Set up Vector Search with Private Service Connect

Stay organized with collections Save and categorize content based on your preferences.

Private Service Connect allows private consumption of services across VPC networks that belong to different groups, teams, projects, and organizations. You can publish and consume services using IP addresses that you define and that are internal to your VPC network, and for Vector Search endpoints to perform vector similarity searches.

Enabling Private Service Connect on a Vector Search endpoint is suited for use cases that:

1. Require low latency and a secure connection to Vector Search serving backends.
2. Have limited IP space for exclusive VPC peering reservation.
3. Need to access the serving backends from multiple user VPC networks.

To learn more about setting up Private Service Connect, go to Private Service Connect Overview in the Virtual Private Cloud (VPC) documentation.

Create the index endpoint

You must enable Private Service Connect when you create your endpoint. This procedure is similar to creating other endpoints in Vertex AI.

PROJECT=PROJECT_ID
VPC_PROJECT=VPC_PROJECT_ID
REGION=us-central1
VERTEX_ENDPOINT=REGION-aiplatform.googleapis.com
curl -H "Content-Type: application/json" \
  -H "Authorization: Bearer `gcloud auth print-access-token`" \
  https://REGION-aiplatform.googleapis.com/v1/projects/PROJECT/locations/REGION/indexEndpoints \
  -d '{
     "displayName": "DISPLAY_NAME",
     "privateServiceConnectConfig": {
       "enablePrivateServiceConnect": true,
       "projectAllowlist": [ "VPC_PROJECT_1", "VPC_PROJECT_2", "VPC_PROJECT_N"]
     }
   }

About index deployment options

You can deploy your index with automatic or manual service connectivity.

• Deploy with Private Service Connect automation: Set up a service connection policy and deploy your indexes. Setting up a service connection policy lets you deploy to a specific network without creating a compute address and forwarding rule each time.
• Deploy with manual connection: Deploy your index and manually create a compute address and forwarding rule. You might choose this option if you need to use multiple IP addresses for the same service attachment URI, although this is not a common use case.

Deploy with Private Service Connect automation

You can set up a service connection policy so that you don't have to manually create a compute address and forwarding rule after each index deployment.

1. First, create a service connection policy that specifies the network, service class, and region to deploy indexes to. This is a one-time setup. If you've already done this, skip to the next procedure.
2. Deploy the index.

Limitations

Automation allows only one IP address per project per network. If you need to use multiple IP addresses, see Manually deploy the index.

Create a service connection policy

You must be a network administrator to create a service connection policy for automating index deployment.

To automate index deployment, follow these steps:

1. Create your service connection policy.

   • PROJECT: The service project where you are creating Vertex AI resources.

   • VPC_PROJECT: The project where your client VPC is. For single VPC setup, this is the same as $PROJECT. For Shared VPC setup, this is the VPC host project.

   • NETWORK_NAME: The name of the network to deploy to, in the format projects//global/networks/.

   • REGION: The network region.

   • PSC_SUBNETS: The Private Service Connect subnets to use.

   gcloud network-connectivity service-connection-policies create <policy_name> \
   --project=<vpc_project> --network=<network_name> # in the format projects/<project_id>/global/networks/<network_name> \
   --service-class=gcp-vertexai --region=<region> --subnets=<psc subnets>
   

2. View your service connection policy.

   gcloud network-connectivity service-connection-policies list --project=<vpc_project> -–region=<region>
   

For more information about service connection policies, go to Configure service connection policies.

Deploy the index

PROJECT=PROJECT
VPC_PROJECTS=VPC_PROJECTS
REGION=REGION
curl -X POST -H "Authorization: Bearer $(gcloud auth print-access-token)"
-H "Content-Type: application/json; charset=utf-8" "https://REGIONAL_ENDPOINT.googleapis.com/v1/projects/PROJECT_NUMBER/locations/REGION/indexEndpoints/INDEX_ENDPOINT_ID:deployIndex"
-d '{
  "deployedIndex": {
    "id": "DEPLOYED_INDEX_ID",
    "index": "projects/PROJECT/locations/us-central1/indexes/INDEX_ID ",
    "displayName": "DISPLAY_NAME",
    "psc_automation_configs": [
      { "project_id": "PROJECT_1", "network": "NETWORK_NAME_1" },
      { "project_id": "PROJECT_2", "network": "NETWORK_NAME_2" },
      { "project_id": "PROJECT_N", "network": "NETWORK_NAME_N" }]
    }
}'

Delete service connection policy

If you need to delete the service connection policy, run the following command:

gcloud network-connectivity service-connection-policies delete --project=<vpc_project> –-region=<region> <policy_name>

For more information about service connection policies, go to Configure service connection policies.

