how to instantiate saqc object with existing flags that are not float

I could not get saqc to generate an object from dataset with existing simple-schemed flags.

From my understanding, if i have data flagged UNFLAGGED or BAD, i should be able to initialise a SaQC object without having to translate the flags to float values by myself by using the simple scheme - since thats the point of the scheme.

I could not find no documentation on how to do that anywhere and i tried the following variants, that i thought should all be valid, but all throw errors instead of SaQC objects:

data = pd.Series([1,2], index=pd.date_range('2000',periods=2,freq='D'), name='data')
flags = ['GOOD', 'BAD']
qc = saqc.SaQC(data=data, flags=flags, scheme='simple')

Doesnt work (AttributeError: 'str' object has no attribute 'items'), i thought maybe its because flags are passed as a list (wich actually is allowed by the signature)

data = pd.Series([1,2], index=pd.date_range('2000',periods=2,freq='D'), name='data')
flags = pd.DataFrame(['BAD', 'UNFLAGGED'], index=data.index, columns=['data'])
qc = saqc.SaQC(data=data, flags=flags, scheme='simple')

That fails with ValueError: dtype must be float or categorical, so i tried:

data = pd.Series([1,2], index=pd.date_range('2000',periods=2,freq='D'), name='data')
flags = pd.DataFrame(['BAD', 'UNFLAGGED'], index=data.index, columns=['data'], dtype='category')
qc = saqc.SaQC(data=data, flags=flags, scheme='simple')

And that fails with: ValueError: Cannot cast object dtype to float64

-> of course, replacing 'BAD' and 'UNFLAGGED' by floats, than works: but i thought the whole point of the schemes is that one does not have to know/care for the inner representation of the schemes flagging values, so that seems to be a little pointless to me.

Also, if there is an easy solution to getting that done, it should be documented more prominently (for example in a documentaiton o the SaQC object itself) - since this is a quite prominently advertised feature and one should not have to try out different interpretations of the errormessages to get the result.