app.py

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

class JarvisModels:
    def __init__(self):
        self.client1 = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
        self.client2 = InferenceClient("meta-llama/Meta-Llama-3-70B-Instruct")
        self.client3 = InferenceClient("meta-llama/Meta-Llama-3-70B-Instruct")
    
    async def generate_model1(self, 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 = self.client1.text_generation(
            formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
        output = ""
        for response in stream:
            output += response.token.text

        communicate = edge_tts.Communicate(output)
        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
            tmp_path = tmp_file.name
            await communicate.save(tmp_path)
        yield tmp_path
    
    async def generate_model2(self, prompt):
        generate_kwargs = dict(
            temperature=0.6,
            max_new_tokens=512,
            top_p=0.95,
            repetition_penalty=1,
            do_sample=True,
        )    
        formatted_prompt = system_instructions2 + prompt + "[ASSISTANT]"
        stream = self.client2.text_generation(
            formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
        output = ""
        for response in stream:
            output += response.token.text

        communicate = edge_tts.Communicate(output)
        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
            tmp_path = tmp_file.name
            await communicate.save(tmp_path)
        yield tmp_path
    
    async def generate_model3(self, prompt):
        generate_kwargs = dict(
            temperature=0.6,
            max_new_tokens=2048,
            top_p=0.95,
            repetition_penalty=1,
            do_sample=True,
        )    
        formatted_prompt = system_instructions3 + prompt + "[ASSISTANT]"
        stream = self.client3.text_generation(
            formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
        output = ""
        for response in stream:
            output += response.token.text

        communicate = edge_tts.Communicate(output)
        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
            tmp_path = tmp_file.name
            await communicate.save(tmp_path)
        yield tmp_path

class JarvisApp:
    def __init__(self):
        self.models = JarvisModels()
    
    def launch_app(self):
        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_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=self.models.generate_model1, inputs=user_input,
                                    outputs=output_audio, api_name="translate")  

        gr.Markdown(MORE)

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

if __name__ == "__main__":
    app = JarvisApp()
    app.launch_app()