Struct DiskANN 

pub struct DiskANN<D>
where
    D: Distance<f32> + Send + Sync + Copy + Clone + 'static,
{
    pub dim: usize,
    pub num_vectors: usize,
    pub max_degree: usize,
    pub distance_name: String,
    /* private fields */
}

Main struct representing a DiskANN index (generic over distance)

Fields

dim: usize

Dimensionality of vectors in the index

num_vectors: usize

Number of vectors in the index

max_degree: usize

Maximum number of edges per node

distance_name: String

Informational: type name of the distance (from metadata)

Implementations

impl<D> DiskANN<D>

where D: Distance<f32> + Send + Sync + Copy + Clone + 'static,

pub fn build_index_default( vectors: &[Vec<f32>], dist: D, file_path: &str, ) -> Result<Self, DiskAnnError>

Build with default parameters: (M=32, L=256, alpha=1.2).

pub fn build_index_with_params( vectors: &[Vec<f32>], dist: D, file_path: &str, p: DiskAnnParams, ) -> Result<Self, DiskAnnError>

Build with a DiskAnnParams bundle.

impl<D> DiskANN<D>

where D: Distance<f32> + Default + Send + Sync + Copy + Clone + 'static,

Extra sugar when your distance type implements Default (most unit-struct metrics do).

pub fn build_index_default_metric( vectors: &[Vec<f32>], file_path: &str, ) -> Result<Self, DiskAnnError>

Build with default params and D::default() metric.

pub fn open_index_default_metric(path: &str) -> Result<Self, DiskAnnError>

Open an index using D::default() as the distance (matches what you built with).

impl<D> DiskANN<D>

where D: Distance<f32> + Send + Sync + Copy + Clone + 'static,

pub fn build_index( vectors: &[Vec<f32>], max_degree: usize, build_beam_width: usize, alpha: f32, dist: D, file_path: &str, ) -> Result<Self, DiskAnnError>

Builds a new index from provided vectors

Arguments

• vectors - The vectors to index (slice of Vec)
• max_degree - Maximum edges per node (M ~ 24-64)
• build_beam_width - Construction L (e.g., 128-400)
• alpha - Pruning parameter (1.2–2.0)
• dist - Any anndists::Distance<f32> (e.g., DistL2)
• file_path - Path of index file

pub fn open_index_with(path: &str, dist: D) -> Result<Self, DiskAnnError>

Opens an existing index file, supplying the distance strategy explicitly.

pub fn search_with_dists( &self, query: &[f32], k: usize, beam_width: usize, ) -> Vec<(u32, f32)>

Searches the index for nearest neighbors using a best-first beam search. Termination rule: continue while the best frontier can still improve the worst in working set. Like search but also returns the distance for each neighbor. Distances are exactly the ones computed during the beam search.

pub fn search(&self, query: &[f32], k: usize, beam_width: usize) -> Vec<u32>

search but only return neighbor ids

pub fn get_vector(&self, idx: usize) -> Vec<f32>

Gets a vector from the index (useful for tests)