@wix/evalforge-evaluator 0.89.0 → 0.90.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/build/index.js +14 -1
- package/build/index.js.map +2 -2
- package/build/index.mjs +14 -1
- package/build/index.mjs.map +2 -2
- package/package.json +5 -5
package/build/index.js
CHANGED
|
@@ -249,6 +249,12 @@ function applyParamsToAssertion(assertion, params) {
|
|
|
249
249
|
}
|
|
250
250
|
};
|
|
251
251
|
}
|
|
252
|
+
if (assertion.type === "time_limit" && params.maxDurationMs !== void 0) {
|
|
253
|
+
return {
|
|
254
|
+
...assertion,
|
|
255
|
+
maxDurationMs: params.maxDurationMs
|
|
256
|
+
};
|
|
257
|
+
}
|
|
252
258
|
if (assertion.type === "skill_was_called" && params.skillNames !== void 0) {
|
|
253
259
|
return {
|
|
254
260
|
...assertion,
|
|
@@ -274,6 +280,12 @@ function resolveSystemAssertion(assertionId, params) {
|
|
|
274
280
|
expectedExitCode: params?.expectedExitCode ?? void 0
|
|
275
281
|
};
|
|
276
282
|
break;
|
|
283
|
+
case "time_limit":
|
|
284
|
+
baseAssertion = {
|
|
285
|
+
type: "time_limit",
|
|
286
|
+
maxDurationMs: params?.maxDurationMs ?? 3e5
|
|
287
|
+
};
|
|
288
|
+
break;
|
|
277
289
|
case "llm_judge":
|
|
278
290
|
baseAssertion = {
|
|
279
291
|
type: "llm_judge",
|
|
@@ -2538,7 +2550,8 @@ async function runScenario(config, evalRunId2, scenario, evalData, template, res
|
|
|
2538
2550
|
fileDiffs: partialResult.fileDiffs?.map((d) => ({
|
|
2539
2551
|
path: d.path,
|
|
2540
2552
|
status: templateFilesMap.get(d.path)
|
|
2541
|
-
}))
|
|
2553
|
+
})),
|
|
2554
|
+
durationMs: partialResult.duration
|
|
2542
2555
|
};
|
|
2543
2556
|
const { "x-wix-ai-gateway-stream": _stream, ...judgeHeaders } = config.aiGatewayHeaders;
|
|
2544
2557
|
const defaultJudgeModel = import_evalforge_types6.DEFAULT_JUDGE_MODEL;
|