Transcription
Hey guys, Gemini CLI just received a ton of new updates this week. So, in today's video, I'm going to break down these seven most interesting updates that I'm sure you guys are going to love. So, we're going to cover everything from some new add-ons that they've just integrated to where the tool can work more intelligently with your editor all the way to some new GitHub integrations to where Gemini CLI is going to be able to review your PRs and actually work on GitHub issues on your behalf. So, we have a ton to cover so let's go ahead and dive on in.
So, the first update I want to show you guys is actually an add-on that allows Gemini CLI to integrate with your editor, such as Cursor or Visual Studio Code. And what makes this add-on so awesome is that now when you're running Gemini CLI, it can fully understand and interact with your different pages that you have open in your project. So, basically, your editor and your CLI are now one, which is awesome. So, what you can do, let me just show you the old way of how things used to work. If you had a question about this page, you would say like what is going on and and then you would call it out. So you always had to reference the page and it would go on and do its thing. So what's interesting now is if you open up a Gemini session, it automatically understands what's happening in your page. So you can say what is going on in this file. Now, what this will do is it automatically under the hood has added this entire page as context thanks to this Gemini CLI companion tool and it automatically knows, oh, I know what page you're on. I can see you're reading this page and it can answer questions about it. What's even cooler is then you can say more specifically is let's actually do this. You can actually highlight code now and you can say what is going on in this section. And what's cool is once again it's all hooked up and it's working like magic now. So you can see it's now going, "Oh, I can see you highlighted this header section." And now it's going to describe to us like, "Oh, you're looking at a section, a div element." And basically what it's saying is like, "Oh, that's the Gemini chat assistant code." So you can see if we scroll this down, yep, this is exactly what we were looking at. So your editor, your what you're highlighting and your actual project text and your CLI is all hooked up cuz that was one of the biggest things of like, man, I love cursor cuz I can just, you know, highlight code and then add it to my chat window and it's automatically here. But now I have all the same awesome features now inside a Gemini CLI. And in order to add this new feature, all you need to do is go into the marketplace button right here. Once you have it open, type in Gemini CLI companion. It might be super far down low because I think it has like 500 600 downloads right now cuz it's a brand new thing. So if you're, you know, having to scroll super far down, don't worry, that's what I had to do. And then you're going to then all you need to do next is you're going to type in IDE. And if this is your first time using it, there'll be an install button. So you'll type in IDE install and this will give you access to this. Then once you're done, you might have to restart the editor. Then you'll type in IDE status and it's going to go, "Yep, I'm connected to cursor. I can see what different files you have open. And so that's everything you need to do to use feature number one. Cool.
So, let's go ahead and hop on to the second new feature that I think you guys will love. All right. So, let's talk about the second coolest update about Gemini CLI. And this is going to be the new omit command that you can submit, which is going to analyze your entire project on your behalf and come up with a tailored Gemini markdown file. Now I'm going to go ahead and run it and I'm going to explain in the background why this is so cool. So, one of the biggest problems when working with AI agents is you have to solve the context problem, which means you only have so many tokens you can fit into every call you make. So, instead of every time you want to go off and make a call where the agent would have to reread through your entire codebase, what we're doing with the Gemini markdown file is just saying, hey, I'm going to do one iteration. I'm going to put as much information in here as I can so that you know in just a second what's going to happen is it's going to include all the core layouts of the project so that in future agent calls it's just going to look up this file and be like oh yeah I have a pretty good idea of like what's happening in the application what I can do and if I want to make changes where I should make them. So yeah, you can see this is a new overall pretty high level of exactly the project. So background what's happening it's a simple Nex.js JS application using Verscell AI SDK where I can actually chat with the Gemini model. So that's actually what's happening. So let us keep going. So we're going to allow it to write the file. Yeah. So it fully made the file. Things are looking good. And you can see it did it. It called out that this is an XJS application. I'm using TypeScript. I am doing an AI assisted power chat with this model. Now the only issue is is it got some of the details wrong. I am not using Gemini 1.5. I'm doing Gemini 2.5 Flash. So, this is kind of one of the like trade-offs of like using Gemini CLI of like it is insanely generous on its tiers. So, let's actually just I want to show this to you really fast because I think this is important to mention. Like with Gemini CLI, you get an insane amount of value on their free tier up to a,000 AI requests per day and it resets. So, in a month, you get 30,000 free calls to Gemini 2.5 Pro, which is crazy. You know, there's no other tool. Cloud Code doesn't do this. Cursor does allow on their free tier a few calls