Local and Fast AI comes to your developer workflow - Full Line Code Completion
HTML-код
- Опубликовано: 5 июн 2024
- Dive into the blog post to learn more - jb.gg/flcc-blog
JetBrains IDEs introduce a new feature that enables developers to autocomplete entire lines of code using locally-run deep learning models. This functionality directly addresses the fundamental constraints of cloud-based alternatives, and ensures data remains on your machine without any need for external server communication. Thus, everyone can complete the code in trains, airplanes, and even ships without the internet, and extra subscriptions.
Full line code completion is available for Java, Kotlin, Python, JavaScript, TypeScript, CSS, PHP, Go, and Ruby, within the corresponding JetBrains IDEs out of the box: IntelliJ IDEA Ultimate, PyCharm Professional, WebStorm, PhpStorm, GoLand, and RubyMine. Expect support for more languages in upcoming releases.
Full line code completion vs. AI Assistant
There are two ways you can benefit from AI functionality in JetBrains IDEs - full line code completion and JetBrains AI Assistant. We appreciate that this might be confusing, so let’s take a closer look at what they have in common and how they differ.
Both full line code completion and JetBrains AI Assistant aim to help you work faster. They both also go beyond the standard completion that has been available in JetBrains IDEs for some time already. However, JetBrains AI Assistant is a separate product in its own right. Its feature set isn’t limited to code completion - it offers you a lot of other things, too, including context-aware smart chat and the ability to generate tests or write documentation.
#JetBrains #fullinecodecompletion #programming #ai 
Наука
This is why I'm subscribing to Jetbrains and have done it for several years now.
Always new useful tools and top notch support
Wow. 4k video. I was thinking like "what the heck? why is the guy so sharp?" 😁 Nice!
Paul is always sharp, even on the lowest bitrates. 😅
@@khalid_abuhakmeh Paul is the dull blade that you use for simple tasks. Paul could barely cut butter!
Thanks, in fact, this was my very very first video with the new camera. I looked at the file size and thought I had left it recording overnight. 😉
@@paulweveritt Incredible quality! I was wondering if it was the camera or perfect lighting. Could you possibly share the camera you use?
@@_WEAPON_ Just before this, JetBrains sent me a legit camera, Sony. And you're right, the lighting makes a big deal.
I've been using this in the EAP version.
working perfectly fine so far
I used it this weekend when doing a Python HTMLParser and once again was freaked out how well it did. Of course, that's a nearly ideal fit.
Looking forward for C# !
Me too! I ❤ JetBrains Rider and this will make it even better. Cheers.
Me too, definitely in my waiting list
@@fakhrulhilal out of curiousity, what kind of .NET apps are you currently working on?
@@khalid_abuhakmeh backend API, and some front end using angular
I am surprised that C# isn't supported yet when it's one of the most important languages in JetBrains' developer tools.
So this works without Jetbrains AI Assistant subscription?
Yes
but you still need their proffesional subscripttion 👎🏿
Great video! Excited to try this out 🎉. Btw, what is the Editor theme / color scheme that you used in the IDE?
Always enjoy this guys videos. Informative yet hilarious :-D
Thanks for the kind words. Why did it take 5 days for me to come look? I've been terrified of the comments! Never sure if I get the infotainment balance right.
You guys are always the best!
That’s how I can justify my need of a GPU and 32gb ram for coding and IDE.
I'm currently porting .net4 app to .net8. I need 3 IDEs open, 16 Gb isn't cutting it.
I'm getting used to this. To be honest it did my head in to start with but I've started to understand how to use it and I really do like this a lot. I'm doing a user interface at the moment and the autocode is very quick for repetative tasks and is pretty accurate overall
Dart (/Flutter) would be excellent too!
that is awesome 🤩
there is no link for press release in the description at the moment
Hi, thanks for the heads-up! We've added a link to the related blog post instead, feel free to check it out.
hey what about Swift and IOS ? please come up also with Swift Programming also specially for app code. Please Please
Its great in phpstorm, i love it
This voice is how i talk to my dog.
Will this come to idea community?
Please support VB as well!
Even better would be to allow us to use our own locally-run model APIs with a chat interface and code completion.
what font and theme do you use
I use the bundled dark and light themes, bundled font, with some zooming to increase size.
It lags on my laptop, needed to downgrade to webstorm v2023.3.6, it seems I need to buy a newer machine for this local AI tools
You can just disable full line code completion in settings, no need to downgrade. If you would submit an issue with details of the performance problems you encountered - it would be very helpful for the team to debug and improve this. Full line completion generation is allocated by limited resources and should not cause any changes in IDE performance
This is amazing 👏
How does compare to jetbrains ai assistent?
