UK Product Safety Recall Radar - Marketplace + Spike Detector

Tracks the official GOV.UK "product safety alerts, reports and recalls" feed (published mainly by the Office for Product Safety and Standards, OPSS) and adds two things the raw feed does not give you:

1. Which recalls trace back to an online marketplace. OPSS often writes "X sold via Amazon" or "sold on eBay" straight into the alert's own title, but that is free text, not a filterable field on GOV.UK. This Actor extracts the marketplace name (Amazon, eBay, Temu, AliExpress, Wish, Shein, Etsy, TikTok Shop, Facebook Marketplace, Vinted, OnBuy, Fruugo, Not on the High Street, or any other "sold via/on/through X" phrasing) so you can filter to just those, run over run.
2. Category spike detection. A rolling per-category count (kept in this Actor's own key-value store across runs) flags a product category - for example "childcare-articles-childrens-equipment" or "hobby-sports- equipment" - that is currently generating an unusually high number of serious/high-risk alerts inside a configurable time window. That is a derived early-warning signal, not just a list of alerts.

Data source

www.gov.uk/api/search.json, filtered to content_store_document_type=product_safety_alert_report_recall - the same public, keyless search index that powers the GOV.UK site's own finder at www.gov.uk/product-safety-alerts-reports-recalls. No API key, no login, no scraping of rendered HTML: every field this Actor uses (title, link, description, publish date, risk level, product category, alert type, corrective-action "measure type") comes back directly from the search API's own fields[] parameters, the same facets shown in that finder's filter sidebar. Licence: Open Government Licence v3.0.

Input

Field	Type	Default	Notes
riskLevels	array of strings	["serious", "high"]	serious, high, medium, low, not-provided
productCategories	array of strings	[] (all)	GOV.UK category slugs, e.g. toys, electrical-products
alertTypes	array of strings	[] (all)	product-safety-alert, product-safety-report, product-recall
keyword	string	""	Free-text search, e.g. a brand name
maxResults	integer	200	Alerts to fetch this run, newest first
rollingWindowDays	integer	30	Window for the category-spike count
spikeThreshold	integer	5	Alerts in the window that mark a category as spiking
onlyNewSinceLastRun	boolean	false	Only push alerts not already returned by an earlier run of this Actor
onlyMarketplaceListings	boolean	false	Only push alerts where a marketplace name was detected

Output (one row per alert)

title, url, published_at, product_alert_type, product_category, risk_level, measure_types, description, marketplace, is_marketplace_listing, event_type (unseen_before / seen_again), category_window_count, category_spike.

Example

Input:

{
  "riskLevels": ["serious", "high"],
  "onlyMarketplaceListings": true,
  "maxResults": 50
}

A resulting row:

{
  "title": "Product Safety Report: Baby Side Sleeper Wedge sold via Amazon (2609-0115)",
  "url": "https://www.gov.uk/product-safety-alerts-reports-recalls/product-safety-report-baby-side-sleeper-wedge-sold-via-amazon-2609-0115",
  "published_at": "2026-09-21T14:24:10Z",
  "product_alert_type": "product-safety-report",
  "product_category": "childcare-articles-childrens-equipment",
  "risk_level": "serious",
  "measure_types": ["removal-of-listing-by-online-marketplace"],
  "description": "Product Safety Report for Baby Side Sleeper Wedge sold via Amazon presenting a serious risk of suffocation.",
  "marketplace": "Amazon",
  "is_marketplace_listing": true,
  "event_type": "unseen_before",
  "category_window_count": 6,
  "category_spike": true
}

Notes

• Pay-per-event: recall-alert-detected is charged once per row pushed to the dataset (filtered-out rows are never charged).
• No scraping, no third-party HTTP library: every request goes straight to www.gov.uk/api/search.json over the standard library's own urllib, with retries for the ordinary transient failures of a free public API (HTTP 429/5xx, connection resets, http.client.IncompleteRead).
• event_type and the category-spike count depend on this Actor's own key-value store, which persists only across runs made under the same Actor/task on the same Apify account - not shared between users.