Gemini real-time

[Shandelier]

Any plans for implementing Gemini real-time API to Agents?

[jadams777]

+1 for this

[CrackheadCode]

+1

[davidzhao]

we are working on it. stay tuned!

[pixelcatgg]

+1 and thank you @davidzhao looking forward to it. something interesting to know is it seems they have a limit of

3 concurrent sessions per API key

is there guidance of how to queue up or rotate api keys? the rates seem pretty low for an actual application

[mpisat]

If everyone enters their own api key for demo, it should work fine. Just like openai key requirement at the demo site.

[jadams777]

Which demo?

[mpisat]

Which demo?

https://playground.livekit.io/
which is outdated I think?

[OrkhanGG]

+1 and thank you @davidzhao looking forward to it. something interesting to know is it seems they have a limit of

3 concurrent sessions per API key

is there guidance of how to queue up or rotate api keys? the rates seem pretty low for an actual application

There might be some workarounds to rotate the keys and bypass API usage limits, but I’m not sure if that complies with Google's policies. Since this is an experimental model, I strongly recommend against using it in production for now. The rates will likely increase once the stable version is released.

[smxthereisonlyone]

Any update here and when it will support video input?

[arthurblake]

Gemini 2 looks EXTREMELY impressive, and our early tests are showing that it's way smarter than OpenAI in our particular business domain. I've been asked to shift over to Gemini MultiModal for my livekit development. So I'm really looking forward to this support...

[laag-deploy]

+1, looking forward to it

[sajithlaldev]

+1, looking forward to it