Please see this blog post blog.jetbrains.com/blog/2024/04/04/full-line-code-completion-in-jetbrains-ides-all-you-need-to-know/#full-line-code-completion-vs.-ai-assistant and let us know if it answers your question!
I'm waiting for Rider to get this update!
No plan for F# support?
That is a lot of effort just for both people that use F# :-)
Full Line Code Completion did not work for me. Waiting about 1 minutes, no full line code completion cames up.
What language and IDE were you using?
Toolbox -> Update All.
I genuinely thought Macaulay Culkin was in the thumbnail 😄
If Macaulay Culkin was portrayed by an elderly hobbit. 😉
Are there any Hardware requirements?
The basic requirements look like this: your processor must be 64-bit. If the processor architecture is x86, it is required that it supports the AVX2 instruction set.
This is the biggest reason I hated CoPilot and liked AI assistant. When full line completions get in the way with bad suggestions and remove the better, more accurate shorter suggestions that already exist, you fall into that pattern of typing a little, waiting a second, fixing bad suggestions instead of writing it correctly to begin with. It's a slow workflow that leads to worse code.
Hi, thanks for your feedback! Could you please elaborate on what you mean by removing shorter suggestions?
Also jetbrains academy is so expensive!
I deactivated this because i could not easily just type less and not get the completion..
This is my money's worth.
Intellisense has been keeping me in VS...
🛋️ it's really comfy over here
copilot or this?
This is local and included in your JetBrains subscription, no extra charge. Does that help you make a decision?
@@khalid_abuhakmeh i already subs to copilot so still need a comparison between them
I have been happily using WS for half a decade and like it but this is just atrociously bad. It gives worse results than llama3 8b, even than gemma. Why on earth would anyone use this as long as Codeium is free?
Thanks for your feedback. Codeium is indeed a nice free AI Code Completion tool. But it's worth noting that Codeium is cloud-based which means that your code is sent to the cloud where code suggestions are generated. Full Line Code Completion works 100% locally on your device (your code is never sent to the Internet). Besides being 100% local, Full Line Code Completion is the only AI Code Completion tool that checks code correctness before showing the suggestions. To reduce computational resource consumption Full Line Code Completion uses just a 100-million parameter model (in 2024.1 version) and it's kind of unfair to compare a 100-million parameter model to 8-billion Llama3 or 2-billion or 7-billion Gemma. While it is possible to run an 8-billion parameter model on a modern laptop, generation will consume a lot of CPU and RAM which is non-practical for most of the use-cases. If you want to try local code completion with really big models you can try the TabbyML plugin.
You have to subscribe for something that works offline? That's illogical 👎🏿
You don't. If you're using one of the mentioned IDEs - Full Line Completion works out of the box. The mentioned IDEs themselves are not free, but Full Line Completion does not increase their price
Another one of those ridiculously expensive subscriptions! Sorry I will pass on this one
Please note that Full Line Code Completion is available out of the box in the mentioned IDEs (no additional subscription needed)
Those "Kevin home alone" thumbnails have to stop...
What if the next video shows you how to thwart a home invasion with paint cans and toy cars. Is it ok then?
I guess in such case I would allow it :)@@khalid_abuhakmeh
@@adamolszewski4054 Please don't let @khalid_abuhakmeh photoshop my face onto Kevin for a thumbnail. Unless I get royalties.
@@adamolszewski4054 be sure to like an subscribe to get that video and other videos we publish. (no garauntees I can get Paul to do the Home Alone video 😅)
@@khalid_abuhakmeh Khalid and I both work remote (PA and VA.) We're alone. A lot. Thankfully each house has dogs. Dogs are the best.
Where is c#? Why not first?
C# is being worked on currently and shouldn't be too long. We appreciate your patience.
@@khalid_abuhakmeh We'll see if it's not long ;-)
@@khalid_abuhakmehAbout how long? Should I wait or try EAP? As I understood, it's available in EAP right now, right?
Full Line Code Completion works on language-specific models. So while the plugin can be installed in Rider, the model for C# is not ready yet. I'll double check and report back as to the status.
IntelliJ Idea Ultimate is the top selling IDE for Jetbrains, so business first. I am happy to be a Java developer. All new features, first and fast.😊
Jetbrains is AWESOME !!!!😅