llmverify/dist/engines/runtime/token-rate.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.

/**
 * Token Rate Engine
 *
 * Monitors tokens-per-second throughput and detects degradation.
 * Lower token rates may indicate provider throttling or issues.
 *
 * WHAT THIS DOES:
 * ✅ Calculates tokens per second
 * ✅ Compares to baseline throughput
 * ✅ Detects throughput degradation
 *
 * WHAT THIS DOES NOT DO:
 * ❌ Predict throughput changes
 * ❌ Identify cause of slowdowns
 * ❌ Account for response complexity
 *
 * @module engines/runtime/token-rate
 * @author Haiec
 * @license MIT
 */
import { CallRecord, EngineResult, BaselineState } from '../../types/runtime';
/**
 * Analyzes token generation rate against baseline.
 *
 * @param call - The call record to analyze
 * @param baseline - Current baseline state
 * @param thresholds - Optional custom thresholds
 * @returns Engine result with token rate analysis
 *
 * @example
 * const result = TokenRateEngine(callRecord, baseline);
 * if (result.status === 'warn') {
 *   console.log('Token rate below normal');
 * }
 */
export declare function TokenRateEngine(call: CallRecord, baseline: Pick<BaselineState, 'avgTokensPerSecond'>, thresholds?: {
    warnRatio?: number;
    errorRatio?: number;
}): EngineResult;