```json { "classifications": [ { "column": "id", "role": "identifier", "confidence": 1.0, "reason": "Primary key with unique values per row (902 distinct in 902 rows)" }, { "column": "dataset_id", "role": "identifier", "confidence": 1.0, "reason": "Foreign key referencing datasets, used for JOINs (95 distinct values)" }, { "column": "batch_id", "role": "identifier", "confidence": 1.0, "reason": "Foreign key referencing batches, used for JOINs (96 distinct values)" }, { "column": "engine", "role": "dimension", "confidence": 1.0, "reason": "Low-cardinality categorical field (2 values: duckdb, whoosh) for grouping/filtering" }, { "column": "query_text", "role": "attribute", "confidence": 0.95, "reason": "High-cardinality free-form text content (767 distinct values), displayed but not aggregated" }, { "column": "description", "role": "attribute", "confidence": 1.0, "reason": "Descriptive free-form text (824 distinct values), displayed but not grouped or aggregated" }, { "column": "model_name", "role": "dimension", "confidence": 1.0, "reason": "Low-cardinality categorical field (2 values) for grouping/filtering by AI model" }, { "column": "latency_ms", "role": "measure", "confidence": 1.0, "reason": "Numeric quantity (0-14872ms) to be aggregated with sum/avg for performance analysis" }, { "column": "created_at", "role": "dimension", "confidence": 1.0, "reason": "Timestamp for temporal grouping, filtering, and time-series analysis" } ] } ```