#writing #fantasy #sciencefiction #sfandf #godot #interactivefiction #inkscript #inkle #gamedev #ollama #dnd #dungeonsanddragons #ttrpg #rpg

Not too long ago, my friends and I sat down and were toying with the idea of making a shared universe with the potential end goal of creating a game. I've been drawn to the idea of doing an interactive fiction RPG hybrid for a while... Something that is kind of crunchy, but a system that lets you do a multitude of things as a player. Text is the easiest way for that to happen, the most powerful graphics processor is going to be your mind and imagination. MUDs and MOOs are sort of the cornerstone of what that can look like in practice, but that requires the player to understand a specific set of syntax to play the game. While not too difficult to learn, it is a barrier to entry.

I began by playing around with Godot and the Ink scripting language. My thought was that it would be easier to start from scratch with Godot and effectively build a framework/UI for Ink and then maybe a separate combat system that shares some state with the Ink-based story scripting... BUT, I've also been playing around with how to run and train LLMs locally. I even bought an M2 Macbook just so that I could run a local AI setup. So far I've been successful in getting Llama 3.2 configured to act as a DM/GM within the 3.5e rules. Last night we tested it out.

The nice thing about LLMs is that you get the text parsing without the need for a uniform syntax that players need to use. They can just talk to it the way they would interact with a DM. And, I imagine this is probably what WotC is working on for D&D proper, but are layering GUI components on top of it to make it even more user friendly.

The AI does have trouble with ending the story/quest. And it seems like no matter how you try to resolve the story, it wants you to keep going. So these are some of things that I need to work on before trying to keep state.