Skip to main content

Embedder

semtree_embed

Trait Embedder 

Source 
pub trait Embedder: Send + Sync {
    // Required methods
    fn embed<'life0, 'life1, 'life2, 'async_trait>(
        &'life0 self,
        texts: &'life1 [&'life2 str],
    ) -> Pin<Box<dyn Future<Output = Result<Vec<Embedding>, EmbedError>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait,
             'life2: 'async_trait;
    fn dimension(&self) -> usize;
    fn model_id(&self) -> &str;

    // Provided methods
    fn embed_one<'life0, 'life1, 'async_trait>(
        &'life0 self,
        text: &'life1 str,
    ) -> Pin<Box<dyn Future<Output = Result<Embedding, EmbedError>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait { ... }
    fn max_batch_size(&self) -> usize { ... }
    fn fingerprint(&self) -> String { ... }
}

Required Methods§

Source

fn embed<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, texts: &'life1 [&'life2 str], ) -> Pin<Box<dyn Future<Output = Result<Vec<Embedding>, EmbedError>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Source

fn dimension(&self) -> usize

Number of dimensions every vector this embedder produces has.

A store built for one dimension cannot hold vectors of another, so an index persists this and refuses to mix them.

Source

fn model_id(&self) -> &str

Stable identifier for the model behind this embedder, e.g. "fastembed:AllMiniLML6V2" or "openai:text-embedding-3-small".

It must stay the same across runs and versions for the same model, and differ whenever the produced vectors would be incompatible. It is the discriminating half of fingerprint.

Provided Methods§

Source

fn embed_one<'life0, 'life1, 'async_trait>( &'life0 self, text: &'life1 str, ) -> Pin<Box<dyn Future<Output = Result<Embedding, EmbedError>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Source

fn max_batch_size(&self) -> usize

Largest number of texts a single embed call should receive. Callers indexing many chunks split their work into groups of at most this size, bounding peak memory (local models) and request size (remote APIs). The default suits the on-device models; raise or lower it to match a backend’s limits.

Source

fn fingerprint(&self) -> String

Fingerprint an index stores to detect that it was built with a different embedder. Re-indexing is required whenever this changes.

Dyn Compatibility§

This trait is dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementors§