Currently make data will execute make_vectorstore.py directly, which drops the whole on disk data base and repopulates it from scratch. It would be nice to cache the results of the embedding for future reuse instead of reusing the api.
Additionally more interactive tools (cli, ui?) to:
populate the vectorstore initially
maintain a vectorstore overtime (e.g. update with just the new released set)
filter suggestions for the 99 and the commander by specific set
would be cool
A place to start be would be the filter_main_sets() function, which has the behaviour and a suitable API, but can be hoisted up to the user level more effectively.
Currently
make datawill executemake_vectorstore.pydirectly, which drops the whole on disk data base and repopulates it from scratch. It would be nice to cache the results of the embedding for future reuse instead of reusing the api.Additionally more interactive tools (cli, ui?) to:
would be cool
A place to start be would be the
filter_main_sets()function, which has the behaviour and a suitable API, but can be hoisted up to the user level more effectively.