Crate cuvs_sys 

Generated raw FFI bindings to the pinned cuVS 26.08 C API.

Published as an Algeon implementation dependency. Applications use algeon-datafusion; workspace code uses this crate for its component contracts. See the API documentation boundary.

Structs

CUstream_st

DLDataType

\brief The data type the tensor can hold. The data type is assumed to follow the native endian-ness. An explicit error message should be raised when attempting to export an array with non-native endianness

DLDevice

\brief A Device for Tensor and operator.

DLManagedTensor

\brief C Tensor object, manage memory of DLTensor. This data structure is intended to facilitate the borrowing of DLTensor by another framework. It is not meant to transfer the tensor. When the borrowing framework doesn’t need the tensor, it should call the deleter to notify the host that the resource is no longer needed.

DLTensor

\brief Plain C Tensor object, does not manage memory.

cudaDataType_t

cuvsAceParams

Parameters for ACE (Augmented Core Extraction) graph build. ACE enables building indexes for datasets too large to fit in GPU memory by:

cuvsAllNeighborsIndexParams

@brief Parameters controlling SNMG all-neighbors build.

cuvsBinaryQuantizer

@brief Defines and stores threshold for quantization upon training

cuvsBinaryQuantizerParams

@brief Binary quantizer parameters.

cuvsBruteForceIndex

@defgroup bruteforce_c_index Bruteforce index @{ / /** @brief Struct to hold address of cuvs::neighbors::brute_force::index and its active trained dtype

cuvsCagraCompressionParams

Parameters for VPQ compression.

cuvsCagraExtendParams

@defgroup cagra_c_extend_params C API for CUDA ANN Graph-based nearest neighbor search @{ / /** @brief Supplemental parameters to extend CAGRA Index

cuvsCagraIndex

@brief Struct holding the CAGRA index storage address and vector element dtype (DLPack-style)

cuvsCagraIndexParams

@brief Supplemental parameters to build CAGRA Index

cuvsCagraSearchParams

@brief Supplemental parameters to search CAGRA index

cuvsDataset

@brief Dataset handle representing owning storage or a non-owning view.

cuvsFilter

@brief Struct to hold address of cuvs::neighbors::prefilter and its type

cuvsHnswAceParams

Parameters for ACE (Augmented Core Extraction) graph build for HNSW. ACE enables building indexes for datasets too large to fit in GPU memory by:

cuvsHnswExtendParams

@defgroup hnsw_c_extend_params Parameters for extending HNSW index @{

cuvsHnswIndex

@brief Struct to hold address of cuvs::neighbors::Hnsw::index and its active trained dtype

cuvsHnswIndexParams

cuvsHnswSearchParams

@defgroup hnsw_c_search_params C API for hnswlib wrapper search params @{

cuvsIvfFlatIndex

@defgroup ivf_flat_c_index IVF-Flat index @{ / /** @brief Struct to hold address of cuvs::neighbors::ivf_flat::index and its active trained dtype

cuvsIvfFlatIndexParams

@defgroup ivf_flat_c_index_params IVF-Flat index build parameters @{ / /** @brief Supplemental parameters to build IVF-Flat Index

cuvsIvfFlatSearchParams

@defgroup ivf_flat_c_search_params IVF-Flat index search parameters @{ / /** @brief Supplemental parameters to search IVF-Flat index

cuvsIvfPqIndex

@defgroup ivf_pq_c_index IVF-PQ index @{ / /** @brief Struct to hold address of cuvs::neighbors::ivf_pq::index and its active trained dtype

cuvsIvfPqIndexParams

@brief Supplemental parameters to build IVF-PQ Index

cuvsIvfPqParams

cuvsIvfPqSearchParams

@defgroup ivf_pq_c_search_params IVF-PQ index search parameters @{ / /** @brief Supplemental parameters to search IVF-PQ index

cuvsIvfSqIndex

@defgroup ivf_sq_c_index IVF-SQ index @{ / /** @brief Struct to hold address of cuvs::neighbors::ivf_sq::index and its active trained dtype

cuvsIvfSqIndexParams

@defgroup ivf_sq_c_index_params IVF-SQ index build parameters @{ / /** @brief Supplemental parameters to build IVF-SQ Index

cuvsIvfSqSearchParams

@defgroup ivf_sq_c_search_params IVF-SQ index search parameters @{ / /** @brief Supplemental parameters to search IVF-SQ index

cuvsKMeansParams

@brief Hyper-parameters for the kmeans algorithm

cuvsNNDescentIndex

@defgroup nn_descent_c_index NN-Descent index @{ / /** @brief Struct to hold address of cuvs::neighbors::nn_descent::index and its active trained dtype

cuvsNNDescentIndexParams

@defgroup nn_descent_c_index_params The nn-descent algorithm parameters. @{ / /** @brief Parameters used to build an nn-descent index

cuvsPcaParams

@brief Parameters for PCA decomposition.

cuvsProductQuantizer

@brief Defines and stores product quantizer upon training

cuvsProductQuantizerParams

@defgroup preprocessing_c_pq C API for Product Quantizer @{ / /** @brief Product quantizer parameters.

cuvsScalarQuantizer

@brief Defines and stores scalar for quantisation upon training

cuvsScalarQuantizerParams

@defgroup preprocessing_c_scalar C API for Scalar Quantizer @{ / /** @brief Scalar quantizer parameters.

cuvsTieredIndex

@defgroup tiered_index_c_index Tiered Index @{ / /** @brief Struct to hold address of cuvs::neighbors::tiered_index::index and its active trained dtype

cuvsTieredIndexParams

@defgroup tiered_c_index_params Tiered Index build parameters @{ / /** @brief Supplemental parameters to build a TieredIndex

cuvsVamanaIndex

@brief Struct to hold address of cuvs::neighbors::vamana::index and its active trained dtype

cuvsVamanaIndexParams

@brief Supplemental parameters to build Vamana Index

Enums

DLDataTypeCode

\brief The type code options DLDataType.

