Message Queues

The Sema Link API uses RabbitMQ for async job dispatch. The API only publishes to queues — it never consumes. Each queue is consumed by a dedicated worker service.

All queues are declared durable on startup. All messages are published with persistent: true so they survive broker restarts.

Queue Reference

Constant	Queue Name	Publisher	Consumer	Purpose
SMS_PRIORITY	sms.priority	API	SMS Worker	OTP and transactional messages — high priority lane
SMS_NORMAL	sms.normal	API	SMS Worker	Standard single sends and due campaign messages
SMS_CONTENT	sms.content	API	SMS Content Worker	Async content moderation for bulk campaigns
DLR_INBOUND	dlr.inbound	DLR Webhook Service	DLR Processor Worker	Raw delivery report callbacks from Celcom Africa
CONTACTS_VALIDATE	contacts.validate	API	Phone Validation Worker	Contact list phone number validation jobs

Connection Behaviour

The RabbitMQ client (src/queue/client.ts) attempts to connect on startup with up to 5 retries and a 2-second delay between attempts. If the broker is unavailable after all retries, the API continues running but queuing is disabled — publish calls will fail silently for that session.

Publishing

The API publishes via the publishToQueue(queue, payload) helper. Payloads are JSON-serialised. The exact payload shape for each queue is defined by the consuming worker.

Worker Services

Workers are separate Node.js processes — they are not part of the semalink-api repo. See Backend Services for details on each worker's responsibilities and message flow.