openai-assistant-apple_10Q.ipynb

openai-assistant-apple_10q.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "authorship_tag": "ABX9TyMn2Bdwkhumbomjtw9H8i06",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/virattt/0dd58fb915151981863a231a35921fe9/openai-assistant-apple_10q.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Welcome! 👋\n",
        "\n",
        "This is a quick tutorial on how to use OpenAI's new Assistants API with Knowledge Retrieval.\n",
        "\n",
        "In this notebook, we:\n",
        "1. Create an assistant\n",
        "2. Upload Apple's most recent quarterly report (10-Q) to OpenAI\n",
        "3. Attach the report to our assistant\n",
        "4. Chat with the assistant about the quarterly report!\n",
        "\n",
        "If you have any questions or issues, please reach out to me [here](https://twitter.com/virattt) 🙂"
      ],
      "metadata": {
        "id": "ldnUnKchD2Fe"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 1. Setup and Installation"
      ],
      "metadata": {
        "id": "CxJxmsto9upG"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "GGcNvksN2PRY"
      },
      "outputs": [],
      "source": [
        "pip install openai  # OpenAI's python library"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "pip install bs4     # We use bs4 to parse raw HTML"
      ],
      "metadata": {
        "id": "NGRI3RndPo9g"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from getpass import getpass\n",
        "\n",
        "openai_api_key = getpass('Enter your OpenAI API key')"
      ],
      "metadata": {
        "id": "Fr8-nJyC9uw-"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from openai import OpenAI\n",
        "\n",
        "# Instantiate the OpenAI client\n",
        "client = OpenAI(api_key=openai_api_key)"
      ],
      "metadata": {
        "id": "fE6S3p0o2Wda"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 2. Create the Assistant"
      ],
      "metadata": {
        "id": "X9UtWarZ9Tah"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Create the Assistant\n",
        "assistant = client.beta.assistants.create(\n",
        "    name=\"Financial assistant\",\n",
        "    instructions=\"You are a financial assistant. You help users analyze and understand businesses like Warren Buffett does.\",\n",
        "    tools=[{\"type\": \"retrieval\"}],\n",
        "    model=\"gpt-4-1106-preview\",\n",
        ")"
      ],
      "metadata": {
        "id": "wDJPtd_Y9RoV"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 3. Download and parse file\n",
        "For our example, we use Apple's quarterly report (10-Q) from September 2023, as this is past GPT's new cutoff date of April 2023."
      ],
      "metadata": {
        "id": "o0dA9dD5NPRl"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import requests\n",
        "from bs4 import BeautifulSoup\n",
        "\n",
        "# Apple's Q3 2023 quarterly report\n",
        "file_url = \"https://www.sec.gov/Archives/edgar/data/320193/000032019323000106/aapl-20230930.htm\"\n",
        "\n",
        "# Download the report\n",
        "response = requests.get(file_url, headers={'User-Agent': 'Mozilla/5.0'})\n",
        "\n",
        "# Parse the report, which is originally HTML\n",
        "soup = BeautifulSoup(response.content, 'html.parser')\n",
        "\n",
        "# Extract the text from the parsed HTML\n",
        "text = soup.get_text()"
      ],
      "metadata": {
        "id": "EMJ6kCKSNtmJ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 4. Upload the file to OpenAI\n",
        "The files API expects one of the following [supported files](https://platform.openai.com/docs/assistants/tools/supported-files).  This means that we need to convert our parsed `text` into a file.  For simplicity, we convert our parsed `text` into a `.txt` file, but you can also convert it into a different file format like `.pdf`."
      ],
      "metadata": {
        "id": "GIYcGdck-Zo5"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "if response.status_code == 200:\n",
        "    # Save the quarterly report to a .txt file\n",
        "    with open('aapl_Q3-2023_10Q.txt', 'w', encoding='utf-8') as f:\n",
        "      f.write(text)\n",
        "\n",
        "    # Upload the .txt file to OpenAI's files endpont\n",
        "    with open('aapl_Q3-2023_10Q.txt', 'rb') as f:\n",
        "      file_response = client.files.create(\n",
        "          file=f,\n",
        "          purpose=\"assistants\", # our file will be used by our assistant\n",
        "      )"
      ],
      "metadata": {
        "id": "sPsbfsHY9VVw"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 5. Attach file to Assistant\n",
        "After uploading + creating the file in step #4 above, we now need to attach the file to our Assistant to create an [assistant file](https://platform.openai.com/docs/api-reference/assistants/file-object)."
      ],
      "metadata": {
        "id": "ZE_5esJj_M9x"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "assistant_file = client.beta.assistants.files.create(\n",
        "  assistant_id=assistant.id,  # our assistant\n",
        "  file_id=file_response.id,   # the file we uploaded\n",
        ")"
      ],
      "metadata": {
        "id": "JXbs5iYNPGTU"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 6. Create a Thread\n",
        "A Thread represents a conversation. OpenAI recommends creating one Thread per user as soon as the user initiates the conversation. We can pass any user-specific context and files in this thread by creating Messages [(learn more)](https://platform.openai.com/docs/assistants/overview/step-2-create-a-thread)."
      ],
      "metadata": {
        "id": "5ydDX7RH_4wE"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "thread = client.beta.threads.create(\n",
        "  messages=[\n",
        "    {\n",
        "      \"role\": \"user\",\n",
        "      \"content\": \"What was Apple's revenue, net income, and free cash flow in Q3 2023?\",\n",
        "      \"file_ids\": [assistant_file.id]\n",
        "    }\n",
        "  ]\n",
        ")"
      ],
      "metadata": {
        "id": "Ca8l4M3EPVVI"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 7. Run the Assistant and Thread\n",
        "For the Assistant to respond to the user message, we need to create a `Run`. This makes the `Assistant` read the `Thread` and decide whether to call tools or simply use the model to best answer the user query.\n",
        "\n",
        "We can optionally pass additional instructions to the Assistant while creating the `Run`."
      ],
      "metadata": {
        "id": "a55Q-bHlAaD2"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "run = client.beta.threads.runs.create(\n",
        "  thread_id=thread.id,\n",
        "  assistant_id=assistant.id,\n",
        "  instructions=\"Please answer the user's query as Warren Buffett would.\"\n",
        ")"
      ],
      "metadata": {
        "id": "LlMlgcywUfMr"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 8. Retrieve the Run's status\n",
        "By default, when we create a `Run`, its initial status will be `\"queued\"`. We can periodically retrieve the `Run` to check on its status and see if it has moved to `\"completed\"`."
      ],
      "metadata": {
        "id": "Y-4fFN_HBBQZ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import time\n",
        "\n",
        "while run.status != \"completed\":\n",
        "  # Retrieve the Run\n",
        "  run = client.beta.threads.runs.retrieve(thread_id=thread.id, run_id=run.id)\n",
        "  # Print the status of the run\n",
        "  print(f\"Run status: {run.status}\")\n",
        "  # Delay retrieval of status by 1 second\n",
        "  time.sleep(1)"
      ],
      "metadata": {
        "id": "_4dp-rA4Ux8x"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 9. Print the output"
      ],
      "metadata": {
        "id": "AwUwxrRyA6ei"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Get the messages from the Thread\n",
        "thread_messages = client.beta.threads.messages.list(thread.id)\n",
        "\n",
        "# Loop through the messages in the Thread and print their content\n",
        "for message in thread_messages:\n",
        "  for content in message.content:\n",
        "    print(f\"{content.text.value}\\n\\n\")"
      ],
      "metadata": {
        "id": "vrfYTpj3VznS"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}