chat_dov.py

import argparse
from dataclasses import asdict
import json
import os
import streamlit as st
from datasets import load_dataset

from data_driven_characters.character import get_character_definition
from data_driven_characters.corpus import (
    get_corpus_summaries,
    load_docs,
)

from data_driven_characters.chatbots import (
    SummaryChatBot,
    RetrievalChatBot,
    SummaryRetrievalChatBot,
)
from data_driven_characters.interfaces import CommandLine, Streamlit


OUTPUT_ROOT = "output"






def create_chatbot(corpus, character_name, chatbot_type, retrieval_docs, summary_type):
    # logging
    corpus_name = os.path.splitext(os.path.basename(corpus))[0]
    output_dir = f"{OUTPUT_ROOT}/{corpus_name}/summarytype_{summary_type}"
    ####  corpus é fixo do Dov Tzamir, carregado em main()
    ####  
    os.makedirs(output_dir, exist_ok=True)
    summaries_dir = f"{output_dir}/summaries"
    character_definitions_dir = f"{output_dir}/character_definitions"
    os.makedirs(character_definitions_dir, exist_ok=True)

    # load docs
    docs = load_docs(corpus_path=corpus, chunk_size=2048, chunk_overlap=64)

    # generate summaries
    corpus_summaries = get_corpus_summaries(
        docs=docs, summary_type=summary_type, cache_dir=summaries_dir
    )

    # get character definition
    character_definition = get_character_definition(
        name=character_name,
        corpus_summaries=corpus_summaries,
        cache_dir=character_definitions_dir,
    )
    print(json.dumps(asdict(character_definition), indent=4))

    # construct retrieval documents
    if retrieval_docs == "raw":
        documents = [
            doc.page_content
            for doc in load_docs(corpus_path=corpus, chunk_size=256, chunk_overlap=16)
        ]
    elif retrieval_docs == "summarized":
        documents = corpus_summaries
    else:
        raise ValueError(f"Unknown retrieval docs type: {retrieval_docs}")

    # initialize chatbot
    if chatbot_type == "summary":
        chatbot = SummaryChatBot(character_definition=character_definition)
    elif chatbot_type == "retrieval":
        chatbot = RetrievalChatBot(
            character_definition=character_definition,
            documents=documents,
        )
    elif chatbot_type == "summary_retrieval":
        chatbot = SummaryRetrievalChatBot(
            character_definition=character_definition,
            documents=documents,
        )
    else:
        raise ValueError(f"Unknown chatbot type: {chatbot_type}")
    exit
    return chatbot


##  python -m streamlit run chat_dov.py -- --corpus data/tzamir.txt --character_name Dov --chatbot_type retrieval --retrieval_docs raw --interface streamlit

def main():

    # parametros fixos para Dov Tzamir, arquivos ja processados , exceto indice que são em memoria
    st.title("Converse com o avatar do Dov Tzamir")
    st.write("Baseado no texto do livro Fragmentos de Memória do Tito")
    st.write(" ")

    chatbot = st.cache_resource(create_chatbot)(
        "data/tzamir.txt",    #args.corpus,  
        "Dov",               #args.character_name,
        "retrieval",         #args.chatbot_type,
        "raw",               #args.retrieval_docs,
        "map_reduce",        #args.summary_type,
    )

    st.write(" ")
    st.write("Digite o seu diálogo aqui finalizando a linha com ENTER")
    st.write("Voce pode continuar o diálogo, apagando sua perguntanda anterior e  digitando aqui novamente")
    openai_api_key = os.environ["OPENAI_API_KEY"] 


    app = Streamlit(chatbot=chatbot)
    app.run()


if __name__ == "__main__":
    main()