{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Function Calling Example"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 0. Import OpenAI API Key\n",
    "- Firstly, load OpenAI API key from `.env` file\n",
    "- Please create a .env file in the root directory of this project\n",
    "and add `OPENAI_API_KEY=sk-*******`\n",
    "or just replace the value of OPENAI_API_KEY below!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from os import environ\n",
    "from dotenv import load_dotenv\n",
    "\n",
    "\n",
    "load_dotenv()\n",
    "OPENAI_API_KEY = environ[\"OPENAI_API_KEY\"]\n",
    "assert OPENAI_API_KEY is not None, \"Please set OPENAI_API_KEY in .env file\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Load function call from FunctionCalls class\n",
    "\n",
    "- `FunctionCalls` class is a collection of functions.\n",
    "- It contains many pre-defined functions.\n",
    "- The functions will be parsed into `FunctionCall` objects only once when Python starts.\n",
    "- You can get `FunctionCall` object by calling `FunctionCalls.get_function_call` method.\n",
    "- The argument of `FunctionCalls.get_function_call` method is a function to parse."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from typing import Literal\n",
    "from app.models.function_calling.base import FunctionCall\n",
    "from app.models.function_calling.functions import FunctionCalls\n",
    "\n",
    "web_search_function: FunctionCall = FunctionCalls.get_function_call(\n",
    "    FunctionCalls.web_search\n",
    ")\n",
    "\n",
    "# `functions` contains a list of all the functions that can be called\n",
    "functions: list[FunctionCall] = [web_search_function]\n",
    "\n",
    "# \"auto\" will automatically select a function based on the prompt\n",
    "function_call: FunctionCall | Literal[\"auto\", \"none\"] = \"auto\"\n",
    "\n",
    "# \"none\" will not call any function\n",
    "function_call = \"none\"\n",
    "\n",
    "# A specific function is forcibly called by passing the function call object.\n",
    "# NOTE: `function_call` must be in the `functions`,\n",
    "# if you define `function_call` as a `FunctionCall` object.\n",
    "function_call = web_search_function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "- function name: web_search\n",
      "- function arguments: {'query_to_search': 'current price of AAPL stock'}\n"
     ]
    }
   ],
   "source": [
    "from app.models.completion_models import FunctionCallParsed\n",
    "from app.utils.function_calling.request import request_function_call\n",
    "\n",
    "# `messages` contains a list of messages to send\n",
    "messages: list[dict[str, str]] = [\n",
    "    {\"role\": \"user\", \"content\": \"What is price of AAPL now?\"}\n",
    "]\n",
    "\n",
    "# `function_call_parsed` is a dictionary containing `name` and `arguments` of the function call\n",
    "# NOTE:\n",
    "function_call_parsed: FunctionCallParsed = await request_function_call(\n",
    "    messages=messages,\n",
    "    functions=functions,\n",
    "    function_call=function_call,\n",
    "    model=\"gpt-3.5-turbo\",\n",
    "    api_key=OPENAI_API_KEY,\n",
    ")\n",
    "\n",
    "print(f\"- function name: {function_call_parsed['name']}\")\n",
    "print(f\"- function arguments: {function_call_parsed.get('arguments')}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from typing import Optional\n",
    "from app.models.completion_models import FunctionCallParsed\n",
    "from app.utils.function_calling.request import request_function_call\n",
    "\n",
    "# `messages` contains a list of messages to send\n",
    "messages: list[dict[str, str]] = [\n",
    "    {\"role\": \"user\", \"content\": \"What is price of AAPL now?\"}\n",
    "]\n",
    "\n",
    "# `function_call_parsed` is a dictionary containing `name` and `arguments` of the function call\n",
    "# NOTE: if `function_call` is \"none\", an error will be raised when requesting a function call.\n",
    "is_error_raised: bool = False\n",
    "try:\n",
    "    function_call_parsed: Optional[\n",
    "        FunctionCallParsed\n",
    "    ] = await request_function_call(\n",
    "        messages=messages,\n",
    "        functions=functions,\n",
    "        function_call=\"none\",\n",
    "        model=\"gpt-3.5-turbo\",\n",
    "        api_key=OPENAI_API_KEY,\n",
    "    )\n",
    "except ValueError:\n",
    "    is_error_raised = True\n",
    "\n",
    "assert is_error_raised"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Define a custom function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "- function name: weather\n",
      "- function arguments: {'location': 'Boston', 'unit': 'celsius'}\n"
     ]
    }
   ],
   "source": [
    "from typing import Annotated\n",
    "from app.utils.function_calling.parser import parse_function_call_from_function\n",
    "from app.utils.function_calling.request import request_function_call\n",
    "\n",
    "\n",
    "def weather(\n",
    "    location: Annotated[\n",
    "        str,\n",
    "        \"The location to get the weather for.\",\n",
    "    ],\n",
    "    unit: Annotated[\n",
    "        str,\n",
    "        [\"celsius\", \"fahrenheit\"],\n",
    "        \"The unit of temperature.\",\n",
    "    ],\n",
    "):\n",
    "    \"\"\"Get the weather for a location.\"\"\"\n",
    "    pass\n",
    "\n",
    "\n",
    "messages: list[dict[str, str]] = [\n",
    "    {\"role\": \"user\", \"content\": \"What’s the weather like in Boston?\"}\n",
    "]\n",
    "function_call: FunctionCall = parse_function_call_from_function(weather)\n",
    "functions: list[FunctionCall] = [function_call]\n",
    "function_call_parsed: FunctionCallParsed = await request_function_call(\n",
    "    messages=messages,\n",
    "    functions=functions,\n",
    "    function_call=function_call,\n",
    "    model=\"gpt-3.5-turbo\",\n",
    "    api_key=OPENAI_API_KEY,\n",
    ")\n",
    "print(f\"- function name: {function_call_parsed['name']}\")\n",
    "print(f\"- function arguments: {function_call_parsed.get('arguments')}\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.11.4"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}