Exact similarity search over large collections of data series is a fundamental operation in modern applications, yet existing solutions are often fragmented, specialized, or tailored to specific execution environments. We present DaiSy (Data series similarity Search library), a unified library for exact data series similarity search that integrates multiple state-of-the-art algorithms within a single, coherent framework. DaiSy is the first library to support exact similarity search across diverse execution environments, including implementations for disk-based, in-memory, GPU-accelerated, and distributed scalable similarity search. The library supports interfaces in both C++ and Python, enabling researchers and practitioners to easily integrate its functionality in a variety of tasks.
ALPHA VERSION: Currently, DaiSy is experimental. The library is still under active development. We welcome suggestions and bug reports.
When using DaiSy, please consider cite the following artifacts:
@misc{daisy2026,
title = {{The DaiSy Library for Fast and Exact, Data Series and Vector Similarity Search}},
author ={Francesca Del Gaudio, Manos Chatzakis, Gayathiri Ravendirane, Botao Peng, Panagiota Faatourou, Themis Palpanas},
year = {2026},
howpublished = {\url{https://github.com/mchatzakis/daisy}}
}We currently support several algorithms for exact similarity search, each optimized for specific use cases and environments. The following table summarizes the key features of each algorithm:
| Algorithm | Description |
|---|---|
| Bruteforce | Naive parallel similarity search implementation |
| Lower Bound Bruteforce | Optimized bruteforce with lower bounding for the distance calculations |
| MESSI | In-memory parallel similarity search |
| PARIS | Disk-based parallel similarity search |
| SING | GPU-accelerated in-memory parallel similarity search |
| Odyssey | Distributed and parallel in-memory similarity search |
- Operating System: Linux, macOS, or Windows
- C++ Compiler: C++14 or higher (GCC 6+, Clang 3.4+, MSVC 2015+)
- CMake: Version 3.15 or higher
Optionally,
- Python: 3.10-3.12
- MPI: Required for Odyssey distributed computing algorithm
- CUDA: Required for SING GPU acceleration algorithm
To download DaiSy, use:
git clone https://github.com/MChatzakis/daisy.git
cd daisy
git submodule update --init --recursiveBased on the available hardware, you can specify the below arguments to enable/disable features.
| Flag | Description | Default | Dependencies |
|---|---|---|---|
BUILD_PYTHON |
Enable Python bindings | OFF |
Python 3.10+ |
BUILD_BENCHMARK |
Build benchmarking tools | OFF |
GoogleBenchmark |
BUILD_TESTS |
Build test suite | OFF |
GoogleTest |
BUILD_DEMO |
Build demonstration applications | ON |
Core library |
BUILD_ODYSSEY |
Enable MPI for distributed computing | OFF |
OpenMPI/MPICH |
BUILD_SING |
Enable CUDA for GPU acceleration | OFF |
CUDA Toolkit |
DEBUG_MSG |
Enable debug output | OFF |
None |
To compile:
mkdir build && cd build
cmake ..
makeIf you intent to use only the Python interface, you can install the library directly from PyPI using pip:
pip install daisy-exact-searchIf you want to use Odyssey, you will need to install mpi:
pip install daisy-exact-search[mpi]Kindly note that we are aware for compatibility issues related to ARM processors (e.g., Apple MX processors). Due to pthread-barriers and SIMD being unavailable on ARM, we currently noticing compilations failling on ARM machines. We are currently working on possible solutions, however we recommend using DaiSy on non-ARM machines for the time being.
We provide several usage examples in both C++ and Python under demos/, demonstrating how to utilize the library for various similarity search tasks.
We provide several troubleshooting guides and extra resources in the docs/ directory.
In this directory, we also provide useful information about how to contribute to the project, and how to implement new algorithms.
Work supported by ΥΠ AI ΘΑ & NextGenerationEU project HARSH (ΥΠ3ΤΑ-0560901).
The logo of DaiSy was designed by Eva Chamilaki.