Microsoft Semantic Kernel - Build LLM Apps in C# DOT NET and Python

Key Takeaways

hey what's up coders welcome to one little coder Microsoft has recently launched an open source package called semantic kernel it's like a framework which helps you integrate large language models within your existing applications for example let's say you've got a gpd4 use case and you already have an application for example a calendar application how do you efficiently integrate the gpt4 inside your calendar application is what you can use semantic kernel for if you heard this and then you thought oh is this related to Lang chain or is this like GPT index exactly Microsoft is coming up with a similar framework what Lang chain has been doing in the python world and Microsoft has started with C sharp I don't know who uses C sharp like I'm not quite familiar with C sharp at this point but it looks like Microsoft is starting with C sharp keeping Microsoft developers in mind and also semantic kernel is coming soon for python so in this video I'm going to show you certain details that I know about semantic kernel I have not tried semantic kernel yet I'm not a c-sharp developer but I'm going to tell you everything that I know about semantic kernel including they've got a free LinkedIn course and how you can build your skills on top of it to start with couple of days back Microsoft put out a block opposing hello semantic kernel and basically what they are saying is semantic kernel is a lightweight SDK like let's say framework that lets you mix conventional programming languages like C sharp and python with large language model AI proms with what kind of things that they can do they can help you with prompt templating which is something that long chain already has got now what is a prompt templating so for example if you want to ask a question why is this why let's say a particular product is bad now instead of having the product hard coded you can have a prompt template in which like you can use like very similar like what you do with if string in Python you can do prompt template there so now that is prompt templating you can do chaining combining multiple responses training is more complicated than multiple responses just planning capabilities so prompt here semantic kernel can help you with prompt templating chaining planning multiple capabilities that's what they have said so now what does it help you do it helps you do certain things for example you want to summarize a Lindy chat that is already on um you know like some platform you can do that or you want to flag an important Next Step that is in Microsoft graph you can do that if you want to plan your next trip along with booking an air ticket you can do that so if you see this is like an automation I don't know how many of you actually know about if this then that there is a very popular application like back in the day um it's a it's an application where you can set rules if this then that so you can set like an automation rule like very similar like with this appear but on the free app version so what you can see here what Microsoft is trying to do is it's trying to bring together that kind of capability within the large language model space for existing developers so they want to capture the existing developers like C sharp developers or python developers who want to integrate large language model capabilities within their application now specifically talking about semantic kernel what is it central to semantic kernel that like the most important thing for semantic kernel is something called a skill now what is this skill going to do the skill can be either of these two so the skill can be a semantic code or a native code now what is the semantic code a semantic code is when you have a large language model prompt an AI prompt for a large language model is semantic code and a native code is like a normal computer code so now a skill is like combination of these functions and these functions could be semantic code or a native code and this combines and then creates the collection of skills the skills that are going to be used for everything else in this now that is skill now the next thing is memory memory is basically context when you have a conversation with charge GPT does charge GPT understand what you said at the start so that context is what memory is and connectors like for example you have got slack in your company you want to use slack data you want to use Google Google sheet data so these are connectors so they are providing you memory connector how you connect this with skill and finally they have something called a planner and we can see see from this picture also they have something called planner now what is a planner planner facilitates you to do complex tasks by taking a user ask and then translating it to skill memory connector so like user would come and say you know I want to do this that that statement from the user is now split into skill memory connector with the help of planner and that's what planner is doing and semantic kernel currently supports even the latest gpd4 and also Azure open a service again I have not tried it myself but it looks quite promising from what is Microsoft doing it almost feels like Microsoft is in like some crazy Nitro mode if you if you're familiar with car race to just ship something every single day and capture this entire large language model so they almost did not capture the browser model or maybe browser was forced on people they had a good operating system but you know still a lot of people love Mac but still Microsoft operating system is quite big they did not do well on the laptop Hardware side they did not do well on the smartphone Hardware side they did not do anything at all on the I mean there was Windows OS but they couldn't capitalize a good Market in the smartphone OS side and search engine we all know it looks like Microsoft wants to really establish themselves as the AI leader in the in the space current space that we have got if you compare it with companies like Google Apple Amazon like the fan companies meta is quite good but generally if you see so semantic kernel looks like your Lang chain equivalent if you are in the Microsoft ecosystem or if you are in let's say C sharp to start with but like I said it's not only C sharp but at this point there is a there is a python preview that is available right now so you can see it's a very simple python code so you can see it's um you import semantic kernel and then use import async IO create the kernel and from there you add some configuration about which model do you want to use um for example fine tuning model like the base model DaVinci which prompting model you want to use open a key after that you create the prompt which is very similar like your prompting template and then you are going to create a semantic function as you can see this is a semantic function it's not a native function so native function would be like simple python code this is semantic function and then you are going to say what do you want to summarize because this is a summarization script and then you run everything and then you get the result so now this is a very simple example of how you can use semantic kernel for a summarization task with open a open GPT 3 in this case it takes wency over 3 GPT 3.5 now this is also coming and they are going to it's currently it's called preview but it's not finalized but this is currently available so semantic kernel is available for C sharp users python users now if you want to get started with semantic kernel I would strongly encourage you to go read this blog post but again they have come up with this LinkedIn course for free I mean if you do not know Microsoft owns LinkedIn so they have come up with this LinkedIn course where you can learn introduction to semantic kernel who is semantic kernel for what is semantic kernel why do you develop something on semantic kernel rather than you know going native getting started with semantic kernel and then some example sample examples I have not taken the course yet but you can actually see that there are already 513 people actively taking this course and uh 221 9 291 people have bookmarked there is an interest in this course and if you if you are let's say a C sharp developer like a DOT net developer then you should definitely look at semantic kernel so overall if you're a dotnet developer if you're a python developer it's really good for you to get familiarized with this particular World which is a large language model world and a lot of people have been seeing the benefit of using Lang chain as opposed to directly doing something with open a so it provides a lot of this what Microsoft is giving so if you want something alternative to Lang chain or if you want to look at something new from a big corporation like Microsoft definitely this is an open source package but you can bet on a development from a company like Microsoft so if you are looking for this kind of large language model based EI application or if you already have an application which you want to include large language model in it to be powered by AI then semantic kernel is something that definitely you should check out I will link all the required links in the YouTube description I hope this video was helpful to you in learning what does Microsoft schematic kernel and how can you use it what do you do with that if you have any question let me know in the comment section otherwise I will see you in another video