/**
 * PDF text extraction.
 *
 * Wraps `unpdf` (pure-ESM, zero native deps) behind a narrow, well-typed
 * boundary so the Read tool never touches pdfjs directly. Swapping the
 * backend later is a one-file change.
 *
 * Responsibilities:
 *   - Parse the caller's `pages` spec and clamp it to the document and
 *     the per-call page cap.
 *   - Extract per-page text.
 *   - Detect "no extractable text" (scanned / image-only PDFs) and
 *     return a structured signal rather than silently-empty output.
 *   - Render the extracted text with `[Page N]` markers so the caller
 *     can line-number it exactly like any other file content.
 *
 * Pure-ish: does no filesystem I/O. Callers are expected to have
 * already loaded the PDF bytes.
 */
export interface PdfExtractionRequest {
    /** PDF bytes. Accepts a Node Buffer or any Uint8Array. */
    data: Buffer | Uint8Array;
    /**
     * Page spec: `"N"`, `"N-M"`, or undefined. When undefined, extracts
     * the first `maxPages` pages starting at page 1.
     */
    pagesSpec?: string | undefined;
    /** Upper bound on how many pages may be extracted in one call. */
    maxPages?: number;
}
export interface PdfExtractionOk {
    kind: 'ok';
    totalPages: number;
    /** First page extracted (1-based, inclusive). */
    firstPage: number;
    /** Last page extracted (1-based, inclusive). */
    lastPage: number;
    /** Per-page text. `pages[i].pageNumber` is 1-based. */
    pages: Array<{
        pageNumber: number;
        text: string;
    }>;
    /**
     * Full rendered text with `[Page N]` markers separating pages, ready
     * to be handed to the same line-numbering pipeline Read uses for
     * plain text files.
     */
    rendered: string;
}
export interface PdfExtractionEmpty {
    kind: 'empty';
    totalPages: number;
    firstPage: number;
    lastPage: number;
    message: string;
}
export interface PdfExtractionInvalidRange {
    kind: 'invalid-range';
    totalPages: number;
    message: string;
}
export interface PdfExtractionError {
    kind: 'error';
    message: string;
}
export type PdfExtractionResult = PdfExtractionOk | PdfExtractionEmpty | PdfExtractionInvalidRange | PdfExtractionError;
/**
 * Default cap on pages extracted per call. Matches the advertised
 * schema default for Read's `pages` param ("max 20 pages per request").
 */
export declare const DEFAULT_MAX_PAGES = 20;
export type PagesSpecResult = {
    ok: true;
    first: number;
    last: number;
} | {
    ok: false;
    message: string;
};
/**
 * Parse a pages-spec string against a document's page count and the
 * per-call cap. Enforces:
 *   - Format: `"N"` or `"N-M"` (base 10, whitespace tolerant)
 *   - 1 <= first <= last <= totalPages
 *   - (last - first + 1) <= maxPages
 *
 * Returns `{ ok: false }` with an actionable message on any failure.
 */
export declare function parsePagesSpec(spec: string, totalPages: number, maxPages: number): PagesSpecResult;
/**
 * Extract text from a PDF buffer. Never throws — all failure modes are
 * returned as structured results so the caller can render them as
 * tool-content messages.
 */
export declare function extractPdfText(req: PdfExtractionRequest): Promise<PdfExtractionResult>;
//# sourceMappingURL=pdf-extractor.d.ts.map