How to call a volatile postgres function with array input args?

[henpanta]

Hi!
I have a table called worker (id, user_id, surname, forename) and the following (tracked as mutation) function:

CREATE OR REPLACE FUNCTION public.insert_records(objects worker[]) 
RETURNS SETOF worker 
LANGUAGE plpgsql VOLATILE 
AS $ function $ BEGIN
 INSERT INTO public.worker 
 SELECT * FROM unnest(objects) RETURNING ret;
 RETURN QUERY select * from public.worker;
END;
$ function $

I'd like to execute a mutation from graphql:

mutation insertWorkers {
  insert_records(args: {objects: [
    {surname: "A", forename:"B", user_id: "1"},
    {surname: "C", forename:"D", user_id: "2"}
  ]}) {
    id
    user_id
  }
}

When I run the mutation, I got this error:

{
  "errors": [
    {
      "extensions": {
        "code": "parse-failed",
        "path": "$.selectionSet.insert_records.args.args.objects"
      },
      "message": "A string is expected for type: _worker"
    }
  ]
}

Why would hasura interpret worker array as _worker string?
Can I use the hasura-generated [worker_insert_input!]! as input args, and make it work with a function?

My goal is to circumvent the max 65535 parameters limit, so I can bulk insert hundreds of thousands of records in the same transaction. I found that UNNEST sql function could 'unpack' an array, so this is the reason to create a new sql function for the bulk insert. If not the method above, what would be the correct way to achieve this goal? Batching the insert would be very complicated due to my insert triggers and other side-effects (not explained above), I try to avoid it.

regards,
Norbi