User repository on GitHub
discord bot for jev that lets it talk
- Forks
- 4
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Demos & Experiments
Jevbot
Discord bot forcing a textless decision model to speak by repeatedly picking the next word through vocab tournaments, a completeness check, and reuse penalties.
- Stars
- 21
- Language
- Python
- Published by
- Community
- Author
- lyramakesmusic
- Last verified
- Oct 2026
About this project
Discord bot that makes TypeSafe's Jev talk — a decision model that "cannot generate text," loomed word-by-word into broken sentences. Jev is a non-autoregressive decision model. It answers questions with calibrated probabilities, not text. This bot gives it a 20K word vocabulary and asks "next word?" repeatedly via tournament sampling until it forms a reply. Tournament sampling: 20K vocab shuffled into 255-word buckets, all scored in parallel. Runoff: Top-2 from each bucket compete in a final round.
Quoted from the project's README. Read the full README.
Works forJev for Classification
Used inJev for Software & DevTools
- Languages
- Python
Highlights
- Builds replies word by word from a 20K-word vocabulary through parallel bucket scoring.
- Runoff round plus a separate completeness question decides when a reply ends.
- Reuse penalties on content words and stopwords keep generation out of repetition loops.
- Vocabulary file doubles as the content filter and ships with slurs removed.
Watch out
Replies are intentionally broken and cost several model calls per word, so expect novelty output rather than fluent chat. 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.
Tens of thousands of words are shuffled into bounded buckets, scored in parallel, and the top two per bucket compete in a final round until a full reply forms.
A separate true-or-false question asks whether the reply is complete, and generation stops when the model judges it done.
Each reply costs on the order of cents because every word needs roughly half a dozen model calls through the proxy API.
Only the last few user messages are included, while the bot own prior output is excluded because it degrades follow-ups; it also answers only mentions and replies.
Run pip install -r requirements.txt, create a .env with DISCORD_TOKEN_JEV and OPENROUTER_API_KEY, enable Message Content Intent in the Discord developer portal, and run python jev_bot.py.
A penalty system multiplies content words by 2.5 per reuse and stopwords by 1.6, which prevents loops like repeated filler words; Jev approximates text through probabilities rather than generating tokens.
vocab.txt is a 20,000-word list taken from bewinxed/jevgpt with slurs removed; words can be added or removed freely, so the vocabulary itself acts as the content filter.
More Jev resources like this
46GitHub stars
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
941GitHub stars
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
TypeLLM/TypeLLMPython45GitHub stars
Drop-in service that exposes Upstage Solar models through the System One shape, accepting state plus noul, choice, or score questions.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
hunkim/solar-mini4-jevPython
hunkim/solar-mini4-jevPython38GitHub stars
Research recipe that fuses two frozen open readers into Jev-style yes/no judgments with a fit-free router, confidence gate, and candidate sets.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
FeiLiuEM/open-medical-jevPython
FeiLiuEM/open-medical-jevPython62GitHub stars
Simulation study that drives one xArm7 apple-to-plate task with Jev, a large reasoning model, and a small model through identical incremental controls.
Jev Demos & Experiments#community#jevbest#catalog-import
40GitHub stars
Open research toolkit that encodes an image once and answers multiple typed choice, noul, and score questions from a shared distribution, plus a local 0.6B text-decision model.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
IamBusy/OpenJev-VisionPython
IamBusy/OpenJev-VisionPythonFrom the Jev guides
Original write-ups on what this entry is for.
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
From the Jev community
Posts from builders shipping with Jev right now.