import { D as DetectorResult, S as ScanInput } from '../types-BpZyHhFT.js';

interface XPost {
    id: string;
    text: string;
    created_at: number;
    reply_count: number;
    like_count: number;
    retweet_count: number;
    in_reply_to_id?: string;
    reply_delta_ms?: number;
}

/**
 * Cadence detector.
 *
 * Autonomous agents post at regular intervals, at all hours of the day, and
 * show no business-hours bias. Humans cluster posts in their waking hours
 * and go quiet during sleep.
 *
 * Signals analyzed:
 *  - Hour-of-day distribution entropy (high entropy = agent-like)
 *  - Business-hours ratio (0.4-0.8 is typical for humans, <0.3 for agents)
 *  - Interval-between-posts standard deviation (agents = low variance)
 *  - Sleep-window detection (5am-9am local dead zone = human)
 */
declare function cadenceDetector(posts: XPost[]): DetectorResult;

interface HeliusTransaction {
    signature: string;
    timestamp: number;
    fee: number;
    programs: string[];
    type: string;
}

/**
 * On-chain cadence detector.
 *
 * Autonomous agents transact in bursts when triggered by external events,
 * and show either perfectly regular cron-like activity or reactive bursts.
 * Humans transact sporadically with long gaps and clustered around waking
 * hours.
 *
 * Signals analyzed:
 *  - Transaction time-of-day distribution
 *  - Burst detection (many txs within seconds = bot)
 *  - Program diversity (agents hit few programs, humans hit many)
 *  - Fee behavior (agents pay consistent priority fees)
 */
declare function onchainDetector(txs: HeliusTransaction[]): DetectorResult;

/**
 * Voice consistency detector.
 *
 * Humans make typos that evolve over time, reuse unique phrases, drift in
 * tone, and use emoji inconsistently. Autonomous agents have frozen voice:
 * stable vocabulary, no typos, consistent phrase templates, predictable
 * emoji usage.
 *
 * Returns a HIGHER score for more consistent (agent-like) voice.
 */
declare function voiceDetector(posts: XPost[]): DetectorResult;

/**
 * Timing anomaly detector.
 *
 * Returns a count of anomalies rather than a 0-100 score. The aggregator
 * normalizes the count to a score (0 anomalies = 100, 10+ = 0).
 *
 * Anomalies flagged:
 *  - Sub-second response times (bot-like)
 *  - Perfect regularity in posting intervals (cron job signature)
 *  - Simultaneous X and on-chain activity within 1 second of each other
 *  - Posts during 4am-7am UTC with high frequency
 *  - Absolute silence gaps > 48h followed by burst activity
 */
declare function timingDetector(posts: XPost[], txs: HeliusTransaction[]): DetectorResult;

/**
 * Cross-source correlation detector.
 *
 * The hardest signal to fake: do the X posting timeline and the on-chain
 * activity timeline correlate? An autonomous agent that is actually running
 * the claimed wallet will show synchronized bursts. A human operator
 * manually posting will show uncorrelated timelines.
 *
 * Score is 0-100 where higher = stronger correlation = more autonomous.
 * Uncorrelated timelines (two independent actors) score near 0, which
 * usually flips the verdict to HYBRID.
 */
declare function correlationDetector(posts: XPost[], txs: HeliusTransaction[], target: ScanInput): DetectorResult;

export { cadenceDetector, correlationDetector, onchainDetector, timingDetector, voiceDetector };