aws-sdk-java-v2-bedrock_skill

Overview

This skill provides reusable patterns and code samples for using Amazon Bedrock with the AWS SDK for Java 2.x. It focuses on foundation model discovery, invocation, streaming responses, embeddings, image generation, and Spring Boot integration. Use it to standardize how your Java applications call Claude, Llama, Titan, Stability Diffusion and other Bedrock-hosted models.

How this skill works

The skill shows how to configure BedrockClient and BedrockRuntimeClient, build provider-specific JSON payloads, and parse responses across different model formats. It includes examples for synchronous invokes, streaming event handlers for real-time output, and embedding extraction for RAG workflows. Spring Boot wiring and bean configuration patterns are provided to make the clients available across services.

When to use it

• When you need to list and inspect available foundation models in Bedrock.
• When invoking Claude, Llama, Titan or third-party models for text generation.
• When implementing streaming responses for real-time UIs or chatbots.
• When producing text embeddings for vector search and RAG systems.
• When integrating generative AI into Spring Boot microservices.

Best practices

• Reuse client instances and configure them as Spring beans to avoid repeated initialization costs.
• Build model-specific payload builders to encapsulate provider differences and reduce parsing bugs.
• Use streaming handlers for long outputs and async clients for non-blocking I/O.
• Never log raw prompts or sensitive outputs; use IAM roles and avoid storing credentials.
• Implement retries with exponential backoff for throttling and validate model responses before use.

Example use cases

• A Spring Boot service that routes prompts to different foundation models based on cost and latency constraints.
• Real-time chat UI that streams model tokens to the browser using the streaming response handler.
• A document search system that generates embeddings with Titan and stores them in a vector index for RAG.
• Automated image generation pipeline using Stable Diffusion models hosted via Bedrock.
• A multi-model experimentation harness that lists available models and compares outputs for evaluation.