docling-core 2.33.1__py3-none-any.whl → 2.34.0__py3-none-any.whl

docling_core/transforms/serializer/base.py

@@ -202,6 +202,16 @@ class BaseDocSerializer(ABC):
         """Hook for strikethrough formatting serialization."""
         ...
 
+    @abstractmethod
+    def serialize_subscript(self, text: str, **kwargs: Any) -> str:
+        """Hook for subscript formatting serialization."""
+        ...
+
+    @abstractmethod
+    def serialize_superscript(self, text: str, **kwargs: Any) -> str:
+        """Hook for superscript formatting serialization."""
+        ...
+
     @abstractmethod
     def serialize_hyperlink(
         self,
@@ -239,6 +249,15 @@ class BaseDocSerializer(ABC):
         """Serialize the item's captions."""
         ...
 
+    @abstractmethod
+    def serialize_annotations(
+        self,
+        item: DocItem,
+        **kwargs: Any,
+    ) -> SerializationResult:
+        """Serialize the item's annotations."""
+        ...
+
     @abstractmethod
     def get_excluded_refs(self, **kwargs: Any) -> set[str]:
         """Get references to excluded items."""
@@ -257,3 +276,18 @@ class BaseSerializerProvider(ABC):
     def get_serializer(self, doc: DoclingDocument) -> BaseDocSerializer:
         """Get a the associated serializer."""
         ...
+
+
+class BaseAnnotationSerializer(ABC):
+    """Base class for annotation serializers."""
+
+    @abstractmethod
+    def serialize(
+        self,
+        *,
+        item: DocItem,
+        doc: DoclingDocument,
+        **kwargs: Any,
+    ) -> SerializationResult:
+        """Serializes the passed annotation."""
+        ...

docling_core/transforms/serializer/common.py

@@ -15,6 +15,7 @@ from pydantic import AnyUrl, BaseModel, ConfigDict, NonNegativeInt, computed_fie
 from typing_extensions import Self, override
 
 from docling_core.transforms.serializer.base import (
+    BaseAnnotationSerializer,
     BaseDocSerializer,
     BaseFallbackSerializer,
     BaseFormSerializer,
@@ -30,6 +31,7 @@ from docling_core.transforms.serializer.base import (
 from docling_core.types.doc.document import (
     DOCUMENT_TOKENS_EXPORT_LABELS,
     ContentLayer,
+    DescriptionAnnotation,
     DocItem,
     DoclingDocument,
     FloatingItem,
@@ -41,9 +43,10 @@ from docling_core.types.doc.document import (
     OrderedList,
     PictureClassificationData,
     PictureDataType,
-    PictureDescriptionData,
     PictureItem,
     PictureMoleculeData,
+    Script,
+    TableAnnotationType,
     TableItem,
     TextItem,
     UnorderedList,
@@ -122,7 +125,9 @@ def _iterate_items(
         yield item
 
 
-def _get_picture_annotation_text(annotation: PictureDataType) -> Optional[str]:
+def _get_annotation_text(
+    annotation: Union[PictureDataType, TableAnnotationType],
+) -> Optional[str]:
     result = None
     if isinstance(annotation, PictureClassificationData):
         predicted_class = (
@@ -132,7 +137,7 @@ def _get_picture_annotation_text(annotation: PictureDataType) -> Optional[str]:
         )
         if predicted_class is not None:
             result = predicted_class.replace("_", " ")
-    elif isinstance(annotation, PictureDescriptionData):
+    elif isinstance(annotation, DescriptionAnnotation):
         result = annotation.text
     elif isinstance(annotation, PictureMoleculeData):
         result = annotation.smi
@@ -211,6 +216,8 @@ class DocSerializer(BaseModel, BaseDocSerializer):
     list_serializer: BaseListSerializer
     inline_serializer: BaseInlineSerializer
 
+    annotation_serializer: BaseAnnotationSerializer
+
     params: CommonParams = CommonParams()
 
     _excluded_refs_cache: dict[str, set[str]] = {}
@@ -449,6 +456,10 @@ class DocSerializer(BaseModel, BaseDocSerializer):
                 res = self.serialize_underline(text=res)
             if formatting.strikethrough:
                 res = self.serialize_strikethrough(text=res)
+            if formatting.script == Script.SUB:
+                res = self.serialize_subscript(text=res)
+            elif formatting.script == Script.SUPER:
+                res = self.serialize_superscript(text=res)
         if params.include_hyperlinks and hyperlink:
             res = self.serialize_hyperlink(text=res, hyperlink=hyperlink)
         return res
@@ -473,6 +484,16 @@ class DocSerializer(BaseModel, BaseDocSerializer):
         """Hook for strikethrough formatting serialization."""
         return text
 
+    @override
+    def serialize_subscript(self, text: str, **kwargs: Any) -> str:
+        """Hook for subscript formatting serialization."""
+        return text
+
+    @override
+    def serialize_superscript(self, text: str, **kwargs: Any) -> str:
+        """Hook for superscript formatting serialization."""
+        return text
+
     @override
     def serialize_hyperlink(
         self,
@@ -505,6 +526,19 @@ class DocSerializer(BaseModel, BaseDocSerializer):
             text_res = ""
         return create_ser_result(text=text_res, span_source=results)
 
