Currently agent types are mostly useless property, as agents looks the same and this typing thing is not helping.
We probably need to think on adopting multiple runners/frameworks like llama-index, crewai, etc. BUT we need to have strong understanding why they are different.
Like there are classical prompt based agents, graph based agents of sort, and event based agents of sort. Maybe there are something else out there what we need to consider.
Currently agent types are mostly useless property, as agents looks the same and this typing thing is not helping. We probably need to think on adopting multiple runners/frameworks like llama-index, crewai, etc. BUT we need to have strong understanding why they are different.
Like there are classical prompt based agents, graph based agents of sort, and event based agents of sort. Maybe there are something else out there what we need to consider.