Better together build and deploy to Azure with GitHub | BRK180

Key Takeaways

[Music] hello how's it going I'm Mandy wayy and I'm super excited to be here today to talk with you I lead our Azure Dev tools team and we build SDK libraries and tools that help any developer working in any programming language be successful on Azure I'm really excited to be here today with Jessica to share things with you about how you can use GitHub and Azure together and hi everyone my name is Jessica Dean I'm a Staff Enterprise advocate here at GitHub and I'm excited to be here with Mandy to talk about GitHub and Azure together kicking it right off we're going to talk a little bit about GitHub GitHub offers the best collaboration tools for you as developers it helps you combine built-in security and AI capabilities so that you can build apps quickly efficiently and confidently it's an Enterprise ready platform that helps you streamline your velopment process by providing everything you need in one easy place it also offers best collaborative tools and has security integrated at every step of your workflow and best of all as we'll see today it's powered by AI now with Azure Azure helps you bring your ideas to Life by leveraging a comprehensive set of tools and resources that again help you build deploy and manage your applications effortlessly it also helps you seamlessly unify your resources by providing an integrated environment that enhances your collaboration and your efficiency it helps you innovate on trust and on top of that it's going to help you again build and connect over to Azure and we're going to learn a little bit about that right now yeah when you're using GitHub and Azure together you get integrated developer workflows where you can access your GitHub resources your Azure resources right from within your IDE like Visual Studio or Visual Studio code we're also working to make it really easy for you to streamline deployments using GitHub access and azured together and then with GitHub actions and GitHub Advanced security you can build an incredible Automation and security at every step of your cicd pipeline So today we're going to take a look at this in to-end developer experience of how we start in our idees where you're coding where you can stay focused on your code connecting to our devops and automation tools like GitHub and Azure devops and then how all of that connects to Azure like I mentioned earlier we work on making this entire Journey successful for any developer in any language so if you're a python developer you can use the tools that you love and our libraries will feel pythonic and idiomatic to you and GitHub co-pilot is there to help the whole way we're going to take a look today at our devops life cycle and how you can use GitHub and Azure along each step of the way we're going to dive into the code phase build deploy and operate and we're going to do this by looking at an actual application we're going to build a rag pattern application retrieval augmented generation you might have heard about it in the previous session this lets us build an AI powered chat assistant for our Koso Outdoor Store this will allow us to build a chat experience where the information that the assistant's bringing back is grounded in our data so when you ask for help about tents you don't just get general information about tents you get information about the tents we have in our store and we're going to deploy all of this using GitHub actions and all of the things we're going to show today are available for you to try the codes on GitHub you can get started by going there and trying it out and we're going to show lots of demos as we work through all of these stages so next I'm going to turn it over to Jessica to start on the code phase thank you Mandy we're going to take a look at how GitHub and Azure work together when we're working with the code phase of our workflow now Mandy mentioned that we're working a lot on tools and how we can make tools working with GitHub repos and get as seamless as possible that's where visual studio and visual studio code comes in we focus on things like seamless authentication so that we can make accessing your GitHub repos from your IDE as easy as possible you can also use extensions like GitHub co-pilot and GitHub co-pilot chat again directly from your ID D directly from your editor without ever having to leave your workflow or your developer inter Loop and we are always working on how we can make our tools and common get workflows like create merge rebase one of my personal favorites of undo less commit without having to go to stack Overflow and look things up we want to make those as easy as possible and we can do that in Visual Studio code another challenge we're trying to solve is how we can make setting up your development environments easier when you're setting up your new systems and your development environments with binaries dependencies depending on the language and framework and versions you're trying to work with that can be very daunting it can be very overwhelming and then trying to scale that across your teams can feel almost impossible at times this is where GitHub code spaces comes in it is your compute in the cloud and one of my favorite ways of using GitHub co-pilot is also in this day and age when we're working with rag applications and large language models maybe the systems we're using aren't as powerful enough as we need this is another way where we can where we can leverage Cloud compute to be able to help run our development environments all codespaces really is is your development environment running in a container in the cloud and you can use your native extensions whether that's GitHub co-pilot GitHub co-pilot chat all of that can be done directly from within