feat: Add support for tool use in anthropic_sdk_dart client

[davidmigloz]

Claude is capable of interacting with external client-side tools and functions, allowing you to equip Claude with your own custom tools to perform a wider variety of tasks.

Refer to the official documentation for more information.

In the following example, we want the model to be able to use our function that return the current weather in a given city:

Map<String, dynamic> _getCurrentWeather(
  final String location,
  final String unit,
) {
  const temperature = 22;
  const weather = 'Sunny';
  return {
    'temperature': unit == 'celsius' ? temperature : (temperature * 9 / 5) + 32,
    'unit': unit,
    'description': weather,
  };
}

To do that, we need to provide the definition of the tool:

const tool = Tool(
  name: 'get_current_weather',
  description: 'Get the current weather in a given location',
  inputSchema: {
    'type': 'object',
    'properties': {
      'location': {
        'type': 'string',
        'description': 'The city and state, e.g. San Francisco, CA',
      },
      'unit': {
        'type': 'string',
        'description': 'The unit of temperature to return',
        'enum': ['celsius', 'fahrenheit'],
      },
    },
    'required': ['location'],
  },
);

Then we can use the tool in the message request:

final request1 = CreateMessageRequest(
  model: Model.model(Models.claude35Sonnet20240620),
  messages: [
    Message(
      role: MessageRole.user,
      content: MessageContent.text(
        'What’s the weather like in Boston right now?',
      ),
    ),
  ],
  tools: [tool],
  toolChoice: ToolChoice(
    type: ToolChoiceType.tool,
    name: tool.name,
  ),
  maxTokens: 1024,
);
final aiMessage1 = await client.createMessage(request: request1);

final toolUse = aiMessage1.content.blocks.firstOrNull;
if (toolUse == null || toolUse is! ToolUseBlock) {
  return;
}

// Call your tool here with the given input
final toolResult = _getCurrentWeather(
  toolUse.input['location'],
  toolUse.input['unit'],
);

final request2 = CreateMessageRequest(
  model: Model.model(Models.claude35Sonnet20240620),
  messages: [
    Message(
      role: MessageRole.user,
      content: MessageContent.text(
        'What’s the weather like in Boston right now in Fahrenheit?',
      ),
    ),
    Message(
      role: MessageRole.assistant,
      content: aiMessage1.content,
    ),
    Message(
      role: MessageRole.user,
      content: MessageContent.blocks([
        Block.toolResult(
          toolUseId: toolUse.id,
          content: ToolResultBlockContent.text(json.encode(toolResult)),
        ),
      ]),
    ),
  ],
  tools: [tool],
  maxTokens: 1024,
);
final aiMessage2 = await client.createMessage(request: request2);

print(aiMessage2.content.text);
// Based on the current weather information for Boston, here's what it's like right now:
//
// The temperature in Boston is 71.6°F (Fahrenheit).
// The weather conditions are described as sunny.

You can also stream the input for a tool:

final stream = client.createMessageStream(request: request);
await for (final res in stream) {
  res.map(
    messageStart: (MessageStartEvent v) {},
    messageDelta: (MessageDeltaEvent v) {},
    messageStop: (MessageStopEvent v) {},
    contentBlockStart: (ContentBlockStartEvent v) {},
    contentBlockDelta: (ContentBlockDeltaEvent v) {
      stdout.write(v.delta.inputJson);
    },
    contentBlockStop: (ContentBlockStopEvent v) {},
    ping: (PingEvent v) {},
  );
}
// {"location": "Boston, MA", "unit": "fahrenheit"}