Transcript for "Hands-On Workshop: Agents & GenAI - 4/29/26":
Hi, Nel. Good morning, good afternoon, good evening. Welcome in welcome in for this agents and gen AI workshop hosted by DataIQ. We'll begin shortly. So, yeah, sit down, enjoy yourself, grab a cup of coffee, water, tea, juice, or whatever's your gig, and we'll begin shortly. And, as you're, getting into the workshop with us, please give me a thumbs up if you can hear me, if you can see me. Can you see us? We are true today. And, if you can hear, our voice and see the screen, please, give us a thumbs up in the in the chat. And, yeah, we're gonna start shortly. Hope you're all excited, about the workshop today. It'll be a very fun walk through about, like, LLM and gen AI, recipes in Dataiku and also building up agents. So it'll be super cool. We're gonna start in about a minute or two. And, yeah, just be ready, and enjoy. Hope you enjoy the session. Welcome in. Our beginning shortly. See a couple, message in the chat already. So, looks like it sounds like people are seeing us, Carlos, and, yeah, hearing us and, seeing the screen. So that's amazing. We're gonna start in about a a minute. And, yeah, welcome in. Yeah. Hi, everyone. Hi. Michael, Rohit, Shibal, Gilson, James, Shane, Steve. Welcome. Welcome. Welcome. And, before we start the hands on workshop, so, you are here today with us with the hands on workshop in agents about agents and GenAI, hosted by DataIQ. And, before we jump into the content, please make sure that you, you are logged into your DataIQ account and, you have accept the invitation. You should have received an email, with the invitation to this, to this space at that I q cloud. If not or if you didn't find it, please log in to the into your account or sign up for a new account with the mail that you have registered and, accept the invitation. So while you do that, we're gonna go through, like, a little bit of, like, the house campaign and, things we're gonna do today. But please make sure you can log in into your account and do accept the invitation. And we're gonna use that space, for the whole workshop today. And, today, I am, hosting the, this workshop about agents in Gen AI for you today, alongside of my friend here, Carlos Luisa, Carlos. Please refer to introduce yourself. Yeah. Sure. Hi, guys. I'm a senior customer success engineer here at Atypo, and I will be your troubleshooting buddy today for this workshop. So while Roma is running the workshop, I'll be here in the background ready to jump in and help you with any issue. So if you have any issue, just, drop it in the chat. That will help you. Amazing. Yeah. So anything, any comment, any error, any problem, any feedback, anything you wanna chat with us, put in the chat. Carlos will be helping us and me. I'll be your, DJ, or, I'll be, leading you through, this workshop today. My name is Luis Rom. I'm part of the solutions engineering team here at DataIQ. We are super happy to join, to have you join us, in this workshop today. There are a couple things. So, like, in this, tool here, you can see, like, on the side of the chat tab, there are docs. There's, like, a, tab that's called docs, and you can see everything, like, all the assets that, that are needed for you, like, to go, to go, like, go along with us with in this workshop. And, also, we have just, released a poll. Please, if you can just fill in, that that's very quickly. Just one question. Fill in a poll for us, and, I'll get it started. And, you also should have received, let me show you with you, like, a link to access the workshop assets. We are going to get into, we're gonna get into that in a minute, but, just be mindful. Like, use these instructions. Use this, portal to, join the data queue cloud space, join your account, join your space, accept the invitation, and all the other assets that you might need you throughout this workshop or after the workshop. Right? So, like, the instructions, the understanding of the use case, the slides that you're seeing today or gonna see today with us, and the post workshop survey, they're all in this page that we built for you. So make sure that you can log in to and, access this, at this page. Right? So without further ado, the agenda today is, I'm gonna present a little bit of, the platform that I queue, give you a demo of what we are building. And that'll probably take, like, ten ten minutes, fifteen minutes. And then we're gonna jump into the fun stuff. So the hands on, workshop on building a multi agent application would be about an hour, twenty minutes, thirty minutes. And then we're gonna wrap up the session with, next steps, questions, q and a, and and a survey. So that's the agenda for today. And, I believe, like, most of you, might have heard about Dataiku or have, some sort of experience. Right? Dataiku is this platform for the AI success, and it's powered basically by three things. So you folks, the business, the people, the data teams, the AI, the engineering teams, the developers. So everybody then is to get access to data and to AI tools, to build enterprise scale projects with, obviously, security and control built in in mind. Right? Second thing that you need for AI success is orchestration. So alongside with the people, the platform, the tools, everything that you need to connect to data, datasets, databases, machine learning models, LLMs, agents across your current infrastructure, but also the infrastructure or architecture that you might have in the future. Right? So you need people and tools in place, to make these AI AI projects, actually take place. And on top of that, you need governance. So you you need to build trust, with your enterprise and governance so you can oversight and control, at every layer. So we can put people together with the tools together in the right way. So that IQ is this orchestration layer that is sitting on top of your infrastructure. So your cloud providers, your data platforms, your legacy systems, your LLMs, your models, you name it. And in one single tool, in one single interface, in one single platform, you can do your AI predictive models, your a generative AI, your predictive AI, your generative AI, and your AGN GKI in one single place. You can build, orchestrate, deploy, and then govern, and trend AI projects. And, obviously, like, you're running your business, you're operating your business, so you need your domain experts. You need, the people that know supply chain, finance, manufacturing, sales, operations, risks, marketing from your company, from your business to be part of this of this journey, to be part of this AI project. So, we are basically linking together the tools and people, in one single place. And we are trusted by, you know, like, more than 750, global enterprises across industries, health tech, FSI, manufacturing. So it's a platform that works and a platform that deliver the value, for, the companies. And a little bit of the like, I wanted to give a little bit of a glimpse of agent success before we jump into the, to the workshop. So is enterprise agents need to be viewed, for reasoning govern governance at scale. And, we have basically, like, three ways of thinking, agent AKI or the agent like era, and DataIQ. One is, how we can build applications. So it's structural reasoning and enterprise decisions. So how can we build applications, agents, that can think the way our organization think, with reasoning paths, with workflows, things like that? Also, how can we orchestrate the intelligence that we have gained across the last couple years about our business, our analytics, our insights, our models, our business rules within the same unified workflow that we are building agents? And, last but not least, how can we have a decision governance built in so I can secure my, my use cases, my projects. I can trace the events, everything that I'm making decisions alongside with my AI projects. How can I control cost, and how can I audit anything that's happening within my my workflows, within my applications? Right? So that's that. Agents, in a nutshell. We have couple of different features, couple different functionalities in which one of the three pillars. So we have in the structure reasoning ways for you to create agents, whether visual agents or very customizable programmable code agents in that IQ, and also ways to expose and ways to collaborate and contribute and share those agents alongside, along with, like, within our teams in in the in the enterprise. Right? So we have agent hub. We have agent, an agent library so we can contribute and share those agents, with your peers. Alongside creating those agents, you need ways to integrate, as I said, to integrate what your business have learned about, your business, about your operations. So semantic models, analytics, predictive models, and