docling 2.51.0__py3-none-any.whl → 2.52.0__py3-none-any.whl

docling/cli/main.py

@@ -48,6 +48,7 @@ from docling.datamodel.base_models import (
 from docling.datamodel.document import ConversionResult
 from docling.datamodel.pipeline_options import (
     AsrPipelineOptions,
+    ConvertPipelineOptions,
     EasyOcrOptions,
     OcrOptions,
     PaginatedPipelineOptions,
@@ -71,8 +72,13 @@ from docling.datamodel.vlm_model_specs import (
 from docling.document_converter import (
     AudioFormatOption,
     DocumentConverter,
+    ExcelFormatOption,
     FormatOption,
+    HTMLFormatOption,
+    MarkdownFormatOption,
     PdfFormatOption,
+    PowerpointFormatOption,
+    WordFormatOption,
 )
 from docling.models.factories import get_ocr_factory
 from docling.pipeline.asr_pipeline import AsrPipeline
@@ -626,10 +632,33 @@ def convert(  # noqa: C901
                 backend=MetsGbsDocumentBackend,
             )
 
+            # SimplePipeline options
+            simple_format_option = ConvertPipelineOptions(
+                do_picture_description=enrich_picture_description,
+                do_picture_classification=enrich_picture_classes,
+            )
+            if artifacts_path is not None:
+                simple_format_option.artifacts_path = artifacts_path
+
             format_options = {
                 InputFormat.PDF: pdf_format_option,
                 InputFormat.IMAGE: pdf_format_option,
                 InputFormat.METS_GBS: mets_gbs_format_option,
+                InputFormat.DOCX: WordFormatOption(
+                    pipeline_options=simple_format_option
+                ),
+                InputFormat.PPTX: PowerpointFormatOption(
+                    pipeline_options=simple_format_option
+                ),
+                InputFormat.XLSX: ExcelFormatOption(
+                    pipeline_options=simple_format_option
+                ),
+                InputFormat.HTML: HTMLFormatOption(
+                    pipeline_options=simple_format_option
+                ),
+                InputFormat.MD: MarkdownFormatOption(
+                    pipeline_options=simple_format_option
+                ),
             }
 
         elif pipeline == ProcessingPipeline.VLM:

docling/datamodel/pipeline_options.py

@@ -135,6 +135,8 @@ class EasyOcrOptions(OcrOptions):
     recog_network: Optional[str] = "standard"
     download_enabled: bool = True
 
+    suppress_mps_warnings: bool = True
+
     model_config = ConfigDict(
         extra="forbid",
         protected_namespaces=(),
@@ -257,11 +259,21 @@ class PipelineOptions(BaseOptions):
     accelerator_options: AcceleratorOptions = AcceleratorOptions()
     enable_remote_services: bool = False
     allow_external_plugins: bool = False
+    artifacts_path: Optional[Union[Path, str]] = None
 
 
-class PaginatedPipelineOptions(PipelineOptions):
-    artifacts_path: Optional[Union[Path, str]] = None
+class ConvertPipelineOptions(PipelineOptions):
+    """Base convert pipeline options."""
+
+    do_picture_classification: bool = False  # True: classify pictures in documents
+
+    do_picture_description: bool = False  # True: run describe pictures in documents
+    picture_description_options: PictureDescriptionBaseOptions = (
+        smolvlm_picture_description
+    )
+
 
+class PaginatedPipelineOptions(ConvertPipelineOptions):
     images_scale: float = 1.0
     generate_page_images: bool = False
     generate_picture_images: bool = False
@@ -293,13 +305,11 @@ class LayoutOptions(BaseModel):
 
 class AsrPipelineOptions(PipelineOptions):
     asr_options: Union[InlineAsrOptions] = asr_model_specs.WHISPER_TINY
-    artifacts_path: Optional[Union[Path, str]] = None
 
 
 class VlmExtractionPipelineOptions(PipelineOptions):
     """Options for extraction pipeline."""
 
-    artifacts_path: Optional[Union[Path, str]] = None
     vlm_options: Union[InlineVlmOptions] = NU_EXTRACT_2B_TRANSFORMERS
 
 
@@ -310,8 +320,6 @@ class PdfPipelineOptions(PaginatedPipelineOptions):
     do_ocr: bool = True  # True: perform OCR, replace programmatic PDF text
     do_code_enrichment: bool = False  # True: perform code OCR
     do_formula_enrichment: bool = False  # True: perform formula OCR, return Latex code
-    do_picture_classification: bool = False  # True: classify pictures in documents
-    do_picture_description: bool = False  # True: run describe pictures in documents
     force_backend_text: bool = (
         False  # (To be used with vlms, or other generative models)
     )
@@ -319,9 +327,6 @@ class PdfPipelineOptions(PaginatedPipelineOptions):
 
     table_structure_options: TableStructureOptions = TableStructureOptions()
     ocr_options: OcrOptions = EasyOcrOptions()
-    picture_description_options: PictureDescriptionBaseOptions = (
-        smolvlm_picture_description
-    )
     layout_options: LayoutOptions = LayoutOptions()
 
     images_scale: float = 1.0

docling/models/base_model.py

@@ -4,7 +4,13 @@ from collections.abc import Iterable
 from typing import Any, Generic, Optional, Protocol, Type, Union
 
 import numpy as np
-from docling_core.types.doc import BoundingBox, DocItem, DoclingDocument, NodeItem
+from docling_core.types.doc import (
+    BoundingBox,
+    DocItem,
+    DoclingDocument,
+    NodeItem,
+    PictureItem,
+)
 from PIL.Image import Image
 from typing_extensions import TypeVar
 
@@ -164,8 +170,17 @@ class BaseItemAndImageEnrichmentModel(
             return None
 
         assert isinstance(element, DocItem)
-        element_prov = element.prov[0]
 
+        # Allow the case of documents without page images but embedded images (e.g. Word and HTML docs)
+        if len(element.prov) == 0 and isinstance(element, PictureItem):
+            embedded_im = element.get_image(conv_res.document)
+            if embedded_im is not None:
+                return ItemAndImageEnrichmentElement(item=element, image=embedded_im)
+            else:
+                return None
+
+        # Crop the image form the page
+        element_prov = element.prov[0]
         bbox = element_prov.bbox
         width = bbox.r - bbox.l
         height = bbox.t - bbox.b
@@ -183,4 +198,14 @@ class BaseItemAndImageEnrichmentModel(
         cropped_image = conv_res.pages[page_ix].get_image(
             scale=self.images_scale, cropbox=expanded_bbox
         )
+
+        # Allow for images being embedded without the page backend or page images
+        if cropped_image is None and isinstance(element, PictureItem):
+            embedded_im = element.get_image(conv_res.document)
+            if embedded_im is not None:
+                return ItemAndImageEnrichmentElement(item=element, image=embedded_im)
+            else:
+                return None
+
+        # Return the proper cropped image
         return ItemAndImageEnrichmentElement(item=element, image=cropped_image)

docling/models/easyocr_model.py

@@ -78,14 +78,17 @@ class EasyOcrModel(BaseOcrModel):
                 download_enabled = False
                 model_storage_directory = str(artifacts_path / self._model_repo_folder)
 
