>>105793365 (OP)
My experience as a webshitter working with various systems, old and new:
>sqlite
I love it in theory and it's ideal for local software, but can't recommend it as the DB for a web backend. Even for small MVPs.
It can handle it from a performance standpoint, but it has a lot of quirks and terrible defaults. Types are just suggestions, PKs can be null, foreign keys do nothing until you issue a special pragma telling sqlite to actually enforce them.
It's really easy to fire up a real DB with Docker nowadays, so just do that instead.
>mysql
I've only seen this used with Wordpress and other legacy PHP stuff. To be honest I haven't looked into it recently, newer versions might be fine. It's not on my radar when deciding what DB to use and I don't think anyone would choose it for greenfield now.
>postgres
Basically the default for new projects now.
Has really good extensions for vector similarity search (pgvector with pgvectorscale), so it's an especially attractive option for AI projects that don't want a separate vector DB.
>sql server
It's got a really good UI which is appealing to non-technical/semi-technical management types, and also integrates well with .NET (Entity Framework). Neither of the above are really worth the licensing fee anymore so it's slowly dying off. Last time I checked, no vector features unless you use the Azure cloud version.