Guardrail AI: Comprehensive Guide for Python Applications

Guardrail AI is an open-source framework specifically designed for implementing safety guardrails in AI applications. It helps ensure AI systems operate within defined boundaries and follow specific guidelines.

What is Guardrail AI?

Guardrail AI provides:

Validation of AI outputs against custom rules

Quality checks for generated content

Bias detection and mitigation

Structured output enforcement

PII detection and redaction

Custom rule creation

Installation

bash

pip install guardrail-ai

# Or with specific components

pip install guardrail-ai[all]

pip install guardrail-ai[pii]

pip install guardrail-ai[quality]

1. Basic Usage Examples

Simple Content Validation

python

from guardrail import Guardrail

from guardrail.validators import ProfanityFilter, ToxicityFilter, PIIFilter

# Initialize guardrail with validators

guardrail = Guardrail(

    validators=[

        ProfanityFilter(),

        ToxicityFilter(threshold=0.8),

        PIIFilter(entities=["EMAIL", "PHONE_NUMBER", "SSN"])

    ]

)

# Validate text

text = "This is a sample text with an email user@example.com"

result = guardrail.validate(text)

print(f"Valid: {result.is_valid}")

print(f"Violations: {result.violations}")

print(f"Sanitized text: {result.sanitized_text}")