Does the engine cheat with heuristics when the API lags?

No. The choice belongs to Jev and there is no heuristic fallback; gravity keeps falling while the query runs, and bad or late answers can still end the game.

Can I trust the confidence number on screen?

Only as the model's self-report: the displayed confidence is what the model returned and does not guarantee a good move.

Is the session cost figure my actual bill?

No, it is an in-memory estimate from tokens the API reports, reset per game and marked partial when calls omit usage; the provider invoice is authoritative.

Is it safe to run with my API key?

Reasonably for local demo use: the key lives in server config, goes only to the API host, never reaches the browser, and server or key files are never served publicly.

How do I run the demo locally?

With Node.js 22 or newer, git clone, copy .env.example to .env and fill in TYPESAFE_API_KEY, then run node --env-file=.env server.cjs and open http://127.0.0.1:8765; no npm install or build step is needed.

Can I play without an API key?

Yes. The Marcelinho joga mode runs without AI and also works by opening index.html directly, with no server or credentials.

How does the model pick a move?

The engine enumerates reachable placements with their lines, holes, and resulting height, the server sends the board and alternatives to the System One endpoint with jev-latest as a choice question, and the browser verifies and executes the chosen path.

CommunityJev Demos & Experiments

Demo Tetris Jev

Arcade Tetris demo where Jev picks each piece's rotation and position: the engine enumerates reachable placements and a local Node server asks one choice question per move.

See highlights

Open github.com

Stars: 1
Language: JavaScript
Published by: Community
Author: marcelomar21
Last verified: Oct 2026

marcelomar21/demo-tetris-jev

User repository on GitHub

View on GitHub

Tetris arcade jogado pelo Jev da TypeSafe AI, com decisões em JSON, antecipação de jogadas e custo por partida.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Uma demo em que o Jev, da TypeSafe AI, joga Tetris: o modelo escolhe a rotação e a posição de cada peça, e o jogo transforma a decisão estruturada em movimentos reais. Interface arcade em Canvas 2D, com a decisão em JSON, confiança, latência e custo estimado da partida visíveis durante o jogo. JavaScript e Node.js, sem dependências externas. Requer Node.js 22 ou superior. Para jogar com IA, configure uma chave da TypeSafe com créditos disponíveis. Edite .env e preencha TYPESAFEAPIKEY. Depois:

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

Tagscommunity github

Works forJev for Classification

Used inJev for Software & DevTools

Languages: JavaScript

Highlights

Shows the decision JSON, confidence, latency, and estimated session cost live beside the Canvas board during play.
Lookahead prepares the next piece while gravity continues; stale predictions are discarded with at most one query in flight.
Runs with no install or build step on recent Node; a manual mode plays without any key, server, or credential.
Key stays server-side bound to localhost with origin checks, and tests run on simulated responses without spending credits.

Watch out

Local game demo; cost and confidence figures are on-screen estimates. 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

jev-playground

Zero-dependency local playground showing six Jev workflows from triage to eval judging with shippable snippets.

Jev Demos & Experiments#community#github

Community

AviroopPaul/jev-playgroundJavaScript

computer-use-cache

Drop-in OpenAI-compatible cache proxy for repeatable agent workflows: exact hits stay local, and with a TypeSafe key Jev judges whether a cached response is semantically reusable for the current request.