Frontmatter MCP

"""Embedding model module for semantic search.""" from typing import Any, cast import numpy as np from numpy.typing import NDArray from sentence_transformers import SentenceTransformer class EmbeddingModel: """Lazy-loading wrapper for sentence-transformers model.""" def __init__(self, name: str) -> None: """Initialize the embedding model wrapper. Args: model_name: Name of the sentence-transformers model to use. """ self._name = name self._model: SentenceTransformer | None = None @property def name(self) -> str: """Get the model name.""" return self._name @property def model(self) -> SentenceTransformer: """Get the model, loading it if necessary.""" if self._model is None: self._load_model() assert self._model is not None return self._model def _load_model(self) -> None: """Load the sentence-transformers model.""" self._model = SentenceTransformer(self._name) @property def is_loaded(self) -> bool: """Check if the model is loaded.""" return self._model is not None def get_dimension(self) -> int: """Get the embedding dimension. Returns: The dimension of the embedding vectors. """ dim = self.model.get_sentence_embedding_dimension() if dim is None: raise RuntimeError("Model does not report embedding dimension") return dim def encode(self, text: str) -> NDArray[np.floating[Any]]: """Encode text to embedding vector. Args: text: Text to encode. Returns: Embedding vector as numpy array. """ return cast(NDArray[np.floating[Any]], self.model.encode(text))