DLDeviceType

RuntimeVersionError

cuvsAllNeighborsAlgo

@brief Graph build algorithm selection.

cuvsBinaryQuantizerThreshold

@defgroup preprocessing_c_binary C API for Binary Quantizer @{ / /** @brief In the cuvsBinaryQuantizerTransform function, a bit is set if the corresponding element in the dataset vector is greater than the corresponding element in the threshold vector. The mean and sampling_median thresholds are calculated separately for each dimension.

cuvsCagraGraphBuildAlgo

@brief Enum to denote which ANN algorithm is used to build CAGRA graph

cuvsCagraHashMode

@brief Enum to denote Hash Mode used while searching CAGRA index

cuvsCagraHnswHeuristicType

@brief A strategy for selecting the graph build parameters based on similar HNSW index parameters.

cuvsCagraSearchAlgo

@brief Enum to denote algorithm used to search CAGRA Index

cuvsDatasetLayout_t

@brief Generic dataset layout kind for C API dataset handles.

cuvsDatasetMemType_t

@brief Memory space holding a C API dataset handle’s data.

cuvsDistanceType

enum to tell how to compute distance

cuvsErrorKind_t

@brief The kind of error from a C API call that passed through the exception boundary.

cuvsError_t

@defgroup error_c cuVS Error Messages @{ / /** @brief An enum denoting error statuses for function calls

cuvsFilterType

@brief Enum to denote filter type.

cuvsHnswHierarchy

@brief Hierarchy for HNSW index when converting from CAGRA index

cuvsIvfPqCodebookGen

@defgroup ivf_pq_c_index_params IVF-PQ index build parameters @{ / /** @brief A type for specifying how PQ codebooks are created

cuvsIvfPqListLayout

@brief A type for specifying the memory layout of IVF-PQ list data

cuvsKMeansInitMethod

@defgroup kmeans_c_params k-means hyperparameters @{

cuvsKMeansType

@brief Type of k-means algorithm.

cuvsLogLevel_t

@brief An enum denoting log levels

cuvsMergeStrategy

@brief Strategy for merging indices.

cuvsNNDescentDistCompDtype

@brief Dtype to use for distance computation

cuvsPcaSolver

@brief Solver algorithm for PCA eigen decomposition.

cuvsTieredIndexANNAlgo

@brief Enum to hold which ANN algorithm is being used in the tiered index

Constants

EXPECTED_CUVS_VERSION

cudaMemcpyDeviceToHost

cudaMemcpyHostToDevice

cudaSuccess

Functions

cudaDeviceSynchronize^⚠

cudaFree^⚠

cudaGetDeviceCount^⚠

cudaGetErrorString^⚠

cudaMalloc^⚠

cudaMemcpy^⚠

cudaStreamCreate^⚠

cudaStreamDestroy^⚠

cudaStreamSynchronize^⚠

cuvsAceParamsCreate^⚠

@brief Allocate ACE params, and populate with default values

cuvsAceParamsDestroy^⚠

@brief De-allocate ACE params

cuvsAllNeighborsBuild^⚠

@brief Build an all-neighbors k-NN graph automatically detecting host vs device dataset.

cuvsAllNeighborsIndexParamsCreate^⚠

@brief Create a default all-neighbors index parameters struct.

cuvsAllNeighborsIndexParamsDestroy^⚠

@brief Destroy an all-neighbors index parameters struct.

cuvsBinaryQuantizerCreate^⚠

@brief Allocate Binary Quantizer and populate with default values

cuvsBinaryQuantizerDestroy^⚠

@brief De-allocate Binary Quantizer

cuvsBinaryQuantizerParamsCreate^⚠

@brief Allocate Binary Quantizer params, and populate with default values

cuvsBinaryQuantizerParamsDestroy^⚠

@brief De-allocate Binary Quantizer params

cuvsBinaryQuantizerTrain^⚠

@brief Trains a binary quantizer to be used later for quantizing the dataset.

cuvsBinaryQuantizerTransform^⚠

@brief Applies binary quantization transform to the given dataset

cuvsBinaryQuantizerTransformWithParams^⚠

@brief Applies binary quantization transform to the given dataset

cuvsBruteForceBuild^⚠

@defgroup bruteforce_c_index_build Bruteforce index build @{ / /** @brief Build a BRUTEFORCE index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCUDA, kDLCUDAHost, kDLCUDAManaged, or kDLCPU. Also, acceptable underlying types are: 1. kDLDataType.code == kDLFloat and kDLDataType.bits = 32 2. kDLDataType.code == kDLFloat and kDLDataType.bits = 16

cuvsBruteForceDeserialize^⚠

Load index from file. The serialization format can be subject to changes, therefore loading an index saved with a previous version of cuvs is not guaranteed to work.

