/* tslint:disable */
/* eslint-disable */

/* auto-generated by NAPI-RS */

/** Config for a reusable [`DocumentConverter`]. */
export interface ConverterOptions {
  /**
   * Named Whisper model preset for audio sources (English-only /
   * Distil-Whisper variants under `.models/asr/<preset>/`).
   */
  asrModel?: string
  /**
   * ASR transcription language for audio/video: a Whisper code (`"en"`,
   * `"de"`, …) or `"auto"` (default) — detected from the first 30 seconds.
   */
  asrLang?: string
  /**
   * Character encoding of text inputs (Markdown, CSV, AsciiDoc, WebVTT,
   * XML, …) — docling's `TextBackendOptions.encoding`: a WHATWG label
   * (`"shift_jis"`, `"koi8-r"`, `"windows-1251"`). Unset = detect (BOM,
   * UTF-8, a statistical guess for a non-Latin encoding — #675 — else
   * windows-1252); bytes it cannot decode fail the conversion.
   */
  encoding?: string
  /**
   * Jupyter notebooks (#673): which cell outputs the document keeps —
   * `"none"` (default, the notebook as written) or `"text"` (stream and
   * `text/plain` results as code blocks after their cell, image outputs
   * as pictures).
   */
  ipynbOutputs?: string
  /**
   * Document metadata (#679): `"none"` (default, nowhere), `"json"` (a
   * `metadata` object in the docling JSON) or `"frontmatter"` (YAML
   * front matter at the top of the Markdown as well).
   */
  metadata?: string
  /**
   * Tracked changes in DOCX / DOC / ODT (#851, Pandoc's
   * `--track-changes`): `"accept"` (insertions shown, deletions left
   * out), `"reject"` (the reverse) or `"all"` (both, typed with author
   * and date in the Pandoc AST). Unset: each reader's own view. DOCX
   * reads all three, DOC accept and reject, ODT accept; a mode a reader
   * lacks is an error.
   */
  trackChanges?: string
  /**
   * DOCX DrawingML drawings (#858): `"mermaid"` (the default) adds a
   * `mermaid` code item after a drawing's picture with the flowchart its
   * labelled shapes and connectors describe — only when every connector
   * names its two end shapes; a drawing laid by eye gets no graph.
   * `"none"` turns the extraction off.
   */
  extractDiagrams?: string
  /**
   * Word (DOCX / DOC): a style's `w:outlineLvl` alone makes it a heading
   * (docling's `use_outline_level_for_headings`, docling#4480; default
   * `true`). `false` for templates whose body styles carry an outline
   * level only for the TOC.
   */
  useOutlineLevelForHeadings?: boolean
  /**
   * Max frames sampled from a video input as timestamped pictures (needs
   * the ffmpeg binary at runtime; `0` = transcript only). Default 8.
   * `4294967295` (`2**32 - 1`, the largest value this field takes) keeps
   * every distinct cut (#647).
   */
  videoFrames?: number
  /**
   * ffmpeg's scene score (0–1) a video frame must exceed to count as a
   * cut (#647). Default 0.27; 0.6 keeps hard cuts only.
   */
  videoSceneThreshold?: number
  /**
   * Downscale each sampled video frame inside ffmpeg so its longer side
   * is at most this many px (#647); `0` / unset = the source resolution.
   */
  videoFrameMaxSide?: number
  /**
   * Drop a sampled video frame whose 64-bit difference hash is within
   * this Hamming distance (0–64) of a frame already kept (#647): the same
   * slide after a fade becomes one picture. Unset = keep every frame;
   * 4–6 is a good distance.
   */
  videoFrameDedupe?: number
  /**
   * XBRL: the directory the instance's taxonomy is read from (docling's
   * `XBRLBackendOptions.taxonomy`); unset = the instance's own directory.
   */
  xbrlTaxonomy?: string
  /**
   * Convert only this PDF page window: `"A-B"` or a single page `"N"`
   * (1-based inclusive — issue #80). Other formats ignore it.
   */
  pages?: string
  /**
   * Per-document budget in seconds for the PDF pipeline (docling's
   * `document_timeout`, #497); unset = unlimited. Checked between pages:
   * once spent, the pages done so far are the document, `status` is
   * `"partial_success"` and `errors` says why.
   */
  documentTimeout?: number
  /**
   * Print progress to stderr (#783, docling's `show_progress`): a line
   * per document, a bar for its pages and each enrichment pass, drawn in
   * place on a terminal. Default `false`. For your own UI pass a
   * `progressCallback` to the async conversions instead.
   */
  showProgress?: boolean
  /**
   * OCR recognition language for scanned PDF/image pages: `"en"` (default;
   * proper Latin word spacing) or `"ch"` (the multilingual
   * docling-conformance model), or a BCP-47 tag for either language —
   * `"en-US"`, `"eng"`, `"zh"`, `"zh-Hans"`, `"zh-TW"`, docling's `iso:`
   * prefix accepted (#388); script and region subtags are ignored. Any
   * other language is an error. Formats that never OCR ignore it.
   */
  ocrLang?: string
  /**
   * Which regions feed the OCR (docling's `OcrMode`, #254): `"default"` |
   * `"full_page"` | `"layout_regions"` | `"pdf_aware_layout_regions"`.
   * `full_page`/`layout_regions` discard the text layer like
   * `forceFullPageOcr`.
   */
  ocrMode?: string
  /**
   * OCR render scale in px per PDF point (docling's `OcrOptions.scale`,
   * #254); unset reads the pipeline's own 2.0 px/pt render (docling's
   * default is 3 = 216 dpi).
   */
  ocrScale?: number
  /**
   * Where a PDF's text layer splits words drawn without a space glyph
   * (#921): `"adaptive"` (default, docling-parse ≥ 7's line-local
   * threshold; also `DOCLING_RS_PDF_WORD_GAP`) | `"legacy"`
   * (docling-parse ≤ 6) | a factor such as `"0.2"` of the two flanking
   * glyphs' mean advance, in (0, 4].
   */
  pdfWordGap?: string
  /**
   * Keep a PDF's typographic punctuation as drawn (#921) — `„ “ ‚ ‘`,
   * en/em dashes, `…` — instead of folding it to ASCII the way
   * docling-parse's sanitizer does (the default; also
   * `DOCLING_RS_PDF_KEEP_TYPOGRAPHY`).
   */
  pdfKeepTypography?: boolean
  /**
   * Report what the PDF reader repaired to read the input (#934) on the
   * result's `pdfRepairs` — a stream whose `/Length` misses its
   * `endstream`, a padded xref. Off by default (also
   * `DOCLING_RS_PDF_REPORT_REPAIRS`); the conversion is unchanged.
