Accelerate software delivery with Gemini and Code Transformations

Key Takeaways

[Music] all right welcome everybody to our session on how you can accelerate software delivery with Gemini and the power of cloud workstations and and uh my name is Marcus graa I'm a product manager within Google Cloud focus on Gemini code assist and I'm also joined by uh devian shat who is another product manager within the Cod team as well as Christian goru product Man U from from Commerce Bank who'll be talking about their uh experience using both Gemini Cod uh code assist and Cloud stations on their development uh Journey all right so rough agenda for today I'll start by giving a high level review of what is Gemini code assist and then talk a bit about how AI geni can help developers under day-to-day development uh life cycle then I'll pass over to Divan who'll be given a deep dive on how we can use code Transformations followed by a live demo of Transformations using Gemini 1.5 with 1 million tokens and you close up with a testimonial from Christian on the Commerce Bank Journey using J code assist inside a cloud work state so let's start so what is Gemini code assist at a very high level Gemini code assist is your AI powered uh developer collaborator Gemini code assist is the evolution of jet AI for developers it is now powered by the latest Gemini uh models so you can essentially expect to get Gemini models for cenation for chat for our most common use cases and also as usual provides the Enterprise capabilities our customers need such as IP protections governance and security controls to Ure you're not only enabling your developers but also meeting your compliance and security guard rails and since our launch we're excited to say that we now have our launch in December 2023 we're not excited to say we have 5,000 business businesses which are actively using Gemini Cod assist on the day today many of you likely are here in the audience so I also wanted to take some time to thank you for all your trust as well as for the great feedback a folks did across uh this journey right so okay so a core principle we follow within Google Cloud developer products is meeting developers where they are right so we all know developers have a myriad of tools they use such as IDs like vs code or intj cic cd2 po request systems the key idea here is we follow the same principle within Gemini code assist and we do so by supporting the most popular IDE and code editors our developers uh use such as Visual Studio code intellig pie charm Cloud work stations Cloud Chell editor and many more with Android Studio coming soon on our pipeline beyond that one key approach which we take within uh Gemini code assist is should take a holistic uh AI assistance approach across the software development life cycle the key idea here is software delivery is not just about writing code in the code editor there's a whole Loop of steps which go from conception to development to testing deployment troubleshooting and maintenance and you can see here how Gemini code assist thinks about each of those steps and provides assistance on steps such as design you can essentially imagine a uh chatbot which allows you to brainstorm and prototype ideas for your new application yet to be conceived you can think about build like C Generations helping as you're writing code testing for example helping ass getting assistance with creating unit test to ensure a high test uh coverage followed by deployment let's say you use terraform or kubernetes resource manager or g-cloud Gemini code assist is fine you know those languages CH sure can also help with your infrastructure as code setup in addition to troubleshooting with capabilities such as Gemini in Firebase or Gemini in Crash lytics you can get assistance with things such as helping understand your crashes anrs and troubleshooting issues you may have with your application right so the key idea here is each of those steps have an opportunity to create friction points within your development life cycle we see essentially here a larger opportunity to enable developers across each of those uh steps right another point I want to mention here is Google was the very first Gemini code assist customer from even before Gemini code assist was broadly available Gemini was approved for production usage within Google with thousands of developers using for day-to-day development this means essentially being able to get a much tighter developer feedback loop with the engineers using that for the day-to-day development and providing feedback on both developer experience as well as providing data points to confirm that you're achieving the actual productivity and uh Effectiveness gain you want to have with such a solution and now the internal usage of Gemini code assist helped essentially improve uh Gemini both in terms of being Enterprise ready as well as optimizing its developer experience to dive deeper on that Google has been spending uh at this point decades investigating ways to optimize developer productivity developer satisfaction and build happy and satisfied teams you can see here some of the Publications which Google published uh recently such as our annual Dora report our softare engineering at Google book and our Sr book the key idea here is you can see Gemini code assist as a delivery mechanism to bring all the capabilities which were validated tried and tested within Google developers uh and ab tested to ensure that we are providing more mature and more tailored product to our development uh teams and moving on I wanted to quickly talk about some new capabilities which were announced uh this week at Cloud next which are now available within Gemini code assist first one is Gemini Pro this is now generally available to all users if you install and use Gemini code assist right now you'll be getting Gemini Pro 1.0 for chat use cases providing you uh assistance being bringing the power of Gemini directly to your code editor with the context of your local files the second capability which you can hear more about on our customization sessions uh tomorrow is code customization