embedder.py

import faiss
from sentence_transformers import SentenceTransformer
import numpy as np
import os

# Set up cache directory in a writable location
cache_dir = os.path.join(os.getcwd(), ".cache")
os.makedirs(cache_dir, exist_ok=True)
os.environ['HF_HOME'] = cache_dir
os.environ['TRANSFORMERS_CACHE'] = cache_dir

# Initialize model as None - will be loaded lazily
_model = None

def preload_model():
    """Preload the sentence transformer model at startup"""
    global _model
    if _model is None:
        print("Preloading sentence transformer model...")
        try:
            _model = SentenceTransformer("all-MiniLM-L6-v2", cache_folder=cache_dir)
            print("Model preloading completed")
        except Exception as e:
            print(f"Error loading model: {e}")
            # Fallback to a different model if the first one fails
            try:
                _model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2", cache_folder=cache_dir)
                print("Fallback model preloading completed")
            except Exception as e2:
                print(f"Error loading fallback model: {e2}")
                raise
    return _model

def get_model():
    """Get the sentence transformer model, loading it lazily if needed"""
    global _model
    if _model is None:
        print("Warning: Model not preloaded, loading now...")
        return preload_model()
    return _model

def build_faiss_index(chunks):
    model = get_model()
    embeddings = model.encode(chunks)
    dimension = embeddings.shape[1]
    index = faiss.IndexFlatL2(dimension)
    index.add(np.array(embeddings))
    return index, chunks