How to Build Production-Ready AI Agents: MongoDB Atlas + Google Vertex AI

0:06My name is Michael Lynn, and I'm a principal developer advocate at Mongo DB, and today I'm really excited to bring to you a great conversation with a a very smart person in the AI space. He works at Google and we're going to get to an introduction shortly. We're going to help you take your applications to the next level. In this episode, as it says in the description, my name is Mike Lynn. I'm going to be joined by Yang Lee. Yang is a staff solutions Architect at Google Cloud, and he's going to be talking about the powerful combination of Mongo DB, Atlas and Google Cloud AI and how this can supercharge your development process.

0:44So we're going to, we're going to be talking a little bit about the theory, we're going to explain some of the product sets and how Mongo DB and Google Cloud work together. And then we're going to get into a demo and show you some really exciting stuff. Yang is a really bright guy, very easy to talk to, very knowledgeable. So stay tuned for that. We'll check and see who's coming into the live. Oh, we've got a couple people coming in. Yeah. So if you come into the chat, maybe let us know where you're joining us from around the world.

1:11I'm I'm joining the live and running it from a little town called Newtown, PA on the East Coast of the United States. And I have a semi rainy day. So. So tell us where you're from, where you're joining from, and let's get right into today's discussion with Yang Lee. Let's bring Yang in. Welcome. Welcome, Yang Lee. It's great to have you on the show. Thanks for having me, Michael. Yeah, terrific. Well, I really appreciate you taking the time today to to help us understand a little bit more about Google Cloud solutions in the AI space and how we work together as companies and with our product sets.

1:44Welcome, Robin. It's great to have you joining us from from India. Terrific. So where in the world are you, Yang? I'm a solution architect from Google Cloud. I had been working with Mongo DB for a long time. Mongo DB is one of the most important partners for Google Cloud and we have the fortunate honor to work together with you guys along those years. We have many important integrations enabling our service to more and more use cases and businesses. And together today, I'll be very happy to talk about how Mongo DB Analyst, together with Google Cloud AI especially can supercharge your applications, make your business a success in the future.

2:23Yeah, yeah. I love that. I'm really excited to have this conversation because, well, I'm a developer and I use both Mongo DB and Google Cloud solutions. So I'm interested to see your perspective as a solutions architect working at Google Cloud, how these things come together and what specific set of products that that we're going to be covering today. So let's let's. Jump right into it. So yeah, when I'm I'm considering developing an application and I want that application to be intelligent. Tell me where we begin.

2:50Let's let's talk about how we begin the process of developing with Mongo DB and Google Cloud. Google Cloud provide a whole set of tools and services to help application developers to build things for your business and otherwise Mongo DB Atlas, arguably the most powerful database for documents that together we enable many use cases. So today what I would like to do is to give you a high level as well as the enough detail for you to start with. So what I'm going to talk about is into generous 3 sections, I would like to go over the Google.

3:22There we go, my services. And so this is the first thing and I'm going to talk about the detail integrations between Mongo DB Analyst and the Google Cloud. So now you know, hey, which services I can I can integrate with, in what fashion and what I can achieve. And then the third, obviously, we're going to create a demo and showcase you how to create a production level NI demo with the support, in this case using Mongo DB Atlas as the rack engine behind it. Fantastic.

3:50That sounds like a great way to spend the next, I don't know, maybe 40 minutes, 45 minutes. So let's let's dive in. Sure. Now let's start with the Google Ji services. You know, Michael, like you said, hey, where should I start? I go to the Google Ji services. I can see all these whole set of tools, right? Sometimes it's confusing. However, what I'd like to do today is to give you a architectural level of how these things are laid out so you know where to start.

4:14And what do they sell for? Perfect, Yeah. The Google Cloud, we deliver the most comprehensive platform for building and adopting the agents and AI applications. And this whole stack is powered by AI hyper computers that we build and these are the performance optimized hardware, some of them designed and created by Google, some by other leading providers in the whole world. We adopt the open communities of softwares and it provide a flexible consumption management for your need.

4:46On top of the data is the data and the models for the data. You know, we support the multimodal models, all kinds of things. And based on that, you'll be able to implement and the leverage the vector search. We're going to talk about that later. More about this which generate the AI insights and power your data science need and the jobs models. Google continuously to invest in and deliver the most powerful AI models world. As of today, we have a Gemini 2.5. We also have the latest image Gen. models as the video generation of the models, right.

