from sentence_transformers import SentenceTransformer
import pandas as pd
from app.config import settings

df = None
model = None

def load_icd_dataset():
    global df, model
    try:
        df = pd.read_pickle(settings.ICD_DATA_PATH)
        model = SentenceTransformer('all-MiniLM-L6-v2')

        if "encoded" not in df.columns:
            df["encoded"] = df["Description"].apply(lambda x: model.encode(x, normalize_embeddings=True))
            df.to_pickle(settings.ICD_DATA_PATH)
        return df, model
    except FileNotFoundError:
        df, model = None, None
        return df, model

# Load dataset once at import
df, model = load_icd_dataset()