Struct EmbeddingsClient 

pub struct EmbeddingsClient { /* private fields */ }

The real Embedder, protocol-aware over the two wire APIs newt speaks: Ollama POST /api/embeddings ({model, prompt} → {embedding: [f32]}) and OpenAI-compatible POST /v1/embeddings ({model, input} → {data: [{embedding: [f32]}]}, e.g. vLLM serving an embedding model). Mirrors the summarizer’s HTTP discipline — a configurable timeout, optional bearer auth, and exponential-backoff retry (embedding a whole repo is many requests; transient failures recover).

Implementations

impl EmbeddingsClient

pub fn new( url: impl Into<String>, model: impl Into<String>, kind: BackendKind, api_key: Option<String>, timeout_secs: u64, retries: u32, ) -> Self

pub async fn embed_batch(&self, texts: &[String]) -> Result<Vec<Vec<f32>>>

Embed several texts (sequential — one request each, deterministic order). Fails fast on the first error so the index never holds a partial set.

Trait Implementations

impl Embedder for EmbeddingsClient

fn embed<'life0, 'life1, 'async_trait>( &'life0 self, text: &'life1 str, ) -> Pin<Box<dyn Future<Output = Result<Vec<f32>>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Embed one text into a vector.