per month, but you know what is that 115eenth of the value on the free tier. So yeah, Gemini CLI, you know, it is not as like oomph like it is. It will have, you know, your results will vary, but you're getting an insane free tier. So there is a trade-off using the tool right now, but I'm sure as soon as the new Gemini 3 comes out, this thing is just going to be an absolute beast of a tool. So yeah, it's already pretty good, but I think when the new models come out, it's going to be even better. Final thing, the Gemini CLI output, like you said, it did okay. But what I like to do, and I recommend you doing, too, is like I would recommend this as a good starting point. And I would just over and over is say like, hey, that was a really good start, but I would like for you to go deeper and actually outline what's happening on each page and how everything connects. And then this is just going to like once again keep beefing up this Gemini document over and over and over again until it does a really good job talking about your project. So, think about this as a starting point, add to it, and then you're going to end up doing a much better job or agents will do a much better job once this file is really beefed up. So, yeah, that's the second coolest feature I've noticed inside of the new update.
So, let's hop over to the next ones and dive in. Oh, and before I forget, if you're interested in turning your app ideas into real world production AI application in days instead of months, you're definitely going to want to check out the new project I've created for you guys called Shipkit. Long story short, Shipkit includes a bunch of real world productionready templates for the most common types of AI applications that I've been paid to build. So, I've tried to put everything I know about AI applications into this project. So, you get a bunch of pre-built templates for chat applications, rag applications. This one's the most powerful. Also, multi- aent applications. In addition to getting pre-built templates, it comes with a course teaching you guys how to go from an idea and walks you through the entire process from end to end. It also includes AI prompts along the way. So agents are doing the entire work for you all along the way. And it comes with three additional real world AI projects where I show you guys how to go from a template, use the course, and create a real world production application. So if you have any questions, I definitely recommend checking out the project down in the description below. But enough of that, back to the video.
So, the third update I've noticed inside of Gemini CLI is they now allow you to manually manage your memories that are applied to all agents in the future. So if you're familiar with cursor rules, it kind of feels like that where you can make a new rule where you're saying like, "Hey, if you're working on a Python file, make sure to like format it this way or add comments this way." So let me show you how you can actually manage this inside of Gemini CLI. So first things first, you can actually type in memory. And what you can do is then let me refresh it just cuz I've made a few quick changes. So you can type in memory. Then you can type in memory show. And what this will do is list out all of the memories that you have in your project so far. Right now, you'll notice it says there's one file which contains all of our memories. And what do you know? It's all coming from our Gemini markdown file. So, yeah. So, you got to see section 2 now is building up in update 3. So, now what I want to show you is how you can actually add in more instructions to make sure that your agents behave more like they should when using Gemini CLI. So, what you can do is type in memory once again. Then you can add a memory and then you just manually start typing it out like regular text. So, we're just going to call this important important rule. And then we're going to say make sure all text inside of our websites that get shown in the UI is all capitalized. So, if a button used to say, and then we're going to show it the old way, it now needs to say hello. So, we're just going to be super explicit so that there's no excuse if it messes up. So now what we can do is we're just going to do one quick memory refresh just to make sure that it now has the new rule. And then now what we can do is do memory show. And what you can see is now we're reading from two files. So pretty cool. And you can tell there's actually two sources now. So source one, this is the new rule we just created. And if you notice, it's actually saving this inside of like if you were to drill deeper into this, this is in my root folder of my user account. So, it's creating a Gemini folder and inside of that it's creating some instructions. Caveat, you need to know this. When you create a rule inside of Gemini, it gets applied to all future projects. So, if I go work on a different project, this same rule is going to apply. So, let's actually test this out. So, we're going to say, please follow my rules and update the current page. So, what this should do is like we just added a memory, so we're going to now actually see what it does. So, you can see, yep, I'm reading the page file. It's cool. Once again, it knows what page I'm on. You can actually press CtrlG to see it in real time and follow this command to see all the files that we have to accept first. So, we're going to allow. But now, you can press Ctrl G in here and you can see in real time what files are being edited. So, G. Yep, you can see active files. We have right now this is, you know, thankfully cuz it's connected to cursor. You can see I'm looking at these pages. You can just press Crl G once again to get out of it. So yeah, you can see it now updated all the text on the page to say Gemini chat assistant. I would have expected it to do more. So hopefully it's going to keep