The topic grids (PM system + wizard) showed extra categories like «سیم»,
«کابل», «ثبثب», «dsfg». Those rows were leftover junk in material_categories
from earlier testing — previously invisible because the frontend hardcoded
the 8-topic list; the API-driven topic grid surfaced them. The rows are
deleted (they had no references) and the seeder now deletes any category
outside the 8 canonical slugs, so a re-seed always yields exactly the same
8 categories as before. Dump refreshed from the clean state.
- backups/procurement.dump regenerated from the clean seeded state (was the
junk test data); README documents both the app seeder and the restore path
- docs/casdoor-auth.md: later-batch changelog (fulfillment split, demo
seeder, API-driven topics, hub-and-spoke, count modal, FOUND ceiling,
semantic colors)
The databases live in Docker volumes (not in git), so teammates on a fresh
boot saw an empty system. backups/procurement.dump + casdoor.dump are exact
pg_dump snapshots of the current demo state (users/roles, projects,
warehouses, quotes with receipts, inventory rows; Casdoor org/app/users);
restore.ps1 / restore.sh replace both DBs with the committed state
(--clean --if-exists, verified by a round-trip restore into throwaway DBs).
README documents how to restore and how to take a fresh dump.
ali_mohaghegh