Bloxly

AI Chat is fully operational again. The issue was caused by a routine infrastructure configuration update on our end that temporarily interrupted connectivity to the AI provider. The fix has been applied and all services are running normally.

The earlier fix resolved the 500 errors on stream startup, but we are now seeing elevated error rates from the LLM provider.

Fix is deployed. Added Flush() to the metrics response writer wrapper.
Monitoring logs and error rates for the next 10 minutes to confirm stability.

Root cause found. The metrics middleware wraps http.ResponseWriter in a custom struct to track Prometheus metrics. That struct was missing a Flush() method.

We are actively developing a fix to this issue.