Gladia - Notice history

A change in our method to scale our available real-time sessions workers prevent those to scale-up, resulting in a lowered capacity and prevented new session from being created.

We updated the needed metrics to get back to our nominal behavior.

Following a recent update to our API specification and documentation, an unexpected cascading impact affected the callback request body for transcription jobs.

While we initially believed the issue had been fully resolved yesterday evening, we’ve since identified that some callbacks continued to fail for certain requests.

Impact period (Europe/Paris time):

• December 10 — from 19:00 to 22:34

• December 11 — from 23:10 to 11:46

During these windows, transcriptions were correctly processed, but the corresponding callbacks were not delivered.

Our team is now resending all impacted callbacks to ensure you receive the expected results.

Importantly, this callback issue only affected customers still using the deprecated callback_url parameter, and did not impact requests using the new callback_config format.

We’ve already identified part of the root cause and are actively monitoring the system to guarantee a permanent fix.

We sincerely apologize for the inconvenience this may have caused and deeply appreciate your patience and understanding as we work to fully stabilize the service.

We are currently not scaling capacity fast enough to meet demand. Performance may continue to be impacted. . Our team is actively adding capacity and monitoring the system closely.

We experienced a temporary API outage caused by the global Cloudflare incident impacting internet routing. A secondary issue in our edge stack prolonged the disruption. Service was restored once the secondary issue fixed.

We experienced degraded batch-processing performance due to an internal overload on our infrastructure. Service performance progressively recovered after rebalancing compute capacity and clearing queues. No data was lost.

We've put a fallback in place and diarization is working as intended. We're monitoring until our main providers get back on track to close the incident.

The issue has been mitigated and normal performance has been restored. All real-time connections are now operating as expected.

Between 03:30 and 11:30 UTC, our real-time (RT) WebSocket service experienced intermittent connection issues due to an infrastructure scaling problem. A temporary workaround was deployed to stabilize performance while the underlying issue was identified and resolved. The root cause was a change in our cloud provider’s API that affected our autoscaling logic. The issue has since been fully resolved, and we have implemented safeguards to prevent recurrence. We sincerely apologize for the disruption and appreciate your patience.

A fix has been deployed. This incident has been resolved.

A workaround is in place to allow more capacity until we fix the root cause.

We have an issue scaling ressources to handle Real Time requests.