Raised when a request is invalid or the user lacks permission. The default status is 400 (Bad Request). Use status=403 explicitly for permiss

Sklearn-compatible wrapper around the Rust mwmbl_rank.RustXGBPipeline. All feature extraction and XGBoost training/inference runs in Rust.

Provider that uses a static set of domains.

Main crawler class that manages Redis connections, URL queues, and crawling operations. This class encapsulates all the crawler function

Provider that implements the built-in spam detection and exclusion rules.

A hyperlink found on a crawled page.

Decorator that applies MMR diversity re-ranking to a wrapped ranker's results. Demotes (never drops) same-domain / near-duplicate results. Delega

Raised when a URL resolves to an address we refuse to fetch.

Provider that combines multiple blacklist providers.

The state of the document in the index. A value of None indicates an organic search result.

Detailed results for a single evaluation job.

Provider that fetches HaGeZi blocklist.

Learning-to-rank ranker. Accepts any model with a predict(DataFrame) -> array interface. The DataFrame passed to predict has columns: qu

A batch as stored/returned by the server, with the user ID replaced by its hash.

A single crawled URL and its result.

Extracted text content from a successfully crawled page.

A parsed search recipe. ``request`` and ``response`` are the raw YAML maps.

Minimal ranker stub exposing the methods MMRRanker delegates to.

Response returned when a new API key is created. The raw `key` value is only returned once — store it securely.

A single API key entry returned by the list endpoint. The raw key value is intentionally omitted.

A batch of crawled pages submitted by a crawler client.

`done` — terminal event sent once when the stream finishes.

`error` — the pipeline crashed; the stream will end.

Summary of an evaluation job, as returned by the list-jobs endpoint.

A dataset as stored by the server, with the user ID replaced by its hash.

Base class for jusText exceptions.

`link_followed` — an outbound link from a crawled page was fetched and added to the candidate pool.

A ranking model that has access to the gold standard and will return back any matching queries in the order they occur in the gold standard

`page_fetched` — a promoted page was crawled.

Object representing one block of text in HTML.

A single search result to be indexed.

A single ranked search result. Carried directly by ``result_promoted`` events and inside the ``results`` list. ``title`` and ``extract`` are

A set of search results submitted for indexing via the results API.

`results` — the current authoritative ranking. Emitted progressively after each source returns, and once more as the final ranking. Clients s

Search response with optional Mwmbl usage metadata.

`source_failed` — a source errored or timed out and contributed nothing.

`source_returned` — a source finished successfully.

`source_started` — a source's query task has been launched (one per source).

Response schema containing vote statistics for multiple URLs.

Statistics for votes on a specific search result.

Response returned after successfully submitting a WASM file for evaluation.