If you need information about the schema of your CSV, JSON, and Parquet data, you can use the NOSREAD_SCHEMA return type in your READ_NOS statement and the schema is detected automatically. The syntax is RETURNTYPE('NOSREAD_SCHEMA').
For Parquet data you can also continue to use the return type NOSREAD_PARQUET_SCHEMA if you are using the FULLSCAN option, as FULLSCAN is not available in NOSREAD_SCHEMA.
See Teradata Vantage™ - SQL Operators and User-Defined Functions, B035-1210, available at https://docs.teradata.com/.