the key idea here is a hybrid approach to make Gemini code assist more contextual to your uh application we do that in two ways first one is expanded local context the key idea here is beyond a single file Gemini code assist now cross the entirety of your local workspace and fetches files that are relevant to the task you have at hand let's say if you're writing unit test it you fetch the content of the file which has the unit test even if the file is not open on a separate tab so this is essentially about ensuring you're getting you're passing the right amount of informations to Gemini to ensure it is providing more accurate and more comprehensive assistance Beyond local customization we also now have remote context essentially currently in private preview which allows you to go beyond your local work space and index the entirety of your code bases even if they're private so let's say you have 50 different repositories with lots of private Frameworks being defined the key idea here is allowing Gemini to ground the assistance on this private code allowing for both more accurate uh AI assistance as well as to gradually nudge your developers towards the best practices by essentially incorporating the best practices which are hardcoded within your company uh code that's two and three is code Transformations that's the main focus of today's session code Transformations the key idea here is focusing on existing code code which has already been written allowing you to select a piece of code and use natural language instructions to modify the code in a given manner for example adding comments to code or make code more readable or uh troubleshoot code which may have low uh bad latency or fix code which has issues that's the key idea of code Transformations which particularly become accentuated once you think about uh using Gemini 1.5 with full code based awareness allow you now to take into account not only a few files but the entirety or a good fraction of your code uh Repository all right let's now shift let's talk about how can uh these capabilities I mentioned before help developers in their day-to-day so when you're building Gemini Cod assist the first thing we did was talk to a extensive number of customers to understand what were the friction points holding them back from having their optimal software delivery velocity and essentially we saw a large number of uh issues categories of challenges those were the first categories is so more often uh being uh reported the first one is this low develop developer onboarding let's say you're developer joining a new team or you're a developer that's unfamiliar with a code base or a developer trying to call a new API for the first time we see here a long time in some cases days spent trying to understand a specific piece of code to a point where you're comfortable Reading Writing and reviewing code for that code base that's the n AI being able to help accelerate that process the second is excessive contact switching so we likely all been there you have 50 tabs open 10 on St overflow five Google search 10 on dogs and then having to juggle between all those different information sources this all creates cognitive overhead which takes you away from the focus on writing your core Logic for a business case you want to solve the key idea here is by bringing all that information to a single surface essentially chatbot within your uh code uh ID or with as in line code comption code Generations we can minimize all those sources of uh uh multitasking and Contex switching the third Point here is high maintenance cost and Tech debt an important point about uh delivery velocity is not just about writing code fast today but having a code base where you can consistently uh maintain and evolve over time one key thing you see here is code bases tend to grow in size and in complexity and Comm pattern you see is uh Tech de essentially code which helped you move faster today but over time starts building a drag productivity we also see here AI by virtal of making things such as improving test coverage or helping better document code being a a lever which can help you minimize this issue and finally excessive repetitive tasks such as writing unit tests writing comments those are all simple tasks which you have to do but given they are well structure are a good fit to be automated or stream line by AI assistance tools let me now talk about what a solution on this space means right so here have a case study from Wayfair one of our partners and customers who have been using Gemini code assist on development teams you can see here in numbers the kind of impact they experienced the first one is fast developer on boarding essentially environment setup in their case so their developers can more quickly get a Dev environment set up a big thing here is by making test generation easier you can have an increased test coverage in their case 48% more tests on their AB experiment and finally not only developers writing code faster are being more productive but more importantly developers 60% of the developers reported being able to focus on work which they consider to be more satisfying work that's also a very interesting statistic which I wanted to share here and moving on here second thing we did is we also did some controlled experiments the key idea here was pick specific control tasks which are representative of those friction points before and run a controlled AB experiment where one cohort gets no AI assistance and the order cohort gets AI assistance what you're able to see here is consistent ly for those specific tasks AI assistance helping uh helping improve and accelerate the time to completion by numbers around the order of 20 to 50% depending on the task so we have here for example modifying existing code writing brand new code or getting help with understanding code or writing unit tests those being tasks where you can see code assist helping you on your day today and finally here one thing I want to dive deeper here is enter price Readiness so those are some of the measures we took within Gemini Cod assist to ensure Gemini is ready