ai_sdk_google 3.0.0

Google Generative AI provider for the AI SDK Dart port.

example/example.md

ai_sdk_google examples

Installation

dart pub add ai_sdk_dart ai_sdk_google

The top-level google factory reads const String.fromEnvironment('GOOGLE_API_KEY'). Use it with fvm dart run --define=GOOGLE_API_KEY=AIza... bin/app.dart, or read Platform.environment['GOOGLE_API_KEY'] yourself and pass apiKey: to GoogleGenerativeAIProvider in server and CLI apps.

---

Text generation

import 'package:ai_sdk_dart/ai_sdk_dart.dart';
import 'package:ai_sdk_google/ai_sdk_google.dart';

final result = await generateText(
  model: google('gemini-2.0-flash'),
  prompt: 'What is the speed of light?',
);
print(result.text);
print('tokens used: ${result.usage}');

---

Streaming

import 'dart:io';

final result = await streamText(
  model: google('gemini-2.0-flash'),
  prompt: 'Explain how neural networks learn.',
);
await for (final chunk in result.textStream) {
  stdout.write(chunk);
}

---

System prompt

final result = await generateText(
  model: google('gemini-2.0-flash'),
  system: 'You are a helpful assistant that replies only in haiku.',
  prompt: 'Tell me about Flutter.',
);
print(result.text);

instructions is the current canonical top-level instruction parameter; system remains a supported fallback and is used above for continuity with older code. Prefer instructions in new code:

final result = await generateText(
  model: google('gemini-2.0-flash'),
  instructions: 'You are a helpful assistant that replies only in haiku.',
  prompt: 'Tell me about Flutter.',
);
print(result.text);

---

Reasoning

The core reasoning: parameter (LanguageModelV4Reasoning, from package:ai_sdk_provider) works with Google models — it's mapped internally to Gemini's thinkingConfig/thinkingBudget:

import 'package:ai_sdk_provider/ai_sdk_provider.dart';

final result = await generateText(
  model: google('gemini-2.5-pro'),
  reasoning: LanguageModelV4Reasoning.high,
  prompt: 'Solve step by step: if 3x + 5 = 20, what is x?',
);
print(result.text);
print(result.reasoningText);

Signed reasoning replay

Google's reasoning parts carry an opaque thoughtSignature in providerOptions['google']['thoughtSignature']. Don't read or construct it yourself — carry result.responseMessages through unchanged as history and it round-trips automatically:

const question = ModelMessage(
  role: ModelMessageRole.user,
  content: 'Solve step by step: if 3x + 5 = 20, what is x?',
);
final first = await generateText(
  model: google('gemini-2.5-pro'),
  reasoning: LanguageModelV4Reasoning.high,
  messages: [question],
);

final second = await generateText(
  model: google('gemini-2.5-pro'),
  reasoning: LanguageModelV4Reasoning.high,
  messages: [
    question,
    ...first.responseMessages.map(ModelMessage.fromProvider),
    const ModelMessage(role: ModelMessageRole.user, content: 'Now check your work.'),
  ],
);
print(second.text);

---

Structured output

final result = await generateText<Map<String, dynamic>>(
  model: google('gemini-2.0-flash'),
  prompt: 'Return the capital and timezone of Japan as JSON.',
  output: Output.object(
    schema: Schema<Map<String, dynamic>>(
      jsonSchema: const {
        'type': 'object',
        'properties': {
          'capital': {'type': 'string'},
          'timezone': {'type': 'string'},
        },
      },
      fromJson: (json) => json,
    ),
  ),
);
print(result.output); // {capital: Tokyo, timezone: Asia/Tokyo}

---

Tools (multi-step agent)

final result = await generateText(
  model: google('gemini-2.0-flash'),
  prompt: 'What is the weather in Tokyo?',
  maxSteps: 5,
  tools: {
    'getWeather': tool<Map<String, dynamic>, String>(
      description: 'Get current weather for a city.',
      inputSchema: Schema<Map<String, dynamic>>(
        jsonSchema: const {
          'type': 'object',
          'properties': {'city': {'type': 'string'}},
          'required': ['city'],
        },
        fromJson: (json) => json,
      ),
      execute: (input, _) async => 'Sunny, 28°C in ${input['city']}',
    ),
  },
);
print(result.text);

---

Embeddings

final a = await embed(
  model: google.embedding('text-embedding-004'),
  value: 'Hello, world!',
);
final b = await embed(
  model: google.embedding('text-embedding-004'),
  value: 'Hi there!',
);
print('similarity: ${cosineSimilarity(a.embedding, b.embedding)}');

---

Multimodal (image + text)

import 'dart:io';
import 'package:ai_sdk_provider/ai_sdk_provider.dart';

final imageBytes = await File('photo.jpg').readAsBytes();

final result = await generateText(
  model: google('gemini-2.0-flash'),
  messages: [
    ModelMessage.parts(
      role: ModelMessageRole.user,
      parts: [
        LanguageModelV4ImagePart(
          image: DataContentBytes(imageBytes),
          mediaType: 'image/jpeg',
        ),
        const LanguageModelV4TextPart(text: 'What is in this image?'),
      ],
    ),
  ],
);
print(result.text);

---

Custom API key

final myGoogle = GoogleGenerativeAIProvider(apiKey: 'AIza...');
final result = await generateText(
  model: myGoogle('gemini-2.0-flash'),
  prompt: 'Hello!',
);

---

Runnable example apps

• examples/basic — Dart CLI covering all core APIs
• examples/advanced_app — Flutter app with Google provider, multimodal, and embeddings