repo
stringlengths
7
90
file_url
stringlengths
81
315
file_path
stringlengths
4
228
content
stringlengths
0
32.8k
language
stringclasses
1 value
license
stringclasses
7 values
commit_sha
stringlengths
40
40
retrieved_at
stringdate
2026-01-04 14:38:15
2026-01-05 02:33:18
truncated
bool
2 classes
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/tesseract_ocr_model.py
docling/models/tesseract_ocr_model.py
from __future__ import annotations import logging from pathlib import Path from typing import Iterable, Optional, Type from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.doc.page import TextCell from docling.datamodel.accelerator_options import AcceleratorOptions from docling.datamodel.base_models import Page from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options import ( OcrOptions, TesseractOcrOptions, ) from docling.datamodel.settings import settings from docling.models.base_ocr_model import BaseOcrModel from docling.utils.ocr_utils import ( map_tesseract_script, parse_tesseract_orientation, tesseract_box_to_bounding_rectangle, ) from docling.utils.profiling import TimeRecorder _log = logging.getLogger(__name__) class TesseractOcrModel(BaseOcrModel): def __init__( self, enabled: bool, artifacts_path: Optional[Path], options: TesseractOcrOptions, accelerator_options: AcceleratorOptions, ): super().__init__( enabled=enabled, artifacts_path=artifacts_path, options=options, accelerator_options=accelerator_options, ) self.options: TesseractOcrOptions self._is_auto: bool = "auto" in self.options.lang self.scale = 3 # multiplier for 72 dpi == 216 dpi. self.reader = None self.script_readers: dict[str, tesserocr.PyTessBaseAPI] = {} if self.enabled: install_errmsg = ( "tesserocr is not correctly installed. " "Please install it via `pip install tesserocr` to use this OCR engine. " "Note that tesserocr might have to be manually compiled for working with " "your Tesseract installation. The Docling documentation provides examples for it. " "Alternatively, Docling has support for other OCR engines. See the documentation: " "https://docling-project.github.io/docling/installation/" ) missing_langs_errmsg = ( "tesserocr is not correctly configured. No language models have been detected. " "Please ensure that the TESSDATA_PREFIX envvar points to tesseract languages dir. " "You can find more information how to setup other OCR engines in Docling " "documentation: " "https://docling-project.github.io/docling/installation/" ) try: import tesserocr except ImportError: raise ImportError(install_errmsg) try: tesseract_version = tesserocr.tesseract_version() except Exception: raise ImportError(install_errmsg) _, self._tesserocr_languages = tesserocr.get_languages() if not self._tesserocr_languages: raise ImportError(missing_langs_errmsg) # Initialize the tesseractAPI _log.debug("Initializing TesserOCR: %s", tesseract_version) lang = "+".join(self.options.lang) if any(lang.startswith("script/") for lang in self._tesserocr_languages): self.script_prefix = "script/" else: self.script_prefix = "" tesserocr_kwargs = { "init": True, "oem": tesserocr.OEM.DEFAULT, } self.osd_reader = None if self.options.path is not None: tesserocr_kwargs["path"] = self.options.path # Set main OCR reader with configurable PSM main_psm = ( self.options.psm if self.options.psm is not None else tesserocr.PSM.AUTO ) if lang == "auto": self.reader = tesserocr.PyTessBaseAPI(psm=main_psm, **tesserocr_kwargs) else: self.reader = tesserocr.PyTessBaseAPI( lang=lang, psm=main_psm, **tesserocr_kwargs, ) # OSD reader must use PSM.OSD_ONLY for orientation detection self.osd_reader = tesserocr.PyTessBaseAPI( lang="osd", psm=tesserocr.PSM.OSD_ONLY, **tesserocr_kwargs ) self.reader_RIL = tesserocr.RIL def __del__(self): if self.reader is not None: # Finalize the tesseractAPI self.reader.End() for script in self.script_readers: self.script_readers[script].End() def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: if not self.enabled: yield from page_batch return for page_i, page in enumerate(page_batch): assert page._backend is not None if not page._backend.is_valid(): yield page else: with TimeRecorder(conv_res, "ocr"): assert self.reader is not None assert self.osd_reader is not None assert self._tesserocr_languages is not None ocr_rects = self.get_ocr_rects(page) all_ocr_cells = [] for ocr_rect_i, ocr_rect in enumerate(ocr_rects): # Skip zero area boxes if ocr_rect.area() == 0: continue high_res_image = page._backend.get_page_image( scale=self.scale, cropbox=ocr_rect ) local_reader = self.reader self.osd_reader.SetImage(high_res_image) doc_orientation = 0 osd = self.osd_reader.DetectOrientationScript() # No text, or Orientation and Script detection failure if osd is None: _log.error( "OSD failed for doc (doc %s, page: %s, " "OCR rectangle: %s)", conv_res.input.file, page_i, ocr_rect_i, ) # Skipping if OSD fail when in auto mode, otherwise proceed # to OCR in the hope OCR will succeed while OSD failed if self._is_auto: continue else: doc_orientation = parse_tesseract_orientation( osd["orient_deg"] ) if doc_orientation != 0: high_res_image = high_res_image.rotate( -doc_orientation, expand=True ) if self._is_auto: script = osd["script_name"] script = map_tesseract_script(script) lang = f"{self.script_prefix}{script}" # Check if the detected language is present in the system if lang not in self._tesserocr_languages: msg = f"Tesseract detected the script '{script}' and language '{lang}'." msg += " However this language is not installed in your system and will be ignored." _log.warning(msg) else: if script not in self.script_readers: import tesserocr self.script_readers[script] = ( tesserocr.PyTessBaseAPI( path=self.reader.GetDatapath(), lang=lang, psm=self.options.psm if self.options.psm is not None else tesserocr.PSM.AUTO, init=True, oem=tesserocr.OEM.DEFAULT, ) ) local_reader = self.script_readers[script] local_reader.SetImage(high_res_image) boxes = local_reader.GetComponentImages( self.reader_RIL.TEXTLINE, True ) cells = [] for ix, (im, box, _, _) in enumerate(boxes): # Set the area of interest. Tesseract uses Bottom-Left for the origin local_reader.SetRectangle( box["x"], box["y"], box["w"], box["h"] ) # Extract text within the bounding box text = local_reader.GetUTF8Text().strip() confidence = local_reader.MeanTextConf() left, top = box["x"], box["y"] right = left + box["w"] bottom = top + box["h"] bbox = BoundingBox( l=left, t=top, r=right, b=bottom, coord_origin=CoordOrigin.TOPLEFT, ) rect = tesseract_box_to_bounding_rectangle( bbox, original_offset=ocr_rect, scale=self.scale, orientation=doc_orientation, im_size=high_res_image.size, ) cells.append( TextCell( index=ix, text=text, orig=text, from_ocr=True, confidence=confidence, rect=rect, ) ) # del high_res_image all_ocr_cells.extend(cells) # Post-process the cells self.post_process_cells(all_ocr_cells, page) # DEBUG code: if settings.debug.visualize_ocr: self.draw_ocr_rects_and_cells(conv_res, page, ocr_rects) yield page @classmethod def get_options_type(cls) -> Type[OcrOptions]: return TesseractOcrOptions
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/rapid_ocr_model.py
docling/models/rapid_ocr_model.py
import logging from collections.abc import Iterable from pathlib import Path from typing import Literal, Optional, Type, TypedDict import numpy from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.doc.page import BoundingRectangle, TextCell from docling.datamodel.accelerator_options import AcceleratorDevice, AcceleratorOptions from docling.datamodel.base_models import Page from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options import ( OcrOptions, RapidOcrOptions, ) from docling.datamodel.settings import settings from docling.models.base_ocr_model import BaseOcrModel from docling.utils.accelerator_utils import decide_device from docling.utils.profiling import TimeRecorder from docling.utils.utils import download_url_with_progress _log = logging.getLogger(__name__) _ModelPathEngines = Literal["onnxruntime", "torch"] _ModelPathTypes = Literal[ "det_model_path", "cls_model_path", "rec_model_path", "rec_keys_path", "font_path" ] class _ModelPathDetail(TypedDict): url: str path: str class RapidOcrModel(BaseOcrModel): _model_repo_folder = "RapidOcr" # from https://github.com/RapidAI/RapidOCR/blob/main/python/rapidocr/default_models.yaml # matching the default config in https://github.com/RapidAI/RapidOCR/blob/main/python/rapidocr/config.yaml # and naming f"{file_info.engine_type.value}.{file_info.ocr_version.value}.{file_info.task_type.value}" _default_models: dict[ _ModelPathEngines, dict[_ModelPathTypes, _ModelPathDetail] ] = { "onnxruntime": { "det_model_path": { "url": "https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/v3.4.0/onnx/PP-OCRv4/det/ch_PP-OCRv4_det_infer.onnx", "path": "onnx/PP-OCRv4/det/ch_PP-OCRv4_det_infer.onnx", }, "cls_model_path": { "url": "https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/v3.4.0/onnx/PP-OCRv4/cls/ch_ppocr_mobile_v2.0_cls_infer.onnx", "path": "onnx/PP-OCRv4/cls/ch_ppocr_mobile_v2.0_cls_infer.onnx", }, "rec_model_path": { "url": "https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/v3.4.0/onnx/PP-OCRv4/rec/ch_PP-OCRv4_rec_infer.onnx", "path": "onnx/PP-OCRv4/rec/ch_PP-OCRv4_rec_infer.onnx", }, "rec_keys_path": { "url": "https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/v2.0.7/paddle/PP-OCRv4/rec/ch_PP-OCRv4_rec_infer/ppocr_keys_v1.txt", "path": "paddle/PP-OCRv4/rec/ch_PP-OCRv4_rec_infer/ppocr_keys_v1.txt", }, "font_path": { "url": "https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/v3.4.0/resources/fonts/FZYTK.TTF", "path": "fonts/FZYTK.TTF", }, }, "torch": { "det_model_path": { "url": "https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/v3.4.0/torch/PP-OCRv4/det/ch_PP-OCRv4_det_infer.pth", "path": "torch/PP-OCRv4/det/ch_PP-OCRv4_det_infer.pth", }, "cls_model_path": { "url": "https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/v3.4.0/torch/PP-OCRv4/cls/ch_ptocr_mobile_v2.0_cls_infer.pth", "path": "torch/PP-OCRv4/cls/ch_ptocr_mobile_v2.0_cls_infer.pth", }, "rec_model_path": { "url": "https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/v3.4.0/torch/PP-OCRv4/rec/ch_PP-OCRv4_rec_infer.pth", "path": "torch/PP-OCRv4/rec/ch_PP-OCRv4_rec_infer.pth", }, "rec_keys_path": { "url": "https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/v3.4.0/paddle/PP-OCRv4/rec/ch_PP-OCRv4_rec_infer/ppocr_keys_v1.txt", "path": "paddle/PP-OCRv4/rec/ch_PP-OCRv4_rec_infer/ppocr_keys_v1.txt", }, "font_path": { "url": "https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/v3.4.0/resources/fonts/FZYTK.TTF", "path": "fonts/FZYTK.TTF", }, }, } def __init__( self, enabled: bool, artifacts_path: Optional[Path], options: RapidOcrOptions, accelerator_options: AcceleratorOptions, ): super().__init__( enabled=enabled, artifacts_path=artifacts_path, options=options, accelerator_options=accelerator_options, ) self.options: RapidOcrOptions self.scale = 3 # multiplier for 72 dpi == 216 dpi. if self.enabled: try: from rapidocr import EngineType, RapidOCR # type: ignore except ImportError: raise ImportError( "RapidOCR is not installed. Please install it via `pip install rapidocr onnxruntime` to use this OCR engine. " "Alternatively, Docling has support for other OCR engines. See the documentation." ) # Decide the accelerator devices device = decide_device(accelerator_options.device) use_cuda = str(AcceleratorDevice.CUDA.value).lower() in device use_dml = accelerator_options.device == AcceleratorDevice.AUTO intra_op_num_threads = accelerator_options.num_threads gpu_id = 0 if use_cuda and ":" in device: gpu_id = int(device.split(":")[1]) _ALIASES = { "onnxruntime": EngineType.ONNXRUNTIME, "openvino": EngineType.OPENVINO, "paddle": EngineType.PADDLE, "torch": EngineType.TORCH, } backend_enum = _ALIASES.get(self.options.backend, EngineType.ONNXRUNTIME) det_model_path = self.options.det_model_path cls_model_path = self.options.cls_model_path rec_model_path = self.options.rec_model_path rec_keys_path = self.options.rec_keys_path font_path = self.options.font_path if artifacts_path is not None: det_model_path = ( det_model_path or artifacts_path / self._model_repo_folder / self._default_models[backend_enum.value]["det_model_path"]["path"] ) cls_model_path = ( cls_model_path or artifacts_path / self._model_repo_folder / self._default_models[backend_enum.value]["cls_model_path"]["path"] ) rec_model_path = ( rec_model_path or artifacts_path / self._model_repo_folder / self._default_models[backend_enum.value]["rec_model_path"]["path"] ) rec_keys_path = ( rec_keys_path or artifacts_path / self._model_repo_folder / self._default_models[backend_enum.value]["rec_keys_path"]["path"] ) font_path = ( font_path or artifacts_path / self._model_repo_folder / self._default_models[backend_enum.value]["font_path"]["path"] ) for model_path in ( rec_keys_path, cls_model_path, rec_model_path, rec_keys_path, font_path, ): if model_path is None: continue if not Path(model_path).exists(): _log.warning(f"The provided model path {model_path} is not found.") params = { # Global settings (these are still correct) "Global.text_score": self.options.text_score, "Global.font_path": self.options.font_path, # "Global.verbose": self.options.print_verbose, # Detection model settings "Det.model_path": det_model_path, "Det.use_cuda": use_cuda, "Det.use_dml": use_dml, "Det.intra_op_num_threads": intra_op_num_threads, # Classification model settings "Cls.model_path": cls_model_path, "Cls.use_cuda": use_cuda, "Cls.use_dml": use_dml, "Cls.intra_op_num_threads": intra_op_num_threads, # Recognition model settings "Rec.model_path": rec_model_path, "Rec.font_path": self.options.rec_font_path, "Rec.rec_keys_path": rec_keys_path, "Rec.use_cuda": use_cuda, "Rec.use_dml": use_dml, "Rec.intra_op_num_threads": intra_op_num_threads, "Det.engine_type": backend_enum, "Cls.engine_type": backend_enum, "Rec.engine_type": backend_enum, "EngineConfig.paddle.use_cuda": use_cuda, "EngineConfig.paddle.gpu_id": gpu_id, "EngineConfig.torch.use_cuda": use_cuda, "EngineConfig.torch.gpu_id": gpu_id, } if self.options.rec_font_path is not None: _log.warning( "The 'rec_font_path' option for RapidOCR is deprecated. Please use 'font_path' instead." ) user_params = self.options.rapidocr_params if user_params: _log.debug("Overwriting RapidOCR params with user-provided values.") params.update(user_params) self.reader = RapidOCR( params=params, ) @staticmethod def download_models( backend: _ModelPathEngines, local_dir: Optional[Path] = None, force: bool = False, progress: bool = False, ) -> Path: if local_dir is None: local_dir = settings.cache_dir / "models" / RapidOcrModel._model_repo_folder local_dir.mkdir(parents=True, exist_ok=True) # Download models for model_type, model_details in RapidOcrModel._default_models[backend].items(): output_path = local_dir / model_details["path"] if output_path.exists() and not force: continue output_path.parent.mkdir(exist_ok=True, parents=True) buf = download_url_with_progress(model_details["url"], progress=progress) with output_path.open("wb") as fw: fw.write(buf.read()) return local_dir def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: if not self.enabled: yield from page_batch return for page in page_batch: assert page._backend is not None if not page._backend.is_valid(): yield page else: with TimeRecorder(conv_res, "ocr"): ocr_rects = self.get_ocr_rects(page) all_ocr_cells = [] for ocr_rect in ocr_rects: # Skip zero area boxes if ocr_rect.area() == 0: continue high_res_image = page._backend.get_page_image( scale=self.scale, cropbox=ocr_rect ) im = numpy.array(high_res_image) result = self.reader( im, use_det=self.options.use_det, use_cls=self.options.use_cls, use_rec=self.options.use_rec, ) if result is None or result.boxes is None: _log.warning("RapidOCR returned empty result!") continue result = list( zip(result.boxes.tolist(), result.txts, result.scores) ) del high_res_image del im if result is not None: cells = [ TextCell( index=ix, text=line[1], orig=line[1], confidence=line[2], from_ocr=True, rect=BoundingRectangle.from_bounding_box( BoundingBox.from_tuple( coord=( (line[0][0][0] / self.scale) + ocr_rect.l, (line[0][0][1] / self.scale) + ocr_rect.t, (line[0][2][0] / self.scale) + ocr_rect.l, (line[0][2][1] / self.scale) + ocr_rect.t, ), origin=CoordOrigin.TOPLEFT, ) ), ) for ix, line in enumerate(result) ] all_ocr_cells.extend(cells) # Post-process the cells self.post_process_cells(all_ocr_cells, page) # DEBUG code: if settings.debug.visualize_ocr: self.draw_ocr_rects_and_cells(conv_res, page, ocr_rects) yield page @classmethod def get_options_type(cls) -> Type[OcrOptions]: return RapidOcrOptions
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/api_vlm_model.py
docling/models/api_vlm_model.py
from collections.abc import Iterable from concurrent.futures import ThreadPoolExecutor from typing import Union import numpy as np from PIL.Image import Image from docling.datamodel.base_models import Page, VlmPrediction, VlmStopReason from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options_vlm_model import ApiVlmOptions from docling.exceptions import OperationNotAllowed from docling.models.base_model import BaseVlmPageModel from docling.models.utils.generation_utils import GenerationStopper from docling.utils.api_image_request import ( api_image_request, api_image_request_streaming, ) from docling.utils.profiling import TimeRecorder class ApiVlmModel(BaseVlmPageModel): # Override the vlm_options type annotation from BaseVlmPageModel vlm_options: ApiVlmOptions # type: ignore[assignment] def __init__( self, enabled: bool, enable_remote_services: bool, vlm_options: ApiVlmOptions, ): self.enabled = enabled self.vlm_options = vlm_options if self.enabled: if not enable_remote_services: raise OperationNotAllowed( "Connections to remote services is only allowed when set explicitly. " "pipeline_options.enable_remote_services=True, or using the CLI " "--enable-remote-services." ) self.timeout = self.vlm_options.timeout self.concurrency = self.vlm_options.concurrency self.params = { **self.vlm_options.params, "temperature": self.vlm_options.temperature, } def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: page_list = list(page_batch) if not page_list: return original_order = page_list[:] valid_pages = [] for page in page_list: assert page._backend is not None if page._backend.is_valid(): valid_pages.append(page) # Process valid pages in batch if valid_pages: with TimeRecorder(conv_res, "vlm"): # Prepare images and prompts for batch processing images = [] prompts = [] pages_with_images = [] for page in valid_pages: assert page.size is not None hi_res_image = page.get_image( scale=self.vlm_options.scale, max_size=self.vlm_options.max_size ) # Only process pages with valid images if hi_res_image is not None: images.append(hi_res_image) prompt = self._build_prompt_safe(page) prompts.append(prompt) pages_with_images.append(page) # Use process_images for the actual inference if images: # Only if we have valid images with TimeRecorder(conv_res, "vlm_inference"): predictions = list(self.process_images(images, prompts)) # Attach results to pages for page, prediction in zip(pages_with_images, predictions): page.predictions.vlm_response = prediction # Yield pages preserving original order for page in original_order: yield page def process_images( self, image_batch: Iterable[Union[Image, np.ndarray]], prompt: Union[str, list[str]], ) -> Iterable[VlmPrediction]: """Process raw images without page metadata.""" images = list(image_batch) # Handle prompt parameter if isinstance(prompt, str): prompts = [prompt] * len(images) elif isinstance(prompt, list): if len(prompt) != len(images): raise ValueError( f"Prompt list length ({len(prompt)}) must match image count ({len(images)})" ) prompts = prompt def _process_single_image(image_prompt_pair): image, prompt_text = image_prompt_pair # Convert numpy array to PIL Image if needed if isinstance(image, np.ndarray): if image.ndim == 3 and image.shape[2] in [3, 4]: from PIL import Image as PILImage image = PILImage.fromarray(image.astype(np.uint8)) elif image.ndim == 2: from PIL import Image as PILImage image = PILImage.fromarray(image.astype(np.uint8), mode="L") else: raise ValueError(f"Unsupported numpy array shape: {image.shape}") # Ensure image is in RGB mode if image.mode != "RGB": image = image.convert("RGB") stop_reason = VlmStopReason.UNSPECIFIED if self.vlm_options.custom_stopping_criteria: # Instantiate any GenerationStopper classes before passing to streaming instantiated_stoppers = [] for criteria in self.vlm_options.custom_stopping_criteria: if isinstance(criteria, GenerationStopper): instantiated_stoppers.append(criteria) elif isinstance(criteria, type) and issubclass( criteria, GenerationStopper ): instantiated_stoppers.append(criteria()) # Skip non-GenerationStopper criteria (should have been caught in validation) # Streaming path with early abort support page_tags, num_tokens = api_image_request_streaming( image=image, prompt=prompt_text, url=self.vlm_options.url, timeout=self.timeout, headers=self.vlm_options.headers, generation_stoppers=instantiated_stoppers, **self.params, ) else: # Non-streaming fallback (existing behavior) page_tags, num_tokens, stop_reason = api_image_request( image=image, prompt=prompt_text, url=self.vlm_options.url, timeout=self.timeout, headers=self.vlm_options.headers, **self.params, ) page_tags = self.vlm_options.decode_response(page_tags) input_prompt = prompt_text if self.vlm_options.track_input_prompt else None return VlmPrediction( text=page_tags, num_tokens=num_tokens, stop_reason=stop_reason, input_prompt=input_prompt, ) with ThreadPoolExecutor(max_workers=self.concurrency) as executor: yield from executor.map(_process_single_image, zip(images, prompts))
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/base_ocr_model.py
docling/models/base_ocr_model.py
import copy import logging from abc import abstractmethod from collections.abc import Iterable from pathlib import Path from typing import TYPE_CHECKING, List, Optional, Type import numpy as np from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.doc.page import TextCell from PIL import Image, ImageDraw from rtree import index from docling.datamodel.accelerator_options import AcceleratorOptions from docling.datamodel.base_models import Page from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options import OcrOptions from docling.datamodel.settings import settings from docling.models.base_model import BaseModelWithOptions, BasePageModel _log = logging.getLogger(__name__) class BaseOcrModel(BasePageModel, BaseModelWithOptions): def __init__( self, *, enabled: bool, artifacts_path: Optional[Path], options: OcrOptions, accelerator_options: AcceleratorOptions, ): # Make sure any delay/error from import occurs on ocr model init and not first use from scipy.ndimage import binary_dilation, find_objects, label self.enabled = enabled self.options = options # Computes the optimum amount and coordinates of rectangles to OCR on a given page def get_ocr_rects(self, page: Page) -> List[BoundingBox]: from scipy.ndimage import binary_dilation, find_objects, label BITMAP_COVERAGE_TRESHOLD = 0.75 assert page.size is not None def find_ocr_rects(size, bitmap_rects): image = Image.new( "1", (round(size.width), round(size.height)) ) # '1' mode is binary # Draw all bitmap rects into a binary image draw = ImageDraw.Draw(image) for rect in bitmap_rects: x0, y0, x1, y1 = rect.as_tuple() x0, y0, x1, y1 = round(x0), round(y0), round(x1), round(y1) draw.rectangle([(x0, y0), (x1, y1)], fill=1) np_image = np.array(image) # Dilate the image by 10 pixels to merge nearby bitmap rectangles structure = np.ones( (20, 20) ) # Create a 20x20 structure element (10 pixels in all directions) np_image = binary_dilation(np_image > 0, structure=structure) # Find the connected components labeled_image, num_features = label( np_image > 0 ) # Label black (0 value) regions # Find enclosing bounding boxes for each connected component. slices = find_objects(labeled_image) bounding_boxes = [ BoundingBox( l=slc[1].start, t=slc[0].start, r=slc[1].stop - 1, b=slc[0].stop - 1, coord_origin=CoordOrigin.TOPLEFT, ) for slc in slices ] # Compute area fraction on page covered by bitmaps area_frac = np.sum(np_image > 0) / (size.width * size.height) return (area_frac, bounding_boxes) # fraction covered # boxes if page._backend is not None: bitmap_rects = page._backend.get_bitmap_rects() else: bitmap_rects = [] coverage, ocr_rects = find_ocr_rects(page.size, bitmap_rects) # return full-page rectangle if page is dominantly covered with bitmaps if self.options.force_full_page_ocr or coverage > max( BITMAP_COVERAGE_TRESHOLD, self.options.bitmap_area_threshold ): return [ BoundingBox( l=0, t=0, r=page.size.width, b=page.size.height, coord_origin=CoordOrigin.TOPLEFT, ) ] # return individual rectangles if the bitmap coverage is above the threshold elif coverage > self.options.bitmap_area_threshold: return ocr_rects else: # overall coverage of bitmaps is too low, drop all bitmap rectangles. return [] # Filters OCR cells by dropping any OCR cell that intersects with an existing programmatic cell. def _filter_ocr_cells( self, ocr_cells: List[TextCell], programmatic_cells: List[TextCell] ) -> List[TextCell]: # Create R-tree index for programmatic cells p = index.Property() p.dimension = 2 idx = index.Index(properties=p) for i, cell in enumerate(programmatic_cells): idx.insert(i, cell.rect.to_bounding_box().as_tuple()) def is_overlapping_with_existing_cells(ocr_cell): # Query the R-tree to get overlapping rectangles possible_matches_index = list( idx.intersection(ocr_cell.rect.to_bounding_box().as_tuple()) ) return ( len(possible_matches_index) > 0 ) # this is a weak criterion but it works. filtered_ocr_cells = [ rect for rect in ocr_cells if not is_overlapping_with_existing_cells(rect) ] return filtered_ocr_cells def post_process_cells(self, ocr_cells: List[TextCell], page: Page) -> None: r""" Post-process the OCR cells and update the page object. Updates parsed_page.textline_cells directly since page.cells is now read-only. """ # Get existing cells from the read-only property existing_cells = page.cells # Combine existing and OCR cells with overlap filtering final_cells = self._combine_cells(existing_cells, ocr_cells) assert page.parsed_page is not None # Update parsed_page.textline_cells directly page.parsed_page.textline_cells = final_cells page.parsed_page.has_lines = len(final_cells) > 0 # When force_full_page_ocr is used, PDF-extracted word/char cells are # unreliable. Filter out cells where from_ocr=False, keeping any OCR- # generated cells. This ensures downstream components (e.g., table # structure model) fall back to OCR-extracted textline cells. if self.options.force_full_page_ocr: page.parsed_page.word_cells = [ c for c in page.parsed_page.word_cells if c.from_ocr ] page.parsed_page.char_cells = [ c for c in page.parsed_page.char_cells if c.from_ocr ] page.parsed_page.has_words = len(page.parsed_page.word_cells) > 0 page.parsed_page.has_chars = len(page.parsed_page.char_cells) > 0 def _combine_cells( self, existing_cells: List[TextCell], ocr_cells: List[TextCell] ) -> List[TextCell]: """Combine existing and OCR cells with filtering and re-indexing.""" if self.options.force_full_page_ocr: combined = ocr_cells else: filtered_ocr_cells = self._filter_ocr_cells(ocr_cells, existing_cells) combined = list(existing_cells) + filtered_ocr_cells # Re-index in-place for i, cell in enumerate(combined): cell.index = i return combined def draw_ocr_rects_and_cells(self, conv_res, page, ocr_rects, show: bool = False): image = copy.deepcopy(page.image) scale_x = image.width / page.size.width scale_y = image.height / page.size.height draw = ImageDraw.Draw(image, "RGBA") # Draw OCR rectangles as yellow filled rect for rect in ocr_rects: x0, y0, x1, y1 = rect.as_tuple() y0 *= scale_x y1 *= scale_y x0 *= scale_x x1 *= scale_x shade_color = (255, 255, 0, 40) # transparent yellow draw.rectangle([(x0, y0), (x1, y1)], fill=shade_color, outline=None) # Draw OCR and programmatic cells for tc in page.cells: x0, y0, x1, y1 = tc.rect.to_bounding_box().as_tuple() y0 *= scale_x y1 *= scale_y x0 *= scale_x x1 *= scale_x if y1 <= y0: y1, y0 = y0, y1 color = "magenta" if tc.from_ocr else "gray" draw.rectangle([(x0, y0), (x1, y1)], outline=color) if show: image.show() else: out_path: Path = ( Path(settings.debug.debug_output_path) / f"debug_{conv_res.input.file.stem}" ) out_path.mkdir(parents=True, exist_ok=True) out_file = out_path / f"ocr_page_{page.page_no:05}.png" image.save(str(out_file), format="png") @abstractmethod def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: pass @classmethod @abstractmethod def get_options_type(cls) -> Type[OcrOptions]: pass
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/base_layout_model.py
docling/models/base_layout_model.py
from __future__ import annotations from abc import ABC, abstractmethod from collections.abc import Iterable, Sequence from typing import Type from docling.datamodel.base_models import LayoutPrediction, Page from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options import BaseLayoutOptions from docling.models.base_model import BaseModelWithOptions, BasePageModel class BaseLayoutModel(BasePageModel, BaseModelWithOptions, ABC): """Shared interface for layout models.""" @classmethod @abstractmethod def get_options_type(cls) -> Type[BaseLayoutOptions]: """Return the options type supported by this layout model.""" @abstractmethod def predict_layout( self, conv_res: ConversionResult, pages: Sequence[Page], ) -> Sequence[LayoutPrediction]: """Produce layout predictions for the provided pages.""" def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page], ) -> Iterable[Page]: pages = list(page_batch) predictions = self.predict_layout(conv_res, pages) for page, prediction in zip(pages, predictions): page.predictions.layout = prediction yield page
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/page_assemble_model.py
docling/models/page_assemble_model.py
import logging import re from collections.abc import Iterable from typing import List import numpy as np from pydantic import BaseModel from docling.datamodel.base_models import ( AssembledUnit, ContainerElement, FigureElement, Page, PageElement, Table, TextElement, ) from docling.datamodel.document import ConversionResult from docling.models.base_model import BasePageModel from docling.models.layout_model import LayoutModel from docling.utils.profiling import TimeRecorder _log = logging.getLogger(__name__) class PageAssembleOptions(BaseModel): pass class PageAssembleModel(BasePageModel): def __init__(self, options: PageAssembleOptions): self.options = options def sanitize_text(self, lines): if len(lines) <= 1: return " ".join(lines) for ix, line in enumerate(lines[1:]): prev_line = lines[ix] if prev_line.endswith("-"): prev_words = re.findall(r"\b[\w]+\b", prev_line) line_words = re.findall(r"\b[\w]+\b", line) if ( len(prev_words) and len(line_words) and prev_words[-1].isalnum() and line_words[0].isalnum() ): lines[ix] = prev_line[:-1] else: lines[ix] += " " sanitized_text = "".join(lines) # Text normalization sanitized_text = sanitized_text.replace("⁄", "/") # noqa: RUF001 sanitized_text = sanitized_text.replace("’", "'") # noqa: RUF001 sanitized_text = sanitized_text.replace("‘", "'") # noqa: RUF001 sanitized_text = sanitized_text.replace("“", '"') sanitized_text = sanitized_text.replace("”", '"') sanitized_text = sanitized_text.replace("•", "·") return sanitized_text.strip() # Strip any leading or trailing whitespace def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: for page in page_batch: assert page._backend is not None if not page._backend.is_valid(): yield page else: with TimeRecorder(conv_res, "page_assemble"): assert page.predictions.layout is not None # assembles some JSON output page by page. elements: List[PageElement] = [] headers: List[PageElement] = [] body: List[PageElement] = [] for cluster in page.predictions.layout.clusters: # _log.info("Cluster label seen:", cluster.label) if cluster.label in LayoutModel.TEXT_ELEM_LABELS: textlines = [ cell.text.replace("\x02", "-").strip() for cell in cluster.cells if len(cell.text.strip()) > 0 ] text = self.sanitize_text(textlines) text_el = TextElement( label=cluster.label, id=cluster.id, text=text, page_no=page.page_no, cluster=cluster, ) elements.append(text_el) if cluster.label in LayoutModel.PAGE_HEADER_LABELS: headers.append(text_el) else: body.append(text_el) elif cluster.label in LayoutModel.TABLE_LABELS: tbl = None if page.predictions.tablestructure: tbl = page.predictions.tablestructure.table_map.get( cluster.id, None ) if not tbl: # fallback: add table without structure, if it isn't present tbl = Table( label=cluster.label, id=cluster.id, text="", otsl_seq=[], table_cells=[], cluster=cluster, page_no=page.page_no, ) elements.append(tbl) body.append(tbl) elif cluster.label == LayoutModel.FIGURE_LABEL: fig = None if page.predictions.figures_classification: fig = page.predictions.figures_classification.figure_map.get( cluster.id, None ) if not fig: # fallback: add figure without classification, if it isn't present fig = FigureElement( label=cluster.label, id=cluster.id, text="", data=None, cluster=cluster, page_no=page.page_no, ) elements.append(fig) body.append(fig) elif cluster.label in LayoutModel.CONTAINER_LABELS: container_el = ContainerElement( label=cluster.label, id=cluster.id, page_no=page.page_no, cluster=cluster, ) elements.append(container_el) body.append(container_el) page.assembled = AssembledUnit( elements=elements, headers=headers, body=body ) yield page
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/code_formula_model.py
docling/models/code_formula_model.py
import re from collections.abc import Iterable from pathlib import Path from typing import List, Literal, Optional, Tuple, Union import numpy as np from docling_core.types.doc import ( CodeItem, DocItemLabel, DoclingDocument, NodeItem, TextItem, ) from docling_core.types.doc.labels import CodeLanguageLabel from PIL import Image from pydantic import BaseModel from transformers import AutoModelForImageTextToText, AutoProcessor from docling.datamodel.accelerator_options import AcceleratorDevice, AcceleratorOptions from docling.datamodel.base_models import ItemAndImageEnrichmentElement from docling.models.base_model import BaseItemAndImageEnrichmentModel from docling.models.utils.hf_model_download import download_hf_model from docling.utils.accelerator_utils import decide_device class CodeFormulaModelOptions(BaseModel): """ Configuration options for the CodeFormulaModel. Attributes ---------- kind : str Type of the model. Fixed value "code_formula". do_code_enrichment : bool True if code enrichment is enabled, False otherwise. do_formula_enrichment : bool True if formula enrichment is enabled, False otherwise. """ kind: Literal["code_formula"] = "code_formula" do_code_enrichment: bool = True do_formula_enrichment: bool = True class CodeFormulaModel(BaseItemAndImageEnrichmentModel): """ Model for processing and enriching documents with code and formula predictions. Attributes ---------- enabled : bool True if the model is enabled, False otherwise. options : CodeFormulaModelOptions Configuration options for the CodeFormulaModel. code_formula_model : CodeFormulaPredictor The predictor model for code and formula processing. Methods ------- __init__(self, enabled, artifacts_path, accelerator_options, code_formula_options) Initializes the CodeFormulaModel with the given configuration options. is_processable(self, doc, element) Determines if a given element in a document can be processed by the model. __call__(self, doc, element_batch) Processes the given batch of elements and enriches them with predictions. """ _model_repo_folder = "docling-project--CodeFormulaV2" elements_batch_size = 5 images_scale = 1.67 # = 120 dpi, aligned with training data resolution expansion_factor = 0.18 def __init__( self, enabled: bool, artifacts_path: Optional[Path], options: CodeFormulaModelOptions, accelerator_options: AcceleratorOptions, ): """ Initializes the CodeFormulaModel with the given configuration. Parameters ---------- enabled : bool True if the model is enabled, False otherwise. artifacts_path : Path Path to the directory containing the model artifacts. options : CodeFormulaModelOptions Configuration options for the model. accelerator_options : AcceleratorOptions Options specifying the device and number of threads for acceleration. """ self.enabled = enabled self.options = options if self.enabled: self.device = decide_device( accelerator_options.device, supported_devices=[AcceleratorDevice.CPU, AcceleratorDevice.CUDA], ) if artifacts_path is None: artifacts_path = self.download_models() else: artifacts_path = artifacts_path / self._model_repo_folder self._processor = AutoProcessor.from_pretrained( artifacts_path, ) self._model_max_length = self._processor.tokenizer.model_max_length self._model = AutoModelForImageTextToText.from_pretrained( artifacts_path, device_map=self.device ) self._model.eval() @staticmethod def download_models( local_dir: Optional[Path] = None, force: bool = False, progress: bool = False, ) -> Path: return download_hf_model( repo_id="docling-project/CodeFormulaV2", revision="main", local_dir=local_dir, force=force, progress=progress, ) def is_processable(self, doc: DoclingDocument, element: NodeItem) -> bool: """ Determines if a given element in a document can be processed by the model. Parameters ---------- doc : DoclingDocument The document being processed. element : NodeItem The element within the document to check. Returns ------- bool True if the element can be processed, False otherwise. """ return self.enabled and ( (isinstance(element, CodeItem) and self.options.do_code_enrichment) or ( isinstance(element, TextItem) and element.label == DocItemLabel.FORMULA and self.options.do_formula_enrichment ) ) def _extract_code_language(self, input_string: str) -> Tuple[str, Optional[str]]: """Extracts a programming language from the beginning of a string. This function checks if the input string starts with a pattern of the form ``<_some_language_>``. If it does, it extracts the language string and returns a tuple of (remainder, language). Otherwise, it returns the original string and `None`. Args: input_string (str): The input string, which may start with ``<_language_>``. Returns: Tuple[str, Optional[str]]: A tuple where: - The first element is either: - The remainder of the string (everything after ``<_language_>``), if a match is found; or - The original string, if no match is found. - The second element is the extracted language if a match is found; otherwise, `None`. """ pattern = r"^<_([^_>]+)_>\s*(.*)" match = re.match(pattern, input_string, flags=re.DOTALL) if match: language = str(match.group(1)) # the captured programming language remainder = str(match.group(2)) # everything after the <_language_> return remainder, language else: return input_string, None def _get_code_language_enum(self, value: Optional[str]) -> CodeLanguageLabel: """ Converts a string to a corresponding `CodeLanguageLabel` enum member. If the provided string does not match any value in `CodeLanguageLabel`, it defaults to `CodeLanguageLabel.UNKNOWN`. Args: value (Optional[str]): The string representation of the code language or None. Returns: CodeLanguageLabel: The corresponding enum member if the value is valid, otherwise `CodeLanguageLabel.UNKNOWN`. """ if not isinstance(value, str): return CodeLanguageLabel.UNKNOWN try: return CodeLanguageLabel(value) except ValueError: return CodeLanguageLabel.UNKNOWN def _get_prompt(self, label: str) -> str: """ Constructs the prompt for the model based on the input label. Parameters ---------- label : str The type of input, either 'code' or 'formula'. Returns ------- str The constructed prompt including necessary tokens and query. Raises ------ NotImplementedError If the label is not 'code' or 'formula'. """ if label == "code": query = "<code>" elif label == "formula": query = "<formula>" else: raise NotImplementedError("Label must be either code or formula") messages = [ { "role": "user", "content": [{"type": "image"}, {"type": "text", "text": query}], }, ] prompt = self._processor.apply_chat_template( messages, add_generation_prompt=True ) return prompt def _post_process(self, texts: list[str]) -> list[str]: """ Processes a list of text strings by truncating at '<end_of_utterance>' and removing a predefined set of unwanted substrings. Parameters ---------- texts : list[str] A list of strings to be post-processed. Returns ------- list[str] A list of cleaned strings with specified substrings removed and truncated at '<end_of_utterance>' if present. """ to_remove = ["</code>", "</formula>", "<loc_0><loc_0><loc_500><loc_500>"] def clean_text(text: str) -> str: idx = text.find("<end_of_utterance>") if idx != -1: text = text[:idx] for token in to_remove: if token in text: text = text.replace(token, "") return text.lstrip() return [clean_text(t) for t in texts] def __call__( self, doc: DoclingDocument, element_batch: Iterable[ItemAndImageEnrichmentElement], ) -> Iterable[NodeItem]: """ Processes the given batch of elements and enriches them with predictions. Parameters ---------- doc : DoclingDocument The document being processed. element_batch : Iterable[ItemAndImageEnrichmentElement] A batch of elements to be processed. Returns ------- Iterable[Any] An iterable of enriched elements. """ if not self.enabled: for element in element_batch: yield element.item return labels: List[str] = [] images: List[Union[Image.Image, np.ndarray]] = [] elements: List[TextItem] = [] for el in element_batch: elements.append(el.item) # type: ignore[arg-type] labels.append(el.item.label) # type: ignore[attr-defined] images.append(el.image) prompts = [self._get_prompt(label) for label in labels] inputs = self._processor( text=prompts, images=images, return_tensors="pt", ) inputs = inputs.to(self.device) gen_kwargs = dict( max_new_tokens=self._model_max_length - inputs.input_ids.shape[1], use_cache=True, do_sample=False, ) generated_ids = self._model.generate(**inputs, **gen_kwargs) outputs = self._processor.batch_decode( generated_ids[:, inputs.input_ids.shape[1] :], skip_special_tokens=False ) outputs = self._post_process(outputs) for item, output in zip(elements, outputs): if isinstance(item, CodeItem): output, code_language = self._extract_code_language(output) item.code_language = self._get_code_language_enum(code_language) item.text = output yield item
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/base_table_model.py
docling/models/base_table_model.py
from __future__ import annotations from abc import ABC, abstractmethod from collections.abc import Iterable, Sequence from typing import Type from docling.datamodel.base_models import Page, TableStructurePrediction from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options import BaseTableStructureOptions from docling.models.base_model import BaseModelWithOptions, BasePageModel class BaseTableStructureModel(BasePageModel, BaseModelWithOptions, ABC): """Shared interface for table structure models.""" enabled: bool @classmethod @abstractmethod def get_options_type(cls) -> Type[BaseTableStructureOptions]: """Return the options type supported by this table model.""" @abstractmethod def predict_tables( self, conv_res: ConversionResult, pages: Sequence[Page], ) -> Sequence[TableStructurePrediction]: """Produce table structure predictions for the provided pages.""" def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page], ) -> Iterable[Page]: if not getattr(self, "enabled", True): yield from page_batch return pages = list(page_batch) predictions = self.predict_tables(conv_res, pages) for page, prediction in zip(pages, predictions): page.predictions.tablestructure = prediction yield page
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/layout_model.py
docling/models/layout_model.py
import copy import logging import warnings from collections.abc import Sequence from pathlib import Path from typing import List, Optional, Union import numpy as np from docling_core.types.doc import DocItemLabel from PIL import Image from docling.datamodel.accelerator_options import AcceleratorOptions from docling.datamodel.base_models import BoundingBox, Cluster, LayoutPrediction, Page from docling.datamodel.document import ConversionResult from docling.datamodel.layout_model_specs import DOCLING_LAYOUT_V2, LayoutModelConfig from docling.datamodel.pipeline_options import LayoutOptions from docling.datamodel.settings import settings from docling.models.base_layout_model import BaseLayoutModel from docling.models.utils.hf_model_download import download_hf_model from docling.utils.accelerator_utils import decide_device from docling.utils.layout_postprocessor import LayoutPostprocessor from docling.utils.profiling import TimeRecorder from docling.utils.visualization import draw_clusters _log = logging.getLogger(__name__) class LayoutModel(BaseLayoutModel): TEXT_ELEM_LABELS = [ DocItemLabel.TEXT, DocItemLabel.FOOTNOTE, DocItemLabel.CAPTION, DocItemLabel.CHECKBOX_UNSELECTED, DocItemLabel.CHECKBOX_SELECTED, DocItemLabel.SECTION_HEADER, DocItemLabel.PAGE_HEADER, DocItemLabel.PAGE_FOOTER, DocItemLabel.CODE, DocItemLabel.LIST_ITEM, DocItemLabel.FORMULA, ] PAGE_HEADER_LABELS = [DocItemLabel.PAGE_HEADER, DocItemLabel.PAGE_FOOTER] TABLE_LABELS = [DocItemLabel.TABLE, DocItemLabel.DOCUMENT_INDEX] FIGURE_LABEL = DocItemLabel.PICTURE FORMULA_LABEL = DocItemLabel.FORMULA CONTAINER_LABELS = [DocItemLabel.FORM, DocItemLabel.KEY_VALUE_REGION] def __init__( self, artifacts_path: Optional[Path], accelerator_options: AcceleratorOptions, options: LayoutOptions, ): from docling_ibm_models.layoutmodel.layout_predictor import LayoutPredictor self.options = options device = decide_device(accelerator_options.device) layout_model_config = options.model_spec model_repo_folder = layout_model_config.model_repo_folder model_path = layout_model_config.model_path if artifacts_path is None: artifacts_path = ( self.download_models(layout_model_config=layout_model_config) / model_path ) else: if (artifacts_path / model_repo_folder).exists(): artifacts_path = artifacts_path / model_repo_folder / model_path elif (artifacts_path / model_path).exists(): warnings.warn( "The usage of artifacts_path containing directly " f"{model_path} is deprecated. Please point " "the artifacts_path to the parent containing " f"the {model_repo_folder} folder.", DeprecationWarning, stacklevel=3, ) artifacts_path = artifacts_path / model_path self.layout_predictor = LayoutPredictor( artifact_path=str(artifacts_path), device=device, num_threads=accelerator_options.num_threads, ) @classmethod def get_options_type(cls) -> type[LayoutOptions]: return LayoutOptions @staticmethod def download_models( local_dir: Optional[Path] = None, force: bool = False, progress: bool = False, layout_model_config: LayoutModelConfig = LayoutOptions().model_spec, # use default ) -> Path: return download_hf_model( repo_id=layout_model_config.repo_id, revision=layout_model_config.revision, local_dir=local_dir, force=force, progress=progress, ) def draw_clusters_and_cells_side_by_side( self, conv_res, page, clusters, mode_prefix: str, show: bool = False ): """ Draws a page image side by side with clusters filtered into two categories: - Left: Clusters excluding FORM, KEY_VALUE_REGION, and PICTURE. - Right: Clusters including FORM, KEY_VALUE_REGION, and PICTURE. Includes label names and confidence scores for each cluster. """ scale_x = page.image.width / page.size.width scale_y = page.image.height / page.size.height # Filter clusters for left and right images exclude_labels = { DocItemLabel.FORM, DocItemLabel.KEY_VALUE_REGION, DocItemLabel.PICTURE, } left_clusters = [c for c in clusters if c.label not in exclude_labels] right_clusters = [c for c in clusters if c.label in exclude_labels] # Create a deep copy of the original image for both sides left_image = page.image.copy() right_image = page.image.copy() # Draw clusters on both images draw_clusters(left_image, left_clusters, scale_x, scale_y) draw_clusters(right_image, right_clusters, scale_x, scale_y) # Combine the images side by side combined_width = left_image.width * 2 combined_height = left_image.height combined_image = Image.new("RGB", (combined_width, combined_height)) combined_image.paste(left_image, (0, 0)) combined_image.paste(right_image, (left_image.width, 0)) if show: combined_image.show() else: out_path: Path = ( Path(settings.debug.debug_output_path) / f"debug_{conv_res.input.file.stem}" ) out_path.mkdir(parents=True, exist_ok=True) out_file = out_path / f"{mode_prefix}_layout_page_{page.page_no:05}.png" combined_image.save(str(out_file), format="png") def predict_layout( self, conv_res: ConversionResult, pages: Sequence[Page], ) -> Sequence[LayoutPrediction]: # Convert to list to ensure predictable iteration pages = list(pages) # Separate valid and invalid pages valid_pages = [] valid_page_images: List[Union[Image.Image, np.ndarray]] = [] for page in pages: assert page._backend is not None if not page._backend.is_valid(): continue assert page.size is not None page_image = page.get_image(scale=1.0) assert page_image is not None valid_pages.append(page) valid_page_images.append(page_image) # Process all valid pages with batch prediction batch_predictions = [] if valid_page_images: with TimeRecorder(conv_res, "layout"): batch_predictions = self.layout_predictor.predict_batch( # type: ignore[attr-defined] valid_page_images ) # Process each page with its predictions layout_predictions: list[LayoutPrediction] = [] valid_page_idx = 0 for page in pages: assert page._backend is not None if not page._backend.is_valid(): existing_prediction = page.predictions.layout or LayoutPrediction() page.predictions.layout = existing_prediction layout_predictions.append(existing_prediction) continue page_predictions = batch_predictions[valid_page_idx] valid_page_idx += 1 clusters = [] for ix, pred_item in enumerate(page_predictions): label = DocItemLabel( pred_item["label"].lower().replace(" ", "_").replace("-", "_") ) # Temporary, until docling-ibm-model uses docling-core types cluster = Cluster( id=ix, label=label, confidence=pred_item["confidence"], bbox=BoundingBox.model_validate(pred_item), cells=[], ) clusters.append(cluster) if settings.debug.visualize_raw_layout: self.draw_clusters_and_cells_side_by_side( conv_res, page, clusters, mode_prefix="raw" ) # Apply postprocessing processed_clusters, processed_cells = LayoutPostprocessor( page, clusters, self.options ).postprocess() # Note: LayoutPostprocessor updates page.cells and page.parsed_page internally with warnings.catch_warnings(): warnings.filterwarnings( "ignore", "Mean of empty slice|invalid value encountered in scalar divide", RuntimeWarning, "numpy", ) conv_res.confidence.pages[page.page_no].layout_score = float( np.mean([c.confidence for c in processed_clusters]) ) conv_res.confidence.pages[page.page_no].ocr_score = float( np.mean([c.confidence for c in processed_cells if c.from_ocr]) ) prediction = LayoutPrediction(clusters=processed_clusters) page.predictions.layout = prediction if settings.debug.visualize_layout: self.draw_clusters_and_cells_side_by_side( conv_res, page, processed_clusters, mode_prefix="postprocessed" ) layout_predictions.append(prediction) return layout_predictions
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/picture_description_vlm_model.py
docling/models/picture_description_vlm_model.py
import sys import threading from collections.abc import Iterable from pathlib import Path from typing import Optional, Type, Union from PIL import Image from transformers import AutoModelForImageTextToText from docling.datamodel.accelerator_options import AcceleratorOptions from docling.datamodel.pipeline_options import ( PictureDescriptionBaseOptions, PictureDescriptionVlmOptions, ) from docling.models.picture_description_base_model import PictureDescriptionBaseModel from docling.models.utils.hf_model_download import ( HuggingFaceModelDownloadMixin, ) from docling.utils.accelerator_utils import decide_device # Global lock for model initialization to prevent threading issues _model_init_lock = threading.Lock() class PictureDescriptionVlmModel( PictureDescriptionBaseModel, HuggingFaceModelDownloadMixin ): @classmethod def get_options_type(cls) -> Type[PictureDescriptionBaseOptions]: return PictureDescriptionVlmOptions def __init__( self, enabled: bool, enable_remote_services: bool, artifacts_path: Optional[Union[Path, str]], options: PictureDescriptionVlmOptions, accelerator_options: AcceleratorOptions, ): super().__init__( enabled=enabled, enable_remote_services=enable_remote_services, artifacts_path=artifacts_path, options=options, accelerator_options=accelerator_options, ) self.options: PictureDescriptionVlmOptions if self.enabled: if artifacts_path is None: artifacts_path = self.download_models(repo_id=self.options.repo_id) else: artifacts_path = Path(artifacts_path) / self.options.repo_cache_folder self.device = decide_device(accelerator_options.device) try: import torch from transformers import AutoModelForVision2Seq, AutoProcessor except ImportError: raise ImportError( "transformers >=4.46 is not installed. Please install Docling with the required extras `pip install docling[vlm]`." ) # Initialize processor and model with _model_init_lock: self.processor = AutoProcessor.from_pretrained(artifacts_path) self.model = AutoModelForImageTextToText.from_pretrained( artifacts_path, device_map=self.device, dtype=torch.bfloat16, _attn_implementation=( "flash_attention_2" if self.device.startswith("cuda") and accelerator_options.cuda_use_flash_attention2 else "sdpa" ), ) if sys.version_info < (3, 14): self.model = torch.compile(self.model) # type: ignore else: self.model.eval() self.provenance = f"{self.options.repo_id}" def _annotate_images(self, images: Iterable[Image.Image]) -> Iterable[str]: from transformers import GenerationConfig # Create input messages messages = [ { "role": "user", "content": [ {"type": "image"}, {"type": "text", "text": self.options.prompt}, ], }, ] # TODO: do batch generation for image in images: # Prepare inputs prompt = self.processor.apply_chat_template( messages, add_generation_prompt=True ) inputs = self.processor(text=prompt, images=[image], return_tensors="pt") inputs = inputs.to(self.device) # Generate outputs generated_ids = self.model.generate( **inputs, generation_config=GenerationConfig(**self.options.generation_config), ) generated_texts = self.processor.batch_decode( generated_ids[:, inputs["input_ids"].shape[1] :], skip_special_tokens=True, ) yield generated_texts[0].strip()
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/easyocr_model.py
docling/models/easyocr_model.py
import logging import warnings import zipfile from collections.abc import Iterable from pathlib import Path from typing import List, Optional, Type import numpy from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.doc.page import BoundingRectangle, TextCell from docling.datamodel.accelerator_options import AcceleratorDevice, AcceleratorOptions from docling.datamodel.base_models import Page from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options import ( EasyOcrOptions, OcrOptions, ) from docling.datamodel.settings import settings from docling.models.base_ocr_model import BaseOcrModel from docling.utils.accelerator_utils import decide_device from docling.utils.profiling import TimeRecorder from docling.utils.utils import download_url_with_progress _log = logging.getLogger(__name__) class EasyOcrModel(BaseOcrModel): _model_repo_folder = "EasyOcr" def __init__( self, enabled: bool, artifacts_path: Optional[Path], options: EasyOcrOptions, accelerator_options: AcceleratorOptions, ): super().__init__( enabled=enabled, artifacts_path=artifacts_path, options=options, accelerator_options=accelerator_options, ) self.options: EasyOcrOptions self.scale = 3 # multiplier for 72 dpi == 216 dpi. if self.enabled: try: import easyocr except ImportError: raise ImportError( "EasyOCR is not installed. Please install it via `pip install easyocr` to use this OCR engine. " "Alternatively, Docling has support for other OCR engines. See the documentation." ) if self.options.use_gpu is None: device = decide_device(accelerator_options.device) # Enable easyocr GPU if running on CUDA, MPS use_gpu = any( device.startswith(x) for x in [ AcceleratorDevice.CUDA.value, AcceleratorDevice.MPS.value, ] ) else: warnings.warn( "Deprecated field. Better to set the `accelerator_options.device` in `pipeline_options`. " "When `use_gpu and accelerator_options.device == AcceleratorDevice.CUDA` the GPU is used " "to run EasyOCR. Otherwise, EasyOCR runs in CPU." ) use_gpu = self.options.use_gpu download_enabled = self.options.download_enabled model_storage_directory = self.options.model_storage_directory if artifacts_path is not None and model_storage_directory is None: download_enabled = False model_storage_directory = str(artifacts_path / self._model_repo_folder) 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( detection_models: List[str] = ["craft"], recognition_models: List[str] = ["english_g2", "latin_g2"], local_dir: Optional[Path] = None, force: bool = False, progress: bool = False, ) -> Path: # Models are located in https://github.com/JaidedAI/EasyOCR/blob/master/easyocr/config.py from easyocr.config import ( detection_models as det_models_dict, recognition_models as rec_models_dict, ) if local_dir is None: local_dir = settings.cache_dir / "models" / EasyOcrModel._model_repo_folder local_dir.mkdir(parents=True, exist_ok=True) # Collect models to download download_list = [] for model_name in detection_models: if model_name in det_models_dict: download_list.append(det_models_dict[model_name]) for model_name in recognition_models: if model_name in rec_models_dict["gen2"]: download_list.append(rec_models_dict["gen2"][model_name]) # Download models for model_details in download_list: buf = download_url_with_progress(model_details["url"], progress=progress) with zipfile.ZipFile(buf, "r") as zip_ref: zip_ref.extractall(local_dir) return local_dir def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: if not self.enabled: yield from page_batch return for page in page_batch: assert page._backend is not None if not page._backend.is_valid(): yield page else: with TimeRecorder(conv_res, "ocr"): ocr_rects = self.get_ocr_rects(page) all_ocr_cells = [] for ocr_rect in ocr_rects: # Skip zero area boxes if ocr_rect.area() == 0: continue high_res_image = page._backend.get_page_image( scale=self.scale, cropbox=ocr_rect ) im = numpy.array(high_res_image) 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 cells = [ TextCell( index=ix, text=line[1], orig=line[1], from_ocr=True, confidence=line[2], rect=BoundingRectangle.from_bounding_box( BoundingBox.from_tuple( coord=( (line[0][0][0] / self.scale) + ocr_rect.l, (line[0][0][1] / self.scale) + ocr_rect.t, (line[0][2][0] / self.scale) + ocr_rect.l, (line[0][2][1] / self.scale) + ocr_rect.t, ), origin=CoordOrigin.TOPLEFT, ) ), ) for ix, line in enumerate(result) if line[2] >= self.options.confidence_threshold ] all_ocr_cells.extend(cells) # Post-process the cells self.post_process_cells(all_ocr_cells, page) # DEBUG code: if settings.debug.visualize_ocr: self.draw_ocr_rects_and_cells(conv_res, page, ocr_rects) yield page @classmethod def get_options_type(cls) -> Type[OcrOptions]: return EasyOcrOptions
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/__init__.py
docling/models/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/readingorder_model.py
docling/models/readingorder_model.py
from pathlib import Path from docling_core.types.doc import ( DocItemLabel, DoclingDocument, DocumentOrigin, GroupLabel, NodeItem, ProvenanceItem, RefItem, RichTableCell, TableData, ) from docling_core.types.doc.document import ContentLayer from docling_ibm_models.list_item_normalizer.list_marker_processor import ( ListItemMarkerProcessor, ) from docling_ibm_models.reading_order.reading_order_rb import ( PageElement as ReadingOrderPageElement, ReadingOrderPredictor, ) from pydantic import BaseModel, ConfigDict from docling.datamodel.base_models import ( BasePageElement, Cluster, ContainerElement, FigureElement, Table, TextElement, ) from docling.datamodel.document import ConversionResult from docling.utils.profiling import ProfilingScope, TimeRecorder class ReadingOrderOptions(BaseModel): model_config = ConfigDict(protected_namespaces=()) model_names: str = "" # e.g. "language;term;reference" class ReadingOrderModel: def __init__(self, options: ReadingOrderOptions): self.options = options self.ro_model = ReadingOrderPredictor() self.list_item_processor = ListItemMarkerProcessor() def _assembled_to_readingorder_elements( self, conv_res: ConversionResult ) -> list[ReadingOrderPageElement]: elements: list[ReadingOrderPageElement] = [] page_no_to_pages = {p.page_no: p for p in conv_res.pages} for element in conv_res.assembled.elements: page_height = page_no_to_pages[element.page_no].size.height # type: ignore bbox = element.cluster.bbox.to_bottom_left_origin(page_height) text = element.text or "" elements.append( ReadingOrderPageElement( cid=len(elements), ref=RefItem(cref=f"#/{element.page_no}/{element.cluster.id}"), text=text, page_no=element.page_no, page_size=page_no_to_pages[element.page_no].size, label=element.label, l=bbox.l, r=bbox.r, b=bbox.b, t=bbox.t, coord_origin=bbox.coord_origin, ) ) return elements def _add_child_elements( self, element: BasePageElement, doc_item: NodeItem, doc: DoclingDocument ): child: Cluster for child in element.cluster.children: c_label = child.label c_bbox = child.bbox.to_bottom_left_origin( doc.pages[element.page_no + 1].size.height ) c_text = " ".join( [ cell.text.replace("\x02", "-").strip() for cell in child.cells if len(cell.text.strip()) > 0 ] ) c_prov = ProvenanceItem( page_no=element.page_no + 1, charspan=(0, len(c_text)), bbox=c_bbox ) if c_label == DocItemLabel.LIST_ITEM: # TODO: Infer if this is a numbered or a bullet list item l_item = doc.add_list_item(parent=doc_item, text=c_text, prov=c_prov) self.list_item_processor.process_list_item(l_item) elif c_label == DocItemLabel.SECTION_HEADER: doc.add_heading(parent=doc_item, text=c_text, prov=c_prov) else: doc.add_text(parent=doc_item, label=c_label, text=c_text, prov=c_prov) def _create_rich_cell_group( self, element: BasePageElement, doc: DoclingDocument, table_item: NodeItem ) -> RefItem: """Create a group containing all child elements for a rich table cell.""" group_name = f"rich_cell_group_{len(doc.tables)}_0_0" group_element = doc.add_group( label=GroupLabel.UNSPECIFIED, name=group_name, parent=table_item, ) # Add all child elements to the group self._add_child_elements(element, group_element, doc) return group_element.get_ref() def _readingorder_elements_to_docling_doc( self, conv_res: ConversionResult, ro_elements: list[ReadingOrderPageElement], el_to_captions_mapping: dict[int, list[int]], el_to_footnotes_mapping: dict[int, list[int]], el_merges_mapping: dict[int, list[int]], ) -> DoclingDocument: id_to_elem = { RefItem(cref=f"#/{elem.page_no}/{elem.cluster.id}").cref: elem for elem in conv_res.assembled.elements } cid_to_rels = {rel.cid: rel for rel in ro_elements} origin = DocumentOrigin( mimetype="application/pdf", filename=conv_res.input.file.name, binary_hash=conv_res.input.document_hash, ) doc_name = Path(origin.filename).stem out_doc: DoclingDocument = DoclingDocument(name=doc_name, origin=origin) for page in conv_res.pages: page_no = page.page_no + 1 size = page.size assert size is not None, "Page size is not initialized." out_doc.add_page(page_no=page_no, size=size) current_list = None skippable_cids = { cid for mapping in ( el_to_captions_mapping, el_to_footnotes_mapping, el_merges_mapping, ) for lst in mapping.values() for cid in lst } page_no_to_pages = {p.page_no: p for p in conv_res.pages} for rel in ro_elements: if rel.cid in skippable_cids: continue element = id_to_elem[rel.ref.cref] page_height = page_no_to_pages[element.page_no].size.height # type: ignore if isinstance(element, TextElement): if element.label == DocItemLabel.CODE: cap_text = element.text prov = ProvenanceItem( page_no=element.page_no + 1, charspan=(0, len(cap_text)), bbox=element.cluster.bbox.to_bottom_left_origin(page_height), ) code_item = out_doc.add_code(text=cap_text, prov=prov) if rel.cid in el_to_captions_mapping.keys(): for caption_cid in el_to_captions_mapping[rel.cid]: caption_elem = id_to_elem[cid_to_rels[caption_cid].ref.cref] new_cap_item = self._add_caption_or_footnote( caption_elem, out_doc, code_item, page_height ) code_item.captions.append(new_cap_item.get_ref()) if rel.cid in el_to_footnotes_mapping.keys(): for footnote_cid in el_to_footnotes_mapping[rel.cid]: footnote_elem = id_to_elem[ cid_to_rels[footnote_cid].ref.cref ] new_footnote_item = self._add_caption_or_footnote( footnote_elem, out_doc, code_item, page_height ) code_item.footnotes.append(new_footnote_item.get_ref()) else: new_item, current_list = self._handle_text_element( element, out_doc, current_list, page_height ) if rel.cid in el_merges_mapping.keys(): for merged_cid in el_merges_mapping[rel.cid]: merged_elem = id_to_elem[cid_to_rels[merged_cid].ref.cref] self._merge_elements( element, merged_elem, new_item, page_height ) elif isinstance(element, Table): # Check if table has no structure prediction if element.num_rows == 0 and element.num_cols == 0: # Only create 1x1 table if there are children to put in it if element.cluster.children: # Create minimal 1x1 table with rich cell containing all children tbl_data = TableData(num_rows=1, num_cols=1, table_cells=[]) else: # Create empty table with no structure tbl_data = TableData(num_rows=0, num_cols=0, table_cells=[]) else: tbl_data = TableData( num_rows=element.num_rows, num_cols=element.num_cols, table_cells=element.table_cells, ) prov = ProvenanceItem( page_no=element.page_no + 1, charspan=(0, 0), bbox=element.cluster.bbox.to_bottom_left_origin(page_height), ) tbl = out_doc.add_table( data=tbl_data, prov=prov, label=element.cluster.label ) if rel.cid in el_to_captions_mapping.keys(): for caption_cid in el_to_captions_mapping[rel.cid]: caption_elem = id_to_elem[cid_to_rels[caption_cid].ref.cref] new_cap_item = self._add_caption_or_footnote( caption_elem, out_doc, tbl, page_height ) tbl.captions.append(new_cap_item.get_ref()) if rel.cid in el_to_footnotes_mapping.keys(): for footnote_cid in el_to_footnotes_mapping[rel.cid]: footnote_elem = id_to_elem[cid_to_rels[footnote_cid].ref.cref] new_footnote_item = self._add_caption_or_footnote( footnote_elem, out_doc, tbl, page_height ) tbl.footnotes.append(new_footnote_item.get_ref()) # Handle case where table has no structure prediction but has children if ( element.num_rows == 0 and element.num_cols == 0 and element.cluster.children ): # Create rich cell containing all child elements rich_cell_ref = self._create_rich_cell_group(element, out_doc, tbl) # Create rich table cell spanning the entire 1x1 table rich_cell = RichTableCell( text="", # Empty text since content is in the group row_span=1, col_span=1, start_row_offset_idx=0, end_row_offset_idx=1, start_col_offset_idx=0, end_col_offset_idx=1, column_header=False, row_header=False, ref=rich_cell_ref, ) out_doc.add_table_cell(table_item=tbl, cell=rich_cell) # TODO: Consider adding children of Table. elif isinstance(element, FigureElement): cap_text = "" prov = ProvenanceItem( page_no=element.page_no + 1, charspan=(0, len(cap_text)), bbox=element.cluster.bbox.to_bottom_left_origin(page_height), ) pic = out_doc.add_picture(prov=prov) if rel.cid in el_to_captions_mapping.keys(): for caption_cid in el_to_captions_mapping[rel.cid]: caption_elem = id_to_elem[cid_to_rels[caption_cid].ref.cref] new_cap_item = self._add_caption_or_footnote( caption_elem, out_doc, pic, page_height ) pic.captions.append(new_cap_item.get_ref()) if rel.cid in el_to_footnotes_mapping.keys(): for footnote_cid in el_to_footnotes_mapping[rel.cid]: footnote_elem = id_to_elem[cid_to_rels[footnote_cid].ref.cref] new_footnote_item = self._add_caption_or_footnote( footnote_elem, out_doc, pic, page_height ) pic.footnotes.append(new_footnote_item.get_ref()) self._add_child_elements(element, pic, out_doc) elif isinstance(element, ContainerElement): # Form, KV region label = element.label group_label = GroupLabel.UNSPECIFIED if label == DocItemLabel.FORM: group_label = GroupLabel.FORM_AREA elif label == DocItemLabel.KEY_VALUE_REGION: group_label = GroupLabel.KEY_VALUE_AREA container_el = out_doc.add_group(label=group_label) self._add_child_elements(element, container_el, out_doc) return out_doc def _add_caption_or_footnote(self, elem, out_doc, parent, page_height): assert isinstance(elem, TextElement) text = elem.text prov = ProvenanceItem( page_no=elem.page_no + 1, charspan=(0, len(text)), bbox=elem.cluster.bbox.to_bottom_left_origin(page_height), ) new_item = out_doc.add_text( label=elem.label, text=text, prov=prov, parent=parent ) return new_item def _handle_text_element(self, element, out_doc, current_list, page_height): cap_text = element.text prov = ProvenanceItem( page_no=element.page_no + 1, charspan=(0, len(cap_text)), bbox=element.cluster.bbox.to_bottom_left_origin(page_height), ) label = element.label if label == DocItemLabel.LIST_ITEM: if current_list is None: current_list = out_doc.add_group(label=GroupLabel.LIST, name="list") # TODO: Infer if this is a numbered or a bullet list item new_item = out_doc.add_list_item( text=cap_text, enumerated=False, prov=prov, parent=current_list ) self.list_item_processor.process_list_item(new_item) elif label == DocItemLabel.SECTION_HEADER: current_list = None new_item = out_doc.add_heading(text=cap_text, prov=prov) elif label == DocItemLabel.FORMULA: current_list = None new_item = out_doc.add_text( label=DocItemLabel.FORMULA, text="", orig=cap_text, prov=prov ) else: current_list = None content_layer = ContentLayer.BODY if element.label in [DocItemLabel.PAGE_HEADER, DocItemLabel.PAGE_FOOTER]: content_layer = ContentLayer.FURNITURE new_item = out_doc.add_text( label=element.label, text=cap_text, prov=prov, content_layer=content_layer, ) return new_item, current_list def _merge_elements(self, element, merged_elem, new_item, page_height): assert isinstance(merged_elem, type(element)), ( "Merged element must be of same type as element." ) assert merged_elem.label == new_item.label, ( "Labels of merged elements must match." ) prov = ProvenanceItem( page_no=merged_elem.page_no + 1, charspan=( len(new_item.text) + 1, len(new_item.text) + 1 + len(merged_elem.text), ), 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. new_item.prov.append(prov) def __call__(self, conv_res: ConversionResult) -> DoclingDocument: with TimeRecorder(conv_res, "reading_order", scope=ProfilingScope.DOCUMENT): page_elements = self._assembled_to_readingorder_elements(conv_res) # Apply reading order sorted_elements = self.ro_model.predict_reading_order( page_elements=page_elements ) el_to_captions_mapping = self.ro_model.predict_to_captions( sorted_elements=sorted_elements ) el_to_footnotes_mapping = self.ro_model.predict_to_footnotes( sorted_elements=sorted_elements ) el_merges_mapping = self.ro_model.predict_merges( sorted_elements=sorted_elements ) docling_doc: DoclingDocument = self._readingorder_elements_to_docling_doc( conv_res, sorted_elements, el_to_captions_mapping, el_to_footnotes_mapping, el_merges_mapping, ) return docling_doc