   */
  pdfReportRepairs?: boolean
  /**
   * RapidOCR's text-line orientation handling (#666): oriented line
   * crops, tall lines turned 90°, upside-down lines turned by the 0/180
   * classifier (`.models/ocr_textline_ori.onnx`). Unset = off unless
   * `DOCLING_RS_OCR_TEXTLINE_ORIENTATION` is set.
   */
  ocrTextlineOrientation?: boolean
  /**
   * Picture crops (and page images) in px per PDF point — docling's
   * `images_scale` (#520), 0.1–4.0; unset keeps the pipeline's 2.0 px/pt
   * render. The JSON picture `dpi` is 72·scale (#519).
   */
  imagesScale?: number
  /**
   * Keep each page's render as the JSON `pages[n].image` — docling's
   * `generate_page_images` (#520). Default `false`.
   */
  pageImages?: boolean
  /**
   * Which OCR engine reads scanned pages (#460): `"ppocr"` (default, the
   * built-in PP-OCRv3 recognizer) | `"tesseract"` (the system `tesseract`
   * binary). Under Tesseract `ocrLang` is its language list — tessdata
   * stems (`"deu+fra"`) or BCP-47 tags.
   */
  ocrEngine?: string
  /**
   * Email (.eml/.msg): append an Attachments section — names and content
   * types only, never the payload (#251). Default `false`.
   */
  listAttachments?: boolean
  /**
   * Omit empty cells from sparse XLSX/XLS table grids (#271; docling.rs
   * extension). Default `false`.
   */
  skipEmptyCells?: boolean
  /**
   * Unpadded `| a | b |` Markdown tables, all formats (#271; docling.rs
   * extension). Default `false`.
   */
  compactTables?: boolean
  /**
   * Render pictures' alt text (DOCX/PPTX/XLSX, #676) in Markdown and
   * HTML; `strict` implies it, JSON has it always. Default `false`.
   */
  pictureAltText?: boolean
  /**
   * EBCDIC (#252): copybook layout as inline `EbcdicLayout` JSON or a
   * file path; defaults to the `<stem>.layout.json` sidecar.
   */
  ebcdicLayout?: string
  /**
   * Keep layout + TableFormer, never OCR — docling's `do_ocr=False`, its
   * CLI's `--no-ocr` (#244; since 2.0 the meaning of `noOcr`, #611).
   * Structured output survives; text that exists only as pixels (scanned
   * pages, text inside images) comes back empty. Default `false`.
   */
  noOcr?: boolean
  /** `noOcr` under its pre-2.0 name, still read. */
  skipOcr?: boolean
  /**
   * Skip the whole PDF ML stack and read the embedded text layer only —
   * the CLI's `--text-layer-only` (what `noOcr` meant before 2.0, #611).
   * Default `false`. `"auto"` decides per document: the text layer alone
   * when it carries the document (no model touched), the full pipeline —
   * OCR included — when it does not.
   */
  textLayerOnly?: boolean | string
  /**
   * The password of an encrypted PDF or Office document — .docx/.xlsx/
   * .pptx/.doc/.xls/.ppt (#611, #625).
   */
  password?: string
  /**
   * `password` under its pre-#625 name (docling's PDF-only
   * `--pdf-password`), still read; set one of the two.
   */
  pdfPassword?: string
  /**
   * Skip TableFormer — tables come from the layout model's geometry
   * instead (the CLI's `--no-table-former`, #577). Default `false`.
   */
  noTableFormer?: boolean
  /**
   * OCR every PDF page even when it carries an embedded text layer
   * (docling's `force_full_page_ocr`) — for text layers that exist but lie.
   * Default `false`.
   */
  forceFullPageOcr?: boolean
  /**
   * Keep every detected picture as a picture: disable the demotion of
   * uncaptioned dense-text "picture" regions into paragraphs (the escape
   * hatch for image-extraction workflows, #173). Default `false`.
   */
  noTextPanels?: boolean
  /**
   * Infer PDF/image section-header levels after assembly (docling's
   * `HeadingHierarchyModel`, #302): PDF bookmarks are authoritative, then
   * legal/outline numbering, then font style. Default `false` — headings
   * then keep the flat detected level.
   */
  headingHierarchy?: boolean
  /**
   * Classify every PDF/image picture with DocumentFigureClassifier
   * (docling's `do_picture_classification`, #423): the 26-class prediction
   * lands on the JSON picture item's `classification`. Needs
   * `.models/picture_classifier.onnx`; a missing model warns and skips the
   * pass. Default `false`.
   */
  doPictureClassification?: boolean
  /**
   * Rewrite detected code blocks (and detect their language) with the
   * CodeFormulaV2 VLM (docling's `do_code_enrichment`, #423). Needs
   * `.models/code_formula/`; an autoregressive decode per code region —
   * seconds each on CPU. Default `false`.
   */
  doCodeEnrichment?: boolean
  /**
   * Decode display formulas to LaTeX with CodeFormulaV2 (docling's
   * `do_formula_enrichment`, #423): Markdown renders `$$latex$$` instead
   * of the formula placeholder comment. Same model and cost as
   * `doCodeEnrichment`. Default `false`.
   */
  doFormulaEnrichment?: boolean
  /**
   * OCR the pictures embedded in non-PDF documents — DOCX/PPTX
   * screenshots, HTML figures, sampled video frames — with the pipeline's
   * OCR models and attach the text to the picture as docling's description
   * annotation (#645): Markdown prints it between the caption and the
   * image placeholder, the JSON picture item carries `meta.description` +
   * the `description` annotation, chunks include it. Needs the OCR models
   * (`.models/ocr_rec*`, `ocr_det.onnx`; or Tesseract under `ocrEngine`);
   * with `skipOcr` / `textLayerOnly` it warns and reads nothing. PDF/image
   * pages are OCR'd by the pipeline itself. Default `false`.
   */
  doPictureOcr?: boolean
  /**
   * With `doPictureOcr`: only read the pictures the DocumentFigureClassifier
   * labels as one of these comma-separated classes
   * (`"screenshot_from_computer,screenshot_from_manual"`); unset = every
   * picture. Needs `.models/picture_classifier.onnx`.
   */
  pictureOcrClasses?: string
  /**
   * With `doPictureOcr`: skip pictures whose smaller side is under this
   * many pixels (icons, bullets). Default 32
   * (`DOCLING_RS_PICTURE_OCR_MIN_SIDE`).
   */
  pictureOcrMinSide?: number
  /**
   * The picture OCR's resolution in px per picture pixel under
   * RapidOCR's 2000 px cap (#666): `1.0` reads pictures at their native
   * pixels; unset keeps the image-input rule.
   */
  pictureOcrScale?: number
  /**
   * `false` drops the embedded image bytes from every picture after the
   * enrichment pass (#645): slim JSON/DCLX, placeholder-only Markdown, the
   * OCR text kept. Default `true`.
