/**
 * Incremental crawl diff — compute stable content hashes and skip unchanged screens.
 *
 * Each crawled page gets a content fingerprint (murmurhash-free, pure JS).
 * On the next crawl of the same project, pages with the same fingerprint are
 * marked SKIPPED instead of re-processed, drastically reducing crawl time and
 * LLM token spend for sites with mostly-static content.
 *
 * Design: "same content hash" means the meaningful text hasn't changed.
 * Layout/CSS/JS changes that don't affect visible text are ignored intentionally.
 */
export interface ContentHash {
    url: string;
    hash: string;
    textLength: number;
    computedAt: string;
}
export interface DiffResult {
    url: string;
    changed: boolean;
    previousHash?: string;
    currentHash: string;
    reason: 'new' | 'changed' | 'unchanged';
}
/**
 * Compute a stable 64-bit-ish hash from text content.
 * djb2 variant — fast, no deps, collision-resistant enough for content fingerprinting.
 */
export declare function hashContent(text: string): string;
/**
 * Normalize text before hashing — strips whitespace-only lines, trims each line,
 * collapses runs of blank lines. Makes hash stable across minor formatting changes.
 */
export declare function normalizeText(text: string): string;
export declare function computeContentHash(url: string, markdownText: string): ContentHash;
/**
 * Compare a page's current content against prior hashes.
 * Returns { changed: false } for unchanged pages — caller should skip re-processing.
 */
export declare function diffContent(url: string, currentMarkdown: string, priorHashes: Map<string, string>): DiffResult;
/**
 * Build a hash map from prior screen records (from DB query result).
 * screenRecords: Array<{ url: string; elementHash?: string | null }>
 */
export declare function buildPriorHashMap(screenRecords: Array<{
    url: string;
    elementHash?: string | null;
}>): Map<string, string>;
/**
 * Filter a batch of discovered URLs against prior hashes.
 * Returns { toProcess, toSkip } — caller enqueues toProcess, logs toSkip count.
 *
 * Note: filtering at URL level only works if the prior crawl already visited the URL.
 * For new URLs, they're always in toProcess.
 */
export declare function partitionByChange(urls: string[], priorHashes: Map<string, string>): {
    toProcess: string[];
    toSkip: string[];
};