toolings, that those agents will need. And that, like, it will offer you, like, a different way, like, different ways to create all of those things in this in this one single interface. And, also with security and governance in mind, so ways for you to explore and audit every action the agents are taking, security controls, guardrails, and dashboards so we can actually, like, manage monitor and manage, what the agents are. Well, agent activities. Right? So that's agents in a nutshell, for us. We're gonna jump into the fun part on into the hands on workshop. I wanna do, like, a, post check-in. Like, if you have any problems, anything, any any any anything at all, like, just post in the chat. We'll be happy to help, Carlos and I. And, what I'm gonna build today is, imagine that you're part of this, company we call MeridianScos, part of the support team, this company. And you want to build a customer three sixty support agent to help support teams like yourself, to jump between, like, systems speech together, customer contacts, understand what what is the customer that we're talking like, it's sitting in front of you and we're talking we're speaking with in the line. So how can I build an agent, a customer three sixty agent, support agent, that can help you with help you with your, day to day activities? So how can I create this agent in a single interface to surface, for example, customer profiles, order history, answer policy and product catalog questions, and help me through with, refund processing? Right? So this is what we're gonna build. We're gonna build an agent that is very, that's very real, very practical. And, so let's check it out. We're gonna start our workshop. Again, housekeeping. We have this Highspot portal that we have put together for you guys, so please make sure you have access to it. We have all the assets in the workshop in in here in this, in this portal. And, one of these one of the assets is this instruction, workshop participant instructions. That's the one, fifty pages. It's gonna take, I promise, less than one hour, maybe one hour. And, so I'm gonna reference this workshop all the way through the, all the way through the the walk through. And, why I'm telling you that is, like, it is very important that you follow along with me also with this you don't need to read all the instructions here, but, follow with me along with these instructions because part of the naming of the datasets, instructions that we're gonna give to the tools, like, to the agents, they'll be here. And, we're gonna save a lot of time using this, this material. Right? So without further ado, this is what we're gonna build in that IQ today. This is, the whole workflow. We're gonna build, basically, a way for our agent here. So this agent, customer three sixty support agent, connecting through different tools, different capabilities. So we are enabling this this agent to connect through different tools, different capabilities, and also with another agent that is so all these tools, all these agents, are leveraging, like, a whole like, a a whole, data pipeline. So I have, the input data about customers, orders, reviews. I have, knowledge bases, knowledge banks with product catalog information, with policy information about my company. And, we're gonna do, you know, like, some data preparation so we can bring all this knowledge from datasets, structured datasets, and unstructured data from PDFs into this, one single agent. Right? So that's what what we're gonna build today. And, but before, we start so you should have already accepted the invitation by now, hopefully. If you didn't find it, please post in the chat so Carlos and I can help you with that. And, I believe you're familiar with Dataiku Cloud. This is the launchpad. This is where, the administrators, of your company or Dataiku administrators at your company are setting up extensions, plug ins, connections to third party providers, and they'll be able to audit trail and understand usage and this, in our platform. And we put a note for you, also, an instance, a design instance where we're gonna spend the next hour together, to build this pipeline, to build this use case with you. Right? So, so please open up the DataIQ cloud link that you have received either by mail or you have opened, this portal here, DataIQ cloud. So please open up this. Familiarize yourself with the the interface and find the node design and click in open instance. Right? So we're gonna go inside of the open instance. You might not be seeing all the projects that I'm seeing now, but you go on the left menu here, you go on projects. You should have inside of the sandbox, you should have a folder with your name. If you don't find it, please send us a message in the chat, but it should be there. You should find it. If you don't find it, please post in the chat, and, we'll try to help you as as quickly as possible. I am gonna use the workshops page. And, once you found, your names, project folder, and, you see we have, like, two projects in here. Right? We have the starter. So we have the starter here and the completed. So the completed is the reference for you to compare what I've done alongside of me and Carlo today after the workshop, or maybe you wanna, check it out, really quickly. But, this is the reference what we're gonna end up doing, throughout this session today. So do not open the completed. Let's just start with the, GenAI starter. Right? So open up the starter, and I think I don't see any messages. Carlos, does this say anything? I think everybody has access to, the folder. So I will have you. cancel later. Alright. Thank you, Carlos. And, once we are so familiarize yourself with the, with the with the launchpad, open up the projects folder, find your name inside of the sandbox page, inside of the sandbox repository, and, go inside of the workshops and find a starter. I see a couple folks don't have the, access yet, but you're gonna you're gonna you're gonna get it. I'm sure, we're gonna find out find out a way to give access to everybody. And, click on the starter, and we're gonna start with this project here. You see, we don't have so I see, so how many people we have here don't don't have the, like, access to their folder? We have two, three, four, five. Only they started. That's fine. So if you don't have the completed folder, that's fine, Trived. We're gonna complete throughout this workshop. We're gonna get it done. And, so see, the workflow that we are seeing now like, the flow that we're seeing now is a little bit different than what, we were, seeing in the beginning of the of the session today, which is fine. It's not a 100% complete yet. And, that's what we're gonna do throughout this workshop. So we want to build or classify and structure reviews from customers using LLM, or more, more precisely LLM recipes, Gen AI recipes in Dataiku. We're gonna join and prepare datasets to create a unified customer lookup table. We're gonna embed documents into knowledge banks. We have some examples here. So I already have, like, policy docs embedded, so we're gonna do another exercise. Then with all the things that we have done so far, we're gonna create agent tools. You're gonna create a simple visual agent. I'm gonna show you what is a structure agent what a structure agent means, on that IQ, and then we're gonna publish our new agent, brand new agent into agent hub. Right? So, this is what we're gonna do today. And, and if you're seeing the instructions, I'm gonna show you. If you're seeing the instructions, we are gonna go through six laps, if I'm not mistaken. And, I'm gonna refer the laps as we go through this session so we can find out, you know, like, inside of the instructions where we are at, and you can catch up with me if you're a little, going a little bit behind. Or if you're going farther and, you know, like, you're faster than me, that's fine too. You can locate yourself where we are, inside of this inside of the project. Right? So, let's familiar familiarize familiarize ourselves with the data here. So in the input data, zone, we have customers, we have orders, and we have reviews. All those datasets here are already prebuilt, already preloaded for you. Like, they're coming from a s three AWS s three bucket. But that IQ connects that natively to virtually any data warehouse, any database, any processing engine, any LLM, including Snowflake, BigQuery, Redshift, pros Postgres, Spark. So whatever your organization is using, this is an example. Whatever your organization is using, you'll be able to connect to, and, you're gonna see, like, an icon like, a blue icon as this one for a, particular dataset. So in customers, I have the general, information about a customer. So customer ID, first and last name, email, region. So, you know, like, the information details about