for the production usage a common question you get from customers is Will Gemini code assist or the underlying models be trained on my code or on my responses and the short answer is no we do not train our models on our customer data we do not even store the data used for the prompts Gemini Cod assist is strictly a stateless service the data comes in we St in memory of course for serving but it don't persist the data on any of our back ends the key idea here is to further increase uh data leakage protections to not even persist customer data in our backends the next thing here is IP protection our foundational models are train or permissively licensed uh code and even when a response Gemini provides citing its sources we provide what we call as recitations which tells you where that uh which parts of the code are matching our training data set the source the URL of that code as well as the license it falls under so let's say for example if you're getting code which happens to recite take overflow or GitHub you'll be able to know where it came from to take a more informed uh decision about whether or not to incorporate the code and provide the license attribution for the case where it's needed you can also go as far as enabling blocking of all those citations to take a more aggressive posture in terms of Ip protection and that's it for the high level overview I I'll hand over to uh Divan to give you folks a deep dive on code Transformations on gini code assist thanks Marcus good evening everyone my name is Divan chhat I'm the product manager on the Gemini code assist team now developers who have been using Gemini code assist have already seen great improvements in their productivity when it comes to like doing repetitive simple tasks like creating a function or documenting a code but what about some of the tasks that needs to be done like it's complicated for example moving a legacy Java 8 to 21 or creating test cases for a backend service so today we are announcing a new feature to Gemini code assist called code Transformations it's it's publicly available in Cloud workstations and Cloud shell editor now with code Transformations you can provide or ask Gemini natural language prompts for analyzing refactoring or optimizing your code now we have been asking developers what are some of the pain points and one of the things that they have mentioned is that it's disruptive when they are in the developer flow and now they have to ask Gemini questions well we are also introducing incode transformation is the inline text box that helps you ask Gemini questions right in your code doesn't disrupt your development flow at all additionally we are also introducing Smart Suggestions smart commands that's going to help you quickly get started with certain tasks like generating test cases or explain the code and don't have to figure out what a natural language prompt would be but we wanted to take this one step ahead today we also adding the support for full code based awareness that is backed by Gemini's 1.5 models that includes industry standards and industrywide 1 million tokens Now 1 million token in perspective is 30,000 lines of code that's a lot now coupling this with code transformation you can perform complex tasks not just on a single file or a selected piece of code but on the entirety of your repository available in preview on cloud workstations and shell editor and we'll have the details on how you can enroll but how about let's look at this in a live demo so here I am in my favorite ID Visual Studio code that is running on cloud workstation and today I'm going to assume a developer persona as a developer I've just inherited this repository I don't know the contents of it what are its functionalities so why don't we ask Gemini now as Gemini is doing its magic let's go ahead and quickly glance over the repository I can clearly see it's a java8 sample file it has some of the features that were introduced at that point in time Java operator nason's JavaScript engine stream a util do date so on and so forth let's see what Gemini has returned back and perfect it's confirming my assumptions the repository contains a mix of java and JavaScript files demonstrating various features of java 8 and N nashon JavaScript engine but it didn't stop there it has also gone ahead and given me the breakdown of each and every file and its functionalities isn't this awesome also it's given me a breakdown of JavaScript files too and what we did here is we sent the whole code base down to to Gemini it looked at the whole code and sent me this analysis but did I mention it's Java 8 isn't it really old what about we convert some of the files here from java 8 to 21 let's look at one of the folders Lambda it has three files Lambda 1 threading Java interface in Lambda 1 I can clearly see that it's using some of the deprecated AP I util do date similarly in threading I can see that it's using suspend and stop deicated methods what if I tell Gemini to convert all the files in this Lambda folder from java 8 to 21 and also I'm going to tell Gemini to provide me only the refactored code and the reason why I'm saying this is because the interface 1.5 interface 1. Java doesn't need any refactoring it's adhering to Java 21 standards let's give it a few more seconds and there you go Gemini has given me the refactored code for Lambda 1 threading Java and it's also given me the explanation at the bottom that basically says what key changes it performed on those two files it replace date with local date amazing right and threading Java it basically replace the synchronized block with re-entrant lock it also repls the Stop and the suspend but do I trust the refactored code that Gemini has provided does it alter my functionality of the code let's put that to test as well so here I have Lambda 1. Java file I'm going to quickly run this file and what this file does is lists down the presenters in the natural order um Cloud next date and Google phones and its price points let's go ahead and apply the refactored code that g had provided me and I'm going to go going to go ahead and run this