Threads
Loading user profile...
I build multi tenant apps with Clerk. Solves a lot of problems out of the box.
Sure. What country are you in?
Polar on top off Stripe for VAT legality. Challenge is syncing your client with whatever is thrown back by the API. You must consider all payload possibilities. I did an implementation with monthly subscription and top-tokens with extensive testing and I'm still not 100% confident to push it to production. One slip up with people's money and your toast.
How are you handling Stripe payload?
First time for everything
console.log for debugging but that's about it. Then delete for production.
Realtime. Real expensive.
No thanks.
Build it from scratch.
[Realtime document ](https://gist.github.com/AlexDjangoX/0a81aba0bcc669251c6a9d2b9e2ce5c5) This is a document from my working version of Supabase realtime integrated with Clerk.
Send me a DM. I'll look for it in past commit history.
With great difficulty
git checkout -b database-upgrade 6 hrs. It's all boilerplate. Nothing new or novel. You don't have to write any syntax. Those days are over. If you have a full suit of tests prior to migration. Excellent. Migrate. run tests. Have LLM analyse prior to migration and run tests. Have it do the migration. Update tests. Your migrating from a working state? The job is ✔ done already. It becomes LLM's context. Use them tools.
Estimated time - 6 hours.
Real time is very resource hungry. Have you experienced that?
Learn the skill of using LLM's - you still need engineering skills, but LLM's allow you to iterate faster. Its a skill that devs must master to keep up. You cannot trust LLM's to think for you but you certainly can use them as a very effective tool. For a data base set up ask about one-to-one and one-to-many relationships of models. Quizz the LLM about the relationships. You know in natural language what you want. Prompt the LLM. Get a final set up. Take that set up and ask.anotger LLM to explain all the relationships and see if it lines up with your mental model. It's a lot better than asking random people here. LLM's are the feature of Engineering. Your not being lazy. You are upskilling.
Dude. LLM's are really good at this kind of thing. Even the free tier of ChatGPT.
Look into Stream
Convex has out of the box features which are very difficult to implement in Supabase. Especially if you want live updates.
I always use Supabase but Convex seems to be a better option especially if you want live updates.
No
No.
I used it on free tier.
Too resource hungry.
Realtime 😪
On a monolithic, multi-tenant application I upgraded: NextJS 14 to NextJS 16. Clerk 5 to Clerk 6. Tailwind 3 to Tailwind 4. Prisma 6 to Prisma 7. Also bumped up Lexical editor to latest. Why not? Those ridiculously long NextJS 14 compile times were doing my head in. NextJS 16 compile time are near instant....until you start using reactCompiler, compile times went up a smidgeon Upgrade, surprisingly, was uneventful. No blocks. Smooth as.
On a monolithic, multi-tenant application I upgraded: NextJS 14 to NextJS 16. Clerk 5 to Clerk 6. Tailwind 3 to Tailwind 4. Prisma 6 to Prisma 7. Also bumped up Lexical editor to latest. Why not? Those ridiculously long NextJS 14 compile times were doing my head in. NextJS 16 compile time are near instant....until you start using reactCompiler, compile times went up a smidgeon Upgrade, surprisingly, was uneventful. No blocks. Smooth as.
Ouch.
I farmed this out to Clerk, creating a multi-tenat Tutors platform. Everything is handled in middleware and tenant warpper in server actions. Tech stack - NextJS, Supabase, Zuplo API Gateway and Clerk leveraging public and private meradata.
We need to see the code to understand the symptoms.