a customer. I have their orders. So the order placed by the customer, which customer, category, items bought, which day they bought, how much they spend, things related to the orders in general. And I have the reviews. Customers have posted if they have posted. Because, so the review, they posted about a certain order. So ratings, one to five stars, and the text. So this product was great, or I didn't like this product at all or it didn't work for me. Right? So, at this point, we are probably all together. We are in the flow already. We are fam familiarize familiarize it familiarizing ourselves with the, with the interface. You can also explore through the preview here. You don't need to jump inside of the dataset. You can just, like, go across the icons and see, with the preview button here in the bottom, all the same columns that we have explored in the beginning of the of this session. And, again, like, we have all these zones. So ways to organize my pipeline are already previewed. So I have the input data zone. I have anything related to knowledge banks and the refund agent, and I have a data preparation zone. Let me see the chat. How is it going? How is it going, Carlos? It seems that a lot of people don't have access to the your folder, so I'm trying to help. Chad is also. helping some of them. Thank you. Thank you, folks. Yeah. Alright. But some some other question to is to to. slow down a bit while we are, like, troubleshooting all their issues. So, let's go slow down a bit just now and so we can help them because a lot of people don't have access to it for for some reason. Okay. Five Samsung's time. More recipes. So I'm just checking out the chat here. If you see this project, it's called workshop auto admin, so you don't have to use it. You have to use the one that is called hands on workshop agents and. That's the one that Luis. is using. So Yeah. so if you see any other project, that's not the project that you should use. Yeah. So inside of the folder, with your name, should only use the starter project. So that's the hands on workshop agents and Gen AI starter. So this one with your name. Right? So just click on it. And, I was just explaining a little bit, like, what what like, give an overview of what we have inside of this project so far. So no worries if you are not following that along, too much. So there's so Chrissy said there's no other project in our sandbox folder called hands on workshop agents. Is there any project inside your folder, Chrissy? Can you go back to the the main page so you can show them how to access your folders? So once you get to that instance, you should go to the left side where it says projects, and you should see a project folder with your name. Yeah. And, Travati, if you've been to any other webinar workshop before before and have, like, multiple spaces, yeah, make sure you go to your, That to your space. So space one or two agents, 04/29/2026. Everybody that hasn't been hasn't been able to access the work like, the project yet, just make sure that in the launchpad, you are inside of this space here. Yeah. Make sure that you are in that day, the '29. That's the one we're using. Yeah. But I assume everybody else is already catch up with me. If, yeah, I guess everybody else is, caught up with the content so far. I think, like, most of us have access to the folder, have revealed the starter. It's already inside of the starter. I'm gonna go along. And, Carlos, please, work with the folks, in the chat, and, let me know, like, if there is, like, a major issue or anything like that. I'm gonna go, show, like, in the lab one. So we're gonna we are just, like, walking through, the zones and the components that we have already previewed for you in this project. So I mentioned about the inputs, the sources of our, data that are gonna work here. We have created a couple recipes already. So, for example, we have this, this this grow by, by key, recipe that's in yellow that is computing, orders and, you know, like, some metrics such as, like, dates, amount, average, order amount, delivery status, things like that per customer. So we're doing, like, an aggregation here. They discuss, called compute customer metrics. So we have that. We also have a join, so we are joining. Once we have calculated, some metrics about, the orders, we have joined, that information with, customers. So at the end, what I have is a customer enriched, dataset with all the customer information and the orders placed by each one of the customers, right, alongside with some metrics. If I go back to the flow, I wanna touch a little bit on this, pink zone here that is the policy knowledge base and refund agent. And we have built a knowledge base, so an embedding out from the, policy documents. So my business have, has policy documents about how we treat and how we support our, customers, throughout, like, any support interaction. And I have, embedded already for all of you. I have already embedded this, unstructured knowledge and structured data into a and embedded. Right? So one agent already is already using this as a tool. Right? So, and we're gonna do shortly, something pretty similar to what we've done for policy talks. Throughout this workshop, we're gonna do four product talks, which is the catalog of products that we have in the company. Right? Alright. So, we have this. So So we have the input data. We are familiarized with the, with the flow that we have in DataQ. We have the, so input data. We have data preparation. We have policy, knowledge bank. And the first thing that I wanna do is, inside these reviews, I have the rating. I have the review text, but it would be amazing to have some sort of, like, emotion sentiment analysis out of this text. I wanna have a sense of, you know, like, whether these reviews were positive, neutral, or negative. So, what I wanna do is, leverage our, Jane AI recipes, LLM recipes to classify this text. So, this is where we are gonna start doing, like, some hands on work, now. So, let's go back to the flow. So if you're not in the flow, go back to the flow. Reset you, you know, like, to so we can see, like, all of the components in that IQ. Click on reviews. So this is the first dataset we're gonna work on. So click on reviews. Click on actions if this panel is not expanded. So open up the panel. Click on actions. And then down here in the pink recipes, the gen AI recipes, you should call them recipes back in the days, we're gonna select classify text. So this is powered by LLMs. And, with these recipes, you can do task specific classes or any sort of classification tasks, like sentiment analysis and emotion analysis. What I'm gonna do is do a sentiment analysis now. So click on sentiment analysis. And please review that we have your reviews as input dataset, and, we're gonna change the name of this, output dataset. So the output dataset won't be reviews classified. It'll be reviews with sentiment. I'm gonna keep all the other things as it is, as a standard as it came in. So storage and format, keep as it is, and hit create recipe. Now that actually is gonna prompt me a a, an interface for me, like, to fill in the settings for this recipe. I'm gonna select the LLM that is gonna be used, to classify the text, the revealed text that I saw that I just saw in the dataset. So I'm gonna use GPT 4.1 mini. In your company, in your enterprise, you're gonna use, whatever LLM that your company allowed you to or know, like, is, available to you. So, and it'll be exactly the same exercise with a different element. Right? So I'm gonna select the GPT point 4.1 mini. I'm gonna select the column that I'm gonna use to classify this text. So I'm gonna use review text. And I'm gonna keep the other two things here, the other two fields as as they are. So tasks is I wanna run a sentiment analysis, plus equal minus, which means positive, neutral, negative. And I just want one result. I just want to have the most likely class. So I just wanna I instead of having the probability, I just want to have whether this review was a positive, a neutral, or a negative, review made by the customer made by the, my customer. Right? And, actually, I can check. If I click on show prompt, I can check how is this, how that IQ is interacting with the LLM behind, behind the hood, behind the scene. So, this is the query that that IQ is using for each one of the lines, and this is what, what like, the field that we're passing through and some examples of, you know, like, oh, this is a text that it should classify as negative. This is a text that it should classify as positive and, neutral. Right? So clicking cancel, it is just an example. We'll keep, edit it, this as is, and we are gonna run this first recipe. And, what I'm expecting out of this is now having a output dataset, a new dataset with a new column