Implementation:Protectai Llm guard Output Regex

Knowledge Sources	Protectai_Llm_guard
Domains	Pattern_Matching, Content_Filtering
Last Updated	2026-02-14 12:00 GMT

Overview

Regex is an output scanner that detects or enforces regex patterns in LLM responses by delegating to the input-side InputRegex scanner.

Description

The Regex output scanner is a thin wrapper around the corresponding input scanner InputRegex. It applies a list of regular expression patterns to the LLM output and determines whether the output should be flagged based on pattern matches. The is_blocked parameter controls the behavior: when True, the scanner flags outputs that match any of the patterns (blocking mode); when False, it flags outputs that do not match any of the patterns (requiring mode). The match_type parameter controls the regex matching strategy (e.g., SEARCH for partial matches). When redact is enabled, matched patterns are replaced in the output text.

Usage

Use this scanner when you need to enforce specific text patterns in LLM outputs. Common use cases include blocking outputs that contain sensitive patterns (e.g., credit card numbers, social security numbers, API keys), ensuring outputs match required formats (e.g., structured response templates), and redacting sensitive information that matches known patterns.

Code Reference

Source Location

• Repository: Protectai_Llm_guard
• File: llm_guard/output_scanners/regex.py
• Lines: 1-42

Signature

class Regex(Scanner):
    def __init__(
        self,
        patterns: list[str],
        *,
        is_blocked: bool = True,
        match_type: MatchType | str = MatchType.SEARCH,
        redact: bool = True,
    ) -> None: ...

    def scan(self, prompt: str, output: str) -> tuple[str, bool, float]: ...

Import

from llm_guard.output_scanners import Regex

I/O Contract

Inputs

Name	Type	Required	Description
prompt	str	Yes	The input prompt
output	str	Yes	The LLM output to scan against regex patterns

Constructor Parameters

Name	Type	Required	Default	Description
patterns	list[str]	Yes	N/A	List of regex patterns to match against the output
is_blocked	bool	No	True	If True, block matching outputs; if False, require matches
match_type	str	No	MatchType.SEARCH	Regex matching strategy (SEARCH for partial, FULL_MATCH for complete)
redact	bool	No	True	Whether to redact matched patterns in the output

Outputs

Name	Type	Description
sanitized_output	str	The output with matched patterns optionally redacted
is_valid	bool	Whether the output passed the scan
risk_score	float	Risk score (-1.0 to 1.0)

Usage Examples

Basic Usage

from llm_guard.output_scanners import Regex

# Block outputs containing credit card-like patterns
scanner = Regex(
    patterns=[r"\b\d{4}[-\s]?\d{4}[-\s]?\d{4}[-\s]?\d{4}\b"],
    is_blocked=True,
    redact=True,
)

prompt = "What is my account info?"
output = "Your card number is 4532-1234-5678-9012 and your balance is $500."

sanitized_output, is_valid, risk_score = scanner.scan(prompt, output)

print(sanitized_output)  # Card number will be redacted
print(f"Valid: {is_valid}, Risk: {risk_score}")

Require Pattern Match

from llm_guard.output_scanners import Regex

# Require output to contain a specific format (e.g., ticket number)
scanner = Regex(
    patterns=[r"TICKET-\d{6}"],
    is_blocked=False,
    redact=False,
)

prompt = "Create a support ticket"
output = "Your ticket has been created: TICKET-123456"

sanitized_output, is_valid, risk_score = scanner.scan(prompt, output)
# is_valid will be True because the required pattern is present

Related Pages

• Principle:Protectai_Llm_guard_Regex_Pattern_Matching