choose-graph-or-not

name	choose-graph-or-not
description	Walk the graph-vs-relational decision tree and return a one-screen verdict — LPG, RDF, stay relational, stay RAG, or stay KV — plus the neighbor hand-off. Reach for this before any labels or Cypher.

Skill: choose-graph-or-not

Invoked by: graph-data-modeler (primary).

When to invoke: "should this be a graph?"; "do we need Neo4j?"; "LPG or RDF?"; any request to introduce a graph store.

Output: one-screen verdict + the tree leaf + a neighbor hand-off if the answer is no.

Procedure

1. Read the workload. What question is asked at read time? What is rolled back at write time?
2. Traverse ../../knowledge/graph-vs-relational-decision-tree.md. Start from the database-engineering “specialized engine” leaf. Do not skip to labels.
3. If relational / KV / document / RAG: stop. Name the owner (database-engineering, ai-rag-engineering, or stay put). Do not sketch an LPG “anyway.”
4. If LPG: proceed to ../model-property-graph/SKILL.md.
5. If RDF: say SPARQL 1.1 + shared IRIs; do not invent SPARQL 1.2.
6. If retrieval-graph: stop unless a BM25 / no-graph baseline has lost. Hand III.a to memory-architect-lead. Then ../construct-retrieval-graph/SKILL.md.

Guardrails

• Multi-row transactions + ad-hoc SQL + known schema → not a graph.
• “We might want a graph later” is not a yes.
• GraphQL and Microsoft Graph API are not this tree.