Setting up AI powered analysis environment on Windows

In this screencast, I'll show you how to set up an AI powered data analysis environment on Windows. I normally use a Mac, so there might be some weirdness here when I try to navigate myself around Windows using a Mac keyboard and a Windows computer. This is also a remote desktop connection, so not my real computer. So, it might be that I accidentally hit Mac keyboard shortcuts and do something with my own computer instead of this remote accessed computer. So the computer that we have is a pretty basic uh desktop computer. So there's 16 gigs of memory and there is no uh GPU to speak of. So we will be running a large language model locally with with the CPU that's probably going to be very slow. Uh but that is something that a lot of people would have to deal with if they run to run this locally. But if you do data analysis then the speed mightn't actually matter that much. For example, if you classify survey responses that you have in your Excel sheet using a model that runs on your local computer, you can just leave it running overnight. So, uh if it takes a minute to respond, then uh it might not be such a big deal. So, and I'm also using Gemini here because uh we we need to do some things in PowerShell to to set up things. And the reason why we need to do it in shell instead of just point and click is that I don't have an administrative right to this computer. So this is like uh the worst case scenario for installing anything. It's a shared computer uh fairly tight security policy and no admin rights. So I will have to install everything in my home directory. And uh if you have admin rights to your computer when you install this and you want to install uh things uh for all the users which is a lot simpler. I'll uh explain the differences as we go. And the overall plan is is here in notepad. Um so we're going to install first LM Studio then we're going to install model. We're going to test a few settings and then we install NodeJS. So this is the hardest part because it's we need to install it using PowerShell and and just copying the files into the right place because uh install it with the installer would require administrative rights which I don't have. Then uh the rest should be fairly easy Gemini we install in node then we test it then we install the position uh data science editor that's for R and Python coding. Uh it's like a more modern version of R Studio built on Visual Studio Code. Uh then we install R and then we install Python and we we test a few things and I marked here uh install location because I need to install NodeJS manually. Just copy the files. So I'll mark where I put them here in my notes. So let's get started. And we don't need to know the details of the computer anymore. So let's just go to install studio and now because I don't have admin rights I will just install it for myself and now I will make note of this install path because I wrote the NodeJS in the same place. It can go anywhere in the computer, but I think it's it's nice to just have everything in one place. So, that's where I install all the programs. So, we're going to install it. It's now installed and we run it. So, now that we have LM Studio installed, the next step is to install a model. And you install models from the search. And what I recommend is to pick here uh have this only include staff picks that fit on my device. This to be on because it may ensures that you only load models or download models that actually work on your computer. And this is important because if you try to load a model that's too large, it either fails to load or it can try to load and then it crashes your computer or makes your computer like completely unresponsive because it starts to use a hard drive to extend the memory which is really slow. Um, we're going to take a model from the Quen 3 family and we will go for something small because this computer is is not very powerful and also because the internet connection is pretty slow. And let's see what we have. And I want to have this uh let's see non-thinking model. We want to have a hybrid model instead of a of a non-thinking. So uh Quen 3 4B this is uh a hybrid model. So we can uh enable and disable thinking. So if you want to learn how these models work and play around then uh this uh hybrid model will be useful. So we're going to download it and if you want to have a smaller variant of the model you can pick the variant here. So there are different uh quantizations and quantization basically means that it's uh the the model weights the matrices the numbers are stored with less precision. So if you have less precision then the quality is a bit weaker but then again it it runs a bit faster and it fits uh into smaller space. We'll take the one that is recommended. If the only reason why you would ever want to pick something that is not recommended is if you have a a very very small memory in your computer like if you have something like 8 GB and you are running something else be except uh in addition to LM Studio then you might go for even more smaller model like you might go for 1.7B and then here there is uh this this uh this model is even more aggressively quanticized. So it is uh like uh 30 to 40% smaller than the default model. But we'll go for the 4 billion parameter model because I think that's going to work pretty well on this computer at least on a reasonable speed and uh we'll start the download. All right, the model is now downloaded and we will just click on using chat. So that starts loading the model and what we'll do now we'll we'll test the model and um then we'll adjust some settings and just see if we can break the model. So, we're going to create a new chat and uh let's switch thinking off because I want to just see how it works. So, let's do hi and let's see it responds. Okay. And uh we're going to start a new chat and just play around with the settings a bit. So, we go new chat and let's do this uh small exercise of