v26.08.01

Full Changelog: v26.08.00...v26.08.01

v26.08.00

What's Changed

🚨 Breaking Changes

• Auto select CAGRA build algorithm for hnsw::build by @tfeher in #1719
• Track tensor and index lifetimes in the Rust DLPack bindings by @yan-zaretskiy in #2245
• wheels: build CUDA 13 wheels with latest CTK (13.3.0) by @jameslamb in #2305
• [FEA] Multi-node Out of Core Streaming KMeans API by @tarang-jain in #2066
• Rust: builder-based params, per-module errors by @yan-zaretskiy in #2284
• [KMeans] Rename streaming_batch_size param by @tarang-jain in #2328
• Update DLPack to 1.0 by @divyegala in #2354
• Apply ABI Breaking Changes to KMeans Params (release branch) by @tarang-jain in #2329
• Revert "Update DLPack to 1.0 (#2354)" by @bdice in #2366
• [REVIEW] New Dataset API Clarifying Ownership by @HowardHuang1 in #1846

🐛 Bug Fixes

• [CI nightly] Temporarily disable conda-python-tests codecov reporting by @aamijar in #2274
• Exclude SPDX copyright lines from .clang-format by @achirkin in #2275
• Fix FastIntDiv tparam by @aamijar in #2278
• CAGRA: fix concurrent initialization and usage of dataset descriptor by @achirkin in #2237
• Revert "FIX: disable warpspeed scan (#2062)" by @bdice in #2074
• Predict IVF-PQ FP16 overflow and auto-switch to FP32 by @huuanhhuyn in #2246
• Update .clang-format to fix copyright and style check by @HowardHuang1 in #2304
• Fix batched silhouette score stream ordering by @csadorf in #2316
• fix(pre-commit): normalize SPDX handling by @bdice in #2318
• Remove fragile explicit import that breaks with jextract/CUDA changes by @imotov in #2124
• Remove trailing new lines from GTest names by @robertmaynard in #2312
• Update Comparison Branch for ABI Stability Git Workflow by @tarang-jain in #2331
• CAGRA ACE build: avoid rounding down graph degree during sub-index build by @achirkin in #2335
• [BUG] Fix CAGRA search recall with a graph built by NN Descent by @enp1s0 in #819
• Fix IVF extend list resize bug by @qwertyforce in #2297
• Use batched unfused 1-NN for KMeans on Blackwell by @viclafargue in #2346
• Fix prefilter bitmap host buffer under-allocation in Java bindings by @imotov in #2375
• Fix incorrect distances tensor type in cuvsCagraSearch by @imotov in #2372

📖 Documentation

• Update cuvs-bench docs to show both CLI and Python API usage by @jnke2016 in #2084
• doc: fix better the python api example by @yihong0618 in #2080
• docs: RAFT->cuVS in issue templates by @jameslamb in #2097
• Updating Fern and fixing a couple small links by @cjnolet in #2210
• Adding nvidia as the "global-theme" by @cjnolet in #2216
• remove pypi.nvidia.com from install docs by @jameslamb in #2223
• Add code of conduct by @aamijar in #2272
• Readme: ivf-rabitq paper in references by @aamijar in #1877
• contributing.md link to docs by @aamijar in #2119
• Enable multi-source docs for cuVS Fern instance by @aschilling-nv in #2277
• Remove legacy RAPIDS Doxygen theme assets by @bdice in #2286

🚀 New Features

• Add JIT-LTO based filter UDF support for CAGRA by @dantegd in #2132
• Support IVF-RaBitQ in cuVS Library by @Stardust-SJF in #1866
• Improve CAGRA-Q performance and add support for PQ_LEN=8 by @enp1s0 in #1533
• Expose raft::memory_tracking_resources to C/Python/Java/Rust by @achirkin in #2073
• Scripts for Billion-Scale Synthetic Data Generation by @jinsolp in #2247
• Add CAGRA build heuristics independent of HNSW parameters by @achirkin in #2345
• CAGRA Bloom Filter by @divyegala in #2236

🛠️ Improvements

• Enforce smaller wheel sizes by @KyleFromNVIDIA in #2089
• Require CMake 4.0 by @KyleFromNVIDIA in #2078
• Use L2 for IVF-PQ coarse clustering when the main metric is IP by @yan-zaretskiy in #2076
• Improve devcontainer startup times by @trxcllnt in #2109
• fix(visibility): add symbol visibility controls to cuvs by @vyasr in #2101
• [REVIEW] Add OpenSearch backend to cuvs-bench by @jrbourbeau in #2012
• Forward merge fix by @cjnolet in #2140
• Add SECURITY.md by @jameslamb in #2139
• merge release/26.06 into main by @jameslamb in #2146
• IVF-SQ Python API by @viclafargue in #1913
• fix(ci): fix configuration for breaking change notification workflow by @jameslamb in #2151
• [FEA] JIT LTO Pairwise Distances by @tarang-jain in #2099
• uint64 header support for cuvs_bench by @jinsolp in #2130
• refactor: switch to rapids-artifact-name for consistent artifact naming by @gforsyth in #2136
• Build and test with CUDA 13.3.0 by @bdice in #2226
• Require C++20 for C++ CUDA examples by @HowardHuang1 in #2244
• Use main shared-workflows branch by @bdice in #2259
• Switch to public RAFT numpy_serialize APIs by @julianmi in #2038
• fix(java): use Maven mirror, remove unnecessary plugins, other CI-stability improvements by @jameslamb in #2253
• Update CODEOWNERS team names for rapidsai->NVIDIA migration by @josephine-wolf-oberholtzer in #2261
• Re-enable verify-codeowners pre-commit check by @josephine-wolf-oberholtzer in #2270
• Temporarily disable conda-python-tests codecov reporting by @josephine-wolf-oberholtzer in #2271
• Upgrade faiss to 1.14.3 by @aamijar in #2134
• Update CuPy and NumPy dependency specs by @bdice in #2260
• Harden multi-GPU ANN tests for larger GPU counts by @viclafargue in #2280
• SNMG Batched KMeans Python API by @viclafargue in #2154
• Re-enable Python code coverage by @josephine-wolf-oberholtzer in #2290
• fix(deps): try to dynamically load nvrtc from wheel by @gforsyth in #2299
• switch from 'cuda-python' to specific components (e.g. 'cuda-bindings') by @jameslamb in #2294
• Switch Product Quantization VQ to mean() when n_centers = 1 by @lowener in #2250
• Native row-major PCA by @aamijar in #2125
• Add build option to overwrite libraft logging level by @huuanhhuyn in #2276
• Add missing memory estimates by @huuanhhuyn in #2255
• update actions/checkout to fix zizmor by @arhag23 in #2327
• Empty commit to trigger a build by @bdice in #2341
• Improve balanced k-means rebalancing by @anaruse in #2222
• All-neighbors graph on host memory by @jinsolp in #2313
• Multi partition cagra search by @jamxia155 in #2035

