Skip to content

Provider · TypeSafe

Run Jev decisions through INFRO.

Jev, TypeSafe’s System One model, answers typed questions — yes or no, a choice, a score — with probabilities, and writes no text. INFRO puts those decisions behind the same gateway key, spend limits and request records as your LLM calls, on your own TypeSafe account. What Jev is.

Request a 14-day trialRead the docs

Point TypeSafe’s SDK at INFRO

INFRO serves TypeSafe’s endpoint at the same path, with the same request body and model names. TypeSafe’s SDKs read their key and base URL from the environment, so pointing one at INFRO changes two variables and no code: the base URL, and an INFRO gateway key in place of the TypeSafe key. Your TypeSafe key stays in your INFRO connection, encrypted, and is sent only to TypeSafe.

jev-latest, the SDKs’ default, and jev-preview work as they do on TypeSafe and follow its releases; pin a version with TYPESAFE_DEFAULT_MODEL, as TypeSafe advises. Each response’s model names the version that answered.

The System One reference has the request, the response and every error.

TypeSafe’s SDK, pointed at INFRO

Removed: TYPESAFE_API_KEY=<your TypeSafe key>Added: TYPESAFE_API_KEY=<your INFRO gateway key>Added: TYPESAFE_BASE_URL=https://api.infro.ioAdded: TYPESAFE_DEFAULT_MODEL=jev-1.13.0  # optional: pin a version

The same request with curl

curl https://api.infro.io/v1/systemone \
  -H "Authorization: Bearer $INFRO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "jev-1.13.0",
    "state": "I was charged twice. Please refund one of them.",
    "questions": {
      "refund_request": { "type": "noul", "instructions": "Is the customer asking for a refund?" }
    }
  }'

What INFRO does with each decision

  1. Key scope

    The gateway key is checked first. A key can be limited to TypeSafe, to named models, to a project and to an environment.

  2. Spend limit

    Before TypeSafe is called, the request’s size in bytes is reserved against every spend limit that applies to it — organization, project, key, member and end customer. Over a blocking limit the request is refused, and TypeSafe is never called.

  3. Your TypeSafe connection

    The request goes to TypeSafe on your own key. With several TypeSafe keys connected, requests are spread across them, and a key that is rate limited or refused passes the request to the next one.

  4. Answers, cost and a record

    TypeSafe’s answers come back unchanged, with the version that answered and usage.cost: the estimated TypeSafe charge at TypeSafe’s list rate, or at the rate you negotiated. Every decision is recorded with its route, connection, attempts and estimated cost, and its metadata can be exported to your OpenTelemetry collector.

What a response carries

200 OK

{
  "id": "req_…",
  "model": "jev-1.13.0",
  "answers": {
    "refund_request": { "type": "noul", "noul": 0.91 }
  },
  "usage": {
    "input_tokens": 1000,
    "output_tokens": 0,
    "cost": 0.000042,
    "price_source": "list"
  },
  "route": "primary",
  "route_name": null,
  "provider": "typesafe",
  "connection": "conn_…"
}

cost is the estimated TypeSafe charge, at its list rate or the rate you negotiated. When the version that answered is one INFRO’s catalog does not carry yet, or TypeSafe reports no token counts, cost is null — unpriced, never $0.

Who bills what

TypeSafe bills your TypeSafe account for Jev, at TypeSafe’s rates. Its published list rate for Jev 1.13 is $0.042 per million input tokens, with output free (docs.typesafe.ai/models, checked 6 October 2026).

INFRO charges a subscription for the control plane and takes 0% markup on model spend. Nothing INFRO sells is priced per token. See plans.

Worth knowing before you start

  • Jev answers on POST /v1/systemone, not on chat completions; an OpenAI client cannot call it.
  • INFRO guardrails inspect chat completions only. They do not read Jev requests, and a guardrail policy set to block endpoints it cannot inspect refuses POST /v1/systemone.
  • TypeSafe sets its own limits — 64k tokens per request, and rate limits it adjusts as demand changes (docs.typesafe.ai/models, checked 6 October 2026). INFRO returns TypeSafe’s 429 with its Retry-After.
  • Request and answer bodies are stored only if your organization turned prompt logging on.
  • The console Playground does not run Jev yet.

Field guides on Jev

Questions

Do I need a TypeSafe account?
Yes. INFRO runs Jev on your own TypeSafe account, under your key, and TypeSafe bills you directly; INFRO is not part of that transaction.
Does INFRO change Jev’s answers?
No. TypeSafe’s answers are returned as TypeSafe sent them. INFRO adds the version that answered, the estimated cost, the route and the connection beside them.
Can some traffic keep going straight to TypeSafe?
Yes. A TypeSafe connection in INFRO changes nothing about your TypeSafe account. Requests sent straight to TypeSafe keep working; INFRO does not record them.

Put one Jev workflow through INFRO.

Start on your own TypeSafe account, or see the product with our team first.

Request a 14-day trialTalk to us

Are you TypeSafe? Get your Connect to INFRO button