that is the predation with the sentiment analysis from based on that review text, based based on that column with the review from the customer, saying it was a positive, it was a negative, or it was a neutral, experience. Right? Review. Okay. And once it's done, we're gonna review, the dataset, the output dataset. Let me, double check how we are doing, in the chat. We are not that far along in the in the instructions. So if you if you are following along with the instructions, we are in lab one, and we are about, I'm gonna show you. So we are about here in page nine ten. Right? And then IQ has just computed the, the new dataset. So I'm gonna click and explore this dataset, reviews with sentiment. If you want, you can go back to the flow, and you should see a new output dataset here called reviews with sentiment. Double click in it, and then you see the prediction column with the positive, negative, and neutral classification. Right? So pretty cool. This is the first thing that we wanted to do, in this, in this lab. We just wrap up, I think, the first, the first lab. We're gonna go to Lab 2 now, and start building the data foundation. And what I wanna do is start enriching reviews, with orders, with customers so I can use, the information that I already have about my orders, my customers, and the things that I've just learned from, like, sentiment analysis and the reviews as as a tool as a capability for my, agents. So they will be turning to tools, later on the pro in the process. And, also, embed the catalog of products. So the PDF that we have about the products that we have in the catalog and turn them into knowledge banks. So, also, they can then be turned into tools so the agents can use all this knowledge that we have, in our company. Right? So let's get back to the, to the flow. So, if everybody's with me, go back to the flow, and, we're gonna do a join. We're gonna enrich the reviews that we have just created. So, if you go back to the join oh, sorry. We're gonna enrich the orders. So click on orders in in the flow. Click on orders. Here, expand the menu, the panel here, the actions, and look up for the join recipe. This is my join recipe. So click on join recipe. Then I can just gonna show you a couple of things for you to set up. So we are setting up orders with with review sentiments. So I I want to put together all the orders that I have from my customers. And if there is any review alongside with these orders, I want to have those reviews and the sentiment that I have just found, with this dataset. And I wanna build, like, a whole big final dataset that will have orders with the sentiment. Right? So, the second dataset will be my reviews with sentiment. So look up for reviews with sentiment, and I'm doing a doing a change in this name. So if you go to the instructions, we have a name a different name for this. So orders reveal join. Keep the other things, the other configurations as as they are, and hit create recipe. And I noticed, that Dataiku has proposed the, the join. So we're doing a left join. So we wanna keep all the orders. And if there is any review, we're gonna put together that review with that particular order. You have many options of joins, in data queue. You can data queue so it's morally selected what is the key for this joint transformation, but it could have join on like any other field if you find there is a better field for you to do it. You can preselect columns or delete columns if you want. Don't touch it. We're gonna keep as it is. And, we can review what the output dataset will look like. So we have all the columns, so we have the orders. We have the customer ID so we can relay discuss this order to a particular customer. We have information about the order, and we have, the reviews and the sentiment about that particular order. Right? Let's see the chat. We are all together here, I think. Once you review this and we are cool with that, we can run the recipe. And, again, like, as we ran the, the previous prep recipe, you should have a an output recipe or an output dataset with all the, orders joined with their reviews, if there is any. So, see, like, I have lines with reviews, and I have lines, without reviews. So now I have everything in one dataset. If I get back to my flow, now I can see that I'm putting together orders and reviews, with this sentiment. Right? So that's great. I think that's, that's great for us, like, to, for this preparation. Now what I wanna do is remember, I mentioned all the policies about refund, about, my company, they are already embedded, as a vector here, so I can use that as a tool for my agent. But I don't have anything for, the products. And, and that's really important because I want to use that catalog of products, the information that I have about my products as part of the knowledge that my agent will have so I can deal Me and the support team, I can have this agent to assist me while I'm talking to this customer. Right? So, let's go back to the flow and, can we set this on here? So let's get back to the flow. And inside of the input data zone, you have products. So we have product docs. Let's go inside of the product docs. So just review a little bit of what is inside of it. So I have all the PDFs about apparels, electronics, home kitchen, appliances. So, like, anything related to the products that I have in my company, all PDFs. Right? So that's what we have as a the unstructured polish that I mentioned to you. So let's get back to the flow, and, I'm gonna embed all of these files, into a, into an embedding, into a vector that I can use with my eight. So click on product docs, get back to the flow, click on product docs, and you see, like, that IQ propose you a couple of things that you can do with your, with this particular component, with this particular icon, and the interface. Let's make sure that we are under visual recipes, and we click on embedded document. So we're gonna embed this document into a knowledge bank. Right? Do not click on extract content. Click on embed documents. So click on embed documents. Now I have, the same look and feel. Basically, what is the input and what is the output? I have the product docs. I could have increased the information this agent would have when using leveraging this knowledge bank by adding metadata if I wanted to. This is optional. We are leaving as it is. And we are gonna keep the name, the embedding model, the vector stores, but we have you know, like, it's good to know that we have the, ability to change all of those things, according to, your use case in your company. So let's create, let's create the recipe as it is. And, for the embeddings, we have basically two approaches. We have text first extraction. So if your document is basically like a huge text file, you can use Text First Extraction to extract extract text from these documents, like, from PDFs, from PowerPoint presentations, docs, docx, HTML files, things like that. We have visual first extraction. So we can use VLM VLMs, visual language models, to extract information from documents that may include text but also include visual elements like charts, graphics, tables, anything related to to images. And we can also create custom rules, which is for particular extensions, I want to use text extraction first. Or for particular sort of extensions, I want to use visual arts extractions. In this case, we're gonna keep text first ex extraction. Our documents are pretty simple or just text, and I'm gonna run. And, hopefully, everybody's with me. We are in lab two. And in lab two now. And we just ran the, let me show you. So we just ran the knowledge base for the product docs catalog. So we are, there is a question. How can you create that folder of inputs with your PDFs? So can you go back to the flow so you can. Yeah. show how how to create the the folder? Yeah. So, I I didn't see the name and the, who put the, put the message the question in chat, but you go to add items. So in your not in this workshop right now, but, like, in a future project, you're gonna go to your, flow, add item, connect, or create. And then, you have in that IQ manage, you can you have the, the option to create a folder and put files in. It. So that's what I've done here. We have created two folders, product docs and policy docs, and then you can import, add file. See, like, you can add files from, like, different places into this folder. You You can also connect to, you know, like, FTP servers and things like that to pull, documents from a particular, fiber storage if if you have that in place. Right? Hope that makes sense. Hope that clarifies a little bit, and thank you for the question. Now I think we should be it's running. It can take a little bit of time, to run this recipe here, but it shouldn't be that long. It's already finishing, closing