chapters transcript notes
click any line to jump to that moment in the video
0:22 Hello and welcome back to another episode of Agentic Thinking. this week Matias is traveling I'm going to do my best as I can to give you some more insights around Agentic or AI and building things with AI. I'm 0:34 flying solo today just going to be me. However, we're going to jump in today. if you are following the explicit measures podcast, there are a number of conversations we've had around 0:45 less build, less thinking, more building. This is a tool that has been built by Alex Powers and I'm going to do a little bit of discovery. , we have Alex slated for later this week to kind 0:55 of help us tell us how he built this tool, go into a bit more of how he built the , experience here. But what I'd to do is I'd to in a real world example. I've been wanting to spend some time with this tool that Alex 1:06 has built. It's on the Microsoft , GitHub page. I want to just dive into this. , I want to explore this task flows builder tool and see what we can get and how we can , create or have it 1:17 build things for us. , that being said, I'm going to live demo this thing. I'm going to throw a little bit of music here in the background because I'll be thinking and looking at this and working with my agents to get everything running 1:28 and hopefully we'll get something with tax flows built. I'm guessing Alex is probably somewhere on the internet. He may see this and and drop into the chat. That would be amazing. but that being said, let's go over to my desktop 1:39 and I'll show you where we're going to start from. All , let me get the desktop started here and I'm just going to throw some music on the background here as , just in case I get quiet and am thinking on the program. That way 1:50 it's not totally dead silence here on the mic as . All , we're going to start off. , I want to bring up the project here. , I'm going to bring in here a browser window. This 2:01 is the project that we're going to go through today. It's from Microsoft. It's called Fabric Task Flows. And Alex has built this one. And there's a really nice little intro video in here 2:11 describing what is your business use case. You describe to this co-pilot agent and it has a a series of libraries here. And then you describe 2:21 what you want. I have a source system. It has this data. I'm going to bring in data from Excel. I'm bringing things in from SharePoint. What would you recommend agent to build inside Microsoft Fabric? , we hear a lot of 2:32 the times questions from people there's many things to build, there's pipelines, there's notebooks, all these different experiences. What is a recommended way to build something? this is a tool to help you and guide you 2:42 through this. that being said, I want to jump in and go through the tool. this GitHub repo, you can go clone this if you go to the page 2:52 here. And I'll put this link in the chat description as , just in case you want to see this as . This is also in the chat in case you want to go find out and play with this solution on your own. To clone this, you can go 3:04 grab the clone button here in the upper hand corner and then you click on the HTTPS link here for this repo. I've already done this. I've 3:14 already cloned it down to my Visual Studio Code. If you were going to go do that, you would click on this GitHub icon here, and there should be a little icon here that says you want to clone the 3:25 project. , I'll show you how it works. Let me just start over here with a clean VS Code. And as I tell Matias all the time, make it larger you can see what's going on. 3:36 if I wanted to clone this project, I would go here to clone, click the little copy button, copy the URL. I'm going to move this window out of the way. I'll go back here to Visual Studio Code. 3:47 And then here I'll click on the source control button on the hand side. It's going to ask me to clone a repository. I'm going to click on clone. I will then plop in the URL. 3:59 clone from GitHub. I'll hit enter. It's going to say, would you to open this one? , it would typically give you a folder location of where you're going to clone this thing from and show you where it would would exist. I've 4:10 already cloned it, it's just asking me if I want want to already open it. Since I already cloned this project, I can just open it directly. And here's the project. That's a little bit large for me to look at. I'm going to 4:20 zoom out here just a touch. , here is the project inside Visual Studio Code. I could read through the read me here and I get a gist 4:31 on what this is doing. I can also click this little button in the upper hand corner. Let's see if I can zoom in here a little bit. This little icon here because this is a readme file. It is in 4:42 markdown and you can easily read the file with a clean view. here's what I'll do. I'll put the preview here and we can read through this together. get a from a problem to production in 4:53 minutes. That's the gist here. and it talks through what who it's for, how the design