adk_litertlm

English | 한국어 | 日本語 | 中文

LiteRT-LM model integration for ADK Dart packages.

This package provides a LiteRtModel implementation that connects LiteRT-LM on-device models to the ADK Dart agent runtime, enabling fully offline, on-device AI agent execution.

Features

  • On-device inference: Run agents entirely on-device using LiteRT-LM models.
  • ADK runtime integration: Drop-in BaseLlm implementation compatible with LlmAgent and the full ADK pipeline.
  • Streaming support: Real-time token-by-token streaming responses.

Installation

dart pub add adk_litertlm

Or add to your pubspec.yaml:

dependencies:
  adk_litertlm: ^2026.8.17+1

Getting Started

import 'package:adk_litertlm/adk_litertlm.dart';

final model = LiteRtModel(modelPath: 'path/to/model.tflite');
final agent = LlmAgent(name: 'local-agent', model: model);

ADK 2.0 Compatibility

This package is aligned with ADK 2.0, supporting v2 Workflows (declarative node-graph scheduling) and Managed Agent integration for hybrid on-device + cloud execution.

Libraries

adk_litertlm
LiteRT-LM model integration for ADK.