up. And once this job is finished, then I'll have the embedding. Right? So, like, if I see I now I have, like, a folder with the product docs, embedded images, and the, the knowledge bank. So if I go back to my flow, now I see product docs embedded and a folder with the embedded images that, this recipe got from the, like, from the job. Right? Okay. Now I see, you know, I have a couple of different things in this in, like, in in this input data zone, and, I'm not liking it. I want to organize a little bit better, this workflow, and, that's what I'm gonna do. I wanted to put what is data preparation to data preparation zone and put whatever is, like, agent and policy knowledge data knowledge bank into that particular zone. Right? So if you go back to your flow, you're just gonna go click through. So left click on your mouse or your touchpad. And, so just, like, click through all the components that you wanna move to that particular zone. So I'm gonna start first with the data preparation. So, I have this classified text recipe that I wanna put in the data preparation zone alongside with reviews, with sentiment, the join, and the orders, reviews join. So, I selected four components here. And on the right or maybe if you right click, you're gonna see different menus. But on the right, in that particular panel, actions, I have the move to flow zone. So I'm gonna click on that, and then I'm gonna select that data preparation. Right? So I'm gonna move all the four components into the data preparation zone. Alright? So now I have everything that I just selected into the data preparation zone. And I'm gonna do the same thing that I've done for, the data preparation. I'm gonna do for policy. So I'm gonna select the embedding. So the embedding, the product embedding, the images folder, and the and the embedding. So the embedding recipe, the embedding, and the images folder. And I'm gonna do the same thing. So I'm gonna move, to a different zone. And now instead of the data preparation, I'm moving to policy knowledge bank and refund agents. So let's move it to there. And, yeah, I'm feeling a little bit more organized. I'm feeling better now. It's pretty cool. Now I have everything in place that are, like, where I wanted to. And, I think now we can start creating the agent. Right? And, before, creating the agent itself, what I wanna do is build the capabilities, as we call the tools. So this agent can look up in dataset for information, can search knowledge banks, can use predictive models, search in the web, or use any sort of, like, third party integration, and, you know, like, be be be useful, to react, to act, to automate tasks, and to do the job that you want it to do. And we have a menu of agent or agentic AI, or agents flows, in that I feel that's the one just right on the side of the flow, so this agent, where you can do, like, the whole agentic workflow within, within that IQ. So I can, prompt engineering, engineer instructions or prompts, into prompt studio. I can create semantic models. I can do reviews for agents, see the traces or the decisions the agents are are making. But we're we're gonna spend the next few minutes now. So we are starting lab three now. We are creating agent tools, the capabilities that I just mentioned. So let's go to agent tools. So click on agent tools. Let's see. I have already two tools created. So I have a refund agent that is query another agent, and I have a knowledge bank, the policy docs. Right? And, I'm gonna build with you, like, three new tools. So two dataset lookups and one knowledge bank. And how I do that is I have this new agent tool here where I can access out of the blocks out of the box preview, two recipes that, visually, you can just select whatever you need, your agent to do, and then enable this agent to do, you know, like, sort this sort of function. So knowledge bank search, dataset lookup, query, and the other agent, connect to MCP servers, or even use, like, a custom, custom program that you might have. Right? So the first recipe we're gonna do is, a dataset lookup. So look up for the dataset lookup, tool, and, we are gonna give it a name. We are gonna name it lookup customer. So lookup that's the first thing that I want my agent to, to know how to do. So lookup customer. Let's hit create. And then I'm prompted with, you know, like, the settings for this particular tool. So I'm gonna select the dataset. So the source for any lookup customer what is the source for this lookup customer? Is the customer enriched. So we created that together. Let's find the dataset customer enriched here. Okay? And, I have preconfigs here, so I have to tell whatever, whoever, whichever it's pretty difficult. So whoever, whichever. I'm, personifying agents here. So whichever agent is using, this tool, what is the column that this agent can search through this dataset, to look up and retrieve information for us? So the lookup column is basically what is the searchable variable. So the searchable variable in this case will be customer ID. Cool. And I will leave this data column, so the retrieval columns empty, because if I leave them empty, then I can retrieve all columns from this dataset. And I'm gonna keep single record. I'm gonna explain the retrieval mode in a minute and, keep the user, column to as it is. And I'm gonna give an instruction to this agent. And if you go back to the instructions, I'm gonna, update where we are here. So we are in Lab 3. Boom. Create agent tools. And, if we scroll down till this point, we're gonna see our first instruction. And, now it's gonna be pretty helpful, these instructions, because I'm just gonna copy and paste the instruction for any agent to be using this particular tool. So, I'm gonna go through this for you, with you, the first. So this is the first tool I'm gonna go with the instructions for you, with you. And also the prerequisite, I need a customer ID. So this tool looks up for a single customer. So it's a single record and return two groups of information. One, customer details with all these columns and customer metrics with all these columns. Right? That's pretty amazing. So lab three, around page 24 in the instructions, I just copy and paste the instructions into this agent tool, and I'm gonna save the tool. And I'm gonna go back to agent tools so I can create a second tool. I wanna do a make a pause here, how we are doing, how is the chat. Give me a thumbs up if everybody like, whoever is with me caught up with me. Give a thumbs up in the chat, please. I wanna see how we are doing, so far. We are in Lab 3 now. Joseph, Philip, thank you very much. You can give it a run test. Yeah. No problem. There is a that's amazing. Thank you. Alright. So, we're gonna build a second tool that is, retrieving or a dataset lookup for orders and reviews. So we're gonna do pretty much the same process that we have just created for the lookup customer, but now for orders and reviews. Right? So let's click on agent new agent tool. So if you're back in the flow, you can go to the menu here, agent tools, and then new agent tool. Look up for, dataset lookup. Boom. And, you have a name for this second tool. I'm gonna use lookup orders and reviews and hit create. Okay. Same look and feel. I'm gonna select the dataset, orders orders. So look up for orders review join. That's the one that we have. All the orders and the reviews that go alongside with that order. And if, there is any, right, like, there's any, there'll be a review with the order and the sentiment of that order. Right? Of that review. Sorry. Look up column. Now we're gonna select two columns. So I want this age the agents that are using this tool to be able to search by order ID or customer ID. Again, I'm gonna leave the data columns empty because I want any agent to be able to retrieve any column. And, remember that, on the last tool, we have used a single record for the retrieval mode. Basically, the retrieval mode is whether I want the agent to retrieve one one line or multiple lines. One record or multiple records from this dataset. So now I want to retrieve or I want any agent using this tool to retrieve multiple orders and reviews. Right? And I'm gonna set up the, max records to eight. So maximum of eight acre eight records, at a time. Okay. Then I'm gonna, leave access data as as is. And, if you go to page page 27, we have, a new instruction, and I'm copy and paste from that particular page. So I'm going here. We are around page 27. And, again, like, I'm giving an instruction for whichever agent using this tool. What is the prerequisite, its customer ID, and what is this tool made to paid for? And, what is the result? What is the, structure result, the structure outcome of this agent, using this tool? Alright? And then I'm gonna I'm gonna save. So sorry. So we've done we've