Are the throughput and cost-saving benchmark numbers real?

They are in-repo procedures: tens of thousands of judgments against human labels with scripts included, reporting identical agreement for packed and unpacked arms within a stated interval rather than proof of equivalence.

What happens to the speedup if the provider starts billing by tokens?

It evaporates. Packing wins because the ceiling counts requests; the durable saving is the once-carried question cutting billed tokens, which survives even if the rate card is rewritten.

Can I use this to make one urgent judgment come back faster?

Do not. Packing makes a single waiting item slower, so call the API directly for interactive answers and reserve this for queues of thousands.

What is the catch before I pack user reports or long documents?

Use unpacked mode for hostile content, since packing mixes many unrelated items into one context, and measure long rows on your own data because depth falls off with length.

What does the guidance-once option change?

It stops repeating the question under every packed item. A live 32-item request dropped from 4,598 to 1,777 billed tokens, but it is off by default because it costs about 0.2 points of agreement.

How long can a single item be?

A single item is refused over 20,000 characters, and a request holds about 98,000 characters of items in total, so longer rows mean shallower packs; measure long rows on your own data because throughput does not degrade gracefully.

Can I resume a million-row job if it is killed?

Use checkpoint: it is built for re-running a taxonomy change over history, covering a million rows, and it survives being killed.

CommunityJev Demos & Experiments

Jev Ultralightspeed

Python bulk-classification helper that packs many independent short texts and one judgment question into fewer Jev requests, aimed at queues of tickets, logs, or labels.

See highlights

Open github.com

Stars: 12
Language: Python
Published by: Community
Author: collapseindex
Last verified: Oct 2026

collapseindex/jev-ultralightspeed

User repository on GitHub

View on GitHub

BRRRRRRRRRRRRRRRRRRRRRR

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

About this project

v0.24.0 · Apache-2.0 · no required dependencies. You have a pile of text and one question about each. Fifty thousand support tickets to triage. A quarter of reviews to sort by sentiment. A month of logs to flag. A column to backfill on a table you already have. This answers the question for all of them, with TypeSafe's Jev, in the time it takes to get a coffee.

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

Tagscommunity awesomejev catalog-import

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Python

Highlights

Single classify call over a pile of texts with triage and checkpoint helpers for million-row jobs.
Measured equal agreement between packed and one-per-request arms with full rerun scripts.
Guidance-once option that stops repeating the question under every packed item.
Length-based depth table so queue shape, not hope, decides expected throughput.

Watch out

Throughput arithmetic assumes a request-counted ceiling and collapses if pricing moves to tokens; hostile text should use unpacked mode since packing places many strangers in one context. 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

open-jev-typed-decision-engine

Open 150M encoder that reproduces the typed noul, choice, and score interface in one forward pass, trainable on a free GPU and runnable locally for nothing.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

intikhab49/open-jev-typed-decision-enginePython

TypeLLM

941

Library that adds schema-guaranteed typed outputs to existing autoregressive LLMs without changing weights, served through SGLang or a hosted API.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

TypeLLM/TypeLLMPython

jevify

Toolkit that makes a model behind an OpenAI-compatible endpoint or running in-process answer typed choice, score, and yes/no questions from next-token probabilities in Jev API shape.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

fidecastro/jevifyPython

jev-forge

164

End-to-end toolkit that synthesizes decision data, trains and calibrates candidate scorers, evaluates fixed splits, and serves Jev-compatible inference for interactive web decisions.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

zwliJay/jev-forgePython

jev-kit

Claude Code kit that guards PreToolUse calls with a free pre-filter plus one Jev question for ambiguous cases, and adds sub-agent sizing, rechecks, search, daemon speedups, and monitoring.

Jev Demos & Experiments#community#awesomejev#catalog-import

Community

jonathanavis96/jev-kitPython

stuntd

Self-hosted proxy that records typed decisions, distils them into small heads on a frozen encoder, and answers locally when confident.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

bladedevoff/stuntdPython

From the Jev guides

Original write-ups on what this entry is for.

Per-Tick Decisions in Games

Sep 28, 2026

A game tick is a state plus a legal action list. How Clash Royale, Minecraft, and the earlier Mario, Pokémon, StarCraft, and drone loops ask Jev one step at a time.

Jev Demos & Experiments#gaming#agents#control

Read article

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai