Retrieval service (RAG) architecture

An answer endpoint over your own documents: vector store, retrieval step, generation step, cached responses.

AWSragpgvectorretrievalknowledge base
Answer endpoint10M requestsQuestion receivedHTTP requestRetrieve contextTop-k over the vectorstoreAnswer with citationsGrounded generationvector_searchSimilarity searchVector storedb.t4g.large +pgvectorSource documents1 TB standardAnswer cachecache.t4g.smallBudget capguardrailArcFoundry runtimeruntime
10 nodes — 4 infrastructure and 6 agent, on one graph. Drawn by the same layout the product uses.
What is in it

Every resource, and what it costs.

Projections from August 2026 list prices for always-on resources. Connect an account and these become the figures your provider actually bills.

Resources in the Retrieval service (RAG) template with projected monthly cost.
NodeTypeWhat it isProjected
Answer endpointNetwork10M requests$10/mo
Vector storeDatabasedb.t4g.large + pgvector$105/mo
Source documentsStorage1 TB standard$12/mo
Answer cacheDatabasecache.t4g.small$25/mo
Question receivedTriggerHTTP request
Retrieve contextAgent stepTop-k over the vector storebilled per token
Answer with citationsAgent stepGrounded generationbilled per token
vector_searchToolSimilarity search
Budget capGuardrailHalts runs over $400/moenforced at runtime
ArcFoundry runtimeRuntime

Agent steps are priced from provider-reported token usage once the agent runs, not estimated. Guardrails cost nothing and are the reason a runaway agent cannot.

Open Retrieval service (RAG) on the canvas.

It loads as an editable graph. Connect an account or instrument an agent and the projected figures above become measured ones.