JavaScript and deployment investigator
Record browser errors, failed assets and rendered content on one public page, then compare the evidence after a deployment change.
Open Website projects, then select a website or add and verify one. Your project and run evidence stay private.
Inspect what the browser observed
From a website project, open the JavaScript and deployment investigator. Save one public page, choose a viewport and optionally state the visible text or CSS selector you expect.
Each manual run uses a fresh browser context. Its evidence separates browser errors, failed resource requests, rendered content, audit-policy restrictions and possible explanations. Review the event sequence alongside the screenshot.
Repeat the same page check
Review a healthy run before approving it as a baseline. After a change, run the saved page check again to see new, existing and resolved observations where the evidence is comparable.
A disappeared error is only useful evidence of resolution when the relevant capture completed. Missing evidence stays unavailable.
Worked example: HTTP 200 with a missing script
Illustrative summary of a controlled deployment fixture. Healthy and failing states were deliberately created for testing, not collected from a customer site.
Saved page check
Expect the visible text Deployment ready and the element #page-marker on the same public page.
- Healthy state
- The application script loads and the expected text appears. The bounded browser capture records no observed deployment issues. That result covers the observed page state and window, so later interactions can still need a separate check.
- Controlled missing-script failure
- The document returns HTTP 200 and the expected element is present, but its expected text is absent. A dynamically imported script returns HTTP 404 and the browser reports an import failure. The report flags the observed issues for review.
- Incomplete measurement
- If the request, error or rendering capture is incomplete, the report marks that coverage unavailable. It cannot establish that a previously failed script recovered, even if no matching error was retained.
Observation and possible explanation
The failed resource and import error refer to the same script. That relationship supports a possible explanation for the missing content. It does not prove the root cause.
Check the deployed asset path, release manifest and server response. After correcting a confirmed problem, repeat the same page check and compare completed evidence. A request blocked by the audit policy remains a measurement restriction.
What an investigation can cover
One public page is observed in a fixed desktop or mobile viewport. The browser does not sign in, fill forms, submit requests that change state or fetch source maps. Screenshots show the captured public viewport and may contain visible page information.
The investigator uses the existing bounded diagnostic queue. A request limit, timeout, cancellation or policy block can leave part of the evidence unavailable. A failed measurement does not prove a website defect.
The latest five finished investigations are retained for up to 30 days, plus the approved baseline and active work. Keep hostname verification in place, download needed evidence before it expires and review it before sharing.