Hi everyone,
I’m building a personal voice assistant using the Gemini Live API (google-genai SDK, client.aio.live.connect()). It was working fine for several days.
What happened:
Due to a bug in my own reconnect logic, my app went into a rapid reconnect loop and made roughly 6,000+ connection attempts over a couple of hours (peaking at 40+ attempts per minute) after hitting a rate limit. This started producing:
1011 None. You exceeded your current quota, please check your plan and billing details.
I checked the AI Studio Usage and Rate Limit dashboards, and my actual RPM/TPM/RPD were all well within the free-tier limits shown there — so this didn’t look like a genuine quota breach, more like a soft throttle.
About a day later, the error changed to:
1008 None. Your project has been denied access. Please contact support.
This has now persisted for more than 2 days (last successful connection: Sep 24, 9:11 PM). It has NOT gone away with:
- A brand-new API key in the same project
- A completely new Google Cloud project with a fresh key
- An API key generated from a different Google account entirely
- Different Live-capable models (
gemini-3.1-flash-live-preview,gemini-3.8-live) — confirmed viaclient.models.list()that both are listed as available withbidiGenerateContentsupport for my key, but connecting to either gives the same “denied access” error
Since it reproduces across accounts, projects, and models, it doesn’t look like an account or model-specific issue — I suspect it may be tied to my IP/network, or some backend flag triggered by the reconnect burst that hasn’t cleared.
Questions:
- Has anyone else seen this specific “1008 denied access” error after a burst of rapid reconnects?
- Is this something that self-resolves, or does it need manual intervention from the Gemini API team?
- Is there a better channel to get this looked at (e.g., a support ticket, or does enabling billing bypass this kind of restriction)?
Happy to share more logs/details if useful. Thanks in advance.