My deployment has been stuck in the Provision stage for over 18 hours. The Republish button does not start a new build — every attempt still shows the same deployment timestamp. Cancelling and republishing does not work.
The error shown is: operator class "int4_ops" does not accept data type text on a CREATE UNIQUE INDEX statement that Replit’s migration system is generating incorrectly.
I have confirmed:
-
The index already exists correctly in the production database with no operator classes
-
The Drizzle schema, snapshot, and migration file have all been corrected
-
The broken SQL is not present anywhere in my codebase
-
Replit’s deployment system appears to be reading a cached migration state from the original failed deployment 18 hours ago
-
The Republish button is not triggering a fresh build — every attempt shows “Started about 18 hours ago by Emma”
I need you to force-clear the stuck deployment so a fresh build can start. My live site is currently frozen on an old version and I cannot ship any updates.