denselinkage.blocking.DenseBlockingIndex

class denselinkage.blocking.DenseBlockingIndex(*, searchable: SearchableIndex, embedder: Embedder, records_by_id: dict[str, Record], top_k: int, similarity_threshold: float)[source]

Bases: BlockingIndex

Immutable artifact built by DenseBlocker for one reference set.

Owns the reference records’ SearchableIndex and embedder, and generates CandidatePair objects for a query record set. Each pair is oriented record_a = indexed (left/reference) record, record_b = query record. top_k / similarity_threshold default to the originating spec’s values and may be overridden per query.

property searchable: SearchableIndex

The built nearest-neighbour index over the reference vectors.

property embedder: Embedder

The embedder that encodes queries (and encoded the reference set).

property records: Mapping[str, Record]

The indexed reference records, keyed by id (read-only view).

property top_k: int

Default neighbours per query (overridable per query()).

property similarity_threshold: float

Default minimum similarity for a candidate (overridable per query).