47_LangChainPDF.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "6e9a4f41",
   "metadata": {},
   "source": [
    "<div style=\"background-color:#000;\"><img src=\"pqn.png\"></img></div>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fd709f5c",
   "metadata": {},
   "source": [
    "This code integrates LangChain library functionalities to process and query PDF documents using OpenAI's language model. It loads a PDF file, splits it into pages, and stores these pages in a vector database (ChromaDB). It then creates a toolkit to interact with the vector store and uses an agent executor to query the database based on user input. This is useful for extracting and querying information from structured documents like financial reports."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "58bade97",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ce47c00a",
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain.llms import OpenAI\n",
    "from langchain.document_loaders import PyPDFLoader\n",
    "from langchain.vectorstores import Chroma\n",
    "from langchain.agents.agent_toolkits import (\n",
    "    create_vectorstore_agent,\n",
    "    VectorStoreToolkit,\n",
    "    VectorStoreInfo,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "18b8f141",
   "metadata": {},
   "source": [
    "Set the OpenAI API key for authentication"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7a0ddde1",
   "metadata": {},
   "outputs": [],
   "source": [
    "os.environ[\"OPENAI_API_KEY\"] = \"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\""
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9cb3ca7c",
   "metadata": {},
   "source": [
    "Create an instance of the OpenAI language model with specified parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6ebf63b0",
   "metadata": {},
   "outputs": [],
   "source": [
    "llm = OpenAI(temperature=0.1, verbose=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e5e50726",
   "metadata": {},
   "source": [
    "Initialize a PDF loader for the specified file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "681ccbf2",
   "metadata": {},
   "outputs": [],
   "source": [
    "loader = PyPDFLoader(\"apple.pdf\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "41e538eb",
   "metadata": {},
   "source": [
    "Split the PDF into individual pages for processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "adf3f45c",
   "metadata": {},
   "outputs": [],
   "source": [
    "pages = loader.load_and_split()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fbbc170b",
   "metadata": {},
   "source": [
    "Load the split pages into a Chroma vector database for efficient querying"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "964eaa4a",
   "metadata": {},
   "outputs": [],
   "source": [
    "store = Chroma.from_documents(pages, collection_name=\"annualreport\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dbd9b7f5",
   "metadata": {},
   "source": [
    "Create a VectorStoreInfo object to hold metadata about the vector store"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4fd45821",
   "metadata": {},
   "outputs": [],
   "source": [
    "vectorstore_info = VectorStoreInfo(\n",
    "    name=\"apple\",\n",
    "    description=\"Apple quarterly consolidated financials\",\n",
    "    vectorstore=store,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9db55c3d",
   "metadata": {},
   "source": [
    "Convert the vector store information into a toolkit for LangChain"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "40f81c28",
   "metadata": {},
   "outputs": [],
   "source": [
    "toolkit = VectorStoreToolkit(vectorstore_info=vectorstore_info)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "288a7fd0",
   "metadata": {},
   "source": [
    "Create an agent executor that uses the language model and toolkit for querying"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2bee8094",
   "metadata": {},
   "outputs": [],
   "source": [
    "agent_executor = create_vectorstore_agent(llm=llm, toolkit=toolkit, verbose=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "961d1da5",
   "metadata": {},
   "source": [
    "Prompt the user to enter a search term for querying the document"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "445ea906",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = input(\"Enter your search term: \")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c6b36ac3",
   "metadata": {},
   "outputs": [],
   "source": [
    "if prompt:\n",
    "\n",
    "    # Pass the user input to the agent executor for processing\n",
    "\n",
    "    response = agent_executor.run(prompt)\n",
    "\n",
    "    # Print the response from the language model to the screen\n",
    "\n",
    "    print(response)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8a4cac29",
   "metadata": {},
   "source": [
    "<a href=\"https://pyquantnews.com/\">PyQuant News</a> is where finance practitioners level up with Python for quant finance, algorithmic trading, and market data analysis. Looking to get started? Check out the fastest growing, top-selling course to <a href=\"https://gettingstartedwithpythonforquantfinance.com/\">get started with Python for quant finance</a>. For educational purposes. Not investment advise. Use at your own risk."
   ]
  }
 ],
 "metadata": {
  "jupytext": {
   "cell_metadata_filter": "-all",
   "main_language": "python",
   "notebook_metadata_filter": "-all"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}