4º. Building a personal knowledge graph with just PostgreSQL (no Neo4j needed)

At some point during development, we needed to answer questions that search couldn't: "Who's connected to this project?" "What links this email to tomorrow's meeting?" "Which people keep appearing together across my notes?" Semantic search finds records by meaning. Full-text search finds them by keywords. But neither can traverse relationships. For that, you need a graph. The obvious choice was Neo4j, or at least Apache AGE (which adds Cypher queries to PostgreSQL). We evaluated both and went with... two tables and WITH RECURSIVE. This post explains why, and shows the actual schema, queries, and API that power the knowledge graph. Why not a graph database? We seriously considered Apache AGE. It adds OpenCypher support directly inside PostgreSQL — no separate service, same database. The competitive analysis even had it as "Innovation #2: Bi-temporal personal knowledge graph via Apache AGE." We rejected it for three reasons: Scale. This is a personal system. One user, maybe a family. We'