EntityAI Kotlin DSL

EntityAI is a proof-of-concept library designed to solve a specific problem: writing Artificial Intelligence logic in Minecraft’s native Java API is verbose, fragile, and difficult to maintain.

This project is the result of “over-engineering” a university assignment.

The task was simple: implement basic character behaviors using a provided (and quite limited) AI library for Unity. Instead of following the easy path, my team and I decided to build a Minecraft Mod.

Since we couldn’t use the provided Unity tools (and Minecraft’s internal AI system is notoriously verbose and fragile), I had to architect my own solution. I engineered a high-level Kotlin DSL (Domain Specific Language) that replaces the spaghetti code of native Minecraft development with a clean, declarative syntax for Behavior Trees and State Machines.

The Kotlin DSL

The core innovation is the use of Kotlin Type-Safe Builders and Higher-Order Functions to create a readable syntax for AI logic. Instead of chaining objects manually, the API allows defining behavior hierarchies naturally.

ai.activities += TreeActivity("default", ai, rootLoopUnconditional {
    runAndWait {
        or {
            and {
                findNearestLivingEntities()
                or {
                    // Find enemy in a distance of 32 blocks.
                    findAttackTargetIfNotFound(32.0f) { it is EnderVillager }
                    findAttackTargetIfNotFound(32.0f) {
                        it is PlayerEntity && !it.isCreative && !it.isSpectator
                    }
                }
                // walkToEntity is a macro: generates more nodes under the hood.
                walkToEntity(MemoryTypes.ATTACK_TARGET, 1.5f, 1.0f, 32.0f)
                succeeder { attack() }
            }
            and {
                findRandomWalkTarget(1.0f)
                timed(40, 60) { walkToTarget() } // 40-60 ticks. 1 tick = 0.05s
            }
        }
    }
})

Architecture: The “Activity” System

The library introduces an architecture based on Activities. Unlike standard Minecraft entities that run a single flat list of goals, EntityAI entities run a Hierarchical State Machine.

Use Case: Magic End

The library was validated by building Magic End, a mod that introduces a new ecosystem to the End dimension.

• Complex Societies: Implemented as Endervilles, villages where entities have professions (Farmer, Guardian, Shaman) and interact socially (sharing food, calling for help).
• Evolution Mechanics: The Void Worm entity uses the State Machine to track its “kill count” and physically evolves into a Void Snake boss after consuming enough enemies.

Technical Challenges