VecFlow/vecflow/pyproject.toml at main · Supercomputing-System-AI-Lab/VecFlow · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
# SPDX-License-Identifier: Apache-2.0
#
# Modern PEP 517 packaging for the VecFlow Python wrapper. Drives the CMake
# build via scikit-build-core so a single `pip install .` produces a wheel
# that includes the pybind11 module compiled against the conda-installed
# libcuvs-vecflow. Supersedes setup.py — pyproject.toml takes priority when
# both exist.

[build-system]
requires = [
    "scikit-build-core[pyproject]>=0.10",
    "pybind11>=2.11",
]
build-backend = "scikit_build_core.build"

[project]
name = "vecflow"
description = "VecFlow: A High-Performance Vector Data Management System for Filtered-Search on GPUs"
readme = "README.md"
requires-python = ">=3.11"
license = { text = "Apache-2.0" }
authors = [
    { name = "SSAIL lab, University of Illinois at Urbana-Champaign" },
]
dynamic = ["version"]
dependencies = [
    "numpy>=1.19",
]
classifiers = [
    "Development Status :: 3 - Alpha",
    "Programming Language :: Python :: 3",
    "Programming Language :: Python :: 3.11",
    "Programming Language :: Python :: 3.12",
    "Programming Language :: Python :: 3.13",
    "Programming Language :: Python :: 3.14",
    "Programming Language :: C++",
    "Operating System :: POSIX :: Linux",
    "Topic :: Scientific/Engineering",
]

[project.optional-dependencies]
gpu = ["cupy-cuda12x>=10"]

[project.urls]
Homepage   = "https://supercomputing-system-ai-lab.github.io/projects/vecflow/"
Repository = "https://github.com/Supercomputing-System-AI-Lab/VecFlow"

# ── scikit-build-core driver ──────────────────────────────────────────────
# The CMake project at vecflow/CMakeLists.txt (already exists) does
# pybind11_add_module(vecflow ...) and find_package(cuvs); we just hand it off.
[tool.scikit-build]
cmake.version  = ">=3.30"
ninja.version  = ">=1.10"
build-dir      = "build/{wheel_tag}"
wheel.packages = ["vecflow"]

# Single-file source of truth: vecflow/VERSION
[tool.scikit-build.metadata.version]
provider = "scikit_build_core.metadata.regex"
input    = "VERSION"
regex    = '^(?P<value>\d+\.\d+\.\d+)$'

# When find_package(cuvs) runs inside a conda env with libcuvs-vecflow
# installed, it picks up $CONDA_PREFIX/lib/cmake/cuvs/cuvs-config.cmake
# automatically — no override needed. CPM only kicks in for source builds.
[tool.scikit-build.cmake.define]
CMAKE_VERBOSE_MAKEFILE = "ON"