Datafor uses the AI Agent service for conversational analysis, semantic search, vector indexes, and scheduled vector jobs. An administrator can enable the service and configure the public and internal endpoints from one page.

1. Open the AI Agent settings

1. Sign in to the Datafor console with an administrator account.
2. Go to Settings → System → AI Agent (Preview).
3. Open the AI Agent tab.

2. Enable the service

Turn on Enable AI Agent service. This allows Datafor to use the AI Agent for AI-assisted analysis, vector indexes, and scheduled vector jobs.

3. Configure the service endpoints

The current settings page has two endpoint fields:

Field	Used by	Configuration
AI Agent Server Address	The web console and browser clients	Enter the public URL that users' browsers can reach.
Internal Server Address of AI Agent	The Datafor backend	Keep it aligned with the public endpoint unless the backend must use a separate private address.

In the locally verified deployment, both fields are empty and the AI Assistant uses the unified Datafor service successfully. The UI shows http://localhost:28080/datafor/ai as a placeholder, not as a value that must be copied to every environment.

If your deployment supplies explicit endpoints, use the addresses provided by the deployment administrator. The public address must be reachable by browser clients. Use a different internal address only when the Datafor backend has a separate server-to-server route.

4. Save and verify

1. Click Save changes.
2. Return to Home.
3. Open AI Agent.
4. Confirm that the AI Assistant page loads and that New Chat, History, Common Questions, Ops, and LLM are available.

To prepare for many users submitting questions together, follow Managing High Concurrency. Worker count is configured in the installed Agent's server-side .env file; the endpoint settings on this page do not change analysis capacity.

Troubleshooting

Symptom	Check
AI Assistant does not open	Confirm that Enable AI Agent service is on and the configured public endpoint is reachable from the browser.
Backend AI requests fail but the page loads	Check Internal Server Address of AI Agent and server-to-server network access.
Save changes is disabled	The form has no unsaved changes. Modify a setting only when the deployment configuration needs to change.
The AI Assistant loads, but the progress title does not change until the answer appears	Datafor is behind a reverse proxy that buffers the AI Agent stream. Disable buffering for /datafor/ai/ as described in Deploying Datafor Behind Nginx.