AWS Sdk Java V2 Bedrock

AWS SDK for Java 2.x - Amazon Bedrock

When to Use

Use this skill when:

• Listing and inspecting foundation models on Amazon Bedrock
• Invoking foundation models for text generation (Claude, Llama, Titan)
• Generating images with AI models (Stable Diffusion)
• Creating text embeddings for RAG applications
• Implementing streaming responses for real-time generation
• Working with multiple AI providers through unified API
• Integrating generative AI into Spring Boot applications
• Building AI-powered chatbots and assistants

Overview

Amazon Bedrock provides access to foundation models from leading AI providers through a unified API. This skill covers patterns for working with various models including Claude, Llama, Titan, and Stability Diffusion using AWS SDK for Java 2.x.

Instructions

Follow these steps to work with Amazon Bedrock:

1. Set Up AWS Credentials - Configure credentials with appropriate Bedrock permissions
2. Enable Bedrock Access - Request access to specific foundation models in the AWS Console
3. Add Dependencies - Include bedrock and bedrockruntime dependencies
4. Create Clients - Instantiate BedrockClient for management and BedrockRuntimeClient for invocation
5. List Models - Query available foundation models and their capabilities
6. Invoke Models - Build proper payloads for each model provider's format
7. Handle Streaming - Implement streaming response handlers for real-time generation
8. Integrate with Spring - Configure beans and services for enterprise applications

Quick Start

Dependencies

<!-- Bedrock (model management) -->
<dependency>
    <groupId>software.amazon.awssdk</groupId>
    <artifactId>bedrock</artifactId>
</dependency>

<!-- Bedrock Runtime (model invocation) -->
<dependency>
    <groupId>software.amazon.awssdk</groupId>
    <artifactId>bedrockruntime</artifactId>
</dependency>

<!-- For JSON processing -->
<dependency>
    <groupId>org.json</groupId>
    <artifactId>json</artifactId>
    <version>20231013</version>
</dependency>

Basic Client Setup

import software.amazon.awssdk.regions.Region;
import software.amazon.awssdk.services.bedrock.BedrockClient;
import software.amazon.awssdk.services.bedrockruntime.BedrockRuntimeClient;

// Model management client
BedrockClient bedrockClient = BedrockClient.builder()
    .region(Region.US_EAST_1)
    .build();

// Model invocation client
BedrockRuntimeClient bedrockRuntimeClient = BedrockRuntimeClient.builder()
    .region(Region.US_EAST_1)
    .build();

Core Patterns

Model Discovery

import software.amazon.awssdk.services.bedrock.model.*;
import java.util.List;

public List<FoundationModelSummary> listFoundationModels(BedrockClient bedrockClient) {
    return bedrockClient.listFoundationModels().modelSummaries();
}

Model Invocation

import software.amazon.awssdk.core.SdkBytes;
import software.amazon.awssdk.services.bedrockruntime.model.*;
import org.json.JSONObject;

public String invokeModel(BedrockRuntimeClient client, String modelId, String prompt) {
    JSONObject payload = createPayload(modelId, prompt);

    InvokeModelResponse response = client.invokeModel(request -> request
        .modelId(modelId)
        .body(SdkBytes.fromUtf8String(payload.toString())));

    return extractTextFromResponse(modelId, response.body().asUtf8String());
}

private JSONObject createPayload(String modelId, String prompt) {
    if (modelId.startsWith("anthropic.claude")) {
        return new JSONObject()
            .put("anthropic_version", "bedrock-2023-05-31")
            .put("max_tokens", 1000)
            .put("messages", new JSONObject[]{
                new JSONObject().put("role", "user").put("content", prompt)
            });
    } else if (modelId.startsWith("amazon.titan")) {
        return new JSONObject()
            .put("inputText", prompt)
            .put("textGenerationConfig", new JSONObject()
                .put("maxTokenCount", 512)
                .put("temperature", 0.7));
    } else if (modelId.startsWith("meta.llama")) {
        return new JSONObject()
            .put("prompt", "[INST] " + prompt + " [/INST]")
            .put("max_gen_len", 512)
            .put("temperature", 0.7);
    }
    throw new IllegalArgumentException("Unsupported model: " + modelId);
}

Streaming Responses

public void streamResponse(BedrockRuntimeClient client, String modelId, String prompt) {
    JSONObject payload = createPayload(modelId, prompt);

