Durable workflows are generally available

Workflows let you split a long job into steps that survive restarts, deploys, and crashes. Each step is checkpointed, so a failure retries that step instead of the whole run.

What’s included

• Retries with backoff per step, with a maximum attempt count
• Timeouts per step and per run
• Step-level logs in the dashboard and through cloud logs --workflow
• Manual replay from any failed step

Get started

import { workflow, step } from "@cloud/workflows";

export default workflow("send-invoices", async () => {
  const accounts = await step("load", () => loadAccounts());
  await step("send", { retries: 5, timeout: "2m" }, () => sendAll(accounts));
});

Workflows are billed as regular compute. There is no extra charge for checkpoints.