from sentence_transformers import SentenceTransformer
from PIL import Image
from fastapi import UploadFile
from typing import List, Optional
import torch
import os

# Set multiple environment variables to redirect all caching
os.environ["TRANSFORMERS_CACHE"] = "/app/cache"
os.environ["HF_HOME"] = "/app/cache"
os.environ["XDG_CACHE_HOME"] = "/app/cache"  # Additional variable for broader compatibility

# Ensure cache directory exists (redundant with Dockerfile but added for safety)
cache_dir = "/app/cache"
if not os.path.exists(cache_dir):
    os.makedirs(cache_dir, exist_ok=True)

# Load model with custom cache directory
model = SentenceTransformer("clip-ViT-B-32")

def get_text_embedding(text: str) -> Optional[List[float]]:
    try:
        embedding = model.encode(text, convert_to_tensor=True).cpu().numpy().tolist()
        return embedding
    except Exception as e:
        print(f"Error generating text embedding: {e}")
        return None

def get_image_embedding(image_file: UploadFile) -> Optional[List[float]]:
    try:
        image = Image.open(image_file.file).convert("RGB").resize((224, 224))
        embedding = model.encode(image, convert_to_tensor=True).cpu().numpy().tolist()
        return embedding
    except Exception as e:
        print(f"Error generating image embedding: {e}")
        return None