Consider methods of directing the conversation flow. Like {{guidance}} or nemoguardrails.

It's great to see this kind of infrastructure work to integrate a variety of agents into whatever system you're building. Keep up the good work! I may crib some notes from this for use in non-Python projects.

Cool - if I'm understanding correctly it's basically an API between different classes of "agents"?

One piece of feedback: the project sounds useful once I read into it a bit more but the headline is confusing since "unifying" can mean many different things.

Hmm in what sense did you think it was meant? Curious. To me it seemed pretty clear but I'm guessing I'm just naive to some corner of the industry.

I had no idea what "unifying" meant at first. The best guess I can make is something that lets humans interact with AI and computer systems, which is super vague.

Maybe I'm just to used to all the marketing speak out there where "unifying" has been co-opted to mean basically nothing. e.g. "We unify technology with human potential!"

Good points. Perhaps “cooperate” is clearer? Or “Connecting”?

Tho at a certain point we’ve gotta reclaim the fun academic terms from the MBAs!

I think what's missing from the title is "actor framework" - omitting it is kinda burying the lede, since the main idea seems to be fitting AI and human workflow orchestration into an actor framework, which is a well known paradigm with a name that immediately evokes the intent of the library.

thank you for this feedback! I was considering something like that. I'm going to make some updates for clarity and mentioning the actor model approach more prominently will definitely be part of it.

thanks for that feedback. If you guys saw my messaging just two weeks ago it was even more obtuse...

I'll work on it some more :)

Awesome stuff - this is the kind of high-level thinking I love to see. I think your "do one thing and do it well" focus is commendable, and the subjective choices you've taken to get there are all A+ to me. This isn't turning out to be a very useful comment but wanted to let you know that someone who's also thinking about abstract frameworks for agent ensembles thinks you're right on the money.

A bit busy right now but will probably try to contribute if I can! I know very little about the low-level workings of Python and whatever the "GIL" is, but my single-process ensemble ended up using `AsyncIO` instead of `Thread` because of difficulty running a Flask server in a child thread; do you know if that would fix your CPU blocking concerns? Mostly asking so I know if I have the same problems and just haven't discovered them yet lol

One very subjective opinion: The README is great but very long - I would consider putting either/both of the code samples from the bottom at the top, so someone experienced with the basic concepts can take a quick look and understand the gist immediately. Your language at the top is of course all accurate, but it's such an abstract project that even an accurate description leaves a lot of questions, IMO.

P.S. I can't believe you got the github username "operand" that's sick

thanks so much! I went to sleep and saw that this post actually blew up a bit :)

I'm taking in all of this feedback. I hear you on the long readme. I'm thinking I'll break out the walkthrough into its own doc.

Regarding the multithreading. It's something I have to still look into. I'm not sure that using asyncio will get around it.

Commend the investment in time on this and I will see if I can contribute. From a product lens it's great to see higher level thinking on how to incorporate agents. Feels like we're moving beyond the 'cool demos' phase now.

Looks very nice and I adhere to the concepts quite a lot !

one thing you could add : tha ability to save messages sent throughout the Agency/system to a log. That would grant you instant replay ability.

100% The only reason I haven't yet is it just hasn't been important. Great first issue for a PR! ;)

Ye old blackboard.