Game Jam Entries
Two video games I made on different teams as part of the CSU Game Dev club for submission into game jam competitions.
Github
ViewThese are the two games I made during my first semester of game developer club.
A game jam is a game development competition where a team of people compete to make a video game in a limited time frame. There’s also generally a theme that the game has to follow and teams aren’t allowed to use any premade assets or code.
Spacekeep
The first game I worked on was for the Ludum Dare 54 game jam.
This was a 72 hour jam with the theme "Limited Space".
I worked on a team with five other people, three of which were programmers.
Our interpretation of this theme was limited inventory space, so we settled on creating an inventory management game.
We took inspiration from the game Dome Keeper and decided to set the game in space with evil aliens attacking.
Most of my work on the game was related to mining. In order to upgrade their ship, the player had to venture out into space and mine asteroids for resources. The chance of valuable materials dropping also increases with the distance from the mothership.
I wanted mining to be a little cartoonish, so I made the asteroids bounce around when they were hit. To get this effect, I created a mini explosion whenever the player and an asteroid collide. A strong force is applied to both objects, but I made the ship immovable so it isn't affected. I also added a high drag to the asteroid so it comes to a stop quickly after being hit.
For ore types, I defined four scriptable objects with different properties. Then, when an asteroid is destroyed, it spawns three ore chunk prefabs, randomly picking an ore type for each. Depending on the tier of the asteroid, the chance of getting a higher tier ore type is increased. In this video the different colors represent different ore types:
The last contribution I made was to enemy spawning. Originally, we were trying to implement Boids to make the enemies move in a flocking pattern, but we ran out of time. Instead, I built a simple system that spawns enemies at predefined points near the edge of the map and moves them slowly towards the center.
Verminium
The second game jam I participated in was the Spooktober Fest Jam. It was an internal game jam for the CSU Game Dev Club that lasted three weeks. This time, I worked with two other people and the theme was "Tiny Spooky Creatures" since it was being held in October.
Verminium is about a rat in an abandoned city that collects trinkets and trades them for gear. The unique mechanic we came up with was that all the NPCs are GPT agents. GPT is the same LLM (Large Language Model) that powers ChatGPT. This meant that interactions between the player and NPCs were unscripted and unique on each playthrough.
I was in charge of the trading system and bringing the AI NPCs to life.
To protect my API keys I had to set up a middleman server that would handle requests between the game and the OpenAI API. I used Express.js to create the server and Langchain to make my API calls more manageable.
I wrote a boilerplate prompt that would be used as a template for interactions. This helped the AI get into character and explained their role.
This template was then filled with the player's question, the NPC and player items, and a summary of the chat history up to that point.
This is what one template looked like:
{personality}
========
You are bartering with the player given the following information about the current state of the trade.
========
These are YOUR ITEMS (the number after each item is the value of the item):
{npcItems}
========
These are THE PLAYERS ITEMS (the number after each item is the value of the item):
{playerItems}
========
This is a summary of the previous messages between you and the player (if blank, this is the first interaction):
{currentSummary}
========
Here's the user's NEW message:
{userMessage}
========
NEVER mention the chat history and NEVER mention the value numbers.
NEVER refuse a trade that has balanced value numbers.
NEVER refuse to trade one of your items.
NEVER continue to barter if the player has agreed to your offer.
NEVER change your mind if the player has agreed to your offer.
If the chat history and player message suggest that the player has agreed, ONLY say "Thank your for your business"
The response to this prompt was then chained into a followup prompt to generate a summary for the chat history and update a trade array. In this summary stage the AI also picked whether the iteraction should end. If the chat was ended then the most recent trade array would be used in game to update the player and NPC's inventory. The most difficult part of this process was getting GPT to follow the rules of the game because it often acted irrationally or out of character.
Flycatcher - Waitlist Builder
Build your custom branded waitlist, collect insights about your subscribers, and send email campaign...
TOS Chat - Custom AI Chatbot
Custom GPT chatbot that allows the user to 'chat' with terms of service and privacy policy agreement...
Object Recognition in Autonomous Rover
Adding a human detection system to a food delivery robot that serves boarding students on campus who...
All Posts