Get Started

Vald is a highly scalable distributed fast approximate nearest neighbor dense vector search engine.
Vald is designed and implemented based on Cloud-Native architecture.

This tutorial shows how to deploy and run the Vald components on your Kubernetes cluster. And, Fashion-MNIST is used as an example of a dataset.


The below image shows the architecture image about the deployment result of Get Started.
The 4 kinds of components, Vald LB Gateway, Vald Discoverer, Vald Agent, and Vald Index Manager will be deployed to the Kubernetes. For more information about Vald’s architecture, please refer to Architecture.

The 5 steps to Get Started with Vald:

  1. Check and Satisfy the Requirements
  2. Prepare Kubernetes Cluster
  3. Deploy Vald on Kubernetes Cluster
  4. Run Example Code
  5. Cleanup


  • Kubernetes: v1.19 ~
  • Go: v1.15 ~
  • Helm: v3 ~
  • libhdf5 (only required for get started)

Helm is used to deploying Vald on your Kubernetes and HDF5 is used to decode the sample data file to run the example.
If Helm or HDF5 is not installed, please install Helm and HDF5.

Installation command for Helm
curl | bash
Installation command for HDF5
# yum
yum install -y hdf5-devel

# apt
apt-get install libhdf5-serial-dev

# homebrew
brew install hdf5

Prepare the Kubernetes Cluster

This tutorial requires the Kubernetes cluster.
Vald runs on public Cloud Kubernetes Services such as GKE, EKS. In the sense of trying to Get Started, k3d or kind are easy Kubernetes tools to use.

This tutorial uses Kubernetes Ingress and Kubernetes Metrics Server for running Vald.
Please make sure these functions are available.

The configuration of Kubernetes Ingress is depended on your Kubernetes cluster’s provider. Please refer to on yourself.

In the following example, we create the Kubernetes cluster using k3d, that the internal port 80 (where the traefik ingress controller is listening on) is exposed on the host system.

k3d cluster create -p 8081:80@loadbalancer

The way to deploy Kubernetes Metrics Service is here:

kubectl apply -f && \
kubectl wait -n kube-system --for=condition=ready pod -l k8s-app=metrics-server --timeout=600s

Deploy Vald on Kubernetes Cluster

This chapter shows how to deploy Vald using Helm and run it on your Kubernetes cluster.
In this tutorial, you will deploy the basic configuration of Vald that is consisted of vald-agent-ngt, vald-lb-gateway, vald-discoverer, and vald-manager-index.

For vald-discoverer to work correctly, if you deploy multiple Vald clusters in the same Kubernetes cluster, please deploy one Vald cluster in one namespace.
  1. Clone the repository

    git clone && \
    cd vald
  2. Confirm which cluster to deploy

    kubectl cluster-info
  3. Edit Configurations

    Set the parameters for connecting to the vald-lb-gateway through Kubernetes ingress from the external network. Please set these parameters.

    vim example/helm/values.yaml
    ## vald-lb-gateway settings
          enabled: true
          # TODO: Set your ingress host.
          host: localhost
          # TODO: Set annotations which you have to set for your Ingress resource.

    If you decided to use port-forward instead of ingress, please set to false.

  4. Deploy Vald using Helm

    Add vald repo into the helm repo.

    helm repo add vald

    Deploy vald on your Kubernetes cluster.

    helm install vald vald/vald --values example/helm/values.yaml
  5. Verify

    When finish deploying Vald, you can check the Vald’s pods status following command.

    kubectl get pods
    Example output
    If the deployment is successful, all Vald components should be running.
    NAME                                       READY   STATUS      RESTARTS   AGE
    vald-agent-ngt-0                           1/1     Running     0          7m12s
    vald-agent-ngt-1                           1/1     Running     0          7m12s
    vald-agent-ngt-2                           1/1     Running     0          7m12s
    vald-agent-ngt-3                           1/1     Running     0          7m12s
    vald-agent-ngt-4                           1/1     Running     0          7m12s
    vald-discoverer-7f9f697dbb-q44qh           1/1     Running     0          7m11s
    vald-lb-gateway-6b7b9f6948-4z5md           1/1     Running     0          7m12s
    vald-lb-gateway-6b7b9f6948-68g94           1/1     Running     0          6m56s
    vald-lb-gateway-6b7b9f6948-cvspq           1/1     Running     0          6m56s
    vald-manager-index-74c7b5ddd6-jrnlw        1/1     Running     0          7m12s
    kubectl get ingress
    Example output
    NAME                      CLASS     HOSTS       ADDRESS        PORTS   AGE
    vald-lb-gateway-ingress   traefik   localhost   80      7m43s
    kubectl get svc
    Example output
    NAME                 TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)             AGE
    kubernetes           ClusterIP    <none>        443/TCP             9m29s
    vald-agent-ngt       ClusterIP   None         <none>        8081/TCP,3001/TCP   8m48s
    vald-discoverer      ClusterIP   None         <none>        8081/TCP,3001/TCP   8m48s
    vald-manager-index   ClusterIP   None         <none>        8081/TCP,3001/TCP   8m48s
    vald-lb-gateway      ClusterIP   None         <none>        8081/TCP,3001/TCP   8m48s

Run Example Code

In this chapter, you will execute insert, search, and delete vectors to your Vald cluster using the example code.
The Fashion-MNIST is used as a dataset for indexing and search query.

