@rulvar/evals 1.28.0 → 1.29.0

package/dist/index.d.ts

@@ -227,7 +227,10 @@ interface RubricGraderOptions {
   name?: string;
   /**
   * Minimum fraction of criteria that must pass; default 1 (all).
-  * The fraction is also reported as the verdict score.
+  * The fraction is also reported as the verdict score. Must be a
+  * finite fraction in [0, 1]: anything else throws a ConfigError at
+  * construction, because an out of range threshold silently passes
+  * or fails every verdict (v1.28.0 review P2).
   */
   passThreshold?: number;
 }
@@ -400,6 +403,13 @@ interface SweepPool {
   models: SweepModel[];
   cases: SweepCase[];
 }
+/**
+* The claim bands. Both effective values must be finite fractions in
+* [0, 1] with weakness strictly below strength (so the bands are
+* ordered and an uninformative mid band exists); runSweepMatrix
+* rejects anything else with a ConfigError before any engine, store,
+* or envelope activity.
+*/
 interface SweepThresholds {
   /** passRate at or above emits a strength claim; default 0.9. */
   strength: number;

package/dist/index.js

@@ -352,9 +352,15 @@ function goldenGrader(expected, options = {}) {
 }
 //#endregion
 //#region src/graders/rubric.ts
+/**
+* Rubric grader: scores the output against
+* declared, named criteria with per-criterion verdicts. Criteria are pure
+* predicates; the grader is host-side and involves no engine calls.
+*/
 function rubricGrader(criteria, options = {}) {
 	const name = options.name ?? "rubric";
 	const threshold = options.passThreshold ?? 1;
+	if (!Number.isFinite(threshold) || threshold < 0 || threshold > 1) throw new ConfigError(`rubricGrader: passThreshold must be a finite fraction between 0 and 1 inclusive; got ${String(threshold)}`);
 	return {
 		name,
 		grade(context) {
@@ -801,6 +807,8 @@ async function runSweepMatrix(pool, options) {
 		...SWEEP_THRESHOLD_DEFAULTS,
 		...options.thresholds
 	};
+	for (const [field, value] of [["strength", thresholds.strength], ["weakness", thresholds.weakness]]) if (!Number.isFinite(value) || value < 0 || value > 1) throw new ConfigError(`runSweepMatrix: thresholds.${field} must be a finite fraction between 0 and 1 inclusive; got ${String(value)}`);
+	if (!(thresholds.weakness < thresholds.strength)) throw new ConfigError(`runSweepMatrix: thresholds.weakness (${String(thresholds.weakness)}) must be strictly below thresholds.strength (${String(thresholds.strength)}) so the bands are ordered and an uninformative mid band exists`);
 	if (options.envelope !== void 0 && options.suite?.budgetUsd === void 0) throw new ConfigError("runSweepMatrix: an aggregate envelope requires suite.budgetUsd (the per-target ceiling); unbounded targets under an envelope would be unaccountable");
 	const byTaskClass = /* @__PURE__ */ new Map();
 	for (const entry of pool.cases) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@rulvar/evals",
-  "version": "1.28.0",
+  "version": "1.29.0",
   "description": "Rulvar evals: eval cases, golden outputs, rubric and judge graders, matrix sweeps, canary fingerprint.",
   "type": "module",
   "license": "Apache-2.0",
@@ -22,8 +22,8 @@
     "access": "public"
   },
   "dependencies": {
-    "@rulvar/core": "1.28.0",
-    "@rulvar/testing": "1.28.0"
+    "@rulvar/core": "1.29.0",
+    "@rulvar/testing": "1.29.0"
   },
   "devDependencies": {
     "@types/node": "^22.20.0",