docling 2.36.1__py3-none-any.whl → 2.38.0__py3-none-any.whl

docling/models/page_preprocessing_model.py

@@ -2,7 +2,7 @@ import re
 import warnings
 from collections.abc import Iterable
 from pathlib import Path
-from typing import Optional
+from typing import Literal, Optional
 
 import numpy as np
 from PIL import ImageDraw
@@ -17,7 +17,6 @@ from docling.utils.profiling import TimeRecorder
 
 class PagePreprocessingOptions(BaseModel):
     images_scale: Optional[float]
-    create_parsed_page: bool
 
 
 class PagePreprocessingModel(BasePageModel):
@@ -66,10 +65,8 @@ class PagePreprocessingModel(BasePageModel):
     def _parse_page_cells(self, conv_res: ConversionResult, page: Page) -> Page:
         assert page._backend is not None
 
-        page.cells = list(page._backend.get_text_cells())
-
-        if self.options.create_parsed_page:
-            page.parsed_page = page._backend.get_segmented_page()
+        page.parsed_page = page._backend.get_segmented_page()
+        assert page.parsed_page is not None
 
         # Rate the text quality from the PDF parser, and aggregate on page
         text_scores = []

docling/models/rapid_ocr_model.py

@@ -134,7 +134,7 @@ class RapidOcrModel(BaseOcrModel):
                             all_ocr_cells.extend(cells)
 
                     # Post-process the cells
-                    page.cells = self.post_process_cells(all_ocr_cells, page.cells)
+                    self.post_process_cells(all_ocr_cells, page)
 
                 # DEBUG code:
                 if settings.debug.visualize_ocr:

docling/models/readingorder_model.py

@@ -124,7 +124,7 @@ class ReadingOrderModel:
             page_no = page.page_no + 1
             size = page.size
 
-            assert size is not None
+            assert size is not None, "Page size is not initialized."
 
             out_doc.add_page(page_no=page_no, size=size)
 
@@ -334,12 +334,12 @@ class ReadingOrderModel:
             "Labels of merged elements must match."
         )
         prov = ProvenanceItem(
-            page_no=element.page_no + 1,
+            page_no=merged_elem.page_no + 1,
             charspan=(
                 len(new_item.text) + 1,
                 len(new_item.text) + 1 + len(merged_elem.text),
             ),
-            bbox=element.cluster.bbox.to_bottom_left_origin(page_height),
+            bbox=merged_elem.cluster.bbox.to_bottom_left_origin(page_height),
         )
         new_item.text += f" {merged_elem.text}"
         new_item.orig += f" {merged_elem.text}"  # TODO: This is incomplete, we don't have the `orig` field of the merged element.

docling/models/tesseract_ocr_cli_model.py

@@ -99,12 +99,12 @@ class TesseractOcrCliModel(BaseOcrModel):
 
         return name, version
 
-    def _run_tesseract(self, ifilename: str, osd: pd.DataFrame):
+    def _run_tesseract(self, ifilename: str, osd: Optional[pd.DataFrame]):
         r"""
         Run tesseract CLI
         """
         cmd = [self.options.tesseract_cmd]
-        if self._is_auto:
+        if self._is_auto and osd is not None:
             lang = self._parse_language(osd)
             if lang is not None:
                 cmd.append("-l")
@@ -231,6 +231,7 @@ class TesseractOcrCliModel(BaseOcrModel):
                                 fname = image_file.name
                                 high_res_image.save(image_file)
                             doc_orientation = 0
+                            df_osd: Optional[pd.DataFrame] = None
                             try:
                                 df_osd = self._perform_osd(fname)
                                 doc_orientation = _parse_orientation(df_osd)
@@ -305,7 +306,7 @@ class TesseractOcrCliModel(BaseOcrModel):
                             all_ocr_cells.append(cell)
 
                     # Post-process the cells
-                    page.cells = self.post_process_cells(all_ocr_cells, page.cells)
+                    self.post_process_cells(all_ocr_cells, page)
 
                 # DEBUG code:
                 if settings.debug.visualize_ocr:

docling/models/tesseract_ocr_model.py

@@ -235,7 +235,7 @@ class TesseractOcrModel(BaseOcrModel):
                         all_ocr_cells.extend(cells)
 
                     # Post-process the cells
-                    page.cells = self.post_process_cells(all_ocr_cells, page.cells)
+                    self.post_process_cells(all_ocr_cells, page)
 
