import type { DocumentSource, FileInput } from './index.js';
import type { FileLoaderOptions } from './text.js';
/**
 * Extracts a PDF's text: the whole document as one string, or one string per page. Wrap the parser
 * you already use — `pdfjs-dist`, `unpdf`, `pdf-parse`, or an OCR service — so no PDF library is a
 * dependency of this package.
 */
export type PdfTextExtractor = (bytes: Uint8Array, source: string) => Promise<string | readonly string[]> | string | readonly string[];
/** Options for `loadPdf()`. */
export interface PdfLoaderOptions extends FileLoaderOptions {
    /** Your PDF parser. */
    extract: PdfTextExtractor;
    /**
     * One document per page when the extractor returns pages, with `page` in the metadata, so a
     * citation can name the page. On by default; off joins the pages into one document.
     */
    splitPages?: boolean;
}
/** Raised when a file given to `loadPdf()` is not a PDF. */
export declare class PdfLoaderError extends Error {
    /** The file that failed. */
    readonly source: string;
    constructor(message: string, 
    /** The file that failed. */
    source: string);
}
/**
 * PDF files, through the parser you inject. A file that does not start with the PDF signature is
 * refused before it reaches the parser, so a mislabelled upload fails with a clear error.
 */
export declare function loadPdf(inputs: FileInput | Iterable<FileInput>, options: PdfLoaderOptions): AsyncGenerator<DocumentSource>;