Library / n8nWorkflow library
These workflows are hand-authored examples on fictional data, not exports from a client system.
Illustrative endpoints. No connected accounts. Review and configure the JSON before running it in n8n.
Trigger A booking webhook receives a contact event.
Systems · illustrative n8n · GoHighLevel contact API · Owner notification webhook
Why it is shaped this way
- Dedupe key: the incoming eventId, not the customer’s name.
- Upsert: the receiving CRM matches on email in this fictional contract.
- Retry: none automatically; failed requests go to a person for review.
- Error branch: an HTTP error is carried into the explicit failure check.
- Alert: notify the owner; sample static dedupe needs durable storage in production.
What to change before running
- Replace the contact and notification endpoints and configure credentials.
- Match incoming fields and the eventId dedupe key to your source contract.
- Replace sample static dedupe with durable storage and test the failure branch.
Trigger A missed-call webhook receives a call event.
Systems · illustrative n8n · GoHighLevel call webhook · SMS provider · Callback queue
Why it is shaped this way
- Dedupe key: callId makes repeated webhook deliveries a single attempt.
- Send: the illustrative SMS endpoint accepts a fictional callback number.
- Retry: none automatically, to avoid duplicate customer messages.
- Error branch: failed sends create a callback alert instead of a success record.
- Alert: a person follows up; credentials and a durable call ledger are required.
What to change before running
- Replace the SMS and callback endpoints and configure sending credentials.
- Map callId and the caller number, then replace sample dedupe with durable storage.
- Set your message and contact permissions; test duplicate calls and SMS failure routing.
Trigger An hourly schedule checks the failed-event queue.
Systems · illustrative n8n · Failed-event API · Owner notification webhook
Why it is shaped this way
- Dedupe key: eventId collapses repeats inside each hourly digest.
- Window: the example feed contract returns failures from the previous hour.
- Retry: none automatically; an unavailable feed also raises an alert.
- Error branch: feed errors become a digest so missing data stays visible.
- Alert: one owner digest; delivery failures remain in the execution log.
What to change before running
- Replace the queue and owner-notification endpoints and configure credentials.
- Choose the schedule, timezone and failed-event fields for your receiving system.
- Add a last-seen checkpoint and test empty queues, read errors and notification errors.