- Search connected data stores using natural language queries
- Return grounded answers with citations from your indexed content
- Support follow-up questions with multi-turn conversations
- Build search-first interfaces with ranked results, filters, and facets
- Inspect which data stores are connected to your Gemini Enterprise setup
Why and when to use it
Use Gemini Enterprise when your organization already uses Gemini Enterprise or Vertex AI Search as the layer that unifies internal data sources. This connector is a good fit when you want Lovable to build on top of that existing search layer instead of calling each source system directly. Gemini Enterprise works best for search and answer experiences such as internal knowledge search, support assistants, and research tools. It is not the right choice when you need direct write access to the source systems themselves, such as sending email, creating Jira issues, or modifying files in external tools.Common use cases and example apps
| Example app | Example prompt | Description |
|---|---|---|
| Internal knowledge assistant | Build a chat interface that answers employee questions using our Gemini Enterprise engine and always shows citations. | Create a grounded Q&A experience on top of your indexed company content. The app asks Gemini Enterprise for grounded answers and displays source-backed responses with follow-up questions. |
| Enterprise search portal | Build a search page for our Gemini Enterprise engine with filters, result cards, and a document preview panel. | Turn Gemini Enterprise into a structured search UI. The app returns ranked search results from connected data stores and lets users browse sources directly. |
| Support knowledge tool | Build a support workspace where agents can search our help content and get a cited answer before replying to customers. | Help teams find trusted answers faster. The app combines search and grounded answer generation so agents can verify what the source documents actually say. |
| Cross-source research assistant | Build a research assistant that searches across the data stores connected to our Gemini Enterprise app and summarizes the findings with citations. | Search across blended enterprise sources from one interface. The app uses your existing Gemini Enterprise setup to retrieve and summarize results across multiple connected data stores. |
| Source explorer | Build an admin page that shows which data stores are connected to our Gemini Enterprise engine and lets users search within specific sources. | Make the search configuration visible inside your app. The app can list connected data stores, show their types, and target search experiences to selected sources. |
Requirements and billing
Workspace admins and owners can create Gemini Enterprise connections. Before connecting Gemini Enterprise, make sure you have:- A Google Cloud project with a Gemini Enterprise or Vertex AI Search engine
- Data stores already connected to that engine in Google Cloud
- A Google account that can access the Gemini Enterprise engine and the connected content you want the app to use
- Your GCP project ID
- Your Engine ID
- Your Location (
global,us, oreu) - Lovable workspace admin or owner role
Gemini Enterprise usage, quotas, and billing are managed by Google Cloud. Requests made through this connector count toward your Google Cloud project, not Lovable.
How Gemini Enterprise connections work
- Each connection stores the GCP project ID, Location, and Engine ID for one Gemini Enterprise engine.
- You authorize the connection with a Google account, and Lovable stores that access securely in the connector gateway with automatic refresh.
- Your app can use the connection across linked projects in the workspace.
- You can create multiple Gemini Enterprise connections if you need different engines, environments, or Google accounts.
- For Google Workspace and supported third-party sources connected through Gemini Enterprise, retrieval respects the access of the Google account used for the connection.
- Search results and grounded answers depend on both your Gemini Enterprise engine configuration and the access of the Google account used for the connection.
Setup and configuration
Open the Gemini Enterprise connector
Go to Connectors → App connectors and select Gemini Enterprise.
Enter your Gemini Enterprise details
Fill in the connection fields:
- GCP project ID: the Google Cloud project where your Gemini Enterprise engine lives
- Location:
global,us, oreu - Engine ID: the Gemini Enterprise engine identifier
Review the requested Google permissions
If you open Advanced settings, you can review the Google permissions requested for the connection. The default selection is appropriate for most Gemini Enterprise setups.
What to ask Lovable after connecting
Use prompts that make the intended experience clear. Gemini Enterprise supports both classic search interfaces and grounded chat-style answers.Limitations and troubleshooting
- Each connection is tied to the Google account that authorized it. What your app can retrieve depends on that account’s access.
- This connector supports Gemini Enterprise search and grounded answers. It does not expose raw APIs for systems behind Gemini Enterprise, such as Jira REST or Microsoft Graph.
- Gemini Enterprise actions that are only available in the Google Cloud console are not available through this connector.
- Website data stores are not supported for Gemini Enterprise search and assistant apps.
- If you see a 403 error, confirm the Google account can access the engine and the relevant data sources.
- If you see a 404 error, confirm the GCP project ID, Location, and Engine ID exactly match the engine in Google Cloud.
- If you see a 429 error, review quotas and rate limits in your Google Cloud project.
How to unlink projects from a connection
Editors and above can remove specific projects from a connection without deleting the connection entirely. The connection will remain available for other projects. To unlink projects:
When unlinked, those projects will no longer have access to through this connection. If a project needs again, you can link it to any available connection.