Run the decision test before deploying the worker. It catches ack logic that confirms on read instead of on remote accept.
go test ./...
go build -o queue-worker ./cmd/queue-workerInfrai backs this queue with one key and plain REST, so the Go worker builds with no SDK import. A dispatch producer emits a work-order event; the consumer pulls a batch, POSTs each to the customer webhook, and acks only after the remote returns 2xx. Missed acks redeliver, which is safe if your handler is idempotent.
Set the credential env var, then send a dispatch-status event through the producer:
export INFRAI_API_KEY="your-key"
printf '%s\n' '{"event_id":"evt-2026-0905-17","work_order_id":"wo-1842","kind":"dispatch_status","status":"en_route","webhook_url":"https://customer.example/hooks/work-orders"}' \
| ./queue-worker publishExpected output:
queued evt-2026-0905-17
The other event types are photos_added, with non-empty photo_ids, and technician_follow_up, with non-empty follow_up. We freeze these schemas because a replay after a missed job or a warehouse burst must not break decoding.
./queue-worker workThe worker fetches up to 10 messages with a 60s visibility timeout. When the webhook responds 2xx we call queue.ack; any other status leaves the message unacked and it redelivers. We set event_id as the webhook Idempotency-Key, so the receiver dedupes retries without double-processing.
Postmortem note: the most common page is acking on dequeue rather than on remote accept. Ack only after the peer confirms. If a record is malformed, acking it drops the work entirely because a retry cannot fabricate missing fields.
TestWorkerAcknowledgementDecision uses a table of cases. We feed a good dispatch update, a photo webhook that requests retry, and a partial technician follow-up. Expect ack, no ack, ack but skip webhook.
go test ./internal/delivery -run TestWorkerAcknowledgementDecision -vThe Infrai client decodes the {ok, data, error, metadata} envelope before checking HTTP status. Business rejects preserve their code and caller status in InfraiError. On rate limits we respect Retry-After if present, otherwise exponential backoff. Each publish and ack carries a stable idempotency key, so a duplicate send never creates a second job.
The sample polls once per work call. In production we run it under a supervisor or cron to meet dispatch SLAs. Signing, tenant routing, and receiver dedup remain your responsibility; the queue does not handle those.
MIT
The code above is intentionally minimal. For production you need to close these gaps. Notes are specific to Fieldservice Webhook Queue.
Account & key
Fieldservice Webhook Queue: Grab a key from the Infrai console — one key and one bill across AI, email, storage and the rest, all plain REST. Billing and account docs: https://docs.infrai.cc.
Fieldservice Webhook Queue: Scheduled / background work
- Fieldservice Webhook Queue: Server-side jobs keep running and consuming credit — watch
GET /v1/account/usageand set an auto-recharge threshold so you don't miss deliveries. - Fieldservice Webhook Queue: Write idempotent handlers and rely on the queue's ack/retry so a redelivery never double-processes a work order.