Skip to content

ACTIONS_AND_TASKS.md

Latest commit

 

History

History
55 lines (40 loc) · 2.9 KB

ACTIONS_AND_TASKS.md

File metadata and controls

55 lines (40 loc) · 2.9 KB

Will Cycle: Intentions and Executions

Design Philosophy

In Caosmos, the behavior engine is based on a hierarchy where the Action is the catalyst for all change. A citizen does not simply "be" in a task; they enter it through a validated action.

This "Will Cycle" leverages the non-deterministic nature of AI models (controlled by their temperature). We interpret the unpredictability of these models as Free Will: the capacity of a citizen to react differently to the same stimuli, creating an organic and living simulation.

1. Actions (The Catalyst)

An Action is the unit of intention that the system (AI or User) sends to the citizen. It is the "Verb" that initiates any interaction with the world.

  • Entry Point: All AI decisions are first manifested as an ActionRequest.
  • Validation and Gatekeeping: Before any change occurs, an ActionHandler validates if the action is physically possible (e.g., Are you near? Do you have the tool?).
  • Dual Natures: An action can have two natures depending on its purpose:
    • Atomic Action: Executes instantaneously and finishes (e.g., PICKUP to collect an object).
    • Generating Action: If validation is positive, the action "installs" a persistent Task in the citizen ( e.g., WORK or EXPLORE).

2. Tasks (The Continuous Execution)

A Task is the persistent state a citizen enters after a generating action. It represents a flow of logic that stays alive across multiple server "ticks", implementing the executeOnTick interface.

  • State of Being: While a task is active, the citizen changes their cognitive state (e.g., MOVING, BUSY).
  • Direct Modification: Unlike other systems, Tasks do not generate internal Actions. The task manipulates the citizen's state directly:
    • Calculates movement step-by-step (Lerp) and updates position (setPosition).
    • Gradually consumes energy and hunger by interacting with the citizen's biology.
    • Evaluates when the goal has been reached (e.g., reaching the exploration limit).
  • Interruption: Tasks return an ActiveTask on each tick and can be interrupted by external events (danger) or replaced by new actions.

3. The Flow of Reality

  1. Intention (ActionRequest): The AI decides: "I want to explore the North".
  2. Handling (ActionHandler): The engine validates: "Can this citizen start exploring now?".
  3. Activation (Task): If valid, an ExploreTask is instantiated and assigned.
  4. Physical Simulation: On each tick, the ExploreTask mathematically calculates the new position, moves the citizen 0.1 meters, reduces their energy, and checks for collisions, without generating new "Actions".
  5. Completion: The task finishes on its own (goal met) or is replaced by a new action.

Important

Remember: No Action, No Task. The Task is the persistent consequence of a will expressed through an Action.