VectorQuery

Namespace

Rowles.LeanCorpus.Search.Queries

Assembly

Rowles.LeanCorpus.dll

ANN search over vector data using HNSW when available, with a flat SIMD fallback.

public sealed class VectorQuery : Query, IEquatable<Query>

Inheritance

object

Query

VectorQuery

Implements

IEquatable<Query>

Inherited Members

Query.Boost

Query.Equals(Query)

Query.Rewrite()

Query.CreateWeight(IndexSearcher)

Extension Methods

QueryExtensions.WithBoost<T>(T, float)

QueryExtensions.And(Query, Query)

QueryExtensions.Not(Query, Query)

QueryExtensions.Or(Query, Query)

VectorQuery(string, float[], int, int, int, Query?)

Initialises a new VectorQuery for the given field and query vector.

EfSearch

HNSW search-time candidate pool size (the ef parameter). Larger values increase recall at the cost of latency. Defaults to max(64, 4 * topK).

Field

Gets the single field this query targets, or an empty value for fieldless and multi-field queries.

Filter

Optional pre-filter restricting candidates to documents whose IDs satisfy the predicate.

OversamplingFactor

Exact-rerank oversampling factor. The HNSW shortlist returns topK * OversamplingFactor candidates which are then exactly rescored. Default: 1 (no oversampling).

QueryVector

Gets the query vector used to find approximate nearest neighbours.

TopK

Gets the maximum number of nearest-neighbour results to return.

CosineSimilarity(ReadOnlySpan<float>, ReadOnlySpan<float>)

Computes cosine similarity between two vectors using SIMD where available.

Equals(object?)

Determines whether the specified object is equal to the current object.

GetHashCode()

Serves as the default hash function.

Visit(QueryVisitor)

Visits this query and its children.