/**
 * Source Authority
 *
 * Classifies a URL by domain into a trust tier so downstream
 * consumers can rank or filter sources by how authoritative they
 * are. Used by:
 *
 *   - The research engine, to prioritize official sources when
 *     multiple sources conflict
 *   - The browse response (via verification engine), so any caller
 *     can apply their own "minimum authority" policy without
 *     reimplementing the classifier
 *   - The future /v1/verify endpoint, where AI-citation validation
 *     can reject or downweight non-authoritative sources
 *
 * ## Categories
 *
 * Categories are ordered by authority. Each carries a `score` in
 * [0..1] for callers that want a continuous signal:
 *
 *   - government       (.95) — national/federal government domains
 *   - institutional    (.90) — major libraries, archives, museums
 *                              (loc.gov, archives.gov, britishmuseum.org)
 *   - academic         (.85) — universities (.edu, .ac.uk, .ac.jp)
 *   - scientific       (.85) — peer-reviewed journals, preprint
 *                              servers (nature.com, arxiv.org, pubmed)
 *   - intergovernmental (.85) — UN, EU, WHO, OECD
 *   - reference        (.70) — Wikipedia, Britannica, Encyclopedia.com
 *   - news             (.70) — established mainstream news
 *   - official         (.70) — generic .org / .gov.* not in the
 *                              specific lists
 *   - commercial       (.50) — generic for-profit sites
 *   - unknown          (.50) — anything that doesn't classify
 *
 * Categories are NOT mutually exclusive in principle (a domain
 * could be both government and academic) but the classifier returns
 * the highest-authority match it finds.
 *
 * ## Updating the lists
 *
 * The phrase lists are intentionally curated, not exhaustive. Add
 * domains as real-world traffic surfaces gaps. The classifier is
 * a fast first pass — for high-stakes decisions, callers should
 * combine this with their own domain allowlist.
 */
/**
 * Authority categories, ordered roughly by trust tier.
 */
export type AuthorityCategory = 'government' | 'institutional' | 'academic' | 'scientific' | 'intergovernmental' | 'reference' | 'news' | 'official' | 'commercial' | 'unknown';
/**
 * Authority classification result returned by `classifySource`.
 */
export interface AuthorityClassification {
    category: AuthorityCategory;
    /** Score in [0..1]. Higher = more authoritative. */
    score: number;
    /** Domain that was matched (for diagnostics). */
    matchedDomain: string;
    /** Optional country code for government sources. */
    country?: string;
    /**
     * Human-readable reason explaining why this category was assigned.
     * Useful for surfacing in UIs or for debugging unexpected
     * classifications.
     */
    reason: string;
}
/**
 * Classify a URL by its domain into an authority category.
 *
 * The classifier walks the lists in order of authority (highest
 * first) and returns on the first match. So a domain that's both
 * government and reference (rare but possible) classifies as
 * government, the higher tier.
 *
 * Walk order (load-bearing for some classifications):
 *
 *   1. Country-specific government (so we can attribute the country)
 *   2. Institutional archives — runs BEFORE the generic .gov TLD
 *      check so loc.gov, archives.gov, ndl.go.jp etc. get their
 *      richer "institutional" classification instead of the
 *      anonymous "generic government" tag.
 *   3. Intergovernmental
 *   4. Scientific publishers — same reason as institutional:
 *      pubmed.ncbi.nlm.nih.gov, ncbi.nlm.nih.gov should classify
 *      as scientific even though they're also under .gov.
 *   5. Academic (explicit list, then TLD patterns)
 *   6. Generic .gov / .gob / .gouv / .govt — fallback for any
 *      remaining government domain that didn't get a richer tag
 *   7. Reference encyclopedias
 *   8. News
 *   9. Generic .org / .edu fallback (.edu is already caught at
 *      step 5, this is the .org branch)
 *  10. Unknown — neutral default
 *
 * Within equal-score tiers (intergovernmental / scientific /
 * academic, all 0.85), the walk order above determines which
 * category label wins on a domain that could match multiple. The
 * `score` field is the same regardless, so consumers filtering by
 * score see identical behavior. Don't build logic on
 * "scientific always wins over academic" — it's walk-order
 * dependent and could change if the lists are reordered.
 *
 * Returns `unknown` (score 0.5) for any domain that doesn't match
 * any of the curated lists. This is a deliberate safe default —
 * callers reading the result should treat unknown as "no signal,"
 * not "low quality."
 */
export declare function classifySource(url: string): AuthorityClassification;
/**
 * Returns true if the URL meets a minimum authority score. Cheap
 * convenience wrapper for the common "filter by trust tier" path.
 */
export declare function meetsMinimumAuthority(url: string, minScore: number): boolean;
/**
 * Exposed for tests and for callers that want to extend the
 * classifier with their own domains.
 */
export declare const _internalForTests: {
    GOVERNMENT_DOMAINS: Record<string, readonly string[]>;
    INSTITUTIONAL_ARCHIVES: readonly string[];
    ACADEMIC_DOMAINS: readonly string[];
    ACADEMIC_TLDS: readonly string[];
    SCIENTIFIC_DOMAINS: readonly string[];
    INTERGOVERNMENTAL_DOMAINS: readonly string[];
    REFERENCE_DOMAINS: readonly string[];
    NEWS_DOMAINS: readonly string[];
    CATEGORY_SCORES: Record<AuthorityCategory, number>;
};
//# sourceMappingURL=source-authority.d.ts.map