Put stub for experimental format export

Signed-off-by: Christoph Auer <[email protected]>

docling/datamodel/document.py

@@ -3,7 +3,7 @@
 from pathlib import Path, PurePath
 from typing import ClassVar, Dict, Iterable, List, Optional, Tuple, Type, Union
 
-from docling_core.types import BaseCell, BaseText
+from docling_core.types import BaseCell, BaseText, DoclingDocument
 from docling_core.types import Document as DsDocument
 from docling_core.types import DocumentDescription as DsDocumentDescription
 from docling_core.types import FileInfoObject as DsFileInfoObject
@@ -60,6 +60,8 @@
     ),
 )
 
+_EMPTY_DOCLING_DOC = DoclingDocument(description={}, file_info={})  # TODO: Stub
+
 
 class InputDocument(BaseModel):
     file: PurePath = None
@@ -137,6 +139,7 @@ class ConvertedDocument(BaseModel):
     assembled: AssembledUnit = AssembledUnit()
 
     output: DsDocument = _EMPTY_DOC
+    experimental: DoclingDocument = _EMPTY_DOCLING_DOC
 
     def _to_ds_document(self) -> DsDocument:
         title = ""

docling/document_converter.py

@@ -289,4 +289,4 @@ def _assemble_doc(self, conv_res: ConversionResult):
             elements=all_elements, headers=all_headers, body=all_body
         )
 
-        conv_res.output = self.glm_model(conv_res)
+        conv_res.output, conv_res.experimental = self.glm_model(conv_res)

docling/models/ds_glm_model.py

@@ -1,10 +1,14 @@
 import copy
 import random
+from typing import Tuple
 
 from deepsearch_glm.nlp_utils import init_nlp_model
-from deepsearch_glm.utils.doc_utils import to_legacy_document_format
+from deepsearch_glm.utils.doc_utils import (
+    to_docling_document,
+    to_legacy_document_format,
+)
 from deepsearch_glm.utils.load_pretrained_models import load_pretrained_nlp_models
-from docling_core.types import BaseText
+from docling_core.types import BaseText, DoclingDocument
 from docling_core.types import Document as DsDocument
 from docling_core.types import Ref
 from docling_core.types.experimental.base import BoundingBox, CoordOrigin
@@ -25,7 +29,9 @@ def __init__(self, config):
         model = init_nlp_model(model_names=self.model_names)
         self.model = model
 
-    def __call__(self, conv_res: ConversionResult) -> DsDocument:
+    def __call__(
+        self, conv_res: ConversionResult
+    ) -> Tuple[DsDocument, DoclingDocument]:
         ds_doc = conv_res._to_ds_document()
         ds_doc_dict = ds_doc.model_dump(by_alias=True)
 
@@ -34,6 +40,7 @@ def __call__(self, conv_res: ConversionResult) -> DsDocument:
             glm_doc, ds_doc_dict, update_name_label=True
         )
 
+        docling_doc: DoclingDocument = to_docling_document(glm_doc)  # Experimental
         exported_doc = DsDocument.model_validate(ds_doc_dict)
 
         # DEBUG code:
@@ -84,4 +91,4 @@ def draw_clusters_and_cells(ds_document, page_no):
         # draw_clusters_and_cells(ds_doc, 0)
         # draw_clusters_and_cells(exported_doc, 0)
 
-        return exported_doc
+        return (exported_doc, docling_doc)

examples/batch_convert.py

@@ -4,6 +4,8 @@
 from pathlib import Path
 from typing import Iterable
 
+import yaml
+
 from docling.datamodel.base_models import ConversionStatus, PipelineOptions
 from docling.datamodel.document import ConversionResult, DocumentConversionInput
 from docling.document_converter import DocumentConverter
@@ -30,6 +32,14 @@ def export_documents(
             with (output_dir / f"{doc_filename}.json").open("w") as fp:
                 fp.write(json.dumps(conv_res.render_as_dict()))
 
+            # Export Docling document format to YAML (experimental):
+            with (output_dir / f"{doc_filename}.yaml").open("w") as fp:
+                fp.write(
+                    yaml.safe_dump(
+                        conv_res.experimental.model_dump(mode="json", by_alias=True)
+                    )
+                )
+
             # Export Text format:
             with (output_dir / f"{doc_filename}.txt").open("w") as fp:
                 fp.write(conv_res.render_as_text())