Normalized Schema Generator

An anywidget-based Jupyter widget for Microsoft Fabric notebooks that visually configures a normalized schema JSON config for Process Mining data import.

Replaces manual JSON authoring by providing dropdowns, toggles, and inline forms to select lakehouses, tables, columns, and FK/PK relationships, then auto-generates the dataSource section of the config.

Install

pip install normalized-schema-generator

Usage

In a Fabric notebook

from normalized_schema_generator import NormalizedSchemaWidget

w = NormalizedSchemaWidget()
w

The widget auto-detects the Fabric environment and populates lakehouses, tables, and columns using notebookutils and Spark.

After configuring your schema, retrieve the generated JSON:

print(w.config_json)

Local development

Outside Fabric, the widget uses mock data automatically — no notebookutils or Spark needed:

pip install normalized-schema-generator jupyterlab
jupyter lab

Features

• Lakehouse selection from the current workspace
• Data source toggle: Delta Tables or Files (CSV / Parquet)
• Events table (required): column checklist with data type badges, include/exclude toggles
• Cases table (optional): same column management, with required Events-to-Cases FK relationship
• Reference management: add/edit/remove FK joins to lookup tables, with target column selection
• Auto-generated JSON: full inputDataBinding config with dataSource, connection properties, and dataset definitions
• Copy to clipboard for the generated JSON
• Fluent 2 styling matching the Fabric design language

JSON output

The widget generates a complete normalized import config following the Process Mining Normalized Import specification:

• dataSourceSchemaType: 1 (Normalized)
• dataSourceType: 2 (OneLake)
• Proper datasets[] with Kind 0 (Event), 1 (Case), 2 (Join)
• FK columns excluded from Columns[] (exposed via Join[].ExportName)
• No nested joins on lookup datasets

License

MIT