   */
  keepPictureImages?: boolean
  /**
   * Describe every picture — the embedded pictures of DOCX/PPTX/XLSX/
   * HTML, a PDF's figures, video frames, an image input as a whole —
   * through an OpenAI-compatible vision endpoint (#670); the answer is the
   * picture's description annotation, where `doPictureOcr` puts its text.
   * `pictureOcrMinSide` / `pictureOcrClasses` select the pictures here
   * too; identical images are sent once; an endpoint error warns and the
   * conversion goes on. Default `false`.
   */
  doPictureDescription?: boolean
  /**
   * The vision endpoint (`/v1` base or the full `chat/completions` URL);
   * falls back to `vlmEndpoint`, then `$DOCLING_RS_VLM_ENDPOINT`.
   */
  pictureDescriptionEndpoint?: string
  /** Model name; falls back to `vlmModel`, then `$DOCLING_RS_VLM_MODEL`. */
  pictureDescriptionModel?: string
  /** Bearer token; falls back to `vlmApiKey`, then `$DOCLING_RS_VLM_API_KEY`. */
  pictureDescriptionApiKey?: string
  /** The instruction sent with every picture; unset = the mode's prompt. */
  pictureDescriptionPrompt?: string
  /**
   * `"caption"` (default; MarkItDown's caption prompt) | `"ocr"`
   * (markitdown-ocr's text-extraction prompt).
   */
  pictureDescriptionMode?: string
  /** Pictures in flight at once (default 1). */
  pictureDescriptionConcurrency?: number
  /**
   * Redact personal data from the converted document before any export
   * (#621): e-mail, phone, card numbers (Luhn), IBANs (mod-97), IP
   * addresses, URL credentials, national IDs, and — with the NER model
   * under `.models/ner/` — names, organizations and locations; every
   * string of the document model is rewritten, so Markdown, JSON, DCLX
   * and chunks all come out clean. The result's `redaction` carries the
   * counts per label. A docling.rs extension; default `false`.
   */
  redactPii?: boolean
  /**
   * `"label"` (default: `[EMAIL]`) | `"pseudonym"` (`[EMAIL_1]`, one
   * number per distinct value) | `"fixed:<text>"`.
   */
  redactMode?: string
  /**
   * Comma-separated kinds to redact (`"email,phone,credit_card,iban,
   * ip_address,url_credentials,national_id,person,organization,location,
   * address"`); unset = every kind.
   */
  redactKinds?: string
  /**
   * Extra patterns as `NAME=REGEX` lines (newline-separated); the match
   * (or capture group 1) is redacted as `[NAME]`.
   */
  redactPattern?: string
  /**
   * Embedded images and page renders: `"drop"` (default) | `"box_out"`
   * (OCR and paint over the lines carrying a value) | `"keep"`.
   */
  redactImages?: string
  /**
   * With `redactPii` (#694): comma-separated terms never redacted even
   * when a detector flags them.
   */
  redactAllowTerms?: string
  /** Comma-separated terms always redacted, as `[REDACTED]`. */
  redactDenyTerms?: string
  /**
   * Regexes, one per line, a flagged span must not match in full to be
   * redacted — the document's own part numbers and record IDs.
   */
  redactAllowPatterns?: string
  /** The NER detector's minimum confidence, 0–1 (default 0.85). */
  redactNerMinScore?: number
  /**
   * `"standard"` (default) or `"vlm"` (#77): replace the whole ONNX stack —
   * layout, OCR, TableFormer — with a remote OpenAI-compatible vision
   * endpoint, which converts each rendered page on its own. PDF and image
   * inputs only; any other format is rejected rather than silently falling
   * back, matching the CLI's `--pipeline vlm`.
   *
   * Selecting it explicitly is deliberate: the `DOCLING_RS_VLM_*` variables
   * below are fallbacks, never triggers. A stale `DOCLING_RS_VLM_ENDPOINT`
   * left in the environment must not silently route every PDF over the
   * network.
   *
   * By the same rule the `vlm_*` options below are read **only** under
   * `"vlm"`. Passing them with the standard pipeline is not an error and
   * does not switch pipelines — they are ignored, exactly as the CLI drops
   * a stray `--vlm-endpoint` given without `--pipeline vlm`. They configure
   * the VLM; `pipeline` alone selects it.
   */
  pipeline?: string
  /**
   * VLM server: the base `/v1` URL or the full `…/chat/completions` one —
   * the suffix is appended when missing. Falls back to
   * `$DOCLING_RS_VLM_ENDPOINT`; required when `pipeline` is `"vlm"`.
   */
  vlmEndpoint?: string
  /**
   * VLM model name as the server knows it (e.g. `"granite-docling"`).
   * Falls back to `$DOCLING_RS_VLM_MODEL`; required when `pipeline` is `"vlm"`.
   */
  vlmModel?: string
  /**
   * Bearer token for the VLM endpoint; local servers (LM Studio, Ollama,
   * vLLM) need none. Falls back to `$DOCLING_RS_VLM_API_KEY`.
   */
  vlmApiKey?: string
  /**
   * Instruction sent with every page image, overriding docling's
   * DocLang-eliciting default. Falls back to `$DOCLING_RS_VLM_PROMPT`.
   */
  vlmPrompt?: string
  /**
   * `max_tokens` for each page completion. Default `8192` — a dense page of
   * DocLang runs long, and a truncated answer loses the tail of the page.
   */
  vlmMaxTokens?: number
  /**
   * Emit cleaner, more conformant Markdown (code-fence languages preserved,
   * no inline-run spacing artifacts) instead of docling's byte-for-byte
   * legacy output. Markdown only. Default `false`.
   */
  strict?: boolean
  /**
   * For HTML/EPUB/MHTML/JATS, resolve external `<img src>` (data: URIs, local
   * files, http(s) URLs, EPUB/MHTML archive parts, JATS `<graphic>` files)
   * and embed the bytes. Off by default; when on,
   * http(s) URLs are fetched over the network — enable only for trusted input.
   * The alias of `imageSources: "remote"` (#646).
   */
  fetchImages?: boolean
  /**
   * Which image references resolve (#646): `"none"` (default — every
   * picture a placeholder), `"embedded"` (`data:` URIs and parts of the
   * same container: EPUB/MHTML entries, an email's `cid:` attachments —
   * no filesystem, no network), `"local"` (plus files under the source
   * file's directory, never an absolute path or one that escapes it),
   * `"remote"` (plus http(s) fetches). HTML, EPUB, MHTML, JATS, AsciiDoc,
   * ODF, Markdown and email bodies.
   */
  imageSources?: string
  /**
   * Hosts a `"remote"` image fetch may reach (#646): exact names or
   * `*.suffix` wildcards; redirects are held to the same list. Unset =
   * any host.
   */
  imageHosts?: Array<string>
  /** Largest image that resolves, in bytes (#646; default 32 MiB). */
  maxImageBytes?: number
  /**
   * Images resolved per document; the rest stay placeholders (#646;
   * default unlimited).