digging into the rest of the files and hopefully find out all the other changes that need to happen. Hopefully the control G a few times didn't mess with it, but you know, once again, it shouldn't break the CLI tool. So we'll see. We're going to give it a few more seconds to see if it's actually able to update the rest of the different text on the page. So, we'll come back in a second once it's done. Awesome. So, as you can see, it was able to go through and update, making sure everything was capital. So, this was a super straightforward task and it was able to knock all of it out. So, yeah, it's all looking good. So, what I want to show you next is just one final thing going back to our different memories is I really found much better results when updating this format to say like important rule or something just like that because anytime I was just saying like, hey, you know, please make sure this happens, it wouldn't do it. But the second I started to add phrases like important rule or like you must do this, it did much better. So, just, you know, keep common best practices for prompt engineering for much better results. So, with that out of the way, let's go ahead and hop over to update 4.
All right, so the fourth major feature that was just added to Gemini CLI is that you can now easily view all of your stats. I think some of this might have been available in the past, but now it's a lot more detailed than it used to be. So, specifically, you can look at models and tool calls so you can see exactly how they've been used in each session. So, after we have just ran our chat rules, we can now come in here and actually just look for models. So you can see, okay, I can see you're using this model. I can see how many tokens you're used, what was cached, and you can see a pretty detailed view. You can also go deeper into it and then actually look at tools. So you can now see like what tools were used, which ones weren't used. So in this case, it was just updating this, you know, page.tsx file. So everything in here was pretty much just a replace command over and over again. Speaking of tools, you can always see what tools are available in Gemini CLI. So if you just type in tools you can see all the major tools that are happening in here. So you know we can have everything from read to files to read folders to save to search text and write file. So in our case it was like a write file with a replace is basically what happened. So all around super new feature that they added in. Just want to show you two. If you head out though one thing that I do wish they would add. So, if you come back into Gemini is if right now if I run stats, you know, it's for a brand new session. So, like I think it would be cool to like be able to run this and say like, hey, for this session and globally cuz I would just love to check in to see like have I spent am I near my budget? Am I not out of my budget? You know, for my daily quota. I think that would be something really cool to add. So, half of this is a review, half of this is a wish list. Yeah, that's the fourth new edition. much more detailed dive into exactly what is happening with your model so you can keep up to date with exactly what's going on.
So, let's get that out of the way and hop over to feature number five. So, the fifth and really powerful update you can add inside a Gemini CLI is that it now allows you to work with multiple directories and workspaces all in one session. So let me show you why this is so important. Let's imagine that we're working on this chat application internet money. Well, the only problem is we only have code for this in another project that we've already worked on. So, we don't want to manually copy everything over from this project over to this one because we've already built it once. Let AI do all the hard work of exploring that other codebase and bringing it into ours. So, here's exactly how you can do that. If you type in directory, what you can do is first see what you know files you have access to in directories. We only have our current application. Well, now what we can do is add in an additional directory. So when we do this you can now see great we have two different directories in our current workspace. So now we can do cool things like this say like please bring over the stripe.ts file to this project. Now what this is going to do is first off it's going to analyze what's going on. It's going to find our chat SAS application. It's going to find that Stripe file that we told it to and we're going to say yeah please go ahead actually bring this file over here. You can now see cool we have the exact same one over here. And now you can see we can keep going down the path as well. So you can see it's like digging through and figuring out okay well in order to use that file I can see it was importing something from an environment variable file. So what I'm now going to do is look through that environment variable file created over here and start bringing that one in as well. So you can see it's like doing a really good job of understanding our other project and then bringing in those changes to our current project. So we're not going to like I don't want to install and go through all that process. But yeah, this is one of the greatest parts about AI is it really comes down to a a model and context problem. These models are getting insanely smart. So that's really not a problem as much as it used to be. But context, we need to have the right information in the right spots. And one of the best ways to do that is just to have pre-built or real world applications. And this allows you to mix and match directories, grab some code from that project, grab some code from that project, merge it all together to make sure you guys can build awesome projects in no time. So yeah, I thought this was a really cool new feature. Really like that they added this.
