Struct NemotronEmbedder 

pub struct NemotronEmbedder { /* private fields */ }

NVIDIA Embed Nemotron 8B embedder using realizar’s GGUF infrastructure

Produces 4096-dimensional embeddings from a Llama 3.1 8B-based model. Supports asymmetric retrieval with query/passage prefixes.

Requires the nemotron feature to be enabled.

Example

use trueno_rag::embed::{NemotronEmbedder, NemotronConfig, Embedder};

let config = NemotronConfig::new("models/NV-Embed-v2-Q4_K.gguf")
    .with_gpu(true);
let embedder = NemotronEmbedder::new(config)?;

let query_emb = embedder.embed_query("What is machine learning?")?;
let doc_emb = embedder.embed_document("Machine learning is a branch of AI...")?;

Implementations

impl NemotronEmbedder

pub fn new(config: NemotronConfig) -> Result<Self>

Create a new Nemotron embedder from configuration

Errors

Returns an error if:

• The model file doesn’t exist or can’t be read
• The model is not a valid GGUF file
• The model architecture is not compatible

pub fn config(&self) -> &NemotronConfig

Get the configuration

Trait Implementations

impl Debug for NemotronEmbedder

Available on crate feature nemotron only.

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

Formats the value using the given formatter.

impl Embedder for NemotronEmbedder

Available on crate feature nemotron only.

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

Embed a single text

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

Batch embed multiple texts

fn dimension(&self) -> usize

Get embedding dimension

fn model_id(&self) -> &str

Get model identifier

fn embed_query(&self, query: &str) -> Result<Vec<f32>>

Embed a query (may use query prefix)

fn embed_document(&self, document: &str) -> Result<Vec<f32>>

Embed a document (may use document prefix)

fn embed_chunks(&self, chunks: &mut [Chunk]) -> Result<()>

Embed chunks and update them in place