/** * Evaluation Harness, built-in benchmark suites. * * These suites are the stable benchmark set checked in CI. They exercise the * PRODUCTION scoring and budget-evaluation code paths (PerfMonitor, * scoreScenario, the gate) end to end and deterministically. * * HONESTY NOTE, what is real vs. synthetic here: * - REAL: each scenario's `durationMs` is its own wall-clock runtime * (`Date.now() - t0`), an actual measurement of this process. * - SYNTHETIC PLACEHOLDERS: the render-cycle timings (`syntheticRenderCycles`), * the injected SLO/queue/overhead metrics passed to `PerfMonitor.evaluate`, * and the token/cost figures are fixtures chosen to drive the scoring path * down known branches, they are NOT captured from a live workload. Every * such value below is flagged inline with `synthetic placeholder`. * * This module deliberately does not present fabricated timings/costs as real * measurements. A future suite that captures live workload metrics can add * genuinely-measured scenarios alongside these. * * Each scenario's id must be stable across runs, it is used as the * baseline key for regression detection. */ import type { EvalScenario } from './types.js'; /** * All built-in benchmark suites. * * Suite names are stable, used as keys in baselines. */ export declare const BUILTIN_SUITES: Record; /** Flat list of all built-in scenarios (across all suites). */ export declare const ALL_SCENARIOS: EvalScenario[]; //# sourceMappingURL=suites.d.ts.map