Retrieval Ranking & Relevance Control
Rank retrieved memories and context by relevance and freshness before presenting to the reasoner, not just top-K by cosine similarity.
This page is the complete instruction page for one pattern called "Retrieval Ranking & Relevance Control." It explains the problem this pattern solves, the idea behind the solution, when you should (and should not) use it, and what happens afterward, both the good effects and the costs.
This matters because building AI agents is not just about making them clever. It is also about making them safe and predictable. Following a well-tested pattern like this one helps avoid common mistakes, and shows you exactly which safety rules and regulations it connects to, listed under "Standards Mesh" on this page.
Context
Vector similarity alone produces retrieval results that are topically close but practically irrelevant. Stale, low-quality, or off-context memories confuse reasoning.
Problem
Naive similarity retrieval returns "related" but not "relevant" content.
Forces
- Recall vs precision
- Freshness vs coverage
Solution
Rank retrievals by a composite score: similarity × freshness × source trust × task-relevance. Apply hybrid retrieval (dense + sparse). Use cross-encoder re-ranking for top-K refinement. Cap context by relevance threshold, not fixed K.
Where this pattern acts in the two lifecycles
05 Observe. Tool result ingestion, state update. 7 of 61 catalogue patterns act at this stage.
05 Operate & Monitor. Runtime telemetry, drift detection. 18 of 61 catalogue patterns act at this stage.
Stages and counts from the catalogue data. The Runtime lens is the loop the agent runs each turn; the AgentOps lens is the lifecycle the team runs the agent through. See the Matrix for every pattern at each stage.
Applicability
- RAG agents
- Memory-heavy agents
- Agents with large or heterogeneous knowledge stores
Anti-Patterns
- Fixed top-K without relevance check
- Single-stage dense-only retrieval at scale
Consequences
- +Higher reasoning quality
- +Reduced context waste
- −Ranking pipeline complexity
- −Tuning required per corpus
