Forward-merge release/25.12 into main

.devcontainer/Dockerfile

@@ -1,3 +1,6 @@
+# SPDX-FileCopyrightText: Copyright (c) 2023-2025, NVIDIA CORPORATION. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
 # syntax=docker/dockerfile:1.5
 
 ARG BASE
@@ -42,6 +45,7 @@ ENV HISTFILE="/home/coder/.cache/._bash_history"
 ENV AWS_ROLE_ARN="arn:aws:iam::279114543810:role/nv-gha-token-sccache-devs"
 ENV SCCACHE_REGION="us-east-2"
 ENV SCCACHE_BUCKET="rapids-sccache-devs"
+ENV SCCACHE_S3_USE_PREPROCESSOR_CACHE_MODE=true
 ENV SCCACHE_IDLE_TIMEOUT=0
 
 ###

.github/workflows/build.yaml

@@ -39,8 +39,10 @@ jobs:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
       date: ${{ inputs.date }}
+      node_type: cpu16
       script: ci/build_cpp.sh
       sha: ${{ inputs.sha }}
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
 
   rocky8-clib-standalone-build:
     secrets: inherit
@@ -63,6 +65,7 @@ jobs:
       artifact-name: "libcuvs_c_${{ matrix.cuda_version }}.tar.gz"
       file_to_upload: "libcuvs_c.tar.gz"
       sha: ${{ inputs.sha }}
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   rust-build:
     needs: cpp-build
     secrets: inherit
@@ -84,6 +87,7 @@ jobs:
       node_type: "gpu-l4-latest-1"
       script: "ci/build_rust.sh"
       sha: ${{ inputs.sha }}
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   go-build:
     needs: cpp-build
     secrets: inherit
@@ -127,6 +131,7 @@ jobs:
       artifact-name: "cuvs-java-cuda${{ matrix.cuda_version }}"
       file_to_upload: "java/cuvs-java/target/"
       sha: ${{ inputs.sha }}
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   python-build:
     needs: [cpp-build]
     secrets: inherit
@@ -137,6 +142,7 @@ jobs:
       date: ${{ inputs.date }}
       script: ci/build_python.sh
       sha: ${{ inputs.sha }}
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   upload-conda:
     needs: [cpp-build, python-build]
     secrets: inherit
@@ -173,11 +179,13 @@ jobs:
       branch: ${{ inputs.branch }}
       sha: ${{ inputs.sha }}
       date: ${{ inputs.date }}
+      node_type: cpu16
       script: ci/build_wheel_libcuvs.sh
       # build for every combination of arch and CUDA version, but only for the latest Python
       matrix_filter: group_by([.ARCH, (.CUDA_VER|split(".")|map(tonumber)|.[0])]) | map(max_by(.PY_VER|split(".")|map(tonumber)))
       package-name: libcuvs
       package-type: cpp
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   wheel-publish-libcuvs:
     needs: wheel-build-libcuvs
     secrets: inherit
@@ -198,8 +206,10 @@ jobs:
       branch: ${{ inputs.branch }}
       sha: ${{ inputs.sha }}
       date: ${{ inputs.date }}
+      node_type: cpu8
       script: ci/build_wheel_cuvs.sh
       package-name: cuvs
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
       package-type: python
   wheel-publish-cuvs:
     needs: wheel-build-cuvs

