User repository on GitHub
.NET SDK for the TypeSafe AI platform
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
Jev directory
CommunityJev Repos & SDKs
TypeSafeAI for .NET
.NET SDK for the System One API with typed question sets, HttpClientFactory and dependency-injection wiring, plus Microsoft.Extensions.AI guardrail, routing, tool, and evaluator adapters.
- Stars
- 6
- Language
- C#
- License
- APACHE-2.0
- Reading time
- 3 min
- Published by
- Community
- Author
- Hawxy
- Last verified
- Sep 2026
About this project
A .NET SDK for the TypeSafe AI System One API. Ask small, typed judgments about text or structured state and get calibrated probabilities back that your code can act on: Noul: is this true? Returns the probability of yes. Choice: which of these labels? Returns the most likely label plus a probability per label and a confidence. Score: where on this scale? Returns the expected level plus a probability per level and a confidence.
Quoted from the project's README. Read the full README.
Highlights
- A QuestionSet returns typed handles: Noul answers give a probability, Choice an enum label, Score a level with its legend.
- Retries and per-attempt timeouts run on a Polly resilience pipeline with an injectable TimeProvider for tests.
- Two packages: the core TypeSafeAI client and TypeSafeAI.Extensions.AI for guardrail, routing, tool, and evaluator middleware.
- Targets net8.0 and net10.0, and the package is trim and NativeAOT safe.
Quickstart
csharp
using TypeSafeAI;
using var client = new TypeSafeClient("YOUR_API_KEY");
var q = new QuestionSet();
var urgent = q.Noul("Does this convey urgency?");
var category = q.Choice<TicketCategory>("What is this ticket about?");
var result = await client.SystemOneAsync("Payouts have been failing for 3 days.", q);
double p = result.Get(urgent).Probability;Watch out
Apache-2.0. A community SDK, not affiliated with TypeSafe; needs an API key, and live tests skip without TYPESAFE_API_KEY.
Questions about this entry
Short answers drawn from the linked project or page.
A community SDK, and the README states it is not affiliated with or endorsed by TypeSafe. It ships as two NuGet packages: TypeSafeAI for the core client with typed question sets, retries, HttpClientFactory and DI wiring, and TypeSafeAI.Extensions.AI for the Microsoft.Extensions.AI middleware, targeting net8.0 and net10.0 under Apache-2.0.
One per state. SystemOneManyAsync runs a request per state in parallel, bounded by maxConcurrency, and returns results in input order, so a thousand rows is a thousand separate requests. There is no packing of many states into one call here, only parallel execution of individual ones.
The core package is described as trim and AOT safe, and the state helpers have a reflection catch. TypeSafeContent.FromObject takes your own types through reflection, and the README points you at the JsonTypeInfo overload for AOT, or at FromNode with a JsonObject to stay reflection-free.
Bodies are only written at Trace level, which is off by default. Every call emits an Activity tagged with model, request id, status, and attempt, and you add it to OpenTelemetry with AddSource(TypeSafeDiagnostics.ActivitySourceName); log events flow through ILogger<TypeSafeClient> when the client comes from DI.
Call AddTypeSafeClient with a configuration section or a delegate. It registers ITypeSafeClient and TypeSafeClient as typed HTTP clients and returns the IHttpClientBuilder, so you can still add your own handlers. Configuration is applied first and your delegate second, so the delegate wins on conflict.
Four things around a chat client that TypeSafe itself is not, since it returns judgments rather than text. You get guardrail middleware that blocks or reviews before the model sees the message and attaches a GuardrailOutcome on passes, a router that returns null to fall through to the inner client on low confidence, an AIFunction tool bridge the model can call with a state string, and an IEvaluator adapter for Microsoft.Extensions.AI.Evaluation.
RequestOptions carries Model, Timeout, MaxRetries, ExtraHeaders and ExtraBody, so MaxRetries = 0 is how you fail fast on a latency-sensitive path, and ExtraBody forwards an experimental flag forward. Defaults are a 10-second per-attempt timeout and 2 retries on 408, 429, 5xx, connection and timeout failures, all running on a Polly pipeline driven by an injectable TimeProvider so tests can use a fake clock.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Facebook: Intake Router keeps diagnosis out of the answer space at the front desk.
FacebookClinicnotes
X: OpenRouter Decision Model Rankings
X: herdr + jev, an agent router
X: A group chat bot with decision room routing
More Jev resources like this
A Juspay TypeScript SDK that treats Jev as a third inference type beside generate and stream: providers declare which kinds they serve, and tryDecide returns typed boolean, choice, or score answers, null when no decision provider is set.
Jev Repos & SDKs#community#typescript#sdk
Community
juspay/neurolinkTypeScript
juspay/neurolinkTypeScriptA filesystem-first framework for durable AI agents whose auto model router defaults to Jev through Vercel AI Gateway and whose evaluate helper asks typed Choice, Score, and Boolean questions inside tools.
Jev Repos & SDKs#community#typescript#agents
Community
vercel/eveTypeScript
vercel/eveTypeScriptAn unofficial .NET client for TypeSafe AI's System One API and the Jev model, bringing official JS and Python SDK behaviour (retries, errors, key handling) to netstandard2.0 and net10.0 with typed answers, dependency injection and OpenTelemetry tracing.
Jev Repos & SDKs#community#github#csharp
Community
sanamhub/typesafe-dotnetC#
sanamhub/typesafe-dotnetC#Community .NET SDK for the System One API: typed noul, choice, and score questions with structured, confidence-scored answers.
Jev Repos & SDKs#community#sdk#csharp
Community
saibimajdi/typesafeai-dotnet-sdkC#
saibimajdi/typesafeai-dotnet-sdkC#Unofficial LlamaIndex reranker and router built on the official Python SDK: JevRerank scores retrieved passages and JevSingleSelector chooses which tool handles a query.
Jev Repos & SDKs#community#python#sdk
Community
WiktorB2004/llama-index-jevPython
WiktorB2004/llama-index-jevPythonDependency-free Go 1.24+ client for Jev Noul, Choice, and Score questions: ChoiceOf[T] and ScoreOf[L] return the caller's own types and reject labels the question never offered, with validation before sending, retries that honor Retry-After, OpenRouter support, and eleven tested examples.
Jev Repos & SDKs#community#go#sdk
Community
HomayoonAlimohammadi/jev-sdk-goGo
HomayoonAlimohammadi/jev-sdk-goGoFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.