OpenJev (now SemIf) is not Jev
OpenJev is an independent project (publicly discussed as later renamed SemIf) that explores a Jev-shaped interface on open weights: state + runtime criteria in, option probabilities out, without asking the local model to write JSON.
This page is for builders crossing from curious to production-minded Jev use. jev.pro is an independent unofficial field guide — not affiliated with TypeSafe AI. Cover intents; do not clone rival IA. Confirm facts on docs.typesafe.ai.
Open-cluster pages are independent field-guide notes. Replicas and third-party interfaces mentioned anywhere on jev.pro are not Jev and not endorsed.
TypeSafe’s documentation index does not list OpenJev as a product, SDK, or supported runtime. Treat it as third-party research.
Hard lines
- Not affiliated with TypeSafe AI.
- Not affiliated with jev.pro.
- Does not ship TypeSafe’s closed model or RLCD training.
- MIT-licensed code and upstream model licenses are the author’s concern, not ours.
We will not paste third-party latency or agreement tables here. Those numbers are not TypeSafe docs, and we have not reproduced them.
Why the page exists
People search “open Jev.” The accurate sentence is: you can study an open interface clone; you cannot download Jev. Production TypeSafe access remains the console API or documented gateways.
If you evaluate SemIf, label results as “open-model interface clone,” never as “Jev scores.”
What a SemIf eval is allowed to claim
You may report agreement between an open-weight interface and hosted Jev on your labels. You may not:
- Call the numbers “Jev scores,” “Jev calibration,” or “System One.”
- Import TypeSafe workflow-eval multiples (they have cited high-end speed/cost figures) onto a SemIf chart.
- Reuse floors fitted on
jev-1.13.0as if calibration transferred. - Treat MIT-licensed code as a TypeSafe SDK (
TYPESAFE_API_KEYis not a SemIf concept).
Official: hosted Jev is accessed via the console API (POST https://api.typesafe.ai/v1/systemone) or documented gateways. We will not paste third-party latency or agreement tables here. Those numbers are not TypeSafe docs, and we have not reproduced them. See public evals for how this site treats TypeSafe’s own published evals — still no jev.pro leaderboard.
Noul 0.5 on an open model is not official “yes/no equally likely.” That calibration sentence is about Jev. Different weights, different meaning.
When not to use this page’s advice
Do not use SemIf as a sandbox for production Jev thresholds if you care about Jev. Do not cite this page as TypeSafe’s position (their index does not list the project). Do not ask jev.pro for keys to “compare the two” — we issue none. Cover the search intent; do not photocopy a rival project’s README.
See also
FAQ
Can I use SemIf as a sandbox for Jev thresholds?
Not if you care about Jev. Different weights, different calibration.
Does TypeSafe endorse it?
Their index does not list it. We do not either.
Cover ≠ clone?
This page exists so we cover the intent without photocopying a rival project.
Can I put SemIf and Jev in one Grafana board?
Only with a processor label so nobody reads open-model rows as Jev.
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.