Struct NearestNeighborQueryPlanner 

pub struct NearestNeighborQueryPlanner {
    pub config: PlannerConfig,
}

Plans k-NN queries over a heterogeneous set of HNSW shards.

Fields

config: PlannerConfig

Planner configuration.

Implementations

impl NearestNeighborQueryPlanner

pub fn new(config: PlannerConfig) -> Self

Create a new planner with the given configuration.

pub fn plan( &self, query_vec: &[f32], k: usize, shards: &[ShardInfo], ) -> QueryPlan

Produce an optimised QueryPlan for the given query vector and k.

Algorithm

1. Compute a stable query_id via FNV-1a over the query bytes.
2. Filter shards by latency budget and minimum vector count.
3. Optionally sort local shards to the front.
4. Limit to max_fanout shards.
5. Choose the execution strategy based on the mix of local/remote shards.

pub fn explain(&self, plan: &QueryPlan) -> String

Return a human-readable description of the plan.

pub fn replan_on_failure( &self, plan: &QueryPlan, failed_peer: &str, ) -> QueryPlan

Produce a revised plan after failed_peer could not be reached.

All shards owned by failed_peer are removed from the original plan’s shard list, and the strategy is recomputed from the survivors.