Structured output - Yelu AI API

Structured output makes model responses easier to parse and integrate. Depending on the selected model, use JSON object mode or a strict JSON schema.

Request a JSON object

Tell the model to produce JSON and set response_format. Some models require an explicit instruction containing the word “JSON.”

curl https://api.yelu.ai/v1/chat/completions \
  -H "Authorization: Bearer $YELU_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o-mini",
    "messages": [
      {"role":"system","content":"Return valid JSON with keys summary and priority."},
      {"role":"user","content":"The checkout API returns 503 for five percent of requests."}
    ],
    "response_format": {"type":"json_object"},
    "temperature": 0
  }'

const response = await fetch('https://api.yelu.ai/v1/chat/completions', {
  method: 'POST',
  headers: { Authorization: `Bearer ${process.env.YELU_API_KEY}`, 'Content-Type': 'application/json' },
  body: JSON.stringify({
    model: 'gpt-4o-mini',
    messages: [
      { role: 'system', content: 'Return valid JSON with keys summary and priority.' },
      { role: 'user', content: 'The checkout API returns 503 for five percent of requests.' },
    ],
    response_format: { type: 'json_object' },
    temperature: 0,
  }),
});
if (!response.ok) throw new Error(await response.text());
const result = JSON.parse((await response.json()).choices[0].message.content);
console.log(result);

import json
import os
import requests

response = requests.post(
    "https://api.yelu.ai/v1/chat/completions",
    headers={"Authorization": f"Bearer {os.environ['YELU_API_KEY']}"},
    json={
        "model": "gpt-4o-mini",
        "messages": [
            {"role": "system", "content": "Return valid JSON with keys summary and priority."},
            {"role": "user", "content": "The checkout API returns 503 for five percent of requests."},
        ],
        "response_format": {"type": "json_object"},
        "temperature": 0,
    },
    timeout=60,
)
response.raise_for_status()
result = json.loads(response.json()["choices"][0]["message"]["content"])
print(result)

Use JSON schema

On models that support strict structured output, send response_format.type = json_schema with a named schema:

{
  "type": "json_schema",
  "json_schema": {
    "name": "incident_triage",
    "strict": true,
    "schema": {
      "type": "object",
      "properties": {
        "summary": {"type": "string"},
        "priority": {"type": "string", "enum": ["low", "medium", "high", "critical"]}
      },
      "required": ["summary", "priority"],
      "additionalProperties": false
    }
  }
}

Validate at the boundary

Even strict output can be interrupted by content filtering, output limits, network failure, or provider-specific behavior. Your application should:

check the HTTP response and finish status;
parse JSON with size and depth limits;
validate against the same schema used in the request;
reject unknown fields when they are unsafe;
apply domain validation, such as ranges and cross-field rules;
handle refusals and truncated output separately from schema failures.

JSON mode versus schema

Capability	JSON object	JSON schema
Valid JSON syntax	Intended	Intended
Required keys and types	Prompt-dependent	Enforced when strict support is available
Model availability	Broader	Narrower
Best use	Flexible extraction and prototypes	Production data contracts

Never interpolate model-produced JSON into SQL, shell commands, HTML, or authorization rules without context-specific escaping and validation.

​Request a JSON object

​Use JSON schema

​Validate at the boundary

​JSON mode versus schema

Request a JSON object

Use JSON schema

Validate at the boundary

JSON mode versus schema