Skip to content
Puzzak edited this page Nov 22, 2025 · 2 revisions

What is this?

  1. Is it official?
    • No. The application is developed by a third-party developer. I am not affiliated (unfortunately) with Google or DeepMind.
  2. Does Gemini run in the app?
    • No. The app is just an interface for the Google AI Core, which runs Gemini Nano in sandbox environment on your phone.
  3. What does the app do then?
    • Basically it "talks with AI Core" on your behalf. What it does is:
      1. Prepares the data for AI (compiles your request with instructions, resources and context data (if you enabled it)
      2. Establishes connection with AI Core (sends compiled data, receives generated content or content stream)
      3. Stores and displays generated content (saves conversation history and displays it in the app's chat)
  4. Do I need the app to chat with Gemini?
    • Maybe? If you want to chat with Gemini Flash or Pro you will need to use the official Google's Gemini app. As of moment of development, no other app does use AI Core. You can compile your own app though, using this as a base or create a new using the original plugin.

How does it work?

  1. Does this app work offline?
    • Yes. Both app and AI Core don't need internet to work. The app will use internet to refresh translations and prompt data but it will work just fine without this - and AI Core doesn't need internet to generate content, but it will update models from time to time in the background.
  2. Does the AI support photos or other media?
    • Yes, Gemini Nano is multimodal, which means it does have ability to "see" and maybe generate images.
    • No, the app (1.0.1) does not support anything but text. It may change one day, but no promises here.
  3. Are my chats safe?
  4. Does the app manage Gemini or AI Core?
    • No, the app can only request AI Core to download Gemini Nano if it is not downloaded (usually this will happen on the first launch or after AI Core data is deleted). The app doesn't store Gemini and can only send your request and receive the response, everything else (including censorship of outputs) is controlled by AI Core.

Clone this wiki locally