Vector Index Management

Overview

The Vector Index Management endpoints allow you to configure and manage vector indexes for your graphs. Vector indexes are essential for efficient similarity-based searches and significantly improve the performance of vector operations. These endpoints provide functionality for:

Reading vector index configuration and statistics
Enabling vector indexes with custom parameters
Rebuilding indexes to optimize performance
Deleting indexes when no longer needed

Vector indexes are particularly important for:

Accelerating vector similarity searches
Supporting large-scale vector operations
Optimizing memory usage for vector data
Enabling real-time vector search capabilities

Read Configuration

Retrieve the current vector index configuration for a graph using GET: /v1.0/tenants/{tenant-guid}/graphs/{graph-guid}/vectorindex/config. This endpoint returns the current index settings including type, parameters, and status information.

curl --location --request GET 'http://localhost:8701/v1.0/tenants/00000000-0000-0000-0000-000000000000/graphs/00000000-0000-0000-0000-000000000000/vectorindex/config' \
--header 'Authorization: Bearer litegraphadmin'

import { LiteGraphSdk } from "litegraphdb";

var api = new LiteGraphSdk(
  "http://localhost:8701/",
  "<Tenant-Guid>",
  "*******"
);

const readVectorIndexConfig = async () => {
  try {
    const data = await api.Graph.readVectorIndexConfig(guid);
    console.log(data, "check data");
  } catch (err) {
    console.log("err:", JSON.stringify(err));
  }
};

import litegraph

sdk = litegraph.configure(
    endpoint="http://localhost:8701",
    tenant_guid="Tenant-Guid",
    access_key="******"
)

def read_config():
    config = litegraph.VectorIndex.get_config(graph_guid="graph-guid")
    print(config)

read_config()

Response

{
  "VectorIndexType": "HnswSqlite",
  "VectorIndexFile": "graph-00000000-0000-0000-0000-000000000000-hnsw.db",
  "VectorDimensionality": 384,
  "VectorIndexM": 16,
  "VectorIndexEf": 50,
  "VectorIndexEfConstruction": 200
}

Read Statistics

Retrieve performance statistics and metrics for the vector index using GET: /v1.0/tenants/{tenant-guid}/graphs/{graph-guid}/vectorindex/stats. This endpoint provides insights into index performance, size, and usage metrics.

curl --location --request GET 'http://localhost:8701/v1.0/tenants/00000000-0000-0000-0000-000000000000/graphs/00000000-0000-0000-0000-000000000000/vectorindex/stats' \
--header 'Authorization: Bearer litegraphadmin'

import { LiteGraphSdk } from "litegraphdb";

var api = new LiteGraphSdk(
  "http://localhost:8701/",
  "<Tenant-Guid>",
  "*******"
);

const readVectorIndexStats = async () => {
  try {
    const data = await api.Graph.readVectorIndexStats(guid);
    console.log(data, "check data");
  } catch (err) {
    console.log("err:", JSON.stringify(err));
  }
};

import litegraph

sdk = litegraph.configure(
    endpoint="http://localhost:8701",
    tenant_guid="Tenant-Guid",
    access_key="******",
)

def get_stats():
    stats = litegraph.VectorIndex.get_stats(graph_guid="graph-guid")
    print(stats)

get_stats()

Response

{
  "VectorCount": 0,
  "Dimensions": 384,
  "IndexType": "HnswSqlite",
  "M": 16,
  "EfConstruction": 200,
  "DefaultEf": 50,
  "IndexFile": "indexes\\graph-00000000-0000-0000-0000-000000000000-hnsw.db",
  "IndexFileSizeBytes": 0,
  "EstimatedMemoryBytes": 0,
  "LastRebuildUtc": "2025-09-04T13:25:45.912460Z",
  "IsLoaded": true,
  "DistanceMetric": "Cosine"
}

Enable Vector Index

Enable a vector index for a graph using PUT: /v2.0/tenants/{tenant-guid}/graphs/{graph-guid}/vectorindex/enable. This endpoint allows you to configure and activate vector indexing with custom parameters to optimize search performance.

Configuration Parameters

The enable request supports the following parameters:

VectorIndexType: The type of vector index to use (e.g., "HnswSqlite")
VectorIndexFile: The filename for the index storage
VectorIndexThreshold: Optional threshold value for index operations
VectorDimensionality: The dimensionality of vectors in the index
VectorIndexM: The number of bi-directional links for each node (HNSW parameter)
VectorIndexEf: The size of the dynamic candidate list (HNSW parameter)
VectorIndexEfConstruction: The size of the dynamic candidate list during construction (HNSW parameter)

curl --location --request PUT 'http://localhost:8701/v2.0/tenants/00000000-0000-0000-0000-000000000000/graphs/00000000-0000-0000-0000-000000000000/vectorindex/enable' \
--header 'Authorization: Bearer litegraphadmin' \
--header 'Content-Type: application/json' \
--data-raw '{
    "VectorIndexType": "HnswSqlite",
    "VectorIndexFile": "graph-00000000-0000-0000-0000-000000000000-hnsw.db",
    "VectorIndexThreshold": null,
    "VectorDimensionality": 384,
    "VectorIndexM": 16,
    "VectorIndexEf": 50,
    "VectorIndexEfConstruction": 200
}'

