1.58.0 (2026-11-19)
Snowpark Python API Updates
New Features
DataFrameReader.documents()(Private Preview) reads with aschemaor apromptnow return theAI_COMPLETEtoken usage of each document inINPUT_TOKENSandOUTPUT_TOKENScolumns. A schema field can no longer be named like one of these columns.
Improvements
to_timestamp,to_char/to_varchar,to_binaryandtry_to_binarynow accept the format as aColumnor a literalstr.to_binaryandtry_to_binarynow pass an empty-string format to Snowflake instead of silently ignoring it.
Bug Fixes
- Fixed a bug where
to_binaryandtry_to_binaryraisedTypeErrorwhen the format was aColumn. - Fixed a bug where
DataFrameReader.documents()(Private Preview) reads with a schema rejected.jpg,.jpeg,.png,.gifand.webpimages with a "file type is not supported" error.
Snowpark Local Testing Updates
Bug Fixes
- Fixed a bug where
to_binaryin local testing treated an empty-string format asHEXand failed with an unclear error for an unknown format, instead of raising an invalid binary format error as Snowflake does.