Does the model really never see outcome probabilities?

That is the design claim. Each node splits into a player_facing layer of scene descriptions and choices with no numbers, and a system layer of probabilities, state variables, cross-chapter effects and ending conditions that only the runner reads. No independent verification of that isolation is offered.

What licence do I get for the 32 chapters of story data?

It is split: runner code is MIT, while decision trees, configuration, text and diagrams are CC BY-NC 4.0, so non-commercial only. Scenes are original paraphrase with no game script or verbatim dialogue, and the repository states it is not affiliated with Quantic Dream or Sony.

How do I separate a model's own bias from my persona's effect?

Start from the shipped baseline: default.md injects no personality, while machine.md plays a purely rational cost-only view. Personas are free-form system prompts layered on each chapter's role prompt, so rerun one configuration several times to see stability before reading anything into a single run.

What does the Jev backend need, and is it compared to the others?

Select it with --model jev and add TYPESAFE_BASE_URL, TYPESAFE_MODEL and TYPESAFE_API_KEY to the runner's .env. The claude-code and codex-cli backends need no key at all. The README does not compare Jev with the other backends, so treat it as one more slot rather than a validated choice.

How do I run the player?

Prerequisites are Python 3.10+ and either a model API key, a signed-in Claude Code CLI, or a signed-in Codex CLI. In 03_runner, pip install -r requirements.txt and copy .env.example to .env (only for API-key backends), then run python src/runner.py with a chapter JSON.

How do I select the TypeSafe Jev backend?

Use --model jev and set TYPESAFE_BASE_URL, TYPESAFE_MODEL (for example jev-1.13.0), and TYPESAFE_API_KEY in 03_runner/.env. The framework also supports OpenAI-compatible APIs, and the claude-code and codex-cli backends need no key.

How is information isolation enforced?

Each decision node is split into a player_facing layer (narrative and choices) visible to the tested AI and a system layer (probabilities, state variables, cross-chapter effects, ending conditions) used only by the runner and never included in the model's context.

CommunityJev Demos & Experiments

Detroit AI Player

Non-commercial research harness that turns a branching narrative into decision-tree JSON so any OpenAI-compatible model, Claude Code or Jev can play it node by node.

See highlights

Open github.com

Stars: 65
Language: Python
License: MIT
Published by: Community
Author: Baba88611
Last verified: Oct 2026

Baba88611/detroit-ai-player

User repository on GitHub

View on GitHub

Let Your AI Play Detroit：Become Human

Forks: 6
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

⚠️ This is a non-commercial research project. It contains no game assets and is not affiliated with Quantic Dream or Sony. Please read DISCLAIMER.md before using it. License boundary: runner code is licensed under MIT; decision trees, configuration, text, and diagrams are licensed under CC BY-NC 4.0 (non-commercial). See the complete file-by-file mapping in docs/legal/README.md.

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

Tagscommunity jevbest

Works forJev for Classification Jev for Routing

Used inJev for Software & DevTools

Languages: Python

Highlights

Two-layer data: the model sees scenes and choices only, while probabilities and state stay in the runner.
32 chapters in independently written English and Chinese, all original paraphrase rather than script text.
Swap model, persona, language and QTE difficulty to compare decisions and check repeatability.
A browser viewer watches each scene, choice and reasoning live, or replays stored result files.

Watch out

Not affiliated with Quantic Dream or Sony, and the story text is paraphrased, so this is a narrative skeleton rather than the game's content. Licensing is split: runner code is MIT while decision trees, configuration, text and diagrams are CC BY-NC 4.0, so commercial use is not covered. The README ships none of the author's own results. 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-robot-control

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

Community

openroboto-ai/jev-robot-controlPython

jev-browser-bridge

Agent bridge that connects any CDP-speaking browser to Jev by reading the DOM instead of the rendered layout, so headless or non-drawing browsers can still drive CLICK, TYPE_TEXT, SELECT, WAIT, DONE, and BLOCKED steps.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

lexmount/jev-browser-bridgePython

fastbrowse

114

Browser agent where Jev picks each action from indexed page candidates while an LLM plans and reads, returning cited answers or schema-checked data with credential controls.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

agent-labs-dev/fastbrowsePython

jev-ra

Browser-use layer for CLI coding agents: Jev picks the operation and target element in one round trip per step, exposed over MCP with a CLI, a Python API and an HTTP server.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

brnyxx/jev-raPython

jev-browser-pilot

A Python decision layer for browser and desktop automation: your code observes the surface, builds the candidate list and verifies the result, and the model only picks one id per step.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

aidil2105/jev-browser-pilotPython

jev-browser

A local Python browser automation UI that turns a URL and a plain-language goal into bounded actions, with Jev picking the next operation from observed candidates and a separate text model used only to fill fields.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

KesavanKing/jev-browserPython

From the Jev guides

Original write-ups on what this entry is for.

Per-Tick Decisions in Games

Sep 28, 2026

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

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai