在 Amazon Bedrock 上使用 Anthropic Claude 3.7 Sonnet 的推理功能 - AWS SDK 程式碼範例

文件 AWS SDK 範例 GitHub 儲存庫中有更多可用的 AWS SDK 範例

本文為英文版的機器翻譯版本,如內容有任何歧義或不一致之處,概以英文版為準。

在 Amazon Bedrock 上使用 Anthropic Claude 3.7 Sonnet 的推理功能

下列程式碼範例示範如何在 Amazon Bedrock 上使用 Anthropic Claude 3.7 Sonnet 的推理功能

Java
SDK for Java 2.x
注意

GitHub 上提供更多範例。尋找完整範例,並了解如何在 AWS 程式碼範例儲存庫中設定和執行。

搭配非同步 Bedrock 執行時期用戶端使用 Anthropic Claude 3.7 Sonnet 的推理功能。

import com.example.bedrockruntime.models.anthropicClaude.lib.ReasoningResponse; import software.amazon.awssdk.auth.credentials.DefaultCredentialsProvider; import software.amazon.awssdk.core.document.Document; import software.amazon.awssdk.regions.Region; import software.amazon.awssdk.services.bedrockruntime.BedrockRuntimeAsyncClient; import software.amazon.awssdk.services.bedrockruntime.model.*; import java.util.concurrent.CompletableFuture; /** * This example demonstrates how to use Anthropic Claude 3.7 Sonnet's reasoning capability * with an asynchronous Amazon Bedrock runtime client. * It shows how to: * - Set up the Amazon Bedrock async runtime client * - Create a message * - Configure reasoning parameters * - Send an asynchronous request with reasoning enabled * - Process both the reasoning output and final response */ public class ReasoningAsync { public static ReasoningResponse reasoningAsync() { // Create the Amazon Bedrock runtime client var client = BedrockRuntimeAsyncClient.builder() .credentialsProvider(DefaultCredentialsProvider.create()) .region(Region.US_EAST_1) .build(); // Specify the model ID. For the latest available models, see: // https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html var modelId = "us.anthropic.claude-3-7-sonnet-20250219-v1:0"; // Create the message with the user's prompt var prompt = "Describe the purpose of a 'hello world' program in one line."; var message = Message.builder() .content(ContentBlock.fromText(prompt)) .role(ConversationRole.USER) .build(); // Configure reasoning parameters with a 2000 token budget Document reasoningConfig = Document.mapBuilder() .putDocument("thinking", Document.mapBuilder() .putString("type", "enabled") .putNumber("budget_tokens", 2000) .build()) .build(); try { // Send message and reasoning configuration to the model CompletableFuture<ConverseResponse> asyncResponse = client.converse(request -> request .additionalModelRequestFields(reasoningConfig) .messages(message) .modelId(modelId) ); // Process the response asynchronously return asyncResponse.thenApply(response -> { var content = response.output().message().content(); ReasoningContentBlock reasoning = null; String text = null; // Process each content block to find reasoning and response text for (ContentBlock block : content) { if (block.reasoningContent() != null) { reasoning = block.reasoningContent(); } else if (block.text() != null) { text = block.text(); } } return new ReasoningResponse(reasoning, text); } ).get(); } catch (Exception e) { System.err.printf("Can't invoke '%s': %s", modelId, e.getMessage()); throw new RuntimeException(e); } } public static void main(String[] args) { // Execute the example and display reasoning and final response ReasoningResponse response = reasoningAsync(); System.out.println("\n<thinking>"); System.out.println(response.reasoning().reasoningText()); System.out.println("</thinking>\n"); System.out.println(response.text()); } }

搭配同步 Bedrock 執行時期用戶端使用 Anthropic Claude 3.7 Sonnet 的推理功能。

import com.example.bedrockruntime.models.anthropicClaude.lib.ReasoningResponse; import software.amazon.awssdk.auth.credentials.DefaultCredentialsProvider; import software.amazon.awssdk.core.document.Document; import software.amazon.awssdk.core.exception.SdkClientException; import software.amazon.awssdk.regions.Region; import software.amazon.awssdk.services.bedrockruntime.BedrockRuntimeClient; import software.amazon.awssdk.services.bedrockruntime.model.*; /** * This example demonstrates how to use Anthropic Claude 3.7 Sonnet's reasoning capability * with the synchronous Amazon Bedrock runtime client. * It shows how to: * - Set up the Amazon Bedrock runtime client * - Create a message * - Configure reasoning parameters * - Send a request with reasoning enabled * - Process both the reasoning output and final response */ public class Reasoning { public static ReasoningResponse reasoning() { // Create the Amazon Bedrock runtime client var client = BedrockRuntimeClient.builder() .credentialsProvider(DefaultCredentialsProvider.create()) .region(Region.US_EAST_1) .build(); // Specify the model ID. For the latest available models, see: // https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html var modelId = "us.anthropic.claude-3-7-sonnet-20250219-v1:0"; // Create the message with the user's prompt var prompt = "Describe the purpose of a 'hello world' program in one line."; var message = Message.builder() .content(ContentBlock.fromText(prompt)) .role(ConversationRole.USER) .build(); // Configure reasoning parameters with a 2000 token budget Document reasoningConfig = Document.mapBuilder() .putDocument("thinking", Document.mapBuilder() .putString("type", "enabled") .putNumber("budget_tokens", 2000) .build()) .build(); try { // Send message and reasoning configuration to the model ConverseResponse bedrockResponse = client.converse(request -> request .additionalModelRequestFields(reasoningConfig) .messages(message) .modelId(modelId) ); // Extract both reasoning and final response var content = bedrockResponse.output().message().content(); ReasoningContentBlock reasoning = null; String text = null; // Process each content block to find reasoning and response text for (ContentBlock block : content) { if (block.reasoningContent() != null) { reasoning = block.reasoningContent(); } else if (block.text() != null) { text = block.text(); } } return new ReasoningResponse(reasoning, text); } catch (SdkClientException e) { System.err.printf("ERROR: Can't invoke '%s'. Reason: %s", modelId, e.getMessage()); throw new RuntimeException(e); } } public static void main(String[] args) { // Execute the example and display reasoning and final response ReasoningResponse response = reasoning(); System.out.println("\n<thinking>"); System.out.println(response.reasoning().reasoningText()); System.out.println("</thinking>\n"); System.out.println(response.text()); } }
  • 如需 API 詳細資訊,請參閱《AWS SDK for Java 2.x API 參考》中的 Converse