extensions_genkit 0.3.0

Genkit adapter for the extensions AI abstractions. Plug Genkit models into the ChatClient pipeline — including logging, caching, and function-invocation middleware — without modifying the core packages.

extensions_genkit

Genkit adapter for the extensions AI abstractions.

Use any Genkit model through the existing ChatClient pipeline — logging, caching, function-invocation middleware, and dependency injection — without modifying the core packages.

Features

• GenkitChatClient — streams responses from a Genkit model as ChatResponseUpdate values and surfaces tool requests as FunctionCallContent so the FunctionInvokingChatClient middleware can handle the execution loop.
• AIFunction.toGenkitToolDefinition() — converts an extensions AIFunction to a Genkit ToolDefinition.
• AIFunction.toGenkitTool() — wraps an AIFunction as a Genkit Tool suitable for passing to Genkit.generateStream.
• ServiceCollection.addGenkitChatClient() — registers the client and returns a ChatClientBuilder for attaching middleware.

Getting started

Add the package to your pubspec.yaml:

dependencies:
  extensions_genkit: ^0.1.0

You will also need a Genkit model plugin, for example:

dependencies:
  genkit: ^0.13.0
  genkit_google_genai: ^0.13.0

Usage

Basic streaming

import 'package:extensions/ai.dart';
import 'package:extensions_genkit/extensions_genkit.dart';
import 'package:genkit/genkit.dart';
import 'package:genkit_google_genai/genkit_google_genai.dart';

final ai = Genkit(plugins: [googleAI()]);
final client = GenkitChatClient(
  genkit: ai,
  model: googleAI.gemini('gemini-2.5-flash'),
);

final stream = client.getStreamingResponse(
  messages: [ChatMessage.fromText(ChatRole.user, 'Hello!')],
);

await for (final update in stream) {
  stdout.write(update.text);
}

Dependency injection with middleware

import 'package:extensions/ai.dart';
import 'package:extensions/dependency_injection.dart';
import 'package:extensions_genkit/extensions_genkit.dart';
import 'package:genkit/genkit.dart';
import 'package:genkit_google_genai/genkit_google_genai.dart';

final services = ServiceCollection()
  ..addSingletonInstance<Genkit>(Genkit(plugins: [googleAI()]))
  ..addGenkitChatClient(model: googleAI.gemini('gemini-2.5-flash'))
    .useFunctionInvocation()
    .useLogging();

final sp = services.buildServiceProvider();
final client = sp.getRequiredService<ChatClient>();

Tool calling

Define tools as AIFunction values and pass them via ChatOptions. The FunctionInvokingChatClient middleware (.useFunctionInvocation()) handles the call/respond loop automatically.

final weatherTool = AIFunctionFactory.create(
  name: 'getWeather',
  description: 'Returns the current weather for a location.',
  parametersSchema: {
    'type': 'object',
    'properties': {
      'location': {'type': 'string', 'description': 'City name'},
    },
    'required': ['location'],
  },
  callback: (args, {cancellationToken}) async {
    final location = args['location'] as String;
    return {'temperature': 22, 'condition': 'sunny', 'location': location};
  },
);

final response = await client.getResponse(
  messages: [ChatMessage.fromText(ChatRole.user, "What's the weather in Paris?")],
  options: ChatOptions(tools: [weatherTool]),
);

print(response.text);

Additional information

• extensions package
• Genkit Dart documentation
• File an issue