file again and there you go the functionality has remained the same it has gone ahead and updated the deprecated class that I was using with local date and I can do this same process for threading Java as well imagine the amount of time that Gemini is saving and improving my productivity I can do this for all the folders in the repository in fact I can do this for the whole repository itself isn't that great now let's do one thing let's switch to a different kind of an application a web based application built on python you're going to still assume a developer Persona and I'm again inheriting this repository now I can clearly see that there is no documentation the readme file is missing so how about we actually go ahead and ask Gemini to generate a detailed readme.md file that contains information on each file and let's say couple of ways of deploying this application oops deploying the application locally now as Gemini is doing its magic let's go ahead and glance at the repository so it I can see it has a backend service a front end service front end has some static files HTML so on and so forth back end has a back dop that has basically Services let's see what Gemini has returned back and awesome Gemini has given me the file breakdown what the functionality of each each of these files are how to run the application locally a step-by-step approach and also it's recommending me how to leverage cloud data store just in case if I want to do that I like this readme file I'm just going to go ahead copy the contents of this and dump it into a readme file now of course not perfect but I can clearly see that Gemini has oops not here let's do it here and I can copy the content here of course not perfect but we can itrade on this and add more content but let's dive deeper into this documentation use case when I had opened the back. pi I saw that it's missing API docs it's one of the ways of understanding what are the services in a particular python web application how about I ask Gemini to create API docs for all the function in this file and in order to do so I'm going to use an inline text box how do I trigger that command I on Mac control I on Windows it opens up a text box right in my development flow I'm going to say to Gemini add API docs for all the functions in the file let's give it a couple of seconds with inline textbox you also get the capability of a diff view you can quickly understand what your original code was against the refactored suggested code from Gemini and and I like what I'm seeing here it identified get messages and add messages as a function it also identified that add message requires expects but not in get message so I like this I'm going to go ahead and approve these changes and there you go it's that simple now let's go ahead head and take a tester's poner let's be a tester for a day for a tester one of the important task is to create a test plan in order to do so you've got to understand what the web application does what are its functionalities the backend the front-end Services what about I ask Gemini to actually create a test plan for me create a test case plan for the web application in the repository and here again we are using the full codebase awareness that that uh that is supported here with Gemini code assist let's give it a couple of seconds and there you go we have a test case plan created for the front end service and the backend service but it didn't stop there it's also given me test cases for front end and specifically how to display the duration on the on the web page for the back end it looked at those two functions get message and add message and given me test case scenarios for those two additionally it's also given me recommendations on what tools and Frameworks to use P test also it identify that I'm using mongodb in the back end for database and therefore it's recommending me mock as the mocking framework to mock the DB operations isn't that amazing but the job is not yet done there's one more step left for a tester is to create a test uh test case scenarios so let's go ahead and um ask Gemini to give me a set of test case uh test cases for back dop using the recommendation which is py test framework let's give it a couple of seconds and voila we have a test case File created for our backend service in fact it's gone ahead and testing couple of scenarios empty messages how to get a message adding a message and just in case an invalid data is there with this I can run the commands which are provided in the in the response as well so it's given me the the step-by-step approach to running these test cases and isn't this fantastic like I'm already sold I don't know about you all with code Transformations and full code base awareness Gemini can perform these complex tasks not just on a single file or a selected piece of code but also on entirety of the repository I'm going to hand it over to Christian who's going to basically talk about how Gemini code assist helped Commerce Bank thanks a lot D this was a really impressive demonstration so um now we're going to jump back to the slides and um before we going back to the topic of cloud workstations and Gemini I first want to introduce uh coms Bank to you so I'm Christian Gawker I'm vice president head of cyber zenter of Excellence at com bank and comment Bank itself is a bank from Germany and comment bank is the leading bank for the German middl and is a strong partner for almost 11 million private and small business customers comment bank is also um has a client Centric portfolio of financial services in two segments this is private and um small business customers as well as corporate clients what you can see in the middle is our Focus business model it's based on growth excellence and responsibility and I think the most interesting point of this list is the one exactly in the middle of this slide shaping the digital and sustainable transformation and I guess we're working on the on the digital one for today okay next I'm also part of uh BDA the big data and advanced analytics team of Commerce Bank and um we are roughly 500 colleagues and we have more or less