Debugging GOAP Plans with Visualizers

In this lesson, you will learn how to create a custom Editor window that exposes the inner workings of your GOAP planner, transforming abstract node chains into readable visual data.

The Problem with Black-Box Planning

When using A* search to generate a GOAP plan, the logic happens in a split second. If an NPC stands still or performs an unexpected action, it is difficult to know if the planner found an invalid path, failed to find a goal, or if the "World State" is incorrectly reporting conditions. By visualizing the plan, we turn this "black box" into a transparent sequence of state transitions.

Building the Visualizer

To visualize a plan, you need to access the Queue<GoapAction> currently being executed by your agent. We can use a custom EditorWindow to iterate through this queue and display the path taken.

[CustomEditor(typeof(GoapAgent))]
public class GoapVisualizer : Editor {
    public override void OnInspectorGUI() {
        base.OnInspectorGUI();
        GoapAgent agent = (GoapAgent)target;

        if (agent.CurrentPlan != null) {
            EditorGUILayout.LabelField("Current Plan Steps:", EditorStyles.boldLabel);
            foreach (var action in agent.CurrentPlan) {
                EditorGUILayout.LabelField($"> {action.name}");
            }
        }
    }
}

Visualizing State Transitions

Beyond the action list, the most powerful debugging tool is visualizing the State Chain. During the planning phase, your algorithm generates a series of "World States" representing what the world should look like after each action.

You should cache the WorldState associated with each node in your A* search. By drawing these as nodes in a GUI window using Handles.DrawLine or GUI.Box, you can see exactly where a precondition failed. If an NPC cannot find a "Get Wood" action, your visualizer will show you the exact state (e.g., HasAxe = False) that blocked the chain, saving you hours of stepping through code.

Key Takeaway: By caching the plan and its associated state transitions into a custom Inspector or Editor Window, you convert invisible planning logic into a step-by-step diagnostic roadmap that highlights exactly why an AI chose—or failed to choose—a specific action.