app.py

import os
import streamlit as st
from langchain.chat_models import ChatOpenAI
from langchain.document_loaders import DirectoryLoader, BSHTMLLoader
from langchain.text_splitter import CharacterTextSplitter
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.vectorstores import Chroma
from langchain.chains import RetrievalQA


st.title('👩🏻‍💻 Query Shakespeare\'s Plays 🎭')

placeholder = st.empty()
openai_api_key = placeholder.text_input(
    'OpenAI API Key',
    type='password',
    help='This app requires a functioning OpenAI API key in order to work.',
    placeholder='Your OpenAI API Key'
)

if openai_api_key == '':
    st.stop()
else:
    placeholder.empty()

# Load model
llm = ChatOpenAI(temperature=0, model_name='gpt-3.5-turbo', openai_api_key=openai_api_key)
embeddings = OpenAIEmbeddings(openai_api_key=openai_api_key)

# Load data
loader = DirectoryLoader('./shakespeare/html/', loader_cls=lambda path: BSHTMLLoader(path, bs_kwargs={'features': 'html.parser'}))
data = loader.load()

# Chunk text
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0, separator='\n')
documents = text_splitter.split_documents(data)

# Vectorize
db = Chroma.from_documents(documents, embeddings)

# QA Chain
shakespeare_qa = RetrievalQA.from_chain_type(llm=llm, chain_type='stuff', retriever=db.as_retriever())

# Query
col1, col2 = st.columns(2)
with col1:
    response = shakespeare_qa.run(st.text_input(
        'Query',
        help='Ask a question pertaining to Shakespeare\'s plays.',
        placeholder='What is the name of Hamlet\'s mother?'
    ))
with col2:
    st.write(f'📖 {response}')