Lighthouse Audit History Block for Svelte

Edit RAW, newest first: each run has an id, an ISO date, the four scores and its fails, one audit:size string per failing audit (milliseconds, elements or a CLS number, as the audit's unit in AUDITS says). The three lists, the changes, the header and the bars are all worked out from those strings.

An audit is Newly failing if it fails in the chosen run and not in the baseline, Fixed if it failed in the baseline and passes now, and Still failing if it fails in both. The baseline is the run before by default; the select offers the other runs, so you can compare with last month's run too. The oldest run has no baseline and lists its audits as failing.

The performance audits that fail in the chosen run, largest first, up to four. Each bar is the audit's impact in milliseconds on a scale from 0 to 800 ms shared by every run, so a smaller bar in a later run really is smaller. The number is also in text beside the name.

Everything. The runs are typed, nothing is audited and nothing is requested. The chosen run and baseline are local state. In your app store each run's Lighthouse result and map it to the same fields.

The list becomes a select with the date and Performance score, the scores sit in a two-column grid and the three lists and the bars scroll inside one region of fixed height, so choosing a run never moves the page.