I am attempting to utilize the Go Vertex AI SDK in order to use some of the functionalities on Vertex AI. Specifically, I want to query from a deployed matching engine index via Go.

Here is my implementation:

package main

import (
    "fmt"

    "context"

    "regexp"

    aiplatformpb "cloud.google.com/go/aiplatform/apiv1/aiplatformpb"

    aiplatform "cloud.google.com/go/aiplatform/apiv1"

    "google.golang.org/api/option"
    "google.golang.org/protobuf/types/known/structpb"
)

func embedTexts() {
    ...
}

func main() {
    ctx := context.Background()

    query_string_arr := []string{"Some text to be used as a query for nearest neighbors"}
    var project_id = "<project id>"
    var location = "us-central1"

    // Get text embeddings for query
    embeddings, err := embedTexts(project_id, location, query_string_arr)
    if err != nil {
        fmt.Println(err.Error())
    } else {
        fmt.Println("The length of the embedding vector is:")
        fmt.Println(len(embeddings[0]))
    }

    // ensure that we are hitting the correct API endpoint for vector search client
    apiEndpoint := fmt.Sprintf("%s-aiplatform.googleapis.com:443", location)

    // initialize matching engine client
    c, err := aiplatform.NewMatchClient(ctx, option.WithEndpoint(apiEndpoint))
    if err != nil {
        fmt.Println(err.Error())
    } else {
        fmt.Println("Vector Search Client successfully initialized!")
    }

    // // initialize structs to create request body

    //// 1. initialize the index datapoint restriction struct
    restricts_params := &aiplatformpb.IndexDatapoint_Restriction{
        Namespace: "<some namespace>",
        AllowList: []string{"<some value to allow for namespace>"},
    }

    //// 2. initialize the index datapoint struct
    query_datapoint := &aiplatformpb.IndexDatapoint{
        DatapointId:   "<Placeholder>",
        FeatureVector: embeddings[0],
        Restricts:     []*aiplatformpb.IndexDatapoint_Restriction{restricts_params},
    }

    //// 3. initialize the query struct
    find_nbrs_query := &aiplatformpb.FindNeighborsRequest_Query{
        Datapoint:     query_datapoint,
        NeighborCount: 50,
    }

    //// 4. initialize the nearest neighbors request struct
    req := &aiplatformpb.FindNeighborsRequest{
        IndexEndpoint:       "projects/<project number>/locations/us-central1/indexEndpoints/<vector search index endpoint id>",
        DeployedIndexId:     "<vector search index id>",
        Queries:             []*aiplatformpb.FindNeighborsRequest_Query{find_nbrs_query},
        ReturnFullDatapoint: true,
    }

    // send request to vector search
    resp, err := c.FindNeighbors(ctx, req)
    if err != nil {
        fmt.Println("Failing at vector search request time.")
        fmt.Println(err.Error())
    } else {
        fmt.Println("Vector Search Request successfully sent!")
    }

    defer c.Close()

    for _, neighbor := range resp.GetNearestNeighbors() {
        fmt.Printf("Neighbor ID:%sn", neighbor.Id)
    }
}

However, I’ve come across a particular error that I cannot seem to resolve.

rpc error: code = Unimplemented desc = Operation is not implemented, or supported, or enabled.

This is failing at the snippet of code where I am calling c.FindNeighbors. This seems to indicate that:

• The endpoint I am using, us-central1-aiplatform.googleapis.com:443, is not the correct endpoint to be using for calling the vector search index endpoint.
• The method I am using is incorrect somehow (although the compiler did not throw any errors at me).

In either of these cases, what methods should I be using? Or what endpoint should I be hitting? This is a rather open-ended question, but I am not sure how to proceed (I am also rather new to Go so please give some pointers as well if I seem to misunderstand something in my code snippet!)

Side note: I have also checked that I have full access to my GCP environment. In fact, my POST requests from CLI work as intended. I am just struggling with making the gRPC call via Go.