Use response_format when your application needs predictable JSON instead of prose. The gateway forwards the schema through the OpenAI-compatible Chat Completions route to a model that supports structured output.
Generate Schema-Constrained JSON
A matching response looks like:
Choosing a Model
Use a model that advertises support for response_format. The request remains OpenAI-compatible even when the gateway routes it to another supported provider.
json_schema constrains the response to your schema. json_object only asks
for valid JSON and does not enforce required fields or value types.
Structured responses are logged in Requests like other chat completions, including the request schema and returned JSON.
See Chat Completions for the full request reference.