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:

• metricType (MetricType)

  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:

Index

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)