                 # DEBUG code:
                 if settings.debug.visualize_ocr:

docling/models/vlm_models_inline/hf_transformers_model.py

@@ -99,6 +99,7 @@ class HuggingFaceTransformersVlmModel(BasePageModel, HuggingFaceModelDownloadMix
             self.vlm_model = model_cls.from_pretrained(
                 artifacts_path,
                 device_map=self.device,
+                torch_dtype=self.vlm_options.torch_dtype,
                 _attn_implementation=(
                     "flash_attention_2"
                     if self.device.startswith("cuda")
@@ -122,7 +123,9 @@ class HuggingFaceTransformersVlmModel(BasePageModel, HuggingFaceModelDownloadMix
                 with TimeRecorder(conv_res, "vlm"):
                     assert page.size is not None
 
-                    hi_res_image = page.get_image(scale=self.vlm_options.scale)
+                    hi_res_image = page.get_image(
+                        scale=self.vlm_options.scale, max_size=self.vlm_options.max_size
+                    )
 
                     # Define prompt structure
                     prompt = self.formulate_prompt()

docling/models/vlm_models_inline/mlx_model.py

@@ -73,7 +73,9 @@ class HuggingFaceMlxModel(BasePageModel, HuggingFaceModelDownloadMixin):
                 with TimeRecorder(conv_res, f"vlm-mlx-{self.vlm_options.repo_id}"):
                     assert page.size is not None
 
-                    hi_res_image = page.get_image(scale=self.vlm_options.scale)
+                    hi_res_image = page.get_image(
+                        scale=self.vlm_options.scale, max_size=self.vlm_options.max_size
+                    )
                     if hi_res_image is not None:
                         im_width, im_height = hi_res_image.size
 