of checking how the temperature affects the response. So, all right. So, let's see the default settings what it does and it's it's okay speed. So, I think this is uh about the size of the model that this this computer can handle. I could go for the 8 billion parameter model, but it would probably be substantially slower. If I wanted to do like offline data analysis where I code data or like zero responses on my computer, then the speed wouldn't matter because I would be able to just run it overnight. Okay, so this is uh what it does. And if we want to make it a bit more creative, we [clears throat] can adjust the temperature to increase the randomness of the response. So, we're going to take uh disable these. And this top case sampling means how many different tokens it considers when it picks the token that it produces next. So, we're going to put it to 50. And then let's increase temperature to uh let's go 1 1.5 and and see. And we're going to branch the discussion and regenerate response. So, now it should be a lot more creative. But still still stay coherent. Hard work and creativity. Okay, I like that. Uh so so this is uh it's not much different. Let's let's ramp it up and we're going to branch again. So we're going to set temperature to three. So so larger number is more randomness and top k means how many different tokens it considers. So let's go uh with temperature three. We can now we can see a bit more creativity like vibrant national independence traditions. sound a culture of blueberries a modern social media presence. That's that's a interesting list of three and uh we can see now that it's uh it's starting to break apart a bit like there's this is not grammatically correct and that's also not grammatically correct. Uh if we ramp up this to let's say let's say 10 and uh let's put 200 here. So it consists a wide variety of tokens and picks the response randomly. This is probably going to entirely break the model. So uh it starts to Yeah. So it it no longer can can even uh write grammatically correct things. So when you increase the randomness uh what you can do is to make it more creative, make it more kind of surprising what it produces. But when you increase the randomness too much, then the model loses all coherence. So if you if you put the temperature to zero then uh it means that it all there's like not much randomness and we we can stop responding if we put this top case sampling to one. So that means that it only considers the most likely token. And if we branch the discussion, we generate the response. You can see that leaks, forests, traditions, music. This is pretty boring. So it kind of like states the most obvious thing. And uh another interesting thing is that if we remove randomness uh altogether, if we branch again and we regenerate, it will regenerate the same response. So uh if you think about a model that is is completely deterministic, it's pretty boring. Like you can think about a model uh that is trained to do jokes. If the model always does the same joke, then it's not a very good joke model. Okay, that is set up and uh we need to do do we're going to reset all this stuff and reset to defaults and the next thing that we do is to go to the developer and we we launch the server. So this means that uh this LM studio can now be accessed from other programs and we leave it running and we'll later access LM studio from positron editor. Okay. So, so this this is now done. So, so uh [clears throat] next uh is installing NodeJ GS and this is where probably need Geminina's help and we're going to go to NodeJS and this is like an an execution environment that you need to uh to run certain software. [clears throat] So, there are three ways to install it. Uh this is if you have Docker or some some other um environment. We don't so we're not going to do that. Or even if we did, I wouldn't know if this computer has one. Uh this the easiest. If you have admin rights, you pick Windows installer, but I don't have admin rights, so I have to install it manually. And what I do here is that I I open the standalone binary. And this is like just the files of NodeJS JS. And then I will need to put them somewhere where I can somewhere where the computer can access them. And then uh I will now open a new new uh tab or new window. So let's do new file explorer. Yes, this one. And then I go to my install location. So I just copy that and then then I create a new folder called NodeJS. So this is you can create this anywhere in your computer but but because it install Adam Studio here I'm just going to install everything everything here. So it's uh it's going to be NodeJS and then this is the NodeJS files. I'll choose them all and I copy them here. And again, this is something if you don't have admin rights, you need to just copy this. Otherwise, you would just run this installer and it installs everything. All right, it's here. And the next thing is that we need to set path. And the path is is uh a variable that the computer stores that tells the computer where to find programs. And um I I'm not very good with using using Windows. So I will now use Gemini to help me out. And um I will tell it that I'm running Windows 10 and then I have NodeJS in this uh this location and how do I set the path? And I've tried this before and Gemini has this option of thinking and fast. And I found that thinking is much better for this because the fast gives me kind of like things that might work if I have admin rights. But uh yeah, and and the reason why the the backslash direction was incorrect is that I don't know how to type the the correct direction on a non-Mac keyboard. So my keyboard, my Mac keyboard wouldn't wouldn't uh type it properly. So, so we do uh Windows key uh and then we type env and this requires admin rights. But we can do that. And okay, number one, option [clears throat] number one didn't uh didn't work. And let's say let's tell it that this didn't work. And the reason why I'm doing this uh on camera with