```json { "table_name": "pipeline_stage_status", "table_type_tags": [ {"tag": "event", "confidence": 0.90}, {"tag": "path", "confidence": 0.30} ], "primary_type": "event", "insufficient_evidence": false, "entity": { "columns": ["dataset_id", "batch_id"], "unique_per_row": false }, "time_axis": { "columns": ["started_at"], "kind": "event_time" }, "reasoning": "The composite entity (dataset_id, batch_id) repeats across rows (dataset_id RPUV 12.83, batch_id RPUV 6.95), ordered by started_at (Uniqueness 1.000). Each row records a pipeline stage execution — a discrete action (download, bronze, classify, etc.) that happened to the batch. The rows are bare occurrences to count and analyze, not state transitions that replay into a meaningful entity state. The id column (Uniqueness 1.000) is an artificial occurrence identifier, not the entity. Event is the primary reading; path receives minor confidence because the rows do chronicle a batch's progression through stages, though no tracked balance accumulates." } ```