fix(web-probe): analyze sample timing series

This commit is contained in:
Codex
2026-06-20 16:36:13 +00:00
parent cbae56de19
commit 93b9158119
2 changed files with 172 additions and 4 deletions
+1
View File
@@ -87,6 +87,7 @@ export function hwlabNodeWebProbeHelp(): Record<string, unknown> {
"observe start registers a local UniDesk-side observer id under .state/web-observe/index.json; after start, prefer observe status|command|stop|collect|analyze <id> instead of repeating --node/--lane/--state-dir.",
"observe command actions are explicit user/control actions and are appended to control.jsonl; use --type newSession/selectProvider/sendPrompt/goto/screenshot/mark/stop and keep prompt text out of issue comments by citing textHash/textBytes.",
"observe analyze is offline-only: it reads artifact JSONL and writes analysis/report.md plus analysis/report.json without accessing Workbench APIs or driving the browser.",
"observe analyze scans every sampled DOM point, extracts Workbench timing text such as 总耗时/total and 最近 N 秒/分前, writes sampleMetrics.timeline, and flags backwards/frozen/stale/diagnostic timing anomalies instead of relying on status tail output.",
"Use recordStep(name, data) or fetchApiMatrix(paths) to keep structured partial evidence when a later step fails.",
"Use reloadStable(), gotoCurrentStable(), or safeReload() for bounded retries around page reload/current-URL navigation jitter such as ERR_NETWORK_CHANGED.",
"Playwright page.evaluate accepts one serializable argument; use page.evaluate(({ a, b }) => ..., { a, b }) or safeEvaluate(fn, { a, b }).",