Struct MatryoshkaIndex 

pub struct MatryoshkaIndex {
    pub config: MatryoshkaConfig,
    /* private fields */
}

Matryoshka embedding index supporting adaptive-dimension search.

Stores embeddings at full dimensionality but can search at any prefix dimension for a speed-accuracy trade-off.

Fields

config: MatryoshkaConfig

Index configuration.

Implementations

impl MatryoshkaIndex

pub fn new(config: MatryoshkaConfig) -> Result<Self>

Create a new Matryoshka index.

Errors

Returns an error if supported_dims is empty, any dimension is zero, or any dimension exceeds full_dim.

pub fn insert( &mut self, id: VectorId, embedding: Vec<f32>, metadata: Option<HashMap<String, Value>>, ) -> Result<()>

Insert a full-dimensional embedding into the index.

Errors

Returns an error if the embedding dimension does not match full_dim.

pub fn len(&self) -> usize

Return the number of stored vectors.

pub fn is_empty(&self) -> bool

Check if the index is empty.

pub fn search( &self, query: &[f32], dim: usize, top_k: usize, ) -> Result<Vec<SearchResult>>

Search at a specific dimension by truncating embeddings to the first dim components.

Arguments

• query - Full-dimensional (or at least dim-dimensional) query vector.
• dim - The truncation dimension to use for search.
• top_k - Number of results to return.

Errors

Returns an error if dim exceeds the query length or full_dim.

pub fn funnel_search( &self, query: &[f32], top_k: usize, funnel_config: &FunnelConfig, ) -> Result<Vec<SearchResult>>

Two-phase funnel search: coarse filter at low dimension, rerank at full dimension.

1. Search at funnel_config.filter_dim for candidate_multiplier * top_k candidates.
2. Rerank those candidates at full_dim.
3. Return the top top_k.

Errors

Returns an error if the query is shorter than full_dim.

pub fn cascade_search( &self, query: &[f32], top_k: usize, dims: &[usize], reduction_factor: f32, ) -> Result<Vec<SearchResult>>

Multi-stage cascade search through multiple dimensions.

Searches through dimensions in ascending order, progressively narrowing candidates. At each stage, the candidate set is reduced by the reduction_factor.

Trait Implementations

impl Clone for MatryoshkaIndex

fn clone(&self) -> MatryoshkaIndex

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for MatryoshkaIndex

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for MatryoshkaIndex

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for MatryoshkaIndex

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.