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/auto_ocr_model.py
docling/models/auto_ocr_model.py
import logging import sys from collections.abc import Iterable from pathlib import Path from typing import Optional, Type from docling.datamodel.accelerator_options import AcceleratorOptions from docling.datamodel.base_models import Page from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options import ( EasyOcrOptions, OcrAutoOptions, OcrMacOptions, OcrOptions, RapidOcrOptions, ) from docling.models.base_ocr_model import BaseOcrModel from docling.models.easyocr_model import EasyOcrModel from docling.models.ocr_mac_model import OcrMacModel from docling.models.rapid_ocr_model import RapidOcrModel _log = logging.getLogger(__name__) class OcrAutoModel(BaseOcrModel): def __init__( self, enabled: bool, artifacts_path: Optional[Path], options: OcrAutoOptions, accelerator_options: AcceleratorOptions, ): super().__init__( enabled=enabled, artifacts_path=artifacts_path, options=options, accelerator_options=accelerator_options, ) self.options: OcrAutoOptions self._engine: Optional[BaseOcrModel] = None if self.enabled: if "darwin" == sys.platform: try: from ocrmac import ocrmac self._engine = OcrMacModel( enabled=self.enabled, artifacts_path=artifacts_path, options=OcrMacOptions( bitmap_area_threshold=self.options.bitmap_area_threshold, force_full_page_ocr=self.options.force_full_page_ocr, ), accelerator_options=accelerator_options, ) _log.info("Auto OCR model selected ocrmac.") except ImportError: _log.info("ocrmac cannot be used because ocrmac is not installed.") if self._engine is None: try: import onnxruntime from rapidocr import EngineType, RapidOCR # type: ignore self._engine = RapidOcrModel( enabled=self.enabled, artifacts_path=artifacts_path, options=RapidOcrOptions( backend="onnxruntime", bitmap_area_threshold=self.options.bitmap_area_threshold, force_full_page_ocr=self.options.force_full_page_ocr, ), accelerator_options=accelerator_options, ) _log.info("Auto OCR model selected rapidocr with onnxruntime.") except ImportError: _log.info( "rapidocr cannot be used because onnxruntime is not installed." ) if self._engine is None: try: import easyocr self._engine = EasyOcrModel( enabled=self.enabled, artifacts_path=artifacts_path, options=EasyOcrOptions( bitmap_area_threshold=self.options.bitmap_area_threshold, force_full_page_ocr=self.options.force_full_page_ocr, ), accelerator_options=accelerator_options, ) _log.info("Auto OCR model selected easyocr.") except ImportError: _log.info("easyocr cannot be used because it is not installed.") if self._engine is None: try: import torch from rapidocr import EngineType, RapidOCR # type: ignore self._engine = RapidOcrModel( enabled=self.enabled, artifacts_path=artifacts_path, options=RapidOcrOptions( backend="torch", bitmap_area_threshold=self.options.bitmap_area_threshold, force_full_page_ocr=self.options.force_full_page_ocr, ), accelerator_options=accelerator_options, ) _log.info("Auto OCR model selected rapidocr with torch.") except ImportError: _log.info( "rapidocr cannot be used because rapidocr or torch is not installed." ) if self._engine is None: _log.warning("No OCR engine found. Please review the install details.") def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: if not self.enabled or self._engine is None: yield from page_batch return yield from self._engine(conv_res, page_batch) @classmethod def get_options_type(cls) -> Type[OcrOptions]: return OcrAutoOptions
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/picture_description_base_model.py
docling/models/picture_description_base_model.py
from abc import abstractmethod from collections.abc import Iterable from pathlib import Path from typing import List, Optional, Type, Union from docling_core.types.doc import ( DoclingDocument, NodeItem, PictureItem, ) from docling_core.types.doc.document import ( # TODO: move import to docling_core.types.doc PictureDescriptionData, ) from PIL import Image from docling.datamodel.accelerator_options import AcceleratorOptions from docling.datamodel.pipeline_options import ( PictureDescriptionBaseOptions, ) from docling.models.base_model import ( BaseItemAndImageEnrichmentModel, BaseModelWithOptions, ItemAndImageEnrichmentElement, ) class PictureDescriptionBaseModel( BaseItemAndImageEnrichmentModel, BaseModelWithOptions ): images_scale: float = 2.0 def __init__( self, *, enabled: bool, enable_remote_services: bool, artifacts_path: Optional[Union[Path, str]], options: PictureDescriptionBaseOptions, accelerator_options: AcceleratorOptions, ): self.enabled = enabled self.options = options self.provenance = "not-implemented" def is_processable(self, doc: DoclingDocument, element: NodeItem) -> bool: return self.enabled and isinstance(element, PictureItem) def _annotate_images(self, images: Iterable[Image.Image]) -> Iterable[str]: raise NotImplementedError def __call__( self, doc: DoclingDocument, element_batch: Iterable[ItemAndImageEnrichmentElement], ) -> Iterable[NodeItem]: if not self.enabled: for element in element_batch: yield element.item return images: List[Image.Image] = [] elements: List[PictureItem] = [] for el in element_batch: assert isinstance(el.item, PictureItem) describe_image = True # Don't describe the image if it's smaller than the threshold if len(el.item.prov) > 0: prov = el.item.prov[0] # PictureItems have at most a single provenance page = doc.pages.get(prov.page_no) if page is not None: page_area = page.size.width * page.size.height if page_area > 0: area_fraction = prov.bbox.area() / page_area if area_fraction < self.options.picture_area_threshold: describe_image = False if describe_image: elements.append(el.item) images.append(el.image) outputs = self._annotate_images(images) for item, output in zip(elements, outputs): item.annotations.append( PictureDescriptionData(text=output, provenance=self.provenance) ) yield item @classmethod @abstractmethod def get_options_type(cls) -> Type[PictureDescriptionBaseOptions]: pass
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/page_preprocessing_model.py
docling/models/page_preprocessing_model.py
import re import warnings from collections.abc import Iterable from pathlib import Path from typing import Literal, Optional import numpy as np from PIL import ImageDraw from pydantic import BaseModel from docling.datamodel.base_models import Page from docling.datamodel.document import ConversionResult from docling.datamodel.settings import settings from docling.models.base_model import BasePageModel from docling.utils.profiling import TimeRecorder class PagePreprocessingOptions(BaseModel): images_scale: Optional[float] skip_cell_extraction: bool = ( False # Skip text cell extraction for VLM-only processing ) class PagePreprocessingModel(BasePageModel): def __init__(self, options: PagePreprocessingOptions): self.options = options # Pre-compiled regex patterns for efficiency self.GLYPH_RE = re.compile(r"GLYPH<[0-9A-Fa-f]+>") self.SLASH_G_RE = re.compile(r"(?:/G\d+){2,}") self.FRAG_RE = re.compile(r"\b[A-Za-z](?:/[a-z]{1,3}\.[a-z]{1,3}){2,}\b") self.SLASH_NUMBER_GARBAGE_RE = re.compile( r"(?:/\w+\s*){2,}" ) # Two or more "/token " sequences def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: for page in page_batch: assert page._backend is not None if not page._backend.is_valid(): yield page else: with TimeRecorder(conv_res, "page_parse"): page = self._populate_page_images(page) if not self.options.skip_cell_extraction: page = self._parse_page_cells(conv_res, page) yield page # Generate the page image and store it in the page object def _populate_page_images(self, page: Page) -> Page: # default scale page.get_image( scale=1.0 ) # puts the page image on the image cache at default scale images_scale = self.options.images_scale # user requested scales if images_scale is not None: page._default_image_scale = images_scale page.get_image( scale=images_scale ) # this will trigger storing the image in the internal cache return page # Extract and populate the page cells and store it in the page object def _parse_page_cells(self, conv_res: ConversionResult, page: Page) -> Page: assert page._backend is not None 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 = [] for c in page.cells: score = self.rate_text_quality(c.text) text_scores.append(score) with warnings.catch_warnings(): warnings.filterwarnings( "ignore", "Mean of empty slice", RuntimeWarning, "numpy" ) conv_res.confidence.pages[page.page_no].parse_score = float( np.nanquantile( text_scores, q=0.10 ) # To emphasise problems in the parse_score, we take the 10% percentile score of all text cells. ) # DEBUG code: def draw_text_boxes(image, cells, show: bool = False): draw = ImageDraw.Draw(image.copy()) for c in cells: x0, y0, x1, y1 = ( c.to_bounding_box().l, c.to_bounding_box().t, c.to_bounding_box().r, c.to_bounding_box().b, ) draw.rectangle([(x0, y0), (x1, y1)], outline="red") if show: image.show() else: out_path: Path = ( Path(settings.debug.debug_output_path) / f"debug_{conv_res.input.file.stem}" ) out_path.mkdir(parents=True, exist_ok=True) out_file = out_path / f"cells_page_{page.page_no:05}.png" image.save(str(out_file), format="png") if settings.debug.visualize_cells: draw_text_boxes(page.get_image(scale=1.0), page.cells) return page def rate_text_quality(self, text: str) -> float: # Hard errors: if any of these patterns are found, return 0.0 immediately. blacklist_chars = ["�"] if ( any(text.find(c) >= 0 for c in blacklist_chars) or self.GLYPH_RE.search(text) or self.SLASH_G_RE.search(text) or self.SLASH_NUMBER_GARBAGE_RE.match( text ) # Check if text is mostly slash-number pattern ): return 0.0 penalty = 0.0 # Apply a penalty only if the fragmented words pattern occurs at least three times. frag_matches = self.FRAG_RE.findall(text) if len(frag_matches) >= 3: penalty += 0.1 * len(frag_matches) # Additional heuristic: if the average token length is below 2, add a penalty. # tokens = text.split() # if tokens and (sum(map(len, tokens)) / len(tokens)) < 2: # penalty += 0.2 return max(1.0 - penalty, 0.0)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/tesseract_ocr_cli_model.py
docling/models/tesseract_ocr_cli_model.py
import csv import io import logging import os import subprocess import tempfile from collections.abc import Iterable from pathlib import Path from subprocess import DEVNULL, PIPE, Popen from typing import List, Optional, Tuple, Type import pandas as pd from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.doc.page import TextCell from docling.datamodel.accelerator_options import AcceleratorOptions from docling.datamodel.base_models import Page from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options import ( OcrOptions, TesseractCliOcrOptions, ) from docling.datamodel.settings import settings from docling.models.base_ocr_model import BaseOcrModel from docling.utils.ocr_utils import ( map_tesseract_script, parse_tesseract_orientation, tesseract_box_to_bounding_rectangle, ) from docling.utils.profiling import TimeRecorder _log = logging.getLogger(__name__) class TesseractOcrCliModel(BaseOcrModel): def __init__( self, enabled: bool, artifacts_path: Optional[Path], options: TesseractCliOcrOptions, accelerator_options: AcceleratorOptions, ): super().__init__( enabled=enabled, artifacts_path=artifacts_path, options=options, accelerator_options=accelerator_options, ) self.options: TesseractCliOcrOptions self.scale = 3 # multiplier for 72 dpi == 216 dpi. self._name: Optional[str] = None self._version: Optional[str] = None self._tesseract_languages: Optional[List[str]] = None self._script_prefix: Optional[str] = None self._is_auto: bool = "auto" in self.options.lang if self.enabled: try: self._get_name_and_version() self._set_languages_and_prefix() except Exception as exc: raise RuntimeError( f"Tesseract is not available, aborting: {exc} " "Install tesseract on your system and the tesseract binary is discoverable. " "The actual command for Tesseract can be specified in `pipeline_options.ocr_options.tesseract_cmd='tesseract'`. " "Alternatively, Docling has support for other OCR engines. See the documentation." ) def _get_name_and_version(self) -> Tuple[str, str]: if self._name is not None and self._version is not None: return self._name, self._version # type: ignore cmd = [self.options.tesseract_cmd, "--version"] proc = Popen(cmd, stdout=PIPE, stderr=PIPE) stdout, stderr = proc.communicate() proc.wait() # HACK: Windows versions of Tesseract output the version to stdout, Linux versions # to stderr, so check both. version_line = ( (stdout.decode("utf8").strip() or stderr.decode("utf8").strip()) .split("\n")[0] .strip() ) # If everything else fails... if not version_line: version_line = "tesseract XXX" name, version = version_line.split(" ") self._name = name self._version = version return name, version def _run_tesseract(self, ifilename: str, osd: Optional[pd.DataFrame]): r""" Run tesseract CLI """ cmd = [self.options.tesseract_cmd] if self._is_auto and osd is not None: lang = self._parse_language(osd) if lang is not None: cmd.append("-l") cmd.append(lang) elif self.options.lang is not None and len(self.options.lang) > 0: cmd.append("-l") cmd.append("+".join(self.options.lang)) if self.options.path is not None: cmd.append("--tessdata-dir") cmd.append(self.options.path) # Add PSM option if specified in the configuration if self.options.psm is not None: cmd.extend(["--psm", str(self.options.psm)]) cmd += [ifilename, "stdout", "tsv"] _log.info("command: {}".format(" ".join(cmd))) output = subprocess.run(cmd, stdout=PIPE, stderr=DEVNULL, check=True) # _log.info(output) # Decode the byte string to a regular string decoded_data = output.stdout.decode("utf-8") # _log.info(decoded_data) # Read the TSV file generated by Tesseract df_result = pd.read_csv( io.StringIO(decoded_data), quoting=csv.QUOTE_NONE, sep="\t" ) # Display the dataframe (optional) # _log.info("df: ", df.head()) # Filter rows that contain actual text (ignore header or empty rows) df_filtered = df_result[ df_result["text"].notna() & (df_result["text"].apply(str).str.strip() != "") ] return df_filtered def _perform_osd(self, ifilename: str) -> pd.DataFrame: r""" Run tesseract in PSM 0 mode to detect the language """ cmd = [self.options.tesseract_cmd] cmd.extend(["--psm", "0", "-l", "osd", ifilename, "stdout"]) _log.info("command: {}".format(" ".join(cmd))) output = subprocess.run(cmd, capture_output=True, check=True) decoded_data = output.stdout.decode("utf-8") df_detected = pd.read_csv( io.StringIO(decoded_data), sep=":", header=None, names=["key", "value"] ) return df_detected def _parse_language(self, df_osd: pd.DataFrame) -> Optional[str]: assert self._tesseract_languages is not None scripts = df_osd.loc[df_osd["key"] == "Script"].value.tolist() if len(scripts) == 0: _log.warning("Tesseract cannot detect the script of the page") return None script = map_tesseract_script(scripts[0].strip()) lang = f"{self._script_prefix}{script}" # Check if the detected language has been installed if lang not in self._tesseract_languages: msg = f"Tesseract detected the script '{script}' and language '{lang}'." msg += " However this language is not installed in your system and will be ignored." _log.warning(msg) return None _log.debug( f"Using tesseract model for the detected script '{script}' and language '{lang}'" ) return lang def _set_languages_and_prefix(self): r""" Read and set the languages installed in tesseract and decide the script prefix """ # Get all languages cmd = [self.options.tesseract_cmd] cmd.append("--list-langs") _log.info("command: {}".format(" ".join(cmd))) output = subprocess.run(cmd, stdout=PIPE, stderr=DEVNULL, check=True) decoded_data = output.stdout.decode("utf-8") df_list = pd.read_csv(io.StringIO(decoded_data), header=None) self._tesseract_languages = df_list[0].tolist()[1:] # Decide the script prefix if any(lang.startswith("script/") for lang in self._tesseract_languages): script_prefix = "script/" else: script_prefix = "" self._script_prefix = script_prefix def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: if not self.enabled: yield from page_batch return for page_i, page in enumerate(page_batch): assert page._backend is not None if not page._backend.is_valid(): yield page else: with TimeRecorder(conv_res, "ocr"): ocr_rects = self.get_ocr_rects(page) all_ocr_cells = [] for ocr_rect_i, ocr_rect in enumerate(ocr_rects): # Skip zero area boxes if ocr_rect.area() == 0: continue high_res_image = page._backend.get_page_image( scale=self.scale, cropbox=ocr_rect ) try: with tempfile.NamedTemporaryFile( suffix=".png", mode="w+b", delete=False ) as image_file: 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) except subprocess.CalledProcessError as exc: _log.error( "OSD failed (doc %s, page: %s, " "OCR rectangle: %s, processed image file %s):\n %s", conv_res.input.file, page_i, ocr_rect_i, image_file, exc.stderr, ) # Skipping if OSD fail when in auto mode, otherwise proceed # to OCR in the hope OCR will succeed while OSD failed if self._is_auto: continue if doc_orientation != 0: high_res_image = high_res_image.rotate( -doc_orientation, expand=True ) high_res_image.save(fname) try: df_result = self._run_tesseract(fname, df_osd) except subprocess.CalledProcessError as exc: _log.error( "tesseract OCR failed (doc %s, page: %s, " "OCR rectangle: %s, processed image file %s):\n %s", conv_res.input.file, page_i, ocr_rect_i, image_file, exc.stderr, ) continue finally: if os.path.exists(fname): os.remove(fname) # _log.info(df_result) # Print relevant columns (bounding box and text) for ix, row in df_result.iterrows(): text = row["text"] conf = row["conf"] left, top = float(row["left"]), float(row["top"]) right = left + float(row["width"]) bottom = top + row["height"] bbox = BoundingBox( l=left, t=top, r=right, b=bottom, coord_origin=CoordOrigin.TOPLEFT, ) rect = tesseract_box_to_bounding_rectangle( bbox, original_offset=ocr_rect, scale=self.scale, orientation=doc_orientation, im_size=high_res_image.size, ) cell = TextCell( index=ix, text=str(text), orig=str(text), from_ocr=True, confidence=conf / 100.0, rect=rect, ) all_ocr_cells.append(cell) # Post-process the cells self.post_process_cells(all_ocr_cells, page) # DEBUG code: if settings.debug.visualize_ocr: self.draw_ocr_rects_and_cells(conv_res, page, ocr_rects) yield page @classmethod def get_options_type(cls) -> Type[OcrOptions]: return TesseractCliOcrOptions def _parse_orientation(df_osd: pd.DataFrame) -> int: # For strictly optimal performance with invariant dataframe format: mask = df_osd["key"].to_numpy() == "Orientation in degrees" orientation_val = df_osd["value"].to_numpy()[mask][0] orientation = parse_tesseract_orientation(orientation_val.strip()) return orientation
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/ocr_mac_model.py
docling/models/ocr_mac_model.py
import logging import sys import tempfile from collections.abc import Iterable from pathlib import Path from typing import Optional, Type from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.doc.page import BoundingRectangle, TextCell from docling.datamodel.accelerator_options import AcceleratorOptions from docling.datamodel.base_models import Page from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options import ( OcrMacOptions, OcrOptions, ) from docling.datamodel.settings import settings from docling.models.base_ocr_model import BaseOcrModel from docling.utils.profiling import TimeRecorder _log = logging.getLogger(__name__) class OcrMacModel(BaseOcrModel): def __init__( self, enabled: bool, artifacts_path: Optional[Path], options: OcrMacOptions, accelerator_options: AcceleratorOptions, ): super().__init__( enabled=enabled, artifacts_path=artifacts_path, options=options, accelerator_options=accelerator_options, ) self.options: OcrMacOptions self.scale = 3 # multiplier for 72 dpi == 216 dpi. if self.enabled: if "darwin" != sys.platform: raise RuntimeError("OcrMac is only supported on Mac.") install_errmsg = ( "ocrmac is not correctly installed. " "Please install it via `pip install ocrmac` to use this OCR engine. " "Alternatively, Docling has support for other OCR engines. See the documentation: " "https://docling-project.github.io/docling/installation/" ) try: from ocrmac import ocrmac except ImportError: raise ImportError(install_errmsg) self.reader_RIL = ocrmac.OCR def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: if not self.enabled: yield from page_batch return for page in page_batch: assert page._backend is not None if not page._backend.is_valid(): yield page else: with TimeRecorder(conv_res, "ocr"): ocr_rects = self.get_ocr_rects(page) all_ocr_cells = [] for ocr_rect in ocr_rects: # Skip zero area boxes if ocr_rect.area() == 0: continue high_res_image = page._backend.get_page_image( scale=self.scale, cropbox=ocr_rect ) with tempfile.NamedTemporaryFile( suffix=".png", mode="w" ) as image_file: fname = image_file.name high_res_image.save(fname) boxes = self.reader_RIL( fname, recognition_level=self.options.recognition, framework=self.options.framework, language_preference=self.options.lang, ).recognize() im_width, im_height = high_res_image.size cells = [] for ix, (text, confidence, box) in enumerate(boxes): x = float(box[0]) y = float(box[1]) w = float(box[2]) h = float(box[3]) x1 = x * im_width y2 = (1 - y) * im_height x2 = x1 + w * im_width y1 = y2 - h * im_height left = x1 / self.scale top = y1 / self.scale right = x2 / self.scale bottom = y2 / self.scale cells.append( TextCell( index=ix, text=text, orig=text, from_ocr=True, confidence=confidence, rect=BoundingRectangle.from_bounding_box( BoundingBox.from_tuple( coord=(left, top, right, bottom), origin=CoordOrigin.TOPLEFT, ) ), ) ) # del high_res_image all_ocr_cells.extend(cells) # Post-process the cells self.post_process_cells(all_ocr_cells, page) # DEBUG code: if settings.debug.visualize_ocr: self.draw_ocr_rects_and_cells(conv_res, page, ocr_rects) yield page @classmethod def get_options_type(cls) -> Type[OcrOptions]: return OcrMacOptions
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/factories/ocr_factory.py
docling/models/factories/ocr_factory.py
import logging from docling.models.base_ocr_model import BaseOcrModel from docling.models.factories.base_factory import BaseFactory logger = logging.getLogger(__name__) class OcrFactory(BaseFactory[BaseOcrModel]): def __init__(self, *args, **kwargs): super().__init__("ocr_engines", *args, **kwargs)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/factories/picture_description_factory.py
docling/models/factories/picture_description_factory.py
import logging from docling.models.factories.base_factory import BaseFactory from docling.models.picture_description_base_model import PictureDescriptionBaseModel logger = logging.getLogger(__name__) class PictureDescriptionFactory(BaseFactory[PictureDescriptionBaseModel]): def __init__(self, *args, **kwargs): super().__init__("picture_description", *args, **kwargs)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/factories/table_factory.py
docling/models/factories/table_factory.py
from docling.models.base_table_model import BaseTableStructureModel from docling.models.factories.base_factory import BaseFactory class TableStructureFactory(BaseFactory[BaseTableStructureModel]): def __init__(self, *args, **kwargs): super().__init__("table_structure_engines", *args, **kwargs)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/factories/__init__.py
docling/models/factories/__init__.py
import logging from functools import lru_cache from docling.models.factories.layout_factory import LayoutFactory from docling.models.factories.ocr_factory import OcrFactory from docling.models.factories.picture_description_factory import ( PictureDescriptionFactory, ) from docling.models.factories.table_factory import TableStructureFactory logger = logging.getLogger(__name__) @lru_cache def get_ocr_factory(allow_external_plugins: bool = False) -> OcrFactory: factory = OcrFactory() factory.load_from_plugins(allow_external_plugins=allow_external_plugins) logger.info("Registered ocr engines: %r", factory.registered_kind) return factory @lru_cache def get_picture_description_factory( allow_external_plugins: bool = False, ) -> PictureDescriptionFactory: factory = PictureDescriptionFactory() factory.load_from_plugins(allow_external_plugins=allow_external_plugins) logger.info("Registered picture descriptions: %r", factory.registered_kind) return factory @lru_cache def get_layout_factory(allow_external_plugins: bool = False) -> LayoutFactory: factory = LayoutFactory() factory.load_from_plugins(allow_external_plugins=allow_external_plugins) logger.info("Registered layout engines: %r", factory.registered_kind) return factory @lru_cache def get_table_structure_factory( allow_external_plugins: bool = False, ) -> TableStructureFactory: factory = TableStructureFactory() factory.load_from_plugins(allow_external_plugins=allow_external_plugins) logger.info("Registered table structure engines: %r", factory.registered_kind) return factory
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/factories/layout_factory.py
docling/models/factories/layout_factory.py
from docling.models.base_layout_model import BaseLayoutModel from docling.models.factories.base_factory import BaseFactory class LayoutFactory(BaseFactory[BaseLayoutModel]): def __init__(self, *args, **kwargs): super().__init__("layout_engines", *args, **kwargs)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/factories/base_factory.py
docling/models/factories/base_factory.py
import enum import logging from abc import ABCMeta from typing import Generic, Optional, Type, TypeVar from pluggy import PluginManager from pydantic import BaseModel from docling.datamodel.pipeline_options import BaseOptions from docling.models.base_model import BaseModelWithOptions A = TypeVar("A", bound=BaseModelWithOptions) logger = logging.getLogger(__name__) class FactoryMeta(BaseModel): kind: str plugin_name: str module: str class BaseFactory(Generic[A], metaclass=ABCMeta): default_plugin_name = "docling" def __init__(self, plugin_attr_name: str, plugin_name=default_plugin_name): self.plugin_name = plugin_name self.plugin_attr_name = plugin_attr_name self._classes: dict[Type[BaseOptions], Type[A]] = {} self._meta: dict[Type[BaseOptions], FactoryMeta] = {} @property def registered_kind(self) -> list[str]: return [opt.kind for opt in self._classes.keys()] def get_enum(self) -> enum.Enum: return enum.Enum( self.plugin_attr_name + "_enum", names={kind: kind for kind in self.registered_kind}, type=str, module=__name__, ) @property def classes(self): return self._classes @property def registered_meta(self): return self._meta def create_instance(self, options: BaseOptions, **kwargs) -> A: try: _cls = self._classes[type(options)] return _cls(options=options, **kwargs) except KeyError: raise RuntimeError(self._err_msg_on_class_not_found(options.kind)) def create_options(self, kind: str, *args, **kwargs) -> BaseOptions: for opt_cls, _ in self._classes.items(): if opt_cls.kind == kind: return opt_cls(*args, **kwargs) raise RuntimeError(self._err_msg_on_class_not_found(kind)) def _err_msg_on_class_not_found(self, kind: str): msg = [] for opt, cls in self._classes.items(): msg.append(f"\t{opt.kind!r} => {cls!r}") msg_str = "\n".join(msg) return f"No class found with the name {kind!r}, known classes are:\n{msg_str}" def register(self, cls: Type[A], plugin_name: str, plugin_module_name: str): opt_type = cls.get_options_type() if opt_type in self._classes: raise ValueError( f"{opt_type.kind!r} already registered to class {self._classes[opt_type]!r}" ) self._classes[opt_type] = cls self._meta[opt_type] = FactoryMeta( kind=opt_type.kind, plugin_name=plugin_name, module=plugin_module_name ) def load_from_plugins( self, plugin_name: Optional[str] = None, allow_external_plugins: bool = False ): plugin_name = plugin_name or self.plugin_name plugin_manager = PluginManager(plugin_name) plugin_manager.load_setuptools_entrypoints(plugin_name) for plugin_name, plugin_module in plugin_manager.list_name_plugin(): plugin_module_name = str(plugin_module.__name__) # type: ignore if not allow_external_plugins and not plugin_module_name.startswith( "docling." ): logger.warning( f"The plugin {plugin_name} will not be loaded because Docling is being executed with allow_external_plugins=false." ) continue attr = getattr(plugin_module, self.plugin_attr_name, None) if callable(attr): logger.info("Loading plugin %r", plugin_name) config = attr() self.process_plugin(config, plugin_name, plugin_module_name) def process_plugin(self, config, plugin_name: str, plugin_module_name: str): for item in config[self.plugin_attr_name]: try: self.register(item, plugin_name, plugin_module_name) except ValueError: logger.warning("%r already registered", item)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/utils/__init__.py
docling/models/utils/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/utils/generation_utils.py
docling/models/utils/generation_utils.py
import logging import re import sys from abc import abstractmethod from typing import List from transformers import StoppingCriteria _log = logging.getLogger(__name__) class GenerationStopper: """ Base interface for stopping logic. - should_stop(s): True to stop given the current decoded text window. - lookback_tokens(): how many tokens should be considered (default: sys.maxsize). """ @abstractmethod def should_stop(self, s: str) -> bool: pass def lookback_tokens(self) -> int: return sys.maxsize class DocTagsRepetitionStopper(GenerationStopper): """ Detects repetitive <tag>...<loc_x><loc_y><loc_w><loc_h>text</tag> blocks, but only when repeats are **consecutive** and both tag & inner text are identical. Performance: - Heavy check runs every N calls (default 32). - Only decodes the last LOOKBACK_TOKENS tokens per sequence (default 200). """ def __init__(self, *, N: int = 32, lookback_tokens: int = 200): self.N = max(1, int(N)) self._lookback_tokens = max(1, int(lookback_tokens)) self._call_count = 0 # <tag> ... <loc_x><loc_y><loc_w><loc_h> text ... </tag> self._PATTERN = re.compile( r""" <(?P<tag>[a-zA-Z0-9_]+)>\s* (?P<prefix>.*?)? <loc_(?P<x>\d+)><loc_(?P<y>\d+)><loc_(?P<w>\d+)><loc_(?P<h>\d+)> (?P<text>.*?) </(?P=tag)> """, re.DOTALL | re.VERBOSE, ) # --- small helper --- def _regular(self, vals: List[int]) -> bool: """3+ strictly increasing values with ~regular spacing (±20%).""" if len(vals) < 3: return False diffs = [b - a for a, b in zip(vals, vals[1:])] if any(d <= 0 for d in diffs): return False mean = sum(diffs) / len(diffs) tol = 0.2 * mean return all(abs(d - mean) <= tol for d in diffs) def should_stop(self, s: str) -> bool: """ Trip only on **consecutive** runs (no other matched blocks between) of ≥3 items with the same <tag> and identical inner text, where within that run we see: - any exact duplicate (x,y,w,h), or - stable X/W with regular Y progression, or - stable Y/H with regular X progression. """ # Stream matches and evaluate runs on-the-fly to stay compact and fast. prev_tag = prev_text = None run = [] # list of (x,y,w,h) def run_repetitive(boxes: List[tuple]) -> bool: if len(boxes) < 3: return False # duplicates? if len(set(boxes)) < len(boxes): return True xs, ys, ws, hs = zip(*boxes) x_stable = all(x == xs[0] for x in xs) y_stable = all(y == ys[0] for y in ys) w_stable = all(w == ws[0] for w in ws) h_stable = all(h == hs[0] for h in hs) # horizontal (down the page): X/W stable, Y regular if (x_stable or w_stable) and self._regular(list(ys)): return True # vertical (across): Y/H stable, X regular if (y_stable or h_stable) and self._regular(list(xs)): return True return False for m in self._PATTERN.finditer(s): tag, text = m.group("tag"), m.group("text") box = ( int(m.group("x")), int(m.group("y")), int(m.group("w")), int(m.group("h")), ) if prev_tag == tag and prev_text == text: run.append(box) # consecutive same-tag+text else: # evaluate previous run before starting a new one if run_repetitive(run): return True prev_tag, prev_text = tag, text run = [box] # check the last run return run_repetitive(run) class HFStoppingCriteriaWrapper(StoppingCriteria): """ Adapts any GenerationStopper to HuggingFace Transformers. Decodes exactly min(seq_len, stopper.lookback_tokens()) tokens from the end. """ def __init__( self, tokenizer, stopper: GenerationStopper, *, skip_special_tokens: bool = False, ): self.tokenizer = tokenizer self.stopper = stopper self.skip_special_tokens = skip_special_tokens def __call__(self, input_ids, scores, **kwargs) -> bool: lb = max(1, int(self.stopper.lookback_tokens())) for seq in input_ids: # (batch, seq_len) window = seq[-lb:] # slicing handles lb > len(seq) try: text = self.tokenizer.decode( window, skip_special_tokens=self.skip_special_tokens ) except Exception as e: _log.info(f"Decoding failed for stopping check: {e}") continue try: if self.stopper.should_stop(text): _log.info( "HF wrapper: stopping due to TextStopper.should_stop==True" ) return True except Exception as e: _log.info(f"Error in TextStopper.should_stop: {e}") continue return False
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/utils/hf_model_download.py
docling/models/utils/hf_model_download.py
import logging from pathlib import Path from typing import Optional _log = logging.getLogger(__name__) def download_hf_model( repo_id: str, local_dir: Optional[Path] = None, force: bool = False, progress: bool = False, revision: Optional[str] = None, ) -> Path: from huggingface_hub import snapshot_download from huggingface_hub.utils import disable_progress_bars if not progress: disable_progress_bars() download_path = snapshot_download( repo_id=repo_id, force_download=force, local_dir=local_dir, revision=revision, ) return Path(download_path) class HuggingFaceModelDownloadMixin: @staticmethod def download_models( repo_id: str, local_dir: Optional[Path] = None, force: bool = False, progress: bool = False, revision: Optional[str] = None, ) -> Path: return download_hf_model( repo_id=repo_id, local_dir=local_dir, force=force, progress=progress, revision=revision, )
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/plugins/defaults.py
docling/models/plugins/defaults.py
def ocr_engines(): from docling.models.auto_ocr_model import OcrAutoModel from docling.models.easyocr_model import EasyOcrModel from docling.models.ocr_mac_model import OcrMacModel from docling.models.rapid_ocr_model import RapidOcrModel from docling.models.tesseract_ocr_cli_model import TesseractOcrCliModel from docling.models.tesseract_ocr_model import TesseractOcrModel return { "ocr_engines": [ OcrAutoModel, EasyOcrModel, OcrMacModel, RapidOcrModel, TesseractOcrModel, TesseractOcrCliModel, ] } def picture_description(): from docling.models.picture_description_api_model import PictureDescriptionApiModel from docling.models.picture_description_vlm_model import PictureDescriptionVlmModel return { "picture_description": [ PictureDescriptionVlmModel, PictureDescriptionApiModel, ] } def layout_engines(): from docling.experimental.models.table_crops_layout_model import ( TableCropsLayoutModel, ) from docling.models.layout_model import LayoutModel return { "layout_engines": [ LayoutModel, TableCropsLayoutModel, ] } def table_structure_engines(): from docling.models.table_structure_model import TableStructureModel return { "table_structure_engines": [ TableStructureModel, ] }
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/plugins/__init__.py
docling/models/plugins/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/vlm_models_inline/nuextract_transformers_model.py
docling/models/vlm_models_inline/nuextract_transformers_model.py
import logging import sys import time from collections.abc import Iterable from pathlib import Path from typing import Any, Optional, Union import numpy as np from PIL.Image import Image from transformers import AutoModelForImageTextToText, AutoProcessor, GenerationConfig from docling.datamodel.accelerator_options import ( AcceleratorOptions, ) from docling.datamodel.base_models import VlmPrediction, VlmStopReason from docling.datamodel.pipeline_options_vlm_model import InlineVlmOptions from docling.models.base_model import BaseVlmModel from docling.models.utils.hf_model_download import ( HuggingFaceModelDownloadMixin, ) from docling.utils.accelerator_utils import decide_device _log = logging.getLogger(__name__) # Source code from https://huggingface.co/numind/NuExtract-2.0-8B def process_all_vision_info(messages, examples=None): """ Process vision information from both messages and in-context examples, supporting batch processing. Args: messages: List of message dictionaries (single input) OR list of message lists (batch input) examples: Optional list of example dictionaries (single input) OR list of example lists (batch) Returns: A flat list of all images in the correct order: - For single input: example images followed by message images - For batch input: interleaved as (item1 examples, item1 input, item2 examples, item2 input, etc.) - Returns None if no images were found """ try: from qwen_vl_utils import fetch_image, process_vision_info except ImportError: raise ImportError( "qwen-vl-utils is required for NuExtractTransformersModel. " "Please install it with: pip install qwen-vl-utils" ) from qwen_vl_utils import fetch_image, process_vision_info # Helper function to extract images from examples def extract_example_images(example_item): if not example_item: return [] # Handle both list of examples and single example examples_to_process = ( example_item if isinstance(example_item, list) else [example_item] ) images = [] for example in examples_to_process: if ( isinstance(example.get("input"), dict) and example["input"].get("type") == "image" ): images.append(fetch_image(example["input"])) return images # Normalize inputs to always be batched format is_batch = messages and isinstance(messages[0], list) messages_batch = messages if is_batch else [messages] is_batch_examples = ( examples and isinstance(examples, list) and (isinstance(examples[0], list) or examples[0] is None) ) examples_batch = ( examples if is_batch_examples else ([examples] if examples is not None else None) ) # Ensure examples batch matches messages batch if provided if examples and len(examples_batch) != len(messages_batch): if not is_batch and len(examples_batch) == 1: # Single example set for a single input is fine pass else: raise ValueError("Examples batch length must match messages batch length") # Process all inputs, maintaining correct order all_images = [] for i, message_group in enumerate(messages_batch): # Get example images for this input if examples and i < len(examples_batch): input_example_images = extract_example_images(examples_batch[i]) all_images.extend(input_example_images) # Get message images for this input input_message_images = process_vision_info(message_group)[0] or [] all_images.extend(input_message_images) return all_images if all_images else None class NuExtractTransformersModel(BaseVlmModel, HuggingFaceModelDownloadMixin): def __init__( self, enabled: bool, artifacts_path: Optional[Path], accelerator_options: AcceleratorOptions, vlm_options: InlineVlmOptions, ): self.enabled = enabled self.vlm_options = vlm_options if self.enabled: import torch self.device = decide_device( accelerator_options.device, supported_devices=vlm_options.supported_devices, ) _log.debug(f"Available device for NuExtract VLM: {self.device}") self.max_new_tokens = vlm_options.max_new_tokens self.temperature = vlm_options.temperature repo_cache_folder = vlm_options.repo_id.replace("/", "--") if artifacts_path is None: artifacts_path = self.download_models( repo_id=self.vlm_options.repo_id, revision=self.vlm_options.revision, ) elif (artifacts_path / repo_cache_folder).exists(): artifacts_path = artifacts_path / repo_cache_folder self.processor = AutoProcessor.from_pretrained( artifacts_path, trust_remote_code=vlm_options.trust_remote_code, use_fast=True, ) self.processor.tokenizer.padding_side = "left" self.vlm_model = AutoModelForImageTextToText.from_pretrained( artifacts_path, device_map=self.device, dtype=self.vlm_options.torch_dtype, _attn_implementation=( "flash_attention_2" if self.device.startswith("cuda") and accelerator_options.cuda_use_flash_attention2 else "sdpa" ), trust_remote_code=vlm_options.trust_remote_code, ) if sys.version_info < (3, 14): self.vlm_model = torch.compile(self.vlm_model) # type: ignore else: self.vlm_model.eval() # Load generation config self.generation_config = GenerationConfig.from_pretrained(artifacts_path) def process_images( self, image_batch: Iterable[Union[Image, np.ndarray]], prompt: Union[str, list[str]], ) -> Iterable[VlmPrediction]: """ Batched inference for NuExtract VLM using the specialized input format. Args: image_batch: Iterable of PIL Images or numpy arrays prompt: Either: - str: Single template used for all images - list[str]: List of templates (one per image, must match image count) """ import torch from PIL import Image as PILImage # Normalize images to RGB PIL pil_images: list[Image] = [] for img in image_batch: if isinstance(img, np.ndarray): if img.ndim == 3 and img.shape[2] in (3, 4): pil_img = PILImage.fromarray(img.astype(np.uint8)) elif img.ndim == 2: pil_img = PILImage.fromarray(img.astype(np.uint8), mode="L") else: raise ValueError(f"Unsupported numpy array shape: {img.shape}") else: pil_img = img if pil_img.mode != "RGB": pil_img = pil_img.convert("RGB") pil_images.append(pil_img) if not pil_images: return # Normalize templates (1 per image) if isinstance(prompt, str): templates = [prompt] * len(pil_images) else: if len(prompt) != len(pil_images): raise ValueError( f"Number of templates ({len(prompt)}) must match number of images ({len(pil_images)})" ) templates = prompt # Construct NuExtract input format inputs = [] for pil_img, template in zip(pil_images, templates): input_item = { "document": {"type": "image", "image": pil_img}, "template": template, } inputs.append(input_item) # Create messages structure for batch processing messages = [ [ { "role": "user", "content": [x["document"]], } ] for x in inputs ] # Apply chat template to each example individually texts = [ self.processor.tokenizer.apply_chat_template( messages[i], template=x["template"], tokenize=False, add_generation_prompt=True, ) for i, x in enumerate(inputs) ] # Process vision inputs using qwen-vl-utils image_inputs = process_all_vision_info(messages) # Process with the processor processor_inputs = self.processor( text=texts, images=image_inputs, padding=True, return_tensors="pt", **self.vlm_options.extra_processor_kwargs, ) processor_inputs = {k: v.to(self.device) for k, v in processor_inputs.items()} # Generate gen_kwargs = { **processor_inputs, "max_new_tokens": self.max_new_tokens, "generation_config": self.generation_config, **self.vlm_options.extra_generation_config, } if self.temperature > 0: gen_kwargs["do_sample"] = True gen_kwargs["temperature"] = self.temperature else: gen_kwargs["do_sample"] = False start_time = time.time() with torch.inference_mode(): generated_ids = self.vlm_model.generate(**gen_kwargs) generation_time = time.time() - start_time # Trim generated sequences input_len = processor_inputs["input_ids"].shape[1] trimmed_sequences = generated_ids[:, input_len:] # Decode with the processor/tokenizer decoded_texts: list[str] = self.processor.batch_decode( trimmed_sequences, skip_special_tokens=True, clean_up_tokenization_spaces=False, ) # Optional logging num_tokens = None if generated_ids.shape[0] > 0: # type: ignore # Todo: confirm num tokens is actually from first item, code was already like this num_tokens = int(generated_ids[0].shape[0]) _log.debug( f"Generated {num_tokens} tokens in {generation_time:.2f}s " f"for batch size {generated_ids.shape[0]}." # type: ignore ) for text in decoded_texts: # Apply decode_response to the output text decoded_text = self.vlm_options.decode_response(text) yield VlmPrediction( text=decoded_text, generation_time=generation_time, num_tokens=num_tokens, stop_reason=VlmStopReason.UNSPECIFIED, )
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/vlm_models_inline/hf_transformers_model.py
docling/models/vlm_models_inline/hf_transformers_model.py
import importlib.metadata import logging import sys import time from collections.abc import Iterable from pathlib import Path from typing import Any, Optional, Union import numpy as np from PIL.Image import Image from transformers import StoppingCriteria, StoppingCriteriaList, StopStringCriteria from docling.datamodel.accelerator_options import ( AcceleratorOptions, ) from docling.datamodel.base_models import Page, VlmPrediction, VlmStopReason from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options_vlm_model import ( InlineVlmOptions, TransformersModelType, TransformersPromptStyle, ) from docling.models.base_model import BaseVlmPageModel from docling.models.utils.generation_utils import ( GenerationStopper, HFStoppingCriteriaWrapper, ) from docling.models.utils.hf_model_download import ( HuggingFaceModelDownloadMixin, ) from docling.utils.accelerator_utils import decide_device from docling.utils.profiling import TimeRecorder _log = logging.getLogger(__name__) class HuggingFaceTransformersVlmModel(BaseVlmPageModel, HuggingFaceModelDownloadMixin): def __init__( self, enabled: bool, artifacts_path: Optional[Path], accelerator_options: AcceleratorOptions, vlm_options: InlineVlmOptions, ): self.enabled = enabled self.vlm_options = vlm_options if self.enabled: import torch from transformers import ( AutoModel, AutoModelForCausalLM, AutoModelForImageTextToText, AutoModelForVision2Seq, AutoProcessor, BitsAndBytesConfig, GenerationConfig, ) transformers_version = importlib.metadata.version("transformers") if ( self.vlm_options.repo_id == "microsoft/Phi-4-multimodal-instruct" and transformers_version >= "4.52.0" ): raise NotImplementedError( f"Phi 4 only works with transformers<4.52.0 but you have {transformers_version=}. Please downgrage running pip install -U 'transformers<4.52.0'." ) self.device = decide_device( accelerator_options.device, supported_devices=vlm_options.supported_devices, ) _log.debug(f"Available device for VLM: {self.device}") self.use_cache = vlm_options.use_kv_cache self.max_new_tokens = vlm_options.max_new_tokens self.temperature = vlm_options.temperature repo_cache_folder = vlm_options.repo_id.replace("/", "--") if artifacts_path is None: artifacts_path = self.download_models( self.vlm_options.repo_id, revision=self.vlm_options.revision ) elif (artifacts_path / repo_cache_folder).exists(): artifacts_path = artifacts_path / repo_cache_folder self.param_quantization_config: Optional[BitsAndBytesConfig] = None if vlm_options.quantized: self.param_quantization_config = BitsAndBytesConfig( load_in_8bit=vlm_options.load_in_8bit, llm_int8_threshold=vlm_options.llm_int8_threshold, ) model_cls: Any = AutoModel if ( self.vlm_options.transformers_model_type == TransformersModelType.AUTOMODEL_CAUSALLM ): model_cls = AutoModelForCausalLM elif ( self.vlm_options.transformers_model_type == TransformersModelType.AUTOMODEL_VISION2SEQ ): model_cls = AutoModelForVision2Seq elif ( self.vlm_options.transformers_model_type == TransformersModelType.AUTOMODEL_IMAGETEXTTOTEXT ): model_cls = AutoModelForImageTextToText self.processor = AutoProcessor.from_pretrained( artifacts_path, trust_remote_code=vlm_options.trust_remote_code, revision=vlm_options.revision, ) self.processor.tokenizer.padding_side = "left" self.vlm_model = model_cls.from_pretrained( artifacts_path, device_map=self.device, dtype=self.vlm_options.torch_dtype, _attn_implementation=( "flash_attention_2" if self.device.startswith("cuda") and accelerator_options.cuda_use_flash_attention2 else "sdpa" ), trust_remote_code=vlm_options.trust_remote_code, revision=vlm_options.revision, ) if sys.version_info < (3, 14): self.vlm_model = torch.compile(self.vlm_model) # type: ignore else: self.vlm_model.eval() # Load generation config self.generation_config = GenerationConfig.from_pretrained( artifacts_path, revision=vlm_options.revision ) def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: page_list = list(page_batch) if not page_list: return valid_pages = [] invalid_pages = [] for page in page_list: assert page._backend is not None if not page._backend.is_valid(): invalid_pages.append(page) else: valid_pages.append(page) # Process valid pages in batch if valid_pages: with TimeRecorder(conv_res, "vlm"): # Prepare images and prompts for batch processing images = [] user_prompts = [] pages_with_images = [] for page in valid_pages: assert page.size is not None hi_res_image = page.get_image( scale=self.vlm_options.scale, max_size=self.vlm_options.max_size ) # Only process pages with valid images if hi_res_image is not None: images.append(hi_res_image) # Define prompt structure user_prompt = self._build_prompt_safe(page) user_prompts.append(user_prompt) pages_with_images.append(page) # Use process_images for the actual inference if images: # Only if we have valid images with TimeRecorder(conv_res, "vlm_inference"): predictions = list(self.process_images(images, user_prompts)) # Attach results to pages for page, prediction in zip(pages_with_images, predictions): page.predictions.vlm_response = prediction # Yield all pages (valid and invalid) for page in invalid_pages: yield page for page in valid_pages: yield page def process_images( self, image_batch: Iterable[Union[Image, np.ndarray]], prompt: Union[str, list[str]], ) -> Iterable[VlmPrediction]: """ Batched inference for Hugging Face Image-Text-to-Text VLMs (e.g., SmolDocling / SmolVLM). - Lets the processor handle all padding & batching for text+images. - Trims generated sequences per row using attention_mask (no pad-id fallbacks). - Keeps your formulate_prompt() exactly as-is. """ import numpy as np import torch from PIL import Image as PILImage # -- Normalize images to RGB PIL pil_images: list[Image] = [] for img in image_batch: if isinstance(img, np.ndarray): if img.ndim == 3 and img.shape[2] in (3, 4): pil_img = PILImage.fromarray(img.astype(np.uint8)) elif img.ndim == 2: pil_img = PILImage.fromarray(img.astype(np.uint8), mode="L") else: raise ValueError(f"Unsupported numpy array shape: {img.shape}") else: pil_img = img if pil_img.mode != "RGB": pil_img = pil_img.convert("RGB") pil_images.append(pil_img) if not pil_images: return # -- Normalize prompts (1 per image) if isinstance(prompt, str): user_prompts = [prompt] * len(pil_images) else: if len(prompt) != len(pil_images): raise ValueError( f"Number of prompts ({len(prompt)}) must match number of images ({len(pil_images)})" ) user_prompts = prompt # Use your prompt formatter verbatim if self.vlm_options.transformers_prompt_style == TransformersPromptStyle.NONE: inputs = self.processor( pil_images, return_tensors="pt", padding=True, # pad across batch for both text and vision **self.vlm_options.extra_processor_kwargs, ) else: prompts: list[str] = [self.formulate_prompt(p) for p in user_prompts] # -- Processor performs BOTH text+image preprocessing + batch padding (recommended) inputs = self.processor( text=prompts, images=pil_images, return_tensors="pt", padding=True, # pad across batch for both text and vision **self.vlm_options.extra_processor_kwargs, ) inputs = {k: v.to(self.device) for k, v in inputs.items()} # -- Optional stopping criteria stopping_criteria_list: StoppingCriteriaList = StoppingCriteriaList() # Add string-based stopping criteria if self.vlm_options.stop_strings: stopping_criteria_list.append( StopStringCriteria( stop_strings=self.vlm_options.stop_strings, tokenizer=self.processor.tokenizer, ) ) # Add custom stopping criteria if self.vlm_options.custom_stopping_criteria: for criteria in self.vlm_options.custom_stopping_criteria: # If it's a class (not an instance), determine the type and handle accordingly if isinstance(criteria, type): # Check if it's a GenerationStopper class if issubclass(criteria, GenerationStopper): # Instantiate GenerationStopper and wrap it stopper_instance = criteria() wrapped_criteria = HFStoppingCriteriaWrapper( self.processor.tokenizer, stopper_instance ) stopping_criteria_list.append(wrapped_criteria) elif issubclass(criteria, StoppingCriteria): # It's a StoppingCriteria class, instantiate with tokenizer criteria_instance = criteria(self.processor.tokenizer) stopping_criteria_list.append(criteria_instance) elif isinstance(criteria, GenerationStopper): # Wrap GenerationStopper instances in HFStoppingCriteriaWrapper wrapped_criteria = HFStoppingCriteriaWrapper( self.processor.tokenizer, criteria ) stopping_criteria_list.append(wrapped_criteria) else: # If it's already an instance of StoppingCriteria, use it directly stopping_criteria_list.append(criteria) stopping_criteria = ( StoppingCriteriaList(stopping_criteria_list) if stopping_criteria_list else None ) # -- Filter out decoder-specific keys from extra_generation_config decoder_keys = { "skip_special_tokens", "clean_up_tokenization_spaces", "spaces_between_special_tokens", } generation_config = { k: v for k, v in self.vlm_options.extra_generation_config.items() if k not in decoder_keys } decoder_config = { k: v for k, v in self.vlm_options.extra_generation_config.items() if k in decoder_keys } # -- Generate (Image-Text-to-Text class expects these inputs from processor) gen_kwargs = { **inputs, "max_new_tokens": self.max_new_tokens, "use_cache": self.use_cache, "generation_config": self.generation_config, **generation_config, } if self.temperature > 0: gen_kwargs["do_sample"] = True gen_kwargs["temperature"] = self.temperature else: gen_kwargs["do_sample"] = False if stopping_criteria is not None: gen_kwargs["stopping_criteria"] = stopping_criteria start_time = time.time() with torch.inference_mode(): generated_ids = self.vlm_model.generate(**gen_kwargs) generation_time = time.time() - start_time input_len = inputs["input_ids"].shape[1] # common right-aligned prompt length trimmed_sequences = generated_ids[:, input_len:] # only newly generated tokens # -- Decode with the processor/tokenizer (skip specials, keep DocTags as text) decode_fn = getattr(self.processor, "batch_decode", None) if decode_fn is None and getattr(self.processor, "tokenizer", None) is not None: decode_fn = self.processor.tokenizer.batch_decode if decode_fn is None: raise RuntimeError( "Neither processor.batch_decode nor tokenizer.batch_decode is available." ) decoded_texts: list[str] = decode_fn( trimmed_sequences, **decoder_config, ) # -- Clip off pad tokens from decoded texts pad_token = self.processor.tokenizer.pad_token if pad_token: decoded_texts = [text.rstrip(pad_token) for text in decoded_texts] # -- Optional logging num_tokens = None if generated_ids.shape[0] > 0: num_tokens = int(generated_ids[0].shape[0]) _log.debug( f"Generated {num_tokens} tokens in {generation_time:.2f}s " f"for batch size {generated_ids.shape[0]}." ) for i, text in enumerate(decoded_texts): input_prompt = ( prompts[i] if self.vlm_options.track_input_prompt and prompts else None ) # Apply decode_response to the output text decoded_text = self.vlm_options.decode_response(text) yield VlmPrediction( text=decoded_text, generation_time=generation_time, num_tokens=num_tokens, stop_reason=VlmStopReason.UNSPECIFIED, input_prompt=input_prompt, )
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/vlm_models_inline/mlx_model.py
docling/models/vlm_models_inline/mlx_model.py
import logging import sys import threading import time from collections.abc import Iterable from pathlib import Path from typing import Optional, Union import numpy as np from PIL.Image import Image from transformers import StoppingCriteria from docling.datamodel.accelerator_options import ( AcceleratorOptions, ) from docling.datamodel.base_models import ( Page, VlmPrediction, VlmPredictionToken, VlmStopReason, ) from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options_vlm_model import InlineVlmOptions from docling.models.base_model import BaseVlmPageModel from docling.models.utils.generation_utils import GenerationStopper from docling.models.utils.hf_model_download import ( HuggingFaceModelDownloadMixin, ) from docling.utils.profiling import TimeRecorder _log = logging.getLogger(__name__) # Global lock for MLX model calls - MLX models are not thread-safe # All MLX models share this lock to prevent concurrent MLX operations _MLX_GLOBAL_LOCK = threading.Lock() class HuggingFaceMlxModel(BaseVlmPageModel, HuggingFaceModelDownloadMixin): def __init__( self, enabled: bool, artifacts_path: Optional[Path], accelerator_options: AcceleratorOptions, vlm_options: InlineVlmOptions, ): self.enabled = enabled self.vlm_options = vlm_options self.max_tokens = vlm_options.max_new_tokens self.temperature = vlm_options.temperature if self.enabled: try: from mlx_vlm import generate, load, stream_generate # type: ignore from mlx_vlm.prompt_utils import apply_chat_template # type: ignore from mlx_vlm.utils import load_config # type: ignore except ImportError: if sys.version_info < (3, 14): raise ImportError( "mlx-vlm is not installed. Please install it via `pip install mlx-vlm` to use MLX VLM models." ) else: raise ImportError( "mlx-vlm is not installed. It is not yet available on Python 3.14." ) repo_cache_folder = vlm_options.repo_id.replace("/", "--") self.apply_chat_template = apply_chat_template self.stream_generate = stream_generate # PARAMETERS: if artifacts_path is None: artifacts_path = self.download_models( self.vlm_options.repo_id, revision=self.vlm_options.revision, ) elif (artifacts_path / repo_cache_folder).exists(): artifacts_path = artifacts_path / repo_cache_folder ## Load the model self.vlm_model, self.processor = load(artifacts_path) self.config = load_config(artifacts_path) # Validate custom stopping criteria - MLX doesn't support HF StoppingCriteria if self.vlm_options.custom_stopping_criteria: for criteria in self.vlm_options.custom_stopping_criteria: if isinstance(criteria, StoppingCriteria): raise ValueError( f"MLX models do not support HuggingFace StoppingCriteria instances. " f"Found {type(criteria).__name__}. Use GenerationStopper instead." ) elif isinstance(criteria, type) and issubclass( criteria, StoppingCriteria ): raise ValueError( f"MLX models do not support HuggingFace StoppingCriteria classes. " f"Found {criteria.__name__}. Use GenerationStopper instead." ) def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: page_list = list(page_batch) if not page_list: return valid_pages = [] invalid_pages = [] for page in page_list: assert page._backend is not None if not page._backend.is_valid(): invalid_pages.append(page) else: valid_pages.append(page) # Process valid pages in batch if valid_pages: with TimeRecorder(conv_res, f"vlm-mlx-{self.vlm_options.repo_id}"): # Prepare images and prompts for batch processing images = [] user_prompts = [] pages_with_images = [] for page in valid_pages: assert page.size is not None hi_res_image = page.get_image( scale=self.vlm_options.scale, max_size=self.vlm_options.max_size ) # Only process pages with valid images if hi_res_image is not None: images.append(hi_res_image) # Define prompt structure user_prompt = self._build_prompt_safe(page) user_prompts.append(user_prompt) pages_with_images.append(page) # Use process_images for the actual inference if images: # Only if we have valid images predictions = list(self.process_images(images, user_prompts)) # Attach results to pages for page, prediction in zip(pages_with_images, predictions): page.predictions.vlm_response = prediction # Yield all pages (valid and invalid) for page in invalid_pages: yield page for page in valid_pages: yield page def process_images( self, image_batch: Iterable[Union[Image, np.ndarray]], prompt: Union[str, list[str]], ) -> Iterable[VlmPrediction]: """Process raw images without page metadata. Args: image_batch: Iterable of PIL Images or numpy arrays prompt: Either: - str: Single prompt used for all images - list[str]: List of prompts (one per image, must match image count) Raises: ValueError: If prompt list length doesn't match image count. """ # Convert image batch to list for length validation image_list = list(image_batch) if len(image_list) == 0: return # Handle prompt parameter if isinstance(prompt, str): # Single prompt for all images user_prompts = [prompt] * len(image_list) elif isinstance(prompt, list): # List of prompts (one per image) if len(prompt) != len(image_list): raise ValueError( f"Number of prompts ({len(prompt)}) must match number of images ({len(image_list)})" ) user_prompts = prompt else: raise ValueError(f"prompt must be str or list[str], got {type(prompt)}") # MLX models are not thread-safe - use global lock to serialize access with _MLX_GLOBAL_LOCK: _log.debug("MLX model: Acquired global lock for thread safety") for image, user_prompt in zip(image_list, user_prompts): # Convert numpy array to PIL Image if needed if isinstance(image, np.ndarray): if image.ndim == 3 and image.shape[2] in [3, 4]: # RGB or RGBA array from PIL import Image as PILImage image = PILImage.fromarray(image.astype(np.uint8)) elif image.ndim == 2: # Grayscale array from PIL import Image as PILImage image = PILImage.fromarray(image.astype(np.uint8), mode="L") else: raise ValueError( f"Unsupported numpy array shape: {image.shape}" ) # Ensure image is in RGB mode (handles RGBA, L, etc.) if image.mode != "RGB": image = image.convert("RGB") # Use the MLX chat template approach like in the __call__ method formatted_prompt = self.apply_chat_template( self.processor, self.config, user_prompt, num_images=1 ) # Stream generate with stop strings and custom stopping criteria support start_time = time.time() _log.debug("start generating ...") tokens: list[VlmPredictionToken] = [] output = "" # Use stream_generate for proper stop string handling for token in self.stream_generate( self.vlm_model, self.processor, formatted_prompt, [image], # MLX stream_generate expects list of images max_tokens=self.max_tokens, verbose=False, temp=self.temperature, ): # Collect token information if len(token.logprobs.shape) == 1: tokens.append( VlmPredictionToken( text=token.text, token=token.token, logprob=token.logprobs[token.token], ) ) elif ( len(token.logprobs.shape) == 2 and token.logprobs.shape[0] == 1 ): tokens.append( VlmPredictionToken( text=token.text, token=token.token, logprob=token.logprobs[0, token.token], ) ) else: _log.warning( f"incompatible shape for logprobs: {token.logprobs.shape}" ) output += token.text # Check for any configured stop strings if self.vlm_options.stop_strings: if any( stop_str in output for stop_str in self.vlm_options.stop_strings ): _log.debug("Stopping generation due to stop string match") break # Check for custom stopping criteria (GenerationStopper instances) if self.vlm_options.custom_stopping_criteria: for criteria in self.vlm_options.custom_stopping_criteria: # Handle both instances and classes of GenerationStopper if isinstance(criteria, GenerationStopper): stopper = criteria elif isinstance(criteria, type) and issubclass( criteria, GenerationStopper ): stopper = criteria() # Determine the text window to check based on lookback_tokens lookback_tokens = stopper.lookback_tokens() # Check only the last N characters worth of text # This is a simplified approach - in practice, you might want to # decode the last N tokens from the token list for more accuracy text_to_check = ( output[-lookback_tokens:] if len(output) > lookback_tokens else output ) try: if stopper.should_stop(text_to_check): _log.info( f"Stopping generation due to GenerationStopper: {type(stopper).__name__}" ) break except Exception as e: _log.warning( f"Error in GenerationStopper.should_stop: {e}" ) continue else: # note: for-else idiom continue # Only executed if the inner loop didn't break break # Break the outer loop if any stopper triggered generation_time = time.time() - start_time _log.debug( f"{generation_time:.2f} seconds for {len(tokens)} tokens ({len(tokens) / generation_time:.1f} tokens/sec)." ) # Apply decode_response to the output before yielding decoded_output = self.vlm_options.decode_response(output) input_prompt = ( formatted_prompt if self.vlm_options.track_input_prompt else None ) yield VlmPrediction( text=decoded_output, generation_time=generation_time, generated_tokens=tokens, num_tokens=len(tokens), stop_reason=VlmStopReason.UNSPECIFIED, input_prompt=input_prompt, ) _log.debug("MLX model: Released global lock")