Gemini is that how you set the path might depend on the Windows version and what kind of uses us use us uh user rights you have to the computer. I'm just kind of like showing uh the the process how you can you can fix this yourself. And uh yeah, so this all this uh tells you how to figure it out yourself. We're going to use PowerShell. And uh so so this uh script writes the the path to to our environment. And um let's do PowerShell. And this is uh typically when you work with large science models to troubleshoot uh computer issues uh using the the PowerShell or terminal in Mac is very handy because you can just copy paste what you see from from the from text and then the model understands text and can help you. So, we're going to copy it here. And uh Okay. And now we can do uh you must close and reopen the command promp from code editor. So, um we're [snorts] going to open it. And then we we do powers again. Not there. and then type node v. That will tell us the nodejs version. So that is now installed. So we have now nodejs installed. And the next thing that we need to do is to install gemini cli. And uh it's important that you you uh close the powershell and reopen for two reasons. One is that the path uh only takes effect after you relaunch the uh the terminal or what whatever command prompt whatever you call it. And the second one is that you want to ensure that it works after restarting because we want to uh have the the the path set correctly also in positron and if we have to fix it here every time we start then it's not a very usable solution. So that works. Then we go um Gemini CLI and there is an installation instruction here. So we will uh do this install globally with npm. So this is with with NodeJS and then we we paste it here [clears throat] and this is something that uh happens because of the security policy of the computer. So uh what we now need to do is we tell Gemini that uh I am trying to install Gemini I get this error and then it'll fix it. So again, this is like uh a Windows Power user might remember what settings they need to change, but I'm not one of those. So I'll just uh I'll just do this. And again, this is like uh I would use the thinking model here so that the model gives us the best quality responses. It's really frustrating if the model gives us some a lot of different things to try and and then it fails. So we want the model to spend a bit more time trying to figure out what will be the best solution for us given all the data that we have provided and uh permanent fix try this first. Uh, all right. We're going to try that first and then so that allows us to uh install and uh we do yes but not yes always. And then we uh [clears throat] run that again. And now it should work. Yes. Now there's another security issue or or policy issue. So as I said this is like a worst case uh uh scenario because this this is uh a computer class computer. So it's it's very constrained what you can do with this computer. Um, but if you can get it uh get something installed on on on this kind of computer, then it's going to be much easier because uh if you have your own computer that you are running, then uh the security policy might not be as as strict as on this computer. [snorts] And now it gives us another command that uh changes the the installation policy. So you might want to pause the video and read all the descriptions. So I've done this before, so I'm not I don't stop and read this. Uh we're going to do this one because we want to be able to run the Gemini directly instead of uh using this Gemini. We don't want to have this this uh this problem. So we'll just do this security bypass. Yes. And then we'll install. And now Gemini CLA is installed. The next thing is that we we do is that we try it. So we do Gemini. All right. Now it's loaded and it asks us to log in. And we're going to log in with Google because I have the Gemini Pro account. And um we're going to use Microsoft Edge because that is what we have here. And we're going to then authenticate. And that's done. And [clears throat] now there is we know that it works, but there's not a bit of a problem. And the problem is that we are now on in our home data folder. And uh if you uh if you're in a folder then Gemini CLI has access to all files in that folder. So before we start testing it we will create a new folder and we'll change to that folder. But uh before I do that I'll adjust some settings. So, uh the the setting first setting that you uh you type / settings uh the first setting that we change is uh preview features and and this allows you to run the latest models. So, we're going to switch that to true. And this gives us access to Gemini 3. And then we do model. And uh Gemini 3 is now enabled. There used to be a wait list. So, if you don't have it here, then it might tell you that you need to uh go to a wait list, but I'm not sure if the weight list is valid anymore. It was valid when uh Gemini 3 was launching. So, we're going to go for for the pro models. And this is just for testing and showing how it works. uh if you use this uh for for data analysis then I would recommend that you go for the auto because uh that saves you saves you tokens and uh I I pay for per month for this Gemini and that means that there's like a a quota that I get to use it and if I run uh over my quota then it automatically switches me to uh to the flashlight model which is the cheapest model and I don't want to spend all my ext expensive model time for simple problems and I would go for auto but let's go for Gemini 3 Pro just because I want to ensure that all these demos work that I'm going to do with with single shot instead of trying and fixing so we go for that and then we we quit and now we need to do uh another uh new folder so so we will go to uh we have this PC and this is my data and uh [snorts] well There is uh there's a I'm going to create a new folder and let's call it test to delete me. I'm just going to call it delete me because then I will