Jev AI Hub
Start Learning

Developers

Jev on OpenRouter and Vercel AI Gateway

Call TypeSafe Jev through OpenRouter or Vercel AI Gateway when you do not have a TypeSafe key yet, or already bill through those platforms.

Published
Sep 20, 2026
Updated
Sep 20, 2026
Last verified
Sep 20, 2026

Quick answer

Jev is still TypeSafe's model. OpenRouter lists ~typesafe/jev-latest and documents POST /api/alpha/decisions plus a TypeSafe-compatible /api/v1/systemone. Vercel AI Gateway uses model ID typesafe-ai/jev with the AI SDK experimental_evaluate helper. Prefer the official TypeSafe API when you have a key.

OpenRouter Jev and Vercel Jev are access paths, not a second model. You still send state + typed questions. You still get Choice / Score / Noul. The key and base URL change.

Use this page if you searched jev openrouter, openrouter jev, vercel jev, or vercel ai gateway.

Official path first

If you already have a TypeSafe key, call https://api.typesafe.ai/v1/systemone as in the API guide. Gateways are useful when:

  • you are on the TypeSafe waitlist and already have an OpenRouter or Vercel account
  • your team bills all models through one gateway
  • you want gateway logging / budgets

OpenRouter

OpenRouter's TypeSafe SDK guide (accessed 2026-09-20):

  • Point the official SDK at OpenRouter. The SDK appends /v1/systemone, so the URL becomes https://openrouter.ai/api/v1/systemone.
  • Bare IDs map into OpenRouter's namespace: jev-latest~typesafe/jev-latest, jev-1.13typesafe/jev-1.13.
  • OpenRouter also documents POST https://openrouter.ai/api/alpha/decisions with "model": "~typesafe/jev-latest".
curl https://openrouter.ai/api/v1/systemone \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "jev-latest",
    "state": "I was charged twice for my subscription.",
    "questions": {
      "refund": {
        "type": "noul",
        "instructions": "Is the customer asking for money back?"
      }
    }
  }'

OpenRouter's published response extras (id, provider, usage.cost) are gateway fields. Official TypeSafe responses are model, answers, usage.

Vercel AI Gateway

Vercel documents Jev as an evaluation model:

Last verified 2026-09-20
Model IDtypesafe-ai/jev
HelperAI SDK experimental_evaluate
Pricing listed$0.042 / 1M input tokens, no output charge
import { experimental_evaluate as evaluate } from "ai";

const result = await evaluate({
  model: "typesafe-ai/jev",
  state: "The support agent issued a full refund to the customer.",
  questions: {
    refunded: {
      type: "boolean",
      instructions: "Was a refund issued?",
    },
  },
});

Vercel's guide uses type: "boolean" in the AI SDK wrapper. The TypeSafe HTTP primitive remains Noul. Check which client you are in before you copy criteria fields.

Local Vercel OIDC tokens expire; Vercel's own guide says to re-run vercel env pull after a 401.

When to use a gateway

You already operate OpenRouter or Vercel and want Jev on the same bill.

When not to

You are debugging question wording. Use TypeSafe's official Playground / API so gateway mapping is not another variable.

Common mistakes

  • Sending Jev a chat messages array because “OpenRouter is OpenAI-compatible”.
  • Comparing gateway token counts to TypeSafe's usage without reading each vendor's definition.
  • Assuming waitlist access is solved forever by a gateway. Availability and terms are the gateway's, then TypeSafe's.

Next: How to use Jev or Getting Started.

FAQ

Can I use Jev on OpenRouter?

Yes. OpenRouter documents a Decisions API and a TypeSafe-compatible /api/v1/systemone. Model IDs include ~typesafe/jev-latest and typesafe/jev-1.13.

What is the Vercel Jev model ID?

Vercel's AI Gateway catalog lists typesafe-ai/jev. Calls go through experimental_evaluate, not a chat completion.

Sources

  1. TypeSafe SDK IntegrationOpenRouter · accessed 2026-09-20 · documentation
  2. Jev LatestOpenRouter · accessed 2026-09-20 · documentation
  3. How to classify, route, and score with Jev and AI SDKVercel · accessed 2026-09-20 · documentation
  4. Jev on AI GatewayVercel · accessed 2026-09-20 · documentation