+    @override
+    def serialize_annotations(
+        self,
+        item: DocItem,
+        **kwargs: Any,
+    ) -> SerializationResult:
+        """Serialize the item's annotations."""
+        return self.annotation_serializer.serialize(
+            item=item,
+            doc=self.doc,
+            **kwargs,
+        )
+
     def _get_applicable_pages(self) -> Optional[list[int]]:
         pages = {
             item.prov[0].page_no: ...

docling_core/transforms/serializer/doctags.py

@@ -7,6 +7,7 @@ from pydantic import BaseModel
 from typing_extensions import override
 
 from docling_core.transforms.serializer.base import (
+    BaseAnnotationSerializer,
     BaseDocSerializer,
     BaseFallbackSerializer,
     BaseFormSerializer,
@@ -17,12 +18,14 @@ from docling_core.transforms.serializer.base import (
     BaseTableSerializer,
     BaseTextSerializer,
     SerializationResult,
+    Span,
 )
 from docling_core.transforms.serializer.common import (
     CommonParams,
     DocSerializer,
     create_ser_result,
 )
+from docling_core.types.doc.base import BoundingBox
 from docling_core.types.doc.document import (
     CodeItem,
     DocItem,
@@ -38,6 +41,7 @@ from docling_core.types.doc.document import (
     PictureItem,
     PictureMoleculeData,
     PictureTabularChartData,
+    ProvenanceItem,
     TableItem,
     TextItem,
     UnorderedList,
@@ -414,6 +418,39 @@ class DocTagsListSerializer(BaseModel, BaseListSerializer):
 class DocTagsInlineSerializer(BaseInlineSerializer):
     """DocTags-specific inline group serializer."""
 
+    def _get_inline_location_tags(
+        self, doc: DoclingDocument, item: InlineGroup, params: DocTagsParams
+    ) -> SerializationResult:
+
+        prov: Optional[ProvenanceItem] = None
+        boxes: list[BoundingBox] = []
+        doc_items: list[DocItem] = []
+        for it, _ in doc.iterate_items(root=item):
+            if isinstance(it, DocItem):
+                for prov in it.prov:
+                    boxes.append(prov.bbox)
+                    doc_items.append(it)
+        if prov is None:
+            return create_ser_result()
+
+        bbox = BoundingBox.enclosing_bbox(boxes=boxes)
+
+        # using last seen prov as reference for page dims
+        page_w, page_h = doc.pages[prov.page_no].size.as_tuple()
+
+        loc_str = DocumentToken.get_location(
+            bbox=bbox.to_top_left_origin(page_h).as_tuple(),
+            page_w=page_w,
+            page_h=page_h,
+            xsize=params.xsize,
+            ysize=params.ysize,
+        )
+
+        return SerializationResult(
+            text=loc_str,
+            spans=[Span(item=it) for it in doc_items],
+        )
+
     @override
     def serialize(
         self,
@@ -428,12 +465,23 @@ class DocTagsInlineSerializer(BaseInlineSerializer):
         """Serializes the passed item."""
         my_visited = visited if visited is not None else set()
         params = DocTagsParams(**kwargs)
-        parts = doc_serializer.get_parts(
-            item=item,
-            list_level=list_level,
-            is_inline_scope=True,
-            visited=my_visited,
-            **kwargs,
+        parts: List[SerializationResult] = []
+        if params.add_location:
+            inline_loc_tags_ser_res = self._get_inline_location_tags(
+                doc=doc,
+                item=item,
+                params=params,
+            )
+            parts.append(inline_loc_tags_ser_res)
+            params.add_location = False  # suppress children location serialization
+        parts.extend(
+            doc_serializer.get_parts(
+                item=item,
+                list_level=list_level,
+                is_inline_scope=True,
+                visited=my_visited,
+                **{**kwargs, **params.model_dump()},
+            )
         )
         wrap_tag = DocumentToken.INLINE.value
         delim = _get_delim(params=params)
@@ -460,6 +508,15 @@ class DocTagsFallbackSerializer(BaseFallbackSerializer):
         return create_ser_result()
 
 
+class DocTagsAnnotationSerializer(BaseAnnotationSerializer):
+    """DocTags-specific annotation serializer."""
+
+    @override
+    def serialize(self, *, item: DocItem, **kwargs: Any) -> SerializationResult:
+        """Serializes the item's annotations."""
+        return create_ser_result()
+
+
 class DocTagsDocSerializer(DocSerializer):
     """DocTags-specific document serializer."""
 
