The LLM Adventure Game Framework (LAG) is an infinite text adventure game framework that uses an LLM agent to deliver incomparable gameplay flexibility and player agency, built with Python 3 and Godot 4. By leveraging both Large Language Models and Image Generation Models, the framework responds to open-ended player input with unique, contextually relevant text and visuals which creates a truly personalized gaming experience in the spirit of late-era text adventures.

This project was inspired by studying the history of video games as they exist today. They began as simple text only adventures with obtuse command systems and lacking graphics entirely. From there, games rapidly progress with increasing graphical fidelity landing on the hyper-real systems we see today. Somewhere towards the beginning of that graphical tidal wave were games that married the free form text input with set commands and the graphical systems familiar to those found in point-and-click adventure games.

LAG aims to fit into this niche of game. While generative systems today are not able to carry the entire cycle of a video game its own, when these systems are combine and unified with purpose built code then a truly freeform game can emerge.