This is the reason: if you are using free tier API keys, the only models light enough not to burn all your daily tokens with a smart home command are the 2.5 flash and the 2.5 flash-lite.
Try this: go to the app settings and in Configuration disable Google Search Grounding. Unfortunately, the 2.5 models don’t allow you to manage search and smart home management at the same time.
Okay but how can I make an advance flow? I can’t figure it out
The app has an action card that receives a command for your smart home (including creating an advanced flow), executes it, and returns Gemini’s response as a tag (for example, “I created an advanced flow with trigger xxxxx …”).
How you send the command to the action card is up to you: you can use the Telegram app or Homey’s free-text notification. Here’s an example of a flow using Telegram:
Alternatively, you can add the chat widget to the app’s home page and send commands to Gemini with it:
Remember, however, that Gemini’s free API keys are very limited and don’t allow for complex tasks like creating an advanced flow. For those, you must have an active billing account and use a more powerful model (at least 3.1).
I get:
Error sending Gemini prompt: {“error”:{“code”:404,“message”:“This model models/gemini-2.5-flash is no longer available to new users. Please update your code to use a newer model for the latest features and improvements.”,“status”:“NOT_FOUND”}}
What should I use to use the free tier?
Moreover I don’t see any usage at all in my dashboard:
Google will shutdown all the 2.5 models in october 2026 (source) and has therefore blocked the use of these models for the new APIs (those who were already using them will be able to continue doing so until October).
Unfortunately, they are also the least powerful models, so the smart home controls must be rather basic.
I have released the new version with support for Gemini 3.6 Flash and Gemini 3.5 Flash-Lite. From initial tests, Gemini 3.5 Flash-Lite model seems to be a good cost/performance compromise in managing advanced flows.
What about “Send a multimodal prompt to Gemini (text + image)” ? Do they work with 3.5 Flash-lite?
ciao, the 3.5-flash-lite works with no problem with the card prompt+image. Since you’re using the free tier, you should probably try the 3.1-flash-lite, which costs even less.
Hello! First of all, thank you for this amazing app.
I’ve just installed it on my Homey Pro to monitor my smart home devices (Somfy, RuuviTag, etc.), but I’m getting a “You have exceeded your daily quota” error on my very first prompt using Gemini Flash models.
I’ve investigated the issue and it seems related to a recent Google change. Google is no longer generating API keys starting with AIza.... All new API keys generated from Google AI Studio now start with the prefix AQ.
It looks like the app might be having trouble parsing or authenticating this new AQ. format, causing Google to reject the connection and triggering the generic quota error.
Could you please take a look and update the app to support the new Google API key format? Thank you so much!
Ciao @Nicolas_Crocco,
pleas send me a diagnostic report so I’ll take a look.
Hi Simone,
Diagnostic code
e6b8018f-0e24-4588-98a8-99bec0981abc
Thanks
El El mar, 11 ago 2026 a las 16:57, Simone Di Maio <notifications@athom.discoursemail.com> escribió:
Ciao, the problem is that Google’s free tier quota, which wasn’t particularly generous before, is effectively unusable with the new API Keys. Add to this the fact that the new API Keys can no longer use the less expensive 2.x family models, and the result is that with Gemini’s new Free APIs, the quota is no longer sufficient even for a simple conversational card like “what’s the weather going to be like tomorrow.”
The only way to use Gemini is to have an API Key with pay-as-you-go billing.
However, I recently discovered that all Google AI Pro subscribers have a $10 monthly credit that can also be used for Gemini billing (depending on your app usage, monthly usage can vary from a few cents to 1/2 euro).
In any case, I’ve released version 3.3.7 of the app for testing with support for the new 3.7 Flash model (which is particularly high-performance for creating and managing flows) and added more explanatory errors when the usage limits are reached.
