feat: enable precision control in float serialization (#1914)

* chore: propagate precision control in float serialization

Signed-off-by: Panos Vagenas <[email protected]>

* parametrize float serialization, propagate core updates

Signed-off-by: Panos Vagenas <[email protected]>

* update test float precision

Signed-off-by: Panos Vagenas <[email protected]>

* repin docling-core

Signed-off-by: Panos Vagenas <[email protected]>

---------

Signed-off-by: Panos Vagenas <[email protected]>

Author: vagenas

docling/datamodel/base_models.py

@@ -12,14 +12,22 @@
     Size,
     TableCell,
 )
+from docling_core.types.doc.base import PydanticSerCtxKey, round_pydantic_float
 from docling_core.types.doc.page import SegmentedPdfPage, TextCell
 from docling_core.types.io import (
     DocumentStream,
 )
 
 # DO NOT REMOVE; explicitly exposed from this location
 from PIL.Image import Image
-from pydantic import BaseModel, ConfigDict, Field, computed_field
+from pydantic import (
+    BaseModel,
+    ConfigDict,
+    Field,
+    FieldSerializationInfo,
+    computed_field,
+    field_serializer,
+)
 
 if TYPE_CHECKING:
     from docling.backend.pdf_backend import PdfPageBackend
@@ -142,6 +150,10 @@ class Cluster(BaseModel):
     cells: List[TextCell] = []
     children: List["Cluster"] = []  # Add child cluster support
 
+    @field_serializer("confidence")
+    def _serialize(self, value: float, info: FieldSerializationInfo) -> float:
+        return round_pydantic_float(value, info.context, PydanticSerCtxKey.CONFID_PREC)
+
 
 class BasePageElement(BaseModel):
     label: DocItemLabel
@@ -194,6 +206,16 @@ class FigureElement(BasePageElement):
     predicted_class: Optional[str] = None
     confidence: Optional[float] = None
 
+    @field_serializer("confidence")
+    def _serialize(
+        self, value: Optional[float], info: FieldSerializationInfo
+    ) -> Optional[float]:
+        return (
+            round_pydantic_float(value, info.context, PydanticSerCtxKey.CONFID_PREC)
+            if value is not None
+            else None
+        )
+
 
 class FigureClassificationPrediction(BaseModel):
     figure_count: int = 0

pyproject.toml

@@ -44,7 +44,7 @@ authors = [
 requires-python = '>=3.9,<4.0'
 dependencies = [
   'pydantic (>=2.0.0,<3.0.0)',
-  'docling-core[chunking] (>=2.40.0,<3.0.0)',
+  'docling-core[chunking] (>=2.42.0,<3.0.0)',
   'docling-parse (>=4.0.0,<5.0.0)',
   'docling-ibm-models (>=3.6.0,<4)',
   'filetype (>=1.2.0,<2.0.0)',