Build With Gemini CLI | Google AI Tools | Generative AI

Key Takeaways

What's up everybody? Welcome back to another video within the generative man tool series. Where you guys have been? It's been a while. Long time no see. And before I start this video, I have a question for you. Are you someone who likes being in control or who likes to be controlled? That didn't sound right, did it? No, it didn't. Okay, let me rephrase that. As a developer, do you like being in control and use your terminal and use different extensions within your VS Code or you like to control things from the outside like use an external terminal just like I do? If you've been watching my videos, you know I'm a big Warp fan and I love using Warp, but for some time now, Gemini CLI has become like a consolidated part of my workflow. And as a developer, I love it. I've been wanting to make this video for a while because I've been giving a lot of sessions lately on, you know, uh, how you can use Gemini CLA effectively, how you can become a 3x developer and that's that's a very personal opinion and experience as well because I feel like my productivity has really improved using Gemini CLI and I wanted to share like top four to five things that you can do or you know enable you to use it or adv advocate you guys to use it so that you can be a 3x developer or maybe 2x or whatever. Uh it's a variable thing. All right, let's get our hands dirty. Uh let's try Gemini. And this is the first look of the CLI. Now, you'd be using Gemini 2.5 Pro, which is probably Google's best model up till now. And they're very generous with this model as well, especially within the CLI. Not just you're exposed to this model, but you also get like a thousand requests per day, which is a lot. I've never talked it. So, I think it's it's it's very generous. Also, uh it's multimodel. You can work with images, you can work with documents, you can work with your code bases, obviously. Uh no-brainer. Uh it has a very well structured MCP integration process, which I'll demonstrate in this video as well. And there's a lot of cool stuff you can do. For example, uh you can access your shell commands and you can also translate those shell commands through natural language. So you don't have to remember everything. For example, uh I have this folder over here uh which which has uh you know Mr. Bean's images and as you can see some of them are JPEGs. Let me see. Okay, we have JPEGs and we have uh GIF WEBP. So there are different you know diverse kind of types and if I had to do it manually and if I just wanted all of them to be converted into PNGs again a manual process I'll go to the websites online to find a suitable uh you know application which can convert it for me and hopefully doesn't take credit card or stuff. So, but now you can do it manually just by using, you know, Gemini CLI. So, let's try to do that. So, I'm currently in the bean folder. Uh, I'm calling this the bean folder. Nice. All right. So, convert all the images into PNG. Let's see. Now I love these little you know uh identifying image files. So they have like random very cool oneliners as well. So this is what I was talking about. So this is an automated shell command which would actually take up all types of images and translate them to PNG. So yeah, I'm going to allow it for once. It's going to go through a process. So basically RM it's removing all those files probably. So it's just going to keep the PGs. All images have been successfully converted. And there we go. We have all of them in PNGs. So I think you so for these redundant small task where you don't have to feel the burden of remembering different shell commands. I think uh a great interface for you know nontechnical users to actually uh you know do their daily work and even something like this like uh create well create a new folder and move all the files in there. Okay. What would you like to name the new folder? Bean omega releasing the cogs of the images. This is what I was talking about. So there you go. It's running the make directory command of being omega and would be transferring all those files into this folder. All right. So it's done. And there you go. Perfect. Right. Awesome. All right. So as a coder I review PRs or I would review code or I would want to conduct more like a self review now that I have these LLMs and to suggest me the best practices or to pinpoint you know not so best practices within my code right that would be one agenda as a developer. So I have a repository which says worst repository ever. So this is more like a fast API endpoints just for the sake of demonstration. Why I'm calling it the worst repository ever is because it has this endpoint where you have to uh let me zoom this up for a bit for you guys. All right. So you have this endpoint where you have to search for an item ID and item ID again is stored over here as data just plain old JSON with an ID name and description. Perfect. Now what it does it inefficiently searches the item for an array like if you want to read something or edit something. So it actually reads it and and that's it. it moves on to another loop. Now, although if you want to edit something or if you want to read something, you won't just store the item like this and go on to another loop to actually do that. So, it's kind of inefficient and I want this to be improved or at least I want to I want my CLI to report this and not to, you know, be okay with it. So, something like that. So, let's go ahead and uh check out some of the stuff cool stuff that we can do. All right. So, I'm going to type in again Gemini. One thing that I forgot to mention is like you should have uh node version 20 installed 20 or above uh for this to work. Just a little side note. So, right, I'm in my worst repository ever. And one of the cool things is is you can which is also like one of my favorite things while using Gemini CLI is you can search you can use the Google search. In this case I want to search all the best practices that you can do within fast API or within Python and I want it to review my code. So this is something like I would want to do. So I would type in search and you can do the same. search for the best practices. So, how how can it use search? Obviously, it exists as a tool. So, if you've been following my channel or you know about LLMs in general, then you know that you know you can use it as a tool or you know and MCP situation something like that. So search for the best practices in Python or fast API and review the code within their repository. [Music] Okay, let's