Scaffold CRUD

Given a table, scaffold-crud generates a bundle of endpoints that cover the standard CRUD operations. Available from the console and as a REST endpoint:

POST /api/namespaces/:slug/databases/:dbSlug/endpoints/scaffold-crud
Content-Type: application/json

{
  "table": "orders",
  "operations": ["list", "get", "create", "update", "delete"]
}

operations is optional — the default is all five.

What it produces

For a table orders with columns (id INTEGER PK, customer_id INTEGER, total REAL, status TEXT):

Slug	Kind	SQL	Output
list-orders	query	SELECT * FROM orders LIMIT ? OFFSET ?	rows
get-order	query	SELECT * FROM orders WHERE id = ?	first_row
create-order	mutation	INSERT INTO orders (customer_id, total, status) VALUES (?, ?, ?)	rows_written
update-order	mutation	UPDATE orders SET customer_id = ?, total = ?, status = ? WHERE id = ?	rows_written
delete-order	mutation	DELETE FROM orders WHERE id = ?	rows_written

Field types are inferred from the column types (INTEGER → integer, REAL/NUMERIC → number, TEXT/CLOB → text, etc.). Primary-key columns are excluded from create inputs (the DB autoincrements). NOT NULL columns become required: true.

Idempotent on re-run

If a slug already exists (e.g. you ran scaffold-crud once and added a column), the new attempt returns:

{
  "ok": true,
  "data": {
    "created": [/* newly created endpoints */],
    "skipped": [{ "slug": "list-orders", "reason": "Slug already exists" }]
  }
}

You won’t lose existing endpoints. To regenerate, delete them first.

When to scaffold vs hand-write

Scaffold-crud is the right answer for boilerplate plumbing — list this, get-by-id that, basic create/update/delete. As soon as you want joins, computed columns, soft deletes, RBAC beyond $user_id, or non-trivial filtering, write the spec by hand.

You can also call this from the agent: “scaffold CRUD for the orders table” → propose_scaffold_crud → confirm → run.