1:1 mentoring with Big Tech AI engineers
RAG & MCP

Advanced RAG Patterns

Every advanced pattern is one decision wrapped around a search call and one more model call to pay for it: search as a tool, what the router and grader papers trained, fallbacks, multi-hop, and choosing by failure.

Last updated

After this section you can

  • Trace one search tool call, from the query to a cited answer
  • Choose a retrieval pattern from a failure you measured
  • Count the model calls and tokens a pattern adds
14

Advanced RAG Patterns

Count what each model decision costs, and keep only the patterns that earn their calls.

Key idea

A pattern is one search wrapped in decisions. A model decision costs a call, so add a pattern only when a measured failure calls for it, and keep it only if it wins on that slice of the golden set.

Every pattern is one search with model decisions around it
Route search or answer? +1 call One search Grade the chunks +1 batched, or +0 Answer Next hop? +1 call per hop rewrite, +1 call no search

Amber boxes are model decisions. The dashed path skips the search. Wider than your screen? Scroll sideways.

Related

More in RAG & MCP

Get full access to all 149 sections with code examples, diagrams, and interactive animations.

Unlock Premium