created two tools so far, two dataset lookups. And, if I go back to my agent tools, I have, now the capabilities of looking up for orders and reveals customers. I have a way to connect to agents, and I have a way to consult to check knowledge bank on policy docs. But I don't have the product catalog, so I'm gonna create a third tool. That's the last tool, that is gonna allow any agent using this tool to consult product catalog docs, via knowledge bank. So remember, we created that together, the knowledge bank for products. We're gonna create a new agent tool that is a knowledge bank search. So look for knowledge bank search and, promote the same experience we have, seen so far creating a agent tool. So give it a different name, so, knowledge bank dash product docs, and hit create. Okay. Now the interface is a little different because, we're not looking up for datasets. We're looking up for knowledge banks. And I have two now. Right? I have the policy docs embedded that was previewed before this, this session today, and we have the product docs. So let's select the product docs. That's the one we've, we've done together. And, we're not going, like, that deep, like, in all the configurations that we have, but you can configure, like, what is the retrieval mode, similarity score, sources of metadata, and, all things related to this knowledge bank search to this exact exercise. So leave all the settings at their defaults. That's fine. And, on page 29, you see there is a description. Now this is, you need to copy and paste this, from, the the instructions again. So, let's use that description. Copy and paste that description into the description of the tool here. So, I am in page 29 here. Searches the Meridian co product catalog knowledge base that contains information about products such as electronics, apparel, home and kitchen, etcetera. And, so this is the description. Copy paste what the tools the tool is made for, and I'm gonna save it too. And, with if you are if we are all together now, we should have five tools created. Right? So we should have five tools created, and, we should be good to go to lab four. That is actually, like, creating the agent. Hope everybody's excited. I see Matthew, asking for pasting the agent instructions in the chat. I don't know if you can do that for us, Carlos. Or even if you can I know we have already shared the, this page, this portal, but if we can share the the instruction link, I think that'll also help? Yeah. Ryan, accepted us out, so it's okay. you. Thank you, Ryan. Everybody's together, everybody. It's catch up with, it's caught up with work here, I guess. We're going to Lab 4. Lab 4. Let me show you. Lab 4 around page 30. And now we are gonna build the customer three sixty agent. And before going to that, like, we're gonna I'm gonna show you, like, the remember, we had a, agent already prebuilt here. We didn't talk about it. Let it slide, but I'm gonna show you what is what is his refund agent about. And, and after that, we are gonna create this, customer three sixty agent that is gonna leverage the refund agent that is already previewed for you and all the other tools that we have created, to actually be my assistant, when I'm supporting my customers, which are with, like, with with, whichever question they have, in a call, right, in a support, flow. So let's go back to my flow. Let's go back to the flow. And we, I spent couple minutes in the agent tools. We created a couple tools here. Right? Like, the three tools two dataset lookups and, and the knowledge bank search. Now you're gonna spend a little bit of time on the agents and Gen AI models. That's where, actually, you can create your agents. Your you can create the logic, under the hood for your agents to work, and you can relate, relate, you know, but, like, link tools to those agents and link agents to other agents. Right? And you can notice here, I have already a refund agent that is previewed, and, I'm gonna go inside of it. And this agent is, is very special. It's very different than the agent that we're gonna build, in following minutes because this agent is some, it's an agent that we call structured agent. And a structured agent is, an agent that, obeys, follows a particular structure, a particular deterministic workflow and, steps in a process. Right? And those steps, we call blocks. So you see, like, we have blocks here, and this agent has a start to finish. And this agent is kinda, like, going step by step and verifying or, like, take, like, these steps and the outputs. Respect a certain criteria and, so they can move, to the next step. Right? So here, this agent is basically, leveraging a model, a foundational model in in LLM, to check today's date, to the to check customer information, to check whether this customer information, alongside with, with the order that person placed, is eligible or not according to our business policies. And, if, you know, like, if this order this person is reaching out to me through this refund age agent is eligible, then I can approve and dispatch a, refund request or not. I can just, like, reject or, you know, and give a little bit more details on whether, like, this was rejected or not. So this is what the structure agent is doing. This is an advanced agent type built from an order sequence of visual blocks that you're seeing here, including logical branches, reflection steps, agent bend over. So we can hand over things from agent to another, and it'll give you full transparency, in a deterministic workflow, in a deterministic way of, saying, I need this agent to behave in a very particular way. Right? So that's it. This is the agent that we have already prebuilt. So let's go back to agent and Gen AI models, and I'm gonna build, we're gonna build together a, the customer three sixty support agent, a new agent. So let's click on new agent on the top menu. So, again, if you're back in the flow, it might be seeing here, see be it might be seeing something like this now. So get back to this menu here, agent and Gen AI models, the first, part of the menu. Click on new agent. And, just so you know, the refund agent is a structural visual agent. If you wanna do, like, after this workshop, a structural visual agent, now you know where it is. If you you if you're specializing, like, a scripting language for agents like Gantt Langraf or, you know, like, you want to build a fully controllable and customizable, pre specific agent, then you can use code agents. You can code your own agent at that IQ. What we're gonna do now is abstract away all the complexities of, building with code an agent, with the visual agent with the simple visual way visual agent. So that's what I'm gonna do. So let's click on visual simple visual agent, and, we're gonna give it a name. Let me just, get the name on the instructions. We are at about we are on the page 33 now. So let's give it a name, customer three sixty support agent. Let's click click create, and, we're gonna configure that together. So let's select the LN that's gonna power, this agent. So I'm gonna use, GPT4DotDotODot1Mini. Okay? I have my instructions in the document. So if I go to lab five, boom, boom, boom, page 33. No. It's too far. So if I go here, I have my 33. I have my instructions for this agent. So I'm gonna copy and paste these instructions here. Boom. Now I have my workflow. And, so this is done. Now I'm gonna so I have the instructions for this agent to to work, and I'm gonna add the tools. Right? So I have five tools. Gonna select one by one. So the first one is lookup customer. Okay? And notice, when I'm adding the tools here, you're gonna see a preview of what is this agent leveraging. So, I added the first tool is lookup customer, then lookup orders and reviews. See, I added, like, a second tool. Add a third tool that is the knowledge bank policy docs. Great. And you see the views, the preview is getting updated. Fourth tool, knowledge bank product docs. And last but not least, the query another agent. So calling another agent. And you see this agent, is also, referenced in the in the viewport. So two tools that we have previewed for you and three tools that we you and I, we have, built together during this workshop. Right? And you can preview what's, going on with, like, this workflow, with this agent. Yeah. Clicking save. If you see your customer three sixty support agent connecting to five tools, portals, and an agent. We should be good to go. Save it, and let's get back to the flow. Okay? I'll do a pause, see how, how about how how about how is everybody doing today so far? Let me see the chat. Thank you, Carlos, for the instructions. Great. Amazing. Thank you. Thank you for your feedback, Theresa. Alright. And as we did for, like, I think in lab one, let's move a couple components here, to a new flow zone. I