docling/pipeline/asr_pipeline.py

@@ -0,0 +1,253 @@
+import logging
+import os
+import re
+from io import BytesIO
+from pathlib import Path
+from typing import List, Optional, Union, cast
+
+from docling_core.types.doc import DoclingDocument, DocumentOrigin
+
+# import whisper  # type: ignore
+# import librosa
+# import numpy as np
+# import soundfile as sf  # type: ignore
+from docling_core.types.doc.labels import DocItemLabel
+from pydantic import BaseModel, Field, validator
+
+from docling.backend.abstract_backend import AbstractDocumentBackend
+from docling.backend.noop_backend import NoOpBackend
+
+# from pydub import AudioSegment  # type: ignore
+# from transformers import WhisperForConditionalGeneration, WhisperProcessor, pipeline
+from docling.datamodel.accelerator_options import (
+    AcceleratorOptions,
+)
+from docling.datamodel.base_models import (
+    ConversionStatus,
+    FormatToMimeType,
+)
+from docling.datamodel.document import ConversionResult, InputDocument
+from docling.datamodel.pipeline_options import (
+    AsrPipelineOptions,
+)
+from docling.datamodel.pipeline_options_asr_model import (
+    InlineAsrNativeWhisperOptions,
+    # AsrResponseFormat,
+    InlineAsrOptions,
+)
+from docling.datamodel.pipeline_options_vlm_model import (
+    InferenceFramework,
+)
+from docling.datamodel.settings import settings
+from docling.pipeline.base_pipeline import BasePipeline
+from docling.utils.accelerator_utils import decide_device
+from docling.utils.profiling import ProfilingScope, TimeRecorder
+
+_log = logging.getLogger(__name__)
+
+
+class _ConversationWord(BaseModel):
+    text: str
+    start_time: Optional[float] = Field(
+        None, description="Start time in seconds from video start"
+    )
+    end_time: Optional[float] = Field(
+        None, ge=0, description="End time in seconds from video start"
+    )
+
+
+class _ConversationItem(BaseModel):
+    text: str
+    start_time: Optional[float] = Field(
+        None, description="Start time in seconds from video start"
+    )
+    end_time: Optional[float] = Field(
+        None, ge=0, description="End time in seconds from video start"
+    )
+    speaker_id: Optional[int] = Field(None, description="Numeric speaker identifier")
+    speaker: Optional[str] = Field(
+        None, description="Speaker name, defaults to speaker-{speaker_id}"
+    )
+    words: Optional[list[_ConversationWord]] = Field(
+        None, description="Individual words with time-stamps"
+    )
+
+    def __lt__(self, other):
+        if not isinstance(other, _ConversationItem):
+            return NotImplemented
+        return self.start_time < other.start_time
+
+    def __eq__(self, other):
+        if not isinstance(other, _ConversationItem):
+            return NotImplemented
+        return self.start_time == other.start_time
+
+    def to_string(self) -> str:
+        """Format the conversation entry as a string"""
+        result = ""
+        if (self.start_time is not None) and (self.end_time is not None):
+            result += f"[time: {self.start_time}-{self.end_time}] "
+
+        if self.speaker is not None:
+            result += f"[speaker:{self.speaker}] "
+
+        result += self.text
+        return result
+
+
+class _NativeWhisperModel:
+    def __init__(
+        self,
+        enabled: bool,
+        artifacts_path: Optional[Path],
+        accelerator_options: AcceleratorOptions,
+        asr_options: InlineAsrNativeWhisperOptions,
+    ):
+        """
+        Transcriber using native Whisper.
+        """
+        self.enabled = enabled
+
+        _log.info(f"artifacts-path: {artifacts_path}")
+        _log.info(f"accelerator_options: {accelerator_options}")
+
+        if self.enabled:
+            try:
+                import whisper  # type: ignore
+            except ImportError:
+                raise ImportError(
+                    "whisper is not installed. Please install it via `pip install openai-whisper` or do `uv sync --extra asr`."
+                )
+            self.asr_options = asr_options
+            self.max_tokens = asr_options.max_new_tokens
+            self.temperature = asr_options.temperature
+
+            self.device = decide_device(
+                accelerator_options.device,
+                supported_devices=asr_options.supported_devices,
+            )
+            _log.info(f"Available device for Whisper: {self.device}")
+
+            self.model_name = asr_options.repo_id
+            _log.info(f"loading _NativeWhisperModel({self.model_name})")
+            if artifacts_path is not None:
+                _log.info(f"loading {self.model_name} from {artifacts_path}")
+                self.model = whisper.load_model(
+                    name=self.model_name,
+                    device=self.device,
+                    download_root=str(artifacts_path),
+                )
+            else:
+                self.model = whisper.load_model(
+                    name=self.model_name, device=self.device
+                )
+
+            self.verbose = asr_options.verbose
+            self.timestamps = asr_options.timestamps
+            self.word_timestamps = asr_options.word_timestamps
+
+    def run(self, conv_res: ConversionResult) -> ConversionResult:
+        audio_path: Path = Path(conv_res.input.file).resolve()
+
+        try:
+            conversation = self.transcribe(audio_path)
+
+            # Ensure we have a proper DoclingDocument
+            origin = DocumentOrigin(
+                filename=conv_res.input.file.name or "audio.wav",
+                mimetype="audio/x-wav",
+                binary_hash=conv_res.input.document_hash,
+            )
+            conv_res.document = DoclingDocument(
+                name=conv_res.input.file.stem or "audio.wav", origin=origin
+            )
+
+            for citem in conversation:
+                conv_res.document.add_text(
+                    label=DocItemLabel.TEXT, text=citem.to_string()
+                )
+
+            conv_res.status = ConversionStatus.SUCCESS
+            return conv_res
+
+        except Exception as exc:
+            _log.error(f"Audio tranciption has an error: {exc}")
+
+        conv_res.status = ConversionStatus.FAILURE
+        return conv_res
+
+    def transcribe(self, fpath: Path) -> list[_ConversationItem]:
+        result = self.model.transcribe(
+            str(fpath), verbose=self.verbose, word_timestamps=self.word_timestamps
+        )
+
+        convo: list[_ConversationItem] = []
+        for _ in result["segments"]:
+            item = _ConversationItem(
+                start_time=_["start"], end_time=_["end"], text=_["text"], words=[]
+            )
+            if "words" in _ and self.word_timestamps:
+                item.words = []
+                for __ in _["words"]:
+                    item.words.append(
+                        _ConversationWord(
+                            start_time=__["start"],
+                            end_time=__["end"],
+                            text=__["word"],
+                        )
+                    )
+            convo.append(item)
+
+        return convo
+
+
+class AsrPipeline(BasePipeline):
+    def __init__(self, pipeline_options: AsrPipelineOptions):
+        super().__init__(pipeline_options)
+        self.keep_backend = True
+
+        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,
+                accelerator_options=pipeline_options.accelerator_options,
+                asr_options=asr_options,
+            )
+        else:
+            _log.error(f"No model support for {self.pipeline_options.asr_options}")
+
+    def _determine_status(self, conv_res: ConversionResult) -> ConversionStatus:
+        status = ConversionStatus.SUCCESS
+        return status
+
+    @classmethod
+    def get_default_options(cls) -> AsrPipelineOptions:
+        return AsrPipelineOptions()
+
+    def _build_document(self, conv_res: ConversionResult) -> ConversionResult:
+        _log.info(f"start _build_document in AsrPipeline: {conv_res.input.file}")
+        with TimeRecorder(conv_res, "doc_build", scope=ProfilingScope.DOCUMENT):
+            self._model.run(conv_res=conv_res)
+
+        return conv_res
+
+    @classmethod
+    def is_backend_supported(cls, backend: AbstractDocumentBackend):
+        return isinstance(backend, NoOpBackend)

docling/pipeline/base_pipeline.py

@@ -193,6 +193,17 @@ class PaginatedPipeline(BasePipeline):  # TODO this is a bad name.
                 )
                 raise e
 