5:21We also support the models from the partners as well as the models from our open source community. Then it's the Vertex AI suite and the GAIAPIS and the AI Studio. The Vertex AI is a whole set of package providing and supporting the whole life cycle of building applications of AI and genuine applications and agents. Putting the model garden, model builder, agent builder and the genuine APIs is literally exposing the AI models through the API so that you can call into directly and leverage the advanced capabilities. And for the AI Studio, it's an integrated environment for you to easily experiment, develop, try out your AI applications. And then it's the, it's the integrated Gemini Assistant, which is integrated into the Google Cloud as well as the Google Workspace.

6:15For example, using Gmail, you probably already use the Gemini Assistant, right? To summarize your emails to help you get a quick easy answers for the emails, so on and so forth. And then it's the AI application that agents, we know that more and more companies are starting to build their own AI applications and agents. So for that, we provide Google prebuild agents you can leverage easily. We also list the partner build and custom build agents that you can pick and choose and like a Lego block build more advanced features, workflows based on your business need quickly, right? That's the Google agent space.

6:52We announced the Google agent space earlier this year during Google Next as of today that it provided enterprise level search and also whole set of tool set such as in the notebook LM. We also provide agent gallery within agent space so that you can pick and choose agents easily and also in this your own agent ancient space and Vertex AI, two of the most cornerstones of Google AI services suite. So I would like to drill more into them. Vertex AI as you can see that we have the model garden, the model garden list, the latest Google models, all the models we have and the continuously advanced models from Google. We have these partner enlisted models, right? Mongo DB can create a model, other partners can create a model as well as the other open models from the industry so that from the model garden you you can pick and choose and leverage them into your own pipelines.

7:46So this would these could be models from any producer. From to be Google from any producer, we need to have all the most advanced models listed in our model that you can use. So I'm trying to place this like in some in some frame of reference like this would sort of be like like Hugging Face where you've got other models, well, models indexed and and information about the models and this is where you can go to consume them, correct? Michael, interesting.

8:13The Hugging Face itself is part of the model garden, so that then you have a Hugging Face subcategory and then you go there, you can see all those Hugging Face models and you can see other models and that's not in Hugging Face. So it's a combination of all these. Great, great. That's good to know. There is the model builder. For whatever reason, he wants to build their own model or customize the model, right? The model builder provides the whole set of the tools and the capabilities for you to do so, starting from the Prompt Engineering tuning notebooks all the way to building up the pipelines to run it and also monitor how your model is running. OK, that's the model builder.

8:51And then on the top is the agent builder. You know we do want to build more animal agents, right? The agent builder provide the whole functionalities and the whole lifespan for you to build, run and operate the agents with the agent development kit. You know the name tells a lot. Agent Development Kit it's out-of-the-box to provide functionalities for you to build agents very easily. For example, that we are now also announced that the A to a agent to agent protocol, right? Agent Kid supports A to A out-of-the-box, so it's much easier for you to create an A to a compatible agent using Agent Development Kid.

9:27Yeah, yeah, Agents seem to be such a huge focus these days. We've kind of moved from RAG into into the agent space and then combining those things. Tell me a little bit more about Agent Builder and the user experience, the developer experience. Is it is it a graphical user interface or is it more like Lang chain where we're where we're using Python or some language to to instantiate agents the agent. Builder starts with very friendly UI that you can pick and choose and it's effectively that if you are not a very technical person, you can literally talk to the agent builder.

10:00You describe what you want. He said, hey, I want agent to do such and such for me using natural language and agent will be able to create it for you. Meanwhile, if you are a very tech savvy person, you're like a Python And going to the coding, you can create the code and line by line create agent yourself. So it'll provide a whole spectrum of capabilities for you to create and run your agent. And so that's the agent ADK agent, right? And once it is there, we also provide the agent engine that you can run your agent on top of. So that's the agent engine.

10:32And once your agency is mature, you know, sometimes you want to run it yourself or if you want to expose it, let's say self agent right, in a public fashion in the agent garden where other agents are also listed that you can pick and choose and yourself. So other people can build on top of imagine that you have agent causing to other agents and do complex work flows, right? And that's where the A to A comes into shine, because they talk natural language.

10:56That's A to A between them, right? Yeah. I love the standards based approach and I love the the A to a spec, the the protocol, yeah. Then they basically speak of the same language and you don't need to translate between the agents. If you need a rack capabilities from Mongo DB, Alice or other data sources, it's easily created within the agent builder also right? And he wants to expose an API of your agent. You can expose that and manage it data connectors to other data sources. You can operate your agent and you can have the grounding. You know you, it's very similar to the rack. The grounding of the agents and the tracing analytics is all supported within the agent builder. So that's not set.

