/**
 * The vocabulary for doc URLs: normalize one for matching, join two to build
 * one, and ask whether one sits under a base. Every layer that touches URL
 * space goes through here, because the alternative is what this replaced: the
 * same trailing-slash strip written with three different regexes and the same
 * segment-boundary rule written three times, quietly disagreeing.
 *
 * Pure string work over a URL's path space. Nothing here knows about docs,
 * versions or pages; the domain meaning lives in `core/`, which calls in.
 */
/**
 * Strip a trailing slash so route matching is consistent regardless of the
 * consumer's SvelteKit `trailingSlash` setting: `/docs/intro/` and `/docs/intro`
 * resolve to the same content entry. The root stays `/`.
 */
export declare function normalizePath(pathname: string): string;
/**
 * True when nothing follows a base, or what follows starts a new segment rather
 * than continuing the last one. `?` and `#` count, so the rule holds for a link
 * href as well as a pathname; that is what stops `/docs` from matching
 * `/docsmith`.
 *
 * Exported for the one caller that has already sliced the base off itself: the
 * archive link rewriter matches its base with a regex and needs the remainder
 * afterwards, so it checks the boundary rather than re-testing the whole URL.
 */
export declare function atBoundary(rest: string): boolean;
/** True when `url` is `base` or sits beneath it on a segment boundary. */
export declare function under(url: string, base: string): boolean;
/**
 * The one segment directly below `base`, which is what decides who owns a URL:
 * an archived version owns `<docs base>/<id>/…`, and a header link owns its
 * whole first-segment section. `undefined` when `url` is not under `base` at
 * all, or when the base is the end of it, so callers must handle "no segment"
 * rather than interpolate it into a path.
 */
export declare function firstSegmentUnder(url: string, base: string): string | undefined;
/**
 * Join two URL pieces with exactly one slash between them, whichever of them
 * already carries it. Use it for every URL built from a configured origin or
 * base, so a trailing slash in someone's config can never double up.
 *
 * Only the seam between the two is touched, never the rest of either: rescanning
 * would turn `https://x.dev` into `https:/x.dev`. Nor is the result normalized,
 * so `join(origin, '/')` still ends in a slash and a sitemap can carry the home
 * page.
 *
 * An empty `left` returns `right` untouched, which is what an unconfigured
 * origin needs: `llms.txt` and the copy-page menu fall back to site-relative
 * links rather than emitting a stray leading slash. Callers that must not degrade
 * to a relative URL, like `<link rel="canonical">`, have to check the origin
 * themselves.
 */
export declare function join(left: string, right: string): string;