Set Request.Output when the assistant must produce exactly one JSON object that can later be extracted or decoded into a concrete struct.

API surface

request := inference.Request{
	Model: model.CustomModel(
		model.ProviderName("acme"), model.APIFormatOpenAI,
		"https://api.example.test", "chat-1",
		model.WithStructuredOutput(),
	),
	Output: &inference.OutputSchema{
		Name: "answer",
		Description: "The concise answer.",
		Schema: json.RawMessage(`{
			"type":"object",
			"properties":{"answer":{"type":"string"}},
			"required":["answer"],
			"additionalProperties":false
		}`),
		Strict: true,
	},
}
if err := inference.ValidateRequestFeatures(request); err != nil {
	panic(err)
}

Validation runs ValidateOutputSchema first. The model must advertise Caps.StructuredOutput; if ordinary tools are also present it must additionally advertise Caps.StructuredOutputWithTools. The reserved internal tool name _looprig_final_output cannot be supplied as an ordinary tool name.

Strict is part of the provider-neutral contract and is passed to codecs that support a strict response format. Extraction still checks the response representation and finish reason; a valid request does not guarantee valid model output.

Proof

Source: inference/client.go, inference/output.go
Tests: inference/client_test.go, inference/output_test.go