docling API reference

176 public APIs from docling (docling-project/docling) — 104 classes, 16 functions, 56 methods. Signatures extracted by static analysis of the actual source.

Repository: docling-project/docling

Kind	Count
Classes	104
Functions	16
Methods	56

API list

classdocling.backend.abstract_backend.DeclarativeDocumentBackend

DeclarativeDocumentBackend.

classdocling.backend.abstract_backend.PaginatedDocumentBackend

PaginatedDocumentBackend.

classdocling.backend.boxnote_backend.BoxNoteDocumentBackend

Declarative backend for Box Notes (.boxnote files).

funcdocling.backend.docx.drawingml.utils.convert_with_libreoffice(input_path:str | Path, output_path:str | Path) -> None

Convert a DOCX/PPTX file to PDF via LibreOffice.

funcdocling.backend.docx.drawingml.utils.get_libreoffice_cmd(raise_if_unavailable:bool=False) -> Optional[str]

Return the libreoffice cmd and optionally test it.

classdocling.backend.docx.latex.omml.Pr

Properties element processor for OMML elements.

methoddocling.backend.docx.latex.omml.Pr.do_brk(elm:_Element) -> str

Process line break element.

methoddocling.backend.docx.latex.omml.Pr.do_common(elm:_Element) -> None

Process common property elements.

funcdocling.backend.docx.latex.omml.escape_latex(strs:str) -> str

Escape special LaTeX characters in a string.

funcdocling.backend.docx.latex.omml.get_val(key:str | None, default:str, store:dict | None=CHR) -> str

Get a value from a dictionary store or return the key/default.

funcdocling.backend.docx.latex.omml.load(stream:Any) -> Iterator[str]

Load and convert OMML elements from a stream.

funcdocling.backend.docx.latex.omml.load_string(string:str) -> Iterator[str]

Load and convert OMML elements from a string.

classdocling.backend.docx.latex.omml.oMath2Latex

Convert OMML oMath elements to LaTeX format.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_acc(elm:_Element) -> str

Process accent element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_bar(elm:_Element) -> str

Process bar element (overline/underline).

methoddocling.backend.docx.latex.omml.oMath2Latex.do_d(elm:_Element) -> str

Process delimiter element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_eqarr(elm:_Element) -> str

Process equation array element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_f(elm:_Element) -> str

Process fraction element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_fname(elm:_Element) -> str

Process function name element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_func(elm:_Element) -> str

Process function application element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_lim(elm:_Element) -> str

Process limit content element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_limlow(elm:_Element) -> str

Process lower limit element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_limupp(elm:_Element) -> str

Process upper limit element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_m(elm:_Element) -> str

Process matrix element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_mr(elm:_Element) -> str

Process matrix row element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_nary(elm:_Element) -> str

Process n-ary operator element (e.g., sum, product, integral).

methoddocling.backend.docx.latex.omml.oMath2Latex.do_r(elm:_Element) -> str

Process run element containing text.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_rad(elm:_Element) -> str

Process radical (root) element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_spre(elm:_Element) -> None

Process pre-sub-superscript element (not supported yet).

methoddocling.backend.docx.latex.omml.oMath2Latex.do_ssub(elm:_Element) -> str

Process subscript element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_ssubsup(elm:_Element) -> str

Process combined sub-superscript element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_ssup(elm:_Element) -> str

Process superscript element.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_sub(elm:_Element) -> str

Process standalone subscript content.

methoddocling.backend.docx.latex.omml.oMath2Latex.do_sup(elm:_Element) -> str

Process standalone superscript content.

methoddocling.backend.docx.latex.omml.oMath2Latex.latex() -> str

Get the LaTeX representation.

methoddocling.backend.docx.latex.omml.oMath2Latex.process_unicode(s:str) -> str

Process Unicode character and convert to LaTeX.

methoddocling.backend.docx.latex.omml.oMath2Latex.process_unknow(elm:_Element, stag:str) -> Any | None

Handle unknown element types.

classdocling.backend.image_backend.ImageDocumentBackend

Image-native backend that bypasses pypdfium2.

classdocling.backend.msexcel_backend.ExcelCell

Represents an Excel cell.

classdocling.backend.msexcel_backend.ExcelTable

Represents an Excel table on a worksheet.