-            self.reader = easyocr.Reader(
-                lang_list=self.options.lang,
-                gpu=use_gpu,
-                model_storage_directory=model_storage_directory,
-                recog_network=self.options.recog_network,
-                download_enabled=download_enabled,
-                verbose=False,
-            )
+            with warnings.catch_warnings():
+                if self.options.suppress_mps_warnings:
+                    warnings.filterwarnings("ignore", message=".*pin_memory.*MPS.*")
+                self.reader = easyocr.Reader(
+                    lang_list=self.options.lang,
+                    gpu=use_gpu,
+                    model_storage_directory=model_storage_directory,
+                    recog_network=self.options.recog_network,
+                    download_enabled=download_enabled,
+                    verbose=False,
+                )
 
     @staticmethod
     def download_models(
@@ -147,7 +150,14 @@ class EasyOcrModel(BaseOcrModel):
                             scale=self.scale, cropbox=ocr_rect
                         )
                         im = numpy.array(high_res_image)
-                        result = self.reader.readtext(im)
+
+                        with warnings.catch_warnings():
+                            if self.options.suppress_mps_warnings:
+                                warnings.filterwarnings(
+                                    "ignore", message=".*pin_memory.*MPS.*"
+                                )
+
+                            result = self.reader.readtext(im)
 
                         del high_res_image
                         del im

docling/models/picture_description_vlm_model.py

@@ -67,7 +67,7 @@ class PictureDescriptionVlmModel(
                 self.model = AutoModelForImageTextToText.from_pretrained(
                     artifacts_path,
                     device_map=self.device,
-                    torch_dtype=torch.bfloat16,
+                    dtype=torch.bfloat16,
                     _attn_implementation=(
                         "flash_attention_2"
                         if self.device.startswith("cuda")

docling/models/vlm_models_inline/hf_transformers_model.py

@@ -112,7 +112,7 @@ class HuggingFaceTransformersVlmModel(BaseVlmPageModel, HuggingFaceModelDownload
             self.vlm_model = model_cls.from_pretrained(
                 artifacts_path,
                 device_map=self.device,
-                torch_dtype=self.vlm_options.torch_dtype,
+                dtype=self.vlm_options.torch_dtype,
                 _attn_implementation=(
                     "flash_attention_2"
                     if self.device.startswith("cuda")

docling/models/vlm_models_inline/nuextract_transformers_model.py

@@ -144,7 +144,7 @@ class NuExtractTransformersModel(BaseVlmModel, HuggingFaceModelDownloadMixin):
             self.vlm_model = AutoModelForImageTextToText.from_pretrained(
                 artifacts_path,
                 device_map=self.device,
-                torch_dtype=self.vlm_options.torch_dtype,
+                dtype=self.vlm_options.torch_dtype,
                 _attn_implementation=(
                     "flash_attention_2"
                     if self.device.startswith("cuda")

docling/pipeline/asr_pipeline.py

@@ -208,25 +208,13 @@ class AsrPipeline(BasePipeline):
 
         self.pipeline_options: AsrPipelineOptions = pipeline_options
 
-        artifacts_path: Optional[Path] = None
-        if pipeline_options.artifacts_path is not None:
-            artifacts_path = Path(pipeline_options.artifacts_path).expanduser()
-        elif settings.artifacts_path is not None:
-            artifacts_path = Path(settings.artifacts_path).expanduser()
-
-        if artifacts_path is not None and not artifacts_path.is_dir():
-            raise RuntimeError(
-                f"The value of {artifacts_path=} is not valid. "
-                "When defined, it must point to a folder containing all models required by the pipeline."
-            )
-
         if isinstance(self.pipeline_options.asr_options, InlineAsrNativeWhisperOptions):
             asr_options: InlineAsrNativeWhisperOptions = (
                 self.pipeline_options.asr_options
             )
             self._model = _NativeWhisperModel(
                 enabled=True,  # must be always enabled for this pipeline to make sense.
-                artifacts_path=artifacts_path,
+                artifacts_path=self.artifacts_path,
                 accelerator_options=pipeline_options.accelerator_options,
                 asr_options=asr_options,
             )

docling/pipeline/base_extraction_pipeline.py

@@ -1,19 +1,33 @@
 import logging
 from abc import ABC, abstractmethod
+from pathlib import Path
 from typing import Optional
 
 from docling.datamodel.base_models import ConversionStatus, ErrorItem
 from docling.datamodel.document import InputDocument
 from docling.datamodel.extraction import ExtractionResult, ExtractionTemplateType
-from docling.datamodel.pipeline_options import BaseOptions
+from docling.datamodel.pipeline_options import BaseOptions, PipelineOptions
+from docling.datamodel.settings import settings
 
 _log = logging.getLogger(__name__)
 
 
 class BaseExtractionPipeline(ABC):
-    def __init__(self, pipeline_options: BaseOptions):
+    def __init__(self, pipeline_options: PipelineOptions):
         self.pipeline_options = pipeline_options
 
+        self.artifacts_path: Optional[Path] = None
+        if pipeline_options.artifacts_path is not None:
+            self.artifacts_path = Path(pipeline_options.artifacts_path).expanduser()
+        elif settings.artifacts_path is not None:
+            self.artifacts_path = Path(settings.artifacts_path).expanduser()
+
+        if self.artifacts_path is not None and not self.artifacts_path.is_dir():
+            raise RuntimeError(
+                f"The value of {self.artifacts_path=} is not valid. "
+                "When defined, it must point to a folder containing all models required by the pipeline."
+            )
+
     def execute(
         self,
         in_doc: InputDocument,
@@ -54,5 +68,5 @@ class BaseExtractionPipeline(ABC):
 
     @classmethod
     @abstractmethod
-    def get_default_options(cls) -> BaseOptions:
+    def get_default_options(cls) -> PipelineOptions:
         pass

docling/pipeline/base_pipeline.py

@@ -4,7 +4,8 @@ import time
 import traceback
 from abc import ABC, abstractmethod
 from collections.abc import Iterable
-from typing import Any, Callable, List
+from pathlib import Path
+from typing import Any, Callable, List, Optional
 
 from docling_core.types.doc import NodeItem
 
@@ -20,9 +21,19 @@ from docling.datamodel.base_models import (
     Page,
 )
 from docling.datamodel.document import ConversionResult, InputDocument
-from docling.datamodel.pipeline_options import PdfPipelineOptions, PipelineOptions
+from docling.datamodel.pipeline_options import (
+    ConvertPipelineOptions,
+    PdfPipelineOptions,
+    PipelineOptions,
+)
 from docling.datamodel.settings import settings
 from docling.models.base_model import GenericEnrichmentModel
+from docling.models.document_picture_classifier import (
+    DocumentPictureClassifier,
+    DocumentPictureClassifierOptions,
+)
+from docling.models.factories import get_picture_description_factory
+from docling.models.picture_description_base_model import PictureDescriptionBaseModel
 from docling.utils.profiling import ProfilingScope, TimeRecorder
 from docling.utils.utils import chunkify
 