New Contributors

• @yihong0618 made their first contribution in #2080
• @HowardHuang1 made their first contribution in #2244
• @Stardust-SJF made their first contribution in #1866
• @josephine-wolf-oberholtzer made their first contribution in #2261
• @aschilling-nv made their first contribution in #2277
• @huuanhhuyn made their first contribution in #2246
• @qwertyforce made their first contribution in #2297

Full Changelog: v26.08.00a...v26.08.00

v26.06.00

What's Changed

🚨 Breaking Changes

• Default to static linking of libcudart by @bdice in #1627
• Remove JIT+LTO fragment database by @KyleFromNVIDIA in #1927
• Use static cudart by @KyleFromNVIDIA in #1931
• Always build with JIT+LTO by @KyleFromNVIDIA in #1923
• Migrate RMM usage to CCCL MR design by @bdice in #1990
• Exposition of KMeans param object for PQ in C++ by @lowener in #2005
• [Cleanup] Combine Batched and Regular KMeans Impl by @tarang-jain in #2015
• Preserve input memory location for NN Descent by @jinsolp in #1928

🐛 Bug Fixes

• Fix CCCL compilation error by @viclafargue in #1963
• Forward-merge release/26.04 into main by @KyleFromNVIDIA in #1971
• Forward-merge release/26.04 into main by @KyleFromNVIDIA in #1980
• Remove dangling pointers in JIT Fragments by @divyegala in #1988
• Add head_rev to cuvs recipe by @KyleFromNVIDIA in #1993
• Fix potential OOB access in CAGRA search when graph size < dataset size by @irina-resh-nvda in #1780
• Fix MG kmeans intertia_check n_iters by @aamijar in #2020
• Fix cuvs_bench pytest pareto assert by @aamijar in #2027
• Fix nightly build matrix by @KyleFromNVIDIA in #2054
• Fix vulnerable index deserialization by @lowener in #2068
• Fix symbol export kmeans by @aamijar in #2070
• Fix argmin/argmax based on the distance type by @achirkin in #2016
• Remove unneeded request for CUDA device link phase by @robertmaynard in #2077
• Update Faiss and DiskANN Patch to Use C++20 by @tarang-jain in #1796
• Fix brute force Rust index dataset lifetime by @yan-zaretskiy in #2083
• Fix segfault cuvs bench by @aamijar in #2088
• Fix cagra::optimize modifying the state of raft::resources by @achirkin in #2103
• Add direct target dependency when embedding fatbins by @KyleFromNVIDIA in #2106
• Fix check for PQ vectorized load by @lowener in #2107
• Fix workspace usage by @mfoerste4 in #2135
• Add missing visibility controls in IVF SQ by @divyegala in #2141

📖 Documentation

• Elaborate on fragment architecture in JIT+LTO documentation by @KyleFromNVIDIA in #1991
• Add Cluster and Distance sections to C documentation by @lowener in #1955
• Adding CAGRA merge to the documentation by @viclafargue in #1942
• [Doc Update] CAGRA Memory Footprint by @singhmanas1 in #1300
• Align docs with pluggable benchmark API by @jnke2016 in #1891
• Add docs for cagra mem usage with NN Descent build algo by @jinsolp in #2000
• Fix minor typos in cuvs-bench source build docs by @jrbourbeau in #2006
• Fix cuvs-bench docker images in docs by @jrbourbeau in #2003
• Update JIT+LTO guide to reflect new automatic embedding system by @KyleFromNVIDIA in #2045
• [Docs] Convert Sphinx docs to Fern by @cjnolet in #2067
• Add UDF Usage and Developer docs by @divyegala in #2030
• [DOC] Adding API guides for core cuVS types by @cjnolet in #2117

🚀 New Features

• [REVIEW] Add L1 support to NN-Descent by @yan-zaretskiy in #1898
• PCA C and Python API by @aamijar in #1987
• Introduce UDF Architecture by @divyegala in #1804
• JIT LTO Cagra Search by @divyegala in #1807
• Expose supported brute force metrics in all_neighbors by @jinsolp in #1827
• [REVIEW] Generalize and improve cagra::optimize by @mfoerste4 in #1830
• IVF-SQ C++ API by @viclafargue in #1865

🛠️ Improvements

• Use PQ API in CAGRA-Q + SCANN by @lowener in #1746
• Speed up recall calculation in cuVS Bench for large top-K by @jamxia155 in #1816
• Update codespell Version in pre-commit-config by @tarang-jain in #1920
• Forward-merge release/26.04 into main by @gforsyth in #1936
• Refactor StaticFatbinFragmentEntry to use tags by @KyleFromNVIDIA in #1970
• Replace cudaMemcpy2DAsync Calls with raft::copy_matrix by @tarang-jain in #1976
• update pip devcontainers' base image tags by @trxcllnt in #1985
• Refactor instantiation matrices to generate at build time by @KyleFromNVIDIA in #1984
• Add option to enable "sve" optimization level on armv9 by @LizYou in #1121
• Improve cuvs-bench doc and add executable dir option by @tfeher in #681
• Enforce type safety in JIT+LTO launcher by @KyleFromNVIDIA in #1997
• Add KDE kernel by @Intron7 in #1915
• Coderabbit integration by @benfred in #1908
• Refactor fatbin registration to use common input file by @KyleFromNVIDIA in #2008
• Update to clang 20.1.8 by @bdice in #2009
• JIT+LTO IVF-PQ compute similarity by @KyleFromNVIDIA in #1957
• Refactor JIT+LTO kernels by @KyleFromNVIDIA in #2021
• feat(rust): add serialize/deserialize support for CAGRA index by @zbennett10 in #1840
• Use new compute-matrix workflow by @KyleFromNVIDIA in #2034
• Reuse minClusterAndDistance Helper for Balanced KMeans by @tarang-jain in #2001
• feat(rust): add search_with_filter to CAGRA Index by @jamie8johnson in #2019
• [REVIEW] Drop extra copy in get_last_error_text by @jakirkham in #2044
• FIX: disable warpspeed scan by @mfoerste4 in #2062
• Use token.rapids.nvidia.com when issuing S3 bucket creds in devcontainers by @trxcllnt in #2047
• Remove NO_CUDART_DEP property by @KyleFromNVIDIA in #2065
• Switch the remaining C++17 components to C++20 by @achirkin in #2063
• fix(ci): resolve all zizmor findings and add zizmor pre-commit checks by @gforsyth in #2053
• fix(ci): declare explicit secrets in publish-rust.yaml by @gforsyth in #2069
• [REVIEW] Rewrite cuvs-sys build to discover pre-installed cuVS via cmake-package by @yan-zaretskiy in #2022
• Fix symbol export by @vyasr in #2052
• fix(ci): add explicit actions: write permission for telemetry-summarize
  by @gforsyth in #2075
• [REVIEW] Improve 1-NN performance with split GEMM/reduction kernels on Blackwell by @vinaydes in #1768
• Build and test with CUDA 13.2.0 by @bdice in #2072
• Centralize shared utilities across benchmark backends by @jnke2016 in #2040
• Persistent CAGRA: benchmark group and bad config warnings by @achirkin in #2091
• Multi-GPU Batched KMeans by @viclafargue in #2017
• IVF-SQ C API by @viclafargue in #1910
• skip CuPy 14.1.0 by @jameslamb in #2142