import { LiteGraphSdk } from "litegraphdb";

var api = new LiteGraphSdk(
  "http://localhost:8701/",
  "<Tenant-Guid>",
  "*******"
);

const enableVectorIndex = async () => {
  try {
    const data = await api.Graph.enableVectorIndex(guid, {
      VectorIndexType: "HnswSqlite",
      VectorIndexFile: "graph-00000000-0000-0000-0000-000000000000-hnsw.db",
      VectorIndexThreshold: null,
      VectorDimensionality: 384,
      VectorIndexM: 16,
      VectorIndexEf: 50,
      VectorIndexEfConstruction: 200,
    });
    console.log(data, "check data");
  } catch (err) {
    console.log("err:", JSON.stringify(err));
  }
};

import litegraph

sdk = litegraph.configure(
    endpoint="http://localhost:8701",
    tenant_guid="Tenant-Guid",
    access_key="******",
)

def enable_vector_index():
    vector_index = litegraph.VectorIndex.enable(
        graph_guid="00000000-0000-0000-0000-000000000000",
        config=litegraph.HnswLiteVectorIndexModel(
            VectorIndexType="HnswSqlite",
            VectorIndexFile="graph-00000000-0000-0000-0000-000000000000-hnsw.db",
            VectorDimensionality=384,
            M=16,
            DefaultEf=50,
            EfConstruction=200
        )
    )
    print(vector_index)

enable_vector_index()

Response

{
  "VectorIndexType": "HnswSqlite",
  "VectorIndexFile": "graph-00000000-0000-0000-0000-000000000000-hnsw.db",
  "VectorDimensionality": 384,
  "VectorIndexM": 16,
  "VectorIndexEf": 50,
  "VectorIndexEfConstruction": 200
}

Rebuild Vector Index

Rebuild the vector index to optimize performance and update the index structure using POST: /v2.0/tenants/{tenant-guid}/graphs/{graph-guid}/vectorindex/rebuild. This operation is useful when the index becomes fragmented or when you want to optimize performance after significant data changes.

curl --location --request POST 'http://localhost:8701/v2.0/tenants/00000000-0000-0000-0000-000000000000/graphs/00000000-0000-0000-0000-000000000000/vectorindex/rebuild' \
--header 'Authorization: Bearer litegraphadmin' \
--data-raw ''

import { LiteGraphSdk } from "litegraphdb";

var api = new LiteGraphSdk(
  "http://localhost:8701/",
  "<Tenant-Guid>",
  "*******"
);

const rebuildVectorIndex = async () => {
  try {
    const data = await api.Graph.rebuildVectorIndex(guid);
    console.log(data, "check data");
  } catch (err) {
    console.log("err:", JSON.stringify(err));
  }
};

import litegraph

sdk = litegraph.configure(
    endpoint="http://localhost:8701",
    tenant_guid="Tenant-Guid",
    access_key="******",
)

def rebuild_vector_index():
    vector_index = litegraph.VectorIndex.rebuild(
        graph_guid="graph-guid",

    )
    print(vector_index)

rebuild_vector_index()

Response

Upon successful rebuild, the API returns a 200 OK status code. No response body is returned for this operation.

Delete Vector Index

Remove the vector index for a graph using DELETE: /v2.0/tenants/{tenant-guid}/graphs/{graph-guid}/vectorindex. This operation permanently removes the index and all associated data. Use this when you no longer need vector search capabilities for the graph.

curl --location --request DELETE 'http://localhost:8701/v2.0/tenants/00000000-0000-0000-0000-000000000000/graphs/00000000-0000-0000-0000-000000000000/vectorindex' \
--header 'Authorization: Bearer litegraphadmin' \
--data-raw ''

import { LiteGraphSdk } from "litegraphdb";

var api = new LiteGraphSdk(
  "http://localhost:8701/",
  "<Tenant-Guid>",
  "*******"
);

const deleteVectorIndex = async () => {
  try {
    const data = await api.Graph.deleteVectorIndex(guid);
    console.log(data, "check data");
  } catch (err) {
    console.log("err:", JSON.stringify(err));
  }
};

import litegraph

sdk = litegraph.configure(
    endpoint="http://localhost:8701",
    tenant_guid="Tenant-Guid",
    access_key="******",
)

def delete_vector_index():
    litegraph.VectorIndex.delete(graph_guid="graph-guid")
    print("Vector index deleted")

delete_vector_index()

Response

Upon successful delete, the API returns a 200 OK status code. No response body is returned for this operation.

Next Steps

After managing vector indexes, you can:

Perform optimized vector similarity searches
Monitor index performance and statistics
Implement vector-based recommendation systems
Build semantic search applications
Optimize vector operations for large-scale deployments