11:35As you can see, we provide for people to easily create, monitor, experiment and iterate right for the development and that is the agent space. Agent space provides a whole set of tools out-of-the-box for you to deploy, monitor, run your agent. It's effectively a place that you can run your agent pipeline and monitor and continue integrating. That's why it's called agent space, right? So notebook is one example of the tools. You can create all kinds of agents around it, such as you can generate ideas, do deep research or any other functionalities within your business flow. And similarly, we also have agent gallery within agent space that you can pick other people's agent. You can enlist their own stuff over there. It also has its own agent designer, which is by powered by agent builder at the bottom.

12:23So that within agent, you can directly design and iterate your agent easy whilst your pipeline is also running over there. So that's more what the agent space and vertex AI, OK overall, eventually what the Google AI services, we want to bring the generative AI to the real world experiences in a quick, efficient and secure fashion, right? Starting with the models, first party, third party, you can easily find, customize the models and transform your business processes with the portfolio of AI agents and the applications. It's good and you can leverage easily the platform we provide that's a single integrated development platform, the Vertex AI for all your needs.

13:05It's powered by the trend too and service your workload powered by the superb infrastructure provided by Google, right AI hyper computers. And we do want to, from Visa perspective, we want you to be able to create and deploy AI services that's really useful and set your customer need so that your business can provide, thrive and become very successful in the future. Yeah. OK. So as we went over the Google Cloud generative AI services, now let's dive deeper into how Mongo DB Alias is integrated with various Google services. OK, And what the options.

13:42This is the high level of Mongo DB Alias integration with Google Cloud. ODB has been working with GCP for a long time in a very proactive and successful fashion. We had to be delivered many, many integrated, powering the support of all kinds of customer scenarios, right? As of today, Mongo DB Alice is available in 38 Google Cloud regions. This enables Mongo DB Alice to serve people around the world, customers around the world in a quick, productive and easy fashion. OK Mongo DB is also named Partner of the Year for three consecutive years by Google Cloud Marketplace. This alone showcases the success and impact of Mongo DB Alice integration with GCP services.

14:26Mongo DB others is deeply integrated with all the major GCP services such as BQGKE, Data flow and the JNI services. Anyway, working to deepening those integrations and widening the services to support these cases, right? Mongo DB is enlisted on Google Cloud marketplace. With a few clicks you can subscribe and start using Mongo DB Alice. It's fully integrated with Google Billing Monitoring log. That's. That's absolutely critical. I mean, one of the most frustrating things is, is this multi point shop where you, you need to establish relationships and accounts with multiple vendors and, and making sure that you're, you're your thresholds are, are set right and, and your billing is, is correct. And it's nice to know that we've got 1 central point for billing. That's right.

15:17Now let's look deeper into how Mongo DB fits into the GCP services and the GCP ecosystem. OK, As you can see that Mongo DB is being one of the top document DBS in the whole world. It fits nicely with Bikori who is a great data analytics engine here and also integrates easily with Vertex AI for the AI and ML workflows. And further, if you need, you can put the BI insights and generated from looker. So as you can, the MongoDB fits nicely within the GCP services and this is the whole thing is governed by fundamental services within GCP, such as the data plaques that's for data management operations and also machine learning operations trying to operate your AI services applications as well as our agent within Vertex AI. And this is a more detailed technical diagram showcasing how these or where these services are connected and how the data flows.

16:14From the left side, you can see that we have all kinds of data sources. It could be so a sales force, IoT or any other things, right? For Google Cloud, you can create easily pops up ingestion data pipeline, right? Once the data is there, it can be further directed into, for example, data flow for ETR processing or the cloud storage. You want to build a data data self and you can further apply the data proc to process the data depending on whatever your business need. And once the data is also can be pumped into BQ for analytics. And in this whole process, Mongo DB can integrate with cloud pops up directly with showcase later data flow, BQ, Cloud storage and as well as the BI reporting tools such as Looker and analytics as a vertex AI right or Cloud logging, monitoring, billing is all integrated with Mongo DB SIM. So it feels as if a first party service of the GCP services. OK, Bigquery is one of the top offerings from GCP. So what I would like to do here is to showcase a little bit more detail about how MongoDB Atlas literally integrate with Bigquery.

