Is there a way to clear cached content from a Google Generative AI API session?
I set up three configurations for Google Generative AI on HA using the same API key. One to rewrite announcements for style and variety. A second to use with VPE as a voice assistant with internet search. And a third to use with VPE for function calling to control devices with assist.
The announcement styling worked great until I set up the others for my new VPE. I also switched to Gemini Flash 2.5 preview near the same time. Now, the announcements seem to be re-written as if their intent was function calling (“I can’t find a device called, ‘Good Morning’…”), even though the prompt is different and unique.
I wonder if the cached conversation for Google Generative AI is not being cleared. Does this sound like my problem?
I have tried to return to my previous model (Gemini Flash 2.0) for the announcement workload without improvement.
For context, this problem emerged around May 8th.
Implicit caching is enabled by default for all Gemini 2.5 models. We automatically pass on cost savings if your request hits caches. There is nothing you need to do in order to enable this. It is effective as of May 8th, 2025. The minimum input token count for context caching is 1,024 for 2.5 Flash and 2,048 for 2.5 Pro.
1 post - 1 participant