DuckDB

DuckDB is an in-process SQL OLAP database management system designed to be embedded in applications and analytics workflows. It runs inside the host process as a library or a single-file binary, so there is no separate database server to manage. DuckDB uses a columnar storage layout and vectorized execution to accelerate analytical queries and provides ACID-compliant transactions with support for standard SQL.

The architecture emphasizes analytical performance and ease of integration. It features a columnar storage engine with automatic parallelism, a modern query planner, and a code-generating execution path (using LLVM) to accelerate expressions and operators. DuckDB supports common analytical capabilities such as window functions, joins, and aggregations, along with efficient access to data formats. It can operate in-memory or combine in-memory processing with on-disk storage and supports durable transactions.

DuckDB offers broad language and ecosystem bindings to facilitate use in data science and applications. It provides bindings for Python (duckdb-python), R, Java, and C++, and is commonly used in Jupyter notebooks and data-analysis pipelines. It can read Parquet, CSV, and other formats directly, enabling queries over external data without a separate import step, and it can return results to familiar data structures in pandas and R data frames.

Open source and licensing information. DuckDB is released under an open-source license (MIT) and has been developed with contributions from a growing community. It is widely adopted for embedded analytics, data exploration, ETL tasks, and ad-hoc analysis within applications and research workflows, offering a compact, easily embeddable solution for analytical workloads.