Ninja Tokens
Ninja Tokens is a lightweight coordination layer for AI. It solves the classic "too many agents do the same thing at once" problem by giving targets a token budget and letting agents request or queue access to actions through a fair and deterministic system.
Under the hood, each target owns a per-tag budget and queue; agents request a token, get granted immediately if capacity exists, or are enqueued and later granted as capacity frees up. Priority and FIFO tiebreakers keep outcomes predictable, and TTL policies are enforced opportunistically. No ticks or timers, so the system scales cleanly.
Features Overview
Tag-based tokens and budgets: Use Gameplay Tags to define independent action channels (e.g.,
AI.Token.Type.Attack) with per-target capacities and costs.Compatible with Behavior Trees and State Trees: Includes tasks to request and return tokens, and decorators/conditions to check token availability.
Fair, predictable queuing: Highest priority wins (useful for elites, bosses and the such); ties honor arrival order. If a high-priority request doesn’t fit, a smaller one that fits can proceed, no blocking.
TTL without ticking: Requests with time-to-live are expired on later interactions (new requests/returns). No timers, no per-frame overhead, allowing the system to scale further.
Simple integration points: Blueprint & C++ delegates and other extension and integration points, so the system can be easily extended or integrated with other modules.
Dedicated diagnostics: Integrated with the Visual Logger and Unreal Insights.