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/vlm_models_inline/__init__.py
docling/models/vlm_models_inline/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/models/vlm_models_inline/vllm_model.py
docling/models/vlm_models_inline/vllm_model.py
import logging import sys import time from collections.abc import Iterable from pathlib import Path from typing import Any, Dict, Optional, Union import numpy as np from PIL.Image import Image from docling.datamodel.accelerator_options import AcceleratorOptions from docling.datamodel.base_models import ( Page, VlmPrediction, VlmPredictionToken, VlmStopReason, ) from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options_vlm_model import ( InlineVlmOptions, TransformersPromptStyle, ) from docling.models.base_model import BaseVlmPageModel from docling.models.utils.hf_model_download import HuggingFaceModelDownloadMixin from docling.utils.accelerator_utils import decide_device from docling.utils.profiling import TimeRecorder _log = logging.getLogger(__name__) class VllmVlmModel(BaseVlmPageModel, HuggingFaceModelDownloadMixin): """ vLLM-backed vision-language model that accepts PIL images (or numpy arrays) via vLLM's multi_modal_data, with prompt formatting handled by formulate_prompt(). """ # --------- Allowlist of vLLM args --------- # SamplingParams (runtime generation controls) _VLLM_SAMPLING_KEYS = { # Core "max_tokens", "temperature", "top_p", "top_k", # Penalties "presence_penalty", "frequency_penalty", "repetition_penalty", # Stops / outputs "stop", "stop_token_ids", "skip_special_tokens", "spaces_between_special_tokens", # Search / length "n", "best_of", "length_penalty", "early_stopping", # Misc "logprobs", "prompt_logprobs", "min_p", "seed", } # LLM(...) / EngineArgs (engine/load-time controls) _VLLM_ENGINE_KEYS = { # Model/tokenizer/impl "tokenizer", "tokenizer_mode", "download_dir", # Parallelism / memory / lengths "tensor_parallel_size", "pipeline_parallel_size", "gpu_memory_utilization", "max_model_len", "max_num_batched_tokens", "kv_cache_dtype", "dtype", # Quantization (coarse switch) "quantization", # Multimodal limits "limit_mm_per_prompt", # Execution toggles "enforce_eager", } def __init__( self, enabled: bool, artifacts_path: Optional[Path], accelerator_options: AcceleratorOptions, vlm_options: InlineVlmOptions, ): self.enabled = enabled self.vlm_options: InlineVlmOptions = vlm_options self.llm = None self.sampling_params = None self.processor = None # used for CHAT templating in formulate_prompt() self.device = "cpu" self.max_new_tokens = vlm_options.max_new_tokens self.temperature = vlm_options.temperature if not self.enabled: return from transformers import AutoProcessor try: from vllm import LLM, SamplingParams except ImportError: if sys.version_info < (3, 14): raise ImportError( "vllm is not installed. Please install it via `pip install vllm`." ) else: raise ImportError( "vllm is not installed. It is not yet available on Python 3.14." ) # Device selection self.device = decide_device( accelerator_options.device, supported_devices=vlm_options.supported_devices ) _log.debug(f"Available device for VLM: {self.device}") # Resolve artifacts path / cache folder repo_cache_folder = vlm_options.repo_id.replace("/", "--") if artifacts_path is None: artifacts_path = self.download_models( self.vlm_options.repo_id, revision=self.vlm_options.revision ) elif (artifacts_path / repo_cache_folder).exists(): artifacts_path = artifacts_path / repo_cache_folder # --------- Strict split & validation of extra_generation_config --------- extra_cfg = self.vlm_options.extra_generation_config load_cfg = {k: v for k, v in extra_cfg.items() if k in self._VLLM_ENGINE_KEYS} gen_cfg = {k: v for k, v in extra_cfg.items() if k in self._VLLM_SAMPLING_KEYS} unknown = sorted( k for k in extra_cfg.keys() if k not in self._VLLM_ENGINE_KEYS and k not in self._VLLM_SAMPLING_KEYS ) if unknown: _log.warning( "Ignoring unknown extra_generation_config keys for vLLM: %s", unknown ) # --------- Construct LLM kwargs (engine/load-time) --------- llm_kwargs: Dict[str, Any] = { "model": str(artifacts_path), "model_impl": "transformers", "limit_mm_per_prompt": {"image": 1}, "revision": self.vlm_options.revision, "trust_remote_code": self.vlm_options.trust_remote_code, **load_cfg, } if self.device == "cpu": llm_kwargs.setdefault("enforce_eager", True) else: llm_kwargs.setdefault( "gpu_memory_utilization", 0.3 ) # room for other models # Quantization (kept as-is; coarse) if self.vlm_options.quantized and self.vlm_options.load_in_8bit: llm_kwargs.setdefault("quantization", "bitsandbytes") # Initialize vLLM LLM self.llm = LLM(**llm_kwargs) # Initialize processor for prompt templating (needed for CHAT style) self.processor = AutoProcessor.from_pretrained( artifacts_path, trust_remote_code=self.vlm_options.trust_remote_code, revision=self.vlm_options.revision, ) # --------- SamplingParams (runtime) --------- self.sampling_params = SamplingParams( temperature=self.temperature, max_tokens=self.max_new_tokens, stop=(self.vlm_options.stop_strings or None), **gen_cfg, ) def __call__( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: # If disabled, pass-through if not self.enabled: for page in page_batch: yield page return page_list = list(page_batch) if not page_list: return # Preserve original order original_order = page_list[:] # Separate valid/invalid valid_pages: list[Page] = [] invalid_pages: list[Page] = [] for page in page_list: assert page._backend is not None if page._backend.is_valid(): valid_pages.append(page) else: invalid_pages.append(page) if valid_pages: with TimeRecorder(conv_res, "vlm"): images: list[Image] = [] user_prompts: list[str] = [] pages_with_images: list[Page] = [] for page in valid_pages: assert page.size is not None hi_res_image = page.get_image( scale=self.vlm_options.scale, max_size=self.vlm_options.max_size, ) if hi_res_image is None: continue images.append(hi_res_image) # Define prompt structure user_prompt = self._build_prompt_safe(page) user_prompts.append(user_prompt) pages_with_images.append(page) if images: with TimeRecorder(conv_res, "vlm_inference"): predictions = list(self.process_images(images, user_prompts)) for page, prediction in zip(pages_with_images, predictions): page.predictions.vlm_response = prediction # Yield in original order for page in original_order: yield page def process_images( self, image_batch: Iterable[Union[Image, np.ndarray]], prompt: Union[str, list[str]], ) -> Iterable[VlmPrediction]: """Process images in a single batched vLLM inference call.""" import numpy as np from PIL import Image as PILImage # -- Normalize images to RGB PIL pil_images: list[Image] = [] for img in image_batch: if isinstance(img, np.ndarray): if img.ndim == 3 and img.shape[2] in (3, 4): pil_img = PILImage.fromarray(img.astype(np.uint8)) elif img.ndim == 2: pil_img = PILImage.fromarray(img.astype(np.uint8), mode="L") else: raise ValueError(f"Unsupported numpy array shape: {img.shape}") else: pil_img = img if pil_img.mode != "RGB": pil_img = pil_img.convert("RGB") pil_images.append(pil_img) if not pil_images: return # Normalize prompts if isinstance(prompt, str): user_prompts = [prompt] * len(pil_images) elif isinstance(prompt, list): if len(prompt) != len(pil_images): raise ValueError( f"Number of prompts ({len(prompt)}) must match number of images ({len(pil_images)})" ) user_prompts = prompt else: raise ValueError(f"prompt must be str or list[str], got {type(prompt)}") # Format prompts prompts: list[str] = [self.formulate_prompt(up) for up in user_prompts] # Build vLLM inputs llm_inputs = [ {"prompt": p, "multi_modal_data": {"image": im}} for p, im in zip(prompts, pil_images) ] # Generate assert self.llm is not None and self.sampling_params is not None start_time = time.time() outputs = self.llm.generate(llm_inputs, sampling_params=self.sampling_params) # type: ignore generation_time = time.time() - start_time # Optional debug if outputs: try: num_tokens_within_batch = len(outputs[0].outputs[0].token_ids) _log.debug( f"Generated {num_tokens_within_batch} tokens for batch in {generation_time:.2f}s." ) except Exception: num_tokens_within_batch = 0 # Emit predictions for i, output in enumerate(outputs): text = output.outputs[0].text if output.outputs else "" stop_reason = ( VlmStopReason.END_OF_SEQUENCE if output.outputs[0].stop_reason else VlmStopReason.LENGTH ) generated_tokens = [ VlmPredictionToken(token=int(t)) for t in output.outputs[0].token_ids ] num_tokens = len(generated_tokens) if not self.vlm_options.track_generated_tokens: generated_tokens = [] input_prompt = prompts[i] if self.vlm_options.track_input_prompt else None _log.debug(f"VLM generated response carries input prompt: {input_prompt}") decoded_text = self.vlm_options.decode_response(text) yield VlmPrediction( text=decoded_text, generation_time=generation_time, num_tokens=num_tokens, stop_reason=stop_reason, generated_tokens=generated_tokens, input_prompt=input_prompt, )
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/api_image_request.py
docling/utils/api_image_request.py
import base64 import json import logging from io import BytesIO from typing import Dict, List, Optional, Tuple import requests from PIL import Image from pydantic import AnyUrl from docling.datamodel.base_models import OpenAiApiResponse, VlmStopReason from docling.models.utils.generation_utils import GenerationStopper _log = logging.getLogger(__name__) def api_image_request( image: Image.Image, prompt: str, url: AnyUrl, timeout: float = 20, headers: Optional[dict[str, str]] = None, **params, ) -> Tuple[str, Optional[int], VlmStopReason]: img_io = BytesIO() image = ( image.copy() ) # Fix for inconsistent PIL image width/height to actual byte data image = image.convert("RGBA") good_image = True try: image.save(img_io, "PNG") except Exception as e: good_image = False _log.error(f"Error, corrupter PNG of size: {image.size}: {e}") if good_image: try: image_base64 = base64.b64encode(img_io.getvalue()).decode("utf-8") messages = [ { "role": "user", "content": [ { "type": "image_url", "image_url": { "url": f"data:image/png;base64,{image_base64}" }, }, { "type": "text", "text": prompt, }, ], } ] payload = { "messages": messages, **params, } headers = headers or {} r = requests.post( str(url), headers=headers, json=payload, timeout=timeout, ) if not r.ok: _log.error(f"Error calling the API. Response was {r.text}") # image.show() # r.raise_for_status() api_resp = OpenAiApiResponse.model_validate_json(r.text) generated_text = api_resp.choices[0].message.content.strip() num_tokens = api_resp.usage.total_tokens stop_reason = ( VlmStopReason.LENGTH if api_resp.choices[0].finish_reason == "length" else VlmStopReason.END_OF_SEQUENCE ) return generated_text, num_tokens, stop_reason except Exception as e: _log.error(f"Error, could not process request: {e}") return "", 0, VlmStopReason.UNSPECIFIED else: return "", 0, VlmStopReason.UNSPECIFIED def api_image_request_streaming( image: Image.Image, prompt: str, url: AnyUrl, *, timeout: float = 20, headers: Optional[dict[str, str]] = None, generation_stoppers: list[GenerationStopper] = [], **params, ) -> Tuple[str, Optional[int]]: """ Stream a chat completion from an OpenAI-compatible server (e.g., vLLM). Parses SSE lines: 'data: {json}\\n\\n', terminated by 'data: [DONE]'. Accumulates text and calls stopper.should_stop(window) as chunks arrive. If stopper triggers, the HTTP connection is closed to abort server-side generation. """ img_io = BytesIO() image.save(img_io, "PNG") image_b64 = base64.b64encode(img_io.getvalue()).decode("utf-8") messages = [ { "role": "user", "content": [ { "type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_b64}"}, }, {"type": "text", "text": prompt}, ], } ] payload = { "messages": messages, "stream": True, # <-- critical for SSE streaming **params, } # Debug: Log the payload to verify temperature is included _log.debug(f"API streaming request payload: {json.dumps(payload, indent=2)}") # Some servers require Accept: text/event-stream for SSE. # It's safe to set it; OpenAI-compatible servers tolerate it. hdrs = {"Accept": "text/event-stream", **(headers or {})} # Try to force temperature via header if server ignores payload parameter if "temperature" in params: hdrs["X-Temperature"] = str(params["temperature"]) # Stream the HTTP response with requests.post( str(url), headers=hdrs, json=payload, timeout=timeout, stream=True ) as r: if not r.ok: _log.error( f"Error calling the API {url} in streaming mode. Response was {r.text}" ) r.raise_for_status() full_text = [] for raw_line in r.iter_lines(decode_unicode=True): if not raw_line: # keep-alives / blank lines continue if not raw_line.startswith("data:"): # Some proxies inject comments; ignore anything not starting with 'data:' continue data = raw_line[len("data:") :].strip() if data == "[DONE]": break try: obj = json.loads(data) except json.JSONDecodeError: _log.debug("Skipping non-JSON SSE chunk: %r", data[:200]) continue # OpenAI-compatible delta format # obj["choices"][0]["delta"]["content"] may be None or missing (e.g., tool calls) try: delta = obj["choices"][0].get("delta") or {} piece = delta.get("content") or "" except (KeyError, IndexError) as e: _log.debug("Unexpected SSE chunk shape: %s", e) piece = "" # Try to extract token count num_tokens = None try: if "usage" in obj: usage = obj["usage"] num_tokens = usage.get("total_tokens") except Exception as e: num_tokens = None _log.debug("Usage key not included in response: %s", e) if piece: full_text.append(piece) for stopper in generation_stoppers: # Respect stopper's lookback window. We use a simple string window which # works with the GenerationStopper interface. lookback = max(1, stopper.lookback_tokens()) window = "".join(full_text)[-lookback:] if stopper.should_stop(window): # Break out of the loop cleanly. The context manager will handle # closing the connection when we exit the 'with' block. # vLLM/OpenAI-compatible servers will detect the client disconnect # and abort the request server-side. return "".join(full_text), num_tokens return "".join(full_text), num_tokens
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/layout_postprocessor.py
docling/utils/layout_postprocessor.py
import bisect import logging import sys from collections import defaultdict 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, Page from docling.datamodel.pipeline_options import LayoutOptions _log = logging.getLogger(__name__) class UnionFind: """Efficient Union-Find data structure for grouping elements.""" def __init__(self, elements): self.parent = {elem: elem for elem in elements} self.rank = dict.fromkeys(elements, 0) def find(self, x): if self.parent[x] != x: self.parent[x] = self.find(self.parent[x]) # Path compression return self.parent[x] def union(self, x, y): root_x, root_y = self.find(x), self.find(y) if root_x == root_y: return if self.rank[root_x] > self.rank[root_y]: self.parent[root_y] = root_x elif self.rank[root_x] < self.rank[root_y]: self.parent[root_x] = root_y else: self.parent[root_y] = root_x self.rank[root_x] += 1 def get_groups(self) -> dict[int, list[int]]: """Returns groups as {root: [elements]}.""" groups = defaultdict(list) for elem in self.parent: groups[self.find(elem)].append(elem) return groups class SpatialClusterIndex: """Efficient spatial indexing for clusters using R-tree and interval trees.""" def __init__(self, clusters: list[Cluster]): p = index.Property() p.dimension = 2 self.spatial_index = index.Index(properties=p) self.x_intervals = IntervalTree() self.y_intervals = IntervalTree() self.clusters_by_id: dict[int, Cluster] = {} for cluster in clusters: self.add_cluster(cluster) def add_cluster(self, cluster: Cluster): bbox = cluster.bbox self.spatial_index.insert(cluster.id, bbox.as_tuple()) self.x_intervals.insert(bbox.l, bbox.r, cluster.id) self.y_intervals.insert(bbox.t, bbox.b, cluster.id) self.clusters_by_id[cluster.id] = cluster def remove_cluster(self, cluster: Cluster): self.spatial_index.delete(cluster.id, cluster.bbox.as_tuple()) del self.clusters_by_id[cluster.id] def find_candidates(self, bbox: BoundingBox) -> set[int]: """Find potential overlapping cluster IDs using all indexes.""" spatial = set(self.spatial_index.intersection(bbox.as_tuple())) x_candidates = self.x_intervals.find_containing( bbox.l ) | self.x_intervals.find_containing(bbox.r) y_candidates = self.y_intervals.find_containing( bbox.t ) | self.y_intervals.find_containing(bbox.b) return spatial.union(x_candidates).union(y_candidates) def check_overlap( self, bbox1: BoundingBox, bbox2: BoundingBox, overlap_threshold: float, containment_threshold: float, ) -> bool: """Check if two bboxes overlap sufficiently.""" if bbox1.area() <= 0 or bbox2.area() <= 0: return False iou = bbox1.intersection_over_union(bbox2) containment1 = bbox1.intersection_over_self(bbox2) containment2 = bbox2.intersection_over_self(bbox1) return ( iou > overlap_threshold or containment1 > containment_threshold or containment2 > containment_threshold ) class Interval: """Helper class for sortable intervals.""" def __init__(self, min_val: float, max_val: float, id: int): self.min_val = min_val self.max_val = max_val self.id = id def __lt__(self, other): if isinstance(other, Interval): return self.min_val < other.min_val return self.min_val < other class IntervalTree: """Memory-efficient interval tree for 1D overlap queries.""" def __init__(self): self.intervals: list[Interval] = [] # Sorted by min_val def insert(self, min_val: float, max_val: float, id: int): interval = Interval(min_val, max_val, id) bisect.insort(self.intervals, interval) def find_containing(self, point: float) -> set[int]: """Find all intervals containing the point.""" pos = bisect.bisect_left(self.intervals, point) result = set() # Check intervals starting before point for interval in reversed(self.intervals[:pos]): if interval.min_val <= point <= interval.max_val: result.add(interval.id) else: break # Check intervals starting at/after point for interval in self.intervals[pos:]: if point <= interval.max_val: if interval.min_val <= point: result.add(interval.id) else: break return result class LayoutPostprocessor: """Postprocesses layout predictions by cleaning up clusters and mapping cells.""" # Cluster type-specific parameters for overlap resolution OVERLAP_PARAMS = { "regular": {"area_threshold": 1.3, "conf_threshold": 0.05}, "picture": {"area_threshold": 2.0, "conf_threshold": 0.3}, "wrapper": {"area_threshold": 2.0, "conf_threshold": 0.2}, } WRAPPER_TYPES = { DocItemLabel.FORM, DocItemLabel.KEY_VALUE_REGION, DocItemLabel.TABLE, DocItemLabel.DOCUMENT_INDEX, } SPECIAL_TYPES = WRAPPER_TYPES.union({DocItemLabel.PICTURE}) CONFIDENCE_THRESHOLDS = { DocItemLabel.CAPTION: 0.5, DocItemLabel.FOOTNOTE: 0.5, DocItemLabel.FORMULA: 0.5, DocItemLabel.LIST_ITEM: 0.5, DocItemLabel.PAGE_FOOTER: 0.5, DocItemLabel.PAGE_HEADER: 0.5, DocItemLabel.PICTURE: 0.5, DocItemLabel.SECTION_HEADER: 0.45, DocItemLabel.TABLE: 0.5, DocItemLabel.TEXT: 0.5, # 0.45, DocItemLabel.TITLE: 0.45, DocItemLabel.CODE: 0.45, DocItemLabel.CHECKBOX_SELECTED: 0.45, DocItemLabel.CHECKBOX_UNSELECTED: 0.45, DocItemLabel.FORM: 0.45, DocItemLabel.KEY_VALUE_REGION: 0.45, DocItemLabel.DOCUMENT_INDEX: 0.45, } LABEL_REMAPPING = { # DocItemLabel.DOCUMENT_INDEX: DocItemLabel.TABLE, DocItemLabel.TITLE: DocItemLabel.SECTION_HEADER, } def __init__( self, page: Page, clusters: list[Cluster], options: LayoutOptions ) -> None: """Initialize processor with page and clusters.""" self.cells = page.cells self.page = page self.page_size = page.size self.all_clusters = clusters self.options = options self.regular_clusters = [ c for c in clusters if c.label not in self.SPECIAL_TYPES ] self.special_clusters = [c for c in clusters if c.label in self.SPECIAL_TYPES] # Build spatial indices once self.regular_index = SpatialClusterIndex(self.regular_clusters) self.picture_index = SpatialClusterIndex( [c for c in self.special_clusters if c.label == DocItemLabel.PICTURE] ) self.wrapper_index = SpatialClusterIndex( [c for c in self.special_clusters if c.label in self.WRAPPER_TYPES] ) def postprocess(self) -> tuple[list[Cluster], list[TextCell]]: """Main processing pipeline.""" self.regular_clusters = self._process_regular_clusters() self.special_clusters = self._process_special_clusters() # Remove regular clusters that are included in wrappers contained_ids = { child.id for wrapper in self.special_clusters if wrapper.label in self.SPECIAL_TYPES for child in wrapper.children } self.regular_clusters = [ c for c in self.regular_clusters if c.id not in contained_ids ] # Combine and sort final clusters final_clusters = self._sort_clusters( self.regular_clusters + self.special_clusters, mode="id" ) # Conditionally process cells if not skipping cell assignment if not self.options.skip_cell_assignment: for cluster in final_clusters: cluster.cells = self._sort_cells(cluster.cells) # Also sort cells in children if any 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]: """Process regular clusters with iterative refinement.""" clusters = [ c for c in self.regular_clusters if c.confidence >= self.CONFIDENCE_THRESHOLDS[c.label] ] # Apply label remapping for cluster in clusters: if cluster.label in self.LABEL_REMAPPING: cluster.label = self.LABEL_REMAPPING[cluster.label] # Conditionally assign cells to clusters if not self.options.skip_cell_assignment: # Initial cell assignment clusters = self._assign_cells_to_clusters(clusters) # Remove clusters with no cells (if keep_empty_clusters is False), # but always keep clusters with label DocItemLabel.FORMULA if not self.options.keep_empty_clusters: clusters = [ cluster for cluster in clusters if cluster.cells or cluster.label == DocItemLabel.FORMULA ] # Handle orphaned cells unassigned = self._find_unassigned_cells(clusters) if unassigned and self.options.create_orphan_clusters: next_id = max((c.id for c in self.all_clusters), default=0) + 1 orphan_clusters = [] for i, cell in enumerate(unassigned): conf = cell.confidence orphan_clusters.append( Cluster( id=next_id + i, label=DocItemLabel.TEXT, bbox=cell.to_bounding_box(), confidence=conf, cells=[cell], ) ) clusters.extend(orphan_clusters) # Iterative refinement prev_count = len(clusters) + 1 for _ in range(3): # Maximum 3 iterations if prev_count == len(clusters): break prev_count = len(clusters) clusters = self._adjust_cluster_bboxes(clusters) clusters = self._remove_overlapping_clusters(clusters, "regular") return clusters def _process_special_clusters(self) -> list[Cluster]: special_clusters = [ c for c in self.special_clusters if c.confidence >= self.CONFIDENCE_THRESHOLDS[c.label] ] 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 special_clusters = [ cluster for cluster in special_clusters if not ( cluster.label == DocItemLabel.PICTURE and cluster.bbox.area() / page_area > 0.90 ) ] for special in special_clusters: contained = [] for cluster in self.regular_clusters: containment = cluster.bbox.intersection_over_self(special.bbox) if containment > 0.8: contained.append(cluster) if contained: # Sort contained clusters by minimum cell ID: contained = self._sort_clusters(contained, mode="id") special.children = contained # Adjust bbox only for Form and Key-Value-Region, not Table or Picture if special.label in [DocItemLabel.FORM, DocItemLabel.KEY_VALUE_REGION]: special.bbox = BoundingBox( l=min(c.bbox.l for c in contained), t=min(c.bbox.t for c in contained), r=max(c.bbox.r for c in contained), b=max(c.bbox.b for c in contained), ) # Conditionally collect cells from children if not self.options.skip_cell_assignment: all_cells = [] for child in contained: all_cells.extend(child.cells) special.cells = self._deduplicate_cells(all_cells) special.cells = self._sort_cells(special.cells) else: special.cells = [] picture_clusters = [ c for c in special_clusters if c.label == DocItemLabel.PICTURE ] picture_clusters = self._remove_overlapping_clusters( picture_clusters, "picture" ) wrapper_clusters = [ c for c in special_clusters if c.label in self.WRAPPER_TYPES ] wrapper_clusters = self._remove_overlapping_clusters( wrapper_clusters, "wrapper" ) return picture_clusters + wrapper_clusters def _handle_cross_type_overlaps(self, special_clusters) -> list[Cluster]: """Handle overlaps between regular and wrapper clusters before child assignment. In particular, KEY_VALUE_REGION proposals that are almost identical to a TABLE should be removed. """ wrappers_to_remove = set() for wrapper in special_clusters: if wrapper.label not in self.WRAPPER_TYPES: continue # only treat KEY_VALUE_REGION for now. for regular in self.regular_clusters: if regular.label == DocItemLabel.TABLE: # Calculate overlap overlap_ratio = wrapper.bbox.intersection_over_self(regular.bbox) conf_diff = wrapper.confidence - regular.confidence # If wrapper is mostly overlapping with a TABLE, remove the wrapper if ( overlap_ratio > 0.9 and conf_diff < 0.1 ): # self.OVERLAP_PARAMS["wrapper"]["conf_threshold"]): # 80% overlap threshold wrappers_to_remove.add(wrapper.id) break # Filter out the identified wrappers special_clusters = [ cluster for cluster in special_clusters if cluster.id not in wrappers_to_remove ] return special_clusters def _should_prefer_cluster( self, candidate: Cluster, other: Cluster, params: dict ) -> bool: """Determine if candidate cluster should be preferred over other cluster based on rules. Returns True if candidate should be preferred, False if not.""" # Rule 1: LIST_ITEM vs TEXT if ( candidate.label == DocItemLabel.LIST_ITEM and other.label == DocItemLabel.TEXT ): # Check if areas are similar (within 20% of each other) area_ratio = candidate.bbox.area() / other.bbox.area() area_similarity = abs(1 - area_ratio) < 0.2 if area_similarity: return True # Rule 2: CODE vs others if candidate.label == DocItemLabel.CODE: # Calculate how much of the other cluster is contained within the CODE cluster containment = other.bbox.intersection_over_self(candidate.bbox) if containment > 0.8: # other is 80% contained within CODE return True # If no label-based rules matched, fall back to area/confidence thresholds area_ratio = candidate.bbox.area() / other.bbox.area() conf_diff = other.confidence - candidate.confidence if ( area_ratio <= params["area_threshold"] and conf_diff > params["conf_threshold"] ): return False return True # Default to keeping candidate if no rules triggered rejection def _select_best_cluster_from_group( self, group_clusters: list[Cluster], params: dict, ) -> Cluster: """Select best cluster from a group of overlapping clusters based on all rules.""" current_best = None for candidate in group_clusters: should_select = True for other in group_clusters: if other == candidate: continue if not self._should_prefer_cluster(candidate, other, params): should_select = False break if should_select: if current_best is None: current_best = candidate else: # If both clusters pass rules, prefer the larger one unless confidence differs significantly if ( candidate.bbox.area() > current_best.bbox.area() and current_best.confidence - candidate.confidence <= params["conf_threshold"] ): current_best = candidate return current_best if current_best else group_clusters[0] def _remove_overlapping_clusters( self, clusters: list[Cluster], cluster_type: str, overlap_threshold: float = 0.8, containment_threshold: float = 0.8, ) -> list[Cluster]: if not clusters: return [] spatial_index = ( self.regular_index if cluster_type == "regular" else self.picture_index if cluster_type == "picture" else self.wrapper_index ) # Map of currently valid clusters valid_clusters = {c.id: c for c in clusters} uf = UnionFind(valid_clusters.keys()) params = self.OVERLAP_PARAMS[cluster_type] for cluster in clusters: candidates = spatial_index.find_candidates(cluster.bbox) candidates &= valid_clusters.keys() # Only keep existing candidates candidates.discard(cluster.id) for other_id in candidates: if spatial_index.check_overlap( cluster.bbox, valid_clusters[other_id].bbox, overlap_threshold, containment_threshold, ): uf.union(cluster.id, other_id) result = [] for group in uf.get_groups().values(): if len(group) == 1: result.append(valid_clusters[group[0]]) continue group_clusters = [valid_clusters[cid] for cid in group] best = self._select_best_cluster_from_group(group_clusters, params) # Simple cell merging - no special cases for cluster in group_clusters: if cluster != best: best.cells.extend(cluster.cells) best.cells = self._deduplicate_cells(best.cells) best.cells = self._sort_cells(best.cells) result.append(best) return result def _select_best_cluster( self, clusters: list[Cluster], area_threshold: float, conf_threshold: float, ) -> Cluster: """Iteratively select best cluster based on area and confidence thresholds.""" current_best = None for candidate in clusters: should_select = True for other in clusters: if other == candidate: continue area_ratio = candidate.bbox.area() / other.bbox.area() conf_diff = other.confidence - candidate.confidence if area_ratio <= area_threshold and conf_diff > conf_threshold: should_select = False break if should_select: if current_best is None or ( candidate.bbox.area() > current_best.bbox.area() and current_best.confidence - candidate.confidence <= conf_threshold ): current_best = candidate return current_best if current_best else clusters[0] def _deduplicate_cells(self, cells: list[TextCell]) -> list[TextCell]: """Ensure each cell appears only once, maintaining order of first appearance.""" seen_ids = set() unique_cells = [] for cell in cells: if cell.index not in seen_ids: seen_ids.add(cell.index) unique_cells.append(cell) return unique_cells def _assign_cells_to_clusters( self, clusters: list[Cluster], min_overlap: float = 0.2 ) -> list[Cluster]: """Assign cells to best overlapping cluster.""" for cluster in clusters: cluster.cells = [] for cell in self.cells: if not cell.text.strip(): continue best_overlap = min_overlap best_cluster = None for cluster in clusters: if cell.rect.to_bounding_box().area() <= 0: continue overlap_ratio = cell.rect.to_bounding_box().intersection_over_self( cluster.bbox ) if overlap_ratio > best_overlap: best_overlap = overlap_ratio best_cluster = cluster if best_cluster is not None: best_cluster.cells.append(cell) # Deduplicate cells in each cluster after assignment for cluster in clusters: cluster.cells = self._deduplicate_cells(cluster.cells) return clusters def _find_unassigned_cells(self, clusters: list[Cluster]) -> list[TextCell]: """Find cells not assigned to any cluster.""" assigned = {cell.index for cluster in clusters for cell in cluster.cells} return [ cell for cell in self.cells if cell.index not in assigned and cell.text.strip() ] def _adjust_cluster_bboxes(self, clusters: list[Cluster]) -> list[Cluster]: """Adjust cluster bounding boxes to contain their cells.""" for cluster in clusters: if not cluster.cells: continue cells_bbox = BoundingBox( l=min(cell.rect.to_bounding_box().l for cell in cluster.cells), t=min(cell.rect.to_bounding_box().t for cell in cluster.cells), r=max(cell.rect.to_bounding_box().r for cell in cluster.cells), b=max(cell.rect.to_bounding_box().b for cell in cluster.cells), ) if cluster.label == DocItemLabel.TABLE: # For tables, take union of current bbox and cells bbox cluster.bbox = BoundingBox( l=min(cluster.bbox.l, cells_bbox.l), t=min(cluster.bbox.t, cells_bbox.t), r=max(cluster.bbox.r, cells_bbox.r), b=max(cluster.bbox.b, cells_bbox.b), ) else: cluster.bbox = cells_bbox return clusters def _sort_cells(self, cells: list[TextCell]) -> list[TextCell]: """Sort cells in native reading order.""" return sorted(cells, key=lambda c: (c.index)) def _sort_clusters( self, clusters: list[Cluster], mode: str = "id" ) -> list[Cluster]: """Sort clusters in reading order (top-to-bottom, left-to-right).""" if mode == "id": # sort in the order the cells are printed in the PDF. return sorted( clusters, key=lambda cluster: ( ( min(cell.index for cell in cluster.cells) if cluster.cells else sys.maxsize ), cluster.bbox.t, cluster.bbox.l, ), ) elif mode == "tblr": # Sort top-to-bottom, then left-to-right ("row first") return sorted( clusters, key=lambda cluster: (cluster.bbox.t, cluster.bbox.l) ) elif mode == "lrtb": # Sort left-to-right, then top-to-bottom ("column first") return sorted( clusters, key=lambda cluster: (cluster.bbox.l, cluster.bbox.t) ) else: return clusters
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/export.py
docling/utils/export.py
import logging from collections.abc import Iterable from typing import Any, Dict, List, Tuple, Union from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.legacy_doc.base import BaseCell, BaseText, Ref, Table from docling.datamodel.document import ConversionResult, Page _log = logging.getLogger(__name__) def generate_multimodal_pages( doc_result: ConversionResult, ) -> Iterable[Tuple[str, str, List[Dict[str, Any]], List[Dict[str, Any]], Page]]: label_to_doclaynet = { "title": "title", "table-of-contents": "document_index", "subtitle-level-1": "section_header", "checkbox-selected": "checkbox_selected", "checkbox-unselected": "checkbox_unselected", "caption": "caption", "page-header": "page_header", "page-footer": "page_footer", "footnote": "footnote", "table": "table", "formula": "formula", "list-item": "list_item", "code": "code", "figure": "picture", "picture": "picture", "reference": "text", "paragraph": "text", "text": "text", } content_text = "" page_no = 0 start_ix = 0 end_ix = 0 doc_items: List[Tuple[int, Union[BaseCell, BaseText]]] = [] doc = doc_result.legacy_document def _process_page_segments(doc_items: list[Tuple[int, BaseCell]], page: Page): segments = [] for ix, item in doc_items: item_type = item.obj_type label = label_to_doclaynet.get(item_type, None) if label is None or item.prov is None or page.size is None: continue bbox = BoundingBox.from_tuple( tuple(item.prov[0].bbox), origin=CoordOrigin.BOTTOMLEFT ) new_bbox = bbox.to_top_left_origin(page_height=page.size.height).normalized( page_size=page.size ) new_segment = { "index_in_doc": ix, "label": label, "text": item.text if item.text is not None else "", "bbox": new_bbox.as_tuple(), "data": [], } if isinstance(item, Table): table_html = item.export_to_html() new_segment["data"].append( { "html_seq": table_html, "otsl_seq": "", } ) segments.append(new_segment) return segments def _process_page_cells(page: Page): cells: List[dict] = [] if page.size is None: return cells for cell in page.cells: new_bbox = ( cell.rect.to_bounding_box() .to_top_left_origin(page_height=page.size.height) .normalized(page_size=page.size) ) is_ocr = cell.from_ocr ocr_confidence = cell.confidence cells.append( { "text": cell.text, "bbox": new_bbox.as_tuple(), "ocr": is_ocr, "ocr_confidence": ocr_confidence, } ) return cells def _process_page(): page_ix = page_no - 1 page = doc_result.pages[page_ix] page_cells = _process_page_cells(page=page) page_segments = _process_page_segments(doc_items=doc_items, page=page) content_md = doc.export_to_markdown( main_text_start=start_ix, main_text_stop=end_ix ) # No page-tagging since we only do 1 page at the time content_dt = doc.export_to_document_tokens( main_text_start=start_ix, main_text_stop=end_ix, add_page_index=False ) return content_text, content_md, content_dt, page_cells, page_segments, page if doc.main_text is None: return for ix, orig_item in enumerate(doc.main_text): item = doc._resolve_ref(orig_item) if isinstance(orig_item, Ref) else orig_item if item is None or item.prov is None or len(item.prov) == 0: _log.debug(f"Skipping item {orig_item}") continue item_page = item.prov[0].page # Page is complete if page_no > 0 and item_page > page_no: yield _process_page() start_ix = ix doc_items = [] content_text = "" page_no = item_page end_ix = ix doc_items.append((ix, item)) if item.text is not None and item.text != "": content_text += item.text + " " if len(doc_items) > 0: yield _process_page()
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/accelerator_utils.py
docling/utils/accelerator_utils.py
import logging from typing import List, Optional from docling.datamodel.accelerator_options import AcceleratorDevice _log = logging.getLogger(__name__) def decide_device( accelerator_device: str, supported_devices: Optional[List[AcceleratorDevice]] = None ) -> str: r""" Resolve the device based on the acceleration options and the available devices in the system. Rules: 1. AUTO: Check for the best available device on the system. 2. User-defined: Check if the device actually exists, otherwise fall-back to CPU """ import torch device = "cpu" has_cuda = torch.backends.cuda.is_built() and torch.cuda.is_available() has_mps = torch.backends.mps.is_built() and torch.backends.mps.is_available() if supported_devices is not None: if has_cuda and AcceleratorDevice.CUDA not in supported_devices: _log.info( f"Removing CUDA from available devices because it is not in {supported_devices=}" ) has_cuda = False if has_mps and AcceleratorDevice.MPS not in supported_devices: _log.info( f"Removing MPS from available devices because it is not in {supported_devices=}" ) has_mps = False if accelerator_device == AcceleratorDevice.AUTO.value: # Handle 'auto' if has_cuda: device = "cuda:0" elif has_mps: device = "mps" elif accelerator_device.startswith("cuda"): if has_cuda: # if cuda device index specified extract device id parts = accelerator_device.split(":") if len(parts) == 2 and parts[1].isdigit(): # select cuda device's id cuda_index = int(parts[1]) if cuda_index < torch.cuda.device_count(): device = f"cuda:{cuda_index}" else: _log.warning( "CUDA device 'cuda:%d' is not available. Fall back to 'CPU'.", cuda_index, ) elif len(parts) == 1: # just "cuda" device = "cuda:0" else: _log.warning( "Invalid CUDA device format '%s'. Fall back to 'CPU'", accelerator_device, ) else: _log.warning("CUDA is not available in the system. Fall back to 'CPU'") elif accelerator_device == AcceleratorDevice.MPS.value: if has_mps: device = "mps" else: _log.warning("MPS is not available in the system. Fall back to 'CPU'") elif accelerator_device == AcceleratorDevice.CPU.value: device = "cpu" else: _log.warning( "Unknown device option '%s'. Fall back to 'CPU'", accelerator_device ) _log.info("Accelerator device: '%s'", device) return device
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/glm_utils.py
docling/utils/glm_utils.py
import re from pathlib import Path from typing import List import pandas as pd from docling_core.types.doc import ( BoundingBox, CoordOrigin, DocItemLabel, DoclingDocument, DocumentOrigin, GroupLabel, ProvenanceItem, Size, TableCell, TableData, ) from docling_core.types.doc.document import ContentLayer def resolve_item(paths, obj): """Find item in document from a reference path""" if len(paths) == 0: return obj if paths[0] == "#": return resolve_item(paths[1:], obj) try: key = int(paths[0]) except Exception: key = paths[0] if len(paths) == 1: if isinstance(key, str) and key in obj: return obj[key] elif isinstance(key, int) and key < len(obj): return obj[key] else: return None elif len(paths) > 1: if isinstance(key, str) and key in obj: return resolve_item(paths[1:], obj[key]) elif isinstance(key, int) and key < len(obj): return resolve_item(paths[1:], obj[key]) else: return None else: return None def _flatten_table_grid(grid: List[List[dict]]) -> List[dict]: unique_objects = [] seen_spans = set() for sublist in grid: for obj in sublist: # Convert the spans list to a tuple of tuples for hashing spans_tuple = tuple(tuple(span) for span in obj["spans"]) if spans_tuple not in seen_spans: seen_spans.add(spans_tuple) unique_objects.append(obj) return unique_objects def to_docling_document(doc_glm, update_name_label=False) -> DoclingDocument: # noqa: C901 origin = DocumentOrigin( mimetype="application/pdf", filename=doc_glm["file-info"]["filename"], binary_hash=doc_glm["file-info"]["document-hash"], ) doc_name = Path(origin.filename).stem doc: DoclingDocument = DoclingDocument(name=doc_name, origin=origin) for page_dim in doc_glm["page-dimensions"]: page_no = int(page_dim["page"]) size = Size(width=page_dim["width"], height=page_dim["height"]) doc.add_page(page_no=page_no, size=size) if "properties" in doc_glm: props = pd.DataFrame( doc_glm["properties"]["data"], columns=doc_glm["properties"]["headers"] ) else: props = pd.DataFrame() current_list = None for ix, pelem in enumerate(doc_glm["page-elements"]): ptype = pelem["type"] span_i = pelem["span"][0] span_j = pelem["span"][1] if "iref" not in pelem: # print(json.dumps(pelem, indent=2)) continue iref = pelem["iref"] if re.match("#/figures/(\\d+)/captions/(.+)", iref): # print(f"skip {iref}") continue if re.match("#/tables/(\\d+)/captions/(.+)", iref): # print(f"skip {iref}") continue path = iref.split("/") obj = resolve_item(path, doc_glm) if obj is None: current_list = None print(f"warning: undefined {path}") continue if ptype == "figure": current_list = None text = "" caption_refs = [] for caption in obj["captions"]: text += caption["text"] for nprov in caption["prov"]: npaths = nprov["$ref"].split("/") nelem = resolve_item(npaths, doc_glm) if nelem is None: # print(f"warning: undefined caption {npaths}") continue span_i = nelem["span"][0] span_j = nelem["span"][1] cap_text = caption["text"][span_i:span_j] # doc_glm["page-elements"].remove(nelem) prov = ProvenanceItem( page_no=nelem["page"], charspan=tuple(nelem["span"]), bbox=BoundingBox.from_tuple( nelem["bbox"], origin=CoordOrigin.BOTTOMLEFT ), ) caption_obj = doc.add_text( label=DocItemLabel.CAPTION, text=cap_text, prov=prov ) caption_refs.append(caption_obj.get_ref()) prov = ProvenanceItem( page_no=pelem["page"], charspan=(0, len(text)), bbox=BoundingBox.from_tuple( pelem["bbox"], origin=CoordOrigin.BOTTOMLEFT ), ) pic = doc.add_picture(prov=prov) pic.captions.extend(caption_refs) _add_child_elements(pic, doc, obj, pelem) elif ptype == "table": current_list = None text = "" caption_refs = [] item_label = DocItemLabel(pelem["name"]) for caption in obj["captions"]: text += caption["text"] for nprov in caption["prov"]: npaths = nprov["$ref"].split("/") nelem = resolve_item(npaths, doc_glm) if nelem is None: # print(f"warning: undefined caption {npaths}") continue span_i = nelem["span"][0] span_j = nelem["span"][1] cap_text = caption["text"][span_i:span_j] # doc_glm["page-elements"].remove(nelem) prov = ProvenanceItem( page_no=nelem["page"], charspan=tuple(nelem["span"]), bbox=BoundingBox.from_tuple( nelem["bbox"], origin=CoordOrigin.BOTTOMLEFT ), ) caption_obj = doc.add_text( label=DocItemLabel.CAPTION, text=cap_text, prov=prov ) caption_refs.append(caption_obj.get_ref()) table_cells_glm = _flatten_table_grid(obj["data"]) table_cells = [] for tbl_cell_glm in table_cells_glm: if tbl_cell_glm["bbox"] is not None: bbox = BoundingBox.from_tuple( tbl_cell_glm["bbox"], origin=CoordOrigin.BOTTOMLEFT ) else: bbox = None is_col_header = False is_row_header = False is_row_section = False if tbl_cell_glm["type"] == "col_header": is_col_header = True elif tbl_cell_glm["type"] == "row_header": is_row_header = True elif tbl_cell_glm["type"] == "row_section": is_row_section = True table_cells.append( TableCell( row_span=tbl_cell_glm["row-span"][1] - tbl_cell_glm["row-span"][0], col_span=tbl_cell_glm["col-span"][1] - tbl_cell_glm["col-span"][0], start_row_offset_idx=tbl_cell_glm["row-span"][0], end_row_offset_idx=tbl_cell_glm["row-span"][1], start_col_offset_idx=tbl_cell_glm["col-span"][0], end_col_offset_idx=tbl_cell_glm["col-span"][1], text=tbl_cell_glm["text"], bbox=bbox, column_header=is_col_header, row_header=is_row_header, row_section=is_row_section, ) ) tbl_data = TableData( num_rows=obj.get("#-rows", 0), num_cols=obj.get("#-cols", 0), table_cells=table_cells, ) prov = ProvenanceItem( page_no=pelem["page"], charspan=(0, 0), bbox=BoundingBox.from_tuple( pelem["bbox"], origin=CoordOrigin.BOTTOMLEFT ), ) tbl = doc.add_table(data=tbl_data, prov=prov, label=item_label) tbl.captions.extend(caption_refs) elif ptype in [DocItemLabel.FORM.value, DocItemLabel.KEY_VALUE_REGION.value]: label = DocItemLabel(ptype) group_label = GroupLabel.UNSPECIFIED if label == DocItemLabel.FORM: group_label = GroupLabel.FORM_AREA elif label == DocItemLabel.KEY_VALUE_REGION: group_label = GroupLabel.KEY_VALUE_AREA container_el = doc.add_group(label=group_label) _add_child_elements(container_el, doc, obj, pelem) elif "text" in obj: text = obj["text"][span_i:span_j] type_label = pelem["type"] name_label = pelem["name"] if update_name_label and len(props) > 0 and type_label == "paragraph": prop = props[ (props["type"] == "semantic") & (props["subj_path"] == iref) ] if len(prop) == 1 and prop.iloc[0]["confidence"] > 0.85: name_label = prop.iloc[0]["label"] prov = ProvenanceItem( page_no=pelem["page"], charspan=(0, len(text)), bbox=BoundingBox.from_tuple( pelem["bbox"], origin=CoordOrigin.BOTTOMLEFT ), ) label = DocItemLabel(name_label) if label == DocItemLabel.LIST_ITEM: if current_list is None: current_list = doc.add_group(label=GroupLabel.LIST, name="list") # TODO: Infer if this is a numbered or a bullet list item doc.add_list_item( text=text, enumerated=False, prov=prov, parent=current_list ) elif label == DocItemLabel.SECTION_HEADER: current_list = None doc.add_heading(text=text, prov=prov) elif label == DocItemLabel.CODE: current_list = None doc.add_code(text=text, prov=prov) elif label == DocItemLabel.FORMULA: current_list = None doc.add_text(label=DocItemLabel.FORMULA, text="", orig=text, prov=prov) elif label in [DocItemLabel.PAGE_HEADER, DocItemLabel.PAGE_FOOTER]: current_list = None doc.add_text( label=DocItemLabel(name_label), text=text, prov=prov, content_layer=ContentLayer.FURNITURE, ) else: current_list = None doc.add_text(label=DocItemLabel(name_label), text=text, prov=prov) return doc def _add_child_elements(container_el, doc, obj, pelem): payload = obj.get("payload") if payload is not None: children = payload.get("children", []) for child in children: c_label = DocItemLabel(child["label"]) c_bbox = BoundingBox.model_validate(child["bbox"]).to_bottom_left_origin( doc.pages[pelem["page"]].size.height ) c_text = " ".join( [ cell["text"].replace("\x02", "-").strip() for cell in child["cells"] if len(cell["text"].strip()) > 0 ] ) c_prov = ProvenanceItem( page_no=pelem["page"], charspan=(0, len(c_text)), bbox=c_bbox ) if c_label == DocItemLabel.LIST_ITEM: # TODO: Infer if this is a numbered or a bullet list item doc.add_list_item(parent=container_el, text=c_text, prov=c_prov) elif c_label == DocItemLabel.SECTION_HEADER: doc.add_heading(parent=container_el, text=c_text, prov=c_prov) else: doc.add_text( parent=container_el, label=c_label, text=c_text, prov=c_prov )
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/ocr_utils.py
docling/utils/ocr_utils.py
from typing import Optional, Tuple from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.doc.page import BoundingRectangle from docling.utils.orientation import CLIPPED_ORIENTATIONS, rotate_bounding_box def map_tesseract_script(script: str) -> str: r""" """ if script == "Katakana" or script == "Hiragana": script = "Japanese" elif script == "Han": script = "HanS" elif script == "Korean": script = "Hangul" return script def parse_tesseract_orientation(orientation: str) -> int: # Tesseract orientation is [0, 90, 180, 270] clockwise, bounding rectangle angles # are [0, 360[ counterclockwise parsed = int(orientation) if parsed not in CLIPPED_ORIENTATIONS: msg = ( f"invalid tesseract document orientation {orientation}, " f"expected orientation: {sorted(CLIPPED_ORIENTATIONS)}" ) raise ValueError(msg) parsed = -parsed parsed %= 360 return parsed def tesseract_box_to_bounding_rectangle( bbox: BoundingBox, *, original_offset: Optional[BoundingBox] = None, scale: float, orientation: int, im_size: Tuple[int, int], ) -> BoundingRectangle: # box is in the top, left, height, width format, top left coordinates rect = rotate_bounding_box(bbox, angle=orientation, im_size=im_size) rect = BoundingRectangle( r_x0=rect.r_x0 / scale, r_y0=rect.r_y0 / scale, r_x1=rect.r_x1 / scale, r_y1=rect.r_y1 / scale, r_x2=rect.r_x2 / scale, r_y2=rect.r_y2 / scale, r_x3=rect.r_x3 / scale, r_y3=rect.r_y3 / scale, coord_origin=CoordOrigin.TOPLEFT, ) if original_offset is not None: if original_offset.coord_origin is not CoordOrigin.TOPLEFT: msg = f"expected coordinate origin to be {CoordOrigin.TOPLEFT.value}" raise ValueError(msg) if original_offset is not None: rect.r_x0 += original_offset.l rect.r_x1 += original_offset.l rect.r_x2 += original_offset.l rect.r_x3 += original_offset.l rect.r_y0 += original_offset.t rect.r_y1 += original_offset.t rect.r_y2 += original_offset.t rect.r_y3 += original_offset.t return rect
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/utils.py
docling/utils/utils.py
import hashlib from io import BytesIO from itertools import islice from pathlib import Path from typing import List, Union import requests from tqdm import tqdm def chunkify(iterator, chunk_size): """Yield successive chunks of chunk_size from the iterable.""" if isinstance(iterator, List): iterator = iter(iterator) for first in iterator: # Take the first element from the iterator yield [first, *list(islice(iterator, chunk_size - 1))] def create_file_hash(path_or_stream: Union[BytesIO, Path]) -> str: """Create a stable page_hash of the path_or_stream of a file""" block_size = 65536 hasher = hashlib.sha256(usedforsecurity=False) def _hash_buf(binary_stream): buf = binary_stream.read(block_size) # read and page_hash in chunks while len(buf) > 0: hasher.update(buf) buf = binary_stream.read(block_size) if isinstance(path_or_stream, Path): with path_or_stream.open("rb") as afile: _hash_buf(afile) elif isinstance(path_or_stream, BytesIO): _hash_buf(path_or_stream) return hasher.hexdigest() def create_hash(string: str): hasher = hashlib.sha256(usedforsecurity=False) hasher.update(string.encode("utf-8")) return hasher.hexdigest() def download_url_with_progress(url: str, progress: bool = False) -> BytesIO: buf = BytesIO() with requests.get(url, stream=True, allow_redirects=True) as response: total_size = int(response.headers.get("content-length", 0)) progress_bar = tqdm( total=total_size, unit="B", unit_scale=True, unit_divisor=1024, disable=(not progress), ) for chunk in response.iter_content(10 * 1024): buf.write(chunk) progress_bar.update(len(chunk)) progress_bar.close() buf.seek(0) return buf
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/model_downloader.py
docling/utils/model_downloader.py
import logging from pathlib import Path from typing import Optional from docling.datamodel.layout_model_specs import DOCLING_LAYOUT_V2 from docling.datamodel.pipeline_options import ( LayoutOptions, granite_picture_description, smolvlm_picture_description, ) from docling.datamodel.settings import settings from docling.datamodel.vlm_model_specs import ( GRANITEDOCLING_MLX, GRANITEDOCLING_TRANSFORMERS, SMOLDOCLING_MLX, SMOLDOCLING_TRANSFORMERS, ) from docling.models.code_formula_model import CodeFormulaModel from docling.models.document_picture_classifier import DocumentPictureClassifier from docling.models.easyocr_model import EasyOcrModel from docling.models.layout_model import LayoutModel from docling.models.picture_description_vlm_model import PictureDescriptionVlmModel from docling.models.rapid_ocr_model import RapidOcrModel from docling.models.table_structure_model import TableStructureModel from docling.models.utils.hf_model_download import download_hf_model _log = logging.getLogger(__name__) def download_models( output_dir: Optional[Path] = None, *, force: bool = False, progress: bool = False, with_layout: bool = True, with_tableformer: bool = True, with_code_formula: bool = True, with_picture_classifier: bool = True, with_smolvlm: bool = False, with_granitedocling: bool = False, with_granitedocling_mlx: bool = False, with_smoldocling: bool = False, with_smoldocling_mlx: bool = False, with_granite_vision: bool = False, with_rapidocr: bool = True, with_easyocr: bool = False, ): if output_dir is None: output_dir = settings.cache_dir / "models" # Make sure the folder exists output_dir.mkdir(exist_ok=True, parents=True) if with_layout: _log.info("Downloading layout model...") LayoutModel.download_models( local_dir=output_dir / LayoutOptions().model_spec.model_repo_folder, force=force, progress=progress, ) if with_tableformer: _log.info("Downloading tableformer model...") TableStructureModel.download_models( local_dir=output_dir / TableStructureModel._model_repo_folder, force=force, progress=progress, ) if with_picture_classifier: _log.info("Downloading picture classifier model...") DocumentPictureClassifier.download_models( local_dir=output_dir / DocumentPictureClassifier._model_repo_folder, force=force, progress=progress, ) if with_code_formula: _log.info("Downloading code formula model...") CodeFormulaModel.download_models( local_dir=output_dir / CodeFormulaModel._model_repo_folder, force=force, progress=progress, ) if with_smolvlm: _log.info("Downloading SmolVlm model...") download_hf_model( repo_id=smolvlm_picture_description.repo_id, local_dir=output_dir / smolvlm_picture_description.repo_cache_folder, force=force, progress=progress, ) if with_granitedocling: _log.info("Downloading GraniteDocling model...") download_hf_model( repo_id=GRANITEDOCLING_TRANSFORMERS.repo_id, local_dir=output_dir / GRANITEDOCLING_TRANSFORMERS.repo_cache_folder, force=force, progress=progress, ) if with_granitedocling_mlx: _log.info("Downloading GraniteDocling MLX model...") download_hf_model( repo_id=GRANITEDOCLING_MLX.repo_id, local_dir=output_dir / GRANITEDOCLING_MLX.repo_cache_folder, force=force, progress=progress, ) if with_smoldocling: _log.info("Downloading SmolDocling model...") download_hf_model( repo_id=SMOLDOCLING_TRANSFORMERS.repo_id, local_dir=output_dir / SMOLDOCLING_TRANSFORMERS.repo_cache_folder, force=force, progress=progress, ) if with_smoldocling_mlx: _log.info("Downloading SmolDocling MLX model...") download_hf_model( repo_id=SMOLDOCLING_MLX.repo_id, local_dir=output_dir / SMOLDOCLING_MLX.repo_cache_folder, force=force, progress=progress, ) if with_granite_vision: _log.info("Downloading Granite Vision model...") download_hf_model( repo_id=granite_picture_description.repo_id, local_dir=output_dir / granite_picture_description.repo_cache_folder, force=force, progress=progress, ) if with_rapidocr: for backend in ("torch", "onnxruntime"): _log.info(f"Downloading rapidocr {backend} models...") RapidOcrModel.download_models( backend=backend, local_dir=output_dir / RapidOcrModel._model_repo_folder, force=force, progress=progress, ) if with_easyocr: _log.info("Downloading easyocr models...") EasyOcrModel.download_models( local_dir=output_dir / EasyOcrModel._model_repo_folder, force=force, progress=progress, ) return output_dir
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/__init__.py
docling/utils/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/locks.py
docling/utils/locks.py
import threading pypdfium2_lock = threading.Lock()
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/profiling.py
docling/utils/profiling.py
import time from datetime import datetime from enum import Enum from typing import TYPE_CHECKING, List import numpy as np from pydantic import BaseModel from docling.datamodel.settings import settings if TYPE_CHECKING: from docling.datamodel.document import ConversionResult class ProfilingScope(str, Enum): PAGE = "page" DOCUMENT = "document" class ProfilingItem(BaseModel): scope: ProfilingScope count: int = 0 times: List[float] = [] start_timestamps: List[datetime] = [] def avg(self) -> float: return np.average(self.times) # type: ignore def std(self) -> float: return np.std(self.times) # type: ignore def mean(self) -> float: return np.mean(self.times) # type: ignore def percentile(self, perc: float) -> float: return np.percentile(self.times, perc) # type: ignore class TimeRecorder: def __init__( self, conv_res: "ConversionResult", key: str, scope: ProfilingScope = ProfilingScope.PAGE, ): if settings.debug.profile_pipeline_timings: if key not in conv_res.timings.keys(): conv_res.timings[key] = ProfilingItem(scope=scope) self.conv_res = conv_res self.key = key def __enter__(self): if settings.debug.profile_pipeline_timings: self.start = time.monotonic() self.conv_res.timings[self.key].start_timestamps.append(datetime.utcnow()) return self def __exit__(self, *args): if settings.debug.profile_pipeline_timings: elapsed = time.monotonic() - self.start self.conv_res.timings[self.key].times.append(elapsed) self.conv_res.timings[self.key].count += 1
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/orientation.py
docling/utils/orientation.py
from typing import Tuple from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.doc.page import BoundingRectangle CLIPPED_ORIENTATIONS = [0, 90, 180, 270] def rotate_bounding_box( bbox: BoundingBox, angle: int, im_size: Tuple[int, int] ) -> BoundingRectangle: # The box is left top width height in TOPLEFT coordinates # Bounding rectangle start with r_0 at the bottom left whatever the # coordinate system. Then other corners are found rotating counterclockwise bbox = bbox.to_top_left_origin(im_size[1]) left, top, width, height = bbox.l, bbox.t, bbox.width, bbox.height im_w, im_h = im_size angle = angle % 360 if angle == 0: return BoundingRectangle.from_bounding_box(bbox) elif angle == 90: r_x0 = top + height r_y0 = im_w - left r_x1 = r_x0 r_y1 = r_y0 - width r_x2 = r_x1 - height r_y2 = r_y1 r_x3 = r_x2 r_y3 = r_y0 elif angle == 180: r_x0 = im_w - left r_y0 = im_h - (top + height) r_x1 = r_x0 - width r_y1 = r_y0 r_x2 = r_x1 r_y2 = r_y1 + height r_x3 = r_x0 r_y3 = r_y2 elif angle == 270: r_x0 = im_h - (top + height) r_y0 = left r_x1 = r_x0 r_y1 = r_y0 + width r_x2 = r_x1 + height r_y2 = r_y1 r_x3 = r_x2 r_y3 = r_y0 else: msg = ( f"invalid orientation {angle}, expected values in:" f" {sorted(CLIPPED_ORIENTATIONS)}" ) raise ValueError(msg) rectangle = BoundingRectangle( r_x0=r_x0, r_y0=r_y0, r_x1=r_x1, r_y1=r_y1, r_x2=r_x2, r_y2=r_y2, r_x3=r_x3, r_y3=r_y3, coord_origin=CoordOrigin.TOPLEFT, ) return rectangle
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/utils/visualization.py
docling/utils/visualization.py
from docling_core.types.doc import DocItemLabel from PIL import Image, ImageDraw, ImageFont from PIL.ImageFont import FreeTypeFont from docling.datamodel.base_models import Cluster def draw_clusters( image: Image.Image, clusters: list[Cluster], scale_x: float, scale_y: float ) -> None: """ Draw clusters on an image """ draw = ImageDraw.Draw(image, "RGBA") # Create a smaller font for the labels font: ImageFont.ImageFont | FreeTypeFont try: font = ImageFont.truetype("arial.ttf", 12) except OSError: # Fallback to default font if arial is not available font = ImageFont.load_default() for c_tl in clusters: all_clusters = [c_tl, *c_tl.children] for c in all_clusters: # Draw cells first (underneath) cell_color = (0, 0, 0, 40) # Transparent black for cells for tc in c.cells: cx0, cy0, cx1, cy1 = tc.rect.to_bounding_box().as_tuple() cx0 *= scale_x cx1 *= scale_x cy0 *= scale_x cy1 *= scale_y draw.rectangle( [(cx0, cy0), (cx1, cy1)], outline=None, fill=cell_color, ) # Draw cluster rectangle x0, y0, x1, y1 = c.bbox.as_tuple() x0 *= scale_x x1 *= scale_x y0 *= scale_x y1 *= scale_y if y1 <= y0: y1, y0 = y0, y1 if x1 <= x0: x1, x0 = x0, x1 cluster_fill_color = (*list(DocItemLabel.get_color(c.label)), 70) cluster_outline_color = ( *list(DocItemLabel.get_color(c.label)), 255, ) draw.rectangle( [(x0, y0), (x1, y1)], outline=cluster_outline_color, fill=cluster_fill_color, ) # Add label name and confidence label_text = f"{c.label.name} ({c.confidence:.2f})" # Create semi-transparent background for text text_bbox = draw.textbbox((x0, y0), label_text, font=font) text_bg_padding = 2 draw.rectangle( [ ( text_bbox[0] - text_bg_padding, text_bbox[1] - text_bg_padding, ), ( text_bbox[2] + text_bg_padding, text_bbox[3] + text_bg_padding, ), ], fill=(255, 255, 255, 180), # Semi-transparent white ) # Draw text draw.text( (x0, y0), label_text, fill=(0, 0, 0, 255), # Solid black font=font, )
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/image_backend.py
docling/backend/image_backend.py
import logging from io import BytesIO from pathlib import Path from typing import Iterable, List, Optional, Union from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.doc.page import ( BoundingRectangle, PdfPageBoundaryType, PdfPageGeometry, SegmentedPdfPage, TextCell, ) from PIL import Image from docling.backend.abstract_backend import AbstractDocumentBackend from docling.backend.pdf_backend import PdfDocumentBackend, PdfPageBackend from docling.datamodel.backend_options import PdfBackendOptions from docling.datamodel.base_models import InputFormat, Size from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) class _ImagePageBackend(PdfPageBackend): def __init__(self, image: Image.Image): self._image: Optional[Image.Image] = image self.valid: bool = self._image is not None def is_valid(self) -> bool: return self.valid def get_text_in_rect(self, bbox: BoundingBox) -> str: # No text extraction from raw images without OCR return "" def get_segmented_page(self) -> SegmentedPdfPage: # Return empty segmented page with proper dimensions for raw images assert self._image is not None page_size = self.get_size() bbox = BoundingBox( l=0.0, t=0.0, r=float(page_size.width), b=float(page_size.height), coord_origin=CoordOrigin.BOTTOMLEFT, ) dimension = PdfPageGeometry( angle=0.0, rect=BoundingRectangle.from_bounding_box(bbox), boundary_type=PdfPageBoundaryType.CROP_BOX, art_bbox=bbox, bleed_bbox=bbox, crop_bbox=bbox, media_bbox=bbox, trim_bbox=bbox, ) return SegmentedPdfPage( dimension=dimension, char_cells=[], word_cells=[], textline_cells=[], has_chars=False, has_words=False, has_lines=False, ) def get_text_cells(self) -> Iterable[TextCell]: # No text cells on raw images return [] def get_bitmap_rects(self, scale: float = 1) -> Iterable[BoundingBox]: # For raw images, the entire page is a bitmap assert self._image is not None page_size = self.get_size() full_page_bbox = BoundingBox( l=0.0, t=0.0, r=float(page_size.width), b=float(page_size.height), coord_origin=CoordOrigin.TOPLEFT, ) if scale != 1: full_page_bbox = full_page_bbox.scaled(scale=scale) yield full_page_bbox def get_page_image( self, scale: float = 1, cropbox: Optional[BoundingBox] = None ) -> Image.Image: assert self._image is not None img = self._image if cropbox is not None: # Expected cropbox comes in TOPLEFT coords in our pipeline if cropbox.coord_origin != CoordOrigin.TOPLEFT: # Convert to TOPLEFT relative to current image height cropbox = cropbox.to_top_left_origin(img.height) left, top, right, bottom = cropbox.as_tuple() left = max(0, round(left)) top = max(0, round(top)) right = min(img.width, round(right)) bottom = min(img.height, round(bottom)) img = img.crop((left, top, right, bottom)) if scale != 1: new_w = max(1, round(img.width * scale)) new_h = max(1, round(img.height * scale)) img = img.resize((new_w, new_h)) return img def get_size(self) -> Size: assert self._image is not None return Size(width=self._image.width, height=self._image.height) def unload(self): # Help GC and free memory self._image = None class ImageDocumentBackend(PdfDocumentBackend): """Image-native backend that bypasses pypdfium2. Notes: - Subclasses PdfDocumentBackend to satisfy pipeline type checks. - Intentionally avoids calling PdfDocumentBackend.__init__ to skip the image→PDF conversion and any pypdfium2 usage. - Handles multi-page TIFF by extracting frames eagerly to separate Image objects to keep thread-safety when pages process in parallel. """ def __init__( self, in_doc: InputDocument, path_or_stream: Union[BytesIO, Path], options: PdfBackendOptions = PdfBackendOptions(), ): # Bypass PdfDocumentBackend.__init__ to avoid image→PDF conversion AbstractDocumentBackend.__init__(self, in_doc, path_or_stream, options) self.options: PdfBackendOptions = options if self.input_format not in {InputFormat.IMAGE}: raise RuntimeError( f"Incompatible file format {self.input_format} was passed to ImageDocumentBackend." ) # Load frames eagerly for thread-safety across pages self._frames: List[Image.Image] = [] try: img = Image.open(self.path_or_stream) # type: ignore[arg-type] # Handle multi-frame and single-frame images # - multiframe formats: TIFF, GIF, ICO # - singleframe formats: JPEG (.jpg, .jpeg), PNG (.png), BMP, WEBP (unless animated), HEIC frame_count = getattr(img, "n_frames", 1) if frame_count > 1: for i in range(frame_count): img.seek(i) self._frames.append(img.copy().convert("RGB")) else: self._frames.append(img.convert("RGB")) except Exception as e: raise RuntimeError(f"Could not load image for document {self.file}") from e def is_valid(self) -> bool: return len(self._frames) > 0 def page_count(self) -> int: return len(self._frames) def load_page(self, page_no: int) -> _ImagePageBackend: if not (0 <= page_no < len(self._frames)): raise IndexError(f"Page index out of range: {page_no}") return _ImagePageBackend(self._frames[page_no]) @classmethod def supported_formats(cls) -> set[InputFormat]: # Only IMAGE here; PDF handling remains in PDF-oriented backends return {InputFormat.IMAGE} @classmethod def supports_pagination(cls) -> bool: return True def unload(self): super().unload() self._frames = []
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/docling_parse_v4_backend.py
docling/backend/docling_parse_v4_backend.py
