How hard can it be to deploy Traefik anyway?
ฝัง
- เผยแพร่เมื่อ 1 เม.ย. 2023
- In this video, we explore how ChatGPT, an AI-powered language model, can be used to deploy the popular nginx and traefik containers for web server and load balancing functionalities.
Is ChatGPT up to the task? Let's find out!
===
🎙️ selfhosted.show podcast
📱 twitter @ironicbadger
🦣 mastondon techhub.social/@ironicbadger
📓 blog.ktz.me
💾 perfectmediaserver.com
🧑🏽💻 github.com/ironicbadger
Intro music: Joe Ressington
Great to see you starting a TH-cam channel now, Alex. Always enjoyed you (and still do) on the Selfhosted Show. Best of luck!
Loved the Video, Great Stuff!!
You really need to know what you are doing working with it, but it's scary close.
This is the kind of thing I've noticed with ChatGPT-generated code, 1) it will get off and debugging that WITH ChatGPT becomes a circular nightmare (you saw how it kept putting some of those incorrect things back in even when you said it didn't work or wanted it in a difference section) 2) It gets close, but those edge cases and last troubleshooting steps are now even HARDER if you don't have the context (imagine someone without any knowledge of how the traefik documentation works trying to debug the letsencrypt issues) and 3) It's mostly providing the basic boiler plate examples rather than pushing some of the correct default configs for most use cases. Either way, I don't see my job as being in jeopardy, especially when you can't even give appropriate hints to make it work properly.
Totally agree! I’ve been using traefik for a few years now so I knew where to look but that basic understanding from the operator isn’t there then like you say, it’s arguably worse than nothing at all.
@@ktzsystems we used it to generate some (arguably) boilerplate java code, and it was implementing the interface improperly, and just kept going between different packages flipped one way, then back, then one way, then back. I was hoping it would be able to deal with things like maven compiler error messages, etc. But it struggles when the message is "package doesn't exist", this even feeding it a skeleton of a working example.
Right? After I finished recording I spent another 30 mins trying to spoon feed it the answers but ultimately it was a futile exercise.
It surely won’t be that long before the models can actually make usable code though.
@@ktzsystems I mean, most of the code IS usable, it's just the stuff that's broken requires more and more experience and expertise to debug than potentially learning it from the beginning (at least in my opinion)
With references to THHGTTG and Rammstein, this is an instant sub.
Hope to see lots of content!
Don’t panic.
@@ktzsystems
Harmless?
Mostly.
Reminds me of voice recognition. It was the next great thing, in the nineties.... And it's still not ready for prime time except where the choices are limited.
From over here in the rocking chair, it might well be decades before real progress happens.
we are all doomed