Skip to content

Override System

The override system lets you tune action, goal, belief, and transition parameters on a per-agent basis without creating a new Brain asset. This is how you create agent variations from a shared brain template.

Why Overrides?

A single Brain asset can be shared by many agents. But you might want:

  • A veteran guard with higher detection range
  • A cowardly enemy with a lower attack priority
  • A specific agent that ignores certain transitions

Overrides change parameters without changing the brain's structure. The planner topology (actions, goals, beliefs, transitions) stays the same — only the numbers change.

Four Override Types

Override Type What You Can Change What Stays the Same
Action Strategy, settings, cost, post-processes Preconditions, effects, planner topology
Goal Priority, validity conditions Desired state, transitions
Belief Settings (range, threshold, etc.) Same bit in TagMask128, same Evaluate() logic
Transition Enabled/disabled, condition Source/target state

Action Overrides

Swap the strategy, change settings, or adjust cost for a specific agent:

var context = agent.GetContext<GoapAgentContext>();

// Override an action's cost
context.SetActionOverride(actionInstance, new GoapActionOverrideEntry
{
    Strategy = customStrategy,
    Settings = customSettings,
    Cost = 5f
});

// Remove the override (revert to brain defaults)
context.RemoveActionOverride(actionInstance);

Use case: A wounded agent uses a slower, more cautious movement strategy with higher cost, making the planner prefer other actions when available.

Goal Overrides

Change a goal's priority or validity conditions for a specific agent:

context.SetGoalOverride(goalInstance, new GoapGoalOverrideEntry
{
    Priority = 10f  // This agent really wants this goal
});

context.RemoveGoalOverride(goalInstance);

Use case: A boss enemy has "Attack" at priority 10 while normal enemies have it at priority 3.

Belief Overrides

Tune belief parameters (detection range, thresholds, etc.) per-agent:

// Via the BeliefSettings container
var beliefSettings = context.BeliefSettings;

// Set an override
beliefSettings.SetOverride(beliefInstanceId, "CustomRange", new MyBeliefSettings
{
    TargetKeyId = keyId,
    Threshold = 20f  // This agent detects from farther away
});

// Check for override
if (beliefSettings.TryGetOverride(beliefInstanceId, out IBeliefSettings settings))
{
    // Use overridden settings
}

// Remove override
beliefSettings.RemoveOverride(beliefInstanceId);

Evaluation Priority

  1. Per-agent override (from BeliefSettings container) — highest priority
  2. Instance DefaultSettings — lowest priority

Use case: A veteran guard has a proximity belief with Range = 20f instead of the default Range = 10f.

Note

Belief overrides change parameters, not the belief's existence in the planner. The belief still occupies the same bit in TagMask128 and uses the same Evaluate() method.

Transition Overrides

Enable, disable, or change the condition of an FSM transition for a specific agent:

// Disable a transition for this agent
context.SetTransitionOverride(transitionId, new GoapTransitionOverrideEntry
{
    Enabled = false
});

// Re-enable it
context.ClearTransitionOverride(transitionId);

Use case: A berserker agent has the "flee to safety" transition disabled — it never retreats.

Identifying Targets by InstanceId

Overrides target specific instances by their InstanceId (a SerializableGuid), not by SO reference.

Warning

Two action instances can share the same strategy SO but have different InstanceIds and different settings. Always identify override targets by InstanceId, not by SO reference.

Bake to New Brain

If you've configured overrides during Play mode and want to make them permanent, use the Bake button in the GOAP Hub. This creates a new Brain asset with the overrides baked in as default values.

This workflow is useful for:

  • Prototyping agent variations during Play mode
  • Creating specialized brains from a base template
  • Converting runtime tuning into permanent configuration

What's Next