Decision API converter: OpenAI, Gemini and System One (Jev, Clef)
Paste a decision request and pick the target API. Get the converted request, a curl snippet with placeholder keys, and a warning for every step that loses something: option order, criteria, limits, unsupported input.
Runs in your browser. Nothing you paste is uploaded.
curl
How the answers come back
| Question | OpenAI | Gemini | Jev / Clef |
|---|---|---|---|
| Yes/no | {type: "predicate", name, probability} | {type: "noul", noul (0.01–0.99), confidence} | {type: "noul", noul} |
| Pick one | {type: "choice", name, choice, probabilities: [{value, probability}], confidence} | {type: "choice", choice, probabilities: {option: p}, confidence} | {type: "choice", choice, confidence, probabilities: {…}} |
| Ordered rubric | {type: "score", name, score, probabilities: [{value, label, probability}], confidence} | {type: "score", score, legend: {"0": …}, probabilities: {"0": p}, confidence} | {type: "score", score, confidence, legend, probabilities} |
| Refusal | {type: "refusal", name} (name is null for an unnamed question) | not documented | not documented |
| Container | array, in question order | map keyed by question id | map keyed by question name (Clef wraps the response in result: unconfirmed) |
17 warning types. Most link to the provider's docs; format facts come from the sources listed on each format pair page (last checked 2026-11-19).
How to use it
- Paste a decision request body. The format is detected: a
questionsarray is OpenAI, aquestionsmap withinputis Gemini, andstateis System One. Or pick it in From. - Pick the target in To: OpenAI, Gemini, Jev or Clef.
- Read the warnings, grouped by question, then Copy JSON or Copy curl. The curl uses
$ENV_VARplaceholders, never a real key.
How the converter works
It reads your request into one shared shape (yes/no, pick one, ordered rubric, plus the input), then writes it out in the target format. Each step that can’t carry something across adds a warning, most with a link to the provider’s docs:
- Yes/no criteria. OpenAI predicates have no criteria, so Gemini’s or System One’s definitions are folded into
instructions. - Option order. Gemini sorts option keys alphabetically. OpenAI keeps an array in your order.
- Rubrics. OpenAI levels are
{label, description}; Gemini and System One take a list of strings, so labels and descriptions are merged or split. - Limits. Question, option and level counts are checked against each target’s documented limits, and question keys are sanitized to the target’s allowed characters.
- Input. OpenAI
input_jsonand conversation history are serialized to text for Gemini; images and other media need a manual step; Jev’sboxquestion has no equivalent elsewhere.
Not sure a decision API fits? Turn your classifier schema into decision questions and see which fields still need an LLM.
How we check this
Format details come from the providers’ docs (quoted or noted on each format pair page), checked 2026-11-19. Where the docs don’t confirm something (the Gemini REST path, the Jev auth header and question limit, Clef’s response wrapper), the converter says so instead of guessing.
FAQ
What's the difference between `predicate` and `noul`?
Both are yes/no questions. OpenAI calls it predicate and answers with a probability. Gemini and System One call it noul and answer with a noul value (0.01–0.99 on Gemini). OpenAI predicates have no criteria field, so the converter folds Gemini’s or System One’s yes/no definitions into instructions and warns you.
Why did my option order change when converting to Gemini?
Gemini takes options as a criteria map and sorts the keys alphabetically in UTF-8 byte order, so your order isn’t kept. If the order carries meaning, use a score question instead.
Can I send conversation history to Gemini Decisions?
No. Google’s docs say conversation history, thought and tool-call blocks aren’t supported and to serialize multi-turn context as text. The converter does that and warns you.
Is the Gemini Decisions REST path confirmed?
Not yet. The SDK call is documented, but the REST path we use comes from third-party posts. The curl snippet says so in a comment. Check Google’s API reference before you rely on it.
Is what I paste sent anywhere?
No. Conversion runs in your browser and the page makes no network requests with your input. It isn’t stored or put in the URL.