from transformers import BlipProcessor, BlipForConditionalGeneration
from langchain.chains import LLMChain
from langchain.schema import BaseOutputParser
from PIL import Image
import torch

# Define a simple Output Parser
class CaptionParser(BaseOutputParser):
    def parse(self, text: str):
        return text.strip()

# LangChain-compatible VLM wrapper
class BLIPImageCaptioning:
    def __init__(self):
        self.device = "cuda" if torch.cuda.is_available() else "cpu"
        self.processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base", use_auth_token=None)
        self.model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base", use_auth_token=None).to(self.device)

    def predict(self, image_path: str) -> str:
        raw_image = Image.open(image_path).convert('RGB')
        inputs = self.processor(raw_image, return_tensors="pt").to(self.device)
        out = self.model.generate(**inputs)
        caption = self.processor.decode(out[0], skip_special_tokens=True)
        return caption

# Use the BLIP model via LangChain
class ImageCaptionChain:
    def __init__(self):
        self.captioner = BLIPImageCaptioning()
        self.output_parser = CaptionParser()

    def run(self, image_path: str):
        caption = self.captioner.predict(image_path)
        return self.output_parser.parse(caption)

# ----------- Run Example -------------
if __name__ == "__main__":
    image_path = r"images\sample.jpg"  # Replace with your image path
    chain = ImageCaptionChain()
    caption = chain.run(image_path)
    print("Generated Caption:", caption)