@@ -473,6 +530,8 @@ class DocTagsDocSerializer(DocSerializer):
     list_serializer: BaseListSerializer = DocTagsListSerializer()
     inline_serializer: BaseInlineSerializer = DocTagsInlineSerializer()
 
+    annotation_serializer: BaseAnnotationSerializer = DocTagsAnnotationSerializer()
+
     params: DocTagsParams = DocTagsParams()
 
     @override

docling_core/transforms/serializer/html.py

@@ -21,6 +21,7 @@ from pydantic import AnyUrl, BaseModel
 from typing_extensions import override
 
 from docling_core.transforms.serializer.base import (
+    BaseAnnotationSerializer,
     BaseDocSerializer,
     BaseFallbackSerializer,
     BaseFormSerializer,
@@ -35,7 +36,7 @@ from docling_core.transforms.serializer.base import (
 from docling_core.transforms.serializer.common import (
     CommonParams,
     DocSerializer,
-    _get_picture_annotation_text,
+    _get_annotation_text,
     create_ser_result,
 )
 from docling_core.transforms.serializer.html_styles import (
@@ -47,6 +48,7 @@ from docling_core.types.doc.base import ImageRefMode
 from docling_core.types.doc.document import (
     CodeItem,
     ContentLayer,
+    DescriptionAnnotation,
     DocItem,
     DoclingDocument,
     FloatingItem,
@@ -59,7 +61,9 @@ from docling_core.types.doc.document import (
     ListItem,
     NodeItem,
     OrderedList,
+    PictureClassificationData,
     PictureItem,
+    PictureMoleculeData,
     PictureTabularChartData,
     SectionHeaderItem,
     TableCell,
@@ -758,14 +762,7 @@ class HTMLFallbackSerializer(BaseFallbackSerializer):
     """HTML-specific fallback serializer."""
 
     @override
-    def serialize(
-        self,
-        *,
-        item: NodeItem,
-        doc_serializer: "BaseDocSerializer",
-        doc: DoclingDocument,
-        **kwargs: Any,
-    ) -> SerializationResult:
+    def serialize(self, *, item: NodeItem, **kwargs: Any) -> SerializationResult:
         """Fallback serializer for items not handled by other serializers."""
         if isinstance(item, DocItem):
             return create_ser_result(
@@ -777,6 +774,42 @@ class HTMLFallbackSerializer(BaseFallbackSerializer):
             return create_ser_result()
 
 
+class HTMLAnnotationSerializer(BaseModel, BaseAnnotationSerializer):
+    """HTML-specific annotation serializer."""
+
+    def serialize(
+        self,
+        *,
+        item: DocItem,
+        doc: DoclingDocument,
+        **kwargs: Any,
+    ) -> SerializationResult:
+        """Serializes the passed annotation to HTML format."""
+        res_parts: list[SerializationResult] = []
+        for ann in item.get_annotations():
+            if isinstance(
+                ann,
+                (PictureClassificationData, DescriptionAnnotation, PictureMoleculeData),
+            ):
+                if ann_text := _get_annotation_text(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,
+                    )
+                    res_parts.append(ann_ser_res)
+
+        return create_ser_result(
+            text=" ".join([r.text for r in res_parts if r.text]),
+            span_source=res_parts,
+        )
+
+
 class HTMLDocSerializer(DocSerializer):
     """HTML-specific document serializer."""
 
@@ -790,6 +823,8 @@ class HTMLDocSerializer(DocSerializer):
     list_serializer: BaseListSerializer = HTMLListSerializer()
     inline_serializer: BaseInlineSerializer = HTMLInlineSerializer()
 
+    annotation_serializer: BaseAnnotationSerializer = HTMLAnnotationSerializer()
+
     params: HTMLParams = HTMLParams()
 
     @override
@@ -812,6 +847,16 @@ class HTMLDocSerializer(DocSerializer):
         """Apply HTML-specific strikethrough serialization."""
         return f"<del>{text}</del>"
 
+    @override
+    def serialize_subscript(self, text: str, **kwargs: Any) -> str:
+        """Apply HTML-specific subscript serialization."""
+        return f"<sub>{text}</sub>"
+
+    @override
+    def serialize_superscript(self, text: str, **kwargs: Any) -> str:
+        """Apply HTML-specific superscript serialization."""
+        return f"<sup>{text}</sup>"
+
     @override
     def serialize_hyperlink(
         self,
@@ -968,20 +1013,13 @@ class HTMLDocSerializer(DocSerializer):
                     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)
+            if isinstance(item, (PictureItem, TableItem)):
+                ann_res = self.serialize_annotations(
+                    item=item,
+                    **kwargs,
+                )
+                if ann_res.text:
+                    results.append(ann_res)
 
         text_res = params.caption_delim.join([r.text for r in results])
         if text_res:

docling_core/transforms/serializer/markdown.py

@@ -15,6 +15,7 @@ from tabulate import tabulate
 from typing_extensions import override
 
 from docling_core.transforms.serializer.base import (
+    BaseAnnotationSerializer,
     BaseDocSerializer,
     BaseFallbackSerializer,
     BaseFormSerializer,
@@ -29,7 +30,7 @@ from docling_core.transforms.serializer.base import (
 from docling_core.transforms.serializer.common import (
     CommonParams,
     DocSerializer,
-    _get_picture_annotation_text,
+    _get_annotation_text,
     _PageBreakSerResult,
     create_ser_result,
 )
@@ -37,6 +38,7 @@ from docling_core.types.doc.base import ImageRefMode
 from docling_core.types.doc.document import (
     CodeItem,
     ContentLayer,
+    DescriptionAnnotation,
     DocItem,
     DoclingDocument,
     FloatingItem,
@@ -48,7 +50,9 @@ from docling_core.types.doc.document import (
     KeyValueItem,
     NodeItem,
     OrderedList,
+    PictureClassificationData,
     PictureItem,
+    PictureMoleculeData,
     PictureTabularChartData,
     SectionHeaderItem,
     TableItem,
@@ -58,6 +62,23 @@ from docling_core.types.doc.document import (
 )
 
