Changelog

4.0.1 — 2026-07-17

Changed

• Documentation: expanded README — API docs, signup/dashboard links, badges, and links to the other official clients.

4.0.0 — unreleased

Complete rewrite of the official Python client. The previous 3.x releases were generated from the OpenAPI specification; 4.0 is a hand-authored, idiomatic Python package with a much smaller public surface.

Breaking changes

There is no migration shim — 3.x and 4.x are different libraries that happen to share a name on PyPI. Stay on webscraping_ai == 3.2.1 if you cannot update your call sites yet.

• The class-per-tag layout (AIApi, HTMLApi, SelectedHTMLApi, TextApi, AccountApi) is gone. Use a single Client (or AsyncClient) instead.
• All endpoint methods now live directly on the client and use snake_case names matching the URL path:
  • Client.html(...), Client.text(...)
  • Client.selected(...), Client.selected_multiple(...)
  • Client.question(...), Client.fields(...)
  • Client.account()
• The configuration object is gone. Pass api_key (and optionally base_url, timeout, http_client) directly to Client(...).
• The transport stack moved from urllib3 to httpx. The only runtime dependency is now httpx.
• Errors moved from a single ApiException to a typed hierarchy mirroring the documented HTTP status codes: BadRequestError (400), PaymentRequiredError (402), AuthenticationError (403), RateLimitError (429), ServerError (500), GatewayTimeoutError (504), plus APITimeoutError and APIConnectionError for transport failures. They all share a WebScrapingAIError base class.

Added

• AsyncClient — async/await counterpart to Client, sharing the same method signatures and error hierarchy.
• Native support for the API's mixed query-string encodings (deepObject for headers / fields, form+explode for selectors, flat for everything else) via a custom encoder.
• py.typed marker; the package ships with type information.