Skip to content

TESTING_GUIDE.md

Latest commit

 

History

History
349 lines (282 loc) · 9.58 KB

TESTING_GUIDE.md

File metadata and controls

349 lines (282 loc) · 9.58 KB

🧪 Contact Center AI - Testing Guide

This guide provides comprehensive instructions for testing the Contact Center Agentic Flow System as a customer interacting with AI support agents.

🚀 Quick Start

Prerequisites

  • Python 3.11+
  • Running Contact Center AI application (on port 8000)
  • Dependencies installed (pip install -r requirements.txt)

1. Start the Application

# Make sure the application is running
python -m src.main

The application should be listening on http://localhost:8000

2. Run Customer Test Interface

# Interactive testing
python test_customer_interaction.py

# Or run automated scenarios
python test_customer_interaction.py
# Select option 2 when prompted

📋 Available Testing Methods

Method 1: Interactive Test Client (Recommended)

The easiest way to test customer interactions:

python test_customer_interaction.py

Features:

  • ✅ Automatic authentication
  • ✅ Pre-built test scenarios
  • ✅ Interactive conversation flow
  • ✅ Real-time agent responses
  • ✅ Follow-up message support

Method 2: Direct API Testing with cURL

Step 1: Get Authentication Token

curl -X POST "http://localhost:8000/token" \
  -H "Content-Type: application/x-www-form-urlencoded" \
  -d "username=johndoe&password=secret"

Example response:

{
  "access_token": "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9...",
  "token_type": "bearer"
}

Step 2: Start a Conversation

curl -X POST "http://localhost:8000/api/v1/conversations" \
  -H "Authorization: Bearer YOUR_TOKEN_HERE" \
  -H "Content-Type: application/json" \
  -d '{
    "customer_id": "CUST_001",
    "channel": "web",
    "initial_message": "I need help with my account",
    "priority": "medium"
  }'

Step 3: Send Follow-up Messages

curl -X POST "http://localhost:8000/api/v1/conversations/CONVERSATION_ID/messages" \
  -H "Authorization: Bearer YOUR_TOKEN_HERE" \
  -H "Content-Type: application/json" \
  -d '{
    "content": "I cant access my billing information"
  }'

Method 3: Python Script Testing

import asyncio
import httpx

async def test_conversation():
    base_url = "http://localhost:8000"
    
    # 1. Authenticate
    async with httpx.AsyncClient() as client:
        auth_response = await client.post(
            f"{base_url}/token",
            data={"username": "johndoe", "password": "secret"}
        )
        token = auth_response.json()["access_token"]
        headers = {"Authorization": f"Bearer {token}"}
        
        # 2. Start conversation
        conversation_data = {
            "customer_id": "TEST_001",
            "channel": "web", 
            "initial_message": "Hello, I need technical support",
            "priority": "high"
        }
        
        conv_response = await client.post(
            f"{base_url}/api/v1/conversations",
            json=conversation_data,
            headers=headers
        )
        
        conversation = conv_response.json()
        print(f"Agent Response: {conversation['response']}")
        
        # 3. Send follow-up
        message_data = {"content": "My internet keeps disconnecting"}
        
        msg_response = await client.post(
            f"{base_url}/api/v1/conversations/{conversation['conversation_id']}/messages",
            json=message_data,
            headers=headers
        )
        
        message = msg_response.json()
        print(f"Agent Response: {message['content']}")

# Run the test
asyncio.run(test_conversation())

🎯 Test Scenarios

Scenario 1: General Support (Tier 1)

Initial Message: "What are your business hours?" Expected:

  • Routed to Tier 1 Support Agent
  • Quick FAQ-style response
  • High confidence score (>0.8)

Scenario 2: Technical Issue (Tier 2)

Initial Message: "My internet connection keeps dropping every few minutes" Expected:

  • Routed to Technical Support Agent
  • Request for diagnostic information
  • Offer troubleshooting steps

Follow-up: "I've already restarted my router twice" Expected:

  • Advanced diagnostic suggestions
  • Possible escalation triggers

Scenario 3: Billing Question (Billing Agent)

Initial Message: "I was charged twice for the same service this month" Expected:

  • Routed to Billing Agent
  • Request for account verification
  • Investigation of duplicate charges

Scenario 4: Sales Inquiry (Sales Agent)

Initial Message: "I'm interested in upgrading to your premium plan" Expected:

  • Routed to Sales Agent
  • Product information and pricing
  • Upselling opportunities

Scenario 5: Complex Issue (Escalation)