 
+def _get_annotation_ser_result(
+    ann_kind: str, ann_text: str, mark_annotation: bool, doc_item: DocItem
+):
+    return create_ser_result(
+        text=(
+            (
+                f'<!--<annotation kind="{ann_kind}">-->'
+                f"{ann_text}"
+                f"<!--<annotation/>-->"
+            )
+            if mark_annotation
+            else ann_text
+        ),
+        span_source=doc_item,
+    )
+
+
 class MarkdownParams(CommonParams):
     """Markdown-specific serialization parameters."""
 
@@ -136,6 +157,49 @@ class MarkdownTextSerializer(BaseModel, BaseTextSerializer):
         return create_ser_result(text=text, span_source=res_parts)
 
 
+class MarkdownAnnotationSerializer(BaseModel, BaseAnnotationSerializer):
+    """Markdown-specific annotation serializer."""
+
+    def serialize(
+        self,
+        *,
+        item: DocItem,
+        doc: DoclingDocument,
+        **kwargs: Any,
+    ) -> SerializationResult:
+        """Serialize the item's annotations."""
+        params = MarkdownParams(**kwargs)
+
+        res_parts: list[SerializationResult] = []
+        for ann in item.get_annotations():
+            if isinstance(
+                ann,
+                (
+                    PictureClassificationData,
+                    DescriptionAnnotation,
+                    PictureMoleculeData,
+                ),
+            ):
+                if ann_text := _get_annotation_text(ann):
+                    ann_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_res)
+        return create_ser_result(
+            text="\n\n".join([r.text for r in res_parts if r.text]),
+            span_source=item,
+        )
+
+
 class MarkdownTableSerializer(BaseTableSerializer):
     """Markdown-specific table item serializer."""
 
@@ -149,6 +213,7 @@ class MarkdownTableSerializer(BaseTableSerializer):
         **kwargs: Any,
     ) -> SerializationResult:
         """Serializes the passed item."""
+        params = MarkdownParams(**kwargs)
         res_parts: list[SerializationResult] = []
 
         cap_res = doc_serializer.serialize_captions(
@@ -159,6 +224,16 @@ class MarkdownTableSerializer(BaseTableSerializer):
             res_parts.append(cap_res)
 
         if item.self_ref not in doc_serializer.get_excluded_refs(**kwargs):
+
+            if params.include_annotations:
+
+                ann_res = doc_serializer.serialize_annotations(
+                    item=item,
+                    **kwargs,
+                )
+                if ann_res.text:
+                    res_parts.append(ann_res)
+
             rows = [
                 [
                     # make sure that md tables are not broken
@@ -214,22 +289,12 @@ class MarkdownPictureSerializer(BasePictureSerializer):
 
         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)
+                ann_res = doc_serializer.serialize_annotations(
+                    item=item,
+                    **kwargs,
+                )
+                if ann_res.text:
+                    res_parts.append(ann_res)
 
             img_res = self._serialize_image_part(
                 item=item,
@@ -257,7 +322,7 @@ class MarkdownPictureSerializer(BasePictureSerializer):
                     res_parts.append(
                         create_ser_result(text=md_table_content, span_source=item)
                     )
-        text_res = "\n\n".join([r.text for r in res_parts])
+        text_res = "\n\n".join([r.text for r in res_parts if r.text])
 
         return create_ser_result(text=text_res, span_source=res_parts)
 
@@ -471,6 +536,8 @@ class MarkdownDocSerializer(DocSerializer):
     list_serializer: BaseListSerializer = MarkdownListSerializer()
     inline_serializer: BaseInlineSerializer = MarkdownInlineSerializer()
 
+    annotation_serializer: BaseAnnotationSerializer = MarkdownAnnotationSerializer()
+
     params: MarkdownParams = MarkdownParams()
 
     @override

docling_core/types/doc/document.py

@@ -15,7 +15,7 @@ import warnings
 from enum import Enum
 from io import BytesIO
 from pathlib import Path
-from typing import Any, Dict, Final, List, Literal, Optional, Tuple, Union
+from typing import Any, Dict, Final, List, Literal, Optional, Sequence, Tuple, Union
 from urllib.parse import unquote
 
 import pandas as pd
@@ -30,6 +30,7 @@ from pydantic import (
     computed_field,
     field_validator,
     model_validator,
+    validate_call,
 )
 from tabulate import tabulate
 from typing_extensions import Annotated, Self, deprecated
@@ -53,7 +54,7 @@ _logger = logging.getLogger(__name__)
 