Deploy with manual connection

Deploy the index and create a forwarding rule in your VPC project.

Deploy the index

Now that the index is ready, in this step, you deploy the index to the endpoint you created with Private Service Connect enabled.

gcloud ai index-endpoints deploy-index INDEX_ENDPOINT_ID \
    --deployed-index-id=DEPLOYED_INDEX_ID \
    --display-name=DEPLOYED_INDEX_ENDPOINT_NAME \
    --index=INDEX_ID \
    --region=LOCATION \
    --project=PROJECT_ID

gcloud ai index-endpoints deploy-index INDEX_ENDPOINT_ID `
    --deployed-index-id=DEPLOYED_INDEX_ID `
    --display-name=DEPLOYED_INDEX_ENDPOINT_NAME `
    --index=INDEX_ID `
    --region=LOCATION `
    --project=PROJECT_ID

gcloud ai index-endpoints deploy-index INDEX_ENDPOINT_ID ^
    --deployed-index-id=DEPLOYED_INDEX_ID ^
    --display-name=DEPLOYED_INDEX_ENDPOINT_NAME ^
    --index=INDEX_ID ^
    --region=LOCATION ^
    --project=PROJECT_ID

POST https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/indexEndpoints/INDEX_ENDPOINT_ID:deployIndex

{
 "deployedIndex": {
   "id": "DEPLOYED_INDEX_ID",
   "index": "projects/PROJECT_ID/locations/LOCATION/indexes/INDEX_ID",
   "displayName": "DEPLOYED_INDEX_ENDPOINT_NAME"
 }
}

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/indexEndpoints/INDEX_ENDPOINT_ID:deployIndex"

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/indexEndpoints/INDEX_ENDPOINT_ID:deployIndex" | Select-Object -Expand Content

{
 "name": "projects/PROJECT_NUMBER/locations/LOCATION/indexEndpoints/INDEX_ENDPOINT_ID/operations/OPERATION_ID",
 "metadata": {
   "@type": "type.googleapis.com/google.cloud.aiplatform.v1.DeployIndexOperationMetadata",
   "genericMetadata": {
     "createTime": "2022-10-19T17:53:16.502088Z",
     "updateTime": "2022-10-19T17:53:16.502088Z"
   },
   "deployedIndexId": "DEPLOYED_INDEX_ID"
 }
}

provider "google" {
  region = "us-central1"
}

resource "google_vertex_ai_index_endpoint_deployed_index" "default" {
  depends_on        = [google_vertex_ai_index_endpoint.default]
  index_endpoint    = google_vertex_ai_index_endpoint.default.id
  index             = google_vertex_ai_index.default.id
  deployed_index_id = "deployed_index_for_psc"
}

resource "google_vertex_ai_index_endpoint" "default" {
  display_name = "sample-endpoint"
  description  = "A sample index endpoint with Private Service Connect enabled"
  region       = "us-central1"
  private_service_connect_config {
    enable_private_service_connect = true
    project_allowlist = [
      data.google_project.project.project_id,
    ]
  }
}

data "google_project" "project" {}

# Cloud Storage bucket name must be unique
resource "random_id" "default" {
  byte_length = 8
}

# Create a Cloud Storage bucket
resource "google_storage_bucket" "bucket" {
  name                        = "vertex-ai-index-bucket-${random_id.default.hex}"
  location                    = "us-central1"
  uniform_bucket_level_access = true
}

