Skip to main content
Structured Output lets you configure compatible models to enforce output format and structure using optional schema to adhere to. Structured Outputs can follow a predefined JSON Schema:
  1. This ensures responses are type-safe, making integration within rigorous systems easy.
  2. Models also don’t need extra prompting to adhere to the defined schema, making it easy to build.

​
Configuring Response Format

There are two types of Structured Outputs available within Prompts:
  • JSON Mode, enforcing an output in JSON.
  • JSON Schema, enforcing an output in a predefined schema.
JSON mode is a more basic available Response Format. It only ensures the output is a valid JSON, while JSON Schema enforces a predefined schema, making it more rigorous and safe for integrating models in your pipeline.
To configure a Prompt’s output, look for the Response Format configuration:

Choose the desired Response Format.

Structured Output configuration will only appear when the chosen model is compatible. To find compatible models, refer to your LLM providers’ documentation.

​
Configuring a JSON Schema

When selecting a JSON Schema Response Format, you need to define the Schema which the model follows.
Response Format Json Schema Pn

Choose an existing schema or create a new one.

When creating a new Schema, the following modal opens, letting you configure the JSON Schema payload.
Json Schema Configuration Pn

Once saved within your project, a schema can be reused between multiple prompt configurations.

To learn more about JSON Schema definitions, refer to the Official Documentation.