a huge part of of the data of the bank running our infrastructure and uh we have uh around 500 colleagues and uh we love being Pioneers meaning we started back in 2019 working together with Google cloud and what I find quite remarkable when you're using Google cloud services as of today we started back in 2019 working together with Google and quite once per month maybe a little bit often even um we we commit code to Google and it gets accepted and then our code ends up in the code base of G Cloud so it might be that when you use G Cloud uh code then there might be a contribution from Comet Bank in there then um within BDA my team cyber Center of Excellence we are fostering a secure scalable and standardized public cloud and create the infrastructure and framework to become a cloud first business this means we are responsible in the unit for data protection information security and Cloud this means mainly public cloud and for public Cloud um we are bringing the business case of the bank into for example Google cloud and bringing a use case into into the cloud means starts from engineering goes over architecture but the very interesting part is that Google Cloud provides already a quite secure foundation for the financial industry which is quite heavily regulated in in Germany Europe and on a global level and we build something what we call the cloud Enterprise Suite which makes brings the cloud security to Next Level to a trustworthy level we call it in order to make it easy for our financial applications to directly migrate to the cloud and run on top of Google Cloud and I will show some you some examples how we are using the cloud Enterprise weite in combination with Cloud workstations and code assist so something what we use heavily are Cloud workstations and um Cloud workstations are great but the way Marcos is delivering Cloud workstations usually doesn't fit our needs so what we are doing we are going to customize the cloud workstations so luckily there are a lot of knobs we can we can we can change and switch so Cloud workstations with c images um provide us exactly the the amount of detail we need in order to fulfill our cases what we do with Cloud workstations but what are we doing exactly with Cloud workstations so in the end I want to show you three cases the first one we call it Cloud developer workstations these are the cloud workstations we are using for Developers for uh Engineers so they're using wish Studio code they're using intell J to code in terraform python Java or JavaScript and this then directly embedded in our C pipeline next we have the cloud operator workstation the cloud operator workstation is tailored towards staging so staging means we have an developer has now um created an application in in development environment and we want to Stage it into test and into production and this usually is gated by an operator and the operator is using a cloud workstation and in our case with Visual Studio code or osss code and and runs terraform in order to Stage the application and by the way in this case in order to enable this uh we also need to make sure that the cloud work station runs without connectivity to the Internet so it runs completely um on our on our uh uh uh premises and how we we Define uh Cloud workstation scope last but not least the cloud analytics workstation this is for the data scientists who are working in a lab environment and use a lot of data connect the gpus to the cloud Workstation to gain the value now you see the right hand column is empty and the right hand column is for a very important thing which is also very important for the finan industry this is for security and one reason why we go and went with Cloud workstations is because it supports all our checkboxes customer manage encryption key asset inventory service level agreements VPC VPC service controls access approval access transparency GE locations or we can host it in Europe we can secure it with our Commerce Bank keys and this is what we in the end need so to sum up Cloud workstations um we are using it quite heavily um for integration into the CSD pipeline customization to address these three different uh needs security to be regulatory compliant and of course the feature itself having a workstation on the cloud I want to give you a small glimpse on how we are embedding Cloud workstations in our CSD pipeline so you see on the right hand side our Cloud native cicd pipeline so it starts with Git git assed on under Commerce Bank uh responsibility but everything from git is directly embedded in the cloud workstations and then Cloud workstations is connected to assured open source software from Google or to a custom uh repository and on the other hand we have also access to to images in the artifact registry what you see on the on the bottom to pull these images into Cloud workstations to them for example or to to change them and everything from the from the cloud workstations is get them sent to the to to Cloud build or to to build the application or to orchestrate for example to do Code Compliance and testing with for example Zar Cube and if that everything is successful it gets pushed to artifact registry this is the iPod and then from the artifact registry gets taken to the into the cd part where the operator is already waiting with his Cloud workstation now next I I want to show you a different use case how we leveraged both the workstation and our CSD pipeline in order to to bring something to to our teams which we find quite remarkable so we call it 1D 1D because our goal was we have an employee and the employee gets on board to commment bank get on gets on on to to our teams so legal check is done but now how much time should pass until until this developer can start working and our goal was to be below 24 hours and actually this was possible to to achieve with the cloud workstations so we need one day to onboard a developer to get him fully on on on track to get him onto the cloud work station he gets all the permissions he needs uh and usually it doesn't need 24 hours it needs just some some minutes there but um our goal was to be