of the solution works. You can discover, design, test, sign off, deploy, validate, and 5:03 document. . , I'm looking here pretty much for the quick start here. prerequisites, Python 3.11 plus and GitHub copilot with agent mode turned 5:14 on. . , this pipeline will generate agent prompts and we'll paste these back into co-pilot chat. We can advance and go next through the progress through the phases. , and 5:26 it says a you'll get a shared see the shared workflow guide MD for full pipeline experience. , here's what I'm going to do. I'm going this is really good documentation. There's a lot of things in this. , I'm lazy 5:38 and I'm going to move VS Code out of the way here. And , what I'm going to do instead is I'm going to go directly into a terminal. Hopefully, I can zoom this in a little bit bigger here. There we go. 5:49 this is just a straight terminal that I'm going to use on my computer. I don't really that. I'm going to move it here a bit more front and center here. And what I want to do is I want to first navigate to the directory where this 6:00 lives. We're going to try that first. I know I'm going to go to cd repos. That's a folder on my desktop where I've cloned this to. And I believe I 6:11 want to go into the specific folder here, which is fabric task flows. again, I'll go change directory again and it will be called fabric dash task 6:23 flows as . And there we go. I'm in the project folder. This is the same folder structure I was looking at earlier. because I'm in this folder, there are and I'll show you the folder 6:34 structure again. There are there's an a folder for GitHub. There's agents and hooks and prompts. all kinds of really interesting folder structures up here, 6:44 which if I run copilot from the command line here. , I'm going to run copilot the CLI version here. And what this will do is it'll pick up everything in this folder structure and be able to leverage 6:54 these agents and read me files. , I'm going to just type in the word copilot and hit run. What this should do, if I'm got everything configured correctly here, this should spin up my C-pilot CLI 7:06 license. There it goes. And it just moved it away from my desktop. Let's bring it back here. Oh, it's making me sign in on a 7:17 different screen. I'll do that . And let's see if I can change desktops here. Back to my ring one. 7:27 . Hopefully we're back. I signed in on a different screen. , it should be presenting me back here with Copilot. sure. I'll allow that. 7:39 . it's my co-pilot is up and running. it it shows me here I am in the current folder repo fabrics task flow. And I'm going to just ask co-pilot what is this repo about? What can I do here? 7:50 what does this repo do? Just going to give it ask it some generic questions just I can get a feel for how this project is supposed to work. 8:00 We'll give this a little test here and see what it runs. 8:15 All , listing. , here I'm going to quickly zoom in. Notice that it's listing some files here. , it's reading some files about this. , it's looking through the shared workflow guide. It probably found this because it 8:25 read the readme. , the readme file initially is where it starts. A lot of times when I'm working with agents, you'll notice that if you don't have a readme or you don't have a - definfined readme with instructions or 8:35 details, that's usually where agents are going to start. They're going to start first in the readme folder and then they're going to work their way through all the different other files. in the readme, it said go look at shared workflow guide. It read that and then it 8:46 looked for GitHub agents and skills. it's it's learning what's going on in the repo here and then it's reading through other areas here. found the advisor agent and other items. . it's going to provide me a 8:57 little summary here. this is a documentation driven accelerator for Microsoft fabric projects. All , great. In practice, it contains at fabric advisor plus skills to route your 9:09 work through phases. , great. To give us a guided system for here's our data problem to here's a solution and how we deploy it. , great. let's see here. I'm going to go a little 9:20 bit further here. I'm not quite sure what commands I need to use to get started here. There's some information on the readme page currently, ? Python and there's a shared scripts python. py. , I'm going to ask, how do 9:31 I get started with this this project? How do I get started with this project? one thing I will probably come back to after this one. I'm going to ask a question here. We'll see what it says. 9:42 I may need to turn on a bit more permissive mode here to make it allow to create files and things on my machine. let's ask the question 9:55 and we're going to let it think through how to resolve this problem. good. It's reading back through the read me file again. 10:09 It should tell me what I should run next for command. One thing I'm not quite sure, there are commands in here and I'm trying to figure out what's the command to get started with. 10:22 I'm looking through the repo here on on a separate window to see. . , fastest path to use the pipeline. Install Python. Yep, I already have Python install on my computer. Start the project. All , Python. 