app.py

import openai
import gradio as gr
import os
from gtts import gTTS
from pdfminer.high_level import extract_text


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

from transformers import pipeline
p = pipeline("automatic-speech-recognition",model="openai/whisper-tiny")

def text_to_speech(text):
    myobj = gTTS(text=text, lang='en', slow=False) 
    myobj.save("test.wav") 
    return 'test.wav'



def transcribe(audio):
    text = p(audio)["text"]
    return text
    
def sentiment(text):
    response = openai.Completion.create(
    model="text-davinci-003",
    prompt=f"calssify the text into below sentiment category\n\ntext :  {text}\"\n\n['positive','negative','neutral']\n",
    temperature=1,
    max_tokens=256,
    top_p=1,
    frequency_penalty=0,
    presence_penalty=0)
    return response.choices[0].text.strip()
    
 
messages = [
    {"role": "system", 
     "content": "you name is Rebecca and you are a Pepsico call center assistant and your job is to take the order from the customer and also analysis the sentiment of the customer"}
]

def chatbot(input1,input2):
    if input1:
        input =input1
        input = transcribe(input)
        messages.append({"role": "user", "content": input})
        if sentiment(input) == 'negative':
            reply = "Sorry for any inconvinience. We are tranfering your call."
        else:
            chat = openai.ChatCompletion.create(
            model="gpt-3.5-turbo",
            messages=messages,
            temperature=0.2,
            max_tokens=320,
            top_p=1,
            frequency_penalty=0,
            presence_penalty=0)
            reply = chat.choices[0].message.content        
    
        messages.append({"role": "assistant", "content": reply})
        return reply
    elif input2:
        input =input2
        messages.append({"role": "user", "content": input})
        if sentiment(input) == 'negative':
            reply = "Sorry for any inconvinience. We are tranfering your call."
        else:
            chat = openai.ChatCompletion.create(
            model="gpt-3.5-turbo",
            messages=messages,
            temperature=0.2,
            max_tokens=320,
            top_p=1,
            frequency_penalty=0,
            presence_penalty=0)
            reply = chat.choices[0].message.content        
    
        messages.append({"role": "assistant", "content": reply})
        return text_to_speech(reply)
        

        
##inputs = gr.inputs.Textbox(lines=7, label="Chat with PepsiCo AI assitant")

#inputs= gr.Audio(source="microphone", type="filepath")
        
inputs= [ gr.Audio(source="microphone", type="filepath"),
          gr.inputs.Textbox(lines=7)]


#outputs = gr.outputs.Textbox(label="Reply")

gr.Interface(fn= chatbot, 
             inputs= inputs, 
             outputs= 'audio', 
             title="PepsiCo-chatbot",
             description="Give your order",
             theme="compact").launch()


# gr.Interface(
#     fn=transcribe, 
#     inputs= inputs, 
#     outputs="text"

# ).launch()