The generator CLI

dbt-4ps-generator reads a bc2adls CDM manifest and writes dbt staging models. Install it with uv tool install dbt-4ps-generator, run it ephemerally with uvx dbt-4ps-generator, or from a checkout with uv run dbt-4ps-generator inside dbt_4ps_staging_generator/.

generate

dbt-4ps-generator generate \
  --manifest _cdm/deltas.manifest.cdm.json \
  --output-directory example_dbt_project/models/staging/4ps

For every entity in the manifest it writes:

• stg_4ps__<table>.sql — a SELECT ... FROM STREAM READ_FILES(...) model with the column types from the CDM metadata and columns renamed to snake_case (CreditLimitLCY-20 → credit_limit_lcy)
• _4ps__models.yaml — model + column descriptions, and a dbt_utils.unique_combination_of_columns test on each table's primary key

Option	Default	Purpose
--manifest	required	Path to deltas.manifest.cdm.json; entity files are resolved relative to it
--output-directory	required	Where the .sql models and schema yaml are written
--model-prefix	stg_4ps__	Prefix for generated model names
--schema-name	models	Base name of the schema yaml
--schema-prefix	_4ps__	Prefix for the schema yaml
--file-metadata-column / --no-file-metadata-column	on	Add the _metadata column (source file info) to every model

download

dbt-4ps-generator download \
  --volume-path /Volumes/<catalog>/<schema>/<volume> \
  --output-directory ./cdm

Downloads all .json files (the manifest and entity definitions) from the Unity Catalog volume bc2adls exports to. Authentication uses the Databricks SDK's default chain: set DATABRICKS_HOST and DATABRICKS_TOKEN, or use a profile from ~/.databrickscfg (DATABRICKS_CONFIG_PROFILE).

Type mapping

CDM dataFormat values map to Databricks SQL types:

CDM	SQL
String, Guid, Time, Duration	string
Int32	int
Int64	bigint
Decimal	decimal(precision,scale) from the CDM traits, falling back to string when precision is missing
Boolean	boolean
Date	date
DateTime, DateTimeOffset	timestamp

Unknown formats fall back to string.