ChatGoogleGenerativeAI

Wrapper around Google AI for Developers API (aka Gemini API).

Setup

To use ChatGoogleGenerativeAI you need to have an API key. You can get one here.

The following models are available:

• gemini-1.5-flash:
  • text / image / audio -> text model
  • Max input token: 1048576
  • Max output tokens: 8192
• gemini-1.5-pro: text / image -> text model
  • text / image / audio -> text model
  • Max input token: 1048576
  • Max output tokens: 8192
• gemini-pro-vision:
  • text / image -> text model
  • Max input token: 12288
  • Max output tokens: 4096
• gemini-1.0-pro (or gemini-pro):
  • text -> text model
  • Max input token: 30720
  • Max output tokens: 2048

Mind that this list may not be up-to-date. Refer to the documentation for the updated list.

Usage

final apiKey = Platform.environment['GOOGLEAI_API_KEY'];

final chatModel = ChatGoogleGenerativeAI(
  apiKey: apiKey,
  defaultOptions: ChatGoogleGenerativeAIOptions(
    model: 'gemini-1.5-pro',
    temperature: 0,
  ),
);

final chatPrompt = ChatPromptTemplate.fromTemplates([
  (ChatMessageType.system, 'You are a helpful assistant that translates {input_language} to {output_language}.'),
  (ChatMessageType.human, 'Text to translate:\n{text}'),
]);

final chain = chatPrompt | chatModel | StringOutputParser();

final res = await chain.invoke({
  'input_language': 'English',
  'output_language': 'French',
  'text': 'I love programming.',
});
print(res);
// -> 'J'adore programmer.'

Multimodal support

final apiKey = Platform.environment['GOOGLEAI_API_KEY'];

final chatModel = ChatGoogleGenerativeAI(
  apiKey: apiKey,
  defaultOptions: ChatGoogleGenerativeAIOptions(
    model: 'gemini-1.5-pro',
    temperature: 0,
  ),
);
final res = await chatModel.invoke(
  PromptValue.chat([
    ChatMessage.human(
      ChatMessageContent.multiModal([
        ChatMessageContent.text('What fruit is this?'),
        ChatMessageContent.image(
          mimeType: 'image/jpeg',
          data: base64.encode(
            await File('./bin/assets/apple.jpeg').readAsBytes(),
          ),
        ),
      ]),
    ),
  ]),
);
print(res.output.content);
// -> 'That is an apple.'

Streaming

final apiKey = Platform.environment['GOOGLEAI_API_KEY'];

final promptTemplate = ChatPromptTemplate.fromTemplates(const [
  (ChatMessageType.system, 'You are a helpful assistant that replies only with numbers in order without any spaces or commas.'),
  (ChatMessageType.human, 'List the numbers from 1 to {max_num}'),
]);

final chatModel = ChatGoogleGenerativeAI(
  apiKey: apiKey,
  defaultOptions: const ChatGoogleGenerativeAIOptions(
    model: 'gemini-1.5-pro',
    temperature: 0,
  ),
);

final chain = promptTemplate.pipe(chatModel).pipe(StringOutputParser());

final stream = chain.stream({'max_num': '30'});
await stream.forEach(print);
// 1
// 2345678910111213
// 1415161718192021
// 222324252627282930 

Tool calling

ChatGoogleGenerativeAI supports tool calling.

Check the docs for more information on how to use tools.

Example:

const tool = ToolSpec(
  name: 'get_current_weather',
  description: 'Get the current weather in a given location',
  inputJsonSchema: {
    'type': 'object',
    'properties': {
      'location': {
        'type': 'string',
        'description': 'The city and state, e.g. San Francisco, CA',
      },
    },
    'required': ['location'],
  },
);
final chatModel = ChatGoogleGenerativeAI(
  defaultOptions: ChatGoogleGenerativeAIOptions(
    model: 'gemini-1.5-pro',
    temperature: 0,
    tools: [tool],
  ),
);
final res = await model.invoke(
  PromptValue.string('What’s the weather like in Boston and Madrid right now in celsius?'),
);