classdocling.backend.noop_backend.NoOpBackend

A no-op backend that only validates input existence.

classdocling.backend.opendocument_backend.OdtDocumentBackend

Backend for OpenDocument Text (``.odt``) files.

funcdocling.backend.utils.image_resource_loader.validate_url_safety(url:str) -> None

Reject URLs that resolve to a non-public IP address.

classdocling.backend.webvtt_backend.WebVTTDocumentBackend

Declarative backend for WebVTT (.vtt) files.

classdocling.backend.xml.uspto_backend.PatentUspto

Parser of patent documents from the US Patent Office.

methoddocling.backend.xml.uspto_backend.PatentUspto.parse(patent_content:str) -> DoclingDocument | None

Parse a USPTO patent.

classdocling.backend.xml.uspto_backend.PatentUsptoAppV1.PatentHandler

SAX ContentHandler for patent documents.

classdocling.backend.xml.uspto_backend.PatentUsptoGrantAps.Field

Represent a field in a patent APS document.

classdocling.backend.xml.uspto_backend.PatentUsptoGrantAps.Section

Represent a section in a patent APS document.

classdocling.backend.xml.uspto_backend.PatentUsptoGrantV2.PatentHandler

SAX ContentHandler for patent documents.

classdocling.backend.xml.uspto_backend.PatentUsptoIce.PatentHandler

SAX ContentHandler for patent documents.

classdocling.datamodel.accelerator_options.AcceleratorDevice

Devices to run model inference

classdocling.datamodel.backend_options.BaseBackendOptions

Common options for all declarative document backends.

classdocling.datamodel.backend_options.EbcdicBackendOptions

Options specific to the EBCDIC backend.

classdocling.datamodel.backend_options.EbcdicField

A single fixed-width field of an EBCDIC record.

classdocling.datamodel.backend_options.EbcdicFieldType

Storage type of a field inside an EBCDIC record.

classdocling.datamodel.backend_options.EbcdicLayout

Parsing rules for an EBCDIC data file.

classdocling.datamodel.backend_options.EbcdicRecordLayout

Field layout of one record schema, i.e.

classdocling.datamodel.backend_options.EpubBackendOptions

Options specific to the EPUB backend.

classdocling.datamodel.backend_options.HTMLBackendOptions

Options specific to the HTML backend.

classdocling.datamodel.backend_options.LatexBackendOptions

Options specific to the LaTeX backend.

classdocling.datamodel.backend_options.MarkdownBackendOptions

Options specific to the Markdown backend.

classdocling.datamodel.backend_options.MetsGbsBackendOptions

Options specific to the METS-GBS document backend.

classdocling.datamodel.backend_options.MsExcelBackendOptions

Options specific to the MS Excel backend.

classdocling.datamodel.backend_options.MsPowerpointBackendOptions

Options specific to the MS PowerPoint backend.

classdocling.datamodel.backend_options.MsWordBackendOptions

Options specific to the MS Word backend.

classdocling.datamodel.backend_options.PdfBackendOptions

Backend options for pdf document backends.

classdocling.datamodel.backend_options.XBRLBackendOptions

Options specific to the XBRL backend.

classdocling.datamodel.base_models.ErrorItem

Structured error information from document conversion.

classdocling.datamodel.base_models.InputFormat

A document format supported by document backend parsers.

classdocling.datamodel.document.InputDocument

A document as an input of a Docling conversion.

funcdocling.datamodel.document.build_invalid_input_errors(in_doc:'InputDocument') -> list[ErrorItem]

Build the ErrorItem list for an invalid input document.

classdocling.datamodel.extraction.ExtractedPageData

Data model for extracted content from a single page.

classdocling.datamodel.extraction.ExtractionResult

Result of document extraction.

classdocling.datamodel.pipeline_options.BaseLayoutOptions

Base options for document layout analysis models.

classdocling.datamodel.pipeline_options.BaseOptions

Base class for all pipeline option models.

classdocling.datamodel.pipeline_options.BaseTableStructureOptions

Base options for table structure extraction models.

classdocling.datamodel.pipeline_options.ConvertPipelineOptions

Base configuration for document conversion pipelines.

