So I have tried to set the db value in the return of the recipe to False, but still a SQlite Database is created and data is stored therein. We dont want to store any data in a persistent object due to pricavy concerns, we instead we store the labeled data directly in a cloud storage (using a callback).
Is it possivle at all to run the backend without a db as said in the docs?
I have just double checked and we indeed removed the possibility to disable to the built-in DB functionality when working on one of the internal refactors but we have not updated this part of the docs. Sorry about that!
However, given it's actually useful to have Prodigy without the DB connection, we are happy to put it back. In the meantime you could use the before_db callback to remove the information from the Prodigy task before saving to DB. This would result in garbage database that you'd need to delete, but it would serve as workaround.
For example, for the NER annotation this callback could look like this:
def before_db(examples):
for eg in examples:
eg["text"] = ""
eg["spans"] = []
eg["tokens"] = []
eg["meta"]={}
return examples
