Using ACE Framework

[ScottzCodez]

You should consider using the ACE Framework here on Github, to keep the agent Aligned, so it will never go astray.
ACE (Autonomous Cognitive Entity) repository https://github.com/daveshap/ACE_Framework

[DataBassGit]

Hi Scottz,

We are part of the team that worked on this project. We wrote the AceAF demo, which you can see here: https://github.com/daveshap/ACE_Framework/tree/main/demos/AceAF

In our demo, AgentForge is a subcomponent of ACE. We found that ACE takes a lot of tokens to run properly. When inference speeds go up, and token costs go down, it will become an interesting system. Until then, there's just too much overhead to be practical.

In the meantime, you can use a system like ETHOS to keep your agent aligned It sits between the LLM and the memory system to check and make sure an agent follows a specific code of ethics. In this case, it uses Dave's Heuristic Imperatives: https://github.com/anselale/ETHOS