noulYes-or-no
Returns a 0-1 probability of “true” for the state — fits binary decisions and threshold gating.
"type": "noul", "criteria": { "true": "meets the criteria", "false": "does not meet" }
The aggregated home of System One decision models — Jev, Decider, Drex, d1, Clef and more behind one typesafe-sdk-compatible endpoint. Swap base_url and api_key, then pick any model per request with stable, precisely-billed answers.
https://api.easyjev.onlineFrom latency and billing to failover, built for every System One decision model — not just one.
System One models drop the autoregressive loop — a single forward pass completes each evaluation in milliseconds, a natural fit for high-frequency, real-time decision paths.
Prepaid and usage-based — no subscription, no auto-renewal; a top-up of just $5 is enough to get started.
Every new account is automatically credited $1 in free trial credit on first sign-in — run your first judgments at zero cost.
Fully compatible with the official typesafe-sdk Python package — change only base_url and api_key to integrate, then pick any model per request.
The full range of mainstream System One decision models — Jev, Decider, Drex, d1, Clef and more — all on sale behind one endpoint.
Open the Playground in your browser, feed your own state and questions to a real model, and see decision models at work — no setup required.
System One models are decision models — Jev, Perplexity Decider, Nace.AI Drex, Liquid d1, Cloudflare Clef and more: instead of generating text, they directly return structured judgments for the state and questions you provide. EasyJev hosts them all behind one typesafe-sdk-compatible endpoint.
noulReturns a 0-1 probability of “true” for the state — fits binary decisions and threshold gating.
"type": "noul", "criteria": { "true": "meets the criteria", "false": "does not meet" }
choiceReturns a distribution and top pick across up to 255 candidates — fits routing and classification.
"type": "choice", "criteria": { "a": "description for a", "b": "description for b" }
scoreReturns an interpolable score across 2-10 ordered levels — fits grading and severity assessment.
"type": "score", "criteria": ["low", "mid", "high"]
See a model judge your own state and questions — live, right in the browser.
Open the PlaygroundSystem One models do not generate text — they evaluate state.Use a generative model for writing, summarizing, or conversation. When you need a deterministic yes / no, which-option, or what-level decision, pick a System One model — that is their home turf.
One billing model across the whole catalog. Per-model live prices live in the Model Square.
Every model’s input price, release date and live status — one square, one glance.
View all model pricingCompatible with the official typesafe-sdk — from sign-up to your first judgment call, only two parameters change.
Enter an email, get a code, and you are in — first sign-in creates your account automatically; new users get $1 in free trial credit.
Create a gateway key in the console. The full plaintext is shown only once at creation — store it safely.
Point typesafe-sdk’s base_url and api_key at this site; everything else stays untouched.
from typesafe_sdk import Noul, TypeSafeClient # The only two edits: your sk- key from this site, and our base_url client = TypeSafeClient( api_key="sk-your-key", base_url="https://api.easyjev.online", ) result = client.system_one( state="User asks: can I get a refund for order #1024?", questions={"refund": Noul(instructions="Is this a refund request?")}, ) print(result.nouls["refund"].noul) # probability of "yes", e.g. 0.97
import requests resp = requests.post( "https://api.easyjev.online/v1/systemone", headers={"Authorization": "Bearer sk-your-key"}, json={ "model": "typesafe/jev-latest", "state": {"price": 100}, "questions": { "q1": { "type": "choice", "instructions": "pick the cheapest", "criteria": {"a": "option a", "b": "option b"}, } }, }, ) print(resp.json()["answers"]) # {"q1": {"choice": "a", "confidence": 0.9}}
Billing, top-ups and model boundaries — the answer is probably here.
All amounts are metered in integer quota, where 1,000,000,000 quota = $1 — no floating-point drift. Every model on this site bills output at $0; each model’s input price (e.g. jev-latest at $0.042 per million tokens, Decider V1.2 at $0.02) is shown live on the pricing page. Each call costs input tokens × input price/M + output tokens × output price/M (each term rounded, then summed), settled exactly against the usage returned by upstream.
When the balance is ≤ 0, calls return 402 (insufficient_balance) and are rejected. The check happens before the call: the current call may drive the balance negative, and only the next call gets blocked — so a judgment is never cut off halfway.
Online top-up is live: open the Recharge page, pick a preset amount or enter a custom one, and pay in the hosted checkout — cards, Apple Pay, Google Pay and WeChat Pay are supported. Credit lands in your balance automatically after payment. If a payment does not arrive promptly, email support@easyjev.online with your order number.
Yes. EasyJev serves the whole System One catalog through its own gateway platform: successful response bodies match the official API shape exactly (including usage and unknown fields), so the official SDK works with only base_url and api_key swapped. Existing model names — jev-1.13.0, jev-1.14.0 and the jev-latest alias — keep working unchanged, and every other decision model is selectable per request.
System One models generate no text — they only evaluate state: given state and questions, they return structured judgments — a true/false probability, an option distribution, or an ordinal score — each with a probability usable as confidence. They fit deterministic decision scenarios such as intent routing, moderation pre-filtering and permission checks; use a generative model for writing, summarizing or conversation.
Yes — every new account is automatically credited $1 in free trial credit on first sign-in (the current default; the amount follows the notice shown on the page at sign-up time).