from typing import Dict, List, Any
from transformers import BlipProcessor, BlipForConditionalGeneration
from PIL import Image
import requests

class EndpointHandler():
  def __init__(self, path=""):
    self.processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
    self.model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")

  def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
    """
    data args:
        image_url (:obj: `str`): URL of the image to be captioned
    Return:
        A :obj:`list` | `dict`: will be serialized and returned
    """
    # get inputs
    image_url = data.pop("image_url", None)

    # check if image_url exists
    if image_url is None:
        return [{"error": "image_url not provided"}]

    # get image from URL
    try:
        raw_image = Image.open(requests.get(image_url, stream=True).raw).convert('RGB')
    except:
        return [{"error": "unable to load image from the provided URL"}]

    # unconditional image captioning
    inputs = self.processor(raw_image, return_tensors="pt")

    # generate captions
    out = self.model.generate(**inputs)

    # return the generated captions
    return [{"caption": self.processor.decode(out[0], skip_special_tokens=True)}]