 Uint64 = typing.Annotated[int, Field(ge=0, le=(2**64 - 1))]
 LevelNumber = typing.Annotated[int, Field(ge=1, le=100)]
-CURRENT_VERSION: Final = "1.3.0"
+CURRENT_VERSION: Final = "1.4.0"
 
 DEFAULT_EXPORT_LABELS = {
     DocItemLabel.TITLE,
@@ -85,8 +86,8 @@ DOCUMENT_TOKENS_EXPORT_LABELS.update(
 )
 
 
-class BasePictureData(BaseModel):
-    """BasePictureData."""
+class BaseAnnotation(BaseModel):
+    """Base class for all annotation types."""
 
     kind: str
 
@@ -98,7 +99,7 @@ class PictureClassificationClass(BaseModel):
     confidence: float
 
 
-class PictureClassificationData(BasePictureData):
+class PictureClassificationData(BaseAnnotation):
     """PictureClassificationData."""
 
     kind: Literal["classification"] = "classification"
@@ -106,19 +107,18 @@ class PictureClassificationData(BasePictureData):
     predicted_classes: List[PictureClassificationClass]
 
 
-class PictureDescriptionData(BasePictureData):
-    """PictureDescriptionData."""
+class DescriptionAnnotation(BaseAnnotation):
+    """DescriptionAnnotation."""
 
     kind: Literal["description"] = "description"
     text: str
     provenance: str
 
 
-class PictureMoleculeData(BaseModel):
+class PictureMoleculeData(BaseAnnotation):
     """PictureMoleculeData."""
 
     kind: Literal["molecule_data"] = "molecule_data"
-
     smi: str
     confidence: float
     class_name: str
@@ -126,13 +126,19 @@ class PictureMoleculeData(BaseModel):
     provenance: str
 
 
-class PictureMiscData(BaseModel):
-    """PictureMiscData."""
+class MiscAnnotation(BaseAnnotation):
+    """MiscAnnotation."""
 
     kind: Literal["misc"] = "misc"
     content: Dict[str, Any]
 
 
+# deprecated aliases:
+BasePictureData = BaseAnnotation
+PictureDescriptionData = DescriptionAnnotation
+PictureMiscData = MiscAnnotation
+
+
 class ChartLine(BaseModel):
     """Represents a line in a line chart.
 
@@ -196,7 +202,7 @@ class ChartPoint(BaseModel):
     value: Tuple[float, float]
 
 
-class PictureChartData(BaseModel):
+class PictureChartData(BaseAnnotation):
     """Base class for picture chart data.
 
     Attributes:
@@ -381,10 +387,10 @@ class PictureTabularChartData(PictureChartData):
 
 PictureDataType = Annotated[
     Union[
+        DescriptionAnnotation,
+        MiscAnnotation,
         PictureClassificationData,
-        PictureDescriptionData,
         PictureMoleculeData,
-        PictureMiscData,
         PictureTabularChartData,
         PictureLineChartData,
         PictureBarChartData,
@@ -818,6 +824,18 @@ class DocItem(
         )
         return page_image.crop(crop_bbox.as_tuple())
 
+    def get_annotations(self) -> Sequence[BaseAnnotation]:
+        """Get the annotations of this DocItem."""
+        return []
+
+
+class Script(str, Enum):
+    """Text script position."""
+
+    BASELINE = "baseline"
+    SUB = "sub"
+    SUPER = "super"
+
 
 class Formatting(BaseModel):
     """Formatting."""
@@ -826,6 +844,7 @@ class Formatting(BaseModel):
     italic: bool = False
     underline: bool = False
     strikethrough: bool = False
+    script: Script = Script.BASELINE
 
 
 class TextItem(DocItem):
@@ -1182,6 +1201,19 @@ class PictureItem(FloatingItem):
         text = serializer.serialize(item=self).text
         return text
 
+    def get_annotations(self) -> Sequence[BaseAnnotation]:
+        """Get the annotations of this PictureItem."""
+        return self.annotations
+
+
+TableAnnotationType = Annotated[
+    Union[
+        DescriptionAnnotation,
+        MiscAnnotation,
+    ],
+    Field(discriminator="kind"),
+]
+
 
 class TableItem(FloatingItem):
     """TableItem."""
@@ -1192,6 +1224,8 @@ class TableItem(FloatingItem):
         DocItemLabel.TABLE,
     ] = DocItemLabel.TABLE
 
+    annotations: List[TableAnnotationType] = []
+
     def export_to_dataframe(self) -> pd.DataFrame:
         """Export the table as a Pandas DataFrame."""
         if self.data.num_rows == 0 or self.data.num_cols == 0:
@@ -1438,6 +1472,15 @@ class TableItem(FloatingItem):
         text = serializer.serialize(item=self).text
         return text
 