    InvokeModelWithResponseStreamRequest streamRequest =
        InvokeModelWithResponseStreamRequest.builder()
            .modelId(modelId)
            .body(SdkBytes.fromUtf8String(payload.toString()))
            .build();

    client.invokeModelWithResponseStream(streamRequest,
        InvokeModelWithResponseStreamResponseHandler.builder()
            .onEventStream(stream -> {
                stream.forEach(event -> {
                    if (event instanceof PayloadPart) {
                        PayloadPart payloadPart = (PayloadPart) event;
                        String chunk = payloadPart.bytes().asUtf8String();
                        processChunk(modelId, chunk);
                    }
                });
            })
            .build());
}

Text Embeddings

public double[] createEmbeddings(BedrockRuntimeClient client, String text) {
    String modelId = "amazon.titan-embed-text-v1";

    JSONObject payload = new JSONObject().put("inputText", text);

    InvokeModelResponse response = client.invokeModel(request -> request
        .modelId(modelId)
        .body(SdkBytes.fromUtf8String(payload.toString())));

    JSONObject responseBody = new JSONObject(response.body().asUtf8String());
    JSONArray embeddingArray = responseBody.getJSONArray("embedding");

    double[] embeddings = new double[embeddingArray.length()];
    for (int i = 0; i < embeddingArray.length(); i++) {
        embeddings[i] = embeddingArray.getDouble(i);
    }

    return embeddings;
}

Spring Boot Integration

@Configuration
public class BedrockConfiguration {

    @Bean
    public BedrockClient bedrockClient() {
        return BedrockClient.builder()
            .region(Region.US_EAST_1)
            .build();
    }

    @Bean
    public BedrockRuntimeClient bedrockRuntimeClient() {
        return BedrockRuntimeClient.builder()
            .region(Region.US_EAST_1)
            .build();
    }
}

@Service
public class BedrockAIService {

    private final BedrockRuntimeClient bedrockRuntimeClient;

    @Value("${bedrock.default-model-id:anthropic.claude-sonnet-4-5-20250929-v1:0}")
    private String defaultModelId;

    public BedrockAIService(BedrockRuntimeClient bedrockRuntimeClient) {
        this.bedrockRuntimeClient = bedrockRuntimeClient;
    }

    public String generateText(String prompt) {
        return generateText(prompt, defaultModelId);
    }

    public String generateText(String prompt, String modelId) {
        Map<String, Object> payload = createPayload(modelId, prompt);
        String payloadJson = new ObjectMapper().writeValueAsString(payload);

        InvokeModelResponse response = bedrockRuntimeClient.invokeModel(
            request -> request
                .modelId(modelId)
                .body(SdkBytes.fromUtf8String(payloadJson)));

        return extractTextFromResponse(modelId, response.body().asUtf8String());
    }
}

Basic Usage Example

BedrockRuntimeClient client = BedrockRuntimeClient.builder()
    .region(Region.US_EAST_1)
    .build();

String prompt = "Explain quantum computing in simple terms";
String response = invokeModel(client, "anthropic.claude-sonnet-4-5-20250929-v1:0", prompt);
System.out.println(response);

Best Practices

Model Selection

• Claude 4.5 Sonnet: Best for complex reasoning, analysis, and creative tasks
• Claude 4.5 Haiku: Fast and affordable for real-time applications
• Claude 3.7 Sonnet: Most advanced reasoning capabilities
• Llama 3.1: Latest generation open-source alternative, good for general tasks
• Titan: AWS native, cost-effective for simple text generation

Performance Optimization

• Reuse client instances (don't create new clients for each request)
• Use async clients for I/O operations
• Implement streaming for long responses
• Cache foundation model lists

Security

• Never log sensitive prompt data
• Use IAM roles for authentication (never access keys)
• Implement rate limiting for public applications
• Sanitize user inputs to prevent prompt injection

Error Handling

• Implement retry logic for throttling (exponential backoff)
• Handle model-specific validation errors
• Validate responses before processing
• Use proper exception handling for different error types

Cost Optimization

• Use appropriate max_tokens limits
• Choose cost-effective models for simple tasks
• Cache embeddings when possible
• Monitor usage and set budget alerts

