Implementation:Guardrails ai Guardrails Guard For Pydantic

Knowledge Sources	Guardrails Guardrails Guard API
Domains	Factory_Pattern, Structured_Output
Last Updated	2026-02-14 00:00 GMT

Overview

Concrete factory classmethod for creating a Guard from a Pydantic model provided by the guardrails package.

Description

The Guard.for_pydantic classmethod accepts a Pydantic BaseModel class (or List of models), extracts its JSON schema and embedded validators via pydantic_model_to_schema, and returns a Guard instance configured for structured output generation. It supports optional output formatters including the jsonformer constrained decoding backend.

Usage

Use this classmethod when you have a Pydantic model defining the desired output structure. Pass the model class to get a Guard ready for structured output generation.

Code Reference

Source Location

• Repository: guardrails
• File: guardrails/guard.py
• Lines: L453-508

Signature

@classmethod
def for_pydantic(
    cls,
    output_class: ModelOrListOfModels,
    *,
    reask_messages: Optional[List[Dict]] = None,
    messages: Optional[List[Dict]] = None,
    name: Optional[str] = None,
    description: Optional[str] = None,
    output_formatter: Optional[Union[str, BaseFormatter]] = None,
):
    """Create a Guard instance using a Pydantic model to specify the output schema.

    Args:
        output_class: The pydantic model that describes the desired output structure.
        messages: A list of messages to give to the LLM.
        reask_messages: A list of messages to use during reasks.
        name: A unique name for this Guard.
        description: A description for this Guard.
        output_formatter: 'none' (default), 'jsonformer', or a Guardrails Formatter.
    """

Import

from guardrails import Guard

I/O Contract

Inputs

Name	Type	Required	Description
output_class	ModelOrListOfModels	Yes	Pydantic BaseModel class or List[BaseModel]
messages	Optional[List[Dict]]	No	Default LLM messages
reask_messages	Optional[List[Dict]]	No	Messages for re-ask prompts
name	Optional[str]	No	Unique guard name (defaults to gr-{id})
description	Optional[str]	No	Guard description
output_formatter	Optional[Union[str, BaseFormatter]]	No	"none" (default), "jsonformer", or custom BaseFormatter

Outputs

Name	Type	Description
guard	Guard[Dict] or Guard[List]	Guard instance with output schema, validators, and output type configured

Usage Examples

Basic Structured Output

from pydantic import BaseModel, Field
from guardrails import Guard

class Person(BaseModel):
    name: str = Field(description="Person's full name")
    age: int = Field(description="Person's age")

guard = Guard.for_pydantic(output_class=Person)
result = guard(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Generate a person profile."}],
)
print(result.validated_output)  # {"name": "...", "age": ...}

With Jsonformer Constrained Decoding

guard = Guard.for_pydantic(
    output_class=Person,
    output_formatter="jsonformer",
)

Related Pages

Implements Principle

• Principle:Guardrails_ai_Guardrails_Guard_Pydantic_Factory

Requires Environment

• Environment:Guardrails_ai_Guardrails_Python_3_10_Runtime