Struct VectorSearch 

pub struct VectorSearch { /* private fields */ }

Builder for vector similarity search operations

Supports both brute-force search on RecordBatches and ANN search on Lance datasets.

Implementations

impl VectorSearch

pub fn new(column: &str) -> Self

Create a new VectorSearch builder for the specified column

Arguments

• column - Name of the vector column in the data

Example

let search = VectorSearch::new("embedding");

pub fn query_vector(self, vec: Vec<f32>) -> Self

Set the query vector for similarity search

Arguments

• vec - The query vector (must match dimension of data vectors)

pub fn metric(self, metric: DistanceMetric) -> Self

Set the distance metric

Arguments

• metric - Distance metric (L2, Cosine, or Dot)

pub fn top_k(self, k: usize) -> Self

Set the number of results to return

Arguments

• k - Maximum number of results

pub fn include_distance(self, include: bool) -> Self

Whether to include distance scores in the output

Arguments

• include - If true, adds a distance column to results

pub fn distance_column_name(self, name: &str) -> Self

Set the name for the distance column

Arguments

• name - Column name for distance values (default: “_distance”)

pub fn column(&self) -> &str

Get the column name

pub fn get_query_vector(&self) -> Option<&[f32]>

Get the query vector if set

pub fn get_metric(&self) -> &DistanceMetric

Get the distance metric

pub fn get_top_k(&self) -> usize

Get the top_k value

pub async fn search(&self, data: &RecordBatch) -> Result<RecordBatch>

Perform brute-force vector search on a RecordBatch

This method computes distances for all vectors in the batch and returns the top-k results sorted by distance (ascending).

Arguments

• data - RecordBatch containing the vector column

Returns

A new RecordBatch with the top-k rows, optionally including a distance column

Example

let results = VectorSearch::new("embedding")
    .query_vector(query_vec)
    .metric(DistanceMetric::Cosine)
    .top_k(10)
    .search(&candidates)
    .await?;

pub async fn search_lance(&self, dataset: &Dataset) -> Result<RecordBatch>

Perform ANN vector search on a Lance dataset

This method uses Lance’s native ANN search via scan().nearest(), which leverages vector indices (IVF_PQ, IVF_HNSW, etc.) when available.

Arguments

• dataset - Lance dataset with vector column

Returns

A RecordBatch with the top-k nearest neighbors

Example

let dataset = lance::Dataset::open("data.lance").await?;
let results = VectorSearch::new("embedding")
    .query_vector(query_vec)
    .metric(DistanceMetric::L2)
    .top_k(10)
    .search_lance(&dataset)
    .await?;

Trait Implementations

impl Clone for VectorSearch

fn clone(&self) -> VectorSearch

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for VectorSearch

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.