import logging import math import os logger = logging.getLogger(__name__) _model = None _LOCAL_MODEL_PATH = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "models") def _get_model(): global _model if _model is None: logger.info("Loading embedding model ...") from fastembed import TextEmbedding local_path = _LOCAL_MODEL_PATH if os.path.isdir(local_path) and os.path.isfile(os.path.join(local_path, "model_optimized.onnx")): logger.info("Loading from local path: %s", local_path) _model = TextEmbedding(model_name=local_path, cache_dir=os.path.dirname(local_path)) else: os.environ.setdefault("HF_ENDPOINT", "https://hf-mirror.com") _model = TextEmbedding("BAAI/bge-small-zh-v1.5") logger.info("Embedding model loaded") return _model def compute_embedding(text: str) -> list[float]: vec = list(_get_model().embed(text))[0] return vec.tolist() def cosine_similarity(a: list[float], b: list[float]) -> float: dot = sum(x * y for x, y in zip(a, b)) na = math.sqrt(sum(x * x for x in a)) nb = math.sqrt(sum(y * y for y in b)) if na == 0 or nb == 0: return 0.0 return dot / (na * nb)