Closed iwilltry42 closed 2 months ago
So is it expected that when using archive, user shouldn't be ingesting files since those file ingestion won't be made into archive anyway?
Yeah, if you want to "update" an archive, you would do import
-> ingest
-> export
. That's essentially, because the archive doesn't contain information about the embeddings function.
I guess we could somehow serialize it with a very limited parameter set that makes sense for our setup, but I'm not sure it's worth the effort.
knowledge export <dataset> --output foo.zip
knowledge retrieve -d <dataset> --archive foo.zip "some question"
knowledge list-datasets --archive foo.zip
knowledge get-dataset <dataset> --archive foo.zip
knowledge import foo.zip
Depends On (currently using fork): https://github.com/philippgille/chromem-go/pull/88
NOTE: there are quite a few rough edges and missing pieces, e.g. the server-mode version of this is not yet implemented.. we'll follow up on this in the future!