Hey team, is it possible to retrieve a total relevance score instead of the text match score?

We don't calculate any additional scores besides text relevance score. Results are first scored based on text relevance, and then sorted based on other criteria mentioned in sort_by - these are simple sorts, and don't affect the score.

I'm overriding the text match sort and I'm seeing higher scores at the bottom of the query results.

This is because

_eval(status:active),_text_match:desc

explicitly places active status results on top, ignoring text relevance score. Then within each bucket (1) results that match the filter (2) results that don't match the filter, the value for the

eval

is tied, so then the text match score is used to break the tie. You probably want something like

_text_match(buckets: 10):desc, _eval(status:active)

to achieve what you're looking to do. The bucketing mechanism is described here: https://typesense.org/docs/guide/ranking-and-relevance.html#ranking-based-on-relevance-and-popularity

Thank you Jason! I see so a query like

_text_match(buckets: 10):desc, _eval(status:active)

would produce results with both active and inactive, ordered by text relevance score

Yup

Hey Jason, would it be possible to retrieve the score in this

text_match_info

? It would be quite useful knowing the relative difference between results, and further modifying it after processing results. I'm currently querying by

text_match_type:max_weight

I'm mainly looking to assign scores to my results, if fetching the score is not possible do you recommend another method to assign scores to results?

The search API response should have that exact field

So I'm using the go sdk and

TextMatch  *int64

is the only score I see

We might have to add that to the go library to parse from the http response from Typesense. Could you open a github issue in the go sdk repo?

Yup will do, just opened an issue

We'll look into this this week and keep you posted in the Github issue.

Thanks again Jason!