Feature: Get labels and vectors from index

[leoplusx]

Describe what you are looking for

I'd like to be able to access all labels stored inside an index. I'd also like to be able to access vectors from the index by id.

hnswlib has these methods:

• index.get_items(ids)
• get_ids_list()

I find them useful, and am missing them in usearch.

Can you contribute to the implementation?

• I can contribute

Is your feature request specific to a certain interface?

Python bindings

Contact Details

No response

Is there an existing issue for this?

• I have searched the existing issues

Code of Conduct

• I agree to follow this project's Code of Conduct

[ashvardanian]

Consider done. I will add it in the next couple of days ;)

[ashvardanian]

🎉 This issue has been resolved in version 0.14.0 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀

[ashvardanian]

@leoplusx I managed to implement those today and they are already available in the latest release. You can use index.labels property to get all labels, use index[x] to get one vector under the x label, and check if it's present with the simple in operator. More enhancements coming soon!

[leoplusx]

Actually, it doesn't seem to work. At least in Python:

Label values all seem to be 0.

Even though, when I run a search, it's clear that there actually are non-zero labels in the index.

Code example:

matches, distances, count  = iu.search(emb_np, k=10)
matches
> array([36517119, 38332614, 38332473, 38332625, 35783650, 38332484,
       36786081, 36927136, 38642553, 37563334])

iu[36517119]
> IndexError: Couldn't find key.

len(iu.labels)
> 3361000

len(set(iu.labels))
> 1

iu.labels[:10]
> array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0])

[ashvardanian]

Oh, wow. The new tests cover it, so it must have been solved, but not yet released. Will push the new release sooner.