within one day and now we have the developer there being happy he can work on this first day so what's is he going to do on day two well on day two he's thinking about production so he has written his code in development and now the question is how does he get his code from development to production and the answer is simple you have seen it you use the cloud native CSD pipeline again you have the operators using the cloud workstation and then we are staging the software from Dev into production with the necessary quality gates in between and so this allows us the benefit is of course um security so imagine you have your a banking application and something goes wrong you need to patch it you need to be very reactive you cannot wait until um some some machines have booted up and you have set up all the the permissions you have install all the right libraries all the dependencies there's no time for that you to press a button and then start working and um so this is one of them the other one is time to Market of course we want to react to the needs of the customers as quickly as possible and in the end of course we also want to make our developers happy yeah because our developers um should not worry about which packages they need to set up of course they should have and all of them have wishes which packages they need but in the end they should not set it up so um the combination of Technology people and processes makes it make it possible to have this 1D 1D based on cloud workstations and our Cloud native CSD pipeline I have put a quote on this slide which is quite U recent quote from February this year from a um senior development engineer at comment bank and he was quite hesitant switching to to Cloud workstations and the moment he were using it he came to me to me to my desk and said hey Cloud workstations is by far the most effective tool in the last 10 years when it comes to increasing development Effectiveness and he said developing became fun again so this was quite quite interesting so um and what was also interesting um we did the customer satisfaction survey on the cloud workstations to make sure that what we bring to the bank and bring to the to the developers actually makes sense and makes the the increase the the effectiveness and while we were thinking about how to structure the customer satisfaction score already developers started adopting the cloud workstation so they were much faster adopting the new tech technology that we could uh validate that uh the the effectiveness and the the scalability of the platform and so it only took uh like three months until 95% of the developers switch to Cloud workstations which without without us forcing them to do it so this was quite a a positive um experience now changing from cloud workstations to ai ai assisted development something we have we've heard a lot in the last 30 minutes and in the last two days of course and um we in Commerce Bank see that AI has a high potential in many different areas in in the business in the development and risk much more however it is framed by regulatory requirements as that Comet Bank coming from Germany being a part of the European Union you probably have heard of the EU European Union AI artificial intelligence act uh which regulates or tries to regulate and address the risks of artificial intelligence so um when we think about using AI there's governance in place there's Security in place trustworthiness of AI is is a huge question and the point is why we have solved or at least we have guided lines for traditional technology like let's say databases or source code now ai brings new new uh new flavor to the to these to this to this terms to this taxonomy so um it's it's it's not trivial to solve these issues or to solve these these points but I think we are on a on a good way and together with the regulators and together with our teams internally I think we cannot do everything as of today but we can start and one thing with respect of development is we can start with techn techology and business driven exploration to innovate and create value so this means um we want to rapidly ideate I want to have a website ready in two minutes I need an developer to onboard in one day but I want to have a website in two minutes should be possible I want to drive Innovation I want to use modern Cloud Technologies I want that our developers use cloud serverless um services in order to get the application running as quickly as possible so let me give you an example with code Transformations I want to to show you how to get my idea to code base in minutes so should this should not take more than 2 minutes so the idea is to utilize Cloud workstations and terminite code assist as you have seen right right now live in the in the demo I want to generate a structure for an application the back end the front end I want to test it with unit tests and even more I want to containerize it again put it into a serverless environment and then take it from there to Manufacturing and two things I haven't mentioned yet but which are very important for us what you can see on the slide is um what we call the cloud service security assessment so we assess every uh cloud service making sure that we can indeed use it and there are two elements which are new for Gemini or for in general AI one is IP identification so making sure there's a a um legal safety when using the results from the II and the second is data governance this is what marus already mentioned so we need to be sure when we input our comments Bank values into the into the prompt of the AI then um it's only used to process a um a response but not to train further models so let's get started first thing I'm doing I'm I'm telling um Gemini I want to build a web application uh I'm going to upload this upload the CSV file and please um front end should be implemented with react back end with flask and please give me the folder structure few seconds later I get the the uh the structure you see here on the screen the front end in react back end and flask and Google Cloud Storage because the CSV I'm going to upload needs to rest somewhere