fix: DoclingDocument model validator should deal with any raw input

Signed-off-by: Cesar Berrospi Ramis <[email protected]>

Author: ceberam

docling_core/types/doc/document.py

@@ -2308,12 +2308,12 @@ class DoclingDocument(BaseModel):
 
     @model_validator(mode="before")
     @classmethod
-    def transform_to_content_layer(cls, data: dict) -> dict:
+    def transform_to_content_layer(cls, data: Any) -> Any:
         """transform_to_content_layer."""
         # Since version 1.1.0, all NodeItems carry content_layer property.
         # We must assign previous page_header and page_footer instances to furniture.
         # Note: model_validators which check on the version must use "before".
-        if "version" in data and data["version"] == "1.0.0":
+        if isinstance(data, dict) and data.get("version", "") == "1.0.0":
             for item in data.get("texts", []):
                 if "label" in item and item["label"] in [
                     DocItemLabel.PAGE_HEADER.value,

test/test_docling_doc.py

@@ -9,7 +9,7 @@
 import yaml
 from PIL import Image as PILImage
 from PIL import ImageDraw
-from pydantic import AnyUrl, ValidationError
+from pydantic import AnyUrl, BaseModel, ValidationError
 
 from docling_core.types.doc.base import BoundingBox, CoordOrigin, ImageRefMode, Size
 from docling_core.types.doc.document import (  # BoundingBox,
@@ -1156,6 +1156,14 @@ def test_upgrade_content_layer_from_1_0_0():
     assert doc.version == CURRENT_VERSION
     assert doc.texts[0].content_layer == ContentLayer.FURNITURE
 
+    # test that transform_to_content_layer model validator can handle any data type
+    class ContentOutput(BaseModel):
+        content: str | DoclingDocument
+
+    co = ContentOutput.model_validate_json('{"content": "Random string with version"}')
+    assert co
+    assert isinstance(co.content, str)
+
 
 def test_version_doc():