Elevated errors when using Google Gemini models with Voice Agent

Updates timeline

1. Investigating Jun 17, 07:17 PM UTC

   We are currently investigating elevated error rates affecting Voice Agent requests that use Google Gemini models. During this period, some requests to Google models (including gemini-2.5-flash-lite) may fail or time out. Other Voice Agent functionality and non-Google models are not affected. Our team is actively working to identify the cause and restore full service. We will provide an update shortly. To avoid downtime, please define multiple LLM providers (https://developers.deepgram.com/docs/voice-agent-llm-models#using-multiple-llm-providers) in your Voice Agent configuration.

2. Identified Jun 17, 08:54 PM UTC

   We have identified the cause of elevated error rates affecting Voice Agent requests that use Google Gemini models. A recent change on Google's side to how API keys operate is causing requests to fail, and our team is rolling out updated credentials to resolve it. Until the update is complete, customers may see intermittent failures for Gemini-model requests — concentrated in the first half of each hour, with most requests succeeding in the second half. Other Voice Agent models are unaffected. We are actively working on the fix and will update when resolved.

3. Identified Jun 23, 03:09 PM UTC

   Continuing to monitor; root cause confirmed as the Google API key change.

4. Resolved Jun 23, 03:09 PM UTC

   This incident has been resolved.