/**
 * Document formats this package can extract text from. Added in 1.1:
 * `docx`, `pptx`, `xlsx` join the original `pdf`. All Office formats
 * are handled via the optional `officeparser` peer dependency.
 */
type DocumentFormat = 'pdf' | 'docx' | 'pptx' | 'xlsx';
interface IndexedDocument {
    id: string;
    url: string;
    title: string;
    text: string;
    /**
     * Source document format. For PDF this is `'pdf'`; for the Office
     * formats added in 1.1 it's `'docx'`, `'pptx'`, or `'xlsx'`. Optional
     * for back-compat with rows produced by 1.0.x (which were all PDF and
     * didn't carry a format field).
     */
    format?: DocumentFormat;
    /**
     * For PDFs: page count (populated). For DOCX/PPTX/XLSX: not surfaced
     * by `officeparser`'s text API, so `pages` is undefined for all
     * Office formats. (A future patch could re-read the ZIP to count
     * slides / sheets, but the cost is rarely worth it — consumers
     * who need it can call the ZIP inspector directly.)
     */
    pages?: number;
    extractedAt?: string;
}
/**
 * Back-compat alias. v1.0.x exported `IndexedPdf`; 1.1+ uses the
 * format-agnostic `IndexedDocument` name. The shape is identical.
 */
type IndexedPdf = IndexedDocument;
interface ExtractOptions {
    cacheDir?: string;
    fetchTimeout?: number;
    /**
     * Maximum number of bytes accepted from the document download. Defaults
     * to 32 MB. Responses with a declared `Content-Length` larger than this
     * are rejected before reading the body. For streaming responses without
     * a `Content-Length`, the read is aborted once the cap is exceeded.
     *
     * Applies to all formats (PDF, DOCX, PPTX, XLSX).
     */
    maxBytes?: number;
    /**
     * Maximum number of characters retained from the parsed document text.
     * Defaults to 5,000,000 (5 MB). Defends against compression-bomb-style
     * documents whose flate streams decompress to hundreds of megabytes.
     * Anything beyond the cap is truncated and a warning is logged.
     */
    maxExtractedTextChars?: number;
    /**
     * For Office formats (DOCX, PPTX, XLSX), the maximum total uncompressed
     * archive size accepted from the input. Defaults to 100 MB.
     *
     * The Office formats are ZIP archives of XML. A few-KB malicious input
     * can inflate to multi-GB on extraction, sitting between the existing
     * `maxBytes` (compressed input cap) and `maxExtractedTextChars`
     * (post-extraction text cap) — `officeparser` materializes the
     * inflated XML in memory before we see the text. This option closes
     * that window: the ZIP central directory is inspected up-front; if the
     * declared total uncompressed size exceeds the cap, the document is
     * rejected without invoking the parser.
     *
     * Set to `Infinity` to disable. Has no effect on PDFs (their
     * compression bound is `maxExtractedTextChars`).
     *
     * Added in 1.2. Closes the deferral from the 2026-05-17 v1.1 audit.
     */
    maxInflatedArchiveBytes?: number;
    fetch?: typeof fetch;
    cache?: 'use' | 'bypass' | 'refresh';
    mergePages?: boolean;
    /**
     * When `true`, parse and fetch errors include the full underlying error
     * message and the unscrubbed URL is printed in logs. When `false`
     * (default), only the URL origin and a categorized error tag are logged
     * — protecting internal URLs and document state from CI log enumeration.
     */
    debug?: boolean;
    /**
     * Override format detection (URL-extension + magic-byte sniff). When set,
     * the byte stream is parsed with the named extractor regardless of the
     * URL extension. Useful for URLs without a recognizable extension (e.g.
     * a CMS attachment URL like `/api/files/abc123`) when the caller knows
     * the format from out-of-band context.
     *
     * Added in 1.1.
     */
    format?: DocumentFormat;
}
interface IndexPdfsOptions extends ExtractOptions {
    concurrency?: number;
    /**
     * Maximum number of URLs accepted by a single `indexPdfs` /
     * `indexDocuments` call. Defaults to 5,000. Anything above the cap is
     * truncated and a warning is logged. Defends against a misconfigured
     * or malicious sitemap enqueueing millions of fetch requests.
     *
     * Set to `Infinity` to disable the cap. Most ICJIA-style sites
     * (research-heavy government CMS) fit well under 5,000.
     *
     * Added in 1.2 (closes I6 from the 1.0.2 audit).
     */
    maxUrls?: number;
}
/** Back-compat alias for `IndexPdfsOptions`. */
type IndexDocumentsOptions = IndexPdfsOptions;
type UrlOrEntry = string | {
    url: string;
    title?: string;
    id?: string;
};
interface DiscoveredDocument {
    url: string;
    title: string;
    format?: DocumentFormat;
}
/** Back-compat alias for `DiscoveredDocument`. */
type DiscoveredPdf = DiscoveredDocument;

export type { DiscoveredDocument as D, ExtractOptions as E, IndexDocumentsOptions as I, UrlOrEntry as U, DiscoveredPdf as a, DocumentFormat as b, IndexPdfsOptions as c, IndexedDocument as d, IndexedPdf as e };