I see similar behavior in AI Studio in the “stream realtime” tab > “Talk to Gemini” . It can hold a conversation for about a minute and then I get an error: “something went wrong” and the convo ends.
Even we are facing this issue.
Till friday it was working fine. and suddenly over weekend it started giving internal 500 error from Gemini Live Service through bidicontent api endpoint.
It is really impacting the working solution no changes were made to the code.
Any solution or alternative for this?
was it easy lift and shift? which framework were you using?
as the session.updates events & params,etc are to be configured on OAI Realtime @Ajitesh_Abhishek
I mean I hear your frustration but to be fair it literally has “experimental” in the model name..hah..they don’t even call it alpha or beta.. it’s super far from GA. I probably shouldn’t have put it in my production app but it was too good not to try. I’m hopeful they’ll fix it and keep making it more stable, but yeah for now I’m also exploring gpt4o-realtime.
I use the model gemini-2.0-flash-exp-image-generation to generate images. It was working fine yesterday. Now the response code is always 503 and 500.
When editing images, although the response code is 200 and the text message is generated normally, there is no “.inlineData” property in “content.parts”
Unable to generate or edit images at this time. Is Gemini crashed?
Hey @abalabala, Since it’s a experimental model, it can sometimes overloaded hence you are seeing 503 and 500 error. Gemini doesn’t crashed. It’s an intermittent issue. It will resolve by itself in sometime.