import requests
import json
import torch
from PIL import Image
from transformers import AutoProcessor, AutoModelForCausalLM 

class EndpointHandler():
    def __init__(self, path=""):
        self.device = "cuda:0" if torch.cuda.is_available() else "cpu"
        self.torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

        self.model = AutoModelForCausalLM.from_pretrained("microsoft/Florence-2-large", torch_dtype=self.torch_dtype, trust_remote_code=True).to(self.device)
        self.processor = AutoProcessor.from_pretrained("microsoft/Florence-2-large", trust_remote_code=True)


    def predict_image(self, url, task, prompt):
        image = Image.open(requests.get(url, stream=True).raw)

        inputs = self.processor(text=task + prompt, images=image, return_tensors="pt").to(self.device, self.torch_dtype)

        generated_ids = self.model.generate(
            input_ids=inputs["input_ids"],
            pixel_values=inputs["pixel_values"],
            max_new_tokens=4096,
            num_beams=3,
            do_sample=False
        )
        generated_text = self.processor.batch_decode(generated_ids, skip_special_tokens=False)[0]

        parsed_answer = self.processor.post_process_generation(generated_text, task=task, image_size=(image.width, image.height))
        return parsed_answer
        

    def __call__(self, event):
        if "inputs" not in event:
            return {
                "statusCode": 400,
                "body": json.dumps("Error: Please provide an 'inputs' parameter."),
            }  

        inputs = event["inputs"]
        url = inputs["url"]
        task = inputs["task"]
        prompt = inputs["prompt"]
        parsed_answer = self.predict_image(url, task, prompt)

        return {
                "statusCode": 200,
                "body": json.dumps(parsed_answer),
            }