from typing import Dict, List, Any
from transformers import CLIPTokenizer, CLIPModel
import numpy as np


class EndpointHandler:
    def __init__(self, path=""):
        # load the model
        repo_id = "openai/clip-vit-large-patch14-336"
        self.model = CLIPModel.from_pretrained(repo_id)
        self.tokenizer = CLIPTokenizer.from_pretrained(repo_id)

    def __call__(self, data: Dict[str, Any]) -> List[float]:
        """
         data args:
              inputs (:obj: `str` | `PIL.Image` | `np.array`)
              kwargs
        Return:
              A :obj:`list` | `dict`: will be serialized and returned
        """
        # compute the embedding of the input
        query = data["inputs"]
        inputs = self.tokenizer(query, padding=True, return_tensors="pt")
        text_features = self.model.get_text_features(**inputs)
        text_features = text_features.detach().numpy()
        input_embedding = text_features[0]

        # normalize the embedding
        input_embedding = input_embedding / np.linalg.norm(input_embedding)

        return input_embedding.tolist()