@@ -36,6 +47,18 @@ class BasePipeline(ABC):
         self.build_pipe: List[Callable] = []
         self.enrichment_pipe: List[GenericEnrichmentModel[Any]] = []
 
+        self.artifacts_path: Optional[Path] = None
+        if pipeline_options.artifacts_path is not None:
+            self.artifacts_path = Path(pipeline_options.artifacts_path).expanduser()
+        elif settings.artifacts_path is not None:
+            self.artifacts_path = Path(settings.artifacts_path).expanduser()
+
+        if self.artifacts_path is not None and not self.artifacts_path.is_dir():
+            raise RuntimeError(
+                f"The value of {self.artifacts_path=} is not valid. "
+                "When defined, it must point to a folder containing all models required by the pipeline."
+            )
+
     def execute(self, in_doc: InputDocument, raises_on_error: bool) -> ConversionResult:
         conv_res = ConversionResult(input=in_doc)
 
@@ -108,15 +131,58 @@ class BasePipeline(ABC):
     def is_backend_supported(cls, backend: AbstractDocumentBackend):
         pass
 
-    # def _apply_on_elements(self, element_batch: Iterable[NodeItem]) -> Iterable[Any]:
-    #    for model in self.build_pipe:
-    #        element_batch = model(element_batch)
-    #
-    #    yield from element_batch
 
+class ConvertPipeline(BasePipeline):
+    def __init__(self, pipeline_options: ConvertPipelineOptions):
+        super().__init__(pipeline_options)
+        self.pipeline_options: ConvertPipelineOptions
 
-class PaginatedPipeline(BasePipeline):  # TODO this is a bad name.
-    def __init__(self, pipeline_options: PipelineOptions):
+        # ------ Common enrichment models working on all backends
+
+        # Picture description model
+        if (
+            picture_description_model := self._get_picture_description_model(
+                artifacts_path=self.artifacts_path
+            )
+        ) is None:
+            raise RuntimeError(
+                f"The specified picture description kind is not supported: {pipeline_options.picture_description_options.kind}."
+            )
+
+        self.enrichment_pipe = [
+            # Document Picture Classifier
+            DocumentPictureClassifier(
+                enabled=pipeline_options.do_picture_classification,
+                artifacts_path=self.artifacts_path,
+                options=DocumentPictureClassifierOptions(),
+                accelerator_options=pipeline_options.accelerator_options,
+            ),
+            # Document Picture description
+            picture_description_model,
+        ]
+
+    def _get_picture_description_model(
+        self, artifacts_path: Optional[Path] = None
+    ) -> Optional[PictureDescriptionBaseModel]:
+        factory = get_picture_description_factory(
+            allow_external_plugins=self.pipeline_options.allow_external_plugins
+        )
+        return factory.create_instance(
+            options=self.pipeline_options.picture_description_options,
+            enabled=self.pipeline_options.do_picture_description,
+            enable_remote_services=self.pipeline_options.enable_remote_services,
+            artifacts_path=artifacts_path,
+            accelerator_options=self.pipeline_options.accelerator_options,
+        )
+
+    @classmethod
+    @abstractmethod
+    def get_default_options(cls) -> ConvertPipelineOptions:
+        pass
+
+
+class PaginatedPipeline(ConvertPipeline):  # TODO this is a bad name.
+    def __init__(self, pipeline_options: ConvertPipelineOptions):
         super().__init__(pipeline_options)
         self.keep_backend = False
 

docling/pipeline/extraction_vlm_pipeline.py

@@ -1,7 +1,6 @@
 import inspect
 import json
 import logging
-from pathlib import Path
 from typing import Optional
 
 from PIL.Image import Image
@@ -16,7 +15,10 @@ from docling.datamodel.extraction import (
     ExtractionResult,
     ExtractionTemplateType,
 )
-from docling.datamodel.pipeline_options import BaseOptions, VlmExtractionPipelineOptions
+from docling.datamodel.pipeline_options import (
+    PipelineOptions,
+    VlmExtractionPipelineOptions,
+)
 from docling.datamodel.settings import settings
 from docling.models.vlm_models_inline.nuextract_transformers_model import (
     NuExtractTransformersModel,
@@ -35,22 +37,10 @@ class ExtractionVlmPipeline(BaseExtractionPipeline):
         self.accelerator_options = pipeline_options.accelerator_options
         self.pipeline_options: VlmExtractionPipelineOptions
 
-        artifacts_path: Optional[Path] = None
-        if pipeline_options.artifacts_path is not None:
-            artifacts_path = Path(pipeline_options.artifacts_path).expanduser()
-        elif settings.artifacts_path is not None:
-            artifacts_path = Path(settings.artifacts_path).expanduser()
-
-        if artifacts_path is not None and not artifacts_path.is_dir():
-            raise RuntimeError(
-                f"The value of {artifacts_path=} is not valid. "
-                "When defined, it must point to a folder containing all models required by the pipeline."
-            )
-
         # Create VLM model instance
         self.vlm_model = NuExtractTransformersModel(
             enabled=True,
-            artifacts_path=artifacts_path,  # Will download automatically
+            artifacts_path=self.artifacts_path,  # Will download automatically
             accelerator_options=self.accelerator_options,
             vlm_options=pipeline_options.vlm_options,
         )
@@ -194,11 +184,14 @@ class ExtractionVlmPipeline(BaseExtractionPipeline):
             class ExtractionTemplateFactory(ModelFactory[template]):  # type: ignore
                 __use_examples__ = True  # prefer Field(examples=...) when present
                 __use_defaults__ = True  # use field defaults instead of random values
+                __check_model__ = (
+                    True  # setting the value to avoid deprecation warnings
+                )
 
             return ExtractionTemplateFactory.build().model_dump_json(indent=2)  # type: ignore
         else:
             raise ValueError(f"Unsupported template type: {type(template)}")
 
     @classmethod
-    def get_default_options(cls) -> BaseOptions:
+    def get_default_options(cls) -> PipelineOptions:
         return VlmExtractionPipelineOptions()

docling/pipeline/simple_pipeline.py

@@ -6,21 +6,21 @@ from docling.backend.abstract_backend import (
 )
 from docling.datamodel.base_models import ConversionStatus
 from docling.datamodel.document import ConversionResult
-from docling.datamodel.pipeline_options import PipelineOptions
-from docling.pipeline.base_pipeline import BasePipeline
+from docling.datamodel.pipeline_options import ConvertPipelineOptions
+from docling.pipeline.base_pipeline import ConvertPipeline
 from docling.utils.profiling import ProfilingScope, TimeRecorder
 
 _log = logging.getLogger(__name__)
 
 
-class SimplePipeline(BasePipeline):
+class SimplePipeline(ConvertPipeline):
     """SimpleModelPipeline.
 
