Memory · Software component

Memory Retriever

Software componentMemoryCognition & Memoryarc:MemoryRetriever

A retrieval component that dispatches a memory query to the memory type suited to it and returns filtered results.

Responsibility. Retrieves relevant memories from the store appropriate to the query type.

Also known as: Memory query router, Memory retrieval mechanism, Memory relevance filter, Episodic retrieval, Memory-augmented retrieval hook, Episodic-to-working retrieval, Semantic-to-working augmentation

Direct neighbourhood (hover for relationship types)

Relationships

is configured by structural

Memory Retrieval Policy Ch5.7 Ch10.1

invokes dependency

is invoked by dependency

reads dependency

sends data to dynamic

is constrained by control

Context Budget Allocator Ch5.9
Context Window Manager abstract Ch5.7

is evaluated by assurance

Evaluation Harness Ch3.7

Design guidance

SHOULD route temporal/event queries to episodic storage and conceptual queries to semantic vector search.
MAY query memory repeatedly until sufficient context is gathered.
MUST filter retrieved memories for relevance; unfiltered memory retrieval can perform worse than no memory.
SHOULD treat retrieval as a ranking problem over relevance, recency and importance rather than pure similarity search.
SHOULD bound the number of retrieved memories by token budget and adapt it to available context.
SHOULD filter by context metadata (segment, time window, system version) so lessons come from contextually matched episodes.
SHOULD place only the most relevant episodes into working memory to avoid saturation.
SHOULD prefer distilled semantic knowledge over episodic or document retrieval when it offers better value per token.

Quantitative guidance

As stated by the sources; verify before use.

Removing relevance filtering dropped retrieval accuracy from 88.7% to 64.3% (research on cognitive memory systems); typically 20-30% performance drop (Ch3.7).
Similarity threshold often 0.7-0.8 or top-k with k=3-10 (Ch5.7).
Memory-augmented ReAct typically injects k=3-5 episodes (Ch5.7).
Budget example: ~20% of window, 10K tokens / 300 tokens per episode ~30 max; 3-10 episodes optimal in practice; stop when similarity < 0.75 (Ch5.7).
Twenty detailed episode summaries may consume ~30K tokens; summaries typically 200-500 tokens (Ch5.7).

Classification

Patterns: Similarity search with contextual filteringAgentic RAG (iterative retrieval)Query-type routingRetrieval-augmented reasoning (prior traces as scaffolding)Content-based retrievalTemporal (recency) retrievalContextual retrieval via metadata filtersCausal retrieval (filter by desired outcomes)Trajectory-level retrievalRelevance-threshold early stopHybrid relevance scoring: semantic similarity, temporal proximity, causal relevanceCompression of retrieved episodes to key facts, decisions and outcomes before injectionMemory integration for planning: retrieve, apply, store (Ref5.07)Graduated memory lookup
Quality attributes: Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)Performance efficiency (ISO/IEC 25010)
Risks mitigated: Retrieval noiseIrrelevant context contaminationContext window overflowStale or misleading high-similarity episodes

Sources

Ch1.4: T. Nguyen, "Memory and Perception Systems," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 1.4. ISBN: 9798244538229.
Ch3.7: T. Nguyen, "Tool Usage Auditing," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.7. ISBN: 9798244538229.
Ch3.10: T. Nguyen, "Efficiency Metrics," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.10. ISBN: 9798244538229.
Ch4.1: T. Nguyen, "Introduction to AI Agent Deployment and Scaling," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 4.1. ISBN: 9798244538229.
Ch5.1: T. Nguyen, "Chain-of-Thought (CoT) Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.1. ISBN: 9798244538229.
Ch5.7: T. Nguyen, "Episodic Memory," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.7. ISBN: 9798244538229.
Ch5.8: T. Nguyen, "Semantic Memory," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.8. ISBN: 9798244538229.
Ch5.9: T. Nguyen, "Working Memory," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.9. ISBN: 9798244538229.
Ch10.1: T. Nguyen, "Conversational UI," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 10.1. ISBN: 9798244538229.
Ref5.04: C. Stryker, "What is AI agent memory?," IBM Think. Accessed: Sep. 27, 2026. [Online]. Available: https://www.ibm.com/think/topics/ai-agent-memory
Ref5.05: Model Context Protocol, "Model Context Protocol servers," GitHub repository. Accessed: Sep. 27, 2026. [Online]. Available: https://github.com/modelcontextprotocol/servers
Ref5.08: X. Huang et al., "Understanding the planning of LLM agents: A survey," arXiv:2402.02716, 2024.