Get desktop application:
View/edit binary Protocol Buffers messages
DiskAnn Index
Used in:
Graph spec version
Graph file
r parameter
alpha parameter
L parameter
/ Entry points to the graph
Flat Index
Used in:
(message has no fields)
Inverted Index File Metadata.
Used in:
Centroids of partitions. `dimension * num_partitions` of float32s. Deprecated, use centroids_tensor instead.
[deprecated = true];
File offset of each partition.
Number of records in the partition.
Tensor of centroids. `num_partitions * dimension` of float32s.
KMeans loss.
The unique index name in the dataset.
Columns to be used to build the index.
The version of the dataset this index was built from.
The [`IndexType`] of the index.
/ Index implementation details.
The type of an index.
Used in:
Vector index
Product Quantization.
Used in:
The number of bits to present a centroid.
Number of sub vectors.
Vector dimension
Codebook. `dimension * 2 ^ num_bits` of float32s.
Tensor of codebook. `2 ^ num_bits * dimension` of floats.
Used in:
,Data shape, [dim1, dim2, ...]
Data buffer
Used in:
A transform matrix to apply to a vector or vectors.
Used in:
The file offset the matrix is stored
Data shape of the matrix, [rows, cols].
Transform type.
Transform type
Used in:
Vector Index Metadata
Used in:
Index specification version.
Vector dimension;
Composed vector index stages. For example, `IVF_PQ` index type can be expressed as: ```no_run,ignore let stages = vec![Ivf{}, PQ{num_bits: 8, num_sub_vectors: 16}] ```
Vector distance metrics type
One stage in the vector index pipeline.
Used in:
Flat index
`IVF` - Inverted File
Product Quantization
Transformer
DiskANN
Metric Type for Vector Index
Used in:
L2 (Euclidean) Distance
Cosine Distance
Dot Product
Hamming Distance