cuvsBruteForceIndexCreate^⚠

@brief Allocate BRUTEFORCE index

cuvsBruteForceIndexDestroy^⚠

@brief De-allocate BRUTEFORCE index

cuvsBruteForceSearch^⚠

@defgroup bruteforce_c_index_search Bruteforce index search @{ / /** @brief Search a BRUTEFORCE index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCUDA, kDLCUDAHost, kDLCUDAManaged. It is also important to note that the BRUTEFORCE index must have been built with the same type of queries, such that index.dtype.code == queries.dl_tensor.dtype.code Types for input are: 1. queries: kDLDataType.code == kDLFloat and kDLDataType.bits = 32 or kDLDataType.bits = 16 2. neighbors: kDLDataType.code == kDLUInt and kDLDataType.bits = 32 3. distances: kDLDataType.code == kDLFloat and kDLDataType.bits = 32

cuvsBruteForceSerialize^⚠

@defgroup bruteforce_c_index_serialize BRUTEFORCE C-API serialize functions @{ / /** Save the index to file. The serialization format can be subject to changes, therefore loading an index saved with a previous version of cuvs is not guaranteed to work.

cuvsCagraBuild^⚠

@brief Build a CAGRA index from a dataset handle. Acceptable underlying types are: 1. kDLDataType.code == kDLFloat and kDLDataType.bits = 32 2. kDLDataType.code == kDLFloat and kDLDataType.bits = 16 3. kDLDataType.code == kDLInt and kDLDataType.bits = 8 4. kDLDataType.code == kDLUInt and kDLDataType.bits = 8

cuvsCagraCompressionParamsCreate^⚠

@brief Allocate CAGRA Compression params, and populate with default values

cuvsCagraCompressionParamsDestroy^⚠

@brief De-allocate CAGRA Compression params

cuvsCagraDeserializeGraph^⚠

Load the CAGRA graph from file without retaining a serialized dataset.

cuvsCagraDeserializeGraphAndDataset^⚠

Load the CAGRA graph and dataset from file.

cuvsCagraExtend^⚠

@brief Extend a CAGRA index with a caller-owned pre-concatenated padded dataset.

cuvsCagraExtendParamsCreate^⚠

@brief Allocate CAGRA Extend params, and populate with default values

cuvsCagraExtendParamsDestroy^⚠

@brief De-allocate CAGRA Extend params

cuvsCagraIndexCreate^⚠

@brief Allocate CAGRA index

cuvsCagraIndexDestroy^⚠

@brief De-allocate CAGRA index

cuvsCagraIndexFromArgs^⚠

Load index from a dataset and graph

cuvsCagraIndexGetDataset^⚠

@brief Returns a view of the CAGRA dataset

cuvsCagraIndexGetDims^⚠

@brief Get dimension of the CAGRA index

cuvsCagraIndexGetGraph^⚠

@brief Returns a view of the CAGRA graph

cuvsCagraIndexGetGraphDegree^⚠

@brief Get graph degree of the CAGRA index

cuvsCagraIndexGetSize^⚠

@brief Get size of the CAGRA index

cuvsCagraIndexParamsCreate^⚠

@brief Allocate CAGRA Index params, and populate with default values

cuvsCagraIndexParamsDestroy^⚠

@brief De-allocate CAGRA Index params

cuvsCagraIndexParamsFromDataset^⚠

@brief Create CAGRA index parameters heuristically tuned for a dataset

cuvsCagraIndexParamsFromHnswParams^⚠

@brief Create CAGRA index parameters similar to an HNSW index

cuvsCagraMerge^⚠

@brief Merge multiple CAGRA indices into a single CAGRA index.

cuvsCagraSearch^⚠

@defgroup cagra_c_index_search C API for CUDA ANN Graph-based nearest neighbor search @{ / /** @brief Search a CAGRA index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCUDA, kDLCUDAHost, kDLCUDAManaged. It is also important to note that the CAGRA Index must have been built with the same type of queries, such that index.dtype.code == queries.dl_tensor.dtype.code Types for input are: 1. queries: a. kDLDataType.code == kDLFloat and kDLDataType.bits = 32 b. kDLDataType.code == kDLFloat and kDLDataType.bits = 16 c. kDLDataType.code == kDLInt and kDLDataType.bits = 8 d. kDLDataType.code == kDLUInt and kDLDataType.bits = 8 2. neighbors: kDLDataType.code == kDLUInt and kDLDataType.bits = 32 or kDLDataType.code == kDLInt and kDLDataType.bits = 64 3. distances: kDLDataType.code == kDLFloat and kDLDataType.bits = 32

cuvsCagraSearchMultiPartition^⚠

@brief Search multiple CAGRA index partitions concurrently and return the global top-k per query.

cuvsCagraSearchParamsCreate^⚠

@brief Allocate CAGRA search params, and populate with default values

cuvsCagraSearchParamsDestroy^⚠

@brief De-allocate CAGRA search params

cuvsCagraSerializeGraph^⚠

@defgroup cagra_c_index_serialize CAGRA C-API serialize functions @{ / /** Save the CAGRA graph to file without its dataset.

cuvsCagraSerializeGraphAndDataset^⚠

Save the CAGRA graph and its attached dataset to file.

cuvsCagraSerializeToHnswlib^⚠

Save the CAGRA index to file in hnswlib format. NOTE: The saved index can only be read by the hnswlib wrapper in cuVS, as the serialization format is not compatible with the original hnswlib.

cuvsCagraUpdateDataset^⚠

@brief Update a CAGRA index with a device-padded dataset.

