We are running an annotation process with multiple users.
The dataset that we are using contains subsets of related examples (e.g. sentences coming from a given document). We were wondering if you can think of any way to tell prodigy to send all of the examples pertaining to a particular document ID to the same user;
Do we have any control mechanisms to define such rules?
Hi! At the moment, the stream is a simple Python generator, so there's no direct mechanism for Prodigy to communicate the name of the current user session (which is only decided at runtime) back to each stream. Otherwise, all streams would have to be functions.
So the most straightforward way I can think of would currently be to just run separate instances and give each user a dedicated dataset and/or name you provide when you start the server. Within the stream, you can periodically check what's already in the datasets, and based on who uses the instance, the stream can decide whether to send out an example or not.