We have been checking every 5 minutes for 49 days. Earlier days are grey because we have no data, not because anything failed. A problem is published once two consecutive checks confirm it.
Incident History
Knowledge and search degraded by AI provider latency
The knowledge service health check reported degraded performance in two windows because the OpenAI API, which we use for embeddings, took longer than 3 seconds to answer. Searches that needed new embeddings could have been slower during those windows. It recovered on its own once the provider's latency went back to normal.
- Health checks start reporting degraded: the OpenAI check hits its 3 s timeout.
- Back to normal.
- Degraded again, for the same reason.
- Provider latency back to normal; resolved.
Intermittent failures reaching several services
Our external checks failed intermittently for the agent platform, knowledge and search, the branch avatar and the dashboard for about four hours. In the same window the platform API returned 25 server errors and some responses took up to 7 seconds, while traffic kept arriving at its usual volume. The root cause was not confirmed: services that share no infrastructure (the dashboard is served by a CDN) failed at the same time, which points to the network path between our checks and the cloud provider.
- First failed checks and first server errors on the platform API.
- Most errors of the day (13 in 15 minutes) while checks keep failing intermittently.
- Last errors; checks back to normal.