Model catestrophic memory loss

So this issue is a well-documented issue with LLMs if you continously prompt a model and the history buffer starts to overflow. Perhaps a feature could be added where we can manually record important events in journal so the AI does not forget. I.e. if I killed a character 3 days ago the engine should not assign a quest again where i need to meet the character. This journal can then be incorporated into the state engine or context window.