17:26OK, MongoDB is very good at storing the transactional data, the documents and also drive the real time business value from it through the aggregations, charts, your SQL interfaces and so forth, right? Meanwhile, the advantage of Bigquery is arguably Bigquery is the most powerful data and analytics platform and service in the whole world. And Corey parabytes of data in seconds, it's a very quick. Meanwhile, BQ also provides BQML that is powered by Gemini and our latest models that you can do data analytics for prediction workflows as well as the data insights generation workflows, right? There's another thing that's very powerful for powerful BQ is that within Q console, you can have a conversational interaction with your data.

18:18You talk to your BQ using natural language and then Q will convert that into SQL and query your data correspondingly. So that even with people who do not have too much technical skills, they can achieve that. This also enables the users to quickly iterate your thoughts, experiment your workflows so that eventually you will know exactly, exactly what kind of a processing and the data insights you need. And then you can literally write the SQL. By the way, if you already drafted the SQL for you, you probably just need to review and take several places.

18:50And then you just apply that into a function, into a routine job, right? And the whole thing set up. Quick question about the integration between EQ and and data that resides in Atlas. Is there an ETL process? Is there a connector? How does that work? How do they work together? We have multiple ways to put them physically together and this is one example here again we have multiple ways. So this is the one example a single time Mongo DB engine waiting it right.

19:19Mongo DB generate this CDC stream, the change data stream, right? Yeah, the pops up can be easy to connected to the CDC stream. You literally just create pops up topic with the schema and then you create a BQ subscription. It's your BQ instance. You create a subscription, then the whole flow is set up just using these two simple configs. What's going to happen with this is then every time anything changes in Mongo DB, it generate this CDC stream pumping pops up and then the pops up depending on the subscription it pumps further into the BQ table.

19:52So on your side, you continuously have the latest data or the latest information, which enables you to leverage the advanced analytics capabilities and BQML on your data, generate business insights. And OK, so this is one example. And here is another example. Let's say that you are a data scientist and you use Python notebooks, right? We from the BQ side, we have a BQ Big Frames and integration with MongoDB Atlas. This the BQ Big Frames is enables big queries, scalable data warehouse capabilities into the popular Python data science community. The ecosystem within your Python notebooks, you can use open library to connect and read from Mongo DB alerts easily. So this now you have the data from one side right And with big frames you can connect to BQ and it can also translate your need into BQ queries directly through the big frames. So now and and also leverage the BQML which is powered by Vertice AI.

20:52So now that you are data scientist, you do not need to leave Python notebooks. You will have access to both science and you will have the business insights from the DQ to power your business decisions and your data science jobs. So that's an import here, OK. Great. Yeah. DQ data stream is also connected with Mongo DB Atlas. The data stream shipped the support for Mongo recently and Data stream is a simple and seamless and serverless secure data streaming service within Google. It out-of-the-box connects with all the major Google services. So by data streaming supporting Mongo DB Alice, Mongo DB effectively can connect to all the major services within Google very easy.

21:38So what we need to do is that it was simply in data stream. You have a config right? Connect the connect streams of your Mongo DB Alice and other several things and then hold it up. Very simple. Once the Mongo DB Alice data such as CDC streams or any other data is pumped into the data stream, you can then redirect the data into GCP services such as the data flow where you can do the batching and the streaming ETL or the BQ. So you see, now we have another way to update into BQ or the DS where you can build the data lake.

22:11So look at these scenarios. By doing this, you effectively enabled the streaming analytics workflows, the data housing and the lake house workflows, and the RAG engine and the Geni or agentic workflows. For most businesses, they also run their own application where the third party or they develop themselves right within GCB ecosystem. We also provide a very easy way for Mongo DB Alice who connected to these applications. What how we achieve this is within Google we provide all kinds of application integrations and connectors whilst PIC private service connect right within these. For example, let's say you have a rational database within your application and you can enlist that and connected to a pops up pipeline where the Mongo DB now subscribe to the popsup pipeline.

23:02Data will flow into Mongo DB right in a in a very good fashion. Meanwhile, Mongo DB Analyst is able to connect with additional data sources directly depending on what you need. With all this data within Mongo DB Atlas, now you can visualize and analysis your data using all the tools or any tool of your preference such as smoker, right. Once you have the business insights, you can pump it back to the applications through those connectors. Now within your application you can surface all kinds of data insides data flow results to power your business. Yeah, I talk about a way to, to enable intelligence in your application.

