Skip to main content

Connect to Google Vertex AI

Connect to Google Vertex AI and use Gemini models for AI-powered apps and automations in Retool.

Google Vertex AI is Google Cloud's managed AI platform that provides access to the Gemini family of multimodal models within your own GCP infrastructure. Unlike the Gemini API accessed via Google AI Studio, Vertex AI runs inside your GCP project, giving you enterprise-grade security, data residency, and IAM-based access management. You can create a Google Vertex AI resource to build AI-powered apps and workflows that leverage Gemini's capabilities for text generation, data analysis, and image understanding.

What you can do with Google Vertex AI in Retool​

After you create a Google Vertex AI resource in Retool, you can:

  • Generate, summarize, and classify text using Gemini's language models.
  • Build conversational, multi-turn chat experiences with context-aware responses.
  • Analyze images to generate text or caption images.

Supported models​

Google Vertex AI resources require you to bring your own key for authentication. Retool-managed keys are not supported.

The following models are available with your own API key:

ModelResource queriesAgent support
Gemini 2.5 Flash
Gemini 2.5 Pro
Gemini 3 Flash (Preview)
Gemini 3.1 Flash-Lite (Preview)
Gemini 3.1 Pro (Preview)
Gemini 3.5 Flash

Specify the model ID in your queries. Refer to Google's Vertex AI model documentation for the complete list of available model versions.

Before you begin​

Before creating a Google Vertex AI resource, you need:

  • Retool permissions: Ability to create and manage resources in your Retool organization.
  • Google Cloud project: An active GCP project with the Vertex AI API enabled.
  • Service account: A Google Cloud service account with the Vertex AI User role (roles/aiplatform.user) or equivalent permissions.
  • Service account JSON key: A downloaded JSON key file for the service account.

Create a Google Vertex AI resource​

Create a resource to connect Retool to Google Vertex AI and configure authentication. Once connected, you can then select it when writing queries to make use of its available models.

Follow these steps to create a Google Vertex AI resource in Retool.

Create a new resource​

Navigate to Resources in the main navigation and click Create new > Resource and search for Google Vertex AI. Then, click the Google Vertex AI tile to create a new resource.

Configure connection settings​

Specify a name for the resource that identifies it within your organization. Include a description that can provide more context to users and Assist about how to use the resource.

Configure authentication​

Bring your own key only

This resource requires a Google Cloud service account. Retool-managed keys are not available for Google Vertex AI.

Google Vertex AI resources authenticate using a Google Cloud service account JSON key.

Google Vertex AI configuration
Google Vertex AI configuration

Project ID​

Your Google Cloud project identifier (e.g., my-gcp-project). This is the project where the Vertex AI API is enabled.

Location​

The GCP region where your Vertex AI requests are processed. Select a region close to your users or data for lower latency. The default is us-central1. Refer to Google's locations documentation for the full list of supported regions.

Service account key​

Paste the full contents of your Google Cloud service account JSON key file. Retool uses this key to authenticate requests to the Vertex AI API.

To generate a service account key:

  1. Open the Service accounts page in Google Cloud Console.
  2. Select your project and find or create a service account with the Vertex AI User role.
  3. Click the service account, open the Keys tab, and click Add key > Create new key.
  4. Select JSON and download the key file.
  5. Open the file and paste its entire contents into the Service account key field.
Create a dedicated service account for Retool

Create a service account specifically for Retool with only the permissions it requires. All credentials are encrypted and stored securely.

Save the resource​

Click Create resource to save your Google Vertex AI resource. The resource is now available for use in apps and workflows.

Note

AI resources do not have a Test connection button. To verify your configuration, create a new app and ask the agent to test the resource for you.

Interact with Google Vertex AI data​

Once you've created a Google Vertex AI resource, you can interact with Gemini models using Retool.

Caution

Google Vertex AI isn't yet supported in the new app builder. Use a classic app, workflow, or agent instead.

For classic apps, workflows, and agents, refer to the AI queries documentation to learn how to write a resource query.

Best practices​

Follow these best practices to keep your Google Vertex AI usage secure and cost-efficient.

Security​

  • Use a dedicated service account for Retool: Create a service account specifically for Retool with only the Vertex AI User role, so you can revoke or rotate access without disrupting other integrations.
  • Rotate service account keys regularly: Roll your service account JSON key periodically from the Service accounts page, and immediately if a key is ever exposed.
  • Use resource environments: Organizations on an Enterprise plan can configure multiple resource environments to maintain separate service accounts for production, staging, and development.

Manage token usage and cost​

  • Choose the smallest capable model: Use a faster, cheaper model (e.g. Gemini Flash) for simple tasks like classification or extraction, and reserve Pro models for tasks that need deeper reasoning.
  • Set output token limits: Configure a maximum output token limit appropriate to the task to avoid paying for unnecessarily long responses.
  • Trim conversation history: For chat use cases, truncate or summarize older messages instead of sending the full history on every request. Vertex AI bills for all input tokens, including message history.
  • Monitor usage: Track token consumption and spend in the Google Cloud Console to catch unexpected usage spikes early.