Awesome Jev / TypeSafe
The field guide to typed decisions.
Explore the docs, live builds, and independent tests behind Jev. It turns context into a probability, choice, or score; your code decides what happens next.
One call, three typed answers. In TypeSafe's documented support-ticket example, Jev chooses the technical team (0.85 probability), scores frustration at level 1 on a 0–2 rubric, and gives urgency a 1.0 Noul probability. The example response is from
jev-1.13.0; application code still decides when to route or escalate.
| I want to… | Go here |
|---|---|
| Understand the idea in 2 minutes | See where Jev fits, try the policy threshold, then read the introduction and the three primitives |
| Make my first typed call | Copy the runnable example, shape your own question, then explore the official SDKs |
| See it work live | Play Jev Chess, try Typewriter's 16 judgments or Jevtown's simulated audience; browse more applications |
| Test the claims | Read what independent tests found, inspect JevBench's cross-model results, then browse independent evaluations and TypeSafe's own evals |
Download the JSON directory · Use with a coding agent · Suggest a resource · Join the builder community
Independent community project. This repository is not affiliated with or endorsed by TypeSafe AI. Community entries are labeled by section; inclusion is not a claim that TypeSafe has reviewed or approved them.
Last updated: 2026-09-21. Links and project descriptions change; please report a stale entry.
Contents
Start here
- Introduction — What Jev is, how System One models differ from text-generation models, and the Choice, Score, and Noul primitives.
- Quick start — The shortest path from an API key to a typed decision in Python or JavaScript.
- How to build with TypeSafe — Design guidance for decomposing a workflow into narrow judgments while keeping policy and side effects in code.
- TypeSafe Console — Create keys and inspect live Jev requests.
Choose the right tool
In a support workflow, separate the work before choosing a model. This is a practical design rule based on the TypeSafe introduction linked above, not a performance claim:
| What the step needs | Use | Example |
|---|---|---|
| Apply an explicit rule to known fields | Code | Check an account flag or enforce a routing threshold. |
| Judge messy context with a bounded answer | Jev | Choose billing, technical, or other for a ticket, with probabilities. |
| Produce prose or work through an open-ended task | Text LLM | Draft the reply after the route is chosen. |
Code still validates the answer and owns the action. Measure Jev's error and abstention rates on your own cases before automating a consequential step.
One state can answer several focused questions in the same request. Pick the answer shape your code can use directly:
| Question shape | Use it for | What comes back |
|---|---|---|
| Noul | A clear yes/no claim, such as “Does this message request a refund?” | A number from 0 to 1: the probability of yes. |
| Choice | Selecting from named options, such as billing, technical, or sales. | The selected option, a probability for every option, and confidence. |
| Score | An ordered rubric, such as calm, concerned, or angry. | A position on your rubric, probabilities over its levels, and confidence. |
Ask independent questions together. Set thresholds, fallback behavior, and side effects in application code.
Make your first decision
Install the official JavaScript SDK with npm install @typesafe-ai/sdk (Node.js 20+), set TYPESAFE_API_KEY in your environment, save this as first-decision.mjs, then run node first-decision.mjs:
import { choice, noul, TypeSafeClient } from '@typesafe-ai/sdk';
const { answers } = await new TypeSafeClient().systemOne({
state: { ticket: 'I was charged twice. Please refund the extra payment.' },
questions: {
team: choice('Which team should handle this ticket?', {
billing: 'Payments and refunds',
technical: 'Bugs and integrations',
other: 'None of the above',
}),
refund: noul('Does the customer explicitly request a refund?'),
},
});
const team = answers.team.choice;
const probability = answers.team.probabilities[team];
const action = team !== 'other' && probability >= 0.9
? `route to ${team}` : 'send to review';
console.log({ team, probability, refundProbability: answers.refund.noul, action });
Jev returns the typed answers; the 0.9 routing rule is ordinary application code. It is an illustrative threshold, not a measured or recommended operating point. The ticket text is sent to TypeSafe's API; use a synthetic ticket for this first call.
Shape a typed question
Start with one state and a question whose answer your code can use. This synthetic support report can be asked as a Choice, Noul, or Score. On the live site, edit the fields and copy a JavaScript SDK call. The designer runs in your browser without making a model request; running the copied code later sends the state to TypeSafe.
| Design input | Synthetic example |
|---|---|
| State text | The PDF upload fails with a 500 error. I need it before today's deadline. |
| Choice question | Which team should handle this report? |
| Choice options | technical=Failures and integrations; support=Account and usage help; other=Neither team |
| Noul question | Does the message explicitly mention a deadline? |
| Score question | How much does the reported issue block the user's work? |
| Score levels | Cosmetic; Workaround available; Blocks the task |
Keep the state short, describe the options so they do not overlap, and include a no-match option when the task allows it. Choose thresholds and actions only after measuring your own labelled cases.
Try a policy threshold
The documented support-ticket example above selects technical with probability 0.85. In this illustrative policy, a ticket routes automatically only