cuvsDatasetCreate^⚠

@brief Create an empty owning dataset handle.

cuvsDatasetDestroy^⚠

@brief Destroy a dataset handle created by a cuvsDatasetMake* function.

cuvsDatasetGetDtype^⚠

@brief Get the element dtype of a dataset handle.

cuvsDatasetGetIsOwning^⚠

@brief Get whether a dataset handle owns its backing storage.

cuvsDatasetGetLayout^⚠

@brief Get the layout of a dataset handle.

cuvsDatasetGetMemType^⚠

@brief Get the memory residency of a dataset handle.

cuvsDatasetMakePadded^⚠

@brief Create an owning padded dataset in the requested memory space.

cuvsDatasetMakePaddedView^⚠

@brief Create a non-owning padded dataset view from a host- or device-resident tensor.

cuvsDatasetMakeStandardView^⚠

@brief Create a non-owning standard dataset view from a host- or device-resident tensor.

cuvsDeviceIdGet^⚠

@brief Get the id of the device associated with this cuvsResources_t

cuvsGetLastErrorKind^⚠

@brief Returns the calling thread’s error kind for the most recent C API call that passed through the exception boundary. Returns CUVS_ERROR_KIND_NONE after success.

cuvsGetLastErrorText^⚠

@brief Returns a string describing the last seen error on this thread, or NULL if the last function succeeded.

cuvsGetLogLevel^⚠

@brief Returns the current log level

cuvsHnswAceParamsCreate^⚠

@brief Allocate HNSW ACE params, and populate with default values

cuvsHnswAceParamsDestroy^⚠

@brief De-allocate HNSW ACE params

cuvsHnswBuild^⚠

@brief Build an HNSW index using ACE (Augmented Core Extraction) algorithm.

cuvsHnswDeserialize^⚠

Load hnswlib index from file which was serialized from a HNSW index. NOTE: When hierarchy is NONE, the loaded hnswlib index is immutable, and only be read by the hnswlib wrapper in cuVS, as the serialization format is not compatible with the original hnswlib. Experimental, both the API and the serialization format are subject to change.

cuvsHnswExtend^⚠

@brief Add new vectors to an HNSW index NOTE: The HNSW index can only be extended when the hierarchy is CPU when converting from a CAGRA index.

cuvsHnswExtendParamsCreate^⚠

@brief Allocate HNSW extend params, and populate with default values

cuvsHnswExtendParamsDestroy^⚠

@brief De-allocate HNSW extend params

cuvsHnswFromCagra^⚠

@brief Convert a CAGRA Index to an HNSW index. NOTE: When hierarchy is: 1. NONE: This method uses the filesystem to write the CAGRA index in /tmp/<random_number>.bin before reading it as an hnswlib index, then deleting the temporary file. The returned index is immutable and can only be searched by the hnswlib wrapper in cuVS, as the format is not compatible with the original hnswlib. 2. CPU: The returned index is mutable and can be extended with additional vectors. The serialized index is also compatible with the original hnswlib library.

cuvsHnswFromCagraWithDataset^⚠

cuvsHnswIndexCreate^⚠

@brief Allocate HNSW index

cuvsHnswIndexDestroy^⚠

@brief De-allocate HNSW index

cuvsHnswIndexParamsCreate^⚠

@brief Allocate HNSW Index params, and populate with default values

cuvsHnswIndexParamsDestroy^⚠

@brief De-allocate HNSW Index params

cuvsHnswSearch^⚠

@defgroup hnsw_c_index_search C API for CUDA ANN Graph-based nearest neighbor search @{ / /** @brief Search a HNSW index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCPU, kDLCUDAHost, or kDLCUDAManaged. It is also important to note that the HNSW Index must have been built with the same type of queries, such that index.dtype.code == queries.dl_tensor.dtype.code Supported types for input are: 1. queries: a. kDLDataType.code == kDLFloat and kDLDataType.bits = 32 b. kDLDataType.code == kDLInt and kDLDataType.bits = 8 c. kDLDataType.code == kDLUInt and kDLDataType.bits = 8 2. neighbors: kDLDataType.code == kDLUInt and kDLDataType.bits = 64 3. distances: kDLDataType.code == kDLFloat and kDLDataType.bits = 32 NOTE: When hierarchy is NONE, the HNSW index can only be searched by the hnswlib wrapper in cuVS, as the format is not compatible with the original hnswlib.

cuvsHnswSearchParamsCreate^⚠

@brief Allocate HNSW search params, and populate with default values

cuvsHnswSearchParamsDestroy^⚠

@brief De-allocate HNSW search params

cuvsHnswSerialize^⚠

@brief Serialize a CAGRA index to a file as an hnswlib index NOTE: When hierarchy is NONE, the saved hnswlib index is immutable and can only be read by the hnswlib wrapper in cuVS, as the serialization format is not compatible with the original hnswlib. However, when hierarchy is CPU, the saved hnswlib index is compatible with the original hnswlib library.

cuvsIvfFlatBuild^⚠

@defgroup ivf_flat_c_index_build IVF-Flat index build @{ / /** @brief Build a IVF-Flat index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCUDA, kDLCUDAHost, kDLCUDAManaged, or kDLCPU. Also, acceptable underlying types are: 1. kDLDataType.code == kDLFloat and kDLDataType.bits = 32 2. kDLDataType.code == kDLInt and kDLDataType.bits = 8 3. kDLDataType.code == kDLUInt and kDLDataType.bits = 8

cuvsIvfFlatDeserialize^⚠

Load index from file.

