Struct Qwen3Provider 

pub struct Qwen3Provider { /* private fields */ }

Qwen3 embedding provider backed by candle-nn.

Unlike FastembedProvider (ONNX), this uses candle for pure Rust GPU/CPU inference. The embed() method takes &self, so no Mutex is needed.

Implementations

impl Qwen3Provider

pub fn new(model_id: &str) -> Result<Self>

Create a new provider using CPU with F32 precision.

Downloads the model from HuggingFace on first call (cached locally).

pub fn with_options( model_id: &str, device: Device, dtype: DType, max_length: usize, ) -> Result<Self>

Create with custom device (GPU), dtype, and max sequence length.

pub fn model_id(&self) -> &str

The HuggingFace model repo ID.

Trait Implementations

impl EmbeddingProvider for Qwen3Provider

fn dim(&self) -> usize

The dimensionality of the embedding vectors.

fn name(&self) -> &str

Provider name for display.

fn embed(&self, text: &str) -> Result<EmbeddingResult>

Embed a single text string.

fn embed_batch(&self, texts: &[&str]) -> Result<Vec<EmbeddingResult>>

Embed multiple texts in batch.