adk_dart 0.1.0

Core Dart port of Agent Development Kit (ADK) runtime primitives.

adk_dart

adk_dart is a Dart-first port of the Agent Development Kit (ADK), using ref/adk-python as the primary reference implementation.

Current Scope (v0.1.0)

Implemented core runtime layers:

• Agent runtime: BaseAgent, LlmAgent (Agent alias), Context, InvocationContext
• Runner runtime: Runner, InMemoryRunner
• Session runtime: BaseSessionService, InMemorySessionService, session state prefixes (app:, user:, temp:)
• Event model: Event, EventActions, final-response detection
• LLM model abstraction: BaseLlm, LlmRequest, LlmResponse, LLMRegistry
• Tool runtime: BaseTool, FunctionTool, BaseToolset, tool-callback execution path
• LLM flow: BaseLlmFlow, SingleFlow, AutoFlow, function-call orchestration

This is a functional core port, not full parity yet.

Quick Start

import 'package:adk_dart/adk_dart.dart';

class EchoModel extends BaseLlm {
  EchoModel() : super(model: 'echo');

  @override
  Stream<LlmResponse> generateContent(LlmRequest request, {bool stream = false}) async* {
    yield LlmResponse(content: Content.modelText('echo response'));
  }
}

Future<void> main() async {
  final agent = Agent(name: 'root_agent', model: EchoModel());
  final runner = InMemoryRunner(agent: agent);

  final session = await runner.sessionService.createSession(
    appName: runner.appName,
    userId: 'user_1',
    sessionId: 'session_1',
  );

  await for (final event in runner.runAsync(
    userId: 'user_1',
    sessionId: session.id,
    newMessage: Content.userText('hello'),
  )) {
    print(event.content?.parts.first.text);
  }
}

Test

dart test
dart analyze

Porting Roadmap

Planned next parity layers from adk-python:

• Plugins and callback parity hardening
• Live/Bidi execution path (run_live semantics)
• App-level compaction/resumability edge cases
• Additional built-in tools / model integrations
• A2A, evaluation, telemetry, memory, artifact backends