The example code is implemented in Go and using vald-client-go, one of the official Vald client libraries, for requesting to Vald cluster. Vald provides multiple language client libraries such as Go, Java, Node.js, Python, etc. If you are interested, please refer to SDKs.

  1. Port Forward(option)

    If you do not use Kubernetes Ingress, port-forward is required to make requests from your local environment.

    kubectl port-forward deployment/vald-lb-gateway 8081:8081
  2. Download dataset

    Download Fashion-MNIST that is used as a dataset for indexing and search query.

    Move to the working directory

    cd example/client

    Download Fashion-MNIST testing dataset

  3. Run Example

    We use example/client/main.go to run the example.
    This example will insert and index 400 vectors into the Vald from the Fashion-MNIST dataset via gRPC. And then after waiting for indexing, it will request for searching the nearest vector 10 times. You will get the 10 nearest neighbor vectors for each search query.
    Run example codes by executing the below command.

    go run main.go
    The detailed explanation of example code is here
    This will execute 6 steps.
    1. init

      • Import packages

        example code
        package main
        import (
      • Set variables

        • The constant number of training datasets and test datasets.

          example code
          const (
              insertCount = 400
              testCount = 20
        • The variables for configuration.

          example code
          const (
              datasetPath         string
              grpcServerAddr      string
              indexingWaitSeconds uint
      • Recognition parameters.

        example code
        func init() {
            flag.StringVar(&datasetPath, "path", "fashion-mnist-784-euclidean.hdf5", "set dataset path")
            flag.StringVar(&grpcServerAddr, "addr", "", "set gRPC server address")
            flag.UintVar(&indexingWaitSeconds, "wait", 60, "set indexing wait seconds")
    2. load

      • Loading from Fashion-MNIST dataset and set id for each vector that is loaded. This step will return the training dataset, test dataset, and ids list of ids when loading is completed with success.

        example code
        ids, train, test, err := load(datasetPath)
        if err != nil {
    3. Create the gRPC connection and Vald client with gRPC connection.

      example code
      ctx := context.Background()
      conn, err := grpc.DialContext(ctx, grpcServerAddr, grpc.WithTransportCredentials(insecure.NewCredentials()))
      if err != nil {
      client := vald.NewValdClient(conn)
    4. Insert and Index

      • Insert and Indexing 400 training datasets to the Vald agent.

        example code
        for i := range ids [:insertCount] {
            _, err := client.Insert(ctx, &payload.Insert_Request{
                Vector: &payload.Object_Vector{
                    Id:     ids[i],
                    Vector: train[i],
                Config: &payload.Insert_Config{
                    SkipStrictExistCheck: true,
            if err != nil {
            if i%10 == 0 {
                glg.Infof("Inserted %d", i)
      • Wait until indexing finish.

        example code
        wt := time.Duration(indexingWaitSeconds) * time.Second
        glg.Infof("Wait %s for indexing to finish", wt)
    5. Search

      • Search 10 neighbor vectors for each 20 test datasets and return a list of the neighbor vectors.

      • When getting approximate vectors, the Vald client sends search config and vector to the server via gRPC.

        example code
        glg.Infof("Start search %d times", testCount)
        for i, vec := range test[:testCount] {
            res, err := client.Search(ctx, &payload.Search_Request){
                Vector: vec,
                Config: &payload.Search_Config{
                    Num: 10,
                    Radius: -1,
                    Epsilon: 0.1,
                    Timeout: 100000000,
            if err != nil {
            b, _ := json.MarshalIndent(res.GetResults(), "", " ")
            glg.Infof("%d - Results : %s\n\n", i+1, string(b))
            time.Sleep(1 * time.Second)
    6. Remove

      • Remove 200 indexed training datasets from the Vald agent.

        example code
        for i := range ids [:removeCount] {
            _, err := client.Remove(ctx, &payload.Remove_Request{
                Id: &payload.Object_ID{
                    Id: ids[i],
            if err != nil {
            if i%10 == 0 {
                glg.Infof("Removed %d", i)
    7. Flush

      • Remove all remaining training datasets from the Vald agent.

        example code
        _, err := client.Flush(ctx, &payload.Flush_Request{})
        if err != nil {


In the last, you can remove the deployed Vald Cluster by executing the below command.

helm uninstall vald

Next Steps

Congratulation! You completely entered the Vald World!

If you want, you can try other tutorials such as:

For more information, we recommend you to check:

See also