AWS Bedrock RAG Pipeline for Internal Docs Q&A Built Entirely with Terraform

A developer has published a production-ready retrieval-augmented generation (RAG) pipeline on AWS Bedrock, designed to help employees and support agents quickly find answers from internal documents. The architecture routes PDF, HTML, and TXT files stored in Amazon S3 through a Bedrock Knowledge Base, which handles chunking and embedding, before storing vectors in OpenSearch Serverless. The system uses the RetrieveAndGenerate API to return grounded answers with source citations. The entire infrastructure is managed via Terraform, with embedding and generation models defined as variables, allowing teams to swap or upgrade models by changing a single line in a tfvars file. The setup targets one of the most common enterprise pain points: policy documents and FAQs accumulating in shared drives while staff repeatedly ask the same questions through informal channels.