@m4trix/evals 0.16.0 → 0.17.0
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/dist/cli-simple.cjs +8 -0
- package/dist/cli-simple.cjs.map +1 -1
- package/dist/cli-simple.js +8 -0
- package/dist/cli-simple.js.map +1 -1
- package/dist/cli.cjs +8 -0
- package/dist/cli.cjs.map +1 -1
- package/dist/cli.js +8 -0
- package/dist/cli.js.map +1 -1
- package/dist/index.cjs +8 -0
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.ts +19 -1
- package/dist/index.js +8 -0
- package/dist/index.js.map +1 -1
- package/package.json +1 -1
package/dist/cli-simple.js
CHANGED
|
@@ -561,6 +561,7 @@ function processOneTestCase(task, testCaseItem, totalEvaluations, publishEvent,
|
|
|
561
561
|
const reruns = typeof testCaseItem.testCase.getReruns === "function" ? testCaseItem.testCase.getReruns() : 1;
|
|
562
562
|
const rerunPassed = [];
|
|
563
563
|
for (let r = 0; r < reruns; r++) {
|
|
564
|
+
const evaluatorRunId = `run-${randomUUID()}`;
|
|
564
565
|
const started = Date.now();
|
|
565
566
|
const evaluatorScores = [];
|
|
566
567
|
let testCaseError;
|
|
@@ -587,6 +588,11 @@ function processOneTestCase(task, testCaseItem, totalEvaluations, publishEvent,
|
|
|
587
588
|
input: testCaseItem.testCase.getInput(),
|
|
588
589
|
ctx,
|
|
589
590
|
output,
|
|
591
|
+
meta: {
|
|
592
|
+
triggerId: task.triggerId,
|
|
593
|
+
runId: evaluatorRunId,
|
|
594
|
+
datasetId: task.datasetId
|
|
595
|
+
},
|
|
590
596
|
logDiff,
|
|
591
597
|
log
|
|
592
598
|
})
|
|
@@ -1046,6 +1052,7 @@ var EffectRunner = class {
|
|
|
1046
1052
|
(sum, tc) => sum + (typeof tc.testCase.getReruns === "function" ? tc.testCase.getReruns() : 1),
|
|
1047
1053
|
0
|
|
1048
1054
|
);
|
|
1055
|
+
const triggerId = request.triggerId ?? `trg-${randomUUID()}`;
|
|
1049
1056
|
const runId = `run-${randomUUID()}`;
|
|
1050
1057
|
const artifactPath = createArtifactPath(
|
|
1051
1058
|
this.config.artifactDirectory,
|
|
@@ -1087,6 +1094,7 @@ var EffectRunner = class {
|
|
|
1087
1094
|
await Effect.runPromise(
|
|
1088
1095
|
Queue.offer(this.runQueue, {
|
|
1089
1096
|
runId,
|
|
1097
|
+
triggerId,
|
|
1090
1098
|
datasetId: request.datasetId,
|
|
1091
1099
|
dataset: dataset.dataset,
|
|
1092
1100
|
evaluators: selectedEvaluators,
|