app.py

import gradio as gr
from transformers import BlipProcessor, BlipForConditionalGeneration
from PIL import Image

print("Loading BLIP Processor and Model...")
# 1. Load the specific components directly (Bypasses the buggy pipeline names)
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
print("Model loaded successfully!")

def generate_prompt(input_img):
    if input_img is None:
        return "Please upload an image."
    try:
        # 2. Convert to RGB to prevent transparent PNG crashes
        clean_image = input_img.convert('RGB')
        
        # 3. Process the image into numbers the AI understands
        inputs = processor(clean_image, return_tensors="pt")
        
        # 4. Generate the text (max_new_tokens forces a detailed description)
        output = model.generate(**inputs, max_new_tokens=75)
        
        # 5. Decode the numbers back into human-readable text
        generated_text = processor.decode(output[0], skip_special_tokens=True)
        
        return generated_text
    except Exception as e:
        print(f"Error processing image: {e}")
        return f"System Error: {str(e)}"

# Build the User Interface
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown(
        """
        # 🔍 Image to Simple Text Interrogator
        Upload an image to reverse-engineer its contents into a simple Single line Text.
        """
    )
    
    with gr.Row():
        img_input = gr.Image(type="pil", label="Upload Image")
        text_output = gr.Textbox(label="Generated AI Prompt", lines=4, interactive=False, show_copy_button=True)
    
    btn = gr.Button("Analyze Image", variant="primary")
    
    # Connect the button and declare the API name for the blog
    btn.click(
        fn=generate_prompt, 
        inputs=img_input, 
        outputs=text_output, 
        api_name="get_prompt"
    )

# Launch with strict queuing to protect the CPU
if __name__ == "__main__":
    demo.queue(default_concurrency_limit=1).launch(server_name="0.0.0.0", server_port=7860, ssr_mode=False)