I started playing with Sonic recently, and it’s a really nice lightweight full-text search system. It doesn’t have any distributed functionality, but you could potentially run it using a side-car pattern for super-fast full-text search.

You can see my usage here, https://github.com/entone/web_archive_viewer

This is a perfectly valid question for the Ecto category btw.

I’m a big proponent of a polyglot database approach, and choosing the right tool for the job. I think pushing your text data to a purpose built DB allows you to make the right choices with regards to your SQL architecture without sacrificing anything for your search functionality.

Same thing for telemetry data and logs.

Instead of duplicating the columns through concatenation, why not use a trigram index for the columns that you need?

Also another issue I had when using postgresql’s full text search, was the minimum characters required in the search query.

I ended up using trigram indexes instead for my text searching. Though ymmv depending on length of text that you are searching. For short words like in your skills table, a trigram index might work better