neural-sparse-cpp is a C++ library for high-performance sparse vector similarity search, developed as part of the OpenSearch Project. It provides multiple index types for nearest neighbor search over sparse vectors.
Key features include:
- Multiple index types: inverted index, SEISMIC, and SEISMIC with scalar quantization
- Scalar quantization support for reduced memory usage
- SIMD-optimized distance computations (AVX2, AVX512, NEON, SVE)
- ID mapping and ID selector filtering
- Index serialization (save/load)
- Python bindings via SWIG
The library is licensed under the Apache v2.0 License and is part of the OpenSearch Software Foundation, a project of The Linux Foundation.
- Project Website
- OpenSearch Software Foundation
- Downloads
- Documentation
- Need help? Try Forums
- Talk to other developers? Slack
- Communications
- Project Principles
- Contributing to OpenSearch
- Proposing Features
- Onboarding Guide
- Maintainer Responsibilities
- Release Management
- Organization Admins
- Repo Maintainers
- Issue Triage
- Security
The OpenSearch Project style guidelines and OpenSearch terms documents provide style standards and terminology to be observed when creating OpenSearch Project content.
This project has adopted the Amazon Open Source Code of Conduct. For more information see the Code of Conduct FAQ, or contact conduct@opensearch.foundation with any additional questions or comments.
This project is licensed under the Apache v2.0 License.
Copyright OpenSearch Contributors. See NOTICE for details.