inngest-durable-functions

Call inngest.createFunction with a unique id, a triggers array containing events or cron schedules, and an async handler receiving event and step. Wrap all side effects like API calls and database writes in step.run() so they are memoized and retried independently.

Each step retries automatically with exponential backoff, defaulting to 4 retries per step, configurable via the retries option. Throw NonRetriableError for permanent failures like missing users, or RetryAfterError to respect API rate limits with custom timing.

Yes, add a cron trigger like { cron: "0 9 * * *" } to the triggers array, optionally with a timezone prefix such as TZ=Europe/Paris. You can combine up to 10 event and cron triggers on a single function.

Each function run allows a maximum of 1,000 steps, 4MB of returned data per step, and 32MB of combined run state. If you hit these limits, split work into smaller functions connected via step.invoke() or step.sendEvent().

Steps re-execute when their step IDs change, because Inngest hashes step IDs as state identifiers for memoization. Keep step IDs stable across deploys; changing an ID intentionally forces re-execution of that step.

Yes, the extendedTracesMiddleware from inngest/experimental integrates with OpenTelemetry providers like Sentry and auto-instruments libraries such as pg, redis, express, and winston. Pass a structured logger to the Inngest client for correlated function logs.