your GitHub codes space and gab co-pilot is your pair programmer it's always there to help with things like converting your comments to code Autos suggesting tests that maybe you're struggling with trying to figure out what makes the most sense or even showing Alternatives and different ways of doing things one of my personal favorites of using GitHub co-pilot is actually serving as an educational tool a little bit about my background I came from the kubernetes world I came from Helm I came from containers and when I started working even on this session I could barely spell llm I could barely spell Ai and I know that that's probably a common joke but I promise you I had so many conversations with co-pilot of what does this project do how do I get started what do I need and the best part is is it made approaching something that felt so big and so overwhelming and made approaching that easy and today I want to show you how easy it is to use co-pilot chat to use Visual Studio code and to use code spaces all right so I'm starting off where I'm in my repo here this is just a fork of the Koso chat repo that you can get started playing with today and all you need to get started with code spaces is your repo you would simply navigate over to code and if you're looking at it right off the bat you would see local where you could go ahead and clone or you could click on the code spaces tab now in my case I already have code spaces that are pre-provisioned in the event that I didn't I would see a big green button right here that would prompt me to create a code space alternatively another way I could click to create a code space is I could click this plus sign right here where I could create a codespace on my main branch or I could click the little dots where I can actually create a codespace configuration and this will help me set up what's called a Dev container this is where I get to start defining all the extensions all the binaries everything that I need for my application now as I mentioned I do already have a code space and I could choose to open that one of two ways one I can open it in the browser or I can open in my ID if I open it directly in the browser you'll notice it's Visual Studio code in my browser I even have custom dot files I'm super big in power level 9k 10K and I'm able to have access to all of my same development environment directly in the browser but it's running in the cloud now just to show you I'm on a Mac I can type LSB relase Dasa and take a look right here this is a Debian based container I'm using a Linux container on my Mac because again it's running in the cloud Now using Code spaces in the browser is awesome it makes it very easy to get up and running but I can also open it directly in Visual Studio code and you'll notice in Visual Studio code it's my same instance it has my same extensions uh a new extension that just got launched is prompt flow we'll talk a little bit about that later on I have prompt flow extensions provisioned I have GitHub co-pilot chat all of that comes ready to go and now as I'm learning how to get started with this particular project I can navigate to my co-pilot chat and I can ask it what does this project do now you'll notice that it starts off with at workspace and I'll go ahead and just zoom in here while it's thinking at workspace is taking a look at the files in my repo it's taking a look at my GitHub co-pilot worksspace to analyze what does this file do it's going to read the readme it's going to read the infrastructure files and it's going to make sure that the answer it gives me is grounded in real data I'm going to drag the chat over to the right hand side because I like to have chat up with my Explorer with my terminal I like to kind of be able to do multiple things at once and I can see that the chat has given me six references and if I drop this down it's using the files in the Explorer to give me an actual answer it tells me it uses the promp flow extension it runs in Visual Studio code it uses Azure AI search it helps me become familiar with this project one of the things I also like is it prompts me for other questions that maybe I didn't think of asking that's really powerful if maybe again I'm not even sure what to ask or how to prompt co-pilot now in my case I do know exactly what I want to ask I want to ask where is the code that manages getting customers from chat I'm curious how our chat application that we're working on actually builds and again it's using at workspace to take a look at where those files live what's in the repo and it's going to make sure that it provides an answer that's based on the data that's in this particular project repository you can see it's currently filtering to the most relevant information and scrolling up here it gives me live responses again it uses four references and I can see there's some python files in here but scrolling down it tells me the customer information is located in the get customer function in this particular file and if I click it and I'll go ahead and drop my terminal down I can see that I have it opens right up I can see the get customer function now another way that I like to use co-pilot is let's say that I'm new and I don't quite understand what this code does right from within the editor I can hit command command I or if you're on Windows control I and I can just use slash commands and say slash explain and then that way I don't even have to think of how do I phrase this how do I ask this I'm not quite sure what I'm even searching for I can just say explain what this code does and it gives me an actual explanation explaining that this is connecting to Cosmos DB it tells me there's an endpoint there's a key it shows how this works now I like to set the next person up for Success I'm going