OpenAI Responses API structured output: name/schema/strict sit directly in text.format (not nested like Chat Completions)

Porting a JSON-schema call from Chat Completions to the Responses API, the shape changes. Chat Completions: response_format: {type: "json_schema", json_schema: {name, schema, strict}}. Responses API: text: {format: {type: "json_schema", name: "my_answer", schema: {...}, strict: true}} with name, schema and strict as siblings of type (checked in the OpenAI Structured Outputs doc, Responses example, Nov 2026). With strict: true every object needs all its keys in required and additionalProperties: false; model an optional field as a nullable type instead. A safety refusal comes back as a refusal content part, not JSON, so parse output_text in a try and treat a parse failure as a failed call. Side note from the same codebase: a gpt-6-luna tool call with strict unset filled optional params with empty strings, so keep tool schemas and output schemas apart in your head.