On 2 October, Phaseo experienced widespread disruption across the website and API. Customers encountered slow pages, pages failing to load, and unreliable API access.

We identified excessive database load, primarily from scheduled jobs that aggregate usage and analytics data. These jobs competed with the database resources needed to serve customer requests.

We reduced expensive background work and improved database queries and reporting jobs. Service initially recovered on 3 October, but intermittent problems continued through 9 October. We then increased database compute capacity, after which the website and API returned to normal operation based on our checks.

We have also added automatic availability checks for the homepage, Models page, and documentation. Confirmed failures now alert us through incident.io and publish an initial Investigating update on our status page.

We’re sorry for the disruption and the uncertainty during recovery. We’ll continue monitoring service stability and reviewing background workloads to reduce the risk of recurrence.