This guide will walk you through integrating Looker with Generative AI Studio (Vertex AI) via Cloud Functions using the Looker Action API. Users can use the Looker Explore to examine data, then send the Looker query to a Cloud Function, specifying the model prompt and parameters on the form submissions.

The cloud function included in this demo is used to communicate from Looker to Vertex AI via the Action API:

Installation:

Installation can be done either via manual steps or Terraform.

Manual Installation

Before following the steps below, make sure you have enabled the Secret Manager API, Cloud Build API, Cloud Functions API, Cloud Run Admin API, Artifact Registry API, and the Vertex AI API. It will take a few minutes after enabling this APIs for it to propagate through the systems.

Also make sure you have a Sendgrid or Mailgun account and API key to use for sending emails. You can create a Sendgrid developer account from the GCP marketplace.

Use Cloud Shell or the gcloud CLI for the following steps.

The variables to modify are:

• PROJECT - ID you want to deploy the Cloud Functions to
• EMAIL_SENDER - Email address of the sender
• MODEL_VARIANT - Default Vertex AI model to use if not selected in form (default: gemini-2.5-flash)
• CALL_LIMIT - Request quota limit (default: 50)
• ONE_MINUTE - Time window for quota in seconds (default: 60)
• OUTPUT_TOKEN_LIMIT - Max output tokens (default: 8192)

1. Set the variables below:

   ACTION_LABEL="Vertex AI"
   ACTION_NAME="vertex-ai"
   REGION="us-central1"
   PROJECT="my-project-id"
   EMAIL_SENDER="my-sender-email-address@foo.com"
   MODEL_VARIANT="gemini-2.5-flash"
   CALL_LIMIT="50"
   ONE_MINUTE="60"
   OUTPUT_TOKEN_LIMIT="8192"
   
   

2. Clone this repo

   git clone https://github.com/looker-open-source/vertex-ai-actions
   cd vertex-ai-actions/
   

3. Create a .env.yaml with variables:

   printf "ACTION_LABEL: ${ACTION_LABEL}\nACTION_NAME: ${ACTION_NAME}\nREGION: ${REGION}\nPROJECT: ${PROJECT}\nEMAIL_SENDER: ${EMAIL_SENDER}\nGEN2_ROUTER: true\nMODEL_VARIANT: ${MODEL_VARIANT}\nCALL_LIMIT: ${CALL_LIMIT}\nONE_MINUTE: ${ONE_MINUTE}\nOUTPUT_TOKEN_LIMIT: ${OUTPUT_TOKEN_LIMIT}" > .env.yaml
   

4. Generate the LOOKER_AUTH_TOKEN secret. The auth token secret can be any randomly generated string. You can generate such a string with the openssl command:

   LOOKER_AUTH_TOKEN="`openssl rand -hex 64`"
   

5. Add the Auth Token and Sendgrid API key as Secrets, then create a Service Account to run the Cloud Functions and give it access to the Secrets:

   SENDGRID_API_KEY="copy your sendgrid api key here"
   
   printf ${SENDGRID_API_KEY} | gcloud secrets create SENDGRID_API_KEY --data-file=- --replication-policy=user-managed --locations=${REGION} --project=${PROJECT}
   
   printf ${LOOKER_AUTH_TOKEN} | gcloud secrets create LOOKER_AUTH_TOKEN --data-file=- --replication-policy=user-managed --locations=${REGION} --project=${PROJECT}
   
   gcloud iam service-accounts create vertex-ai-actions-cloud-function --display-name="Vertex AI Actions Cloud Functions" --project=${PROJECT}
   
   SERVICE_ACCOUNT_EMAIL=vertex-ai-actions-cloud-function@${PROJECT}.iam.gserviceaccount.com
   
   eval gcloud projects add-iam-policy-binding ${PROJECT} --member=serviceAccount:${SERVICE_ACCOUNT_EMAIL} --role='roles/cloudfunctions.invoker'
   
   eval gcloud projects add-iam-policy-binding ${PROJECT} --member=serviceAccount:${SERVICE_ACCOUNT_EMAIL} --role='roles/aiplatform.user'
   
   eval gcloud projects add-iam-policy-binding ${PROJECT} --member=serviceAccount:${SERVICE_ACCOUNT_EMAIL} --role='roles/secretmanager.secretAccessor'
   
   eval gcloud secrets add-iam-policy-binding SENDGRID_API_KEY --member=serviceAccount:${SERVICE_ACCOUNT_EMAIL} --role='roles/secretmanager.secretAccessor' --project=${PROJECT}
   
   eval gcloud secrets add-iam-policy-binding LOOKER_AUTH_TOKEN --member=serviceAccount:${SERVICE_ACCOUNT_EMAIL} --role='roles/secretmanager.secretAccessor' --project=${PROJECT}
   

6. Deploy the cloud function for the action hub (this may take a few minutes):

   gcloud functions deploy ${ACTION_NAME} --entry-point action_handler --env-vars-file .env.yaml --trigger-http --runtime=python311 --allow-unauthenticated --gen2 --memory=8192MB --timeout=540s --region=${REGION} --project=${PROJECT} --service-account ${SERVICE_ACCOUNT_EMAIL} --set-secrets 'LOOKER_AUTH_TOKEN=LOOKER_AUTH_TOKEN:latest,SENDGRID_API_KEY=SENDGRID_API_KEY:latest'
   

7. Copy the Action Hub URL (action_list endpoint) and the LOOKER_AUTH_TOKEN to input into Looker:

   echo Action Hub URL: $(gcloud functions describe ${ACTION_NAME} --region=${REGION} --format='value(serviceConfig.uri)')
   echo LOOKER_AUTH_TOKEN: $LOOKER_AUTH_TOKEN
   

8. In Looker, go to the Admin > Actions page and click Add Action Hub

   • Enter the Action Hub URL
   • click Configure Authorization and enter the LOOKER_AUTH_TOKEN value for the Authorization Token and click Enable
   • Toggle the Enabled button and click Save

Troubleshooting:

If the action build fails, you will receive an email notification. Go to the Admin > Scheduler History page to view the error message returned from the Action or use scheduled_plan System Activity Explore:

• Explore query to see details on action executions:

  https://${YOUR_LOOKER_DOMAIN}/explore/system__activity/scheduled_plan?fields=scheduled_job.id,scheduled_job.created_time,scheduled_plan_destination.action_type,scheduled_plan_destination.format,scheduled_job.status,scheduled_plan.run_once,scheduled_plan_destination.parameters,scheduled_job.status_detail&f[scheduled_plan_destination.action_type]=vertex-ai&sorts=scheduled_job.created_time+desc&limit=500

Deployment via Terraform

Alternatively, you can manage the deployment using Terraform. A terraform/ directory is provided with the configuration.

1. Navigate to the terraform directory:

   cd terraform

2. Initialize Terraform:

   terraform init

3. Create a terraform.tfvars file with your configuration:

   project_id        = "your-project-id"
   region            = "us-central1"
   email_sender      = "noreply@yourdomain.com"
   looker_auth_token = "your-looker-token"
   mailgun_api_token = "your-mailgun-token" # Or sendgrid_api_key
   mailgun_domain    = "your-mailgun-domain"

4. Plan and apply:

   terraform plan
   terraform apply

5. See the Manual Installation section for instructions on how to configure Looker's action hub with the output Cloud Run url.