Okay, so now let's dive into feature number six. And this one is I'm excited they've added it, but I wish they went a little bit further. So let me show it to you. So right now you can actually save and manage your conversation history. So if you do for/ chat, you can actually see all of your saved conversations. The problem is it doesn't look like Gemini automatically saves your conversations. So that means every time you're interacting with Gemini CLI and you're done, it's just gone and it's gone for good. So what you have to do is you can manually start to save your conversations. So, you can actually save this conversation and we'll just call this, you know, a Stripe and we'll call it Stripe import because I was in the middle of working on that. And now you can see, oop, sorry, we're going to call this chat save. Now, cool. We now have a new different conversation. Sorry, my this is killing me. So, yeah, you can now see save conversations. So, this is awesome because now if we actually just completely quit this command. So, we can actually just get out of here. And now if we ever hop back into Gemini's CLI, we can now do once again chat list and then fantastic, we can resume our conversation. So resume and then we're just going to call it strike. Fantastic. We're back to where we were and we can continue to talk to our models. So this feels very similar to, you know, clawed code. They have a resume functionality to where you can just see all your previous conversations. And when it comes to cursor because you have that side panel and you can easily view all of your previous conversations. Only recommendation I wish they would add is I don't know why we don't just automatically save all these conversations locally. It's all getting saved on our local computer. It would just be awesome for every new session to automatically get named and saved. So, it is cool that we now have this functionality and I love how much progress they're making because this is one of the only open- source projects where we can get to see what's going on and contribute. So, I think that is awesome. But I would just love to uh to see that come in to where hey, why not just go ahead and autosave it so I could automatically see all possible conversations that I've had instead of only the ones I manually saved.
Now, that is update number six. So, let's dive into the final and greatest one, which is now the ability for you to actually start to use Gemini CLI inside of GitHub to do some magic. So, let's hop over to the final update.
All right, guys. So, let's dive into the seventh feature that was just added to Gemini CLI. And I'm going to go ahead and give a heads up. It is working but not working at the same time. And you're going to see what I mean. But don't worry, they are on a weekly cadence dropping brand new fixes. So heck, by the time you're seeing this, it's probably all going to be working. But I just want to show it because it's such a cool concept and I think you'll absolutely love it in your developer life cycle workflow. So let's dive in. So the seventh feature is now that Gemini CLI directly integrates with GitHub through the Gemini CLI GitHub actions. And basically what this automatically does for you is anytime you're making a change to your codebase and push it to GitHub, it's automatically going to do a PR review. It's going to triage your issues. It's going to, you know, answer any questions if you tag Gemini CLI. And you can do all sorts of more advanced workflows based on like what you want to happen inside of your codebase. So getting this set up is so straightforward. So let's actually show you how to do that. So, what we're going to do is go over to Gemini CLI. And what you can do is you can type in setup GitHub. Once you do this, what it's going to do is it's going to open up a bunch of different windows. So, we're just going to say allow always. And this is going to open up multiple windows. So, let's explain what you need to do in each one of these to get things working. So, first things first, it's going to take you to GitHub. And obviously, you need to have this project pushed to GitHub before you can take any action. So, I have mine saved right here. And then what you'll notice is down here at the bottom there's repository secrets. You want to add your Gemini API key. You can get that by going to AI Studio and yes, specifically a studio.google.com and then when you go to your dashboard, you're just going to click create an API key and that's exactly what you're going to go paste over there. Okay, cool. So that's the first thing. Then what you want to do next is once that's set up under the hood, it's going to then walk you through how to save that Gemini API secret. Like don't worry, there's instructions. It's going to show you exactly how to like take that key, exactly what page to put it in on GitHub. And then, you know, we're pretty much done at that point. So, what you'll notice after running this command is it's going to update your project. So in this case, it's updated our GitHub workflows to include multiple different workflows that are automatically going to get triggered under different conditions. So if you're new to GitHub actions and workflows, basically what happens is like there's a bunch of triggers. So like, hey, anytime there is an issue that was just opened or reopened, you know, or there was a, you know, an issue comment, basically we're just doing some checks. The second we understand we once we've met these criterias then we're going to start performing some different jobs. So this is exactly like each job that gets pulled up here there's specific instructions. Don't worry this is pretty complex. You don't need to know how all this actually works. But I do want to show you in action because that's the part that's really cool. So what we're going to do is I'm going to make a new branch. Get checkout.b. And I'm on test I'm already