+    @validate_call
+    def add_annotation(self, annotation: TableAnnotationType) -> None:
+        """Add an annotation to the table."""
+        self.annotations.append(annotation)
+
+    def get_annotations(self) -> Sequence[BaseAnnotation]:
+        """Get the annotations of this TableItem."""
+        return self.annotations
+
 
 class GraphCell(BaseModel):
     """GraphCell."""
@@ -1776,6 +1819,18 @@ class DoclingDocument(BaseModel):
             item.parent = parent_ref
 
             self.form_items.append(item)
+
+        elif isinstance(item, (UnorderedList, OrderedList, InlineGroup)):
+            item_label = "groups"
+            item_index = len(self.groups)
+
+            cref = f"#/{item_label}/{item_index}"
+
+            item.self_ref = cref
+            item.parent = parent_ref
+
+            self.groups.append(item)
+
         else:
             raise ValueError(f"Item {item} is not supported for insertion")
 
@@ -2111,8 +2166,8 @@ class DoclingDocument(BaseModel):
         :param parent: Optional[NodeItem]:  (Default value = None)
 
         """
-        if not parent:
-            parent = self.body
+        if not isinstance(parent, (OrderedList, UnorderedList)):
+            raise ValueError("ListItem's parent must be a list group")
 
         if not orig:
             orig = text
@@ -2267,6 +2322,7 @@ class DoclingDocument(BaseModel):
         parent: Optional[NodeItem] = None,
         label: DocItemLabel = DocItemLabel.TABLE,
         content_layer: Optional[ContentLayer] = None,
+        annotations: Optional[list[TableAnnotationType]] = None,
     ):
         """add_table.
 
@@ -2284,7 +2340,11 @@ class DoclingDocument(BaseModel):
         cref = f"#/tables/{table_index}"
 
         tbl_item = TableItem(
-            label=label, data=data, self_ref=cref, parent=parent.get_ref()
+            label=label,
+            data=data,
+            self_ref=cref,
+            parent=parent.get_ref(),
+            annotations=annotations or [],
         )
         if prov:
             tbl_item.prov.append(prov)
@@ -2301,7 +2361,7 @@ class DoclingDocument(BaseModel):
 
     def add_picture(
         self,
-        annotations: List[PictureDataType] = [],
+        annotations: Optional[List[PictureDataType]] = None,
         image: Optional[ImageRef] = None,
         caption: Optional[Union[TextItem, RefItem]] = None,
         prov: Optional[ProvenanceItem] = None,
@@ -2310,7 +2370,7 @@ class DoclingDocument(BaseModel):
     ):
         """add_picture.
 
-        :param data: List[PictureData]: (Default value = [])
+        :param data: Optional[List[PictureData]]: (Default value = None)
         :param caption: Optional[Union[TextItem:
         :param RefItem]]:  (Default value = None)
         :param prov: Optional[ProvenanceItem]:  (Default value = None)
@@ -2324,7 +2384,7 @@ class DoclingDocument(BaseModel):
 
         fig_item = PictureItem(
             label=DocItemLabel.PICTURE,
-            annotations=annotations,
+            annotations=annotations or [],
             image=image,
             self_ref=cref,
             parent=parent.get_ref(),
@@ -3589,6 +3649,52 @@ class DoclingDocument(BaseModel):
 
             return (GraphData(cells=cells, links=links), overall_prov)
 
+        def _add_text(
+            full_chunk: str,
+            bbox: Optional[BoundingBox],
+            pg_width: int,
+            pg_height: int,
+            page_no: int,
+            tag_name: str,
+            doc_label: DocItemLabel,
+            doc: DoclingDocument,
+            parent: Optional[NodeItem],
+        ):
+            # For everything else, treat as text
+            text_content = extract_inner_text(full_chunk)
+            element_prov = (
+                ProvenanceItem(
+                    bbox=bbox.resize_by_scale(pg_width, pg_height),
+                    charspan=(0, len(text_content)),
+                    page_no=page_no,
+                )
+                if bbox
+                else None
+            )
+
+            content_layer = ContentLayer.BODY
+            if tag_name in [DocItemLabel.PAGE_HEADER, DocItemLabel.PAGE_FOOTER]:
+                content_layer = ContentLayer.FURNITURE
+
+            if doc_label == DocItemLabel.SECTION_HEADER:
+                # Extract level from tag_name (e.g. "section_level_header_1" -> 1)
+                level = int(tag_name.split("_")[-1])
+                doc.add_heading(
+                    text=text_content,
+                    level=level,
+                    prov=element_prov,
+                    parent=parent,
+                    content_layer=content_layer,
+                )
+            else:
+                doc.add_text(
+                    label=doc_label,
+                    text=text_content,
+                    prov=element_prov,
+                    parent=parent,
+                    content_layer=content_layer,
+                )
+
         # doc = DoclingDocument(name="Document")
         for pg_idx, doctag_page in enumerate(doctag_document.pages):
             page_doctags = doctag_page.tokens
@@ -3623,7 +3729,7 @@ class DoclingDocument(BaseModel):
             tag_pattern = (
                 rf"<(?P<tag>{DocItemLabel.TITLE}|{DocItemLabel.DOCUMENT_INDEX}|"
                 rf"{DocItemLabel.CHECKBOX_UNSELECTED}|{DocItemLabel.CHECKBOX_SELECTED}|"
-                rf"{DocItemLabel.TEXT}|{DocItemLabel.PAGE_HEADER}|"
+                rf"{DocItemLabel.TEXT}|{DocItemLabel.PAGE_HEADER}|{GroupLabel.INLINE}|"
                 rf"{DocItemLabel.PAGE_FOOTER}|{DocItemLabel.FORMULA}|"
                 rf"{DocItemLabel.CAPTION}|{DocItemLabel.PICTURE}|"
                 rf"{DocItemLabel.FOOTNOTE}|{DocItemLabel.CODE}|"
@@ -3648,7 +3754,7 @@ class DoclingDocument(BaseModel):
                     # no closing tag; only the existence of the item is recovered
                     full_chunk = f"<{tag_name}></{tag_name}>"
 
