Struct Embedder 

pub struct Embedder { /* private fields */ }

Loads sentence-transformers/all-MiniLM-L6-v2 once and computes normalized sentence embeddings from it.

Construction fetches the model’s config.json, tokenizer.json, and model.safetensors from the Hugging Face Hub on first use (cached under the platform cache directory afterward), and loads them into a CPU-only candle BERT model. Reuse one Embedder across calls to Self::embed rather than reloading the model per call.

Implementations

impl Embedder

pub fn load() -> Result<Self, EmbedError>

Fetches (or loads from cache) and initializes the embedding model.

Errors

Returns EmbedError if the platform cache directory cannot be resolved, the model files cannot be fetched from the Hugging Face Hub, or the fetched files cannot be parsed/loaded as a BERT model.

pub fn embed(&self, text: &str) -> Result<Vec<f32>, EmbedError>

Computes a 384-dimensional, mean-pooled, L2-normalized sentence embedding for text.

Errors

Returns EmbedError if tokenization or model inference fails.

Trait Implementations

impl Debug for Embedder

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

Formats the value using the given formatter.