{
    "slots": [
        {
            "scope": "dataset",
            "slot": "nature",
            "value": {
                "nature": "observational"
            },
            "evidence": {
                "reasoning": "Rows are physical flower measurements (sepal and petal dimensions) with a species label, taken from the classic Iris dataset published via seaborn-data. No business process is operated here."
            }
        },
        {
            "scope": "dataset",
            "slot": "decisions_served",
            "value": {
                "decisions": [
                    "Which morphological features (petal vs sepal dimensions) best separate the three species, for choosing classifier features",
                    "Whether a new specimen's measurements place it in setosa, versicolor or virginica",
                    "Which classification or clustering method to choose when benchmarking on a small, well-separated dataset"
                ]
            },
            "evidence": {
                "reasoning": "Four continuous measures and a 3-level nominal species column over 150 rows make this a classification and feature-selection dataset. The gold rollup groups the measures by species."
            }
        },
        {
            "scope": "dataset",
            "slot": "levers",
            "value": {
                "levers": []
            },
            "evidence": {
                "reasoning": "The data records measured plant traits that no operator controls, so there is nothing to pull. The _ingestion_* and _batch_id columns are pipeline metadata, not business levers."
            }
        },
        {
            "scope": "dataset",
            "slot": "audience",
            "value": {
                "audience": [
                    "data scientist",
                    "machine learning student or educator",
                    "botanist or biostatistician"
                ]
            },
            "evidence": {
                "reasoning": "Iris is a canonical teaching and benchmarking dataset, and its subject is botanical measurement."
            }
        },
        {
            "scope": "table:iris",
            "slot": "role",
            "value": {
                "role": "fact"
            },
            "evidence": {
                "reasoning": "The detector's fact role is right: each row is one measured specimen with four numeric measures. There are no related tables."
            }
        },
        {
            "scope": "table:iris",
            "slot": "grain",
            "value": {
                "unit": "one row per measured iris flower specimen",
                "key_columns": []
            },
            "evidence": {
                "reasoning": "No column is unique: the measurements repeat (sepal_length has 35 distinct values in 150 rows) and species has only 3 values. The specimen identity is implicit in row position, so there is no natural key."
            }
        },
        {
            "scope": "table:iris",
            "slot": "temporal_posture",
            "value": {
                "posture": "atemporal; the only timestamp is the ingestion time, which is a single load-time value",
                "columns": [
                    "_ingestion_timestamp"
                ]
            },
            "evidence": {
                "reasoning": "_ingestion_timestamp has 1 unique value across all rows, so it records when the load happened and not when the measurements were taken. The measurements carry no time dimension."
            }
        },
        {
            "scope": "table:iris",
            "slot": "column_bindings",
            "value": {
                "bindings": [
                    {
                        "column": "sepal_length",
                        "binding": "measure"
                    },
                    {
                        "column": "sepal_width",
                        "binding": "measure"
                    },
                    {
                        "column": "petal_length",
                        "binding": "measure"
                    },
                    {
                        "column": "petal_width",
                        "binding": "measure"
                    },
                    {
                        "column": "species",
                        "binding": "category"
                    },
                    {
                        "column": "_ingestion_timestamp",
                        "binding": "temporal"
                    },
                    {
                        "column": "_batch_id",
                        "binding": "identifier"
                    },
                    {
                        "column": "_source_file",
                        "binding": "identifier"
                    },
                    {
                        "column": "_source_system",
                        "binding": "category"
                    }
                ]
            },
            "evidence": {
                "reasoning": "_batch_id and _source_file are single-valued lineage identifiers, so identifier fits better than the detector's category. _source_system is a constant label, so category stays."
            }
        }
    ]
}