app.py

import streamlit as st
from urllib.parse import urlparse
from langchain.chat_models import ChatOpenAI
from langchain.embeddings import OpenAIEmbeddings
from langchain.vectorstores import FAISS
from langchain.chains import RetrievalQA
from langchain.chains import ConversationalRetrievalChain
from langchain.prompts import PromptTemplate
from langchain.document_loaders import WebBaseLoader, AsyncHtmlLoader
from langchain.document_transformers import Html2TextTransformer
from langchain.callbacks import get_openai_callback
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.memory import ConversationBufferWindowMemory
from streamlit_chat import message

import asyncio
from langchain.docstore.document import Document

import os
from dotenv import load_dotenv



if 'conversation' not in st.session_state:
    st.session_state['conversation'] = None
if 'messages' not in st.session_state:
    st.session_state['messages'] = []



st.sidebar.title("URL")
input_url = st.sidebar.text_input("Inserisci url:")

#loading openai api keys
load_dotenv()


st.title("Chat with your website 🤖")



question = st.text_area("Ask here:")

#generate the main prompt
prompt_template = """Intruction: You are a website agent that is talking with a human. Use only the chat history and the following information:

{context}

to answer in a helpful manner to the question. If you don't know the answer - say that you don't know. 
Keep your replies short, compassionate and informative. 
{chat_history}

### Input: {question}
### Response:
"""


prompt = PromptTemplate(
    template=prompt_template, input_variables=["context", "question", "chat_history"]
)


memory = ConversationBufferWindowMemory(
    memory_key="chat_history",
    ai_prefix="### Response",
    human_prefix="### Input",
    output_key="answer",
    return_messages=True
)


    



if st.button("Invia", type="primary"):
    
    loader = AsyncHtmlLoader(input_url)
    data = loader.load()
    html2text = Html2TextTransformer()
    docs_transformed = html2text.transform_documents(data)

    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size = 2000,
        chunk_overlap=200,
        separators="\n"
    )

    docs = text_splitter.split_documents(docs_transformed)

    openai_embeddings = OpenAIEmbeddings(model_name="gpt-3.5-turbo")

    vectordb = FAISS.from_documents(
        documents=docs,
        embedding=openai_embeddings)

    retriever = vectordb.as_retriever(search_kwargs={"k": 2})

    relevant_docs = retriever.get_relevant_documents(question)

    if st.session_state['conversation'] is None:
        llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0)

        st.session_state['conversation'] = ConversationalRetrievalChain.from_llm(
            llm,
            chain_type='stuff',
            retriever=retriever,
            memory=memory,
            combine_docs_chain_kwargs={"prompt": prompt},
            verbose=True
        )

    #qa = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=retriever, chain_type_kwargs={"prompt": prompt})

     # Write answer and sources
    answer = st.empty()

    with get_openai_callback() as cb:
        #run the chain and generate response
        response = st.session_state['conversation'](question)
        print(cb)
    
    
    answer.write(response["answer"])
    st.write(relevant_docs)