Intent routing. Classifying user requests across sub-skills and selecting the correct pipeline.
Architecture guidance. Recommending Neptune Database vs Analytics, property graph vs RDF, Gremlin vs openCypher, and when NOT to use Neptune.
Data modeling patterns. Graph schema design for fraud detection, customer 360, service dependencies, access control, supply chain, knowledge graphs, semantic layers.
Query generation. Producing Gremlin, openCypher, and SPARQL queries with correct pagination, parameterization, and optimization patterns.
Connection recipes. Generating SDK connection code (Python, Java, Node.js) with correct VPC, TLS, IAM auth, and public endpoint configuration for both Neptune Database and Analytics.
GraphRAG pipeline design. Designing document ingestion, entity extraction, graph construction, embedding storage, and two-phase retrieval with Neptune Analytics.
Agentic memory architecture. Designing long-term (Neptune) + short-term (DynamoDB) memory systems with graph traversal and vector search.
Infrastructure provisioning — The skill generates AWS CLI commands, CDK stacks, or boto3 code. The user or their CI/CD pipeline executes them.
Data-plane operations — The skill generates Gremlin/openCypher queries. The user runs them against their Neptune endpoint.
LLM calls — For GraphRAG entity extraction and agentic memory, the skill generates prompts and code structure. The user provides their LLM client (Bedrock, OpenAI, etc.).
All code, configurations, CLI commands, and recommendations produced by this skill are AI-generated. Review all outputs before deploying to production environments.