cuvsIvfFlatExtend^⚠

@defgroup ivf_flat_c_index_extend IVF-Flat index extend @{ / /** @brief Extend the index with the new data.

cuvsIvfFlatIndexCreate^⚠

@brief Allocate IVF-Flat index

cuvsIvfFlatIndexDestroy^⚠

@brief De-allocate IVF-Flat index

cuvsIvfFlatIndexGetCenters^⚠

@brief Get the cluster centers corresponding to the lists [n_lists, dim]

cuvsIvfFlatIndexGetDim^⚠

@brief Get the dimensionality of the indexed data

cuvsIvfFlatIndexGetNLists^⚠

@brief Get the number of clusters/inverted lists in the index

cuvsIvfFlatIndexParamsCreate^⚠

@brief Allocate IVF-Flat Index params, and populate with default values

cuvsIvfFlatIndexParamsDestroy^⚠

@brief De-allocate IVF-Flat Index params

cuvsIvfFlatSearch^⚠

@defgroup ivf_flat_c_index_search IVF-Flat index search @{ / /** @brief Search a IVF-Flat index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCUDA, kDLCUDAHost, kDLCUDAManaged. It is also important to note that the IVF-Flat Index must have been built with the same type of queries, such that index.dtype.code == queries.dl_tensor.dtype.code Types for input are: 1. queries: kDLDataType.code == kDLFloat and kDLDataType.bits = 32 2. neighbors: kDLDataType.code == kDLUInt and kDLDataType.bits = 32 3. distances: kDLDataType.code == kDLFloat and kDLDataType.bits = 32

cuvsIvfFlatSearchParamsCreate^⚠

@brief Allocate IVF-Flat search params, and populate with default values

cuvsIvfFlatSearchParamsDestroy^⚠

@brief De-allocate IVF-Flat search params

cuvsIvfFlatSerialize^⚠

@defgroup ivf_flat_c_index_serialize IVF-Flat C-API serialize functions @{ / /** Save the index to file.

cuvsIvfPqBuild^⚠

@defgroup ivf_pq_c_index_build IVF-PQ index build @{ / /** @brief Build a IVF-PQ index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCUDA, kDLCUDAHost, kDLCUDAManaged, or kDLCPU. Also, acceptable underlying types are: 1. kDLDataType.code == kDLFloat and kDLDataType.bits = 32 2. kDLDataType.code == kDLFloat and kDLDataType.bits = 16 3. kDLDataType.code == kDLInt and kDLDataType.bits = 8 4. kDLDataType.code == kDLUInt and kDLDataType.bits = 8

cuvsIvfPqBuildPrecomputed^⚠

@brief Build a view-type IVF-PQ index from device memory precomputed centroids and codebook.

cuvsIvfPqDeserialize^⚠

Load index from file.

cuvsIvfPqExtend^⚠

@defgroup ivf_pq_c_index_extend IVF-PQ index extend @{ / /** @brief Extend the index with the new data.

cuvsIvfPqIndexCreate^⚠

@brief Allocate IVF-PQ index

cuvsIvfPqIndexDestroy^⚠

@brief De-allocate IVF-PQ index

cuvsIvfPqIndexGetCenters^⚠

@brief Get the cluster centers corresponding to the lists in the original space

cuvsIvfPqIndexGetCentersPadded^⚠

@brief Get the padded cluster centers [n_lists, dim_ext] where dim_ext = round_up(dim + 1, 8)

cuvsIvfPqIndexGetCentersRot^⚠

@brief Get the rotated cluster centers [n_lists, rot_dim] where rot_dim = pq_len * pq_dim

cuvsIvfPqIndexGetDim^⚠

Get the dimensionality

cuvsIvfPqIndexGetListIndices^⚠

@brief Get the indices of each vector in a ivf-pq list

cuvsIvfPqIndexGetListSizes^⚠

@brief Get the sizes of each list

cuvsIvfPqIndexGetNLists^⚠

Get the number of clusters/inverted lists

cuvsIvfPqIndexGetPqBits^⚠

Get the bit length of an encoded vector element after compression by PQ.

cuvsIvfPqIndexGetPqCenters^⚠

@brief Get the PQ cluster centers

cuvsIvfPqIndexGetPqDim^⚠

Get the dimensionality of an encoded vector after compression by PQ.

cuvsIvfPqIndexGetPqLen^⚠

Get the Dimensionality of a subspace, i.e. the number of vector components mapped to a subspace

cuvsIvfPqIndexGetRotationMatrix^⚠

@brief Get the rotation matrix [rot_dim, dim] Transform matrix (original space -> rotated padded space)

cuvsIvfPqIndexGetSize^⚠

Get the size of the index

cuvsIvfPqIndexParamsCreate^⚠

@brief Allocate IVF-PQ Index params, and populate with default values

cuvsIvfPqIndexParamsDestroy^⚠

@brief De-allocate IVF-PQ Index params

cuvsIvfPqIndexUnpackContiguousListData^⚠

@brief Unpack n_rows consecutive PQ encoded vectors of a single list (cluster) in the compressed index starting at given offset, not expanded to one code per byte. Each code in the output buffer occupies ceildiv(index.pq_dim() * index.pq_bits(), 8) bytes.

