System One

TypeSafe Jev (System One) decision API.

POST /v1/systemone takes a state to evaluate and a map of typed questions, and returns structured answers with probabilities and confidence per question. It serves the TypeSafe Jev (System One) model with the native TypeSafe schema — request and response shapes are identical. This is not a chat model and cannot be called via /v1/chat/completions.

POST/v1/systemone

Connecting#

Reuse any TypeSafe documentation example as-is — replace https://api.typesafe.ai with https://apirouter.pleum.ai and your TypeSafe key with a plm_ key. The body format is unchanged.

HTTP (Python)

import os

import requests

resp = requests.post(
    "https://apirouter.pleum.ai/v1/systemone",
    headers={
        "Authorization": f"Bearer {os.environ['PLEUM_API_KEY']}",
        "Content-Type": "application/json",
    },
    json={
        "model": "jev-latest",
        "state": "Hi, I've been trying to connect my Stripe account for 3 days and I'm about to churn.",
        "questions": {
            "urgency": {"type": "noul", "instructions": "Does this message express urgency?"},
            "team": {
                "type": "choice",
                "instructions": "Which team should handle this?",
                "criteria": {
                    "billing": "Payment, invoices, subscription issues",
                    "technical": "Bugs, integrations, API errors",
                    "sales": None,
                },
            },
        },
    },
)
print(resp.json())

HTTP (TypeScript)

const res = await fetch("https://apirouter.pleum.ai/v1/systemone", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.PLEUM_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "jev-latest",
    state: "Hi, I've been trying to connect my Stripe account for 3 days and I'm about to churn.",
    questions: {
      urgency: { type: "noul", instructions: "Does this message express urgency?" },
      team: {
        type: "choice",
        instructions: "Which team should handle this?",
        criteria: {
          billing: "Payment, invoices, subscription issues",
          technical: "Bugs, integrations, API errors",
          sales: null,
        },
      },
    },
  }),
});

const data = await res.json();
console.log(data.answers.urgency.noul); // e.g. 0.999

Authenticate with a plm_ API key in the Authorization: Bearer header.

Request body#

Parameter	Type	Required	Description
model	string	Required	Model ID, e.g. `jev-latest`.
state	string \| object \| array	Required	Content to evaluate — text, chat logs, records, app state.
questions	map<string, Question>	Required	Keys are question names; values are question objects. Must contain at least one entry. The key is only used to map answers back and is not sent to the model.
questions.*.type	"noul" \| "choice" \| "score"	Required	Question type — see Question types below.
questions.*.instructions	string \| object \| array	Required	The question itself.
questions.*.criteria	object \| array	Optional	Per-type criteria — noul: {true, false} descriptions; choice: map from option to description (null allowed); score: ordered array of level descriptions (minimum two).

request

curl https://apirouter.pleum.ai/v1/systemone \
  -H "Authorization: Bearer $PLEUM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "jev-latest",
    "state": "Hi, I'"'"'ve been trying to connect my Stripe account for 3 days.",
    "questions": {
      "urgency": {"type": "noul", "instructions": "Does this message express urgency?"}
    }
  }'

Question types#

noul — true/false probability. Returns a noul value between 0 and 1.

choice — classification. Returns the top choice, probabilities summing to 1, and derived confidence.

score — rubric scoring. Returns the probability-weighted score (may land between levels), a legend, and probabilities.

Response#

answers returns one answer per question under the same key you chose. Questions are evaluated in parallel and isolation — decompose complex judgments into atomic questions and combine them in one call (official TypeSafe guidance). usage carries measured tokens; cost is a PleumRouter extension (KRW cost, FX rate, markup) and is included only once settlement has finished.

200 OK

{
  "model": "jev-latest",
  "answers": {
    "urgency": {"type": "noul", "noul": 0.999},
    "team": {
      "type": "choice",
      "choice": "billing",
      "probabilities": {"billing": 0.84, "technical": 0.159, "sales": 0.001},
      "confidence": 0.84
    }
  },
  "usage": {"input_tokens": 312, "output_tokens": 48},
  "cost": {"usd": 0.000013104, "krw": 0.1, "fx_rate": 1345.69, "markup_rate": 0.0}
}

Note: Jev is most accurate in English; CJK languages including Korean may be less accurate (official TypeSafe guidance). Write important states and instructions in English where possible.

Billing#

Billing is metered on the measured usage.input_tokens (output tokens are free). A conservative estimate holds credits at request time; settlement captures the measured amount.