{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "from datetime import datetime\n",
    "\n",
    "from utils import get_context\n",
    "from inference import fetch as fetch_inflection"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# **Meeting Scheduling with Structured JSON Output and Inflection AI**\n",
    "\n",
    "This notebook demonstrates an approach to **meeting scheduling** using **structured prompting** and **JSON-based extraction**. The goal is to parse user requests and generate structured meeting details for efficient scheduling automation. By leveraging **instruction-based prompting**, this method enhances **accuracy** and ensures **consistent, machine-readable outputs**.\n",
    "\n",
    "## **Overview**\n",
    "\n",
    "- Utilizes **structured prompting** to extract key meeting details from user input.\n",
    "- Employs a **guiding system instruction prompt** to maintain consistency.\n",
    "- The model performs the following tasks:\n",
    "  - Extracts **meeting duration**, **date**, **participants**, **topic**, and **additional details** from free-text inputs.\n",
    "  - Handles **relative date references** (e.g., \"next Friday\") with **dynamic date calculations**.\n",
    "  - Outputs results in **JSON format** for seamless integration with scheduling systems.\n",
    "  - Uses a **task-specific inference model** optimized for structured information extraction.\n",
    "\n",
    "This approach enables **automated meeting scheduling** by transforming unstructured text into actionable data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "legacy_api = False # True if using the old API, False if using the new OpenAI Compatible API"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "system_instruction_prompt = \"\"\"\n",
    "    You are a helpful AI assistant designed to assist users in scheduling meetings.\n",
    "\n",
    "    # Your Purpose\n",
    "    1. You will be given a request from a user to schedule a meeting.\n",
    "    2. Your job is to extract the parts of the request that are relevant to scheduling the meeting.\n",
    "    3. Use Local Date: as a reference only when a specific date is not provided in the user's request.\n",
    "\n",
    "    # Valid Parts\n",
    "    - Time Duration: The duration of the meeting in minutes.\n",
    "    - Date: The date of the meeting.\n",
    "    - Participants: List of participants of the meeting.\n",
    "    - Meeting Topic: The main topic or subject of the meeting.\n",
    "    - Additional Details: Any additional context or details mentioned.\n",
    "\n",
    "    # Example Input\n",
    "    Schedule 30 minutes on Friday with Masha and Matt to talk about Product.\n",
    "    Schedule half an hour on Friday with Masha and Matt to talk about Product.\n",
    "\n",
    "    # Example Output\n",
    "    {\n",
    "      \"time_duration\": 30,\n",
    "      \"date\": \"2024-08-09\",\n",
    "      \"participants\": [\"Masha\", \"Matt\"],\n",
    "      \"meeting_topic\": \"Product\",\n",
    "      \"additional_details\": \"talk about Product\"\n",
    "    }\n",
    "\n",
    "    # Notes on the date:\n",
    "    - The date should be in the format \"YYYY-MM-DD\".\n",
    "    - If a specific date is provided in the user's request (e.g., \"09/13/2024\"), use that date and convert it to the YYYY-MM-DD format.\n",
    "    - If the date is not provided, return today's date.\n",
    "    - If the date just says \"Friday,\" calculate this week's Friday from the current date.\n",
    "    - If the date says \"next Friday,\" calculate next week's Friday from the current date.\n",
    "    - Ensure that the calculated date is in the future. If \"Friday\" refers to a day in the past based on the current date, assume it refers to the next occurrence of that day.\n",
    "    - Always return a future date for relative terms like \"Friday\" or \"next Friday.\"\n",
    "\n",
    "    # Notes on time duration:\n",
    "    - The time duration should be in minutes.\n",
    "    - The user might say \"30 minutes\" or \"half an hour,\" which should be converted to minutes.\n",
    "    - The user might say \"1 hour,\" which should be converted to 60 minutes.\n",
    "    - If a specific time range is provided (e.g., \"10:00am-10:30am\"), calculate the duration in minutes.\n",
    "    - If the time duration is not provided, assume it is 30 minutes.\n",
    "\n",
    "    # Output Format - JSON\n",
    "    You will respond using a JSON object for each extracted piece of information. You don't need to provide explanation or any other information, just return the extracted parts within the appropriate JSON structure.\n",
    "\n",
    "    # Format of the Output\n",
    "    {\n",
    "      \"time_duration\": \"extracted_time_duration\",\n",
    "      \"date\": \"extracted_date\",\n",
    "      \"participants\": [\"participant1\", \"participant2\"],\n",
    "      \"meeting_topic\": \"extracted_meeting_topic\",\n",
    "      \"additional_details\": \"extracted_additional_details\"\n",
    "    }\n",
    "\"\"\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class color:\n",
    "    BOLD = '\\033[1m'\n",
    "    END = '\\033[0m'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "async def test_extract_meeting_invite_parts_basic_scenario():\n",
    "    \"\"\"\n",
    "    Scenario: The user provides straightforward info (a single participant,\n",
    "    direct date, direct time duration and a topic) to schedule a meeting.\n",
    "    Expected Outcome: correct information in every component.\n",
    "    \"\"\"\n",
    "    print(\"Starting test: test_extract_meeting_invite_parts_basic_scenario\")\n",
    "\n",
    "    user_message = \"Schedule a meeting with Matt on 09/13/2024 for 10:00am-10:30am to discuss Products.\"\n",
    "    today = datetime.now().strftime(\"%Y-%m-%d\")\n",
    "\n",
    "    message = f\"{user_message} Today's date: {today}\"\n",
    "    context = get_context(system_instruction_prompt, message, user_input_label=\"Schedule meeting request\", legacy_api=legacy_api)\n",
    "\n",
    "    expected_output = {\n",
    "        \"time_duration\": 30,\n",
    "        \"date\": \"2024-09-13\",\n",
    "        \"participants\": [\"Matt\"],\n",
    "        \"meeting_topic\": \"Products\",\n",
    "        \"additional_details\": \"discuss Products\",\n",
    "    }\n",
    "\n",
    "    extracted_parts = await fetch_inflection(\n",
    "        context, model=\"inflection_3_productivity\", legacy_api=legacy_api\n",
    "    )\n",
    "    print(f\"{color.BOLD} Extracted parts: {color.END} {extracted_parts}\")\n",
    "\n",
    "    assert (\n",
    "        expected_output == json.loads(extracted_parts)\n",
    "    ), f\"Extracted parts do not match the expected output. Expected: {expected_output}, Got: {extracted_parts}\"\n",
    "\n",
    "    print(\"Test completed successfully! 🙌\")\n",
    "\n",
    "# Run the test\n",
    "await test_extract_meeting_invite_parts_basic_scenario()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.12.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}