   */
  maxImages?: number
  /**
   * Total resolved image bytes per document, in MiB (#646; default
   * unlimited).
   */
  maxImageTotalMb?: number
  /**
   * Smallest image that resolves, in bytes — skips spacer / tracking
   * pixels (#646; default 0).
   */
  minImageBytes?: number
  /**
   * Restrict the converter to these formats (ids like `"md"`, `"pdf"`, or
   * extensions like `".html"`); anything else is rejected. Default: accept all.
   */
  allowedFormats?: Array<string>
}
/** Per-call output options (how to render the converted document). */
export interface OutputOptions {
  /**
   * `"markdown"` (default), `"json"` (docling-core DoclingDocument wire
   * format), `"text"` (plain text, docling's `export_to_text`, #613),
   * `"latex"` (a complete LaTeX document, #317), `"html"` (a
   * complete HTML document, docling-core's `HTMLDocSerializer`, #492 —
   * pictures follow `imageMode` like Markdown) or `"pandoc"` (Pandoc's
   * JSON AST for `pandoc -f json`, #515 — pictures follow `imageMode`) or
   * `"vtt"` (WebVTT subtitles, docling's `--to vtt`, #614: a cue per timed
   * text item — an audio/video transcript's segments, a `.vtt` input's
   * cues).
   */
  to?: string
  /**
   * Picture handling for Markdown: `"placeholder"` (default; `"embedded"`
   * for `to: "pandoc"`, #537), `"embedded"` (base64 data URIs inline), or
   * `"referenced"` (returns image files in `images`). Ignored for JSON,
   * which always embeds images as data URIs.
   */
  imageMode?: string
  /** Directory name used in `referenced` image links. Default `"artifacts"`. */
  artifactsDir?: string
  /**
   * `to: "pandoc"` (#848): the `Header` depth a section heading gets —
   * `"docling"` (default: the title is `Header 1`, a level-n section
   * heading `Header n+1`, docling's Markdown / HTML depth) or `"source"`
   * (`Header n`: a DOCX `Heading 1` is `Header 1`, as `pandoc -f docx`
   * reads it).
   */
  pandocHeadingDepth?: string
  /**
   * Text inserted between pages in Markdown output — docling's
   * `export_to_markdown(page_break_placeholder=…)`, e.g.
   * `"<!-- page break -->"`. A break lands only between two rendered
   * blocks on different pages (PDF/image pages, slides, sheets, DjVu
   * pages) — never first or last, empty pages collapse. Unset: no page
   * breaks, docling's default. Markdown only.
   */
  pageBreakPlaceholder?: string
}
/**
 * All options for the one-shot module-level functions (converter config +
 * output options in a single object).
 */
export interface ConvertOptions {
  strict?: boolean
  fetchImages?: boolean
  /** Image-source tier (#646): `"none"` | `"embedded"` | `"local"` | `"remote"`. */
  imageSources?: string
  /** Remote image host allow-list (#646). */
  imageHosts?: Array<string>
  /** Largest image that resolves, in bytes (#646). */
  maxImageBytes?: number
  /** Images resolved per document (#646). */
  maxImages?: number
  /** Total resolved image bytes per document, in MiB (#646). */
  maxImageTotalMb?: number
  /** Smallest image that resolves, in bytes (#646). */
  minImageBytes?: number
  /** Named Whisper model preset for audio sources. */
  asrModel?: string
  /**
   * ASR transcription language for audio/video: a Whisper code (`"en"`,
   * `"de"`, …) or `"auto"` (default) — detected from the first 30 seconds.
   */
  asrLang?: string
  /**
   * Character encoding of text inputs (docling's
   * `TextBackendOptions.encoding`); unset = detect.
   */
  encoding?: string
  /**
   * Jupyter notebooks (#673): which cell outputs the document keeps —
   * `"none"` (default, the notebook as written) or `"text"` (stream and
   * `text/plain` results as code blocks after their cell, image outputs
   * as pictures).
   */
  ipynbOutputs?: string
  /**
   * Document metadata (#679): `"none"` (default), `"json"` or
   * `"frontmatter"`.
   */
  metadata?: string
  /**
   * Tracked changes (#851): `"accept"` | `"reject"` | `"all"`; see
   * [`ConverterOptions::track_changes`].
   */
  trackChanges?: string
  /**
   * Diagram extraction (#858): `"mermaid"` (default) | `"none"`; see
   * [`ConverterOptions::extract_diagrams`].
   */
  extractDiagrams?: string
  /**
   * Outline-level headings (docling#4480; default `true`); see
   * [`ConverterOptions::use_outline_level_for_headings`].
   */
  useOutlineLevelForHeadings?: boolean
  /**
   * Max frames sampled from a video input (`0` = transcript only;
   * `4294967295` = every distinct cut, #647).
   */
  videoFrames?: number
  /** Scene score (0–1) a video frame must exceed to be a cut (#647). */
  videoSceneThreshold?: number
  /** Cap on a sampled video frame's longer side in px (#647). */
  videoFrameMaxSide?: number
  /**
   * Difference-hash distance (0–64) under which a sampled video frame is
   * a duplicate and dropped (#647).
   */
  videoFrameDedupe?: number
  /** XBRL taxonomy directory (docling's `XBRLBackendOptions.taxonomy`). */
  xbrlTaxonomy?: string
  /** PDF page window `"A-B"` (or `"N"`), 1-based inclusive (#80). */
  pages?: string
  /**
   * Per-document budget in seconds (docling's `document_timeout`, #497);
   * see `ConverterOptions.documentTimeout`.
   */
  documentTimeout?: number
  /**
   * Print progress to stderr (#783, docling's `show_progress`): a line
   * per document, a bar for its pages and each enrichment pass, drawn in
   * place on a terminal. Default `false`. For your own UI pass a
   * `progressCallback` to the async conversions instead.
   */
  showProgress?: boolean
  /**
   * OCR recognition language for scanned pages: `"en"` (default) | `"ch"`,
   * or a BCP-47 tag for either (`"en-US"`, `"zh-Hans"`, #388).
   */
  ocrLang?: string
  /**
   * Which regions feed the OCR (docling's `OcrMode`, #254): `"default"` |
   * `"full_page"` | `"layout_regions"` | `"pdf_aware_layout_regions"`.
   */
  ocrMode?: string
  /**
   * OCR render scale in px per PDF point (docling's `OcrOptions.scale`,
   * #254); unset reads the pipeline's own 2.0 px/pt render.
   */
  ocrScale?: number
  /** The PDF text layer's word-gap rule (#921). See [`ConverterOptions`]. */
  pdfWordGap?: string
  /**
   * Keep a PDF's typographic punctuation as drawn (#921). See
   * [`ConverterOptions`].
