Change 60.000 ms limit

Hi!

I’m using Gemini a lot to says things on my google speakers, works great but with a little issue.
As soon as it’s a bit longer reply, like telling the weather and stuff it times out.
I doubt it’s more than 60 seconds tbh but it still fails with that error.
In this case it’s Start Flow Regnvarning (RainWarning)

If I put the Start Flow before the “Set Regnvarning Skickad (rainwarning sent) to Yes” it won’t set it to Yes due to the time limit.

This is the flow for generating the text.

The instructions to generate the text is
Act as a voice assistant in a smart home. It has started to rain outside according to a Netatmo rain gauge: Rain now You should tell the residents that it has started to rain, how much it is raining right now and that they need to cover outdoor furniture and put it away outside. Download data from Weatherdata YR.NO JSON and tell them what the forecast is for rain here in Bålsta and when it might stop raining. Give me ONLY the above in the answer, without introduction, quotation marks or follow-up questions. Answer directly with the text to be read out.

Can i change this time limit somewhere?
It’s not an overly complex flow but the creation of the speech does take 20 sec.

@Thyg0d two things that should help:

  1. Keep the slow step away from the rest. Set “Regnvarning skickad” to Yes first and start the speech Flow after it, so the variable never depends on how long the model takes. Your prompt also asks Gemini to download YR.NO data itself, which is probably what makes it slow: passing the forecast in as text (or dropping that part) makes the answer much faster.

  2. Give the AI step its own time limit. I made an app for this, Cortex AI (so I’m biased): its “Ask Cortex a question” card has a “wait at most” of 5–50 s and a fallback text, so the Flow always continues, worst case with “It has started to rain”. It works with Gemini too. And the “brain” you ask can see your Netatmo rain gauge and Homey’s weather forecast for the next hours by itself, so the prompt can be just “Tell us it started to rain, how hard, and when it will stop. Only the text to read out.” You send the Answer token to your Google speakers as now.

App topic: [APP][Pro] Cortex AI – an AI brain for your Flows and your home

Thanks!

Yeah I solved it by getting the data more often from the API and storing the JSON output to be able o query it.

Regarding the timing I changed the whole setup of the ask to Gemini to shorten it.
More or less, it will rain soon and stop at xxyy. get off you lazy ass.. :smiley: