Implementation:Sdv dev SDV Metadata Detect From Dataframes

Knowledge Sources	SDV SDV Documentation
Domains	Data_Science, Schema_Inference
Last Updated	2026-02-14 00:00 GMT

Overview

Concrete tool for automatically detecting metadata schemas from pandas DataFrames, provided by the SDV library.

Description

The Metadata.detect_from_dataframes class method analyzes a dictionary of DataFrames and produces a Metadata object with inferred column sdtypes, primary keys, and optionally foreign key relationships. It iterates over each table, detects column types using heuristic rules, and then optionally runs relationship detection across tables using column name matching.

A companion method Metadata.detect_from_dataframe handles single-table detection.

Usage

Import the Metadata class and call detect_from_dataframes when you have one or more DataFrames and need to create a metadata schema automatically. Use detect_from_dataframe for single-table workflows.

Code Reference

Source Location

• Repository: SDV
• File: sdv/metadata/metadata.py
• Lines: L103-143 (detect_from_dataframes), L146-180 (detect_from_dataframe)

Signature

@classmethod
def detect_from_dataframes(
    cls,
    data,
    infer_sdtypes=True,
    infer_keys='primary_and_foreign',
    foreign_key_inference_algorithm='column_name_match',
):
    """Detect the metadata for all tables in a dictionary of dataframes.

    Args:
        data (dict):
            Dictionary of table names to dataframes.
        infer_sdtypes (bool):
            Whether to infer the sdtypes of each column. Defaults to True.
        infer_keys (str):
            Whether to infer primary and/or foreign keys. Options:
            'primary_and_foreign', 'primary_only', None. Defaults to 'primary_and_foreign'.
        foreign_key_inference_algorithm (str):
            Algorithm for detecting foreign keys. Currently only 'column_name_match'.

    Returns:
        Metadata: A new metadata object with detected sdtypes.
    """

Import

from sdv.metadata import Metadata

I/O Contract

Inputs

Name	Type	Required	Description
data	dict[str, pd.DataFrame]	Yes	Dictionary mapping table names to DataFrames
infer_sdtypes	bool	No	Auto-detect column types (default: True)
infer_keys	str or None	No	Key inference mode: 'primary_and_foreign', 'primary_only', or None
foreign_key_inference_algorithm	str	No	FK detection algorithm (default: 'column_name_match')

Outputs

Name	Type	Description
return value	Metadata	Metadata instance with tables, columns, sdtypes, keys, and relationships

Usage Examples

Multi Table Detection

from sdv.metadata import Metadata
import pandas as pd

# Prepare data as dict of DataFrames
data = {
    'users': pd.DataFrame({
        'user_id': [1, 2, 3],
        'name': ['Alice', 'Bob', 'Charlie'],
        'age': [25, 30, 35]
    }),
    'orders': pd.DataFrame({
        'order_id': [101, 102, 103],
        'user_id': [1, 2, 1],
        'amount': [50.0, 75.0, 100.0]
    })
}

# Auto-detect metadata including relationships
metadata = Metadata.detect_from_dataframes(data)
print(metadata)

Single Table Detection

from sdv.metadata import Metadata
import pandas as pd

df = pd.DataFrame({
    'id': [1, 2, 3],
    'category': ['A', 'B', 'A'],
    'value': [10.5, 20.3, 30.1]
})

metadata = Metadata.detect_from_dataframe(df)
print(metadata)

Related Pages

Implements Principle

• Principle:Sdv_dev_SDV_Metadata_Detection

Requires Environment

• Environment:Sdv_dev_SDV_Python_Runtime