Lighthouse / PageSpeed Insights Audit

Runs Google's official Lighthouse audit on any URL — the same audit Chrome DevTools runs locally — as a hosted, scriptable Actor. This wraps Google's own public PageSpeed Insights v5 API directly; it is not a scrape.

What you get per URL

• Category scores (0-100): performance, accessibility, best-practices, SEO, PWA.
• Core Web Vitals, lab data: this specific run's LCP, CLS, TBT, FCP, Speed Index and Time to Interactive, under fixed simulated network/CPU throttling — reproducible, synthetic.
• Core Web Vitals, field data: real Chrome User Experience Report percentiles for the exact URL, and a site-wide fallback when Google lacks enough traffic on that specific page. This is what real visitors experienced, not a simulation — kept separate from lab data deliberately, since a page can score badly in the lab while real users had a fine experience, or the reverse.
• Ranked opportunities: failing performance audits, sorted by estimated time savings in milliseconds — the fixes worth doing first, not just a wall of raw audit data.

You need your own Google API key

This API requires a key for any real usage. Confirmed directly against the live endpoint:

Request	Result
No key	HTTP 429 — Google's shared anonymous quota, already exhausted globally
Invalid key	HTTP 400 — "API key not valid"
Valid key	Full audit

Getting one is free and takes a minute: in Google Cloud Console, enable the PageSpeed Insights API for a project, then create an API key under Credentials. No billing account is required for this API's free daily quota.

⚠️ Verification status

Every other Actor in this suite was tested against live, real data before shipping. This one could not be — no valid Google API key was available during development, and the invalid-key error path was confirmed live (see the table above) but the success-path response shape was not. The client is built directly against Google's stable, versioned, publicly documented v5 API schema, which has been unchanged for years, so this is lower-risk than reverse-engineering an undocumented site — but it is a documented risk, not a verified one.

Please run it once with your own key and check the output against the schema before relying on it. If anything looks off, the fix is almost certainly in src/scraper.py's field mapping, not the request itself (the request format is confirmed working via the error-path test above).

Input

Field	Type	Notes
urls	array	URLs to audit. Each takes ~15-40 seconds — this audits pages one at a time, it does not bulk-crawl a site.
apiKey	string	Your Google PageSpeed Insights API key. Required.
strategy	enum	mobile or desktop. Google's ranking signals are mobile-first.
categories	array	Any of performance, accessibility, best-practices, seo, pwa. Fewer categories audit faster.
locale	string	Language for audit text, e.g. es, de. Defaults to English.
proxyConfiguration	object	Not needed — Google's own infrastructure fetches the target page, not this Actor.

Known limits

• Per-URL, not per-thousand. This is priced and used per audited page — a batch of 100 URLs is 100 real Lighthouse runs, each taking real time.
• Field data is often absent. core_web_vitals_field is null for any URL without enough real-world Chrome traffic for Google to report on — common for low-traffic or new pages. Check core_web_vitals_field_origin for the site-wide number in that case.
• A failed URL is still a row. If an audit fails (bad key, unreachable URL), that URL's row carries an error field instead of scores, rather than vanishing from the dataset silently.

Local development

pip install -r requirements.txt
python test_local.py https://example.com --api-key YOUR_REAL_KEY --out sample_output.json

There is no sample_output.json committed in this folder, unlike every other Actor in this suite — see the verification status note above for why.