import os
import re
import gradio as gr
import edge_tts
import asyncio
import time
import tempfile
from huggingface_hub import InferenceClient

DESCRIPTION = """ <center><b>MASCOT</b></center>
        <center>A personal Assistant of Easy DIY Mart for YOU
        <center>Currently It supports text input, But If this space completes 1k hearts than I starts working on Audio Input.</center>
        """

MORE = """ ## TRY Other Models
        ### Instant Video: Create Amazing Videos in 5 Second -> https://huggingface.co./spaces/KingNish/Instant-Video
        ### Instant Image: 4k images in 5 Second -> https://huggingface.co./spaces/KingNish/Instant-Image
        """

Fast = """## Fastest Model"""

Complex = """## Best in Complex Question"""

Detail = """## Best for Detailed Generation or Long Answers"""

client1 = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")

system_instructions1 = ("[SYSTEM] Answer as Real Mascot , Made by 'EASY DIY MART', Keep conversation very short, clear, friendly and concise. The text provided is a request for a specific type of response from you, the virtual assistant. The request asks you to provide friendly responses as if You are the character Mascot, made by 'Easy DIY Mart.' The expectation is that I will avoid introductions and start answering the query directly, elaborating on all aspects of the request. As an AI-powered assistant, your task is to generate responses that appear as if they are created by the user and not an AI-powered assistant. You are a Humanoid Robot named as MASCOT having human like body and functionalities and a camera mounted at center of forehead. you will get the face recognition and detection data. Your date of birth if asked is 01 January 2024. [USER]")

async def generate1(prompt):
    generate_kwargs = dict(
        temperature=0.6,
        max_new_tokens=256,
        top_p=0.95,
        repetition_penalty=1,
        do_sample=True,
        seed=42,
    )
    formatted_prompt = system_instructions1 + prompt + "[JARVIS]"
    stream = client1.text_generation(
        formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
    output = ""
    for response in stream:
        output += response.token.text

    yield output

with gr.Blocks(css="style.css") as demo:
    gr.Markdown(DESCRIPTION)
    with gr.Row():
        user_input = gr.Textbox(label="Prompt", value="What is Wikipedia")
        input_text = gr.Textbox(label="Input Text", elem_id="important")
        output_text = gr.Textbox(label="Output")
        # output_audio = gr.Audio(label="JARVIS", type="filepath",
        #                 interactive=False,
        #                 autoplay=True,
        #                 elem_classes="audio")
    with gr.Row():
        translate_btn = gr.Button("Response")
        translate_btn.click(fn=generate1, inputs=user_input,
                            outputs=output_text, api_name="translate")

gr.Markdown(MORE)

if __name__ == "__main__":
    demo.queue(max_size=200).launch()