Skip to main content
The RunAnywhere Flutter SDK runs language, speech, and vision models on the device. Prompts, audio, and images stay on the phone, and everything keeps working offline.

Requirements

Run on a physical device. The Android emulator has no arm64 native libraries for these backends, and Apple MLX only executes on physical iOS hardware.

Packages

runanywhere_qhexrt publishes at 0.20.19 on pub.dev. Four of the five move together; QHexRT does not.

The namespaces

Every capability is a static getter on RunAnywhere. hybrid exists only on Flutter. Streaming is Stream<T>; everything else returns a Future.

Differences worth knowing up front

Dart has no method overloading, so Flutter’s surface diverges from the other SDKs in ways that break copied code:
  • Four generation methods, not two: generate, generateChat, generateStream, generateChatStream.
  • generateStructured takes the schema as a String, not an object.
  • ingestAll(documents) for a list, where the others overload ingest.
  • RunAnywhere.deviceId throws before initialize().
  • No AudioInput.file and no ImageInput.rawRgba.

Where to go next

Installation

pubspec and platform setup

Quick Start

A working chat

Text generation

Prompts and options

Voice agent

Speech in, speech out