.github/workflows/pr.yaml

@@ -144,6 +144,7 @@ jobs:
       build_type: pull-request
       node_type: cpu16
       script: ci/build_cpp.sh
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   conda-cpp-tests:
     needs: [conda-cpp-build, changed-files]
     secrets: inherit
@@ -152,6 +153,7 @@ jobs:
     with:
       build_type: pull-request
       script: ci/test_cpp.sh
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   conda-cpp-checks:
     needs: conda-cpp-build
     secrets: inherit
@@ -174,6 +176,7 @@ jobs:
     with:
       build_type: pull-request
       script: ci/test_python.sh
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   rocky8-clib-standalone-build:
     needs: [checks]
     secrets: inherit
@@ -187,14 +190,15 @@ jobs:
     with:
       build_type: pull-request
       arch: "amd64"
-      date: ${{ inputs.date }}_c
+      date: ${{ inputs.date }}
       container_image: "rapidsai/ci-wheel:25.12-cuda${{ matrix.cuda_version }}-rockylinux8-py3.10"
       node_type: "cpu16"
       # requires_license_builder: false
       script: "ci/build_standalone_c.sh --build-tests"
       artifact-name: "libcuvs_c_${{ matrix.cuda_version }}.tar.gz"
       file_to_upload: "libcuvs_c.tar.gz"
       sha: ${{ inputs.sha }}
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   rocky8-clib-tests:
     needs: [rocky8-clib-standalone-build, changed-files]
     secrets: inherit
@@ -210,7 +214,7 @@ jobs:
       build_type: pull-request
       node_type: "gpu-l4-latest-1"
       arch: "amd64"
-      date: ${{ inputs.date }}_c
+      date: ${{ inputs.date }}
       container_image: "rapidsai/ci-wheel:25.12-cuda${{ matrix.cuda_version }}-rockylinux8-py3.10"
       script: "ci/test_standalone_c.sh"
       sha: ${{ inputs.sha }}
@@ -235,6 +239,7 @@ jobs:
       script: "ci/test_java.sh"
       artifact-name: "cuvs-java-cuda${{ matrix.cuda_version }}"
       file_to_upload: "java/cuvs-java/target/"
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   rust-build:
     needs: [conda-cpp-build, changed-files]
     secrets: inherit
@@ -254,6 +259,7 @@ jobs:
       arch: "amd64"
       container_image: "rapidsai/ci-conda:25.12-cuda${{ matrix.cuda_version }}-ubuntu24.04-py3.13"
       script: "ci/build_rust.sh"
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   go-build:
     needs: [conda-cpp-build, changed-files]
     secrets: inherit
@@ -289,20 +295,24 @@ jobs:
     uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@release/25.12
     with:
       build_type: pull-request
+      node_type: cpu16
       script: ci/build_wheel_libcuvs.sh
       # build for every combination of arch and CUDA version, but only for the latest Python
       matrix_filter: group_by([.ARCH, (.CUDA_VER|split(".")|map(tonumber)|.[0])]) | map(max_by(.PY_VER|split(".")|map(tonumber)))
       package-name: libcuvs
       package-type: cpp
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   wheel-build-cuvs:
     needs: wheel-build-libcuvs
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@release/25.12
     with:
       build_type: pull-request
+      node_type: cpu8
       script: ci/build_wheel_cuvs.sh
       package-name: cuvs
       package-type: python
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   wheel-tests-cuvs:
     needs: [wheel-build-cuvs, changed-files]
     secrets: inherit
@@ -311,6 +321,7 @@ jobs:
     with:
       build_type: pull-request
       script: ci/test_wheel_cuvs.sh
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   devcontainer:
     secrets: inherit
     needs: telemetry-setup

.github/workflows/test.yaml

@@ -41,6 +41,7 @@ jobs:
       date: ${{ inputs.date }}
       script: ci/test_cpp.sh
       sha: ${{ inputs.sha }}
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   conda-python-tests:
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@release/25.12
@@ -50,6 +51,7 @@ jobs:
       date: ${{ inputs.date }}
       script: ci/test_python.sh
       sha: ${{ inputs.sha }}
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   conda-java-tests:
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@release/25.12
@@ -70,6 +72,7 @@ jobs:
       arch: "amd64"
       container_image: "rapidsai/ci-conda:25.12-cuda${{ matrix.cuda_version }}-ubuntu24.04-py3.13"
       script: "ci/test_java.sh"
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   wheel-tests-cuvs:
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@release/25.12
@@ -79,3 +82,4 @@ jobs:
       date: ${{ inputs.date }}
       sha: ${{ inputs.sha }}
       script: ci/test_wheel_cuvs.sh
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN

