What is jevtok-ts and what does it do?

An offline token counter and request accountant for Jev, written in TypeScript for Node.js and Next.js. It is a port of LabGuy94/jevtok that matches its reconstructed encoder for jev-1.13-20260917, with no runtime dependencies.

Why can't I just use a standard tokenizer?

Because Jev does not use standard o200k_base tokenization: the README notes that 'evidence' is three tokens and '2024' is four under Jev's encoder, so a generic tokenizer would misestimate your input budget.

How do I install and use it?

npm install jevtok-ts on Node.js 22+, then getEncoding('jev-1.13'), enc.count('evidence') for a quick check, and estimateRequestTokens(state, questions) to get an estimated usage.input_tokens for a request. ESM and CommonJS both work.

Does it require an API key or network access?

No. Everything runs offline: no Python, network calls, API keys, native extensions or runtime dependencies. The vocabulary and lookup tables are shipped inside the package, which is the point of a local accounting tool.

What encoding and model version does it model?

getEncoding() defaults to the jev-1.13 encoding, reporting model version jev-1.13-20260917 and a vocabulary of 200,019 entries. listEncodingNames() returns ['jev-1.13'] — one encoding, matching the upstream port's target.

What are the request-guard helpers for?

estimateRequestTokens throws on budget violations, analyzeRequestTokens reports counts and all violations without throwing, and createJevGuard builds a reusable check()/assert() over a fixed {state, questions} payload.

What are the limits of jevtok-ts?

It supports Node.js 22+ only — browsers and edge runtimes are not supported — and it is tied to the jev-1.13 encoding, so newer model versions would need a revised encoder. The captured README was truncated before the end of its API table.

Jev directory

CommunityJev Repos & SDKs

Jevtok TS

Offline Jev token counting and request accounting for Node.js and Next.js: an MIT TypeScript port of jevtok that matches the reconstructed encoder for jev-1.13-20260917 and ships its vocabulary inside the package — no keys, network calls or dependencies.

See highlights Open github.com

Stars: 3
Language: TypeScript
License: MIT
Reading time: 2 min
Published by: Community
Author: aarzhaev
Last verified: Oct 2026

aarzhaev/jevtok-ts

User repository on GitHub

View on GitHub

Offline Jev token counting and request accounting for Node.js and Next.js. MIT TypeScript port of LabGuy94/jevtok.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity catalog-import logicrw tokenizer npm

Works forJev for Evaluation Jev for Decision gating

Used inJev for Software & DevTools

Languages: TypeScript JavaScript
Platforms: Next.js
Capabilities: Open source

Highlights

No Python, network calls, API keys, native extensions or runtime dependencies: vocabulary and lookup tables ship inside the npm package.
Ordinary o200k_base tokenization is not interchangeable with Jev — the word 'evidence' is three tokens and '2024' is four.
estimateRequestTokens computes an estimated usage.input_tokens from a state plus noul, choice or score questions and throws on budget violations.
ESM and CommonJS with TypeScript declarations for Node.js 22+, CI-tested on Node 22 and 24; browsers and edge runtimes are not supported.
A TypeScript port of LabGuy94/jevtok; getEncoding() defaults to jev-1.13 with 200,019 vocabulary entries and model version jev-1.13-20260917.

Quickstart

import { getEncoding, estimateRequestTokens } from 'jevtok-ts';

const enc = getEncoding('jev-1.13');
enc.count('evidence'); // 3

const questions = {
  relevant: { type: 'noul' as const, instructions: 'Is this about software?' },
};
const inputTokens = estimateRequestTokens({ title: 'Managed Postgres' }, questions);

Watch out

Copy reviewed against the truncated repo README on 2026-10-15; the capture ends inside the API documentation, so parts of the question-cost semantics were not reviewed. Encoder parity with the upstream jevtok is the port's own claim for one specific model version.

Questions about this entry

Short answers drawn from the linked project or page.

From the Jev guides

Original write-ups on what this entry is for.

The Open Reimplementations: Local Models That Speak System One

Sep 22, 2026

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

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai