handler.py

import torch
import string
from PIL import Image
from torchvision import transforms
import io

class EndpointHandler():
    def __init__(self, path=""):
        # TorchScript model load ho raha hai
        self.model = torch.jit.load(f"{path}/master_brain_2026_final.pt")
        self.model.eval()
        self.chars = string.ascii_uppercase + string.ascii_lowercase + string.digits + "@#$%&="
        self.blank_index = len(self.chars)
        self.preprocess = transforms.Compose([
            transforms.Resize((35, 142)),
            transforms.ToTensor(),
            transforms.Normalize((0.5,), (0.5,))
        ])

    def __call__(self, data):
        inputs = data.pop("inputs", data)
        # Base64 ya raw bytes ko handle karne ke liye
        if isinstance(inputs, str):
            import base64
            inputs = Image.open(io.BytesIO(base64.b64decode(inputs))).convert('RGB')
        else:
            inputs = Image.open(io.BytesIO(inputs)).convert('RGB')
        
        img_tensor = self.preprocess(inputs).unsqueeze(0)
        with torch.no_grad():
            logits = self.model(img_tensor)
            max_indices = torch.argmax(logits, dim=2).squeeze()
            res = []
            prev = -1
            for idx in max_indices:
                val = idx.item()
                if val != self.blank_index and val != prev:
                    res.append(self.chars[val])
                prev = val
        return {"prediction": "".join(res).strip()}