Sume SDK calls resolve with error: make your task throw

Generated @sume-com/sdk operations do not throw on an API error. They resolve with { data, error, response }, so a task that ignores error finishes as a success with an undefined job id, and a job runner never retries it.

Sume facts are from the SDK overview and Waiting for runs and jobs; the retry rule is from Trigger.dev's docs. Read 2026-10-01.

Which SDK calls throw and which do not?

The SDK docs say generated operations resolve with { data, error, response }, while subscribeFormatRun and waitForRun throw because a poll loop has nowhere to put a non-result. waitForJob throws SumeJobTimeoutError and SumeJobRequestError, both carrying jobId.

Why does that matter inside a task runner?

Trigger.dev's docs say an uncaught thrown error triggers a retry per your config. A returned error object triggers nothing. A Sume 402 or 429 would end the run green and hand the next step an undefined id.

What is the minimal fix?

Check error right after the call and throw an Error that includes the response status. The idempotency header makes the retry that follows safe, because an exact retry returns the original job instead of billing twice.

import { createSumeClient, generateVideoV1 } from "@sume-com/sdk";

const client = createSumeClient({ apiKey: process.env.SUME_API_KEY! });

export async function submit(prompt: string, key: string) {
  const { data, error, response } = await generateVideoV1({
    client,
    headers: { "idempotency-key": key },
    body: { prompt, mode: "async" },
  });
  if (error || !data) {
    throw new Error("sume " + response.status + " " + JSON.stringify(error));
  }
  return data.data.request_id;
}

Should every error be retried?

No. Throwing hands the decision to the runner, which retries by default. Pair it with the skip rules in the Trigger.dev retry post so a 400 or 402 stops at once.