     This class is used at the moment for formats / backends
     which produce straight DoclingDocument output.
     """
 
-    def __init__(self, pipeline_options: PipelineOptions):
+    def __init__(self, pipeline_options: ConvertPipelineOptions):
         super().__init__(pipeline_options)
 
     def _build_document(self, conv_res: ConversionResult) -> ConversionResult:
@@ -47,8 +47,8 @@ class SimplePipeline(BasePipeline):
         return ConversionStatus.SUCCESS
 
     @classmethod
-    def get_default_options(cls) -> PipelineOptions:
-        return PipelineOptions()
+    def get_default_options(cls) -> ConvertPipelineOptions:
+        return ConvertPipelineOptions()
 
     @classmethod
     def is_backend_supported(cls, backend: AbstractDocumentBackend):

docling/pipeline/standard_pdf_pipeline.py

@@ -15,18 +15,13 @@ from docling.datamodel.pipeline_options import PdfPipelineOptions
 from docling.datamodel.settings import settings
 from docling.models.base_ocr_model import BaseOcrModel
 from docling.models.code_formula_model import CodeFormulaModel, CodeFormulaModelOptions
-from docling.models.document_picture_classifier import (
-    DocumentPictureClassifier,
-    DocumentPictureClassifierOptions,
-)
-from docling.models.factories import get_ocr_factory, get_picture_description_factory
+from docling.models.factories import get_ocr_factory
 from docling.models.layout_model import LayoutModel
 from docling.models.page_assemble_model import PageAssembleModel, PageAssembleOptions
 from docling.models.page_preprocessing_model import (
     PagePreprocessingModel,
     PagePreprocessingOptions,
 )
-from docling.models.picture_description_base_model import PictureDescriptionBaseModel
 from docling.models.readingorder_model import ReadingOrderModel, ReadingOrderOptions
 from docling.models.table_structure_model import TableStructureModel
 from docling.pipeline.base_pipeline import PaginatedPipeline
@@ -41,18 +36,6 @@ class StandardPdfPipeline(PaginatedPipeline):
         super().__init__(pipeline_options)
         self.pipeline_options: PdfPipelineOptions
 
-        artifacts_path: Optional[Path] = None
-        if pipeline_options.artifacts_path is not None:
-            artifacts_path = Path(pipeline_options.artifacts_path).expanduser()
-        elif settings.artifacts_path is not None:
-            artifacts_path = Path(settings.artifacts_path).expanduser()
-
-        if artifacts_path is not None and not artifacts_path.is_dir():
-            raise RuntimeError(
-                f"The value of {artifacts_path=} is not valid. "
-                "When defined, it must point to a folder containing all models required by the pipeline."
-            )
-
         with warnings.catch_warnings():  # deprecated generate_table_images
             warnings.filterwarnings("ignore", category=DeprecationWarning)
             self.keep_images = (
@@ -63,7 +46,7 @@ class StandardPdfPipeline(PaginatedPipeline):
 
         self.reading_order_model = ReadingOrderModel(options=ReadingOrderOptions())
 
-        ocr_model = self.get_ocr_model(artifacts_path=artifacts_path)
+        ocr_model = self.get_ocr_model(artifacts_path=self.artifacts_path)
 
         self.build_pipe = [
             # Pre-processing
@@ -76,14 +59,14 @@ class StandardPdfPipeline(PaginatedPipeline):
             ocr_model,
             # Layout model
             LayoutModel(
-                artifacts_path=artifacts_path,
+                artifacts_path=self.artifacts_path,
                 accelerator_options=pipeline_options.accelerator_options,
                 options=pipeline_options.layout_options,
             ),
             # Table structure model
             TableStructureModel(
                 enabled=pipeline_options.do_table_structure,
-                artifacts_path=artifacts_path,
+                artifacts_path=self.artifacts_path,
                 options=pipeline_options.table_structure_options,
                 accelerator_options=pipeline_options.accelerator_options,
             ),
@@ -91,37 +74,19 @@ class StandardPdfPipeline(PaginatedPipeline):
             PageAssembleModel(options=PageAssembleOptions()),
         ]
 
-        # Picture description model
-        if (
-            picture_description_model := self.get_picture_description_model(
-                artifacts_path=artifacts_path
-            )
-        ) is None:
-            raise RuntimeError(
-                f"The specified picture description kind is not supported: {pipeline_options.picture_description_options.kind}."
-            )
-
         self.enrichment_pipe = [
             # Code Formula Enrichment Model
             CodeFormulaModel(
                 enabled=pipeline_options.do_code_enrichment
                 or pipeline_options.do_formula_enrichment,
-                artifacts_path=artifacts_path,
+                artifacts_path=self.artifacts_path,
                 options=CodeFormulaModelOptions(
                     do_code_enrichment=pipeline_options.do_code_enrichment,
                     do_formula_enrichment=pipeline_options.do_formula_enrichment,
                 ),
                 accelerator_options=pipeline_options.accelerator_options,
             ),
-            # Document Picture Classifier
-            DocumentPictureClassifier(
-                enabled=pipeline_options.do_picture_classification,
-                artifacts_path=artifacts_path,
-                options=DocumentPictureClassifierOptions(),
-                accelerator_options=pipeline_options.accelerator_options,
-            ),
-            # Document Picture description
-            picture_description_model,
+            *self.enrichment_pipe,
         ]
 
         if (
@@ -158,20 +123,6 @@ class StandardPdfPipeline(PaginatedPipeline):
             accelerator_options=self.pipeline_options.accelerator_options,
         )
 