remember to delete it later. So this is just for testing. And we we're going to then change to that direct directory. So we do cd change directory and then we can we can copy here. There are there is or we can just double click and then here copy address and then we paste it here. And now we are in the test folder and the test folder is empty. So we know that now it's safe to run Gemini here because Geminina can't read anything. And then we'll run Gemini again to see how it works. And now it's all set up. And then we can ask it to do things for us. So, let's let's make a fun game. Uh, program a small game for me. It'll probably present me a plan first and then uh after the plan, it'll just implement the program. It's probably going to make some kind of space shooter or uh a warm game, a snake game. Uh let's see snake game and then it tells about the implementation details and it runs in a browser and that's like the simplest thing to do and let's say that go for it and then it'll program it for us. You can see here that it's going to start writing u files here and it asks for our permission to write it pretty soon. So it'll design the game and then ask for permission to write the file for the game. Okay. So now it's asking permissions and we say that always allow and this is only for this session because we wanted to be able to edit the file. If you were doing something that is uh let's say more serious uh and and you actually knew how to program yourself then you might want to review the changes. So you would uh review what it wants to do and just allow once so that you understand the changes that it does. This will be important like in bigger project but this kind of like wipe coding we just uh always allow and that it created the uh three files and then it's checking uh ls is just checking if files exist we always allow so it validated that it actually was able to write the files and then it tells that doubleclick index html and that opens the Okay. So, this is the game and I'm going to reload it. So, so this is a snake game as the snake grows when I eat these apples. The snake is too fast. Make it make the game easier. You could also tell it to make it multiplayer or or make it uh and now it says that now it uh it changes this settings here for the game. And then when we reload it then it's a bit slower I think. Yeah, it feels that it's easier to play now. All right. So this is Geminina CLI and now we're going to quit it and we close the PowerShell. We don't need that anymore. And um then uh we'll close this. So now we are we are test by coding game here. And now we install the Positron editor. And uh we take a new tab. We take Positron and let's install it. So this is just a basic installer. It tells you that we need to do things before we install Posyetron. So uh it tells us to install uh Python and R. We will install Positron first because then we can use uh Gemini CLI inside Positron to help us [clears throat] troubleshoot any installation issues if they arise. So, we're going to uh download it. And if we don't have R or Python installed, then uh this doesn't really do anything for us. It's just like a text editor, but we can't run any code because we don't have a an environment installed. And we take the user level install. If you have admin rights, then you can take the system level, but I'm I'm only have user level. So, we're going to download it and then run it. And now it's in it's downloaded and we wait for it to become available to us to open file. Yeah. And we accept. I've read this before. And then it installs. Now it installs in the same location as everything. And we install. And uh we can we can have a desktop icon. I like it on Windows. And now it installs and we have to wait. Okay, it's installed and we launch it. So this is the the positron main window and we need to open a folder where we uh we work in and I'm just going to open the same folder that we were using before and it it tries to start a console probably R but we don't have R installed so it's not going to do anything. And we're going to open uh a folder because you always need to work in a folder. And the folder that we have is going to be the test to delete me. So we we open it. And yes, I trust because I'm the author. And then we can just delete delete this stuff. We don't we don't need the game anymore. It's not that great of a game. And we we delete it. Delete permanently. And there's no session running because I don't have R installed. Uh I just wanted to first test that I have uh Gemini CLA working here. So uh the this the terminal is where you run the Gemini. And I'm going to move this terminal to the right uh uh secondary sidebar here. The reason why I'm moving it there is that if I if I want to do some some coding, I want to see the file here on the left and then Gemini on the right. So I discuss about the code with Gemini and uh we can just type Gemini and hope that it works and it doesn't. So so this is the reason um uh we need to we need to uh why why I want to test this first. So there is an issue with the path and then we we go back to Gemini uh the the browser and then I tell that uh I installed Oitron and I'm using terminal in in that software I get this error. we copy then we paste and then it tells us how to fix it. So the the the problem here is that because we we install it manually by copying the NodeJS files just on on the computer. Uh we need to uh adjust it. We need to tell both the PowerShell and Positron where to find Gemini. If you run the installer for NodeJS, then you don't need to go through any of this stuff. Okay, so now it tells us which commands to run. We're just going to copy paste. Let's let's say that let's proceed step by step because we don't want it to uh to write a lot of instructions if the first step fails. So so this will be um setting the paths is is what is makes this complicated. It'll probably ask us to next check with PowerShell because it was working there and then where the path is