# Changelog of Site Speed Audit for Agencies: Mobile, Desktop & Regressions (`hani-tools/site-speed-audit`) Actor

- **URL**: https://apify.com/hani-tools/site-speed-audit/changelog.md
- **Full Actor documentation**: https://apify.com/hani-tools/site-speed-audit.md

## Changelog

### 1.0.0-rc3 (2026-10-19) – private build, not published

- Named history dataset renamed `site-summaries` → `site-speed-audit-site-summaries`. Under Apify's limited permissions an Actor cannot open a named storage that another Actor created, so a generic name could fail the run.
- Writing to the history dataset or opening/saving the baseline store can no longer fail the run (warning only; comparison is skipped).
- Default input / prefill changed from example.com to https://choosealicense.com/ (a light static page that audits reliably within the 45 s limit).
- Actor title shortened to 60 characters (the platform limit is 63).

### 1.0.0-rc2 (2026-10-16) – not yet published

- New input `maxAuditSeconds` (default 45 s, 20–180): hard limit per measurement including browser start-up, enforced by killing the measurement process. Replaces the old "page-load timeout + 90 s (min 120 s)" limit.
- Pages that exceed it get status `timeout_limit` with a plain-English hint to raise `maxAuditSeconds`; they are not charged and not retried.
- `pageLoadTimeoutSecs` default lowered from 60 to 30 s and always kept ≥ 15 s below `maxAuditSeconds`.
- Extra median runs are skipped for a page when a measurement took longer than 36 s or hit the limit (`extraRunsSkipped` field), so extra runs stay profitable.

### 1.0.0 (2026-10-16) – first release candidate (not yet published)

- Bulk lab audits of URL lists, sitemaps (incl. sitemap index and .xml.gz) and domains (robots.txt `Sitemap:` → /sitemap.xml → home-page links).
- Template sampling: max pages per site and max pages per first-level path pattern; include/exclude filters.
- Mobile + desktop in one run; median of 1, 3 or 5 runs (Lighthouse's own median-run selection) with min–max score range.
- Performance, accessibility, best-practices and SEO scores; LCP, FCP, TBT, CLS, Speed Index, TTFB, page weight; lab Core Web Vitals verdict.
- Plain-English top fixes per page and site-wide prioritised fixes ("affects 37/40 pages, ~1.2 s per page").
- Regression detection against the median of the last 3 runs stored in a named key-value store (`site-speed-baselines-<baselineName>`, one record per site), configurable thresholds, optional JSON webhook.
- Performance budgets with pass/fail per audit, CI-friendly `OUTPUT` record and optional failing of the run.
- Branded HTML report per site (agency name, logo URL, accent colour), portfolio overview HTML, LLM-friendly `summary.md`, optional original Lighthouse HTML reports.
- Measurement transparency on every row: Lighthouse and Chromium versions, benchmarkIndex, throttling settings, CPU/memory.
- Honest failure handling: HTTP preflight (404/410, non-HTML, login redirects, DNS/TLS errors, private addresses), robots.txt for discovered pages, per-host politeness delay, sequential audits, hard per-run timeout, limited retries for transient errors. Failed or skipped pages are never charged; the run fails loudly when nothing could be measured.
- Pay-per-event: `page-audit`, `extra-run`, `site-report`; respects the user's maximum charge per run.