---
title: One endpoint, three shapes
description: "POST /v1/systemone runs a Decision alone, Jev questions alone, or Jev answers feeding a Decision."
icon: brain
---

**One endpoint, three shapes.** `POST /v1/systemone` is the only public operation. What you include in the request decides what runs:

| Shape | Send | What runs |
|---|---|---|
| **Decision only** | `decision` + `state` | Your native Decision, locally. No provider is called. |
| **Jev only** | `questions` + `state` | Jev alone. Its answers are the result; no Decision is created or executed. |
| **Jev + Decision** | `decision` + `questions` + `state` | Jev first, then your Decision over its answers and your state. The Decision returns the consolidated result. |

All three shapes cost **$0.042 per 1M input tokens, with output free**. The response header `X-aityx-Mode` reports which shape ran: `defined`, `jev` or `jev_aityx`.

## Decision only

Send a native `decision` with a `state` that matches its input schema. Answers and types are inferred from the model’s decisions. There is no repeated output definition, and no AI reads or fills in your inputs on this path.

The same model and validated facts produce the same answers. If the facts do not match, the call returns `422 invalid_state`; it never falls back to AI interpretation. Supply matching facts, or add Jev questions that establish the missing ones.

## Jev only

Send top-level Jev `questions` with your `state`, and optionally a Jev `model` alias. The request is forwarded to Jev and its answers are returned as the result. No aityx Decision is created, and amounts, dates and conditions are not calculated for you. Jev question types are `noul`, `choice` and `score`.

This shape suits semantic answers your application consumes directly. See [Using Jev with aityx](/docs/coming-from-jev).

## Jev + Decision

Add a native `decision` alongside the questions and one shared `state`. Jev runs first. Its named answers merge with the state — Jev answers override state fields with the same names — and the Decision then calculates the consolidated result your application acts on.

- **Noul** becomes a numeric probability in `[0,1]`.
- **Choice** becomes its selected string.
- **Score** becomes a number.

The API does not threshold or round those values; any cutoff belongs in your Decision’s expressions. A model input can also bind any field of the Jev response with `from: "jev:<JSON Pointer>"`. Jev adds its own inference and network time, so native millisecond timings do not describe the whole chained request.

If Jev succeeds but the Decision then fails, the call returns HTTP `200` with Jev’s answers and a `decision_error`. Check that field before treating the result as a completed Decision. See [Errors](/docs/resources/errors).

## There is no authoring or interpretation endpoint

aityx does not generate Decisions from prompts or documents, and does not interpret free-form input into model facts. You [write the native JSON](/docs/guides/define-by-hand) — by hand or with your own development tools — and send it with each case. To run unstructured input through a Decision, send Jev questions with it, as above.

The console’s **Check** and **Replay** tools are console operations, not extra public API routes. To compare models from code, execute your cases against each definition. See [Compare revised rules](/docs/guides/revise-and-replay).

[Pricing and limits](/docs/resources/pricing-and-limits) explains token accounting and credits.
