Ask your AI (ChatGPT or Codex) to move a record or a whole table out of the way. It goes to the Siglata trash, disappears from queries and comes back exactly as it was if you ask within 30 days.
In a spreadsheet, deleting the wrong row is a scare. Either someone remembers what was there, or the data is gone. When your AI keeps the company's tables, it has to be able to clear out what's left over and, if it gets something wrong, undo it without relying on anyone's memory. Siglata rebuilds this part so your AI looks after the tables from start to finish, with a safe way back.
You ask in plain words. Your ChatGPT sends a whole table to the trash, with its records and formats, or just some of its records. In the trash, they disappear from queries, exports and history, and no one can write to them. Another table can even use the same name.
For 30 days, everything can come back. The table or the records return exactly as they were, without a new revision. After that, Siglata deletes the item for good.
Who can do what follows each person's access. Sending to the trash and restoring need write access. Deleting for good before the deadline works only on what is already in the trash, needs a confirmation, and is limited to whoever created the item, an owner or an admin.
Mogiana Distribuidora, a fictional company, has a Visits table with one record per customer visit. Helena asks ChatGPT:
"Send the three test visits I created yesterday to the trash."
ChatGPT finds the three records and sends them to the trash. The next week, Helena notices one of them was a real visit and asks:
"Restore the visit to Padaria Central from the trash."
The record returns to the Visits table with the same fields and the same revision as before.
Trash for tables and records works today in ChatGPT and Codex.