+            # Filter out uninitialized pages (those with size=None) that may remain
+            # after timeout or processing failures to prevent assertion errors downstream
+            initial_page_count = len(conv_res.pages)
+            conv_res.pages = [page for page in conv_res.pages if page.size is not None]
+
+            if len(conv_res.pages) < initial_page_count:
+                _log.info(
+                    f"Filtered out {initial_page_count - len(conv_res.pages)} uninitialized pages "
+                    f"due to timeout or processing failures"
+                )
+
         return conv_res
 
     def _unload(self, conv_res: ConversionResult) -> ConversionResult:

docling/pipeline/standard_pdf_pipeline.py

@@ -72,7 +72,6 @@ class StandardPdfPipeline(PaginatedPipeline):
             PagePreprocessingModel(
                 options=PagePreprocessingOptions(
                     images_scale=pipeline_options.images_scale,
-                    create_parsed_page=pipeline_options.generate_parsed_pages,
                 )
             ),
             # OCR

docling/utils/layout_postprocessor.py

@@ -8,7 +8,7 @@ from docling_core.types.doc import DocItemLabel, Size
 from docling_core.types.doc.page import TextCell
 from rtree import index
 
-from docling.datamodel.base_models import BoundingBox, Cluster
+from docling.datamodel.base_models import BoundingBox, Cluster, Page
 
 _log = logging.getLogger(__name__)
 
@@ -194,11 +194,11 @@ class LayoutPostprocessor:
         DocItemLabel.TITLE: DocItemLabel.SECTION_HEADER,
     }
 
-    def __init__(self, cells: List[TextCell], clusters: List[Cluster], page_size: Size):
-        """Initialize processor with cells and clusters."""
-        """Initialize processor with cells and spatial indices."""
-        self.cells = cells
-        self.page_size = page_size
+    def __init__(self, page: Page, clusters: List[Cluster]) -> None:
+        """Initialize processor with page and clusters."""
+        self.cells = page.cells
+        self.page = page
+        self.page_size = page.size
         self.all_clusters = clusters
         self.regular_clusters = [
             c for c in clusters if c.label not in self.SPECIAL_TYPES
@@ -240,6 +240,10 @@ class LayoutPostprocessor:
             for child in cluster.children:
                 child.cells = self._sort_cells(child.cells)
 
+        assert self.page.parsed_page is not None
+        self.page.parsed_page.textline_cells = self.cells
+        self.page.parsed_page.has_lines = len(self.cells) > 0
+
         return final_clusters, self.cells
 
     def _process_regular_clusters(self) -> List[Cluster]:
@@ -301,6 +305,7 @@ class LayoutPostprocessor:
         special_clusters = self._handle_cross_type_overlaps(special_clusters)
 
         # Calculate page area from known page size
+        assert self.page_size is not None
         page_area = self.page_size.width * self.page_size.height
         if page_area > 0:
             # Filter out full-page pictures

{docling-2.36.1.dist-info → docling-2.38.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling
-Version: 2.36.1
+Version: 2.38.0
 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
@@ -61,6 +61,8 @@ Requires-Dist: mlx-vlm>=0.1.22; (python_version >= "3.10" and sys_platform == "d
 Provides-Extra: rapidocr
 Requires-Dist: rapidocr-onnxruntime<2.0.0,>=1.4.0; python_version < "3.13" and extra == "rapidocr"
 Requires-Dist: onnxruntime<2.0.0,>=1.7.0; extra == "rapidocr"
+Provides-Extra: asr
+Requires-Dist: openai-whisper>=20240930; extra == "asr"
 Dynamic: license-file
 
 <p align="center">
@@ -93,14 +95,15 @@ Docling simplifies document processing, parsing diverse formats — including ad
 
 ## Features
 
-* 🗂️ Parsing of [multiple document formats][supported_formats] incl. PDF, DOCX, XLSX, HTML, images, and more
+* 🗂️  Parsing of [multiple document formats][supported_formats] incl. PDF, DOCX, PPTX, XLSX, HTML, WAV, MP3, images (PNG, TIFF, JPEG, ...), and more
 * 📑 Advanced PDF understanding incl. page layout, reading order, table structure, code, formulas, image classification, and more
 * 🧬 Unified, expressive [DoclingDocument][docling_document] representation format
-* ↪️ Various [export formats][supported_formats] and options, including Markdown, HTML, and lossless JSON
+* ↪️  Various [export formats][supported_formats] and options, including Markdown, HTML, [DocTags](https://arxiv.org/abs/2503.11576) and lossless JSON
 * 🔒 Local execution capabilities for sensitive data and air-gapped environments
 * 🤖 Plug-and-play [integrations][integrations] incl. LangChain, LlamaIndex, Crew AI & Haystack for agentic AI
 * 🔍 Extensive OCR support for scanned PDFs and images
-* 🥚 Support of several Visual Language Models ([SmolDocling](https://huggingface.co/ds4sd/SmolDocling-256M-preview))
+* 👓 Support of several Visual Language Models ([SmolDocling](https://huggingface.co/ds4sd/SmolDocling-256M-preview))
+* 🎙️  Support for Audio with Automatic Speech Recognition (ASR) models
 * 💻 Simple and convenient CLI
 
