Crate hypembed 

HypEmbed

A pure-Rust, local-first embedding inference library.

HypEmbed loads transformer model weights, tokenizes text, runs a forward pass, and returns normalized embedding vectors — all in safe Rust with zero external ML runtime dependencies.

Quick Start

use hypembed::{Embedder, EmbeddingOptions, PoolingStrategy};

let model = Embedder::load("./model").unwrap();
let options = EmbeddingOptions::default()
    .with_normalize(true)
    .with_pooling(PoolingStrategy::Mean);
let embeddings = model.embed(&["hello world", "rust embeddings"], &options).unwrap();

Re-exports

pub use pipeline::embedder::Embedder;

pub use pipeline::embedder::EmbeddingOptions;

pub use model::pool::PoolingStrategy;

pub use error::HypEmbedError;

Modules

error

model

pipeline

tensor

tokenizer