Structured outputs and JSON

Structured outputs make the model return machine-readable JSON instead of free-form prose, so you can parse the result directly. FluxRouter accepts the standard OpenAI response_format field against https://api.fluxrouter.ai/v1/chat/completions, so you do this exactly as you would calling OpenAI. The only Flux-specific part is the base URL and the flux-auto model.

Two modes#

The response_format field has two structured modes:

• JSON mode ({"type": "json_object"}): the model returns syntactically valid JSON. You still describe the shape you want in the prompt.
• JSON schema ({"type": "json_schema", ...}): the model returns JSON that conforms to a schema you provide, so the keys and types are constrained, not just "valid JSON."

JSON mode#

Ask for JSON and set response_format to json_object. Always state in the prompt what fields you expect.

import json
from openai import OpenAI

client = OpenAI(
    api_key="sk-...",                          # your Flux key
    base_url="https://api.fluxrouter.ai/v1",   # the one line you change
)

response = client.chat.completions.create(
    model="flux-auto",
    response_format={"type": "json_object"},
    messages=[
        {
            "role": "user",
            "content": "Extract the name and age as JSON with keys 'name' and 'age': Ada is 36.",
        }
    ],
)

data = json.loads(response.choices[0].message.content)
print(data)  # {'name': 'Ada', 'age': 36}

JSON schema#

For a guaranteed shape, pass a json_schema. The model is constrained to the schema's properties and types.

from openai import OpenAI

client = OpenAI(
    api_key="sk-...",
    base_url="https://api.fluxrouter.ai/v1",
)

response = client.chat.completions.create(
    model="flux-auto",
    response_format={
        "type": "json_schema",
        "json_schema": {
            "name": "person",
            "schema": {
                "type": "object",
                "properties": {
                    "name": {"type": "string"},
                    "age": {"type": "integer"},
                },
                "required": ["name", "age"],
                "additionalProperties": False,
            },
        },
    },
    messages=[{"role": "user", "content": "Ada is 36 years old."}],
)

print(response.choices[0].message.content)  # {"name":"Ada","age":36}

curl (JSON mode)#

curl https://api.fluxrouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $FLUX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "flux-auto",
    "response_format": { "type": "json_object" },
    "messages": [
      { "role": "user", "content": "Return JSON with keys name and age: Ada is 36." }
    ]
  }'

Make sure your model supports it#

Structured output support depends on the model that serves the request. JSON mode is widely supported; strict json_schema enforcement is supported by fewer models. flux-auto routes to a sensible model, but whether a given mode is honored is a property of the underlying model, not of FluxRouter.

If your application depends on a specific mode, pin a model you have verified supports it. Pass a flux-pinned-* id (for example flux-pinned-gpt-5 or flux-pinned-claude-sonnet) instead of flux-auto. See Models for the full list.

Whatever model serves the request, always validate the JSON you get back (parse it, check required keys) rather than trusting the shape blindly.

Anthropic shape#

The Anthropic-compatible base at https://api.fluxrouter.ai/anthropic does not use response_format. To get structured output there, use Anthropic's standard patterns (a tool whose input schema is your target shape, or prompting for JSON), exactly as you would calling Anthropic directly. See Tools and function calling.