NewSCANNIndex()
This function creates a SCANN (Scalable Nearest Neighbors) index configuration for fast approximate search.
func NewSCANNIndex(metricType MetricType, nlist int, withRawData bool) Index
PARAMETERS:
-
The distance metric type for similarity search (e.g., index.COSINE, index.L2, index.IP).
nlist (int)
The number of cluster units (inverted lists). Higher values speed up search but reduce recall. Typical range: 1-65536.
withRawData (bool)
Whether to store raw vectors alongside the index for reranking.
RETURNS:
An index configuration instance. Pass this to CreateIndex() via the index option.
Example
import (
"github.com/milvus-io/milvus/client/v2/index"
"github.com/milvus-io/milvus/client/v2/milvusclient"
)
// Create index configuration
idx := index.NewSCANNIndex(index.COSINE, 128, true)
// Use with CreateIndex
createIdxOption := milvusclient.NewCreateIndexOption("collection_name", "vector_field", idx)
task, err := client.CreateIndex(ctx, createIdxOption)