Skip to content

Regex Match

The regex_match guardrail validates output against regex patterns.

Import

from pydantic_ai_guardrails.guardrails.output import regex_match

Basic Usage

from pydantic_ai_guardrails import GuardedAgent
from pydantic_ai_guardrails.guardrails.output import regex_match

guarded_agent = GuardedAgent(
    agent,
    output_guardrails=[
        regex_match(pattern=r'^[A-Z]', must_match=True),
    ],
)

Parameters

Parameter	Type	Default	Description
`pattern`	`str`	Required	Regex pattern
`must_match`	`bool`	`True`	If True, output must match. If False, output must not match.
`flags`	`int`	`0`	Regex flags (e.g., `re.IGNORECASE`)

Examples

Must Match

# Output must start with capital letter
guardrail = regex_match(
    pattern=r'^[A-Z]',
    must_match=True,
)

Must Not Match

# Block TODO markers
guardrail = regex_match(
    pattern=r'TODO|FIXME|XXX',
    must_match=False,
)

Case Insensitive

import re

guardrail = regex_match(
    pattern=r'error|warning|fail',
    must_match=False,
    flags=re.IGNORECASE,
)

Format Validation

# Must be a valid date format
guardrail = regex_match(
    pattern=r'^\d{4}-\d{2}-\d{2}$',
    must_match=True,
)

Violation Result

When triggered, returns:

{
    'tripwire_triggered': True,
    'message': 'Output does not match required pattern',
    'severity': 'medium',
    'metadata': {
        'pattern': r'^[A-Z]',
        'must_match': True,
    },
}

Use Cases

Format enforcement: Dates, IDs, codes
Content filtering: Block unwanted phrases
Structure validation: Headers, sections
Quality checks: No placeholder text