juanjo finishes ripping Supabase out of Mike's backend
The fork's database rewrite is done - every backend route now talks to plain Postgres through a different toolkit, with the frontend none the wiser.
Mike originally leaned on Supabase, a hosted service that handled both the database and user logins. juanjo's fork swaps that out for a local Postgres setup, and this stretch of work is the mechanical finish line: every remaining backend file - user lookups, access checks, document storage, chat, projects, workflows, tabular data - has been rewritten against the new database layer.
The interesting trick is that the frontend didn't have to change at all. The team built small translator layers that keep the data shape going to the browser identical, even though the plumbing underneath is completely different. Two bugs surfaced after the sweep - a streaming-response crash and a column-naming mismatch - and both got patched, a useful warning for anyone copying this approach.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?