User repository on GitHub
Shady Town: social-deduction party game for the living room TV, moderated by TypeSafe Jev
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Demos & Experiments
Shady Town
Social-deduction party game where the TV moderates: humans play on their phones, and one fast structured Jev judgment per round turns game state into per-player suspicion meters, with decisions and thresholds left in code.
- Stars
- 0
- Language
- Ruby
- Published by
- Community
- Author
- tpaulshippy
- Last verified
- Oct 2026
About this project
A social-deduction party game for the living room TV. Humans play, the TV moderates. Shady Town turns your living room into a detective show. Everyone plays — no one has to sit out to be the moderator. You act, argue, bluff, and vote on your phones. The TV shows live suspicion meters next to each player's name, like a sports broadcast. Did Ben just lie about being the doctor? The meters twitch. Is the group going in circles? The game nudges you: "new clue needed" or "time to vote."
Quoted from the project's README. Read the full README.
- Languages
- Ruby
Highlights
- One fast structured judgment per round - game state in, suspicion levels out - with decisions and thresholds kept in code, on Jev via ruby_llm-typesafe.
- Meters react to votes and claims rather than volume, so quiet players still move, and the TV pushes the room to vote when arguments start looping.
- bundle exec ruby bin/moderate runs in mock mode with no key needed, and the specs run offline against injected fake answers.
- Eliminated players go on the jury - they can watch and chat but not vote - and the TV ends with a recap of the biggest bluff.
Watch out
Worth stating plainly first: the README carries a name note making clear this is an original game, wording and code inspired by the classic 1986 social-deduction party format, with no affiliation with and no use of any commercial Mafia or Werewolf board, card or video game - if you are looking for a licensed implementation of that format, this is not it. The Jev layer is deliberately thin and the README is honest about it: one fast structured judgment per round, game state in and suspicion levels out, with decisions and thresholds in code, on jev-latest via ruby_llm-typesafe. What the README does not document is the substance behind that: which questions are asked each round, the threshold values behind the Clear, Shady and Busted meters, token cost, latency, or what happens on a low-confidence or failed judgment. It also does not specify TV hardware or browser requirements, and mock mode is the only path that runs without a TypeSafe key. 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.
One structured judgment per round. Game state goes in and suspicion levels come out, with the evaluator mapping those to each player's Clear, Shady or Busted meter plus bluff and stall signals; decisions and thresholds stay in code rather than in the model. It is a modest, deliberately narrow use of Jev via ruby_llm-typesafe at jev-latest. The README does not list the individual questions, so reading lib/shady_town/evaluator.rb is how you would see the actual prompt surface.
Yes for the developer path. bundle exec ruby bin/moderate runs the demo in mock mode with no key needed, and bundle exec rspec runs the tests, which use injected fake answers so they run offline. Live mode is the same command with TYPESAFE_API_KEY exported. This is a Ruby project: lib/shady_town/game_state.rb holds round, alive, votes, claims and events, lib/shady_town/evaluator.rb holds the per-player judgments and bluff and stall signals, and bin/moderate is the demo.
No, and the README has an explicit name note about it. This is an original game inspired by the classic 1986 social-deduction party format - its own name, its own wording and its own code - with no affiliation with and no use of any commercial Mafia or Werewolf board game, card game or video game. Mechanically it keeps the familiar loop: troublemakers take someone out at night, the doctor saves, the detective checks, and the town wins by finding them all while the troublemakers win by outnumbering the town.
That is the design target. Votes and claims move the meters rather than loudness, so a player who has barely spoken with no votes against them stays Clear, while someone whose doctor claim does not add up and is collecting votes slides toward Busted for everyone to see. The TV also reads the room two ways: a bluff signal when stories contradict each other, and a stall signal after minutes of repeated arguments with no new information, which suggests taking 60 more seconds and then voting. The expected setup is 6-10 people, a TV, about 30 minutes, and 3-4 rounds.
Ruby. It is built on TypeSafe's Jev through ruby_llm-typesafe (jev-latest), and it makes one fast structured judgment per round: game state in, suspicion levels out. Decisions and thresholds stay in code, in lib/shady_town/evaluator.rb and game_state.rb.
A per-player suspicion level of Clear, Shady or Busted, plus bluff and stall signals. The room-mood checks light up when stories contradict each other or when arguments repeat without new information, and the meters reflect only what the room has said and done so far.
Bring 6-10 people, a TV and about 30 minutes. One person opens Shady Town on the TV and everyone else joins on a phone with a name, no app install required: scan a link, get a secret role, and play 3-4 rounds of night, day and vote.
More Jev resources like this
8GitHub stars
Claude-driven browser automation where plain-English chat drives Brave or Chrome, Jev makes on-page decisions, and a guard freezes irreversible clicks.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
ranjan2829/AskJevTypeScript
ranjan2829/AskJevTypeScript4GitHub stars
Manifest V3 ad blocker that scores candidate feed elements with Jev's noul primitive to collapse native ads, layered over network rules and a media engine.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
vmendes90/jev-shieldTypeScript
vmendes90/jev-shieldTypeScript1GitHub stars
A Bun and Next.js moderation demo that screens text with one systemOne call — seven hazard Nouls plus a severity Score in parallel — and computes the verdict in code from policy thresholds.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
NorbertBodziony/guard-jevTypeScript
NorbertBodziony/guard-jevTypeScript0GitHub stars
Last Exit is a short cyberpunk checkpoint game where you type a cover story, present a manifest and offer a bribe to an inspector who decides whether to let you through. Each reply makes two Jev calls: one turns your statements into typed observations, the other picks his next action.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
0x963D/last-exitJavaScript
0x963D/last-exitJavaScript0GitHub stars
Sloppy Jev's is a Manifest V3 Chrome extension that blurs AI-generated prose and ads, classifying extracted page text through jev-latest on the TypeSafe API. Plain JavaScript and CSS, no build step, MIT licensed, loaded unpacked.
Jev Demos & Experiments#community#jevbest#catalog-import
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
From 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.