-    def get_picture_description_model(
-        self, artifacts_path: Optional[Path] = None
-    ) -> Optional[PictureDescriptionBaseModel]:
-        factory = get_picture_description_factory(
-            allow_external_plugins=self.pipeline_options.allow_external_plugins
-        )
-        return factory.create_instance(
-            options=self.pipeline_options.picture_description_options,
-            enabled=self.pipeline_options.do_picture_description,
-            enable_remote_services=self.pipeline_options.enable_remote_services,
-            artifacts_path=artifacts_path,
-            accelerator_options=self.pipeline_options.accelerator_options,
-        )
-
     def initialize_page(self, conv_res: ConversionResult, page: Page) -> Page:
         with TimeRecorder(conv_res, "page_init"):
             page._backend = conv_res.input._backend.load_page(page.page_no)  # type: ignore

docling/pipeline/threaded_standard_pdf_pipeline.py

@@ -20,10 +20,14 @@ import itertools
 import logging
 import threading
 import time
+import warnings
 from collections import defaultdict, deque
 from dataclasses import dataclass, field
 from pathlib import Path
-from typing import Any, Iterable, List, Optional, Sequence, Tuple
+from typing import Any, Iterable, List, Optional, Sequence, Tuple, cast
+
+import numpy as np
+from docling_core.types.doc import DocItem, ImageRef, PictureItem, TableItem
 
 from docling.backend.abstract_backend import AbstractDocumentBackend
 from docling.backend.pdf_backend import PdfDocumentBackend
@@ -32,21 +36,16 @@ from docling.datamodel.document import ConversionResult
 from docling.datamodel.pipeline_options import ThreadedPdfPipelineOptions
 from docling.datamodel.settings import settings
 from docling.models.code_formula_model import CodeFormulaModel, CodeFormulaModelOptions
-from docling.models.document_picture_classifier import (
-    DocumentPictureClassifier,
-    DocumentPictureClassifierOptions,
-)
-from docling.models.factories import get_ocr_factory, get_picture_description_factory
+from docling.models.factories import get_ocr_factory
 from docling.models.layout_model import LayoutModel
 from docling.models.page_assemble_model import PageAssembleModel, PageAssembleOptions
 from docling.models.page_preprocessing_model import (
     PagePreprocessingModel,
     PagePreprocessingOptions,
 )
-from docling.models.picture_description_base_model import PictureDescriptionBaseModel
 from docling.models.readingorder_model import ReadingOrderModel, ReadingOrderOptions
 from docling.models.table_structure_model import TableStructureModel
-from docling.pipeline.base_pipeline import BasePipeline
+from docling.pipeline.base_pipeline import ConvertPipeline
 from docling.utils.profiling import ProfilingScope, TimeRecorder
 from docling.utils.utils import chunkify
 
@@ -294,7 +293,7 @@ class RunContext:
 # ──────────────────────────────────────────────────────────────────────────────
 
 
-class ThreadedStandardPdfPipeline(BasePipeline):
+class ThreadedStandardPdfPipeline(ConvertPipeline):
     """High-performance PDF pipeline with multi-threaded stages."""
 
     def __init__(self, pipeline_options: ThreadedPdfPipelineOptions) -> None:
@@ -310,7 +309,7 @@ class ThreadedStandardPdfPipeline(BasePipeline):
     # ────────────────────────────────────────────────────────────────────────
 
     def _init_models(self) -> None:
-        art_path = self._resolve_artifacts_path()
+        art_path = self.artifacts_path
         self.keep_images = (
             self.pipeline_options.generate_page_images
             or self.pipeline_options.generate_picture_images
@@ -337,32 +336,20 @@ class ThreadedStandardPdfPipeline(BasePipeline):
         self.reading_order_model = ReadingOrderModel(options=ReadingOrderOptions())
 
         # --- optional enrichment ------------------------------------------------
-        self.enrichment_pipe = []
-        code_formula = CodeFormulaModel(
-            enabled=self.pipeline_options.do_code_enrichment
-            or self.pipeline_options.do_formula_enrichment,
-            artifacts_path=art_path,
-            options=CodeFormulaModelOptions(
-                do_code_enrichment=self.pipeline_options.do_code_enrichment,
-                do_formula_enrichment=self.pipeline_options.do_formula_enrichment,
+        self.enrichment_pipe = [
+            # Code Formula Enrichment Model
+            CodeFormulaModel(
+                enabled=self.pipeline_options.do_code_enrichment
+                or self.pipeline_options.do_formula_enrichment,
+                artifacts_path=self.artifacts_path,
+                options=CodeFormulaModelOptions(
+                    do_code_enrichment=self.pipeline_options.do_code_enrichment,
+                    do_formula_enrichment=self.pipeline_options.do_formula_enrichment,
+                ),
+                accelerator_options=self.pipeline_options.accelerator_options,
             ),
-            accelerator_options=self.pipeline_options.accelerator_options,
-        )
-        if code_formula.enabled:
-            self.enrichment_pipe.append(code_formula)
-
-        picture_classifier = DocumentPictureClassifier(
-            enabled=self.pipeline_options.do_picture_classification,
-            artifacts_path=art_path,
-            options=DocumentPictureClassifierOptions(),
-            accelerator_options=self.pipeline_options.accelerator_options,
-        )
-        if picture_classifier.enabled:
-            self.enrichment_pipe.append(picture_classifier)
-
-        picture_descr = self._make_picture_description_model(art_path)
-        if picture_descr and picture_descr.enabled:
-            self.enrichment_pipe.append(picture_descr)
+            *self.enrichment_pipe,
+        ]
 
         self.keep_backend = any(
             (
@@ -374,19 +361,6 @@ class ThreadedStandardPdfPipeline(BasePipeline):
         )
 
     # ---------------------------------------------------------------- helpers
-    def _resolve_artifacts_path(self) -> Optional[Path]:
-        if self.pipeline_options.artifacts_path:
-            p = Path(self.pipeline_options.artifacts_path).expanduser()
-        elif settings.artifacts_path:
-            p = Path(settings.artifacts_path).expanduser()
-        else:
-            return None
-        if not p.is_dir():
-            raise RuntimeError(
-                f"{p} does not exist or is not a directory containing the required models"
-            )
-        return p
-
     def _make_ocr_model(self, art_path: Optional[Path]) -> Any:
         factory = get_ocr_factory(
             allow_external_plugins=self.pipeline_options.allow_external_plugins
@@ -398,20 +372,6 @@ class ThreadedStandardPdfPipeline(BasePipeline):
             accelerator_options=self.pipeline_options.accelerator_options,
         )
 
-    def _make_picture_description_model(
-        self, art_path: Optional[Path]
-    ) -> Optional[PictureDescriptionBaseModel]:
-        factory = get_picture_description_factory(
-            allow_external_plugins=self.pipeline_options.allow_external_plugins
-        )
-        return factory.create_instance(
-            options=self.pipeline_options.picture_description_options,
-            enabled=self.pipeline_options.do_picture_description,
-            enable_remote_services=self.pipeline_options.enable_remote_services,
-            artifacts_path=art_path,
-            accelerator_options=self.pipeline_options.accelerator_options,
-        )
-
     # ────────────────────────────────────────────────────────────────────────
     # Build - thread pipeline
     # ────────────────────────────────────────────────────────────────────────
@@ -585,6 +545,86 @@ class ThreadedStandardPdfPipeline(BasePipeline):
                 elements=elements, headers=headers, body=body
             )
             conv_res.document = self.reading_order_model(conv_res)
+
+            # Generate page images in the output
+            if self.pipeline_options.generate_page_images:
+                for page in conv_res.pages:
+                    assert page.image is not None
+                    page_no = page.page_no + 1
+                    conv_res.document.pages[page_no].image = ImageRef.from_pil(
+                        page.image, dpi=int(72 * self.pipeline_options.images_scale)
+                    )
+
+            # Generate images of the requested element types
+            with warnings.catch_warnings():  # deprecated generate_table_images
+                warnings.filterwarnings("ignore", category=DeprecationWarning)
+                if (
+                    self.pipeline_options.generate_picture_images
+                    or self.pipeline_options.generate_table_images
+                ):
+                    scale = self.pipeline_options.images_scale
+                    for element, _level in conv_res.document.iterate_items():
+                        if not isinstance(element, DocItem) or len(element.prov) == 0:
+                            continue
+                        if (
+                            isinstance(element, PictureItem)
+                            and self.pipeline_options.generate_picture_images
+                        ) or (
+                            isinstance(element, TableItem)
+                            and self.pipeline_options.generate_table_images
+                        ):
+                            page_ix = element.prov[0].page_no - 1
+                            page = next(
+                                (p for p in conv_res.pages if p.page_no == page_ix),
+                                cast("Page", None),
+                            )
+                            assert page is not None
+                            assert page.size is not None
+                            assert page.image is not None
+
+                            crop_bbox = (
+                                element.prov[0]
+                                .bbox.scaled(scale=scale)
+                                .to_top_left_origin(
+                                    page_height=page.size.height * scale
+                                )
+                            )
+
+                            cropped_im = page.image.crop(crop_bbox.as_tuple())
+                            element.image = ImageRef.from_pil(
+                                cropped_im, dpi=int(72 * scale)
+                            )
+
+            # Aggregate confidence values for document:
+            if len(conv_res.pages) > 0:
+                with warnings.catch_warnings():
+                    warnings.filterwarnings(
+                        "ignore",
+                        category=RuntimeWarning,
+                        message="Mean of empty slice|All-NaN slice encountered",
+                    )
+                    conv_res.confidence.layout_score = float(
+                        np.nanmean(
+                            [c.layout_score for c in conv_res.confidence.pages.values()]
+                        )
+                    )
+                    conv_res.confidence.parse_score = float(
+                        np.nanquantile(
+                            [c.parse_score for c in conv_res.confidence.pages.values()],
+                            q=0.1,  # parse score should relate to worst 10% of pages.
+                        )
+                    )
+                    conv_res.confidence.table_score = float(
+                        np.nanmean(
+                            [c.table_score for c in conv_res.confidence.pages.values()]
+                        )
+                    )
+                    conv_res.confidence.ocr_score = float(
+                        np.nanmean(
+                            [c.ocr_score for c in conv_res.confidence.pages.values()]
+                        )
+                    )
+
         return conv_res
 
     # ---------------------------------------------------------------- misc

docling/pipeline/vlm_pipeline.py

@@ -54,18 +54,6 @@ class VlmPipeline(PaginatedPipeline):
 
         self.pipeline_options: VlmPipelineOptions
 
-        artifacts_path: Optional[Path] = None
-        if pipeline_options.artifacts_path is not None:
-            artifacts_path = Path(pipeline_options.artifacts_path).expanduser()
-        elif settings.artifacts_path is not None:
-            artifacts_path = Path(settings.artifacts_path).expanduser()
-
-        if artifacts_path is not None and not artifacts_path.is_dir():
-            raise RuntimeError(
-                f"The value of {artifacts_path=} is not valid. "
-                "When defined, it must point to a folder containing all models required by the pipeline."
-            )
-
         # force_backend_text = False - use text that is coming from VLM response
         # force_backend_text = True - get text from backend using bounding boxes predicted by SmolDocling doctags
         self.force_backend_text = (
@@ -89,7 +77,7 @@ class VlmPipeline(PaginatedPipeline):
                 self.build_pipe = [
                     HuggingFaceMlxModel(
                         enabled=True,  # must be always enabled for this pipeline to make sense.
-                        artifacts_path=artifacts_path,
+                        artifacts_path=self.artifacts_path,
                         accelerator_options=pipeline_options.accelerator_options,
                         vlm_options=vlm_options,
                     ),
@@ -98,7 +86,7 @@ class VlmPipeline(PaginatedPipeline):
                 self.build_pipe = [
                     HuggingFaceTransformersVlmModel(
                         enabled=True,  # must be always enabled for this pipeline to make sense.
-                        artifacts_path=artifacts_path,
+                        artifacts_path=self.artifacts_path,
                         accelerator_options=pipeline_options.accelerator_options,
                         vlm_options=vlm_options,
                     ),
@@ -109,7 +97,7 @@ class VlmPipeline(PaginatedPipeline):
                 self.build_pipe = [
                     VllmVlmModel(
                         enabled=True,  # must be always enabled for this pipeline to make sense.
-                        artifacts_path=artifacts_path,
+                        artifacts_path=self.artifacts_path,
                         accelerator_options=pipeline_options.accelerator_options,
                         vlm_options=vlm_options,
                     ),

{docling-2.51.0.dist-info → docling-2.52.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling
-Version: 2.51.0
+Version: 2.52.0
 Summary: SDK and CLI for parsing PDF, DOCX, HTML, and more, to a unified document representation for powering downstream workflows such as gen AI applications.
 Author-email: Christoph Auer <cau@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Maxim Lysak <mly@zurich.ibm.com>, Nikos Livathinos <nli@zurich.ibm.com>, Ahmed Nassar <ahn@zurich.ibm.com>, Panos Vagenas <pva@zurich.ibm.com>, Peter Staar <taa@zurich.ibm.com>
 License-Expression: MIT
@@ -26,7 +26,7 @@ Requires-Python: <4.0,>=3.9
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: pydantic<3.0.0,>=2.0.0
-Requires-Dist: docling-core[chunking]<3.0.0,>=2.42.0
+Requires-Dist: docling-core[chunking]<3.0.0,>=2.48.0
 Requires-Dist: docling-parse<5.0.0,>=4.4.0
 Requires-Dist: docling-ibm-models<4,>=3.9.1
 Requires-Dist: filetype<2.0.0,>=1.2.0
@@ -110,16 +110,20 @@ Docling simplifies document processing, parsing diverse formats — including ad
 * 🔍 Extensive OCR support for scanned PDFs and images
 * 👓 Support of several Visual Language Models ([SmolDocling](https://huggingface.co/ds4sd/SmolDocling-256M-preview))
 * 🎙️ Audio support with Automatic Speech Recognition (ASR) models
+* 🔌 Connect to any agent using the [MCP server](https://docling-project.github.io/docling/usage/mcp/)
 * 💻 Simple and convenient CLI
 
 ### What's new
 * 📤 Structured [information extraction][extraction] \[🧪 beta\]
+* 📑 New layout model (**Heron**) by default, for faster PDF parsing
+* 🔌 [MCP server](https://docling-project.github.io/docling/usage/mcp/) for agentic applications
 
