from __future__ import annotations from pathlib import Path import numpy as np from iris.config import Settings, resolve_device class LocalEmbedder: def __init__(self, settings: Settings) -> None: self.settings = settings self._model = None def encode(self, texts: list[str]) -> np.ndarray: if not texts: return np.empty((0, 0), dtype=np.float32) model = self._get_model() vectors = model.encode( texts, convert_to_numpy=True, normalize_embeddings=True, show_progress_bar=False, ) return np.asarray(vectors, dtype=np.float32) def _get_model(self): if self._model is None: from sentence_transformers import SentenceTransformer self._model = SentenceTransformer( self.settings.embedding_model, device=resolve_device(self.settings.device), ) return self._model def save_embeddings(path: Path, embeddings: np.ndarray) -> None: path.parent.mkdir(parents=True, exist_ok=True) np.save(path, embeddings) def load_embeddings(path: Path) -> np.ndarray: return np.load(path)