import streamlit as st
from PIL import Image
import torch
from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
from qwen_vl_utils import process_vision_info

@st.cache_resource
def load_model():
    processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")
    model = Qwen2VLForConditionalGeneration.from_pretrained(
        "Qwen/Qwen2-VL-7B-Instruct",
        torch_dtype="auto", device_map="auto"
    )
    return processor, model

processor, model = load_model()

def main():
    st.title("Testing Instructions Generator")
    context = st.text_area("Enter any optional context for the test cases", "")
    uploaded_files = st.file_uploader("Upload screenshots", accept_multiple_files=True, type=['png', 'jpg', 'jpeg'])

    if st.button("Describe Testing Instructions") and uploaded_files:
        with st.spinner("Generating testing instructions..."):
            images = [Image.open(file) for file in uploaded_files]
            response = get_test_instructions(context, images)
            st.markdown("## Generated Testing Instructions:")
            st.markdown(response)

def get_test_instructions(context, images):
    generated_instructions = ""

    for idx, image in enumerate(images):
        # Prepare the inputs by processing both the image and text context
        image_inputs, _ = process_vision_info([
            {
                "type": "image",
                "image": image,
            },
            {
                "type": "text",
                "text": context,
            }
        ])
        
        text = processor.apply_chat_template(
            [
                {
                    "role": "user",
                    "content": [
                        {
                            "type": "image",
                            "image": image,
                        },
                        {
                            "type": "text",
                            "text": context,
                        },
                    ],
                }
            ],
            tokenize=False,
            add_generation_prompt=True
        )
        
        inputs = processor(
            text=[text],
            images=image_inputs,
            padding=True,
            return_tensors="pt"
        )
        inputs = inputs.to("cuda")
        
        # Generate description using the model
        generated_ids = model.generate(**inputs, max_new_tokens=128)
        generated_ids_trimmed = [
            out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
        ]
        
        # Decode the generated output to get a readable string
        description = processor.batch_decode(
            generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
        )[0]
        
        generated_instructions += f"""
        ### Test Case {idx + 1}
        **Description:** {description}
        **Pre-conditions:** N/A
        **Testing Steps:** N/A
        **Expected Result:** {description}
        """
    
    return generated_instructions

if __name__ == "__main__":
    main()