{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "'''\n",
    "requirements.txt file contents:\n",
    " \n",
    "langchain==0.0.154\n",
    "PyPDF2==3.0.1\n",
    "python-dotenv==1.0.0\n",
    "streamlit==1.18.1\n",
    "faiss-cpu==1.7.4\n",
    "streamlit-extras\n",
    "'''\n",
    " \n",
    " \n",
    "import streamlit as st\n",
    "from dotenv import load_dotenv\n",
    "import pickle\n",
    "from PyPDF2 import PdfReader\n",
    "from streamlit_extras.add_vertical_space import add_vertical_space\n",
    "from langchain.text_splitter import RecursiveCharacterTextSplitter\n",
    "from langchain.embeddings.openai import OpenAIEmbeddings\n",
    "from langchain.vectorstores import FAISS\n",
    "from langchain.llms import OpenAI\n",
    "from langchain.chains.question_answering import load_qa_chain\n",
    "from langchain.callbacks import get_openai_callback\n",
    "import os\n",
    " \n",
    "# Sidebar contents\n",
    "with st.sidebar:\n",
    "    st.title('🤗💬 LLM Chat App')\n",
    "    st.markdown('''\n",
    "    ## About\n",
    "    This app is an LLM-powered chatbot built using:\n",
    "    - [Streamlit](https://streamlit.io/)\n",
    "    - [LangChain](https://python.langchain.com/)\n",
    "    - [OpenAI](https://platform.openai.com/docs/models) LLM model\n",
    " \n",
    "    ''')\n",
    "    add_vertical_space(5)\n",
    "    st.write('Made with ❤️ by [Prompt Engineer](https://youtube.com/@engineerprompt)')\n",
    " \n",
    "load_dotenv()\n",
    " \n",
    "def main():\n",
    "    st.header(\"Chat with PDF 💬\")\n",
    " \n",
    " \n",
    "    # upload a PDF file\n",
    "    pdf = st.file_uploader(\"Upload your PDF\", type='pdf')\n",
    " \n",
    "    # st.write(pdf)\n",
    "    if pdf is not None:\n",
    "        pdf_reader = PdfReader(pdf)\n",
    "        \n",
    "        text = \"\"\n",
    "        for page in pdf_reader.pages:\n",
    "            text += page.extract_text()\n",
    " \n",
    "        text_splitter = RecursiveCharacterTextSplitter(\n",
    "            chunk_size=1000,\n",
    "            chunk_overlap=200,\n",
    "            length_function=len\n",
    "            )\n",
    "        chunks = text_splitter.split_text(text=text)\n",
    " \n",
    "        # # embeddings\n",
    "        store_name = pdf.name[:-4]\n",
    "        st.write(f'{store_name}')\n",
    "        # st.write(chunks)\n",
    " \n",
    "        if os.path.exists(f\"{store_name}.pkl\"):\n",
    "            with open(f\"{store_name}.pkl\", \"rb\") as f:\n",
    "                VectorStore = pickle.load(f)\n",
    "            # st.write('Embeddings Loaded from the Disk')s\n",
    "        else:\n",
    "            embeddings = OpenAIEmbeddings()\n",
    "            VectorStore = FAISS.from_texts(chunks, embedding=embeddings)\n",
    "            with open(f\"{store_name}.pkl\", \"wb\") as f:\n",
    "                pickle.dump(VectorStore, f)\n",
    " \n",
    "        # embeddings = OpenAIEmbeddings()\n",
    "        # VectorStore = FAISS.from_texts(chunks, embedding=embeddings)\n",
    " \n",
    "        # Accept user questions/query\n",
    "        query = st.text_input(\"Ask questions about your PDF file:\")\n",
    "        # st.write(query)\n",
    " \n",
    "        if query:\n",
    "            docs = VectorStore.similarity_search(query=query, k=3)\n",
    " \n",
    "            llm = OpenAI()\n",
    "            chain = load_qa_chain(llm=llm, chain_type=\"stuff\")\n",
    "            with get_openai_callback() as cb:\n",
    "                response = chain.run(input_documents=docs, question=query)\n",
    "                print(cb)\n",
    "            st.write(response)\n",
    " \n",
    "if __name__ == '__main__':\n",
    "    main()"
   ]
  }
 ],
 "metadata": {
  "language_info": {
   "name": "python"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}