cuvsIvfPqSearch^⚠

@defgroup ivf_pq_c_index_search IVF-PQ index search @{ / /** @brief Search a IVF-PQ index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCUDA, kDLCUDAHost, kDLCUDAManaged. It is also important to note that the IVF-PQ Index must have been built with the same type of queries, such that index.dtype.code == queries.dl_tensor.dtype.code Types for input are: 1. queries: kDLDataType.code == kDLFloat and kDLDataType.bits = 32 or kDLDataType.bits = 16 2. neighbors: kDLDataType.code == kDLUInt and kDLDataType.bits = 32 3. distances: kDLDataType.code == kDLFloat and kDLDataType.bits = 32

cuvsIvfPqSearchParamsCreate^⚠

@brief Allocate IVF-PQ search params, and populate with default values

cuvsIvfPqSearchParamsDestroy^⚠

@brief De-allocate IVF-PQ search params

cuvsIvfPqSerialize^⚠

@defgroup ivf_pq_c_index_serialize IVF-PQ C-API serialize functions @{ / /** Save the index to file.

cuvsIvfPqTransform^⚠

@defgroup ivf_pq_c_index_transform IVF-PQ index transform @{ / /** @brief Transform the input data by applying pq-encoding

cuvsIvfSqBuild^⚠

@defgroup ivf_sq_c_index_build IVF-SQ index build @{ / /** @brief Build an IVF-SQ index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCUDA, kDLCUDAHost, kDLCUDAManaged, or kDLCPU. Also, acceptable underlying types are: 1. kDLDataType.code == kDLFloat and kDLDataType.bits = 32 2. kDLDataType.code == kDLFloat and kDLDataType.bits = 16

cuvsIvfSqDeserialize^⚠

Load index from file.

cuvsIvfSqExtend^⚠

@defgroup ivf_sq_c_index_extend IVF-SQ index extend @{ / /** @brief Extend the index with the new data.

cuvsIvfSqIndexCreate^⚠

@brief Allocate IVF-SQ index

cuvsIvfSqIndexDestroy^⚠

@brief De-allocate IVF-SQ index

cuvsIvfSqIndexGetCenters^⚠

@brief Get the cluster centers corresponding to the lists [n_lists, dim]

cuvsIvfSqIndexGetDim^⚠

Get the dimensionality of the data

cuvsIvfSqIndexGetNLists^⚠

Get the number of clusters/inverted lists

cuvsIvfSqIndexGetSize^⚠

Get the size of the index

cuvsIvfSqIndexParamsCreate^⚠

@brief Allocate IVF-SQ Index params, and populate with default values

cuvsIvfSqIndexParamsDestroy^⚠

@brief De-allocate IVF-SQ Index params

cuvsIvfSqSearch^⚠

@defgroup ivf_sq_c_index_search IVF-SQ index search @{ / /** @brief Search an IVF-SQ index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCUDA, kDLCUDAHost, kDLCUDAManaged. Types for input are: 1. queries: kDLDataType.code == kDLFloat and kDLDataType.bits = 32 or 16 2. neighbors: kDLDataType.code == kDLInt and kDLDataType.bits = 64 3. distances: kDLDataType.code == kDLFloat and kDLDataType.bits = 32

cuvsIvfSqSearchParamsCreate^⚠

@brief Allocate IVF-SQ search params, and populate with default values

cuvsIvfSqSearchParamsDestroy^⚠

@brief De-allocate IVF-SQ search params

cuvsIvfSqSerialize^⚠

@defgroup ivf_sq_c_index_serialize IVF-SQ C-API serialize functions @{ / /** Save the index to file.

cuvsKMeansClusterCost^⚠

@brief Compute cluster cost

cuvsKMeansFit^⚠

@brief Find clusters with k-means algorithm.

cuvsKMeansParamsCreate^⚠

@brief Allocate KMeans params, and populate with default values

cuvsKMeansParamsDestroy^⚠

@brief De-allocate KMeans params

cuvsKMeansPredict^⚠

@brief Predict the closest cluster each sample in X belongs to.

cuvsMatrixCopy^⚠

@brief Copy a matrix

cuvsMatrixSliceRows^⚠

@brief Slices rows from a matrix

cuvsMultiGpuResourcesCreate^⚠

@brief Create an Initialized opaque C handle for C++ type raft::device_resources_snmg for multi-GPU operations

cuvsMultiGpuResourcesCreateWithDeviceIds^⚠

@brief Create an Initialized opaque C handle for C++ type raft::device_resources_snmg for multi-GPU operations with specific device IDs

cuvsMultiGpuResourcesDestroy^⚠

@brief Destroy and de-allocate opaque C handle for C++ type raft::device_resources_snmg

cuvsMultiGpuResourcesSetMemoryPool^⚠

@brief Set a memory pool on all devices managed by the multi-GPU resources

cuvsNNDescentBuild^⚠

@defgroup nn_descent_c_index_build NN-Descent index build @{ / /** @brief Build a NN-Descent index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCUDA, kDLCUDAHost, kDLCUDAManaged, or kDLCPU. Also, acceptable underlying types are: 1. kDLDataType.code == kDLFloat and kDLDataType.bits = 32 2. kDLDataType.code == kDLFloat and kDLDataType.bits = 16 3. kDLDataType.code == kDLInt and kDLDataType.bits = 8 4. kDLDataType.code == kDLUInt and kDLDataType.bits = 8

cuvsNNDescentIndexCreate^⚠

@brief Allocate NN-Descent index

cuvsNNDescentIndexDestroy^⚠

@brief De-allocate NN-Descent index

cuvsNNDescentIndexGetDistances^⚠

@brief Get the distances from a build NN_Descent index

cuvsNNDescentIndexGetGraph^⚠

@brief Get the KNN graph from a built NN-Descent index

cuvsNNDescentIndexParamsCreate^⚠

@brief Allocate NN-Descent Index params, and populate with default values

cuvsNNDescentIndexParamsDestroy^⚠

@brief De-allocate NN-Descent Index params

cuvsPairwiseDistance^⚠

@brief Compute pairwise distances for two matrices

cuvsPcaFit^⚠

@brief Perform PCA fit operation.

