benchmark

You can detect performance regressions by collecting Core Web Vitals like TTFB, FCP, and LCP, then comparing new page load results against stored baselines. This process enables automated PR-level regression detection and trend analysis over time.

The best way to baseline page load times is to run a full audit against target URLs to measure navigation timing and resource sizes. Results are stored under .gstack/benchmark-reports, enabling future comparisons to identify regressions.

Yes, you can run a performance benchmark on a specific subset of pages. The tool supports flexible page selection including auto-discovery, explicit page specification, or using --diff and --trend modes for targeted Core Web Vitals analysis.

Yes, measuring navigation timing and resource sizes requires a GPU-enabled browser context and the browse daemon. These dependencies are necessary to accurately collect Core Web Vitals for baseline performance comparisons.

Trend analysis works by comparing newly collected core timing metrics against prior baseline results stored in .gstack/benchmark-reports. Using the --trend mode, it tracks web performance metric changes across multiple PRs to identify ongoing regressions.