Cloudera Docs

Using an AI visual

The AI visual allows you to explore datasets through natural language queries. You can ask questions by typing or speaking, review detailed execution steps to understand how responses are generated, and share parameters across visuals to keep dashboards synchronized.

Interacting with the AI visual

After configuring the AI visual, you can begin exploring your data through conversational queries. Ask questions about your data using voice or text. The AI visual analyzes the connected dataset and generates responses using natural language. Responses are streamed into the conversation as they are generated.

Adding text input:
  1. Type your question into the input field.
  2. Click SEND to submit the query.

The AI visual processes your question and returns a relevant response based on the dataset. When the response begins streaming, the answer appears progressively in the conversation.

Adding speech input:
  1. Click to ask a question using your voice.
  2. Click SEND to submit the query.

    The AI visual processes your question and returns a relevant response based on the dataset. When the response begins streaming, the answer appears progressively in the conversation.

Adding follow-up queries:

You can continue the conversation by asking follow-up questions to refine or extend the previous query, using either text or voice input.

Canceling a response:

After you ask a question, the SEND button changes to STOP while the question is being processed.

Click STOP to:

  • Stop displaying any additional streamed text immediately
  • Cancel the in-progress request
  • Return to idle state so you can ask another question right away.

If it was a new question, the pending question is removed from the chat and restored in the input box so you can edit and resend it.

If your question was a retry of a failed response, the original question remains in the chat history, and the retry is canceled.

Viewing response details

You can review how the AI visual generated a response. Execution Details provide a transparent, step-by-step explanation of how the query was processed and which data was used. This information can help you:
  • understand how the response was generated
  • verify which data was used
  • troubleshoot unexpected results

To view execution details, click next to the response. The Execution Details modal includes the following sections:

Overview
Displays summary metadata for the response, including:
  • AI approach (SQL / Vector / Previous Data) with description
  • Completion profile
  • Completion model details:
    • ID
    • name (when available)
    • engine
  • Embedding profile
  • Embedding model details (vector approach only):
    • ID
    • name
    • engine
  • Execution summary (completed versus total steps, including failures)
  • Execution duration
  • Data summary (rows × columns)
  • Dataset name
  • Number of applied filters
  • Retry indicator (if the response is a retry, including failed step)
  • Original question
  • Response length (characters)
Execution Steps
Shows a breakdown of how the response was generated:
  • Checking Previous Data determines whether earlier queries already contain the required information.
  • Collecting Dataset Information and Filters shows dataset details and applied dashboard filters.
  • Generating and Validating SQL shows the generated query and visual configuration.
  • Analyzing Results provides the natural language summary of the query results.
  • Vector Search / Generating Response performs embedding-based retrieval and generates the final answer. (vector or text-only paths)

You can expand each step to view the associated data, dataset information, filters, or query results.

To save the execution details, click Download.

Sharing parameters

When querying a dataset with an AI visual, the generated parameters can be shared with other visuals on the dashboard to enable cross-filtering.

Sharing parameters allows you to explore relationships between different data points more efficiently. When a shared parameter changes, all dependent visuals on the dashboard update automatically to reflect the same subset of data. This eliminates the need to adjust each visual individually and helps maintain consistency across the dashboard.

  1. Click next to the response.

    The Share Parameters modal lists the available parameters with column and values information.

  2. Select the parameters you want to share.

    You can use Select All or Select None if multiple parameters are available.

  3. [Optional] Enable the toggle to include null or empty values in shared parameters.

    If enabled, null or empty results from the AI query are included in the shared parameters, allowing other visuals to filter for missing or empty data.

  4. Click SHARE PARAMETERS to apply the changes.

    You can click Cancel to close the modal without applying parameter sharing.