and then uh to fix it in in the terminal. Yes, we do that and that works. The the funny thing is that it it wants to it says powershell here because these are powershell commands but we actually run them inside positron and not in powershell. Uh then we uh we copy paste we we copy paste here what we see now it's telling us to install but we have it working in in powershell. So so this is the powershell and it works. So we we tell it that it works in PowerShell but not in Positron. And why we want to install Gemini CLI or get it working first is again that then we can use Gemini CLI to troubleshoot things inside Positron if if any any things require troubleshooting. So let's let's do that. We run it and then we we do Gemini. Now it finds it because it's it's taking its time to load. And we can we can do that. And we'll tell it that it works. Let's close Positron and let's open Posetron again and make sure that Gemini works. So Gemini seems to work. So we we fix the path once and that that solved the problem. So if it didn't work then you can just uh just move things from just just type here and uh type commands and then give the error message to Gemini or whatever large language model you use. All right the the next thing that we do is that we we install an extension. So there is an extension and let's make that a bit bigger. So there is an extension for Gemini. So, we do Gemini CLI. And I'm not sure if it works on this computer, but we're going to install it. And uh this allows you to uh highlight code and then see Gemini CLI can can see the highlights. And uh [clears throat] we pick yes. Yeah. and it enable. So why probably it thinks that we are running on visual studio code and for that reason it doesn't install automatically. We can uh we can enable it. So we install it manually and then we enable it and now we can highlight now Geminina CLA knows which files we have which file we have open in post which is useful like if you are troubleshooting something then you want to make sure that you and the AI are both working on the same file. All right. So that is now set up and uh we have on our to-do list we have uh positron test Geminina CLI in positron terminal works then we install R and the way we just go here is that we go uh install and then there's R setup and we just go to uh the R website. We go R for Windows and uh this is what you want to do. Uh install R for the first time. Yes. So we install R and then we download R for Windows and we install it only for one user and then we try to uh to start an R session inside Posyron and then we'd use Gemini CLI to do a regression analysis demo for us. So it goes to local folder because we don't have admin rights. And we're going to install everything. We don't want to have a desktop shortcut because I never run R directly. I run it only through Positron. And now it's installed. And now we test it. So we start a new session and we pick R. It's installed already. And that means that we could have used it without installing. Uh but let's let's try to run the or let's just start start a new session and uh we go with R. This was Python installed. And then we can tell pos uh Gemini here that how would we how would we use R if it's installed on our own computer. But most people who who install this probably have admin rights. And let's uh let's ask how do we use a custom installation of R. And let's let's ask Gemini how we would use this installation assuming that I don't have R installed already on the computer. So, uh, how how can I use R located in instead of the systemwide version. So, this would be something that you you would do if you install R on your home folder on a computer that doesn't have R. And I really like now this is uh this makes troubleshooting a lot easier because uh Gemini can access our computer now and uh we allow it to to read what's in that folder and it finds that this is the things that we installed ourselves [snorts] and then it tells us how to do how to use it. So these uh AI tools uh are really nice because uh you can ask the AI tool tool itself how it should be used and this is like how to add it to path and I I'm telling I need to tell it that I want to use it in positron and we allow it to read that file. So that's our setting file and uh we just allow it once because we don't want it to be able to read arbitrary files on our computer. But again, this is something that if you can if R works like it does work like here, then uh there's you don't need to go through any of this stuff. But this is just like to show what I would do if I wanted to run the R from my home folder. Okay. Uh add this your settings. we can ask uh so we can ask the AI to do the change. So it's it's going to add the R session the R that we just installed into the settings. And and this is the reason why you want to have the the the AI installed first. And we allow it to edit. And now we restart Positron. And I can just delete this session and start session. Let's go that one. Okay. And uh then I can ask uh I can do terminal again and then Gemini. So I have Running from my user profile, but it was easy as I was already on this computer. So setting it up was kind of like just to show the worst case. So now we we test it. So our plan was that we uh we have inst uh test by doing a regression analysis. So we do uh program a demonstrate regression in R. visualize the model and it'll run. We'll do an R file that loads some demo data or generate some data, runs a regression and then visualizes the results. And it allows us to uh execute it. We always allow. And this can happen if again the path is not correct. and it will fix it'll it wrote the file now and we are we're okay. So we always allow writing uh our files into our folder and then it asks it tells us that we install R where we have multiple copies of R installed. uh we can just tell it to to run our script and uh let's let's tell that it can find it from from here so that it uses uh uses the installation directory. So