cuvsPcaFitTransform^⚠

@brief Perform PCA fit and transform in a single operation.

cuvsPcaInverseTransform^⚠

@brief Perform PCA inverse transform operation.

cuvsPcaParamsCreate^⚠

@brief Allocate PCA params and populate with default values.

cuvsPcaParamsDestroy^⚠

@brief De-allocate PCA params.

cuvsPcaTransform^⚠

@brief Perform PCA transform operation.

cuvsProductQuantizerBuild^⚠

@brief Builds a product quantizer to be used later for quantizing the dataset.

cuvsProductQuantizerCreate^⚠

@brief Allocate Product Quantizer

cuvsProductQuantizerDestroy^⚠

@brief De-allocate Product Quantizer

cuvsProductQuantizerGetEncodedDim^⚠

@brief Get the encoded dimension of the quantized dataset.

cuvsProductQuantizerGetPqBits^⚠

@brief Get the bit length of the vector element after compression by PQ.

cuvsProductQuantizerGetPqCodebook^⚠

@brief Get the PQ codebook.

cuvsProductQuantizerGetPqDim^⚠

@brief Get the dimensionality of the vector after compression by PQ.

cuvsProductQuantizerGetUseVq^⚠

@brief Get whether VQ is used.

cuvsProductQuantizerGetVqCodebook^⚠

@brief Get the VQ codebook.

cuvsProductQuantizerInverseTransform^⚠

@brief Applies product quantization inverse transform to the given quantized codes

cuvsProductQuantizerParamsCreate^⚠

@brief Allocate Product Quantizer params, and populate with default values

cuvsProductQuantizerParamsDestroy^⚠

@brief De-allocate Product Quantizer params

cuvsProductQuantizerTransform^⚠

@brief Applies product quantization transform to the given dataset

cuvsRMMAlloc^⚠

@brief Allocates device memory using RMM

cuvsRMMAsyncMemoryResourceEnable^⚠

@brief Switches the working memory resource to use stream-ordered asynchronous allocation (cudaMallocAsync / cudaFreeAsync). Unlike the pool resource, this resource returns memory to the stream immediately without blocking the CPU, eliminating device-wide synchronization on deallocation. This is especially beneficial when multiple CAGRA searches run concurrently on separate CUDA streams, because the internal workspace allocations no longer serialize kernel launches. Be aware that this function will change the memory resource for the whole process and the new memory resource will be used until explicitly changed.

cuvsRMMFree^⚠

@brief Deallocates device memory using RMM

cuvsRMMHostAlloc^⚠

@brief Allocates pinned memory on the host using RMM @param[out] ptr Pointer to allocated host memory @param[in] bytes Size in bytes to allocate @return cuvsError_t

cuvsRMMHostFree^⚠

@brief Deallocates pinned memory on the host using RMM @param[in] ptr Pointer to allocated host memory to free @param[in] bytes Size in bytes to deallocate @return cuvsError_t

cuvsRMMMemoryResourceReset^⚠

@brief Resets the memory resource to use the default memory resource (cuda_memory_resource) @return cuvsError_t

cuvsRMMPoolMemoryResourceEnable^⚠

@brief Switches the working memory resource to use the RMM pool memory resource, which will bypass unnecessary synchronizations by allocating a chunk of device memory up front and carving that up for temporary memory allocations within algorithms. Be aware that this function will change the memory resource for the whole process and the new memory resource will be used until explicitly changed.

cuvsRefine^⚠

@defgroup ann_refine_c Approximate Nearest Neighbors Refinement C-API @{ / /** @brief Refine nearest neighbor search.

cuvsResourcesCreate^⚠

@brief Create an Initialized opaque C handle for C++ type raft::resources

cuvsResourcesCreateWithMemoryTracking^⚠

@brief Create an opaque C handle for C++ type raft::resources whose memory allocations are tracked and written as CSV samples from a background thread.

cuvsResourcesDestroy^⚠

@brief Destroy and de-allocate opaque C handle for C++ type raft::resources

cuvsResourcesGetStreamPoolSize^⚠

@brief Get the number of streams in a resources handle’s RAFT CUDA stream pool.

cuvsResourcesGetStreamPoolStream^⚠

@brief Borrow a CUDA stream from a resources handle’s RAFT CUDA stream pool.

cuvsResourcesSetStreamPool^⚠

@brief Install a RAFT CUDA stream pool on a resources handle, replacing any existing pool.

cuvsResourcesSetWorkspacePool^⚠

@brief Configure the temporary workspace on this resources object as an uncapped pool, backed by the current device memory resource. After the initial reservation is allocated on first use, subsequent calls to cuvsRMMAlloc / cuvsRMMFree on the same resources handle hit the pool cache rather than calling cudaMallocAsync / cudaFreeAsync, reducing CUDA context lock contention under concurrent query threads. The pool grows without shrinking: freed allocations are returned to the pool rather than to the device, so the pool’s high-water mark only increases until the resources object is destroyed.

