Ask A Book Questions.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "2d3e92ed",
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain.document_loaders import UnstructuredPDFLoader, OnlinePDFLoader\n",
    "from langchain.text_splitter import RecursiveCharacterTextSplitter\n",
    "import yaml\n",
    "with open('config.yaml', 'r') as f:\n",
    "    config = yaml.safe_load(f)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5166d759",
   "metadata": {},
   "source": [
    "### Load your data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "b4a2d6bf",
   "metadata": {},
   "outputs": [],
   "source": [
    "loader = UnstructuredPDFLoader(\"data/GENERAL NOTIONS OF INDEXABILITY FOR QUEUEING CONTROL AND ASSET MANAGEMENT.pdf\")\n",
    "#loader = OnlinePDFLoader(\"https://projecteuclid.org/journals/annals-of-applied-probability/volume-21/issue-3/General-notions-of-indexability-for-queueing-control-and-asset-management/10.1214/10-AAP705.pdf\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "bcdac23c",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "detectron2 is not installed. Cannot use the hi_res partitioning strategy. Falling back to partitioning with the fast strategy.\n"
     ]
    }
   ],
   "source": [
    "data = loader.load()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "b4fd7c9e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "You have 1 document(s) in your data\n",
      "There are 78242 characters in your document\n"
     ]
    }
   ],
   "source": [
    "print (f'You have {len(data)} document(s) in your data')\n",
    "print (f'There are {len(data[0].page_content)} characters in your document')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8af9b604",
   "metadata": {},
   "source": [
    "### Chunk your data up into smaller documents"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "fb3c6f02",
   "metadata": {},
   "outputs": [],
   "source": [
    "text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=0)\n",
    "texts = text_splitter.split_documents(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "879873a4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Now you have 81 documents\n"
     ]
    }
   ],
   "source": [
    "print (f'Now you have {len(texts)} documents')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "838b2843",
   "metadata": {},
   "source": [
    "### Create embeddings of your documents to get ready for semantic search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "373e695a",
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain.vectorstores import Chroma, Pinecone\n",
    "from langchain.embeddings.openai import OpenAIEmbeddings\n",
    "import pinecone"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "0e093ef3",
   "metadata": {
    "hide_input": false
   },
   "outputs": [],
   "source": [
    "OPENAI_API_KEY = config['openai_key']\n",
    "PINECONE_API_KEY = config['pinecone_key']\n",
    "PINECONE_API_ENV = config['pinecone_env']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "4e0d1c6a",
   "metadata": {},
   "outputs": [],
   "source": [
    "embeddings = OpenAIEmbeddings(openai_api_key=OPENAI_API_KEY)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "0deb2f6a",
   "metadata": {},
   "outputs": [],
   "source": [
    "# initialize pinecone\n",
    "pinecone.init(\n",
    "    api_key=PINECONE_API_KEY,  # find at app.pinecone.io\n",
    "    environment=PINECONE_API_ENV  # next to api key in console\n",
    ")\n",
    "index_name = config['index_name']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "388988ce",
   "metadata": {},
   "outputs": [],
   "source": [
    "docsearch = Pinecone.from_texts([t.page_content for t in texts], embeddings, index_name=index_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3c35dcd9",
   "metadata": {},
   "source": [
    "### Query those docs to get your answer back"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "f051337b",
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain.llms import OpenAI\n",
    "from langchain.chains.question_answering import load_qa_chain"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "6b9b1c03",
   "metadata": {},
   "outputs": [],
   "source": [
    "#llm = OpenAI(model_name='gpt-3.5-turbo',temperature=0, openai_api_key=OPENAI_API_KEY, max_tokens=526)\n",
    "llm = OpenAI(model_name='gpt-4o',temperature=0, openai_api_key=OPENAI_API_KEY)\n",
    "chain = load_qa_chain(llm, chain_type=\"stuff\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "f67ea7c2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'The paper discusses monotonicity in various contexts, including the monotonicity of policies for single-project subproblems in restless bandit problems, the monotonicity of optimal policies for P(W) in service control problems for M/M/1 queues, and the construction of a descending set of h-values with corresponding monotone policies in the computation of station indices.'"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "query = \"Enumerate all that is said about monotonicity in this paper\"\n",
    "docs = docsearch.similarity_search(query, include_metadata=True)\n",
    "chain.run(input_documents=docs, question=query)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}