wanna organize a little bit of this, for, of this flow before and that that'll be pretty simple, I think. So let's, right. So we are back into the flow. Click on customer three sixty support agent. You can right click to it, or you can click on expand actions, expand actions, and then move move to. So let's move this agent to a new zone. So I'm not using a an existing zone. I'm creating a new zone. And I'm gonna call the zone customer three sixty support. Alright. And I'm gonna select green because, I don't know, just because, I think. So put the customer three sixty support. Choose a color that you might like, might prefer, and just hit confirm. And then you can see your customer three six support agent just, like, moving to a new zone. And now we're pretty close to, where we started, in the demo. Right? So this is the workflow that we saw in the beginning of the session today, with all the four flow zones. So the input data, the knowledge bank from our structured data, the data preparation we did on the structured data, and and the agent zone, so the customer three sixty support agent. And what we're gonna do now is have some fun, play with this agent, deploy it, and, see how it works. Right? How does it sound? I think that's pretty exciting. So, there is a there's various ways for you to deploy and expose agents, in Dataiku. We can expose through chat interfaces. We can expose through programmatic interfaces or APIs, and we can expose through something we call agent hub. That is a centralized portal for discovering, building, and chatting with AI agents. So imagine, like, anyone in your company that has access to, to DataIQ, has a license, has a user, in DataIQ, will join, AgentHub. If if that person has the permission to create agents, all these people, all these teams will be creating their own agents. You're gonna be able to evaluate agents across different teams, across different folks. And, you're gonna see activities. You're gonna monitor how, how is the feedback or how these agents are working. And, you can promote those agents that everybody in each corner of their company are building and promote them to the enterprise as an enterprise agent. So agent hub is displays that you can not only expose the agents that you you created, like, as enterprise agents, but also crowdsource agents across your entire company and pick the best and promote them as the, as, like, enterprise agents, throughout the company. I find it pretty cool, and that's what I'm gonna do with like, we're gonna expose this customer three sixty support agent through this agent hub interface. Right? So I'm gonna go to agent, to the agent interface. On the third menu here, we have agent hub, agent chat, and UIs. So let's click on agent hub. And, this is already prebuilt. And, we're gonna click on agent hub. And, the first thing that should pop up for you is a an app configuration needed, and, it'll have, like, a green button saying go to admin settings now. If you don't see it or if you miss the pop up message, the pop up window, don't worry. First, if you see it, click on go to admin settings now, and I'm gonna catch up with you later. If not, you can just don't worry. Just go, like, to your name, in the bottom left of the screen and go to admin settings. And, whether you click on the button or not, went through the admin settings, that's what you should be seeing now. And, you should select that's the only thing we're gonna fix. You should select what l m LLM model to use, to power agents created and shared within agent hub. So please select the GPT four dot o mini and click save. It should be we should be good to go. And that's the only thing we're gonna do here. Right? So in core settings, select GPT four dot ODot one mini and click in save. That's the only thing we're gonna do in core settings. The second thing we're gonna do is go to enterprise agents and remember let me delay this. It didn't say this. So, enterprise agents, they like, we want to expose that customer three six support agent through the agent hub. So everybody has access to agent hub. We'll be able to reuse the agent that we have just created. So that's where what what I'm gonna do. So let's click on enterprise agents. Click on add agent. And, now this will be a challenge. I think that's the most challenging thing. But, across the entire workshop, I don't think we had, difficult things to do now. You're just gonna find your project. Right? So go through all the projects you have here. Find your name. Go through your project. It should always be seeing one project. I see everybody's project, so that's the most challenging thing for me. I have to go through all the projects here. I could only search, but it's also fine. So, we're gonna find our project, and, you're gonna select let me so let me find it. So select your project. So let me find for hands on workshop, so starter. Yeah. Okay. There you go. So I have my project here, and then I'm gonna select the so you should see two agents. You're gonna select the 360 support agent. So let's go to customer three sixty support agent. Select it and add selected. Amazing. So I am configuring the agent that we have just created, the customer three sixty support agent, as an enterprise agent. And then everybody with access again with access to agent hub should be able to reuse this agent across the enterprise. Go back to your instructions. You should see an instruction on how to do, on, yeah, on the instruction of, that you're gonna give to this enterprise agent. It's on page 41. So copy and paste this instruction here into your description. So this is the customer three sixty agent that gives support to representatives in a single interface for everything they need doing a customer interaction. Cool. And, something else I wanna do is give some examples. So people that discovery discover my agent, how they can interact with this agent. Right? So I'm gonna give a couple questions as examples so people, users can use this agent. So, for example, what is so I have a customer that is asking why their wire land wireless headphones only play sound from one ear to, one ear. What should I tell them? Or this customer wants to refund or pull up customer x information. Right? So things that will give our users a sense on how to use these, agents. So I'm just copying pasting all these sample questions from the doc to my agent. Boom. So I'm configuring again, like, that customer. So I'm exposing that customer three sixty support agent through AgentHub. That's the interface that, it's basically a marketplace of agents that you can have all your company building and reusing agents with each other. And and I'm exposing that agent that I was just created in the agent hub as an enterprise agent. So everybody has access to that agent that I've just created. I think it's awesome, especially for my support teams. Right? So and, now I'm just copy and pasting the the sample questions, that you can how we can interact with this, with this agent. Right? And once once once you've done it, save this, this enterprise agent, and it should have something like this. So I have one agent in agent management that is the customer three sixty support agent with the description on how it does, what it does, and with, questions, example questions. Alright? And I'm gonna go back to, agent hub. So if you were here setting up a enterprise agents, you just click on the top left side back to Agent Hub. And then if you go to agent library, you should see customer three sixty support agent. How are we doing, Carlos? You see any anything? Any questions, we we might address? We're doing good since. Alright. Amazing. Give a thumbs up for us, folks in the chat, please. Send a message. Thumbs up. How are we go how are we doing so far? Because we are wrapping up in two minutes. I'm gonna test this agent, right, like, with the case study. Actually, you can test it out before, going to the case study. So just, like, if you click on agent library or if you click on a your enterprise agent, you just go to, to this interface. You have the sample questions here. You see, like, I have couple different questions. I can just ask one of those questions here. I have a customer asking about a air crisp air fryer, and they said it smells like it's burning. Is that normal? Is that normal? And, I'm gonna hit let's go and see how this agent answer, me. And what's pretty interesting, in agent hub is that not only I can interact with this agent, and, you know, I get the results that I need, for my daily work, but, also, I can gain trust. I can gain trust using it and be sure that the information that it's retrieving for me is correct. Right? So if I click on the details so the agent says, okay. So the burn is now from the air crisp air fryer is normal during the first use. Right? So