cuvsScalarQuantizerCreate^⚠

@brief Allocate Scalar Quantizer and populate with default values

cuvsScalarQuantizerDestroy^⚠

@brief De-allocate Scalar Quantizer

cuvsScalarQuantizerInverseTransform^⚠

@brief Perform inverse quantization step on previously quantized dataset

cuvsScalarQuantizerParamsCreate^⚠

@brief Allocate Scalar Quantizer params, and populate with default values

cuvsScalarQuantizerParamsDestroy^⚠

@brief De-allocate Scalar Quantizer params

cuvsScalarQuantizerTrain^⚠

@brief Trains a scalar quantizer to be used later for quantizing the dataset.

cuvsScalarQuantizerTransform^⚠

@brief Applies quantization transform to given dataset

cuvsSelectK^⚠

@brief Select the k smallest values from a flat device array of n candidates.

cuvsSelectKBatched^⚠

@brief Select the k smallest or largest values from each row of a device matrix.

cuvsSetLastErrorText^⚠

@brief Sets a string describing an error seen on the thread. Passing NULL clears any previously seen error message.

cuvsSetLogLevel^⚠

@brief Sets the log level

cuvsStreamGet^⚠

@brief Get the cudaStream_t from a cuvsResources_t

cuvsStreamSet^⚠

@brief Set cudaStream_t on cuvsResources_t to queue CUDA kernels on APIs that accept a cuvsResources_t handle

cuvsStreamSync^⚠

@brief Syncs the current CUDA stream on the resources object

cuvsTieredIndexBuild^⚠

@defgroup tieredindex_c_index_build Tiered index build @{ / /** @brief Build a TieredIndex index with a DLManagedTensor which has underlying DLDeviceType equal to kDLCUDA, kDLCUDAHost, kDLCUDAManaged, or kDLCPU. Also, acceptable underlying types are: 1. kDLDataType.code == kDLFloat and kDLDataType.bits = 32 2. kDLDataType.code == kDLFloat and kDLDataType.bits = 16

cuvsTieredIndexCreate^⚠

@brief Allocate Tiered Index

cuvsTieredIndexDestroy^⚠

@brief De-allocate Tiered index

cuvsTieredIndexExtend^⚠

@} / /** @defgroup tiered_c_index_extend Tiered index extend @{ / /** @brief Extend the index with the new data.

cuvsTieredIndexMerge^⚠

@defgroup tiered_c_index_merge Tiered index merge @{ / /** @brief Merge multiple indices together into a single index

cuvsTieredIndexParamsCreate^⚠

@brief Allocate Tiered Index Params and populate with default values

cuvsTieredIndexParamsDestroy^⚠

@brief De-allocate Tiered Index params

cuvsTieredIndexSearch^⚠

@defgroup tieredindex_c_index_search Tiered index search @{ / /** @brief Search a TieredIndex index with a DLManagedTensor

cuvsVamanaBuild^⚠

@brief Build Vamana index

cuvsVamanaIndexCreate^⚠

@brief Allocate Vamana index

cuvsVamanaIndexDestroy^⚠

@brief De-allocate Vamana index

cuvsVamanaIndexGetDims^⚠

@brief Get the dimension of the index

cuvsVamanaIndexParamsCreate^⚠

@brief Allocate Vamana Index params, and populate with default values

cuvsVamanaIndexParamsDestroy^⚠

@brief De-allocate Vamana Index params

cuvsVamanaSerialize^⚠

@brief Save Vamana index to file

cuvsVersionGet^⚠

@brief Get the version of the cuVS library @param[out] major Major version @param[out] minor Minor version @param[out] patch Patch version @return cuvsError_t

validate_runtime_version

Query and validate the exact native cuVS version required by this binding.

Type Aliases

cudaError_t

cudaMemcpyKind

cudaStream_t

cuvsAceParams_t

cuvsAllNeighborsIndexParams_t

cuvsBinaryQuantizerParams_t

cuvsBinaryQuantizer_t

cuvsBruteForceIndex_t

cuvsCagraCompressionParams_t

cuvsCagraExtendParams_t

cuvsCagraIndexParams_t

cuvsCagraIndex_t

cuvsCagraSearchParams_t

cuvsDataset_t

cuvsHnswAceParams_t

cuvsHnswExtendParams_t

cuvsHnswIndexParams_t

cuvsHnswIndex_t

cuvsHnswSearchParams_t

cuvsIvfFlatIndexParams_t

cuvsIvfFlatIndex_t

cuvsIvfFlatSearchParams_t

cuvsIvfPqIndexParams_t

cuvsIvfPqIndex_t

cuvsIvfPqParams_t

cuvsIvfPqSearchParams_t

cuvsIvfSqIndexParams_t

cuvsIvfSqIndex_t

cuvsIvfSqSearchParams_t

cuvsKMeansParams_t

cuvsNNDescentIndexParams_t

cuvsNNDescentIndex_t

cuvsPcaParams_t

cuvsProductQuantizerParams_t

cuvsProductQuantizer_t

cuvsResources_t

@brief An opaque C handle for C++ type raft::resources

cuvsScalarQuantizerParams_t

cuvsScalarQuantizer_t

cuvsTieredIndexParams_t

cuvsTieredIndex_t

cuvsVamanaIndexParams_t

cuvsVamanaIndex_t