Merge release/25.12 into main (rapidsai#1630)

Addresses open forward merger:
rapidsai#1582

---------

Co-authored-by: Vyas Ramasubramani <[email protected]>
Co-authored-by: Paul Taylor <[email protected]>
Co-authored-by: Nate Rock <[email protected]>
Co-authored-by: Lorenzo Dematté <[email protected]>
Co-authored-by: Divye Gala <[email protected]>
Co-authored-by: Victor Lafargue <[email protected]>
Co-authored-by: Corey J. Nolet <[email protected]>
Co-authored-by: Ben Frederickson <[email protected]>
Co-authored-by: Ben Frederickson <[email protected]>

Author: 9 people

.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
@@ -198,6 +201,7 @@ jobs:
       artifact-name: "libcuvs_c_${{ matrix.cuda_version }}_${{ matrix.arch }}.tar.gz"
       file_to_upload: "libcuvs_c.tar.gz"
       sha: ${{ inputs.sha }}
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   rocky8-clib-tests:
     needs: [rocky8-clib-standalone-build, changed-files]
     secrets: inherit
@@ -241,6 +245,7 @@ jobs:
       script: "ci/test_java.sh"
       artifact-name: "cuvs-java-cuda${{ 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
@@ -260,6 +265,7 @@ jobs:
       arch: "amd64"
       container_image: "rapidsai/ci-conda:26.02-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
@@ -295,20 +301,24 @@ jobs:
     uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@main
     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@main
     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
@@ -317,6 +327,7 @@ 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@main
@@ -50,6 +51,7 @@ jobs:
       date: ${{ inputs.date }}
       script: ci/test_python.sh
       sha: ${{ inputs.sha }}
+      sccache-dist-token-secret-name: GIST_REPO_READ_ORG_GITHUB_TOKEN
   conda-java-tests:
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@main
@@ -70,6 +72,7 @@ jobs:
       arch: "amd64"
       container_image: "rapidsai/ci-conda:26.02-cuda${{ 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@main
@@ -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

