```json { "table_name": "analysis_run_llm_usage", "table_type_tags": [ {"tag": "dimension", "confidence": 0.75}, {"tag": "event", "confidence": 0.30} ], "primary_type": "dimension", "insufficient_evidence": false, "entity": { "columns": ["jobs_id", "model_name"], "unique_per_row": true }, "time_axis": null, "reasoning": "The composite entity (jobs_id, model_name) is unique per row — jobs_id RPUV 2.00 and model_name RPUV 45.20 combine to make each row distinct (113 jobs × ~2 models = 226 rows). Each row describes the LLM usage summary for one model within one analysis job: calls, tokens, cost. The _ingestion_timestamp is a provenance column (constant across all rows). No event timestamp or sequence orders actions; the row exists to catalog what happened during the job, not to record when it happened. Dimension." } ```