Semantic Kernel

#1

Post by overbytecode » Tue, Nov 28, 2023, 2:03 PM UTC

Re: Semantic Kernel

#2

Post by ashvardanian » Tue, Nov 28, 2023, 6:49 PM UTC

I'm surprised by how uncommon SK is compared to LangChain. Microsoft is very active in the space and has a few other related LLM frameworks of different nature - Semantic Memory and Guidance.

Re: Semantic Kernel

#3

Post by __loam » Tue, Nov 28, 2023, 10:04 PM UTC

ashvardanian wrote: post #2

I'm surprised by how uncommon SK is compared to LangChain. Microsoft is very active in the space and has a few other related LLM frameworks of different nature - Semantic Memory and Guidance.

Well you see, Langchain rolls of the tongue better so it has more broad appeal despite being a pile of crap.

Re: Semantic Kernel

#4

Post by nswanberg » Tue, Nov 28, 2023, 10:12 PM UTC

ashvardanian wrote: post #2

I'm surprised by how uncommon SK is compared to LangChain. Microsoft is very active in the space and has a few other related LLM frameworks of different nature - Semantic Memory and Guidance.

Seems nice if you're using c# or java. It also supports python, but for that Simon's llm library is nice because he designed it as both a library and a command line tool: https://github.com/simonw/llm

Re: Semantic Kernel

#5

Post by ren_engineer » Tue, Nov 28, 2023, 10:19 PM UTC

ashvardanian wrote: post #2

I'm surprised by how uncommon SK is compared to LangChain. Microsoft is very active in the space and has a few other related LLM frameworks of different nature - Semantic Memory and Guidance.

They've also got Autogen and Promptflow, seems like Microsoft just has random teams doing their own thing with lots of overlap

Re: Semantic Kernel

#6

Post by outside1234 » Tue, Nov 28, 2023, 10:33 PM UTC

ren_engineer wrote: post #5

ashvardanian wrote: post #2

I'm surprised by how uncommon SK is compared to LangChain. Microsoft is very active in the space and has a few other related LLM frameworks of different nature - Semantic Memory and Guidance.

They've also got Autogen and Promptflow, seems like Microsoft just has random teams doing their own thing with lots of overlap

And all are of dubious quality as well

Re: Semantic Kernel

#7

Post by gtirloni » Tue, Nov 28, 2023, 11:43 PM UTC

Gone are the days of seeing "kernel" in a title and getting juicy OS development content :)

old fart signs out

Re: Semantic Kernel

#8

Post by thund » Tue, Nov 28, 2023, 11:45 PM UTC

gtirloni wrote: post #7

Gone are the days of seeing "kernel" in a title and getting juicy OS development content :) old fart signs out

the days of coding are going to go :-) from assembly to javascript to AI... where no man has gone before

Re: Semantic Kernel

#9

Post by MattEland » Tue, Nov 28, 2023, 11:51 PM UTC

I'm using Semantic Kernel to build more complex AI systems. The technology is still maturing and you deal with things like missing abstractions and the like. The planners are intriguing and promising, and they'll be more manageable once assistants are available to make things more modular. Still, it's the best thing I've seen for building maintainable generative AI solutions.

Re: Semantic Kernel

#10

Post by ycg » Wed, Nov 29, 2023, 12:42 AM UTC

ashvardanian wrote: post #2

I'm surprised by how uncommon SK is compared to LangChain. Microsoft is very active in the space and has a few other related LLM frameworks of different nature - Semantic Memory and Guidance.

Semantic Memory (renamed to Kernel Memory - https://github.com/microsoft/kernel-memory) complements SK. Guidance's features are being absorbed into SK, following the departure of that team from Microsoft. Additionally, we have TypeChat (https://github.com/microsoft/TypeChat), which aims to ensure type-safe responses from LLMs. Most features of Autogen are also being integrated into SK, along with Assistants. SK serves as the orchestration engine powering Microsoft Copilots.