annotate record based on `KNN vector similarity` with annotated data

[davidberenstein1957]

Normally KNN is used for clustering. Assuming un-annotated samples have a high similarity with newly annotated examples, it might be interesting to label them based on this.

MVP

from argilla_plugins.programmatic_labelling import knn

knn(name="dataset", sim_threshold=0.9)
knn.start()

Stretch filtering variables like query could be added to limit the sync.

[davidberenstein1957]

This has some overlap with the extend_matrix reasoning.

[davidberenstein1957]

top_k most similar is already possible with rg.load()