import logging from collections.abc import Iterable from io import BytesIO from pathlib import Path from typing import TYPE_CHECKING, Optional, Union import pypdfium2 as pdfium from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.doc.page import SegmentedPdfPage, TextCell from docling_parse.pdf_parser import DoclingPdfParser, PdfDocument from PIL import Image from pypdfium2 import PdfPage from docling.backend.pdf_backend import PdfDocumentBackend, PdfPageBackend from docling.datamodel.backend_options import PdfBackendOptions from docling.datamodel.base_models import Size from docling.utils.locks import pypdfium2_lock if TYPE_CHECKING: from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) class DoclingParseV4PageBackend(PdfPageBackend): def __init__( self, *, dp_doc: PdfDocument, page_obj: PdfPage, page_no: int, create_words: bool = True, create_textlines: bool = True, keep_chars: bool = False, keep_lines: bool = False, keep_images: bool = True, ): self._ppage = page_obj self._dp_doc = dp_doc self._page_no = page_no self._create_words = create_words self._create_textlines = create_textlines self._keep_chars = keep_chars self._keep_lines = keep_lines self._keep_images = keep_images self._dpage: Optional[SegmentedPdfPage] = None self._unloaded = False self.valid = (self._ppage is not None) and (self._dp_doc is not None) def _ensure_parsed(self) -> None: if self._dpage is not None: return seg_page = self._dp_doc.get_page( self._page_no + 1, keep_chars=self._keep_chars, keep_lines=self._keep_lines, keep_bitmaps=self._keep_images, create_words=self._create_words, create_textlines=self._create_textlines, enforce_same_font=True, ) # In Docling, all TextCell instances are expected with top-left origin. [ tc.to_top_left_origin(seg_page.dimension.height) for tc in seg_page.textline_cells ] [tc.to_top_left_origin(seg_page.dimension.height) for tc in seg_page.char_cells] [tc.to_top_left_origin(seg_page.dimension.height) for tc in seg_page.word_cells] self._dpage = seg_page def is_valid(self) -> bool: return self.valid def get_text_in_rect(self, bbox: BoundingBox) -> str: self._ensure_parsed() assert self._dpage is not None # Find intersecting cells on the page text_piece = "" page_size = self.get_size() scale = ( 1 # FIX - Replace with param in get_text_in_rect across backends (optional) ) for i, cell in enumerate(self._dpage.textline_cells): cell_bbox = ( cell.rect.to_bounding_box() .to_top_left_origin(page_height=page_size.height) .scaled(scale) ) overlap_frac = cell_bbox.intersection_over_self(bbox) if overlap_frac > 0.5: if len(text_piece) > 0: text_piece += " " text_piece += cell.text return text_piece def get_segmented_page(self) -> Optional[SegmentedPdfPage]: self._ensure_parsed() return self._dpage def get_text_cells(self) -> Iterable[TextCell]: self._ensure_parsed() assert self._dpage is not None return self._dpage.textline_cells def get_bitmap_rects(self, scale: float = 1) -> Iterable[BoundingBox]: self._ensure_parsed() assert self._dpage is not None AREA_THRESHOLD = 0 # 32 * 32 images = self._dpage.bitmap_resources for img in images: cropbox = img.rect.to_bounding_box().to_top_left_origin( self.get_size().height ) if cropbox.area() > AREA_THRESHOLD: cropbox = cropbox.scaled(scale=scale) yield cropbox def get_page_image( self, scale: float = 1, cropbox: Optional[BoundingBox] = None ) -> Image.Image: page_size = self.get_size() if not cropbox: cropbox = BoundingBox( l=0, r=page_size.width, t=0, b=page_size.height, coord_origin=CoordOrigin.TOPLEFT, ) padbox = BoundingBox( l=0, r=0, t=0, b=0, coord_origin=CoordOrigin.BOTTOMLEFT ) else: padbox = cropbox.to_bottom_left_origin(page_size.height).model_copy() padbox.r = page_size.width - padbox.r padbox.t = page_size.height - padbox.t with pypdfium2_lock: image = ( self._ppage.render( scale=scale * 1.5, rotation=0, # no additional rotation crop=padbox.as_tuple(), ) .to_pil() .resize( size=(round(cropbox.width * scale), round(cropbox.height * scale)) ) ) # We resize the image from 1.5x the given scale to make it sharper. return image def get_size(self) -> Size: with pypdfium2_lock: return Size(width=self._ppage.get_width(), height=self._ppage.get_height()) # TODO: Take width and height from docling-parse. # return Size( # width=self._dpage.dimension.width, # height=self._dpage.dimension.height, # ) def unload(self): if not self._unloaded and self._dp_doc is not None: self._dp_doc.unload_pages((self._page_no + 1, self._page_no + 2)) self._unloaded = True self._ppage = None self._dpage = None self._dp_doc = None class DoclingParseV4DocumentBackend(PdfDocumentBackend): def __init__( self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path], options: PdfBackendOptions = PdfBackendOptions(), ): super().__init__(in_doc, path_or_stream, options) password = ( self.options.password.get_secret_value() if self.options.password else None ) with pypdfium2_lock: self._pdoc = pdfium.PdfDocument(self.path_or_stream, password=password) self.parser = DoclingPdfParser(loglevel="fatal") self.dp_doc: PdfDocument = self.parser.load( path_or_stream=self.path_or_stream, password=password ) success = self.dp_doc is not None if not success: raise RuntimeError( f"docling-parse v4 could not load document {self.document_hash}." ) def page_count(self) -> int: # return len(self._pdoc) # To be replaced with docling-parse API len_1 = len(self._pdoc) len_2 = self.dp_doc.number_of_pages() if len_1 != len_2: _log.error(f"Inconsistent number of pages: {len_1}!={len_2}") return len_2 def load_page( self, page_no: int, create_words: bool = True, create_textlines: bool = True ) -> DoclingParseV4PageBackend: with pypdfium2_lock: ppage = self._pdoc[page_no] return DoclingParseV4PageBackend( dp_doc=self.dp_doc, page_obj=ppage, page_no=page_no, create_words=create_words, create_textlines=create_textlines, ) def is_valid(self) -> bool: return self.page_count() > 0 def unload(self): super().unload() # Unload docling-parse document first if self.dp_doc is not None: self.dp_doc.unload() self.dp_doc = None # Then close pypdfium2 document with proper locking if self._pdoc is not None: with pypdfium2_lock: try: self._pdoc.close() except Exception: # Ignore cleanup errors pass self._pdoc = None
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/md_backend.py
docling/backend/md_backend.py
import logging import re import warnings from copy import deepcopy from enum import Enum from html import unescape from io import BytesIO from pathlib import Path from typing import Literal, Optional, Union, cast import marko import marko.element import marko.inline from docling_core.types.doc import ( ContentLayer, DocItem, DocItemLabel, DoclingDocument, DocumentOrigin, ListItem, NodeItem, TableCell, TableData, TextItem, ) from docling_core.types.doc.document import Formatting from marko import Markdown from pydantic import AnyUrl, BaseModel, Field, TypeAdapter from typing_extensions import Annotated, override from docling.backend.abstract_backend import ( DeclarativeDocumentBackend, ) from docling.backend.html_backend import HTMLDocumentBackend from docling.datamodel.backend_options import ( HTMLBackendOptions, MarkdownBackendOptions, ) from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) _MARKER_BODY = "DOCLING_DOC_MD_HTML_EXPORT" _START_MARKER = f"#_#_{_MARKER_BODY}_START_#_#" _STOP_MARKER = f"#_#_{_MARKER_BODY}_STOP_#_#" class _PendingCreationType(str, Enum): """CoordOrigin.""" HEADING = "heading" LIST_ITEM = "list_item" class _HeadingCreationPayload(BaseModel): kind: Literal["heading"] = "heading" level: int class _ListItemCreationPayload(BaseModel): kind: Literal["list_item"] = "list_item" enumerated: bool _CreationPayload = Annotated[ Union[ _HeadingCreationPayload, _ListItemCreationPayload, ], Field(discriminator="kind"), ] class MarkdownDocumentBackend(DeclarativeDocumentBackend): def _shorten_underscore_sequences(self, markdown_text: str, max_length: int = 10): # This regex will match any sequence of underscores pattern = r"_+" def replace_match(match): underscore_sequence = match.group( 0 ) # Get the full match (sequence of underscores) # Shorten the sequence if it exceeds max_length if len(underscore_sequence) > max_length: return "_" * max_length else: return underscore_sequence # Leave it unchanged if it is shorter or equal to max_length # Use re.sub to replace long underscore sequences shortened_text = re.sub(pattern, replace_match, markdown_text) if len(shortened_text) != len(markdown_text): warnings.warn("Detected potentially incorrect Markdown, correcting...") return shortened_text @override def __init__( self, in_doc: InputDocument, path_or_stream: Union[BytesIO, Path], options: MarkdownBackendOptions = MarkdownBackendOptions(), ): super().__init__(in_doc, path_or_stream, options) _log.debug("Starting MarkdownDocumentBackend...") # Markdown file: self.path_or_stream = path_or_stream self.valid = True self.markdown = "" # To store original Markdown string self.in_table = False self.md_table_buffer: list[str] = [] self._html_blocks: int = 0 try: if isinstance(self.path_or_stream, BytesIO): text_stream = self.path_or_stream.getvalue().decode("utf-8") # remove invalid sequences # very long sequences of underscores will lead to unnecessary long processing times. # In any proper Markdown files, underscores have to be escaped, # otherwise they represent emphasis (bold or italic) self.markdown = self._shorten_underscore_sequences(text_stream) if isinstance(self.path_or_stream, Path): with open(self.path_or_stream, encoding="utf-8") as f: md_content = f.read() # remove invalid sequences # very long sequences of underscores will lead to unnecessary long processing times. # In any proper Markdown files, underscores have to be escaped, # otherwise they represent emphasis (bold or italic) self.markdown = self._shorten_underscore_sequences(md_content) self.valid = True _log.debug(self.markdown) except Exception as e: raise RuntimeError( f"Could not initialize MD backend for file with hash {self.document_hash}." ) from e return def _close_table(self, doc: DoclingDocument): if self.in_table: _log.debug("=== TABLE START ===") for md_table_row in self.md_table_buffer: _log.debug(md_table_row) _log.debug("=== TABLE END ===") tcells: list[TableCell] = [] result_table = [] for n, md_table_row in enumerate(self.md_table_buffer): data = [] if n == 0: header = [t.strip() for t in md_table_row.split("|")[1:-1]] for value in header: data.append(value) result_table.append(data) if n > 1: values = [t.strip() for t in md_table_row.split("|")[1:-1]] for value in values: data.append(value) result_table.append(data) for trow_ind, trow in enumerate(result_table): for tcol_ind, cellval in enumerate(trow): row_span = ( 1 # currently supporting just simple tables (without spans) ) col_span = ( 1 # currently supporting just simple tables (without spans) ) icell = TableCell( text=cellval.strip(), row_span=row_span, col_span=col_span, start_row_offset_idx=trow_ind, end_row_offset_idx=trow_ind + row_span, start_col_offset_idx=tcol_ind, end_col_offset_idx=tcol_ind + col_span, column_header=trow_ind == 0, row_header=False, ) tcells.append(icell) num_rows = len(result_table) num_cols = len(result_table[0]) self.in_table = False self.md_table_buffer = [] # clean table markdown buffer # Initialize Docling TableData table_data = TableData( num_rows=num_rows, num_cols=num_cols, table_cells=tcells ) # Populate for tcell in tcells: table_data.table_cells.append(tcell) if len(tcells) > 0: doc.add_table(data=table_data) return def _create_list_item( self, doc: DoclingDocument, parent_item: Optional[NodeItem], text: str, enumerated: bool, formatting: Optional[Formatting] = None, hyperlink: Optional[Union[AnyUrl, Path]] = None, ): item = doc.add_list_item( text=text, enumerated=enumerated, parent=parent_item, formatting=formatting, hyperlink=hyperlink, ) return item def _create_heading_item( self, doc: DoclingDocument, parent_item: Optional[NodeItem], text: str, level: int, formatting: Optional[Formatting] = None, hyperlink: Optional[Union[AnyUrl, Path]] = None, ): if level == 1: item = doc.add_title( text=text, parent=parent_item, formatting=formatting, hyperlink=hyperlink, ) else: item = doc.add_heading( text=text, level=level - 1, parent=parent_item, formatting=formatting, hyperlink=hyperlink, ) return item def _iterate_elements( # noqa: C901 self, *, element: marko.element.Element, depth: int, doc: DoclingDocument, visited: set[marko.element.Element], creation_stack: list[ _CreationPayload ], # stack for lazy item creation triggered deep in marko's AST (on RawText) list_ordered_flag_by_ref: dict[str, bool], list_last_item_by_ref: dict[str, ListItem], parent_item: Optional[NodeItem] = None, formatting: Optional[Formatting] = None, hyperlink: Optional[Union[AnyUrl, Path]] = None, ): if element in visited: return # Iterates over all elements in the AST # Check for different element types and process relevant details if ( isinstance(element, marko.block.Heading) or isinstance(element, marko.block.SetextHeading) ) and len(element.children) > 0: self._close_table(doc) _log.debug( f" - Heading level {element.level}, content: {element.children[0].children}" # type: ignore ) if len(element.children) > 1: # inline group will be created further down parent_item = self._create_heading_item( doc=doc, parent_item=parent_item, text="", level=element.level, formatting=formatting, hyperlink=hyperlink, ) else: creation_stack.append(_HeadingCreationPayload(level=element.level)) elif isinstance(element, marko.block.List): has_non_empty_list_items = False for child in element.children: if isinstance(child, marko.block.ListItem) and len(child.children) > 0: has_non_empty_list_items = True break self._close_table(doc) _log.debug(f" - List {'ordered' if element.ordered else 'unordered'}") if has_non_empty_list_items: parent_item = doc.add_list_group(name="list", parent=parent_item) list_ordered_flag_by_ref[parent_item.self_ref] = element.ordered elif ( isinstance(element, marko.block.ListItem) and len(element.children) > 0 and isinstance((child := element.children[0]), marko.block.Paragraph) and len(child.children) > 0 ): self._close_table(doc) _log.debug(" - List item") enumerated = ( list_ordered_flag_by_ref.get(parent_item.self_ref, False) if parent_item else False ) non_list_children: list[marko.element.Element] = [ item for item in child.children if not isinstance(item, marko.block.ListItem) ] if len(non_list_children) > 1: # inline group will be created further down parent_ref: Optional[str] = ( parent_item.self_ref if parent_item else None ) parent_item = self._create_list_item( doc=doc, parent_item=parent_item, text="", enumerated=enumerated, formatting=formatting, hyperlink=hyperlink, ) if parent_ref: list_last_item_by_ref[parent_ref] = cast(ListItem, parent_item) else: creation_stack.append(_ListItemCreationPayload(enumerated=enumerated)) elif isinstance(element, marko.inline.Image): self._close_table(doc) _log.debug(f" - Image with alt: {element.title}, url: {element.dest}") fig_caption: Optional[TextItem] = None if element.title is not None and element.title != "": title = unescape(element.title) fig_caption = doc.add_text( label=DocItemLabel.CAPTION, text=title, formatting=formatting, hyperlink=hyperlink, ) doc.add_picture(parent=parent_item, caption=fig_caption) elif isinstance(element, marko.inline.Emphasis): _log.debug(f" - Emphasis: {element.children}") formatting = deepcopy(formatting) if formatting else Formatting() formatting.italic = True elif isinstance(element, marko.inline.StrongEmphasis): _log.debug(f" - StrongEmphasis: {element.children}") formatting = deepcopy(formatting) if formatting else Formatting() formatting.bold = True elif isinstance(element, marko.inline.Link): _log.debug(f" - Link: {element.children}") hyperlink = TypeAdapter(Optional[Union[AnyUrl, Path]]).validate_python( element.dest ) elif isinstance(element, (marko.inline.RawText, marko.inline.Literal)): _log.debug(f" - RawText/Literal: {element.children}") snippet_text = ( element.children.strip() if isinstance(element.children, str) else "" ) snippet_text = unescape(snippet_text) # Detect start of the table: if "|" in snippet_text or self.in_table: # most likely part of the markdown table self.in_table = True if len(self.md_table_buffer) > 0: self.md_table_buffer[len(self.md_table_buffer) - 1] += snippet_text else: self.md_table_buffer.append(snippet_text) elif snippet_text: self._close_table(doc) if creation_stack: while len(creation_stack) > 0: to_create = creation_stack.pop() if isinstance(to_create, _ListItemCreationPayload): enumerated = ( list_ordered_flag_by_ref.get( parent_item.self_ref, False ) if parent_item else False ) parent_ref = parent_item.self_ref if parent_item else None parent_item = self._create_list_item( doc=doc, parent_item=parent_item, text=snippet_text, enumerated=enumerated, formatting=formatting, hyperlink=hyperlink, ) if parent_ref: list_last_item_by_ref[parent_ref] = cast( ListItem, parent_item ) elif isinstance(to_create, _HeadingCreationPayload): # not keeping as parent_item as logic for correctly tracking # that not implemented yet (section components not captured # as heading children in marko) self._create_heading_item( doc=doc, parent_item=parent_item, text=snippet_text, level=to_create.level, formatting=formatting, hyperlink=hyperlink, ) else: doc.add_text( label=DocItemLabel.TEXT, parent=parent_item, text=snippet_text, formatting=formatting, hyperlink=hyperlink, ) elif isinstance(element, marko.inline.CodeSpan): self._close_table(doc) _log.debug(f" - Code Span: {element.children}") snippet_text = str(element.children).strip() doc.add_code( parent=parent_item, text=snippet_text, formatting=formatting, hyperlink=hyperlink, ) elif ( isinstance(element, (marko.block.CodeBlock, marko.block.FencedCode)) and len(element.children) > 0 and isinstance((child := element.children[0]), marko.inline.RawText) and len(snippet_text := (child.children.strip())) > 0 ): self._close_table(doc) _log.debug(f" - Code Block: {element.children}") doc.add_code( parent=parent_item, text=snippet_text, formatting=formatting, hyperlink=hyperlink, ) elif isinstance(element, marko.inline.LineBreak): if self.in_table: _log.debug("Line break in a table") self.md_table_buffer.append("") elif isinstance(element, marko.block.HTMLBlock): self._html_blocks += 1 self._close_table(doc) _log.debug(f"HTML Block: {element}") if ( len(element.body) > 0 ): # If Marko doesn't return any content for HTML block, skip it html_block = element.body.strip() # wrap in markers to enable post-processing in convert() text_to_add = f"{_START_MARKER}{html_block}{_STOP_MARKER}" doc.add_code( parent=parent_item, text=text_to_add, formatting=formatting, hyperlink=hyperlink, ) else: if not isinstance(element, str): self._close_table(doc) _log.debug(f"Some other element: {element}") if ( isinstance(element, (marko.block.Paragraph, marko.block.Heading)) and len(element.children) > 1 ): parent_item = doc.add_inline_group(parent=parent_item) processed_block_types = ( marko.block.CodeBlock, marko.block.FencedCode, marko.inline.RawText, ) # Iterate through the element's children (if any) if hasattr(element, "children") and not isinstance( element, processed_block_types ): for child in element.children: if ( isinstance(element, marko.block.ListItem) and isinstance(child, marko.block.List) and parent_item and list_last_item_by_ref.get(parent_item.self_ref, None) ): _log.debug( f"walking into new List hanging from item of parent list {parent_item.self_ref}" ) parent_item = list_last_item_by_ref[parent_item.self_ref] self._iterate_elements( element=child, depth=depth + 1, doc=doc, visited=visited, creation_stack=creation_stack, list_ordered_flag_by_ref=list_ordered_flag_by_ref, list_last_item_by_ref=list_last_item_by_ref, parent_item=parent_item, formatting=formatting, hyperlink=hyperlink, ) def is_valid(self) -> bool: return self.valid def unload(self): if isinstance(self.path_or_stream, BytesIO): self.path_or_stream.close() self.path_or_stream = None @classmethod def supports_pagination(cls) -> bool: return False @classmethod def supported_formats(cls) -> set[InputFormat]: return {InputFormat.MD} def convert(self) -> DoclingDocument: _log.debug("converting Markdown...") origin = DocumentOrigin( filename=self.file.name or "file", mimetype="text/markdown", binary_hash=self.document_hash, ) doc = DoclingDocument(name=self.file.stem or "file", origin=origin) if self.is_valid(): # Parse the markdown into an abstract syntax tree (AST) marko_parser = Markdown() parsed_ast = marko_parser.parse(self.markdown) # Start iterating from the root of the AST self._iterate_elements( element=parsed_ast, depth=0, doc=doc, parent_item=None, visited=set(), creation_stack=[], list_ordered_flag_by_ref={}, list_last_item_by_ref={}, ) self._close_table(doc=doc) # handle any last hanging table # if HTML blocks were detected, export to HTML and delegate to HTML backend if self._html_blocks > 0: # export to HTML html_backend_cls = HTMLDocumentBackend html_str = doc.export_to_html() def _restore_original_html(txt, regex): _txt, count = re.subn(regex, "", txt) if count != self._html_blocks: raise RuntimeError( "An internal error has occurred during Markdown conversion." ) return _txt # restore original HTML by removing previously added markers for regex in [ rf"<pre>\s*<code>\s*{_START_MARKER}", rf"{_STOP_MARKER}\s*</code>\s*</pre>", ]: html_str = _restore_original_html(txt=html_str, regex=regex) self._html_blocks = 0 # delegate to HTML backend stream = BytesIO(bytes(html_str, encoding="utf-8")) md_options = cast(MarkdownBackendOptions, self.options) html_options = HTMLBackendOptions( enable_local_fetch=md_options.enable_local_fetch, enable_remote_fetch=md_options.enable_remote_fetch, fetch_images=md_options.fetch_images, source_uri=md_options.source_uri, infer_furniture=False, add_title=False, ) in_doc = InputDocument( path_or_stream=stream, format=InputFormat.HTML, backend=html_backend_cls, filename=self.file.name, backend_options=html_options, ) html_backend_obj = html_backend_cls( in_doc=in_doc, path_or_stream=stream, options=html_options, ) doc = html_backend_obj.convert() else: raise RuntimeError( f"Cannot convert md with {self.document_hash} because the backend failed to init." ) return doc
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/msexcel_backend.py
docling/backend/msexcel_backend.py
import logging from io import BytesIO from pathlib import Path from typing import Annotated, Any, Optional, Union, cast from docling_core.types.doc import ( BoundingBox, ContentLayer, CoordOrigin, DocItem, DocItemLabel, DoclingDocument, DocumentOrigin, GroupLabel, ImageRef, ProvenanceItem, Size, TableCell, TableData, ) from openpyxl import load_workbook from openpyxl.chartsheet.chartsheet import Chartsheet from openpyxl.drawing.image import Image from openpyxl.drawing.spreadsheet_drawing import TwoCellAnchor from openpyxl.worksheet.worksheet import Worksheet from PIL import Image as PILImage from pydantic import BaseModel, Field, NonNegativeInt, PositiveInt from pydantic.dataclasses import dataclass from typing_extensions import override from docling.backend.abstract_backend import ( DeclarativeDocumentBackend, PaginatedDocumentBackend, ) from docling.datamodel.backend_options import MsExcelBackendOptions from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) @dataclass class DataRegion: """Represents the bounding rectangle of non-empty cells in a worksheet.""" min_row: Annotated[ PositiveInt, Field(description="Smallest row index (1-based index).") ] max_row: Annotated[ PositiveInt, Field(description="Largest row index (1-based index).") ] min_col: Annotated[ PositiveInt, Field(description="Smallest column index (1-based index).") ] max_col: Annotated[ PositiveInt, Field(description="Largest column index (1-based index).") ] def width(self) -> PositiveInt: """Number of columns in the data region.""" return self.max_col - self.min_col + 1 def height(self) -> PositiveInt: """Number of rows in the data region.""" return self.max_row - self.min_row + 1 class ExcelCell(BaseModel): """Represents an Excel cell. Attributes: row: The row number of the cell. col: The column number of the cell. text: The text content of the cell. row_span: The number of rows the cell spans. col_span: The number of columns the cell spans. """ row: int col: int text: str row_span: int col_span: int class ExcelTable(BaseModel): """Represents an Excel table on a worksheet. Attributes: anchor: The column and row indices of the upper-left cell of the table (0-based index). num_rows: The number of rows in the table. num_cols: The number of columns in the table. data: The data in the table, represented as a list of ExcelCell objects. """ anchor: tuple[NonNegativeInt, NonNegativeInt] num_rows: int num_cols: int data: list[ExcelCell] class MsExcelDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentBackend): """Backend for parsing Excel workbooks. The backend converts an Excel workbook into a DoclingDocument object. Each worksheet is converted into a separate page. The following elements are parsed: - Cell contents, parsed as tables. If two groups of cells are disconnected between each other, they will be parsed as two different tables. - Images, parsed as PictureItem objects. The DoclingDocument tables and pictures have their provenance information, including the position in their original Excel worksheet. The position is represented by a bounding box object with the cell indices as units (0-based index). The size of this bounding box is the number of columns and rows that the table or picture spans. """ @override def __init__( self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path], options: MsExcelBackendOptions = MsExcelBackendOptions(), ) -> None: """Initialize the MsExcelDocumentBackend object. Parameters: in_doc: The input document object. path_or_stream: The path or stream to the Excel file. options: Backend options for Excel parsing. Raises: RuntimeError: An error occurred parsing the file. """ super().__init__(in_doc, path_or_stream, options) # Initialise the parents for the hierarchy self.max_levels = 10 self.parents: dict[int, Any] = {} for i in range(-1, self.max_levels): self.parents[i] = None self.workbook = None try: if isinstance(self.path_or_stream, BytesIO): self.workbook = load_workbook( filename=self.path_or_stream, data_only=True ) elif isinstance(self.path_or_stream, Path): self.workbook = load_workbook( filename=str(self.path_or_stream), data_only=True ) self.valid = self.workbook is not None except Exception as e: self.valid = False raise RuntimeError( f"MsExcelDocumentBackend could not load document with hash {self.document_hash}" ) from e @override def is_valid(self) -> bool: _log.debug(f"valid: {self.valid}") return self.valid @classmethod @override def supports_pagination(cls) -> bool: return True @override def page_count(self) -> int: if self.is_valid() and self.workbook: return len(self.workbook.sheetnames) else: return 0 @classmethod @override def supported_formats(cls) -> set[InputFormat]: return {InputFormat.XLSX} @override def convert(self) -> DoclingDocument: """Parse the Excel workbook into a DoclingDocument object. Raises: RuntimeError: Unable to run the conversion since the backend object failed to initialize. Returns: The DoclingDocument object representing the Excel workbook. """ origin = DocumentOrigin( filename=self.file.name or "file.xlsx", mimetype="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet", binary_hash=self.document_hash, ) doc = DoclingDocument(name=self.file.stem or "file.xlsx", origin=origin) if self.is_valid(): doc = self._convert_workbook(doc) else: raise RuntimeError( f"Cannot convert doc with {self.document_hash} because the backend failed to init." ) return doc def _convert_workbook(self, doc: DoclingDocument) -> DoclingDocument: """Parse the Excel workbook and attach its structure to a DoclingDocument. Args: doc: A DoclingDocument object. Returns: A DoclingDocument object with the parsed items. """ if self.workbook is not None: # Iterate over all sheets for idx, name in enumerate(self.workbook.sheetnames): _log.info(f"Processing sheet {idx}: {name}") sheet = self.workbook[name] page_no = idx + 1 # do not rely on sheet.max_column, sheet.max_row if there are images page = doc.add_page(page_no=page_no, size=Size(width=0, height=0)) self.parents[0] = doc.add_group( parent=None, label=GroupLabel.SECTION, name=f"sheet: {name}", content_layer=self._get_sheet_content_layer(sheet), ) doc = self._convert_sheet(doc, sheet) width, height = self._find_page_size(doc, page_no) page.size = Size(width=width, height=height) else: _log.error("Workbook is not initialized.") return doc def _convert_sheet( self, doc: DoclingDocument, sheet: Union[Worksheet, Chartsheet] ) -> DoclingDocument: """Parse an Excel worksheet and attach its structure to a DoclingDocument Args: doc: The DoclingDocument to be updated. sheet: The Excel worksheet to be parsed. Returns: The updated DoclingDocument. """ if isinstance(sheet, Worksheet): doc = self._find_tables_in_sheet(doc, sheet) doc = self._find_images_in_sheet(doc, sheet) # TODO: parse charts in sheet return doc def _find_tables_in_sheet( self, doc: DoclingDocument, sheet: Worksheet ) -> DoclingDocument: """Find all tables in an Excel sheet and attach them to a DoclingDocument. Args: doc: The DoclingDocument to be updated. sheet: The Excel worksheet to be parsed. Returns: The updated DoclingDocument. """ if self.workbook is not None: content_layer = self._get_sheet_content_layer(sheet) tables = self._find_data_tables(sheet) treat_singleton_as_text = ( isinstance(self.options, MsExcelBackendOptions) and self.options.treat_singleton_as_text ) for excel_table in tables: origin_col = excel_table.anchor[0] origin_row = excel_table.anchor[1] num_rows = excel_table.num_rows num_cols = excel_table.num_cols if ( treat_singleton_as_text and num_rows == 1 and num_cols == 1 and excel_table.data ): page_no = self.workbook.index(sheet) + 1 doc.add_text( text=excel_table.data[0].text, label=DocItemLabel.TEXT, parent=self.parents[0], prov=ProvenanceItem( page_no=page_no, charspan=(0, 0), bbox=BoundingBox.from_tuple( ( origin_col, origin_row, origin_col + num_cols, origin_row + num_rows, ), origin=CoordOrigin.TOPLEFT, ), ), content_layer=content_layer, ) else: table_data = TableData( num_rows=num_rows, num_cols=num_cols, table_cells=[], ) for excel_cell in excel_table.data: cell = TableCell( text=excel_cell.text, row_span=excel_cell.row_span, col_span=excel_cell.col_span, start_row_offset_idx=excel_cell.row, end_row_offset_idx=excel_cell.row + excel_cell.row_span, start_col_offset_idx=excel_cell.col, end_col_offset_idx=excel_cell.col + excel_cell.col_span, column_header=excel_cell.row == 0, row_header=False, ) table_data.table_cells.append(cell) page_no = self.workbook.index(sheet) + 1 doc.add_table( data=table_data, parent=self.parents[0], prov=ProvenanceItem( page_no=page_no, charspan=(0, 0), bbox=BoundingBox.from_tuple( ( origin_col, origin_row, origin_col + num_cols, origin_row + num_rows, ), origin=CoordOrigin.TOPLEFT, ), ), content_layer=content_layer, ) return doc def _find_true_data_bounds(self, sheet: Worksheet) -> DataRegion: """Find the true data boundaries (min/max rows and columns) in a worksheet. This function scans all cells to find the smallest rectangular region that contains all non-empty cells or merged cell ranges. It returns the minimal and maximal row/column indices that bound the actual data region. Args: sheet: The worksheet to analyze. Returns: A data region representing the smallest rectangle that covers all data and merged cells. If the sheet is empty, returns (1, 1, 1, 1) by default. """ min_row, min_col = None, None max_row, max_col = 0, 0 for cell in sheet._cells.values(): if cell.value is not None: r, c = cell.row, cell.column min_row = r if min_row is None else min(min_row, r) min_col = c if min_col is None else min(min_col, c) max_row = max(max_row, r) max_col = max(max_col, c) # Expand bounds to include merged cells for merged in sheet.merged_cells.ranges: min_row = ( merged.min_row if min_row is None else min(min_row, merged.min_row) ) min_col = ( merged.min_col if min_col is None else min(min_col, merged.min_col) ) max_row = max(max_row, merged.max_row) max_col = max(max_col, merged.max_col) # If no data found, default to (1, 1, 1, 1) if min_row is None or min_col is None: min_row = min_col = max_row = max_col = 1 return DataRegion(min_row, max_row, min_col, max_col) def _find_data_tables(self, sheet: Worksheet) -> list[ExcelTable]: """Find all compact rectangular data tables in an Excel worksheet. Args: sheet: The Excel worksheet to be parsed. Returns: A list of ExcelTable objects representing the data tables. """ bounds: DataRegion = self._find_true_data_bounds( sheet ) # The true data boundaries tables: list[ExcelTable] = [] # List to store found tables visited: set[tuple[int, int]] = set() # Track already visited cells # Limit scan to actual data bounds for ri, row in enumerate( sheet.iter_rows( min_row=bounds.min_row, max_row=bounds.max_row, min_col=bounds.min_col, max_col=bounds.max_col, values_only=False, ), start=bounds.min_row - 1, ): for rj, cell in enumerate(row, start=bounds.min_col - 1): if cell.value is None or (ri, rj) in visited: continue # If the cell starts a new table, find its bounds table_bounds, visited_cells = self._find_table_bounds( sheet, ri, rj, bounds.max_row, bounds.max_col ) visited.update(visited_cells) # Mark these cells as visited tables.append(table_bounds) return tables def _find_table_bounds( self, sheet: Worksheet, start_row: int, start_col: int, max_row: int, max_col: int, ) -> tuple[ExcelTable, set[tuple[int, int]]]: """Determine the bounds of a compact rectangular table. Args: sheet: The Excel worksheet to be parsed. start_row: The row number of the starting cell. start_col: The column number of the starting cell. max_row: Maximum row boundary from true data bounds. max_col: Maximum column boundary from true data bounds. Returns: A tuple with an Excel table and a set of cell coordinates. """ _log.debug("find_table_bounds") table_max_row = self._find_table_bottom(sheet, start_row, start_col, max_row) table_max_col = self._find_table_right(sheet, start_row, start_col, max_col) # Collect the data within the bounds data = [] visited_cells: set[tuple[int, int]] = set() for ri, row in enumerate( sheet.iter_rows( min_row=start_row + 1, # start_row is 0-based but iter_rows is 1-based max_row=table_max_row + 1, min_col=start_col + 1, max_col=table_max_col + 1, values_only=False, ), start_row, ): for rj, cell in enumerate(row, start_col): # Check if the cell belongs to a merged range row_span = 1 col_span = 1 for merged_range in sheet.merged_cells.ranges: if ( merged_range.min_row <= ri + 1 and ri + 1 <= merged_range.max_row and merged_range.min_col <= rj + 1 and rj + 1 <= merged_range.max_col ): row_span = merged_range.max_row - merged_range.min_row + 1 col_span = merged_range.max_col - merged_range.min_col + 1 break if (ri, rj) not in visited_cells: data.append( ExcelCell( row=ri - start_row, col=rj - start_col, text=str(cell.value), row_span=row_span, col_span=col_span, ) ) # Mark all cells in the span as visited for span_row in range(ri, ri + row_span): for span_col in range(rj, rj + col_span): visited_cells.add((span_row, span_col)) return ( ExcelTable( anchor=(start_col, start_row), num_rows=table_max_row + 1 - start_row, num_cols=table_max_col + 1 - start_col, data=data, ), visited_cells, ) def _find_table_bottom( self, sheet: Worksheet, start_row: int, start_col: int, max_row: int ) -> int: """Find the bottom boundary of a table. Args: sheet: The Excel worksheet to be parsed. start_row: The starting row of the table. start_col: The starting column of the table. max_row: Maximum row boundary from true data bounds. Returns: The row index representing the bottom boundary of the table. """ table_max_row: int = start_row for ri, (cell,) in enumerate( sheet.iter_rows( min_row=start_row + 2, max_row=max_row, min_col=start_col + 1, max_col=start_col + 1, values_only=False, ), start_row + 1, ): # Check if the cell is part of a merged range merged_range = next( (mr for mr in sheet.merged_cells.ranges if cell.coordinate in mr), None, ) if cell.value is None and not merged_range: break # Stop if the cell is empty and not merged # Expand table_max_row to include the merged range if applicable if merged_range: table_max_row = max(table_max_row, merged_range.max_row - 1) else: table_max_row = ri return table_max_row def _find_table_right( self, sheet: Worksheet, start_row: int, start_col: int, max_col: int ) -> int: """Find the right boundary of a table. Args: sheet: The Excel worksheet to be parsed. start_row: The starting row of the table. start_col: The starting column of the table. max_col: The actual max column of the table. Returns: The column index representing the right boundary of the table." """ table_max_col: int = start_col for rj, (cell,) in enumerate( sheet.iter_cols( min_row=start_row + 1, max_row=start_row + 1, min_col=start_col + 2, max_col=max_col, values_only=False, ), start_col + 1, ): # Check if the cell is part of a merged range merged_range = next( (mr for mr in sheet.merged_cells.ranges if cell.coordinate in mr), None, ) if cell.value is None and not merged_range: break # Stop if the cell is empty and not merged # Expand table_max_col to include the merged range if applicable if merged_range: table_max_col = max(table_max_col, merged_range.max_col - 1) else: table_max_col = rj return table_max_col def _find_images_in_sheet( self, doc: DoclingDocument, sheet: Worksheet ) -> DoclingDocument: """Find images in the Excel sheet and attach them to the DoclingDocument. Args: doc: The DoclingDocument to be updated. sheet: The Excel worksheet to be parsed. Returns: The updated DoclingDocument. """ if self.workbook is not None: content_layer = self._get_sheet_content_layer(sheet) # Iterate over byte images in the sheet for item in sheet._images: # type: ignore[attr-defined] try: image: Image = cast(Image, item) pil_image = PILImage.open(image.ref) # type: ignore[arg-type] page_no = self.workbook.index(sheet) + 1 anchor = (0, 0, 0, 0) if isinstance(image.anchor, TwoCellAnchor): anchor = ( image.anchor._from.col, image.anchor._from.row, image.anchor.to.col + 1, image.anchor.to.row + 1, ) doc.add_picture( parent=self.parents[0], image=ImageRef.from_pil(image=pil_image, dpi=72), caption=None, prov=ProvenanceItem( page_no=page_no, charspan=(0, 0), bbox=BoundingBox.from_tuple( anchor, origin=CoordOrigin.TOPLEFT ), ), content_layer=content_layer, ) except Exception: _log.error("could not extract the image from excel sheets") return doc @staticmethod def _find_page_size( doc: DoclingDocument, page_no: PositiveInt ) -> tuple[float, float]: left: float = -1.0 top: float = -1.0 right: float = -1.0 bottom: float = -1.0 for item, _ in doc.iterate_items(traverse_pictures=True, page_no=page_no): if not isinstance(item, DocItem): continue for provenance in item.prov: bbox = provenance.bbox left = min(left, bbox.l) if left != -1 else bbox.l right = max(right, bbox.r) if right != -1 else bbox.r top = min(top, bbox.t) if top != -1 else bbox.t bottom = max(bottom, bbox.b) if bottom != -1 else bbox.b return (right - left, bottom - top) @staticmethod def _get_sheet_content_layer(sheet: Worksheet) -> Optional[ContentLayer]: return ( None if sheet.sheet_state == Worksheet.SHEETSTATE_VISIBLE else ContentLayer.INVISIBLE )
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/abstract_backend.py
docling/backend/abstract_backend.py
from abc import ABC, abstractmethod from io import BytesIO from pathlib import Path from typing import TYPE_CHECKING, Union from docling_core.types.doc import DoclingDocument from docling.datamodel.backend_options import ( BackendOptions, BaseBackendOptions, DeclarativeBackendOptions, ) if TYPE_CHECKING: from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument class AbstractDocumentBackend(ABC): @abstractmethod def __init__( self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path], options: BaseBackendOptions = BaseBackendOptions(), ): self.file = in_doc.file self.path_or_stream = path_or_stream self.document_hash = in_doc.document_hash self.input_format = in_doc.format self.options = options @abstractmethod def is_valid(self) -> bool: pass @classmethod @abstractmethod def supports_pagination(cls) -> bool: pass def unload(self): if isinstance(self.path_or_stream, BytesIO): self.path_or_stream.close() self.path_or_stream = None @classmethod @abstractmethod def supported_formats(cls) -> set["InputFormat"]: pass class PaginatedDocumentBackend(AbstractDocumentBackend): """DeclarativeDocumentBackend. A declarative document backend is a backend that can transform to DoclingDocument straight without a recognition pipeline. """ @abstractmethod def page_count(self) -> int: pass class DeclarativeDocumentBackend(AbstractDocumentBackend): """DeclarativeDocumentBackend. A declarative document backend is a backend that can transform to DoclingDocument straight without a recognition pipeline. """ @abstractmethod def __init__( self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path], options: BackendOptions = DeclarativeBackendOptions(), ) -> None: super().__init__(in_doc, path_or_stream, options) @abstractmethod def convert(self) -> DoclingDocument: pass
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/docling_parse_backend.py
docling/backend/docling_parse_backend.py
import logging import random from collections.abc import Iterable from io import BytesIO from pathlib import Path from typing import List, Optional, Union import pypdfium2 as pdfium from docling_core.types.doc import BoundingBox, CoordOrigin, Size from docling_core.types.doc.page import ( BoundingRectangle, SegmentedPdfPage, TextCell, ) from docling_parse.pdf_parsers import pdf_parser_v1 from PIL import Image, ImageDraw from pypdfium2 import PdfPage from docling.backend.pdf_backend import PdfDocumentBackend, PdfPageBackend from docling.backend.pypdfium2_backend import get_pdf_page_geometry from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) class DoclingParsePageBackend(PdfPageBackend): def __init__( self, parser: pdf_parser_v1, document_hash: str, page_no: int, page_obj: PdfPage ): self._ppage = page_obj parsed_page = parser.parse_pdf_from_key_on_page(document_hash, page_no) self.valid = "pages" in parsed_page if self.valid: self._dpage = parsed_page["pages"][0] else: _log.info( f"An error occurred when loading page {page_no} of document {document_hash}." ) def is_valid(self) -> bool: return self.valid def _compute_text_cells(self) -> List[TextCell]: """Compute text cells from docling-parse data.""" cells: List[TextCell] = [] cell_counter = 0 if not self.valid: return cells page_size = self.get_size() parser_width = self._dpage["width"] parser_height = self._dpage["height"] for i in range(len(self._dpage["cells"])): rect = self._dpage["cells"][i]["box"]["device"] x0, y0, x1, y1 = rect if x1 < x0: x0, x1 = x1, x0 if y1 < y0: y0, y1 = y1, y0 text_piece = self._dpage["cells"][i]["content"]["rnormalized"] cells.append( TextCell( index=cell_counter, text=text_piece, orig=text_piece, from_ocr=False, rect=BoundingRectangle.from_bounding_box( BoundingBox( l=x0 * page_size.width / parser_width, b=y0 * page_size.height / parser_height, r=x1 * page_size.width / parser_width, t=y1 * page_size.height / parser_height, coord_origin=CoordOrigin.BOTTOMLEFT, ) ).to_top_left_origin(page_size.height), ) ) cell_counter += 1 return cells def get_text_in_rect(self, bbox: BoundingBox) -> str: if not self.valid: return "" # Find intersecting cells on the page text_piece = "" page_size = self.get_size() parser_width = self._dpage["width"] parser_height = self._dpage["height"] scale = ( 1 # FIX - Replace with param in get_text_in_rect across backends (optional) ) for i in range(len(self._dpage["cells"])): rect = self._dpage["cells"][i]["box"]["device"] x0, y0, x1, y1 = rect cell_bbox = BoundingBox( l=x0 * scale * page_size.width / parser_width, b=y0 * scale * page_size.height / parser_height, r=x1 * scale * page_size.width / parser_width, t=y1 * scale * page_size.height / parser_height, coord_origin=CoordOrigin.BOTTOMLEFT, ).to_top_left_origin(page_height=page_size.height * scale) overlap_frac = cell_bbox.intersection_over_self(bbox) if overlap_frac > 0.5: if len(text_piece) > 0: text_piece += " " text_piece += self._dpage["cells"][i]["content"]["rnormalized"] return text_piece def get_segmented_page(self) -> Optional[SegmentedPdfPage]: if not self.valid: return None text_cells = self._compute_text_cells() # Get the PDF page geometry from pypdfium2 dimension = get_pdf_page_geometry(self._ppage) # Create SegmentedPdfPage return SegmentedPdfPage( dimension=dimension, textline_cells=text_cells, char_cells=[], word_cells=[], has_lines=len(text_cells) > 0, has_words=False, has_chars=False, ) def get_text_cells(self) -> Iterable[TextCell]: return self._compute_text_cells() def get_bitmap_rects(self, scale: float = 1) -> Iterable[BoundingBox]: AREA_THRESHOLD = 0 # 32 * 32 for i in range(len(self._dpage["images"])): bitmap = self._dpage["images"][i] cropbox = BoundingBox.from_tuple( bitmap["box"], origin=CoordOrigin.BOTTOMLEFT ).to_top_left_origin(self.get_size().height) if cropbox.area() > AREA_THRESHOLD: cropbox = cropbox.scaled(scale=scale) yield cropbox def get_page_image( self, scale: float = 1, cropbox: Optional[BoundingBox] = None ) -> Image.Image: page_size = self.get_size() if not cropbox: cropbox = BoundingBox( l=0, r=page_size.width, t=0, b=page_size.height, coord_origin=CoordOrigin.TOPLEFT, ) padbox = BoundingBox( l=0, r=0, t=0, b=0, coord_origin=CoordOrigin.BOTTOMLEFT ) else: padbox = cropbox.to_bottom_left_origin(page_size.height).model_copy() padbox.r = page_size.width - padbox.r padbox.t = page_size.height - padbox.t image = ( self._ppage.render( scale=scale * 1.5, rotation=0, # no additional rotation crop=padbox.as_tuple(), ) .to_pil() .resize(size=(round(cropbox.width * scale), round(cropbox.height * scale))) ) # We resize the image from 1.5x the given scale to make it sharper. return image def get_size(self) -> Size: return Size(width=self._ppage.get_width(), height=self._ppage.get_height()) def unload(self): self._ppage = None self._dpage = None class DoclingParseDocumentBackend(PdfDocumentBackend): def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]): super().__init__(in_doc, path_or_stream) self._pdoc = pdfium.PdfDocument(self.path_or_stream) self.parser = pdf_parser_v1() success = False if isinstance(self.path_or_stream, BytesIO): success = self.parser.load_document_from_bytesio( self.document_hash, self.path_or_stream ) elif isinstance(self.path_or_stream, Path): success = self.parser.load_document( self.document_hash, str(self.path_or_stream) ) if not success: raise RuntimeError( f"docling-parse could not load document with hash {self.document_hash}." ) def page_count(self) -> int: return len(self._pdoc) # To be replaced with docling-parse API def load_page(self, page_no: int) -> DoclingParsePageBackend: return DoclingParsePageBackend( self.parser, self.document_hash, page_no, self._pdoc[page_no] ) def is_valid(self) -> bool: return self.page_count() > 0 def unload(self): super().unload() self.parser.unload_document(self.document_hash) self._pdoc.close() self._pdoc = None
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/docling_parse_v2_backend.py
docling/backend/docling_parse_v2_backend.py
import logging import random from collections.abc import Iterable from io import BytesIO from pathlib import Path from typing import TYPE_CHECKING, List, Optional, Union import pypdfium2 as pdfium from docling_core.types.doc import BoundingBox, CoordOrigin from docling_core.types.doc.page import ( BoundingRectangle, PdfPageBoundaryType, PdfPageGeometry, SegmentedPdfPage, TextCell, ) from docling_parse.pdf_parsers import pdf_parser_v2 from PIL import Image, ImageDraw from pypdfium2 import PdfPage from docling.backend.pdf_backend import PdfDocumentBackend, PdfPageBackend from docling.backend.pypdfium2_backend import get_pdf_page_geometry from docling.datamodel.base_models import Size from docling.utils.locks import pypdfium2_lock if TYPE_CHECKING: from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) class DoclingParseV2PageBackend(PdfPageBackend): def __init__( self, parser: pdf_parser_v2, document_hash: str, page_no: int, page_obj: PdfPage ): self._ppage = page_obj parsed_page = parser.parse_pdf_from_key_on_page(document_hash, page_no) self.valid = "pages" in parsed_page and len(parsed_page["pages"]) == 1 if self.valid: self._dpage = parsed_page["pages"][0] else: _log.info( f"An error occurred when loading page {page_no} of document {document_hash}." ) def is_valid(self) -> bool: return self.valid def _compute_text_cells(self) -> List[TextCell]: """Compute text cells from docling-parse v2 data.""" cells: List[TextCell] = [] cell_counter = 0 if not self.valid: return cells page_size = self.get_size() parser_width = self._dpage["sanitized"]["dimension"]["width"] parser_height = self._dpage["sanitized"]["dimension"]["height"] cells_data = self._dpage["sanitized"]["cells"]["data"] cells_header = self._dpage["sanitized"]["cells"]["header"] for i, cell_data in enumerate(cells_data): x0 = cell_data[cells_header.index("x0")] y0 = cell_data[cells_header.index("y0")] x1 = cell_data[cells_header.index("x1")] y1 = cell_data[cells_header.index("y1")] if x1 < x0: x0, x1 = x1, x0 if y1 < y0: y0, y1 = y1, y0 text_piece = cell_data[cells_header.index("text")] cells.append( TextCell( index=cell_counter, text=text_piece, orig=text_piece, from_ocr=False, rect=BoundingRectangle.from_bounding_box( BoundingBox( l=x0 * page_size.width / parser_width, b=y0 * page_size.height / parser_height, r=x1 * page_size.width / parser_width, t=y1 * page_size.height / parser_height, coord_origin=CoordOrigin.BOTTOMLEFT, ) ).to_top_left_origin(page_size.height), ) ) cell_counter += 1 return cells def get_text_in_rect(self, bbox: BoundingBox) -> str: if not self.valid: return "" # Find intersecting cells on the page text_piece = "" page_size = self.get_size() parser_width = self._dpage["sanitized"]["dimension"]["width"] parser_height = self._dpage["sanitized"]["dimension"]["height"] scale = ( 1 # FIX - Replace with param in get_text_in_rect across backends (optional) ) cells_data = self._dpage["sanitized"]["cells"]["data"] cells_header = self._dpage["sanitized"]["cells"]["header"] for i, cell_data in enumerate(cells_data): x0 = cell_data[cells_header.index("x0")] y0 = cell_data[cells_header.index("y0")] x1 = cell_data[cells_header.index("x1")] y1 = cell_data[cells_header.index("y1")] cell_bbox = BoundingBox( l=x0 * scale * page_size.width / parser_width, b=y0 * scale * page_size.height / parser_height, r=x1 * scale * page_size.width / parser_width, t=y1 * scale * page_size.height / parser_height, coord_origin=CoordOrigin.BOTTOMLEFT, ).to_top_left_origin(page_height=page_size.height * scale) overlap_frac = cell_bbox.intersection_over_self(bbox) if overlap_frac > 0.5: if len(text_piece) > 0: text_piece += " " text_piece += cell_data[cells_header.index("text")] return text_piece def get_segmented_page(self) -> Optional[SegmentedPdfPage]: if not self.valid: return None text_cells = self._compute_text_cells() # Get the PDF page geometry from pypdfium2 dimension = get_pdf_page_geometry(self._ppage) # Create SegmentedPdfPage return SegmentedPdfPage( dimension=dimension, textline_cells=text_cells, char_cells=[], word_cells=[], has_textlines=len(text_cells) > 0, has_words=False, has_chars=False, ) def get_text_cells(self) -> Iterable[TextCell]: return self._compute_text_cells() def get_bitmap_rects(self, scale: float = 1) -> Iterable[BoundingBox]: AREA_THRESHOLD = 0 # 32 * 32 images = self._dpage["sanitized"]["images"]["data"] images_header = self._dpage["sanitized"]["images"]["header"] for row in images: x0 = row[images_header.index("x0")] y0 = row[images_header.index("y0")] x1 = row[images_header.index("x1")] y1 = row[images_header.index("y1")] cropbox = BoundingBox.from_tuple( (x0, y0, x1, y1), origin=CoordOrigin.BOTTOMLEFT ).to_top_left_origin(self.get_size().height) if cropbox.area() > AREA_THRESHOLD: cropbox = cropbox.scaled(scale=scale) yield cropbox def get_page_image( self, scale: float = 1, cropbox: Optional[BoundingBox] = None ) -> Image.Image: page_size = self.get_size() if not cropbox: cropbox = BoundingBox( l=0, r=page_size.width, t=0, b=page_size.height, coord_origin=CoordOrigin.TOPLEFT, ) padbox = BoundingBox( l=0, r=0, t=0, b=0, coord_origin=CoordOrigin.BOTTOMLEFT ) else: padbox = cropbox.to_bottom_left_origin(page_size.height).model_copy() padbox.r = page_size.width - padbox.r padbox.t = page_size.height - padbox.t with pypdfium2_lock: image = ( self._ppage.render( scale=scale * 1.5, rotation=0, # no additional rotation crop=padbox.as_tuple(), ) .to_pil() .resize( size=(round(cropbox.width * scale), round(cropbox.height * scale)) ) ) # We resize the image from 1.5x the given scale to make it sharper. return image def get_size(self) -> Size: with pypdfium2_lock: return Size(width=self._ppage.get_width(), height=self._ppage.get_height()) def unload(self): self._ppage = None self._dpage = None class DoclingParseV2DocumentBackend(PdfDocumentBackend): def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]): super().__init__(in_doc, path_or_stream) with pypdfium2_lock: self._pdoc = pdfium.PdfDocument(self.path_or_stream) self.parser = pdf_parser_v2("fatal") success = False if isinstance(self.path_or_stream, BytesIO): success = self.parser.load_document_from_bytesio( self.document_hash, self.path_or_stream ) elif isinstance(self.path_or_stream, Path): success = self.parser.load_document( self.document_hash, str(self.path_or_stream) ) if not success: raise RuntimeError( f"docling-parse v2 could not load document {self.document_hash}." ) def page_count(self) -> int: # return len(self._pdoc) # To be replaced with docling-parse API len_1 = len(self._pdoc) len_2 = self.parser.number_of_pages(self.document_hash) if len_1 != len_2: _log.error(f"Inconsistent number of pages: {len_1}!={len_2}") return len_2 def load_page(self, page_no: int) -> DoclingParseV2PageBackend: with pypdfium2_lock: return DoclingParseV2PageBackend( self.parser, self.document_hash, page_no, self._pdoc[page_no] ) def is_valid(self) -> bool: return self.page_count() > 0 def unload(self): super().unload() self.parser.unload_document(self.document_hash) with pypdfium2_lock: self._pdoc.close() self._pdoc = None
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/msword_backend.py
docling/backend/msword_backend.py
import logging import re from copy import deepcopy from io import BytesIO from pathlib import Path from typing import Any, Callable, Final, Optional, Union from docling_core.types.doc import ( ContentLayer, DocItemLabel, DoclingDocument, DocumentOrigin, GroupLabel, ImageRef, ListGroup, NodeItem, RefItem, RichTableCell, TableCell, TableData, TableItem, ) from docling_core.types.doc.document import Formatting, Script from docx import Document from docx.document import Document as DocxDocument from docx.oxml.table import CT_Tc from docx.oxml.xmlchemy import BaseOxmlElement from docx.styles.style import ParagraphStyle from docx.table import Table, _Cell from docx.text.hyperlink import Hyperlink from docx.text.paragraph import Paragraph from docx.text.run import Run from lxml import etree from PIL import Image, UnidentifiedImageError from pydantic import AnyUrl from typing_extensions import override from docling.backend.abstract_backend import DeclarativeDocumentBackend from docling.backend.docx.drawingml.utils import ( get_docx_to_pdf_converter, get_pil_from_dml_docx, ) from docling.backend.docx.latex.omml import oMath2Latex from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) class MsWordDocumentBackend(DeclarativeDocumentBackend): _BLIP_NAMESPACES: Final = { "a": "http://schemas.openxmlformats.org/drawingml/2006/main", "r": "http://schemas.openxmlformats.org/officeDocument/2006/relationships", "w": "http://schemas.openxmlformats.org/wordprocessingml/2006/main", "wp": "http://schemas.openxmlformats.org/drawingml/2006/wordprocessingDrawing", "mc": "http://schemas.openxmlformats.org/markup-compatibility/2006", "v": "urn:schemas-microsoft-com:vml", "wps": "http://schemas.microsoft.com/office/word/2010/wordprocessingShape", "w10": "urn:schemas-microsoft-com:office:word", "a14": "http://schemas.microsoft.com/office/drawing/2010/main", } @override def __init__( self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path] ) -> None: super().__init__(in_doc, path_or_stream) self.XML_KEY = ( "{http://schemas.openxmlformats.org/wordprocessingml/2006/main}val" ) self.xml_namespaces = { "w": "http://schemas.microsoft.com/office/word/2003/wordml" } self.blip_xpath_expr = etree.XPath( ".//a:blip", namespaces=MsWordDocumentBackend._BLIP_NAMESPACES ) # self.initialise(path_or_stream) # Word file: self.path_or_stream: Union[BytesIO, Path] = path_or_stream self.valid: bool = False # Initialise the parents for the hierarchy self.max_levels: int = 10 self.level_at_new_list: Optional[int] = None self.parents: dict[int, Optional[NodeItem]] = {} self.numbered_headers: dict[int, int] = {} self.equation_bookends: str = "<eq>{EQ}</eq>" # Track processed textbox elements to avoid duplication self.processed_textbox_elements: list[int] = [] self.docx_to_pdf_converter: Optional[Callable] = None self.docx_to_pdf_converter_init = False self.display_drawingml_warning = True for i in range(-1, self.max_levels): self.parents[i] = None self.level = 0 self.listIter = 0 # Track list counters per numId and ilvl self.list_counters: dict[tuple[int, int], int] = {} # Set starting content layer self.content_layer = ContentLayer.BODY self.history: dict[str, Any] = { "names": [None], "levels": [None], "numids": [None], "indents": [None], } self.docx_obj = self.load_msword_file( path_or_stream=self.path_or_stream, document_hash=self.document_hash ) if self.docx_obj: self.valid = True @override def is_valid(self) -> bool: return self.valid @classmethod @override def supports_pagination(cls) -> bool: return False @override def unload(self): if isinstance(self.path_or_stream, BytesIO): self.path_or_stream.close() self.path_or_stream = None @classmethod @override def supported_formats(cls) -> set[InputFormat]: return {InputFormat.DOCX} @override def convert(self) -> DoclingDocument: """Parses the DOCX into a structured document model. Returns: The parsed document. """ origin = DocumentOrigin( filename=self.file.name or "file", mimetype="application/vnd.openxmlformats-officedocument.wordprocessingml.document", binary_hash=self.document_hash, ) doc = DoclingDocument(name=self.file.stem or "file", origin=origin) if self.is_valid(): assert self.docx_obj is not None doc, _ = self._walk_linear(self.docx_obj.element.body, doc) self._add_header_footer(self.docx_obj, doc) return doc else: raise RuntimeError( f"Cannot convert doc with {self.document_hash} because the backend failed to init." ) @staticmethod def load_msword_file( path_or_stream: Union[BytesIO, Path], document_hash: str ) -> DocxDocument: try: if isinstance(path_or_stream, BytesIO): return Document(path_or_stream) elif isinstance(path_or_stream, Path): return Document(str(path_or_stream)) else: return None except Exception as e: raise RuntimeError( f"MsWordDocumentBackend could not load document with hash {document_hash}" ) from e def _update_history( self, name: str, level: Optional[int], numid: Optional[int], ilevel: Optional[int], ): self.history["names"].append(name) self.history["levels"].append(level) self.history["numids"].append(numid) self.history["indents"].append(ilevel) def _prev_name(self) -> Optional[str]: return self.history["names"][-1] def _prev_level(self) -> Optional[int]: return self.history["levels"][-1] def _prev_numid(self) -> Optional[int]: return self.history["numids"][-1] def _prev_indent(self) -> Optional[int]: return self.history["indents"][-1] def _get_level(self) -> int: """Return the first None index.""" for k, v in self.parents.items(): if k >= 0 and v is None: return k return 0 def _walk_linear( self, body: BaseOxmlElement, doc: DoclingDocument, # parent: ) -> tuple[DoclingDocument, list[RefItem]]: added_elements = [] for element in body: tag_name = etree.QName(element).localname # Check for Inline Images (blip elements) drawing_blip = self.blip_xpath_expr(element) drawingml_els = element.findall( ".//w:drawing", namespaces=MsWordDocumentBackend._BLIP_NAMESPACES ) # Check for textbox content - check multiple textbox formats # Only process if the element hasn't been processed before element_id = id(element) if element_id not in self.processed_textbox_elements: # Modern Word textboxes txbx_xpath = etree.XPath( ".//w:txbxContent|.//v:textbox//w:p", namespaces=MsWordDocumentBackend._BLIP_NAMESPACES, ) textbox_elements = txbx_xpath(element) # No modern textboxes found, check for alternate/legacy textbox formats if not textbox_elements and tag_name in ["drawing", "pict"]: # Additional checks for textboxes in DrawingML and VML formats alt_txbx_xpath = etree.XPath( ".//wps:txbx//w:p|.//w10:wrap//w:p|.//a:p//a:t", namespaces=MsWordDocumentBackend._BLIP_NAMESPACES, ) textbox_elements = alt_txbx_xpath(element) # Check for shape text that's not in a standard textbox if not textbox_elements: shape_text_xpath = etree.XPath( ".//a:bodyPr/ancestor::*//a:t|.//a:txBody//a:t", namespaces=MsWordDocumentBackend._BLIP_NAMESPACES, ) shape_text_elements = shape_text_xpath(element) if shape_text_elements: # Create custom text elements from shape text text_content = " ".join( [t.text for t in shape_text_elements if t.text] ) if text_content.strip(): _log.debug(f"Found shape text: {text_content[:50]}...") # Create a paragraph-like element to process with standard handler level = self._get_level() shape_group = doc.add_group( label=GroupLabel.SECTION, parent=self.parents[level - 1], name="shape-text", content_layer=self.content_layer, ) added_elements.append(shape_group.get_ref()) doc.add_text( label=DocItemLabel.TEXT, parent=shape_group, text=text_content, content_layer=self.content_layer, ) if textbox_elements: # Mark the parent element as processed self.processed_textbox_elements.append(element_id) # Also mark all found textbox elements as processed for tb_element in textbox_elements: self.processed_textbox_elements.append(id(tb_element)) _log.debug( f"Found textbox content with {len(textbox_elements)} elements" ) tbc = self._handle_textbox_content(textbox_elements, doc) added_elements.extend(tbc) # Check for Tables if tag_name == "tbl": try: t = self._handle_tables(element, doc) added_elements.extend(t) except Exception: _log.debug("could not parse a table, broken docx table") # Check for Image elif drawing_blip: pics = self._handle_pictures(drawing_blip, doc) added_elements.extend(pics) # Check for Text after the Image if ( tag_name == "p" and element.find( ".//w:t", namespaces=MsWordDocumentBackend._BLIP_NAMESPACES ) is not None ): te1 = self._handle_text_elements(element, doc) added_elements.extend(te1) # Check for DrawingML elements elif drawingml_els: if ( self.docx_to_pdf_converter is None and self.docx_to_pdf_converter_init is False ): self.docx_to_pdf_converter = get_docx_to_pdf_converter() self.docx_to_pdf_converter_init = True if self.docx_to_pdf_converter is None: if self.display_drawingml_warning: if self.docx_to_pdf_converter is None: _log.warning( "Found DrawingML elements in document, but no DOCX to PDF converters. " "If you want these exported, make sure you have " "LibreOffice binary in PATH or specify its path with DOCLING_LIBREOFFICE_CMD." ) self.display_drawingml_warning = False else: self._handle_drawingml(doc=doc, drawingml_els=drawingml_els) # Check for the sdt containers, like table of contents elif tag_name == "sdt": sdt_content = element.find( ".//w:sdtContent", namespaces=MsWordDocumentBackend._BLIP_NAMESPACES ) if sdt_content is not None: # Iterate paragraphs, runs, or text inside <w:sdtContent>. paragraphs = sdt_content.findall( ".//w:p", namespaces=MsWordDocumentBackend._BLIP_NAMESPACES ) for p in paragraphs: te = self._handle_text_elements(p, doc) added_elements.extend(te) # Check for Text elif tag_name == "p": # "tcPr", "sectPr" te = self._handle_text_elements(element, doc) added_elements.extend(te) else: _log.debug(f"Ignoring element in DOCX with tag: {tag_name}") return doc, added_elements def _str_to_int( self, s: Optional[str], default: Optional[int] = 0 ) -> Optional[int]: if s is None: return None try: return int(s) except ValueError: return default def _split_text_and_number(self, input_string: str) -> list[str]: match = re.match(r"(\D+)(\d+)$|^(\d+)(\D+)", input_string) if match: parts = list(filter(None, match.groups())) return parts else: return [input_string] def _get_numId_and_ilvl( self, paragraph: Paragraph ) -> tuple[Optional[int], Optional[int]]: # Access the XML element of the paragraph numPr = paragraph._element.find( ".//w:numPr", namespaces=paragraph._element.nsmap ) if numPr is not None: # Get the numId element and extract the value numId_elem = numPr.find("w:numId", namespaces=paragraph._element.nsmap) ilvl_elem = numPr.find("w:ilvl", namespaces=paragraph._element.nsmap) numId = numId_elem.get(self.XML_KEY) if numId_elem is not None else None ilvl = ilvl_elem.get(self.XML_KEY) if ilvl_elem is not None else None return self._str_to_int(numId, None), self._str_to_int(ilvl, None) return None, None # If the paragraph is not part of a list def _get_list_counter(self, numid: int, ilvl: int) -> int: """Get and increment the counter for a specific numId and ilvl combination.""" key = (numid, ilvl) if key not in self.list_counters: self.list_counters[key] = 0 self.list_counters[key] += 1 return self.list_counters[key] def _reset_list_counters_for_new_sequence(self, numid: int): """Reset counters when starting a new numbering sequence.""" # Reset all counters for this numid keys_to_reset = [key for key in self.list_counters.keys() if key[0] == numid] for key in keys_to_reset: self.list_counters[key] = 0 def _is_numbered_list(self, numId: int, ilvl: int) -> bool: """Check if a list is numbered based on its numFmt value.""" try: # Access the numbering part of the document if not hasattr(self.docx_obj, "part") or not hasattr( self.docx_obj.part, "package" ): return False numbering_part = None # Find the numbering part for part in self.docx_obj.part.package.parts: if "numbering" in part.partname: numbering_part = part break if numbering_part is None: return False # Parse the numbering XML numbering_root = numbering_part.element namespaces = { "w": "http://schemas.openxmlformats.org/wordprocessingml/2006/main" } # Find the numbering definition with the given numId num_xpath = f".//w:num[@w:numId='{numId}']" num_element = numbering_root.find(num_xpath, namespaces=namespaces) if num_element is None: return False # Get the abstractNumId from the num element abstract_num_id_elem = num_element.find( ".//w:abstractNumId", namespaces=namespaces ) if abstract_num_id_elem is None: return False abstract_num_id = abstract_num_id_elem.get( "{http://schemas.openxmlformats.org/wordprocessingml/2006/main}val" ) if abstract_num_id is None: return False # Find the abstract numbering definition abstract_num_xpath = ( f".//w:abstractNum[@w:abstractNumId='{abstract_num_id}']" ) abstract_num_element = numbering_root.find( abstract_num_xpath, namespaces=namespaces ) if abstract_num_element is None: return False # Find the level definition for the given ilvl lvl_xpath = f".//w:lvl[@w:ilvl='{ilvl}']" lvl_element = abstract_num_element.find(lvl_xpath, namespaces=namespaces) if lvl_element is None: return False # Get the numFmt element num_fmt_element = lvl_element.find(".//w:numFmt", namespaces=namespaces) if num_fmt_element is None: return False num_fmt = num_fmt_element.get( "{http://schemas.openxmlformats.org/wordprocessingml/2006/main}val" ) # Numbered formats include: decimal, lowerRoman, upperRoman, lowerLetter, upperLetter # Bullet formats include: bullet numbered_formats = { "decimal", "lowerRoman", "upperRoman", "lowerLetter", "upperLetter", "decimalZero", } return num_fmt in numbered_formats except Exception as e: _log.debug(f"Error determining if list is numbered: {e}") return False def _get_heading_and_level(self, style_label: str) -> tuple[str, Optional[int]]: parts = self._split_text_and_number(style_label) if len(parts) == 2: parts.sort() label_str: str = "" label_level: Optional[int] = 0 if parts[0].strip().lower() == "heading": label_str = "Heading" label_level = self._str_to_int(parts[1], None) if parts[1].strip().lower() == "heading": label_str = "Heading" label_level = self._str_to_int(parts[0], None) return label_str, label_level return style_label, None def _get_label_and_level(self, paragraph: Paragraph) -> tuple[str, Optional[int]]: if paragraph.style is None: return "Normal", None label: str = paragraph.style.style_id name: str = paragraph.style.name or "" base_style_label: Optional[str] = None base_style_name: Optional[str] = None if isinstance( base_style := getattr(paragraph.style, "base_style", None), ParagraphStyle ): base_style_label = base_style.style_id base_style_name = base_style.name if not label: return "Normal", None if ":" in label: parts = label.split(":") if len(parts) == 2: return parts[0], self._str_to_int(parts[1], None) if "heading" in label.lower(): return self._get_heading_and_level(label) if "heading" in name.lower(): return self._get_heading_and_level(name) if base_style_label and "heading" in base_style_label.lower(): return self._get_heading_and_level(base_style_label) if base_style_name and "heading" in base_style_name.lower(): return self._get_heading_and_level(base_style_name) return label, None @classmethod def _get_format_from_run(cls, run: Run) -> Optional[Formatting]: # The .bold and .italic properties are booleans, but .underline can be an enum # like WD_UNDERLINE.THICK (value 6), so we need to convert it to a boolean is_bold = run.bold or False is_italic = run.italic or False is_strikethrough = run.font.strike or False # Convert any non-None underline value to True is_underline = bool(run.underline is not None and run.underline) is_sub = run.font.subscript or False is_sup = run.font.superscript or False script = Script.SUB if is_sub else Script.SUPER if is_sup else Script.BASELINE return Formatting( bold=is_bold, italic=is_italic, underline=is_underline, strikethrough=is_strikethrough, script=script, ) def _get_paragraph_elements(self, paragraph: Paragraph): """ Extract paragraph elements along with their formatting and hyperlink """ # for now retain empty paragraphs for backwards compatibility: if paragraph.text.strip() == "": return [("", None, None)] paragraph_elements: list[ tuple[str, Optional[Formatting], Optional[Union[AnyUrl, Path]]] ] = [] group_text = "" previous_format = None # Iterate over the runs of the paragraph and group them by format for c in paragraph.iter_inner_content(): if isinstance(c, Hyperlink): text = c.text hyperlink = Path(c.address) format = ( self._get_format_from_run(c.runs[0]) if c.runs and len(c.runs) > 0 else None ) elif isinstance(c, Run): text = c.text hyperlink = None format = self._get_format_from_run(c) else: continue if (len(text.strip()) and format != previous_format) or ( hyperlink is not None ): # If the style changes for a non empty text, add the previous group if len(group_text.strip()) > 0: paragraph_elements.append( (group_text.strip(), previous_format, None) ) group_text = "" # If there is a hyperlink, add it immediately if hyperlink is not None: paragraph_elements.append((text.strip(), format, hyperlink)) text = "" else: previous_format = format group_text += text # Format the last group if len(group_text.strip()) > 0: paragraph_elements.append((group_text.strip(), format, None)) return paragraph_elements def _get_paragraph_position(self, paragraph_element): """Extract vertical position information from paragraph element.""" # First try to directly get the index from w:p element that has an order-related attribute if ( hasattr(paragraph_element, "getparent") and paragraph_element.getparent() is not None ): parent = paragraph_element.getparent() # Get all paragraph siblings paragraphs = [ p for p in parent.getchildren() if etree.QName(p).localname == "p" ] # Find index of current paragraph within its siblings try: paragraph_index = paragraphs.index(paragraph_element) return paragraph_index # Use index as position for consistent ordering except ValueError: pass # Look for position hints in element attributes and ancestor elements for elem in (*[paragraph_element], *paragraph_element.iterancestors()): # Check for direct position attributes for attr_name in ["y", "top", "positionY", "y-position", "position"]: value = elem.get(attr_name) if value: try: # Remove any non-numeric characters (like 'pt', 'px', etc.) clean_value = re.sub(r"[^0-9.]", "", value) if clean_value: return float(clean_value) except (ValueError, TypeError): pass # Check for position in transform attribute transform = elem.get("transform") if transform: # Extract translation component from transform matrix match = re.search(r"translate\([^,]+,\s*([0-9.]+)", transform) if match: try: return float(match.group(1)) except ValueError: pass # Check for anchors or relative position indicators in Word format # 'dist' attributes can indicate relative positioning for attr_name in ["distT", "distB", "anchor", "relativeFrom"]: if elem.get(attr_name) is not None: return elem.sourceline # Use the XML source line number as fallback # For VML shapes, look for specific attributes for ns_uri in paragraph_element.nsmap.values(): if "vml" in ns_uri: # Try to extract position from style attribute style = paragraph_element.get("style") if style: match = re.search(r"top:([0-9.]+)pt", style) if match: try: return float(match.group(1)) except ValueError: pass # If no better position indicator found, use XML source line number as proxy for order return ( paragraph_element.sourceline if hasattr(paragraph_element, "sourceline") else None ) def _collect_textbox_paragraphs(self, textbox_elements): """Collect and organize paragraphs from textbox elements.""" processed_paragraphs = [] container_paragraphs = {} for element in textbox_elements: element_id = id(element) # Skip if we've already processed this exact element if element_id in processed_paragraphs: continue tag_name = etree.QName(element).localname processed_paragraphs.append(element_id) # Handle paragraphs directly found (VML textboxes) if tag_name == "p": # Find the containing textbox or shape element container_id = None for ancestor in element.iterancestors(): if any(ns in ancestor.tag for ns in ["textbox", "shape", "txbx"]): container_id = id(ancestor) break if container_id not in container_paragraphs: container_paragraphs[container_id] = [] container_paragraphs[container_id].append( (element, self._get_paragraph_position(element)) ) # Handle txbxContent elements (Word DrawingML textboxes) elif tag_name == "txbxContent": paragraphs = element.findall(".//w:p", namespaces=element.nsmap) container_id = id(element) if container_id not in container_paragraphs: container_paragraphs[container_id] = [] for p in paragraphs: p_id = id(p) if p_id not in processed_paragraphs: processed_paragraphs.append(p_id) container_paragraphs[container_id].append( (p, self._get_paragraph_position(p)) ) else: # Try to extract any paragraphs from unknown elements paragraphs = element.findall(".//w:p", namespaces=element.nsmap) container_id = id(element) if container_id not in container_paragraphs: container_paragraphs[container_id] = [] for p in paragraphs: p_id = id(p) if p_id not in processed_paragraphs: processed_paragraphs.append(p_id) container_paragraphs[container_id].append( (p, self._get_paragraph_position(p)) ) return container_paragraphs def _handle_textbox_content( self, textbox_elements: list, doc: DoclingDocument, ) -> list[RefItem]: elem_ref: list[RefItem] = [] """Process textbox content and add it to the document structure.""" level = self._get_level() # Create a textbox group to contain all text from the textbox textbox_group = doc.add_group( label=GroupLabel.SECTION, parent=self.parents[level - 1], name="textbox", content_layer=self.content_layer, ) elem_ref.append(textbox_group.get_ref()) # Set this as the current parent to ensure textbox content # is properly nested in document structure original_parent = self.parents[level] self.parents[level] = textbox_group # Collect and organize paragraphs container_paragraphs = self._collect_textbox_paragraphs(textbox_elements) # Process all paragraphs all_paragraphs = [] # Sort paragraphs within each container, then process containers for paragraphs in container_paragraphs.values(): # Sort by vertical position within each container sorted_container_paragraphs = sorted( paragraphs, key=lambda x: ( x[1] is None, x[1] if x[1] is not None else float("inf"), ), ) # Add the sorted paragraphs to our processing list all_paragraphs.extend(sorted_container_paragraphs) # Track processed paragraphs to avoid duplicates (same content and position) processed_paragraphs = set() # Process all the paragraphs for p, position in all_paragraphs: # Create paragraph object to get text content paragraph = Paragraph(p, self.docx_obj) text_content = paragraph.text # Create a unique identifier based on content and position paragraph_id = (text_content, position) # Skip if this paragraph (same content and position) was already processed if paragraph_id in processed_paragraphs: _log.debug( f"Skipping duplicate paragraph: content='{text_content[:50]}...', position={position}" ) continue # Mark this paragraph as processed processed_paragraphs.add(paragraph_id) elem_ref.extend(self._handle_text_elements(p, doc)) # Restore original parent self.parents[level] = original_parent return elem_ref def _handle_equations_in_text(self, element, text): only_texts = [] only_equations = [] texts_and_equations = [] for subt in element.iter(): tag_name = etree.QName(subt).localname if tag_name == "t" and "math" not in subt.tag: if isinstance(subt.text, str): only_texts.append(subt.text)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
