Integration guide · independent

Calling Jev through OpenRouter

OpenRouter is a hosted gateway in front of many model providers, exposed through one API key. TypeSafe is one of those providers, so you can call Jev with an OpenRouter key and no separate TypeSafe account or waitlist.

The important catch: Jev is not a chat model. It is served through a dedicated Decisions route, and ordinary chat-completions SDKs will not work with it. This page lists the model IDs, the request path, and the figures that differ depending on when they were published.

Model IDs and aliases

IDListedWhat it is
typesafe/jev-1.1318 September 2026The original dated hosted build
typesafe/jev-1.1414 October 2026A later hosted build; listed with a 262,144-token context
~typesafe/jev-latestRollingAn alias that always points to the newest Jev build
typesafe/jev-router25 September 2026A routing entry, listed with a 1M-token context — not a separate checkpoint

Pin a dated ID for reproducible behaviour. The alias is convenient but will change underneath you when a new build ships.

The request path

OpenRouter documents two shapes: its own Decisions route, POST /api/alpha/decisions, and a TypeSafe-compatible route, POST /api/v1/systemone. The body carries a model, a state, and a questions map using noul, choice, and score. The response carries the typed answers plus gateway fields such as a generation ID, the provider, and usage with cost.

  • →Do not send these requests to /api/v1/chat/completions — the output type is decisions, not text.
  • →The models API defaults to text models; filter on output_modalities=decisions to see Jev and the other decision entries.
  • →OpenRouter lists other decision models on the same output type, including Clef and Clef-flash. Their schemas are not guaranteed to match.
openrouter.sh
curl -s https://openrouter.ai/api/v1/systemone \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "~typesafe/jev-latest",
    "state": "I was charged twice for the same invoice.",
    "questions": {
      "billing": { "type": "noul", "instructions": "Is this about billing?" }
    }
  }'

Context and pricing: check the date

The gateway listing for Jev 1.14 shows a 262,144-token context and text-and-image input at $0.042 per million input tokens with free output. Earlier materials quoted elsewhere on this site — including the System One API limits referenced in the Jev vs Clef comparison — describe 32k of state within a 64k request budget.

These are different builds and different surfaces, not a contradiction to resolve by picking one. Verify the current limits in the official docs and on the model page before you size a request. The same discipline applies to latency: an independent race through OpenRouter found Jev faster on every task it tested, while Cloudflare's own harness reported the opposite ordering.

Frequently asked questions

Do I need a TypeSafe API key to use Jev on OpenRouter?

+

No. An OpenRouter API key is sufficient for the gateway route. OpenRouter forwards the request to TypeSafe, which is listed as the only provider for these model IDs. Keys for the direct hosted endpoint are issued by TypeSafe separately.

Why is Jev missing from /api/v1/models?

+

That list covers text-producing models by default. Jev's output type is decisions. Add the filter output_modalities=decisions to the models API query to see Jev and the other decision entries.

Can I use the OpenAI SDK unchanged?

+

Not for the decision route. Chat-completions SDKs expect a generated text response. Use the OpenRouter Decisions endpoint or the TypeSafe-compatible /api/v1/systemone route with a client that understands state plus typed questions.

Which model ID should I pin?

+

Pin typesafe/jev-1.13 or typesafe/jev-1.14 when behaviour must be reproducible across deployments. Use ~typesafe/jev-latest when you want the newest build automatically, and accept that its behaviour can change.

Is OpenRouter cheaper than calling TypeSafe directly?

+

The listed rate is the same $0.042 per million input tokens with free output. A gateway changes authentication, billing, and observability — not the underlying list price. Check the model page for any gateway-specific charges before estimating volume.

Keep reading

Sources

Reviewed 2026-11-13. Vendor- and project-reported figures are labelled as such; none were independently reproduced by this site.

Integration guide · reviewed 2026-11-13 · first published 2026-11-13