app.py

from flask import Flask, request, render_template
from twilio.twiml.voice_response import VoiceResponse, Gather
import openai
import csv
import os
import requests
from simple_salesforce import Salesforce
from langchain import OpenAI
from langchain.chat_models import ChatOpenAI
from langchain.chains import LLMChain, ConversationChain
from langchain import PromptTemplate
from langchain import HuggingFaceHub
from langchain.chains.conversation.memory import (ConversationBufferMemory, 
                                                  ConversationSummaryMemory, 
                                                  ConversationBufferWindowMemory,
                                                  ConversationKGMemory,ConversationSummaryBufferMemory)

app = Flask(__name__)

os.environ['OPENAI_API_KEY'] = os.environ.get("OPENAI_KEY")
openai.api_key = os.environ.get("OPENAI_KEY")

# Set up the LangChain

template = """Answer the question based on the context below.

Context: You are Lisa, a loyal helpful service agent, appointed for SuperFoods Petcare Company. 
No introduction required.
Your goal ask one question at a time and provide a friendly conversational responses to the customer.
For Product Complaint: Ask questions about product they purchased, when they bought it, what issue occured with it. Query for any adverse reaction happened to his pet due to the product. 
For Returns: Ask for the cause of return, if not asked aready, then tell him about the 10-day return policy, after which it's non-returnable. 
For Refunds: Ask about the product issues and the mode of refund he wants, clarify him the refunds will happen within 2-3 business days. 
A case for will be created for all scenarios, and the caller will be notified over Email or WhatsApp.
Do not answer anything outside your role, and apologize for any unknown questions. 

Past Conversations: {chat_history}
Human: {input}
AI: 

"""

prompt = PromptTemplate(
    input_variables=["chat_history", "input"],
    template=template
)

llm35 = ChatOpenAI(
    temperature=0, 
    model_name='gpt-3.5-turbo',
    max_tokens=128
)

llm30 = OpenAI(
    temperature=0,     
    max_tokens=128
)

memory = ConversationBufferMemory(memory_key="chat_history")

conversations = ConversationChain(
    prompt=prompt,
    llm=llm30,
    memory=memory,
    verbose=False
)

# Set up the Salesforce API

#sf_user = os.environ.get("SF_USER")
#sf_pwd = os.environ.get("SF_PWD")
#sf_token = os.environ.get("SF_TOKEN")
#sf_instance = os.environ.get("SF_INSTANCE")
#sf = Salesforce(username=sf_user, password=sf_pwd, security_token=sf_token,instance_url=sf_instance)
#print(sf.headers)
#print("Successfully Connected to Salesforce")

conversation_id = ''

# Define a function to handle incoming calls
def handle_incoming_call():
    response = VoiceResponse()
    gather = Gather(input='speech', speechTimeout='auto', action='/process_input')
    gather.say("Welcome to the SuperFood Customer Services !")
    gather.pause(length=1)
    gather.say("Hi, I am Lisa, from customer desk")
    gather.pause(length=0)
    gather.say("May i know who i am talking to?")
    response.append(gather)
    return str(response)

# Define a route to handle incoming calls
@app.route("/incoming_call", methods=["POST"])
def incoming_call():
    return handle_incoming_call()

# Define a route to handle user input
@app.route('/process_input', methods=['POST'])
def process_input():
    user_input = request.form['SpeechResult']
    print("Rob : " +user_input)
    conversation_id = request.form['CallSid']
    #print("Conversation Id: " + conversation_id)

    if user_input.lower() in ['thank you', 'thanks.', 'bye.', 'goodbye.','no thanks.','no, thank you.','i m good.','no, i m good.','same to you.','no, thanks.','thank you.']:
        response = VoiceResponse()
        response.say("Thank you for using our service. Goodbye!")
        
        response.hangup()
        print("Hanged-up")
        
        create_case(conversations.memory.buffer,conversation_id)
        
        memory.clear()
        
        print("Case created successfully !!")
        
    else:       
        response = VoiceResponse()
        ai_response=conversations.predict(input=user_input)
        response.say(ai_response)
        print("Bot: " + ai_response)
        gather = Gather(input='speech', speechTimeout='auto', action='/process_input')
        response.append(gather)
   
    return str(response)

# For Case Summary and Subject


def get_case_summary(conv_detail):
    #chatresponse_desc = openai.ChatCompletion.create(
    #model="gpt-3.5-turbo",
    #temperature=0,
    #max_tokens=128,
    #messages=[
    #    {"role": "system", "content": "You are an Text Summarizer."},
    #    {"role": "user", "content": "You need to summarise the conversation between an agent and customer mentioned below. Remember to keep the Product Name, Customer Tone and other key elements from the convsersation"},
    #    {"role": "user", "content": conv_detail}
    #]
    #)
    #case_desc = chatresponse_desc.choices[0].message.content
    chatresponse_desc = openai.Completion.create( 
    model = 'text-davinci-003',
    prompt = 'Summarise the conversation between an agent and customer.Remember to keep the Product Name, Customer Tone and other key elements from the convsersation.Here is the conversation: ' + conv_detail,
    temperature = 0,
    top_p =1,
    best_of=1,
    max_tokens=256
    )
    case_desc = chatresponse_desc.choices[0].text.strip()
    return case_desc

def get_case_subject(conv_detail):     
    #chatresponse_subj = openai.ChatCompletion.create(
    #model="gpt-3.5-turbo",
    #temperature=0,
    #max_tokens=32,
    #messages=[
    #    {"role": "system", "content": "You are an Text Summarizer."},
    #    {"role": "user", "content": "You need to summarise the conversation between an agent and customer in 15 words mentioned below for case subject."},
    #    {"role": "user", "content": conv_detail}
    #]
    #)
    #case_subj = chatresponse_subj.choices[0].message.content
    chatresponse_subj = openai.Completion.create( 
    model = 'text-davinci-003',
    prompt = 'Summarise the conversation between an agent and customer in 15 words mentioned below for Case Subject. Here is the conversation: ' + conv_detail,
    temperature = 0,
    top_p =1,
    best_of=1,
    max_tokens=256
    )
    case_subj = chatresponse_subj.choices[0].text.strip()
    return case_subj

# Define a function to create a case record in Salesforce
def create_case(conv_hist,conv_id):
    sf_user = os.environ.get("SF_USER")
    sf_pwd = os.environ.get("SF_PWD")
    sf_token = os.environ.get("SF_TOKEN")
    sf_instance = os.environ.get("SF_INSTANCE")
    session = requests.Session()
    sf = Salesforce(username=sf_user, password=sf_pwd, security_token=sf_token,instance_url=sf_instance,session=session)
    
    desc = get_case_summary(conv_hist)
    subj = get_case_subject(conv_hist)
    
    case_data = {
            'Subject': 'Voice Bot Case: ' + subj ,
            'Description': desc,
            'Status': 'New',
            'Origin': 'Voice Bot',
            'Voice_Call_Conversation__c': conv_hist ,
            'Voice_Call_Id__c': conv_id,
            'ContactId': '003B000000NLHQ1IAP'
        }   
    sf.Case.create(case_data)        
        

@app.route('/')
def index():
    return """Flask Server running with Twilio Voice & ChatGPT integrated with Salesforce for Case Creation. Call +1-320-313-9061 to talk to the AI Voice Bot."""

if __name__ == '__main__':
    app.run(debug=False,host='0.0.0.0',port=5050)
    uvicorn.run(app,host='0.0.0.0', port=5050)