New Contributors

• @singhmanas1 made their first contribution in #1300
• @LizYou made their first contribution in #1121
• @jamie8johnson made their first contribution in #2019

Full Changelog: v26.06.00a...v26.06.00

v26.04.00

What's Changed

🚨 Breaking Changes

• Use HNSW GPU Hierarchy by Default by @julianmi in #1617
• Backport "Default to static linking of libcudart" by @KyleFromNVIDIA in #1918
• enforce a floor on libnvjitlink, build wheels with CUDA 13.0.x, test wheels against mix of CTK versions by @jameslamb in #1862

🐛 Bug Fixes

• Add float16 support for CAGRA extend by @lowener in #1620
• Fix C examples CUDAToolkit dependency by @tfeher in #1626
• clang-tidy: remove AnalyzeTemporaryDtors key by @achirkin in #1778
• Fix build.sh: build cuvs_c for examples and fix --gpu-arch parsing by @achirkin in #1779
• Better handling of batching of search in MG replicated mode by @viclafargue in #1718
• Add nvjitlink to cuda-toolkit pip extras for cusparse compatibility by @bdice in #1794
• [REVIEW] Fix: for balanced kmeans use grid.x for adjust_centers to avoid grid.y overflow for >262K centroids by @Nischal1729 in #1805
• Fix thrust header by @aamijar in #1817
• Fix persistent CAGRA regressions by @achirkin in #1800
• Fix thrust header by @aamijar in #1825
• Set cudaFuncAttributeMaxDynamicSharedMemorySize with thread-safety by @mythrocks in #1771
• [REVIEW] Move from thrust::make_counting_iterator to cuda::make_counting_iterator by @mythrocks in #1826
• Fixed cuvs benchmark debug build issue (linker step fail) by @irina-resh-nvda in #1599
• Graph degree equals intermediate graph degree bug fix by @irina-resh-nvda in #1834
• FAISS patch for thrust_counting_iterator.h by @aamijar in #1844
• Use 1D grid calculations in epsilon_neighborhood by @divyegala in #1847
• Fix setting CAGRA graph build algo to iterative search by default by @achirkin in #1864
• Faiss suppress warning 611 by @aamijar in #1879
• Pin faiss to 1.14.0 by @aamijar in #1885
• Make some minor fixes to JIT+LTO functionality by @KyleFromNVIDIA in #1896
• Revert "Default to static linking of libcudart" by @KyleFromNVIDIA in #1925
• Fix hanging pytests by @aamijar in #1924
• Disallow programmatic stream serialization in JIT kernel launches by @divyegala in #1932
• IVF-Flat: fix irrelevant assert in the fused kernel mode by @achirkin in #1941
• [REVIEW] cuVS bench: Fix cudaFuncSetAttribute not being called when CAGRA search switches kernel variants by @irina-resh-nvda in #1851
• Pin openblas for aarch64 by @tarang-jain in #1965
• Null JIT kernel launch config by @divyegala in #1974
• [BUG] Fix Vamana Serialization by @tarang-jain in #1966

📖 Documentation

• remove docs references to cuvs-bench-datasets, install a C compiler in pre-commit env by @jameslamb in #1736
• Adding simple tarball install to build and install docs by @cjnolet in #1868
• Fixing nccl link in tarball install instructions by @cjnolet in #1887
• Add developer and user guides for JIT by @divyegala in #1876
• PCA docs by @aamijar in #1949
• Fix a few typos and .rst link syntax by @jrbourbeau in #1973
• Update docs footer year by @aamijar in #1958
• Doc improvements by @aamijar in #1978

🚀 New Features

• JIT compile interleaved_scan_kernel for CUDA 13 by @divyegala in #1405
• [REVIEW] L1 distance support for iterative search CAGRA build by @yan-zaretskiy in #1831
• [FEA] Inertia Computation for Balanced KMeans and Add Option for Weighted Inertia by @tarang-jain in #1880
• PCA preprocessor by @aamijar in #1808
• [FEA] Add Batching to KMeans by @tarang-jain in #1886

🛠️ Improvements

• Automatic Partition Count Derivation for ACE by @julianmi in #1603
• Add filter for cagra::merge by @benfred in #1496
• Ivf_flat extends golang APIs by @cpegeric in #1600
• Drop Python 3.10 support by @gforsyth in #1748
• tighten wheel size limits, expand CI-skipping logic, other small build changes by @jameslamb in #1751
• Migrate hash strategy to use the new cuco::static_map by @PointKernel in #1462
• Update raft headers by @aamijar in #1763
• remove pip.conf migration code in CI scripts, update CI-skipping rules by @jameslamb in #1760
• Convert non-type template parameters to runtime parameters in CAGRA search to cut binary size by @seunghwak in #1498
• Rename lanczos by @aamijar in #1759
• CI: build with CUDA 13.1.1 by @jameslamb in #1766
• Fixes for stricter compilers by @maxwbuckley in #1703
• fix cpu_search call by including k argument by @benfred in #1785
• Use GHA id-token for sccache-dist auth token by @trxcllnt in #1790
• Remove cagra_optimize.hpp by @aamijar in #1791
• Remove unused use_norms constant from kernel_sm60.cuh by @maxwbuckley in #1705
• Remove unused variable read_idx from query loop by @maxwbuckley in #1706
• remove gitutils by @jameslamb in #1797
• refactor: build wheels and conda packages using Python limited API by @gforsyth in #1788
• Add pluggable backend architecture to cuvs-bench by @jnke2016 in #1536
• Use Specific CCCL Includes by @divyegala in #1806
• Replace thrust::tuple with cuda::std::tuple by @miscco in #1811
• check-nightly-ci: update to new version by @jameslamb in #1813
• check-nightly-ci: remove testing config by @jameslamb in #1824
• Refactor JIT LTO kernel generation by @KyleFromNVIDIA in #1812
• Move test_compute_matrix_product.py to cpp/tests by @KyleFromNVIDIA in #1836
• Drop uses of thrust/functional.h by @miscco in #1835
• Better ncv param spectral embedding edge case by @aamijar in #1828
• Modernize the uses of raft in cuVS by @achirkin in #1837
• refactor(limited api): add explicit wheel.py-api to pyproject.toml
  by @gforsyth in #1852
• Update Cython lower bound pin to 3.2.2 by @vyasr in #1858
• Add support for Python 3.14 by @gforsyth in #1845
• fix(rust): change Index::search to take &self instead of self by @zbennett10 in #1839
• Remove pytest upper bound pin by @vyasr in #1867
• Readme: vecflow paper in references by @aamijar in #1874
• Refactor knn graph build params to remove graph_build_types.hpp by @jinsolp in #1235
• Remove IndexWrapper by @aamijar in #1792
• Use C linkage for JIT LTO kernels by @divyegala in #1909
• Prevent nested parallelism in HNSW bench by @julianmi in #1895
• Add a script to check for breaking C ABI changes by @benfred in #1749
• Improve the IVF-PQ Coarse Batch Size Workspace Estimation by @julianmi in #1937
• ScaNN: Fix AVQ prefetch by @rmaschal in #1899

