```json { "table_name": "table_topology", "table_type_tags": [ {"tag": "dimension", "confidence": 0.85}, {"tag": "misc", "confidence": 0.20} ], "primary_type": "dimension", "insufficient_evidence": false, "entity": { "columns": ["jobs_id", "tables_catalog_id"], "unique_per_row": true }, "time_axis": null, "reasoning": "The composite entity (jobs_id, tables_catalog_id) is unique per row — jobs_id RPUV 2.88 × tables_catalog_id RPUV 1.03 ≈ 2.97, plausibly ≥ 334 rows, and the data preview shows distinct (jobs_id, tables_catalog_id) pairs. The four provenance columns (_ingestion_timestamp, _batch_id, _source_file, _source_system) are ignored per the rules. No usable time axis exists — has_temporal and temporal_columns_json describe the REFERENCED tables' temporal properties, not this table's own event timeline. Each row catalogs metadata about a table within a job: its assigned role, byte size, and column inventories. Composite entity unique per row with no axis → dimension." } ```