How close is this community client to the official SDKs?

When you want the documented Python and JS behavior in Go with only the standard library. It supports Ask, AskRaw, ListModels, per-call model and headers, Extra body fields, and gateway base URLs, but strict response validation can reject shapes the live API changes.

Should I chain one-question calls or batch several questions together?

Ask several at once against the same state rather than chaining calls. Each question is evaluated independently, and Nouls, Choices, and Scores accessors group the answers by kind for direct branching.

What does the default retry policy actually retry?

Two retries from 500 ms with a 5 s cap and 25% jitter on 408, 429, 5xx, connection errors, and timeouts, honoring Retry-After within a 30 s budget. Start from DefaultRetryPolicy, tune MaxRetries or Statuses, or pass an empty policy to disable retries.

Can I run it through OpenRouter or Vercel AI Gateway?

Point WithBaseURL or TYPESAFE_BASE_URL at the gateway and it appends /v1/systemone and /v1/models. Use the gateway key and model ID, keep HTTP/2 enabled on custom transports, and pass gateway headers via WithHeaders or per-request Headers.

What Go version and dependencies does it need?

Go 1.26 or newer, and it uses only the standard library, so there are no third-party modules to pull in. It is a community client that mirrors the feature surface of the official Python and JavaScript SDKs.

What license is it under?

Apache-2.0, and it is a community client rather than an official TypeSafe AI SDK. Response validation is deliberately strict so it fails closed instead of returning partial data.

What API limits does the SDK document?

Text only (a string, JSON object, or array, with no images or audio), about 64k tokens per request with state plus the longest single question within about 32k tokens, and input tokens are billed while output is free.

CommunityJev Repos & SDKs

Jev Go SDK

Dependency-free Go client mirroring the official Python and JS SDKs: send one state with typed Noul, Choice, Score, and Raw questions and get calibrated answers in milliseconds.

See highlights

Open github.com

Stars: 3
Language: Go
License: APACHE-2.0
Published by: Community
Author: ajayk
Last verified: Oct 2026

ajayk/jev-go-sdk

User repository on GitHub

View on GitHub

Dependency-free Go client for TypeSafe AI's System One API and the Jev model

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

A dependency-free Go client for TypeSafe AI's API and its flagship System One model, Jev. It mirrors the feature surface of the official Python and JavaScript SDKs. Jev is not a chat model. You send it a state (text or JSON) and a set of typed questions; it returns one calibrated, probability-bearing answer per question in a few hundred milliseconds. No generated text, no tool loop, no prompt parsing: the answers are values your code can branch on directly.

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

Tagscommunity github

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Go

Highlights

Standard library only on Go 1.26+; state accepts strings or JSON-marshalable values via Content.
Choice supports up to 255 options; Score rubrics return Score, Legend, probabilities, and Level.
Fail-closed validation rejects missing, extra, mismatched, or non-normalized answers.
API key never logged; logger records method, path, status, and duration, masking credential headers.

Watch out

Community client, not official; validation fails closed on shape mismatches, so open an issue with a redacted response. 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

jev

Go Unix CLI that pipes state on stdin into one Jev question — noul, choice, or score — across TypeSafe, Cloudflare, and Vercel providers, printing machine-readable JSON and signaling success by exit code.