CV-Library Jobs Scraper (UK)

Scrapes job listings from CV-Library — one of the UK's largest job boards. A second UK job-search actor in this portfolio alongside reed-jobs-scraper (general listings) and efinancialcareers-jobs-scraper (finance-only) — distinct platform, distinct customer base.

Public data only. No login, no cookies, no browser.

The one thing you need to know before using this

CV-Library's robots.txt disallows every query string except two explicit carve-outs (?jobId= and ?page_number=) — the site's own human-facing search URL (?q=...&geo=...) is not robots-permitted for a generic client. This actor uses the equivalent path-based search URL instead (/{keyword}-jobs-in-{location}), which the site itself also serves and which carries no matching disallow rule. See CRAWLING_METHOD.md §3 for the full policy reasoning.

Separately, 3 of 7 TLS profiles tried in recon get a plain HTTP 403 (not a disguised-200 challenge) — this actor's client pool keeps only the 4 confirmed-clean profiles.

What you get

Three record types share one dataset, told apart by recordType.

JOB — one row per listing

Search rows (listing) already carry title, employer, salary range, location, category, contract type and a highlighted description snippet. Turn on Fetch job detail pages (on by default) to also attach jobDetail, the Schema.org JobPosting block with the full, untruncated description.

SEARCH_SUMMARY — one row per (keyword, location) query

CV-Library's own real match count (totalMatches — a genuine structured field, not a heuristic), pages fetched, and locationApplied — whether the requested location actually narrowed the search upstream.

ERROR — one row per input that failed

So every entry in Searches maps to at least one output row.

Input

Field	What it does
Searches	list of {keyword, location} — location is optional
Fetch job detail pages	adds the full untruncated description (on by default)
Max jobs / max pages per search	pagination caps — this actor computes the exact page count from CV-Library's own total, these are extra safety valves
Max concurrent requests / Min seconds between requests	standard pacing controls

Example

{
  "queries": [
    {"keyword": "developer", "location": "manchester"},
    {"keyword": "nurse"}
  ],
  "includeJobDetails": true,
  "maxItems": 100
}

Notes on reliability

• Location can silently widen upstream, keyword cannot. An unrecognised location does not error — CV-Library quietly falls back to the keyword-only national result set. This actor detects that (the location metadata upstream would otherwise return is simply absent) and reports locationApplied: false with zero rows rather than shipping mislabeled data. An unrecognised keyword, in contrast, answers a clean HTTP 404 and is reported as a not_found error — the two behave oppositely, so both are handled explicitly.
• totalMatches is a real structured field, not a page-copy estimate — confirmed live to change correctly per keyword and per location.
• A de-listed/expired job answers a clean HTTP 404 — the search row is still emitted, with jobDetail: null.

Output envelope

Every record carries _input, _source and _scrapedAt. Upstream field names pass through verbatim under listing (and jobDetail when requested) — no renaming, and the <mark> highlight tags CV-Library wraps around matched keywords in the search snippet are left intact rather than stripped.

See CRAWLING_METHOD.md for the full reverse-engineering trail, including the robots.txt policy reasoning and what was NOT verified this session (category/salary/contract-type filtering).