-                doc_label = tag_to_doclabel.get(tag_name, DocItemLabel.PARAGRAPH)
+                doc_label = tag_to_doclabel.get(tag_name, DocItemLabel.TEXT)
 
                 if tag_name == DocumentToken.OTSL.value:
                     table_data = parse_table_content(full_chunk)
@@ -3671,6 +3777,24 @@ class DoclingDocument(BaseModel):
                     else:
                         doc.add_table(data=table_data, caption=caption)
 
+                elif tag_name == GroupLabel.INLINE:
+                    inline_group = doc.add_inline_group()
+                    content = match.group("content")
+                    common_bbox = extract_bounding_box(content)
+                    for item_match in pattern.finditer(content):
+                        item_tag = item_match.group("tag")
+                        _add_text(
+                            full_chunk=item_match.group(0),
+                            bbox=common_bbox,
+                            pg_width=pg_width,
+                            pg_height=pg_height,
+                            page_no=page_no,
+                            tag_name=item_tag,
+                            doc_label=tag_to_doclabel.get(item_tag, DocItemLabel.TEXT),
+                            doc=doc,
+                            parent=inline_group,
+                        )
+
                 elif tag_name in [DocItemLabel.PICTURE, DocItemLabel.CHART]:
                     caption, caption_bbox = extract_caption(full_chunk)
                     table_data = None
@@ -3820,38 +3944,17 @@ class DoclingDocument(BaseModel):
                         )
                 else:
                     # For everything else, treat as text
-                    text_content = extract_inner_text(full_chunk)
-                    element_prov = (
-                        ProvenanceItem(
-                            bbox=bbox.resize_by_scale(pg_width, pg_height),
-                            charspan=(0, len(text_content)),
-                            page_no=page_no,
-                        )
-                        if bbox
-                        else None
+                    _add_text(
+                        full_chunk=full_chunk,
+                        bbox=bbox,
+                        pg_width=pg_width,
+                        pg_height=pg_height,
+                        page_no=page_no,
+                        tag_name=tag_name,
+                        doc_label=doc_label,
+                        doc=doc,
+                        parent=None,
                     )
-
-                    content_layer = ContentLayer.BODY
-                    if tag_name in [DocItemLabel.PAGE_HEADER, DocItemLabel.PAGE_FOOTER]:
-                        content_layer = ContentLayer.FURNITURE
-
-                    if doc_label == DocItemLabel.SECTION_HEADER:
-                        # Extract level from tag_name (e.g. "section_level_header_1" -> 1)
-                        level = int(tag_name.split("_")[-1])
-                        doc.add_heading(
-                            text=text_content,
-                            level=level,
-                            prov=element_prov,
-                            content_layer=content_layer,
-                        )
-                    else:
-                        doc.add_text(
-                            label=doc_label,
-                            text=text_content,
-                            prov=element_prov,
-                            content_layer=content_layer,
-                        )
-
         return doc
 
     @deprecated("Use save_as_doctags instead.")
@@ -4149,3 +4252,58 @@ class DoclingDocument(BaseModel):
                 raise ValueError("Document hierachy is inconsistent.")
 
         return d
+
+    @model_validator(mode="after")
+    def validate_misplaced_list_items(self):
+        """validate_misplaced_list_items."""
+        # find list items without list parent, putting succesive ones together
+        misplaced_list_items: list[list[ListItem]] = []
+        prev: Optional[NodeItem] = None
+        for item, _ in self.iterate_items(
+            traverse_pictures=True,
+            included_content_layers={c for c in ContentLayer},
+            with_groups=True,  # so that we can distinguish neighboring lists
+        ):
+            if isinstance(item, ListItem) and (
+                item.parent is None
+                or not isinstance(
+                    item.parent.resolve(doc=self), (OrderedList, UnorderedList)
+                )
+            ):
+                # non_group_list_items.append(item)
+                if prev is None or not isinstance(prev, ListItem):  # if new list
+                    misplaced_list_items.append([item])
+                else:
+                    misplaced_list_items[-1].append(item)
+            prev = item
+
+        for curr_list_items in reversed(misplaced_list_items):
+
+            # add group
+            new_group = (
+                OrderedList(self_ref="#")
+                if curr_list_items[0].enumerated
+                else UnorderedList(self_ref="#")
+            )
+            self.insert_item_before_sibling(
+                new_item=new_group,
+                sibling=curr_list_items[0],
+            )
+
+            # delete list items from document (should not be affected by group addition)
+            self.delete_items(node_items=curr_list_items)
+
+            # add list items to new group
+            for li in curr_list_items:
+                self.add_list_item(
+                    text=li.text,
+                    enumerated=li.enumerated,
+                    marker=li.marker,
+                    orig=li.orig,
+                    prov=li.prov[0] if li.prov else None,
+                    parent=new_group,
+                    content_layer=li.content_layer,
+                    formatting=li.formatting,
+                    hyperlink=li.hyperlink,
+                )
+        return self

