Skip to content

Memory Store Selection

Use a Qdrant-style vector index for semantic retrieval and a Neo4j-style graph model for explicit relationships in the demo architecture. This keeps vector retrieval and graph reasoning separate enough to teach clearly.

Tradeoffs

ChoiceBest fitStrengthCost
ChromaLocal development and simple prototypesLow-friction embedding workflowsLess emphasis on production operations
QdrantVector search serviceStrong filtering and deployment storySeparate service to operate
WeaviateHybrid semantic/search workloadsRich AI-native database featuresMore platform surface area
pgvectorPostgreSQL-first teamsKeeps vectors near relational dataScaling and tuning stay tied to Postgres
Neo4jExplicit entity relationshipsMature graph modeling and traversalSeparate graph database to operate
RDF/SPARQLStandards-based semantic dataInteroperable semantic web modelHigher modeling complexity

Architecture Guidance

  • Use vector search for semantic similarity and RAG.
  • Use a graph when relationships are first-class data, not only metadata.
  • Keep document source storage separate from indexes.
  • Store document IDs in both vector payloads and graph nodes so results can be joined.

Example Artifact

Page created by Dr. C. Klukas