on a test branch where I've added in some new functionality but I'm going to make a new one. I'm going to push this branch to GitHub. Then once we hop back over to GitHub land, let's see. So here, right here. Yeah. So you can see we now can complete and create a new branch. So I'm going to say added is loading to. So what you'll notice now is I can create this pull request. And then this is where the magic of AI meets your development life cycle. So here's what's so cool. If you hop over to actions right now, you can see it is triggering a bunch of new workflows based on what we just did. So, we added in a new pull request and you can see it's saying, "All right, Gemini pull request review is happening in action." If you hop back over here, you can see there's a Gemini pull request review. So, it's all working together. And you can actually open this up and see exactly where it's at in its life cycle. So, you can see it's like checking out the code. It's actually pulling down our code from there. It's going to then get the pull request details. But for some reason, there's a few commands in here. Like once it's done the review and it's actually like figured out what's going on, it just for whatever reason right now, like I said, this is the bug. It's not actually pasting the PR review to the pull request. So, you can see I've already ran this a little while ago on a different branch. And you can see it didn't add anything. The only thing we got was something from, you know, the only thing we got so far was something from GitHub actions, but we don't actually have the things from Gemini CLI tool actually working. So, in theory, you should see an extra comment down here that says like, "Hey, here's exactly what I see is going wrong." So, I think this will be fixed hopefully by next week. The other thing that I think is super cool is you are able to actually like write comments to Gemini CLI and it will go off and do a code review. So what you can do is look at file changes and then you can say like we'll find some code real fast and we'll say is this the best way to test is loading. But more importantly you always have to start it off with Gemini CLI. So once you do that then we'll submit the review. Submit review. And then what's cool once again is if we hop back to actions you can now see that in just a second. Yeah. You'll see it should say this is it right here. We have now asked a comment to this model based on our pull request and it's added to the queue. So you can now see as soon as this other one's done, it would start to address our comments. For some reason, it's just getting stuck and hung up right now and not actually performing the pull request. But as soon as it's fixed, this is going to be an awesome way to have AI automatically review your workflow and just make sure you have high quality code. And it's all done in Gemini CLI. So that's all around awesome. and you're using I think at this point you're going to be using your own tokens to get this part working. But yeah, all around pretty cool and I was hoping you guys I was hoping it's going to work so I apologize it doesn't. I will post under the video as soon as I get confirmation from you guys or someone at Google that this is working and I'll let you guys know.
So, now that you've seen the seven new features added to Gemini CLI, what I want to do is just kind of share my current thoughts on the state of Gemini CLI, who should be using it, and then I want to go into some like wishlist areas to where I can talk about like some of the main issues I've seen and what I would love to see done differently. Okay, so who is this for current state? I think Gemini CLI is an awesome beginner CLI tool where you get to try out some of the new awesome like agentic workflows in a CLI and you're going to get an insane free tier limit. So, you're not It doesn't cost you anything to try out this tool. That 1,000 requests per day is insane. And I definitely recommend people to at least try it once just so they can like, you know, get a feel for it. But is this ready? Am I going to use it every day right now? Short answer for me, my current workflows, no. And let me explain why because I think it's like a cool way to like recommend like features that would like I would use this way more if it did these few things. Okay, let's dive in. First things first, I would love if there was some sort of limit down here where I could see where I'm at when it comes to daily quotas and rate limits. Right now, as a developer, when I'm using Gemini CLI, I'll make some progress. After I make some progress, I'll do another command and another command, and eventually I'm like, "Okay, cool. I've come up with a plan. I really want you to go off and do a lot of work." then this long task. Well, what happens to me multiple multiple times is I will run into an issue and I will hit a rate limit specifically and after 6 minutes of the code running the agent running it just dies and it says hey you hit a rate limit you need to go do this. So instead of me retroactively having to like dang it just broke. What would be awesome is if it just said right here just a number of like, hey, you're 30% towards hitting your rate limit for the day. Here's exactly what you can do to not hit this rate limit. Or if you are using an API token, I would expect it to, you know, not hit a rate limit, but it does quite often. So, I would love if it would like proactively say like, hey, once again, you are 60% there towards hitting your rate limit for the day. and just giving more concrete numbers because I love how they tell you the exact context length that they have and where you're at. So you you know on this side when you're almost out of room, but when it comes to rate limiting, you have no idea until you just hit the wall and then you're like, "Dang, I have to literally start over cuz this chest this is I can't do anything now." So that's one.
