About the Maathai_llama Plugin

Flutter plugin wrapping llama.cpp to enable fully offline, on-device chat inference powering Maath:ai.

View Repository Installation

What it is

The Maathai_llama plugin is a Flutter plugin that integrates llama.cpp via native bindings to run large language models entirely on-device. Maath:ai uses this plugin to deliver private, offline conversations with no server dependency.

  • Offline inference on Android and iOS (where supported)
  • Bindings to optimized llama.cpp backends
  • Configurable model selection and context parameters
  • Zero data leaves the device for chat processing

Install and Build

These steps mirror the repository instructions.

  1. Clone and fetch submodules: 
    git clone https://github.com/kipkirui/Maathai_llama
cd Maathai_llama
git submodule update --init --recursive
  2. Install Android build chain prerequisites:
    • Android NDK 26.1.10909125
    • CMake 3.22.1
  3. Fetch Flutter deps and run the example app: 
    cd maathai_llamma
flutter pub get
cd example
flutter run --device-id <device>

See the repo for updates and platform specifics.

Privacy

Inference is performed locally; prompts and responses remain on the device. This aligns with Maath:ai’s edge privacy model.

Resources