fix: retain attempted evaluation outcomes and verify four engineering pilots

This commit is contained in:
Agent Workbench Dev
2026-11-01 10:02:50 +08:00
parent 17c9f9d0cd
commit 196b58ac9b
709 changed files with 38104 additions and 4 deletions
+1 -1
View File
@@ -60,7 +60,7 @@ export function evaluationSampleSeal(batch, sample) {
files: sample.files, commands: sample.commands, children: sample.children };
}
export function evaluationSummary(batch) {
const outcomes = batch.samples.map(sample => sample.report?.summary ?? { original: sample.baselineAttemptedAt ? 'unknown' : 'pending', candidate: ['failed','stopped','unknown'].includes(sample.status) ? sample.status : 'pending' });
const outcomes = batch.samples.map(sample => sample.report?.summary ?? { original: sample.baselineAttemptedAt ? 'unknown' : 'pending', candidate: ['failed','stopped','unknown'].includes(sample.status) ? sample.status : ['candidate_running','candidate_complete','awaiting_review','paused','collecting','adopting','sealed'].includes(sample.status) ? 'unknown' : 'pending' });
const count = (role, status) => outcomes.filter(item => item[role] === status).length;
return { samples: batch.samples.length, finished: batch.samples.filter(item => !!item.report).length,
original: Object.fromEntries(['passed', 'failed', 'stopped', 'unknown', 'pending'].map(status => [status, count('original', status)])),