Skip to content

No Refusals

The no_refusals guardrail detects when the LLM refuses to answer or deflects the question.

Import

from pydantic_ai_guardrails.guardrails.output import no_refusals

Basic Usage

from pydantic_ai_guardrails import GuardedAgent
from pydantic_ai_guardrails.guardrails.output import no_refusals

guarded_agent = GuardedAgent(
    agent,
    output_guardrails=[
        no_refusals(),
    ],
)

Parameters

Parameter	Type	Default	Description
`refusal_patterns`	`list[str] \| None`	Default patterns	Custom refusal patterns

Default Patterns

Detects phrases like:

“I cannot help with that”
“I’m not able to”
“As an AI, I don’t”
“I apologize, but I cannot”
“I’m sorry, but I can’t”
“I don’t have the ability to”

Examples

Default Detection

guardrail = no_refusals()

Custom Patterns

guardrail = no_refusals(
    refusal_patterns=[
        r"I cannot",
        r"I'm unable to",
        r"outside my capabilities",
        r"I don't have access",
    ],
)

Violation Result

When triggered, returns:

{
    'tripwire_triggered': True,
    'message': 'Model refused to answer',
    'severity': 'medium',
    'metadata': {
        'pattern_matched': "I cannot help with that",
    },
    'suggestion': 'Provide a helpful response that addresses the question directly',
}

Use Cases

Completeness: Ensure agent always attempts to help
UX quality: Avoid unhelpful responses
Retry trigger: Use with auto-retry to get better answers

With Auto-Retry

guarded_agent = GuardedAgent(
    agent,
    output_guardrails=[no_refusals()],
    max_retries=2,
)

The agent will retry with feedback to provide a helpful response.