GitLab GitLab Webinars, 18 Jul 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: GitLab - The AI-Powered DevSecOps Platform

Description

Watch the replay to learn about the evolution of the Gitlab platform from DevOps to DevSecOps to the new AI-powered DevSecOps platform!

A

A

We will give everybody about a minute before we get started just so in, and we will start shortly.

A

At this point, I think it's one minute past, so we will get started and uh let me introduce myself: my name is Gregory.

A

Shoppingkov I am based out of uh Ireland in Dublin, and I am currently a customer success, manager and I love doing those webinars, and we also have Justin Conrad, who is also a customer service manager and he'll, be helping us with q a during the session So today we're going to be covering AI power, def tech, Ops platform, so we're going to go through what gitlab has to offer in AI space what kind of vision we have and then obviously feel free to ask questions throughout that day, and hopefully there's a bit of time as well.

A

For you to do this uh so yeah, let's get started for the Q a session. However, you have the Q a option, so feel free to ask your questions there. Okay, so to set the context. Ai is one of the biggest technology changes that happened in decades and when it comes to software development, it fundamentally changes the way the software is developed, secured and operated, and today I'd like to share with you how gitlab differentiated approach to Ai and somewhat existing AI powered features and we're developing in how we went from devops platform platform.

A

And now, on top of that, we have ai Power devsecops Platform. So please just go complicated more and they might experience. In my view, a whole lot. Cooler tell me how we kind of continue with that, a little bit more on AI and ML and how they're becoming well established in software development workflows, including for security testing in code checks. So this year more than half it's about, 65 of developers said they are using yeah ml in testing efforts, so we'll be in the next three years.

A

According to the uh survey that we've done and the use case for yeah in soft development goes beyond code generation. It also goes beyond the developer Persona as well, and we are taking a holistic approach when it comes to integrating AI in the software developer. Life cycle, as we have always done so I could love. We believe AI is transformational and the real opportunity goes Way Beyond, creating code, because it's just a start. Writing code comprises a small fraction of the total time software developers spend going from idea to production according to uh pad lift.

A

As of my third 20th video, that's interesting, because developers spend, on average less than 40 of the time writing code and there's an opportunity to make that 60 of the time as efficient as they spend time coding. This is kind of a shocking uh to benefit me personally.

A

I never thought that developers spent so little time coding and as an example, one of the customers, a large insurance company in Spain, noted that the completion isn't an exciting development for the team, but for them developers writing code is only 10 of the process which puts into a bit of perspective. According to them.

A

It makes about 40 more efficient in only 10 of the process, which puts again a further perspective and highlights the unmet need to increase efficiency throughout uh the whole software life cycle here, but not just the court question, because I think as of right now in the industry, there's a bit of a focus on the co-creation part rather than looking at the whole picture.

A

So let's take a look at what gitlab has to offer and how we're taking that kind of holistic approach and AI has become over the last couple months, Central to our devsecops platform and because that currently is the most comprehensive AI power. Enterprise, that's a cost platform and in our vision to us, is clear that we believe that AI should be integrated throughout the whole software development life cycle, so that everyone can leverage in their roles, not just Developers.

A

Apart from it, we believe this responsible users of AI is first of all privacy first, so that our customers can protect their intellectual property. It's very important and we believe that, when incorporated into single data cops application like gitlab yeah I will amplify the benefits of the platform approach.

A

Teams will have greater visibility in metrics developers will spend less time context, switching, which has always been a problem from tasks to tasks, security, compliance measures will be more robust and Business Leaders will get more efficiency in reducing tools Pro as something that we've seen all around has been a problem and we'll deliver it again through a single application. We believe that AI will result in about 10x improvements in workflow efficiency.

A

So we know that privacy is a concern, especially and at Enterprise level. So that is why we're taking a privacy first approach to designing AI, assist capabilities, I, think that's why we are different in that space. We prioritize uh privacy, so our single application is Enterprise ready, allowing you to scale as you need, without compromising compliance or security. It's already built in so we're building, AI capabilities with API protection, privacy and compliance in mind. As of right.

