Exponential Backoff
Retries should spread over increasing windows with jitter so partner outages do not create traffic spikes.
Webhook Guide
Preflix AI webhooks support asynchronous partner workflows: audit milestones, report availability, letter drafts, and dispute status updates.
Examples are synthetic. Production event delivery requires partner setup, signing secrets, and privacy review.
Setup
Signed events should be verified, queued, and processed idempotently so partner workflows stay reliable during retries or outages.
Use a server-side HTTPS endpoint dedicated to Preflix AI events and avoid sharing it with unrelated product webhooks.
Subscribe only to the events your workflow needs, such as audit completion or dispute status changes.
Recompute the HMAC with the timestamp and raw payload before parsing or queueing any work.
Acknowledge valid events with a fast 2xx response, then process longer work asynchronously.
Event Catalog
Every event should map to a specific partner action, such as fetching a report, updating a case, or notifying an advocate.
| Event | When It Fires | Partner Action |
|---|---|---|
bill.created | A partner bill record is accepted. | Create or update the local case shell. |
bill.document_ready | Required document references are attached. | Show that audit preparation can begin. |
audit.started | Preflix AI begins review. | Move the partner case to active analysis. |
audit.completed | A structured audit report is available. | Fetch the report and notify the user or advocate. |
letter.ready | A requested letter draft is prepared. | Present the draft for review before sending. |
dispute.status_changed | A dispute workflow is updated. | Sync status, deadline, and follow-up tasks. |
Signed Payload
Preflix AI should send timestamped signatures in headers so receivers can reject tampered, stale, or replayed events.
preflixai-timestamp
Unix timestamp used in the signature base string.
preflixai-signature
HMAC digest computed with the endpoint signing secret.
audit.completed{
"id": "evt_7b2",
"type": "audit.completed",
"created_at": "2026-01-15T19:04:12Z",
"data": {
"bill_id": "bill_8f4",
"report_id": "report_33",
"status": "complete"
}
}Retry And Failure Logic
Partners should expect at-least-once delivery and build receivers that tolerate duplicate events, temporary outages, and delayed retries.
Retries should spread over increasing windows with jitter so partner outages do not create traffic spikes.
Store event IDs before work starts so duplicate deliveries do not create duplicate reports, tasks, or notifications.
Repeated failures should land in an operations queue with endpoint, event type, response code, and retry count.
Testing Checklist
A webhook receiver is only ready when security checks, retry handling, logging hygiene, and replay behavior have all been exercised.
Integrate Events With The API