   */
  pdfKeepTypography?: boolean
  /** Report the PDF reader's repairs (#934). See [`ConverterOptions`]. */
  pdfReportRepairs?: boolean
  /**
   * RapidOCR's text-line orientation handling (#666). See
   * [`ConverterOptions`].
   */
  ocrTextlineOrientation?: boolean
  /**
   * Picture crops (and page images) in px per PDF point — docling's
   * `images_scale` (#520), 0.1–4.0; unset keeps the pipeline's 2.0 px/pt
   * render. The JSON picture `dpi` is 72·scale (#519).
   */
  imagesScale?: number
  /**
   * Keep each page's render as the JSON `pages[n].image` — docling's
   * `generate_page_images` (#520). Default `false`.
   */
  pageImages?: boolean
  /**
   * Which OCR engine reads scanned pages (#460): `"ppocr"` (default) |
   * `"tesseract"`.
   */
  ocrEngine?: string
  /**
   * Email (.eml/.msg): append an Attachments section — names and content
   * types only, never the payload (#251). Default `false`.
   */
  listAttachments?: boolean
  /**
   * Omit empty cells from sparse XLSX/XLS table grids (#271). Default
   * `false`.
   */
  skipEmptyCells?: boolean
  /** Unpadded Markdown tables (#271). Default `false`. */
  compactTables?: boolean
  /** Pictures' alt text (#676) in Markdown and HTML. Default `false`. */
  pictureAltText?: boolean
  /**
   * EBCDIC (#252): copybook layout as inline `EbcdicLayout` JSON or a
   * file path; defaults to the `<stem>.layout.json` sidecar.
   */
  ebcdicLayout?: string
  /**
   * Keep layout + TableFormer, never OCR — docling's `do_ocr=False`, its
   * CLI's `--no-ocr` (#244; since 2.0 the meaning of `noOcr`, #611).
   * Structured output survives; text that exists only as pixels (scanned
   * pages, text inside images) comes back empty. Default `false`.
   */
  noOcr?: boolean
  /** `noOcr` under its pre-2.0 name, still read. */
  skipOcr?: boolean
  /**
   * Skip the whole PDF ML stack and read the embedded text layer only —
   * the CLI's `--text-layer-only` (what `noOcr` meant before 2.0, #611).
   * Default `false`. `"auto"` decides per document: the text layer alone
   * when it carries the document (no model touched), the full pipeline —
   * OCR included — when it does not.
   */
  textLayerOnly?: boolean | string
  /**
   * The password of an encrypted PDF or Office document — .docx/.xlsx/
   * .pptx/.doc/.xls/.ppt (#611, #625).
   */
  password?: string
  /**
   * `password` under its pre-#625 name (docling's PDF-only
   * `--pdf-password`), still read; set one of the two.
   */
  pdfPassword?: string
  /**
   * Skip TableFormer — tables come from the layout model's geometry
   * instead (the CLI's `--no-table-former`, #577). Default `false`.
   */
  noTableFormer?: boolean
  /**
   * OCR every PDF page even when it carries a text layer (docling's
   * `force_full_page_ocr`). Default `false`.
   */
  forceFullPageOcr?: boolean
  /**
   * Keep every detected picture as a picture: disable text-panel demotion
   * (#173). Default `false`.
   */
  noTextPanels?: boolean
  /**
   * Infer PDF/image section-header levels after assembly (#302). Default
   * `false`.
   */
  headingHierarchy?: boolean
  /**
   * Opt-in enrichment models (#423): picture classification, code rewrite
   * + language, formula LaTeX. See [`ConverterOptions`]. Default `false`.
   */
  doPictureClassification?: boolean
  doCodeEnrichment?: boolean
  doFormulaEnrichment?: boolean
  /**
   * Picture OCR for non-PDF documents (#645) and its filters. See
   * [`ConverterOptions`]. Defaults `false` / unset / 32 / `true`.
   */
  doPictureOcr?: boolean
  pictureOcrClasses?: string
  pictureOcrMinSide?: number
  pictureOcrScale?: number
  keepPictureImages?: boolean
  /**
   * Describe every picture — the embedded pictures of DOCX/PPTX/XLSX/
   * HTML, a PDF's figures, video frames, an image input as a whole —
   * through an OpenAI-compatible vision endpoint (#670); the answer is the
   * picture's description annotation, where `doPictureOcr` puts its text.
   * `pictureOcrMinSide` / `pictureOcrClasses` select the pictures here
   * too; identical images are sent once; an endpoint error warns and the
   * conversion goes on. Default `false`.
   */
  doPictureDescription?: boolean
  /**
   * The vision endpoint (`/v1` base or the full `chat/completions` URL);
   * falls back to `vlmEndpoint`, then `$DOCLING_RS_VLM_ENDPOINT`.
   */
  pictureDescriptionEndpoint?: string
  /** Model name; falls back to `vlmModel`, then `$DOCLING_RS_VLM_MODEL`. */
  pictureDescriptionModel?: string
  /** Bearer token; falls back to `vlmApiKey`, then `$DOCLING_RS_VLM_API_KEY`. */
  pictureDescriptionApiKey?: string
  /** The instruction sent with every picture; unset = the mode's prompt. */
  pictureDescriptionPrompt?: string
  /**
   * `"caption"` (default; MarkItDown's caption prompt) | `"ocr"`
   * (markitdown-ocr's text-extraction prompt).
   */
  pictureDescriptionMode?: string
  /** Pictures in flight at once (default 1). */
  pictureDescriptionConcurrency?: number
  /** PII redaction (#621) and its settings. See [`ConverterOptions`]. */
  redactPii?: boolean
  redactMode?: string
  redactKinds?: string
  redactPattern?: string
  redactImages?: string
  /**
   * With `redactPii` (#694): comma-separated terms never redacted even
   * when a detector flags them.
   */
  redactAllowTerms?: string
  /** Comma-separated terms always redacted, as `[REDACTED]`. */
  redactDenyTerms?: string
  /**
   * Regexes, one per line, a flagged span must not match in full to be
   * redacted — the document's own part numbers and record IDs.
   */
  redactAllowPatterns?: string
  /** The NER detector's minimum confidence, 0–1 (default 0.85). */
  redactNerMinScore?: number
  /**
   * `"standard"` (default) or `"vlm"` (#77): convert PDF/image pages
   * through a remote OpenAI-compatible vision endpoint instead of the ONNX
   * stack. The `vlm_*` options below take effect only under `"vlm"` and are
   * ignored otherwise. See [`ConverterOptions::pipeline`] for the full
   * contract.
   */
  pipeline?: string
  /**
   * VLM server, base `/v1` or full `…/chat/completions` URL. Falls back to
   * `$DOCLING_RS_VLM_ENDPOINT`.
   */
  vlmEndpoint?: string
  /** VLM model name. Falls back to `$DOCLING_RS_VLM_MODEL`. */
  vlmModel?: string
  /** Bearer token for the VLM endpoint. Falls back to `$DOCLING_RS_VLM_API_KEY`. */
  vlmApiKey?: string
  /**
   * Per-page instruction, overriding docling's default DocLang prompt.
