Implementation:Guardrails ai Guardrails Guard Call Structured

Knowledge Sources	Guardrails
Domains	Validation, Structured_Output
Last Updated	2026-02-14 00:00 GMT

Overview

Concrete method for executing a Guard with structured JSON output parameters provided by the guardrails package.

Description

This is the same Guard.__call__ method used for plain text validation, but invoked with structured output parameters (tools or response_format) that guide the LLM to produce JSON. The Runner internally parses JSON output and validates each field against path-specific validators from the Pydantic schema. The validated_output is a Python dict matching the Pydantic model structure.

Usage

Use this after creating a Guard with Guard.for_pydantic() and generating strategy parameters. Pass tools= or response_format= as keyword arguments.

Code Reference

Source Location

• Repository: guardrails
• File: guardrails/guard.py
• Lines: L745-795

Signature

@trace(name="/guard_call", origin="Guard.__call__")
def __call__(
    self,
    llm_api: Optional[Callable] = None,
    *args,
    prompt_params: Optional[Dict] = None,
    num_reasks: Optional[int] = 1,
    messages: Optional[List[Dict]] = None,
    metadata: Optional[Dict] = None,
    full_schema_reask: Optional[bool] = None,
    **kwargs,  # tools=, response_format= passed here
) -> Union[ValidationOutcome[OT], Iterator[ValidationOutcome[OT]]]:

Import

from guardrails import Guard

I/O Contract

Inputs

Name	Type	Required	Description
llm_api	Optional[Callable]	No	LLM callable or model= kwarg
messages	Optional[List[Dict]]	Yes	Chat messages
tools	list	No	Function calling tools from guard.json_function_calling_tool() (via kwargs)
response_format	dict	No	JSON mode schema from guard.response_format_json_schema() (via kwargs)
num_reasks	Optional[int]	No	Max re-ask attempts (default 1)
full_schema_reask	Optional[bool]	No	Regenerate full schema vs just failed fields on reask

Outputs

Name	Type	Description
result	ValidationOutcome[Dict] or ValidationOutcome[List]	Validated output typed to match the Pydantic model structure

Usage Examples

Complete Structured Output Flow

from pydantic import BaseModel, Field
from guardrails import Guard
from guardrails.hub import ValidLength
from guardrails.types.on_fail import OnFailAction

class Movie(BaseModel):
    title: str = Field(
        description="Movie title",
        json_schema_extra={"validators": [
            ValidLength(min=1, max=200, on_fail=OnFailAction.FIX)
        ]}
    )
    year: int = Field(description="Release year")
    genre: str = Field(description="Primary genre")

guard = Guard.for_pydantic(output_class=Movie)
tools = guard.json_function_calling_tool()

result = guard(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Give me a sci-fi movie."}],
    tools=tools,
)

# result.validated_output is a dict: {"title": "...", "year": ..., "genre": "..."}
movie = Movie(**result.validated_output)

Related Pages

Implements Principle

• Principle:Guardrails_ai_Guardrails_Structured_Guard_Execution