true
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/noop_backend.py
docling/backend/noop_backend.py
import logging from io import BytesIO from pathlib import Path from typing import Set, Union from docling.backend.abstract_backend import AbstractDocumentBackend from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) class NoOpBackend(AbstractDocumentBackend): """ A no-op backend that only validates input existence. Used e.g. for audio files where actual processing is handled by the ASR pipeline. """ def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]): super().__init__(in_doc, path_or_stream) _log.debug(f"NoOpBackend initialized for: {path_or_stream}") # Validate input try: if isinstance(self.path_or_stream, BytesIO): # Check if stream has content self.valid = len(self.path_or_stream.getvalue()) > 0 _log.debug( f"BytesIO stream length: {len(self.path_or_stream.getvalue())}" ) elif isinstance(self.path_or_stream, Path): # Check if file exists self.valid = self.path_or_stream.exists() _log.debug(f"File exists: {self.valid}") else: self.valid = False except Exception as e: _log.error(f"NoOpBackend validation failed: {e}") self.valid = False def is_valid(self) -> bool: return self.valid @classmethod def supports_pagination(cls) -> bool: return False @classmethod def supported_formats(cls) -> Set[InputFormat]: return set(InputFormat)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/html_backend.py
docling/backend/html_backend.py
import base64 import logging import os import re import warnings from contextlib import contextmanager from copy import deepcopy from io import BytesIO from pathlib import Path from typing import Final, Optional, Union, cast from urllib.parse import urljoin, urlparse import requests from bs4 import BeautifulSoup, NavigableString, PageElement, Tag from bs4.element import PreformattedString from docling_core.types.doc import ( DocItem, DocItemLabel, DoclingDocument, DocumentOrigin, GroupItem, GroupLabel, PictureItem, RefItem, RichTableCell, TableCell, TableData, TableItem, TextItem, ) from docling_core.types.doc.document import ContentLayer, Formatting, ImageRef, Script from PIL import Image, UnidentifiedImageError from pydantic import AnyUrl, BaseModel, ValidationError from typing_extensions import override from docling.backend.abstract_backend import ( DeclarativeDocumentBackend, ) from docling.datamodel.backend_options import HTMLBackendOptions from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument from docling.exceptions import OperationNotAllowed _log = logging.getLogger(__name__) DEFAULT_IMAGE_WIDTH = 128 DEFAULT_IMAGE_HEIGHT = 128 # Tags that initiate distinct Docling items _BLOCK_TAGS: Final = { "address", "details", "figure", "footer", "img", "h1", "h2", "h3", "h4", "h5", "h6", "ol", "p", "pre", "summary", "table", "ul", } # Block-level elements that should not appear inside <p> _PARA_BREAKERS = { "address", "article", "aside", "blockquote", "div", "dl", "fieldset", "figcaption", "figure", "footer", "form", "h1", "h2", "h3", "h4", "h5", "h6", "header", "hr", "main", "nav", "ol", "ul", "li", "p", # <p> inside <p> also forces closing "pre", "section", "table", "thead", "tbody", "tfoot", "tr", "td", } _CODE_TAG_SET: Final = {"code", "kbd", "samp"} _FORMAT_TAG_MAP: Final = { "b": {"bold": True}, "strong": {"bold": True}, "i": {"italic": True}, "em": {"italic": True}, "var": {"italic": True}, # "mark", # "small", "s": {"strikethrough": True}, "del": {"strikethrough": True}, "u": {"underline": True}, "ins": {"underline": True}, "sub": {"script": Script.SUB}, "sup": {"script": Script.SUPER}, **{k: {} for k in _CODE_TAG_SET}, } class _Context(BaseModel): list_ordered_flag_by_ref: dict[str, bool] = {} list_start_by_ref: dict[str, int] = {} class AnnotatedText(BaseModel): text: str hyperlink: Union[AnyUrl, Path, None] = None formatting: Union[Formatting, None] = None code: bool = False class AnnotatedTextList(list): def to_single_text_element(self) -> AnnotatedText: current_h = None current_text = "" current_f = None current_code = False for at in self: t = at.text h = at.hyperlink f = at.formatting c = at.code current_text += t.strip() + " " if f is not None and current_f is None: current_f = f elif f is not None and current_f is not None and f != current_f: _log.warning( f"Clashing formatting: '{f}' and '{current_f}'! Chose '{current_f}'" ) if h is not None and current_h is None: current_h = h elif h is not None and current_h is not None and h != current_h: _log.warning( f"Clashing hyperlinks: '{h}' and '{current_h}'! Chose '{current_h}'" ) current_code = c if c else current_code return AnnotatedText( text=current_text.strip(), hyperlink=current_h, formatting=current_f, code=current_code, ) def simplify_text_elements(self) -> "AnnotatedTextList": simplified = AnnotatedTextList() if not self: return self text = self[0].text hyperlink = self[0].hyperlink formatting = self[0].formatting code = self[0].code last_elm = text for i in range(1, len(self)): if ( hyperlink == self[i].hyperlink and formatting == self[i].formatting and code == self[i].code ): sep = " " if not self[i].text.strip() or not last_elm.strip(): sep = "" text += sep + self[i].text last_elm = self[i].text else: simplified.append( AnnotatedText( text=text, hyperlink=hyperlink, formatting=formatting, code=code ) ) text = self[i].text last_elm = text hyperlink = self[i].hyperlink formatting = self[i].formatting code = self[i].code if text: simplified.append( AnnotatedText( text=text, hyperlink=hyperlink, formatting=formatting, code=code ) ) return simplified def split_by_newline(self): super_list = [] active_annotated_text_list = AnnotatedTextList() for el in self: sub_texts = el.text.split("\n") if len(sub_texts) == 1: active_annotated_text_list.append(el) else: for text in sub_texts: sub_el = deepcopy(el) sub_el.text = text active_annotated_text_list.append(sub_el) super_list.append(active_annotated_text_list) active_annotated_text_list = AnnotatedTextList() if active_annotated_text_list: super_list.append(active_annotated_text_list) return super_list class HTMLDocumentBackend(DeclarativeDocumentBackend): @override def __init__( self, in_doc: InputDocument, path_or_stream: Union[BytesIO, Path], options: HTMLBackendOptions = HTMLBackendOptions(), ): super().__init__(in_doc, path_or_stream, options) self.options: HTMLBackendOptions self.soup: Optional[BeautifulSoup] = None self.path_or_stream: Union[BytesIO, Path] = path_or_stream self.base_path: Optional[str] = str(options.source_uri) # Initialize the parents for the hierarchy self.max_levels = 10 self.level = 0 self.parents: dict[int, Optional[Union[DocItem, GroupItem]]] = {} self.ctx = _Context() for i in range(self.max_levels): self.parents[i] = None self.hyperlink: Union[AnyUrl, Path, None] = None self.format_tags: list[str] = [] try: raw = ( path_or_stream.getvalue() if isinstance(path_or_stream, BytesIO) else Path(path_or_stream).read_bytes() ) self.soup = BeautifulSoup(raw, "html.parser") except Exception as e: raise RuntimeError( "Could not initialize HTML backend for file with " f"hash {self.document_hash}." ) from e @override def is_valid(self) -> bool: return self.soup is not None @classmethod @override def supports_pagination(cls) -> bool: return False @override def unload(self): if isinstance(self.path_or_stream, BytesIO): self.path_or_stream.close() self.path_or_stream = None @classmethod @override def supported_formats(cls) -> set[InputFormat]: return {InputFormat.HTML} @override def convert(self) -> DoclingDocument: _log.debug("Starting HTML conversion...") if not self.is_valid(): raise RuntimeError("Invalid HTML document.") origin = DocumentOrigin( filename=self.file.name or "file", mimetype="text/html", binary_hash=self.document_hash, ) doc = DoclingDocument(name=self.file.stem or "file", origin=origin) assert self.soup is not None # set the title as furniture, since it is part of the document metadata title = self.soup.title if title and self.options.add_title: title_text = title.get_text(separator=" ", strip=True) title_clean = HTMLDocumentBackend._clean_unicode(title_text) doc.add_title( text=title_clean, orig=title_text, content_layer=ContentLayer.FURNITURE, ) # remove script and style tags for tag in self.soup(["script", "noscript", "style"]): tag.decompose() # remove any hidden tag for tag in self.soup(hidden=True): tag.decompose() # fix flow content that is not permitted inside <p> HTMLDocumentBackend._fix_invalid_paragraph_structure(self.soup) content = self.soup.body or self.soup # normalize <br> tags for br in content("br"): br.replace_with(NavigableString("\n")) # set default content layer # Furniture before the first heading rule, except for headers in tables header = None # Find all headers first all_headers = content.find_all(["h1", "h2", "h3", "h4", "h5", "h6"]) # Keep only those that do NOT have a <table> in a parent chain clean_headers = [h for h in all_headers if not h.find_parent("table")] # Pick the first header from the remaining if len(clean_headers): header = clean_headers[0] # Set starting content layer self.content_layer = ( ContentLayer.BODY if (not self.options.infer_furniture) or (header is None) else ContentLayer.FURNITURE ) # reset context self.ctx = _Context() self._walk(content, doc) return doc @staticmethod def _fix_invalid_paragraph_structure(soup: BeautifulSoup) -> None: """Rewrite <p> elements that contain block-level breakers. This function emulates browser logic when other block-level elements are found inside a <p> element. When a <p> is open and a block-level breaker (e.g., h1-h6, div, table) appears, automatically close the <p>, emit it, then emit the breaker, and if needed open a new <p> for trailing text. Args: soup: The HTML document. The DOM may be rewritten. """ def _start_para(): nonlocal current_p if current_p is None: current_p = soup.new_tag("p") new_nodes.append(current_p) def _flush_para_if_empty(): nonlocal current_p if current_p is not None and not current_p.get_text(strip=True): # remove empty paragraph placeholder if current_p in new_nodes: new_nodes.remove(current_p) current_p = None paragraphs = soup.select(f"p:has({','.join(tag for tag in _PARA_BREAKERS)})") for p in paragraphs: parent = p.parent if parent is None: continue new_nodes = [] current_p = None for node in list(p.contents): if isinstance(node, NavigableString): text = str(node) node.extract() if text.strip(): _start_para() if current_p is not None: current_p.append(NavigableString(text)) # skip whitespace-only text continue if isinstance(node, Tag): node.extract() if node.name in _PARA_BREAKERS: _flush_para_if_empty() new_nodes.append(node) continue else: _start_para() if current_p is not None: current_p.append(node) continue _flush_para_if_empty() siblings = list(parent.children) try: idx = siblings.index(p) except ValueError: # p might have been removed continue p.extract() for n in reversed(new_nodes): parent.insert(idx, n) @staticmethod def _is_remote_url(value: str) -> bool: parsed = urlparse(value) return parsed.scheme in {"http", "https", "ftp", "s3", "gs"} def _resolve_relative_path(self, loc: str) -> str: abs_loc = loc if self.base_path: if loc.startswith("//"): # Protocol-relative URL - default to https abs_loc = "https:" + loc elif not loc.startswith(("http://", "https://", "data:", "file://")): if HTMLDocumentBackend._is_remote_url(self.base_path): # remote fetch abs_loc = urljoin(self.base_path, loc) elif self.base_path: # local fetch # For local files, resolve relative to the HTML file location abs_loc = str(Path(self.base_path).parent / loc) _log.debug(f"Resolved location {loc} to {abs_loc}") return abs_loc @staticmethod def group_cell_elements( group_name: str, doc: DoclingDocument, provs_in_cell: list[RefItem], docling_table: TableItem, ) -> RefItem: group_element = doc.add_group( label=GroupLabel.UNSPECIFIED, name=group_name, parent=docling_table, ) for prov in provs_in_cell: group_element.children.append(prov) pr_item = prov.resolve(doc) item_parent = pr_item.parent.resolve(doc) if pr_item.get_ref() in item_parent.children: item_parent.children.remove(pr_item.get_ref()) pr_item.parent = group_element.get_ref() ref_for_rich_cell = group_element.get_ref() return ref_for_rich_cell @staticmethod def process_rich_table_cells( provs_in_cell: list[RefItem], group_name: str, doc: DoclingDocument, docling_table: TableItem, ) -> tuple[bool, Union[RefItem, None]]: rich_table_cell = False ref_for_rich_cell = None if len(provs_in_cell) >= 1: # Cell rich cell has multiple elements, we need to group them rich_table_cell = True ref_for_rich_cell = HTMLDocumentBackend.group_cell_elements( group_name, doc, provs_in_cell, docling_table ) return rich_table_cell, ref_for_rich_cell def _is_rich_table_cell(self, table_cell: Tag) -> bool: """Determine whether an table cell should be parsed as a Docling RichTableCell. A table cell can hold rich content and be parsed with a Docling RichTableCell. However, this requires walking through the content elements and creating Docling node items. If the cell holds only plain text, the parsing is simpler and using a TableCell is prefered. Args: table_cell: The HTML tag representing a table cell. Returns: Whether the cell should be parsed as RichTableCell. """ is_rich: bool = True children = table_cell.find_all(recursive=True) # all descendants of type Tag if not children: content = [ item for item in table_cell.contents if isinstance(item, NavigableString) ] is_rich = len(content) > 1 else: annotations = self._extract_text_and_hyperlink_recursively( table_cell, find_parent_annotation=True ) if not annotations: is_rich = bool(item for item in children if item.name == "img") elif len(annotations) == 1: anno: AnnotatedText = annotations[0] is_rich = bool(anno.formatting) or bool(anno.hyperlink) or anno.code return is_rich def parse_table_data( self, element: Tag, doc: DoclingDocument, docling_table: TableItem, num_rows: int, num_cols: int, ) -> Optional[TableData]: for t in cast(list[Tag], element.find_all(["thead", "tbody"], recursive=False)): t.unwrap() _log.debug(f"The table has {num_rows} rows and {num_cols} cols.") grid: list = [[None for _ in range(num_cols)] for _ in range(num_rows)] data = TableData(num_rows=num_rows, num_cols=num_cols, table_cells=[]) # Iterate over the rows in the table start_row_span = 0 row_idx = -1 # We don't want this recursive to support nested tables for row in element("tr", recursive=False): if not isinstance(row, Tag): continue # For each row, find all the column cells (both <td> and <th>) # We don't want this recursive to support nested tables cells = row(["td", "th"], recursive=False) # Check if cell is in a column header or row header col_header = True row_header = True for html_cell in cells: if isinstance(html_cell, Tag): _, row_span = HTMLDocumentBackend._get_cell_spans(html_cell) if html_cell.name == "td": col_header = False row_header = False elif row_span == 1: row_header = False if not row_header: row_idx += 1 start_row_span = 0 else: start_row_span += 1 # Extract the text content of each cell col_idx = 0 for html_cell in cells: if not isinstance(html_cell, Tag): continue # extract inline formulas for formula in html_cell("inline-formula"): math_parts = formula.text.split("$$") if len(math_parts) == 3: math_formula = f"$${math_parts[1]}$$" formula.replace_with(NavigableString(math_formula)) provs_in_cell: list[RefItem] = [] rich_table_cell = self._is_rich_table_cell(html_cell) if rich_table_cell: # Parse table cell sub-tree for Rich Cells content: with self._use_table_cell_context(): provs_in_cell = self._walk(html_cell, doc) group_name = f"rich_cell_group_{len(doc.tables)}_{col_idx}_{start_row_span + row_idx}" rich_table_cell, ref_for_rich_cell = ( HTMLDocumentBackend.process_rich_table_cells( provs_in_cell, group_name, doc, docling_table ) ) # Extracting text text = HTMLDocumentBackend._clean_unicode( self.get_text(html_cell).strip() ) col_span, row_span = self._get_cell_spans(html_cell) if row_header: row_span -= 1 while ( col_idx < num_cols and grid[row_idx + start_row_span][col_idx] is not None ): col_idx += 1 for r in range(start_row_span, start_row_span + row_span): for c in range(col_span): if row_idx + r < num_rows and col_idx + c < num_cols: grid[row_idx + r][col_idx + c] = text if rich_table_cell: rich_cell = RichTableCell( text=text, row_span=row_span, col_span=col_span, start_row_offset_idx=start_row_span + row_idx, end_row_offset_idx=start_row_span + row_idx + row_span, start_col_offset_idx=col_idx, end_col_offset_idx=col_idx + col_span, column_header=col_header, row_header=((not col_header) and html_cell.name == "th"), ref=ref_for_rich_cell, # points to an artificial group around children ) doc.add_table_cell(table_item=docling_table, cell=rich_cell) else: simple_cell = TableCell( text=text, row_span=row_span, col_span=col_span, start_row_offset_idx=start_row_span + row_idx, end_row_offset_idx=start_row_span + row_idx + row_span, start_col_offset_idx=col_idx, end_col_offset_idx=col_idx + col_span, column_header=col_header, row_header=((not col_header) and html_cell.name == "th"), ) doc.add_table_cell(table_item=docling_table, cell=simple_cell) return data def _walk(self, element: Tag, doc: DoclingDocument) -> list[RefItem]: """Parse an XML tag by recursively walking its content. While walking, the method buffers inline text across tags like <b> or <span>, emitting text nodes only at block boundaries. Args: element: The XML tag to parse. doc: The Docling document to be updated with the parsed content. """ added_refs: list[RefItem] = [] buffer: AnnotatedTextList = AnnotatedTextList() def _flush_buffer() -> None: if not buffer: return annotated_text_list: AnnotatedTextList = buffer.simplify_text_elements() parts = annotated_text_list.split_by_newline() buffer.clear() if not "".join([el.text for el in annotated_text_list]): return for annotated_text_list in parts: with self._use_inline_group(annotated_text_list, doc): for annotated_text in annotated_text_list: if annotated_text.text.strip(): seg_clean = HTMLDocumentBackend._clean_unicode( annotated_text.text.strip() ) if annotated_text.code: docling_code2 = doc.add_code( parent=self.parents[self.level], text=seg_clean, content_layer=self.content_layer, formatting=annotated_text.formatting, hyperlink=annotated_text.hyperlink, ) added_refs.append(docling_code2.get_ref()) else: docling_text2 = doc.add_text( parent=self.parents[self.level], label=DocItemLabel.TEXT, text=seg_clean, content_layer=self.content_layer, formatting=annotated_text.formatting, hyperlink=annotated_text.hyperlink, ) added_refs.append(docling_text2.get_ref()) for node in element.contents: if isinstance(node, Tag): name = node.name.lower() if name == "img": _flush_buffer() im_ref3 = self._emit_image(node, doc) if im_ref3: added_refs.append(im_ref3) elif name in _FORMAT_TAG_MAP: _flush_buffer() with self._use_format([name]): wk = self._walk(node, doc) added_refs.extend(wk) elif name == "a": with self._use_hyperlink(node): wk2 = self._walk(node, doc) added_refs.extend(wk2) elif name in _BLOCK_TAGS: _flush_buffer() blk = self._handle_block(node, doc) added_refs.extend(blk) elif node.find(_BLOCK_TAGS): _flush_buffer() wk3 = self._walk(node, doc) added_refs.extend(wk3) else: buffer.extend( self._extract_text_and_hyperlink_recursively( node, find_parent_annotation=True, keep_newlines=True ) ) elif isinstance(node, NavigableString) and not isinstance( node, PreformattedString ): if str(node).strip("\n\r") == "": _flush_buffer() else: buffer.extend( self._extract_text_and_hyperlink_recursively( node, find_parent_annotation=True, keep_newlines=True ) ) _flush_buffer() return added_refs @staticmethod def _collect_parent_format_tags(item: PageElement) -> list[str]: tags = [] for format_tag in _FORMAT_TAG_MAP: this_parent = item.parent while this_parent is not None: if this_parent.name == format_tag: tags.append(format_tag) break this_parent = this_parent.parent return tags @property def _formatting(self): kwargs = {} for t in self.format_tags: kwargs.update(_FORMAT_TAG_MAP[t]) if not kwargs: return None return Formatting(**kwargs) def _extract_text_and_hyperlink_recursively( self, item: PageElement, ignore_list=False, find_parent_annotation=False, keep_newlines=False, ) -> AnnotatedTextList: result: AnnotatedTextList = AnnotatedTextList() # If find_parent_annotation, make sure that we keep track of # any a- or formatting-tag that has been present in the # DOM-parents already. if find_parent_annotation: format_tags = self._collect_parent_format_tags(item) this_parent = item.parent while this_parent is not None: if this_parent.name == "a" and this_parent.get("href"): with self._use_format(format_tags): with self._use_hyperlink(this_parent): return self._extract_text_and_hyperlink_recursively( item, ignore_list ) this_parent = this_parent.parent if isinstance(item, PreformattedString): return AnnotatedTextList() if isinstance(item, NavigableString): text = item.strip() code = any(code_tag in self.format_tags for code_tag in _CODE_TAG_SET) if text: return AnnotatedTextList( [ AnnotatedText( text=text, hyperlink=self.hyperlink, formatting=self._formatting, code=code, ) ] ) if keep_newlines and item.strip("\n\r") == "": return AnnotatedTextList( [ AnnotatedText( text="\n", hyperlink=self.hyperlink, formatting=self._formatting, code=code, ) ] ) return AnnotatedTextList() tag = cast(Tag, item) if not ignore_list or (tag.name not in ["ul", "ol"]): for child in tag: if isinstance(child, Tag) and child.name in _FORMAT_TAG_MAP: with self._use_format([child.name]): result.extend( self._extract_text_and_hyperlink_recursively( child, ignore_list, keep_newlines=keep_newlines ) ) elif isinstance(child, Tag) and child.name == "a": with self._use_hyperlink(child): result.extend( self._extract_text_and_hyperlink_recursively( child, ignore_list, keep_newlines=keep_newlines ) ) else: # Recursively get the child's text content result.extend( self._extract_text_and_hyperlink_recursively( child, ignore_list, keep_newlines=keep_newlines ) ) return result @contextmanager def _use_hyperlink(self, tag: Tag): old_hyperlink: Union[AnyUrl, Path, None] = None new_hyperlink: Union[AnyUrl, Path, None] = None this_href = tag.get("href") if this_href is None: yield None else: if isinstance(this_href, str) and this_href: old_hyperlink = self.hyperlink this_href = self._resolve_relative_path(this_href) # ugly fix for relative links since pydantic does not support them. try: new_hyperlink = AnyUrl(this_href) except ValidationError: new_hyperlink = Path(this_href) self.hyperlink = new_hyperlink try: yield None finally: if new_hyperlink: self.hyperlink = old_hyperlink @contextmanager def _use_format(self, tags: list[str]): if not tags: yield None else: self.format_tags.extend(tags) try: yield None finally: self.format_tags = self.format_tags[: -len(tags)] @contextmanager def _use_inline_group( self, annotated_text_list: AnnotatedTextList, doc: DoclingDocument ): """Create an inline group for annotated texts. Checks if annotated_text_list has more than one item and if so creates an inline group in which the text elements can then be generated. While the context manager is active the inline group is set as the current parent. Args: annotated_text_list (AnnotatedTextList): Annotated text doc (DoclingDocument): Currently used document """ if len(annotated_text_list) > 1: inline_fmt = doc.add_group( label=GroupLabel.INLINE, parent=self.parents[self.level], content_layer=self.content_layer, ) self.parents[self.level + 1] = inline_fmt self.level += 1 try: yield None finally: self.parents[self.level] = None self.level -= 1 else: yield None @contextmanager def _use_details(self, tag: Tag, doc: DoclingDocument): """Create a group with the content of a details tag. While the context manager is active, the hierarchy level is set one level higher as the cuurent parent. Args: tag: The details tag. doc: Currently used document. """ self.parents[self.level + 1] = doc.add_group(
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
true
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/asciidoc_backend.py
docling/backend/asciidoc_backend.py
import logging import re from io import BytesIO from pathlib import Path from typing import Final, Union from docling_core.types.doc import ( DocItemLabel, DoclingDocument, DocumentOrigin, GroupItem, GroupLabel, ImageRef, Size, TableCell, TableData, ) from docling.backend.abstract_backend import DeclarativeDocumentBackend from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) DEFAULT_IMAGE_WIDTH: Final = 128 DEFAULT_IMAGE_HEIGHT: Final = 128 class AsciiDocBackend(DeclarativeDocumentBackend): def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]): super().__init__(in_doc, path_or_stream) self.path_or_stream = path_or_stream try: if isinstance(self.path_or_stream, BytesIO): text_stream = self.path_or_stream.getvalue().decode("utf-8") self.lines = text_stream.split("\n") if isinstance(self.path_or_stream, Path): with open(self.path_or_stream, encoding="utf-8") as f: self.lines = f.readlines() self.valid = True except Exception as e: raise RuntimeError( f"Could not initialize AsciiDoc backend for file with hash {self.document_hash}." ) from e return def is_valid(self) -> bool: return self.valid @classmethod def supports_pagination(cls) -> bool: return False def unload(self): return @classmethod def supported_formats(cls) -> set[InputFormat]: return {InputFormat.ASCIIDOC} def convert(self) -> DoclingDocument: """ Parses the ASCII into a structured document model. """ origin = DocumentOrigin( filename=self.file.name or "file", mimetype="text/asciidoc", binary_hash=self.document_hash, ) doc = DoclingDocument(name=self.file.stem or "file", origin=origin) doc = self._parse(doc) return doc def _parse(self, doc: DoclingDocument): """ Main function that orchestrates the parsing by yielding components: title, section headers, text, lists, and tables. """ in_list = False in_table = False text_data: list[str] = [] table_data: list[str] = [] caption_data: list[str] = [] # parents: dict[int, Union[DocItem, GroupItem, None]] = {} parents: dict[int, Union[GroupItem, None]] = {} # indents: dict[int, Union[DocItem, GroupItem, None]] = {} indents: dict[int, Union[GroupItem, None]] = {} for i in range(10): parents[i] = None indents[i] = None for line in self.lines: # line = line.strip() # Title if self._is_title(line): item = self._parse_title(line) level = item["level"] parents[level] = doc.add_text( text=item["text"], label=DocItemLabel.TITLE ) # Section headers elif self._is_section_header(line): item = self._parse_section_header(line) level = item["level"] parents[level] = doc.add_heading( text=item["text"], level=item["level"], parent=parents[level - 1] ) for k, v in parents.items(): if k > level: parents[k] = None # Lists elif self._is_list_item(line): _log.debug(f"line: {line}") item = self._parse_list_item(line) _log.debug(f"parsed list-item: {item}") level = self._get_current_level(parents) if not in_list: in_list = True parents[level + 1] = doc.add_group( parent=parents[level], name="list", label=GroupLabel.LIST ) indents[level + 1] = item["indent"] elif in_list and item["indent"] > indents[level]: parents[level + 1] = doc.add_group( parent=parents[level], name="list", label=GroupLabel.LIST ) indents[level + 1] = item["indent"] elif in_list and item["indent"] < indents[level]: # print(item["indent"], " => ", indents[level]) while item["indent"] < indents[level]: # print(item["indent"], " => ", indents[level]) parents[level] = None indents[level] = None level -= 1 doc.add_list_item( item["text"], parent=self._get_current_parent(parents) ) elif in_list and not self._is_list_item(line): in_list = False level = self._get_current_level(parents) parents[level] = None # Tables elif line.strip() == "|===" and not in_table: # start of table in_table = True elif self._is_table_line(line): # within a table in_table = True table_data.append(self._parse_table_line(line)) elif in_table and ( (not self._is_table_line(line)) or line.strip() == "|===" ): # end of table caption = None if len(caption_data) > 0: caption = doc.add_text( text=" ".join(caption_data), label=DocItemLabel.CAPTION ) caption_data = [] data = self._populate_table_as_grid(table_data) doc.add_table( data=data, parent=self._get_current_parent(parents), caption=caption ) in_table = False table_data = [] # Picture elif self._is_picture(line): caption = None if len(caption_data) > 0: caption = doc.add_text( text=" ".join(caption_data), label=DocItemLabel.CAPTION ) caption_data = [] item = self._parse_picture(line) size: Size if "width" in item and "height" in item: size = Size(width=int(item["width"]), height=int(item["height"])) else: size = Size(width=DEFAULT_IMAGE_WIDTH, height=DEFAULT_IMAGE_HEIGHT) uri = None if ( "uri" in item and not item["uri"].startswith("http") and item["uri"].startswith("//") ): uri = "file:" + item["uri"] elif ( "uri" in item and not item["uri"].startswith("http") and item["uri"].startswith("/") ): uri = "file:/" + item["uri"] elif "uri" in item and not item["uri"].startswith("http"): uri = "file://" + item["uri"] image = ImageRef(mimetype="image/png", size=size, dpi=70, uri=uri) doc.add_picture(image=image, caption=caption) # Caption elif self._is_caption(line) and len(caption_data) == 0: item = self._parse_caption(line) caption_data.append(item["text"]) elif ( len(line.strip()) > 0 and len(caption_data) > 0 ): # allow multiline captions item = self._parse_text(line) caption_data.append(item["text"]) # Plain text elif len(line.strip()) == 0 and len(text_data) > 0: doc.add_text( text=" ".join(text_data), label=DocItemLabel.PARAGRAPH, parent=self._get_current_parent(parents), ) text_data = [] elif len(line.strip()) > 0: # allow multiline texts item = self._parse_text(line) text_data.append(item["text"]) if len(text_data) > 0: doc.add_text( text=" ".join(text_data), label=DocItemLabel.PARAGRAPH, parent=self._get_current_parent(parents), ) text_data = [] if in_table and len(table_data) > 0: data = self._populate_table_as_grid(table_data) doc.add_table(data=data, parent=self._get_current_parent(parents)) in_table = False table_data = [] return doc @staticmethod def _get_current_level(parents): for k, v in parents.items(): if v is None and k > 0: return k - 1 return 0 @staticmethod def _get_current_parent(parents): for k, v in parents.items(): if v is None and k > 0: return parents[k - 1] return None # ========= Title @staticmethod def _is_title(line): return re.match(r"^= ", line) @staticmethod def _parse_title(line): return {"type": "title", "text": line[2:].strip(), "level": 0} # ========= Section headers @staticmethod def _is_section_header(line): return re.match(r"^==+\s+", line) @staticmethod def _parse_section_header(line): match = re.match(r"^(=+)\s+(.*)", line) marker = match.group(1) # The list marker (e.g., "*", "-", "1.") text = match.group(2) # The actual text of the list item header_level = marker.count("=") # number of '=' represents level return { "type": "header", "level": header_level - 1, "text": text.strip(), } # ========= Lists @staticmethod def _is_list_item(line): return re.match(r"^(\s)*(\*|-|\d+\.|\w+\.) ", line) @staticmethod def _parse_list_item(line): """Extract the item marker (number or bullet symbol) and the text of the item.""" match = re.match(r"^(\s*)(\*|-|\d+\.)\s+(.*)", line) if match: indent = match.group(1) marker = match.group(2) # The list marker (e.g., "*", "-", "1.") text = match.group(3) # The actual text of the list item if marker == "*" or marker == "-": return { "type": "list_item", "marker": marker, "text": text.strip(), "numbered": False, "indent": 0 if indent is None else len(indent), } else: return { "type": "list_item", "marker": marker, "text": text.strip(), "numbered": True, "indent": 0 if indent is None else len(indent), } else: # Fallback if no match return { "type": "list_item", "marker": "-", "text": line, "numbered": False, "indent": 0, } # ========= Tables @staticmethod def _is_table_line(line): return re.match(r"^\|.*\|", line) @staticmethod def _parse_table_line(line): # Split table cells and trim extra spaces return [cell.strip() for cell in line.split("|") if cell.strip()] @staticmethod def _populate_table_as_grid(table_data): num_rows = len(table_data) # Adjust the table data into a grid format num_cols = max(len(row) for row in table_data) data = TableData(num_rows=num_rows, num_cols=num_cols, table_cells=[]) for row_idx, row in enumerate(table_data): # Pad rows with empty strings to match column count # grid.append(row + [''] * (max_cols - len(row))) for col_idx, text in enumerate(row): row_span = 1 col_span = 1 cell = TableCell( text=text, row_span=row_span, col_span=col_span, start_row_offset_idx=row_idx, end_row_offset_idx=row_idx + row_span, start_col_offset_idx=col_idx, end_col_offset_idx=col_idx + col_span, column_header=row_idx == 0, row_header=False, ) data.table_cells.append(cell) return data # ========= Pictures @staticmethod def _is_picture(line): return re.match(r"^image::", line) @staticmethod def _parse_picture(line): """ Parse an image macro, extracting its path and attributes. Syntax: image::path/to/image.png[Alt Text, width=200, height=150, align=center] """ mtch = re.match(r"^image::(.+)\[(.*)\]$", line) if mtch: picture_path = mtch.group(1).strip() attributes = mtch.group(2).split(",") picture_info = {"type": "picture", "uri": picture_path} # Extract optional attributes (alt text, width, height, alignment) if attributes: picture_info["alt"] = attributes[0].strip() if attributes[0] else "" for attr in attributes[1:]: key, value = attr.split("=") picture_info[key.strip()] = value.strip() return picture_info return {"type": "picture", "uri": line} # ========= Captions @staticmethod def _is_caption(line): return re.match(r"^\.(.+)", line) @staticmethod def _parse_caption(line): mtch = re.match(r"^\.(.+)", line) if mtch: text = mtch.group(1) return {"type": "caption", "text": text} return {"type": "caption", "text": ""} # ========= Plain text @staticmethod def _parse_text(line): return {"type": "text", "text": line.strip()}
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/pdf_backend.py
docling/backend/pdf_backend.py
from abc import ABC, abstractmethod from collections.abc import Iterable from io import BytesIO from pathlib import Path from typing import Optional, Set, Union from docling_core.types.doc import BoundingBox, Size from docling_core.types.doc.page import SegmentedPdfPage, TextCell from PIL import Image from docling.backend.abstract_backend import PaginatedDocumentBackend from docling.datamodel.backend_options import PdfBackendOptions from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument class PdfPageBackend(ABC): @abstractmethod def get_text_in_rect(self, bbox: BoundingBox) -> str: pass @abstractmethod def get_segmented_page(self) -> Optional[SegmentedPdfPage]: pass @abstractmethod def get_text_cells(self) -> Iterable[TextCell]: pass @abstractmethod def get_bitmap_rects(self, float: int = 1) -> Iterable[BoundingBox]: pass @abstractmethod def get_page_image( self, scale: float = 1, cropbox: Optional[BoundingBox] = None ) -> Image.Image: pass @abstractmethod def get_size(self) -> Size: pass @abstractmethod def is_valid(self) -> bool: pass @abstractmethod def unload(self): pass class PdfDocumentBackend(PaginatedDocumentBackend): def __init__( self, in_doc: InputDocument, path_or_stream: Union[BytesIO, Path], options: PdfBackendOptions = PdfBackendOptions(), ): super().__init__(in_doc, path_or_stream, options) self.options: PdfBackendOptions if self.input_format not in self.supported_formats(): raise RuntimeError( f"Incompatible file format {self.input_format} was passed to a PdfDocumentBackend. Valid format are {','.join(self.supported_formats())}." ) @abstractmethod def load_page(self, page_no: int) -> PdfPageBackend: pass @abstractmethod def page_count(self) -> int: pass @classmethod def supported_formats(cls) -> Set[InputFormat]: return {InputFormat.PDF} @classmethod def supports_pagination(cls) -> bool: return True
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/__init__.py
docling/backend/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/mets_gbs_backend.py
docling/backend/mets_gbs_backend.py
"""Backend for GBS Google Books schema.""" import logging import tarfile from collections.abc import Iterable from dataclasses import dataclass from enum import Enum from io import BytesIO from pathlib import Path from typing import TYPE_CHECKING, Dict, List, Optional, Set, Tuple, Union from docling_core.types.doc import BoundingBox, CoordOrigin, Size from docling_core.types.doc.page import ( BoundingRectangle, PdfPageBoundaryType, PdfPageGeometry, SegmentedPdfPage, TextCell, ) from lxml import etree from PIL import Image from PIL.Image import Image as PILImage from docling.backend.abstract_backend import PaginatedDocumentBackend from docling.backend.pdf_backend import PdfDocumentBackend, PdfPageBackend from docling.datamodel.base_models import InputFormat if TYPE_CHECKING: from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) def _get_pdf_page_geometry( size: Size, ) -> PdfPageGeometry: boundary_type: PdfPageBoundaryType = PdfPageBoundaryType.CROP_BOX bbox_tuple = (0, 0, size.width, size.height) bbox = BoundingBox.from_tuple(bbox_tuple, CoordOrigin.TOPLEFT) return PdfPageGeometry( angle=0.0, rect=BoundingRectangle.from_bounding_box(bbox), boundary_type=boundary_type, art_bbox=bbox, bleed_bbox=bbox, crop_bbox=bbox, media_bbox=bbox, trim_bbox=bbox, ) class MetsGbsPageBackend(PdfPageBackend): def __init__(self, parsed_page: SegmentedPdfPage, page_im: PILImage): self._im = page_im self._dpage = parsed_page self.valid = parsed_page is not None def is_valid(self) -> bool: return self.valid def get_text_in_rect(self, bbox: BoundingBox) -> str: # Find intersecting cells on the page text_piece = "" page_size = self.get_size() scale = ( 1 # FIX - Replace with param in get_text_in_rect across backends (optional) ) for i, cell in enumerate(self._dpage.textline_cells): cell_bbox = ( cell.rect.to_bounding_box() .to_top_left_origin(page_height=page_size.height) .scaled(scale) ) overlap_frac = cell_bbox.intersection_over_self(bbox) if overlap_frac > 0.5: if len(text_piece) > 0: text_piece += " " text_piece += cell.text return text_piece def get_segmented_page(self) -> Optional[SegmentedPdfPage]: return self._dpage def get_text_cells(self) -> Iterable[TextCell]: return self._dpage.textline_cells def get_bitmap_rects(self, scale: float = 1) -> Iterable[BoundingBox]: AREA_THRESHOLD = 0 # 32 * 32 images = self._dpage.bitmap_resources for img in images: cropbox = img.rect.to_bounding_box().to_top_left_origin( self.get_size().height ) if cropbox.area() > AREA_THRESHOLD: cropbox = cropbox.scaled(scale=scale) yield cropbox def get_page_image( self, scale: float = 1, cropbox: Optional[BoundingBox] = None ) -> Image.Image: page_size = self.get_size() assert ( page_size.width == self._im.size[0] and page_size.height == self._im.size[1] ) if not cropbox: cropbox = BoundingBox( l=0, r=page_size.width, t=0, b=page_size.height, coord_origin=CoordOrigin.TOPLEFT, ) image = self._im.resize( size=(round(page_size.width * scale), round(page_size.height * scale)) ).crop(cropbox.scaled(scale=scale).as_tuple()) return image def get_size(self) -> Size: return Size( width=self._dpage.dimension.width, height=self._dpage.dimension.height ) def unload(self) -> None: if hasattr(self, "_im"): delattr(self, "_im") if hasattr(self, "_dpage"): delattr(self, "_dpage") class _UseType(str, Enum): IMAGE = "image" OCR = "OCR" COORD_OCR = "coordOCR" @dataclass class _FileInfo: file_id: str mimetype: str path: str use: _UseType @dataclass class _PageFiles: image: Optional[_FileInfo] = None ocr: Optional[_FileInfo] = None coordOCR: Optional[_FileInfo] = None def _extract_rect(title_str: str) -> Optional[BoundingRectangle]: """ Extracts bbox from title string like 'bbox 279 177 306 214;x_wconf 97' """ parts = title_str.split(";") for part in parts: part = part.strip() if part.startswith("bbox "): try: coords = part.split()[1:] rect = BoundingRectangle.from_bounding_box( bbox=BoundingBox.from_tuple( tuple(map(int, coords)), origin=CoordOrigin.TOPLEFT ) ) return rect except Exception: return None return None def _extract_confidence(title_str) -> float: """Extracts x_wconf (OCR confidence) value from title string.""" for part in title_str.split(";"): part = part.strip() if part.startswith("x_wconf"): try: return float(part.split()[1]) / 100.0 except Exception: return 1 return 1 class MetsGbsDocumentBackend(PdfDocumentBackend): def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]): super().__init__(in_doc, path_or_stream) self._tar: tarfile.TarFile = ( tarfile.open(name=self.path_or_stream, mode="r:gz") if isinstance(self.path_or_stream, Path) else tarfile.open(fileobj=self.path_or_stream, mode="r:gz") ) self.root_mets: Optional[etree._Element] = None self.page_map: Dict[int, _PageFiles] = {} for member in self._tar.getmembers(): if member.name.endswith(".xml"): file = self._tar.extractfile(member) if file is not None: content = file.read() self.root_mets = self._validate_mets_xml(content) if self.root_mets is not None: break if self.root_mets is None: raise RuntimeError( f"METS GBS backend could not load document {self.document_hash}." ) ns = { "mets": "http://www.loc.gov/METS/", "xlink": "http://www.w3.org/1999/xlink", "xsi": "http://www.w3.org/2001/XMLSchema-instance", "gbs": "http://books.google.com/gbs", "premis": "info:lc/xmlns/premis-v2", "marc": "http://www.loc.gov/MARC21/slim", } file_info_by_id: Dict[str, _FileInfo] = {} for filegrp in self.root_mets.xpath(".//mets:fileGrp", namespaces=ns): use_raw = filegrp.get("USE") try: use = _UseType(use_raw) except ValueError: continue # Ignore unknown USE types for file_elem in filegrp.xpath("./mets:file", namespaces=ns): file_id = file_elem.get("ID") mimetype = file_elem.get("MIMETYPE") flocat_elem = file_elem.find("mets:FLocat", namespaces=ns) href = ( flocat_elem.get("{http://www.w3.org/1999/xlink}href") if flocat_elem is not None else None ) if href is None: continue file_info_by_id[file_id] = _FileInfo( file_id=file_id, mimetype=mimetype, path=href, use=use ) USE_TO_ATTR = { _UseType.IMAGE: "image", _UseType.OCR: "ocr", _UseType.COORD_OCR: "coordOCR", } for div in self.root_mets.xpath('.//mets:div[@TYPE="page"]', namespaces=ns): order_str = div.get("ORDER") if not order_str: continue try: page_no = int(order_str) - 1 # make 0-index pages except ValueError: continue page_files = _PageFiles() for fptr in div.xpath("./mets:fptr", namespaces=ns): file_id = fptr.get("FILEID") file_info = file_info_by_id.get(file_id) if file_info: attr = USE_TO_ATTR.get(file_info.use) if attr: setattr(page_files, attr, file_info) self.page_map[page_no] = page_files def _validate_mets_xml(self, xml_string) -> Optional[etree._Element]: root: etree._Element = etree.fromstring(xml_string) if ( root.tag == "{http://www.loc.gov/METS/}mets" and root.get("PROFILE") == "gbs" ): return root _log.warning(f"The root element is not <mets:mets> with PROFILE='gbs': {root}") return None def _parse_page(self, page_no: int) -> Tuple[SegmentedPdfPage, PILImage]: # TODO: use better fallbacks... image_info = self.page_map[page_no].image assert image_info is not None ocr_info = self.page_map[page_no].coordOCR assert ocr_info is not None image_file = self._tar.extractfile(image_info.path) assert image_file is not None buf = BytesIO(image_file.read()) im: PILImage = Image.open(buf) ocr_file = self._tar.extractfile(ocr_info.path) assert ocr_file is not None ocr_content = ocr_file.read() parser = etree.HTMLParser() ocr_root: etree._Element = etree.fromstring(ocr_content, parser=parser) line_cells: List[TextCell] = [] word_cells: List[TextCell] = [] page_div = ocr_root.xpath("//div[@class='ocr_page']") size = Size(width=im.size[0], height=im.size[1]) if page_div: title = page_div[0].attrib.get("title", "") rect = _extract_rect(title) if rect: size = Size(width=rect.width, height=rect.height) else: _log.error(f"Could not find ocr_page for page {page_no}") im = im.resize(size=(round(size.width), round(size.height))) im = im.convert("RGB") # Extract all ocrx_word spans for ix, word in enumerate(ocr_root.xpath("//span[@class='ocrx_word']")): text = "".join(word.itertext()).strip() title = word.attrib.get("title", "") rect = _extract_rect(title) conf = _extract_confidence(title) if rect: word_cells.append( TextCell( index=ix, text=text, orig=text, rect=rect, from_ocr=True, confidence=conf, ) ) # Extract all ocr_line spans # line: etree._Element for ix, line in enumerate(ocr_root.xpath("//span[@class='ocr_line']")): text = "".join(line.itertext()).strip() title = line.attrib.get("title", "") rect = _extract_rect(title) conf = _extract_confidence(title) if rect: line_cells.append( TextCell( index=ix, text=text, orig=text, rect=rect, from_ocr=True, confidence=conf, ) ) page = SegmentedPdfPage( dimension=_get_pdf_page_geometry(size), textline_cells=line_cells, char_cells=[], word_cells=word_cells, has_textlines=True, has_words=True, has_chars=False, ) return page, im def page_count(self) -> int: return len(self.page_map) def load_page(self, page_no: int) -> MetsGbsPageBackend: # TODO: is this thread-safe? page, im = self._parse_page(page_no) return MetsGbsPageBackend(parsed_page=page, page_im=im) def is_valid(self) -> bool: return self.root_mets is not None and self.page_count() > 0 @classmethod def supported_formats(cls) -> Set[InputFormat]: return {InputFormat.METS_GBS} @classmethod def supports_pagination(cls) -> bool: return True def unload(self) -> None: super().unload() self._tar.close()
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/csv_backend.py
docling/backend/csv_backend.py
import csv import logging import warnings from io import BytesIO, StringIO from pathlib import Path from typing import Set, Union from docling_core.types.doc import DoclingDocument, DocumentOrigin, TableCell, TableData from docling.backend.abstract_backend import DeclarativeDocumentBackend from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) class CsvDocumentBackend(DeclarativeDocumentBackend): content: StringIO def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]): super().__init__(in_doc, path_or_stream) # Load content try: if isinstance(self.path_or_stream, BytesIO): self.content = StringIO(self.path_or_stream.getvalue().decode("utf-8")) elif isinstance(self.path_or_stream, Path): self.content = StringIO(self.path_or_stream.read_text("utf-8")) self.valid = True except Exception as e: raise RuntimeError( f"CsvDocumentBackend could not load document with hash {self.document_hash}" ) from e return def is_valid(self) -> bool: return self.valid @classmethod def supports_pagination(cls) -> bool: return False def unload(self): if isinstance(self.path_or_stream, BytesIO): self.path_or_stream.close() self.path_or_stream = None @classmethod def supported_formats(cls) -> Set[InputFormat]: return {InputFormat.CSV} def convert(self) -> DoclingDocument: """ Parses the CSV data into a structured document model. """ # Detect CSV dialect head = self.content.readline() dialect = csv.Sniffer().sniff(head, ",;\t|:") _log.info(f'Parsing CSV with delimiter: "{dialect.delimiter}"') if dialect.delimiter not in {",", ";", "\t", "|", ":"}: raise RuntimeError( f"Cannot convert csv with unknown delimiter {dialect.delimiter}." ) # Parce CSV self.content.seek(0) result = csv.reader(self.content, dialect=dialect, strict=True) self.csv_data = list(result) _log.info(f"Detected {len(self.csv_data)} lines") # Ensure uniform column length expected_length = len(self.csv_data[0]) is_uniform = all(len(row) == expected_length for row in self.csv_data) if not is_uniform: warnings.warn( f"Inconsistent column lengths detected in CSV data. " f"Expected {expected_length} columns, but found rows with varying lengths. " f"Ensure all rows have the same number of columns." ) # Parse the CSV into a structured document model origin = DocumentOrigin( filename=self.file.name or "file.csv", mimetype="text/csv", binary_hash=self.document_hash, ) doc = DoclingDocument(name=self.file.stem or "file.csv", origin=origin) if self.is_valid(): # Convert CSV data to table if self.csv_data: num_rows = len(self.csv_data) num_cols = max(len(row) for row in self.csv_data) table_data = TableData( num_rows=num_rows, num_cols=num_cols, table_cells=[], ) # Convert each cell to TableCell for row_idx, row in enumerate(self.csv_data): for col_idx, cell_value in enumerate(row): cell = TableCell( text=str(cell_value), row_span=1, # CSV doesn't support merged cells col_span=1, start_row_offset_idx=row_idx, end_row_offset_idx=row_idx + 1, start_col_offset_idx=col_idx, end_col_offset_idx=col_idx + 1, column_header=row_idx == 0, # First row as header row_header=False, ) table_data.table_cells.append(cell) doc.add_table(data=table_data) else: raise RuntimeError( f"Cannot convert doc with {self.document_hash} because the backend failed to init." ) return doc
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/webvtt_backend.py
docling/backend/webvtt_backend.py
import logging import re from io import BytesIO from pathlib import Path from typing import Annotated, ClassVar, Literal, Optional, Union, cast from docling_core.types.doc import ( ContentLayer, DocItemLabel, DoclingDocument, DocumentOrigin, Formatting, GroupLabel, NodeItem, ) from pydantic import BaseModel, ConfigDict, Field, field_validator, model_validator from pydantic.types import StringConstraints from typing_extensions import Self, override from docling.backend.abstract_backend import DeclarativeDocumentBackend from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) class _WebVTTTimestamp(BaseModel): """Model representing a WebVTT timestamp. A WebVTT timestamp is always interpreted relative to the current playback position of the media data that the WebVTT file is to be synchronized with. """ model_config = ConfigDict(regex_engine="python-re") raw: Annotated[ str, Field( description="A representation of the WebVTT Timestamp as a single string" ), ] _pattern: ClassVar[re.Pattern] = re.compile( r"^(?:(\d{2,}):)?([0-5]\d):([0-5]\d)\.(\d{3})$" ) _hours: int _minutes: int _seconds: int _millis: int @model_validator(mode="after") def validate_raw(self) -> Self: m = self._pattern.match(self.raw) if not m: raise ValueError(f"Invalid WebVTT timestamp format: {self.raw}") self._hours = int(m.group(1)) if m.group(1) else 0 self._minutes = int(m.group(2)) self._seconds = int(m.group(3)) self._millis = int(m.group(4)) if self._minutes < 0 or self._minutes > 59: raise ValueError("Minutes must be between 0 and 59") if self._seconds < 0 or self._seconds > 59: raise ValueError("Seconds must be between 0 and 59") return self @property def seconds(self) -> float: """A representation of the WebVTT Timestamp in seconds""" return ( self._hours * 3600 + self._minutes * 60 + self._seconds + self._millis / 1000.0 ) @override def __str__(self) -> str: return self.raw _WebVTTCueIdentifier = Annotated[ str, StringConstraints(strict=True, pattern=r"^(?!.*-->)[^\n\r]+$") ] class _WebVTTCueTimings(BaseModel): """Model representating WebVTT cue timings.""" start: Annotated[ _WebVTTTimestamp, Field(description="Start time offset of the cue") ] end: Annotated[_WebVTTTimestamp, Field(description="End time offset of the cue")] @model_validator(mode="after") def check_order(self) -> Self: if self.start and self.end: if self.end.seconds <= self.start.seconds: raise ValueError("End timestamp must be greater than start timestamp") return self @override def __str__(self): return f"{self.start} --> {self.end}" class _WebVTTCueTextSpan(BaseModel): """Model representing a WebVTT cue text span.""" text: str span_type: Literal["text"] = "text" @field_validator("text", mode="after") @classmethod def validate_text(cls, value: str) -> str: if any(ch in value for ch in {"\n", "\r", "&", "<"}): raise ValueError("Cue text span contains invalid characters") if len(value) == 0: raise ValueError("Cue text span cannot be empty") return value @override def __str__(self): return self.text class _WebVTTCueVoiceSpan(BaseModel): """Model representing a WebVTT cue voice span.""" annotation: Annotated[ str, Field( description=( "Cue span start tag annotation text representing the name of thevoice" ) ), ] classes: Annotated[ list[str], Field(description="List of classes representing the cue span's significance"), ] = [] components: Annotated[ list["_WebVTTCueComponent"], Field(description="The components representing the cue internal text"), ] = [] span_type: Literal["v"] = "v" @field_validator("annotation", mode="after") @classmethod def validate_annotation(cls, value: str) -> str: if any(ch in value for ch in {"\n", "\r", "&", ">"}): raise ValueError( "Cue span start tag annotation contains invalid characters" ) if not value: raise ValueError("Cue text span cannot be empty") return value @field_validator("classes", mode="after") @classmethod def validate_classes(cls, value: list[str]) -> list[str]: for item in value: if any(ch in item for ch in {"\t", "\n", "\r", " ", "&", "<", ">", "."}): raise ValueError( "A cue span start tag class contains invalid characters" ) if not item: raise ValueError("Cue span start tag classes cannot be empty") return value @override def __str__(self): tag = f"v.{'.'.join(self.classes)}" if self.classes else "v" inner = "".join(str(span) for span in self.components) return f"<{tag} {self.annotation}>{inner}</v>" class _WebVTTCueClassSpan(BaseModel): span_type: Literal["c"] = "c" components: list["_WebVTTCueComponent"] @override def __str__(self): inner = "".join(str(span) for span in self.components) return f"<c>{inner}</c>" class _WebVTTCueItalicSpan(BaseModel): span_type: Literal["i"] = "i" components: list["_WebVTTCueComponent"] @override def __str__(self): inner = "".join(str(span) for span in self.components) return f"<i>{inner}</i>" class _WebVTTCueBoldSpan(BaseModel): span_type: Literal["b"] = "b" components: list["_WebVTTCueComponent"] @override def __str__(self): inner = "".join(str(span) for span in self.components) return f"<b>{inner}</b>" class _WebVTTCueUnderlineSpan(BaseModel): span_type: Literal["u"] = "u" components: list["_WebVTTCueComponent"] @override def __str__(self): inner = "".join(str(span) for span in self.components) return f"<u>{inner}</u>" _WebVTTCueComponent = Annotated[ Union[ _WebVTTCueTextSpan, _WebVTTCueClassSpan, _WebVTTCueItalicSpan, _WebVTTCueBoldSpan, _WebVTTCueUnderlineSpan, _WebVTTCueVoiceSpan, ], Field(discriminator="span_type", description="The WebVTT cue component"), ] class _WebVTTCueBlock(BaseModel): """Model representing a WebVTT cue block. The optional WebVTT cue settings list is not supported. The cue payload is limited to the following spans: text, class, italic, bold, underline, and voice. """ model_config = ConfigDict(regex_engine="python-re") identifier: Optional[_WebVTTCueIdentifier] = Field( None, description="The WebVTT cue identifier" ) timings: Annotated[_WebVTTCueTimings, Field(description="The WebVTT cue timings")] payload: Annotated[list[_WebVTTCueComponent], Field(description="The cue payload")] _pattern_block: ClassVar[re.Pattern] = re.compile( r"<(/?)(i|b|c|u|v(?:\.[^\t\n\r &<>.]+)*)(?:\s+([^>]*))?>" ) _pattern_voice_tag: ClassVar[re.Pattern] = re.compile( r"^<v(?P<class>\.[^\t\n\r &<>]+)?" # zero or more classes r"[ \t]+(?P<annotation>[^\n\r&>]+)>" # required space and annotation ) @field_validator("payload", mode="after") @classmethod def validate_payload(cls, payload): for voice in payload: if "-->" in str(voice): raise ValueError("Cue payload must not contain '-->'") return payload @classmethod def parse(cls, raw: str) -> "_WebVTTCueBlock": lines = raw.strip().splitlines() if not lines: raise ValueError("Cue block must have at least one line") identifier: Optional[_WebVTTCueIdentifier] = None timing_line = lines[0] if "-->" not in timing_line and len(lines) > 1: identifier = timing_line timing_line = lines[1] cue_lines = lines[2:] else: cue_lines = lines[1:] if "-->" not in timing_line: raise ValueError("Cue block must contain WebVTT cue timings") start, end = [t.strip() for t in timing_line.split("-->")] end = re.split(" |\t", end)[0] # ignore the cue settings list timings: _WebVTTCueTimings = _WebVTTCueTimings( start=_WebVTTTimestamp(raw=start), end=_WebVTTTimestamp(raw=end) ) cue_text = " ".join(cue_lines).strip() if cue_text.startswith("<v") and "</v>" not in cue_text: # adding close tag for cue voice spans without end tag cue_text += "</v>" stack: list[list[_WebVTTCueComponent]] = [[]] tag_stack: list[Union[str, tuple]] = [] pos = 0 matches = list(cls._pattern_block.finditer(cue_text)) i = 0 while i < len(matches): match = matches[i] if match.start() > pos: stack[-1].append(_WebVTTCueTextSpan(text=cue_text[pos : match.start()])) tag = match.group(0) if tag.startswith(("<i>", "<b>", "<u>", "<c>")): tag_type = tag[1:2] tag_stack.append(tag_type) stack.append([]) elif tag == "</i>": children = stack.pop() stack[-1].append(_WebVTTCueItalicSpan(components=children)) tag_stack.pop() elif tag == "</b>": children = stack.pop() stack[-1].append(_WebVTTCueBoldSpan(components=children)) tag_stack.pop() elif tag == "</u>": children = stack.pop() stack[-1].append(_WebVTTCueUnderlineSpan(components=children)) tag_stack.pop() elif tag == "</c>": children = stack.pop() stack[-1].append(_WebVTTCueClassSpan(components=children)) tag_stack.pop() elif tag.startswith("<v"): tag_stack.append(("v", tag)) stack.append([]) elif tag.startswith("</v"): children = stack.pop() if stack else [] if ( tag_stack and isinstance(tag_stack[-1], tuple) and tag_stack[-1][0] == "v" ): _, voice = cast(tuple, tag_stack.pop()) voice_match = cls._pattern_voice_tag.match(voice) if voice_match: class_string = voice_match.group("class") annotation = voice_match.group("annotation") if annotation: classes: list[str] = [] if class_string: classes = [c for c in class_string.split(".") if c] stack[-1].append( _WebVTTCueVoiceSpan( annotation=annotation.strip(), classes=classes, components=children, ) ) pos = match.end() i += 1 if pos < len(cue_text): stack[-1].append(_WebVTTCueTextSpan(text=cue_text[pos:])) return cls( identifier=identifier, timings=timings, payload=stack[0], ) def __str__(self): parts = [] if self.identifier: parts.append(f"{self.identifier}\n") timings_line = str(self.timings) parts.append(timings_line + "\n") for idx, span in enumerate(self.payload): if idx == 0 and len(self.payload) == 1 and span.span_type == "v": # the end tag may be omitted for brevity parts.append(str(span).removesuffix("</v>")) else: parts.append(str(span)) return "".join(parts) class _WebVTTFile(BaseModel): """A model representing a WebVTT file.""" cue_blocks: list[_WebVTTCueBlock] @staticmethod def verify_signature(content: str) -> bool: if not content: return False elif len(content) == 6: return content == "WEBVTT" elif len(content) > 6 and content.startswith("WEBVTT"): return content[6] in (" ", "\t", "\n") else: return False @classmethod def parse(cls, raw: str) -> "_WebVTTFile": # Normalize newlines to LF raw = raw.replace("\r\n", "\n").replace("\r", "\n") # Check WebVTT signature if not cls.verify_signature(raw): raise ValueError("Invalid WebVTT file signature") # Strip "WEBVTT" header line lines = raw.split("\n", 1) body = lines[1] if len(lines) > 1 else "" # Remove NOTE/STYLE/REGION blocks body = re.sub(r"^(NOTE[^\n]*\n(?:.+\n)*?)\n", "", body, flags=re.MULTILINE) body = re.sub(r"^(STYLE|REGION)(?:.+\n)*?\n", "", body, flags=re.MULTILINE) # Split into cue blocks raw_blocks = re.split(r"\n\s*\n", body.strip()) cues: list[_WebVTTCueBlock] = [] for block in raw_blocks: try: cues.append(_WebVTTCueBlock.parse(block)) except ValueError as e: _log.warning(f"Failed to parse cue block:\n{block}\n{e}") return cls(cue_blocks=cues) def __iter__(self): return iter(self.cue_blocks) def __getitem__(self, idx): return self.cue_blocks[idx] def __len__(self): return len(self.cue_blocks) class WebVTTDocumentBackend(DeclarativeDocumentBackend): """Declarative backend for WebVTT (.vtt) files. This parser reads the content of a WebVTT file and converts it to a DoclingDocument, following the W3C specs on https://www.w3.org/TR/webvtt1 Each cue becomes a TextItem and the items are appended to the document body by the cue's start time. """ @override def __init__(self, in_doc: InputDocument, path_or_stream: Union[BytesIO, Path]): super().__init__(in_doc, path_or_stream) self.content: str = "" try: if isinstance(self.path_or_stream, BytesIO): self.content = self.path_or_stream.getvalue().decode("utf-8") if isinstance(self.path_or_stream, Path): with open(self.path_or_stream, encoding="utf-8") as f: self.content = f.read() except Exception as e: raise RuntimeError( "Could not initialize the WebVTT backend for file with hash " f"{self.document_hash}." ) from e @override def is_valid(self) -> bool: return _WebVTTFile.verify_signature(self.content) @classmethod @override def supports_pagination(cls) -> bool: return False @override def unload(self): if isinstance(self.path_or_stream, BytesIO): self.path_or_stream.close() self.path_or_stream = None @classmethod @override def supported_formats(cls) -> set[InputFormat]: return {InputFormat.VTT} @staticmethod def _add_text_from_component( doc: DoclingDocument, item: _WebVTTCueComponent, parent: Optional[NodeItem] ) -> None: """Adds a TextItem to a document by extracting text from a cue span component. TODO: address nesting """ formatting = Formatting() text = "" if isinstance(item, _WebVTTCueItalicSpan): formatting.italic = True elif isinstance(item, _WebVTTCueBoldSpan): formatting.bold = True elif isinstance(item, _WebVTTCueUnderlineSpan): formatting.underline = True if isinstance(item, _WebVTTCueTextSpan): text = item.text else: # TODO: address nesting text = "".join( [t.text for t in item.components if isinstance(t, _WebVTTCueTextSpan)] ) if text := text.strip(): doc.add_text( label=DocItemLabel.TEXT, text=text, parent=parent, content_layer=ContentLayer.BODY, formatting=formatting, ) @override def convert(self) -> DoclingDocument: _log.debug("Starting WebVTT conversion...") if not self.is_valid(): raise RuntimeError("Invalid WebVTT document.") origin = DocumentOrigin( filename=self.file.name or "file", mimetype="text/vtt", binary_hash=self.document_hash, ) doc = DoclingDocument(name=self.file.stem or "file", origin=origin) vtt: _WebVTTFile = _WebVTTFile.parse(self.content) for block in vtt.cue_blocks: block_group = doc.add_group( label=GroupLabel.SECTION, name="WebVTT cue block", parent=None, content_layer=ContentLayer.BODY, ) if block.identifier: doc.add_text( label=DocItemLabel.TEXT, text=str(block.identifier), parent=block_group, content_layer=ContentLayer.BODY, ) doc.add_text( label=DocItemLabel.TEXT, text=str(block.timings), parent=block_group, content_layer=ContentLayer.BODY, ) for cue_span in block.payload: if isinstance(cue_span, _WebVTTCueVoiceSpan): voice_group = doc.add_group( label=GroupLabel.INLINE, name="WebVTT cue voice span", parent=block_group, content_layer=ContentLayer.BODY, ) voice = cue_span.annotation if classes := cue_span.classes: voice += f" ({', '.join(classes)})" voice += ": " doc.add_text( label=DocItemLabel.TEXT, text=voice, parent=voice_group, content_layer=ContentLayer.BODY, ) for item in cue_span.components: WebVTTDocumentBackend._add_text_from_component( doc, item, voice_group ) else: WebVTTDocumentBackend._add_text_from_component( doc, cue_span, block_group ) return doc