23:39I mean this whole entire life cycle. So I'm wondering where, where would would you in an application that's going to leverage AI, for example, vector search. I I'm assuming you would use MongoDB Atlas for for vector search in this particular architecture. That's right, I'm going to showcase a detailed architecture for that particular as far as I show you a demo of how to do that. Great. Terrific. Oh, by the way, there's another thing I'm going to talk about is that security and private network is 1 top demand from our customers. MongoDB Alice together with Google Cloud VPC will leverage PSC Private Service Connect to achieve that. Yes, see, the consumers can use their own internal IP addresses to access the services without leaving your VPC network. The traffic remains in tied within Google Cloud. So now that you can have a very high confidence of the security, yeah, PSC also provides a service oriented access between consumers and producers with a granular control so that you have a lot of control of who can access what right and how what services can be accessed.

24:47A Google Cloud PSC offers a one way network peering service between a Google Cloud VPC to a Mongo DB Atlas VPC. So this is secure and scalable service connection. You can build the applications on top of it for your business. OK, so secure and performant and. Performant. Yeah, yeah. With with proximate pairings between your applications and your database, Yeah. That's correct. Yeah, great. Another thing very important is Mongo DB together with Google Cloud, we put a lot of emphasis on the startup community. We want to help you grow, wants to help you be successful. If you're qualified, you have access to all the benefits of Mongo DB and Google Cloud through our startups cloud program.

25:32With a credit you from the Google, you can get your first year of Google cloud cover up to 100,000 in credits and other Google wide discounts. If you are focusing on the AI, your AI startup, the total credit can be up to $350,000 credits. Great. And from the Mongo side, Mongo offers up to 10,000 free credit for Atlas and extended the data services. OK. And that will be a very big boost for the startup where they are trying out their business, trying to find the first founding, but at the same time trying to deliver MVP and attractive, Yeah.

26:08So very, very familiar with the startups program. I actually helped build the startups program at Mongo DB and, and I've actually put a, one of my startups through the program and I've enjoyed the benefits of, of participation in the Mongo DB as well as the Google Cloud startups program. So if you're, if you're a developer listening to this, watching this and you haven't an idea, you have a really good idea you think will will take flight and you want to accelerate your, your startup. Definitely take a look at these links, click through and and read up on the benefits as well as the, the requirements, which are fairly, fairly light in terms of, you know, what you have to do to to get in, but can't speak highly enough of the, the partnership between Google and Mongo DB to help startups.

26:54I think we're both very focused on on helping developers innovate and and helping startup to be successful so. It's not there exactly. We also provide a technical and onboarding support, technical devices, and eventually we want to help you build a very good application so that tailored to a customer need help you survive, thrive and become very successful in the future. So if you're interested, feel free to scan the QR code or you can go to cloud.google.com/mongoDB/startups for more details.

27:25Yeah, that's great. Appreciate your partnership there, of course. Terrific. So what's next? So as we went through the Google AI services integrations between Mongo DB Alice, Now let's look at a real example of a productionized GI service with Rack engine powered by Mongo DB Alice. OK, so this is what's happening on the back end. From the left you have a data data sources. It can be anything. It can be data, Mongo DB Alice or any other data sources, right?

27:53Once the data is pumping, the Alice had this trigger. It literally says, oh, something's changed, I need to do something and you can configure. In this case, we configure to send the change the data to the vector embedding service powered by Gemini in Vertex. OK. What it does is effectively it turns into this change into a series of numbers or embeddings representing the semantic meaning of this change. OK. So once that embedding is created, so the vector embeddings and the vector indexes together with the related metadata and the context chunk is stored in the Mongo DB alerts.

28:29Now MongoDB Alice serves both as a document store as well as the vector inside index store. And then the user comes in, user starts to ask a question, a natural language English question for example, right. That question is also pumped into the vector embedding service. So we're going to also embed and vectorize the question. So imagine that now your question is being vectorized, it's come to the MongoDB Alice. In this MongoDB, Alice will be able to do the vector matching or vector search, search for the most closely related vectors and then it will find several.

29:05Together with this, Mongo DB will massage the initial English question together with the found vector embeddings, the vector indexes together with the metadata and the context. This effectively form a much more enriched prompt or question set right. It will be pumped or sent into the Vertex AI or Gemini. Now the Gemini are armed with all these grounded information, the latest information or domain specific information with the questions, it'll be able to provide a much more tailored answer to your user or your workflow. OK, so this is effective.