A

Now, we're not going to a lot of detail later on we're going to go through the features and if you have any questions around how we're doing privacy or compliance feel free to ask us for measure documentation um on post suggestions, for example, are we offering it? This is developers already called acquires completely in the gitlab.com infrastructure and provides the same level of security as any other feature on gitlab.com right, there's also an option to use third-party language models.

A

Currently, we've enabled to use Google's quality language models, so there's an option to switch from one to the other, depending on level security. Compliance so in here you've seen why and what we mean by one workflow, so it empowers every user throughout the whole development lifecycle.

A

So there's quite a lot of things you can do here, there's quite a lot of features. So if you go we're going to go through every single one of it in detail as of right now we start with planning. You can summarize epic summarize issue comments, because they could be quite a long list of them. You can create emerge requests within that. So within the merge requests you can have code suggestions, there's suggested reviewer functionality, which would suggest who would be the best person to review this. You can summarize merge request changes.

A

Then, once you push the code, you can highlight and explain this code, you can you, you can then generate tests and merge requests. But after that, once you've done your scanning. You can explain this particular vulnerability that you might not know how to fix a developer. For example, then you can summarize the merge request, review and then, at the very end,.

A

You can deploy and that there's quite a lot, there's also issues that we that we have in different stages. I'm gonna also tell you what we're working on in the future, so this is sort of another list of what you've just seen.

A

So everything is available right now on gitlab.com and code suggestions is available on uh what this will be available in your future for open better on self-managed, okay, and that is basically the same list along with a couple other things that I haven't mentioned so I think what I haven't mentioned is the features that are going to be interesting, for everybody is gitlab chat, for example, you can like GB chat. You can ask questions in relation to GitHub and say how do I do? How do I set up Runners?

A

So it's going to give you a nice reply with the documentation and tell you how I should to do that. There's also value stream forecasting. So, if you're familiar with the value Wireless streams within gitlab or what they are based on historical data, it will tell you what uh the biosyn podcast is, which is also very great and I.

A

Think everything else that we have already covered in a previous slide, and so, let's go through the rest, so now I think we're going to take a closer look at what what is available so in terms of audition I. Think it's split into too many areas. Ai assisted features integrated into the software brand lifecycle, which is something I mentioned, and also AI Ops and features that will allow customers to build and integrate data. Science, workloads, Within gitlab.

A

So let's take a look at both of them.

A

So, first of all everything that you're seeing here is already available, which is great and it's available on Ultimate and, of course, suggestions is available on premium and ultimate, as well as a additional add-on. Now solution reviewers it helps customers do faster in high quality views by automatically finding the right people to review emergency requests and the feedback I heard from gitlab itself, because obviously we use our own product.

A

We heard that we're pretty good at assigning reviewers beforehand, but with the suggestion we use, we became even better so at least internally. uh We're feeling really great feedback about um this feature. So this review is, is the first user-facing is lab machine learning, power feature IT, leverages projects, contribution, graphs, generate suggestions, and this data already exists within gitlab, including merge requests, metadata source code files and give that user account metadata as well.

A

So one suggested Reviewer is enabled and the data extraction is complete. The new, merge, requests or new uh commissions requests automatically trigger a suggested. Reviewer ml model inference engineering up to five suggested reviewers.

A

The suggestions are contextual to the changes in emerging quests, which, which is I, think quite awesome and I usually uh commits to merge requests, may change the viewer suggestions as well, so you actually would know, depending on the commits you would know, who would be the best person according to the commission, making not just whoever has worked more in that particular project, which I think is at a fine level of detail to that and it's available on Ultimate customers.com and also is using gitlab native model. So for that we're not using any uh third-party tools.

A

This is something that we did open house and we have actually been working on this for about a year and a half. Now, at this point, so this is one of our uh oh. Some of all the special features now I think this is. This is the most famous One. Everyone talks about uh course suggestions. It basically allows developers to write code more efficiently by viewing uh course suggestions as they attack. It helps developers your productivity focus and Innovation.

A

This capability is similar to the one of the co-pilots, so Nicole suggestions are available to users of gitlab on SAS by default and self-managed gitlab when requested so there's a little bit of a process, you have to go through if you would like to have that enabled it's also an EVS code and Microsoft Visual Studio, when you have the corresponding gitlab accession installed and also in a gitlab web IDE, it can be used.