# Create index content
resource "google_storage_bucket_object" "data" {
  name    = "contents/data.json"
  bucket  = google_storage_bucket.bucket.name
  content = <<EOF
{"id": "42", "embedding": [0.5, 1.0], "restricts": [{"namespace": "class", "allow": ["cat", "pet"]},{"namespace": "category", "allow": ["feline"]}]}
{"id": "43", "embedding": [0.6, 1.0], "restricts": [{"namespace": "class", "allow": ["dog", "pet"]},{"namespace": "category", "allow": ["canine"]}]}
EOF
}

resource "google_vertex_ai_index" "default" {
  region       = "us-central1"
  display_name = "sample-index-batch-update"
  description  = "A sample index for batch update"
  labels = {
    foo = "bar"
  }

  metadata {
    contents_delta_uri = "gs://${google_storage_bucket.bucket.name}/contents"
    config {
      dimensions                  = 2
      approximate_neighbors_count = 150
      distance_measure_type       = "DOT_PRODUCT_DISTANCE"
      algorithm_config {
        tree_ah_config {
          leaf_node_embedding_count    = 500
          leaf_nodes_to_search_percent = 7
        }
      }
    }
  }
  index_update_method = "BATCH_UPDATE"

  timeouts {
    create = "2h"
    update = "1h"
  }
}

def vector_search_deploy_index(
    project: str,
    location: str,
    index_name: str,
    index_endpoint_name: str,
    deployed_index_id: str,
) -> None:
    """Deploy a vector search index to a vector search index endpoint.

    Args:
        project (str): Required. Project ID
        location (str): Required. The region name
        index_name (str): Required. The index to update. A fully-qualified index
          resource name or a index ID.  Example:
          "projects/123/locations/us-central1/indexes/my_index_id" or
          "my_index_id".
        index_endpoint_name (str): Required. Index endpoint to deploy the index
          to.
        deployed_index_id (str): Required. The user specified ID of the
          DeployedIndex.
    """
    # Initialize the Vertex AI client
    aiplatform.init(project=project, location=location)

    # Create the index instance from an existing index
    index = aiplatform.MatchingEngineIndex(index_name=index_name)

    # Create the index endpoint instance from an existing endpoint.
    index_endpoint = aiplatform.MatchingEngineIndexEndpoint(
        index_endpoint_name=index_endpoint_name
    )

    # Deploy Index to Endpoint
    index_endpoint = index_endpoint.deploy_index(
        index=index, deployed_index_id=deployed_index_id
    )

    print(index_endpoint.deployed_indexes)

Create a forwarding rule in the VPC project

After index deployment is done, the index endpoint returns a service attachment URI instead of an IP address. You need to create a compute address, as well as a forwarding rule in the VPC project targeting the service attachment using the created compute address. To create a compute address, use the following example:

gcloud compute addresses create ${ADDRESS_NAME:?} \
    --region=${REGION:?} \
    --subnet=${SUBNET_NAME:?} \
    --project=${VPC_PROJECT:?}

To create a forwarding rule targeting the service attachment URI using the created compute address, use the following example:

SERVICE_ATTACHMENT_URI=`gcloud ai index-endpoints describe {INDEX_ENDPOINT_ID}
--format="value(deployedIndexes.privateEndpoints.serviceAttachment)"`

gcloud compute forwarding-rules create ${ENDPOINT_NAME:?} \
    --network=${NETWORK_NAME:?} \
    --address=${ADDRESS_NAME:?} \
    --target-service-attachment=${SERVICE_ATTACHMENT_URI:?} \
    --project=${VPC_PROJECT:?} \
    --region=${REGION:?}

(Optional) Create DNS record for the IP address

If you want to connect and load without memorizing the actual IP address, you can create a DNS record. This step is optional.

DNS_NAME_SUFFIX=matchingengine.vertexai.goog. # Don't forget the "." in the end.
DNS_NAME=${INDEX_ENDPOINT_ID:?}.${REGION:?}.${DNS_NAME_SUFFIX:?}

gcloud dns managed-zones create ${DNS_ZONE_NAME:?} \
    --dns-name=${DNS_NAME_SUFFIX:?} \
    --visibility=private \
    --project=${VPC_PROJECT:?} \
    --region=${REGION:?}

gcloud dns record-sets create ${DNS_NAME:?} \
    --rrdatas=${IP_ADDRESS:?} \
    --type=A --ttl=60 \
    --zone=${DNS_ZONE_NAME:?} \
    --project=${VPC_PROJECT:?} \
    --region=${REGION:?}

Send queries to the index endpoint

Now that you've created an endpoint with Private Service Connect and created the index, you can begin running queries.

To query your index, see Query indexes to get nearest neighbors.

What's next

• Learn how to Update and rebuild your index
• Learn how to Monitor the index endpoint