New Contributors

• @cpegeric made their first contribution in #1600
• @PointKernel made their first contribution in #1462
• @seunghwak made their first contribution in #1498
• @Nischal1729 made their first contribution in #1805
• @jnke2016 made their first contribution in #1536
• @zbennett10 made their first contribution in https://github.com/ra...

v26.02.00

What's Changed

🚨 Breaking Changes

• Use CCCL's mdspan implementation by @divyegala in #1605
• Add filter for cagra::merge by @benfred in #1755

🐛 Bug Fixes

• fix(ci): remove unknown parameter name from rocky8 build job by @gforsyth in #1554
• CMake check for FAISS use in benchmarks by @irina-resh-nvda in #1591
• Fix overflow in preprocess_data_kernel of NN Descent by @jinsolp in #1596
• cmake is missing sparse/gram.cu gtest by @aamijar in #1611
• Correctly specify the ninja-build version to download by @robertmaynard in #1634
• Remove Flaky Python ACE Example by @julianmi in #1641
• Fix integer overflow in get_free_host_memory() on systems with >2TB RAM by @bdice in #1636
• Fix libclang download for Rust, CUDA initialization for C tests by @KyleFromNVIDIA in #1652
• Skip Python Multi-GPU Doctests by @julianmi in #1671
• [Bug] Fix memory allocation test in CAGRA graph optimization by @enp1s0 in #1675
• Fix multi-GPU All Neighbors memory coherence issue on older platforms by @viclafargue in #1713

📖 Documentation

• Clean raft/neighbors comments by @aamijar in #1651
• Clean raft::neighbors comments by @aamijar in #1665

🚀 New Features

• Assign the c/ folder to the the c code ownder group by @robertmaynard in #1573
• Add arm64 builds to the libcuvs_c rocky8 matrix by @robertmaynard in #1570
• [FEA] Enforce tighter link restrictions on libcuvs_c by @robertmaynard in #1614
• [FEA] IVF-PQ Build Factories for Precomputed Centroids and Codebooks by @tarang-jain in #1483
• Add HNSW ACE build method by @julianmi in #1597
• [FEA] C + Python API for IVF-PQ Build Factories with Precomputed Centroids by @tarang-jain in #1664
• Add arm64 builds to the libcuvs_c build.yaml matrix by @robertmaynard in #1701
• pre-built libcuvs_c.so now use the new ABI major/minor values by @robertmaynard in #1708
• Correct base release for cuvs abi 1 major by @robertmaynard in #1724
• Add support for PQ preprocessing API by @lowener in #1278
• [FEA] IVF-PQ to Write Flat PQ Codes by @tarang-jain in #1607
• Enable collation of all license files into pre-built binaries by @robertmaynard in #1679

🛠️ Improvements

• Update FAISS patch for RMM memory resource header migration by @bdice in #1566
• Use strict priority in CI conda tests by @bdice in #1583
• Update FAISS from 1.12.0 to 1.13.0 by @bdice in #1585
• Use strict priority in CI conda tests by @bdice in #1606
• Deduplicate {unpack/pack}_list_data_kernel by @jinsolp in #1609
• Expose NN Descent fp16 data type support to python by @jinsolp in #1616
• Remove alpha specs from non-RAPIDS dependencies by @bdice in #1618
• Enable merge barriers by @KyleFromNVIDIA in #1625
• Add devcontainer fallback for C++ test location by @bdice in #1631
• Dispatch to use fp32 distance computation in NN Descent depending on data dimensions by @jinsolp in #1415
• Prepare cuvs for removal of deprecated raft apis by @aamijar in #1610
• Shorten the Test Duration of test_cagra_ace.py by @julianmi in #1640
• Optional seed spectral embedding by @aamijar in #1639
• FAISS patch for removed raft headers by @aamijar in #1654
• Empty commit to trigger a build by @bdice in #1662
• Spectral Clustering dataset api by @aamijar in #1653
• Hide kernel symbols by @divyegala in #1663
• Update cuVS to c++20 by @divyegala in #1649
• Move faiss_select from raft to cuvs by @aamijar in #1658
• Fix DiskANN Override by @tarang-jain in #1666
• Use SPDX license identifiers in pyproject.toml, bump build dependency floors by @jameslamb in #1667
• Add CUDA 13.1 support by @bdice in #1642
• Remove faiss patch by @aamijar in #1668
• Expose pq list data to python by @benfred in #1428
• Add rust bindings for vamana index by @benfred in #1608
• build and test against CUDA 13.1.0 by @jameslamb in #1677
• [Improvement] Allow Configurable DRAM in DiskANN SSD Wrapper by @tarang-jain in #1598
• prefer CUDA 13.1 devcontainers, react to some cutlass removals in RAFT by @jameslamb in #1686
• Empty commit to trigger a build by @jameslamb in #1694
• Spectral Embedding nnz_t by @aamijar in #1628
• Avoid installing CCCL headers in wheels by @divyegala in #1692
• Use main shared-workflows branch by @jameslamb in #1707
• Add Helper to Create NumPy Files by @julianmi in #1645
• Spectral Embedding with all_neighbors by @aamijar in #1693
• Deduplicate calc_chunk_indices_kernel by @jinsolp in #1657
• wheel builds: react to changes in pip's handling of build constraints by @mmccarty in #1710
• fix(build): build package on merge to release/* branch by @gforsyth in #1733
• Add a transform function to ivf_pq by @benfred in #1732

New Contributors

• @irina-resh-nvda made their first contribution in #1591

Full Changelog: v26.02.00a...v26.02.00

v25.12.00

What's Changed

🚨 Breaking Changes

• Using all_neighbors for mutual reachability by @jinsolp in #1234
• Refactor libcuvs_c header and source locations by @robertmaynard in #1357
• Update cagra C API enums to have more long term stable values by @robertmaynard in #1436
• Require CUDA 12.2+ by @jakirkham in #1476
• Remove mutual_reachability_graph Public API by @tarang-jain in #1481

