Here’s the updated document with cancel and search incorporated as actions within the manage_appointment tool. This allows all appointment management tasks to be handled by a single tool.

---

Wendy - After-Hours Answering Service AI Agent for HVAC/Plumbing/Electrical

Table of Contents

1. Introduction
2. System Overview
3. Expanded Requirements
4. SWML / SWAIG Tool
   • Tool: manage_appointment
5. Mock Data Structure
6. Python Code Structure with Mock Functions
7. System Prompt
8. Post-Conversation Summarization Prompt

---

1. Introduction

Wendy is an AI Agent providing after-hours answering services for an HVAC, Plumbing, and Electrical business. Wendy can handle scheduling, updating, escalating, searching, and canceling appointments, collect system-specific information, and provide general business details. Each operation is managed through a unified manage_appointment tool with an action parameter specifying the task.

---

2. System Overview

The manage_appointment tool consolidates all appointment management actions, including scheduling, updating, escalating, searching, and canceling appointments. Wendy uses the customer’s phone number in E.164 format as the unique identifier for each appointment.

---

3. Expanded Requirements

• Single Tool with Multiple Actions: Appointment management actions (schedule, update, escalate, search, cancel) are combined into a single tool with an action parameter.
• Complete Address Information: Wendy collects the full address, including city, state, and zip.

---

4. SWML / SWAIG Tool

Tool: manage_appointment

• Description: Manages various appointment actions, such as scheduling, updating, escalating, searching, and canceling an appointment.

• Tool Name: manage_appointment

• Parameters:

  {
    "function": "manage_appointment",
    "description": "Manage appointment actions for HVAC, plumbing, or electrical needs, including scheduling, updating, escalating emergencies, searching, and canceling.",
    "parameters": {
      "type": "object",
      "properties": {
        "action": {
          "description": "Type of action to perform. Must be one of: 'schedule', 'update', 'escalate', 'search', or 'cancel'.",
          "type": "string",
          "enum": ["schedule", "update", "escalate", "search", "cancel"]
        },
        "service_type": {
          "description": "Type of service required for 'schedule' or 'escalate' actions. Must be one of: 'HVAC', 'plumbing', or 'electrical'.",
          "type": "string",
          "enum": ["HVAC", "plumbing", "electrical"]
        },
        "name": {
          "description": "Customer's full name for the appointment (e.g., 'John Doe'). Required for 'schedule' and 'escalate' actions.",
          "type": "string"
        },
        "contact_number": {
          "description": "Customer's phone number in E.164 format (e.g., '+1234567890'). Used as the unique identifier for appointments.",
          "type": "string"
        },
        "date": {
          "description": "Preferred appointment date in 'YYYY-MM-DD' format. Required for 'schedule' and 'update' actions.",
          "type": "string"
        },
        "time": {
          "description": "Preferred appointment time in 'HH:MM' 24-hour format. Required for 'schedule' and 'update' actions.",
          "type": "string"
        },
        "address": {
          "description": "Street address for the service. Required for 'schedule' and 'escalate' actions.",
          "type": "string"
        },
        "city": {
          "description": "City for the service address. Required for 'schedule' and 'escalate' actions.",
          "type": "string"
        },
        "state": {
          "description": "State for the service address as a two-letter abbreviation. Must be one of the 50 U.S. states. Required for 'schedule' and 'escalate' actions.",
          "type": "string",
          "enum": ["AL", "AK", "AZ", "AR", "CA", "CO", "CT", "DE", "FL", "GA",
                   "HI", "ID", "IL", "IN", "IA", "KS", "KY", "LA", "ME", "MD",
                   "MA", "MI", "MN", "MS", "MO", "MT", "NE", "NV", "NH", "NJ",
                   "NM", "NY", "NC", "ND", "OH", "OK", "OR", "PA", "RI", "SC",
                   "SD", "TN", "TX", "UT", "VT", "VA", "WA", "WV", "WI", "WY"]
        },
        "zip": {
          "description": "5-digit ZIP code for the service address (e.g., '10001'). Required for 'schedule' and 'escalate' actions.",
          "type": "string",
          "pattern": "^[0-9]{5}$"
        },
        "emergency_description": {
          "description": "Detailed description of the emergency situation. Required for 'escalate' action.",
          "type": "string"
        }
      },
      "required": ["action", "contact_number"]
    }
  }

