Production capability baseline
Real calls against the deployed Sydney service, measured through the repository's release gate—not synthetic component timings.
Generated 2026-07-12T07:02:41Z · service 0.3.31
Methodology
- Three attempts for static HTML, JSON, rendered JavaScript, screenshots, and cache reuse.
- A cold/warm named-session pair and four concurrent duplicate requests.
- Wall time includes the client-to-service network path; engine time is also preserved in the private artifact.
- Fixtures verify expected content and target status, not merely an HTTP 200 from Better Fetch.
- Protected-site and residential-routing probes were excluded from this baseline to avoid presenting a small anti-bot sample as universal coverage.
Interpretation
This is a point-in-time regression baseline, not an uptime SLA or a claim that every website behaves like the controlled fixtures. Live web outcomes vary with target behaviour, geography, rendering, and optional routing.
Current health is available on the status page. API fields such as timing_ms, attempts, blocked, and proxy_used expose the outcome of each customer request.