   * Falls back to `$DOCLING_RS_VLM_PROMPT`.
   */
  vlmPrompt?: string
  /** `max_tokens` per page completion. Default `8192`. */
  vlmMaxTokens?: number
  allowedFormats?: Array<string>
  to?: string
  imageMode?: string
  artifactsDir?: string
  /** See [`OutputOptions::pandoc_heading_depth`] (#848). */
  pandocHeadingDepth?: string
  /** See [`OutputOptions::page_break_placeholder`]. */
  pageBreakPlaceholder?: string
}
/** In-memory input for [`DocumentConverter::convert`] / [`convert`]. */
export interface ConvertInput {
  /** Logical document name (used as the docling document name). */
  name: string
  /** Raw file bytes. */
  data: Buffer
  /**
   * Format id or extension (e.g. `"md"`, `"pdf"`, `".html"`). Omit to infer
   * from an extension on `name`.
   */
  format?: string
}
/** One extracted image file, returned for the `referenced` image mode. */
export interface ImageArtifact {
  /** Path relative to the Markdown file (e.g. `"artifacts/image_000000.png"`). */
  path: string
  /** The image bytes to write at `path`. */
  data: Buffer
}
/** The result of a conversion. */
export interface ConvertResult {
  /** The rendered document: Markdown or JSON, per `to`. */
  content: string
  /** Detected input format id (e.g. `"md"`, `"pdf"`). */
  format: string
  /** `"success"`, `"partial_success"`, or `"failure"`. */
  status: string
  /** The document name. */
  inputName: string
  /**
   * For the `referenced` image mode, the image files to write next to the
   * Markdown; empty otherwise.
   */
  images: Array<ImageArtifact>
  /**
   * The problems the conversion survived — docling's
   * `ConversionResult.errors`; non-empty exactly when `status` is
   * `"partial_success"` (a spent `documentTimeout`, #497; a VLM page
   * that failed or came back cut off, #762/#763).
   */
  errors: Array<ConversionErrorItem>
  /**
   * What the PII pass redacted (#621) when `redactPii` was on: counts
   * per label (`{ EMAIL: 2, PHONE: 1 }`); absent otherwise.
   */
  redaction?: Record<string, number>
  /**
   * What the PDF reader repaired to read the input (#934) when
   * `pdfReportRepairs` was on and the input is a PDF — `[]` for a
   * well-formed file; absent otherwise.
   */
  pdfRepairs?: Array<PdfRepairItem>
}
/** One repair the PDF reader made to read the input (#934). */
export interface PdfRepairItem {
  /**
   * `"stream_length"` (a stream whose declared `/Length` does not end at
   * its `endstream`) | `"xref_entries_padded"`.
   */
  kind: string
  /** The stream's `[object number, generation]`. */
  object?: Array<number>
  /** Byte offset of the object's `N G obj` header. */
  offset?: number
  /** The declared `/Length`; absent when missing or unresolvable. */
  declared?: number
  /**
   * The recovered length; absent when the stream had no unambiguous end
   * and was dropped.
   */
  recovered?: number
}
/**
 * docling's `ErrorItem`: one recorded problem of a conversion that still
 * produced a document.
 */
export interface ConversionErrorItem {
  /**
   * `"document_backend"`, `"model"`, `"doc_assembler"`, `"user_input"`,
   * `"pipeline"`.
   */
  componentType: string
  /**
   * The stage that recorded it (`"pipeline"` for the document budget,
   * `"VlmPipeline"` for a VLM page).
   */
  moduleName: string
  errorMessage: string
  /**
   * The 1-based page the problem belongs to; absent when it is
   * document-scoped.
   */
  pageNo?: number
}
/**
 * One entry of a ZIP archive converted by [`convert_archive`] /
 * [`convert_archive_file`] (#557): its path inside the archive and what
 * became of it.
 */
export interface ArchiveItem {
  /** The entry's path inside the archive (`/`-separated). */
  path: string
  /**
   * `"converted"` (`result` is set), `"skipped"` (`error` says why:
   * unsupported type, nested archive, unsafe path, over a
   * `DOCLING_RS_ZIP_MAX_*` limit) or `"failed"` (`error` is the
   * conversion error; the other entries are unaffected).
   */
  outcome: string
  result?: ConvertResult
  error?: string
}
/**
 * Convert a file on disk. Detects the format from the extension and (for
 * HTML/EPUB/JATS image fetching) resolves relative `<img src>` / `<graphic>`
 * paths against the file's directory.
 */
export declare function convertFile(path: string, options?: ConvertOptions | undefined | null): ConvertResult
/** Convert in-memory bytes. */
export declare function convert(input: ConvertInput, options?: ConvertOptions | undefined | null): ConvertResult
/**
 * Async (Promise-returning) [`convert_file`]. The CPU-bound work runs on the
 * libuv thread pool, keeping the event loop free — use this for PDF/image.
 */
export declare function convertFileAsync(path: string, options?: ConvertOptions | undefined | null, progress?: (err: Error | null, arg: string) => any | undefined | null): Promise<ConvertResult>
/** Async (Promise-returning) [`convert`]. */
export declare function convertAsync(input: ConvertInput, options?: ConvertOptions | undefined | null, progress?: (err: Error | null, arg: string) => any | undefined | null): Promise<ConvertResult>
/**
 * Convert every document inside a ZIP archive on disk (#557): one
 * [`ArchiveItem`] per entry. Throws only when the file is not a readable ZIP.
 */
export declare function convertArchiveFile(path: string, options?: ConvertOptions | undefined | null): Array<ArchiveItem>
/**
 * Convert every document inside an in-memory ZIP archive (#557); `input.name`
 * names it, `input.format` is ignored.
 */
export declare function convertArchive(input: ConvertInput, options?: ConvertOptions | undefined | null): Array<ArchiveItem>
/** Async (Promise-returning) [`convert_archive_file`]. */
export declare function convertArchiveFileAsync(path: string, options?: ConvertOptions | undefined | null): Promise<Array<ArchiveItem>>
/** Async (Promise-returning) [`convert_archive`]. */
export declare function convertArchiveAsync(input: ConvertInput, options?: ConvertOptions | undefined | null): Promise<Array<ArchiveItem>>
/** Options for the chunk* functions. */
export interface ChunkOptions {
  /**
   * `"hierarchical"` (default): one chunk per document item, docling's
   * structure-driven chunker. `"hybrid"`: tokenization-aware refinement —
   * splits oversized chunks and merges undersized same-heading neighbours;
   * requires `tokenizer`.
   */
  chunker?: string
  /**
   * Path to a HuggingFace `tokenizer.json` (e.g. all-MiniLM-L6-v2's) for the
   * hybrid chunker's token counts. When omitted, falls back to
   * `.models/chunk/tokenizer.json` (populated by
   * `scripts/install/download_dependencies.sh`).
