations[0].smi
+            if smi:
+                body += _wrap(text=smi, wrap_tag=DocumentToken.SMILES.value)
+
+            # handle tabular chart data
+            chart_data: Optional[TableData] = None
+            if item.meta and item.meta.tabular_chart:
+                chart_data = item.meta.tabular_chart.chart_data
+            elif tabular_chart_annotations := [
                 ann
                 for ann in item.annotations
                 if isinstance(ann, PictureTabularChartData)
-            ]
-            if len(tabular_chart_annotations) > 0:
+            ]:
+                chart_data = tabular_chart_annotations[0].chart_data
+            if chart_data and chart_data.table_cells:
                 temp_doc = DoclingDocument(name="temp")
-                temp_table = temp_doc.add_table(
-                    data=tabular_chart_annotations[0].chart_data
-                )
+                temp_table = temp_doc.add_table(data=chart_data)
                 otsl_content = temp_table.export_to_otsl(
                     temp_doc, add_cell_location=False
                 )