// the find
zkat/big-brain
Utility AI library for the Bevy game engine
big-brain is a Utility AI library for the Bevy game engine — the classic game AI pattern where actions compete on scored utility values rather than running through fixed state machines or behavior trees. It's for Bevy game developers who want NPCs that respond to world state in a data-driven, composable way without hand-rolling all the scoring logic.
The ECS integration is first-class: Scorers and Actions are real Bevy systems running in parallel, so you get Bevy's scheduling and parallelism for free rather than bolting on a separate AI loop. The derive macros (ScorerBuilder, ActionBuilder) eliminate the boilerplate that usually makes ECS-based AI tedious. Action cancellation and state machine-style continuous behaviors are built in, which most toy implementations skip. Reflection support means bevy_inspector_egui can introspect AI state at runtime, which is practically necessary for debugging utility weights.
Development has moved to Codeberg — this GitHub repo is a read-only archive. Anyone adopting it needs to track the new home, and tooling that points at GitHub (docs links, crates.io source) will quietly go stale. The picker system (FirstToScore, etc.) is thin: there's no built-in support for cooldowns, hysteresis, or weighted-random selection, so you end up writing custom pickers pretty quickly for anything beyond toy examples. No built-in spatial awareness or perception system — you have to wire up 'what can this entity see' yourself before Scorers can do anything useful, and the examples gloss over that entirely.