Module npc_ai

This rule is by far the most significant rule in rtsim to date and governs the behaviour of rtsim NPCs. It uses a novel combinator-based API to express long-running NPC actions in a manner that’s halfway between async/coroutine programming and traditional AI decision trees.

It may feel unintuitive when you first work with it, but trust us: expressing your AI behaviour in this way brings radical advantages and will simplify your code and make debugging exponentially easier.

The fundamental abstraction is that of Actions. Actions, somewhat like core::future::Future, represent a long-running behaviour performed by an NPC. See Action for a deeper explanation of actions and the methods that can be used to combine them together.

NPC actions act upon the NPC’s crate::data::npc::Controller. This type represent the immediate behavioural intentions of the NPC during simulation, such as by specifying a location to walk to, an action to perform, speech to say, or some persistent state to change (like the NPC’s home site).

After brain simulation has occurred, the resulting controller state is passed to either rtsim’s internal NPC simulation rule (crate::rule::simulate_npcs) or, if the chunk the NPC is loaded, are passed to the Veloren server’s agent system which attempts to act in accordance with it.

Modules

airship_ai 🔒

dialogue

movement

util

Structs

DefaultState 🔒

NpcAi

Constants

SIMULATED_TICK_SKIP 🔒

How many ticks should pass between running NPC AI. Note that this only applies to simulated NPCs: loaded NPCs have their AI code run every tick. This means that AI code should be broadly DT-independent.

WALKING_SPEED 🔒

Functions

adventure 🔒

bird_large 🔒

captain 🔒

check_for_enemies 🔒

check_inbox 🔒

choose_plaza 🔒

find_farm 🔒

find_forest 🔒

gather_ingredients 🔒

hired 🔒

humanoid 🔒

hunt_animals 🔒

idle 🔒

monster 🔒

pilot 🔒

react_to_events 🔒

smalltalk_to 🔒

socialize 🔒

talk_to 🔒

tell_site_content 🔒

think 🔒

villager 🔒