10:34 place the generated prompt into the chat. I just want it to run. If you mean getting started as a 10:45 contributor. No, I don't want to contribute. Let run pipeline py manage the flow. Don't edit the pipeline state JSON 10:55 yourself. Yeah, but how do I prompt it? One of the things I'm a little confused on here is how do I get it to start? , let me I'm going to try another thing here. There's 11:05 another thing called commands. , I'm going to slash the commands here. And let's see what we get here on commands. 11:19 I would have liked just a fabric advisor. Let's just type the word fabric. Oh, here we go. Advisor. 11:29 Yeah, that's not what what I was looking for here. All . , maybe we'll just try and run the Python project. it it does say start a start a project by running this Python script. 11:39 let's let's see if it'll run it. Let me do one commands here. I'm going to allow all. I'm going to give it some additional tool pathing here. it's going to be able to use I'm doing 11:50 allow all. This is a step where I'm letting my agent have more control around the repo and the files that it's creating. I'm going to do allow all and turn that on. 12:00 And I can go up here and I can I'm going to ask it to see if it can start a project. please run the Python script. 12:14 and call let's do a very simple project. I think this will be simple. Let's try let's just try and think through loading some files from SharePoint and making a semantic model. and call the project 12:27 shareepoint load. I guess I should probably also describe the business problem. let me just the let me just make a a period here and 12:38 let's just write out the description of the business problem. I want to be able to load a 12:48 series of Excel files directly from SharePoint 13:00 and make a collection of tables that can be used to I guess generate 13:10 multiple Semantic models and many reports 13:22 data will be small in volume and we will have let's just make up a number here 13:35 one file three files per day something that three files per day , we'll let it do that and then we'll just hit run. We'll see what it does. let me move this window up 13:45 here. It's blocking a little bit of the window there. , great. All , let's run that. We'll see if it's able to pick up and run the Python script and start the 13:56 project off. And hopefully this will then inject the Python script will allow me to it'll start grilling me and asking me questions around what I want to build and maybe some questions there. , let's see what what it can come up with 14:07 here. 14:56 . , it's starting to run some skills for me. 15:10 it looks it's able to start the fabric project pipeline. It's trying to run the Python script here. 15:28 it's starting to fill in some of the discovery details. here we go. , , what I think has happened , the Python script has started to run. , we're 15:39 going to start catching the questions that the script is asking us for. . Should I treat these SharePoint Excel load files as a process that runs on a schedule or do you need a refresher, a 15:50 refresher ingestion? , let's just do batch and schedule. That most of the work that I do is usually batch and schedule. , let's start with that one. I'll hit enter on this question. 16:01 it's thinking about that one. , what implementing station do you want for this project? Oh, interesting. , low code, code first or mixed or 16:11 other? Type your answer. , my initial feeling here is to go low code. I'm guessing if I went low code, it would then build out maybe some 16:21 data flows for me to load those SharePoint files. I don't really think I want to go code first. I'm tempted to go with mixed. If anyone's in the chat here, if you're on chat on YouTube, 16:31 do you have a preference on this one? Would you to see one way or the other? we'll get some engagement here from the chat to see if anyone in chat will will present something here as . Let me know what one you want to 16:40 use here. I'll give it a minute or two, just a second or two here for people to maybe make a comment on this one. Do you have a preference? Would you to see one way or the other? 16:54 All . Not seen anything yet on chat. , that being said, let's jump in and we'll do we'll try out the low code first. , I'm click back here in the window. We're going to select the low code and hit enter. 17:13 . , it's captured that note. I've got the missing intake values. Next, I'm saving them and rendering them to discovery summary of your confirmation. . 