7.7.3. OPENAI-03 — Structured Outputs (JSON mode)

JSON mode asks the server to emit valid JSON as the message content. You then decode that content string into your own struct — using the same sscan_json machinery the client uses internally.

7.7.3.1. Requesting JSON

json_object_format() returns the response_format value for JSON mode. Assign it to the request:

var req = ChatCompletionRequest(model = "gpt-4o-mini",
    messages <- [ChatMessage(role = "user",
        content = "Return the answer to 6*7 as JSON: {\"answer\": N}.")])
req.response_format = json_object_format()
let res = chat(client, req)

7.7.3.2. Decoding the content

Declare a struct matching the JSON you expect, then decode the assistant’s content into it. With options rtti, sscan_json maps JSON keys onto the struct fields:

struct Answer {
    answer : int
}

let content = res.response.choices[0].message.content
var parsed : Answer
if (sscan_json(content, parsed)) {
    print("decoded answer = {parsed.answer}\n")
}

This is the bridge from “model produced text” to “typed daslang value” — the model returns a JSON string, and sscan_json turns it into a struct you can use directly.

7.7.3.3. Quick Reference

Function / value	Description
`json_object_format()`	The `response_format` value for JSON mode
`req.response_format = …`	Enable JSON mode on a request
`sscan_json(content, parsed)`	Decode a JSON string into a struct (needs rtti)