it's normal. But why? How did this agent get to this, to this answer? So I had the activity. So I had the the traceability, and I can explore a little bit more further down with Trace Explorer. But I have the traceability of all the things that this agent had to do to think through this question, all the tooling that it has it has to go it had to go through, to make the decision, and the sources or yeah. The source of information where it found the answer for this customer. Right? So, in this particular PDF or in this particular PDF, in this page, it found this information, and that's where, you know, like, this agent is reasoning that, you know, like, this is now is a little bit, normal, for first users. Right? And, I'm gonna explore a little bit further, this agent and, in lab six that is basically how can I actually, like, simulate that I'm live? I have this support agent as my tool to help me in front of a customer. So this is a live scenario. It's a Monday morning. I'm working at my company. I receive a call from a customer. Customer is called Lydia, and, she's frustrated. Her most recent order is delayed. She wants answers. So I need to interact with this support agent to help Lydia during this call. Right? So, the first thing is I need to figure out what what Lydia is doing. Like, what is the who is Lydia? What are the information? What are the orders that Lydia, had? So, I'm gonna ask, hey, agent, support agent, pull up customer x. So they've called and upset and want to review their full history before I respond, to give me a little bit more insights on who I'm dealing with, what is the customer profile, and things like that. So now I have more information about Lydia's profile. She's from Southwest, in this segment. She created her first account in 2020. She made five orders in total, first in 2023, last one, 2026, average purchase price of this. And, so it's pretty interesting. I can get to know her a little bit more while I'm talking speaking with her on the phone. And, and she's telling me okay. So she mentions that there's a previous electronics or, like, previous order that include the headphones that only worked in one year. She never called. She's saying that she never call in in that time, but wants to know if the there is any known issue. Right? So, I'm asking the the agent, the support the customer 360 support agent if any customer has ever mentioned about a a order including headphones that had played sound from one ear only. Right? So I'm gonna play out with this. So the customer has also mentioned that a previous electronics order include headphones. Is that normal? Is that a known issue? And, all the things that I've seen before, details, traceability, activity, whatever step that the agent is doing, I can double check. Right? And that's in interesting right now because, like, the agent is telling me, hey. Before we address anything else, ask her, if the headphone jack or Bluetooth connection is fully seated. If the headphones are wireless, then reset them holding, you know, and then give, like, very precise instructions on how to reset. And if the issue persists, then it might be a case for a replacement. Right? And then it's actually what Lydia is asking me. So let's process a refund. Like, I'm not happy with this, product. So can we get a refund? And let me ask my support agent. Hey, agent. Ken Lydia. It is telling me that she wants a refund on this order, this particular order. Can you check if she's eligible and process it? So, and then I have the tools, you know, to discuss, negotiate, and handle objection in front of my customer, that is based on, you know, true data. It's based on the policies that my company has, datasets that I, build, so all the information that is, that is important, that is, in front of me. Right? So and then this agent is telling me, so the refund request for Lydia order has been successfully proceed processed. The order is eligible for a refund since it was placed twenty five twenty nine days ago, which falls within the policy that is a third day return policy, which is amazing. And, again, like I said, you can see all the details how how this agent reason, and what what what are the tools that this agent uses, throughout this workflow. And I don't know what you got, but, that's pretty surprising. Felt pretty, smooth, pretty easy to build out this, this whole workflow. Hopefully, you see the value and you see, you know, how you can, create those agents pretty easily, pretty quickly, in DirectQ and how we can actually, like, bring value, from day one, from using this platform. Again, this is one platform doing predictive, generative, and agentic AI. We did all of the three generations in one one hour workshop. And, yeah, I think it's pretty amazing. Hope everybody followed along. I think I see a lot of thumbs up. So it looks like sound like, a lot of folks followed along. The ones that didn't follow along for some reason, please try to finish yourself after this workshop. Try to do it again maybe, later today, tomorrow. Review the content. I think it'll be pretty useful for your daily daily, daily work. And, if you have any questions, please put in the channel put in the chat, sorry, or reach out to our teams. And before we close-up the session, you have, like, many different ways to move forward, with, your projects, with your learning, within that IQ. So, this is a agentic and gen AI workshop, but we have plenty work workshops. We We have analytics and machine learning workshops. We have MLOps workshops. So join us, on the next work, hands on workshops, on the next hands on sessions. We'd be happy to host you and, have you with us, to walk you through those exercises. If you haven't joined already or if you haven't seen already our DataIQ Academy, I think it's amazing. You have free courses, free certifications, and, like, anything you need for catching up with that IQ platform and with the market, right, with, AI concepts in general as well. So, this is also an amazing content. And third but not least, we have a that IQ knowledge base, step by step tutorials about document like, about any feature we have in the platform, like connections, recipes. Like, anything you've seen today, it's documented in the portal. So knowledge visit the knowledge, knowledge base, knowledge.dataiku.com. And, before I go to questions, please, we have a survey. Please, please, you just it takes one minute and forty three seconds, I think. Forty two seconds, maybe. Oh, you have just posted in chat. Please, if you can share your feedback with us, how we did, with the session today, if you follow along, how hard was it to, to follow along. Was it, like, too easy? Maybe it was too easy. Did you understand us? So anything, maybe, like, other topics that you wanna see in the next sessions, you know, like, in the next hands on workshop. The team, is doing a great and fantastic work to, you know, have this build, sessions, workshops for, each one of us get better, in the platform. So, yeah, any thoughts, anything? So please share your feedback within, with, with the survey that I just posted in the chat. And, again, thank you very much for for attending this workshop with us. I'm really happy to have ran this workshop with you. Thank you, Carlos, for your support. Was a little bit trickier in the beginning. Yeah? But, I guess it was all good. And, yeah, if y'all have any questions, I think that's the time. Yeah. Thank you, Luis. It was an awesome workshop, and it was a great job. Thank you all. Please fill out the the survey. It will take less than a minute, so we want to get your feedback. It's gonna be good. Thank you so much for attending. Have a great Thank you, everybody. Thank you. If there's any question, post in the. chat. We might address this live. No problem at all. Thank you for the patient as well, the patience, in the beginning of the session. But it sounds like, most of the folks have gone through the the exercise today, so it was amazing. Thank you. We're wrapping up I don't know how long left, but, we're wrapping up in a minute. But, yeah, if you have any questions, that's the time. Put us on the spot. Put us on the radar. We're wrapping up in fourteen, fifteen minutes. So any questions, any thoughts, any feedback? And thank you for the thank yous in the chat. Thank you, Ryan, Michael, Rohit, Matthew, Adobe Jen. Thank you very much. Thank you folks who are in backstage. We have a lot of folks in the backstage. The others don't see, but they have a lot of folks here supporting the workshop. So it is great. Carlos, do you have any questions? No. No? questions from my? Alright. Thank you. very much, folks. Thank you. Yeah. Thank you for your, attention, for your time. And, yeah, see you in the next hands on session. Y'all have. a nice week. you, guys. Thank you.