Data stream processing toolkit (dspt) is intended for processing huge amounts (gigabytes) of data in an efficient manner. The structure of the data files is user-defined, and the queries are written in a declarative style. Currently included algorithms range from simple filtering on predicates to aggregation and sorting. It also includes some convenience classes (e.g. for accessing BerkeleyDB databases). The toolkit does not offer as wide a range of operations as an RDBMS, but some measurements of simpler queries have shown it to be more than 10x faster than PostgreSQL.