Module embeddings 

Embeddings Module

This module provides embedding generation functionality using the Candle ML framework. It includes a default BERT-based embedding generator and a trait for custom implementations.

The EmbeddingGenerator uses the all-MiniLM-L6-v2 model by default, which provides 384-dimensional embeddings suitable for general-purpose text similarity tasks.

Examples

use vectorlite::{EmbeddingGenerator, EmbeddingFunction};

// Create a new embedding generator
let generator = EmbeddingGenerator::new()?;

// Generate embedding for text
let embedding = generator.generate_embedding("Hello world")?;
println!("Generated {}D embedding", embedding.len());
 
// Get the dimension
println!("Embedding dimension: {}", generator.dimension());

Structs

EmbeddingGenerator

BERT-based embedding generator using the Candle framework

Enums

EmbeddingError

Errors that can occur during embedding generation

Traits

EmbeddingFunction

Trait for embedding generation functions

Type Aliases

Result

Result type for embedding operations