Struct VectorQueryOptions 

pub struct VectorQueryOptions {
    pub index_type: IndexType,
    pub quantization: QuantizationKind,
    pub query_dim: Option<u32>,
    pub oversample: u8,
    pub meta_token_budget: Option<u8>,
    pub ef_search: usize,
    pub k: usize,
    pub target_recall: f32,
}

All knobs the vector planner exposes.

This struct is constructed by nodedb-sql’s engine rules from query hints and passed through to nodedb-vector’s search entry-points. Every field has a sensible default so callers can build with ..Default::default().

Fields

index_type: IndexType

Graph index type to use.

quantization: QuantizationKind

Quantization codec for in-graph distance computation.

query_dim: Option<u32>

Matryoshka coarse dimension for adaptive-dim querying.

When Some(d), the first d dimensions are used for coarse-pass traversal; a full-dimension rerank follows over top-ef_search candidates. None means use the full embedding dimension.

oversample: u8

BBQ / RaBitQ rerank oversample multiplier.

The actual rerank candidate count is oversample * ef_search. Ignored when quantization == None.

meta_token_budget: Option<u8>

MetaEmbed Meta Token budget for budgeted MaxSim.

None uses the default token count for the collection.

ef_search: usize

Beam width for HNSW / Vamana search. Must be >= k.

k: usize

Number of nearest neighbours to return.

target_recall: f32

Target recall in [0.0, 1.0]. The planner uses this to pick the quantization tier and oversample ratio when the caller does not specify them explicitly.

Trait Implementations

impl Clone for VectorQueryOptions

fn clone(&self) -> VectorQueryOptions

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for VectorQueryOptions

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

Formats the value using the given formatter.

impl Default for VectorQueryOptions

fn default() -> Self

Returns the “default value” for a type.