to go ahead and start adding a comment here and I'm going to say get and then just like that co-pilot is also going to help me do some of the more mindan tasks that I struggle with which is adding comments to code now that we have that set up I'm going to turn this back over to Mandy who's going to walk us through the build phase of our workflow thanks all right wait for the slide to get up so now we're going to take a look at the build part of our project and in this phase we're looking at how do we take that code that we just got working on our local environment which codes spaces and Dev containers were a huge help in making we had all the right dependencies and everything there but now how do we get that up and running in Azure in the cloud for the first time and this is important for our AI applications because a lot of times you need access to the cloud AI resources very early in your development process and this can be a part where there's a lot of questions how do I deploy this what services do I need how do I configure the services and provision them so one of the tools we've been working on to help with that is the Azure developer CLI it's an open source tool it's focused on a very um Dev Focus focused workflow with things like init provision deploy you stay focused on what you're trying to do and it helps you by configuring and provisioning resources and then deploying your code there's one command ASD up which actually combines a lot of those andit provision and deploy all in one step which can help you get your application up and running ACD empowers us to build endtoend templates or blueprints which is one of which is the example that we're using today and we have many more for different kinds of azure app patterns and AI patterns and these templates include things like the infrastructure as code file terraform or bicep the actual GitHub actions workflow files and the actual application code all together so you can get that up and running see it experiment with it start to customize it use it as a blueprint for where you want to take your application another thing that we're looking at is GitHub co-pilot for Azure so GitHub co-pilot for Azure is an extension for GitHub co-pilot and it can help you with Azure specific topics um you can use it right from within GitHub chat and you can ask things that help you manage your Azure resources explore your Azure resources get guidance learn about the latest documentation and features and also help you with troubleshooting and um deployment to Azure we're really excited at Bill to be announcing that this is in private preview we'd love for you to sign up try it out with us this is where we're going to be learning about what scenarios are the most helpful to you and building this together along the way all right let's take a look at some of these tools and a demo okay so I'm in vs code I'm working on the same codebase that Jessica is working on and because this is an Azure open AI related project one of the things that I need to know is what is the endpoint for my Azure open AI resource I may want to get this early in the process so I can start making API calls and at Azure can help me with this so because I'm logged into Azure in VSS code it's using my subscription information it's going out and looking at my actual resources and it brings back the URL in point for this Azure openi resource which I can start experimenting with um that's super helpful but I also may want to know more information about Azure open Ai and so I could ask you know what models does Azure open AI support right now and this is where we have an integration with Microsoft learn and this is great because it's going and getting the information from Microsoft learn that you know is always the most up todate and bringing back the the exact documentation that you're looking for and one of my favorite things about it is that it goes it finds the information it brings it back in and information you can just read in the chat but it also let that finish it's great you can see we've got the latest models in there and you can see that it brings back the actual references the links so if you want to follow those to go do deeper research or have those for a future reference you've got those right there to help you okay next we want to take a look at getting our application up and running so we're going to ACD up this is where it's going to oh hold on is it even a demo if there's not red text on the screen that's the way you know we're doing it live we're doing it live okay here we go one second I don't know why that's going so slow all right here we go all right so it's going to ask me some information it's going to ask me for my environment name I'm going to name this um MW cons chat it's going to ask me about the subscription that I want to deploy this into and it's going to ask me about the location the region that I want to create these resources in so after I give it all that information my project is initialized now I can start using um all of the different infrastructures code files can start configuring them and start provisioning our resources so you'll see it it's going to go along start putting all of that information together and starting to um compare what's currently in my subscription with what I have specified and start to provision those resources but we're in a demo and we don't have a lot of time so we're not going to wait for that because I also want to show you AGD pipeline config oops all right so AGD pipeline config like we mentioned before this is the one that helps you with your GitHub actions workflow when we cre uh created this template there is a GitHub action workflow called Azure devmo that's included this helps you deploy your application to Azure but if you've ever