Playground
Build a decision in the console without writing code, see the answer, and copy it as code.
The console's Playground (the decision builder) lets you try questions on your own content before you write any code. It runs in your active organisation and counts towards its usage, like any API request.
Run a decision
- Open Playground in the console.
- Choose a model: the public models and any fine-tuned models your organisation has deployed.
- Enter the state: plain text, or JSON when you pick the JSON state format.
- Add one or more questions. Give each a name (letters, digits,
_or-, starting with a letter), a type and instructions:- Yes/no (
noul): a statement or question; the answer is a probability. - Choice: 2 to 255 options, one per line (
key: description). - Score: 2 to 10 levels, one per line, lowest first.
- Yes/no (
- Select Run decision. The answer shows each question's probabilities, its confidence, the model version that answered and the usage.
See Question types and Confidence for how to read the result.
Copy as code
Copy as code turns the request you built into a Python or JavaScript snippet for the TypeSafe SDK. The snippet reads your key from TYPESAFE_API_KEY: create one under API keys. The Playground itself does not need a key and never shows one.
Playground or Try it?
| Playground | API reference Try it | |
|---|---|---|
| Where | Console, signed in | These docs, no sign-in |
| Key | Not needed | Paste an API key; it stays in the browser tab's memory |
| Good for | Designing questions on real content | Checking exact request and response shapes, headers and errors |
Next steps
Last updated
Fine-tune a model with Foundry
Prepare labelled examples, train a private adapter on Sieve-4B, check it against the base model, deploy it and call it as ft:<name>.
Route uncertain answers to review
Act on confident answers automatically and send the doubtful ones to a person, using confidence for choice and score questions and the probability for yes/no.