Configuration

const hookgres = new Hookgres({
  pool, // or connectionString
  schema: "hookgres",
  retrySchedule: DEFAULT_RETRY_SCHEDULE,
  requestTimeoutMs: 15_000,
  disableEndpointAfterMs: 5 * 24 * 60 * 60 * 1000,
  concurrency: 10,
  pollIntervalMs: 1_000,
  lockTimeoutMs: 45_000,
  secretRotationGraceMs: 24 * 60 * 60 * 1000,
  fetch: globalThis.fetch,
  userAgent: "Hookgres/0.1",
  onError: console.error,
});

Option	Default	Description
pool	—	A pg.Pool. Hookgres never closes a pool it didn't create.
connectionString	—	Used to create a pool when pool isn't given. close() ends it.
schema	"hookgres"	Postgres schema for the tables. Lowercase letters, digits and underscores, not starting with a digit, up to 63 characters.
retrySchedule	5s, 5m, 30m, 2h, 5h, 10h, 10h	Waits between attempts, in milliseconds. Attempts = length + 1.
requestTimeoutMs	15000	How long to wait for a response before the attempt counts as failed.
disableEndpointAfterMs	5 days	Disable an endpoint after it has failed this long with no success. null never disables.
concurrency	10	Deliveries each worker claims and sends at once. Also the batch size of runOnce().
pollIntervalMs	1000	How long an idle worker waits before checking again. Sends from the same process wake it sooner.
lockTimeoutMs	requestTimeoutMs + 30s	How long a claimed delivery stays locked before another worker may retry it after a crash. Must be greater than requestTimeoutMs.
secretRotationGraceMs	24 hours	Default grace period for rotateEndpointSecret.
fetch	globalThis.fetch	Custom fetch, for an SSRF guard, a proxy or tests.
userAgent	"Hookgres/0.1"	The user-agent header on every request.
onError	console.error	Called with unexpected worker errors, like a lost database connection. Failed deliveries are not errors.

The constructor throws if neither pool nor connectionString is given, if schema is invalid, if retrySchedule contains a negative or non-finite number, or if lockTimeoutMs isn't greater than requestTimeoutMs.

Several instances

Different schema values give you fully separate queues in the same database, for example one per environment or per product:

const billing = new Hookgres({ pool, schema: "hookgres_billing" });
const platform = new Hookgres({ pool, schema: "hookgres_platform" });

All workers that use the same schema should use the same options, especially retrySchedule.