app.py

import os
import time
import openai
import gradio as gr
from gtts import gTTS
from io import BytesIO
from IPython.display import Audio, display

# Set OpenAI API key
openai.api_key = os.environ.get("OPENAI_API_KEY")

# Set OpenAI GPT-3 model
MODEL = "text-davinci-002"


# Initialize chat history as an empty list
chat_history = []

# Define function to generate speech from text using Google Text-to-Speech (gTTS)
def text_to_speech(text):
    tts = gTTS(text=text)
    mp3 = BytesIO()
    tts.write_to_fp(mp3)
    mp3.seek(0)
    display(Audio(mp3, autoplay=True))

# Define function to get chatbot response
def chat(text):
    # Append user input to chat history
    chat_history.append(f"User: {text}")
    
    # Use OpenAI's GPT-3.5 model to generate chatbot response
    response = openai.Completion.create(
        model=MODEL,
        prompt = r"Conversation with user:\n" + "\n".join(chat_history) + r"\nChatbot:",
        temperature=0.5,
        max_tokens=1024,
        n=1,
        stop=None,
        frequency_penalty=0,
        presence_penalty=0
    ).choices[0].text.strip()
    
    # Append chatbot response to chat history
    chat_history.append(f"Chatbot: {response}")
    
    # Generate speech from chatbot response
    text_to_speech(response)
    
    return response

# Define function to clear chat history
def clear_chat():
    global chat_history
    chat_history = []

# Define interface
interface = gr.Interface(
    chat,
    inputs=["text",
             gr.Audio(source="microphone", type="numpy"), 
           ],
    outputs=["text", gr.outputs.Voice()],
    title="Chatbot with OpenAI's GPT-3.5 Model",
    description="An interactive chatbot using OpenAI's GPT-3.5 model with chat persistence and voice inputs/outputs.",
    theme="default",
    layout="vertical",
    allow_flagging=False,
    allow_screenshot=False,
    allow_download=False,
    show_input=True,
    show_output=True
)

# Run interface
interface.launch()