next given the structure I now need the code of the FL back end let's start with the back end I need a landing page I need to upload the CSV file I need to fetch all my uploaded files again after a few seconds I get the output the app.py what you see on the screen with different landing pages and I get even the HTML pages but what I was missing here that's why I went for for the code transformation is I want to have a real validation that the file which is uploaded is indeed a CSV file I do not want to check if there's one comma in there I want to see that the whole file indeed is a valid file so I asked Gemini coures to do it for me and you see the second uh bubble on the right hand side where CH is actually embedding a CSV library to check for that great let's do the next one front end front end setup so I want again to have my front endend react and I want to display errors and alerts because somebody might try to upload upload images for example I want to see the progress of my upload so we usually have huge CSV files in in in in in the financial industry and huge means can be multiple hundreds of of megabytes or even into the gigabytes and of course I want to see what I have uploaded in the end so I want to display the upload files and again after a few seconds I get the app.js for the react front end with the um HTML code at the bottom I even get a CSS file so I can style my output and I get the index file then which in the end is called when I visit the page now next as we have also seen live right now the the unit testing I also want to test my side so what I'm asking Gemini code assist is give the code give him the code above right you test for the back end of the application so andina gives me two outputs first it lists the tests what it things make sense for my application so you see here four examples for example a valid CSV file and the file route is accessible and then it hands me also the python code for it and what I can do I can just click on the on the on the um take the code into uh my my cloud workstation click on run and check if the code that was created by by code assist is indeed working and yes it worked and now I I'm asking hey can you please put it into a Docker file and I want to run it in serverless of course because I don't want to to maintain a VM put it into Cloud run and artifact registry and it splits me out exactly the code um the two tcloud commands gcloud run deploy with a um container in the in the AR reg registry and now after roughly 2 minutes I'm done you see here the beautiful interface with the German word as so it's indeed screenshot of the zsv file upload so you can select your file there and click on upload after 2 minutes but I didn't want to stop stop here you see on the top seven of seven uh steps I have a last step I want to show you so if you invest one more minute you can also make it nice so you can have the commat bank logo at top you can log in you can select select your upload type single file multiple files you can also do that uh with our internal style guides or with the colors and so on together with uh Cod assist great um last uh topic I have is the the code transformation on the documentation so now we have the use case ready but we have one problem we don't have one use case we have multiple use case with many different developers and every developer documents in a different way so what we want to do we want to streamline the documentation and I think this is where thei can help us a lot and um so right now you have a mixed quality mixed structure and imagine again a new employees joining the question is how can he or she consume the documentation easiest well of course if the documentation is universally structured if it has the same depth if it is uh assigned to the Right audience like for business people or for developers or for just code explanation and again the benefit of of um of U Gemini code assist here is the awareness of the full work space because of the 1 million plus tokens so it can indeed document not only the code but the whole workspace and of course the challenge then remains to that every developer of course uses the same prompt so this is one uh uh uh a challenge which has not yet been solved by AI but maybe that's something for next so to recap we want to have modern Dev Ops which consist of cloud workstations gerate code assist and in Cloud workstations we have the flexible customization and three use case I showed you compliant security without the security we cannot go anywhere we need to integrate it into our processes we want to have a familiar usage so that developers should feel familiar and have fun using it we need managed scalability so we don't do anything Marcos is doing this behind the scenes and uh of course boost efficiency in the amount of um code that we can produce or um faster time to Market same for CH code assist frictionless prototyping as you have seen test generation unified documentation um the easy integration and we see again an immense potential going forward um I mean the technology is quite quite young but we see already a lot of use cases where to apply it and I think from here it only gets a lot better A lot quick very very quick and so and if you want to have the same in your organization um yeah I recommend to to test what you have seen today and uh talk to your to your managers talk to your partners um make sure you stick with the in within the regulatory requirements um talk to your people ask your your developers what they want usually they want great things and if you enable your developers they will enable you so this is uh part from comments bank for cloud work sessions and cod assist handing back over to Dev awesome thanks a lot Christian and this is amazing to thank you so I know we are over time but so one last slide that we have is for you to sign up for all the features that you saw today the code transformation full code based awareness by going and uh um registering yourself in the link or scanning the QR code we'll be there outside to take any questions if you have but thank you all for joining thanks [Music] w