17:23 Persist the intake and render discovery summary. All . Here we go. Wow. All . , let's scroll up here. We can look at a little bit more of what it 17:34 typed out before we approve everything. Make the screen a little bit larger here. I'll make it full screen. All . , let's see what it did here. discovery review SharePoint load 17:45 gave me my problem assessment. Oh, four V's. . , volume, velocity, variety, and versatility. . inferred signals, , unstructured, 17:57 semiructured, SQL analytics. . Recommending that medium keywords was files there. . Batch schedule and analytics reporting. . Conference 18:07 low keywords reports. All . Candidate task flows. All . we have a data analytics SQL endpoint score is four. It's giving me some signals on how it's 18:18 detecting what it liked here. Basic data analytics and a medallion architecture. great. should I proceed with the discovery summary? I think this looks fine to me. 18:30 Yeah, confirm. We'll confirm it and let it keep moving. Interesting that it picked up the data analytics SQL endpoint. I'm not sure 18:41 if that is the lakehouse SQL endpoint or is there something else going on there? It's interesting that it's picking that up. it did figure out I think this is a 18:52 very accurate statement that it is picking up my Excel files are unstructured or semistructured type files. instead of just being a SQL data table, this makes sense to me that 19:03 it's picking up this semi or unstructured language that I'm using. let's see here. Ask the user. 19:13 I've got the confirmed discovery inputs. I'm running the discovery brief and advancing the pipeline. , great. 19:31 Oh, it's using a lot of skills . great. Reading the discovery brief. Working through the architecture handoff. doing an index something around task flows. 19:41 . Resolving architecture decisions from the brief summary. All . Excellent. And I really wish I knew how Alex was building this thing. This is going to be fun to get him to deep dive 19:51 on this a little bit later on. . I've narrowed it down to the SQL analytics endpoint for losers. 20:01 Checking the actual items there. 20:17 All . Still thinking through some things. 20:29 All . , working through the code here a little bit more. Wow, it's doing a lot of things here. Scaffold handoffs. 20:55 Wow, lots of scripts are running. , it's read my diagram generation Python file. 21:05 All . , next it's designing our architecture. Slick. While it's doing this, I'm going to go look over here on my 21:18 GitHub or Copilot. , I'm going to go over here. I'm just bringing up while it's still thinking there, designing things. I'm going to go back over to my project task flow. And I'm not seeing any net new files here. , even though it's 21:27 thinking and building things, I'm not sure where these files are going on my machine yet, cuz this is the folder directory and where we started things from. , I'm not sure where it's saving or capturing those items. , but it 21:39 seems it's doing some work somewhere. All , it's doing some more handoffs. 21:51 Uhoh, it's found a Windows console encoding issue. Great. , it's writing some shell scripts. 22:11 co-pilot. What's it where is it reading these files? That's one thing I can't quite figure out here. it's reading a lot of files here. 22:22 I'm checking the repo treat SQL endpoints before I initial the handoff. 22:36 reading some items here using some searching things. , see. . I kind of guessed it here earlier on. I said it might be finding something about data flows. , the fact that I said 22:46 SharePoint and Excel files, , dataf flow has shown up for the first time inside this list. That's interesting. 23:11 Oh, also just a quick note here as as I'm highlighting this while this agent is running, I have also found recently when I build with things, you notice here on the far hand corner of my terminal window, we have here we 23:22 are using chat GPT 5.4 as our model for serving this information. , I've also found recently when building with projects having higherend models either 23:33 Opus or Sonnet 4.6 six or chatbt 5.4. When we're doing a lot more of these reasoning tasks, it's a lot more effective to use these higherend models. 23:43 things once you have a good description of what you're building for functions or other things, lower-end models can handle some of those items. But for me, using these higherend models for thinking and reasoning seem to 23:54 make a lot more sense. They seem to give me much better results. I've settled the hybrid design. I'm replacing the scaffold with the 24:04 final handoff content. And then I'll regenerate the diagram updated from the YAML. All . 24:15 Delete and create projects SharePoint. All . . that little note here around projects SharePoint load docs architecture. . 24:26 I was asking earlier about I couldn't find where the files were going. apparently this project folder was recently created. And when I bring down VS Code here, We can see in fact it 24:36 has made a projects folder. at the top here projects inside projects we can see that it has all of the project pieces here. And the reason I wasn't seeing notes earlier 24:47 there's a file called the git ignore file. This git ignore file denotes that every file in here is ignored when I do check-ins or checkouts directly to 24:58 this library. because this getit ignore file is here all of these files and you can see they're slightly grayed out compared to everything else. this is shared and decisions and diagrams. 