It seems pretty obvious to me that you chain these guys together. Personality module, planning module, decision making module, internal monologue module.. boatl

It seems pretty obvious to me that you chain these guys together. Personality module, planning module, decision making module, internal monologue module.. boatloads of them all blabbering to each other. How you connect them to each other is an architectural question. My easy experiments show that it doesn’t seem so hard, but what do I know. Sure these guys don’t have memory, but I have one “storage” LLM that accepts vague requests and turns them into SQL, another cleans up the SQL so on and so forth, you need a handful of them each step. Now you have a database. Etc. Does anybody does this?