User repository on GitHub
Watch Jev play Freedoom in a local dashboard. TypeSafe direct and Vercel AI Gateway, inspectable decisions, and bounded spending.
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Demos & Experiments
Jev Doom
Watch Jev play Freedoom in a local dashboard: structured engine observations go in, typed decisions come out, and the UI shows the game, choices, probabilities, navigation, latency and cost under a $15 default cap.
- Stars
- 2
- Language
- Python
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- olivier-motium
- Last verified
- Oct 2026
Works forJev for Realtime control
Used inJev for Gaming
Highlights
- Free preview loads a real engine frame without a provider key; billed runs use TypeSafe direct (TYPESAFE_API_KEY) or Vercel AI Gateway.
- Observable mode sends structured observations — visible entity identities and positions, viewed floor cells, remembered observations with ages — never pixels.
- Evaluator kill, hit and damage counters are withheld from model state, and code builds legal choices and executes the chosen motors.
- A $15 cumulative charged-or-reserved cap persists in runs/budget.json; each request reserves $0.05, and a checked run defaults to 90 seconds.
- Quick start: ./setup.sh and ./run.sh with Python 3.10–3.13 and Node 22+, dashboard at http://127.0.0.1:8767; setup downloads Freedoom 0.13.0.
Quickstart
bash
git clone https://github.com/olivier-motium/jev-doom.git
cd jev-doom
./setup.sh
./run.shWatch out
Copy reviewed against the truncated repo README on 2026-10-15; the excerpt stops inside the what-this-measures section and cuts the assisted-mode description mid-sentence. It is an independent experiment, not an official TypeSafe, Vercel or Doom product, and the README says Windows/WSL2 is untested.
Questions about this entry
Short answers drawn from the linked project or page.
An independent experiment that lets you watch Jev make structured decisions in a real Freedoom game. A local dashboard shows the game, choices, probabilities, navigation, latency and cost, and you can change its standing orders while it plays.
No. The default observable mode sends structured engine observations: visible entity identities and positions, viewed floor cells and wall pieces, and remembered observations with ages. The README calls this not a pixels-only agent; the rendered map is for inspection.
Not for a look: the free preview loads a real engine frame without a key. A real run needs TYPESAFE_API_KEY or AI_GATEWAY_API_KEY in .env, then Check provider makes one billed request before you click Start fresh run.
A $15 cumulative charged-or-reserved cap covers both providers and checks, persists in runs/budget.json, and reserves $0.05 per request. It can stop a run before $15 is actually billed; the README advises also setting provider account limits.
macOS or Linux, Python 3.10–3.13 (3.13 recommended), Node.js 22+, and internet access for setup; some systems need ViZDoom's native libraries. Windows users can try WSL2, but the README says it has not been validated.
Pause or stop it, edit standing orders, switch maps between runs, or select the coherent-tactics controller. Fixed tactics use a deterministic selector but still make timing calls, and the README says that mode is not an unpaid demo.
No. Setup downloads Freedoom 0.13.0 from its official release, and no proprietary Doom assets are needed or included. Game state and orders are sent to the selected provider, so do not put private information in standing orders.
More Jev resources like this
An autonomous ViZDoom agent that decouples Doom's simulation from Jev SystemOne inference: the game runs at 35 ticks/sec while Jev is queried asynchronously at about 10 Hz for macro goals, targets, movement and firing decisions.
Jev Demos & Experiments#community#catalog-import#logicrw
Community
amoghcreator/doom-jevPython
amoghcreator/doom-jevPythonA demo that walks a real Pokémon FireRed ROM: the code reads the overworld out of RAM, works out every reachable destination, and Jev picks where to go — and how to fight wild encounters — through typed decisions, not screenshots.
Jev Demos & Experiments#community#catalog-import#logicrw
Community
daniel4x/jevemonHTML
daniel4x/jevemonHTMLA pygame Flappy Bird that hands the bird to Jev: each new pipe becomes one multiple-choice question about simulated flight paths, and Jev answers with a typed choice plus a probability distribution.
Jev Demos & Experiments#community#catalog-import#logicrw
Community
leftspace89/jevbirdPython
leftspace89/jevbirdPythonTypeSafe's Jev plays the real Steam game Vampire Survivors - it picks the character, stage, every level-up and the walking direction four times a second, with a live decision dashboard. A BepInEx plugin and Python brain; native Linux only.
Jev Demos & Experiments#community#catalog-import#logicrw
A 12,000-neuron fly-brain subgraph of the maleCNS connectome fights Jev in mk.js: the fly simulates spikes in the browser at 10 decisions per second while Jev reads the fight as JSON and answers one typed choice among seven moves.
Jev Demos & Experiments#community#catalog-import#logicrw
Community
lavallee/mk-jev-fly-brainJavaScript
lavallee/mk-jev-fly-brainJavaScriptTwo TypeSafe System One game loops in one Python repo: Dino-Jev drives Chrome's chrome://dino/ dinosaur and Krunker-Jev runs a local Krunker-style arena, both feeding structured state to Jev and logging decisions.
Jev Demos & Experiments#community#catalog-import#logicrw
Community
mittal-parth/jev-experimentsPython
mittal-parth/jev-experimentsPythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.