Updated at

The updated-at block installs the track_updated_at trigger, which sets updated_at on every update, and a repository extension that fills createdAt and updatedAt before the row is sent.

better-supabase add updated-at

gen marks every table that has the columns and installs the extension in the generated schema, so defineSupabase(schema) stamps writes with no .use() call. Rename the columns in the block's options:

blocks: {
  "updated-at": { options: { createdAt: "inserted_at", updatedAt: "modified_at" } },
},

What the extension writesCopy Anchor Link

• Inserts get createdAt and updatedAt.
• Updates (including soft deletes from trash) get updatedAt.
• Upserts that may update an existing row only get updatedAt, so the original createdAt survives. With ignoreDuplicates: true both are set.
• Values you pass yourself are refused with invalid_request, so a client cannot backdate a row. Imports and backfills pass { override: true } to keep them:

await db.customers.create(
  { ...row, createdAt: imported.createdAt },
  { override: true },
);

Generated JSON Schema and OpenAPI documents mark these columns readOnly. In a defineApi document they are also left out of the required fields of request bodies, and a column without a comment gets a description: "Set by the server when the row is created." for createdAt and "Set by the server on every write." for updatedAt.

The time comes from defineSupabase(schema, { now }), a function that returns a Temporal.Instant, which makes tests deterministic. See Temporal. In a runtime without Temporal and without a now option, the extension leaves both columns to the column default and the trigger.

A schema you write by hand installs the extension with .use(). On a generated schema the same call replaces the installed one:

import { updatedAtExtension } from "better-supabase/blocks/updated-at";

const db = defineSupabase(schema, { now: () => clock.now() }).use(
  updatedAtExtension(),
);