25:08 These items at the top here are grayed out because they're being ignored. They're not showing me the changes. it is making documents here. It has the architecture applying. 25:20 it has the architecture handoff document, the discovery brief, the test plan. And if I click into these, it's making these documents as we speak. Here's my test plan that it's 25:30 building, the architecture handoff, and these are all currently in markdown file formats. these, even though I'm looking at them in markdown format and 25:40 syntax, I can easily clip over here to viewing them in a pretty way with , the the formatted version of the text here. 25:50 going through the architecture handoff, it's talking about what items it wants to deploy a lakehouse. a SQL endpoint, 26:00 a SharePoint Excel data flow, a semantic model, and a report. , I don't think it needs to deploy the SQL analytics endpoint. It just auto automatically shows 26:11 up. Interesting that it's got a call out here. The discovery brief again, we can look at that as . 26:21 Take a look at this additionally in the markdown. . , it gave some confidence around how it expected to solve these 26:32 problems with the solution that I had. . , it's showing some of the architectural judgment calls. Interesting. 26:46 Favorite scheduled SharePoint file ingestion over streaming components. . Great. Yeah, I this one here. This one was a good observation. Model is reusable through conformed 26:57 tables and multiple semantic models. Great. All . That was as my as I expected. We'll go back here to the co-pilot 27:07 agent here and see what we're picking up here. Checking the review outcome test plan skills. 27:17 I'm checking the test plan fields I can add the detail. All 27:32 I've got the corrections. It's editing and modifying my files still updating my test plan. 27:44 Oh, the pipeline is at the human signoff gate. I'm pulling the exact signoff context I can present it cleanly. , great. Whoa. All . This is probably just formatted funny because my browser is 27:55 interesting here. Oh, it's showing me a diagram here. , it's this is a diagram view. ingestion wave one 28:06 SharePoint Excel starts with data flows drops down here to lakehouse SQL analytics endpoint comes along for the ride then drops in drops into a 28:16 semantic model and then also drops into serving the report. , seems fairly straightforward. SharePoint Excel files land on a schedule into a shared fabric lakehouse 28:28 from those curated tables. ASQL endpoint feeds semantic models and ports the same data layer can be supported by multiple needs. Great. 28:41 . , asking the user, do you approve this architecture or should I send it back? , I think this did a pretty good job first time through. I'm going to approve it. Let's hit enter here on the approve. 28:52 Let's move on. Deploy to a live fabric workspace or review the artifacts only. 29:03 let's deploy some stuff. Let's see if it'll how far it will let me go here. Let's do it says only update the artifacts recommended. 29:14 What should we do here, chat? I'm going to go live artifacts only. I'll go with the recommended step for . Maybe we'll do this again later and go a different route. 29:42 . we're generating some more artifacts. it's creating things here. You can see here it's also building these different files. no live fabric items. . , 29:52 it kicked out of the project here. , let's go back over to VS Code. We're then I'm going to go research or look at what's going on here. I'll zoom in here to make this a bit larger. 30:02 we've got the documentation section done and we have an entire new section called deploy. And here we have the deployment manifest. Do we have a 30:13 configuration deploy SharePoint power Python file and some other JSON files here as ? , let's look through some of these files 30:24 here. Just a bunch of configs in that one. Here's a configuration. 30:37 It's generated by deploy script. deploy. Whoa. . , this looks this is the SharePoint load project. This look at the this is the actual script here. 30:48 All . Let's see if we can compare it to this one. , it's giving me a little bit of a banner here. It's getting me some credentials here. to sign in, 30:58 we get some off headers. Enter the workspace name. All . , I'm not sure how to use this, but I'm pretty sure my agent knows 31:08 how to use this. , let's go back over to the agent here again. , that we're we're back into Copilot, let's go ask we want to 31:19 specifically ask around the deployment deploy SharePoint. py file. How would I deploy this project using the let's call it 31:30 deploy dash sharepoint load.py file. 31:51 And we're going to ask a little bit of clarity here. see if it can think through this. , what I think it's going to do here is I think it's going to go read that file. It's going to go find the deploy SharePoint py file. It's going to 32:02 understand what it's doing there and then help us understand how to deploy things. . , it's going to read the deployment script and tell me how to 32:12 work it. All . There it is. It's just I thought it was going to do. It went in and read the the Python file directly. 