jev-factorio-agent
Factorio agent on FLE where Jev picks goals, next actions and stuck detection while code owns rules and actuation.
Overview
jev-factorio is a Factorio agent in which Jev makes the fast macro decisions (current goal, next action, and whether the agent is stuck) as typed Choice, Score and Noul questions, while deterministic code owns game rules, option filtering and actuation. The authors describe it as, to their knowledge, the first Jev-driven game agent, running on the Factorio Learning Environment.
A mock backend lets the loop run offline with no game, and with a key it calls jev-latest on the System One endpoint. Optional research logging creates exclusive evidence directories with a secret-safe manifest, SHA-256-chained lifecycle events and a final seal, alongside a decision JSONL log. A hierarchical controller is available, and documentation covers architecture, build plan, latency attribution and autonomous campaign supervision with error-triggered Codex repair.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What goal should the agent pursue now? | — |
| Choice | Which filtered action should it take next? | — |
| Yes / no | Is the agent stuck? | — |
How it works
- Read game state from FLE or the mock backend
- Code filters valid options
- Jev chooses goal and next action and checks for stuck
- Code actuates and logs decisions
Strengths
- Clear split between decisions and game mechanics
- Offline mock mode
- Tamper-evident research logs
Caveats
- Research agent; performance not summarised in the excerpt
- Real games require Factorio and FLE setup
Getting started
`pip install -e .` then `PYTHONPATH=src python -m jev_factorio --backend mock --steps 8 --tick-seconds 0`.At a glance
- Type
- Project
- Status
- research
- Language
- Python
- Runtime
- Factorio Learning Environment
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Jev picks what, code owns how - a System One Factorio agent driven by TypeSafe's Jev on FLE
logicrw/awesome-jev-projects
Frequently asked questions
What does Jev decide?
The agent's goal, next action and whether it is stuck.
Can I run it without the game?
Yes, with the mock backend and no key.
How are runs recorded?
Optional research logging writes sealed, hash-chained evidence directories.
Related projects
AI life-and-civilization simulation where Jev makes every person's next action as an auditable Choice and LLMs only plan.
ProjectprototypeJavaScriptReal-time bot battles where you write a fighter's strategy in plain English and a System One model pilots it several times a second.
ProjectbetaTypeScriptBot that plays Red Alert 2 web skirmishes, with code handling mechanics and a decision model making stance, threat and timing calls.
ServicebetaPythonMultiplayer 3D science-fantasy version of Houston's River Oaks District with Jev-powered NPC companion, chauffeur and smart drive.
ProjectbetaTypeScriptBrowser 3D planetary rover sandbox where a TypeSafe controller chooses the next action from complete candidates, versus a baseline.
ProjectbetaTypeScriptDoom runs as a spacecraft payload over real flight software; Jev pilots from the ground every half second while Claude reviews.
ProjectprototypePython