Vector fields store meaning as numbers, and searching them needs a database that supports vectors. On Pawabase Cloud every project’s database supports them, so you do not need to do anything.

​
If saving a table with a vector field is refused

You will see a clear message that vector fields need vector support in the database, and nothing about your table is changed. This only happens when your project uses a database that does not offer it. What to do:
  • On Pawabase Cloud, contact support with the project name. It is not something to set up yourself.
  • If you brought your own database, ask whoever runs it to turn on vector support for it, then save the table again.

​
If a search says it can’t run

Vector search works only on PostgreSQL databases. A project on another kind of database can keep a vector field’s values, but cannot search them, and says so when you try.

​
If a write is refused after you add a vector field

With Fill in automatically on, a record is saved only when its text can be turned into a vector. If your embedding model or its provider is down or out of credit, the save is refused with a message saying why, rather than saving a record search can never find. Fix the provider (see AI → Providers) or switch to another model, and save again. If the message says the vector size does not match, the model produces a different number of values than the field holds. Make a new field of the right size and fill it again.