doc_mem.py

from colorama import Fore
from langchain.chains import ConversationalRetrievalChain
from langchain.chat_models import ChatOpenAI
from langchain.document_loaders import TextLoader
from langchain.embeddings import HuggingFaceInstructEmbeddings
from langchain.memory import ConversationBufferMemory
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.vectorstores import Chroma


if __name__ == '__main__':
    print("status: loading sales document")
    loader = TextLoader("./docs/sales_doc.txt")
    pages = loader.load_and_split()

    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size=1000,
        chunk_overlap=200,
        length_function=len,
    )

    docs = text_splitter.split_documents(pages)

    # Split documents into chunks
    gds_data_split = docs
    print(len(gds_data_split))

    # Define embedding model
    OPENAI_API_KEY = "sk-"

    embeddings = HuggingFaceInstructEmbeddings(model_name="hkunlp/instructor-xl")
    support_data = gds_data_split
    support_store = Chroma.from_documents(
        support_data, embeddings, collection_name="support"
    )
    print("status: configure llm")
    llm = ChatOpenAI(
        model_name="gpt-3.5-turbo",
        temperature=0,
        openai_api_key=OPENAI_API_KEY,
        max_tokens=1024,
    )

    print("status: confiure chain")
    memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)

    support_qa = ConversationalRetrievalChain.from_llm(
        llm=llm,
        retriever=support_store.as_retriever(),
        verbose=False,
     memory = memory
    )

    query = "Hi, my name is Ragesh?"
    print(Fore.RED + query)
    result = support_qa({"question": query})
    print(Fore.GREEN + result['answer'])

    query = "What is your name?"
    print(Fore.RED + query)
    result = support_qa({"question": query})
    print(Fore.GREEN + result['answer']) # Answering as "Yes, my name is OpenAI language model, "

    query = "Do you have Netflix clone?"
    print(Fore.RED + query)
    result = support_qa({"question": query})
    print(Fore.GREEN + result['answer'])

    query = "What is the price of that?"
    print(Fore.RED + query)
    result = support_qa({"question": query})
    print(Fore.GREEN + result['answer']) # yeah. remembers the last product.