 ### Coming soon
 
 * 📝 Metadata extraction, including title, authors, references & language
 * 📝 Chart understanding (Barchart, Piechart, LinePlot, etc)
 * 📝 Complex chemistry understanding (Molecular structures)
+* 📝 Parsing of Web Video Text Tracks (WebVTT) files
 
 ## Installation
 
@@ -145,7 +149,7 @@ result = converter.convert(source)
 print(result.document.export_to_markdown())  # output: "## Docling Technical Report[...]"
 ```
 
-More [advanced usage options](https://docling-project.github.io/docling/usage/) are available in
+More [advanced usage options](https://docling-project.github.io/docling/usage/advanced_options/) are available in
 the docs.
 
 ## CLI

{docling-2.51.0.dist-info → docling-2.52.0.dist-info}/RECORD

@@ -30,7 +30,7 @@ docling/backend/xml/jats_backend.py,sha256=LPj33EFdi2MRCakkLWrRLlUAc-B-949f8zp5g
 docling/backend/xml/uspto_backend.py,sha256=nyAMr5ht7dclxkVDwsKNeiOhLQrUtRLS8JdscB2AVJg,70924
 docling/chunking/__init__.py,sha256=h83TDs0AuOV6oEPLAPrn9dpGKiU-2Vg6IRNo4cv6GDA,346
 docling/cli/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-docling/cli/main.py,sha256=yRgDoc60wm5py2ztq6-q5BPnxfLJZ1EA8Bce-EcpLPs,30952
+docling/cli/main.py,sha256=K4m7dtnLUM2gqU8n_Mntpc_ODrwWtrjBPTUZakQ8erg,32111
 docling/cli/models.py,sha256=5C3CZz3HZXoCrBl92Is62KMCtUqsZK-oygj1hqzJ8vo,6008
 docling/cli/tools.py,sha256=QhtRxQG0TVrfsMqdv5i7J0_qQy1ZZyWYnHPwJl7b5oY,322
 docling/datamodel/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -40,25 +40,25 @@ docling/datamodel/base_models.py,sha256=vOt895z0GsFirHkkI3hM23e9oyUuz9RXfcGFtoIN
 docling/datamodel/document.py,sha256=ElY7G6FYJ6Bayyw433_tbnxyE47fnQRoBG_mygvOBrA,17370
 docling/datamodel/extraction.py,sha256=7dgvtK5SuvgfB8LHAwS1FwrW1kcMQJuJG0ol8uAQgoQ,1323
 docling/datamodel/layout_model_specs.py,sha256=GSkJ-Z_0PVgwWGi7C7TsxbzRjlrWS9ZrHJjHumv-Z5U,2339
-docling/datamodel/pipeline_options.py,sha256=bwBZoQbjk--5vE7Vz7N6KEbew-b93ge0ez1-cDPlUnQ,11019
+docling/datamodel/pipeline_options.py,sha256=N9g-3FA4hFU8A0uGvPmcy1emBBT4JH6u7CUzl3D-Ta0,11049
 docling/datamodel/pipeline_options_asr_model.py,sha256=7X068xl-qpbyPxC7-TwX7Q6tLyZXGT5h1osZ_xLNLM0,1454
 docling/datamodel/pipeline_options_vlm_model.py,sha256=AcqqThSW74hwQ6x7pazzm57LnJiUqB7gQi5wFayGlbk,2628
 docling/datamodel/settings.py,sha256=c0MTw6pO5be_BKxHKYl4SaBJAw_qL-aapxp-g5HHj1A,2084
 docling/datamodel/vlm_model_specs.py,sha256=8D-bF95EoaD-Wd29lVX094HPJT1gYN393aFmzv7RipQ,8713
 docling/models/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling/models/api_vlm_model.py,sha256=-zisU32pgDRbychyG6-neB0qweNbPaYnLXwiGT7SEdI,2859
-docling/models/base_model.py,sha256=tXFM7zJwF6Kn2EhtaB4QmgK4O2ruv1C7SjdBgM5QKak,6225
+docling/models/base_model.py,sha256=beMGyrpl-yYX3YnLzQkLfxMLxwmDWnbcFhkjbUlWJSU,7146
 docling/models/base_ocr_model.py,sha256=kT8TylASOpPlY60rIG6VL6_eLVsfg5KvEVnZHzDWtR0,8193
 docling/models/code_formula_model.py,sha256=XRugm4EwifLRc-TrAk-glKlktJP-nAPneKh2EOovkJU,11308
 docling/models/document_picture_classifier.py,sha256=9JvoWeH5uQBC7levjM8zptk7UT-b8EQnD-2EnxTjTT4,6202
-docling/models/easyocr_model.py,sha256=ECPBd-48cCw5s935NsPJO_C_1QuK_yAUGloMM77WqIM,7387
+docling/models/easyocr_model.py,sha256=sCWIe2xUYU1uK8g4qkPXh0OkUX-rV6BRd4Deb_na9Y4,7899
 docling/models/layout_model.py,sha256=2D7Ey2Mvtzyq9KARIFLaUZKzsR661h7Zggwn0IM9R3c,9154
 docling/models/ocr_mac_model.py,sha256=y-1DSFDbACHpEwNTfQwzN9ab8r5j5rBFNPtQ48BzsrA,5396
 docling/models/page_assemble_model.py,sha256=TvN1naez7dUodLxpUUBzpuMCpqZBTf6YSpewxgjzmrg,6323
 docling/models/page_preprocessing_model.py,sha256=EmusNexws5ZmR93js_saVU0BedqZ_HIHQeY7lcf52tI,5284
 docling/models/picture_description_api_model.py,sha256=o3EkV5aHW_6WzE_fdj_VRnNCrS_btclO_ZCLAUqrfl0,2377
 docling/models/picture_description_base_model.py,sha256=kLthLhdlgwhootQ4_xhhcAk6A-vso5-qcsFJ3TcYfO0,2991
-docling/models/picture_description_vlm_model.py,sha256=5BJvaF3PHuL9lCVYqPv9krh3h_7YwNSdKYw1EVEj13k,4156
+docling/models/picture_description_vlm_model.py,sha256=Uja_BQSk7F-U1J2hm4yeLguirUzKYv1K8zRyw1IYomY,4150
 docling/models/rapid_ocr_model.py,sha256=7yZC7I1qoC9xC8xJIjTk2c8VFm89RfB6Vr7IDOnr5gs,7102
 docling/models/readingorder_model.py,sha256=bZoXHaSwUsa8niSmJrbCuy784ixCeBXT-RQBUfgHJ4A,14925
 docling/models/table_structure_model.py,sha256=7vO8LisdoqCTsY8X8lsk9d-oD2hVjUtdaWlkMTQxEg0,12518