Common Model IDs

// Claude Models
public static final String CLAUDE_SONNET_4_5 = "anthropic.claude-sonnet-4-5-20250929-v1:0";
public static final String CLAUDE_HAIKU_4_5 = "anthropic.claude-haiku-4-5-20251001-v1:0";
public static final String CLAUDE_OPUS_4_1 = "anthropic.claude-opus-4-1-20250805-v1:0";
public static final String CLAUDE_3_7_SONNET = "anthropic.claude-3-7-sonnet-20250219-v1:0";
public static final String CLAUDE_OPUS_4 = "anthropic.claude-opus-4-20250514-v1:0";
public static final String CLAUDE_SONNET_4 = "anthropic.claude-sonnet-4-20250514-v1:0";
public static final String CLAUDE_3_5_SONNET_V2 = "anthropic.claude-3-5-sonnet-20241022-v2:0";
public static final String CLAUDE_3_5_HAIKU = "anthropic.claude-3-5-haiku-20241022-v1:0";
public static final String CLAUDE_3_OPUS = "anthropic.claude-3-opus-20240229-v1:0";

// Llama Models
public static final String LLAMA_3_3_70B = "meta.llama3-3-70b-instruct-v1:0";
public static final String LLAMA_3_2_90B = "meta.llama3-2-90b-instruct-v1:0";
public static final String LLAMA_3_2_11B = "meta.llama3-2-11b-instruct-v1:0";
public static final String LLAMA_3_2_3B = "meta.llama3-2-3b-instruct-v1:0";
public static final String LLAMA_3_2_1B = "meta.llama3-2-1b-instruct-v1:0";
public static final String LLAMA_4_MAV_17B = "meta.llama4-maverick-17b-instruct-v1:0";
public static final String LLAMA_4_SCOUT_17B = "meta.llama4-scout-17b-instruct-v1:0";
public static final String LLAMA_3_1_405B = "meta.llama3-1-405b-instruct-v1:0";
public static final String LLAMA_3_1_70B = "meta.llama3-1-70b-instruct-v1:0";
public static final String LLAMA_3_1_8B = "meta.llama3-1-8b-instruct-v1:0";
public static final String LLAMA_3_70B = "meta.llama3-70b-instruct-v1:0";
public static final String LLAMA_3_8B = "meta.llama3-8b-instruct-v1:0";

// Amazon Titan Models
public static final String TITAN_TEXT_EXPRESS = "amazon.titan-text-express-v1";
public static final String TITAN_TEXT_LITE = "amazon.titan-text-lite-v1";
public static final String TITAN_EMBEDDINGS = "amazon.titan-embed-text-v1";
public static final String TITAN_IMAGE_GENERATOR = "amazon.titan-image-generator-v1";

// Stable Diffusion
public static final String STABLE_DIFFUSION_XL = "stability.stable-diffusion-xl-v1";

// Mistral AI Models
public static final String MISTRAL_LARGE_2407 = "mistral.mistral-large-2407-v1:0";
public static final String MISTRAL_LARGE_2402 = "mistral.mistral-large-2402-v1:0";
public static final String MISTRAL_SMALL_2402 = "mistral.mistral-small-2402-v1:0";
public static final String MISTRAL_PIXTRAL_2502 = "mistral.pixtral-large-2502-v1:0";
public static final String MISTRAL_MIXTRAL_8X7B = "mistral.mixtral-8x7b-instruct-v0:1";
public static final String MISTRAL_7B = "mistral.mistral-7b-instruct-v0:2";

// Amazon Nova Models
public static final String NOVA_PREMIER = "amazon.nova-premier-v1:0";
public static final String NOVA_PRO = "amazon.nova-pro-v1:0";
public static final String NOVA_LITE = "amazon.nova-lite-v1:0";
public static final String NOVA_MICRO = "amazon.nova-micro-v1:0";
public static final String NOVA_CANVAS = "amazon.nova-canvas-v1:0";
public static final String NOVA_REEL = "amazon.nova-reel-v1:1";

// Other Models
public static final String COHERE_COMMAND = "cohere.command-text-v14";
public static final String DEEPSEEK_R1 = "deepseek.r1-v1:0";
public static final String DEEPSEEK_V3_1 = "deepseek.v3-v1:0";

Examples

Example 1: Simple Text Generation with Claude

public String generateWithClaude(BedrockRuntimeClient client, String prompt) {
    JSONObject payload = new JSONObject()
        .put("anthropic_version", "bedrock-2023-05-31")
        .put("max_tokens", 1000)
        .put("messages", new JSONObject[]{
            new JSONObject().put("role", "user").put("content", prompt)
        });

