Closed Joseph-Cardwell closed 1 year ago
Hey @Joseph-Cardwell!
The parent/child semantic distinction is the same as the plan/execute semantics in Langchain's latest agent.
Here planners can be children of another planner, like you noticed.
Tbh I thought of worker too but it seemed to capitalistic, I thought it'd be more fun to imagine them as parents and children.
I have something you might like... https://github.com/shaman-ai/llambdao. Check out init.py. The repo doesn't work yet but that's how I'm imagining future multi agent systems to be architected.
Hey @Joseph-Cardwell!
The parent/child semantic distinction is the same as the plan/execute semantics in Langchain's latest agent.
Here planners can be children of another planner, like you noticed.
Tbh I thought of worker too but it seemed to capitalistic, I thought it'd be more fun to imagine them as parents and children.
I have something you might like... https://github.com/shaman-ai/llambdao. Check out init.py. The repo doesn't work yet but that's how I'm imagining future multi agent systems to be architected.
I think the node concept is FAR cleaner and produces less counter intuitive artifacts.
Super stoked to build DAOs that execute functions through AI agents!
It seems to me the parent/child relationship isn't really serving a purpose.
Would removing that relationship and simply dealing with them as 'workers' that can have descendants and know if they are descendant, be clearer and remove unnecessary complexity?