RAG Service

Express
   ↓
POST /chat
   ↓
FastAPI
   ↓
RAG
   ↓
Vector DB
   ↓
LLM

We're separating the AI service from the main application so each component has a clear responsibility.