Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.irona.ai/llms.txt

Use this file to discover all available pages before exploring further.

IronaAI allows you to specify the format of the response from the language model using structured outputs. This feature ensures that the output adheres to a defined structure, such as JSON or a Pydantic model, making it easier to parse and integrate into your application.

​
When to use structured outputs

Use structured outputs when you need the language model’s response to follow a specific schema. This is ideal for:
  • Generating structured data for databases
  • Creating consistent API responses
  • Enforcing a predefined output format
  • Structured outputs can be specified as either a JSON object or a Pydantic model.

​
Using JSON object as response format

To request a JSON object, pass response_format={"type": "json_object"} to the create method. The response will be a valid JSON string. 
response = client.completions.create(
    response_format={"type": "json_object"},
    messages=[
        {"role": "system", "content": "You are a helpful assistant designed to output JSON."},
        {"role": "user", "content": "Who won the world series in 2020?"}
    ]
)
print(response.choices[0].message.content)
The response will be a JSON string that you can parse with json.loads() in Python or JSON.parse() in TypeScript.

​
Using Pydantic model as response format

For more complex structures, define a Pydantic model and pass it as the response_format. The response will be an instance of that model. 
from pydantic import BaseModel

class CalendarEvent(BaseModel):
    name: str
    date: str
    participants: list[str]

class EventsList(BaseModel):
    events: list[CalendarEvent]

messages = [{"role": "user", "content": "List 5 important events in the XIX century"}]

resp = client.completions.create(
    messages=messages,
    response_format=EventsList
)
print(resp)
The response will be an EventsList instance, allowing direct access to its attributes. Note: Pydantic models are Python-specific. In TypeScript, use JSON schema or other validation tools.