The Jev HTTP endpoint
TypeSafe documents a single evaluation endpoint for Jev / System One: POST https://api.typesafe.ai/v1/systemone. Chat-completions-style URLs are the wrong tool for typed questions. The other first-party route in the public reference is GET https://api.typesafe.ai/v1/models, which lists aliases your account can send.
This page is for builders wiring TypeSafe Jev (System One) into agents, routing, or evals. jev.pro is an independent unofficial field guide — not affiliated with TypeSafe AI. We do not sell or issue keys. Confirm hosts and paths on docs.typesafe.ai/api.
POST https://api.typesafe.ai/v1/systemone
Authorization: Bearer <API_KEY>
Content-Type: application/json
List names your account can send:
curl https://api.typesafe.ai/v1/models \
-H "Authorization: Bearer $TYPESAFE_API_KEY"
GET /v1/models currently lists aliases. Versioned IDs such as jev-1.13.0 still work on POST even if they are absent from the list. On the snapshot we read, jev-latest and jev-preview resolve to jev-1.13.0 — confirm language on the models page.
What you send and get
The POST body is model, state, and questions. The success body is model, answers, and usage. Choice and Score answers include probabilities plus confidence. Noul answers include noul only. Question ids are not sent to the model. The payload is text only; English is primary.
Documented HTTP failures are 401, 422, 429, and 529. Official SDKs retry 429/529. Do not invent a /health URL; an authenticated GET /v1/models is a cheap ping if you must.
Gateway twin
Vercel documents the same shape under https://ai-gateway.vercel.sh/typesafe/v1/systemone with Gateway auth and model id typesafe-ai/jev. See channels compared. Mixing a TypeSafe bearer key with a Gateway host (or the reverse) is a common 401.
Inventory titles are not extra routes
There is no undocumented second first-party evaluation path on this page. Titles elsewhere on jev.pro that look like /v1/batch, /v1/classify, or a search endpoint are application patterns unless the API reference lists them. Search and re-rank cookbooks still assume you retrieved candidates; Jev judges text you already have.
Production notes
- Pin
jev-1.13.0once thresholds exist. Aliases move when TypeSafe ships. - Vendor-claimed context: 64k tokens for state + all questions; 32k for state + the longest question. Filter unused fields.
- Vendor-claimed list price: $0.042 / million input tokens, output free. 250k tok/s, 1200 RPM, and 70ms–500ms / 40x–200x from the launch post are vendor claims, not SLOs this site measured.
- Do not stream tokens. You receive a complete
answersmap. - Schema-safe ≠ factually correct.
See also
- API hub — first-party routes and pattern pages
- First API call — official quick-start curl
- Models and aliases —
jev-1.13.0versus moving aliases - Evaluate naming — SDK
systemOne, not/v1/evaluate
FAQ
Is /v1/systemone versioned independently of the model?
Yes. Path is /v1/…. Weights are the model field.
Health check URL?
Do not invent one. Use /v1/models as a cheap authenticated ping if you must.
Streaming token endpoint?
Not how System One is documented. You receive a complete answers map.
Can I POST to /v1/search on this host?
Not in the public reference we used. Retrieve first, then evaluate — search pattern.
What this page does not claim
- Official TypeSafe status, logos, or support.
- That jev.pro sells, issues, or proxies API keys.
- Extra first-party HTTP routes beyond
POST /v1/systemoneandGET /v1/modelsunless the API reference lists them. - Invented latency SLOs, search volume, or unofficial prices.
- That a schema-constrained answer is automatically factually correct.
Disclaimer
This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai. Never treat jev.pro as TypeSafe official documentation. We do not sell, issue, or proxy API keys.