260 lines. All 32:33 . , it's got a couple arguments here that we can pass into it. . Interesting. , it's giving me a workspace ID here. 32:47 I'm not quite sure how I sign into my fabric tenant because I have a couple different tenants that I to use. it' be nice for it to allow me to pick the fabric tenant where I want to deploy this into. 33:01 My guess is we're going to need to run the deployment SharePoint load Python script 33:21 on the door. based on the Python that I'm looking at. , I'm going back into the Python to see if I can figure out falling back to browse. get credentials. It's asking me for 33:31 Azure identity not installed. It's going to install the identity item if it needs it. authenticated by Azure CLI. Azure CLI is not available using the device 33:41 flow. Falling back to browser signins. it's got a couple different O patterns it's trying to use here. , AZ login looks it's going to be best. Azure CLI works best if user has a 33:52 login. Best for terminal and CI. . 34:03 Whoa. . We got a lot back here. Run the script. , great. This is what I wanted. the link step. , set the location. Run the Python script. , here is 34:14 the mode. , what does mode mean? My fabric workspace. If you want to run the guided version, just run Python. 34:24 Oh, . 34:34 Choose a single workspace. All . , we're going to just wing it here. We'll see what happens. let's see if we can change directory. , we are currently in the wrong directory. We are under repos fabric tax task flows. We can see 34:44 that here. we can see that we're in the fabric task close item as here. And we want to go a bit further into projects and SharePoint load and the deploy 34:54 a login level here. let's just I'm going to dive a little bit deeper into the folder structure. let's do cd see if I can change directory here 35:04 underscore projects. Let me go there. 35:16 didn't seem to change the directory there. How do I change directory on this thing quickly? Maybe I should just open up a new window. 35:32 I don't think I want to do that. I don't want to think there. Let's get out of that. Escape to cancel. 35:51 I don't think I want to do all this. I think I want just to run the command script on a separate window. I'm going to go to here. Yeah. See, I'm I'm in , it says 36:03 it's in this new PowerShell script here. , let's see if I can just A 36:28 I know a login is a PowerShell. Let's see if I can go to PowerShell here real quick. , let me hit Windows. Let's go to PowerShell. 36:39 I've got a little PowerShell window open here. Make that bigger as we can see that one. All , let me see if I can run this command as it's stated here. First one is the set location. 36:50 Past anyway. it ran the first one, the set location. it did change my directory. it's going to try and run a login. 37:01 Let's see if we can run a login on my computer. All , we've got some browser here. let's jump in. I'm going to pull it off my window here and go find my email 37:11 address. We want to sign in with. All . Hopefully, I'm going to be able to log in. Retrieving tenants and 37:21 subscriptions for the selection. Great. 37:39 I do have a lot of tenants attached to me. let's see here. Let's go. . it's looking for different subscription I have access to. Let's do 37:55 I think I want to type in number one. See if it'll let me do one. 38:16 maybe I logged in with the a login. Hopefully that worked. , let's try running the second command here. let's run the Python. we do need a workspace name here. I think I'm going to need to adjust that 38:26 name. I'm going to grab it all the way up to the first quotation mark. and then go back over to the PowerShell and we're going to try and run that 38:36 here. , let's drop that in. And then I want to add this to SharePoint. 38:46 Let's go underscore load. Let's see what it does. Fabric CI/CD is not installed. Install it . Sure, why not? 39:05 Lots of code ripping. . Oh, there it go. , that was the message in that Python script earlier that we saw. It is checking out the task flow. . Great. SharePoint task flow data 39:15 analytics SQL endpoint. Authenticated the Azure CLI. Great. , created . . , it's already found my tenant. It was able to log me in because I'd 39:25 already previously signed into a login and it's asking me to use this workspace. , let's do a yes on this one. We'll say yeah, go ahead and create the workspace. I don't have one created . 39:36 it created that. Fetching some available capacities. 39:46 let's do number two capacity. That looks it would work for me. . Doing some things. Whoa. It's doing all kinds of stuff. 39:57 Publishing workspace folders. Doing a variable library. it's confirming that it's loading some stuff here. 40:15 inside the variable library, I can see here it's doing publishing the variable library. , I should zoom in differently. Here we can see it's publishing the variable library and it says the target environment of not applicable does not match. You need to 40:26 have actively change changing the set to default set. This is just a function of the variable libraries that it's trying to use here. that's no problem there. we're on to publishing the 40:36 lakehouse. We can see down here lakehouse is being built. 