Change Details

Per discussion in T258978, using SQLite is going to be problematic for the production service. So instead, 1. mwaddlink needs to create and then write to MySQL tables instead of SQLite files during the model training phase 1. mwaddlink needs to read from MySQL tables instead of SQLite when responding to queries While https://pypi.org/project/sqldict/ was suggested as a drop-in replacement for sqlitedict (currently in use), it's probably better to use a different implementation for doing a select from the table(s) and loading that into a python dict.

Per discussion in T258978, using SQLite is going to be problematic for the production service. So instead, 1. mwaddlink needs to create and then write to MySQL tables instead of SQLite files during the model training phase 1. mwaddlink needs to read from MySQL tables instead of SQLite when responding to queries Remaining tasks: - [ ] Training pipeline to write directly to MySQL (?)

Per discussion in T258978, using SQLite is going to be problematic for the production service. So instead, 1. mwaddlink needs to create and then write to MySQL tables instead of SQLite files during the model training phase 1. mwaddlink needs to read from MySQL tables instead of SQLite when responding to queries While https://pypi.org/project/sqldict/ was suggested as a drop-in replacement for sqlitedict (currently in use), it's probably better to use a different implementation for doing a select from the table(s) and loading that into a python dict.Remaining tasks: - [ ] Training pipeline to write directly to MySQL (?)