handler.py

from typing import Dict, Any, List
import torch
import numpy as np
from PIL import Image
import base64
import io
import cv2

# Test importing the problematic packages
from transformers import Sam3Model, Sam3Processor


class EndpointHandler:
    """
    Minimal test handler to isolate dependency loading issues
    This handler imports all the same dependencies but doesn't execute SAM3 inference
    """

    def __init__(self, path: str = ""):
        """
        Initialize the handler - test dependency loading without heavy model loading

        Args:
            path: Path to model weights (unused in this test)
        """
        self.device = "cuda" if torch.cuda.is_available() else "cpu"
        print(f"✅ Test handler initialized successfully on device: {self.device}")

        # Test that we can import the SAM3 classes without loading the model
        print(f"✅ Successfully imported Sam3Model: {Sam3Model}")
        print(f"✅ Successfully imported Sam3Processor: {Sam3Processor}")

        # Test other dependencies
        print(f"✅ PyTorch version: {torch.__version__}")
        print(f"✅ NumPy version: {np.__version__}")
        print(f"✅ PIL (Pillow) available: {Image}")
        print(f"✅ OpenCV available: {cv2.__version__}")

        # Don't actually load the model to avoid memory/download issues
        self.model = None
        self.processor = None

        print("✅ Minimal test handler ready - all dependencies loaded successfully!")

    def __call__(self, data: Dict[str, Any]) -> Dict[str, Any]:
        """
        Minimal test endpoint that returns success without actual inference

        Args:
            data: Input data (will be ignored in this test)

        Returns:
            Simple success response to verify the handler works
        """
        try:
            print("📝 Test handler called with data keys:", list(data.keys()) if data else "No data")

            # Test basic operations with imported libraries
            test_array = np.array([1, 2, 3])
            test_tensor = torch.tensor([1.0, 2.0, 3.0])

            print(f"✅ NumPy test array: {test_array}")
            print(f"✅ PyTorch test tensor: {test_tensor}")
            print(f"✅ Device available: {self.device}")

            # Return a successful test response
            return {
                "status": "success",
                "message": "✅ All dependencies loaded and working correctly!",
                "test_results": {
                    "numpy_test": test_array.tolist(),
                    "torch_test": test_tensor.tolist(),
                    "device": self.device,
                    "torch_version": torch.__version__,
                    "numpy_version": np.__version__,
                    "opencv_version": cv2.__version__,
                    "transformers_classes_available": {
                        "Sam3Model": str(Sam3Model),
                        "Sam3Processor": str(Sam3Processor)
                    }
                },
                "input_data_received": data is not None,
                "handler_type": "minimal_test_handler"
            }

        except Exception as e:
            print(f"❌ Error in test handler: {str(e)}")
            return {
                "status": "error",
                "message": f"Test handler failed: {str(e)}",
                "error_type": type(e).__name__,
                "handler_type": "minimal_test_handler"
            }