Connect your app to Gemini Enterprise

1. 1

   Make sure you have a Google Cloud project with a Gemini Enterprise or Vertex AI Search engine, with data sources already connected to it.

   Your Lovable app will connect to this existing engine to access your company's data.

2. 2

   Ensure you have a Google account that can access the engine and its connected content, and you know the engine's GCP project ID, Location (global, us, or eu), and Engine ID.

   These details are essential for Lovable to correctly identify and connect to your Gemini Enterprise engine.

3. 3

   Go to the 'Connectors' section in Lovable and select 'Gemini Enterprise'.

   This is where you initiate the connection process for Gemini Enterprise.

4. 4

   Click 'Add connection' and fill in the 'Display name', 'GCP project ID', 'Location', and 'Engine ID'.

   This configures the basic details for your connection within Lovable, linking it to your specific Gemini Enterprise engine.

5. 5

   Review the default 'Cloud Platform (full access)' permission scope and leave it selected.

   This permission is required for Lovable to make requests to Gemini Enterprise on your behalf.

6. 6

   Optionally, choose who can use this connection by sharing it with specific workspace members or the entire workspace.

   By default, the connection is private to you; sharing it allows others in your team to use it in their projects.

7. 7

   Click 'Connect', then sign in with your Google account that has access to the engine and authorize the connection.

   This securely links your Google account to Lovable, allowing Lovable to access your Gemini Enterprise engine.

Connect my app to Gemini Enterprise. I want to build an internal knowledge assistant that answers employee questions and always shows citations. The app should ask Gemini Enterprise for grounded answers and display source-backed responses with follow-up questions.

Gemini Enterprise

Google Cloud's platform for searching and getting answers from your company's internal data, often used for internal knowledge bases or support tools.

Vertex AI Search

Another name for Google Cloud's enterprise search platform, which is the same technology as Gemini Enterprise.

Grounded answers

Answers provided by an AI that are directly supported by specific pieces of information from your connected data, often with links to the original sources (citations).

Citations

References or links to the original documents or data sources that an AI used to generate an answer, proving its accuracy and allowing you to check the source.

Data stores

Collections of your company's information, like documents, databases, or internal websites, that Gemini Enterprise can search through.

Engine

In Gemini Enterprise, this refers to the specific setup that processes your data and provides search and answer capabilities.

GCP project ID

A unique identifier for your project within Google Cloud, where your Gemini Enterprise engine is located.

Location

The geographical region where your Gemini Enterprise engine is hosted, which can affect data residency and performance.

Engine ID

A specific identifier for your Gemini Enterprise application within your Google Cloud project.

Scopes

Permissions that you grant to an application (like Lovable) to access specific parts of your Google account or services.

App + chat connector

A type of connection in Lovable that can be used both when you're building your app in the project chat and in your finished, published app.

• Using a Google account that doesn't have access to the Gemini Enterprise engine or its data sources will cause requests to fail.
• Entering incorrect GCP project ID, Location, or Engine ID will prevent the connection from working.
• Forgetting to authorize Lovable with your Google account after setting up the connection will prevent it from functioning.
• Not sharing the connection with other workspace members if they need to use it in their projects.
• Expecting the connector to write data back to your source systems (e.g., send emails or create Jira issues), as it only supports searching and grounded answers.
• Trying to search website data stores, as Gemini Enterprise search and assistant apps do not support them through this connector.
• Assuming that each end-user of your app can log in with their own Google account; the connection uses a single shared Google account.
• Not checking Google Cloud quotas or rate limits if requests are failing with a 429 error, as billing is handled by Google Cloud.