/**
 * MHTML / MHT container parsing (RCL-01).
 *
 * An MHTML archive is an RFC 2557 `multipart/related` MIME message: the page
 * plus every subresource it referenced, each part carrying its own transfer
 * encoding and charset. Chrome's `Page.captureSnapshot` writes them
 * quoted-printable; Internet Explorer and most "save as web archive" tools
 * write base64.
 *
 * Only the ROOT frame is prose. Picking the first `text/html` part instead is
 * how an ad iframe ends up indexed as the article, so the root is resolved
 * from the container's `Snapshot-Content-Location` / `Content-Location` when
 * one is present.
 *
 * Read the file as `latin1` (or pass a Buffer): byte fidelity matters before
 * transfer decoding, and the real charset is only known per part.
 *
 * @module v1/cli/knowledge/mhtml
 */
export interface MhtmlPart {
    /** Lowercased MIME type without parameters, e.g. `text/html`. */
    contentType: string;
    /** Charset from the part's Content-Type, lowercased, or null. */
    charset: string | null;
    /** Content-Location, or null. */
    location: string | null;
    /** Lowercased Content-Transfer-Encoding; `binary` when unstated. */
    encoding: string;
    /** Raw bytes after transfer decoding. */
    body: Buffer;
    headers: Record<string, string>;
}
export interface ParsedMhtml {
    /** The part holding the page itself, or null when there is no HTML part. */
    root: MhtmlPart | null;
    parts: MhtmlPart[];
    /** Container `Snapshot-Content-Location`/`Content-Location` header. */
    snapshotLocation: string | null;
    /** Container `Subject` header — Chrome writes the page title here. */
    subject: string | null;
}
/** Decode a quoted-printable body to bytes: soft line breaks vanish, `=XX`
 *  becomes one byte, everything else is its own latin1 byte. */
export declare function decodeQuotedPrintable(input: string): Buffer;
/** Decode bytes under a MIME charset, falling back to UTF-8 (then latin1)
 *  rather than failing an ingest over an exotic label. */
export declare function decodeCharset(bytes: Buffer, charset: string | null): string;
/**
 * Parse an MHTML container. Anything that is not a well-formed multipart
 * message is treated as a single part, so a plain `.html` file renamed to
 * `.mhtml` still ingests.
 */
export declare function parseMhtml(raw: string | Buffer): ParsedMhtml;
/** Parse a container and return its root frame decoded to a string. */
export declare function mhtmlToHtml(raw: string | Buffer): {
    html: string;
    location: string | null;
    subject: string | null;
};
//# sourceMappingURL=mhtml.d.ts.map