 ### Coming soon

{docling-2.36.1.dist-info → docling-2.38.0.dist-info}/RECORD

@@ -1,21 +1,22 @@
 docling/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-docling/document_converter.py,sha256=bnUA9k1LCuCfNwCsneGQiGCvFdnX8W-vbpnu6U_fuuI,14003
+docling/document_converter.py,sha256=3jWywP_TLy-1PMvjJBUlnTM9FNzpBLRCHYA1RKFvGR4,14333
 docling/exceptions.py,sha256=K1WnCS1leK2JtMB5ewZWKkb0EaijFgl-tRzrO9ntgPM,134
 docling/py.typed,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
 docling/backend/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling/backend/abstract_backend.py,sha256=1lNxzwDTn303aXduPDVmTyXn-5ZIoWMLYqNxANGWmQQ,1658
-docling/backend/asciidoc_backend.py,sha256=W-4MRcID6AU9Ax23q8FwDwGG-OOCrBoqcNf2Ch_WPUc,14041
+docling/backend/asciidoc_backend.py,sha256=RDNLrPJHxROiM7-NQdZn3DdvAyiPAndbSWcZo9PbCKU,14417
 docling/backend/csv_backend.py,sha256=2g9famYG2W-ID9jEdZPxc6O8QGv1vWQfjN8pL-QMBE0,4536
-docling/backend/docling_parse_backend.py,sha256=bVSPmmiVXdCVfe-eLtDhbPQKBjkFR8rZJoRxdWIMdYU,7998
-docling/backend/docling_parse_v2_backend.py,sha256=R4YPCEs72GYg-Xc9VfizPv8QjtGmKOsQzVPNAU2RIK0,9376
-docling/backend/docling_parse_v4_backend.py,sha256=aWh-fd-lnuRGVGC_DG17QUptIsArv5V1gJo8QFbB5Ys,6263
+docling/backend/docling_parse_backend.py,sha256=9rUo1vPxX6QLzGqF-2B2iEYglZg6YQ3Uea00XrLluTg,7918
+docling/backend/docling_parse_v2_backend.py,sha256=3ckTfke8IICjaImlIzc3TRhG7KDuxDDba0AuCEcjA-M,9500
+docling/backend/docling_parse_v4_backend.py,sha256=7tQvpCwpYoq98PNszDkrXaFhy5eWmQqMP4RjWWPLPgw,6197
 docling/backend/html_backend.py,sha256=3K-l5SUAAyqISNEb7nPst_I51xzYOVOkgmwXh3lv9sw,21063
-docling/backend/md_backend.py,sha256=JkY1qTvQFXjKSZGfD-83d-fZelorUG_l6mpJdYGqvX8,17210
+docling/backend/md_backend.py,sha256=ghIU_NSaENKrRu49Dn5GvjYtcAgEU7ZHbf-TeYg49nY,17673
 docling/backend/msexcel_backend.py,sha256=3j0WQfqDpgPXdPMCguefdv7arcNVDedPD6gl54cmLn8,18110
-docling/backend/mspowerpoint_backend.py,sha256=RwqfvvzrtM56L9uf7PR9lvlHJ-LyYGpkS1iVxkTl72Q,17203
-docling/backend/msword_backend.py,sha256=iB2yRg8hXtET2-Wjkv5pq0p9Y1SGQYIVCcWtOtXUILU,44621
+docling/backend/mspowerpoint_backend.py,sha256=0lsb8ZeQFxbDt7jZpSQyk5wYHYa3SP2T2y2dMI-o30o,15216
+docling/backend/msword_backend.py,sha256=C4qs4mQEt1JzonCg5v6_yUxdngzcTzSO9k1ik8_DW5Q,44855
+docling/backend/noop_backend.py,sha256=EOPbD86FzZPX-K_DpNrJh0_lC0bZz--4DpG-OagDNGY,1688
 docling/backend/pdf_backend.py,sha256=KE9TMuFO5WX-o5A_DAd4tEaLi4HMZ4XjKdpllItVkWM,2238
-docling/backend/pypdfium2_backend.py,sha256=fUGRBupwTYftEgdIDWKphA2zdfb-SrUoUGENK6j-q-0,11002
+docling/backend/pypdfium2_backend.py,sha256=8dVniLHgiTdJuDbYr66kPp6Ccv5ZDlqDMEbA2xIfS7U,13370