@@ -73,19 +73,19 @@ docling/models/plugins/defaults.py,sha256=OAHWW2tCcUXSyDMFxV_lXVRjSBJ1n6z-Eb3R8c
 docling/models/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling/models/utils/hf_model_download.py,sha256=scBEfsM4yl7xPzqe7UtPvDh9RfQZQnuOhqQKilYBHls,984
 docling/models/vlm_models_inline/__init__.py,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
-docling/models/vlm_models_inline/hf_transformers_model.py,sha256=G0RpKwdzm5NiqIBHG5nWLwBsrDfDebzErzRkyXppZPw,12134
+docling/models/vlm_models_inline/hf_transformers_model.py,sha256=wcBsUGHuPNoZn9pfQh4cyyZi1bFGCoF30T_MFVGDyyM,12128
 docling/models/vlm_models_inline/mlx_model.py,sha256=VP05v97mqzmaG4o9bOpJcxIlEqvNzAapJ15Zz3E3ACI,10169
-docling/models/vlm_models_inline/nuextract_transformers_model.py,sha256=iWoGF8TgQfOOMqS__tSODcUuDnKTPaK7gIRFum5bPzc,10512
+docling/models/vlm_models_inline/nuextract_transformers_model.py,sha256=jLNtlkMDheUyWot7Oqq-GHQIYzJ0fZrbReq5xCnYb9E,10506
 docling/models/vlm_models_inline/vllm_model.py,sha256=_EnK1nfpAPJky7aRlyp8SUIghiZOQO8AkDN_hHqXLZg,8615
 docling/pipeline/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-docling/pipeline/asr_pipeline.py,sha256=tQkhu9fXdkSuYIL22xzV2YRUlQh-9qktHBbs2qeXhJI,9070
-docling/pipeline/base_extraction_pipeline.py,sha256=aJj7qbppgAelwoaVKB1W-s7kFg_OcXRE64NpIIOxZGE,1905
-docling/pipeline/base_pipeline.py,sha256=Tl_C3adFABNxtE7hX83VSdx-j7D8GRvoFcno5A3Z-YQ,10062
-docling/pipeline/extraction_vlm_pipeline.py,sha256=WIRZygpBJmKjszRsFqW4qfPUZ5Frd_Hqoiysp2dGx8Y,8723
-docling/pipeline/simple_pipeline.py,sha256=TXZOwR7hZRji462ZTIpte0VJjzbxvNVE8dbLFANDhSU,2253
-docling/pipeline/standard_pdf_pipeline.py,sha256=yFishq4Cu01BiBGHk3Irr7ogcTQKeSC0QZImQVAhIaY,12740
-docling/pipeline/threaded_standard_pdf_pipeline.py,sha256=NgdZxpfpElnvCgGlrQ8kSvq44LNzJcc6wOqD-AMrKZ0,26132
-docling/pipeline/vlm_pipeline.py,sha256=PlllB2ZbhuUQilJ5_W8RhuBUbLtXdSjnrYnTVzHF0Vc,15955
+docling/pipeline/asr_pipeline.py,sha256=S55VHLoX3Mgauen1YP-PSUlI0LA1bgTgTkU-eC4U-dg,8481
+docling/pipeline/base_extraction_pipeline.py,sha256=GYrEz83IXv-tdIHjtNWxMBNczFwL8SZyf9vnPJ3STaI,2627
+docling/pipeline/base_pipeline.py,sha256=NPMQDTyis-LgQ4SybY2f5AESZl5PxogF-FRQuCDckXg,12748
+docling/pipeline/extraction_vlm_pipeline.py,sha256=veUOTe8nGdnduZKaGn1RRb-NfU1H6t_EN4QAsb022Zg,8260
+docling/pipeline/simple_pipeline.py,sha256=FSL_ucDd9k0D9DjNKMUkyCULIU8a057dvWfLEPmAc2A,2287
+docling/pipeline/standard_pdf_pipeline.py,sha256=xOge0zP5wli51n_6QLrFHQlwwvsivI7OMt00tht3my4,10479
+docling/pipeline/threaded_standard_pdf_pipeline.py,sha256=i67G5AOW7PIFCe5JS2sdBmPAKvAH6ScxIBhjwOGZcrI,28183
+docling/pipeline/vlm_pipeline.py,sha256=oMcdgzym_UQbVN3bajux_hENY40XGOnb6NU6Kwje2Os,15376
 docling/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling/utils/accelerator_utils.py,sha256=DSajLxVx1JEVT0zt5de26llciLNlVfIDfSa2zYCFJzQ,2909
 docling/utils/api_image_request.py,sha256=_CgdzmPqdsyXmyYUFGLZcXcoH586qC6A1p5vsNbj1Q0,1416
@@ -99,9 +99,9 @@ docling/utils/orientation.py,sha256=jTyLxyT31FlOodZoBMlADHNQK2lAWKYVs5z7pXd_6Cg,
 docling/utils/profiling.py,sha256=YaMGoB9MMZpagF9mb5ndoHj8Lpb9aIdb7El-Pl7IcFs,1753
 docling/utils/utils.py,sha256=kJtIYuzXeOyJHYlxmLAo7dGM5rEsDa1i84qEsUj1nio,1908
 docling/utils/visualization.py,sha256=tY2ylE2aiQKkmzlSLnFW-HTfFyqUUMguW18ldd1PLfo,2868
-docling-2.51.0.dist-info/licenses/LICENSE,sha256=mBb7ErEcM8VS9OhiGHnQ2kk75HwPhr54W1Oiz3965MY,1088
-docling-2.51.0.dist-info/METADATA,sha256=oVomcZQUJIX-QnMmIFggzyjuhea7jbgVXus9bZZkYDU,10886
-docling-2.51.0.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-docling-2.51.0.dist-info/entry_points.txt,sha256=hzVlbeE0aMSTQ9S0-NTYN0Hmgsn6qL_EA2qX4UbkAuY,149
-docling-2.51.0.dist-info/top_level.txt,sha256=vkIywP-USjFyYo1AIRQbWQQaL3xB5jf8vkCYdTIfNic,8
-docling-2.51.0.dist-info/RECORD,,
+docling-2.52.0.dist-info/licenses/LICENSE,sha256=mBb7ErEcM8VS9OhiGHnQ2kk75HwPhr54W1Oiz3965MY,1088
+docling-2.52.0.dist-info/METADATA,sha256=EhUePtqwKQJTgkU9pCtvpWT7wtU-84KXkc48XExkRSQ,11233
+docling-2.52.0.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+docling-2.52.0.dist-info/entry_points.txt,sha256=hzVlbeE0aMSTQ9S0-NTYN0Hmgsn6qL_EA2qX4UbkAuY,149
+docling-2.52.0.dist-info/top_level.txt,sha256=vkIywP-USjFyYo1AIRQbWQQaL3xB5jf8vkCYdTIfNic,8
+docling-2.52.0.dist-info/RECORD,,