'Use google cloud aiplatform with golang
I have a vertex AI modele deployed on an endpoint and want to do some prediction from my app in golang.
To do this I create code inspired by this exemple : https://cloud.google.com/go/docs/reference/cloud.google.com/go/aiplatform/latest/apiv1?hl=en
const file = "MY_BASE64_IMAGE"
func main() {
ctx := context.Background()
c, err := aiplatform.NewPredictionClient(cox)
if err != nil {
log.Printf("QueryVertex NewPredictionClient - Err:%s", err)
}
defer c.Close()
parameters, err := structpb.NewValue(map[string]interface{}{
"confidenceThreshold": 0.2,
"maxPredictions": 5,
})
if err != nil {
log.Printf("QueryVertex structpb.NewValue parameters - Err:%s", err)
}
instance, err := structpb.NewValue(map[string]interface{}{
"content": file,
})
if err != nil {
log.Printf("QueryVertex structpb.NewValue instance - Err:%s", err)
}
reqP := &aiplatformpb.PredictRequest{
Endpoint: "projects/PROJECT_ID/locations/LOCATION_ID/endpoints/ENDPOINT_ID",
Instances: []*structpb.Value{instance},
Parameters: parameters,
}
resp, err := c.Predict(cox, reqP)
if err != nil {
log.Printf("QueryVertex Predict - Err:%s", err)
}
log.Printf("QueryVertex Res:%+v", resp)
}
I put the path to my service account json file on GOOGLE_APPLICATION_CREDENTIALS environment variable. But when I'm run my test app I obtain this error message :
QueryVertex Predict - Err:rpc error: code = Unimplemented desc = unexpected HTTP status code received from server: 404 (Not Found); transport: received unexpected content-type "text/html; charset=UTF-8"
QueryVertex Res:<nil>
Solution 1:[1]
As @DazWilkin suggested, configure the client option to specify the specific regional endpoint with a port 443:
option.WithEndpoint("<region>-aiplatform.googleapis.com:443")
Try like below:
func main() {
ctx := context.Background()
c, err := aiplatform.NewPredictionClient(
ctx,
option.WithEndpoint("<region>-aiplatform.googleapis.com:443"),
)
if err != nil {
log.Printf("QueryVertex NewPredictionClient - Err:%s", err)
}
defer c.Close()
.
.
Solution 2:[2]
I'm unfamiliar with Google's (Vertex?) AI Platform and unable to test this hypothesis but it appears that the API uses location-specific endpoints.
Can you try configuring the client's ClientOption
to specify the specific regional endpoint, i.e.:
url := fmt.Sprintf("https://%s-aiplatform.googleapis.com", location)
opts := []option.ClientOption{
option.WithEndpoint(url),
}
And:
package main
import (
"context"
"fmt"
"log"
"os"
aiplatform "cloud.google.com/go/aiplatform/apiv1"
"google.golang.org/api/option"
aiplatformpb "google.golang.org/genproto/googleapis/cloud/aiplatform/v1"
"google.golang.org/protobuf/types/known/structpb"
)
const file = "MY_BASE64_IMAGE"
func main() {
// Values from the environment
project := os.Getenv("PROJECT")
location := os.Getenv("LOCATION")
endpoint := os.Getenv("ENDPOINT")
ctx := context.Background()
// Configure the client with a region-specific endpoint
url := fmt.Sprintf("https://%s-aiplatform.googleapis.com", location)
opts := []option.ClientOption{
option.WithEndpoint(url),
}
c, err := aiplatform.NewPredictionClient(ctx, opts...)
if err != nil {
log.Fatal(err)
}
defer c.Close()
parameters, err := structpb.NewValue(map[string]interface{}{
"confidenceThreshold": 0.2,
"maxPredictions": 5,
})
if err != nil {
log.Fatal(err)
}
instance, err := structpb.NewValue(map[string]interface{}{
"content": file,
})
if err != nil {
log.Printf("QueryVertex structpb.NewValue instance - Err:%s", err)
}
rqst := &aiplatformpb.PredictRequest{
Endpoint: fmt.Sprintf("projects/%s/locations/%s/endpoints/%s",
project,
location,
endpoint,
),
Instances: []*structpb.Value{
instance,
},
Parameters: parameters,
}
resp, err := c.Predict(ctx, rqst)
if err != nil {
log.Fatal(err)
}
log.Printf("QueryVertex Res:%+v", resp)
}
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
Solution | Source |
---|---|
Solution 1 | |
Solution 2 | DazWilkin |