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/pypdfium2_backend.py
docling/backend/pypdfium2_backend.py
import logging import random from collections.abc import Iterable from io import BytesIO from pathlib import Path from typing import TYPE_CHECKING, List, Optional, Union import pypdfium2 as pdfium import pypdfium2.raw as pdfium_c from docling_core.types.doc import BoundingBox, CoordOrigin, Size from docling_core.types.doc.page import ( BoundingRectangle, PdfPageBoundaryType, PdfPageGeometry, SegmentedPdfPage, TextCell, ) from PIL import Image, ImageDraw from pypdfium2 import PdfTextPage from pypdfium2._helpers.misc import PdfiumError from docling.backend.pdf_backend import PdfDocumentBackend, PdfPageBackend from docling.datamodel.backend_options import PdfBackendOptions from docling.utils.locks import pypdfium2_lock def get_pdf_page_geometry( ppage: pdfium.PdfPage, angle: float = 0.0, boundary_type: PdfPageBoundaryType = PdfPageBoundaryType.CROP_BOX, ) -> PdfPageGeometry: """ Create PdfPageGeometry from a pypdfium2 PdfPage object. Args: ppage: pypdfium2 PdfPage object angle: Page rotation angle in degrees (default: 0.0) boundary_type: The boundary type for the page (default: CROP_BOX) Returns: PdfPageGeometry with all the different bounding boxes properly set """ with pypdfium2_lock: # Get the main bounding box (intersection of crop_box and media_box) bbox_tuple = ppage.get_bbox() bbox = BoundingBox.from_tuple(bbox_tuple, CoordOrigin.BOTTOMLEFT) # Get all the different page boxes from pypdfium2 media_box_tuple = ppage.get_mediabox() crop_box_tuple = ppage.get_cropbox() art_box_tuple = ppage.get_artbox() bleed_box_tuple = ppage.get_bleedbox() trim_box_tuple = ppage.get_trimbox() # Convert to BoundingBox objects using existing from_tuple method # pypdfium2 returns (x0, y0, x1, y1) in PDF coordinate system (bottom-left origin) # Use bbox as fallback when specific box types are not defined media_bbox = ( BoundingBox.from_tuple(media_box_tuple, CoordOrigin.BOTTOMLEFT) if media_box_tuple else bbox ) crop_bbox = ( BoundingBox.from_tuple(crop_box_tuple, CoordOrigin.BOTTOMLEFT) if crop_box_tuple else bbox ) art_bbox = ( BoundingBox.from_tuple(art_box_tuple, CoordOrigin.BOTTOMLEFT) if art_box_tuple else bbox ) bleed_bbox = ( BoundingBox.from_tuple(bleed_box_tuple, CoordOrigin.BOTTOMLEFT) if bleed_box_tuple else bbox ) trim_bbox = ( BoundingBox.from_tuple(trim_box_tuple, CoordOrigin.BOTTOMLEFT) if trim_box_tuple else bbox ) return PdfPageGeometry( angle=angle, rect=BoundingRectangle.from_bounding_box(bbox), boundary_type=boundary_type, art_bbox=art_bbox, bleed_bbox=bleed_bbox, crop_bbox=crop_bbox, media_bbox=media_bbox, trim_bbox=trim_bbox, ) if TYPE_CHECKING: from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) class PyPdfiumPageBackend(PdfPageBackend): def __init__( self, pdfium_doc: pdfium.PdfDocument, document_hash: str, page_no: int ): # Note: lock applied by the caller self.valid = True # No better way to tell from pypdfium. try: self._ppage: pdfium.PdfPage = pdfium_doc[page_no] except PdfiumError: _log.info( f"An exception occurred when loading page {page_no} of document {document_hash}.", exc_info=True, ) self.valid = False self.text_page: Optional[PdfTextPage] = None def is_valid(self) -> bool: return self.valid def _compute_text_cells(self) -> List[TextCell]: """Compute text cells from pypdfium.""" with pypdfium2_lock: if not self.text_page: self.text_page = self._ppage.get_textpage() cells = [] cell_counter = 0 page_size = self.get_size() with pypdfium2_lock: for i in range(self.text_page.count_rects()): rect = self.text_page.get_rect(i) text_piece = self.text_page.get_text_bounded(*rect) x0, y0, x1, y1 = rect cells.append( TextCell( index=cell_counter, text=text_piece, orig=text_piece, from_ocr=False, rect=BoundingRectangle.from_bounding_box( BoundingBox( l=x0, b=y0, r=x1, t=y1, coord_origin=CoordOrigin.BOTTOMLEFT, ) ).to_top_left_origin(page_size.height), ) ) cell_counter += 1 # PyPdfium2 produces very fragmented cells, with sub-word level boundaries, in many PDFs. # The cell merging code below is to clean this up. def merge_horizontal_cells( cells: List[TextCell], horizontal_threshold_factor: float = 1.0, vertical_threshold_factor: float = 0.5, ) -> List[TextCell]: if not cells: return [] def group_rows(cells: List[TextCell]) -> List[List[TextCell]]: rows = [] current_row = [cells[0]] row_top = cells[0].rect.to_bounding_box().t row_bottom = cells[0].rect.to_bounding_box().b row_height = cells[0].rect.to_bounding_box().height for cell in cells[1:]: vertical_threshold = row_height * vertical_threshold_factor if ( abs(cell.rect.to_bounding_box().t - row_top) <= vertical_threshold and abs(cell.rect.to_bounding_box().b - row_bottom) <= vertical_threshold ): current_row.append(cell) row_top = min(row_top, cell.rect.to_bounding_box().t) row_bottom = max(row_bottom, cell.rect.to_bounding_box().b) row_height = row_bottom - row_top else: rows.append(current_row) current_row = [cell] row_top = cell.rect.to_bounding_box().t row_bottom = cell.rect.to_bounding_box().b row_height = cell.rect.to_bounding_box().height if current_row: rows.append(current_row) return rows def merge_row(row: List[TextCell]) -> List[TextCell]: merged = [] current_group = [row[0]] for cell in row[1:]: prev_cell = current_group[-1] avg_height = ( prev_cell.rect.height + cell.rect.to_bounding_box().height ) / 2 if ( cell.rect.to_bounding_box().l - prev_cell.rect.to_bounding_box().r <= avg_height * horizontal_threshold_factor ): current_group.append(cell) else: merged.append(merge_group(current_group)) current_group = [cell] if current_group: merged.append(merge_group(current_group)) return merged def merge_group(group: List[TextCell]) -> TextCell: if len(group) == 1: return group[0] merged_bbox = BoundingBox( l=min(cell.rect.to_bounding_box().l for cell in group), t=min(cell.rect.to_bounding_box().t for cell in group), r=max(cell.rect.to_bounding_box().r for cell in group), b=max(cell.rect.to_bounding_box().b for cell in group), ) assert self.text_page is not None bbox = merged_bbox.to_bottom_left_origin(page_size.height) with pypdfium2_lock: merged_text = self.text_page.get_text_bounded(*bbox.as_tuple()) return TextCell( index=group[0].index, text=merged_text, orig=merged_text, rect=BoundingRectangle.from_bounding_box(merged_bbox), from_ocr=False, ) rows = group_rows(cells) merged_cells = [cell for row in rows for cell in merge_row(row)] for i, cell in enumerate(merged_cells, 1): cell.index = i return merged_cells return merge_horizontal_cells(cells) def get_bitmap_rects(self, scale: float = 1) -> Iterable[BoundingBox]: AREA_THRESHOLD = 0 # 32 * 32 page_size = self.get_size() with pypdfium2_lock: rotation = self._ppage.get_rotation() for obj in self._ppage.get_objects(filter=[pdfium_c.FPDF_PAGEOBJ_IMAGE]): pos = obj.get_pos() if rotation == 90: pos = ( pos[1], page_size.height - pos[2], pos[3], page_size.height - pos[0], ) elif rotation == 180: pos = ( page_size.width - pos[2], page_size.height - pos[3], page_size.width - pos[0], page_size.height - pos[1], ) elif rotation == 270: pos = ( page_size.width - pos[3], pos[0], page_size.width - pos[1], pos[2], ) cropbox = BoundingBox.from_tuple( pos, origin=CoordOrigin.BOTTOMLEFT ).to_top_left_origin(page_height=page_size.height) if cropbox.area() > AREA_THRESHOLD: cropbox = cropbox.scaled(scale=scale) yield cropbox def get_text_in_rect(self, bbox: BoundingBox) -> str: with pypdfium2_lock: if not self.text_page: self.text_page = self._ppage.get_textpage() if bbox.coord_origin != CoordOrigin.BOTTOMLEFT: bbox = bbox.to_bottom_left_origin(self.get_size().height) with pypdfium2_lock: text_piece = self.text_page.get_text_bounded(*bbox.as_tuple()) return text_piece def get_segmented_page(self) -> Optional[SegmentedPdfPage]: if not self.valid: return None text_cells = self._compute_text_cells() # Get the PDF page geometry from pypdfium2 dimension = get_pdf_page_geometry(self._ppage) # Create SegmentedPdfPage return SegmentedPdfPage( dimension=dimension, textline_cells=text_cells, char_cells=[], word_cells=[], has_textlines=len(text_cells) > 0, has_words=False, has_chars=False, ) def get_text_cells(self) -> Iterable[TextCell]: return self._compute_text_cells() def get_page_image( self, scale: float = 1, cropbox: Optional[BoundingBox] = None ) -> Image.Image: page_size = self.get_size() if not cropbox: cropbox = BoundingBox( l=0, r=page_size.width, t=0, b=page_size.height, coord_origin=CoordOrigin.TOPLEFT, ) padbox = BoundingBox( l=0, r=0, t=0, b=0, coord_origin=CoordOrigin.BOTTOMLEFT ) else: padbox = cropbox.to_bottom_left_origin(page_size.height).model_copy() padbox.r = page_size.width - padbox.r padbox.t = page_size.height - padbox.t with pypdfium2_lock: image = ( self._ppage.render( scale=scale * 1.5, rotation=0, # no additional rotation crop=padbox.as_tuple(), ) .to_pil() .resize( size=(round(cropbox.width * scale), round(cropbox.height * scale)) ) ) # We resize the image from 1.5x the given scale to make it sharper. return image def get_size(self) -> Size: with pypdfium2_lock: return Size(width=self._ppage.get_width(), height=self._ppage.get_height()) def unload(self): self._ppage = None self.text_page = None class PyPdfiumDocumentBackend(PdfDocumentBackend): def __init__( self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path], options: PdfBackendOptions = PdfBackendOptions(), ): super().__init__(in_doc, path_or_stream, options) password = ( self.options.password.get_secret_value() if self.options.password else None ) try: with pypdfium2_lock: self._pdoc = pdfium.PdfDocument(self.path_or_stream, password=password) except PdfiumError as e: raise RuntimeError( f"pypdfium could not load document with hash {self.document_hash}" ) from e def page_count(self) -> int: with pypdfium2_lock: return len(self._pdoc) def load_page(self, page_no: int) -> PyPdfiumPageBackend: with pypdfium2_lock: return PyPdfiumPageBackend(self._pdoc, self.document_hash, page_no) def is_valid(self) -> bool: return self.page_count() > 0 def unload(self): super().unload() with pypdfium2_lock: self._pdoc.close() self._pdoc = None
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/mspowerpoint_backend.py
docling/backend/mspowerpoint_backend.py
import logging from io import BytesIO from pathlib import Path from typing import Union from docling_core.types.doc import ( BoundingBox, CoordOrigin, DocItemLabel, DoclingDocument, DocumentOrigin, GroupLabel, ImageRef, ProvenanceItem, Size, TableCell, TableData, ) from docling_core.types.doc.document import ContentLayer from PIL import Image, UnidentifiedImageError from pptx import Presentation from pptx.enum.shapes import MSO_SHAPE_TYPE, PP_PLACEHOLDER from pptx.oxml.text import CT_TextLineBreak from docling.backend.abstract_backend import ( DeclarativeDocumentBackend, PaginatedDocumentBackend, ) from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) class MsPowerpointDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentBackend): def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]): super().__init__(in_doc, path_or_stream) self.namespaces = { "a": "http://schemas.openxmlformats.org/drawingml/2006/main", "c": "http://schemas.openxmlformats.org/drawingml/2006/chart", "p": "http://schemas.openxmlformats.org/presentationml/2006/main", } # Powerpoint file: self.path_or_stream = path_or_stream self.pptx_obj = None self.valid = False try: if isinstance(self.path_or_stream, BytesIO): self.pptx_obj = Presentation(self.path_or_stream) elif isinstance(self.path_or_stream, Path): self.pptx_obj = Presentation(str(self.path_or_stream)) self.valid = True except Exception as e: raise RuntimeError( f"MsPowerpointDocumentBackend could not load document with hash {self.document_hash}" ) from e return def page_count(self) -> int: if self.is_valid(): assert self.pptx_obj is not None return len(self.pptx_obj.slides) else: return 0 def is_valid(self) -> bool: return self.valid @classmethod def supports_pagination(cls) -> bool: return True # True? if so, how to handle pages... def unload(self): if isinstance(self.path_or_stream, BytesIO): self.path_or_stream.close() self.path_or_stream = None @classmethod def supported_formats(cls) -> set[InputFormat]: return {InputFormat.PPTX} def convert(self) -> DoclingDocument: # Parses the PPTX into a structured document model. # origin = DocumentOrigin(filename=self.path_or_stream.name, mimetype=next(iter(FormatToMimeType.get(InputFormat.PPTX))), binary_hash=self.document_hash) origin = DocumentOrigin( filename=self.file.name or "file", mimetype="application/vnd.ms-powerpoint", binary_hash=self.document_hash, ) doc = DoclingDocument( name=self.file.stem or "file", origin=origin ) # must add origin information doc = self.walk_linear(self.pptx_obj, doc) return doc def generate_prov( self, shape, slide_ind, text="", slide_size=Size(width=1, height=1) ): if shape.left: left = shape.left top = shape.top width = shape.width height = shape.height else: left = 0 top = 0 width = slide_size.width height = slide_size.height shape_bbox = [left, top, left + width, top + height] shape_bbox = BoundingBox.from_tuple(shape_bbox, origin=CoordOrigin.BOTTOMLEFT) prov = ProvenanceItem( page_no=slide_ind + 1, charspan=[0, len(text)], bbox=shape_bbox ) return prov def handle_text_elements( self, shape, parent_slide, slide_ind, doc: DoclingDocument, slide_size ): is_list_group_created = False enum_list_item_value = 0 new_list = None doc_label = DocItemLabel.LIST_ITEM prov = self.generate_prov(shape, slide_ind, shape.text.strip(), slide_size) def is_list_item(paragraph): """Check if the paragraph is a list item.""" p = paragraph._element if ( p.find(".//a:buChar", namespaces={"a": self.namespaces["a"]}) is not None ): return (True, "Bullet") elif ( p.find(".//a:buAutoNum", namespaces={"a": self.namespaces["a"]}) is not None ): return (True, "Numbered") elif paragraph.level > 0: # Most likely a sub-list return (True, "None") else: return (False, "None") # Iterate through paragraphs to build up text for paragraph in shape.text_frame.paragraphs: is_a_list, bullet_type = is_list_item(paragraph) p = paragraph._element # Convert line breaks to spaces and accumulate text p_text = "" for e in p.content_children: if isinstance(e, CT_TextLineBreak): p_text += " " else: p_text += e.text if is_a_list: enum_marker = "" enumerated = bullet_type == "Numbered" if not is_list_group_created: new_list = doc.add_list_group( name="list", parent=parent_slide, ) is_list_group_created = True enum_list_item_value = 0 if enumerated: enum_list_item_value += 1 enum_marker = str(enum_list_item_value) + "." doc.add_list_item( marker=enum_marker, enumerated=enumerated, parent=new_list, text=p_text, prov=prov, ) else: # is paragraph not a list item # Assign proper label to the text, depending if it's a Title or Section Header # For other types of text, assign - PARAGRAPH doc_label = DocItemLabel.PARAGRAPH if shape.is_placeholder: placeholder_type = shape.placeholder_format.type if placeholder_type in [ PP_PLACEHOLDER.CENTER_TITLE, PP_PLACEHOLDER.TITLE, ]: # It's a title doc_label = DocItemLabel.TITLE elif placeholder_type == PP_PLACEHOLDER.SUBTITLE: DocItemLabel.SECTION_HEADER # output accumulated inline text: doc.add_text( label=doc_label, parent=parent_slide, text=p_text, prov=prov, ) return def handle_title(self, shape, parent_slide, slide_ind, doc): placeholder_type = shape.placeholder_format.type txt = shape.text.strip() prov = self.generate_prov(shape, slide_ind, txt) if len(txt.strip()) > 0: # title = slide.shapes.title.text if slide.shapes.title else "No title" if placeholder_type in [PP_PLACEHOLDER.CENTER_TITLE, PP_PLACEHOLDER.TITLE]: _log.info(f"Title found: {shape.text}") doc.add_text( label=DocItemLabel.TITLE, parent=parent_slide, text=txt, prov=prov ) elif placeholder_type == PP_PLACEHOLDER.SUBTITLE: _log.info(f"Subtitle found: {shape.text}") # Using DocItemLabel.FOOTNOTE, while SUBTITLE label is not avail. doc.add_text( label=DocItemLabel.SECTION_HEADER, parent=parent_slide, text=txt, prov=prov, ) return def handle_pictures(self, shape, parent_slide, slide_ind, doc, slide_size): # Open it with PIL try: # Get the image bytes image = shape.image image_bytes = image.blob im_dpi, _ = image.dpi pil_image = Image.open(BytesIO(image_bytes)) # shape has picture prov = self.generate_prov(shape, slide_ind, "", slide_size) doc.add_picture( parent=parent_slide, image=ImageRef.from_pil(image=pil_image, dpi=im_dpi), caption=None, prov=prov, ) except (UnidentifiedImageError, OSError) as e: _log.warning(f"Warning: image cannot be loaded by Pillow: {e}") return def handle_tables(self, shape, parent_slide, slide_ind, doc, slide_size): # Handling tables, images, charts if shape.has_table: table = shape.table table_xml = shape._element prov = self.generate_prov(shape, slide_ind, "", slide_size) num_cols = 0 num_rows = len(table.rows) tcells = [] # Access the XML element for the shape that contains the table table_xml = shape._element for row_idx, row in enumerate(table.rows): if len(row.cells) > num_cols: num_cols = len(row.cells) for col_idx, cell in enumerate(row.cells): # Access the XML of the cell (this is the 'tc' element in table XML) cell_xml = table_xml.xpath( f".//a:tbl/a:tr[{row_idx + 1}]/a:tc[{col_idx + 1}]" ) if not cell_xml: continue # If no cell XML is found, skip cell_xml = cell_xml[0] # Get the first matching XML node row_span = cell_xml.get("rowSpan") # Vertical span col_span = cell_xml.get("gridSpan") # Horizontal span if row_span is None: row_span = 1 else: row_span = int(row_span) if col_span is None: col_span = 1 else: col_span = int(col_span) icell = TableCell( text=cell.text.strip(), row_span=row_span, col_span=col_span, start_row_offset_idx=row_idx, end_row_offset_idx=row_idx + row_span, start_col_offset_idx=col_idx, end_col_offset_idx=col_idx + col_span, column_header=row_idx == 0, row_header=False, ) if len(cell.text.strip()) > 0: tcells.append(icell) # Initialize Docling TableData data = TableData(num_rows=num_rows, num_cols=num_cols, table_cells=[]) # Populate for tcell in tcells: data.table_cells.append(tcell) if len(tcells) > 0: # If table is not fully empty... # Create Docling table doc.add_table(parent=parent_slide, data=data, prov=prov) return def walk_linear(self, pptx_obj, doc) -> DoclingDocument: # Units of size in PPTX by default are EMU units (English Metric Units) slide_width = pptx_obj.slide_width slide_height = pptx_obj.slide_height max_levels = 10 parents = {} # type: ignore for i in range(max_levels): parents[i] = None # Loop through each slide for slide_num, slide in enumerate(pptx_obj.slides): slide_ind = pptx_obj.slides.index(slide) parent_slide = doc.add_group( name=f"slide-{slide_ind}", label=GroupLabel.CHAPTER, parent=parents[0] ) slide_size = Size(width=slide_width, height=slide_height) doc.add_page(page_no=slide_ind + 1, size=slide_size) def handle_shapes(shape, parent_slide, slide_ind, doc, slide_size): handle_groups(shape, parent_slide, slide_ind, doc, slide_size) if shape.has_table: # Handle Tables self.handle_tables(shape, parent_slide, slide_ind, doc, slide_size) if shape.shape_type == MSO_SHAPE_TYPE.PICTURE: # Handle Pictures if hasattr(shape, "image"): self.handle_pictures( shape, parent_slide, slide_ind, doc, slide_size ) # If shape doesn't have any text, move on to the next shape if not hasattr(shape, "text"): return if shape.text is None: return if len(shape.text.strip()) == 0: return if not shape.has_text_frame: _log.warning("Warning: shape has text but not text_frame") return # Handle other text elements, including lists (bullet lists, numbered lists) self.handle_text_elements( shape, parent_slide, slide_ind, doc, slide_size ) return def handle_groups(shape, parent_slide, slide_ind, doc, slide_size): if shape.shape_type == MSO_SHAPE_TYPE.GROUP: for groupedshape in shape.shapes: handle_shapes( groupedshape, parent_slide, slide_ind, doc, slide_size ) # Loop through each shape in the slide for shape in slide.shapes: handle_shapes(shape, parent_slide, slide_ind, doc, slide_size) # Handle notes slide if slide.has_notes_slide: notes_slide = slide.notes_slide if notes_slide.notes_text_frame is not None: notes_text = notes_slide.notes_text_frame.text.strip() if notes_text: bbox = BoundingBox(l=0, t=0, r=0, b=0) prov = ProvenanceItem( page_no=slide_ind + 1, charspan=[0, len(notes_text)], bbox=bbox, ) doc.add_text( label=DocItemLabel.TEXT, parent=parent_slide, text=notes_text, prov=prov, content_layer=ContentLayer.FURNITURE, ) return doc
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/xml/__init__.py
docling/backend/xml/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/xml/jats_backend.py
docling/backend/xml/jats_backend.py
import logging import traceback from io import BytesIO from pathlib import Path from typing import Final, Optional, Union, cast from bs4 import BeautifulSoup, NavigableString, Tag from docling_core.types.doc import ( DocItemLabel, DoclingDocument, DocumentOrigin, GroupItem, GroupLabel, NodeItem, TableCell, TableData, TextItem, ) from lxml import etree from typing_extensions import TypedDict, override from docling.backend.abstract_backend import DeclarativeDocumentBackend from docling.backend.html_backend import HTMLDocumentBackend from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) JATS_DTD_URL: Final = ["JATS-journalpublishing", "JATS-archive"] DEFAULT_HEADER_ACKNOWLEDGMENTS: Final = "Acknowledgments" DEFAULT_HEADER_ABSTRACT: Final = "Abstract" DEFAULT_HEADER_REFERENCES: Final = "References" DEFAULT_TEXT_ETAL: Final = "et al." class Abstract(TypedDict): label: str content: str class Author(TypedDict): name: str affiliation_names: list[str] class Citation(TypedDict): author_names: str title: str source: str year: str volume: str page: str pub_id: str publisher_name: str publisher_loc: str class Table(TypedDict): label: str caption: str content: str class XMLComponents(TypedDict): title: str authors: list[Author] abstract: list[Abstract] class JatsDocumentBackend(DeclarativeDocumentBackend): """Backend to parse articles in XML format tagged according to JATS definition. The Journal Article Tag Suite (JATS) is an definition standard for the representation of journal articles in XML format. Several publishers and journal archives provide content in JATS format, including PubMed Central® (PMC), bioRxiv, medRxiv, or Springer Nature. Refer to https://jats.nlm.nih.gov for more details on JATS. The code from this document backend has been developed by modifying parts of the PubMed Parser library (version 0.5.0, released on 12.08.2024): Achakulvisut et al., (2020). Pubmed Parser: A Python Parser for PubMed Open-Access XML Subset and MEDLINE XML Dataset XML Dataset. Journal of Open Source Software, 5(46), 1979, https://doi.org/10.21105/joss.01979 """ @override def __init__( self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path] ) -> None: super().__init__(in_doc, path_or_stream) self.path_or_stream = path_or_stream # Initialize the root of the document hierarchy self.root: Optional[NodeItem] = None self.hlevel: int = 0 self.valid: bool = False try: if isinstance(self.path_or_stream, BytesIO): self.path_or_stream.seek(0) self.tree: etree._ElementTree = etree.parse(self.path_or_stream) doc_info: etree.DocInfo = self.tree.docinfo if doc_info.system_url and any( kwd in doc_info.system_url for kwd in JATS_DTD_URL ): self.valid = True return for ent in doc_info.internalDTD.iterentities(): if ent.system_url and any( kwd in ent.system_url for kwd in JATS_DTD_URL ): self.valid = True return except Exception as exc: raise RuntimeError( f"Could not initialize JATS backend for file with hash {self.document_hash}." ) from exc @override def is_valid(self) -> bool: return self.valid @classmethod @override def supports_pagination(cls) -> bool: return False @override def unload(self): if isinstance(self.path_or_stream, BytesIO): self.path_or_stream.close() self.path_or_stream = None @classmethod @override def supported_formats(cls) -> set[InputFormat]: return {InputFormat.XML_JATS} @override def convert(self) -> DoclingDocument: try: # Create empty document origin = DocumentOrigin( filename=self.file.name or "file", mimetype="application/xml", binary_hash=self.document_hash, ) doc = DoclingDocument(name=self.file.stem or "file", origin=origin) self.hlevel = 0 # Get metadata XML components xml_components: XMLComponents = self._parse_metadata() # Add metadata to the document self._add_metadata(doc, xml_components) # walk over the XML body body = self.tree.xpath("//body") if self.root and len(body) > 0: self._walk_linear(doc, self.root, body[0]) # walk over the XML back matter back = self.tree.xpath("//back") if self.root and len(back) > 0: self._walk_linear(doc, self.root, back[0]) except Exception: _log.error(traceback.format_exc()) return doc @staticmethod def _get_text(node: etree._Element, sep: Optional[str] = None) -> str: skip_tags = ["term", "disp-formula", "inline-formula"] text: str = ( node.text.replace("\n", " ") if (node.tag not in skip_tags and node.text) else "" ) for child in list(node): if child.tag not in skip_tags: # TODO: apply styling according to child.tag when supported by docling-core text += JatsDocumentBackend._get_text(child, sep) if sep: text = text.rstrip(sep) + sep text += child.tail.replace("\n", " ") if child.tail else "" return text def _find_metadata(self) -> Optional[etree._Element]: meta_names: list[str] = ["article-meta", "book-part-meta"] meta: Optional[etree._Element] = None for name in meta_names: node = self.tree.xpath(f".//{name}") if len(node) > 0: meta = node[0] break return meta def _parse_abstract(self) -> list[Abstract]: # TODO: address cases with multiple sections abs_list: list[Abstract] = [] for abs_node in self.tree.xpath(".//abstract"): abstract: Abstract = dict(label="", content="") texts = [] for abs_par in abs_node.xpath("p"): texts.append(JatsDocumentBackend._get_text(abs_par).strip()) abstract["content"] = " ".join(texts) label_node = abs_node.xpath("title|label") if len(label_node) > 0: abstract["label"] = label_node[0].text.strip() abs_list.append(abstract) return abs_list def _parse_authors(self) -> list[Author]: # Get mapping between affiliation ids and names authors: list[Author] = [] meta: Optional[etree._Element] = self._find_metadata() if meta is None: return authors affiliation_names = [] for affiliation_node in meta.xpath(".//aff[@id]"): aff = ", ".join([t for t in affiliation_node.itertext() if t.strip()]) aff = aff.replace("\n", " ") label = affiliation_node.xpath("label") if label: # TODO: once superscript is supported, add label with formatting aff = aff.removeprefix(f"{label[0].text}, ") affiliation_names.append(aff) affiliation_ids_names = dict( zip(meta.xpath(".//aff[@id]/@id"), affiliation_names) ) # Get author names and affiliation names for author_node in meta.xpath( './/contrib-group/contrib[@contrib-type="author"]' ): author: Author = { "name": "", "affiliation_names": [], } # Affiliation names affiliation_ids = [ a.attrib["rid"] for a in author_node.xpath('xref[@ref-type="aff"]') ] for id in affiliation_ids: if id in affiliation_ids_names: author["affiliation_names"].append(affiliation_ids_names[id]) # Name author["name"] = ( author_node.xpath("name/given-names")[0].text + " " + author_node.xpath("name/surname")[0].text ) authors.append(author) return authors def _parse_title(self) -> str: meta_names: list[str] = [ "article-meta", "collection-meta", "book-meta", "book-part-meta", ] title_names: list[str] = ["article-title", "subtitle", "title", "label"] titles: list[str] = [ " ".join( elem.text.replace("\n", " ").strip() for elem in list(title_node) if elem.tag in title_names ).strip() for title_node in self.tree.xpath( "|".join([f".//{item}/title-group" for item in meta_names]) ) ] text = " - ".join(titles) return text def _parse_metadata(self) -> XMLComponents: """Parsing JATS document metadata.""" xml_components: XMLComponents = { "title": self._parse_title(), "authors": self._parse_authors(), "abstract": self._parse_abstract(), } return xml_components def _add_abstract( self, doc: DoclingDocument, xml_components: XMLComponents ) -> None: for abstract in xml_components["abstract"]: text: str = abstract["content"] title: str = abstract["label"] or DEFAULT_HEADER_ABSTRACT if not text: continue parent = doc.add_heading( parent=self.root, text=title, level=self.hlevel + 1 ) doc.add_text( parent=parent, text=text, label=DocItemLabel.TEXT, ) return def _add_authors(self, doc: DoclingDocument, xml_components: XMLComponents) -> None: # TODO: once docling supports text formatting, add affiliation reference to # author names through superscripts authors: list = [item["name"] for item in xml_components["authors"]] authors_str = ", ".join(authors) affiliations: list = [ item for author in xml_components["authors"] for item in author["affiliation_names"] ] affiliations_str = "; ".join(list(dict.fromkeys(affiliations))) if authors_str: doc.add_text( parent=self.root, text=authors_str, label=DocItemLabel.PARAGRAPH, ) if affiliations_str: doc.add_text( parent=self.root, text=affiliations_str, label=DocItemLabel.PARAGRAPH, ) return def _add_citation(self, doc: DoclingDocument, parent: NodeItem, text: str) -> None: if isinstance(parent, GroupItem) and parent.label == GroupLabel.LIST: doc.add_list_item(text=text, enumerated=False, parent=parent) else: doc.add_text(text=text, label=DocItemLabel.TEXT, parent=parent) return def _parse_element_citation(self, node: etree._Element) -> str: citation: Citation = { "author_names": "", "title": "", "source": "", "year": "", "volume": "", "page": "", "pub_id": "", "publisher_name": "", "publisher_loc": "", } _log.debug("Citation parsing started") # Author names names = [] for name_node in node.xpath(".//name"): name_str = ( name_node.xpath("surname")[0].text.replace("\n", " ").strip() + " " + name_node.xpath("given-names")[0].text.replace("\n", " ").strip() ) names.append(name_str) etal_node = node.xpath(".//etal") if len(etal_node) > 0: etal_text = etal_node[0].text or DEFAULT_TEXT_ETAL names.append(etal_text) citation["author_names"] = ", ".join(names) titles: list[str] = [ "article-title", "chapter-title", "data-title", "issue-title", "part-title", "trans-title", ] title_node: Optional[etree._Element] = None for name in titles: name_node = node.xpath(name) if len(name_node) > 0: title_node = name_node[0] break citation["title"] = ( JatsDocumentBackend._get_text(title_node) if title_node is not None else node.text.replace("\n", " ").strip() ) # Journal, year, publisher name, publisher location, volume, elocation fields: list[str] = [ "source", "year", "publisher-name", "publisher-loc", "volume", ] for item in fields: item_node = node.xpath(item) if len(item_node) > 0: citation[item.replace("-", "_")] = ( # type: ignore[literal-required] item_node[0].text.replace("\n", " ").strip() ) # Publication identifier if len(node.xpath("pub-id")) > 0: pub_id: list[str] = [] for id_node in node.xpath("pub-id"): id_type = id_node.get("assigning-authority") or id_node.get( "pub-id-type" ) id_text = id_node.text if id_type and id_text: pub_id.append( id_type.replace("\n", " ").strip().upper() + ": " + id_text.replace("\n", " ").strip() ) if pub_id: citation["pub_id"] = ", ".join(pub_id) # Pages if len(node.xpath("elocation-id")) > 0: citation["page"] = ( node.xpath("elocation-id")[0].text.replace("\n", " ").strip() ) elif len(node.xpath("fpage")) > 0: citation["page"] = node.xpath("fpage")[0].text.replace("\n", " ").strip() if len(node.xpath("lpage")) > 0: citation["page"] += ( "–" + node.xpath("lpage")[0].text.replace("\n", " ").strip() # noqa: RUF001 ) # Flatten the citation to string text = "" if citation["author_names"]: text += citation["author_names"].rstrip(".") + ". " if citation["title"]: text += citation["title"] + ". " if citation["source"]: text += citation["source"] + ". " if citation["publisher_name"]: if citation["publisher_loc"]: text += f"{citation['publisher_loc']}: " text += citation["publisher_name"] + ". " if citation["volume"]: text = text.rstrip(". ") text += f" {citation['volume']}. " if citation["page"]: text = text.rstrip(". ") if citation["volume"]: text += ":" text += citation["page"] + ". " if citation["year"]: text = text.rstrip(". ") text += f" ({citation['year']})." if citation["pub_id"]: text = text.rstrip(".") + ". " text += citation["pub_id"] _log.debug("Citation flattened") return text def _add_equation( self, doc: DoclingDocument, parent: NodeItem, node: etree._Element ) -> None: math_text = node.text math_parts = math_text.split("$$") if len(math_parts) == 3: math_formula = math_parts[1] doc.add_text(label=DocItemLabel.FORMULA, text=math_formula, parent=parent) return def _add_figure_captions( self, doc: DoclingDocument, parent: NodeItem, node: etree._Element ) -> None: label_node = node.xpath("label") label: Optional[str] = ( JatsDocumentBackend._get_text(label_node[0]).strip() if label_node else "" ) caption_node = node.xpath("caption") caption: Optional[str] if len(caption_node) > 0: caption = "" for caption_par in list(caption_node[0]): if caption_par.xpath(".//supplementary-material"): continue caption += JatsDocumentBackend._get_text(caption_par).strip() + " " caption = caption.strip() else: caption = None # TODO: format label vs caption once styling is supported fig_text: str = f"{label}{' ' if label and caption else ''}{caption}" fig_caption: Optional[TextItem] = ( doc.add_text(label=DocItemLabel.CAPTION, text=fig_text) if fig_text else None ) doc.add_picture(parent=parent, caption=fig_caption) return # TODO: add footnotes when DocItemLabel.FOOTNOTE and styling are supported # def _add_footnote_group(self, doc: DoclingDocument, parent: NodeItem, node: etree._Element) -> None: # new_parent = doc.add_group(label=GroupLabel.LIST, name="footnotes", parent=parent) # for child in node.iterchildren(tag="fn"): # text = JatsDocumentBackend._get_text(child) # doc.add_list_item(text=text, parent=new_parent) def _add_metadata( self, doc: DoclingDocument, xml_components: XMLComponents ) -> None: self._add_title(doc, xml_components) self._add_authors(doc, xml_components) self._add_abstract(doc, xml_components) return @staticmethod def parse_table_data(element: Tag) -> Optional[TableData]: # TODO, see how to implement proper support for rich tables from HTML backend nested_tables = element.find("table") if nested_tables is not None: _log.debug("Skipping nested table.") return None # Find the number of rows and columns (taking into account spans) num_rows = 0 num_cols = 0 for row in element("tr"): col_count = 0 is_row_header = True if not isinstance(row, Tag): continue for cell in row(["td", "th"]): if not isinstance(row, Tag): continue cell_tag = cast(Tag, cell) col_span, row_span = HTMLDocumentBackend._get_cell_spans(cell_tag) col_count += col_span if cell_tag.name == "td" or row_span == 1: is_row_header = False num_cols = max(num_cols, col_count) if not is_row_header: num_rows += 1 _log.debug(f"The table has {num_rows} rows and {num_cols} cols.") grid: list = [[None for _ in range(num_cols)] for _ in range(num_rows)] data = TableData(num_rows=num_rows, num_cols=num_cols, table_cells=[]) # Iterate over the rows in the table start_row_span = 0 row_idx = -1 for row in element("tr"): if not isinstance(row, Tag): continue # For each row, find all the column cells (both <td> and <th>) cells = row(["td", "th"]) # Check if cell is in a column header or row header col_header = True row_header = True for html_cell in cells: if isinstance(html_cell, Tag): _, row_span = HTMLDocumentBackend._get_cell_spans(html_cell) if html_cell.name == "td": col_header = False row_header = False elif row_span == 1: row_header = False if not row_header: row_idx += 1 start_row_span = 0 else: start_row_span += 1 # Extract the text content of each cell col_idx = 0 for html_cell in cells: if not isinstance(html_cell, Tag): continue # extract inline formulas for formula in html_cell("inline-formula"): math_parts = formula.text.split("$$") if len(math_parts) == 3: math_formula = f"$${math_parts[1]}$$" formula.replace_with(NavigableString(math_formula)) # TODO: extract content correctly from table-cells with lists text = HTMLDocumentBackend.get_text(html_cell).strip() col_span, row_span = HTMLDocumentBackend._get_cell_spans(html_cell) if row_header: row_span -= 1 while ( col_idx < num_cols and grid[row_idx + start_row_span][col_idx] is not None ): col_idx += 1 for r in range(start_row_span, start_row_span + row_span): for c in range(col_span): if row_idx + r < num_rows and col_idx + c < num_cols: grid[row_idx + r][col_idx + c] = text table_cell = TableCell( text=text, row_span=row_span, col_span=col_span, start_row_offset_idx=start_row_span + row_idx, end_row_offset_idx=start_row_span + row_idx + row_span, start_col_offset_idx=col_idx, end_col_offset_idx=col_idx + col_span, column_header=col_header, row_header=((not col_header) and html_cell.name == "th"), ) data.table_cells.append(table_cell) return data def _add_table( self, doc: DoclingDocument, parent: NodeItem, table_xml_component: Table ) -> None: soup = BeautifulSoup(table_xml_component["content"], "html.parser") table_tag = soup.find("table") if not isinstance(table_tag, Tag): return data = JatsDocumentBackend.parse_table_data(table_tag) # TODO: format label vs caption once styling is supported label = table_xml_component["label"] caption = table_xml_component["caption"] table_text: str = f"{label}{' ' if label and caption else ''}{caption}" table_caption: Optional[TextItem] = ( doc.add_text(label=DocItemLabel.CAPTION, text=table_text) if table_text else None ) if data is not None: doc.add_table(data=data, parent=parent, caption=table_caption) return def _add_tables( self, doc: DoclingDocument, parent: NodeItem, node: etree._Element ) -> None: table: Table = {"label": "", "caption": "", "content": ""} # Content if len(node.xpath("table")) > 0: table_content_node = node.xpath("table")[0] elif len(node.xpath("alternatives/table")) > 0: table_content_node = node.xpath("alternatives/table")[0] else: table_content_node = None if table_content_node is not None: table["content"] = etree.tostring(table_content_node).decode("utf-8") # Caption caption_node = node.xpath("caption") caption: Optional[str] if caption_node: caption = "" for caption_par in list(caption_node[0]): if caption_par.xpath(".//supplementary-material"): continue caption += JatsDocumentBackend._get_text(caption_par).strip() + " " caption = caption.strip() else: caption = None if caption is not None: table["caption"] = caption # Label if len(node.xpath("label")) > 0: table["label"] = node.xpath("label")[0].text try: self._add_table(doc, parent, table) except Exception: _log.warning(f"Skipping unsupported table in {self.file!s}") return def _add_title(self, doc: DoclingDocument, xml_components: XMLComponents) -> None: self.root = doc.add_text( parent=None, text=xml_components["title"], label=DocItemLabel.TITLE, ) return def _walk_linear( self, doc: DoclingDocument, parent: NodeItem, node: etree._Element ) -> str: skip_tags = ["term"] flush_tags = ["ack", "sec", "list", "boxed-text", "disp-formula", "fig"] new_parent: NodeItem = parent node_text: str = ( node.text.replace("\n", " ") if (node.tag not in skip_tags and node.text) else "" ) for child in list(node): stop_walk: bool = False # flush text into TextItem for some tags in paragraph nodes if node.tag == "p" and node_text.strip() and child.tag in flush_tags: doc.add_text( label=DocItemLabel.TEXT, text=node_text.strip(), parent=parent ) node_text = "" # add elements and decide whether to stop walking if child.tag in ("sec", "ack"): header = child.xpath("title|label") text: Optional[str] = None if len(header) > 0: text = JatsDocumentBackend._get_text(header[0]) elif child.tag == "ack": text = DEFAULT_HEADER_ACKNOWLEDGMENTS if text: self.hlevel += 1 new_parent = doc.add_heading( text=text, parent=parent, level=self.hlevel ) elif child.tag == "list": new_parent = doc.add_group( label=GroupLabel.LIST, name="list", parent=parent ) elif child.tag == "list-item": # TODO: address any type of content (another list, formula,...) # TODO: address list type and item label text = JatsDocumentBackend._get_text(child).strip() new_parent = doc.add_list_item(text=text, parent=parent) stop_walk = True elif child.tag == "fig": self._add_figure_captions(doc, parent, child) stop_walk = True elif child.tag == "table-wrap": self._add_tables(doc, parent, child) stop_walk = True elif child.tag == "suplementary-material": stop_walk = True elif child.tag == "fn-group": # header = child.xpath(".//title") or child.xpath(".//label") # if header: # text = JatsDocumentBackend._get_text(header[0]) # fn_parent = doc.add_heading(text=text, parent=new_parent) # self._add_footnote_group(doc, fn_parent, child) stop_walk = True elif child.tag == "ref-list" and node.tag != "ref-list": header = child.xpath("title|label") text = ( JatsDocumentBackend._get_text(header[0]) if len(header) > 0 else DEFAULT_HEADER_REFERENCES ) new_parent = doc.add_heading(text=text, parent=parent) new_parent = doc.add_group( parent=new_parent, label=GroupLabel.LIST, name="list" ) elif child.tag == "element-citation": text = self._parse_element_citation(child) self._add_citation(doc, parent, text) stop_walk = True elif child.tag == "mixed-citation": text = JatsDocumentBackend._get_text(child).strip() self._add_citation(doc, parent, text) stop_walk = True elif child.tag == "tex-math": self._add_equation(doc, parent, child) stop_walk = True elif child.tag == "inline-formula": # TODO: address inline formulas when supported by docling-core stop_walk = True # step into child if not stop_walk: new_text = self._walk_linear(doc, new_parent, child) if not (node.getparent().tag == "p" and node.tag in flush_tags): node_text += new_text if child.tag in ("sec", "ack") and text: self.hlevel -= 1 # pick up the tail text node_text += child.tail.replace("\n", " ") if child.tail else "" # create paragraph if node.tag == "p" and node_text.strip(): doc.add_text(label=DocItemLabel.TEXT, text=node_text.strip(), parent=parent) return "" else: # backpropagate the text return node_text
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/xml/uspto_backend.py
docling/backend/xml/uspto_backend.py
"""Backend to parse patents from the United States Patent Office (USPTO). The parsers included in this module can handle patent grants published since 1976 and patent applications since 2001. The original files can be found in https://bulkdata.uspto.gov. """ import html import logging import re import xml.sax import xml.sax.xmlreader from abc import ABC, abstractmethod from enum import Enum, unique from io import BytesIO from pathlib import Path from typing import Final, Optional, Union from bs4 import BeautifulSoup, Tag from docling_core.types.doc import ( DocItem, DocItemLabel, DoclingDocument, DocumentOrigin, TableCell, TableData, TextItem, ) from docling_core.types.doc.document import LevelNumber from pydantic import NonNegativeInt from typing_extensions import Self, TypedDict, override from docling.backend.abstract_backend import DeclarativeDocumentBackend from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument _log = logging.getLogger(__name__) XML_DECLARATION: Final = '<?xml version="1.0" encoding="UTF-8"?>' @unique class PatentHeading(Enum): """Text of docling headings for tagged sections in USPTO patent documents.""" ABSTRACT = "ABSTRACT", 2 CLAIMS = "CLAIMS", 2 @override def __new__(cls, value: str, _) -> Self: obj = object.__new__(cls) obj._value_ = value return obj @override def __init__(self, _, level: LevelNumber) -> None: self.level: LevelNumber = level class PatentUsptoDocumentBackend(DeclarativeDocumentBackend): @override def __init__( self, in_doc: InputDocument, path_or_stream: Union[BytesIO, Path] ) -> None: super().__init__(in_doc, path_or_stream) self.patent_content: str = "" self.parser: Optional[PatentUspto] = None try: if isinstance(self.path_or_stream, BytesIO): while line := self.path_or_stream.readline().decode("utf-8"): if line.startswith("<!DOCTYPE") or line == "PATN\n": self._set_parser(line) self.patent_content += line elif isinstance(self.path_or_stream, Path): with open(self.path_or_stream, encoding="utf-8") as file_obj: while line := file_obj.readline(): if line.startswith("<!DOCTYPE") or line == "PATN\n": self._set_parser(line) self.patent_content += line except Exception as exc: raise RuntimeError( f"Could not initialize USPTO backend for file with hash {self.document_hash}." ) from exc def _set_parser(self, doctype: str) -> None: doctype_line = doctype.lower() if doctype == "PATN\n": self.parser = PatentUsptoGrantAps() elif "us-patent-application-v4" in doctype_line: self.parser = PatentUsptoIce() elif "us-patent-grant-v4" in doctype_line: self.parser = PatentUsptoIce() elif "us-grant-025" in doctype_line: self.parser = PatentUsptoGrantV2() elif all( item in doctype_line for item in ("patent-application-publication", "pap-v1") ): self.parser = PatentUsptoAppV1() else: self.parser = None @override def is_valid(self) -> bool: return bool(self.patent_content) and bool(self.parser) @classmethod @override def supports_pagination(cls) -> bool: return False @override def unload(self) -> None: return @classmethod @override def supported_formats(cls) -> set[InputFormat]: return {InputFormat.XML_USPTO} @override def convert(self) -> DoclingDocument: if self.parser is not None: doc = self.parser.parse(self.patent_content) if doc is None: raise RuntimeError( f"Failed to convert doc (hash={self.document_hash}, " f"name={self.file.name})." ) doc.name = self.file.name or "file" mime_type = ( "text/plain" if isinstance(self.parser, PatentUsptoGrantAps) else "application/xml" ) doc.origin = DocumentOrigin( mimetype=mime_type, binary_hash=self.document_hash, filename=self.file.name or "file", ) return doc else: raise RuntimeError( f"Cannot convert doc (hash={self.document_hash}, " f"name={self.file.name}) because the backend failed to init." ) class PatentUspto(ABC): """Parser of patent documents from the US Patent Office.""" @abstractmethod def parse(self, patent_content: str) -> Optional[DoclingDocument]: """Parse a USPTO patent. Parameters: patent_content: The content of a single patent in a USPTO file. Returns: The patent parsed as a docling document. """ class PatentUsptoIce(PatentUspto): """Parser of patent documents from the US Patent Office (ICE). The compatible formats are: - Patent Grant Full Text Data/XML Version 4.x ICE (from January 2005) - Patent Application Full Text Data/XML Version 4.x ICE (from January 2005) """ def __init__(self) -> None: """Build an instance of PatentUsptoIce class.""" self.handler = PatentUsptoIce.PatentHandler() self.pattern = re.compile(r"^(<table .*?</table>)", re.MULTILINE | re.DOTALL) def parse(self, patent_content: str) -> Optional[DoclingDocument]: try: xml.sax.parseString(patent_content, self.handler) except xml.sax._exceptions.SAXParseException as exc_sax: _log.error(f"Error in parsing USPTO document: {exc_sax}") return None doc = self.handler.doc if doc: raw_tables = re.findall(self.pattern, patent_content) parsed_tables: list[TableData] = [] _log.debug(f"Found {len(raw_tables)} tables to be parsed with XmlTable.") for table in raw_tables: table_parser = XmlTable(XML_DECLARATION + "\n" + table) try: table_data = table_parser.parse() if table_data: parsed_tables.append(table_data) except Exception as exc_table: _log.error(f"Error in parsing USPTO tables: {exc_table}") if len(parsed_tables) != len(doc.tables): _log.error( f"Number of referenced ({len(doc.tables)}) and parsed " f"({len(parsed_tables)}) tables differ." ) else: for idx, item in enumerate(parsed_tables): doc.tables[idx].data = item return doc class PatentHandler(xml.sax.handler.ContentHandler): """SAX ContentHandler for patent documents.""" APP_DOC_ELEMENT: Final = "us-patent-application" GRANT_DOC_ELEMENT: Final = "us-patent-grant" @unique class Element(Enum): """Represents an element of interest in the patent application document.""" ABSTRACT = "abstract", True TITLE = "invention-title", True CLAIMS = "claims", False CLAIM = "claim", False CLAIM_TEXT = "claim-text", True PARAGRAPH = "p", True HEADING = "heading", True DESCRIPTION = "description", False TABLE = "table", False # to track its position, without text DRAWINGS = "description-of-drawings", True STYLE_SUPERSCRIPT = "sup", True STYLE_SUBSCRIPT = "sub", True MATHS = "maths", False # to avoid keeping formulas @override def __new__(cls, value: str, _) -> Self: obj = object.__new__(cls) obj._value_ = value return obj @override def __init__(self, _, is_text: bool) -> None: self.is_text: bool = is_text @override def __init__(self) -> None: """Build an instance of the patent handler.""" # Current patent being parsed self.doc: Optional[DoclingDocument] = None # Keep track of docling hierarchy level self.level: LevelNumber = 1 # Keep track of docling parents by level self.parents: dict[LevelNumber, Optional[DocItem]] = {1: None} # Content to retain for the current patent self.property: list[str] self.claim: str self.claims: list[str] self.abstract: str self.text: str self._clean_data() # To handle mathematical styling self.style_html = HtmlEntity() @override def startElement(self, tag, attributes): """Signal the start of an element. Args: tag: The element tag. attributes: The element attributes. """ if tag in ( self.APP_DOC_ELEMENT, self.GRANT_DOC_ELEMENT, ): self.doc = DoclingDocument(name="file") self.text = "" self._start_registered_elements(tag, attributes) @override def skippedEntity(self, name): """Receive notification of a skipped entity. HTML entities will be skipped by the parser. This method will unescape them and add them to the text. Args: name: Entity name. """ if self.property: elm_val = self.property[-1] element = self.Element(elm_val) if element.is_text: escaped = self.style_html.get_greek_from_iso8879(f"&{name};") unescaped = html.unescape(escaped) if unescaped == escaped: _log.debug(f"Unrecognized HTML entity: {name}") return if element in ( self.Element.STYLE_SUPERSCRIPT, self.Element.STYLE_SUBSCRIPT, ): # superscripts and subscripts need to be under text elements if len(self.property) < 2: return parent_val = self.property[-2] parent = self.Element(parent_val) if parent.is_text: self.text += self._apply_style(unescaped, elm_val) else: self.text += unescaped @override def endElement(self, tag): """Signal the end of an element. Args: tag: The element tag. """ if tag in ( self.APP_DOC_ELEMENT, self.GRANT_DOC_ELEMENT, ): self._clean_data() self._end_registered_element(tag) @override def characters(self, content): """Receive notification of character data. Args: content: Data reported by the handler. """ if self.property: elm_val = self.property[-1] element = self.Element(elm_val) if element.is_text: if element in ( self.Element.STYLE_SUPERSCRIPT, self.Element.STYLE_SUBSCRIPT, ): # superscripts and subscripts need to be under text elements if len(self.property) < 2: return parent_val = self.property[-2] parent = self.Element(parent_val) if parent.is_text: self.text += self._apply_style(content, elm_val) else: self.text += content def _start_registered_elements( self, tag: str, attributes: xml.sax.xmlreader.AttributesImpl ) -> None: if tag in [member.value for member in self.Element]: # special case for claims: claim lines may start before the # previous one is closed if ( tag == self.Element.CLAIM_TEXT.value and self.property and self.property[-1] == tag and self.text.strip() ): self.claim += " " + self.text.strip() self.text = "" elif tag == self.Element.HEADING.value: level_attr: str = attributes.get("level", "") new_level: int = int(level_attr) if level_attr.isnumeric() else 1 max_level = min(self.parents.keys()) # increase heading level with 1 for title, if any self.level = ( new_level + 1 if (new_level + 1) in self.parents else max_level ) self.property.append(tag) def _end_registered_element(self, tag: str) -> None: if tag in [item.value for item in self.Element] and self.property: current_tag = self.property.pop() self._add_property(current_tag, self.text.strip()) def _add_property(self, name: str, text: str) -> None: if not name or not self.doc: return if name == self.Element.TITLE.value: if text: self.parents[self.level + 1] = self.doc.add_title( parent=self.parents[self.level], text=text, ) self.level += 1 self.text = "" elif name == self.Element.ABSTRACT.value: if self.abstract: heading_text = PatentHeading.ABSTRACT.value heading_level = ( PatentHeading.ABSTRACT.level if PatentHeading.ABSTRACT.level in self.parents else 1 ) abstract_item = self.doc.add_heading( heading_text, level=heading_level, parent=self.parents[heading_level], ) self.doc.add_text( label=DocItemLabel.PARAGRAPH, text=self.abstract, parent=abstract_item, ) elif name == self.Element.CLAIM_TEXT.value: text = re.sub("\\s+", " ", text).strip() if text: self.claim += " " + text self.text = "" elif name == self.Element.CLAIM.value and self.claim: self.claims.append(self.claim.strip()) self.claim = "" elif name == self.Element.CLAIMS.value and self.claims: heading_text = PatentHeading.CLAIMS.value heading_level = ( PatentHeading.CLAIMS.level if PatentHeading.CLAIMS.level in self.parents else 1 ) claims_item = self.doc.add_heading( heading_text, level=heading_level, parent=self.parents[heading_level], ) for text in self.claims: self.doc.add_text( label=DocItemLabel.PARAGRAPH, text=text, parent=claims_item ) elif name == self.Element.PARAGRAPH.value and text: # remove blank spaces added in paragraphs text = re.sub("\\s+", " ", text) if self.Element.ABSTRACT.value in self.property: self.abstract = ( (self.abstract + " " + text) if self.abstract else text ) else: self.doc.add_text( label=DocItemLabel.PARAGRAPH, text=text, parent=self.parents[self.level], ) self.text = "" elif name == self.Element.HEADING.value and text: self.parents[self.level + 1] = self.doc.add_heading( text=text, level=self.level, parent=self.parents[self.level], ) self.level += 1 self.text = "" elif name == self.Element.TABLE.value: # set an empty table as placeholder empty_table = TableData(num_rows=0, num_cols=0, table_cells=[]) self.doc.add_table( data=empty_table, parent=self.parents[self.level], ) def _apply_style(self, text: str, style_tag: str) -> str: """Apply an HTML style to text. Args: text: A string containing plain text. style_tag: An HTML tag name for styling text. If the tag name is not recognized as one of the supported styles, the method will return the original `text`. Returns: A string after applying the style. """ formatted = text if style_tag == self.Element.STYLE_SUPERSCRIPT.value: formatted = html.unescape(self.style_html.get_superscript(text)) elif style_tag == self.Element.STYLE_SUBSCRIPT.value: formatted = html.unescape(self.style_html.get_subscript(text)) return formatted def _clean_data(self) -> None: """Reset the variables from stream data.""" self.property = [] self.claim = "" self.claims = [] self.abstract = "" class PatentUsptoGrantV2(PatentUspto): """Parser of patent documents from the US Patent Office (grants v2.5). The compatible format is: - Patent Grant Full Text Data/XML Version 2.5 (from January 2002 till December 2004) """ @override def __init__(self) -> None: """Build an instance of PatentUsptoGrantV2 class.""" self.handler = PatentUsptoGrantV2.PatentHandler() self.pattern = re.compile(r"^(<table .*?</table>)", re.MULTILINE | re.DOTALL) @override def parse(self, patent_content: str) -> Optional[DoclingDocument]: try: xml.sax.parseString(patent_content, self.handler) except xml.sax._exceptions.SAXParseException as exc_sax: _log.error(f"Error in parsing USPTO document: {exc_sax}") return None doc = self.handler.doc if doc: raw_tables = re.findall(self.pattern, patent_content) parsed_tables: list[TableData] = [] _log.debug(f"Found {len(raw_tables)} tables to be parsed with XmlTable.") for table in raw_tables: table_parser = XmlTable(XML_DECLARATION + "\n" + table) try: table_data = table_parser.parse() if table_data: parsed_tables.append(table_data) except Exception as exc_table: _log.error(f"Error in parsing USPTO tables: {exc_table}") if len(parsed_tables) != len(doc.tables): _log.error( f"Number of referenced ({len(doc.tables)}) and parsed " f"({len(parsed_tables)}) tables differ." ) else: for idx, item in enumerate(parsed_tables): doc.tables[idx].data = item return doc class PatentHandler(xml.sax.handler.ContentHandler): """SAX ContentHandler for patent documents.""" GRANT_DOC_ELEMENT: Final = "PATDOC" CLAIM_STATEMENT: Final = "What is claimed is:" @unique class Element(Enum): """Represents an element of interest in the patent application document.""" PDAT = "PDAT", True # any type of data ABSTRACT = ("SDOAB", False) SDOCL = ("SDOCL", False) TITLE = ("B540", False) CLAIMS = ("CL", False) CLAIM = ("CLM", False) PARAGRAPH = ("PARA", True) HEADING = ("H", True) DRAWINGS = ("DRWDESC", False) STYLE_SUPERSCRIPT = ("SP", False) STYLE_SUBSCRIPT = ("SB", False) STYLE_ITALIC = ("ITALIC", False) CWU = ("CWU", False) # avoid tables, chemicals, formulas TABLE = ("table", False) # to keep track of table positions @override def __new__(cls, value: str, _) -> Self: obj = object.__new__(cls) obj._value_ = value return obj @override def __init__(self, _, is_text: bool) -> None: self.is_text: bool = is_text @override def __init__(self) -> None: """Build an instance of the patent handler.""" # Current patent being parsed self.doc: Optional[DoclingDocument] = None # Keep track of docling hierarchy level self.level: LevelNumber = 1 # Keep track of docling parents by level self.parents: dict[LevelNumber, Optional[DocItem]] = {1: None} # Content to retain for the current patent self.property: list[str] self.claim: str self.claims: list[str] self.paragraph: str self.abstract: str self._clean_data() # To handle mathematical styling self.style_html = HtmlEntity() @override def startElement(self, tag, attributes): """Signal the start of an element. Args: tag: The element tag. attributes: The element attributes. """ if tag == self.GRANT_DOC_ELEMENT: self.doc = DoclingDocument(name="file") self.text = "" self._start_registered_elements(tag, attributes) @override def skippedEntity(self, name): """Receive notification of a skipped entity. HTML entities will be skipped by the parser. This method will unescape them and add them to the text. Args: name: Entity name. """ if self.property: elm_val = self.property[-1] element = self.Element(elm_val) if element.is_text: escaped = self.style_html.get_greek_from_iso8879(f"&{name};") unescaped = html.unescape(escaped) if unescaped == escaped: logging.debug("Unrecognized HTML entity: " + name) return if element in ( self.Element.STYLE_SUPERSCRIPT, self.Element.STYLE_SUBSCRIPT, ): # superscripts and subscripts need to be under text elements if len(self.property) < 2: return parent_val = self.property[-2] parent = self.Element(parent_val) if parent.is_text: self.text += self._apply_style(unescaped, elm_val) else: self.text += unescaped @override def endElement(self, tag): """Signal the end of an element. Args: tag: The element tag. """ if tag == self.GRANT_DOC_ELEMENT: self._clean_data() self._end_registered_element(tag) @override def characters(self, content): """Receive notification of character data. Args: content: Data reported by the handler. """ if self.property: elm_val = self.property[-1] element = self.Element(elm_val) if element.is_text: if element in ( self.Element.STYLE_SUPERSCRIPT, self.Element.STYLE_SUBSCRIPT, ): # superscripts and subscripts need to be under text elements if len(self.property) < 2: return parent_val = self.property[-2] parent = self.Element(parent_val) if parent.is_text: self.text += self._apply_style(content, elm_val) else: self.text += content def _start_registered_elements( self, tag: str, attributes: xml.sax.xmlreader.AttributesImpl ) -> None: if tag in [member.value for member in self.Element]: if ( tag == self.Element.HEADING.value and self.Element.SDOCL.value not in self.property ): level_attr: str = attributes.get("LVL", "") new_level: int = int(level_attr) if level_attr.isnumeric() else 1 max_level = min(self.parents.keys()) # increase heading level with 1 for title, if any self.level = ( new_level + 1 if (new_level + 1) in self.parents else max_level ) self.property.append(tag) def _end_registered_element(self, tag: str) -> None: if tag in [elm.value for elm in self.Element] and self.property: current_tag = self.property.pop() self._add_property(current_tag, self.text) def _add_property(self, name: str, text: str) -> None: if not name or not self.doc: return if name == self.Element.PDAT.value and text: if not self.property: self.text = "" return wrapper = self.property[-1] text = self._apply_style(text, wrapper) if self.Element.TITLE.value in self.property and text.strip(): title = text.strip() self.parents[self.level + 1] = self.doc.add_title( parent=self.parents[self.level], text=title, ) self.level += 1 elif self.Element.ABSTRACT.value in self.property: self.abstract += text elif self.Element.CLAIM.value in self.property: self.claim += text # Paragraph text not in claims or abstract elif ( self.Element.PARAGRAPH.value in self.property and self.Element.CLAIM.value not in self.property and self.Element.ABSTRACT.value not in self.property ): self.paragraph += text # headers except claims statement elif ( self.Element.HEADING.value in self.property and self.Element.SDOCL.value not in self.property and text.strip() ): self.parents[self.level + 1] = self.doc.add_heading( text=text.strip(), level=self.level, parent=self.parents[self.level], ) self.level += 1 self.text = "" elif name == self.Element.CLAIM.value and self.claim.strip(): self.claims.append(self.claim.strip()) self.claim = "" elif name == self.Element.CLAIMS.value and self.claims: heading_text = PatentHeading.CLAIMS.value heading_level = ( PatentHeading.CLAIMS.level if PatentHeading.CLAIMS.level in self.parents else 1 ) claims_item = self.doc.add_heading( heading_text, level=heading_level, parent=self.parents[heading_level], ) for text in self.claims: self.doc.add_text( label=DocItemLabel.PARAGRAPH, text=text, parent=claims_item ) elif name == self.Element.ABSTRACT.value and self.abstract.strip(): abstract = self.abstract.strip() heading_text = PatentHeading.ABSTRACT.value heading_level = ( PatentHeading.ABSTRACT.level if PatentHeading.ABSTRACT.level in self.parents else 1 ) abstract_item = self.doc.add_heading( heading_text, level=heading_level, parent=self.parents[heading_level], ) self.doc.add_text( label=DocItemLabel.PARAGRAPH, text=abstract, parent=abstract_item ) elif name == self.Element.PARAGRAPH.value: paragraph = self.paragraph.strip() if paragraph and self.Element.CLAIM.value not in self.property: self.doc.add_text( label=DocItemLabel.PARAGRAPH, text=paragraph, parent=self.parents[self.level], ) elif self.Element.CLAIM.value in self.property: # we may need a space after a paragraph in claim text self.claim += " " self.paragraph = "" elif name == self.Element.TABLE.value: # set an empty table as placeholder empty_table = TableData(num_rows=0, num_cols=0, table_cells=[]) self.doc.add_table( data=empty_table, parent=self.parents[self.level], ) def _apply_style(self, text: str, style_tag: str) -> str: """Apply an HTML style to text. Args: text: A string containing plain text. style_tag: An HTML tag name for styling text. If the tag name is not recognized as one of the supported styles, the method will return the original `text`. Returns: A string after applying the style. """ formatted = text if style_tag == self.Element.STYLE_SUPERSCRIPT.value: formatted = html.unescape(self.style_html.get_superscript(text)) elif style_tag == self.Element.STYLE_SUBSCRIPT.value: formatted = html.unescape(self.style_html.get_subscript(text)) elif style_tag == self.Element.STYLE_ITALIC.value: formatted = html.unescape(self.style_html.get_math_italic(text)) return formatted def _clean_data(self) -> None: """Reset the variables from stream data.""" self.text = "" self.property = [] self.claim = "" self.claims = [] self.paragraph = "" self.abstract = "" class PatentUsptoGrantAps(PatentUspto): """Parser of patents documents from the US Patent Office (grants APS). The compatible format is: - Patent Grant Full Text Data/APS (from January 1976 till December 2001) """ @unique class Section(Enum): """Represent a section in a patent APS document.""" ABSTRACT = "ABST" SUMMARY = "BSUM" DETAILS = "DETD" CLAIMS = "CLMS" DRAWINGS = "DRWD" @unique class Field(Enum):