done this before when you're setting up your GitHub repository and your Azure subscription to work together there's steps that you have to do and you want to make sure that all of your secrets are stored securely encrypted in GitHub Secret store your variables are handed correctly so if we take a look right now in GitHub actions you can see I don't have any secrets I don't have any variables but if I run a pipeline config okay this is going to okay didn't we in it oh sorry my bad all right EAS pipeline config this is logged into Azure it's going to start to set up that connection for me so it was going to actually create the service principle it's going to take the environment secrets from my project and start putting them into the places in GitHub where they're stored securely stored correctly setting up that um for us so we can watch that for a second and if we are refreshing our secrets over here we'll start to see that these start to get populated as these commands run and this pipeline can fig is working because this is set up for templates right because it has that Azure yaml where it's it knows that this is an Azure application that's predesigned to work with GitHub exactly so it's starting to set these up we can start to see them populating we can see that our variables are getting put in and um the last thing that happens is the uh Secrets get put in so now our pipeline is set up we could start immediately pushing and going and um using our pipeline to deploy one other thing that I wanted to show is that um if I forgot all those commands or I couldn't remember how to use or find the information on ad co-pilot can actually help me with that too I can ask how do I deploy this application and because it has all these different artifacts and is set up as a template it will recommend that you can deploy this using ACD so it's going to bring back the information about using a knit and up and even has the connection to go straight to ACD pipeline config if you're ready to set that up so great I am going to now turn it back to Jessica to talk about deploy and show us some more more things with GitHub actions awesome thank you very much Mandy I love ASD pipeline config because I've been using GitHub actions for a long time and I have a collection of get secret scripts that set up service principles and personal access tokens and a bunch of things that are not secure and let's be real friends don't let friends use personal access tokens okay we are not in this world anymore that's last year we are now on the managed identity age we want to set up things with oidc and and Federated identity and that's one of the things that ASD pipeline config does for us in one command it's an amazing feature and it makes using GitHub actions so much easier than ever before and if you're new to GitHub actions or if you haven't gotten a chance to play with it it's the number one cicd platform on GitHub you can use it to automate customize and ex execute your software development workflows a lot of people hear cicd and hear GitHub actions and think it's for build test deploy and it is you are not wrong but it's also to automate any job and any task that you can think of and we'll look at that today because we've been playing with some new things the way GitHub actions works is it has you can write your own actions so it has actions that you can be able to predefine or steps that you can predefine but there's also a marketplace where you can go look at pre-built actions for you and this is important because when it comes to taking a look at automating our workflow processes Studies have actually found that companies that adopt this have 208 times more frequent code deployment 106 times faster commit to deploy lead times and my favorite statistic is over 2600 times faster recovery from incidents because we can be able to automate that recovery and fail fast now I mentioned that there's that the way GitHub actions works is it uses actions what does that specifically mean it means that it's automating steps that like I said are pre-built from a Marketplace in that Marketplace are Azure specific AC actions we call these GitHub actions for Azure and because they're Azure specific they're actually managed and maintained and owned by the Azure service teams so you can have confidence when you're using these actions in your workflows they're going to work they're regularly updated they're regularly maintained from Azure and in addition to everything we've showed you today with the Azure developer CLI there's even a brand new action called the uh the Azure develop or sorry the Azure open AI for chat action and that will also help you set up your workflow for your chat applications now with workflows I've talked about how you can use it for more than just build test and deploy one of the things that we've been playing with is what other jobs or workflows do we as developers have to do when we're working with llms and rag architecture one of the things we have to do is run our prompt evaluations what are prompt evaluations I'm so glad you asked prompt evaluations are a way that you can help validate the quality of your prompts when you're working with prompt engineering it helps you ensure that you're getting real answers and answers that are relevant and in fact actually do you want to see what that looks like I do and one of the things I really think is interesting about this is that prompt evaluations may be something new that you're learning building these kinds of applications but you're using something that you already know used for other things other kinds of automation GitHub actions yep putting those together all GitHub actions is is the steps you would do manually you're just putting it into a file to be able to run that autom