Databricks provider

providers/databricks.yml contains a top-level provider mapping.

Field	Type	Default	Description
cloud	string	required	azure, aws, or gcp
host	URL	none	Workspace origin
host_env	string	none	Environment variable containing the workspace origin
auth_type	string	pat	pat, databricks-cli, oauth, or azure-managed-identity
token_env	string	none	Environment variable containing a PAT
execution_mode	string	state-only	state-only, live-readonly, or live-apply
connect_timeout_seconds	integer	10	HTTP connection timeout
account_id	string	none	Databricks account identifier for account-level operations
sql_warehouse_id	string	none	Direct SQL warehouse ID
sql_warehouse_id_env	string	none	Environment variable containing the warehouse ID
credential_mode	string	none	Descriptive credential mode

Prefer host_env, sql_warehouse_id_env, and token_env over literal secrets or environment-specific identifiers. If both host and host_env are set, DataMuru uses host unless it is empty or still contains the starter placeholder.

See Authentication and Execution modes.

Metadata inventory

datamuru inventory export --out .datamuru/metadata-inventory.json reuses read-only import discovery to export Databricks catalogs and schemas with stable provider-neutral IDs. For each schema it uses paginated Unity Catalog table listing and read-only table detail calls to emit provider-native tables, views, and columns. List pagination uses the provider maximum of 50 results, stops at the relation-budget sentinel, and omits columns, properties, and usernames. Only relation type, column position, type text, and nullability are retained. It excludes system objects by default, stops above 1,000 relations or 10,000 columns, and does not discover grants or lineage. The command never invokes apply, destroy, data reads, or state mutation.