buildOperations

The operator-facing surface: observability, cache control, SQL composition, multi-PG support, install.

These aren't a fifth pillar — they're the seams that let the four pillars compose cleanly inside Postgres.

In this section

• query_stats pgrdf.stats() observability — JSONB health snapshot.
• layers_clear Cache control — explicit invalidation for the dictionary cache.
• bolt Prepared-plan cache — SPARQL→SQL plan reuse per backend.
• hub Compose with regular SQL — pgrdf.sparql as a set-returning function.
• settings Multi-version Postgres support.
• build Drop-in install.

auto_storiesTraining

A two-pass learning order for operators stewarding a pgRDF deployment:

First pass — get it installed and observable

• build Drop-in install — three artefacts bind-mounted onto a stock postgres:17 image, plus the one required shared_preload_libraries line. The five-minute path.
• settings Multi-version Postgres support — PostgreSQL 17 is the supported head during the stabilization window; the pg14/15/16 matrix is temporarily paused.
• query_stats pgrdf.stats() observability — the JSONB key contract you scrape into Prometheus.

Second pass — tune and compose

• bolt Prepared-plan cache — how SPARQL→SQL plans get reused per backend; how to size it.
• layers_clear Cache control — explicit invalidation primitives for the dictionary and plan caches.
• hub Compose with regular SQL — how to wire pgrdf.sparql into views, ORMs, BI tools, and INSERT INTO ... SELECT pipelines.

Learn more

• school SPEC.pgRDF.INSTALL.v0.2.md — the canonical install spec (per-PG tarball layout, K8s init-container variant).
• school Postgres operational guides — pg_stat_statements, pg_stat_activity, EXPLAIN — the standard tooling you'll combine with pgrdf.stats().