Skip to main content

Testing Templates

Complete guide to testing prompt templates safely before deployment.

Testing Workflow​

Step-by-Step Testing​

Step 1: Open Template​

1. Navigate to GenAI Explorer
2. Search for "GenAiPromptTemplate"
3. Select your template
4. Review template content

Step 2: Fill Variables​

1. Locate the "Input Variables" form
2. Fill in all required fields (marked with *)
3. Add optional field values
4. Verify data types are correct

Step 3: Preview Template​

1. Click "Preview Template" button
2. Wait for execution
3. Review generated output
4. Check token count
5. Verify response quality

Step 4: Test Edge Cases​

Test with problematic inputs:

  • Empty strings
  • Very long text (1000+ characters)
  • Special characters (<>, &, ")
  • Null values
  • Numbers as strings
  • Unicode characters

Testing Checklist​

Happy Path Testing​

✅ Test with:

  • Typical user inputs
  • Common scenarios
  • Standard data formats
  • Expected values

Edge Case Testing​

✅ Test with:

  • Empty required fields (should error)
  • Empty optional fields (should work)
  • Maximum length text
  • Special characters: <>, &, ", '
  • Line breaks and formatting
  • Non-English characters

Error Handling​

❌ Should Fail:

- Missing required variables
- Invalid data types
- Malformed input

✅ Should Succeed:

- Missing optional variables
- Edge case values
- Special characters (escaped)

Test Scenarios​

Scenario 1: Customer Support Template​

Template:

Help {!CustomerName} with their {!IssueType} regarding {!ProductName}.

Test Cases:

TestCustomerNameIssueTypeProductNameExpected
Happy Path"John Smith""billing""Premium"✅ Substitutes all
Empty Optional"Jane Doe""""Basic"✅ Works with empty
Long Name"Christopher...""technical""Enterprise"✅ Handles long text
Special Chars"O'Brien & Co""access""Standard"✅ Escapes properly

Scenario 2: Data Analysis Template​

Template:

Analyze this {!DataType} data: {!DataContent}

Provide:
1. Key trend
2. Recommendation

Test Cases:

TestDataTypeDataContentExpected
Normal"sales""Q1: $100K..."✅ Analyzes data
Large Data"transactions""10KB of data"✅ Handles volume
Empty Data"metrics"""❌ Should error
Invalid Type"unknown""abc123"⚠️ Model interprets

Validation Rules​

Required Fields​

// Required field validation
if (variable.required && !value) {
  throw new Error(`${variable.name} is required`);
}

Test:

  • Leave required field empty → Should error
  • Fill required field → Should succeed

Type Validation​

// Type validation
if (variable.type === "Number" && isNaN(value)) {
  throw new Error(`${variable.name} must be a number`);
}

Test:

  • Enter text in number field → Should error
  • Enter number → Should succeed

Length Validation​

// Max length check
if (value.length > maxLength) {
  throw new Error(`${variable.name} exceeds max length`);
}

Test:

  • Enter 10,000 characters → Check behavior
  • Enter normal length → Should succeed

Testing Best Practices​

✅ Do's​

  • ✅ Test with realistic production data
  • ✅ Test all edge cases before deploying
  • ✅ Verify token counts are reasonable
  • ✅ Test with different AI models if applicable
  • ✅ Check output format matches expectations
  • ✅ Test multiple times for consistency

❌ Don'ts​

  • ❌ Skip edge case testing
  • ❌ Assume variables will always have values
  • ❌ Deploy without checking token counts
  • ❌ Forget to test with special characters
  • ❌ Ignore validation errors
  • ❌ Test only once

Debugging Issues​

Issue: Variables Not Substituting​

Symptoms:

Output shows: "Help {!CustomerName}..."
Instead of: "Help John Smith..."

Solutions:

  1. Check variable name matches exactly
  2. Verify variable is defined in template
  3. Ensure value is provided in form
  4. Check for typos in variable reference

Issue: Unexpected Output​

Symptoms:

  • Output doesn't make sense
  • Wrong information included
  • Formatting broken

Solutions:

  1. Review template content carefully
  2. Check variable substitution
  3. Test with simpler inputs
  4. Verify AI model is correct
  5. Check for conflicting instructions

Issue: High Token Count​

Symptoms:

  • Token count > 1000
  • Cost too high
  • Slow responses

Solutions:

  1. Shorten template text
  2. Remove redundant instructions
  3. Use more concise language
  4. Consider smaller AI model

See: Token Optimization for details

Pre-Deployment Checklist​

Before activating a template:

  • ✅ Tested with realistic data
  • ✅ Tested with edge cases (empty, null, long)
  • ✅ All variables substitute correctly
  • ✅ Token count is reasonable (<800)
  • ✅ Tested with different AI models (if applicable)
  • ✅ Output format matches expectations
  • ✅ Required fields have validation
  • ✅ Optional fields have defaults or handle null
  • ✅ Special characters are handled
  • ✅ Tested multiple times for consistency
  • ✅ Performance is acceptable (<3s)
  • ✅ Cost per execution is within budget

Thorough testing prevents production issues and ensures high-quality AI responses.