I’ve also been hit with this and submitted an appeal - I’m using the integration to create short descriptive notification messages for security camera images.
Ironically I was planning on configuring an override switch to disable the GenAI based notification content anyway, which feels like it could be useful in this scenario (even though my automations fallback to some default notification text if nothing’s returned from the GenAI call.
I have the same issue.
Error generating content: 429 Resource has been exhausted (e.g. check quota)
I checked on Google AI site and the limit never reached. Im on 30% only and Im doing only 4-5 use / day.
That is something different I think so, its some issue with the integration or with the API.
Did you found any solution?
Just to update this… my PAID APi key is getting banned every few days for some kind of violation…
Sometimes i appeal it… other times it gets re-instated without me having to…
truly frustrating but it’s not essential to my security system so i can tolerate it for now… but the fact im paying for it now means i dont seem to be able to go back to a free tier which is VERy annoying.
Kind of good to see that I’m not alone with this issue here 
I’ve had the same message coming in weekly now - the first time I appealed; the 2nd and 3rd time the ban was lifted before I even saw the message coming in. Still annoying, though, that the service is down for a few hours every.
I noticed that the emails were referencing a "quota_location" value: "us-central2" which doesn’t even exist when I go to check on the quota usage under the header Quotas & System Limits for project "Gemini API".
And I don’t get any warnings so I could appeal… And already created new project/tenant and here I am with the same 429 and no errors shown. With free tier it’s not possible to get to support. Maybe will try another google account. If anyone has any ideas - please share.
About every 4 or 5 days I get the error and have to appeal. I ask every time if there is anything I can do to prevent it, but they just accept the appeal and reactivate. It’s really annoying to have a submit an appeal at least once a week.
Google returns 429 on many different errors, the quota thing is a red herring and rarely true if you are on a paid plan and only using it for things like this. They have “throughput subscription plans” and other stuff you can buy to be first in line to use their resources. So, if your request comes in at a time where premium clients are also requesting, they get the resource, you get a 429.
There are several other things that throw code 429 as well. Truth is, multi-million/billion dollar companies are blasting gemini and hogging resources which ends up with us getting 429s. Google doesnt need my $1.00 monthly gemini API revenue, lol.
Wondering if switching to “Paid tier” helps. Has anyone tried? Also, how much is the paid tier?
see further up:
Paid tier just means your requests are slightly more important than free but still far below multi million/billion $ companies requests.
I was seeing the same error on the free tier but had no indications of going over the quota on their web site. Turns out I was making 2 consecutive calls which triggered the error. I had to place a 10 second delay (9 seconds wasn’t enough) between the 2 calls and haven’t had any of the 429 errors since.
That’s the problem though, we don’t have enough info to properly diagnose the root cause of 429. In your specific case, at that time, that was the issue. Maybe tmrw it’s because their doing maintenance, or maybe it’s because some whale is hammering their server farms.
Would sure be nice to have an error code with substance or, heaven forbid, a JSON message with a human readable error message.
I have a few automations. But it’s always the analyzing of the camera snapshots that gives me a 429 Error. My success rate is about 60%. But my other automation/scripts ask for a simple response if the garage door is not closed or the front door is open. I never experience 429 errors, even though they get triggered 5 to 15 times a day. Probably Google API doesn’t like to analyze images at once and complains about a quota exceeded.
I regularly get the 429 error and quite honestly although it works well - when it works, the fact that it errors so often and I can find no easy way to reset it. I will be giving up on it. I will be going back to good old fashioned movement detection on my cameras. When you ask it something, does it matter how long the question is? Is this a factor at all?
I am doing various testing to see why it causes 429 errors. Somehow, Google is complaining that we are exceeding our quota. So, I reduced the number of snapshots from 3 to 2 (1.5 second interval). Then, I shortened the AI prompt message from 5 sentences to 2 short, concise instructions (with the help of ChatGPT). So far, it has worked well for past 24 hours with no 429 errors. Give it a try and see if this helps!
The thing is, we don’t have any idea what we’re doing wrong. So we can’t fix it, just guess and try things. The other problem is, when you reduced your tokens, did that actually make a difference or are you just on a lucky streak and it’ll start throwing 429 tmrw.
Heaven for I’d we expect aulti billion dollar tech company to send a JSON error message with some context.
Fwiw, mine worked for a few months reliably and it’s just recently I’m getting more 429 than I get valid responses. Nothing has changed, 3 images and a text prompt per request, which leads me to believe it’s Google’s end.
As mentioned in this post, it looks like gemini 1.5 and 2.0 actually uses Dynamic Shared Quota. This seems to give a reason for the apparently random 429 quota exceeded messages, when our personal quota is nowhere near any limits.
Resources seem to be shared amongst a region. I don’t believe, from what I’m seeing, it makes any difference if you are on the paid version or not.
I’ll try rog889’s approach to reduce tokens and see if that helps.
Reporting back from my above post:
Reducing the tokens by taking the images submitted from 3 to 2 and simplifying the query substantially (some examples around are unnecessarily long) helps considerably.
I would say that above 95% of transactions now work successfully without an error 429.
I would therefore recommend this approach to all.
I’m getting the same error , “quota exceeded” 60% of the time, other times it works.
Something I have noticed, when i get this I can go into the “google generative AI” and I can talk to the AI , it accepts any query.
IF I simply type " compare images" It responds by stating it cannot compare images.
WHEN THE THING WORKS… I can query it again by typing “compare images” and it would ask me for the images.
I feel like we are stuck in a beta version perhaps?
