Skip to main content
n8n-nodes-nephia ships two nodes:
  • Nephia Trigger — creates a watch when you activate the workflow, deletes it when you deactivate it, and starts a run on every event.
  • Nephia — every documented /v1 pull operation, with the credit cost shown on each one.
Both are generated from the same OpenAPI spec as the SDKs, so a new source or parameter appears in the node’s dropdowns rather than in a changelog.

​
Install

In your n8n instance: Settings → Community nodes → Install, then enter:
Self-hosted instances need N8N_COMMUNITY_PACKAGES_ENABLED=true (the default).

​
Credential

Add a Nephia API credential: Test calls GET /v1/account/credits, which is free.

​
Nephia Trigger

Pick a Source, describe what to watch, and set a refresh interval — the floor differs per source and the field enforces it (Watches). The node owns the watch’s whole life: activating the workflow creates it, deactivating deletes it, and the watch id is kept in the workflow’s static data. Do not also create the watch by hand — you would pay for two.

​
Webhook mode (default)

The watch is created with webhook_url pointing at this node’s n8n webhook URL. Nephia POSTs each event, n8n answers 200 immediately, and the workflow runs after. One event, one item, in the envelope the API documents. Needs an n8n instance reachable from the internet.

​
Poll mode

For instances behind a firewall. The watch is created without a webhook_url, and n8n asks for new events on its own schedule using the free replay endpoint. Nothing has to reach your network. The trade-off is latency: you see an event on the next poll, not the second it happens.

​
Signature verification

Each delivery carries X-Signature (how it is built). Paste your signing secret into the credential and the trigger verifies every delivery against the raw body, rejecting anything that fails with 401. Leave it empty and the trigger accepts any POST to its URL. That URL is unguessable, which is the same protection any n8n webhook has — but if the event will drive something that matters, set the secret.

​
Duplicates

Nephia retries a failed delivery at +0s, +30s and +120s, and n8n retries a failed execution on its own terms. The trigger remembers the event.ids it has seen for 24 hours and drops the repeats, but that is best-effort: if running twice would be expensive, stay idempotent on event.id downstream.

​
When a watch is paused

A paused watch — out of credits, too many delivery failures, paused by support — looks exactly like a quiet one. Both modes re-read the watch once an hour and emit a distinct item when the reason changes:
Branch on type to alert yourself. pausedReason says what unblocks it — see Pause reasons.
In Webhook mode the built-in Poll Times field is used only for this hourly check; events still arrive over HTTP.

​
Nephia (action node)

Pick a Resource and an Operation. The operation dropdown shows the credit cost from the spec, so you can price a run before making it.
  • Required fields sit at the top level; everything else is under Additional Fields.
  • Return All follows pagination — cursor, page + per page, or page, whichever the operation speaks — up to Max Pages (default 5). Every page costs credits; the node logs a warning when the cap stopped it early.
  • Split Into Items (on by default) emits one item per result. Turn it off to get the raw response, cursor included, as a single item.
  • POSTs carry an idempotency-key that is reused across retries, so a retried create never makes a second watch.
  • 429 and upstream 503 RATE_LIMITED are retried with Retry-After; 402 says to top up rather than to try again (Errors).
The Watches resource covers all eight watch routes for all seven sources — useful for listing, pausing, resuming, or replaying events from a workflow that did not create the watch.

​
Templates

Six starting points ship with the package under templates/, and are published on n8n.io:

​
Limits

  • One watch per node. Several watches means several trigger nodes, and your plan’s watch cap applies (Limits).
  • The action node’s AI resource is not included; run an agent step on the watch instead, and read it from event.ai.
  • Copying a workflow to another instance re-creates the watch on first activation, because the old one points at the old URL.