magically now I'm going to go over into Source control and that does not seem to work in there that's fine we can use terminal old school this is again doing it live uh I'm going to go ahead and make sure that I add my amazing commit of adding comments we're going to go ahead and say commit DM add comments normally if this were working with get commit just to point out Source control is I can actually also use co-pilot to help me generate commit messages for me the shorter or the longer I've been working with my commits the shorter and shorter my commit messages get Foo and ARG and blah I can use GitHub co-pilot to help with that and it'll automatically read my changes and help me with a commit message in this way I have to do old school I have to YOLO and do it myself so I'm going to go ahead and hit enter which let's see if that works there we go and then we're going to go ahead and say get push and I'm going to make sure I push the correct Branch now again this is able to work and start firing off a GitHub actions workflow in part because we set up Secrets using the Azure developer CLI I'm also going to use a p I'm going to create a pull request I'm going to make sure I open this into my correct base repository so you can see a lot of people have forked this so we're going to go ahead and that is not let's see all right let's go ahead and refresh this one more time nothing quite like doing it live well you know what we're going to do we're going to switch to a failover video I know it's so sad we try to make sure that we get this uh up and running but let's go ahead and start from this current slide I want to make sure that I can show you always be prepared always be prepared is the moral of the story and let's make sure we get to the right part so we've checked in our code uh you can actually see co-pilot right here is helping us with a commit message and right in here I'm actually able to use also a pull request extension built right into Visual Studio code I can choose the correct base branch in this instance I'm also using co-pilot to help me generate a commit message where it's or not a commit message a pull request summary and then we'll go ahead and Flash Forward here so we can start taking a look at what these what actually happens once you end up pushing that it starts off where you're now seeing something that says two pending checks if I drop this down I can see that I have two workflows that just fired off I have my build and deploy workflow which is deploying my infrastructure out into Azure but I also have an evaluate workflow and this is that prompt flow evaluation if we take a look at what that promp flow evaluation looks like I can see it took about 9 minutes to complete and as part of any cicd I'm going to have build artifacts attached to that so moving let's see if I can move a little faster here if we take a look I can see that the 9 minutes and there's my build artifacts of this is my summary results but what does that actually mean do I have to download the build artifact every single time or is there a way to kind of publish a report where I can get real-time information telling me the quality the relevance the fluency the coherence the groundedness of my prompts so now that I can be sure and confident in our chat application that the answers we're going to get back are something that's going to be actually helpful yeah I love being able to build the summaries and have that so um easily understandable and the summaries are are marked down so I mean you can publish any information to it in fact on our build and deploy we're also publishing some deep information we're publishing links that's actually we can send those links anywhere but we have deep links that send us back over into Azure so we can take a look at our Azure resources and you can see that's loading right now we can see that with this is everything that we get deployed by these templates by the way 27 different deployments and we can take a look at our kosa web application which is opening in a new tab and when we actually communicate with our application how can I be helpful today I can say tell me about your tents we mentioned tents earlier right and we want to make sure that it doesn't get information from every tent in the world I want tents that are specifically sold from the Koso Outdoor Store maybe I mean we're in Seattle right now it's camping season we're in the summer maybe I'm looking for a new summer tent I want to be able to see the stars I want to be able to see the sky I want to be able to get ideas of what tents I can purchase from the this particular store I want a tent with a good rain guard yes especially if it was yesterday because it was raining uh let's go ahead there you go you can see that I get my response back I get there's a twers tent right here and in a second I believe my video will zoom in here as well and you can see that I have product IDs associated with this as well so now I know that these are products from the store it's giv grounded information and it's it's it's our data yeah uh now with that I'm going to actually switch back over to our PowerPoint here live redundancy guys on stage this is it all right so now that our application is up and running and our pipeline has been established we're we're ready to move into production and start thinking about day 100 one of the things we have to really think about and we should be thinking about from the very beginning is also security and I mentioned when it comes to GitHub we have security integrated every step of the way in in every part of the platform one of the things we do is we have code scanning we have secet scanning we focus on supply chain security with things like dependabot alerts and we've known that we have code scanning for a while