Skip to content

JSON Validator

The json_validator guardrail ensures LLM output is valid JSON, optionally matching a schema.

Import

from pydantic_ai_guardrails.guardrails.output import json_validator

Basic Usage

from pydantic_ai_guardrails import GuardedAgent
from pydantic_ai_guardrails.guardrails.output import json_validator

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

Parameters

Parameter	Type	Default	Description
`schema`	`dict \| None`	`None`	JSON Schema to validate against

Examples

Basic JSON Validation

# Just check it's valid JSON
guardrail = json_validator()

Schema Validation

guardrail = json_validator(
    schema={
        'type': 'object',
        'properties': {
            'name': {'type': 'string'},
            'age': {'type': 'integer', 'minimum': 0},
            'email': {'type': 'string', 'format': 'email'},
        },
        'required': ['name', 'age'],
    }
)

Array Schema

guardrail = json_validator(
    schema={
        'type': 'array',
        'items': {
            'type': 'object',
            'properties': {
                'id': {'type': 'integer'},
                'title': {'type': 'string'},
            },
        },
    }
)

Violation Result

When triggered, returns:

{
    'tripwire_triggered': True,
    'message': 'Invalid JSON: Expecting property name',
    'severity': 'high',
    'metadata': {
        'error': 'JSONDecodeError',
        'position': 42,
    },
    'suggestion': 'Ensure output is valid JSON format',
}

Use Cases

API responses: Ensure structured output
Data extraction: Validate extracted entities
Configuration: Generate valid config files
Integration: Prepare data for downstream systems