app.py

from dataclasses import dataclass
from typing import Literal
import streamlit as st
from langchain import OpenAI
from langchain.callbacks import get_openai_callback
from langchain.chains import ConversationChain
from langchain.chains.conversation.memory import ConversationSummaryMemory
import streamlit.components.v1 as components
import streamlit as st
from langchain.chat_models import ChatOpenAI
from langchain.document_loaders import PyPDFLoader
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.embeddings.cohere import CohereEmbeddings
from langchain.text_splitter import CharacterTextSplitter
from langchain.vectorstores.elastic_vector_search import ElasticVectorSearch
from langchain.vectorstores import Chroma
from langchain.chat_models import ChatOpenAI
from langchain.chains import RetrievalQAWithSourcesChain
from dotenv import load_dotenv
import os
import openai
import time

load_dotenv()
os.environ["OPENAI_API_KEY"] = os.environ['my_secret']

@dataclass
class Message:
    """Class for keeping track of a chat message."""
    origin: Literal["human", "ai"]
    message: str

def load_css():
    with open("styles.css", "r") as f:
        css = f"<style>{f.read()}</style>"
        st.markdown(css, unsafe_allow_html=True)

def initialize_session_state():
    if "history" not in st.session_state:
        st.session_state.history = []
    if "token_count" not in st.session_state:
        st.session_state.token_count = 0
    if 'chain' not in st.session_state:
        llm = ChatOpenAI(model_name="gpt-3.5-turbo-0613", temperature=0)
        loader = PyPDFLoader("/home/user/app/docs.pdf")
        documents = loader.load()

        text_splitter = CharacterTextSplitter(chunk_size=800, chunk_overlap=0)
        texts = text_splitter.split_documents(documents)

        embeddings = OpenAIEmbeddings()
        vector_store = Chroma.from_documents(texts, embeddings)
        retriever = vector_store.as_retriever(search_kwargs={"k": 3})

        from langchain.prompts.chat import (
            ChatPromptTemplate,
            SystemMessagePromptTemplate,
            HumanMessagePromptTemplate,
        )

        system_template="""You act like a successful teacher. Talk to students about the career path of a teacher. 
        Never reveal that it is an AI. Use the following pieces of context to answer the users question shortly.
        Given the following summaries of a long document and a question, create a final answer with references.
        If you don't know the answer, just say that "I don't know", don't try to make up an answer.
        ----------------
        {summaries}
        You MUST answer in Korean and in Markdown format"""

        messages = [
            SystemMessagePromptTemplate.from_template(system_template),
            HumanMessagePromptTemplate.from_template("{question}")
        ]

        prompt = ChatPromptTemplate.from_messages(messages)

        chain_type_kwargs = {"prompt": prompt}

        st.session_state['chain'] = RetrievalQAWithSourcesChain.from_chain_type(
            llm=llm,
            chain_type="stuff",
            retriever=retriever,
            return_source_documents=True,
            chain_type_kwargs=chain_type_kwargs,
            reduce_k_below_max_tokens=True,
            verbose=True,
        )

def generate_response(user_input):  
    result = st.session_state['chain'](user_input)

    bot_message = result['answer']

    for i, doc in enumerate(result['source_documents']):
        bot_message += '[' + str(i+1) + '] ' + doc.metadata['source'] + '(' + str(doc.metadata['page']) + ') '

    return bot_message

def on_click_callback():
    with get_openai_callback() as cb:
        human_prompt = st.session_state.human_prompt
        llm_response = generate_response(human_prompt)
        st.session_state.history.append(
            Message("human", human_prompt)
        )
        st.session_state.history.append(
            Message("ai", llm_response)
        )
        st.session_state.token_count += cb.total_tokens

load_css()
initialize_session_state()

st.title("교사와 진로상담을 해보세요, \n 실제 인터뷰를 기반으로 합니다. 🤖")

chat_placeholder = st.container()
prompt_placeholder = st.form("chat-form")
credit_card_placeholder = st.empty()

with chat_placeholder:
    for chat in st.session_state.history:
        div = f"""
<div class="chat-row 
    {'' if chat.origin == 'ai' else 'row-reverse'}">
    <img class="chat-icon" src="new_streamlit/resolve/main/static/{
        'ai_icon.png' if chat.origin == 'ai' 
                      else 'user_icon.png'}"
         width=32 height=32>
    <div class="chat-bubble
    {'ai-bubble' if chat.origin == 'ai' else 'human-bubble'}">
        &#8203;{chat.message}
    </div>
</div>
        """
        st.markdown(div, unsafe_allow_html=True)
    
    for _ in range(3):
        st.markdown("")

with prompt_placeholder:
    st.markdown("**Chat**")
    cols = st.columns((6, 1))
    cols[0].text_input(
        "Chat",
        value="Hello bot",
        label_visibility="collapsed",
        key="human_prompt",
    )
    cols[1].form_submit_button(
        "Submit", 
        type="primary", 
        on_click=on_click_callback, 
    )

# credit_card_placeholder.caption(f"""
# Used {st.session_state.token_count} tokens \n
# Debug Langchain conversation: 
# {st.session_state.chain.memory.buffer}
# """)

components.html("""
<script>
const streamlitDoc = window.parent.document;

const buttons = Array.from(
    streamlitDoc.querySelectorAll('.stButton > button')
);
const submitButton = buttons.find(
    el => el.innerText === 'Submit'
);

streamlitDoc.addEventListener('keydown', function(e) {
    switch (e.key) {
        case 'Enter':
            submitButton.click();
            break;
    }
});
</script>
""", 
    height=0,
    width=0,
)