Two, there's numerous times where it just takes a long time. So, as I'm asking it to go do something, normally I would expect like a task like updating the different, you know, text in the website to be all capitalized. Sometimes if I was not insanely hyper clear on exactly like you need to update this raw text in the area, you know, it would take up to 5 minutes, sometimes 6 minutes. But whenever I nailed down like my instruction giving, sometimes it would take 20 seconds. So, like what's interesting though is in other tools I don't really have to be as prescriptive. It's I'm just like, "Hey, capitalize all the text." And it just goes off and it's done in 20 seconds. So, I'm curious what's happening on the background planning and to do creation and analysis and maybe I'm doing something wrong, too. But yeah, usually when I'm just like, "Hey, go make all the text capitalize," it really will sit there for six minutes going back and forth. Once again, don't know if this is a rate limiting issue to where it's like something's happening and I'm getting throttled. I have no idea. But these are just like things I'm seeing that I would love to see fixed so I could use Gemini CLI way more in my workflow. A few other things that I think would be cool are making certain parts of the application more agentic. For example, as you saw today, let's actually just hop it back open. But you can see, let's just open up a new chat. So you can see with Gemini, if I Let's actually close out of this. if I actually want to do something. So in our case when it came to like chat messages, well like it would be awesome if these were automatically saved and automatically named. So just anywhere where things could be automated to make it just easier to use. That would be awesome. So compress this would be great if it just auto said like hey now looks like a good time to compress. You're close to your window and you just wrapped up a task. Do you want to compress? Because like we are using AI applications after all. So, like why not use AI to help do all the manual tasks? Ideally, I'm just chatting with this tool and everything else is recommended to me through AI. Like, oh, check out this perform this action. Hey, you're almost out of context. Let's compress and open up a new chat. So, all things like that would be awesome. And then finally, another one just that would be great is I would love to see more stat messages so I can see, you know, hey, I've had 10 conversations today. Where am I at? How much have I spent with you guys? like that would be a really cool way for me to actually like have a better idea of like where things are at and going. So yeah, but yeah, they are just quick recap though because like I do want to say like absolutely love the direction they're going. One of the coolest open source tools in the CLI space. Absolutely love what we're doing, but these are a few of my wish list things that I would love to see so that I could use Gemini CLI way more often and get to, you know, actually like contribute more to my codebase. So yeah, those are my wish list items.
So, that's a wrap for this video, guys. I hope you enjoyed seeing all the latest upgrades to Gemini CLI. And if you have any questions, I'd love to hear your feedback down in the comments below. And if you're actually using this tool, I'd love to hear it. Or if you're using Claude Code or Cursor, I'm all ears. I'd love to see what you guys are actually using. Also, if you are interested in taking your ideas to a real world AI production application, you're definitely going to want to check out the latest project I'm working on called Shipkit.ai. And long story short, it is a project and a course all combined to help you take your ideas to a real world application in days instead of weeks. And we do that through a combination of pre-built templates, a bunch of courses, modules, and you get access to all of my templates and prompts and everything I use as a developer to help me crank out real world code in no time. So, you definitely want to check out down link description below so you can hop over and check out the full video so you can see it all in action. But yeah, you guys are awesome and I cannot wait to see you guys into the next video. And I definitely recommend checking out whichever video is popping up on the screen now. Thanks guys. See you in the next one. See you. [Music]