The Future?

[rwjdk]

Hi. Is it possible to get some sort of summary of what the merge of SK and Autogen means and what the future hold?

I've missed the last few Office Hours (and recording are not shared anymore 😕) but it sound like the following is happening if I understod correct:

1. SK will not go away but are entering some sort of maintenance mode?
2. The SK and Autogen teams are working of a new framework to "replace" both? (Does this new thing have a name and a repo we can follow)
3. Assume the new Framework will support both Python and C#

Is above correct or have I missed something?

[moonbox3]

Hi @rwjdk, thanks for your questions.

1. Correct. SK will continue to exist - it's a GA SDK that we will continue to provide support on.
2. Correct. This is in flight - no name or repo to share yet. We'll have more to update in the coming weeks.
3. Correct.

[brainmurphy]

@moonbox3 Where can we read about this? I'm trialling Semantic Kernel here, so very interested in its future.

[brainmurphy]

@moonbox3 Now that I think of it, where does this leave fetures that are not GA yet? Specifically, the Agent Orchestration features, which are labelled as experimental / "under active development"?

[rajyraman]

@moonbox3 - Can you please clarify the first response. Are you saying that SK is entering a maintenance mode i.e. only security patches, no feature releases?

[grafanaKibana]

@moonbox3
To be honest, it’s a big disappointment, especially for someone like me who advocated for using Semantic Kernel over alternatives like LangChain. We invested time and effort building on it, shaping our architecture around SK’s capabilities.

It hasn’t been long since GA, and just as we’ve reached a stable point, Microsoft announces it won’t be developed further. That’s frustrating, not just for the wasted effort, but because it raises doubts about the long-term viability of anything new we might adopt. Why wouldn’t the next framework be dropped in a year, too?

[rwjdk]

@grafanaKibana Is this criticism not a bit to premature until we see the new Framework? It is still AI and it is still the same concept so migration might be as easy as replacing namespaces (and even if it is not, it is properly not too big of a deal to migrate)

I for one like that we are getting a new framework and hope it is not just a 1-1 move, as you can feel the pioneer spirit of SK and the bloat the need for backwards compatibility have given; especially in such a new field (heck: there is like 5 different ways you can make a chatCompletion in the current framework)

So IMHO we should instead look forward to a more modern and lean framework for the future 🤞 and praise the SK/AutoGen team for daring to take this move for the betterment of the usage.

[deepinderdeol]

@moonbox3 more details will be good, not only on the new framework but also recommendations for projects that are using SK.

• Should such projects switch to using Microsoft.Extensions.AI APIs ?
• Are features like function filters etc. supported by extensions AI?