🐛 Bug Fixes

• Allow compilation when OpenMP is disabled by @robertmaynard in #1346
• Deallocation should be noexcept by @bdice in #1416
• ANN_BENCH: Don't throw in noexcept do_deallocate by @achirkin in #1417
• Remove unneeded cutlass public build/install dependency by @robertmaynard in #1411
• cuvs_static properly adds C include dir to target_include_directories by @robertmaynard in #1426
• Fix C/Python serialization for FP16. Add python tests by @lowener in #1429
• Properly guard usage of openmp function calls by @robertmaynard in #1435
• Update nlohmann-json to 3.12.0 by @KyleFromNVIDIA in #1449
• Fix test params for filtered ivf_flat by @aamijar in #1463
• [REVIEW][Java][Bug] Fix CuVSMatrix#getRow to take strides into consideration by @ldematte in #1442
• Adding more tests and I find the compression settings are broken in the Golang API by @maxwbuckley in #1472
• Fix binary quantizer host transform bounds and stream order by @achirkin in #1473
• [BUG] Check if dynamic batching conf is null by @tarang-jain in #1459
• Include cagra search algorithms in libcuvs_static.a by @robertmaynard in #1457
• [Java] Disable flaky test testFloatSerialization by @mythrocks in #1503
• Add libaio to cuvs-bench-cpu, improve bench dependencies by @tarang-jain in #1480
• [Java] Fix CagraBuildAndSearchIT concurrent deallocation errors by @ldematte in #1510
• Pin Cython pre-3.2.0 and PyTest pre-9 by @jakirkham in #1528
• Fix a bug in compute_distance_00_generate.py by @enp1s0 in #1532
• Check stride information in from_dlpack by @benfred in #1458
• refactored update-version.sh to handle new branching strategy by @rockhowse in #1535
• fixed bug with update-version.sh by @rockhowse in #1556
• fix(ci): remove unsupported name parameter from custom job def by @gforsyth in #1560
• Remove need to have rapids_logger headers installed to use clib by @robertmaynard in #1527
• Fix BruteForce serialize test by @lowener in #1568
• [C] Fix: cuvsRMMMemoryResourceReset sets a valid resource by @ldematte in #1540
• Fix typo in date field by @vyasr in #1563
• SNMG ANN build with OpenMP nested parallelism by @viclafargue in #1526

📖 Documentation

• Fix table of content documentation by @lowener in #1427
• Docs Spectral Clustering by @aamijar in #1490
• Use current system architecture in conda environment creation command by @bdice in #1499
• Updating README for release by @cjnolet in #1584

🚀 New Features

• Unify binding headers by @robertmaynard in #1413
• CAGRA: decouple source idx type from graph idx type and add a mapping between them by @achirkin in #1251
• Add SOVERSION information to libcuvs_c by @robertmaynard in #1474
• Spectral Clustering by @aamijar in #1425
• ANN_BENCH: integrate NVTX statistics by @achirkin in #1529
• Add Augmented Core Extraction Algorithm by @julianmi in #1404
• Extend CI to build and test x86 libcuvs_c tarballs by @robertmaynard in #1524
• Introduce libcuvs-headers and libcuvs-static CMake components and conda packages by @divyegala in #1494
• [Java] Bindings, tests and benchmarks for RMM pooled memory by @ldematte in #1453

🛠️ Improvements

• Update RAPIDS_BRANCH, codify changes in update-version.sh by @KyleFromNVIDIA in #1368
• Move eigen_solvers from raft by @aamijar in #1402
• CosineExpanded Distance Metric for CAGRA by @tarang-jain in #197
• Enable sccache-dist connection pool by @trxcllnt in #1431
• Use pinned_host_memory_resource instead of pinned_memory_resource. by @bdice in #1434
• Use main in RAPIDS_BRANCH by @bdice in #1439
• Use main shared-workflows branch by @bdice in #1444
• [Review] ScaNN: Add option for AVQ/Noise Shaping to bfloat16 quantization by @rmaschal in #1354
• Add NVTX annotations to CAGRA knn graph build stage by @achirkin in #1443
• [Review][C] Export the ability to get/set the log level to the C API by @ldematte in #1375
• [Java] Test indexing and serialization with integral (byte) dataset by @ldematte in #1366
• Use SPDX for all copyright headers by @KyleFromNVIDIA in #1446
• [REVIEW][Java] Log level api by @ldematte in #1376
• [Review] [Java] Disable flaky Java tests to reduce CI churn by @mythrocks in #1469
• Refactor rapids_cpm_package_details to rapids_cpm_package_info by @bdice in #1433
• Fix golang test name and add more logging and error checking. by @maxwbuckley in #1471
• Use an explicit std::min for byte alignment calculation by @maxwbuckley in #1465
• Replace custom cuda_pinned_resource with RMM's pinned_host_memory_resource by @bdice in #1466
• Add patch for FAISS memory resources by @bdice in #1477
• Improved CAGRA build parameter heuristics by @achirkin in #1448
• Decouple C++ library from C library by @divyegala in #1488
• Use RAPIDS_BRANCH in cmake-format invocations that need rapids-cmake configs by @bdice in #1475
• eigen tolerance option Spectral Embedding by @aamijar in #1493
• Update to CUDA 13.0.2 by @bdice in #1506
• Single Linkage to Use all_neighbors API to build the KNN graph by @tarang-jain in #1507
• Migrate to new CCCL memory resource interface by @bdice in #1502
• Remove unused headers scann by @lowener in #1508
• Update RMM includes from <rmm/mr/device/*> to <rmm/mr/*> by @bdice in #1538
• Use ruff-check, ruff-format instead of black, flake8 by @KyleFromNVIDIA in #1500
• Set memory pool from Python multi-GPU resource by @viclafargue in #1530
• [Java] Fix format with spotless by @ldematte in #1539
• fix bad version update by @trxcllnt in #1555
• Deduplicate KMeans instantiations by @divyegala in #1565
• Improve memory usage in build_mr_linkage by @jinsolp in #1550
• Use sccache-dist build cluster for conda and wheel builds by @trxcllnt in #1495
• [Java] One PinnedMemoryBuffer per CuVSResourcesImpl by @ldematte in #1441
• [Java] Relax cuVS version matching by @ldematte in #1544

New Contributors

• @maxwbuckley made their first contribution in #1471
• @rockhowse made their first contribution in #1535

Full Changelog: v25.12.00a...v25.12.00

v25.10.00

🚨 Breaking Changes

• [Java] Uniform toHost/toDevice to work across all CuVSMatrix classes (#1328) @ldematte
• Use int64_t for getters in the cagra/ivf_flat c-api (#1272) @benfred
• [REVIEW][Java] Rename destroyIndex() to close(), extend AutoCloseable (#1252) @ldematte
• Removing deprecated batching code specific to NN Descent (#1249) @jinsolp
• MG C API (#1160) @viclafargue

🐛 Bug Fixes

• Update cpp/cmake/config.json for ConfigureTest() (#1385) @KyleFromNVIDIA
• Fix k-means++ root rank race condition (#1359) @csadorf
• Fix inadvertent uses of copy constructor in mdarrays across cuVS (#1330) @achirkin
• Refactor CMakeLists to support static only builds of libcuvs (#1317) @robertmaynard
• Fix pq_compressed file created by SSD Vamana build (#1313) @bkarsin
• Revert "[Java]Binary and scalar quantization (#1104)" (#1274) @mythrocks
• Fix balanced kmeans API (#1268) @enp1s0
• [BUG] NEIGHBORS_ALL_NEIGHBORS_TEST build ignores --no-mg (#1230) @enp1s0
• Add error handling for CPU-only bench (#1203) @lowener
• Make duplicate removal in all neighbors robust to distance drift across batches (#1185) @jinsolp
• Fix single GPU sharded search merge (#1094) @viclafargue
• Processing mutual reachability dist in connect_knn_graph for host data (#1093) @jinsolp
• [Java][Fix] Multithreaded querying fails without synchronization (#1082) @chatman