Initial Message: "I need to cancel my service immediately due to poor performance" Follow-up: "This is unacceptable, I want to speak to a manager" Expected:

  • Escalation to Supervisor Agent
  • Higher priority handling
  • Retention-focused response

🔍 Testing Checklist

Basic Functionality

  • ✅ Application starts successfully
  • ✅ Health check endpoint responds
  • ✅ Authentication works with test credentials
  • ✅ Can start new conversations
  • ✅ Can send follow-up messages
  • ✅ Receives agent responses

Agent Routing

  • ✅ Intent classification accuracy
  • ✅ Appropriate agent assignment
  • ✅ Escalation triggers work
  • ✅ Agent handoffs function properly

Response Quality

  • ✅ Responses are relevant to queries
  • ✅ Response time under 5 seconds
  • ✅ Confidence scores are reasonable
  • ✅ Follow-up suggestions provided

Error Handling

  • ✅ Invalid authentication handled gracefully
  • ✅ Malformed requests return proper errors
  • ✅ System recovers from agent failures
  • ✅ Timeout scenarios handled

🐛 Troubleshooting

Common Issues

1. "Service is not available"

Problem: Cannot connect to the application Solution:

# Check if application is running
curl http://localhost:8000/health

# Start the application if not running
python -m src.main

2. "Authentication failed"

Problem: Cannot get access token Solution:

  • Verify username/password: johndoe / secret
  • Check token endpoint: POST /token
  • Ensure Content-Type header is correct

3. "Conversation not found"

Problem: Cannot send messages to conversation Solution:

  • Verify conversation_id from start_conversation response
  • Check if conversation expired
  • Ensure proper Authorization header

4. "Agent timeout" or slow responses

Problem: Agent takes too long to respond Solution:

  • Check application logs for errors
  • Verify LangGraph integration status
  • Check database connectivity

Debug Mode

Enable debug logging:

export LOG_LEVEL=DEBUG
python -m src.main

📊 Performance Testing

Load Testing with Multiple Conversations

import asyncio
import aiohttp

async def load_test():
    """Test with 10 concurrent conversations"""
    
    async def single_conversation(session, i):
        # Authentication
        auth_data = {"username": "johndoe", "password": "secret"}
        async with session.post("/token", data=auth_data) as resp:
            token = (await resp.json())["access_token"]
            headers = {"Authorization": f"Bearer {token}"}
        
        # Start conversation
        conv_data = {
            "customer_id": f"LOAD_TEST_{i}",
            "channel": "web",
            "initial_message": f"Test message {i}",
            "priority": "medium"
        }
        
        async with session.post("/api/v1/conversations", 
                               json=conv_data, 
                               headers=headers) as resp:
            result = await resp.json()
            print(f"Conversation {i}: {result.get('response', 'No response')}")
    
    async with aiohttp.ClientSession() as session:
        tasks = [single_conversation(session, i) for i in range(10)]
        await asyncio.gather(*tasks)

# Run load test
asyncio.run(load_test())

📈 Monitoring Test Results

Key Metrics to Monitor

  • Response Time: Should be < 5 seconds
  • Accuracy: Intent classification > 85%
  • Escalation Rate: < 20% for simple queries
  • Error Rate: < 1% of requests

View Application Metrics

# Check metrics endpoint
curl -H "Authorization: Bearer YOUR_TOKEN" \
  http://localhost:8000/api/v1/metrics

🎯 Advanced Testing

Custom Test Scenarios

Create your own test scenarios by modifying test_customer_interaction.py:

custom_scenarios = [
    {
        "name": "Complex Technical Issue",
        "initial_message": "My VPN connection fails when connecting from certain locations",
        "follow_ups": [
            "It works fine from home but not from my office",
            "I'm using Windows 11 with the latest updates",
            "The error code is 800-something"
        ]
    }
]

A/B Testing Different Prompts

Test different ways of asking the same question:

  • "I need help with billing" vs "I have a question about my invoice"
  • "My internet is slow" vs "Network performance issues"

📝 Reporting Issues

When reporting issues, please include:

  1. Test scenario being executed
  2. Expected behavior vs actual behavior
  3. Error messages or logs
  4. Steps to reproduce
  5. Environment details (OS, Python version, etc.)

🚀 Next Steps

After basic testing works:

  1. Test WebSocket connections for real-time updates
  2. Test file attachments in conversations
  3. Test conversation history retrieval
  4. Test multi-channel scenarios
  5. Test production deployment with Docker

Happy Testing! 🎉

For additional support, check the application logs or create an issue in the project repository.