Yes, it is very close. BetterBibtex in Zotero installs runs a local server.
When you run the curl script, Zotero opens the red citation picker shown in the picture, and then curl returns a JSON string with all the selected items. No need to install any extra software or to program the search window, but not as well integrated as /zotero either.
The only thing missing is how to translate the numeric id in the string to the alphanumeric item identifier to link back to the item, or maybe the zotero://select links also work with the numeric id. I did not find any documentation. This is all I found: Notes 2020-02-23 Zotero select - Google Docs
BetterBibtex could also easily return the identifier with the query, the website mentions that they can add more fields to the output if desired.
I just looked at the Zettlr solution, Citations - Zettlr Docs, this uses another feature of BetterBibtex, automatic export into a .csl.json file (Step 2). I don’t know if the export is still fast for very large libraries. I tried it on mine and it takes about a minute. The file does not contain the identifier string needed for the zotero://select link either.
What Zotero really needs is a local client API that mirrors all functionality of the web API and more.
Sadly, at the moment Zotero is moving towards a cloud-only tool.
I posted another comment on how Logseq could interface with the new Zotero 6 notes while maintaining privacy: Support for Zotero 6 Notes import - #3 by gax