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
true
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/json/docling_json_backend.py
docling/backend/json/docling_json_backend.py
from io import BytesIO from pathlib import Path from typing import Union from docling_core.types.doc import DoclingDocument from typing_extensions import override from docling.backend.abstract_backend import DeclarativeDocumentBackend from docling.datamodel.base_models import InputFormat from docling.datamodel.document import InputDocument class DoclingJSONBackend(DeclarativeDocumentBackend): @override def __init__( self, in_doc: InputDocument, path_or_stream: Union[BytesIO, Path] ) -> None: super().__init__(in_doc, path_or_stream) # given we need to store any actual conversion exception for raising it from # convert(), this captures the successful result or the actual error in a # mutually exclusive way: self._doc_or_err = self._get_doc_or_err() @override def is_valid(self) -> bool: return isinstance(self._doc_or_err, DoclingDocument) @classmethod @override def supports_pagination(cls) -> bool: return False @classmethod @override def supported_formats(cls) -> set[InputFormat]: return {InputFormat.JSON_DOCLING} def _get_doc_or_err(self) -> Union[DoclingDocument, Exception]: try: json_data: Union[str, bytes] if isinstance(self.path_or_stream, Path): with open(self.path_or_stream, encoding="utf-8") as f: json_data = f.read() elif isinstance(self.path_or_stream, BytesIO): json_data = self.path_or_stream.getvalue() else: raise RuntimeError(f"Unexpected: {type(self.path_or_stream)=}") return DoclingDocument.model_validate_json(json_data=json_data) except Exception as e: return e @override def convert(self) -> DoclingDocument: if isinstance(self._doc_or_err, DoclingDocument): return self._doc_or_err else: raise self._doc_or_err
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/json/__init__.py
docling/backend/json/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/docx/__init__.py
docling/backend/docx/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/docx/latex/omml.py
docling/backend/docx/latex/omml.py
""" Office Math Markup Language (OMML) Adapted from https://github.com/xiilei/dwml/blob/master/dwml/omml.py On 23/01/2025 """ import logging import lxml.etree as ET from pylatexenc.latexencode import UnicodeToLatexEncoder from docling.backend.docx.latex.latex_dict import ( ALN, ARR, BACKSLASH, BLANK, BRK, CHARS, CHR, CHR_BO, CHR_DEFAULT, D_DEFAULT, F_DEFAULT, FUNC, FUNC_PLACE, LIM_FUNC, LIM_TO, LIM_UPP, POS, POS_DEFAULT, RAD, RAD_DEFAULT, SUB, SUP, D, F, M, T, ) OMML_NS = "{http://schemas.openxmlformats.org/officeDocument/2006/math}" _log = logging.getLogger(__name__) def load(stream): tree = ET.parse(stream) for omath in tree.findall(OMML_NS + "oMath"): yield oMath2Latex(omath) def load_string(string): root = ET.fromstring(string) for omath in root.findall(OMML_NS + "oMath"): yield oMath2Latex(omath) def escape_latex(strs): last = None new_chr = [] strs = strs.replace(r"\\", "\\") for c in strs: if (c in CHARS) and (last != BACKSLASH): new_chr.append(BACKSLASH + c) else: new_chr.append(c) last = c return BLANK.join(new_chr) def get_val(key, default=None, store=CHR): if key is not None: return key if not store else store.get(key, key) else: return default class Tag2Method: def call_method(self, elm, stag=None): getmethod = self.tag2meth.get if stag is None: stag = elm.tag.replace(OMML_NS, "") method = getmethod(stag) if method: return method(self, elm) else: return None def process_children_list(self, elm, include=None): """ process children of the elm,return iterable """ for _e in list(elm): if OMML_NS not in _e.tag: continue stag = _e.tag.replace(OMML_NS, "") if include and (stag not in include): continue t = self.call_method(_e, stag=stag) if t is None: t = self.process_unknow(_e, stag) if t is None: continue yield (stag, t, _e) def process_children_dict(self, elm, include=None): """ process children of the elm,return dict """ latex_chars = dict() for stag, t, e in self.process_children_list(elm, include): latex_chars[stag] = t return latex_chars def process_children(self, elm, include=None): """ process children of the elm,return string """ return BLANK.join( ( t if not isinstance(t, Tag2Method) else str(t) for stag, t, e in self.process_children_list(elm, include) ) ) def process_unknow(self, elm, stag): return None class Pr(Tag2Method): text = "" __val_tags = ("chr", "pos", "begChr", "endChr", "type") __innerdict = None # can't use the __dict__ """ common properties of element""" def __init__(self, elm): self.__innerdict = {} self.text = self.process_children(elm) def __str__(self): return self.text def __unicode__(self): return self.__str__(self) def __getattr__(self, name): return self.__innerdict.get(name, None) def do_brk(self, elm): self.__innerdict["brk"] = BRK return BRK def do_common(self, elm): stag = elm.tag.replace(OMML_NS, "") if stag in self.__val_tags: t = elm.get(f"{OMML_NS}val") self.__innerdict[stag] = t return None tag2meth = { "brk": do_brk, "chr": do_common, "pos": do_common, "begChr": do_common, "endChr": do_common, "type": do_common, } class oMath2Latex(Tag2Method): """ Convert oMath element of omml to latex """ _t_dict = T __direct_tags = ("box", "sSub", "sSup", "sSubSup", "num", "den", "deg", "e") u = UnicodeToLatexEncoder( replacement_latex_protection="braces-all", unknown_char_policy="keep", unknown_char_warning=False, ) def __init__(self, element): self._latex = self.process_children(element) def __str__(self): return self.latex.replace(" ", " ") def __unicode__(self): return self.__str__(self) def process_unknow(self, elm, stag): if stag in self.__direct_tags: return self.process_children(elm) elif stag[-2:] == "Pr": return Pr(elm) else: return None @property def latex(self): return self._latex def do_acc(self, elm): """ the accent function """ c_dict = self.process_children_dict(elm) latex_s = get_val( c_dict["accPr"].chr, default=CHR_DEFAULT.get("ACC_VAL"), store=CHR ) return latex_s.format(c_dict["e"]) def do_bar(self, elm): """ the bar function """ c_dict = self.process_children_dict(elm) pr = c_dict["barPr"] latex_s = get_val(pr.pos, default=POS_DEFAULT.get("BAR_VAL"), store=POS) return pr.text + latex_s.format(c_dict["e"]) def do_d(self, elm): """ the delimiter object """ c_dict = self.process_children_dict(elm) pr = c_dict["dPr"] null = D_DEFAULT.get("null") s_val = get_val(pr.begChr, default=D_DEFAULT.get("left"), store=T) e_val = get_val(pr.endChr, default=D_DEFAULT.get("right"), store=T) delim = pr.text + D.format( left=null if not s_val else escape_latex(s_val), text=c_dict["e"], right=null if not e_val else escape_latex(e_val), ) return delim def do_spre(self, elm): """ the Pre-Sub-Superscript object -- Not support yet """ def do_sub(self, elm): text = self.process_children(elm) return SUB.format(text) def do_sup(self, elm): text = self.process_children(elm) return SUP.format(text) def do_f(self, elm): """ the fraction object """ c_dict = self.process_children_dict(elm) pr = c_dict.get("fPr") if pr is None: # Handle missing fPr element gracefully _log.debug("Missing fPr element in fraction, using default formatting") latex_s = F_DEFAULT return latex_s.format( num=c_dict.get("num"), den=c_dict.get("den"), ) latex_s = get_val(pr.type, default=F_DEFAULT, store=F) return pr.text + latex_s.format(num=c_dict.get("num"), den=c_dict.get("den")) def do_func(self, elm): """ the Function-Apply object (Examples:sin cos) """ c_dict = self.process_children_dict(elm) func_name = c_dict.get("fName") return func_name.replace(FUNC_PLACE, c_dict.get("e")) def do_fname(self, elm): """ the func name """ latex_chars = [] for stag, t, e in self.process_children_list(elm): if stag == "r": if FUNC.get(t): latex_chars.append(FUNC[t]) else: _log.warning("Function not supported, will default to text: %s", t) if isinstance(t, str): latex_chars.append(t) elif isinstance(t, str): latex_chars.append(t) t = BLANK.join(latex_chars) return t if FUNC_PLACE in t else t + FUNC_PLACE # do_func will replace this def do_groupchr(self, elm): """ the Group-Character object """ c_dict = self.process_children_dict(elm) pr = c_dict["groupChrPr"] latex_s = get_val(pr.chr) return pr.text + latex_s.format(c_dict["e"]) def do_rad(self, elm): """ the radical object """ c_dict = self.process_children_dict(elm) text = c_dict.get("e") deg_text = c_dict.get("deg") if deg_text: return RAD.format(deg=deg_text, text=text) else: return RAD_DEFAULT.format(text=text) def do_eqarr(self, elm): """ the Array object """ return ARR.format( text=BRK.join( [t for stag, t, e in self.process_children_list(elm, include=("e",))] ) ) def do_limlow(self, elm): """ the Lower-Limit object """ t_dict = self.process_children_dict(elm, include=("e", "lim")) latex_s = LIM_FUNC.get(t_dict["e"]) if not latex_s: raise RuntimeError("Not support lim {}".format(t_dict["e"])) else: return latex_s.format(lim=t_dict.get("lim")) def do_limupp(self, elm): """ the Upper-Limit object """ t_dict = self.process_children_dict(elm, include=("e", "lim")) return LIM_UPP.format(lim=t_dict.get("lim"), text=t_dict.get("e")) def do_lim(self, elm): """ the lower limit of the limLow object and the upper limit of the limUpp function """ return self.process_children(elm).replace(LIM_TO[0], LIM_TO[1]) def do_m(self, elm): """ the Matrix object """ rows = [] for stag, t, e in self.process_children_list(elm): if stag == "mPr": pass elif stag == "mr": rows.append(t) return M.format(text=BRK.join(rows)) def do_mr(self, elm): """ a single row of the matrix m """ return ALN.join( [t for stag, t, e in self.process_children_list(elm, include=("e",))] ) def do_nary(self, elm): """ the n-ary object """ res = [] bo = "" for stag, t, e in self.process_children_list(elm): if stag == "naryPr": # if <m:naryPr> contains no <m:chr>, the n-ary represents an integral bo = get_val(t.chr, default="\\int", store=CHR_BO) else: res.append(t) return bo + BLANK.join(res) def process_unicode(self, s): # s = s if isinstance(s,unicode) else unicode(s,'utf-8') # print(s, self._t_dict.get(s, s), unicode_to_latex(s)) # _str.append( self._t_dict.get(s, s) ) out_latex_str = self.u.unicode_to_latex(s) if ( s.startswith("{") is False and out_latex_str.startswith("{") and s.endswith("}") is False and out_latex_str.endswith("}") ): out_latex_str = f" {out_latex_str[1:-1]} " if "ensuremath" in out_latex_str: out_latex_str = out_latex_str.replace("\\ensuremath{", " ") out_latex_str = out_latex_str.replace("}", " ") if out_latex_str.strip().startswith("\\text"): out_latex_str = f" \\text{{{out_latex_str}}} " return out_latex_str def do_r(self, elm): """ Get text from 'r' element,And try convert them to latex symbols @todo text style support , (sty) @todo \text (latex pure text support) """ _str = [] _base_str = [] found_text = elm.findtext(f"./{OMML_NS}t") if found_text: for s in found_text: out_latex_str = self.process_unicode(s) _str.append(out_latex_str) _base_str.append(s) proc_str = escape_latex(BLANK.join(_str)) base_proc_str = BLANK.join(_base_str) if "{" not in base_proc_str and "\\{" in proc_str: proc_str = proc_str.replace("\\{", "{") if "}" not in base_proc_str and "\\}" in proc_str: proc_str = proc_str.replace("\\}", "}") return proc_str tag2meth = { "acc": do_acc, "r": do_r, "bar": do_bar, "sub": do_sub, "sup": do_sup, "f": do_f, "func": do_func, "fName": do_fname, "groupChr": do_groupchr, "d": do_d, "rad": do_rad, "eqArr": do_eqarr, "limLow": do_limlow, "limUpp": do_limupp, "lim": do_lim, "m": do_m, "mr": do_mr, "nary": do_nary, }
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/docx/latex/latex_dict.py
docling/backend/docx/latex/latex_dict.py
""" Adapted from https://github.com/xiilei/dwml/blob/master/dwml/latex_dict.py On 23/01/2025 """ CHARS = ("{", "}", "_", "^", "#", "&", "$", "%", "~") BLANK = "" BACKSLASH = "\\" ALN = "&" CHR = { # Unicode : Latex Math Symbols # Top accents "\u0300": "\\grave{{{0}}}", "\u0301": "\\acute{{{0}}}", "\u0302": "\\hat{{{0}}}", "\u0303": "\\tilde{{{0}}}", "\u0304": "\\bar{{{0}}}", "\u0305": "\\overbar{{{0}}}", "\u0306": "\\breve{{{0}}}", "\u0307": "\\dot{{{0}}}", "\u0308": "\\ddot{{{0}}}", "\u0309": "\\ovhook{{{0}}}", "\u030a": "\\ocirc{{{0}}}}", "\u030c": "\\check{{{0}}}}", "\u0310": "\\candra{{{0}}}", "\u0312": "\\oturnedcomma{{{0}}}", "\u0315": "\\ocommatopright{{{0}}}", "\u031a": "\\droang{{{0}}}", "\u0338": "\\not{{{0}}}", "\u20d0": "\\leftharpoonaccent{{{0}}}", "\u20d1": "\\rightharpoonaccent{{{0}}}", "\u20d2": "\\vertoverlay{{{0}}}", "\u20d6": "\\overleftarrow{{{0}}}", "\u20d7": "\\vec{{{0}}}", "\u20db": "\\dddot{{{0}}}", "\u20dc": "\\ddddot{{{0}}}", "\u20e1": "\\overleftrightarrow{{{0}}}", "\u20e7": "\\annuity{{{0}}}", "\u20e9": "\\widebridgeabove{{{0}}}", "\u20f0": "\\asteraccent{{{0}}}", # Bottom accents "\u0330": "\\wideutilde{{{0}}}", "\u0331": "\\underbar{{{0}}}", "\u20e8": "\\threeunderdot{{{0}}}", "\u20ec": "\\underrightharpoondown{{{0}}}", "\u20ed": "\\underleftharpoondown{{{0}}}", "\u20ee": "\\underledtarrow{{{0}}}", "\u20ef": "\\underrightarrow{{{0}}}", # Over | group "\u23b4": "\\overbracket{{{0}}}", "\u23dc": "\\overparen{{{0}}}", "\u23de": "\\overbrace{{{0}}}", # Under| group "\u23b5": "\\underbracket{{{0}}}", "\u23dd": "\\underparen{{{0}}}", "\u23df": "\\underbrace{{{0}}}", } CHR_BO = { # Big operators, "\u2140": "\\Bbbsum", "\u220f": "\\prod", "\u2210": "\\coprod", "\u2211": "\\sum", "\u222b": "\\int", "\u222c": "\\iint", "\u222d": "\\iiint", "\u222e": "\\oint", "\u222f": "\\oiint", "\u2230": "\\oiiint", "\u22c0": "\\bigwedge", "\u22c1": "\\bigvee", "\u22c2": "\\bigcap", "\u22c3": "\\bigcup", "\u2a00": "\\bigodot", "\u2a01": "\\bigoplus", "\u2a02": "\\bigotimes", } T = { # Greek letters "\U0001d6fc": "\\alpha ", "\U0001d6fd": "\\beta ", "\U0001d6fe": "\\gamma ", "\U0001d6ff": "\\theta ", "\U0001d700": "\\epsilon ", "\U0001d701": "\\zeta ", "\U0001d702": "\\eta ", "\U0001d703": "\\theta ", "\U0001d704": "\\iota ", "\U0001d705": "\\kappa ", "\U0001d706": "\\lambda ", "\U0001d707": "\\m ", "\U0001d708": "\\n ", "\U0001d709": "\\xi ", "\U0001d70a": "\\omicron ", "\U0001d70b": "\\pi ", "\U0001d70c": "\\rho ", "\U0001d70d": "\\varsigma ", "\U0001d70e": "\\sigma ", "\U0001d70f": "\\ta ", "\U0001d710": "\\upsilon ", "\U0001d711": "\\phi ", "\U0001d712": "\\chi ", "\U0001d713": "\\psi ", "\U0001d714": "\\omega ", "\U0001d715": "\\partial ", "\U0001d716": "\\varepsilon ", "\U0001d717": "\\vartheta ", "\U0001d718": "\\varkappa ", "\U0001d719": "\\varphi ", "\U0001d71a": "\\varrho ", "\U0001d71b": "\\varpi ", # Relation symbols "\u2190": "\\leftarrow ", "\u2191": "\\uparrow ", "\u2192": "\\rightarrow ", "\u2193": "\\downright ", "\u2194": "\\leftrightarrow ", "\u2195": "\\updownarrow ", "\u2196": "\\nwarrow ", "\u2197": "\\nearrow ", "\u2198": "\\searrow ", "\u2199": "\\swarrow ", "\u22ee": "\\vdots ", "\u22ef": "\\cdots ", "\u22f0": "\\adots ", "\u22f1": "\\ddots ", "\u2260": "\\ne ", "\u2264": "\\leq ", "\u2265": "\\geq ", "\u2266": "\\leqq ", "\u2267": "\\geqq ", "\u2268": "\\lneqq ", "\u2269": "\\gneqq ", "\u226a": "\\ll ", "\u226b": "\\gg ", "\u2208": "\\in ", "\u2209": "\\notin ", "\u220b": "\\ni ", "\u220c": "\\nni ", # Ordinary symbols "\u221e": "\\infty ", # Binary relations "\u00b1": "\\pm ", "\u2213": "\\mp ", # Italic, Latin, uppercase "\U0001d434": "A", "\U0001d435": "B", "\U0001d436": "C", "\U0001d437": "D", "\U0001d438": "E", "\U0001d439": "F", "\U0001d43a": "G", "\U0001d43b": "H", "\U0001d43c": "I", "\U0001d43d": "J", "\U0001d43e": "K", "\U0001d43f": "L", "\U0001d440": "M", "\U0001d441": "N", "\U0001d442": "O", "\U0001d443": "P", "\U0001d444": "Q", "\U0001d445": "R", "\U0001d446": "S", "\U0001d447": "T", "\U0001d448": "U", "\U0001d449": "V", "\U0001d44a": "W", "\U0001d44b": "X", "\U0001d44c": "Y", "\U0001d44d": "Z", # Italic, Latin, lowercase "\U0001d44e": "a", "\U0001d44f": "b", "\U0001d450": "c", "\U0001d451": "d", "\U0001d452": "e", "\U0001d453": "f", "\U0001d454": "g", "\U0001d456": "i", "\U0001d457": "j", "\U0001d458": "k", "\U0001d459": "l", "\U0001d45a": "m", "\U0001d45b": "n", "\U0001d45c": "o", "\U0001d45d": "p", "\U0001d45e": "q", "\U0001d45f": "r", "\U0001d460": "s", "\U0001d461": "t", "\U0001d462": "u", "\U0001d463": "v", "\U0001d464": "w", "\U0001d465": "x", "\U0001d466": "y", "\U0001d467": "z", } FUNC = { "sin": "\\sin({fe})", "cos": "\\cos({fe})", "tan": "\\tan({fe})", "arcsin": "\\arcsin({fe})", "arccos": "\\arccos({fe})", "arctan": "\\arctan({fe})", "arccot": "\\arccot({fe})", "sinh": "\\sinh({fe})", "cosh": "\\cosh({fe})", "tanh": "\\tanh({fe})", "coth": "\\coth({fe})", "sec": "\\sec({fe})", "csc": "\\csc({fe})", "mod": "\\mod {fe}", "max": "\\max({fe})", "min": "\\min({fe})", } FUNC_PLACE = "{fe}" BRK = "\\\\" CHR_DEFAULT = { "ACC_VAL": "\\hat{{{0}}}", } POS = { "top": "\\overline{{{0}}}", # not sure "bot": "\\underline{{{0}}}", } POS_DEFAULT = { "BAR_VAL": "\\overline{{{0}}}", } SUB = "_{{{0}}}" SUP = "^{{{0}}}" F = { "bar": "\\frac{{{num}}}{{{den}}}", "skw": r"^{{{num}}}/_{{{den}}}", "noBar": "\\genfrac{{}}{{}}{{0pt}}{{}}{{{num}}}{{{den}}}", "lin": "{{{num}}}/{{{den}}}", } F_DEFAULT = "\\frac{{{num}}}{{{den}}}" D = "\\left{left}{text}\\right{right}" D_DEFAULT = { "left": "(", "right": ")", "null": ".", } RAD = "\\sqrt[{deg}]{{{text}}}" RAD_DEFAULT = "\\sqrt{{{text}}}" ARR = "{text}" LIM_FUNC = { "lim": "\\lim_{{{lim}}}", "max": "\\max_{{{lim}}}", "min": "\\min_{{{lim}}}", } LIM_TO = ("\\rightarrow", "\\to") LIM_UPP = "\\overset{{{lim}}}{{{text}}}" M = "\\begin{{matrix}}{text}\\end{{matrix}}"
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/docx/latex/__init__.py
docling/backend/docx/latex/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/backend/docx/drawingml/utils.py
docling/backend/docx/drawingml/utils.py
import os import shutil import subprocess from pathlib import Path from tempfile import mkdtemp from typing import Callable, Optional import pypdfium2 from docx.document import Document from PIL import Image, ImageChops def get_libreoffice_cmd(raise_if_unavailable: bool = False) -> Optional[str]: """Return the libreoffice cmd and optionally test it.""" libreoffice_cmd = ( shutil.which("libreoffice") or shutil.which("soffice") or ( "/Applications/LibreOffice.app/Contents/MacOS/soffice" if os.path.isfile("/Applications/LibreOffice.app/Contents/MacOS/soffice") else None ) ) if raise_if_unavailable: if libreoffice_cmd is None: raise RuntimeError("Libreoffice not found") # The following test will raise if the libreoffice_cmd cannot be used subprocess.run( [ libreoffice_cmd, "-h", ], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL, check=True, ) return libreoffice_cmd def get_docx_to_pdf_converter() -> Optional[Callable]: """ Detects the best available DOCX to PDF tool and returns a conversion function. The returned function accepts (input_path, output_path). Returns None if no tool is available. """ # Try LibreOffice libreoffice_cmd = get_libreoffice_cmd() if libreoffice_cmd: def convert_with_libreoffice(input_path, output_path): subprocess.run( [ libreoffice_cmd, "--headless", "--convert-to", "pdf", "--outdir", os.path.dirname(output_path), input_path, ], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL, check=True, ) expected_output = os.path.join( os.path.dirname(output_path), os.path.splitext(os.path.basename(input_path))[0] + ".pdf", ) if expected_output != output_path: os.rename(expected_output, output_path) return convert_with_libreoffice ## Space for other DOCX to PDF converters if available # No tools found return None def crop_whitespace(image: Image.Image, bg_color=None, padding=0) -> Image.Image: if bg_color is None: bg_color = image.getpixel((0, 0)) bg = Image.new(image.mode, image.size, bg_color) diff = ImageChops.difference(image, bg) bbox = diff.getbbox() if bbox: left, upper, right, lower = bbox left = max(0, left - padding) upper = max(0, upper - padding) right = min(image.width, right + padding) lower = min(image.height, lower + padding) return image.crop((left, upper, right, lower)) else: return image def get_pil_from_dml_docx( docx: Document, converter: Optional[Callable] ) -> Optional[Image.Image]: if converter is None: return None temp_dir = Path(mkdtemp()) temp_docx = Path(temp_dir / "drawing_only.docx") temp_pdf = Path(temp_dir / "drawing_only.pdf") # 1) Save docx temporarily docx.save(str(temp_docx)) # 2) Export to PDF converter(temp_docx, temp_pdf) # 3) Load PDF as PNG pdf = pypdfium2.PdfDocument(temp_pdf) page = pdf[0] image = crop_whitespace(page.render(scale=2).to_pil()) page.close() pdf.close() shutil.rmtree(temp_dir, ignore_errors=True) return image
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/experimental/__init__.py
docling/experimental/__init__.py
"""Experimental modules for Docling. This package contains experimental features that are under development and may change or be removed in future versions. """
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/experimental/datamodel/table_crops_layout_options.py
docling/experimental/datamodel/table_crops_layout_options.py
"""Internal options for the experimental TableCrops layout model.""" from typing import ClassVar from docling.datamodel.pipeline_options import BaseLayoutOptions __all__ = ["TableCropsLayoutOptions"] class TableCropsLayoutOptions(BaseLayoutOptions): """Options for TableCropsLayoutModel (internal-only).""" kind: ClassVar[str] = "docling_experimental_table_crops_layout"
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/experimental/datamodel/__init__.py
docling/experimental/datamodel/__init__.py
"""Experimental datamodel modules."""
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/experimental/datamodel/threaded_layout_vlm_pipeline_options.py
docling/experimental/datamodel/threaded_layout_vlm_pipeline_options.py
"""Options for the threaded layout+VLM pipeline.""" from typing import Union from pydantic import model_validator from docling.datamodel.layout_model_specs import DOCLING_LAYOUT_HERON from docling.datamodel.pipeline_options import LayoutOptions, PaginatedPipelineOptions from docling.datamodel.pipeline_options_vlm_model import ( ApiVlmOptions, InlineVlmOptions, ResponseFormat, ) from docling.datamodel.vlm_model_specs import GRANITEDOCLING_TRANSFORMERS class ThreadedLayoutVlmPipelineOptions(PaginatedPipelineOptions): """Pipeline options for the threaded layout+VLM pipeline.""" images_scale: float = 2.0 # VLM configuration (will be enhanced with layout awareness by the pipeline) vlm_options: Union[InlineVlmOptions, ApiVlmOptions] = GRANITEDOCLING_TRANSFORMERS # Layout model configuration layout_options: LayoutOptions = LayoutOptions( model_spec=DOCLING_LAYOUT_HERON, skip_cell_assignment=True ) # Threading and batching controls layout_batch_size: int = 4 vlm_batch_size: int = 4 batch_timeout_seconds: float = 2.0 queue_max_size: int = 50 @model_validator(mode="after") def validate_response_format(self): """Validate that VLM response format is DOCTAGS (required for this pipeline).""" if self.vlm_options.response_format != ResponseFormat.DOCTAGS: raise ValueError( f"ThreadedLayoutVlmPipeline only supports DOCTAGS response format, " f"but got {self.vlm_options.response_format}. " f"Please set vlm_options.response_format=ResponseFormat.DOCTAGS" ) return self
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/experimental/models/__init__.py
docling/experimental/models/__init__.py
"""Experimental models for Docling.""" __all__: list[str] = []
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/experimental/models/table_crops_layout_model.py
docling/experimental/models/table_crops_layout_model.py
"""Internal TableCrops layout model that marks full pages as table clusters.""" from __future__ import annotations import warnings from collections.abc import Sequence from pathlib import Path from typing import Optional import numpy as np from docling_core.types.doc import DocItemLabel from docling.datamodel.accelerator_options import AcceleratorOptions from docling.datamodel.base_models import BoundingBox, Cluster, LayoutPrediction, Page from docling.datamodel.document import ConversionResult from docling.experimental.datamodel.table_crops_layout_options import ( TableCropsLayoutOptions, ) from docling.models.base_layout_model import BaseLayoutModel __all__ = ["TableCropsLayoutModel"] class TableCropsLayoutModel(BaseLayoutModel): """Experimental layout model that treats the full page as a table cluster. This is useful in cases where a Docling pipeline is applied to images of table crops only. This model is internal and not part of the stable public interface. """ def __init__( self, artifacts_path: Optional[Path], accelerator_options: AcceleratorOptions, options: TableCropsLayoutOptions, ): self.options = options self.artifacts_path = artifacts_path self.accelerator_options = accelerator_options @classmethod def get_options_type(cls) -> type[TableCropsLayoutOptions]: return TableCropsLayoutOptions def predict_layout( self, conv_res: ConversionResult, pages: Sequence[Page], ) -> Sequence[LayoutPrediction]: layout_predictions: list[LayoutPrediction] = [] for page in pages: if page._backend is None or not page._backend.is_valid(): existing_prediction = page.predictions.layout or LayoutPrediction() layout_predictions.append(existing_prediction) continue clusters = self._build_page_clusters(page) prediction = LayoutPrediction(clusters=clusters) self._update_confidence(conv_res, page, clusters) layout_predictions.append(prediction) return layout_predictions def _build_page_clusters(self, page: Page) -> list[Cluster]: page_size = page.size if page_size is None: return [] bbox = BoundingBox( l=0.0, t=0.0, r=page_size.width, b=page_size.height, ) cluster = Cluster( id=0, label=DocItemLabel.TABLE, bbox=bbox, confidence=1.0, cells=[], ) clusters = [cluster] if not self.options.skip_cell_assignment: page_cells = list(page.cells) cluster.cells = page_cells if not page_cells and not self.options.keep_empty_clusters: clusters = [] return clusters def _update_confidence( self, conv_res: ConversionResult, page: Page, clusters: list[Cluster] ) -> None: """Populate layout and OCR confidence scores for the page.""" with warnings.catch_warnings(): warnings.filterwarnings( "ignore", "Mean of empty slice|invalid value encountered in scalar divide", RuntimeWarning, "numpy", ) conv_res.confidence.pages[page.page_no].layout_score = 1.0 ocr_cells = [cell for cell in page.cells if cell.from_ocr] ocr_confidence = float(np.mean([cell.confidence for cell in ocr_cells])) conv_res.confidence.pages[page.page_no].ocr_score = ocr_confidence
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/experimental/pipeline/threaded_layout_vlm_pipeline.py
docling/experimental/pipeline/threaded_layout_vlm_pipeline.py
"""Threaded Layout+VLM Pipeline ================================ A specialized two-stage threaded pipeline that combines layout model preprocessing with VLM processing. The layout model detects document elements and coordinates, which are then injected into the VLM prompt for enhanced structured output. """ from __future__ import annotations import itertools import logging from pathlib import Path from typing import TYPE_CHECKING, List, Optional, Union, cast from docling_core.types.doc import DoclingDocument from docling_core.types.doc.document import DocTagsDocument from PIL import Image as PILImage if TYPE_CHECKING: from docling_core.types.doc.page import SegmentedPage from docling.backend.abstract_backend import AbstractDocumentBackend from docling.backend.pdf_backend import PdfDocumentBackend from docling.datamodel.base_models import ConversionStatus, Page from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options_vlm_model import ( ApiVlmOptions, InferenceFramework, InlineVlmOptions, ) from docling.datamodel.settings import settings from docling.experimental.datamodel.threaded_layout_vlm_pipeline_options import ( ThreadedLayoutVlmPipelineOptions, ) from docling.models.api_vlm_model import ApiVlmModel from docling.models.base_model import BaseVlmPageModel from docling.models.layout_model import LayoutModel from docling.models.vlm_models_inline.hf_transformers_model import ( HuggingFaceTransformersVlmModel, ) from docling.models.vlm_models_inline.mlx_model import HuggingFaceMlxModel from docling.pipeline.base_pipeline import BasePipeline from docling.pipeline.standard_pdf_pipeline import ( ProcessingResult, RunContext, ThreadedItem, ThreadedPipelineStage, ThreadedQueue, ) from docling.utils.profiling import ProfilingScope, TimeRecorder _log = logging.getLogger(__name__) class ThreadedLayoutVlmPipeline(BasePipeline): """Two-stage threaded pipeline: Layout Model → VLM Model.""" def __init__(self, pipeline_options: ThreadedLayoutVlmPipelineOptions) -> None: super().__init__(pipeline_options) self.pipeline_options: ThreadedLayoutVlmPipelineOptions = pipeline_options self._run_seq = itertools.count(1) # deterministic, monotonic run ids # VLM model type (initialized in _init_models) self.vlm_model: BaseVlmPageModel # Initialize models self._init_models() def _init_models(self) -> None: """Initialize layout and VLM models.""" art_path = self._resolve_artifacts_path() # Layout model self.layout_model = LayoutModel( artifacts_path=art_path, accelerator_options=self.pipeline_options.accelerator_options, options=self.pipeline_options.layout_options, ) # VLM model based on options type # Create layout-aware VLM options internally base_vlm_options = self.pipeline_options.vlm_options class LayoutAwareVlmOptions(type(base_vlm_options)): # type: ignore[misc] def build_prompt( self, page: Optional[SegmentedPage], *, _internal_page: Optional[Page] = None, ) -> str: base_prompt = self.prompt augmented_prompt = base_prompt # In this layout-aware pipeline, _internal_page is always provided if _internal_page is None: return base_prompt if not _internal_page.size: _log.warning( f"Page size not available for page {_internal_page.page_no}. Cannot enhance prompt with layout info." ) return base_prompt if _internal_page.predictions.layout: from docling_core.types.doc.tokens import DocumentToken layout_elements = [] for cluster in _internal_page.predictions.layout.clusters: # Get proper tag name from DocItemLabel tag_name = DocumentToken.create_token_name_from_doc_item_label( label=cluster.label ) # Convert bbox to tuple and get location tokens bbox_tuple = cluster.bbox.as_tuple() location_tokens = DocumentToken.get_location( bbox=bbox_tuple, page_w=_internal_page.size.width, page_h=_internal_page.size.height, ) # Create XML element with DocTags format xml_element = f"<{tag_name}>{location_tokens}</{tag_name}>" layout_elements.append(xml_element) if layout_elements: # Join elements with newlines and wrap in layout tags layout_xml = ( "<layout>" + "\n".join(layout_elements) + "</layout>" ) layout_injection = f"{layout_xml}" augmented_prompt = base_prompt + layout_injection _log.debug( "Enhanced Prompt with Layout Info: %s\n", augmented_prompt ) return augmented_prompt vlm_options = LayoutAwareVlmOptions(**base_vlm_options.model_dump()) if isinstance(base_vlm_options, ApiVlmOptions): self.vlm_model = ApiVlmModel( enabled=True, enable_remote_services=self.pipeline_options.enable_remote_services, vlm_options=vlm_options, ) elif isinstance(base_vlm_options, InlineVlmOptions): if vlm_options.inference_framework == InferenceFramework.TRANSFORMERS: self.vlm_model = HuggingFaceTransformersVlmModel( enabled=True, artifacts_path=art_path, accelerator_options=self.pipeline_options.accelerator_options, vlm_options=vlm_options, ) elif vlm_options.inference_framework == InferenceFramework.MLX: self.vlm_model = HuggingFaceMlxModel( enabled=True, artifacts_path=art_path, accelerator_options=self.pipeline_options.accelerator_options, vlm_options=vlm_options, ) elif vlm_options.inference_framework == InferenceFramework.VLLM: from docling.models.vlm_models_inline.vllm_model import VllmVlmModel self.vlm_model = VllmVlmModel( enabled=True, artifacts_path=art_path, accelerator_options=self.pipeline_options.accelerator_options, vlm_options=vlm_options, ) else: raise ValueError( f"Unsupported VLM inference framework: {vlm_options.inference_framework}" ) else: raise ValueError(f"Unsupported VLM options type: {type(base_vlm_options)}") def _resolve_artifacts_path(self) -> Optional[Path]: """Resolve artifacts path from options or settings.""" 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 _create_run_ctx(self) -> RunContext: """Create pipeline stages and wire them together.""" opts = self.pipeline_options # Layout stage layout_stage = ThreadedPipelineStage( name="layout", model=self.layout_model, batch_size=opts.layout_batch_size, batch_timeout=opts.batch_timeout_seconds, queue_max_size=opts.queue_max_size, ) # VLM stage - now layout-aware through enhanced build_prompt vlm_stage = ThreadedPipelineStage( name="vlm", model=self.vlm_model, batch_size=opts.vlm_batch_size, batch_timeout=opts.batch_timeout_seconds, queue_max_size=opts.queue_max_size, ) # Wire stages output_q = ThreadedQueue(opts.queue_max_size) layout_stage.add_output_queue(vlm_stage.input_queue) vlm_stage.add_output_queue(output_q) stages = [layout_stage, vlm_stage] return RunContext( stages=stages, first_stage=layout_stage, output_queue=output_q ) def _build_document(self, conv_res: ConversionResult) -> ConversionResult: """Build document using threaded layout+VLM pipeline.""" run_id = next(self._run_seq) assert isinstance(conv_res.input._backend, PdfDocumentBackend) backend = conv_res.input._backend # Initialize pages start_page, end_page = conv_res.input.limits.page_range pages: List[Page] = [] images_scale = self.pipeline_options.images_scale for i in range(conv_res.input.page_count): if start_page - 1 <= i <= end_page - 1: page = Page(page_no=i) if images_scale is not None: page._default_image_scale = images_scale page._backend = backend.load_page(i) if page._backend and page._backend.is_valid(): page.size = page._backend.get_size() conv_res.pages.append(page) pages.append(page) if not pages: conv_res.status = ConversionStatus.FAILURE return conv_res total_pages = len(pages) ctx = self._create_run_ctx() for st in ctx.stages: st.start() proc = ProcessingResult(total_expected=total_pages) fed_idx = 0 batch_size = 32 try: while proc.success_count + proc.failure_count < total_pages: # Feed pages to first stage while fed_idx < total_pages: ok = ctx.first_stage.input_queue.put( ThreadedItem( payload=pages[fed_idx], run_id=run_id, page_no=pages[fed_idx].page_no, conv_res=conv_res, ), timeout=0.0, ) if ok: fed_idx += 1 if fed_idx == total_pages: ctx.first_stage.input_queue.close() else: break # Drain results from output out_batch = ctx.output_queue.get_batch(batch_size, timeout=0.05) for itm in out_batch: if itm.run_id != run_id: continue if itm.is_failed or itm.error: proc.failed_pages.append( (itm.page_no, itm.error or RuntimeError("unknown error")) ) else: assert itm.payload is not None proc.pages.append(itm.payload) # Handle early termination if not out_batch and ctx.output_queue.closed: missing = total_pages - (proc.success_count + proc.failure_count) if missing > 0: proc.failed_pages.extend( [(-1, RuntimeError("pipeline terminated early"))] * missing ) break finally: for st in ctx.stages: st.stop() ctx.output_queue.close() self._integrate_results(conv_res, proc) return conv_res def _integrate_results( self, conv_res: ConversionResult, proc: ProcessingResult ) -> None: """Integrate processing results into conversion result.""" page_map = {p.page_no: p for p in proc.pages} # Track failed pages for cleanup failed_page_nos = {fp for fp, _ in proc.failed_pages} # Collect pages that will be removed (failed pages) for resource cleanup pages_to_remove = [p for p in conv_res.pages if p.page_no in failed_page_nos] conv_res.pages = [ page_map.get(p.page_no, p) for p in conv_res.pages if p.page_no in page_map or not any(fp == p.page_no for fp, _ in proc.failed_pages) ] if proc.is_complete_failure: conv_res.status = ConversionStatus.FAILURE elif proc.is_partial_success: conv_res.status = ConversionStatus.PARTIAL_SUCCESS else: conv_res.status = ConversionStatus.SUCCESS # Clean up resources for failed pages that were removed for p in pages_to_remove: if p._backend is not None: p._backend.unload() p._image_cache = {} # Clean up parsed_page if it exists (it's Optional[SegmentedPdfPage]) if p.parsed_page is not None: del p.parsed_page p.parsed_page = None # Clean up images if not needed for remaining pages if not self.pipeline_options.generate_page_images: for p in conv_res.pages: p._image_cache = {} def _assemble_document(self, conv_res: ConversionResult) -> ConversionResult: """Assemble final document from VLM predictions.""" from docling_core.types.doc import DocItem, ImageRef, PictureItem from docling.datamodel.pipeline_options_vlm_model import ResponseFormat with TimeRecorder(conv_res, "doc_assemble", scope=ProfilingScope.DOCUMENT): # Response format validation is done in ThreadedLayoutVlmPipelineOptions # This check is kept as a safety net, but should never trigger if validation works if ( self.pipeline_options.vlm_options.response_format != ResponseFormat.DOCTAGS ): raise RuntimeError( f"Unsupported VLM response format {self.pipeline_options.vlm_options.response_format}. Only DOCTAGS format is supported." ) conv_res.document = self._turn_dt_into_doc(conv_res) # Generate images of the requested element types if self.pipeline_options.generate_picture_images: # Create mapping from page_no to Page object since pages may be non-continuous page_map = {p.page_no: p for p in conv_res.pages} 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 ): page_no = element.prov[0].page_no page = page_map.get(page_no) if page is None: _log.warning( f"Page {page_no} not found in conversion result for picture element. Skipping image generation." ) continue 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) ) return conv_res def _turn_dt_into_doc(self, conv_res: ConversionResult) -> DoclingDocument: """Convert DOCTAGS response format to DoclingDocument.""" doctags_list = [] image_list = [] for page in conv_res.pages: # Only include pages that have both an image and VLM predictions if page.image and page.predictions.vlm_response: predicted_doctags = page.predictions.vlm_response.text image_list.append(page.image) doctags_list.append(predicted_doctags) doctags_list_c = cast(List[Union[Path, str]], doctags_list) image_list_c = cast(List[Union[Path, PILImage.Image]], image_list) doctags_doc = DocTagsDocument.from_doctags_and_image_pairs( doctags_list_c, image_list_c ) document = DoclingDocument.load_from_doctags(doctag_document=doctags_doc) return document @classmethod def get_default_options(cls) -> ThreadedLayoutVlmPipelineOptions: return ThreadedLayoutVlmPipelineOptions() @classmethod def is_backend_supported(cls, backend: AbstractDocumentBackend) -> bool: return isinstance(backend, PdfDocumentBackend) def _determine_status(self, conv_res: ConversionResult) -> ConversionStatus: return conv_res.status def _unload(self, conv_res: ConversionResult) -> None: for p in conv_res.pages: if p._backend is not None: p._backend.unload() if conv_res.input._backend: conv_res.input._backend.unload()
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/experimental/pipeline/__init__.py
docling/experimental/pipeline/__init__.py
"""Experimental pipeline modules."""
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/cli/models.py
docling/cli/models.py
import logging import warnings from enum import Enum from pathlib import Path from typing import Annotated, Optional import typer from rich.console import Console from rich.logging import RichHandler from docling.datamodel.settings import settings from docling.models.utils.hf_model_download import download_hf_model from docling.utils.model_downloader import download_models warnings.filterwarnings(action="ignore", category=UserWarning, module="pydantic|torch") warnings.filterwarnings(action="ignore", category=FutureWarning, module="easyocr") console = Console() err_console = Console(stderr=True) app = typer.Typer( name="Docling models helper", no_args_is_help=True, add_completion=False, pretty_exceptions_enable=False, ) class _AvailableModels(str, Enum): LAYOUT = "layout" TABLEFORMER = "tableformer" CODE_FORMULA = "code_formula" PICTURE_CLASSIFIER = "picture_classifier" SMOLVLM = "smolvlm" GRANITEDOCLING = "granitedocling" GRANITEDOCLING_MLX = "granitedocling_mlx" SMOLDOCLING = "smoldocling" SMOLDOCLING_MLX = "smoldocling_mlx" GRANITE_VISION = "granite_vision" RAPIDOCR = "rapidocr" EASYOCR = "easyocr" _default_models = [ _AvailableModels.LAYOUT, _AvailableModels.TABLEFORMER, _AvailableModels.CODE_FORMULA, _AvailableModels.PICTURE_CLASSIFIER, _AvailableModels.RAPIDOCR, ] @app.command("download") def download( output_dir: Annotated[ Path, typer.Option( ..., "-o", "--output-dir", help="The directory where to download the models.", ), ] = (settings.cache_dir / "models"), force: Annotated[ bool, typer.Option(..., help="If true, the download will be forced.") ] = False, models: Annotated[ Optional[list[_AvailableModels]], typer.Argument( help="Models to download (default behavior: a predefined set of models will be downloaded).", ), ] = None, all: Annotated[ bool, typer.Option( ..., "--all", help="If true, all available models will be downloaded (mutually exclusive with passing specific models).", show_default=True, ), ] = False, quiet: Annotated[ bool, typer.Option( ..., "-q", "--quiet", help="No extra output is generated, the CLI prints only the directory with the cached models.", ), ] = False, ): if models and all: raise typer.BadParameter( "Cannot simultaneously set 'all' parameter and specify models to download." ) if not quiet: logging.basicConfig( level=logging.INFO, format="[blue]%(message)s[/blue]", datefmt="[%X]", handlers=[RichHandler(show_level=False, show_time=False, markup=True)], ) to_download = models or (list(_AvailableModels) if all else _default_models) output_dir = download_models( output_dir=output_dir, force=force, progress=(not quiet), with_layout=_AvailableModels.LAYOUT in to_download, with_tableformer=_AvailableModels.TABLEFORMER in to_download, with_code_formula=_AvailableModels.CODE_FORMULA in to_download, with_picture_classifier=_AvailableModels.PICTURE_CLASSIFIER in to_download, with_smolvlm=_AvailableModels.SMOLVLM in to_download, with_granitedocling=_AvailableModels.GRANITEDOCLING in to_download, with_granitedocling_mlx=_AvailableModels.GRANITEDOCLING_MLX in to_download, with_smoldocling=_AvailableModels.SMOLDOCLING in to_download, with_smoldocling_mlx=_AvailableModels.SMOLDOCLING_MLX in to_download, with_granite_vision=_AvailableModels.GRANITE_VISION in to_download, with_rapidocr=_AvailableModels.RAPIDOCR in to_download, with_easyocr=_AvailableModels.EASYOCR in to_download, ) if quiet: typer.echo(output_dir) else: typer.secho(f"\nModels downloaded into: {output_dir}.", fg="green") console.print( "\n", "Docling can now be configured for running offline using the local artifacts.\n\n", "Using the CLI:", f"`docling --artifacts-path={output_dir} FILE`", "\n", "Using Python: see the documentation at <https://docling-project.github.io/docling/usage>.", ) @app.command("download-hf-repo") def download_hf_repo( models: Annotated[ list[str], typer.Argument( help="Specific models to download from HuggingFace identified by their repo id. For example: docling-project/docling-models .", ), ], output_dir: Annotated[ Path, typer.Option( ..., "-o", "--output-dir", help="The directory where to download the models.", ), ] = (settings.cache_dir / "models"), force: Annotated[ bool, typer.Option(..., help="If true, the download will be forced.") ] = False, quiet: Annotated[ bool, typer.Option( ..., "-q", "--quiet", help="No extra output is generated, the CLI prints only the directory with the cached models.", ), ] = False, ): if not quiet: logging.basicConfig( level=logging.INFO, format="[blue]%(message)s[/blue]", datefmt="[%X]", handlers=[RichHandler(show_level=False, show_time=False, markup=True)], ) for item in models: typer.secho(f"\nDownloading {item} model from HuggingFace...") download_hf_model( repo_id=item, # would be better to reuse "repo_cache_folder" property: https://github.com/docling-project/docling/blob/main/docling/datamodel/pipeline_options_vlm_model.py#L76 # but creating options objects seams like an overkill local_dir=output_dir / item.replace("/", "--"), force=force, progress=(not quiet), ) if quiet: typer.echo(output_dir) else: typer.secho(f"\nModels downloaded into: {output_dir}.", fg="green") click_app = typer.main.get_command(app) if __name__ == "__main__": app()
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/cli/tools.py
docling/cli/tools.py
import typer from docling.cli.models import app as models_app app = typer.Typer( name="Docling helpers", no_args_is_help=True, add_completion=False, pretty_exceptions_enable=False, ) app.add_typer(models_app, name="models") click_app = typer.main.get_command(app) if __name__ == "__main__": app()
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/cli/main.py
docling/cli/main.py
import importlib import logging import platform import re import sys import tempfile import time import warnings from collections.abc import Iterable from pathlib import Path from typing import Annotated, Dict, List, Optional, Type import rich.table import typer from docling_core.transforms.serializer.html import ( HTMLDocSerializer, HTMLOutputStyle, HTMLParams, ) from docling_core.transforms.visualizer.layout_visualizer import LayoutVisualizer from docling_core.types.doc import ImageRefMode from docling_core.utils.file import resolve_source_to_path from pydantic import TypeAdapter from rich.console import Console from docling.backend.docling_parse_backend import DoclingParseDocumentBackend from docling.backend.docling_parse_v2_backend import DoclingParseV2DocumentBackend from docling.backend.docling_parse_v4_backend import DoclingParseV4DocumentBackend from docling.backend.image_backend import ImageDocumentBackend from docling.backend.mets_gbs_backend import MetsGbsDocumentBackend from docling.backend.pdf_backend import PdfDocumentBackend from docling.backend.pypdfium2_backend import PyPdfiumDocumentBackend from docling.datamodel.accelerator_options import AcceleratorDevice, AcceleratorOptions from docling.datamodel.asr_model_specs import ( WHISPER_BASE, WHISPER_BASE_MLX, WHISPER_BASE_NATIVE, WHISPER_LARGE, WHISPER_LARGE_MLX, WHISPER_LARGE_NATIVE, WHISPER_MEDIUM, WHISPER_MEDIUM_MLX, WHISPER_MEDIUM_NATIVE, WHISPER_SMALL, WHISPER_SMALL_MLX, WHISPER_SMALL_NATIVE, WHISPER_TINY, WHISPER_TINY_MLX, WHISPER_TINY_NATIVE, WHISPER_TURBO, WHISPER_TURBO_MLX, WHISPER_TURBO_NATIVE, AsrModelType, ) from docling.datamodel.backend_options import PdfBackendOptions from docling.datamodel.base_models import ( ConversionStatus, FormatToExtensions, InputFormat, OutputFormat, ) from docling.datamodel.document import ConversionResult, DoclingVersion from docling.datamodel.pipeline_options import ( AsrPipelineOptions, ConvertPipelineOptions, OcrAutoOptions, OcrOptions, PaginatedPipelineOptions, PdfBackend, PdfPipelineOptions, PipelineOptions, ProcessingPipeline, TableFormerMode, TableStructureOptions, TesseractCliOcrOptions, TesseractOcrOptions, VlmPipelineOptions, ) from docling.datamodel.settings import settings from docling.datamodel.vlm_model_specs import ( GOT2_TRANSFORMERS, GRANITE_VISION_OLLAMA, GRANITE_VISION_TRANSFORMERS, GRANITEDOCLING_MLX, GRANITEDOCLING_TRANSFORMERS, GRANITEDOCLING_VLLM, SMOLDOCLING_MLX, SMOLDOCLING_TRANSFORMERS, SMOLDOCLING_VLLM, VlmModelType, ) 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 from docling.pipeline.vlm_pipeline import VlmPipeline warnings.filterwarnings(action="ignore", category=UserWarning, module="pydantic|torch") warnings.filterwarnings(action="ignore", category=FutureWarning, module="easyocr") _log = logging.getLogger(__name__) console = Console() err_console = Console(stderr=True) ocr_factory_internal = get_ocr_factory(allow_external_plugins=False) ocr_engines_enum_internal = ocr_factory_internal.get_enum() DOCLING_ASCII_ART = r""" ████ ██████ ███░░██░░░░░██████ ████████░░░░░░░░████████████ ████████░░░░░░░░░░░░░░░░░░████████ ██████░░░░░░░░░░░░░░░░░░░░░░░░░░██████ ██████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░█████ ██████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░█████ ██████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░██████ ██████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░██████ ██████░░░░░░░ ░░░░░░░░░░░░░░░░░░░░░░ ░░░░░░░██████ ██████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░██████ ██████░░░░░░ ░░░░░░░░░░░░░░░ ░░░░░░██████ ███▒██░░░░░ ████ ░░░░░░░░░░░░ ████ ░░░░░██▒███ ███▒██░░░░░░ ████ ░░░░░░░░░░░░ ████ ░░░░░██▒████ ███▒██░░░░░░ ██ ██ ░░░░░░░░░░░░ ██ ██ ░░░░░██▒▒███ ███▒███░░░░░ ██ ░░░░████░░░░ ██ ░░░░░██▒▒███ ████▒▒██░░░░░░ ░░░███▒▒▒▒███░░░ ░░░░░░░██▒▒████ ████▒▒██░░░░░░░░░░░░░░░░░█▒▒▒▒▒▒▒▒▒▒█░░░░░░░░░░░░░░░░███▒▒████ ████▒▒▒██░░░░░░░░░░░░█████ ▒▒▒▒▒▒ ██████░░░░░░░░░░░██▒▒▒████ ███▒▒▒▒██░░░░░░░░███▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒███░░░░░░░░██▒▒▒▒███ ███▒▒▒▒▒███░░░░░░██▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒██░░░░░░███▒▒▒▒▒███ ████▒▒▒▒▒████░░░░░░██████████████████████░░░░░░████▒▒▒▒▒████ ███▒▒▒▒▒▒▒▒████░░░░░░░░░░░░░░░░░░░░░░░░░░░████▒▒▒▒▒▒▒▒▒███ ████▒▒▒▒▒▒▒▒███░░░░░████████████████████████▒▒▒▒▒▒▒▒▒████ ████▒▒▒▒▒▒██░░░░░░█ █░░░░░██▒▒▒▒▒▒████ ████▒▒▒▒█░░░░░░░█ D O C L I N G █░░░░░░░░██▒▒▒████ ████▒▒██░░░░░░█ █░░░░░░░░░░█▒▒████ ██████░░░░░░█ D O C L I N G █░░░░░░░░░░░██████ ████░░░░░█ █░░░░░░░░░░░░████ █████░░█ D O C L I N G █░░░░░░░░░░░█████ █████ █░░░░░░░░████████ ██ D O C L I N G █░░░░░░░░█████ █ █░░░████████ █████████████████████████████ """ app = typer.Typer( name="Docling", no_args_is_help=True, add_completion=False, pretty_exceptions_enable=False, ) def logo_callback(value: bool): if value: print(DOCLING_ASCII_ART) raise typer.Exit() def version_callback(value: bool): if value: v = DoclingVersion() print(f"Docling version: {v.docling_version}") print(f"Docling Core version: {v.docling_core_version}") print(f"Docling IBM Models version: {v.docling_ibm_models_version}") print(f"Docling Parse version: {v.docling_parse_version}") print(f"Python: {v.py_impl_version} ({v.py_lang_version})") print(f"Platform: {v.platform_str}") raise typer.Exit() def show_external_plugins_callback(value: bool): if value: ocr_factory_all = get_ocr_factory(allow_external_plugins=True) table = rich.table.Table(title="Available OCR engines") table.add_column("Name", justify="right") table.add_column("Plugin") table.add_column("Package") for meta in ocr_factory_all.registered_meta.values(): if not meta.module.startswith("docling."): table.add_row( f"[bold]{meta.kind}[/bold]", meta.plugin_name, meta.module.split(".")[0], ) rich.print(table) raise typer.Exit() def export_documents( conv_results: Iterable[ConversionResult], output_dir: Path, export_json: bool, export_yaml: bool, export_html: bool, export_html_split_page: bool, show_layout: bool, export_md: bool, export_txt: bool, export_doctags: bool, image_export_mode: ImageRefMode, ): success_count = 0 failure_count = 0 for conv_res in conv_results: if conv_res.status == ConversionStatus.SUCCESS: success_count += 1 doc_filename = conv_res.input.file.stem # Export JSON format: if export_json: fname = output_dir / f"{doc_filename}.json" _log.info(f"writing JSON output to {fname}") conv_res.document.save_as_json( filename=fname, image_mode=image_export_mode ) # Export YAML format: if export_yaml: fname = output_dir / f"{doc_filename}.yaml" _log.info(f"writing YAML output to {fname}") conv_res.document.save_as_yaml( filename=fname, image_mode=image_export_mode ) # Export HTML format: if export_html: fname = output_dir / f"{doc_filename}.html" _log.info(f"writing HTML output to {fname}") conv_res.document.save_as_html( filename=fname, image_mode=image_export_mode, split_page_view=False ) # Export HTML format: if export_html_split_page: fname = output_dir / f"{doc_filename}.html" _log.info(f"writing HTML output to {fname}") if show_layout: ser = HTMLDocSerializer( doc=conv_res.document, params=HTMLParams( image_mode=image_export_mode, output_style=HTMLOutputStyle.SPLIT_PAGE, ), ) visualizer = LayoutVisualizer() visualizer.params.show_label = False ser_res = ser.serialize( visualizer=visualizer, ) with open(fname, "w") as fw: fw.write(ser_res.text) else: conv_res.document.save_as_html( filename=fname, image_mode=image_export_mode, split_page_view=True, ) # Export Text format: if export_txt: fname = output_dir / f"{doc_filename}.txt" _log.info(f"writing TXT output to {fname}") conv_res.document.save_as_markdown( filename=fname, strict_text=True, image_mode=ImageRefMode.PLACEHOLDER, ) # Export Markdown format: if export_md: fname = output_dir / f"{doc_filename}.md" _log.info(f"writing Markdown output to {fname}") conv_res.document.save_as_markdown( filename=fname, image_mode=image_export_mode ) # Export Document Tags format: if export_doctags: fname = output_dir / f"{doc_filename}.doctags" _log.info(f"writing Doc Tags output to {fname}") conv_res.document.save_as_doctags(filename=fname) else: _log.warning(f"Document {conv_res.input.file} failed to convert.") if _log.isEnabledFor(logging.INFO): for err in conv_res.errors: _log.info( f" [Failure Detail] Component: {err.component_type}, " f"Module: {err.module_name}, Message: {err.error_message}" ) failure_count += 1 _log.info( f"Processed {success_count + failure_count} docs, of which {failure_count} failed" ) def _split_list(raw: Optional[str]) -> Optional[List[str]]: if raw is None: return None return re.split(r"[;,]", raw) @app.command(no_args_is_help=True) def convert( # noqa: C901 input_sources: Annotated[ List[str], typer.Argument( ..., metavar="source", help="PDF files to convert. Can be local file / directory paths or URL.", ), ], from_formats: List[InputFormat] = typer.Option( None, "--from", help="Specify input formats to convert from. Defaults to all formats.", ), to_formats: List[OutputFormat] = typer.Option( None, "--to", help="Specify output formats. Defaults to Markdown." ), show_layout: Annotated[ bool, typer.Option( ..., help="If enabled, the page images will show the bounding-boxes of the items.", ), ] = False, headers: str = typer.Option( None, "--headers", help="Specify http request headers used when fetching url input sources in the form of a JSON string", ), image_export_mode: Annotated[ ImageRefMode, typer.Option( ..., help="Image export mode for the document (only in case of JSON, Markdown or HTML). With `placeholder`, only the position of the image is marked in the output. In `embedded` mode, the image is embedded as base64 encoded string. In `referenced` mode, the image is exported in PNG format and referenced from the main exported document.", ), ] = ImageRefMode.EMBEDDED, pipeline: Annotated[ ProcessingPipeline, typer.Option(..., help="Choose the pipeline to process PDF or image files."), ] = ProcessingPipeline.STANDARD, vlm_model: Annotated[ VlmModelType, typer.Option(..., help="Choose the VLM model to use with PDF or image files."), ] = VlmModelType.GRANITEDOCLING, asr_model: Annotated[ AsrModelType, typer.Option(..., help="Choose the ASR model to use with audio/video files."), ] = AsrModelType.WHISPER_TINY, ocr: Annotated[ bool, typer.Option( ..., help="If enabled, the bitmap content will be processed using OCR." ), ] = True, force_ocr: Annotated[ bool, typer.Option( ..., help="Replace any existing text with OCR generated text over the full content.", ), ] = False, tables: Annotated[ bool, typer.Option( ..., help="If enabled, the table structure model will be used to extract table information.", ), ] = True, ocr_engine: Annotated[ str, typer.Option( ..., help=( f"The OCR engine to use. When --allow-external-plugins is *not* set, the available values are: " f"{', '.join(o.value for o in ocr_engines_enum_internal)}. " f"Use the option --show-external-plugins to see the options allowed with external plugins." ), ), ] = OcrAutoOptions.kind, ocr_lang: Annotated[ Optional[str], typer.Option( ..., help="Provide a comma-separated list of languages used by the OCR engine. Note that each OCR engine has different values for the language names.", ), ] = None, psm: Annotated[ Optional[int], typer.Option( ..., help="Page Segmentation Mode for the OCR engine (0-13).", ), ] = None, pdf_backend: Annotated[ PdfBackend, typer.Option(..., help="The PDF backend to use.") ] = PdfBackend.DLPARSE_V4, pdf_password: Annotated[ Optional[str], typer.Option(..., help="Password for protected PDF documents") ] = None, table_mode: Annotated[ TableFormerMode, typer.Option(..., help="The mode to use in the table structure model."), ] = TableFormerMode.ACCURATE, enrich_code: Annotated[ bool, typer.Option(..., help="Enable the code enrichment model in the pipeline."), ] = False, enrich_formula: Annotated[ bool, typer.Option(..., help="Enable the formula enrichment model in the pipeline."), ] = False, enrich_picture_classes: Annotated[ bool, typer.Option( ..., help="Enable the picture classification enrichment model in the pipeline.", ), ] = False, enrich_picture_description: Annotated[ bool, typer.Option(..., help="Enable the picture description model in the pipeline."), ] = False, artifacts_path: Annotated[ Optional[Path], typer.Option(..., help="If provided, the location of the model artifacts."), ] = None, enable_remote_services: Annotated[ bool, typer.Option( ..., help="Must be enabled when using models connecting to remote services." ), ] = False, allow_external_plugins: Annotated[ bool, typer.Option( ..., help="Must be enabled for loading modules from third-party plugins." ), ] = False, show_external_plugins: Annotated[ bool, typer.Option( ..., help="List the third-party plugins which are available when the option --allow-external-plugins is set.", callback=show_external_plugins_callback, is_eager=True, ), ] = False, abort_on_error: Annotated[ bool, typer.Option( ..., "--abort-on-error/--no-abort-on-error", help="If enabled, the processing will be aborted when the first error is encountered.", ), ] = False, output: Annotated[ Path, typer.Option(..., help="Output directory where results are saved.") ] = Path("."), verbose: Annotated[ int, typer.Option( "--verbose", "-v", count=True, help="Set the verbosity level. -v for info logging, -vv for debug logging.", ), ] = 0, debug_visualize_cells: Annotated[ bool, typer.Option(..., help="Enable debug output which visualizes the PDF cells"), ] = False, debug_visualize_ocr: Annotated[ bool, typer.Option(..., help="Enable debug output which visualizes the OCR cells"), ] = False, debug_visualize_layout: Annotated[ bool, typer.Option( ..., help="Enable debug output which visualizes the layour clusters" ), ] = False, debug_visualize_tables: Annotated[ bool, typer.Option(..., help="Enable debug output which visualizes the table cells"), ] = False, version: Annotated[ Optional[bool], typer.Option( "--version", callback=version_callback, is_eager=True, help="Show version information.", ), ] = None, document_timeout: Annotated[ Optional[float], typer.Option( ..., help="The timeout for processing each document, in seconds.", ), ] = None, num_threads: Annotated[int, typer.Option(..., help="Number of threads")] = 4, device: Annotated[ AcceleratorDevice, typer.Option(..., help="Accelerator device") ] = AcceleratorDevice.AUTO, docling_logo: Annotated[ Optional[bool], typer.Option( "--logo", callback=logo_callback, is_eager=True, help="Docling logo" ), ] = None, page_batch_size: Annotated[ int, typer.Option( "--page-batch-size", help=f"Number of pages processed in one batch. Default: {settings.perf.page_batch_size}", ), ] = settings.perf.page_batch_size, ): log_format = "%(asctime)s\t%(levelname)s\t%(name)s: %(message)s" if verbose == 0: logging.basicConfig(level=logging.WARNING, format=log_format) elif verbose == 1: logging.basicConfig(level=logging.INFO, format=log_format) else: logging.basicConfig(level=logging.DEBUG, format=log_format) settings.debug.visualize_cells = debug_visualize_cells settings.debug.visualize_layout = debug_visualize_layout settings.debug.visualize_tables = debug_visualize_tables settings.debug.visualize_ocr = debug_visualize_ocr settings.perf.page_batch_size = page_batch_size if from_formats is None: from_formats = list(InputFormat) parsed_headers: Optional[Dict[str, str]] = None if headers is not None: headers_t = TypeAdapter(Dict[str, str]) parsed_headers = headers_t.validate_json(headers) with tempfile.TemporaryDirectory() as tempdir: input_doc_paths: List[Path] = [] for src in input_sources: try: # check if we can fetch some remote url source = resolve_source_to_path( source=src, headers=parsed_headers, workdir=Path(tempdir) ) input_doc_paths.append(source) except FileNotFoundError: err_console.print( f"[red]Error: The input file {src} does not exist.[/red]" ) raise typer.Abort() except IsADirectoryError: # if the input matches to a file or a folder try: local_path = TypeAdapter(Path).validate_python(src) if local_path.exists() and local_path.is_dir(): for fmt in from_formats: for ext in FormatToExtensions[fmt]: for path in local_path.glob(f"**/*.{ext}"): if path.name.startswith("~$") and ext == "docx": _log.info( f"Ignoring temporary Word file: {path}" ) continue input_doc_paths.append(path) for path in local_path.glob(f"**/*.{ext.upper()}"): if path.name.startswith("~$") and ext == "docx": _log.info( f"Ignoring temporary Word file: {path}" ) continue input_doc_paths.append(path) elif local_path.exists(): if not local_path.name.startswith("~$") and ext == "docx": _log.info(f"Ignoring temporary Word file: {path}") continue input_doc_paths.append(local_path) else: err_console.print( f"[red]Error: The input file {src} does not exist.[/red]" ) raise typer.Abort() except Exception as err: err_console.print(f"[red]Error: Cannot read the input {src}.[/red]") _log.info(err) # will print more details if verbose is activated raise typer.Abort() if to_formats is None: to_formats = [OutputFormat.MARKDOWN] export_json = OutputFormat.JSON in to_formats export_yaml = OutputFormat.YAML in to_formats export_html = OutputFormat.HTML in to_formats export_html_split_page = OutputFormat.HTML_SPLIT_PAGE in to_formats export_md = OutputFormat.MARKDOWN in to_formats export_txt = OutputFormat.TEXT in to_formats export_doctags = OutputFormat.DOCTAGS in to_formats ocr_factory = get_ocr_factory(allow_external_plugins=allow_external_plugins) ocr_options: OcrOptions = ocr_factory.create_options( # type: ignore kind=ocr_engine, force_full_page_ocr=force_ocr, ) ocr_lang_list = _split_list(ocr_lang) if ocr_lang_list is not None: ocr_options.lang = ocr_lang_list if psm is not None and isinstance( ocr_options, (TesseractOcrOptions, TesseractCliOcrOptions) ): ocr_options.psm = psm accelerator_options = AcceleratorOptions(num_threads=num_threads, device=device) # pipeline_options: PaginatedPipelineOptions pipeline_options: PipelineOptions format_options: Dict[InputFormat, FormatOption] = {} pdf_backend_options: Optional[PdfBackendOptions] = PdfBackendOptions( password=pdf_password ) if pipeline == ProcessingPipeline.STANDARD: pipeline_options = PdfPipelineOptions( allow_external_plugins=allow_external_plugins, enable_remote_services=enable_remote_services, accelerator_options=accelerator_options, do_ocr=ocr, ocr_options=ocr_options, do_table_structure=tables, do_code_enrichment=enrich_code, do_formula_enrichment=enrich_formula, do_picture_description=enrich_picture_description, do_picture_classification=enrich_picture_classes, document_timeout=document_timeout, ) if isinstance( pipeline_options.table_structure_options, TableStructureOptions ): pipeline_options.table_structure_options.do_cell_matching = ( True # do_cell_matching ) pipeline_options.table_structure_options.mode = table_mode if image_export_mode != ImageRefMode.PLACEHOLDER: pipeline_options.generate_page_images = True pipeline_options.generate_picture_images = ( True # FIXME: to be deprecated in version 3 ) pipeline_options.images_scale = 2 backend: Type[PdfDocumentBackend] if pdf_backend == PdfBackend.DLPARSE_V1: backend = DoclingParseDocumentBackend pdf_backend_options = None elif pdf_backend == PdfBackend.DLPARSE_V2: backend = DoclingParseV2DocumentBackend pdf_backend_options = None elif pdf_backend == PdfBackend.DLPARSE_V4: backend = DoclingParseV4DocumentBackend # type: ignore elif pdf_backend == PdfBackend.PYPDFIUM2: backend = PyPdfiumDocumentBackend # type: ignore else: raise RuntimeError(f"Unexpected PDF backend type {pdf_backend}") pdf_format_option = PdfFormatOption( pipeline_options=pipeline_options, backend=backend, # pdf_backend backend_options=pdf_backend_options, ) # METS GBS options mets_gbs_options = pipeline_options.model_copy() mets_gbs_options.do_ocr = False mets_gbs_format_option = PdfFormatOption( pipeline_options=mets_gbs_options, 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 # Use image-native backend for IMAGE to avoid pypdfium2 locking image_format_option = PdfFormatOption( pipeline_options=pipeline_options, backend=ImageDocumentBackend, backend_options=pdf_backend_options, ) format_options = { InputFormat.PDF: pdf_format_option, InputFormat.IMAGE: image_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: pipeline_options = VlmPipelineOptions( enable_remote_services=enable_remote_services, ) if vlm_model == VlmModelType.GRANITE_VISION: pipeline_options.vlm_options = GRANITE_VISION_TRANSFORMERS elif vlm_model == VlmModelType.GRANITE_VISION_OLLAMA: pipeline_options.vlm_options = GRANITE_VISION_OLLAMA elif vlm_model == VlmModelType.GOT_OCR_2: pipeline_options.vlm_options = GOT2_TRANSFORMERS elif vlm_model == VlmModelType.SMOLDOCLING: pipeline_options.vlm_options = SMOLDOCLING_TRANSFORMERS if sys.platform == "darwin": try: import mlx_vlm pipeline_options.vlm_options = SMOLDOCLING_MLX except ImportError: if sys.version_info < (3, 14): _log.warning( "To run SmolDocling faster, please install mlx-vlm:\n" "pip install mlx-vlm" ) else: _log.warning( "You can run SmolDocling faster with MLX support, but it is unfortunately not yet available on Python 3.14." ) elif vlm_model == VlmModelType.GRANITEDOCLING: pipeline_options.vlm_options = GRANITEDOCLING_TRANSFORMERS if sys.platform == "darwin": try: import mlx_vlm pipeline_options.vlm_options = GRANITEDOCLING_MLX except ImportError: if sys.version_info < (3, 14): _log.warning( "To run GraniteDocling faster, please install mlx-vlm:\n" "pip install mlx-vlm" )
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
true
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/cli/__init__.py
docling/cli/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/chunking/__init__.py
docling/chunking/__init__.py
# # Copyright IBM Corp. 2024 - 2024 # SPDX-License-Identifier: MIT # from docling_core.transforms.chunker.base import BaseChunk, BaseChunker, BaseMeta from docling_core.transforms.chunker.hierarchical_chunker import ( DocChunk, DocMeta, HierarchicalChunker, ) from docling_core.transforms.chunker.hybrid_chunker import HybridChunker
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/pipeline/extraction_vlm_pipeline.py
docling/pipeline/extraction_vlm_pipeline.py
import inspect import json import logging from typing import Optional from PIL.Image import Image from pydantic import BaseModel from docling.backend.abstract_backend import PaginatedDocumentBackend from docling.backend.pdf_backend import PdfDocumentBackend from docling.datamodel.base_models import ConversionStatus, ErrorItem, VlmStopReason from docling.datamodel.document import InputDocument from docling.datamodel.extraction import ( ExtractedPageData, ExtractionResult, ExtractionTemplateType, ) 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, ) from docling.pipeline.base_extraction_pipeline import BaseExtractionPipeline from docling.utils.accelerator_utils import decide_device _log = logging.getLogger(__name__) class ExtractionVlmPipeline(BaseExtractionPipeline): def __init__(self, pipeline_options: VlmExtractionPipelineOptions): super().__init__(pipeline_options) # Initialize VLM model with default options self.accelerator_options = pipeline_options.accelerator_options self.pipeline_options: VlmExtractionPipelineOptions # Create VLM model instance self.vlm_model = NuExtractTransformersModel( enabled=True, artifacts_path=self.artifacts_path, # Will download automatically accelerator_options=self.accelerator_options, vlm_options=pipeline_options.vlm_options, ) def _extract_data( self, ext_res: ExtractionResult, template: Optional[ExtractionTemplateType] = None, ) -> ExtractionResult: """Extract data using the VLM model.""" try: # Get images from input document using the backend images = self._get_images_from_input(ext_res.input) if not images: ext_res.status = ConversionStatus.FAILURE ext_res.errors.append( ErrorItem( component_type="extraction_pipeline", module_name=self.__class__.__name__, error_message="No images found in document", ) ) return ext_res # Use provided template or default prompt if template is not None: prompt = self._serialize_template(template) else: prompt = "Extract all text and structured information from this document. Return as JSON." # Process all images with VLM model start_page, end_page = ext_res.input.limits.page_range for i, image in enumerate(images): # Calculate the actual page number based on the filtered range page_number = start_page + i try: predictions = list(self.vlm_model.process_images([image], prompt)) if predictions: # Parse the extracted text as JSON if possible, otherwise use as-is extracted_text = predictions[0].text extracted_data = None vlm_stop_reason: VlmStopReason = predictions[0].stop_reason if ( vlm_stop_reason == VlmStopReason.LENGTH or vlm_stop_reason == VlmStopReason.STOP_SEQUENCE ): ext_res.status = ConversionStatus.PARTIAL_SUCCESS try: extracted_data = json.loads(extracted_text) except (json.JSONDecodeError, ValueError): # If not valid JSON, keep extracted_data as None pass # Create page data with proper structure page_data = ExtractedPageData( page_no=page_number, extracted_data=extracted_data, raw_text=extracted_text, # Always populate raw_text ) ext_res.pages.append(page_data) else: # Add error page data page_data = ExtractedPageData( page_no=page_number, extracted_data=None, errors=["No extraction result from VLM model"], ) ext_res.pages.append(page_data) except Exception as e: _log.error(f"Error processing page {page_number}: {e}") page_data = ExtractedPageData( page_no=page_number, extracted_data=None, errors=[str(e)] ) ext_res.pages.append(page_data) except Exception as e: _log.error(f"Error during extraction: {e}") ext_res.errors.append( ErrorItem( component_type="extraction_pipeline", module_name=self.__class__.__name__, error_message=str(e), ) ) return ext_res def _determine_status(self, ext_res: ExtractionResult) -> ConversionStatus: """Determine the status based on extraction results.""" if ext_res.pages and not any(page.errors for page in ext_res.pages): return ( ConversionStatus.PARTIAL_SUCCESS if ext_res.status == ConversionStatus.PARTIAL_SUCCESS else ConversionStatus.SUCCESS ) else: return ConversionStatus.FAILURE def _get_images_from_input(self, input_doc: InputDocument) -> list[Image]: """Extract images from input document using the backend.""" images = [] try: backend = input_doc._backend assert isinstance(backend, PdfDocumentBackend) # Use the backend's pagination interface page_count = backend.page_count() # Respect page range limits, following the same pattern as PaginatedPipeline start_page, end_page = input_doc.limits.page_range _log.info( f"Processing pages {start_page}-{end_page} of {page_count} total pages for extraction" ) for page_num in range(page_count): # Only process pages within the specified range (0-based indexing) if start_page - 1 <= page_num <= end_page - 1: try: page_backend = backend.load_page(page_num) if page_backend.is_valid(): # Get page image at a reasonable scale page_image = page_backend.get_page_image( scale=self.pipeline_options.vlm_options.scale ) images.append(page_image) else: _log.warning(f"Page {page_num + 1} backend is not valid") except Exception as e: _log.error(f"Error loading page {page_num + 1}: {e}") except Exception as e: _log.error(f"Error getting images from input document: {e}") return images def _serialize_template(self, template: ExtractionTemplateType) -> str: """Serialize template to string based on its type.""" if isinstance(template, str): return template elif isinstance(template, dict): return json.dumps(template, indent=2) elif isinstance(template, BaseModel): return template.model_dump_json(indent=2) elif inspect.isclass(template) and issubclass(template, BaseModel): from polyfactory.factories.pydantic_factory import ModelFactory 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) -> PipelineOptions: return VlmExtractionPipelineOptions()
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/pipeline/standard_pdf_pipeline.py
docling/pipeline/standard_pdf_pipeline.py
"""Thread-safe, production-ready PDF pipeline ================================================ A self-contained, thread-safe PDF conversion pipeline exploiting parallelism between pipeline stages and models. * **Per-run isolation** - every :py:meth:`execute` call uses its own bounded queues and worker threads so that concurrent invocations never share mutable state. * **Deterministic run identifiers** - pages are tracked with an internal *run-id* instead of relying on :pyfunc:`id`, which may clash after garbage collection. * **Explicit back-pressure & shutdown** - producers block on full queues; queue *close()* propagates downstream so stages terminate deterministically without sentinels. * **Minimal shared state** - heavyweight models are initialised once per pipeline instance and only read by worker threads; no runtime mutability is exposed. * **Strict typing & clean API usage** - code is fully annotated and respects *coding_rules.md*. """ from __future__ import annotations 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, Callable, 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 from docling.datamodel.base_models import ( AssembledUnit, ConversionStatus, DoclingComponentType, ErrorItem, Page, ) 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.factories import ( get_layout_factory, get_ocr_factory, get_table_structure_factory, ) from docling.models.page_assemble_model import PageAssembleModel, PageAssembleOptions from docling.models.page_preprocessing_model import ( PagePreprocessingModel, PagePreprocessingOptions, ) from docling.models.readingorder_model import ReadingOrderModel, ReadingOrderOptions from docling.pipeline.base_pipeline import ConvertPipeline from docling.utils.profiling import ProfilingScope, TimeRecorder from docling.utils.utils import chunkify _log = logging.getLogger(__name__) # ────────────────────────────────────────────────────────────────────────────── # Helper data structures # ────────────────────────────────────────────────────────────────────────────── @dataclass class ThreadedItem: """Envelope that travels between pipeline stages.""" payload: Optional[Page] run_id: int # Unique per *execute* call, monotonic across pipeline instance page_no: int conv_res: ConversionResult error: Optional[Exception] = None is_failed: bool = False @dataclass class ProcessingResult: """Aggregated outcome of a pipeline run.""" pages: List[Page] = field(default_factory=list) failed_pages: List[Tuple[int, Exception]] = field(default_factory=list) total_expected: int = 0 @property def success_count(self) -> int: return len(self.pages) @property def failure_count(self) -> int: return len(self.failed_pages) @property def is_partial_success(self) -> bool: return 0 < self.success_count < self.total_expected @property def is_complete_failure(self) -> bool: return self.success_count == 0 and self.failure_count > 0 class ThreadedQueue: """Bounded queue with blocking put/ get_batch and explicit *close()* semantics.""" __slots__ = ("_closed", "_items", "_lock", "_max", "_not_empty", "_not_full") def __init__(self, max_size: int) -> None: self._max: int = max_size self._items: deque[ThreadedItem] = deque() self._lock = threading.Lock() self._not_full = threading.Condition(self._lock) self._not_empty = threading.Condition(self._lock) self._closed = False # ---------------------------------------------------------------- put() def put(self, item: ThreadedItem, timeout: Optional[float] | None = None) -> bool: """Block until queue accepts *item* or is closed. Returns *False* if closed.""" with self._not_full: if self._closed: return False start = time.monotonic() while len(self._items) >= self._max and not self._closed: if timeout is not None: remaining = timeout - (time.monotonic() - start) if remaining <= 0: return False self._not_full.wait(remaining) else: self._not_full.wait() if self._closed: return False self._items.append(item) self._not_empty.notify() return True # ------------------------------------------------------------ get_batch() def get_batch( self, size: int, timeout: Optional[float] | None = None ) -> List[ThreadedItem]: """Return up to *size* items. Blocks until ≥1 item present or queue closed/timeout.""" with self._not_empty: start = time.monotonic() while not self._items and not self._closed: if timeout is not None: remaining = timeout - (time.monotonic() - start) if remaining <= 0: return [] self._not_empty.wait(remaining) else: self._not_empty.wait() batch: List[ThreadedItem] = [] while self._items and len(batch) < size: batch.append(self._items.popleft()) if batch: self._not_full.notify_all() return batch # ---------------------------------------------------------------- close() def close(self) -> None: with self._lock: self._closed = True self._not_empty.notify_all() self._not_full.notify_all() # -------------------------------------------------------------- property @property def closed(self) -> bool: return self._closed class ThreadedPipelineStage: """A single pipeline stage backed by one worker thread.""" def __init__( self, *, name: str, model: Any, batch_size: int, batch_timeout: float, queue_max_size: int, postprocess: Optional[Callable[[ThreadedItem], None]] = None, timed_out_run_ids: Optional[set[int]] = None, ) -> None: self.name = name self.model = model self.batch_size = batch_size self.batch_timeout = batch_timeout self.input_queue = ThreadedQueue(queue_max_size) self._outputs: list[ThreadedQueue] = [] self._thread: Optional[threading.Thread] = None self._running = False self._postprocess = postprocess self._timed_out_run_ids = ( timed_out_run_ids if timed_out_run_ids is not None else set() ) # ---------------------------------------------------------------- wiring def add_output_queue(self, q: ThreadedQueue) -> None: self._outputs.append(q) # -------------------------------------------------------------- lifecycle def start(self) -> None: if self._running: return self._running = True self._thread = threading.Thread( target=self._run, name=f"Stage-{self.name}", daemon=False ) self._thread.start() def stop(self) -> None: if not self._running: return self._running = False self.input_queue.close() if self._thread is not None: # Give thread 2s to finish naturally before abandoning self._thread.join(timeout=15.0) if self._thread.is_alive(): _log.warning( "Stage %s thread did not terminate within 15s. " "Thread is likely stuck in a blocking call and will be abandoned (resources may leak).", self.name, ) # ------------------------------------------------------------------ _run def _run(self) -> None: try: while self._running: batch = self.input_queue.get_batch(self.batch_size, self.batch_timeout) if not batch and self.input_queue.closed: break processed = self._process_batch(batch) self._emit(processed) except Exception: # pragma: no cover - top-level guard _log.exception("Fatal error in stage %s", self.name) finally: for q in self._outputs: q.close() # ----------------------------------------------------- _process_batch() def _process_batch(self, batch: Sequence[ThreadedItem]) -> list[ThreadedItem]: """Run *model* on *batch* grouped by run_id to maximise batching.""" groups: dict[int, list[ThreadedItem]] = defaultdict(list) for itm in batch: groups[itm.run_id].append(itm) result: list[ThreadedItem] = [] for rid, items in groups.items(): # If run_id is timed out, skip processing but pass through items as-is # This allows already-completed work to flow through while aborting new work if rid in self._timed_out_run_ids: for it in items: it.is_failed = True if it.error is None: it.error = RuntimeError("document timeout exceeded") result.extend(items) continue good: list[ThreadedItem] = [i for i in items if not i.is_failed] if not good: result.extend(items) continue try: # Filter out None payloads and ensure type safety pages_with_payloads = [ (i, i.payload) for i in good if i.payload is not None ] if len(pages_with_payloads) != len(good): # Some items have None payloads, mark all as failed for it in items: it.is_failed = True it.error = RuntimeError("Page payload is None") result.extend(items) continue pages: List[Page] = [payload for _, payload in pages_with_payloads] processed_pages = list(self.model(good[0].conv_res, pages)) # type: ignore[arg-type] if len(processed_pages) != len(pages): # strict mismatch guard raise RuntimeError( f"Model {self.name} returned wrong number of pages" ) for idx, page in enumerate(processed_pages): result.append( ThreadedItem( payload=page, run_id=rid, page_no=good[idx].page_no, conv_res=good[idx].conv_res, ) ) except Exception as exc: _log.error( "Stage %s failed for run %d: %s", self.name, rid, exc, exc_info=True ) for it in items: it.is_failed = True it.error = exc result.extend(items) return result # -------------------------------------------------------------- _emit() def _emit(self, items: Iterable[ThreadedItem]) -> None: for item in items: if self._postprocess is not None: self._postprocess(item) for q in self._outputs: if not q.put(item): _log.error("Output queue closed while emitting from %s", self.name) class PreprocessThreadedStage(ThreadedPipelineStage): """Pipeline stage that lazily loads PDF backends just-in-time.""" def __init__( self, *, batch_timeout: float, queue_max_size: int, model: Any, timed_out_run_ids: Optional[set[int]] = None, ) -> None: super().__init__( name="preprocess", model=model, batch_size=1, batch_timeout=batch_timeout, queue_max_size=queue_max_size, timed_out_run_ids=timed_out_run_ids, ) def _process_batch(self, batch: Sequence[ThreadedItem]) -> list[ThreadedItem]: groups: dict[int, list[ThreadedItem]] = defaultdict(list) for itm in batch: groups[itm.run_id].append(itm) result: list[ThreadedItem] = [] for rid, items in groups.items(): # If run_id is timed out, skip processing but pass through items as-is # This allows already-completed work to flow through while aborting new work if rid in self._timed_out_run_ids: for it in items: it.is_failed = True if it.error is None: it.error = RuntimeError("document timeout exceeded") result.extend(items) continue good = [i for i in items if not i.is_failed] if not good: result.extend(items) continue try: pages_with_payloads: list[tuple[ThreadedItem, Page]] = [] for it in good: page = it.payload if page is None: raise RuntimeError("Page payload is None") if page._backend is None: backend = it.conv_res.input._backend assert isinstance(backend, PdfDocumentBackend), ( "Threaded pipeline only supports PdfDocumentBackend." ) page_backend = backend.load_page(page.page_no) page._backend = page_backend if page_backend.is_valid(): page.size = page_backend.get_size() pages_with_payloads.append((it, page)) pages = [payload for _, payload in pages_with_payloads] processed_pages = list( self.model(good[0].conv_res, pages) # type: ignore[arg-type] ) if len(processed_pages) != len(pages): raise RuntimeError( "PagePreprocessingModel returned unexpected number of pages" ) for idx, processed_page in enumerate(processed_pages): result.append( ThreadedItem( payload=processed_page, run_id=rid, page_no=good[idx].page_no, conv_res=good[idx].conv_res, ) ) except Exception as exc: _log.error("Stage preprocess failed for run %d: %s", rid, exc) for it in items: it.is_failed = True it.error = exc result.extend(items) return result @dataclass class RunContext: """Wiring for a single *execute* call.""" stages: list[ThreadedPipelineStage] first_stage: ThreadedPipelineStage output_queue: ThreadedQueue timed_out_run_ids: set[int] = field(default_factory=set) # ────────────────────────────────────────────────────────────────────────────── # Main pipeline # ────────────────────────────────────────────────────────────────────────────── class StandardPdfPipeline(ConvertPipeline): """High-performance PDF pipeline with multi-threaded stages.""" def __init__(self, pipeline_options: ThreadedPdfPipelineOptions) -> None: super().__init__(pipeline_options) self.pipeline_options: ThreadedPdfPipelineOptions = pipeline_options self._run_seq = itertools.count(1) # deterministic, monotonic run ids # initialise heavy models once self._init_models() # ──────────────────────────────────────────────────────────────────────── # Heavy-model initialisation & helpers # ──────────────────────────────────────────────────────────────────────── def _init_models(self) -> None: art_path = self.artifacts_path self.keep_images = ( self.pipeline_options.generate_page_images or self.pipeline_options.generate_picture_images or self.pipeline_options.generate_table_images ) self.preprocessing_model = PagePreprocessingModel( options=PagePreprocessingOptions( images_scale=self.pipeline_options.images_scale ) ) self.ocr_model = self._make_ocr_model(art_path) layout_factory = get_layout_factory( allow_external_plugins=self.pipeline_options.allow_external_plugins ) self.layout_model = layout_factory.create_instance( options=self.pipeline_options.layout_options, artifacts_path=art_path, accelerator_options=self.pipeline_options.accelerator_options, ) table_factory = get_table_structure_factory( allow_external_plugins=self.pipeline_options.allow_external_plugins ) self.table_model = table_factory.create_instance( options=self.pipeline_options.table_structure_options, enabled=self.pipeline_options.do_table_structure, artifacts_path=art_path, accelerator_options=self.pipeline_options.accelerator_options, ) self.assemble_model = PageAssembleModel(options=PageAssembleOptions()) self.reading_order_model = ReadingOrderModel(options=ReadingOrderOptions()) # --- optional 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, ), *self.enrichment_pipe, ] self.keep_backend = any( ( self.pipeline_options.do_formula_enrichment, self.pipeline_options.do_code_enrichment, self.pipeline_options.do_picture_classification, self.pipeline_options.do_picture_description, ) ) # ---------------------------------------------------------------- helpers def _make_ocr_model(self, art_path: Optional[Path]) -> Any: factory = get_ocr_factory( allow_external_plugins=self.pipeline_options.allow_external_plugins ) return factory.create_instance( options=self.pipeline_options.ocr_options, enabled=self.pipeline_options.do_ocr, artifacts_path=art_path, accelerator_options=self.pipeline_options.accelerator_options, ) def _release_page_resources(self, item: ThreadedItem) -> None: page = item.payload if page is None: return if not self.keep_images: page._image_cache = {} if not self.keep_backend and page._backend is not None: page._backend.unload() page._backend = None if not self.pipeline_options.generate_parsed_pages: page.parsed_page = None # ──────────────────────────────────────────────────────────────────────── # Build - thread pipeline # ──────────────────────────────────────────────────────────────────────── def _create_run_ctx(self) -> RunContext: opts = self.pipeline_options timed_out_run_ids: set[int] = set() preprocess = PreprocessThreadedStage( batch_timeout=opts.batch_polling_interval_seconds, queue_max_size=opts.queue_max_size, model=self.preprocessing_model, timed_out_run_ids=timed_out_run_ids, ) ocr = ThreadedPipelineStage( name="ocr", model=self.ocr_model, batch_size=opts.ocr_batch_size, batch_timeout=opts.batch_polling_interval_seconds, queue_max_size=opts.queue_max_size, timed_out_run_ids=timed_out_run_ids, ) layout = ThreadedPipelineStage( name="layout", model=self.layout_model, batch_size=opts.layout_batch_size, batch_timeout=opts.batch_polling_interval_seconds, queue_max_size=opts.queue_max_size, timed_out_run_ids=timed_out_run_ids, ) table = ThreadedPipelineStage( name="table", model=self.table_model, batch_size=opts.table_batch_size, batch_timeout=opts.batch_polling_interval_seconds, queue_max_size=opts.queue_max_size, timed_out_run_ids=timed_out_run_ids, ) assemble = ThreadedPipelineStage( name="assemble", model=self.assemble_model, batch_size=1, batch_timeout=opts.batch_polling_interval_seconds, queue_max_size=opts.queue_max_size, postprocess=self._release_page_resources, timed_out_run_ids=timed_out_run_ids, ) # wire stages output_q = ThreadedQueue(opts.queue_max_size) preprocess.add_output_queue(ocr.input_queue) ocr.add_output_queue(layout.input_queue) layout.add_output_queue(table.input_queue) table.add_output_queue(assemble.input_queue) assemble.add_output_queue(output_q) stages = [preprocess, ocr, layout, table, assemble] return RunContext( stages=stages, first_stage=preprocess, output_queue=output_q, timed_out_run_ids=timed_out_run_ids, ) # --------------------------------------------------------------------- build def _build_document(self, conv_res: ConversionResult) -> ConversionResult: """Stream-build the document while interleaving producer and consumer work. Note: If a worker thread gets stuck in a blocking call (model inference or PDF backend load_page/get_size), that thread will be abandoned after a brief wait (15s) during cleanup. The thread continues running until the blocking call completes, potentially holding resources (e.g., pypdfium2_lock). """ run_id = next(self._run_seq) assert isinstance(conv_res.input._backend, PdfDocumentBackend) # Collect page placeholders; backends are loaded lazily in preprocess stage start_page, end_page = conv_res.input.limits.page_range pages: list[Page] = [] for i in range(conv_res.input.page_count): if start_page - 1 <= i <= end_page - 1: page = Page(page_no=i) conv_res.pages.append(page) pages.append(page) if not pages: conv_res.status = ConversionStatus.FAILURE return conv_res total_pages: int = len(pages) ctx: RunContext = self._create_run_ctx() for st in ctx.stages: st.start() proc = ProcessingResult(total_expected=total_pages) fed_idx: int = 0 # number of pages successfully queued batch_size: int = 32 # drain chunk start_time = time.monotonic() timeout_exceeded = False input_queue_closed = False try: while proc.success_count + proc.failure_count < total_pages: # Check timeout if ( self.pipeline_options.document_timeout is not None and not timeout_exceeded ): elapsed_time = time.monotonic() - start_time if elapsed_time > self.pipeline_options.document_timeout: _log.warning( f"Document processing time ({elapsed_time:.3f}s) " f"exceeded timeout of {self.pipeline_options.document_timeout:.3f}s" ) timeout_exceeded = True ctx.timed_out_run_ids.add(run_id) if not input_queue_closed: ctx.first_stage.input_queue.close() input_queue_closed = True # Break immediately - don't wait for in-flight work break # 1) feed - try to enqueue until the first queue is full if not input_queue_closed: while fed_idx < total_pages: ok = ctx.first_stage.input_queue.put( ThreadedItem( payload=pages[fed_idx], run_id=run_id, page_no=pages[fed_idx].page_no, conv_res=conv_res, ), timeout=0.0, # non-blocking try-put ) if ok: fed_idx += 1 if fed_idx == total_pages: ctx.first_stage.input_queue.close() input_queue_closed = True else: # queue full - switch to draining break # 2) drain - pull whatever is ready from the output side out_batch = ctx.output_queue.get_batch(batch_size, timeout=0.05) for itm in out_batch: if itm.run_id != run_id: continue if itm.is_failed or itm.error: proc.failed_pages.append( (itm.page_no, itm.error or RuntimeError("unknown error")) ) else: assert itm.payload is not None proc.pages.append(itm.payload) # 3) failure safety - downstream closed early if not out_batch and ctx.output_queue.closed: missing = total_pages - (proc.success_count + proc.failure_count) if missing > 0: proc.failed_pages.extend( [(-1, RuntimeError("pipeline terminated early"))] * missing ) break # Mark remaining pages as failed if timeout occurred if timeout_exceeded: completed_page_nos = {p.page_no for p in proc.pages} | { fp for fp, _ in proc.failed_pages } for page in pages[fed_idx:]: if page.page_no not in completed_page_nos: proc.failed_pages.append( (page.page_no, RuntimeError("document timeout exceeded")) ) finally: for st in ctx.stages: st.stop() ctx.output_queue.close() self._integrate_results(conv_res, proc, timeout_exceeded=timeout_exceeded) return conv_res # ---------------------------------------------------- integrate_results() def _integrate_results( self, conv_res: ConversionResult, proc: ProcessingResult, timeout_exceeded: bool = False, ) -> None: page_map = {p.page_no: p for p in proc.pages} # Only keep pages that successfully completed processing conv_res.pages = [ page_map[p.page_no] for p in conv_res.pages if p.page_no in page_map ] # Add error details from failed pages for page_no, error in proc.failed_pages: page_label = f"Page {page_no + 1}" if page_no >= 0 else "Unknown page" error_msg = str(error) if error else "" error_item = ErrorItem( component_type=DoclingComponentType.PIPELINE, module_name=self.__class__.__name__, error_message=f"{page_label}: {error_msg}" if error_msg else page_label, ) conv_res.errors.append(error_item) if timeout_exceeded and proc.total_expected > 0: # Timeout exceeded: set PARTIAL_SUCCESS if any pages were attempted conv_res.status = ConversionStatus.PARTIAL_SUCCESS elif proc.is_complete_failure: conv_res.status = ConversionStatus.FAILURE elif proc.is_partial_success: conv_res.status = ConversionStatus.PARTIAL_SUCCESS else: conv_res.status = ConversionStatus.SUCCESS if not self.keep_images: for p in conv_res.pages: p._image_cache = {} for p in conv_res.pages: if not self.keep_backend and p._backend is not None: p._backend.unload() if not self.pipeline_options.generate_parsed_pages: del p.parsed_page p.parsed_page = None # ---------------------------------------------------------------- assemble def _assemble_document(self, conv_res: ConversionResult) -> ConversionResult: elements, headers, body = [], [], [] with TimeRecorder(conv_res, "doc_assemble", scope=ProfilingScope.DOCUMENT): for p in conv_res.pages: if p.assembled: elements.extend(p.assembled.elements) headers.extend(p.assembled.headers) body.extend(p.assembled.body) conv_res.assembled = AssembledUnit( 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
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
true
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/pipeline/threaded_standard_pdf_pipeline.py
docling/pipeline/threaded_standard_pdf_pipeline.py
from docling.pipeline.standard_pdf_pipeline import StandardPdfPipeline class ThreadedStandardPdfPipeline(StandardPdfPipeline): """Backwards compatible import for ThreadedStandardPdfPipeline."""
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/pipeline/vlm_pipeline.py
docling/pipeline/vlm_pipeline.py
import logging import re from io import BytesIO from pathlib import Path from typing import List, Optional, Union, cast from docling_core.types.doc import ( BoundingBox, ContentLayer, DocItem, DoclingDocument, ImageRef, PictureItem, ProvenanceItem, TextItem, ) from docling_core.types.doc.base import ( BoundingBox, Size, ) from docling_core.types.doc.document import DocTagsDocument from PIL import Image as PILImage from docling.backend.abstract_backend import AbstractDocumentBackend from docling.backend.html_backend import HTMLDocumentBackend from docling.backend.md_backend import MarkdownDocumentBackend from docling.backend.pdf_backend import PdfDocumentBackend from docling.datamodel.base_models import InputFormat, Page from docling.datamodel.document import ConversionResult, InputDocument from docling.datamodel.pipeline_options import ( VlmPipelineOptions, ) from docling.datamodel.pipeline_options_vlm_model import ( ApiVlmOptions, InferenceFramework, InlineVlmOptions, ResponseFormat, ) from docling.datamodel.settings import settings from docling.models.api_vlm_model import ApiVlmModel from docling.models.vlm_models_inline.hf_transformers_model import ( HuggingFaceTransformersVlmModel, ) from docling.models.vlm_models_inline.mlx_model import HuggingFaceMlxModel from docling.pipeline.base_pipeline import PaginatedPipeline from docling.utils.profiling import ProfilingScope, TimeRecorder _log = logging.getLogger(__name__) class VlmPipeline(PaginatedPipeline): def __init__(self, pipeline_options: VlmPipelineOptions): super().__init__(pipeline_options) self.keep_backend = True self.pipeline_options: VlmPipelineOptions # 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 = ( pipeline_options.force_backend_text and pipeline_options.vlm_options.response_format == ResponseFormat.DOCTAGS ) self.keep_images = self.pipeline_options.generate_page_images if isinstance(pipeline_options.vlm_options, ApiVlmOptions): self.build_pipe = [ ApiVlmModel( enabled=True, # must be always enabled for this pipeline to make sense. enable_remote_services=self.pipeline_options.enable_remote_services, vlm_options=cast(ApiVlmOptions, self.pipeline_options.vlm_options), ), ] elif isinstance(self.pipeline_options.vlm_options, InlineVlmOptions): vlm_options = cast(InlineVlmOptions, self.pipeline_options.vlm_options) if vlm_options.inference_framework == InferenceFramework.MLX: self.build_pipe = [ HuggingFaceMlxModel( enabled=True, # must be always enabled for this pipeline to make sense. artifacts_path=self.artifacts_path, accelerator_options=pipeline_options.accelerator_options, vlm_options=vlm_options, ), ] elif vlm_options.inference_framework == InferenceFramework.TRANSFORMERS: self.build_pipe = [ HuggingFaceTransformersVlmModel( enabled=True, # must be always enabled for this pipeline to make sense. artifacts_path=self.artifacts_path, accelerator_options=pipeline_options.accelerator_options, vlm_options=vlm_options, ), ] elif vlm_options.inference_framework == InferenceFramework.VLLM: from docling.models.vlm_models_inline.vllm_model import VllmVlmModel self.build_pipe = [ VllmVlmModel( enabled=True, # must be always enabled for this pipeline to make sense. artifacts_path=self.artifacts_path, accelerator_options=pipeline_options.accelerator_options, vlm_options=vlm_options, ), ] else: raise ValueError( f"Could not instantiate the right type of VLM pipeline: {vlm_options.inference_framework}" ) self.enrichment_pipe = [ # Other models working on `NodeItem` elements in the DoclingDocument ] def initialize_page(self, conv_res: ConversionResult, page: Page) -> Page: with TimeRecorder(conv_res, "page_init"): images_scale = self.pipeline_options.images_scale if images_scale is not None: page._default_image_scale = images_scale page._backend = conv_res.input._backend.load_page(page.page_no) # type: ignore if page._backend is not None and page._backend.is_valid(): page.size = page._backend.get_size() if self.force_backend_text: page.parsed_page = page._backend.get_segmented_page() return page def extract_text_from_backend( self, page: Page, bbox: Union[BoundingBox, None] ) -> str: # Convert bounding box normalized to 0-100 into page coordinates for cropping text = "" if bbox: if page.size: if page._backend: text = page._backend.get_text_in_rect(bbox) return text def _assemble_document(self, conv_res: ConversionResult) -> ConversionResult: with TimeRecorder(conv_res, "doc_assemble", scope=ProfilingScope.DOCUMENT): if ( self.pipeline_options.vlm_options.response_format == ResponseFormat.DOCTAGS ): conv_res.document = self._turn_dt_into_doc(conv_res) elif ( self.pipeline_options.vlm_options.response_format == ResponseFormat.MARKDOWN ): conv_res.document = self._turn_md_into_doc(conv_res) elif ( self.pipeline_options.vlm_options.response_format == ResponseFormat.HTML ): conv_res.document = self._turn_html_into_doc(conv_res) else: raise RuntimeError( f"Unsupported VLM response format {self.pipeline_options.vlm_options.response_format}" ) # Generate images of the requested element types if self.pipeline_options.generate_picture_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 ): page_ix = element.prov[0].page_no - 1 page = conv_res.pages[page_ix] 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) ) return conv_res def _turn_dt_into_doc(self, conv_res) -> DoclingDocument: doctags_list = [] image_list = [] for page in conv_res.pages: predicted_doctags = "" img = PILImage.new("RGB", (1, 1), "rgb(255,255,255)") if page.predictions.vlm_response: predicted_doctags = page.predictions.vlm_response.text if page.image: img = page.image image_list.append(img) doctags_list.append(predicted_doctags) doctags_list_c = cast(List[Union[Path, str]], doctags_list) image_list_c = cast(List[Union[Path, PILImage.Image]], image_list) doctags_doc = DocTagsDocument.from_doctags_and_image_pairs( doctags_list_c, image_list_c ) conv_res.document = DoclingDocument.load_from_doctags( doctag_document=doctags_doc ) # If forced backend text, replace model predicted text with backend one if page.size: if self.force_backend_text: scale = self.pipeline_options.images_scale for element, _level in conv_res.document.iterate_items(): if not isinstance(element, TextItem) or len(element.prov) == 0: continue crop_bbox = ( element.prov[0] .bbox.scaled(scale=scale) .to_top_left_origin(page_height=page.size.height * scale) ) txt = self.extract_text_from_backend(page, crop_bbox) element.text = txt element.orig = txt return conv_res.document def _turn_md_into_doc(self, conv_res): def _extract_markdown_code(text): """ Extracts text from markdown code blocks (enclosed in triple backticks). If no code blocks are found, returns the original text. Args: text (str): Input text that may contain markdown code blocks Returns: str: Extracted code if code blocks exist, otherwise original text """ # Regex pattern to match content between triple backticks # This handles multiline content and optional language specifier pattern = r"^```(?:\w*\n)?(.*?)```(\n)*$" # Search with DOTALL flag to match across multiple lines mtch = re.search(pattern, text, re.DOTALL) if mtch: # Return only the content of the first capturing group return mtch.group(1) else: # No code blocks found, return original text return text page_docs = [] for pg_idx, page in enumerate(conv_res.pages): predicted_text = "" if page.predictions.vlm_response: predicted_text = page.predictions.vlm_response.text + "\n\n" predicted_text = _extract_markdown_code(text=predicted_text) response_bytes = BytesIO(predicted_text.encode("utf8")) out_doc = InputDocument( path_or_stream=response_bytes, filename=conv_res.input.file.name, format=InputFormat.MD, backend=MarkdownDocumentBackend, ) backend = MarkdownDocumentBackend( in_doc=out_doc, path_or_stream=response_bytes, ) page_doc = backend.convert() # Modify provenance in place for all items in the page document for item, level in page_doc.iterate_items( with_groups=True, traverse_pictures=True, included_content_layers=set(ContentLayer), ): if isinstance(item, DocItem): item.prov = [ ProvenanceItem( page_no=pg_idx + 1, bbox=BoundingBox( t=0.0, b=0.0, l=0.0, r=0.0 ), # FIXME: would be nice not to have to "fake" it charspan=[0, 0], ) ] # Add page metadata to the page document before concatenation if page.image is not None: pg_width = page.image.width pg_height = page.image.height else: pg_width = 1 pg_height = 1 page_doc.add_page( page_no=pg_idx + 1, size=Size(width=pg_width, height=pg_height), image=ImageRef.from_pil(image=page.image, dpi=72) if page.image else None, ) page_docs.append(page_doc) final_doc = DoclingDocument.concatenate(docs=page_docs) return final_doc def _turn_html_into_doc(self, conv_res): def _extract_html_code(text): """ Extracts text from markdown code blocks (enclosed in triple backticks). If no code blocks are found, returns the original text. Args: text (str): Input text that may contain markdown code blocks Returns: str: Extracted code if code blocks exist, otherwise original text """ # Regex pattern to match content between triple backticks # This handles multiline content and optional language specifier pattern = r"^```(?:\w*\n)?(.*?)```(\n)*$" # Search with DOTALL flag to match across multiple lines mtch = re.search(pattern, text, re.DOTALL) if mtch: # Return only the content of the first capturing group return mtch.group(1) else: # No code blocks found, return original text return text page_docs = [] for pg_idx, page in enumerate(conv_res.pages): predicted_text = "" if page.predictions.vlm_response: predicted_text = page.predictions.vlm_response.text + "\n\n" predicted_text = _extract_html_code(text=predicted_text) response_bytes = BytesIO(predicted_text.encode("utf8")) out_doc = InputDocument( path_or_stream=response_bytes, filename=conv_res.input.file.name, format=InputFormat.HTML, backend=HTMLDocumentBackend, ) backend = HTMLDocumentBackend( in_doc=out_doc, path_or_stream=response_bytes, ) page_doc = backend.convert() # Modify provenance in place for all items in the page document for item, level in page_doc.iterate_items( with_groups=True, traverse_pictures=True, included_content_layers=set(ContentLayer), ): if isinstance(item, DocItem): item.prov = [ ProvenanceItem( page_no=pg_idx + 1, bbox=BoundingBox( t=0.0, b=0.0, l=0.0, r=0.0 ), # FIXME: would be nice not to have to "fake" it charspan=[0, 0], ) ] # Add page metadata to the page document before concatenation if page.image is not None: pg_width = page.image.width pg_height = page.image.height else: pg_width = 1 pg_height = 1 page_doc.add_page( page_no=pg_idx + 1, size=Size(width=pg_width, height=pg_height), image=ImageRef.from_pil(image=page.image, dpi=72) if page.image else None, ) page_docs.append(page_doc) # Concatenate all page documents to preserve hierarchy final_doc = DoclingDocument.concatenate(docs=page_docs) return final_doc @classmethod def get_default_options(cls) -> VlmPipelineOptions: return VlmPipelineOptions() @classmethod def is_backend_supported(cls, backend: AbstractDocumentBackend): return isinstance(backend, PdfDocumentBackend)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/pipeline/__init__.py
docling/pipeline/__init__.py
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/pipeline/legacy_standard_pdf_pipeline.py
docling/pipeline/legacy_standard_pdf_pipeline.py
import logging import warnings from pathlib import Path from typing import Optional, 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 from docling.datamodel.base_models import AssembledUnit, Page from docling.datamodel.document import ConversionResult from docling.datamodel.layout_model_specs import LayoutModelConfig 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.factories import ( get_layout_factory, get_ocr_factory, get_table_structure_factory, ) from docling.models.page_assemble_model import PageAssembleModel, PageAssembleOptions from docling.models.page_preprocessing_model import ( PagePreprocessingModel, PagePreprocessingOptions, ) from docling.models.readingorder_model import ReadingOrderModel, ReadingOrderOptions from docling.pipeline.base_pipeline import PaginatedPipeline from docling.utils.model_downloader import download_models from docling.utils.profiling import ProfilingScope, TimeRecorder _log = logging.getLogger(__name__) class LegacyStandardPdfPipeline(PaginatedPipeline): def __init__(self, pipeline_options: PdfPipelineOptions): super().__init__(pipeline_options) self.pipeline_options: PdfPipelineOptions with warnings.catch_warnings(): # deprecated generate_table_images warnings.filterwarnings("ignore", category=DeprecationWarning) self.keep_images = ( self.pipeline_options.generate_page_images or self.pipeline_options.generate_picture_images or self.pipeline_options.generate_table_images ) self.reading_order_model = ReadingOrderModel(options=ReadingOrderOptions()) ocr_model = self.get_ocr_model(artifacts_path=self.artifacts_path) layout_factory = get_layout_factory( allow_external_plugins=self.pipeline_options.allow_external_plugins ) layout_model = layout_factory.create_instance( options=pipeline_options.layout_options, artifacts_path=self.artifacts_path, accelerator_options=pipeline_options.accelerator_options, ) table_factory = get_table_structure_factory( allow_external_plugins=self.pipeline_options.allow_external_plugins ) table_model = table_factory.create_instance( options=pipeline_options.table_structure_options, enabled=pipeline_options.do_table_structure, artifacts_path=self.artifacts_path, accelerator_options=pipeline_options.accelerator_options, ) self.build_pipe = [ # Pre-processing PagePreprocessingModel( options=PagePreprocessingOptions( images_scale=pipeline_options.images_scale, ) ), # OCR ocr_model, # Layout model layout_model, # Table structure model table_model, # Page assemble PageAssembleModel(options=PageAssembleOptions()), ] self.enrichment_pipe = [ # Code Formula Enrichment Model CodeFormulaModel( enabled=pipeline_options.do_code_enrichment or pipeline_options.do_formula_enrichment, 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, ), *self.enrichment_pipe, ] if ( self.pipeline_options.do_formula_enrichment or self.pipeline_options.do_code_enrichment or self.pipeline_options.do_picture_classification or self.pipeline_options.do_picture_description ): self.keep_backend = True @staticmethod def download_models_hf( local_dir: Optional[Path] = None, force: bool = False ) -> Path: warnings.warn( "The usage of LegacyStandardPdfPipeline.download_models_hf() is deprecated " "use instead the utility `docling-tools models download`, or " "the upstream method docling.utils.models_downloader.download_all()", DeprecationWarning, stacklevel=3, ) output_dir = download_models(output_dir=local_dir, force=force, progress=False) return output_dir def get_ocr_model(self, artifacts_path: Optional[Path] = None) -> BaseOcrModel: factory = get_ocr_factory( allow_external_plugins=self.pipeline_options.allow_external_plugins ) return factory.create_instance( options=self.pipeline_options.ocr_options, enabled=self.pipeline_options.do_ocr, 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 if page._backend is not None and page._backend.is_valid(): page.size = page._backend.get_size() return page def _assemble_document(self, conv_res: ConversionResult) -> ConversionResult: all_elements = [] all_headers = [] all_body = [] with TimeRecorder(conv_res, "doc_assemble", scope=ProfilingScope.DOCUMENT): for p in conv_res.pages: if p.assembled is not None: for el in p.assembled.body: all_body.append(el) for el in p.assembled.headers: all_headers.append(el) for el in p.assembled.elements: all_elements.append(el) conv_res.assembled = AssembledUnit( elements=all_elements, headers=all_headers, body=all_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 @classmethod def get_default_options(cls) -> PdfPipelineOptions: return PdfPipelineOptions() @classmethod def is_backend_supported(cls, backend: AbstractDocumentBackend): return isinstance(backend, PdfDocumentBackend)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/pipeline/asr_pipeline.py
docling/pipeline/asr_pipeline.py
import logging import os import re import sys import tempfile from io import BytesIO from pathlib import Path from typing import TYPE_CHECKING, 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 ( InlineAsrMlxWhisperOptions, 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: if sys.version_info < (3, 14): raise ImportError( "whisper is not installed. Please install it via `pip install openai-whisper` or do `uv sync --extra asr`." ) else: raise ImportError( "whisper is not installed. Unfortunately its dependencies are not yet available for Python 3.14." ) 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: # Access the file path from the backend, similar to how other pipelines handle it path_or_stream = conv_res.input._backend.path_or_stream # Handle both Path and BytesIO inputs temp_file_path: Optional[Path] = None if isinstance(path_or_stream, BytesIO): # For BytesIO, write to a temporary file since whisper requires a file path suffix = Path(conv_res.input.file.name).suffix or ".wav" with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tmp_file: tmp_file.write(path_or_stream.getvalue()) temp_file_path = Path(tmp_file.name) audio_path = temp_file_path elif isinstance(path_or_stream, Path): audio_path = path_or_stream else: raise RuntimeError( f"ASR pipeline requires a file path or BytesIO stream, but got {type(path_or_stream)}" ) 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() ) return conv_res except Exception as exc: _log.error(f"Audio tranciption has an error: {exc}") conv_res.status = ConversionStatus.FAILURE return conv_res finally: # Clean up temporary file if created if temp_file_path is not None and temp_file_path.exists(): try: temp_file_path.unlink() except Exception as e: _log.warning( f"Failed to delete temporary file {temp_file_path}: {e}" ) 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 _MlxWhisperModel: def __init__( self, enabled: bool, artifacts_path: Optional[Path], accelerator_options: AcceleratorOptions, asr_options: InlineAsrMlxWhisperOptions, ): """ Transcriber using MLX Whisper for Apple Silicon optimization. """ self.enabled = enabled _log.info(f"artifacts-path: {artifacts_path}") _log.info(f"accelerator_options: {accelerator_options}") if self.enabled: try: import mlx_whisper # type: ignore except ImportError: raise ImportError( "mlx-whisper is not installed. Please install it via `pip install mlx-whisper` or do `uv sync --extra asr`." ) self.asr_options = asr_options self.mlx_whisper = mlx_whisper self.device = decide_device( accelerator_options.device, supported_devices=asr_options.supported_devices, ) _log.info(f"Available device for MLX Whisper: {self.device}") self.model_name = asr_options.repo_id _log.info(f"loading _MlxWhisperModel({self.model_name})") # MLX Whisper models are loaded differently - they use HuggingFace repos self.model_path = self.model_name # Store MLX-specific options self.language = asr_options.language self.task = asr_options.task self.word_timestamps = asr_options.word_timestamps self.no_speech_threshold = asr_options.no_speech_threshold self.logprob_threshold = asr_options.logprob_threshold self.compression_ratio_threshold = asr_options.compression_ratio_threshold 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"MLX Audio transcription has an error: {exc}") conv_res.status = ConversionStatus.FAILURE return conv_res def transcribe(self, fpath: Path) -> list[_ConversationItem]: """ Transcribe audio using MLX Whisper. Args: fpath: Path to audio file Returns: List of conversation items with timestamps """ result = self.mlx_whisper.transcribe( str(fpath), path_or_hf_repo=self.model_path, language=self.language, task=self.task, word_timestamps=self.word_timestamps, no_speech_threshold=self.no_speech_threshold, logprob_threshold=self.logprob_threshold, compression_ratio_threshold=self.compression_ratio_threshold, ) convo: list[_ConversationItem] = [] # MLX Whisper returns segments similar to native Whisper for segment in result.get("segments", []): item = _ConversationItem( start_time=segment.get("start"), end_time=segment.get("end"), text=segment.get("text", "").strip(), words=[], ) # Add word-level timestamps if available if self.word_timestamps and "words" in segment: item.words = [] for word_data in segment["words"]: item.words.append( _ConversationWord( start_time=word_data.get("start"), end_time=word_data.get("end"), text=word_data.get("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 self._model: Union[_NativeWhisperModel, _MlxWhisperModel] if isinstance(self.pipeline_options.asr_options, InlineAsrNativeWhisperOptions): native_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=self.artifacts_path, accelerator_options=pipeline_options.accelerator_options, asr_options=native_asr_options, ) elif isinstance(self.pipeline_options.asr_options, InlineAsrMlxWhisperOptions): mlx_asr_options: InlineAsrMlxWhisperOptions = ( self.pipeline_options.asr_options ) self._model = _MlxWhisperModel( enabled=True, # must be always enabled for this pipeline to make sense. artifacts_path=self.artifacts_path, accelerator_options=pipeline_options.accelerator_options, asr_options=mlx_asr_options, ) else: _log.error(f"No model support for {self.pipeline_options.asr_options}") def _has_text(self, document: "DoclingDocument") -> bool: """ Helper method to check if the document contains any transcribed text. A transcription is considered non-empty if the .texts list contains items with actual, non whitespace content. """ if not document or not document.texts: return False for item in document.texts: if item.text and item.text.strip(): return True return False def _determine_status(self, conv_res: ConversionResult) -> ConversionStatus: """Determines the final status of ASR Conversion based on its result.""" if conv_res.status == ConversionStatus.FAILURE or conv_res.errors: return ConversionStatus.FAILURE if not self._has_text(conv_res.document): _log.warning( "ASR conversion resulted in an empty document." f"File: {conv_res.input.file.name}" ) return ConversionStatus.PARTIAL_SUCCESS return ConversionStatus.SUCCESS @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)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/pipeline/base_extraction_pipeline.py
docling/pipeline/base_extraction_pipeline.py
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, PipelineOptions from docling.datamodel.settings import settings _log = logging.getLogger(__name__) class BaseExtractionPipeline(ABC): 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, raises_on_error: bool, template: Optional[ExtractionTemplateType] = None, ) -> ExtractionResult: ext_res = ExtractionResult(input=in_doc) try: ext_res = self._extract_data(ext_res, template) ext_res.status = self._determine_status(ext_res) except Exception as e: ext_res.status = ConversionStatus.FAILURE error_item = ErrorItem( component_type="extraction_pipeline", module_name=self.__class__.__name__, error_message=str(e), ) ext_res.errors.append(error_item) if raises_on_error: raise e return ext_res @abstractmethod def _extract_data( self, ext_res: ExtractionResult, template: Optional[ExtractionTemplateType] = None, ) -> ExtractionResult: """Subclass must populate ext_res.pages/errors and return the result.""" raise NotImplementedError @abstractmethod def _determine_status(self, ext_res: ExtractionResult) -> ConversionStatus: """Subclass must decide SUCCESS/PARTIAL_SUCCESS/FAILURE based on ext_res.""" raise NotImplementedError @classmethod @abstractmethod def get_default_options(cls) -> PipelineOptions: pass
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/pipeline/base_pipeline.py
docling/pipeline/base_pipeline.py
import functools import logging import time import traceback from abc import ABC, abstractmethod from collections.abc import Iterable from pathlib import Path from typing import Any, Callable, List, Optional from docling_core.types.doc import NodeItem from docling.backend.abstract_backend import ( AbstractDocumentBackend, PaginatedDocumentBackend, ) from docling.backend.pdf_backend import PdfDocumentBackend from docling.datamodel.base_models import ( ConversionStatus, DoclingComponentType, ErrorItem, Page, ) from docling.datamodel.document import ConversionResult, InputDocument 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 _log = logging.getLogger(__name__) class BasePipeline(ABC): def __init__(self, pipeline_options: PipelineOptions): self.pipeline_options = pipeline_options self.keep_images = False 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) _log.info(f"Processing document {in_doc.file.name}") try: with TimeRecorder( conv_res, "pipeline_total", scope=ProfilingScope.DOCUMENT ): # These steps are building and assembling the structure of the # output DoclingDocument. conv_res = self._build_document(conv_res) conv_res = self._assemble_document(conv_res) # From this stage, all operations should rely only on conv_res.output conv_res = self._enrich_document(conv_res) conv_res.status = self._determine_status(conv_res) except Exception as e: conv_res.status = ConversionStatus.FAILURE if not raises_on_error: error_item = ErrorItem( component_type=DoclingComponentType.PIPELINE, module_name=self.__class__.__name__, error_message=str(e), ) conv_res.errors.append(error_item) else: raise RuntimeError(f"Pipeline {self.__class__.__name__} failed") from e finally: self._unload(conv_res) return conv_res @abstractmethod def _build_document(self, conv_res: ConversionResult) -> ConversionResult: pass def _assemble_document(self, conv_res: ConversionResult) -> ConversionResult: return conv_res def _enrich_document(self, conv_res: ConversionResult) -> ConversionResult: def _prepare_elements( conv_res: ConversionResult, model: GenericEnrichmentModel[Any] ) -> Iterable[NodeItem]: for doc_element, _level in conv_res.document.iterate_items(): prepared_element = model.prepare_element( conv_res=conv_res, element=doc_element ) if prepared_element is not None: yield prepared_element with TimeRecorder(conv_res, "doc_enrich", scope=ProfilingScope.DOCUMENT): for model in self.enrichment_pipe: for element_batch in chunkify( _prepare_elements(conv_res, model), model.elements_batch_size, ): for element in model( doc=conv_res.document, element_batch=element_batch ): # Must exhaust! pass return conv_res @abstractmethod def _determine_status(self, conv_res: ConversionResult) -> ConversionStatus: pass def _unload(self, conv_res: ConversionResult): pass @classmethod @abstractmethod def get_default_options(cls) -> PipelineOptions: pass @classmethod @abstractmethod def is_backend_supported(cls, backend: AbstractDocumentBackend): pass class ConvertPipeline(BasePipeline): def __init__(self, pipeline_options: ConvertPipelineOptions): super().__init__(pipeline_options) self.pipeline_options: ConvertPipelineOptions # ------ 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 def _apply_on_pages( self, conv_res: ConversionResult, page_batch: Iterable[Page] ) -> Iterable[Page]: for model in self.build_pipe: page_batch = model(conv_res, page_batch) yield from page_batch def _build_document(self, conv_res: ConversionResult) -> ConversionResult: if not isinstance(conv_res.input._backend, PaginatedDocumentBackend): raise RuntimeError( f"The selected backend {type(conv_res.input._backend).__name__} for {conv_res.input.file} is not a paginated backend. " f"Can not convert this with a paginated PDF pipeline. " f"Please check your format configuration on DocumentConverter." ) # conv_res.status = ConversionStatus.FAILURE # return conv_res total_elapsed_time = 0.0 with TimeRecorder(conv_res, "doc_build", scope=ProfilingScope.DOCUMENT): for i in range(conv_res.input.page_count): start_page, end_page = conv_res.input.limits.page_range if (start_page - 1) <= i <= (end_page - 1): conv_res.pages.append(Page(page_no=i)) try: total_pages_processed = 0 # Iterate batches of pages (page_batch_size) in the doc for page_batch in chunkify( conv_res.pages, settings.perf.page_batch_size ): start_batch_time = time.monotonic() # 1. Initialise the page resources init_pages = map( functools.partial(self.initialize_page, conv_res), page_batch ) # 2. Run pipeline stages pipeline_pages = self._apply_on_pages(conv_res, init_pages) for p in pipeline_pages: # Must exhaust! # Cleanup cached images if not self.keep_images: p._image_cache = {} # Cleanup page backends if not self.keep_backend and p._backend is not None: p._backend.unload() if ( isinstance(self.pipeline_options, PdfPipelineOptions) and not self.pipeline_options.generate_parsed_pages ): del p.parsed_page p.parsed_page = None end_batch_time = time.monotonic() total_elapsed_time += end_batch_time - start_batch_time if ( self.pipeline_options.document_timeout is not None and total_elapsed_time > self.pipeline_options.document_timeout ): _log.warning( f"Document processing time ({total_elapsed_time:.3f} seconds) exceeded the specified timeout of {self.pipeline_options.document_timeout:.3f} seconds" ) conv_res.status = ConversionStatus.PARTIAL_SUCCESS break total_pages_processed += len(page_batch) _log.debug( f"Finished converting pages {total_pages_processed}/{len(conv_res.pages)} time={end_batch_time:.3f}" ) except Exception as e: conv_res.status = ConversionStatus.FAILURE trace = "\n".join( traceback.format_exception(type(e), e, e.__traceback__) ) _log.warning( f"Encountered an error during conversion of document {conv_res.input.document_hash}:\n" f"{trace}" ) 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: for page in conv_res.pages: if page._backend is not None: page._backend.unload() if conv_res.input._backend: conv_res.input._backend.unload() return conv_res def _determine_status(self, conv_res: ConversionResult) -> ConversionStatus: status = conv_res.status if status in [ ConversionStatus.PENDING, ConversionStatus.STARTED, ]: # preserves ConversionStatus.PARTIAL_SUCCESS status = ConversionStatus.SUCCESS for page in conv_res.pages: if page._backend is None or not page._backend.is_valid(): conv_res.errors.append( ErrorItem( component_type=DoclingComponentType.DOCUMENT_BACKEND, module_name=type(page._backend).__name__, error_message=f"Page {page.page_no} failed to parse.", ) ) status = ConversionStatus.PARTIAL_SUCCESS return status # Initialise and load resources for a page @abstractmethod def initialize_page(self, conv_res: ConversionResult, page: Page) -> Page: pass
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
docling-project/docling
https://github.com/docling-project/docling/blob/7a4593f41f14ffce3871b11a92fdd56d391ffcd0/docling/pipeline/simple_pipeline.py
docling/pipeline/simple_pipeline.py
import logging from docling.backend.abstract_backend import ( AbstractDocumentBackend, DeclarativeDocumentBackend, ) from docling.datamodel.base_models import ConversionStatus from docling.datamodel.document import ConversionResult 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(ConvertPipeline): """SimpleModelPipeline. This class is used at the moment for formats / backends which produce straight DoclingDocument output. """ def __init__(self, pipeline_options: ConvertPipelineOptions): super().__init__(pipeline_options) def _build_document(self, conv_res: ConversionResult) -> ConversionResult: if not isinstance(conv_res.input._backend, DeclarativeDocumentBackend): raise RuntimeError( f"The selected backend {type(conv_res.input._backend).__name__} for {conv_res.input.file} is not a declarative backend. " f"Can not convert this with simple pipeline. " f"Please check your format configuration on DocumentConverter." ) # conv_res.status = ConversionStatus.FAILURE # return conv_res # Instead of running a page-level pipeline to build up the document structure, # the backend is expected to be of type DeclarativeDocumentBackend, which can output # a DoclingDocument straight. with TimeRecorder(conv_res, "doc_build", scope=ProfilingScope.DOCUMENT): conv_res.document = conv_res.input._backend.convert() return conv_res def _determine_status(self, conv_res: ConversionResult) -> ConversionStatus: # This is called only if the previous steps didn't raise. # Since we don't have anything else to evaluate, we can # safely return SUCCESS. return ConversionStatus.SUCCESS @classmethod def get_default_options(cls) -> ConvertPipelineOptions: return ConvertPipelineOptions() @classmethod def is_backend_supported(cls, backend: AbstractDocumentBackend): return isinstance(backend, DeclarativeDocumentBackend)
python
MIT
7a4593f41f14ffce3871b11a92fdd56d391ffcd0
2026-01-04T14:39:31.148558Z
false
psf/black
https://github.com/psf/black/blob/c3cc5a95d4f72e6ccc27ebae23344fce8cc70786/scripts/make_width_table.py
scripts/make_width_table.py
"""Generates a width table for Unicode characters. This script generates a width table for Unicode characters that are not narrow (width 1). The table is written to src/black/_width_table.py (note that although this file is generated, it is checked into Git) and is used by the char_width() function in src/black/strings.py. You should run this script when you upgrade wcwidth, which is expected to happen when a new Unicode version is released. The generated table contains the version of wcwidth and Unicode that it was generated for. In order to run this script, you need to install the latest version of wcwidth. You can do this by running: pip install -U wcwidth """ import sys from collections.abc import Iterable from os.path import basename, dirname, join import wcwidth # type: ignore[import-not-found] def make_width_table() -> Iterable[tuple[int, int, int]]: start_codepoint = -1 end_codepoint = -1 range_width = -2 for codepoint in range(0, sys.maxunicode + 1): width = wcwidth.wcwidth(chr(codepoint)) if width <= 1: # Ignore narrow characters along with zero-width characters so that # they are treated as single-width. Note that treating zero-width # characters as single-width is consistent with the heuristics built # on top of str.isascii() in the str_width() function in strings.py. continue if start_codepoint < 0: start_codepoint = codepoint range_width = width elif width != range_width or codepoint != end_codepoint + 1: yield (start_codepoint, end_codepoint, range_width) start_codepoint = codepoint range_width = width end_codepoint = codepoint if start_codepoint >= 0: yield (start_codepoint, end_codepoint, range_width) def main() -> None: table_path = join(dirname(__file__), "..", "src", "black", "_width_table.py") with open(table_path, "w") as f: f.write(f"""# Generated by {basename(__file__)} # wcwidth {wcwidth.__version__} # Unicode {wcwidth.list_versions()[-1]} from typing import Final WIDTH_TABLE: Final[list[tuple[int, int, int]]] = [ """) for triple in make_width_table(): f.write(f" {triple!r},\n") f.write("]\n") if __name__ == "__main__": main()
python
MIT
c3cc5a95d4f72e6ccc27ebae23344fce8cc70786
2026-01-04T14:40:23.735327Z
false
psf/black
https://github.com/psf/black/blob/c3cc5a95d4f72e6ccc27ebae23344fce8cc70786/scripts/migrate-black.py
scripts/migrate-black.py
#!/usr/bin/env python3 # check out every commit added by the current branch, blackify them, # and generate diffs to reconstruct the original commits, but then # blackified import logging import os import sys from subprocess import PIPE, Popen, check_output, run def git(*args: str) -> str: return check_output(["git", *args]).decode("utf8").strip() def blackify(base_branch: str, black_command: str, logger: logging.Logger) -> int: current_branch = git("branch", "--show-current") if not current_branch or base_branch == current_branch: logger.error("You need to check out a feature branch to work on") return 1 if not os.path.exists(".git"): logger.error("Run me in the root of your repo") return 1 merge_base = git("merge-base", "HEAD", base_branch) if not merge_base: logger.error( f"Could not find a common commit for current head and {base_branch}" ) return 1 commits = git( "log", "--reverse", "--pretty=format:%H", f"{merge_base}~1..HEAD" ).split() for commit in commits: git("checkout", commit, f"-b{commit}-black") check_output(black_command, shell=True) git("commit", "-aqm", "blackify") git("checkout", base_branch, f"-b{current_branch}-black") for last_commit, commit in zip(commits, commits[1:], strict=False): allow_empty = ( b"--allow-empty" in run(["git", "apply", "-h"], stdout=PIPE).stdout ) quiet = b"--quiet" in run(["git", "apply", "-h"], stdout=PIPE).stdout git_diff = Popen( [ "git", "diff", "--binary", "--find-copies", f"{last_commit}-black..{commit}-black", ], stdout=PIPE, ) git_apply = Popen( [ "git", "apply", ] + (["--quiet"] if quiet else []) + [ "-3", "--intent-to-add", ] + (["--allow-empty"] if allow_empty else []) + [ "-", ], stdin=git_diff.stdout, ) if git_diff.stdout is not None: git_diff.stdout.close() git_apply.communicate() git("commit", "--allow-empty", "-aqC", commit) for commit in commits: git("branch", "-qD", f"{commit}-black") return 0 if __name__ == "__main__": import argparse parser = argparse.ArgumentParser() parser.add_argument("base_branch") parser.add_argument("--black_command", default="black -q .") parser.add_argument("--logfile", type=argparse.FileType("w"), default=sys.stdout) args = parser.parse_args() logger = logging.getLogger(__name__) logger.addHandler(logging.StreamHandler(args.logfile)) logger.setLevel(logging.INFO) sys.exit(blackify(args.base_branch, args.black_command, logger))
python
MIT
c3cc5a95d4f72e6ccc27ebae23344fce8cc70786
2026-01-04T14:40:23.735327Z
false