Request parameters
These fields are the ones you fill in on the Playground.
| Field | Type | Meaning |
|---|---|---|
model | string | Fixed to typesafe/jev-1.13. The Playground cannot switch to another billed model. |
state | string / object / array | Shared context for every question. Use text, or JSON for tickets, records and nested fields. |
questions | object | Map of unique question IDs to question objects. IDs are not sent to the model. |
questions[id] | object | One typed judgment. Put the full meaning in instructions, not in the ID. |
type | choice / score / noul | Single option, ordered score, or yes-probability. |
instructions | string / object / array | The complete question. The Playground form sends a string. |
criteria | depends on type | Choice options, Score levels, or optional Noul true/false descriptions. |
Choice
Object: option name → description, 2–255 options. An empty description is sent as null, which means “use the option name only”. Include an other / none-of-the-above option when the list may not cover every input.
Score
Array of 2–10 descriptive levels, ordered low to high. Level numbers start at 0. The returned score is a probability-weighted average and can land between levels. Three levels are 0–2, not 1–3.
Noul
Criteria may be omitted. If you fill them in, provide both true and false. The result noul is the probability of yes — not a separate confidence, and not an automatic boolean.
temperature, top_p, max_tokens and stream are not Playground fields.
Response
A run returns result and elapsedMs. result.answers is keyed by the same question IDs you sent. Typed output guarantees the field shape, not factual correctness.
- Choice: choice, probabilities, confidence.
- Score: score, legend, probabilities, confidence.
- Noul: noul (probability of yes). There is no separate confidence field.
- usage includes input_tokens and output_tokens, and may include cost in USD. If cost is missing, this site does not estimate a charge.
- elapsedMs is the time from sending the run to receiving the result, including validation. It is not pure model inference time.