embedding drift with domain classifier

[jiayaopeng]

Hi Hi,

I would like to know if the embedding drift can still use domain classifier to detect? I saw the thread here but the link is not available anymore. I tried to

• shape the embedding data with one element of the embedding one column as showed here;
• currently I was using embedding_report = Report([DataDriftPreset(num_method="wasserstein", num_threshold=0.1)]) to test out but of course when I try to push the report to our self-host evidently instance, it gave me error evidently.errors.EvidentlyError: Request Entity Too Large probably due to the number of columns from the embeddings

Question:

1. after the migration, where can I find the docs as shown in the issue link?
2. is there a classification method implemented for detecting embedding drift(embeddings, not raw text column like described here) (I did saw [there]"(https://github.com/evidentlyai/evidently/blob/main/src/evidently/presets/drift.py#L27) is embedding_drift_method parameter but it is not clear how we use that)?
3. is there other ways to detect embedding drift?

thank you