DiskANN is a disk-based approximate nearest neighbor search technology designed for very large datasets. It uses the Vamana graph algorithm to provide efficient vector indexing and retrieval with disk storage under limited memory.
Index build parameters
Parameter | Type | Description |
proxima.diskann.builder.pq_dimension_scale | Int | The PQ quantization ratio. PQ dimension = vector dimension / pq_dimension_scale. Default: 8. |
proxima.diskann.builder.max_degree | Int | The maximum number of outgoing edges of a node in the DiskANN graph. A larger value produces a more accurate graph but increases graph construction time. |
proxima.diskann.builder.list_size | Int | The size of the candidate set for edge selection during DiskANN graph construction. A larger value produces a more accurate graph but increases graph construction time. |
proxima.diskann.builder.thread_count | Int | The number of threads used to build the index. |
Index search parameters
Parameter | Type | Description |
proxima.diskann.searcher.list_size | Int | The size of the result candidate set. A larger value increases recall but also increases query time. |
proxima.diskann.searcher.io_limit | Int | The disk I/O limit per query. At most io_limit disk read operations are performed. This value mainly affects the number of Vamana graph traversal operations. A larger value increases recall, the number of I/O operations, and query time. |
proxima.diskann.searcher.beam_search_width | Int | The number of parallel I/O operations. |