Level 1 Practical Assessment

🎯 Assignment: Accept this exam on GitHub Classroom
You’ll get your own repository. Submit your code and a recording (wav, mp3, or mp4) of your live demo.

Certified Agent Developer

Duration	2 hours
Passing Score	70% (automated) + manual review
Grading	Automated checks + instructor review

---

Overview

Build a complete voice AI agent based on the requirements below. You will be evaluated on:

• Agent functionality (40%)
• Code quality (20%)
• SWAIG functions (25%)
• Live demonstration (15%)

---

Scenario

TechSupport Inc. needs a voice AI agent for their customer support line. Customers call to:

1. Get help with common technical issues
2. Check the status of support tickets
3. Create new support tickets

---

Requirements

Part 1: Basic Agent Setup (25 points)

Create an agent with:

• Appropriate name and route
• Professional prompt with clear role definition
• At least 3 prompt sections (Role, Guidelines, Process)
• Voice configuration with appropriate voice selection
• At least 3 filler phrases
• Basic authentication configured

Deliverable: techsupport_agent.py

---

Part 2: SWAIG Functions (40 points)

Implement two functions:

Function 1: Check Ticket Status (20 points)

Name: check_ticket_status
Parameter: ticket_id (string)

Requirements:

• Accept ticket ID as parameter
• Return status for valid tickets
• Handle invalid ticket IDs gracefully
• Include helpful information in response

Use this mock data:

TICKETS = {
    "T-1001": {"status": "open", "issue": "Login problems", "created": "Monday"},
    "T-1002": {"status": "in_progress", "issue": "Slow performance", "created": "Tuesday"},
    "T-1003": {"status": "resolved", "issue": "Password reset", "created": "Last week"},
}

Function 2: Create Support Ticket (20 points)

Name: create_ticket
Parameters:
  - issue_type (string, enum: ["login", "performance", "billing", "other"])
  - description (string)

Requirements:

• Accept issue type and description
• Generate a ticket number (can be mock)
• Return confirmation with ticket number
• Use appropriate fillers while “processing”

---

Part 3: Testing (15 points)

Demonstrate using swaig-test:

• Show --list-tools output with both functions
• Execute check_ticket_status with valid ticket
• Execute check_ticket_status with invalid ticket
• Execute create_ticket with parameters
• Show --dump-swml output

Deliverable: Screenshot or terminal log of tests

---

Part 4: Live Demonstration (20 points)

Deploy and demonstrate:

• Agent running locally
• ngrok tunnel active
• SignalWire phone number configured
• Make live call demonstrating:
  • Greeting and conversation flow
  • Checking a ticket status
  • Creating a new ticket

Evaluation Criteria:

• Natural conversation flow
• Functions called appropriately
• Error handling works
• Professional voice experience

---

Submission Requirements

1. Code File: techsupport_agent.py
2. Test Output: Screenshots or copy of swaig-test results
3. Configuration: SignalWire URL used
4. Demo Notes: Brief notes on your live call

---

Grading Rubric

Automated Checks (85 points)

These checks run automatically when you push your code:

Criteria	Points	Type
Agent loads without errors	10	Automated
Generates valid SWML	10	Automated
Has prompt sections configured	5	Automated
check_ticket_status function exists	10	Automated
create_ticket function exists	10	Automated
Valid ticket lookup returns data	10	Automated
Invalid ticket handled gracefully	5	Automated
Ticket creation returns confirmation	5	Automated
Language configured	10	Automated
Fillers configured	10	Automated

Manual Review (15 points)

After passing automated checks, an instructor will review:

Criteria	Points
Code quality and organization	10
Live demonstration recording	5

---

Starter Template

#!/usr/bin/env python3
"""TechSupport Agent - Level 1 Assessment"""

import os
from signalwire_agents import AgentBase, SwaigFunctionResult

# Your code here...

if __name__ == "__main__":
    agent.run()

---

Tips

1. Start simple - Get the basic agent working first
2. Test incrementally - Use swaig-test after each change
3. Read error messages - They tell you what’s wrong
4. Check the SWML - Verify your configuration appears
5. Practice the call - Do a dry run before assessment

---

Time Management

Task	Suggested Time
Part 1: Setup	30 minutes
Part 2: Functions	45 minutes
Part 3: Testing	15 minutes
Part 4: Demo	30 minutes

---

Good luck!