Data Infrastructure

Published 2026-04-20·Updated 2026-10-15·2 min read

AI documentation guide

Overview

For RAG pipelines and long-term memory that rely on vector search, that vector search infrastructure must be operated reliably. Other retrieval or memory backends that do not use vector search follow their own requirements. This section covers Milvus-based vector database operations. Future documents on Feature Store operations, Knowledge Graph infrastructure, and other data layer components will be added.

What This Section Covers

• Milvus Vector Database -- EKS-based Milvus cluster architecture, index selection (HNSW, IVF, DiskANN), and sharding/replication operations

Related Documents

• Domain Customization (LoRA + RAG) -- The role of vector search in RAG pipelines and selection criteria
• Ragas Evaluation Framework -- Evaluating vector search quality (Context Precision, Recall)
• Monitoring Stack Setup -- Platform observability including the data layer

Document List

Milvus Vector DB

Deploying Milvus vector database on Amazon EKS and integrating with RAG pipelines