29:39What? Welcome to demo as of today. OK, great. Now let me switch to the demos we have here. So this is the demo we have on the back end. We already created a Mongo DB, Alice DB. We pumped in some data. The data we put in is a grocery information. Imagine yourself, you're CEO of a grocery store, right? And this is our database. It contains all the grocers you have. I already asked the two questions saying, hey, hey, good morning. And what kind of products you have and tell me that you can search for grocery products and then tell me what you want.

30:09OK, so now let's try it. I would say this. So as a teenager, for example, who is active in sports, what products would you recommend? And tell me why? As you can see that our agent is working and find a similar product. He said, oh, OK, here's some few product for the teenager, you know, active in sport energy bars, you have this and sports energy gel and energy speak. It tells also what's the reason behind it, right? Convenient source for energy, protein and so forth, right?

30:39You can do you can take it easily. So you see this is conversational experience. You just ask and it tells you what it does. OK, the more advanced examples here, OK, let's say that for Hispanic family, I want to tailor to my Hispanic customer base, right? What products what the best for their diet. Let's see the result here. OK, Salsa. I love salsa, by the way. Yeah, that's also fitting the, well, Corn nuts, It tells you. Hey, crunchy stuff made from corn.

31:07Spanish flavor, Right. So I also have a lot of customers from Japan and the Japanese community. How about them? Do I have anything in stock? Let's see what I found. Oh, Ramya. Yes, Like noodles, right? Japanese cookie, rice wine, wasabi powder, roasted things. OK, so that sounds very good. Let's make it even more general things from here. I was told many times that I need to lose weight. OK, Anything you can recommend A few products, slim and detox.

31:37OK, I'll try that One is mint protein bar. OK, it's later saying go to the gym and eat healthy, right? This is good. I also like salad by the way, to be more specific. OK, any recommendation for the oil for salad and tell me the reason Extra virgin oil see some oil, OK, healthy benefits, numerous those kind of things. See, now you can see that this is just the one example of what people can achieve simply by connecting Mongo DB others as the Wreck engine and together with the Gemini model.

32:10What here is that we wrote the agent and we wrote the agent using ADK Google Agent Development Kit and I'm running and hosting the agent locally on my laptop for this demo purpose. But for your production level jobs and pipelines, you can easily host it on the GCB platform, on the VM or other computational resources. OK, yeah, that's it for the demo. And after this showcases what we can achieve together with the integrations. Yeah, fantastic. And, and how long did it take to build something like that? Did did you build this demo?

32:42I built this together with my colleague. We this, we are quite experienced because we build similar things. You know, quite often it took us about two or three hours to create this for the first time. You probably need to go through things probably one or two days, but it become quicker and quicker. Yeah. Meanwhile, that from a Google side, we are increasing making it easier for the customers to build it. For example, in Vertex AI agent builder, right, you can talk to the agent builder and describe what you need. Imagine your CEO, you're saying, hey, I want an agent to do blah and blah and blah.

33:15The more detail you created, you gave it to the agent agent, it will better create it for you. So we are making it much easier for the people to create it. So let's say you got me. I'm in. I want to start building something very similar to what I just saw. Where do I go and how do I begin? You start with Mongo DB, you know the web page and you just search for GDP integrations and it will direct you to the page. We also have similar pages on the Google side you go to versus AI and Mongo DB. We'll have all kinds of demos tutorial for you to get created. I'm just going to jump and grab a link quick. We go, There we go.

33:51OK, I'm not going to be able to find that, but you can go to mongodb.com, start from there and find your way to the Atlas Cloud database as a service. It's free to get started. You don't have to put a credit card in. You can deploy a free tier cluster right away and and keep it for for however long you like. And then once we get the the database up and running, then we're going to want to go check out the the Google cloud services. Is there a short link or or a link that you want to send people to?

34:19I don't have the link right here, but it's very easy. I meanwhile, another thing I would like add is that GSV and the MongoDB, our technical teams, we routinely post very detailed technical, technical guidances to help the people that they can follow exactly the code and the process to create their own integrations as well as AI agent and applications. So search for these and you'll find a whole bunch of resources. Great. Yep. Terrific. Well, it's been a pleasure talking with you. Is there anything else you want to share with the audience before we wrap things up?

34:48You know, I'm very happy to for the opportunity to present our integrations here. You know, with the most 2 powerful platforms working together, I hope that our collaboration and diverse will be able to serve your business. Try us out. I believe it will make your business even more successful in the future. Fantastic. Well Yang, you'll have to visit us once again and we'll stay stay in touch. I plan on diving in and taking a look at the demo that you created and and maybe building something similar on my own so.

35:15Of course. Thank you.