A

So the code suggestion is a generative, artificial intelligence model and what you'll need is your personal access token and you will then have a secure, API connection to gitlab.com and this API connection security transmit app contacts window from your IDE to your Editor to code suggestions in gitlab hosted service, which calls either Google vertex AI Kodi apis or endangered suggestion is transmitted back to your IDE editor. So this is the way it's working and currently we're using Google's Cloud Red X AI Kodi API models as well. Along with our own we've been developing.

A

We saw that there's an option for using both so explain this code. It allows you to get up to speed quickly on call. You have been given to secure and operate right so deciping, the source code of a new software project could be in daunting and it is time consuming.

A

So in this case the call could be fully documented or maybe written in a program, language that isn't familiar to the developer. Whoever is reviewing it, and even if the developer is familiar with the program language, the code may be too complex or difficult to understand so and in this case it's a helpful tool. But actually you highlight it and actually explains what it does. So, in this case we're using a 30 third party model and it's available for ultimate customers. On.Com currently.

A

This so generate give commands in SLI, so it allows you to discover, recall and execute the mini git commands using uh natural language when you need them on your command line, so this is kind of fairly simple I. Don't think that requires a whole lot of explanation. It's also available for users that are members of at least one project in Ultimate, tier on.com, and it's also a third-party model.

A

So this thing is this: is the coolest one we have.

A

With them explain, this availability will allow users to identify and effective way to fix similar ability by combining a basic vulnerability information with insights derived from their own code, so gitlab surfaces, vulnerabilities that contain relevant information are whether more often users aren't sure where to start. So it takes time to research and synthesize information. That is, you know that you have within that vulnerability, but what to do with it essentially and, moreover, figuring out how to fix it. The given ability can also be difficult.

A

Synthetic wise collaboration with security uh and the stock World collaboration within the team and so again to help teams identify an effective way to fix vulnerability, Within the context of their specific code base. We've released this feature that provides gitlab AI system availability recommendations, leveraging the explanatory power of large language models.

A

This capability combines basic vulnerability information with insights derived from the customer's code, to explain the vulnerability in context. It demonstrates how it can be exploited and provide an example to fix it as well. It's available for ultimate customers.com and we're using Google model for that.

A

This one I think has been based on my experience, also her customer favorites, because it helps security and QA teams. Members write regression tests that prevent security problems from occurring in the future, so sometimes what I'm seeing is the uh tests aren't always being created for the code that's going to be merged, so I think this is a great addition for a lot of companies that are at least I'm talking to so in proposing changes. The new features through merge requests, obviously, is great, but companies are always.

A

You know what about the tests and I think this is something that will fill the gap and that sometimes can be the hardest part rights of any changes you make. So sometimes developers not sure where to start writing these tests. Maybe the test doesn't cover all the scenarios that need to be tested. That is a on its own, a huge I think task, and maybe you just need a second opinion as well.

A

I think I, guess that would believe that we can use generative AI in large language models to help provide our own test coverage for the proposed changes. So the viewers and others can have the confidence and quality of the changes that they've just submitted. So it's available for ultimate customers.com and then also we're using third-party model.

A

So GitHub issues are pretty much Essential for team collaboration and they serve as the sort of Truth for teams to align on a problem definition and a scope of work for ongoing efforts, and that seems to collaborate on issues to refine them. The volume of comments grows which issues with many comments. It can be challenging to understand the status of the glance and you might need to specific time reading comments to get an overview of decisions made so far.

A

So we have a gitlab issues that are super long, maybe five, four years, and there could be hundreds of comments. I'm personally very excited about that, because it gives me a summary of what has been talked about and actually I'm very quickly, updated on what the context is so also using a third party model. Here and available for ultimate customers, on.com.

A

Now so merge request, uh summarize measure Quest changes. So again they are another point of the central point of collaboration for code changes in gitlab and they often contain a variety of changes across many files and services within a project and also in merge requests, communicate an intent of change as it relates to an issue being resolved, but also might not describe what they change was to achieve that.

A

So other view, Cycles progressed the Queen's status image request can become out of sync with the realities of the proposed changes, and sometimes developers also aren't very good at updating that so we're using again large language models to help provide our own summaries of merge requests and as proposal changes. So the reviewers in order so obviously can spend more time discussing the actual changes and less than keeping descriptions updated as well. So for that we use it in third-party model and available for ultimate customers.

