Ask anything about AWS. The system retrieves relevant chunks from official AWS documentation, then generates a grounded answer with source citations.
A full-stack RAG (Retrieval-Augmented Generation) system that answers questions about AWS services using actual AWS documentation as its knowledge base. Semantic search over indexed docs, grounded answers via Claude, source citations back to original pages.
Questions are embedded with Titan Embeddings v2, matched against a Pinecone vector index of AWS documentation chunks, and answered by Claude Sonnet 4.6 — all routed through Amazon Bedrock. The frontend is a single static HTML file served from S3 via CloudFront with Origin Access Control.
| LLM | Claude Sonnet 4.6 via Amazon Bedrock |
| Embeddings | Amazon Titan Embeddings v2 |
| Vector DB | Pinecone (free tier) |
| Compute | AWS Lambda + API Gateway (REST) |
| Frontend | Static HTML on S3 + CloudFront (OAC) |
| Region | us-east-2 |
End-to-end request flow from browser to answer.
User → CloudFront → API Gateway → Lambda
├── Embed question (Titan v2 via Bedrock)
├── Search Pinecone (top 5 chunks)
├── Build prompt with context
├── Generate answer (Claude via Bedrock)
└── Return response + sources
CloudFront uses Origin Access Control (OAC) — the S3 bucket blocks all public access and serves only signed CloudFront requests.
The Lambda enforces a strict Access-Control-Allow-Origin header pinned to the CloudFront domain.
Question text is sanitized before embedding, and source URLs are validated as http(s):// before being rendered as clickable links.