import re
import torch
from fastapi import FastAPI, File, UploadFile
from transformers import DonutProcessor, VisionEncoderDecoderModel
from PIL import Image
from io import BytesIO

app = FastAPI()

processor = DonutProcessor.from_pretrained("naver-clova-ix/donut-base-finetuned-cord-v2", use_fast=False, cache_dir="new_cache_dir/")

model = VisionEncoderDecoderModel.from_pretrained("naver-clova-ix/donut-base-finetuned-cord-v2", cache_dir="new_cache_dir/")

device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)

task_prompt = "<s_cord-v2>"
decoder_input_ids = processor.tokenizer(task_prompt, add_special_tokens=False, return_tensors="pt").input_ids

def generateOutput(fileData):
    pil_image = Image.open(BytesIO(fileData))
    pil_image.resize((800, 600))
    pixel_values = processor(pil_image, return_tensors="pt").pixel_values

    outputs = model.generate(
        pixel_values.to(device),
        decoder_input_ids=decoder_input_ids.to(device),
        max_length=model.decoder.config.max_position_embeddings,
        pad_token_id=processor.tokenizer.pad_token_id,
        eos_token_id=processor.tokenizer.eos_token_id,
        use_cache=True,
        bad_words_ids=[[processor.tokenizer.unk_token_id]],
        return_dict_in_generate=True,
    )
    return outputs

@app.post("/ocr/")
async def analyze_image(file: UploadFile = File(...)):
    content = await file.read()
    outputs = generateOutput(content)
    sequence = processor.batch_decode(outputs.sequences)[0]
    sequence = sequence.replace(processor.tokenizer.eos_token, "").replace(processor.tokenizer.pad_token, "")
    sequence = re.sub(r"<.*?>", "", sequence, count=1).strip()  # remove first task start token
    return processor.token2json(sequence)