autoGPT and this project

i wonder if there is a usecase to use autogpt + a MB agent to restrict source data to know metadata providers and then import everything available

1 Like

Unless you can check the data before importing, that would be a terrible idea. Even for tools like a-tisket, it’s essential to manually check everything and not just enter the way it is.


not even if the metadata from apple, spotify, lastfm, deezer is all the same?

As @wtfislibrious points out, the current import scripts have humans checking the data, and errors still get through. Especially with dates.

A bot would potentially make a huge mess. “autoGPT” would do nothing special that the current script doesn’t do. AI is not a magic fix. It stands for “Automated Idiocy” and would create a high error percentage.


Not even potentially.
Many years ago, we had an editor who made their own bot scripts that scraped a Japanese discography site and automatically imported the data into MB. These imports left out important chunks of data (e.g. labels) and created a lot of duplicate entities. The sheer volume of imports was too much for human editors to efficiently clean up. This went on for months and spanned three accounts (that we know of). The person responsible was banned and the accounts locked, and to this day there are still entities in the database left in the state they were imported in (mostly semi-obscure Japanese stuff).


dont u think a large language model would know that “2023 CAT music” and “CAT music” is the same? and on top of that u can place any limit like dont add artists, labels and focus it on filling missing data

Do you have anything that could be tested to see what it can find?

1 Like

There’s labels that share names and you have to look into the logo being used or the label’s address. Using AI as it currently exists would be terrible as that’s the type of stuff it would mess up. Not to mention that AI often makes up info.

Someday I do think AI will be useful for filling data in, but we’re definitely not there yet.

1 Like