CHANGELOG.md

@@ -1,3 +1,100 @@
+# cuvs 25.12.00 (10 Dec 2025)
+
+### 🚨 Breaking Changes
+* Using `all_neighbors` for mutual reachability by @jinsolp in https://github.com/rapidsai/cuvs/pull/1234
+* Refactor libcuvs_c header and source locations by @robertmaynard in https://github.com/rapidsai/cuvs/pull/1357
+* Update cagra C API enums to have more long term stable values by @robertmaynard in https://github.com/rapidsai/cuvs/pull/1436
+* Require CUDA 12.2+ by @jakirkham in https://github.com/rapidsai/cuvs/pull/1476
+* Remove mutual_reachability_graph Public API by @tarang-jain in https://github.com/rapidsai/cuvs/pull/1481
+### 🐛 Bug Fixes
+* Allow compilation when OpenMP is disabled by @robertmaynard in https://github.com/rapidsai/cuvs/pull/1346
+* Deallocation should be noexcept by @bdice in https://github.com/rapidsai/cuvs/pull/1416
+* ANN_BENCH: Don't throw in noexcept do_deallocate by @achirkin in https://github.com/rapidsai/cuvs/pull/1417
+* Remove unneeded cutlass public build/install dependency by @robertmaynard in https://github.com/rapidsai/cuvs/pull/1411
+* cuvs_static properly adds C include dir to target_include_directories by @robertmaynard in https://github.com/rapidsai/cuvs/pull/1426
+* Fix C/Python serialization for FP16. Add python tests by @lowener in https://github.com/rapidsai/cuvs/pull/1429
+* Properly guard usage of openmp function calls  by @robertmaynard in https://github.com/rapidsai/cuvs/pull/1435
+* Update nlohmann-json to 3.12.0 by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1449
+* Fix test params for filtered ivf_flat by @aamijar in https://github.com/rapidsai/cuvs/pull/1463
+* [REVIEW][Java][Bug] Fix `CuVSMatrix#getRow` to take strides into consideration by @ldematte in https://github.com/rapidsai/cuvs/pull/1442
+* Adding more tests and I find the compression settings are broken in the Golang API by @maxwbuckley in https://github.com/rapidsai/cuvs/pull/1472
+* Fix binary quantizer host transform bounds and stream order by @achirkin in https://github.com/rapidsai/cuvs/pull/1473
+* [BUG] Check if dynamic batching conf is null by @tarang-jain in https://github.com/rapidsai/cuvs/pull/1459
+* Include cagra search algorithms in libcuvs_static.a by @robertmaynard in https://github.com/rapidsai/cuvs/pull/1457
+* [Java] Disable flaky test `testFloatSerialization` by @mythrocks in https://github.com/rapidsai/cuvs/pull/1503
+* Add libaio to cuvs-bench-cpu, improve bench dependencies by @tarang-jain in https://github.com/rapidsai/cuvs/pull/1480
+* [Java] Fix `CagraBuildAndSearchIT` concurrent deallocation errors by @ldematte in https://github.com/rapidsai/cuvs/pull/1510
+* Pin Cython pre-3.2.0 and PyTest pre-9 by @jakirkham in https://github.com/rapidsai/cuvs/pull/1528
+* Fix a bug in compute_distance_00_generate.py by @enp1s0 in https://github.com/rapidsai/cuvs/pull/1532
+* Check stride information in from_dlpack by @benfred in https://github.com/rapidsai/cuvs/pull/1458
+* refactored update-version.sh to handle new branching strategy by @rockhowse in https://github.com/rapidsai/cuvs/pull/1535
+* fixed bug with update-version.sh by @rockhowse in https://github.com/rapidsai/cuvs/pull/1556
+* fix(ci): remove unsupported `name` parameter from custom job def by @gforsyth in https://github.com/rapidsai/cuvs/pull/1560
+* Remove need to have rapids_logger headers installed to use clib by @robertmaynard in https://github.com/rapidsai/cuvs/pull/1527
+* Fix BruteForce serialize test by @lowener in https://github.com/rapidsai/cuvs/pull/1568
+* [C] Fix: `cuvsRMMMemoryResourceReset` sets a valid resource by @ldematte in https://github.com/rapidsai/cuvs/pull/1540
+* Fix typo in date field by @vyasr in https://github.com/rapidsai/cuvs/pull/1563
+* SNMG ANN build with OpenMP nested parallelism by @viclafargue in https://github.com/rapidsai/cuvs/pull/1526
+### 📖 Documentation
+* Fix table of content documentation by @lowener in https://github.com/rapidsai/cuvs/pull/1427
+* Docs Spectral Clustering by @aamijar in https://github.com/rapidsai/cuvs/pull/1490
+* Use current system architecture in conda environment creation command by @bdice in https://github.com/rapidsai/cuvs/pull/1499
+* Updating README for release by @cjnolet in https://github.com/rapidsai/cuvs/pull/1584
+### 🚀 New Features
+* Unify binding headers by @robertmaynard in https://github.com/rapidsai/cuvs/pull/1413
+* CAGRA: decouple source idx type from graph idx type and add a mapping between them by @achirkin in https://github.com/rapidsai/cuvs/pull/1251
+* Add SOVERSION information to libcuvs_c by @robertmaynard in https://github.com/rapidsai/cuvs/pull/1474
+* Spectral Clustering by @aamijar in https://github.com/rapidsai/cuvs/pull/1425
+* ANN_BENCH: integrate NVTX statistics by @achirkin in https://github.com/rapidsai/cuvs/pull/1529
+* Add Augmented Core Extraction Algorithm by @julianmi in https://github.com/rapidsai/cuvs/pull/1404