we install everything here. Run it for me using our found at and it tries to find our script from that folder. So, we allow it to always find stuff. Yes. And now it finds the the R R our R script command that allows it to run R. Uh let's ask are there any workarounds? I know that it can run it. So that sometimes the AI kind of like is overly cautious. So uh or we can we can just go here also and just click on the on run and it runs the R file. But I want to use the AI to run it because um it's important that the AI can run things uh itself because then it can uh do uh a test do an R file and test it and then fix it. So it was able to run it and uh then I can I can ask it to to remember where R is located and it has this Gemini MD file that where it stores what it calls memories and uh then it remembers yeah then it remembers where the R executive is located and now I don't have to tell it anymore when I start R Of course, like if you have it installed as an administrator, then it's available systemwide and you would uh need to tell Gemini where to locate it. But this is like a worst case install. All right, so we were able to do it and here's the the regression plot. So that the regression line and uh that's the R code. So it generated some synthetic data and then um then run a regression analysis for us. All right. So the the next thing is Python and uh you can install this with install Python is to do um goes to store and Microsoft store and then uh we we search for Python and let's go 3.13 the most recent one and we get it. So we install it and this is almost certainly already installed on the system and we we have this now done this example. So we do clear so we we clear the context window so that we can um we can start start over. Okay. And uh then we could ask Gemini to locate Python. Um but it can it can probably do it itself. But if it doesn't if it can't find it, you can just tell that I install it through Microsoft Store and uh find the version that I installed through Microsoft Store. So we can do that as an exercise. So, so let's say that I I I I just installed Python using and then it locates where the the Python is. And again, this is like normally it's installed in a systemwide and and it just knows where it is. But I want to specifically uh uh find it just to be use that specific version. So we don't need these anymore. So we can we can delete and then do the final demo and it checks that it works. And this is where I just install it. So this is my my my local folder and um then uh we can ask that to remember it. So it writes it into memory again. So we don't need to tell it anymore. But this is again you don't normally need to do this because we would just install it systemwide and it just works. Yes. So we have Python installed now. And the final thing in our plan was to uh program a connection to local LM. And then we can we can tell it that um um I have LM studio running code a small script in Python that connects to LM Studio. sends a test prompt and streams the response. You would normally want to test that connections work and and this would be useful for data analys like if you have some survey responses or some some short text fragments to code or you can also all like do even a like a crawler to get for example news items uh from any news website and then use the local lm to classify them and uh yeah now we can tell uh that run it for me so I can see that it works. We can also just just open it here and then press this to run it. But what I wanted to do is uh to to test it because it uses the open AI library which I don't have installed on my computer and it would just produce an error if I try to run it and uh for this reason uh you want to have the model code test and then fix and then when it's fixed then you start running it yourself. So we always allow using pip which is a package installer for Python and we we allow and uh then it installs that for us and if we open LM studio we can see here that uh the model is is generating now. So uh this is the uh it's generating here. So this comes from our own LM studio and it's thinking because it's uh we didn't disable thinking. So this kind of setup would allow you to uh first of all code any data that you have on your computer uh for free and also avoid the the privacy problems of sending resource data to claude or uh anthropic or open AAI or Google which typically is not allowed by GDPR regulation. I should have asked it to do something uh shorter because this model is really slow on this computer. Okay, I I think this is enough. We can just just kill it and uh we can see that it stopped. All right, so that completes um u our setup. So we installed LM Studio a model. We tested settings. Then we installed Node.js. We installed Gemini CLI. Uh we tested by coding in game. Then we installed Positron. We test Gemini CLI in terminal. We installed R. We told Posetron to use our RR installation even if it was uh present on the computer. And then we installed Python and we told uh Posyron to our Gemini to use our Python instead of the one built in in the system. So the uh what makes this uh process uh challenging is there are like three things. One is that there's quite a lot of small software components to install. The second one is that your computer might have security policies and uh you need you need to adjust them to be able to run Geminina CLI. And the third one is that you need to set the path variable to to point to correct places so that the uh Gemini can be found in the terminal and Gemini can find the tools that you have installed. Uh fortunately the the process can be simplified quite a lot when you use PowerShell then uh you just you just have Gemini or some other LLM running and use the the maximum thinking and then you just uh tell what you see in terminal when you have an error and then uh the uh the AI will tell you how to fix it. So it's kind of like your own IT support. So this concludes the setup and hope uh you find it useful.