/**
 * Dependency-free HTML → readable Markdown-ish text (RCL-01).
 *
 * WHY NO DEPENDENCY — the capture path has to work in an `npx monomind`
 * install with nothing extra resolved, and the extension already runs
 * Readability against the live DOM. What lands here is the fallback for an
 * archive nobody cleaned: a small, predictable extractor beats a large one we
 * cannot install.
 *
 * WHY BOILERPLATE STRIPPING IS THE POINT — a site's nav, header, footer and
 * cookie banner are byte-identical on every page. Leave them in and every
 * document from that site chunks into near-identical text, so their embeddings
 * cluster on the chrome rather than the article and recall collapses. Dropping
 * a little real content is a far cheaper error than keeping the chrome.
 *
 * @module v1/cli/knowledge/html-extract
 */
import { decodeEntities } from './html-tags.js';
export { decodeEntities };
export interface HtmlExtractOptions {
    /** Drop nav/header/footer/aside/cookie-banner containers. Default true. */
    stripBoilerplate?: boolean;
}
export interface ExtractedHtml {
    /** `<title>` text, or null when the document has none. */
    title: string | null;
    /** Markdown-ish body text; headings as ATX, list items as `- `. */
    text: string;
}
/**
 * Parse `html` into a title and clean block text.
 *
 * Unbalanced markup is expected — real archives are full of it — so closing a
 * tag pops back to its last matching open and a never-closed element simply
 * stays open until its parent closes. Nothing here throws.
 */
export declare function extractHtmlDocument(html: string, opts?: HtmlExtractOptions): ExtractedHtml;
/**
 * Clean text for an HTML document, with the `<title>` promoted to an `# h1`
 * when the page has no top-level heading of its own.
 */
export declare function htmlToMarkdown(html: string, opts?: HtmlExtractOptions): string;
//# sourceMappingURL=html-extract.d.ts.map