40:51 All . we're on to a demo semantic model. Publishing a semantic model here. 41:11 Data flows will be not checked for dependencies. . All . Deployment summary populating variable library blah blah blah. . Town found 10 variables. . Deployment summary. There should 41:23 be a project. The project is called that. Yep. The task flow is named this. We have six items across five ways. And the workspace is called SharePoint 41:33 load. All . Oh, look. It even gives you a link to it. That's nice. let me pull up a browser window that we can go use. and I'll bring up a brand new browser window here. And hopefully 41:43 when I click this link, it'll put it in this browser window. , let me move that off the screen here. And I'm going to to click on this link. 41:59 It's signing me into a browser window. 42:10 I changed my browser here. Let me move it back here. hopefully this is back . we're back onto our screen here. And I believe I'm going to bring this back down . this is my tenant. 42:21 We can see here that I have the SharePoint load item. the item has been created. The workspace has been created. this is a workspace in my tenant. it's been associated with a 42:33 premium subscription. Awesome. Has a little bit of a a note there as . And it built a bunch of folders. in the folder it has a folder for report analytics the configuration for the variable 42:43 library ingestion using the dataf flow gen 2 processing the data using the semantic model and storing the data using 42:54 lakehouse and the SQL analytics endpoint. Let's just quickly go back to the I'm going to go into the analytics folder and let's go to the report here. I'm 43:04 going to just say view lineage. There it goes. Report links to the semantic model. Excellent. I think that's a pretty 43:15 substantial demo of that. We just asked it to build a simple folder structure and it went from thinking about it, ask me a couple questions, and we have a full ingestion, storage, 43:28 processing the data, and a report page on top of it. , we could probably go into each one of these items. I'm assuming there's nothing there. I didn't give it any reference to what was in the lakehouse. 43:39 if we open up some of the individual items here, we can see there are no tables here connected . Let's go back to the SharePoint load folder. 43:51 Let's go into the ingestion step here. Let's see. It's got the SharePoint Excel data flow. let me look in this data flow and see what it built inside the data flow. 44:06 . , it's pretty much empty and blank here. , it made just the item. It didn't bring in any sample elements here. , this is where I would have to go figure out how to import it from here directly from SharePoint. 44:16 let me close that out. There there was one item in here that I didn't quite understand. There was one item in the workspace that was named the 44:31 see here what do we have left in this list here. Let's go back up to the SharePoint load configuration. This is the part I didn't understand what it was building here. 44:41 let's just see what the variable library looks . . Interesting. it's already picked up a 44:52 it made up a general string here. What is that for? Leadership report semantic model lakehouse raw lakehouse. . Interesting. . 45:04 it made some variables here. Environment name dev deployment timestamp. . Interesting. This seems it's pretty set up pretty already. It's 45:14 using these variable libraries to help me inject different elements. It knows there's a semantic model. It knows there's a lakehouse for this library. Has a data flow reference. , it's a 45:24 gooid of the data flow. I I would assume that this 80c is the same gooid of the data flow. , it's referencing that as . if I go back to ingestion and click 45:35 on the data flow itself, I would assume yes, we do have it. That's really smart. We'll see. Let me see if I can zoom in there a little further. 45:45 the gooid of the data flow is being shown in the other item. you can see that it's referencing this item. pretty neat. I think this was a really good example here. with that being said, I think we'll go ahead and 45:55 wrap. We hope you enjoyed this little quick demo of the task flow fabric task flows tool that Alex Powers has built. Just a quick little demo of how to use 46:05 it, get started with it, start talking to it. hopefully that helps you go through your flow on your own. With that being said, thank you much for another Agentic Thinking podcast. This was more of a demo for this time, but 46:16 hope you enjoyed this episode and found some things you could learn from here as . we'll also post this one back up on our YouTube channel with a nice portrait light widescreen version of this as , you can see everything 46:26 in not portrait mode. Thank you all much and we'll see you next time.