swiftbursteli

Supabase devs said they were working on longer running jobs. I dont think they ever got around to it. Railway runs 6h jobs for me no problem. The thing about your example use case is that edge is more than capable of doing that. its the longer more intensive runs - sequential data aggregation, processing, etc. Supabase cant handle that and you need an app, a server, or a microservice for this. Or just migrate to a new platform that has it. run a scheduled cron job on whatever it is you need and it'll work fine. $5 vps and fire away. I don't think this will take off - not to dissuade you, but maybe to have you shift the request to get the eyes of SB devs. I wish it would get bundled into supabase natively because as a standalone app you're probably gonna want a lot more than just long-running background tasks to get ppl to add yet another CLI tool to the list. I would really, really like to see this like an "automations" panel in SB. but we probably wont get it even if there's already a tool out there for it

the username is literally no-thought. we're talking to a brainworm. because i like this idea though, If i was dev then i'd create two buckets and have raw CSVs and processed CSVs. You can simply append \_processed to the filetype after an LLM processes it. Id consider a bridge storage too - for stuff in progress. That's if you're running functions on supabase. If its handled server-side then just process it & push the changes sequentially. Depending on the service too. It would probably make sense to bring the processed CSV into postgres. Say its like a medical record situation, probably 30-60 days is when your customers need the processed data. after that you can send the data from postgres back to storage for cheaper storage costs.

If you DO go selfhosted like I did. Do not look at anything else: just setup supabase. It’s literally the best thing on the market and there’s nothing that comes close. For managed, multi project management when you’re a full stack eng/ the whole team I would use railway. I found it to be easier to setup the webserver to work - just like you want it to. I don’t have to write in deno or rely on some magic box to figure something out… just throw an Opus 4.8 at a node.js project and tell it to wire up your webserver. Pros for supa: \- Supabase offers secure REST and graphQL endpoints for every data table. \- Their UI is second to none. \- I think their magic link and OAuth is more foolproof \- Basically makes a 2-tier architecture. The BaaS and the client. Pros for Railway: \- Simpler pricing, cheaper for same use case (speaking from my own experience only here) \- can deploy anything with 2-clicks. \- it’s 3-tier arch (data, webserver, client). So most of your work will lay on the webserver. The webserver can handle anything - stripe, twillio, sentry, etc.

Nice. Why not use realtime?

Assume everything is ai now. The internet as we remember it is largely done for

That’s a pretty good protocol. Yeah there’s templates everywhere online for sites - you’re basically reskin the template and ship a backend to that. Recently I’ve been having a ton of fun with SQlite. Check it out. You probably don’t need Postgres either for most of those cases.

Download coolify on a vps 1 click deploy supabase Be shocked at how similar it is to the real thing. Minus the pricing shenanigans with disk IO and concurrent worker stuff... its such a blessing for self hosting

I have had a bad experience on Supabase cloud which led to multiple downtimes and silent failures in Dec of last year. Disk I/O limits, concurrent runners just kept staying up, randomly dropping, or persisting (I kept trying to pool, back off, do all these tricks to avoid it from happening. None of it worked). Self hosted supabase posgres on Coolify. Same thing Cloudflare at the front. Had config issues with Coolify but resolved. If going self-hosted, supabase is the best OSS product full-stop. If going managed, I much prefer railway.