Idea: Separate the events memory and chat memory and add time
Could try separating the events memory and chat memory. So that conversational memory could remain for longer than the server events, or vice versa
Would be better to add a timestamp to each message (more tokens), since they will be added / removed separately from memory
Might also adjust one to be bigger than the other
However, GPT might just respond to the last message, either events or chat
Maybe there could be a small mix of events and chat, which would later move into memory behind