v0.2.0

Added

• ScrapeDoProxyClient and AsyncScrapeDoProxyClient — route requests through Scrape.do's Proxy Mode (proxy.scrape.do:8080). Same request/response surface as the API-mode clients (execute / request / get / post), minus execute_from_url (no equivalent in proxy mode). The async variant is backed by httpx.AsyncClient and uses asyncio.sleep for retry pauses.

• Per-(api_token, parameters) httpx.Client / httpx.AsyncClient pool with bounded LRU eviction (max_pooled_clients=16 default, configurable). Two requests with the same parameters reuse the same TCP / TLS / HTTP-2 connection; the cookie jar on each pooled client is cleared after every request (Scrape.do owns the cookie lifecycle via setCookies / scrape.do-cookies / sessionId, so pooling is purely a transport concern).

• PreparedScrapeDoRequest.to_proxy_httpx_kwargs() — serializes the same data model into httpx kwargs that target the destination URL directly (the API token and Scrape.do parameters live in the proxy URL's userinfo segment, not the request).

• RequestParameters.to_proxy_url() — generates a Scrape.do Proxy-Mode connection string template (http://{api_token}:<params>@proxy.scrape.do:8080) for use with the proxy clients or with third-party tooling (Playwright / Selenium / curl).

• RequestParameters.validate_proxy_params() — cross-validates Proxy-Mode-specific parameter quirks (customHeaders defaulting to true server-side, setCookies interaction, render-mode discouragement).

• SCRAPE_DO_CA_PATH and DEFAULT_PROXY_SSL_CONTEXT in scrape_do.constants — the bundled Scrape.do CA cert and an ssl.SSLContext preloaded with system CAs plus the bundled CA. Default verify source for the proxy-mode clients so HTTPS targets validate correctly through Scrape.do's MITM step without disabling TLS verification. VERIFY_X509_STRICT is cleared so chain validation accepts Scrape.do's self-signed root (which omits the optional AKI extension); all other verification checks remain intact.

• Scrape.do's CA certificate bundled with the wheel under scrape_do.data so the SDK ships everything needed for proxy-mode TLS verification.

• Public re-exports for ScrapeDoProxyClient and AsyncScrapeDoProxyClient in scrape_do/__init__.py.

• AsyncScrapeDoClient backed by httpx.AsyncClient. Near-1:1 of the synchronous client (smart routing, retry strategy, session validation, event hooks), with every IO-bound method async/await. Sleeps between retries use asyncio.sleep rather than time.sleep.

• AsyncClientEventHooks TypedDict and AsyncSessionValidator type alias. Both are async-only — hooks return Awaitable[None] and validators return Awaitable[bool], so they can perform I/O while the request executes.

• Public re-exports for AsyncScrapeDoClient, AsyncClientEventHooks, and AsyncSessionValidator in scrape_do/__init__.py.

• ScrapeDoResponse.json(raw_response=True, **kwargs) -> Any convenience method. With raw_response=True (default) it shortcuts to httpx_response.json(); with raw_response=False it returns json.loads(self.text, **kwargs) so the post-envelope path is reachable without manual parsing.

• Example block in the package-level docstring at src/scrape_do/__init__.py showcasing a typical request flow.

Fixed

• ScrapeDoClient.post() now forwards the session_validator argument to request(). Previously the argument was accepted but silently ignored on POST calls. get() was unaffected.

v0.1.1

Added

• Curated public re-exports in scrape_do/__init__.py so common imports work as from scrape_do import ScrapeDoClient, RequestParameters, ... rather than digging into submodules.

• py.typed PEP 561 marker so downstream type-checkers (mypy, pyright) consume the package's type hints.

• Trove classifiers in package metadata — PyPI's "Python" sidebar and shields.io's pypi/pyversions badge now populate correctly.

Removed

• Empty scrape_do/namespaces/ placeholder folder (was scaffolding from before the roadmap solidified; will be replaced by plugins/ in 0.4+).

Documentation

• Planned package layout added to ROADMAP.

v0.1.0

Initial release. Synchronous client surface.

Added

• ScrapeDoClient synchronous client with request(), get(), post(), execute(), and execute_from_url() methods.

• Smart routing in ScrapeDoClient.request(): accepts kwargs, a pre-built RequestParameters, or a raw api.scrape.do URL — exactly one configuration shape per call.

• Automatic retries on Scrape.do gateway errors (429 / 502 / 510) with configurable backoff strategy (static float or callable). Default is jittered exponential.

• session_validator callback (SyncSessionValidator) for sticky-session rotation detection — when present and session_id is set, the validator decides whether to raise RotatedSessionError.

• SDK-native event hooks via SyncClientEventHooks TypedDict: request / response / retry lifecycle, distinct from httpx transport-level hooks.

• Pydantic-validated RequestParameters covering the full Scrape.do API parameter surface, including browser-action models (ClickAction, WaitAction, FillAction, ExecuteAction, ScreenShotAction, scrolling, request-completion waits).

• ScrapeDoResponse wrapper exposing the parsed JSON envelope, network requests, websocket frames, action results, screenshots, frames, plus a raw status_code passthrough.

• Cookie isolation between sequential requests on the underlying httpx.Client (prevents cross-request bleed).

• Exception hierarchy: ScrapeDoError (base), APIConnectionError, TargetError, RotatedSessionError, plus the API-layer AuthenticationError, BadRequestError, RateLimitError, ServerError, and AuthenticationThrottleError.

• Default request timeout raised to 60 seconds (from httpx's 5s default) to accommodate browser rendering and proxy round-trips.