 docling/backend/docx/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling/backend/docx/latex/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling/backend/docx/latex/latex_dict.py,sha256=tFJp4ScT_AkY2ON7nLEa560p601Jq2glcZvMKxxjn7w,6593
@@ -27,36 +28,38 @@ docling/backend/xml/jats_backend.py,sha256=ghGi9bHjx3BvaOtmzLw86-wZy4UxpQPOPQL4e
 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=fDGjepShl6KO_BdA6qUNyNBoCjqZUKRnmmkzesGtvVU,27202
+docling/cli/main.py,sha256=D2gEoArnQ2yQ9BesH9CkxZbYQyhZRGgjjNWYqmRRUtU,29617
 docling/cli/models.py,sha256=9yLGp6QRJGpR86U3SjmWAXDt3MvBaJLLY4xDVdsu3O8,4160
 docling/cli/tools.py,sha256=QhtRxQG0TVrfsMqdv5i7J0_qQy1ZZyWYnHPwJl7b5oY,322
 docling/datamodel/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling/datamodel/accelerator_options.py,sha256=wv6dOFTVAwr9onkE-0pfUqX_fDb6gX53iPPE6o8nKjI,2511
-docling/datamodel/base_models.py,sha256=bkooSG4brZy2jt2dndkin3DHvfZ5HFp0C94yBGmCWeI,10568
-docling/datamodel/document.py,sha256=vPwiVU5zWCKbVYMq-TSmb7LTjijrqJq0FyAgDBa0XGA,16154
-docling/datamodel/pipeline_options.py,sha256=iMuwsa77hkAgjJWXBRAFEQGw9tGNMDQrPnSvE5mirNs,9081
-docling/datamodel/pipeline_options_vlm_model.py,sha256=-ZPAp2uSKMatDbjZPv9chT587B1aftfDVmi_FDb2aw8,1997
+docling/datamodel/asr_model_specs.py,sha256=L7ETXsUKVbPsVcPLhEIMxQjd4UzMGZBVsy74CLsZBkU,2181
+docling/datamodel/base_models.py,sha256=L35qXLmADZQNEzBC0M6K2xrfLyqrTqDlbPD6E6DkWMc,11146
+docling/datamodel/document.py,sha256=CA_dgt4V_phze5HXpfgfKNBKd1cPC1o3WE_IENX63EM,16252
+docling/datamodel/pipeline_options.py,sha256=N7my7hmvuX6EzlujHeF6RObPSrG_HjN_nfPzILTqP-E,9479
+docling/datamodel/pipeline_options_asr_model.py,sha256=7X068xl-qpbyPxC7-TwX7Q6tLyZXGT5h1osZ_xLNLM0,1454
+docling/datamodel/pipeline_options_vlm_model.py,sha256=rtDMVtKFZbgQD269w8FvHMXEhdRBrsA4rVYk6A-M-b4,2063
 docling/datamodel/settings.py,sha256=ajMz7Ao2m0ZGYkfArqTDDbiF89O408mtgeh06PUi0MA,1900
 docling/datamodel/vlm_model_specs.py,sha256=--jZexGeu-s_lWp7y_WwWEf6CD1J4XqADrS1-OY_pWM,4737
 docling/models/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-docling/models/api_vlm_model.py,sha256=w3P1wOsr3JvZsawbK1Z4uwnD5ehUMbcKGkyhcX83Okc,2738
-docling/models/base_model.py,sha256=Zx_nByGYkubTvvYiQxwiB6P8lc7wOD4ZTC2QIw6vCEg,2950
-docling/models/base_ocr_model.py,sha256=c6a2QzZnAMfQECQDz1JASecl_Z2F3i6P3ax6kHWcz6o,7221
+docling/models/api_vlm_model.py,sha256=GDDJGAia4SJjK7JFxsZy5oEU-D8yQo8Kb3NvvPbTvT0,2820
+docling/models/base_model.py,sha256=NNjIapqCruAEAWR-CCdsNgXc2QkwiPYAcaQ_ZYe1W28,2978
+docling/models/base_ocr_model.py,sha256=HtrefTq9Zy4UnUInMchPv0tbobiA7CQU5VUauKJD7IU,8006
 docling/models/code_formula_model.py,sha256=5uWh-eI-Ejmv3DujKJoKKgJBuvPLokt7AJ_ybt8VHEw,11373
 docling/models/document_picture_classifier.py,sha256=fkJLV7pMy3v6iNwOzVb6zdBU1dGtBM1ARHLIRPfoAG4,6124
