from smolagents import Tool
from transformers import BlipProcessor, BlipForConditionalGeneration
from PIL import Image
import torch

class CaptionImageTool(Tool):
    name = "caption_image_tool"
    description = "Caption an image using a free Hugging Face template."
    inputs = {
        "image_path": {
            "type": "string",
            "description": "The path of the local image file to elaborate"
        }
    }
    output_type = "string"

    def __init__(self):
        super().__init__()
        self.device = "cuda" if torch.cuda.is_available() else "cpu"
        self.model = "Salesforce/blip-image-captioning-base"
        self.processor = BlipProcessor.from_pretrained(self.model)
        self.model = BlipForConditionalGeneration.from_pretrained(self.model).to(self.device)

    def forward(self, image_path: str) -> str:
        try:
            image = Image.open(image_path).convert('RGB')
            inputs = self.processor(image, return_tensors="pt").to(self.device)
            out = self.model.generate(**inputs)
            caption = self.processor.decode(out[0], skip_special_tokens=True)
            return "Image caption: " + caption
        except Exception as e:
            return f"Error caption_image is not working properly, error: {e}, please skip this tool"