API · Last updated 2026-09-21

Jev API testing

The documented way to see Jev is the Playground and then official curl / SDK samples. jev.pro cannot run your tests against TypeSafe and will not accept a key.

This page is for builders who searched an extra “endpoint” name while wiring Jev / System One. jev.pro is an independent unofficial field guide — not affiliated with TypeSafe AI. We do not sell or issue keys. Confirm first-party routes on docs.typesafe.ai/api.

The public API reference documents one evaluation route — POST https://api.typesafe.ai/v1/systemone — plus GET /v1/models. This title is not a separate first-party URL in that reference. Treat the capability as something you implement around the official call. There is no documented /v1/test sandbox route.

Steps

1. Settle wording in the Playground.
2. Freeze a JSON fixture (state + questions + pin).
3. Assert policy actions in CI, not only argmax labels.
4. Contract-test 401/422 with bad keys/bodies against your wrapper mocks if you lack a shared sandbox key.
5. Never commit live secrets.

What to assert

A fixture is model, state, and questions. Pin jev-1.13.0 so jev-latest moving does not fail the suite. On the snapshot we read, jev-latest and jev-preview resolve to that id — confirm. Record expected actions (hold, route to billing, refuse publish) derived from:

• Noul bands (Noul has no confidence)
• Choice/Score probabilities and confidence
• Your floors — illustrations in docs are not defaults

Do not average Nouls across questions into a single “accuracy.” Questions are independent. Schema-safe ≠ factually correct; a golden choice can still be a bad policy if the state was incomplete.

Python tests use typesafe-sdk (TypeSafeClient / AsyncTypeSafeClient, TYPESAFE_API_KEY, Python ≥ 3.10). JS tests use @typesafe-ai/sdk on Node 20+ (systemOne). Official SDKs retry 429/529; CI should not treat a single 429 as a product regression if you forgot backoff. Documented errors only: 401, 422, 429, 529.

Live calls spend input tokens (vendor claim: $0.042 / million input, output free). Prefer mocks for 401/422; use a shared paid key only for a small nightly pin-replay.

Pitfalls

• Golden-file tests that fail when jev-latest moves — pin the id.
• Averaging Nouls across questions as a single “accuracy.”
• Pointing CI at jev.pro. We do not proxy evaluations.
• Committing TYPESAFE_API_KEY in GitHub Actions logs.

See also

• API hub — the POST your fixtures hit
• Playground — wording before you freeze JSON
• First API call — official curl
• Testing guide — offline labels

FAQ

Title used to say “Api Testing” — official name?

No. This is a testing pattern page.

Can I point CI at jev.pro?

We do not proxy evaluations.

Offline eval?

Labeled JSONL + frozen questions — offline evaluation.

Should CI retry 422?

No. Fix the body. Retry 429/529 only.

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/systemone and GET /v1/models unless the API reference lists them.
• Invented latency SLOs, search volume, or unofficial prices. List price and rate limits are vendor claims — confirm on models.
• 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.