My team has been using models/gemini-2.0-flash-exp with the multimodal live websocket. We are streaming audio from our frontend client to the gemini websocket (wss://generativelanguage.googleapis.com/ws/google.ai.generativelanguage.v1beta.GenerativeService.BidiGenerateContent). Today, without making any changes to our codebase, the models have begun freezing when attempting to make tool calls. This issue is apparent in Google AI Studio as well for the multimodal models with tool use.
The google search built in tools is working fine, but any custom tools we provide don’t get called.
Are there any recent changes that have broken this? Is there an ETA on a fix?
Please let me know if you require more information.
Just signed up here to say - yes, facing the same issue. I asked a moderator on Discord and this is what they had to say:
Since I don’t work much with live APIs, I can’t confirm if there’s an issue. However, given that many people are experiencing the problem in discussion you posted, it is likely real.
Unfortunately, it’s hard to provide a specific ETA for these fixes. I’m sure that someone in this area will make every effort to resolve it as soon as possible since this functionality is critical for many applications, so try again in few hours to check if its resolved. Sorry for any inconvenience!
Here’s to hoping that it gets fixed in a few hours.
I have tried to reproduce this issue, by making a simple tool both programmatically and in AI studio (function calling) with live API. I cant seem to reproduce this issue now,
are you guys still facing this issue ??
I hope they fix it soon.