```json { "table_name": "batches", "table_type_tags": [ {"tag": "dimension", "confidence": 0.90}, {"tag": "path", "confidence": 0.15} ], "primary_type": "dimension", "insufficient_evidence": false, "entity": { "columns": ["batch_id"], "unique_per_row": true }, "time_axis": { "columns": ["started_at", "completed_at"], "kind": "event_time" }, "reasoning": "batch_id has Uniqueness 1.000, making each row unique information about one batch execution. The data preview shows pairs of rows (base batch + _gold variant) with distinct IDs, and sibling tables (jobs, datasets) likely reference batch_id as the catalog of processing runs. started_at and completed_at are lifecycle timestamps describing when each batch ran, not an axis ordering state transitions of a repeating entity. The status column (ready/running/completed) records the final outcome, not a sequence of transitions with their timestamps — there is no path through states, just attributes of the batch execution. This is a catalog of batch runs: dimension." } ```