OpenAI API: Structured Outputs (JSON "schema")

 Structured Outputs - OpenAI API

Structured Outputs is available in two forms in the OpenAI API:

1. When using function calling
2. When using a json_schema response format

Structured Outputs via response_format are more suitable when you want to indicate a structured schema for use when the model responds to the user, rather than when the model calls a tool.

Structured Outputs with response_format: {type: "json_schema", ...} is only supported with the gpt-4o-mini, gpt-4o-mini-2024-07-18, and gpt-4o-2024-08-06 model snapshots and later.

Welcome to Pydantic - Pydantic

Pydantic is the most widely used data validation library for Python.

The name "Pydantic" is a portmanteau of "Py" and "pedantic."

zod - npm

TypeScript-first schema validation with static type inference

Structured Outputs - OpenAI API

JSON mode is a more basic version of the Structured Outputs feature. While JSON mode ensures that model output is valid JSON, Structured Outputs reliably matches the model's output to the schema you specify.

Structured Outputs - OpenAI API

Structured Outputs supports a subset of the JSON Schema language.

Supported types

The following types are supported for Structured Outputs:

• String
• Number
• Boolean
• Integer
• Object
• Array
• Enum
• anyOf

Introduction to Structured Outputs | OpenAI Cookbook