Structured output

Define a schema and create the agent

from pydantic import BaseModel, Field
from veadk import Agent

class Ticket(BaseModel):
    summary: str = Field(description="One-line summary of the issue.")
    category: str = Field(description="One of: billing, bug, feature_request, other.")
    priority: str = Field(description="One of: low, medium, high.")

agent = Agent(
    name="ticket_extractor",
    instruction="Extract a support ticket from the user's message.",
    output_schema=Ticket,
)

Parse the result

run returns a JSON string that matches the schema, which you can parse directly with Pydantic:

from veadk import Runner

raw = await Runner(agent=agent, app_name="structured_output").run(
    messages="The billing page crashes every time I open it — third time now, please fix it.",
    session_id="demo-session",
)
ticket = Ticket.model_validate_json(raw)

Use Ark-native structured output

Adding enable_responses=True to the above makes VeADK use Volcengine Ark’s Responses API, converting output_schema into Ark’s native json_schema (strict: true) enforced on the model side. This is more reliable than the default approach, which relies on prompting the model to produce JSON.

agent = Agent(
    name="ticket_extractor",
    instruction="Extract a support ticket from the user's message.",
    output_schema=Ticket,
    enable_responses=True,
)

This relies on Ark’s Responses API and therefore requires an Ark model.

When output_schema is set, the agent returns only the structured result and cannot call tools or transfer to sub-agents.

Overview

Get started

Core components

Observability

Extensions

Frontend

Security

Deployment

Define a schema and create the agent

Parse the result

Use Ark-native structured output

​Define a schema and create the agent

​Parse the result

​Use Ark-native structured output

Define a schema and create the agent

Parse the result

Use Ark-native structured output