Search bar algorithm not perfect

@artur does this suggestion by @Rekrt offer a path to a improved search in Softr?

What I would actually love though, is Softr implementing a nice UI to bring vector-based search to our Airtables/Data.

So this would mean that Softr:

  1. Lets us set the fields to be taken into account for search
  2. Lets us map an Airtable field where the vector for that record can be stored
  3. Request a vector for the record with help of the OpenAI API
  4. Writes the vector to our database
  5. Runs the users search queries in our Softr app through the vector API
  6. Calls the Airtable API based on the query vector, and searches the Airtable field containing the record vectors so that it get’s returned the closest matching records in our data source.

@artur Kindly following up here.

@Rekrt, to achieve the functionality you described, we would need to change the search functionality and first tokenized the input into multiple inputs and then perform search with OR on all of those tokens but the results will be fuzzy and we will end up with ordering issues and so on… I don’t see a quick fix here yet

1 Like

Any chance you can share or point me to a full FAQ on all the search operators you can use when building queries? As in, do typical boolean queries work? What about the typical google operators like - or “xx”? I’m just looking for a complete understanding of how the search bar works so we can properly relay this to our users. Thanks!