Behaviors

Giving agents agency and specifying laws of the universe
Agents can (and typically do) have behaviors:
{
    ...
    behaviors: ['eats_fruit.js', 'messages_friends_fruit.py', 'has_friends.js']
}
Behaviors allow agents to exhibit agency in the simulation. An agent can have any number of behaviors.
Behaviors are pure functions in the form of: (current_state, context) => {next_state}
To run your simulation with hCloud, you'll need to define Behavior Keys for each behavior in your simulation.
The exact semantics vary by programming language, but in spirit every behavior is a pure function which receives a given agent's state, the world context visible to that agent, and produces its next state.
During a timestep an agent passes its state and context to its associated behaviors, modifying its state
Most behaviors output a single state with the same agent_id as they received. For example, the following behavior code causes an agent to move along the x-axis:
JavaScript
Python
JavaScript
const behavior = (state, context) => {
    state.modify("position", position => [position[0] + 1, position[1]]);
}
Python
def behavior(state, context):
  def behavior(state, context):
    position = state.get("position")
    state.set("position", [position[0]+1, position[1]])
  
Agents can use behaviors to create new agents by sending a message to the special hash agent_id. This is covered more in-depth in Messages.
Networks
Composable behaviors
Last updated 4 weeks ago