README.md

@@ -1,7 +1,5 @@
 # <div align="left"><img src="https://rapids.ai/assets/images/rapids_logo.png" width="90px"/>&nbsp;cuVS: Vector Search and Clustering on the GPU</div>
 
-> [!note]
-> cuVS is a new library mostly derived from the approximate nearest neighbors and clustering algorithms in the [RAPIDS RAFT](https://github.com/rapidsai/raft) library of machine learning and data mining primitives. As of version 24.10 (Release in October 2024), cuVS contains the most fully-featured versions of the approximate nearest neighbors and clustering algorithms from RAFT. The algorithms which have been migrated over to cuVS will be removed from RAFT in version 24.12 (released in December 2024).
 
 ## Contents
 
@@ -67,7 +65,7 @@ There are several benefits to using cuVS and GPUs for vector search, including
 6. Multiple language support
 7. Building blocks for composing new or accelerating existing algorithms
 
-In addition to the items above, cuVS takes on the burden of keeping non-trivial accelerated code up to date as new NVIDIA architectures and CUDA versions are released. This provides a delightful development experience, guaranteeing that any libraries, databases, or applications built on top of it will always be getting the best performance and scale.
+In addition to the items above, cuVS shoulders the burden of keeping non-trivial accelerated code up to date as new NVIDIA architectures and CUDA versions are released. This provides a delightful development experience, guaranteeing that any libraries, databases, or applications built on top of it will always be getting the best performance and scale.
 
 ## cuVS Technology Stack
 
@@ -93,7 +91,7 @@ It is recommended to use [mamba](https://conda.github.io/conda-libmamba-solver/u
 conda install -c rapidsai -c conda-forge cuvs
 ```
 
-The cuVS Python package can also be `installed through pip <https://docs.rapids.ai/install#pip>`_.
+The cuVS Python package can also be installed through [pip](https://docs.rapids.ai/install#pip>).
 
