import gradio as gr
from sentence_transformers import SentenceTransformer
from transformers import AutoProcessor, AutoModelForVision2Seq
from PIL import Image
import torch
from torchvision import io
from typing import Dict
from datetime import datetime
import numpy as np
import base64
import os, stat, io

# Load the model in half-precision on the available device(s)
model = AutoModelForVision2Seq.from_pretrained(
    "./SmolVLM-500M-Instruct",
    torch_dtype=torch.bfloat16,
    _attn_implementation="eager"
)
processor = AutoProcessor.from_pretrained("./SmolVLM-500M-Instruct")

def array_to_image(image_array):
    if image_array is None:
        raise ValueError("No image provided. Please upload an image before submitting.")
    # Convert numpy array to PIL Image
    image = Image.fromarray(np.uint8(image_array)).convert("RGB")
    
    return image

def generate_embeddings(text):
    model = SentenceTransformer('./all-MiniLM-L6-v2')
    embeddings = model.encode(sentences)
    return embeddings

def describe_image(image_array):
    image = array_to_image(image_array)
    
    messages = [
        {
            "role": "user",
            "content": [
                {
                    "type": "image",
                },
                {"type": "text", "text": "Make a very detailed description of the image."},
            ],
        }
    ]

    prompt = processor.apply_chat_template(messages, add_generation_prompt=True)

    inputs = processor(text=prompt, images=[image], return_tensors="pt")

    # Inference: Generation of the output
    generated_ids = model.generate(**inputs, max_new_tokens=500)
    output_ids = [
        generated_ids[len(input_ids) :]
        for input_ids, generated_ids in zip(inputs.input_ids, generated_ids)
    ]
    output_text = processor.batch_decode(
        output_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True
    )   
    # Extract the detailed description from the response
    return output_text, generate_embeddings(output_text)

# Create a Gradio interface
iface = gr.Interface(
    fn=describe_image,
    inputs=gr.Image(),
    outputs=[gr.Textbox(label="Description"), gr.JSON(label="Embeddings")],
    title="Image Description with Qwen Model",
    description="Upload an image to get a detailed description using the Qwen2-VL-7B-Instruct model."
)

# Launch the app
#iface.launch(share=True)
iface.launch()