● Start
Requirements.
What you need to build and what phones it runs on.
The app compiles a C++ AI engine, so it needs the native toolchains of both platforms.
01To build
| Tool | Version | For |
|---|---|---|
| Flutter / Dart | 3.41+ / 3.11+ | Both platforms |
| Xcode + CocoaPods | Xcode 16 or newer (tested with Xcode 27.1), CocoaPods 1.16 | iOS (macOS with Apple silicon recommended) |
| CMake | 3.22.1 (Android SDK) and 3.22+ on the Mac for the iOS build | Compiling llama.cpp and whisper.cpp |
| Android SDK / NDK | compileSdk 36, NDK 28.2.13676358 | Android |
| Node.js | 22+ | Only for the optional console |
02Phones
| Minimum | Notes | |
|---|---|---|
| Android | 8.0 (API 26), 64-bit ARM (arm64-v8a) | Runs on the CPU. |
| iOS | 16.0, any iPhone that runs it | Uses Metal on the phone. In the iOS simulator the engine always uses the CPU (the simulator's Metal gives wrong results for these kernels). |
| Memory | 4 GB RAM; 6 GB for the 1.7B models | The app refuses a download the phone cannot run and says why. |
| Storage | The model size + 10 % | 0.06 GB (voice) to 1.3 GB (largest default chat model). |
How it was tested
The item was tested on the iOS simulator (iPhone 17, iOS 26) and the Android emulator (arm64, API 35), and the native engine on the build Mac. Speed depends heavily on the phone; this documentation gives no speed figures.