   */
  tokenizer?: string
  /**
   * The hybrid chunker's token budget per chunk. Default `256` (docling's
   * default for the MiniLM embedding model).
   */
  maxTokens?: number
  /**
   * Merge undersized peer chunks with the same headings (hybrid only).
   * Default `true`, matching docling.
   */
  mergePeers?: boolean
}
/** One chunk record — the analogue of docling's `DocChunk`. */
export interface Chunk {
  /** The chunk body (markdown-flavoured text, same as docling's `DocChunk.text`). */
  text: string
  /**
   * The heading path above the chunk, outermost first; absent for content
   * above any heading.
   */
  headings?: Array<string>
  /**
   * JSON-pointer refs of the document items the chunk was built from
   * (`"#/texts/12"`, `"#/tables/0"`, …).
   */
  docItems: Array<string>
  /**
   * The embedding-ready rendering: heading path + text, newline-joined
   * (docling's `chunker.contextualize(chunk)`).
   */
  contextualized: string
}
/**
 * Chunk a file on disk with docling's chunkers: convert it, then run the
 * hierarchical (default) or hybrid chunker over the document.
 */
export declare function chunkFile(path: string, options?: ChunkOptions | undefined | null): Array<Chunk>
/**
 * Async (Promise-returning) [`chunk_file`]; conversion + chunking run on the
 * libuv thread pool.
 */
export declare function chunkFileAsync(path: string, options?: ChunkOptions | undefined | null): Promise<Array<Chunk>>
/** Chunk in-memory bytes (same contract as [`convert`], then chunk). */
export declare function chunk(input: ConvertInput, options?: ChunkOptions | undefined | null): Array<Chunk>
/** Async (Promise-returning) [`chunk`]. */
export declare function chunkAsync(input: ConvertInput, options?: ChunkOptions | undefined | null): Promise<Array<Chunk>>
/**
 * Chunk an already-converted document, passed as docling-core JSON (the
 * `content` of a `convert*` call with `to: "json"`) — so a document converted
 * once (e.g. through the warm PDF `Pipeline`) can be chunked without
 * re-converting.
 */
export declare function chunkDocument(documentJson: string, options?: ChunkOptions | undefined | null): Array<Chunk>
/** Async (Promise-returning) [`chunk_document`]. */
export declare function chunkDocumentAsync(documentJson: string, options?: ChunkOptions | undefined | null): Promise<Array<Chunk>>
/**
 * Chunk a file and stream each chunk as the chunkers produce it — no
 * all-chunks array is materialized, and the first chunk reaches JS while the
 * rest of the document is still being chunked.
 *
 * `callback` is invoked as `(err, chunk)`: once per chunk with `chunk` a
 * `Chunk`, once with `chunk === null` at the end, or once with a non-null
 * `err` on failure. Prefer the `streamFileChunks` async-generator wrapper in
 * JS over calling this directly.
 */
export declare function chunkFileStreaming(path: string, callback: (err: Error | null, arg: Chunk | undefined | null) => any, options?: ChunkOptions | undefined | null): void
/**
 * Streaming [`chunk`]: chunk in-memory bytes, pushing each chunk through the
 * callback (same contract as [`chunk_file_streaming`]). Prefer the
 * `streamChunks` async-generator wrapper in JS.
 */
export declare function chunkStreaming(input: ConvertInput, callback: (err: Error | null, arg: Chunk | undefined | null) => any, options?: ChunkOptions | undefined | null): void
/**
 * Streaming [`chunk_document`]: chunk an already-converted docling-core JSON
 * document, pushing each chunk through the callback (same contract as
 * [`chunk_file_streaming`]). Prefer the `streamDocumentChunks`
 * async-generator wrapper in JS.
 */
export declare function chunkDocumentStreaming(documentJson: string, callback: (err: Error | null, arg: Chunk | undefined | null) => any, options?: ChunkOptions | undefined | null): void
/** One attachment of an `.eml` / Outlook `.msg`, from [`email_attachments`]. */
export interface EmailAttachment {
  /** Position among the message's attachments. */
  index: number
  /**
   * A safe file name, unique within the message (base name only, control
   * and bidi-format characters removed; `attachment-N` when the message
   * declares none, `<subject>.eml` for a forwarded message; a repeated
   * name gets `-2`, `-3`, …) — safe to write into one directory as it is.
   */
  name: string
  /** The declared media type (`type/subtype`), if any. */
  contentType?: string
  /**
   * The format id it converts as (`"pdf"`, `"docx"`, `"email"`, …), from
   * its extension, else its media type, else the bytes; absent when
   * `skipped` says why it will not convert.
   */
  format?: string
  /** Payload size in bytes (0 without a payload). */
  size: number
  /**
   * An image the message shows inline (part of the HTML body) rather than
   * a file to open; a PDF disposed `inline` is still `false` (#564).
   */
  inline: boolean
  /**
   * Why it is not converted, when it is not: no payload (a reference, an
   * OLE object), over a limit, a nested archive, an unsupported type.
   */
  skipped?: string
  /**
   * The payload, when kept: every attachment with a payload within the
   * limits — also an unsupported type or an archive (hand a `.zip` to
   * your own extractor). Absent over a limit or without a payload.
   */
  data?: Buffer
}
/**
 * Bounds on what [`email_attachments`] keeps, in bytes. Unset = the archive
 * defaults: 10 000 attachments, 256 MiB each, 1 GiB in all.
 */
export interface EmailAttachmentOptions {
  maxEntries?: number
  maxEntrySize?: number
  maxTotalSize?: number
}
/**
 * The attachments of an in-memory `.eml` / `.msg` with their payloads. A
 * forwarded message is an `.eml` entry whose `data` is the nested message.
 * Convert one with `convert({ name: att.name, data: att.data, format: att.format })`
 * — `format` carries what the extension cannot (a `scan.bin` sent as
 * `application/pdf`, a nameless sniffed part; #564).
 */
export declare function emailAttachments(input: ConvertInput, options?: EmailAttachmentOptions | undefined | null): Array<EmailAttachment>
/** [`email_attachments`] for a message file on disk. */
export declare function emailAttachmentsFile(path: string, options?: EmailAttachmentOptions | undefined | null): Array<EmailAttachment>
/**
 * Async (Promise-returning) [`email_attachments`]: a large `.msg` parses off
 * the event loop (#564).
 */
export declare function emailAttachmentsAsync(input: ConvertInput, options?: EmailAttachmentOptions | undefined | null): Promise<Array<EmailAttachment>>
/** Async (Promise-returning) [`email_attachments_file`]. */
export declare function emailAttachmentsFileAsync(path: string, options?: EmailAttachmentOptions | undefined | null): Promise<Array<EmailAttachment>>
/** The list of supported input format ids. */
export declare function supportedFormats(): Array<string>
/**
 * Detect a format id from a filename or extension (e.g. `"report.pdf"` →
 * `"pdf"`). Returns `null` for unknown extensions.
