import { d as IndexedDocument } from './types-DZE2UBNA.js';

/**
 * `@icjia/pdf-search-index/pagefind` — build-step helper for the
 * [Pagefind](https://pagefind.app/) search engine.
 *
 * **When to use this instead of `/fuse` or `/flexsearch`.** Pagefind is
 * fundamentally different from both: it operates on a chunked
 * on-demand index rather than an in-memory index loaded up-front. For
 * five-figure+ corpora it's the only viable option that doesn't pay
 * the full-index download cost on first load.
 *
 *   < 2,500 docs       → Fuse.js
 *   2,500 – 10,000 docs → FlexSearch
 *   **10,000+ docs**     → **Pagefind**
 *
 * **How Pagefind works.** Pagefind expects an HTML site to crawl. It
 * walks the rendered HTML, builds its inverted index, and partitions it
 * into chunks the browser loads only when a query needs them. The
 * resulting bundle on the wire is tiny (the `pagefind.js` client is
 * ~50 KB gz; individual query chunks are 5–20 KB each).
 *
 * **What this helper does.** Our package outputs JSON document rows.
 * Pagefind expects HTML pages. `emitPagefindHTML` is the bridge:
 * given `IndexedDocument[]` rows, it writes one HTML page per document
 * into a configured output directory. Each page contains the document's
 * extracted text inside a `<main data-pagefind-body>` wrapper. After
 * running this helper, the consumer invokes the Pagefind CLI against
 * the output dir, which produces the chunked index files.
 *
 * **Usage:**
 *
 * ```ts
 * // build step (Node)
 * import { emitPagefindHTML } from '@icjia/pdf-search-index/pagefind';
 * import { indexDocuments } from '@icjia/pdf-search-index';
 *
 * const rows = await indexDocuments(urls);
 * await emitPagefindHTML(rows, {
 *   outDir: 'public/pagefind-source',
 *   // Each row becomes pageFind-source/<id>.html.
 * });
 * ```
 *
 * Then add to `package.json` scripts:
 *
 * ```json
 * "build:search": "node build-index.mjs && pagefind --site public --output-subdir _pagefind"
 * ```
 *
 * Pagefind crawls `public/` (which includes the emitted
 * `pagefind-source/*.html`) and writes the chunked index to
 * `public/_pagefind/`. Consumer page loads that via the standard
 * `<script src="/_pagefind/pagefind.js">` Pagefind UI library.
 *
 * **Security.** The emitted HTML inlines the document text via
 * `safeJSONForHTML` — wait, HTML isn't JSON. We use direct HTML escape
 * so PDF text containing `<script>` etc. can't break out. Path-jail on
 * `outDir` is the caller's responsibility; pass an absolute or
 * project-relative path you trust.
 *
 * Added in 1.3.
 */

interface EmitPagefindHTMLOptions {
    /**
     * Directory to write the per-document HTML pages into. Must be inside
     * the same `public/` (or equivalent static-asset) directory that
     * Pagefind will crawl at build time. Relative paths resolve against
     * `process.cwd()`.
     */
    outDir: string;
    /**
     * Optional base URL to embed as a link on each emitted page so users
     * who land directly can navigate to the original document. If unset,
     * the row's `url` is used verbatim.
     */
    baseUrl?: string;
    /**
     * If set, restricts the outDir resolution to stay inside this jail
     * (typically the project's `public/` or `dist/` directory). Any path
     * that would escape the jail is rejected with a clear error. This is
     * analogous to the Astro adapter's `endpoint` jail (C5 fix from
     * 1.0.2). When unset, no jail check is applied — the caller is
     * trusted to pass a sane outDir.
     */
    publicDirJail?: string;
}
interface EmitPagefindHTMLResult {
    /** Absolute path to the output directory after resolution. */
    outDirAbs: string;
    /** Number of HTML pages emitted. */
    pagesEmitted: number;
    /** Filenames written (relative to outDir). */
    filenames: string[];
}
/**
 * Emit one HTML page per `IndexedDocument` row into `outDir`. After
 * this runs, the consumer invokes the Pagefind CLI against the parent
 * static-asset directory to produce the chunked index.
 *
 * Returns the resolved absolute outDir, the page count, and the
 * filename list (useful for debug / verification).
 */
declare function emitPagefindHTML(rows: readonly IndexedDocument[], options: EmitPagefindHTMLOptions): Promise<EmitPagefindHTMLResult>;

export { type EmitPagefindHTMLOptions, type EmitPagefindHTMLResult, IndexedDocument, emitPagefindHTML };