fix: search distance calculation - Githubissues

oasysai / oasysdb

Hybrid vector database with flexible SQL storage engine & multi-index support.

https://docs.oasysdb.com/

Apache License 2.0

357 stars 12 forks source link

fix: search distance calculation #78

Closed edwinkys closed 6 months ago

edwinkys commented 6 months ago

Purpose

This PR is to fix the bug in this Google Colab notebook: https://colab.research.google.com/drive/1zQtpXkkWASt_37aiA7m5ZZcMqBnUGxOZ?usp=sharing

In short, the search result distance returned is that of Euclidean instead of the configured Cosine.

Cause

https://github.com/oasysai/oasysdb/blob/b04772b5dc9fc27faab3e2c42a1f39885f9c2628/src/func/collection.rs#L297

https://github.com/oasysai/oasysdb/blob/b04772b5dc9fc27faab3e2c42a1f39885f9c2628/src/func/utils.rs#L337-L349

Testing

[x] I have tested this PR locally.
[x] I added tests to cover my changes, if not applicable, I have added a reason why.

I added a new test in Python to check the result.

Chore checklist

[x] I have updated the documentation accordingly.
[x] I have added comments to most of my code, particularly in hard-to-understand areas.