Smart Agents

[Cormanz]

This aims to resolve #25 by proposing a new Smart Agent solution called Adept.

Dynamic Agent

The Dynamic Agent is similar to the base REACT (Reason-Execute-Action-Criticism-Thought) loop that SmartGPT already employs.

The key differences are:

• It doesn't specify direct tools, instead specifying an instruction for the Static Agent.
• It (will) self-evaluate itself periodically to maintain focus.

Static Agent

The Static Agent plans out each tool and its purpose that would be used to complete its instruction. Then, it will traverse the plan until it is done, one step at a time, one tool at a time.