-docling/models/easyocr_model.py,sha256=bTK-AQYc-WTzX8SRoMRwVjqlMigaJKGloaLUcH6RCKU,7406
-docling/models/layout_model.py,sha256=KdGhS4EMWKP6BwlhUJ0mdbhk2Fc78qwzqEZbTxyrbFM,8508
-docling/models/ocr_mac_model.py,sha256=CJOwz9h84crvZd3kQMLxYntpXz-1w2eLDjhGUnGIwMQ,5415
+docling/models/easyocr_model.py,sha256=ECPBd-48cCw5s935NsPJO_C_1QuK_yAUGloMM77WqIM,7387
+docling/models/layout_model.py,sha256=EJuRXW0rFdnNPS5AifdEsr812EATUqAioeMCVjw8PL0,8460
+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=8cdhR9n3zcC8JxDen8WdPBx_GNk_5VICeHJo1-kP518,5186
+docling/models/page_preprocessing_model.py,sha256=x8MI4mvjizqEqAb5511dtrNRCJSb-lSmwHw0tmHPFiI,5103
 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=7LeCx9ZdPxsmWJ468OtxCdAkH48A1HD0iwH9cs_7-1Q,3800
-docling/models/rapid_ocr_model.py,sha256=miTPn1YTWKtXUuddiVv0SjgkuNWHXCW3CZ6epDUmKjI,5935
-docling/models/readingorder_model.py,sha256=S9ru2ApY9sE-Uue3hptWHmbmElwo36bUbAikxCFpHYs,14574
+docling/models/rapid_ocr_model.py,sha256=AMdc66s_iWO4p6nQ0LNjQMUYVxrDSxMyLNPpjPYt6N8,5916
+docling/models/readingorder_model.py,sha256=QHb5fyiqmxU8lg4W5IzdukqHPh6V7rNw_57O4-z-Az4,14615
 docling/models/table_structure_model.py,sha256=dQf6u_zn5fHCkHzmTwYfCbRtZCBddsyAM0WNVBUUQzk,12473
-docling/models/tesseract_ocr_cli_model.py,sha256=oQZVWXQ6wRrFonRFwbWeW9nJ9FLQZdzSWErOp0mEff0,12698
-docling/models/tesseract_ocr_model.py,sha256=AjrZNwgVbV0IbzBJwI35YP0KxvqWJWJE0v_lgHJiQrk,10606
+docling/models/tesseract_ocr_cli_model.py,sha256=qcM3-n7Z_dm1CGBhVUcNr2XT41iXnU32zk4RqKHBl9I,12775
+docling/models/tesseract_ocr_model.py,sha256=9DPAE7XP7smej7HYhr7mdwpuxSjAcv_GPrYZG3bb1RA,10587
 docling/models/factories/__init__.py,sha256=x_EM5dDg_A3HBcBYzOoqwmA2AFLtJ1IzYDPX-R1A-Sg,868
 docling/models/factories/base_factory.py,sha256=MfWIljMETi5aaVR-6qLTelW8u1gwDAQsOwg3fu7O4Qc,4028
 docling/models/factories/ocr_factory.py,sha256=G5RkmkKvkl-ihpo6qSj8WC77VdlVSQ1s0ekwUX2ILts,316
@@ -66,19 +69,20 @@ docling/models/plugins/defaults.py,sha256=qslXGnRX07Z3GGttNriqaox0v0vXp4zs4KLurH
 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=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-docling/models/vlm_models_inline/hf_transformers_model.py,sha256=SXSu6spu8zNCsrD32RU_irLs59ltF6PqbLVfpjDujmE,7285
-docling/models/vlm_models_inline/mlx_model.py,sha256=CFe1UNxQufZd5K4iaOW3HsplQBPb_1cENf3KIwWUSWw,5702
+docling/models/vlm_models_inline/hf_transformers_model.py,sha256=w9_N4ccjmYYK5yYQou0LSMGaj6gs8l0hULvXbkfYXSQ,7425
+docling/models/vlm_models_inline/mlx_model.py,sha256=qpyi6fGHm0vPqW2yeTsRBKOTTshNJ1LAPbH1SBDp8Y8,5784
 docling/pipeline/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-docling/pipeline/base_pipeline.py,sha256=DnuxAf7EQusdSRae0QUVth-0f2mSff8JZjX-2vazk00,8751
