main_st.py

"""
Data Scientist.: Dr.Eddy Giusepe Chirinos Isidro

Run script: 
           $ streamlit run main_st.py
"""
import sys
import tiktoken
import streamlit as st

import openai
import os
from dotenv import load_dotenv, find_dotenv
_ = load_dotenv(find_dotenv()) # read local .env file
openai.api_key  = os.environ['OPENAI_API_KEY']

# Definimos a função para contar tokens:
def count_tokens(text):
    encoding = tiktoken.encoding_for_model("gpt-3.5-turbo")
    return len(encoding.encode(text))

# Defina o limite máximo de tokens:
MAX_MEMORY_TOKENS = 100

# Configurar o estado da sessão (session state) para o histórico de conversas  ---> AQUI MUDA! porque estamos usando STREAMLIT 🤗
if "conversation_history" not in st.session_state:
    st.session_state.conversation_history = [
        {"role": "system", "content": "Você é um assistente prestativo."}
    ]

def chatbot_response(chat_input, placeholder_response):
    # Anexe a entrada do usuário ao histórico de conversas:
    st.session_state.conversation_history.append({"role": "user", "content": chat_input})
    
    # Calculamos o total de TOKENS no histórico de conversas:
    total_tokens = sum(count_tokens(message["content"]) for message in st.session_state.conversation_history)

    # Remove the oldest message from conversation history if total tokens exceed the maximum limit
    while total_tokens > MAX_MEMORY_TOKENS:
        if len(st.session_state.conversation_history) > 2:
            removed_message = st.session_state.conversation_history.pop(1)
            total_tokens -= count_tokens(removed_message["content"])
        else:
            break

    # Faça chamadas de API para OpenAI com o histórico de conversas e use respostas de streaming:
    response = openai.ChatCompletion.create(
        model="gpt-3.5-turbo", # or "gpt-3.5-turbo" or "gpt-4"
        messages=st.session_state.conversation_history,
        stream=True,
    )

    # Processe a resposta da API:
    assistant_response = ""
    for chunk in response:
        if "role" in chunk["choices"][0]["delta"]:
            continue
        elif "content" in chunk["choices"][0]["delta"]:
            r_text = chunk["choices"][0]["delta"]["content"]
            assistant_response += r_text
            placeholder_response.markdown(assistant_response, unsafe_allow_html=True)

    # Anexe a resposta do assistente ao histórico da conversa:
    st.session_state.conversation_history.append({"role": "assistant", "content": assistant_response})

# App Streamlit:
st.title("🤗 Chatbot 🤗")
st.sidebar.header("Data Scientist.: Dr.Eddy Giusepe Chirinos Isidro")
st.sidebar.write("[GitHub: EddyGiusepe](https://github.com/EddyGiusepe)")
st.sidebar.write("[e-mail: eddychirinos.unac@gmail.com]()")
user_input = st.text_input("Você:")
response_button = st.button("Enviar")
placeholder_response = st.empty()

if response_button:
    chatbot_response(user_input, placeholder_response)