📖 Documentation

• Add Vamana C rst Docs to toctree (#1399) @tarang-jain
• Regroup overloads in docs (#1377) @lowener
• [DOCS] Update IVF Docs with Supported Distance Metrics (#1365) @tarang-jain
• Add note about Linux-only support for cuVS pre-compiled packages (#1306) @stic
• Documentation Fixes (#1301) @benfred
• Docs for spectral embedding (#1299) @aamijar
• Improve NN Descent documentation (#1246) @jinsolp
• Add NN.extend and kmeans to Python doc (#1217) @lowener

🚀 New Features

• [REVIEW] [Java] Option to build fat-jars with native dependencies included (#1296) @mythrocks
• Updates needed when building with CUDA 13 (#1219) @robertmaynard
• [Java] Add CAGRA index graph accessor/build from graph (host memory) (#1216) @ldematte
• [Feat] Add Dockerfile for reproducible installation (#1195) @rhdong
• Vamana C / Python API (#1112) @tarang-jain
• Mutual Reachability in all-neighbors API (#1016) @jinsolp
• CAGRA Build + DiskANN Search cuvs-bench Wrapper (#899) @tarang-jain

🛠️ Improvements

• Patch FAISS for missing Thrust includes (#1398) @bdice
• Empty commit to trigger a build (#1363) @msarahan
• [Review][Java] Refactor: extract interface from CuVSMatrixBaseImpl (#1361) @ldematte
• use CUDA 13.0.1 CI images (#1353) @jameslamb
• Treat warnings as errors in doc builds + MG ANNs Python API doc update (#1350) @viclafargue
• [Java] Support row strides in CuVSMatrix (#1345) @ldematte
• [Java] Adding tests to use CuVSDeviceMatrix (device memory) directly as a CagraIndex input dataset (#1340) @ldematte
• Improve performance of assigning clusters in batched all_neighbors (#1336) @jinsolp
• Configure repo for automatic release notes generation (#1334) @AyodeAwe
• [Java] Buffered device matrix builder (#1332) @ldematte
• Remove UCX-Py/UCXX from release script (#1331) @pentschev
• [Java] Uniform toHost/toDevice to work across all CuVSMatrix classes (#1328) @ldematte
• [Java] Add libcuvs <-> cuvs-java version check (#1327) @ldematte
• update dependencies: use cuda-toolkit wheels (#1326) @jameslamb
• Add cagra.extend bindings for python (#1324) @benfred
• [Java] Make cudaGetDeviceProperties compatible with CUDA 12 and 13 based on symbol presence (#1323) @ldematte
• Use branch-25.10 again (#1319) @jameslamb
• [Review][Java] Add detailed error message for libcuvs load failure to UnsupportedProvider/UnsupportedOperationExceptions (#1316) @ldematte
• [Java] Add reason(s) to UnsupportedProvider/UnsupportedOperationExceptions (#1314) @ldematte
• MG Python API (#1307) @viclafargue
• Fix redundant memset (#1305) @vinaydes
• Expose pq-centers to C and Python (#1303) @benfred
• Port raft::neighbors::epsilon_neighborhood to cuvs (#1294) @aamijar
• ScaNN: Overlapped gather for AVQ (#1286) @rmaschal
• Update rapids-dependency-file-generator (#1285) @KyleFromNVIDIA
• [Review][Java] Expand CagraIndex#search to accept more types (int8 and int32) (#1283) @ldematte
• All neighbors C and Python bindings (#1282) @viclafargue
• Build and test with CUDA 13.0.0 (#1273) @jameslamb
• Use int64_t for getters in the cagra/ivf_flat c-api (#1272) @benfred
• [Java][C] Expose GPUInfo (#1267) @ldematte
• Add rust bindings for kmeans (#1266) @benfred
• [REVIEW] Add a public API for CAGRA graph optimize (#1260) @abc99lr
• Fix debug build (#1258) @lowener
• Run cargo fmt in the pre-commit hooks (#1255) @benfred
• [REVIEW][Java] Rename destroyIndex() to close(), extend AutoCloseable (#1252) @ldematte
• Removing deprecated batching code specific to NN Descent (#1249) @jinsolp
• Pin Latest Faiss Version (#1247) @tarang-jain
• Use build cluster in devcontainers (#1240) @trxcllnt
• Use rapids_cuda_enable_fatbin_compression (#1239) @robertmaynard
• [Improvement] Replace Calls to thrust::transform with raft::linalg (#1238) @tarang-jain
• [Java] CuVSMatrix for device memory (#1232) @ldematte
• Drop log level for CAGRA trace messages (#1229) @mythrocks
• Update rapids_config to handle user defined branch name (#1227) @robertmaynard
• Reduce noise in Java tests (#1226) @mythrocks
• Update another instance of rapids-build-backend (#1225) @KyleFromNVIDIA
• Fix missing sync_stream in ScaNN build (#1224) @rmaschal
• [Java] Exception-safe RMM Allocations (#1215) @mythrocks
• Update rapids-build-backend to 0.4.1 (#1214) @KyleFromNVIDIA
• Removing filter templates in ivfpq for binary size reduction (#1211) @jinsolp
• ci(labeler): update labeler action to @v5 (#1210) @gforsyth
• [Java] Add decorator for synchronized CuVSResource access (#1209) @ldematte
• Remove unused templated instances (#1204) @vinaydes
• Spectral Embedding precomputed graph api (#1197) @aamijar
• cuvs RAPIDS dependency now use the new rapids-cmake branch info (#1189) @robertmaynard
• Allow latest OS in devcontainers (#1169) @bdice
• MG C API (#1160) @viclafargue
• Merge support for tiered index (#1155) @benfred
• Update build infra to support new branching strategy (#1143) @robertmaynard
• [Java]Binary and scalar quantization (#1104) @punAhuja
• Improve Vamana index build performance and recall (#1032) @bkarsin
• Use GCC 14 in conda builds. (#1030) @vyasr
• Update SNMG ANN API testing (#1023) @viclafargue

[NIGHTLY] v25.12.00

🔗 Links

• Development Branch
• Compare with main branch

🚨 Breaking Changes

• Update cagra C API enums to have more long term stable values (#1436) @robertmaynard
• Refactor libcuvs_c header and source locations (#1357) @robertmaynard
• Using all_neighbors for mutual reachability (#1234) @jinsolp

🐛 Bug Fixes

• Properly guard usage of openmp function calls (#1435) @robertmaynard
• Fix C/Python serialization for FP16. Add python tests (#1429) @lowener
• cuvs_static properly adds C include dir to target_include_directories (#1426) @robertmaynard
• ANN_BENCH: Don't throw in noexcept do_deallocate (#1417) @achirkin
• Deallocation should be noexcept (#1416) @bdice
• Remove unneeded cutlass public build/install dependency (#1411) @robertmaynard
• Allow compilation when OpenMP is disabled (#1346) @robertmaynard

