NewIvfFlatIndex()

This function creates an IVF_FLAT index configuration that partitions vectors into clusters for balanced accuracy and speed.

Go

func NewIvfFlatIndex(metricType MetricType, nlist int) 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.

RETURNS:

Index

An index configuration instance. Pass this to CreateIndex() via the index option.

Example

Go

import (
	"github.com/milvus-io/milvus/client/v2/index"
	"github.com/milvus-io/milvus/client/v2/milvusclient"
)

// Create index configuration
idx := index.NewIvfFlatIndex(index.COSINE, 128)

// Use with CreateIndex
createIdxOption := milvusclient.NewCreateIndexOption("collection_name", "vector_field", idx)
task, err := client.CreateIndex(ctx, createIdxOption)