Structured output

Structured Output ensures LLM responses strictly adhere to developer-defined JSON schemas with 100% reliability.

Engineers use Structured Output to eliminate the unpredictability of natural language responses. By mapping model outputs to a specific JSON schema (via tools like OpenAI's SDK or LangChain), the system guarantees that every field, type, and constraint is met. This precision enables seamless integration into automated pipelines, data extraction workflows, and programmatic interfaces without the need for retry loops or manual parsing.

https://platform.openai.com/docs/guides/structured-outputs