> ## Documentation Index
> Fetch the complete documentation index at: https://docs.k16solutions.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Use the DataX assistant

> Ask questions across catalog metadata, glossary definitions, and Snowflake data.

The DataX assistant helps you discover relevant entities, interpret glossary context, plan analyses, and work with Snowflake data. Use it to accelerate exploration while you remain responsible for validating every result.

<Note>
  **Assistant** appears only when AI assistant access is enabled for your workspace. The assistant
  can use only the catalog and data context available to your account.
</Note>

## Start a conversation

Open **Analytics** > **Assistant**, then enter a specific request in **Explore data…**.

Good prompts include the goal, population, time window, grouping, and desired output. For example:

```text theme={"system"}
Find the catalog entities I would need to compare enrollment counts by department and term. Explain the grain and join path before drafting SQL.
```

You can also begin with one of the suggested prompts. Keep related follow-ups in the same conversation so the assistant can use the existing context.

## Manage context

Open **AI chat settings** to control **Include blueprints**. Turn it on when client and K16 blueprint entities should be available to catalog and schema tools. Turn it off when you want the assistant to focus on non-blueprint sources.

The settings menu also lets you choose whether **Enter** or **Shift+Enter** submits a prompt.

Use the conversation rail to start a new conversation, switch between prior conversations, search history, or delete a conversation you no longer need.

## Ask for verifiable work

Structure requests so you can check the response:

1. Ask the assistant to name the entities and columns it plans to use.
2. Ask it to state the expected grain and join keys.
3. Request narrow SQL with a bounded date range.
4. Run the SQL in a [worksheet](/datax/analytics/worksheets).
5. Compare the output with catalog descriptions and glossary definitions.

<Warning>
  The assistant can produce incorrect SQL or interpretations. Do not publish, share, or act on its
  output until you verify the source entities, access rules, calculations, and returned data.
</Warning>

## Improve a weak response

* Replace broad terms such as “recent” with exact dates.
* Name the intended metric or link to its glossary term.
* Specify whether you need discovery, an explanation, SQL, or a result summary.
* Ask the assistant to call out assumptions and missing catalog context.
* Start a new conversation when prior context is steering the answer in the wrong direction.

For SQL error help inside a worksheet, see [Export and troubleshoot worksheets](/datax/analytics/export-and-troubleshoot).