RAG Demo — architecture

Clean RAG on AWS. Interactive host: rag.automatico.llc.

Path

Documents in S3 are chunked and embedded, stored in Qdrant, then retrieved for an answer with citations.

flowchart LR
  User[User] --> Edge[rag.automatico.llc]
  Edge --> Ask[Ask / Search]
  Ask --> Agents[Agents]
  Agents --> Ingest[Ingest / retrieve]
  Ingest --> S3[S3 corpus]
  Ingest --> Qdrant[Vector index]
  Agents --> Model[Generation]
  Model --> User
        

What comes back for a demo

Kept while pausedS3 corpus documents
Restored on requestAuth, API, retrieval index, ask/search UI
Not kept idleLogin pool, compute, and the vector service