    InvokeModelResponse response = client.invokeModel(InvokeModelRequest.builder()
        .modelId("anthropic.claude-sonnet-4-5-20250929-v1:0")
        .body(SdkBytes.fromUtf8String(payload.toString()))
        .build());

    JSONObject responseBody = new JSONObject(response.body().asUtf8String());
    return responseBody.getJSONArray("content")
        .getJSONObject(0)
        .getString("text");
}

Example 2: Streaming Response

public void streamResponse(BedrockRuntimeClient client, String modelId, String prompt) {
    JSONObject payload = new JSONObject()
        .put("anthropic_version", "bedrock-2023-05-31")
        .put("max_tokens", 500)
        .put("messages", new JSONObject[]{
            new JSONObject().put("role", "user").put("content", prompt)
        });

    InvokeModelWithResponseStreamRequest request = InvokeModelWithResponseStreamRequest.builder()
        .modelId(modelId)
        .body(SdkBytes.fromUtf8String(payload.toString()))
        .build();

    client.invokeModelWithResponseStream(request,
        InvokeModelWithResponseStreamResponseHandler.builder()
            .onEventStream(stream -> stream.forEach(event -> {
                if (event instanceof PayloadPart) {
                    String chunk = ((PayloadPart) event).bytes().asUtf8String();
                    System.out.print(chunk);
                }
            }))
            .build());
}

Example 3: Spring Boot Service

@Service
public class BedrockService {

    private final BedrockRuntimeClient client;
    private final ObjectMapper mapper;

    @Value("${bedrock.model:anthropic.claude-sonnet-4-5-20250929-v1:0}")
    private String modelId;

    public String generate(String prompt) {
        try {
            Map<String, Object> payload = Map.of(
                "anthropic_version", "bedrock-2023-05-31",
                "max_tokens", 1000,
                "messages", List.of(Map.of(
                    "role", "user",
                    "content", prompt
                ))
            );

            InvokeModelResponse response = client.invokeModel(
                InvokeModelRequest.builder()
                    .modelId(modelId)
                    .body(SdkBytes.fromUtf8String(mapper.writeValueAsString(payload)))
                    .build()
            );

            return extractText(response.body().asUtf8String());
        } catch (Exception e) {
            throw new RuntimeException("Bedrock invocation failed", e);
        }
    }
}

See the examples directory for comprehensive usage patterns.

Advanced Topics

See the Advanced Topics for:

• Multi-model service patterns
• Advanced error handling with retries
• Batch processing strategies
• Performance optimization techniques
• Custom response parsing

Model Reference

See the Model Reference for:

• Detailed model specifications
• Payload/response formats for each provider
• Performance characteristics
• Model selection guidelines
• Configuration templates

Testing Strategies

See the Testing Strategies for:

• Unit testing with mocked clients
• Integration testing with LocalStack
• Performance testing
• Streaming response testing
• Test data management

Related Skills

• aws-sdk-java-v2-core - Core AWS SDK patterns
• langchain4j-ai-services-patterns - LangChain4j integration
• spring-boot-dependency-injection - Spring DI patterns
• spring-boot-test-patterns - Spring testing patterns

Constraints and Warnings

• Cost Management: Bedrock API calls incur charges per token; implement usage monitoring and set budget alerts to avoid unexpected costs.
• Model Access: Foundation models must be explicitly enabled in the AWS Console before use; verify model availability in your region.
• Rate Limits: Bedrock has per-model and account-level throttling limits; implement exponential backoff for production workloads.
• Region Availability: Not all models are available in all regions; check model availability before deployment.
• Payload Size: Maximum payload size varies by model; for large documents, consider chunking strategies.
• Streaming Complexity: Streaming responses require careful handling of partial content and error recovery.
• Security: Never embed credentials in code; use IAM roles for EC2/Lambda, environment variables for local development.
• Prompt Injection: Sanitize user inputs to prevent prompt injection attacks that could manipulate model behavior.
• Data Privacy: Prompts and responses may be logged by AWS; review data handling policies for sensitive applications.

References

• AWS Bedrock User Guide
• AWS SDK for Java 2.x Documentation
• Bedrock API Reference
• AWS SDK Examples
• Official AWS Examples
• Supported Models
• Runtime Examples