now right it can take a look at our code it can make helpful suggestions it uses code ql it can help us address problems earlier in our development workflow but most recently we added the generative AI functionality to co-pilot to make it even better this uses generative AI to provide editable suggestions that fix detected vulnerabilities right in our PLL request and it helps you as developers understand targeted recommendations of how to get this out faster quicker and more securely that's amazing that it can help both detect the vulnerability and then also suggest the fix and then I can use co-pilot to just do the fix um and so that's amazing one thing that I was also thinking about is with our Koso chat application yep as we start to add more people have new contributors if they're not familiar with our best practices for handling Secrets if they don't understand how to use the GitHub Secret store like is there something that GitHub Advanced security does to help us with that there absolutely is I got you we have ai powered secret scanning and AI powered secret scanning actually scans for over 200 different token types and it's supported by a partner program of service providers to help you detect leak Secrets at scale and this is great because it can let you know when you do have a secret that's compromised amazing but what if I want to stop it from like ever even getting into to my repo can it help with that it absolutely can it has something that's called push protection which will also prevent you from pushing commits that have secrets or information that you don't want in your giab repo that you don't want Push it'll prevent you from pushing that in the first place and I can actually show you what that looks like can I see it yeah yes I'm GNA switch the right repo all right okay now we got it we can do text support so in the settings feature of your repository right under code security and Analysis is if you scroll down there's an entire section for secret scanning here now I want to talk a little bit about just a few of these things this is literally a you click it and you enable it and you're ready to go what that means is it'll automatically verify if a secret is valid it'll scan for nonprovider patterns you can also Define custom patterns but my favorite is this push protection that like I said is going to block commits that contain supported secrets and it's literally just a single button so what happens is when I'm in my edit and I I'm going to be the problem hi it's me I'm the problem I'm the developer that hasn't figured out managed identity and you know what that's a tomorrow problem in the meantime I'm going to use this credential today and you can see I have don't worry this credential isn't actually real but I have this credential that I've hardcoded because I'm just going to do some quick testing I'm not going to don't worry about SE Ops it's fine we don't need to do a SE four or anything or SE zero uh but what I am going to do is I'm going to go ahead and try to commit this and I want everyone to see this commit message if uh I think it should be do not try this at home it should be probably or it should be high it's me I'm the problem but what's the worst that can happen great we committed it now I'm going to go ahead and try to push and I mentioned earlier like red text and demos you never want to see that actually you do in this case I really do because GitHub Advanced security is going to save me from myself you can see that the remote rejected it it said it failed to push and if I scroll up it tells me the commit that it pushed it tells me the access token or the secret that it pushed and it also gives me the path of exactly where that lives so now it helps Empower me as a developer to know you know what I do need to follow Security's rules I do need to use Azure key ball or managed identity and not be using real keys and it helps you as the project manager and the developer also be confident that you don't have to worry about new developers coming in and risking leaking important secrets that's amazing I like how GitHub Advanced security kind of has your back on that it's great all right so we've had a great time today talking with you about GitHub and Azure hopefully you've learned some new things along the way hopefully we have it fun and authentic with live demos um we're excited to be here and we will be hanging out in the back with any questions that you might want to follow up on um some of the things that you get with working with GitHub and Azure together are like we said the integrated developer workflows code spaces co-pilot everything right in your IDE and editor so you can stay focused on your code and then the Streamline deployments bringing together GitHub actions and Azure tools like the developer C ASD pipeline config the provided actions for you to help make that easy to bring together also the GitHub actions for Azure which are out of the- box actions you can use and then all the great things that you can build into your pipeline through automation like prompt evaluations taking GI up actions using them for a new purpose so you always know how your prompts are performing and then also the code scanning and GitHub Advanced security features so like we mentioned at the beginning everything's here for you to try out um you can follow these links to get to the samples to explore more application templates that we have sign up for the preview and explore more about co-pilot and security and um before we end I really want to thank all the amazing teams whose work is in this talk from both GitHub and azer and all the teams that have made all of this possible it's a really great set of tools and we hope that you check it out out thanks for being here [Applause] today e for