Question 1

Is the example response on the front page a real API call?

Accepted Answer

No, it reproduces TypeSafe's published quickstart response for the documented model version: Choice technical at 0.85 selected probability, Score 1 on a 0-2 frustration rubric, and Noul 1.0 for urgency. The page labels it as a documentation example so you can see the response shape before spending a key.

Question 2

Is this another list that just links to blog posts?

Accepted Answer

It links mostly to runnable things: a downloadable JSON directory, an installable agent skill, JavaScript and Python starters that need only a TypeSafe key, and curated project entries that state what decision Jev actually makes. Listings carry their own limitations and source links rather than endorsements.

Question 3

Should I trust the probabilities, or is it just a nice front end?

Accepted Answer

It tells you to measure error and abstention rates on your own cases before automating a consequential step, and warns that schema-valid output is not the same as a correct decision. Its routing thresholds are labeled illustrative rather than measured operating points, and it points at independent test results instead of its own.

Question 4

Does TypeSafe review the projects listed here?

Accepted Answer

No relationship is claimed. It states it is an independent community project, not affiliated with or endorsed by TypeSafe AI, and that community entries are labeled by section where inclusion is not a claim of review or approval. The list itself is MIT-licensed while each linked project keeps its own license.

Question 5

What do I get on the first visit?

Accepted Answer

A live directory to browse, three featured builds you can open directly, a recently curated block, and a what-I-want-to-do table that routes you in one step: understand the idea in two minutes, make your first typed call, see it work live, or test the claims. The featured previews are Jev Chess, TypeSafe Typewriter and Jevtown, and the guidance is to read a listing before opening a preview.

Question 6

How do I shape a typed question and set a threshold?

Accepted Answer

One state can answer several focused questions in the same request, so ask independent questions together. Noul returns 0 to 1 as the probability of yes; Choice returns the selected option plus a probability for every option and confidence; Score returns a position on your rubric plus probabilities over its levels and confidence. Thresholds, fallback behaviour and side effects stay in application code.

Question 7

How do I add my own project to it?

Accepted Answer

There is a CONTRIBUTING guide with a build-a-listing walkthrough, plus a report-a-stale-entry template for links that have rotted and a releases feed for updates. Listings are grouped by section and carry their own limitations and source links, and the README frames the recently curated block as places to explore rather than a ranking or endorsement.

Awesome Jev / TypeSafe

About this project

Highlights

Quickstart

Questions about this entry

From the Jev guides

From the Jev AI model community