classdocling.datamodel.pipeline_options.EasyOcrOptions

Configuration for EasyOCR engine.

classdocling.datamodel.pipeline_options.LayoutOptions

Deprecated.

classdocling.datamodel.pipeline_options.LayoutPostprocessorOptions

Stage options for ``LayoutPostprocessingModel``.

classdocling.datamodel.pipeline_options.NemotronOcrOptions

Configuration for NVIDIA Nemotron OCR.

classdocling.datamodel.pipeline_options.OcrMode

How to generate the input for the OCR model

classdocling.datamodel.pipeline_options.PipelineOptions

Base configuration for document processing pipelines.

classdocling.datamodel.pipeline_options.TableStructureOptions

Options for the table structure (TableFormer V1).

classdocling.datamodel.pipeline_options.TableStructureV2Options

Options for the table structure (TableFormer V2).

classdocling.datamodel.pipeline_options.VideoPipelineOptions

Configuration options for the video pipeline.

classdocling.datamodel.pipeline_options.VlmConvertOptions

Configuration for VLM-based document conversion.

funcdocling.datamodel.pipeline_options.normalize_pdf_backend(backend:PdfBackend) -> PdfBackend

Normalize deprecated backend enum values to current ones.

classdocling.datamodel.pipeline_options_vlm_model.BaseVlmOptions

Base configuration for vision-language models.

method

docling.datamodel.pipeline_options_vlm_model.BaseVlmOptions.build_prompt(page:Optional['SegmentedPage'], *_internal_page:Optional['Page']=None) -> str

Build the prompt for VLM inference.

classdocling.datamodel.service.chunking.ChunkerType

Choice of the chunkers available in Docling.

classdocling.datamodel.service.chunking.HierarchicalChunkerOptions

Configuration options for the HierarchicalChunker.

classdocling.datamodel.service.chunking.HybridChunkerOptions

Configuration options for the HybridChunker.

classdocling.datamodel.service.responses.ConfidenceScores

JSON-safe snapshot of docling confidence scores.

classdocling.datamodel.service.responses.ZipArchiveResult

Container for a zip archive of the conversion.

classdocling.datamodel.stage_model_specs.ApiModelConfig

API-specific model configuration.

methoddocling.datamodel.stage_model_specs.ApiModelConfig.merge_with(base_params:Dict[str, Any]) -> ApiModelConfig

Merge with base parameters.

classdocling.datamodel.stage_model_specs.EngineModelConfig

Engine-specific model configuration.

method

docling.datamodel.stage_model_specs.EngineModelConfig.merge_with(base_repo_id:str, base_revision:str='main') -> EngineModelConfig

Merge with base configuration.

classdocling.datamodel.stage_model_specs.ImageClassificationModelSpec

Specification for an image-classification model.

method

docling.datamodel.stage_model_specs.ImageClassificationModelSpec.get_repo_id(engine_type:ImageClassificationEngineType) -> str

Get repository ID for specific engine.

method

docling.datamodel.stage_model_specs.ImageClassificationModelSpec.get_revision(engine_type:ImageClassificationEngineType) -> str

Get revision for specific engine.

classdocling.datamodel.stage_model_specs.ObjectDetectionModelSpec

Specification for an object detection model.

methoddocling.datamodel.stage_model_specs.ObjectDetectionModelSpec.get_repo_id(engine_type:ObjectDetectionEngineType) -> str

Get repository ID for specific engine.

methoddocling.datamodel.stage_model_specs.ObjectDetectionModelSpec.get_revision(engine_type:ObjectDetectionEngineType) -> str

Get revision for specific engine.

classdocling.datamodel.stage_model_specs.StagePresetMixin

Mixin for stage options classes that support presets.

methoddocling.datamodel.stage_model_specs.StagePresetMixin.get_preset(preset_id:str) -> StageModelPreset

Get a specific preset.

methoddocling.datamodel.stage_model_specs.StagePresetMixin.list_preset_ids() -> List[str]

List all preset IDs for this stage.

methoddocling.datamodel.stage_model_specs.StagePresetMixin.list_presets() -> List[StageModelPreset]

List all presets for this stage.

