You drop all their knowledge into a file, like .txt. or .html and it uses RAG to learn everything in the file, so if it's a vendor things like shop inventory, prices, or if it's a storyteller/questgiver you can give it knowledge the player must uncover, etc.
It uses the LLM for natural sounding conversations and instructions ("speak in the style of..." etc.)
“event[0]” is a singleplayer sci-fi RPG where you communicate with an AI on a stranded spaceship, essentially a chatbot, to progress the game and learn its story. Your goal is to convince the AI to let you go home. The chatbot detects semantic tags, intents and sentiments from your input, then procedurally generate output in the form of text and game mechanics depending on your request, the context (what part of the game you’re at), and short- or long-term memory.
LLMs weren’t as big of a thing back in 2016 so this game used a rule-based conversation system. Even then, many people praised the AI for how natural and relevant its output felt. There was even an unintentional, secret ending in game due to a bug in how the AI is programmed. I think making a game like this but plug the chat interface into a LLM would create chance for more emergent gameplay and help the game feel more real.
Game: https://store.steampowered.com/app/470260/Event0/
Developer interview: https://www.gamedeveloper.com/design/q-a-ocelot-society-on-b...