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

/* auto-generated by NAPI-RS */

export interface JsLiteParseConfig {
  /** OCR language code (e.g., "eng", "fra"). */
  ocrLanguage?: string
  /** Whether OCR is enabled. */
  ocrEnabled?: boolean
  /** HTTP OCR server URL. If set, uses HTTP OCR instead of Tesseract. */
  ocrServerUrl?: string
  /**
   * Extra HTTP headers sent with every request to `ocrServerUrl`
   * (e.g. `{ Authorization: "Bearer <token>" }`).
   */
  ocrServerHeaders?: Record<string, string>
  /** Path to tessdata directory for Tesseract. */
  tessdataPath?: string
  /** Maximum number of pages to parse. */
  maxPages?: number
  /** Specific pages to parse (e.g., "1-5,10,15-20"). */
  targetPages?: string
  /** DPI for rendering pages (used for OCR and screenshots). */
  dpi?: number
  /** Output format: "json", "text", or "markdown". */
  outputFormat?: string
  /** Keep very small text that would normally be filtered out. */
  preserveVerySmallText?: boolean
  /** Password for encrypted/protected documents. */
  password?: string
  /** Suppress progress output. */
  quiet?: boolean
  /** Number of concurrent OCR workers (default: CPU cores - 1). */
  numWorkers?: number
  /**
   * How to surface raster images in markdown output: "off", "placeholder"
   * (default — emits `![](image_pN_K.png)` references with no bytes), or
   * "embed" (also returns each image's PNG bytes on `images`).
   */
  imageMode?: string
  /**
   * Render hyperlink annotations as `[text](url)` in markdown output
   * (default true). Set false for plain anchor text.
   */
  extractLinks?: boolean
  /**
   * Whether a systemic OCR failure aborts the whole parse (default true).
   * Set false to keep already-recovered native text and return partial
   * results when OCR is unavailable, instead of rejecting.
   */
  ocrFailureFatal?: boolean
  /**
   * OCR request-hedging schedule (ms). Empty/unset = no hedging. Multiple
   * delays (e.g. `[0, 5000, 10000]`) fire duplicate requests per attempt and
   * take the first success — lower tail latency at the cost of extra load.
   */
  ocrHedgeDelaysMs?: Array<number>
  /**
   * Emit per-word sub-boxes on each text item (`TextItem.words`). Default
   * false. Word boxes roughly double the text-item payload, so enable only
   * for word-level bbox attribution.
   */
  emitWordBoxes?: boolean
  /**
   * Restrict output to a page sub-region. Each field is the fraction of the
   * page cropped from that side; a text item survives only if it lies
   * entirely inside the remaining rectangle. Unset keeps the whole page.
   */
  cropBox?: JsCropBox
  /**
   * Drop diagonal text (rotation >2° off the nearest right angle). Default
   * false. Use to exclude rotated watermarks/stamps from the output.
   */
  skipDiagonalText?: boolean
}
/**
 * A page sub-region as the fraction cropped from each side (top-left origin,
 * each in `[0, 1]`).
 */
export interface JsCropBox {
  top: number
  right: number
  bottom: number
  left: number
}
/** One word's sub-box within a `JsTextItem`, in the same viewport space. */
export interface JsWordBox {
  text: string
  x: number
  y: number
  width: number
  height: number
}
export interface JsTextItem {
  text: string
  x: number
  y: number
  width: number
  height: number
  fontName?: string
  fontSize?: number
  confidence?: number
  /** Rotation in degrees (viewport space). Defaults to 0 when omitted. */
  rotation?: number
  /**
   * Per-word sub-boxes for attribution. Empty for items with no word split
   * (e.g. OCR-sourced or single-token items).
   */
  words: Array<JsWordBox>
}
/**
 * A vector-graphic primitive supplied by an external extractor. `kind` selects
 * the variant: `"stroke"` (uses `x1/y1/x2/y2`) or `"rect"` (uses
 * `x/y/width/height`). Coordinates are viewport space (top-left origin, 72
 * DPI), matching the text items. `has_fill`/`has_stroke` carry the paint
 * intent even when no color is known, so ruled-table edge detection still
 * treats a colorless stroked rect as stroked.
