import PIL.Image
import gradio as gr
import base64
import time
import os
import google.generativeai as genai

import pathlib

txt_model = genai.GenerativeModel('gemini-pro')
vis_model = genai.GenerativeModel('gemini-pro-vision')

import os

GOOGLE_API_KEY=os.getenv('GOOGLE_API_KEY')

genai.configure(api_key=GOOGLE_API_KEY)

# Image to Base 64 Converter
def image_to_base64(image_path):
    with open(image_path, 'rb') as img:
        encoded_string = base64.b64encode(img.read())
    return encoded_string.decode('utf-8')

# Function that takes User Inputs and displays it on ChatUI
def query_message(history,txt,img):
    if not img:
        history += [(txt,None)]
        return history
    base64 = image_to_base64(img)
    data_url = f"data:image/jpeg;base64,{base64}"
    history += [(f"{txt} ![]({data_url})", None)]
    return history

# Function that takes User Inputs, generates Response and displays on Chat UI
def llm_response(history,text,img):
    if not img:
        response = txt_model.generate_content(text)
        history += [(None,response.text)]
        return history

    else:
        img = PIL.Image.open(img)
        response = vis_model.generate_content([text,img])
        history += [(None,response.text)]
        return history

# Interface Code- Selector method

def sentence_builder(animal, place):
    return f"""how many {animal}s from the {place} are shown in the picture?"""

# gradio block
    
with gr.Blocks(theme='snehilsanyal/scikit-learn') as app:
    gr.Markdown("## MM COT ##")
    with gr.Row():
        image_box = gr.Image(type="filepath")
    
        chatbot = gr.Chatbot(
            scale = 2,
            height=750
        )
    text_box = gr.Dropdown(
                ["what is in the image", "provide alternative title for the image", "how many birds can be seen in the picture?"], label="Animal", info="Will add more animals later!"
            )

    btn = gr.Button("Submit")
    clicked = btn.click(query_message,
                        [chatbot,text_box,image_box],
                        chatbot
                        ).then(llm_response,
                                [chatbot,text_box,image_box],
                                chatbot
                                )
    gr.Markdown("""
    # Multimodal Chain-of-Thought Reasoning in Language Models
    
    <h5 align="center"><i>"Imagine learning a textbook without figures or tables."</i></h5>
    
    Multimodal-CoT incorporates vision features in a decoupled training framework. The framework consists of two training stages: (i) rationale generation and (ii) answer inference. Both stages share the same model architecture but differ in the input and output.
    """)

    
app.queue()
app.launch()