OneSparse brings sparse linear algebra to PostgreSQL, unlocking high-performance AI, machine learning and deep graph analysis to your relational data.
With native support for sparse graph computation, it powers advanced workloads like Graph Retrieval-Augmented Generation (GraphRAG), Graph Neural Networks (GNNs), and Sparse Distributed Representations, all using standard SQL and running efficiently on CPUs, GPUs, and next-generation sparse accelerators.
AI Models and Agent Frameworks are becoming increasingly sparse, and sparity is the future of AI. Today's dense models run up against the physical limits of current hardware quickly, fueling an ongoing paradigm shift in hardware design. OneSparse is uniquely positioned to take advantage of the newest and future sparse hardware coming from NVIDIA and AMD.
Financial transactions form edges in immense graphs. Until large scale graph API like the OneSparse came along, financial analysis was bottlenecked by the limits of SQL and low level edge-centric APIs. OneSparse can load and analyze financial graphs with hundreds of billions of edges using a simple API.
Modern security threats come from many distributed attack vectors. OneSparse has been used at MIT Lincoln Laboratories to do threat detection and response on billions of IP packets per day to secure global networks and mitigate threats.
Model output vectors represent highly compressed semantic knowledge, but associating vectors with each other and pre-trained models is a complex sparse problem. OneSparse greatly accelerates sparse vector classification and search over exisisting data without leaving your SQL database.
OneSparse is the only Postgres graph database with GAP benchmark results.
Benchmarks below were run on a 48-core AMD EPYC cloud server using graphs from the GAP Benchmark Suite, the industry standard for evaluating graph algorithm performance. Datasets are drawn from the SuiteSparse Matrix Collection, a widely used repository of real-world sparse graphs.
OneSparse brings large scale graph analysis to Postgres. Users can pick from a large library of preexisting algorithms like those used in GAP, or roll their own using our powerful Linear Algebra framework.
OneSparse's core product provides high performance sparse linear algebra on the most powerful open-source relational database in the world: PostgreSQL.
GitHub DocsThe reference implementation of the GraphBLAS API is a foundational library for all of OneSparse and many other software frameworks like Python and MATLAB. OneSparse helps support the development of SuiteSparse and contribute important use cases to its future development.
WebsiteLAGraph is a suite of advanced, high performance graph algorithms implemented with the GraphBLAS by world-class experts in sparse graph computing. These battle-tested algorithms come built into OneSparse, quickly accelerating analysis with minimal setup.
GithubPostgres is the most popular database in the world, and Python is the gold standard for ML, AI and other data science programming languages. Our team has decades of Python programming experience, and we plan very deep integration of OneSparse with the native python-graphblas library.
GithubOneSparse is powered by the SuiteSparse GraphBLAS library which is the high performance reference implementation of the GraphBLAS Organization.
Michel is a Python and Postgres developer and author of many libraries and extensions. He has decades of experience with complex data problems from large scale text search to machine learning and graph analysis. He is a member of the GraphBLAS C API Committee and long time contributor to sparse linear algebra with the GraphBLAS.
David A. Bader is a Distinguished Professor and founder of the Department of Data Science in the Ying Wu College of Computing and Director of the Institute for Data Science at the New Jersey Institute of Technology. He is a 2025 recipient of the 35 HPC Legends Award.
Dr. Timothy A. Davis is a Professor in the Computer Science and Engineering Department at Texas A&M University and a Fellow of SIAM, ACM, and IEEE. He serves as an associate editor for the ACM Transactions on Mathematical Software. In 2018, he received the Walston Chubb Award for Innovation from Sigma Xi.