Skip to main content

Work with Minder

Where: sidebar → Home. The assistant is called Minder.

The chat service must be configured and reachable in your environment. Its answers and available tools depend on the current workspace, your access, and the deployed service capabilities.

Choose the scope

Start a new conversation or open chat history. Select a database when your question concerns a particular dataset. Database selection narrows context; it does not grant access to otherwise restricted data.

Ask a specific question

Describe the table, intended outcome, filters, and assumptions. For example:

Inspect the columns available in the training database.
Help me write a read-only query for orders by status.
Explain your assumptions before proposing any changes.

Replace training with a database you can access. Do not paste database passwords, setup links, or MFA secrets into the conversation.

Check the work

Inspect the assistant's response, SQL, chart data, and run details where offered. Validate key results against the catalog or an independently reviewed query. An answer is not proof that a requested backend action completed.

Review proposed actions

When a pending-action card appears, read its input, affected resources, and status before choosing Approve or Reject. Do not approve an unexpected destructive action. Expired or failed actions need review rather than repeated blind retries.

Charts and follow-up questions​

Where available, open the chart editor to review its title, type, field mappings, and underlying SQL. Refresh query-backed charts after checking the query. Chart export offers PNG, SVG, or Vega-Lite JSON in the reviewed UI; treat exported data and specifications according to your organization's sharing rules.

Follow up with a narrower question when an answer is ambiguous. If the connection is lost, use the reconnect/retry controls and check the existing conversation before sending a duplicate request.

The generic chat file-attachment control is marked as coming soon in the reviewed composer. PDF upload is a separate Documents workflow. Do not assume all document formats or every agent integration are supported.