Project

NPC AI Chatbot

Dialogue-driven NPC prototype combining behaviour trees with text generation.

Unity C# Behaviour Trees Scriptable Objects

Code

Overview

Solo project and my first substantial work using Python. A local NPC chat system where behaviour-driven decision logic determines when an NPC speaks while a local text layer generates responses. The project is designed to run fully offline (no paid APIs) and can be installed on a personal computer. The focus was on learning AI logic and enabling local execution rather than building polished art assets.

Features

• Behaviour-tree driven decision making for when to speak or stay silent.
• Blackboard-style memory system for tracking player context and world state.
• Pluggable dialogue modules for different NPC archetypes.
• Simple debug UI to visualise the active branch in the behaviour tree.

My contribution

• Authored the behaviour tree framework used by all NPCs in the prototype.
• Designed the data layout for blackboard entries and dialogue modules.
• Implemented profiling hooks to understand performance cost of AI updates.
• Iterated on UX to make behaviour debugging fast for designers.

Code snippets

Behaviour tree node evaluation

public abstract class BtNode
{
    public abstract BtState Tick(BtContext context);
}

public sealed class SequenceNode : BtNode
{
    private readonly BtNode[] _children;

    public override BtState Tick(BtContext context)
    {
        foreach (var child in _children)
        {
            var state = child.Tick(context);
            if (state != BtState.Success)
                return state;
        }
        return BtState.Success;
    }
}

Repository

Open on GitHub