Page MenuHomePhabricator

Decide where to store past labeling data
Open, LowPublic

Description

The machine vision project involves generating depicts suggestions for Commons images through one or more machine vision providers, and then allowing users to review these suggestions for accuracy. The review data is valuable for further machine vision training, and per T227353: Define the lifecycle for MV-generated depicts suggestion data Product wants to keep it long-term. There isn't a strong reason to keep it in the production databases though, if there is a cheaper alternative. (See T227351: Define DB table schemas for MV depicts suggestions for size projections.)

Event Timeline

What other storage options do we have? We don't really have any "cold" storage for infrequently needed data, AFAIK.