+docling/pipeline/asr_pipeline.py,sha256=tQkhu9fXdkSuYIL22xzV2YRUlQh-9qktHBbs2qeXhJI,9070
+docling/pipeline/base_pipeline.py,sha256=14yQrDjsojl4RgbBjKFSEfVBYR_sULZfBI1uDzFLi8Y,9331
 docling/pipeline/simple_pipeline.py,sha256=TXZOwR7hZRji462ZTIpte0VJjzbxvNVE8dbLFANDhSU,2253
-docling/pipeline/standard_pdf_pipeline.py,sha256=itCZPj7nMFAQtAlStfmWthpCIHZFUm9W5uTgvVi6PkQ,12738
+docling/pipeline/standard_pdf_pipeline.py,sha256=2Hqg2wnAXfbZbLUOQrRus8PMEuZ549jR1mfR86-CAB4,12659
 docling/pipeline/vlm_pipeline.py,sha256=IrjDbajCPmUPep_jATKNiABST4tQ8mvpkQz9mtBQ8qQ,15279
 docling/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling/utils/accelerator_utils.py,sha256=Fww4UiTiuIB91iuPgUZTy-DYpCGRMI8YuCYKhFb0gjA,2905
 docling/utils/api_image_request.py,sha256=_CgdzmPqdsyXmyYUFGLZcXcoH586qC6A1p5vsNbj1Q0,1416
 docling/utils/export.py,sha256=VwVUnYDk3mhGmISDbVm306fwpGNnoojouStBD4UajXI,4673
 docling/utils/glm_utils.py,sha256=TKOWQqWAHsX_w4fvoAA7_2xCi_urhnp1DsmjY8_sk5w,12274
-docling/utils/layout_postprocessor.py,sha256=3WCmkPsPJ80xfWzAUeWb5L9BmuwJ79ztctvbbUs8AfI,24068
+docling/utils/layout_postprocessor.py,sha256=laTPGGj-hv16Zh1TRcn8NK0POKs7d3jeaV1pRR_TjIU,24228
 docling/utils/locks.py,sha256=RzqQtD5UispgV71pGN_nU6GYfeN11BN0Sh_Dq9ycqGo,52
 docling/utils/model_downloader.py,sha256=6TDxFOvMRYT8JyYyaQS_wXMJzNga61ImY3sFdks66qM,4004
 docling/utils/ocr_utils.py,sha256=AOaDAHr5S74d-IRVR_LKhKynUTIurAwLJ3wNeY58gPA,2326
@@ -86,9 +90,9 @@ docling/utils/orientation.py,sha256=xXlOfowL54FKwjsTFrM7y3ogk1wChLNn_-u74tYIf1s,
 docling/utils/profiling.py,sha256=YaMGoB9MMZpagF9mb5ndoHj8Lpb9aIdb7El-Pl7IcFs,1753
 docling/utils/utils.py,sha256=kJtIYuzXeOyJHYlxmLAo7dGM5rEsDa1i84qEsUj1nio,1908
 docling/utils/visualization.py,sha256=tY2ylE2aiQKkmzlSLnFW-HTfFyqUUMguW18ldd1PLfo,2868
-docling-2.36.1.dist-info/licenses/LICENSE,sha256=mBb7ErEcM8VS9OhiGHnQ2kk75HwPhr54W1Oiz3965MY,1088
-docling-2.36.1.dist-info/METADATA,sha256=0Sl0LfCopUXdEd6mm2kqRpMgFoq6nvZBUXlIKeIqY_E,10036
-docling-2.36.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-docling-2.36.1.dist-info/entry_points.txt,sha256=hzVlbeE0aMSTQ9S0-NTYN0Hmgsn6qL_EA2qX4UbkAuY,149
-docling-2.36.1.dist-info/top_level.txt,sha256=vkIywP-USjFyYo1AIRQbWQQaL3xB5jf8vkCYdTIfNic,8
-docling-2.36.1.dist-info/RECORD,,
+docling-2.38.0.dist-info/licenses/LICENSE,sha256=mBb7ErEcM8VS9OhiGHnQ2kk75HwPhr54W1Oiz3965MY,1088
+docling-2.38.0.dist-info/METADATA,sha256=vT8Zko4wD8iyKUjLAJ83Cm7ntscjEk5ojHvcJXlvT5A,10273
+docling-2.38.0.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+docling-2.38.0.dist-info/entry_points.txt,sha256=hzVlbeE0aMSTQ9S0-NTYN0Hmgsn6qL_EA2qX4UbkAuY,149
+docling-2.38.0.dist-info/top_level.txt,sha256=vkIywP-USjFyYo1AIRQbWQQaL3xB5jf8vkCYdTIfNic,8
+docling-2.38.0.dist-info/RECORD,,