Struct LshIndex 

pub struct LshIndex {
    pub vectors: Vec<Vec<f64>>,
    pub buckets: Vec<LshBucket>,
    pub hashers: Vec<LshHasher>,
    pub dim: usize,
    pub num_tables: usize,
    pub num_hashes: usize,
}

Approximate nearest-neighbour index using Locality-Sensitive Hashing.

Fields

vectors: Vec<Vec<f64>>

All inserted vectors (indexed by position in this Vec).

buckets: Vec<LshBucket>

One bucket table per LSH table.

hashers: Vec<LshHasher>

One hasher per LSH table.

dim: usize

Dimensionality of the vectors.

num_tables: usize

Number of hash tables.

num_hashes: usize

Number of hash bits per table.

Implementations

impl LshIndex

pub fn new(dim: usize, num_tables: usize, num_hashes: usize, seed: u64) -> Self

Create a new LSH index.

• dim — vector dimensionality
• num_tables — number of independent hash tables (more → higher recall)
• num_hashes — number of hash bits per table (more → higher precision)
• seed — XorShift-64 seed for reproducibility

pub fn insert(&mut self, id: usize, vector: &[f64])

Insert a vector with the given id into all hash tables.

pub fn cosine_similarity(a: &[f64], b: &[f64]) -> f64

Compute cosine similarity between two vectors.

Returns 0.0 if either vector has zero magnitude.

pub fn search(&self, query: &[f64], k: usize) -> Vec<(usize, f64)>

Search for the top-k approximate nearest neighbours of query.

Collects candidates from all tables, deduplicates them, computes cosine similarity, and returns the top-k sorted by descending similarity. If the query itself was indexed (same vector), it may appear in results.

pub fn len(&self) -> usize

Return the number of indexed vectors.

pub fn is_empty(&self) -> bool

Return true if no vectors have been indexed.

pub fn clear(&mut self)

Remove all indexed vectors and clear all hash tables.