My Agents are using Gemini but getting this error. Looks like there was an incident raised and marked as resolved yesterday but it's still a problem today. https://cdn.discordapp.com/attachments/1463839912007569520/1463839912188055756/image.png?ex=6979e227&is=697890a7&hm=3b6372447e3b9e641bf6aad05c4de166d3e8db4ed4f97c19805cdcbccf264e4e& https://cdn.discordapp.com/attachments/1463839912007569520/1463839912670265517/image.png?ex=6979e228&is=697890a8&hm=eff69de37aa7fb8bd435f105ffe99d453211cfbcbed0787d9612efc211880b56&

Actionable Suggestion: To work around the ongoing Gemini provider errors, update your agent configuration to use a different model, such as GPT-4.1. For example:

json
"model": {
  "provider": "openai",
  "model": "gpt-4.1"
}

This should help your agents avoid the intermittent Gemini failures while the provider issue persists. See more details in the [call ended reason documentation](https://docs.vapi.ai/calls/call-ended-reason). Source: - [Call ended reason documentation](https://docs.vapi.ai/calls/call-ended-reason)

Thanks for the details. This behavior is consistent with an ongoing intermittent issue on Google/Gemini infrastructure, which can affect several parts of the pipeline. Here’s a quick breakdown and the available workarounds: LLM (Gemini models): Gemini models (including Flash 2.5) may intermittently fail. Workarounds: • Switch to a different model (e.g. GPT‑4.1) • Or bring your own Google API key via Dashboard → Settings → Integrations → Custom Credentials Transcription (STT): Gemini‑based transcription may intermittently fail. Workarounds: • Switch primary/fallback transcriber to another provider (e.g. Deepgram Nova 3) • Or use your own Google API key Voicemail Detection: May intermittently fail if the provider is set to google. Workarounds: • Switch provider to openai or twilio (if using Twilio telephony) • Or disable voicemail detection and use the Voicemail tool instead Query Tool: May intermittently fail since it relies on Google infrastructure. Workarounds (higher effort): • Use a custom knowledge base, or • Replicate the behavior via a function tool Structured Outputs: Gemini models may intermittently fail. Workaround: • Switch to another provider such as OpenAI or Anthropic Speech‑to‑Speech: Gemini speech‑to‑speech may intermittently fail. Workaround: • Switch to a different provider (e.g. OpenAI) We understand this is disruptive and are continuing to monitor the situation. In the meantime, the above workarounds should help keep things stable. Let us know if you’d like help switching any of these configurations.

Hi team! Alex here. I'm trying to follow the workaround mentioned here which says we can bring and use our own Gemini key. But my question about the path mentioned, there is no exist given path:

Dashboard → Settings → Integrations → Custom Credentials

I belive the documentation is kind of update. Which place will be the correct one? https://cdn.discordapp.com/attachments/1463839912007569520/1464003761965105163/image.png?ex=6973e340&is=697291c0&hm=37287f4a76285827fc07af32d1aa562190b3f7a2895af3072abc05af287b1375&

Message marked as helpful by @Alex! 🎉

@User ^

I tried gemini and had so many weird issues. Eventually when I switched to openai I had to redo my prompts because the overengineered prompting for gemini was redundant. OpenAI is very reliable

Hey! Thanks for reporting this. We experienced rate limiting from Google's Vertex AI API that affected Gemini models, transcription, voicemail detection, and related features. You can track the full details here: https://status.vapi.ai/incident/809181 The issue has been resolved on Google's end. If you're still experiencing problems with Gemini, could you try again and let us know? If issues persist, switching to an alternative model (like GPT-4 or Claude) or using your own Google API key are options that can help avoid shared rate limits. Let me know if you run into anything else!

Status page still says it is down BTW. But I'm testing it right now.

You're correct! To add your own Google API key for Gemini, go to Org Settings → Integrations in your Vapi dashboard, select the Google integration, and enter your API key in the provided field. This section is designed for secure credential management and supports Bearer Token authentication for Google services. After saving, your agents will use your key for Gemini requests, helping avoid shared rate limits. For more details, see the [August 30, 2025 changelog](https://docs.vapi.ai/changelog/2025/8/30). Source: - [August 30, 2025 Changelog](https://docs.vapi.ai/changelog/2025/8/30)

This is the error I'm getting. I'm using my own API Key as mentioned above. https://cdn.discordapp.com/attachments/1463839912007569520/1464048320879263867/image.png?ex=69740cc0&is=6972bb40&hm=58bad4fa0b1c7770c563e2ae2a70b41fa6d15cc2ad544bbad47c8431e24411c1&

Hi Alex, could you please provide us with the Call ID?