🚀 New Features

• Update cagra C API enums to have more long term stable values (#1436) @robertmaynard
• Unify binding headers (#1413) @robertmaynard
• Refactor libcuvs_c header and source locations (#1357) @robertmaynard

🛠️ Improvements

• Use pinned_host_memory_resource instead of pinned_memory_resource. (#1434) @bdice
• Enable sccache-dist connection pool (#1431) @trxcllnt
• Move eigen_solvers from raft (#1402) @aamijar
• Update RAPIDS_BRANCH, codify changes in update-version.sh (#1368) @KyleFromNVIDIA
• Using all_neighbors for mutual reachability (#1234) @jinsolp
• CosineExpanded Distance Metric for CAGRA (#197) @tarang-jain

v25.08.00

🚨 Breaking Changes

• [Java] Using functions for mapping (#1007) @ldematte
• Remove CUDA 11 from dependencies.yaml (#962) @KyleFromNVIDIA
• Common knn graph build params (#949) @jinsolp
• stop uploading packages to downloads.rapids.ai (#940) @jameslamb

🐛 Bug Fixes

• [BUG] Fix Large Batch Preprocessing for NN Descent (#1166) @tarang-jain
• Fix TieredIndexIT, Scoped resource access change was missed (#1156) @punAhuja
• Fix for hellinger metric (#1128) @viclafargue
• Fix go CI failures (#1114) @benfred
• Conditionally post process distances in NN Descent for use with distance epilogue (#1108) @jinsolp
• [BUG] Disable flaky cpp test AnnIVFFlatTestF_half.AnnIVFFlat/21 (#1092) @julianmi
• [BUG] Fix device access check in build_sorted_mst (#1083) @tarang-jain
• Pin Faiss Version to Last Stable Release (#1029) @tarang-jain
• Fix verify-copyright precommit hook (#1017) @benfred
• Adding GH_TOKEN pass-through to summarize job (#1011) @msarahan
• Reuse <version_config.h> instead of separate version constants in c_api.h (#1003) @mythrocks
• IVF-PQ coarse search: fix integer overflow and avoid excessive batch sizes (#999) @achirkin
• Fix C compile error in tiered_index.h (#996) @mythrocks
• Fix for NN Descent negative or NaN distances (#994) @jinsolp
• Adding int64 search for MG CAGRA (#975) @viclafargue
• [ANN_BENCH] Fix OpenMP Sections in Wrappers (#911) @tarang-jain
• [BUG] cuvs-bench data_export fixes (#863) @tarang-jain

📖 Documentation

• [ANN_BENCH] [DOCS] Add Vamana / DiskANN to cuvs-bench Docs (#1164) @tarang-jain
• Add Golang build instructions to build.rst (#1116) @mythrocks
• Minor correction for Java test instructions (#1081) @mythrocks
• add docs on CI workflow inputs (#1075) @jameslamb
• [REVIEW][Java] Add instructions on how to run a single test (#1004) @ldematte

🚀 New Features

• Binary CAGRA with NN Descent (#1133) @tarang-jain
• CAGRA build heuristics for HNSW (#1125) @achirkin
• BitwiseHamming distance for NN Descent (#1101) @jinsolp
• ReachabilityPostProcess distance epilogue for NN Descent (#1073) @jinsolp
• Add brute force to all_neighbors API (#1062) @jinsolp
• [Java] Support for tiered index (#1028) @punAhuja
• Enable build_sorted_mst with data on host memory (#997) @jinsolp
• Add C-API method to fetch cuVS version from Java. (#935) @mythrocks
• Spectral Embedding (#871) @aamijar
• [FEA] Build Single Linkage API (#820) @tarang-jain

🛠️ Improvements

• Check shape is initialized in cuvsMatrixSliceRows (#1193) @benfred
• Add error checking for our C example code (#1171) @benfred
• Removing all references to CUDA 11 from codebase (#1150) @cjnolet
• [Java]Fixed TieredIndexParams creation/destruction, calling cuvsTieredIndexParamsCreate (#1147) @punAhuja
• Fix update-version (#1135) @AyodeAwe
• Reducing binary size in ivf_pq by deduplicating kernels (#1130) @jinsolp
• fix(docker): use versioned -latest tag for all rapidsai images (#1129) @gforsyth
• SCaNN Index build (#1120) @rmaschal
• [Java] Extend Dataset to work as an output data container (#1111) @ldematte
• [Java] Fix HNSW params allocation (#1110) @ldematte
• [Java] Fix CAGRA params allocation (#1109) @ldematte
• [Java] Fix POM (#1106) @ldematte
• Enforce java codeowners for managing contents of java/ (#1103) @cjnolet
• Reduce binary size of refine functions (#1095) @tfeher
• [Java] Introduce scoped resource access (#1089) @ldematte
• Expose graph and dataset accessors for CAGRA to C/Python (#1086) @benfred
• Add warning for unused GPU when n_clusters < n_ranks in batch all_neighbors (#1072) @jinsolp
• [Java] Tidy up MemorySegments lifecycle (#1069) @ldematte
• [Java] Refactor SearchResults implementation classes (#1067) @ldematte
• Use CUDA 12.9 in Conda, Devcontainers, Spark, GHA, etc. (#1063) @jakirkham
• Exporting changed graph_build_params namespace into all_neighbors (#1060) @jinsolp
• hnsw::from_cagra: avoid allocating the graph twice (#1057) @achirkin
• Improve memory efficiency for returning NN Descent distances (#1053) @jinsolp
• Deprecation notice for batching code specific to NN Descent (#1052) @jinsolp
• Remove nvidia and dask channels (#1050) @vyasr
• [Java] Enforce Java code format standard (#1049) @narangvivek10
• Make rust publish run after conda upload (#1047) @AyodeAwe
• refactor(cuda11): remove cuda11-only conda channels, cleanup docs (#1046) @gforsyth
• [Review][Java] Fix random segabort/segfault/double free problems (#1045) @ldematte
• [Java] Add Dataset based on MemorySegment (#1034) @ldematte
• [Java] Add Java API benchmarks (#1033) @ldematte
• fix(cli): Add validation for search-mode parameter (#1026) @mayani-nv
• [Java] Encapsulate on-heap float arrays into Dataset (#1024) @ldematte
• refactor(shellcheck): fix all remaining warnings/errors (#1019) @gforsyth
• Instantiate only specific RAFT linewise kernels (#1018) @aamijar
• Update nightly CI check to allow 30 days of failing nightly build. (#1008) @cjnolet
• [Java] Using functions for mapping (#1007) @ldematte
• [REVIEW][Java] Refactor CagraBuildAndSearchIT to explicitly express different execution modes (#1006) @ldematte
• [REVIEW][Java] Add jextract artifacts to gitignore (#1005) @ldematte
• Remove pytest pin (#998) @vyasr
• [java] Utility function for cudaMemcpy (#983) @mythrocks
• [java] Copy pom.xml to target/ on build (#981) @mythrocks
• Remove CUDA 11 from dependencies.yaml (#962) @KyleFromNVIDIA
• refactor(rattler): remove cuda11 options and general cleanup (#961) @gforsyth
• Remove CUDA 11 devcontainers and update CI scripts (#960) @bdice
• Common knn graph build params (#949) @jinsolp
• stop uploading packages to downloads.rapids.ai (#940) @jameslamb
• Instantiate only specific RAFT reduction kernels (#925) @divyegala
• [Java] Replace the C wrapper layer with the Panama API-based Java implementation (#916) @narangvivek10
• Forward-merge branch-25.06 into branch-25.08 (#909) @gforsyth
• Forward-merge branch-25.06 into branch-25.08 (#897) @gforsyth
• Serialize Vamana index with SSD sector alignment per MSFT DiskANN format, generate quantized dataset for integration with DiskANN (#846) @jamxia155
• Add mean and median threshold modes to binary quantization (#716) @enp1s0