{docling_core-2.33.1.dist-info → docling_core-2.34.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling-core
-Version: 2.33.1
+Version: 2.34.0
 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>

{docling_core-2.33.1.dist-info → docling_core-2.34.0.dist-info}/RECORD

@@ -26,12 +26,12 @@ docling_core/transforms/chunker/tokenizer/base.py,sha256=2gOBQPYJYC0iWXOgMG3DiNP
 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=mkajw0QRL--WgVL42Vlp2e2PuUQVh79D6EKP4_3YKy0,18112
-docling_core/transforms/serializer/doctags.py,sha256=mEmRWVuebcG5pZcR1_HX146cyUk0_FjaLQtMXSgh9hs,17870
-docling_core/transforms/serializer/html.py,sha256=_HN1WFKH_WJkxtZrmvm1a6-UDxsEGt_ChWdUysS1qjY,35843
+docling_core/transforms/serializer/base.py,sha256=ZFIiZeplL-QbBs9EDUb1awqxapQ23PsApVetJtAs7Vs,6891
+docling_core/transforms/serializer/common.py,sha256=WP-qO-woidrKyvZ56m0vlKMysoLrMzzZtHSCIwsl3ek,19119
+docling_core/transforms/serializer/doctags.py,sha256=PuAExlP-2HxcDSP_R_phtYQU0yKBW94RrPgb85IUxck,19905
+docling_core/transforms/serializer/html.py,sha256=KiywrroYBS3yk07gQizlmk3oqkXg_NpFwE0VF31_Z-I,37112
 docling_core/transforms/serializer/html_styles.py,sha256=-jBwS4EU7yfKoz0GSoxhwx90OmIKieO6TwPw57IuxcA,4692
-docling_core/transforms/serializer/markdown.py,sha256=ussKqIptiKPTCRNjy3edjap4DOsy52no-FLSeAyv9S0,18759
+docling_core/transforms/serializer/markdown.py,sha256=wfMNrjA4wMehWLCejAhEN1eQPRixUO1SyL6ojkKkzZY,20614
 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=ndXquBrOKTFQApIJ5s2-zstj3xlVKRbJDSId0KOQnUg,14817
-docling_core/types/doc/document.py,sha256=rdevCAZDpMPzPlZmAtiucvBM8h_AjuIZpQDaqjpknl0,142796
+docling_core/types/doc/document.py,sha256=VKZg1VT-H8gTXybgY6lRlcKKR3f6mFDB9UzcrLtII5I,148197
 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
@@ -73,9 +73,9 @@ 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-2.33.1.dist-info/licenses/LICENSE,sha256=2M9-6EoQ1sxFztTOkXGAtwUDJvnWaAHdB9BYWVwGkIw,1087
-docling_core-2.33.1.dist-info/METADATA,sha256=tib261Wc010Z2y6_lgKcXdO2OKPG8pdf2n1CoIYSDBA,6453
-docling_core-2.33.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-docling_core-2.33.1.dist-info/entry_points.txt,sha256=ER4zROQWkFMHIrY-oqY5E4HeCcCIg8dLkNztYGxdb7c,59
-docling_core-2.33.1.dist-info/top_level.txt,sha256=O-tcXpGiurlud-1ZxMq1b-OmrfAVA4sajcgWU32RtfA,13
-docling_core-2.33.1.dist-info/RECORD,,
+docling_core-2.34.0.dist-info/licenses/LICENSE,sha256=2M9-6EoQ1sxFztTOkXGAtwUDJvnWaAHdB9BYWVwGkIw,1087
+docling_core-2.34.0.dist-info/METADATA,sha256=853af3C8OZrbXzZqYFhfDfu-gtG4m7my-6wqzCir_cg,6453
+docling_core-2.34.0.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+docling_core-2.34.0.dist-info/entry_points.txt,sha256=ER4zROQWkFMHIrY-oqY5E4HeCcCIg8dLkNztYGxdb7c,59
+docling_core-2.34.0.dist-info/top_level.txt,sha256=O-tcXpGiurlud-1ZxMq1b-OmrfAVA4sajcgWU32RtfA,13
+docling_core-2.34.0.dist-info/RECORD,,