methoddocling.datamodel.stage_model_specs.StagePresetMixin.register_preset(preset:StageModelPreset) -> None

classdocling.datamodel.stage_model_specs.VlmModelSpec

Specification for a VLM model.

methoddocling.datamodel.stage_model_specs.VlmModelSpec.get_api_params(engine_type:VlmEngineType) -> Dict[str, Any]

Get API parameters for a specific engine.

methoddocling.datamodel.stage_model_specs.VlmModelSpec.get_engine_config(engine_type:VlmEngineType) -> EngineModelConfig

Get EngineModelConfig for a specific engine type.

methoddocling.datamodel.stage_model_specs.VlmModelSpec.get_repo_id(engine_type:VlmEngineType) -> str

Get the repository ID for a specific engine.

methoddocling.datamodel.stage_model_specs.VlmModelSpec.get_revision(engine_type:VlmEngineType) -> str

Get the model revision for a specific engine.

methoddocling.datamodel.stage_model_specs.VlmModelSpec.is_engine_supported(engine_type:VlmEngineType) -> bool

Check if an engine is supported by this model.

classdocling.datamodel.vlm_engine_options.ApiVlmEngineOptions

Options for API-based VLM services.

classdocling.datamodel.vlm_engine_options.VllmCudaGraphMode

CUDA graph capture mode for the vLLM v1 engine.

classdocling.document_converter.LatexFormatOption

Format options for LaTeX documents.

classdocling.document_extractor.DocumentExtractor

Standalone extractor class.

classdocling.document_extractor.ExtractionFormatOption

Per-format configuration for extraction.

classdocling.exceptions.DocumentLoadError

A backend could not parse the input bytes into a document.

classdocling.models.base_layout_model.BaseLayoutModel

Shared interface for layout models.

classdocling.models.base_ocr_model.BaseOcrModel

Base class for all OCR models.

classdocling.models.base_table_model.BaseTableStructureModel

Shared interface for table structure models.

methoddocling.models.inference_engines.common.kserve_v2_client_base.KserveV2Client.close() -> None

Release client resources.

classdocling.models.inference_engines.common.kserve_v2_http.KserveV2InferResponse

KServe v2 infer response payload.

funcdocling.models.inference_engines.vlm._utils.normalize_image_to_pil(image:Union[Image.Image, np.ndarray]) -> Image.Image

Convert any image format to RGB PIL Image.

func

docling.models.inference_engines.vlm._utils.preprocess_image_batch(images:List[Union[Image.Image, np.ndarray]]) -> List[Image.Image]

Preprocess a batch of images to RGB PIL Images.

classdocling.models.inference_engines.vlm.base.BaseVlmEngine

Abstract base class for VLM inference engines.

methoddocling.models.inference_engines.vlm.base.BaseVlmEngine.cleanup() -> None

Clean up resources (optional).

methoddocling.models.inference_engines.vlm.base.BaseVlmEngine.predict(input_data:VlmEngineInput) -> VlmEngineOutput

Run inference on a single input.

method

docling.models.inference_engines.vlm.base.BaseVlmEngine.predict_batch(input_batch:List[VlmEngineInput]) -> List[VlmEngineOutput]

Run inference on a batch of inputs.

classdocling.models.inference_engines.vlm.base.BaseVlmEngineOptions

Base configuration for VLM inference engines.

classdocling.models.inference_engines.vlm.base.VlmEngineInput

Input to a VLM inference engine.

classdocling.models.inference_engines.vlm.base.VlmEngineOutput

Output from a VLM inference engine.

classdocling.models.inference_engines.vlm.base.VlmEngineType

Types of VLM inference engines available.

methoddocling.models.inference_engines.vlm.base.VlmEngineType.is_api_variant(engine_type:'VlmEngineType') -> bool

Check if an engine type is an API variant.

classdocling.models.inference_engines.vlm.mlx_engine.MlxVlmEngine

MLX engine for VLM inference on Apple Silicon.

methoddocling.models.inference_engines.vlm.mlx_engine.MlxVlmEngine.cleanup() -> None

Clean up model resources.

methoddocling.models.inference_engines.vlm.mlx_engine.MlxVlmEngine.initialize() -> None

Initialize the MLX model and processor.

