app.py

import os
import openai
import gradio as gr
import time

openai.api_key = os.getenv("OPENAI_API_KEY")

messages = [{"role": "system", "content": 'You are my assistant, keep your answers brief. You may answer either in turkish or english depending on the previous input.'}]

def speak(audio):
    global messages

    os.rename(audio, audio + ".mp3")
    audiofile = open(audio + ".mp3", "rb")
    transcript = openai.Audio.transcribe("whisper-1", audiofile)             #Ses çevirisi

    messages.append({"role": "user", "content": transcript["text"]})          #Çevirilen sesin mesaja dönüşmesi

    response = openai.ChatCompletion.create(                                  #Chatgpt
    model="gpt-3.5-turbo",
    messages=messages,
    temperature=0.7,
    max_tokens=128,
    top_p=1,
    frequency_penalty=0,
    presence_penalty=0
    )
    
    system_message = response["choices"][0]["message"]
    messages.append(system_message)

    print(system_message['content'])

    chattranscript = ""
    for message in messages:
        if message['role'] != 'system':
            chattranscript += message['role'] + ": " + message['content'] + "\n\n"

    return chattranscript

output = gr.Interface(
    fn=speak, 
    inputs=gr.Audio(source="microphone", type="filepath")
    ,outputs="text")

output.launch()