Dungeon Director: AI Dungeon Master

• 

  staring window

• 

  game view

• 

  ai predicting and controlling difficulty

• 

  overview of game by ai

Inspiration

Most dungeon games have static levels where enemies and traps follow predictable patterns. I wanted to explore the idea of an AI that actively tries to stop the player in real time, similar to a dynamic dungeon master in tabletop RPGs. The inspiration came from games like roguelikes and AI-driven systems where the environment reacts to the player instead of remaining fixed.

What it does

Dungeon Director is a roguelike survival game where an AI Director controls the dungeon while the player tries to escape.

The player must collect keys, avoid traps and enemies, and reach the exit. Meanwhile, the AI Director dynamically changes the dungeon by triggering events such as:

• Gravity inversion
• Darkness mode
• Enemy speed boosts
• Control reversal
• Chaos mode
• Spawning enemies
• Shifting walls
• Spike traps

This creates a constantly changing environment where the player must adapt quickly to survive.

How we built it

The game was built as a browser-based roguelike using HTML, CSS, and JavaScript.

Key systems include:

• Grid-based dungeon engine for movement and tile interaction
  
• Director AI event system that dynamically triggers gameplay changes
  
• Enemy AI pathfinding that hunts the player
  
• Procedural dungeon elements such as spikes, keys, and enemies
  
• Real-time HUD and AI log panel that shows the Director’s actions
  
• Particle and glitch visual effects for feedback and immersion

The AI Director uses a weighted event system to determine when and how to disrupt the player.

Challenges we ran into

One of the biggest challenges was balancing the AI Director so that the game remains challenging but not impossible.

If the AI triggered too many events, the player would lose immediately. If it triggered too few, the game became predictable.

Another challenge was designing a clean interface that shows what the AI is doing, so the player understands why the environment is changing.

Accomplishments that we're proud of

What we learned

Through this project we learned about:

• Designing adaptive gameplay systems
• Balancing procedural and AI-driven mechanics
• Creating responsive browser-based games

• Managing real-time events and game loops in JavaScript

  What's next for Dungeon Director: AI Dungeon Master

  Future improvements could include:

• Machine learning–based AI behavior adaptation

• More enemy types and boss encounters

• Procedural dungeon generation

• Multiplayer survival mode

• Leaderboards and global challenges

The long-term vision is to explore how AI can act as a game designer during gameplay, constantly reshaping the player’s experience.

Built With

• ai
• api
• browser
• canvas
• css3
• development
• game
• generation
• html5
• javascript
• procedural
• roguelike
• systems