Skip to content

TypeSafe.Jev

C# client for Jev, TypeSafe AI's System One model. Typed Noul, Choice and Score questions go out with one state in a single round trip and come back with a value, a probability distribution and a confidence number. Choice questions can be built from an enum, so the compiler finds the switch branch you forgot.

Stars
1
Language
C#
License
MIT
Published by
Community
Author
RavenValentin
Last verified
Oct 2026

About this project

Ask an AI a typed question. Get an answer your switch statement can use. Jev is TypeSafe AI's System One model. You hand it some state and a set of small typed questions, and it answers each one with a value, a probability distribution and a confidence score. No prompt engineering, no "please respond only with JSON", no stripping code fences off the reply. Your enum is the question. Declare the options once, get that enum back, and let the compiler find the switch branch you forgot when you add a member.

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

Highlights

  • Three question types: Noul for a 0-to-1 yes/no probability, Choice over named options, Score across 2 to 10 ordered levels.
  • Question.Choice<TEnum> uses your enum as the option list; [Description] becomes the rubric and As<TEnum>() throws on an unknown name.
  • One dependency, Microsoft.Extensions.Logging.Abstractions; CI reads the packed .nuspec and fails on a second, and runs a native binary each build.
  • Retries on 408, 429 and 5xx with jittered backoff, Retry-After, per-attempt timeouts and a whole-call budget, with ILogger, ActivitySource and Meter.

Watch out

Everything goes through the hosted TypeSafe System One API, so anything beyond the offline tests needs a TYPESAFE_API_KEY. The enum convenience is only as good as the wire mapping behind it: member names travel as snake_case, so renaming TechnicalIssue changes what the model is shown, and an option with no [Description] is sent with an explicit null. The README asserts trim and Native AOT safety, the single-dependency rule and offline testability rather than demonstrating numbers, and says nothing about whether the probabilities are calibrated. Raw JsonObject questions are the escape hatch and only their structure is checked, so a schema mistake comes back as an API error rather than a compile error. 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

45GitHub stars
Go SDK plus a pipeline-friendly CLI for typed decisions, covering single questions, batched question maps, retries, and line-oriented filtering over stdin.
Jev Repos & SDKs#community#jevbest#catalog-import
3GitHub stars
.NET port of the JavaScript SDK that batches Choice, Score, and Noul questions in one call, resolved by explicit options, environment, then defaults.
Jev Repos & SDKs#community#awesomejev#catalog-import
5GitHub stars
Zero-dependency Node bridge that exposes Jev through OpenAI-style chat and responses endpoints with streaming, plus a CLI, a code-audit tool, and an auto-fix loop.
Jev Repos & SDKs#community#awesomejev#catalog-import
4GitHub stars
.NET client for System One where you declare choice, score, and yes-or-no questions in C# and get typed answers with probabilities and confidence back, installable as one NuGet package.
Jev Repos & SDKs#community#awesomejev#catalog-import
3GitHub stars
Unofficial .NET 10 HTTP client for TypeSafe Jev and the Jev-compatible local Laya server that posts the shared /v1/systemone body and returns typed Choice, Score and Noul answers, with no runtime package dependencies.
Jev Repos & SDKs#community#jevbest#catalog-import
3GitHub stars
A community-maintained .NET 10 client for the official TypeSafe Jev API, covering the Choice, Score and Noul primitives with Microsoft.Extensions.AI composition, model discovery and ten runnable samples. Released as tentative 0.6.0 with live compatibility unverified.
Jev Repos & SDKs#community#awesomejev#catalog-import

From the Jev guides

Original write-ups on what this entry is for.

Closed weights have not stopped the ecosystem from rebuilding the shape in the open. API-compatible servers, retrained heads on small models, and inference-time tricks that skip generation entirely.
Jev Repos & SDKs#open-models#research#local
Read article
Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai