v4.3.0

Changed

• search_engine and search_engine_batch use a SERP API zone instead
  of Web Unlocker. New SERP_ZONE env var (default mcp_serp); the
  zone is created on startup if missing.
• Bing is the default engine. Both engines return JSON:
  {organic: [{link, title, description}]}.
• geo_location is renamed to country, and cursor to page (a
  number from 0, 10 results per page). The old names return an
  invalid_input error.
• New time_range: hour, day, week, month, year, any
  (default). Bing supports only day, week and year.
• A search that returns no data gives a service_unavailable error.
• search_engine_batch stops at 45 s and returns the searches that
  finished; the rest get a timeout error.
• Each zone is created on its own, so one failed zone no longer skips
  the others.
• Fetch API (scrape_as_markdown, scrape_batch) request errors use
  the same codes and texts as the hosted server: rate_limited (was
  busy), service_unavailable (was service_error), invalid_input
  (was invalid_url), quota_exceeded (new, usage limit reached) and a
  generic auth_failed (also on 407). A site limiting requests is now
  site_rate_limited.

Removed

• Yandex.