method

docling.models.inference_engines.vlm.mlx_engine.MlxVlmEngine.predict_batch(input_batch:List[VlmEngineInput]) -> List[VlmEngineOutput]

Run inference on a batch of inputs.

classdocling.models.inference_engines.vlm.vllm_engine.VllmVlmEngine

vLLM engine for high-throughput VLM inference.

methoddocling.models.inference_engines.vlm.vllm_engine.VllmVlmEngine.cleanup() -> None

Clean up vLLM resources.

methoddocling.models.inference_engines.vlm.vllm_engine.VllmVlmEngine.initialize() -> None

Initialize the vLLM engine.

classdocling.models.stages.layout.layout_model.LayoutModel

Deprecated.

classdocling.models.stages.ocr.nemotron_ocr_model.NemotronOcrModel

Wrapper for Nvidia's nemotron-ocr-v2 model

classdocling.models.utils.generation_utils.GenerationStopper

Base interface for stopping logic.

classdocling.pipeline.simple_pipeline.SimplePipeline

SimpleModelPipeline.

classdocling.pipeline.standard_pdf_pipeline.ProcessingResult

Aggregated outcome of a pipeline run.

classdocling.pipeline.standard_pdf_pipeline.RunContext

Wiring for a single *execute* call.

classdocling.pipeline.standard_pdf_pipeline.ThreadedItem

Envelope that travels between pipeline stages.

classdocling.service_client._async_client.AsyncDoclingServiceClient

Native async client for docling-serve.

methoddocling.service_client.client.DoclingServiceClient.close() -> None

Release HTTP resources owned by this client.

classdocling.service_client.client.RawServiceResult

Typed wrapper for non-JSON result payloads.

classdocling.service_client.exceptions.DoclingServiceClientError

Base error for all client SDK failures.

classdocling.service_client.exceptions.ServiceError

Raised for non-retryable HTTP service errors.

classdocling.service_client.exceptions.TaskExecutionError

Raised when task-level orchestration fails.

classdocling.service_client.exceptions.TaskNotFoundError

Raised when a task id is unknown to the service.

classdocling.service_client.job.ConversionJob

Long-lived handle for a submitted docling-serve task.

classdocling.service_client.watchers.AsyncStatusWatcher

Protocol for async job status watchers.

classdocling.service_client.watchers.StatusWatcher

Protocol for job status watchers.

funcdocling.utils.code_language.detect_code_language(text:str, hint:str | None=None) -> CodeLanguageLabel

Detect the programming language of a code block.

classdocling.utils.layout_postprocessor.Interval

Helper class for sortable intervals.

methoddocling.utils.layout_postprocessor.IntervalTree.find_containing(point:float) -> set[int]

Find all intervals containing the point.

methoddocling.utils.layout_postprocessor.UnionFind.get_groups() -> dict[int, list[int]]

Returns groups as {root: [elements]}.

classdocling.utils.speaker_diarization.DiarizationResult

Output of speaker diarization.

classdocling.utils.speaker_diarization.SpeakerSegment

A time segment attributed to a single speaker.

func

docling.utils.speaker_diarization.diarize(wav_path:Path, num_speakers:int | None=None, accelerator_device:str='auto') -> DiarizationResult

Run speaker diarization on a WAV file.

funcdocling.utils.utils.create_file_hash(path_or_stream:Union[BytesIO, Path]) -> str

Create a stable page_hash of the path_or_stream of a file

classdocling.utils.video_frame_sampling.VideoFrame

A single sampled video frame with its timestamp.

classdocling.utils.video_frame_sampling.VideoFrameSamplingMode

Frame sampling strategy for the video pipeline.

classdocling.utils.video_frame_sampling.VideoScene

A contiguous time window treated as one scene.

func

docling.utils.visualization.draw_clusters(image:Image.Image, clusters:list[Cluster], scale_x:float, scale_y:float) -> None

Draw clusters on an image

funcdocling.utils.vlm_utils.strip_stop_strings(texts:list[str], stop_strings:list[str]) -> list[str]

Strip stop strings from decoded texts.

About this data

These signatures were extracted from the public source of docling-project/docling using Python's ast module. Argument names, default values, type annotations and return types are taken verbatim from the code. Implementation bodies are never stored. See how it works for details.

Back to all 805 libraries