Видео

Hmm. I think they run sequentially. But don’t remember the codebase well enough. You could hit ai rate limits if you do parallel though. Didn’t see anything about other reporting libs. You can suggest on github. Or check out the logger file in their repo.

It’s very new

Glad you enjoyed it!

Oh nice. They fixed it? Got a link to that update?

@@elie2222 this video is the fix

@@elie2222 this video is the fix

@@elie2222 test

test

Glad it was helpful

Ya. That would be awesome. Would need to find the time to dig into it

I don’t understand the question

Hey. It’s possible srcbook works differently now. They’ve changed quite a lot since the video came out and now offer a hosted solution.

Yes

Thanks!

Thanks!

Glad you enjoyed it!

Not sure :( what’s the error?

@@elie2222 I dm you on discord

Glad it was helpful!

Well it depends how much you want to customize. The easiest is to just edit the prompt you’re sending to Bolt. But if you want full control you can edit the prompt file they use. Or you can go a step further and start editing the codebase. Eg. You might decide to add another type of bolt action to the prompt. But then the parser/ui/actions would also need to be able to handle that new action type.

Thanks!

Glad you enjoyed!

Thanks!

This one can do backend if that wasn’t clear! I demoed frontend only but it can write NodeJS code just as well.

😂 Next error actions gave me some more issues since this tutorial btw. One annoyance is that if you time out a function it doesn't give you an error, but just returns undefined. The other one is that libraries like next-safe-action can be very helpful. I haven't used it yet, but basically use the same ideas. And maybe I mentioned that in this video (forgetting now 😂)

Thank you for that comment. I was not sure if i was going crazy.

n8n !

Thanks for the suggestion! That's a good one! I did a video on their pitch deck half a year back, but have yet to do on their core repo, and there's good ideas to learn from in there.

Glad it was helpful!

Np

You can join the Discord if you like. Link in the video description

I’m sure they’ll get it fixed. They’re working on it actively. Also you can watch the video without having to use it. It’s about the ideas behind srcbook and how you could build something similar yourself.

Responded to you in Discord!

Got the same error man!!

I’ll get one of the founders to answer. It worked for me but they’ve made updates since

Comment from founder: 1) we’ve seen this when the API key isn’t valid. Try generating a new one and try again 2) you can join our Discord (discord.gg/shDEGBSe2d) and share logs I can troubleshoot further Or you could post an issue here: github.com/srcbookdev/srcbook

@@ajamit5006 Comment from founder: 1) we’ve seen this when the API key isn’t valid. Try generating a new one and try again 2) you can join our Discord (discord.gg/shDEGBSe2d) and share logs I can troubleshoot further Or you could post an issue here: github.com/srcbookdev/srcbook

@@elie2222 its broken currently for me as well. I added open ai and anthropic keys, deleted, re-added, and the api key passes its verification check. errno: -4058, code: 'ENOENT', syscall: 'spawn npm', path: 'npm', spawnargs: [ 'install', '--include=dev'

Thanks for that! I’ll fix that up!

Fixed!

This seems to do it from a quick search: codebeautify.org/json-to-jsonl-converter Any LLM will also do it well.

i was reading the code on your repo and also found really interesting your observability integrations with sentry and posthog... by any means you think of producing some video around this? i think it'd be great since most tutorials doesn't seem to get to this production-level concerns

Great. I have more thoughts on it at this point. Next-safe-action is a solid option for it. Midday GitHub has an example of how that’s done if you’re interested. Wraps actions with sentry, does zod parsing, and also checks auth.

Was considering it. The things to note there: Sentry and posthog are mostly automatic. The thing you need to do manually with sentry is the server action error collection. The rest should just work automatically. And then occasionally I’ll more explicitly capture an error that comes up. The other thing I do is ignore certain errors. There’s a few ways to do it and my code here is a bit messy. There was a point I tried to really cut down my sentry errors. But feels impossible to completely clear everything. For posthog it’s all automatic. I do put it into the error logging a little to try see how certain things impact users but I haven’t even looked at the ones I put in. But will be there for me if I ever decide to look.

Glad you liked it!

❤️

Good point. You could use onTouchStart, or when the a link moves into the viewport, you could preload the images. Or you could just not optimize on mobile. Pages would still load fine.

it still has prefetch when a next link comes on view

Screen studio. There’s an affiliate link at the bottom of the description!

Nice!

Glad you liked it!

Glad to hear that!

If you use next/link it defaults to prefetch. This is going a step further and loads images. If you don’t care about the extra performance then don’t do it. It’s a demo. No one telling you that you have to implement your site in the same way. But if you want extra speed this is a way to do it.

It’s not just bad for bandwidth, but for an e-commerce site, where conversion rates are usually higher than on a regular blog or freebie site, it’s acceptable. We’ve seen a big drop when our and client's sites slows down, our CTR once fell to 30%, when it’s usually around 80-90%, never dropping below 70%. So, load speed is crucial. Even after implementing these methods, not just for Next.js sites but for most of our clients, there’s been no noticeable increase in bandwidth costs. Try working as a digital marketer, and you’ll see how cheap bandwidth is compared to the impact on CTR and average session duration.

@@elie2222 no you are not getting my point. There is no argument against prefetching assets to speed up your site. But what they have done is overload everything to the max and claim speed. The original site is in production and it's that fast to interact. Nextjs can do well similarly although slightly heavier on Js. But this "demo" is a mockery of the attention to detail they've done in the original site. You might as well load the entire database in the browser and claim 0ms network latency.

🤯 nice

Thanks!

The benefit is clear in that it avoids the image flash. Whether that’s important is up to you. The serverless endpoints aren’t going to be cooked. They’ll handle this just fine. There’s some cost to it but that could be minimal dependent on the business value driven. If it leads to an extra 5% in sales or a few dollars in revenue then it’s worth paying the extra cents to make the extra server calls. There are studies by Amazon that show that extra speed leads to increased sales.

These days, most e-commerce sites handle their assets with a CDN, so it won’t put any strain on your server.

Glad you enjoyed!