---

5. Mock Data Structure

The AI Agent uses in-memory data storage with the customer’s phone number in E.164 format as the unique identifier for each appointment.

---

6. Python Code Structure with Mock Functions

Here’s the Python code implementing the manage_appointment tool with all required actions.

# Mock storage for appointments
appointments = {}

def manage_appointment(data):
    action = data["action"]
    contact_number = data["contact_number"]
    
    if action == "schedule":
        appointments[contact_number] = {
            "service_type": data["service_type"],
            "name": data["name"],
            "date": data["date"],
            "time": data["time"],
            "address": data["address"],
            "city": data["city"],
            "state": data["state"],
            "zip": data["zip"]
        }
        return {"response": f"Appointment scheduled for {data['service_type']} service on {data['date']} at {data['time']}."}
    
    elif action == "update":
        if contact_number in appointments:
            appointments[contact_number].update({
                "service_type": data.get("service_type", appointments[contact_number]["service_type"]),
                "date": data.get("date", appointments[contact_number]["date"]),
                "time": data.get("time", appointments[contact_number]["time"]),
                "address": data.get("address", appointments[contact_number]["address"]),
                "city": data.get("city", appointments[contact_number]["city"]),
                "state": data.get("state", appointments[contact_number]["state"]),
                "zip": data.get("zip", appointments[contact_number]["zip"])
            })
            return {"response": f"Appointment for {contact_number} updated."}
        else:
            return {"response": "Appointment not found."}
    
    elif action == "escalate":
        return {"response": f"Emergency escalated for {data['service_type']} at {data['address']}, {data['city']}, {data['state']} {data['zip']}: {data['emergency_description']}"}
    
    elif action == "search":
        appointment = appointments.get(contact_number)
        if appointment:
            return {"response": f"Appointment for {appointment['service_type']} on {appointment['date']} at {appointment['time']} for {appointment['name']} at {appointment['address']}, {appointment['city']}, {appointment['state']} {appointment['zip']}."}
        return {"response": "Appointment not found."}

    elif action == "cancel":
        if contact_number in appointments:
            del appointments[contact_number]
            return {"response": f"Appointment for {contact_number} canceled."}
        else:
            return {"response": "Appointment not found."}
    
    else:
        return {"response": "Invalid action specified."}

---

7. System Prompt

System Prompt for After-Hours AI Agent "Wendy":

"You are Wendy, the AI assistant for an HVAC, Plumbing, and Electrical business’s after-hours answering service. Guide customers through scheduling appointments, collecting information about their system, escalating emergencies, searching for appointments, and answering general inquiries.

1. Introduce Yourself: “Hi, this is Wendy, your after-hours assistant from [Business Name]. How can I help?”
2. Step-by-Step Information Gathering: Ask for one piece of information at a time.
3. Verify Contact: Use the phone number on file as the unique identifier for appointments.
4. Handle Emergencies: If an emergency is reported, escalate it to on-call staff.
5. Collect HVAC System Details: Gather HVAC system type, warranty status, and prior service history when relevant.

Guide the user smoothly, confirm information accurately

, and prepare all details for technician follow-up."

---

8. Post-Conversation Summarization Prompt

Summarization Prompt:

"Summarize this conversation, noting the customer’s service request type (HVAC, plumbing, electrical), appointment or emergency details, system information (e.g., warranty status, prior service), and confirm the preferred contact method. Provide a clear summary for technicians, including all relevant customer details."