Access · Last updated 2026-09-21

Jev workspace setup checklist

A Jev workspace is your repo’s wiring: secret injection, a pinned model id, a questions module, and a place to log traces. TypeSafe’s console is where you mint a key; the workspace is yours. Jev (TypeSafe System One) does not ship a workspace product we will invent because a rival IA used that word.

This page is for builders who already heard of Jev and are ready to put a smoke call in CI. jev.pro is an independent unofficial field guide — not affiliated with TypeSafe AI. We do not sell, issue, or proxy API keys. Confirm SDKs and model ids on docs.typesafe.ai. For official versus unofficial reading order, see accessing the unofficial Jev documentation guide.

Steps

1. Confirm access (get access) and store the secret in a vault — never in git. Keys come from console.typesafe.ai after access. Official HTTP sends Authorization: Bearer. Official SDKs read TYPESAFE_API_KEY.
2. Choose first-party HTTP/SDK or a documented gateway. Write the model id for that door in config. Gateway uses a Gateway credential or OIDC and typesafe-ai/jev — not a TypeSafe first-party alias.
3. Add a smoke call: one Noul on a tiny fixture state. See first API call. A 401 means fix the secret; do not retry 401.
4. Put instructions, criteria, and floors in one module so diffs stay reviewable. Dual-stack teams keep one JSON body; smoke Python and TypeScript if both languages ship.
5. Log response.model, usage.input_tokens, and the full probabilities map from day one.
6. Filter state before POST. English is the primary training language; images are not accepted. Unused text still bills as input — vendor list price for Jev 1.13 is currently claimed at $0.042/M input, output free.

Pitfalls

• Waiting on a workspace CLI from this site. Public docs we used start at the Playground, curl, and official SDKs. jev.pro cannot provision a workspace.
• Leaving jev-latest in place after you start tuning floors. Pin jev-1.13.0 (or the current versioned id). Aliases move.
• Shared laptops plus a committed .env. That is how keys leak. Rotate.
• Cloning a rival “workspace” screen as if it were TypeSafe. If TypeSafe adds org features, read their docs — do not invent them here.
• Pointing a sandbox key at production tools. Keep experiments in sandbox access.

What belongs in the repo versus the vault

Repo: questions, criteria, floors, fixture state (synthetic), the pin, and the door name. Vault: the bearer or gateway secret. Runtime logs: resolved model, token usage, probabilities — not the raw key and not more customer PII than you need.

TypeSafe’s models page also claims 250k tok/s and 1200 RPM; treat those as vendor claims that can change. They are not workspace settings you toggle in git.

Onboard the rest of the team from the same module — team onboarding — so reviewers do not edit criteria live in the Playground without a diff.

See also

• Access hub
• Unofficial documentation access
• Sandbox experiments
• Team onboarding
• How to get access

FAQ

Does TypeSafe ship a workspace CLI?

Public docs we used start at the Playground, curl, and official SDKs. Do not wait on an unofficial CLI from this site.

Python and TypeScript in one workspace?

Yes. Keep one JSON body; dual-stack smoke if both languages ship.

Can jev.pro provision the workspace?

No.

What this page does not claim

This page does not claim TypeSafe ships a workspace product or CLI, that we can provision repos, or that jev-latest is a frozen SHA. It does not issue keys. It does not freeze vendor price or RPM.

Disclaimer

This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai.