@m4trix/evals 0.29.0 → 0.31.0
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/README.md +1 -1
- package/dist/cli-simple.cjs +19 -6
- package/dist/cli-simple.cjs.map +1 -1
- package/dist/cli-simple.js +19 -6
- package/dist/cli-simple.js.map +1 -1
- package/dist/cli.cjs +9 -1
- package/dist/cli.cjs.map +1 -1
- package/dist/cli.js +9 -1
- package/dist/cli.js.map +1 -1
- package/dist/index.cjs +7 -0
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.ts +15 -0
- package/dist/index.js +7 -0
- package/dist/index.js.map +1 -1
- package/package.json +1 -1
package/dist/cli.cjs
CHANGED
|
@@ -1731,6 +1731,8 @@ function processOneEvaluation(task, unit, totalEvaluations, publishEvent, persis
|
|
|
1731
1731
|
output,
|
|
1732
1732
|
meta: {
|
|
1733
1733
|
triggerId: task.triggerId,
|
|
1734
|
+
triggerTimestamp: task.triggerTimestamp,
|
|
1735
|
+
triggeredAt: new Date(task.triggerTimestamp).toISOString(),
|
|
1734
1736
|
runId: evaluatorRunId,
|
|
1735
1737
|
datasetName: task.dataset.getDisplayLabel(),
|
|
1736
1738
|
testCaseId: testCaseItem.id,
|
|
@@ -2211,6 +2213,7 @@ var EffectRunner = class {
|
|
|
2211
2213
|
const globalConcurrency = Math.max(1, request.globalConcurrency);
|
|
2212
2214
|
const sem = effect.Effect.unsafeMakeSemaphore(globalConcurrency);
|
|
2213
2215
|
const triggerId = request.triggerId ?? `trg-${crypto.randomUUID()}`;
|
|
2216
|
+
const triggerTimestamp = request.triggerTimestamp ?? Date.now();
|
|
2214
2217
|
const snapshots = [];
|
|
2215
2218
|
for (const job of request.jobs) {
|
|
2216
2219
|
snapshots.push(
|
|
@@ -2218,6 +2221,7 @@ var EffectRunner = class {
|
|
|
2218
2221
|
datasetId: job.datasetId,
|
|
2219
2222
|
evaluatorIds: job.evaluatorIds,
|
|
2220
2223
|
triggerId,
|
|
2224
|
+
triggerTimestamp,
|
|
2221
2225
|
maxConcurrency: this.config.maxConcurrency ?? 1,
|
|
2222
2226
|
globalEvaluationSemaphore: sem,
|
|
2223
2227
|
runConfigName: job.runConfigName,
|
|
@@ -2255,6 +2259,7 @@ var EffectRunner = class {
|
|
|
2255
2259
|
datasetId: request.datasetId,
|
|
2256
2260
|
evaluatorIds: request.evaluatorIds,
|
|
2257
2261
|
triggerId: request.triggerId,
|
|
2262
|
+
triggerTimestamp: request.triggerTimestamp ?? Date.now(),
|
|
2258
2263
|
maxConcurrency: request.concurrency ?? this.config.maxConcurrency ?? 1,
|
|
2259
2264
|
repetitions: request.repetitions,
|
|
2260
2265
|
runConfigName,
|
|
@@ -2282,6 +2287,7 @@ var EffectRunner = class {
|
|
|
2282
2287
|
const totalEvaluations = selectedTestCases.length * repetitions;
|
|
2283
2288
|
const runConfigTags = [...params.runConfigTags ?? []];
|
|
2284
2289
|
const triggerId = params.triggerId ?? `trg-${crypto.randomUUID()}`;
|
|
2290
|
+
const triggerTimestamp = params.triggerTimestamp ?? Date.now();
|
|
2285
2291
|
const runId = `run-${crypto.randomUUID()}`;
|
|
2286
2292
|
const artifactPath = createArtifactPath(this.config.artifactDirectory, params.datasetId, runId);
|
|
2287
2293
|
const snapshot = {
|
|
@@ -2325,6 +2331,7 @@ var EffectRunner = class {
|
|
|
2325
2331
|
effect.Queue.offer(this.runQueue, {
|
|
2326
2332
|
runId,
|
|
2327
2333
|
triggerId,
|
|
2334
|
+
triggerTimestamp,
|
|
2328
2335
|
datasetId: params.datasetId,
|
|
2329
2336
|
dataset: dataset.dataset,
|
|
2330
2337
|
evaluators: selectedEvaluators,
|
|
@@ -3103,7 +3110,8 @@ function EvalsCliApp({ data, args, runner }) {
|
|
|
3103
3110
|
void runner.runDatasetWith({
|
|
3104
3111
|
datasetId: selectedDataset.id,
|
|
3105
3112
|
evaluatorIds: clampedState.selectedEvaluatorIds,
|
|
3106
|
-
...PROGRAMMATIC_RUN_CONFIG
|
|
3113
|
+
...PROGRAMMATIC_RUN_CONFIG,
|
|
3114
|
+
triggerTimestamp: Date.now()
|
|
3107
3115
|
}).then((snapshot) => {
|
|
3108
3116
|
setRuntimeMessage(
|
|
3109
3117
|
`Started ${snapshot.runId} on ${selectedDataset.name} (${snapshot.totalTestCases} cases).`
|