Question 1

Is Jev actually playing Snake, or is the engine doing the work?

Accepted Answer

Jev is making the movement decision, within a deliberately thin set of guardrails. The engine owns rules and state transitions and rejects an impossible 180-degree reversal, but it does not pathfind or override Jev's choice to avoid a wall, obstacle or body collision. The python engine is the same in human and Jev mode; only the source of each requested direction changes.

Question 2

What happens when the API is slow or fails mid-game?

Accepted Answer

The snake holds its current direction for that tick and the miss is displayed and logged, with the completed decision applied at the next available tick. Jev inference runs asynchronously between movement ticks and cannot slow the board, so physical speed stays identical for Human and Jev comparisons. That makes latency limits visible instead of hidden, and it is the fallback_keep_direction policy being used, not the model quietly pathfinding.

Question 3

Can I reproduce an exact run later?

Accepted Answer

Yes, by seed. Food, obstacles and decoys are randomized by the recorded seed, so a run is both varied and exactly reproducible, and each tick log captures the full current state plus the seed-derived environment. Choose Human or Jev, a versioned difficulty and a seed in the browser UI, or pass --seed, --width, --height, --obstacles, --tick-rate and --model to the pygame client.

Question 4

How do I know a Jev score means anything?

Accepted Answer

The README is blunt that Snake has no official widely adopted benchmark and that a single high score does not prove ability. Every run is identified by environment version, named difficulty preset, exact seed, controller type, score, survival ticks and terminal reason, so you compare controllers over the same fixed seed suite and report score and survival distributions instead of quoting one run.

Question 5

How do I set it up and run it locally?

Accepted Answer

Nothing goes in globally. From the project directory create and activate a virtual environment with python3 -m venv .venv, then install requirements with python -m pip install -r requirements.txt. The browser baseline is python -m snake_game.web served on http://127.0.0.1:8000, and the pygame client is python -m snake_game.main with a mode flag. The repo issues those instructions without installing anything for you.

Question 6

What are the difficulty presets, and how do decoys work?

Accepted Answer

All three use the same 20 by 20 board and the same 1.5 moves per second, so Human and Jev runs stay comparable. Easy has no obstacles or decoys, Medium has 8 obstacles and 3 amber decoy fruits, and Hard has 20 obstacles and 7 decoys. Red fruit scores and touching an amber decoy ends the run, and decoys reshuffle every time real fruit is eaten to add uncertainty without changing speed.

Question 7

Can I test the engine with no key and no network?

Accepted Answer

Yes. The engine tests need neither pygame nor network access, just python -m pytest. Jev mode is the part that needs credentials: TYPESAFE_API_KEY in the project-local .env or in the server process environment, with the browser never receiving it. So the rules and state transitions are testable offline, while only the movement decisions need the API.

Jev Snake — iammusham

About this project

Highlights

Questions about this entry

More Jev resources like this

From the Jev guides

From the Jev community