A

Now summarize, my merge request review so during the course of review or merge requests, you might sometimes leave too many comments. This particular comments may have a specific information about things that need to be changed or contact of. Why you're leaving feedback on the proposed changes?

A

If you've left a lot of comments, it might be too hard to remember right everything that you said and what the order the author should look at to resolve your feedback changes. You recommend making there's too many of them great. We have a nice summary I think this is very similar to the one with the issue summaries, also using third-party model and available for ultimate customers now and gitlab chat. So because everyone can contribute that gitlab and as a platform, we offer a variety of features.

A

Sometimes it's hard to know everything will gitlab is capable of side person uses use this feature as well, because I need information, changes, features, change and sometimes I need a better way of looking things up and gitlab chat. Helps with that and it's a great feature. So we use AI to create a chatbot that answers how to questions about gitlab products.

A

It responds with an explanation around links to your documentation, so it has allowed to save a lot of time for us, and this is the evaluation for casting where the prices can kind of predict deployment frequency based on historical data. It's also available for ultimate customers on.com and we're using a third-party model.

A

I think we're now finished with what gitlab is currently offering and moving towards what our AI modelops vision is. So we're going to take a closer at our roadmap for model Ops and what we are looking to release soon. So what is model Ops? It is Data Ops and ml Ops together, so model op is everything you need to do to add. Ai add to your applications, just like we're, adding it to gitlab. We know.

A

Customers, like you, are adding it to the software as well, and we want to support Enterprises in adding Ai and ml to the applications that you're building and deploying using gitlab, and we started building the functionality in 2021 as a partnership with one of our Enterprise customers and Border gitlab community.

A

We recently expanded this by natively supporting ml flow, enabling Ai and ml expectation Within gitlab, and earlier this year we introduced an integration with ML flow, a popular open source tool for data science, experiment tracking our goal. It's very simple. We wanted to help you be more productive and efficient and enable customers to ship AI with their applications faster with higher confidence and to be all managed. Within gitlab in gitlab 13.9, which was I, think February 2021.

A

We introduced support for self-hosted GPU Runners, which enables any gitlab, for instance, to connect uh self-hosted GPU enable Runners to make them available to any gitlab pipeline as well. What was that done? We wanted to help you be more productive and efficient in enable customers to ship AI with their applications faster with higher confidence as well and recently with 16.0.

A

We released SAS GPU Runners, which enable customers again to easily and cost effectively access, GPU enabled Hardware to the same goal: vehicle cheap, AI, faster within applications and later this year, we're planning to introduce a model registry to allowing customers to store the version, deploy and track the help of the AIML models immediately Within gitlab, and so we want to make sure that you're, productive and efficient and enable you to ship AI features with your applications faster and, on the left hand, side. Also, everything that is available today is everything that we've gone through.

A

Github chat by Lucian forecasting and, on the right hand, side it's everything that we are currently working on and there will be more, but this is just a a kind of a small list of what we are going to release.

A

So I'm personally excited about the vulnerability report summary because I think that is going to be a great addition to uh functionality for security, at least, but everything else is quite interesting as well.

A

So I think at this point we have quickly covered through everything that gitlab currently offers, and hopefully you have learned a little bit here and there about what we offer, what the features are like, but I think you're seeing the trend here. Most of the features are available on gitlab.com and there's only code suggestions that can be enabled um or self-managed.

A

So, if you have any questions, please do ask other than that I hope. I hope this was a useful 30 minutes for you and uh yeah more than happy to answer any questions either. Not thank you for coming. I appreciate your time and uh yeah enjoy the rest of your day. So we're gonna hang out here for about five minutes and take any questions.

A

Hi, everyone welcome Jerry. Thank you for coming.

A

Thanks John, thank you for coming.

A

I'm, just gonna stop sharing the screen, but if you, if you have any questions, please do ask no problem. Man thanks for coming.

A

All right, I think we're at this point we're good to go. um Thank you, everyone again and we are going to be closing down the webinar, so thank you, Justin for uh coming on and helping so I'm going to be closing it down bye. Everyone bye.