Webhook Delivery Reliability — Retry + Acknowledgment
I'm running Omi with a custom webhook integration (OpenClaw AI assistant framework on a Mac Mini). Transcripts stream through a local proxy that routes them into a persistent AI agent with memory, home automation, and ambient awareness.
The problem: there's no retry mechanism or delivery acknowledgment for webhooks. If my gateway is briefly down (reboot, deploy, crash), transcripts are silently lost forever. There's no way to know something was missed.
What I'd like:
- Retry with backoff when the webhook endpoint returns non-200 or is unreachable
- A short buffer (even 5-10 min) of unacknowledged payloads
- Delivery receipt / acknowledgment protocol so the sender knows the payload was received
- Optionally, a dead-letter queue or "missed webhooks" endpoint to retrieve what was lost
This is the biggest reliability gap for anyone building serious integrations on top of Omi. The device is great at capturing — but if the pipe leaks, the data is gone.
Log in to comment and vote
Comments1
Aryan
Oct 4
Hi @samshields-oc , your concern is valid: a brief outage shouldn’t silently leave an integration with gaps.
Since you posted, we’ve merged bounded retries for failed developer-webhook requests and internal retention for some deliveries that exhaust those retries. That is progress, but it isn’t yet the delivery history and self-service replay flow you asked for, so we’re keeping this request open.
Are you using the real-time transcript webhook, the completed-conversation webhook, or both? If you’re still seeing missed events, an approximate outage time and the event type would help us investigate the current behavior without sharing transcript contents publicly.