 */
export declare function formatFromName(name: string): string | null
/**
 * The format a buffer evidently is, by its content alone (#672 — the
 * engine's `InputFormat::detect`): the format id, or `null` for bytes
 * without a signature (prose, Markdown, CSV, random bytes). Signatures
 * only: PDF, the Office / ODF / EPUB / iWork packages, the OLE files,
 * RTF, HTML, the raster images, DjVu, audio and video containers,
 * WordPerfect, WebVTT, DocTags, SVG, AbiWord, XML, JSON and Jupyter
 * notebooks, MIME messages, LaTeX. What `convert({ name, data })` does
 * for a name with no or an unknown extension.
 */
export declare function detectFormat(data: Buffer): string | null
/**
 * A reusable converter. Holds config (strict / fetch-images / allowed formats)
 * so you can convert many documents without re-parsing options each time —
 * the analogue of the Rust `DocumentConverter`.
 */
export declare class DocumentConverter {
  constructor(options?: ConverterOptions | undefined | null)
  /** Convert a file on disk (sync). */
  convertFile(path: string, options?: OutputOptions | undefined | null): ConvertResult
  /** Convert in-memory bytes (sync). */
  convert(input: ConvertInput, options?: OutputOptions | undefined | null): ConvertResult
  /** Async (Promise-returning) file conversion (runs off the event loop). */
  convertFileAsync(path: string, options?: OutputOptions | undefined | null, progress?: (err: Error | null, arg: string) => any | undefined | null): Promise<ConvertResult>
  /** Async (Promise-returning) bytes conversion (runs off the event loop). */
  convertAsync(input: ConvertInput, options?: OutputOptions | undefined | null, progress?: (err: Error | null, arg: string) => any | undefined | null): Promise<ConvertResult>
  /**
   * Convert every document inside a ZIP archive on disk (#557), with this
   * converter's config per entry (sync).
   */
  convertArchiveFile(path: string, options?: OutputOptions | undefined | null): Array<ArchiveItem>
  /** Convert every document inside an in-memory ZIP archive (#557) (sync). */
  convertArchive(input: ConvertInput, options?: OutputOptions | undefined | null): Array<ArchiveItem>
  /** Async (Promise-returning) archive conversion from disk. */
  convertArchiveFileAsync(path: string, options?: OutputOptions | undefined | null): Promise<Array<ArchiveItem>>
  /** Async (Promise-returning) archive conversion from bytes. */
  convertArchiveAsync(input: ConvertInput, options?: OutputOptions | undefined | null): Promise<Array<ArchiveItem>>
  /**
   * Stream a file's Markdown in chunks, in document order, as conversion
   * progresses (the headline win for PDF, whose pages convert in parallel).
   *
   * `callback` is invoked as `(err, chunk)`: once per Markdown chunk with
   * `chunk` a string, once with `chunk === null` at the end, or once with a
   * non-null `err` on failure. Every image mode streams here, `referenced`
   * included (its links resolve against `artifactsDir`) — unlike the warm
   * [`Pipeline`]'s streaming, which rejects it. Prefer the
   * `streamFileMarkdown` async-generator wrapper in JS over calling this
   * directly.
   */
  convertFileStreaming(path: string, callback: (err: Error | null, arg: string | undefined | null) => any, options?: OutputOptions | undefined | null): void
}
/**
 * A reusable PDF/image pipeline that keeps the ONNX models (layout, OCR,
 * TableFormer) loaded across calls — the analogue of the Rust `Pipeline`. Use
 * this instead of the per-call `convertFile` when converting many PDFs/images:
 * the one-shot functions rebuild the pipeline (reloading every model) each
 * call, whereas this loads them once.
 *
 * Handles `pdf` and `image` inputs (the ML pipeline). Models load lazily on
 * first use, so constructing a `Pipeline` is cheap; the first conversion pays
 * the model-load cost. Synchronous and single-threaded — reuse one instance
 * for a sequence of documents (e.g. behind a job queue).
 */
export declare class Pipeline {
  /**
   * Construct the pipeline. `strict` (cleaner Markdown), the three
   * enrichment switches (`doPictureClassification`, `doCodeEnrichment`,
   * `doFormulaEnrichment`, #423) and every PDF/image option the one-shot
   * calls honour — `ocrEngine`, `ocrLang`, `ocrMode`, `ocrScale`, `pdfWordGap`,
   * `pdfKeepTypography`,
   * `noOcr`, `textLayerOnly`, `password`, `forceFullPageOcr`,
   * `noTextPanels`, `headingHierarchy`,
   * `pages` — are read here and apply to every conversion on this
   * instance, validated exactly as `DocumentConverter` validates them
   * (#471; before, only `strict` and the enrichment switches were read
   * and a `por+eng` Tesseract pipeline silently ran PP-OCR English).
   * `fetchImages` / `allowedFormats` and the non-PDF options don't apply
   * to the PDF/image pipeline.
   */
  constructor(options?: ConverterOptions | undefined | null)
  /** Convert a PDF or image file, reusing the warm models. */
  convertFile(path: string, options?: OutputOptions | undefined | null): ConvertResult
  /** Convert PDF or image bytes, reusing the warm models. */
  convert(input: ConvertInput, options?: OutputOptions | undefined | null): ConvertResult
  /**
   * Async (Promise-returning) file conversion on the warm pipeline. The
   * CPU-bound work runs on the libuv thread pool, keeping the event loop
   * free; calls on the same instance run one at a time (the models are
   * mutable sessions), so overlapping Promises queue in submission order.
   */
  convertFileAsync(path: string, options?: OutputOptions | undefined | null, progress?: (err: Error | null, arg: string) => any | undefined | null): Promise<ConvertResult>
  /** Async (Promise-returning) bytes conversion on the warm pipeline. */
  convertAsync(input: ConvertInput, options?: OutputOptions | undefined | null, progress?: (err: Error | null, arg: string) => any | undefined | null): Promise<ConvertResult>
  /**
   * Stream a PDF's Markdown in chunks through the warm pipeline, in document
   * order, as pages finish converting (an image converts in one step and
   * arrives as a single chunk).
   *
   * `callback` is invoked as `(err, chunk)`: once per Markdown chunk with
   * `chunk` a string, once with `chunk === null` at the end, or once with a
   * non-null `err` on failure. Only `placeholder` / `embedded` image modes
   * stream; `referenced` is rejected. Prefer the `streamFileMarkdown`
   * async-generator wrapper in JS over calling this directly.
   */
  convertFileStreaming(path: string, callback: (err: Error | null, arg: string | undefined | null) => any, options?: OutputOptions | undefined | null): void
}