app.py

from pyChatGPT import ChatGPT
import gradio as gr
import os, json
from loguru import logger
import random
from transformers import pipeline
import torch

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

whisper = gr.Interface.load(name="spaces/sanchit-gandhi/whisper-large-v2")
openai_chatgpt = gr.Interface.load(name="spaces/anzorq/chatgpt-demo")

#input_message.submit([input_message, history], [input_message, chatbot, history])
def translate_or_transcribe(audio, task):
    text_result = whisper(audio, None, task, fn_index=0)
    return text_result
    
def get_response_from_chatbot(text, chat_history):
    r = openai_chatgpt(message, chat_history)
    response = "Sorry, the chatGPT queue is full. Please try again in some time"
    return response

def chat(message, chat_history):      
    out_chat = []
    if chat_history != '':
        out_chat = json.loads(chat_history)
    response = get_response_from_chatbot(message, chat_history)
    out_chat.append((message, response))
    chat_history = json.dumps(out_chat)
    logger.info(f"out_chat_: {len(out_chat)}")
    return out_chat, chat_history



with gr.Blocks(title='Talk to chatGPT') as demo:
    gr.Markdown("## Talk to chatGPT with your voice in your native language ! ##")
    
    with gr.Group(elem_id="page_2") as page_2:        
        with gr.Row(elem_id="chat_row"):
            chatbot = gr.Chatbot(elem_id="chat_bot", visible=False).style(color_map=("green", "blue"))
            chatbot1 = gr.Chatbot(elem_id="chat_bot1").style(color_map=("green", "blue"))    
        with gr.Row():
            prompt_input_audio = gr.Audio(
                                        source="microphone",
                                        type="filepath",
                                        label="Record Audio Input",
                                        
                                    )
            translate_btn = gr.Button("Check Whisper first ? 👍")
            
        whisper_task = gr.Radio(["translate", "transcribe"], value="transcribe", show_label=False)
        with gr.Row(elem_id="prompt_row"):
            prompt_input = gr.Textbox(lines=2, label="Input text",show_label=True)
            chat_history = gr.Textbox(lines=4, label="prompt", visible=False)
            submit_btn = gr.Button(value = "Send to chatGPT",elem_id="submit-btn").style(
                    margin=True,
                    rounded=(True, True, True, True),
                    width=100
                )
            
             

        translate_btn.click(fn=translate_or_transcribe, 
                         inputs=[prompt_input_audio,whisper_task], 
                         outputs=prompt_input
                        )
        
          
        submit_btn.click(fn=chat, 
                             inputs=[prompt_input, chat_history], 
                             outputs=[chatbot, chat_history],
                            )
    gr.HTML('''
         <p>Note: Please be aware that audio records from iOS devices will not be decoded as expected by Gradio. For the best experience, record your voice from a computer instead of your smartphone ;)</p>
        <div class="footer">
                    <p>Whisper Model by <a href="https://github.com/openai/whisper" style="text-decoration: underline;" target="_blank">OpenAI</a> -   
                    <a href="https://chat.openai.com/chat" target="_blank">chatGPT</a> by <a href="https://openai.com/" style="text-decoration: underline;" target="_blank">OpenAI</a> 
                    </p>
        </div>
        ''')
    gr.Markdown("![visitor badge](https://visitor-badge.glitch.me/badge?page_id=RamAnanth1.chatGPT_voice)")

demo.launch(debug = True)