Graph-Constrained Vector Retriever

Software componentKnowledge & DataKnowledge & Dataarc:GraphConstrainedVectorRetriever

A hybrid retriever that first selects semantically similar candidates by vector search, then filters them by graph relationship constraints such as temporal proximity, causal connection, entity overlap or interaction pattern.

Responsibility. Filters vector-similar candidates by graph relationship constraints in a two-stage query.

Also known as: Two-stage vector-then-graph retrieval, Hybrid episodic retrieval

Variant of Hybrid Retriever abstract

Direct neighbourhood (hover for relationship types)

Relationships

is invoked by dependency

Memory Retriever Ch5.7

reads dependency

Knowledge Graph Store abstract Ch5.7
Vector Index Store abstract Ch5.7

Design guidance

SHOULD maintain consistency between the vector store and graph store it queries.

Quantitative guidance

As stated by the sources; verify before use.

Example: stage one returns 50 candidate episodes; graph filtering narrows to a 3-episode sequence over two weeks for one customer (Ch5.7).

Classification

Patterns: Two-stage retrieval (vector candidates, graph filter)
Technologies: PineconeNeo4j
Quality attributes: Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)
Risks mitigated: Embedding blind spotsLoss of temporal/causal structure in pure similarity search

Sources

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.