see. So now I think the very first step Yeah. So the very first step would be the Google search. So search the web for Python best practices. And so this is like one of those features when I said at the start of my video like do you like being in control is it's plainly indicating towards the kind of access you can have just by sitting on the CLI. So yeah it's kind of cool. So it's done doing the Google search for Python best practices. Now it's searching for fast API best practices. And then once it has all the context it's it'll probably proceed to review our code. So yeah, one of the changes that it did. So within my code, it added HTTP exception in the imports and I used it here, but I never imported it. So it sort of got this. So this one was right. And then it made a few more changes. Uh not sure what this is. Maybe some indentation or stuff. All righty. So I think it pretty much nuked everything I did. It really was the worst repository ever. So now you can see a better reform code. Now you have a special root element with a proper welcome to fast API application. Should have done that. And returns all the items with optional filtering. And it has a particular service that it uses to forget item. Beforehand it wasn't. So let's go to services. Wow. So get item. Now you have a single for loop. Genius. Perfect. So you have get item with I think a uh what is this for? This is for Okay. Get all items and then you have a specific items. And then you can also create item. Perfect. It also wrote a couple of test cases for us. uh locked our requirement txt. Whoa, that's a lot of stuff. And it made a separate file for the data. So what's next? What what else can you do again? And now you've written the code, you've made the changes uh the changes that you could make to the best of your efforts and then you ask Gemini CLI or any other model to review your code and now you have all these changes in place. So now the very next step would be to create a PR out of it or you know work upon existing PRs or existing issues right. So how can you basically interact with GitHub again with MCP servers uh you know available in most of the uh you know visual studio agents and everywhere else. So they're pretty much available here in Gemini CLI as well. As you can see there's one MCB server connected. So I've already done that kind of an integration for this particular repo. Or to do that you need a Gemini folder. So this is more like a configuration folder and within this folder you need a settings.json file which is more like a list of MCP servers. You can add as many MCP servers as you want. So here I have the official integration for the GitHub server and you need a particular token for that which would be your GitHub access token. You can get it uh you know by going to github.com and within your settings you'll find a section where you can create you know fine grain or classic tokens. So you can create one over there. Uh you can specify timeouts. Uh so this is the default value. I didn't set it. Uh so this is more like a recommended value for all the uh MCP servers. So let's see it in action. So here as you can see using one MCP server and you can hit CtrlT uh to toggle the view. So once you do that, here is the list of all the things that you can do uh with your GitHub MCP integration. You can add comments, you can add issues just through natural language. All you have to do is state u a particular you know PR number or the issue number and even if you want to create a new issue or a workflow or anything. So uh obviously you can limit these options uh if you don't if you're working in an organization if and if you're using an organizational account. So you can limit it. Uh but I think it's pretty much it like if I if I wanted to specify list all the uh branches for this repo. So as you can see this is again an automated shell command uh that you can get translated through natural language. So uh we're on the main branch and these are the two remote branches. Perfect. Specified earlier in this video that Gemini CLI also happens to be multimodelled. So you can read files uh you can process those files within your CLI or you can even process images. So, one example of that would be let's say I have this cool little design right out of Figma and I want to generate a simple HTML page or even a React application out of it. Let's do that. And uh you have an image test dot is it png? It's png. Test.png. I want you to create this image. Here we go. Let's see. Fingers crossed. All right. So, we have everything generated. We have a HTML file, a sty CSS. So, this is what our design looks like over here. This is what our design looks like in reality within the HTML page. Let me zoom it out for you guys. Okay. So, let's compare. So, I got pretty much most of the things right. Right. Sort of. All right. So, these are more like four to five things that you can do on daily basis or at least these are a part of my developer workflow. Uh but you might have a question like what's so special about Gemini CLI? What are the big takeaways? Because most of the things that you've shown uh are already doable by using different agents. Okay. So, yes, MCP connections are all over the place. You'll find them everywhere. So, it's not a big takeaway. Uh there are certain models out there which provide you multimodel capabilities. We can debate over their accuracies, their benchmarks and different stuff. Uh but I've been more content uh with using Gemini 2.5 Pro. I believe I've been more satisfied with the results. So it's more like a mediocre big takeaway but it's more biased. Um one of the big takeaways is the Google search which comes in by default while you're using Gemini CLI. I have been using warp this terminal warp for a long time and I have been using uh different other agents out there but for some reason I find Google search more which you have access to uh it's more precise it's more straightforward it's it's I like it more. Yeah. So that's one big takeaway for me and obviously other big takeaway is the model itself. So Gemini 2.5 Pro is one of the best models out there with one of the best context windows out there with like 1 million tokens. So and Gemini CLI is still the most generous CLI out there I believe in terms of the kind of request model that it provides you. So yeah, these are two three points from my side. So let me know if you have any comments, if you have tried it, if there's certain things that I missed, do comment it down below in the comment section. And thank you so much for watching this