bedrock

The Bedrock skill provides a comprehensive knowledge base for integrating Amazon Bedrock into software agent workflows. It is designed for developers, cloud architects, and AI engineers working on generative AI applications. This skill covers the technical implementation of foundation models via the AWS SDK for Python (boto3) and the AWS CLI, focusing on high-efficiency patterns that reduce context overhead while maintaining robust performance. It serves as an internal reference for interacting with models like Claude, Titan, Llama, Mistral, and Stable Diffusion, ensuring that agents can reliably handle inference, conversational history, and data processing tasks.

• Support for text generation with streaming capabilities for low-latency responses.

• Implementation guides for generating vector embeddings using models like Titan Embed Text.

• Advanced conversational patterns including system prompt management and history tracking.

• Configuration logic for model access requests, region-specific provisioning, and EULA requirements.

• Best practices for switching between On-Demand, Provisioned Throughput, and Batch Inference pricing models.

• Integration strategies for RAG (Retrieval-Augmented Generation) architectures.

• Use this skill to generate valid JSON payloads for invoke-model calls to Bedrock Runtime.

• The skill includes specific boilerplate code for handling Claude 3 multi-modal inputs, including base64 encoded images.

• It provides troubleshooting guidance for common errors related to model IDs, content-type mismatches, and IAM permission scope for bedrock:InvokeModel.

• Useful for automating infrastructure as code snippets or troubleshooting environment-specific access issues.

• Ensure all requests provide valid JSON structures as defined by the bedrock-runtime schema.