Regex Check

Validate outputs using regex guardrails. Catch unsafe, malformed, or non-compliant responses automatically.

Objective This metric validates responses against predefined regex patterns, ensuring compliance with specific formats (e.g., email, date, or phone formats).

Interpretation A higher score indicates that the response does not comply with the required regex format or fails validation. A lower (or zero) score indicates compliance.

Code Execution

metrics = [
    {
        "name": "Regex Check",
        "config": {
            "model": "gpt-4o-mini",
            "provider": "openai"
        },
        "column_name": "your-column-identifier",
        "schema_mapping": schema_mapping
    }
]

Example

• Prompt: “Enter your email address.”

• Context: “Valid email format: [email protected]”

• Response: “myemail[at]example.com”

• Metric Output: {"score": 1, "reason": "Response does not match the required email regex format."}