v25.06.01

🚨 Breaking Changes

• [Java] Adding support for prefiltering in CAGRA (#870) @punAhuja

🐛 Bug Fixes

• 25.06 hotfix for SNMG (#1027) @viclafargue
• Update rust/cuvs-sys (#990) @benfred
• Fix rust publishing step (#984) @benfred
• Fix failing CAGRA merge google tests for 25.06 (#974) @rhdong
• Fix for recent NCCL resource update (#968) @viclafargue
• Revert "Fix kmeans::predict argument order (#915)" (#951) @divyegala
• IVF-PQ tests: fix segfault when accessing empty lists (#933) @achirkin
• batch_load_iterator shall copy data located in host memory (#926) @tfeher
• Fix kmeans::predict argument order (#915) @enp1s0
• ANN_BENCH: fix the reported core count (#896) @achirkin
• CUVS_ANN_BENCH_SINGLE_EXE: find the mg library (#890) @achirkin
• [FIX] Revert negated InnerProduct distance for NN Descent (#859) @jinsolp
• IVF-PQ tests: fix segfault when accessing empty lists (#838) @achirkin
• [CuVS-Java] Automate panama bindings generation, Include IVF_PQ parameters in CAGRA index parameters and other changes (#831) @narangvivek10
• Ann-bench: fix unsafe lazy blobs (#828) @achirkin
• Fix test_brute_force conversion (#821) @lowener
• [FEA] New algos and updates/corrections to Faiss cuvs-bench (#677) @tarang-jain

📖 Documentation

• All-neighbors API docs (#944) @jinsolp
• cagra.rst: fitered -> filtered (#866) @eli-b
• [DOCS] Fix Integration Docs (Faiss) (#782) @tarang-jain

🚀 New Features

• Add IVF-PQ support in CAGRA index params in Python (#918) @lowener
• [Feat] Expose C API for CAGRA merge (#860) @rhdong
• Use NCCL wheels from PyPI for CUDA 12 builds (#827) @divyegala
• Add support for half in CAGRA+HNSW (#813) @lowener
• [CI] Enable Java test in CI workflow (#805) @rhdong
• Wrapper for all-neighbors knn graph building (#785) @jinsolp
• Add support of half dtype in IVF-FLAT (#730) @lowener
• IVF-PQ: low-precision coarse search (#715) @achirkin
• [Feat] Add support of logical merge in Cagra (#713) @rhdong

🛠️ Improvements

• Update score calculation for CAGRA-Q instance selection (#938) @enp1s0
• [FEA] Use Native Brute Force for Sparse Pairwise KNN (#927) @tarang-jain
• use 'rapids-init-pip' in wheel CI, other CI changes (#917) @jameslamb
• ANN_BENCH: Avoid repeated calls to raft::get_device_for_address in CAGRA search (#908) @achirkin
• [Java] New off-heap Dataset support for CAGRA and Bruteforce (#902) @chatman
• Finish CUDA 12.9 migration and use branch-25.06 workflows (#901) @bdice
• Update to clang 20 (#898) @bdice
• get Java artifacts from GitHub Actions artifact store (#893) @jameslamb
• Quote head_rev in conda recipes (#892) @bdice
• [Java] Exposing merge API for multiple CAGRA indices (#891) @punAhuja
• Expose ivf-flat centers to python/c (#888) @benfred
• CUDA 12.9 use updated compression flags (#887) @robertmaynard
• Expose ivf-pq centers to python/c (#881) @benfred
• Accept host inputs in python for ivf-pq build and extend (#880) @benfred
• Add tiered_index support (#879) @benfred
• Exclude librmm.so from auditwheel (#878) @bdice
• update.version.sh: remove broken reference, skip most CI on PRs that only modify update-version.sh (#875) @jameslamb
• Add support for Python 3.13 (#874) @gforsyth
• chore: lower wheel size threshold (#872) @gforsyth
• [Java] Adding support for prefiltering in CAGRA (#870) @punAhuja
• Change snmg index to use updated multi gpu resource API (#869) @jinsolp
• run shellcheck on all files, other small pre-commit updates (#865) @jameslamb
• Fix IVF PQ build metric for CAGRA (#862) @lowener
• ANN_BENCH: Expose parallel_mode parameter of FAISS CPU IVF implementation (#861) @achirkin
• Specify matplotlib version (#839) @benfred
• Use random tmp names for index files in tests (#837) @achirkin
• Download build artifacts from Github for CI (#834) @VenkateshJaya
• Add NN-Descent return_distances functionality to python/C (#833) @benfred
• Optimize hnsw::from_cagra<GPU> (#826) @achirkin
• Use vendored RAPIDS.cmake in example code. (#824) @bdice
• refactor(rattler): enable strict channel priority for builds (#823) @gforsyth
• Reduce device memory usage for CAGRA's graph optimization process (2-hop detour counting) (#822) @anaruse
• [cuvs_bench] distinguish search label from build label in data_export.py (#818) @jiangyinzuo
• Vendor RAPIDS.cmake (#816) @bdice
• Update libcuvs libraft ver to 25.06 in conda env (#808) @jinsolp
• Moving NN Descent class and struct declarations to nn_descent_gnnd.hpp (#803) @jinsolp
• Remove @rapidsai/cuvs-build-codeowners (#783) @KyleFromNVIDIA
• Moving wheel builds to specified location and uploading build artifacts to Github (#777) @VenkateshJaya
• Remove unused raft cagra header in add_nodes.cuh (#741) @jiangyinzuo
• Expose kmeans to python (#729) @benfred
• Update cuvs to properly create a NCCL::NCCL target (#720) @robertmaynard
• Optimize euclidean distance in host refine phase (#689) @anstellaire
• Moving MG functions into unified API + raft::device_resources_snmg as device resource type for MG functions (#454) @viclafargue
• Moving random ball cover (#218) @cjnolet