Trait VectorIndex

Source

pub trait VectorIndex: Send + Sync {
    // Required methods
    fn add(&self, id: Uuid, vector: &[f32]) -> Result<()>;
    fn remove(&self, id: Uuid) -> Result<()>;
    fn search<'life0, 'life1, 'async_trait>(
        &'life0 self,
        query: &'life1 [f32],
        limit: usize,
    ) -> Pin<Box<dyn Future<Output = Result<Vec<(Uuid, f32)>>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait;
    fn filtered_search<'life0, 'life1, 'life2, 'async_trait>(
        &'life0 self,
        query: &'life1 [f32],
        limit: usize,
        filter: &'life2 (dyn Fn(Uuid) -> bool + Send + Sync),
    ) -> Pin<Box<dyn Future<Output = Result<Vec<(Uuid, f32)>>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait,
             'life2: 'async_trait;
    fn save(&self, path: &Path) -> Result<()>;
    fn load(&self, path: &Path) -> Result<()>;
    fn len(&self) -> usize;

    // Provided method
    fn is_empty(&self) -> bool { ... }
}

Expand description

A pluggable approximate-nearest-neighbour index over memory embeddings.

search / filtered_search are async (v0.5.18). The PostgreSQL (pgvector) backend runs a real sqlx query on the ambient Tokio runtime, so the read path .awaits it directly instead of bridging through a block_on — which, inside the server/CLI #[tokio::main] runtime, could panic (“Cannot start a runtime from within a runtime”) or deadlock. The in-memory USearch backend does synchronous CPU work inside its async method, so it needs no runtime and assumes no runtime flavor. add / remove / save / load / len remain synchronous.