 ```bash
 # CUDA 13
@@ -114,7 +112,10 @@ conda install -c rapidsai-nightly -c conda-forge cuvs=25.12 cuda-version=13.0
 conda install -c rapidsai-nightly -c conda-forge cuvs=25.12 cuda-version=12.9
 ```
 
-cuVS also has `pip` wheel packages that can be installed. Please see the [Build and Install Guide](https://docs.rapids.ai/api/cuvs/nightly/build/) for more information on installing the available cuVS packages and building from source.
+> [!NOTE]
+> If compiled binary size is a concern, please note that the cuVS builds for CUDA 13 are roughly half the size of CUDA 12 builds. This is a result of improved compression rates in the newer supported CUDA drivers. We will be adopting the newer drivers for CUDA 12 builds in Spring of 2026, which will ultimately bring them down to roughly the size of the CUDA 13 builds. In the meantime, the NVIDIA cuVS team is continuing to shave down the binary sizes for all supported CUDA versions. If binary size is an issue for you, please consider linking to cuVS statically either by building from source or using pre-built `libcuvs-static` conda package.
+
+Please see the [Build and Install Guide](https://docs.rapids.ai/api/cuvs/nightly/build/) for more information on installing the available cuVS packages and building from source.
 
 ## Getting Started
 

c/CMakeLists.txt

@@ -33,7 +33,11 @@ option(CUVSC_STATIC_CUVS_LIBRARY "Link against statical version of libcuvs" OFF)
 # Check if cuVS is already available. If so, it is the user's responsibility to ensure that the
 # CMake package is also available at build time of the Python cuvs package.
 if(NOT TARGET cuvs::cuvs)
-  find_package(cuvs "${RAPIDS_VERSION}" REQUIRED)
+  if(CUVSC_STATIC_CUVS_LIBRARY)
+    find_package(cuvs "${RAPIDS_VERSION}" REQUIRED COMPONENTS cuvs_static)
+  else()
+    find_package(cuvs "${RAPIDS_VERSION}" REQUIRED COMPONENTS cuvs_shared)
+  endif()
 else()
   set(BUILDING_FROM_CUVS ON)
 endif()
@@ -167,40 +171,29 @@ if(PROJECT_IS_TOP_LEVEL)
   include(CPack)
 
   # Add CUDAToolkit as an export dependency
-  rapids_export_package(INSTALL CUDAToolkit cuvs-c-exports)
-  rapids_export_package(BUILD CUDAToolkit cuvs-c-exports)
+  rapids_export_package(INSTALL CUDAToolkit cuvs-exports)
+  rapids_export_package(BUILD CUDAToolkit cuvs-exports)
 
   install(
     TARGETS cuvs_c
     DESTINATION ${lib_dir}
-    COMPONENT cuvs_c
+    COMPONENT c_api
     EXPORT cuvs-c-exports
   )
-
-  rapids_export(
-    INSTALL cuvs_c
-    VERSION "${RAPIDS_VERSION}"
-    EXPORT_SET cuvs-c-exports
-    GLOBAL_TARGETS cuvs_c
-    NAMESPACE cuvs::
-  )
-  rapids_export(
-    BUILD cuvs_c
-    VERSION "${RAPIDS_VERSION}"
-    EXPORT_SET cuvs-c-exports
-    GLOBAL_TARGETS cuvs_c
-    NAMESPACE cuvs::
-  )
   install(
     DIRECTORY include/cuvs
-    COMPONENT cuvs_c
+    COMPONENT c_api
     DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
   )
   install(
     FILES ${CMAKE_CURRENT_BINARY_DIR}/include/cuvs/core/c_config.h
-    COMPONENT cuvs_c
+    COMPONENT c_api
     DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/cuvs/core/
   )
+
+  include(../cpp/cmake/modules/generate_cuvs_export.cmake)
+  generate_cuvs_export(CLIB COMPONENTS c_api EXPORT_SETS cuvs-c-exports)
+
 endif()
 
 # ##################################################################################################

ci/build_cpp.sh

@@ -5,7 +5,6 @@
 set -euo pipefail
 
 source rapids-configure-sccache
-
 source rapids-date-string
 
 export CMAKE_GENERATOR=Ninja
@@ -14,8 +13,7 @@ rapids-print-env
 
 rapids-logger "Begin cpp build"
 
-sccache --zero-stats
-
+sccache --stop-server 2>/dev/null || true
 
 RAPIDS_PACKAGE_VERSION=$(rapids-generate-version)
 export RAPIDS_PACKAGE_VERSION
@@ -35,6 +33,7 @@ rattler-build build --recipe conda/recipes/libcuvs \
                     "${RATTLER_CHANNELS[@]}"
 
 sccache --show-adv-stats
+sccache --stop-server >/dev/null 2>&1 || true
 
 # remove build_cache directory
 rm -rf "$RAPIDS_CONDA_BLD_OUTPUT_DIR"/build_cache

ci/build_java.sh

@@ -4,6 +4,12 @@
 
 set -euo pipefail
 
+source rapids-configure-sccache
+
+export SCCACHE_S3_KEY_PREFIX="cuvs-java/${RAPIDS_CONDA_ARCH}/cuda${RAPIDS_CUDA_VERSION%%.*}/maven/objects-cache"
+export SCCACHE_S3_PREPROCESSOR_CACHE_KEY_PREFIX="cuvs-java/${RAPIDS_CONDA_ARCH}/cuda${RAPIDS_CUDA_VERSION%%.*}/maven/preprocessor-cache"
+export SCCACHE_S3_USE_PREPROCESSOR_CACHE_MODE=true
+
 # TODO: Remove this argument-handling when build and test workflows are separated,
 #       and test_java.sh no longer calls build_java.sh
 #       ref: https://github.com/rapidsai/cuvs/issues/868
@@ -49,5 +55,8 @@ export RAPIDS_CUDA_MAJOR
 
 bash ./build.sh java "${EXTRA_BUILD_ARGS[@]}"
 
+sccache --show-adv-stats
+sccache --stop-server >/dev/null 2>&1 || true
+
 rapids-logger "Test script exiting with value: $EXITCODE"
 exit ${EXITCODE}