Why use this SDK instead of a plain sttp POST with upickle?

Raw HTTP leaves validation, probability bounds, and answer typing to you. This client validates before sending, models every probability as a bounded type, and lets you look answers up with the question value so the type follows.

Will a 3.3-built artifact work if my project is on a newer Scala 3?

Yes by design. It is published against Scala 3.3 LTS because TASTy is forward-compatible but not backward, so an artifact built on 3.3 stays readable from 3.3 onward instead of shutting out LTS projects.

Who decides the yes threshold for a noul answer?

Deliberately. A 0.51 and a 0.99 are very different answers, so there is no Boolean on the answer itself. You call isYes with the default or your own threshold, and choiceIfConfident gates routing on peaked distributions.

Why does passing a Seq of plain strings fail to compile?

Implicit conversions apply to one argument at a time and do not reach inside collections. Map first with labels.map(Content.text) or rows.map(ChoiceOption.fromPair), or splat a converted sequence into the varargs factory.

How do I add it and what does it require?

Add io.github.ticofab %% scala-jev-sdk to libraryDependencies. It targets Scala 3.3+ (LTS) and JDK 17+, and its only dependencies are sttp-client4-core and upickle. It is published against 3.3 on purpose so TASTy stays readable from 3.3 through 3.9 and beyond.

How do I look up an answer?

You look an answer up with the question value itself and the type follows: answers.get(isUrgent) returns an Option[NoulAnswer], not an Any out of a Map. Choice answers come back as your own type when you build the question with choiceOf[Team], so a routing decision lands as a Team rather than a string to re-parse.

What is the confidence-gated routing pattern here?

choiceIfConfident(0.9) returns None when the distribution is too close to call, so you can act automatically when it is peaked and fall through to a human or a slower model when it is not. There is deliberately no Boolean on a noul answer; isYes() takes an explicit threshold.

CommunityJev Repos & SDKs

Scala Jev SDK

Scala 3 client for Jev that sends one state with typed noul, choice, and score questions over your own sttp backend. Answers come back typed with probabilities, usable from Future, blocking Identity, cats-effect, ZIO, Monix, and Pekko.

See highlights

Open github.com

Stars: 11
Language: Scala
License: APACHE-2.0
Published by: Community
Author: ticofab
Last verified: Oct 2026

ticofab/scala-jev-sdk

User repository on GitHub

View on GitHub

Scala SDK for Jev. No effect system bundled.

Forks: 1
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

A Scala client for Jev, TypeSafe AI's System One model — the model that answers typed questions about a piece of state instead of generating text. No effect system. You hand it an sttp backend and the client speaks whatever that backend speaks — Future, blocking Identity with no wrapper at all, or cats-effect, ZIO, Monix and Pekko if that is what your codebase already runs on. Scala 3.3+ (LTS), JDK 17+. The only dependencies are sttp-client4-core and upickle.

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Scala

Highlights

Needs Scala 3.3 LTS and JDK 17, with only sttp-client4-core and upickle as dependencies.
You supply the sttp backend, so the effect type follows what your codebase already runs on.
Answers are looked up with the question value itself, with choices returnable as your own type.
Probabilities are a bounded type, with confidence-gated routing and JSON-capable Content state.

Watch out

You supply and close the sttp backend; client construction fails on a blank key, empty model name, or non-positive timeout. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

go-jev

Go SDK plus a pipeline-friendly CLI for typed decisions, covering single questions, batched question maps, retries, and line-oriented filtering over stdin.