 */
export interface JsGraphic {
  /** "stroke" or "rect". Anything else is dropped. */
  kind: string
  x1?: number
  y1?: number
  x2?: number
  y2?: number
  x?: number
  y?: number
  width?: number
  height?: number
  /** Whether the path is filled. Drives Rect `fill` presence. */
  hasFill?: boolean
  /** Whether the path is stroked. Drives Rect `stroke` presence. */
  hasStroke?: boolean
  /** Fill color as ARGB hex (e.g. "ff000000"). May be absent even when filled. */
  fillColor?: string
  /** Stroke color as ARGB hex. May be absent even when stroked. */
  strokeColor?: string
  /** Stroke line width in points. */
  lineWidth?: number
}
/**
 * A page of pre-extracted text supplied by an external extractor. Coordinates
 * are viewport space (top-left origin, 72 DPI). `graphics` enables ruled-table
 * and horizontal-rule detection; struct nodes are still unsupported on this
 * path, so tagged-heading detection remains unavailable until they are added.
 */
export interface JsPageInput {
  pageNumber: number
  pageWidth: number
  pageHeight: number
  textItems: Array<JsTextItem>
  graphics?: Array<JsGraphic>
}
export interface JsParsedPage {
  pageNum: number
  width: number
  height: number
  text: string
  markdown: string
  textItems: Array<JsTextItem>
}
export interface JsParseResult {
  pages: Array<JsParsedPage>
  text: string
  images: Array<JsExtractedImage>
}
export interface JsExtractedImage {
  id: string
  page: number
  format: string
  bytes: Buffer
}
export interface JsScreenshotResult {
  pageNum: number
  width: number
  height: number
  imageBuffer: Buffer
}
export interface JsPageComplexityStats {
  pageNumber: number
  textLength: number
  textCoverage: number
  hasSubstantialImages: boolean
  imageBlockCount: number
  imageCoverage: number
  largestImageCoverage: number
  fullPageImage: boolean
  uncoveredVectorArea?: number
  isGarbled: boolean
  pageArea: number
  needsOcr: boolean
  reasons: Array<string>
}
/** Search text items for phrase matches, returning merged items with combined bounding boxes. */
export declare function searchItems(items: Array<JsTextItem>, phrase: string, caseSensitive?: boolean | undefined | null): Array<JsTextItem>
/** Main LiteParse parser class. */
export declare class LiteParse {
  /**
   * Create a new LiteParse instance with optional configuration.
   * Any fields not provided will use defaults.
   */
  constructor(config?: JsLiteParseConfig | undefined | null)
  /** Parse a document. Accepts a file path (string) or raw PDF bytes (Buffer). */
  parse(input: string | Buffer): Promise<JsParseResult>
  /**
   * Parse from pre-extracted pages, skipping PDFium text extraction.
   *
   * The caller supplies pages already populated with text items in viewport
   * space (top-left origin, 72 DPI). Runs only grid projection + the
   * configured output formatter, so it never loads PDFium. Use when an
   * external extractor owns text extraction (e.g. to keep its own
   * font-recovery pipeline).
   */
  parsePages(pages: Array<JsPageInput>): JsParseResult
  /**
   * Determine per-page complexity. Returns one entry per parsed page with
   * signals (text coverage, images, garbled text, vector area) and a
   * `needsOcr` verdict — a cheap pre-OCR check to decide whether a document
   * needs advanced parsing. Accepts a file path (string) or raw PDF bytes.
   */
  isComplex(input: string | Buffer): Promise<Array<JsPageComplexityStats>>
  /**
   * Take screenshots of document pages. Returns PNG image buffers.
   *
   * Non-PDF files are automatically converted to PDF before rendering when
   * LibreOffice/ImageMagick are available.
   */
  screenshot(input: string | Buffer, pageNumbers?: Array<number> | undefined | null): Promise<Array<JsScreenshotResult>>
  /** Get the current configuration. */
  get config(): JsLiteParseConfig
}