+* Extend CI to build and test x86 libcuvs_c tarballs by @robertmaynard in https://github.com/rapidsai/cuvs/pull/1524
+* Introduce `libcuvs-headers` and `libcuvs-static` CMake components and conda packages by @divyegala in https://github.com/rapidsai/cuvs/pull/1494
+* [Java] Bindings, tests and benchmarks for RMM pooled memory by @ldematte in https://github.com/rapidsai/cuvs/pull/1453
+### 🛠️ Improvements
+* Update `RAPIDS_BRANCH`, codify changes in `update-version.sh` by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1368
+* Move eigen_solvers from raft by @aamijar in https://github.com/rapidsai/cuvs/pull/1402
+* CosineExpanded Distance Metric for CAGRA by @tarang-jain in https://github.com/rapidsai/cuvs/pull/197
+* Enable `sccache-dist` connection pool by @trxcllnt in https://github.com/rapidsai/cuvs/pull/1431
+* Use pinned_host_memory_resource instead of pinned_memory_resource. by @bdice in https://github.com/rapidsai/cuvs/pull/1434
+* Use main in RAPIDS_BRANCH by @bdice in https://github.com/rapidsai/cuvs/pull/1439
+* Use main shared-workflows branch by @bdice in https://github.com/rapidsai/cuvs/pull/1444
+* [Review] ScaNN: Add option for AVQ/Noise Shaping to bfloat16 quantization by @rmaschal in https://github.com/rapidsai/cuvs/pull/1354
+* Add NVTX annotations to CAGRA knn graph build stage by @achirkin in https://github.com/rapidsai/cuvs/pull/1443
+* [Review][C] Export the ability to get/set the log level to the C API by @ldematte in https://github.com/rapidsai/cuvs/pull/1375
+* [Java] Test indexing and serialization with integral (byte) dataset by @ldematte in https://github.com/rapidsai/cuvs/pull/1366
+* Use SPDX for all copyright headers by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1446
+* [REVIEW][Java] Log level api by @ldematte in https://github.com/rapidsai/cuvs/pull/1376
+* [Review] [Java] Disable flaky Java tests to reduce CI churn by @mythrocks in https://github.com/rapidsai/cuvs/pull/1469
+* Refactor rapids_cpm_package_details to rapids_cpm_package_info by @bdice in https://github.com/rapidsai/cuvs/pull/1433
+* Fix golang test name and add more logging and error checking. by @maxwbuckley in https://github.com/rapidsai/cuvs/pull/1471
+* Use an explicit std::min for byte alignment calculation by @maxwbuckley in https://github.com/rapidsai/cuvs/pull/1465
+* Replace custom cuda_pinned_resource with RMM's pinned_host_memory_resource by @bdice in https://github.com/rapidsai/cuvs/pull/1466
+* Add patch for FAISS memory resources by @bdice in https://github.com/rapidsai/cuvs/pull/1477
+* Improved CAGRA build parameter heuristics by @achirkin in https://github.com/rapidsai/cuvs/pull/1448
+* Decouple C++ library from C library by @divyegala in https://github.com/rapidsai/cuvs/pull/1488
+* Use `RAPIDS_BRANCH` in cmake-format invocations that need rapids-cmake configs by @bdice in https://github.com/rapidsai/cuvs/pull/1475
+* eigen `tolerance` option Spectral Embedding by @aamijar in https://github.com/rapidsai/cuvs/pull/1493
+* Update to CUDA 13.0.2 by @bdice in https://github.com/rapidsai/cuvs/pull/1506
+* Single Linkage to Use all_neighbors API to build the KNN graph by @tarang-jain in https://github.com/rapidsai/cuvs/pull/1507
+* Migrate to new CCCL memory resource interface by @bdice in https://github.com/rapidsai/cuvs/pull/1502
+* Remove unused headers scann by @lowener in https://github.com/rapidsai/cuvs/pull/1508
+* Update RMM includes from `<rmm/mr/device/*>` to `<rmm/mr/*>` by @bdice in https://github.com/rapidsai/cuvs/pull/1538
+* Use ruff-check, ruff-format instead of black, flake8 by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1500
+* Set memory pool from Python multi-GPU resource by @viclafargue in https://github.com/rapidsai/cuvs/pull/1530
+* [Java] Fix format with spotless by @ldematte in https://github.com/rapidsai/cuvs/pull/1539
+* fix bad version update by @trxcllnt in https://github.com/rapidsai/cuvs/pull/1555
+* Deduplicate KMeans instantiations by @divyegala in https://github.com/rapidsai/cuvs/pull/1565
+* Improve memory usage in `build_mr_linkage` by @jinsolp in https://github.com/rapidsai/cuvs/pull/1550
+* Use `sccache-dist` build cluster for conda and wheel builds by @trxcllnt in https://github.com/rapidsai/cuvs/pull/1495
+* [Java] One PinnedMemoryBuffer per CuVSResourcesImpl by @ldematte in https://github.com/rapidsai/cuvs/pull/1441
+* [Java] Relax cuVS version matching by @ldematte in https://github.com/rapidsai/cuvs/pull/1544
+
+## New Contributors
+* @maxwbuckley made their first contribution in https://github.com/rapidsai/cuvs/pull/1471
+* @rockhowse made their first contribution in https://github.com/rapidsai/cuvs/pull/1535
+
+**Full Changelog**: https://github.com/rapidsai/cuvs/compare/v25.12.00a...release/25.12
+
 # cuvs 25.10.00 (8 Oct 2025)
 
 ## 🚨 Breaking Changes

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=26.02 cuda-version=13.0
 conda install -c rapidsai-nightly -c conda-forge cuvs=26.02 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