llmverify/dist/engines/runtime/latency.d.ts

AI Output Verification Toolkit — Local-first LLM safety, hallucination detection, PII redaction, prompt injection defense, and runtime monitoring. Zero telemetry. OWASP LLM Top 10 aligned.

/**
 * Latency Engine
 *
 * Monitors LLM response latency and detects anomalies.
 * Uses baseline comparison to identify performance degradation.
 *
 * WHAT THIS DOES:
 * ✅ Compares current latency to established baseline
 * ✅ Detects sudden latency spikes
 * ✅ Provides normalized anomaly score
 *
 * WHAT THIS DOES NOT DO:
 * ❌ Predict future latency
 * ❌ Identify root cause of latency issues
 * ❌ Account for network conditions
 *
 * @module engines/runtime/latency
 * @author Haiec
 * @license MIT
 */
import { CallRecord, EngineResult, BaselineState } from '../../types/runtime';
/**
 * Analyzes latency of an LLM call against baseline.
 *
 * @param call - The call record to analyze
 * @param baseline - Current baseline state
 * @param thresholds - Optional custom thresholds
 * @returns Engine result with latency analysis
 *
 * @example
 * const result = LatencyEngine(callRecord, baseline);
 * if (result.status === 'error') {
 *   console.log('Latency spike detected');
 * }
 */
export declare function LatencyEngine(call: CallRecord, baseline: Pick<BaselineState, 'avgLatencyMs'>, thresholds?: {
    warnRatio?: number;
    errorRatio?: number;
}): EngineResult;