why do we need 'y' col in predict df?

[limitspro]

in docs of predict func:
dataframe or dict of dataframes containing column ds, y with data

isn't that we do not have y when we predict? could anyone help? really confuse me