If you don't see a need to massively scale horizontally from the get go, SQL. (And if you need to ask the question, you are likely not equipped to handle parallelism/distributed issues coming out of not having ACID guarantees)
A good way to estimate that is if you'll be mostly writing to the database instead of reading from it.
You can also go half-way with things like object storage in Postgres (jsonb fields), so that's usually a no-risk choice for a project in an early phase of development.
A good way to estimate that is if you'll be mostly writing to the database instead of reading from it.
You can also go half-way with things like object storage in Postgres (jsonb fields), so that's usually a no-risk choice for a project in an early phase of development.