@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.mjs
CHANGED
|
@@ -229,6 +229,12 @@ function applyParamsToAssertion(assertion, params) {
|
|
|
229
229
|
}
|
|
230
230
|
};
|
|
231
231
|
}
|
|
232
|
+
if (assertion.type === "time_limit" && params.maxDurationMs !== void 0) {
|
|
233
|
+
return {
|
|
234
|
+
...assertion,
|
|
235
|
+
maxDurationMs: params.maxDurationMs
|
|
236
|
+
};
|
|
237
|
+
}
|
|
232
238
|
if (assertion.type === "skill_was_called" && params.skillNames !== void 0) {
|
|
233
239
|
return {
|
|
234
240
|
...assertion,
|
|
@@ -254,6 +260,12 @@ function resolveSystemAssertion(assertionId, params) {
|
|
|
254
260
|
expectedExitCode: params?.expectedExitCode ?? void 0
|
|
255
261
|
};
|
|
256
262
|
break;
|
|
263
|
+
case "time_limit":
|
|
264
|
+
baseAssertion = {
|
|
265
|
+
type: "time_limit",
|
|
266
|
+
maxDurationMs: params?.maxDurationMs ?? 3e5
|
|
267
|
+
};
|
|
268
|
+
break;
|
|
257
269
|
case "llm_judge":
|
|
258
270
|
baseAssertion = {
|
|
259
271
|
type: "llm_judge",
|
|
@@ -2528,7 +2540,8 @@ async function runScenario(config, evalRunId2, scenario, evalData, template, res
|
|
|
2528
2540
|
fileDiffs: partialResult.fileDiffs?.map((d) => ({
|
|
2529
2541
|
path: d.path,
|
|
2530
2542
|
status: templateFilesMap.get(d.path)
|
|
2531
|
-
}))
|
|
2543
|
+
})),
|
|
2544
|
+
durationMs: partialResult.duration
|
|
2532
2545
|
};
|
|
2533
2546
|
const { "x-wix-ai-gateway-stream": _stream, ...judgeHeaders } = config.aiGatewayHeaders;
|
|
2534
2547
|
const defaultJudgeModel = DEFAULT_JUDGE_MODEL;
|