Struct VectorScanOperator 

pub struct VectorScanOperator { /* private fields */ }

A scan operator that finds nodes by vector similarity.

This operator performs k-nearest neighbor search on vector embeddings stored in node properties. It can use an HNSW index for O(log n) search or fall back to brute-force O(n) search.

Output Schema

Returns a DataChunk with two columns:

1. Node - The matched node ID
2. Float64 - The distance/similarity score

Example

use grafeo_core::execution::operators::VectorScanOperator;

let query = vec![0.1f32, 0.2, 0.3];
let mut scan = VectorScanOperator::brute_force(store, "embedding", query, 10, DistanceMetric::Cosine);

while let Some(chunk) = scan.next()? {
    for i in 0..chunk.row_count() {
        let node_id = chunk.column(0).get_node_id(i);
        let distance = chunk.column(1).get_float64(i);
        println!("Node {:?} at distance {}", node_id, distance);
    }
}

Implementations

impl VectorScanOperator

pub fn brute_force( store: Arc<LpgStore>, property: impl Into<String>, query: Vec<f32>, k: usize, metric: DistanceMetric, ) -> Self

Creates a new vector scan operator for brute-force search.

This is suitable for small datasets (< 10K vectors) where index overhead isn’t worth it.

Arguments

• store - The LPG store to scan
• property - The property name containing vector embeddings
• query - The query vector
• k - Number of nearest neighbors to return
• metric - Distance metric to use

pub fn with_label(self, label: impl Into<String>) -> Self

Sets a label filter for brute-force search.

pub fn with_ef(self, ef: usize) -> Self

Sets the search ef parameter (accuracy vs speed tradeoff).

Higher values give more accurate results but slower search. Default is 64. For production use, 50-200 is typical.

pub fn with_min_similarity(self, threshold: f32) -> Self

Sets a minimum similarity threshold.

Results with similarity below this value will be filtered out. For cosine similarity, this should be in [-1, 1].

pub fn with_max_distance(self, threshold: f32) -> Self

Sets a maximum distance threshold.

Results with distance above this value will be filtered out.

pub fn with_chunk_capacity(self, capacity: usize) -> Self

Sets the chunk capacity for output batches.

Trait Implementations

impl Operator for VectorScanOperator

fn next(&mut self) -> OperatorResult

Pulls the next batch of data. Returns None when exhausted.

fn reset(&mut self)

Resets to initial state so you can iterate again.

fn name(&self) -> &'static str

Returns a name for debugging/explain output.