# Changelog of Google Patent Details Scraper (`cliqtomedia/google-patents-scraper`) Actor

- **URL**: https://apify.com/cliqtomedia/google-patents-scraper/changelog.md
- **Full Actor documentation**: https://apify.com/cliqtomedia/google-patents-scraper.md

## Changelog

### Unreleased — TZ-10 details-only fallback

- Rejected the active search contract after the five-control reliability gate
  produced 0/5 usable runs with no Dataset rows or `RUN_SUMMARY` records.
- Limited active inputs to publication numbers and canonical Google Patents
  detail URLs, with a direct-detail cap of 100 saved rows.
- Kept the search/recovery implementation and deterministic tests dormant for
  provenance; the production entrypoint rejects search controls and does not
  create a search transport for valid detail-only input.

### Unreleased — TZ-10 reliability hardening

- Added bounded search-page recovery for challenge, rate-limit and transient
  failures. Residential search starts each run with a fresh session, retries a
  failed page at most once through a new session, and caps recovery attempts at
  two per run. Recovery counters are recorded in `RUN_SUMMARY` without storing
  session identifiers.

### 2026-09-19 — TZ-09 limited release preparation

- Set the runtime and schemas to a global 100 saved-row hard cap.
- Made the default run small and end-to-end: 10 search results with metadata
  detail enrichment.
- Kept Apify Residential US for search, direct HTTPS for detail and explicit
  `BLOCKED` handling for challenge responses.
- Persisted the Residential search runtime setting in `.actor/actor.json` so
  Cloud runs do not silently fall back to direct search.
- Updated README, Store copy, examples, sample output, pricing and rollback
  materials for the limited private product.
- Fresh default, direct-detail and 50-row controls passed; the first 100-row
  control hit a source challenge and the limited certification remains blocked.

### 2026-09-17 — TZ-07 search-access remediation

- Compared exact search request shapes and bounded live source evidence for
  the 10/20-row successes versus the blocked 50-row controls.
- Recorded that `pageSize` alone is not a proven cause; the allowed direct
  route remains unstable/protected and certification stays blocked.
- Reconciled live final-certification spend without changing source access,
  visibility, icon, pricing or publication state.
- Built the exact merged-main private `0.3.10` candidate and passed the
  default/detail/50/100 gates; the advertised 500-row control remained
  blocked by a source challenge before candidates.

### 2026-09-16 — Search and detail improvements

- Search public Google Patents records by keyword with filters, sorting and
  safe page and result limits.
- Fetch known patent details by publication number or canonical Google Patents
  URL, with metadata or full detail from the same public page.
- Keep publication identity, source links, family values and search totals
  separate from the number of saved rows.
- Record clear outcomes for empty results, missing pages, invalid input,
  source challenges, temporary errors and parser changes.
- Save repeated publications once and keep a run summary with input and row
  counts.

### 2026-09-14 — Detail fields and source safety

- Added bounded claims, descriptions, citations, family data and application
  identity when the source page provides them.
- Added source and legal-value notes so status and expiration fields are not
  presented as legal advice.
- Added source links without downloading PDF or image assets.