Make match graph communities a first-class objects:
User can browse detected communities
User can assign a name to communities
User can edit communities (add/evict videos)
User can browse videos for each community
User can see community for each vide on video-details page
There is a background task to auto-detect communities.
Problem:
After each detection a new set of communities will be generated.
How to match existing communities with a new ones?
Some communities may split, some may merge, some may completely disappear. How to detect and how to handle these situations?
What if two named communities are about to be merged?
What if named community is about to split?
Possible solution:
We can have two different methods of file groupping:
Auto-detected communities, which are volatile and which will be replaced each time when the corresponding pipeline task is executed.
User-defined file collections, which are stable, manually created, and has some additional attributes (name, description or other metada).
User can create a collection from a community. In this case community will remain as is, but a new named collection will be created with the same content.
Make match graph communities a first-class objects:
Problem:
Possible solution: