Skip to content

Getting the most out of Maia🔗

If you're new to Maia, check out the Maia overview to learn more about what it can do.


Prompting basics🔗

To interact with Maia:

  • Type your request in natural language into the text box at the bottom of the Maia panel.
  • Press Enter or click Send.
  • Maia responds in real time, showing results directly in the Designer.

Depending on your request, Maia may:

  • Add new pipelines to the canvas.
  • Modify or highlight components.
  • Delete components and update your pipeline view.
  • Display quick follow-up suggestions to continue the task.

You stay in full control. Maia won't make irreversible changes without your approval.

Writing effective prompts🔗

Maia performs best when your prompts are clear, specific, and actionable.

  • Be specific: "Join customers and orders on customer_id" is better than "Join some tables".
  • Provide names: Refer to pipelines, tables, or components by name when possible.
  • Break up complex tasks: Use multiple prompts for multi-step processes.
  • Reference selections: Selecting components before prompting helps Maia understand context.
  • Try variations: If something doesn't work, rephrase your request or give Maia more detail.

Example prompts🔗

  • Get data from the 'customer_orders' table and only show orders over $100.
  • Join the 'reviews' and 'customers' tables and write it to a new table called review_customers.
  • Run the current pipeline and tell me if there are any issues.
  • Calculate a margin column as (revenue - cost) / revenue and store it as 'margin'.
  • Can you explain what the pipelines do that are in the ascend/ directory?
  • Filter out all invalid emails and write the results to 'sanitized_emails'.
  • Connect to my Salesforce instance and extract contact records into Snowflake.
  • Run the 'daily_sales' pipeline and commit the changes to Git.
  • Create a visualization showing the number of orders per week.
  • Find any components that use the 'customer_id' parameter.
  • Look for existing Salesforce components I can copy the configuration from.
  • Find all Python scripts in the 'scripts' folder.
  • What pipeline files mention the 'orders' table?
  • Move my 'customers' pipeline into the 'customer-data' folder.
  • Make a copy of the 'weekly_orders' pipeline and rename it to 'monthly_orders'.

Maia prompt bar🔗

The Maia prompt bar is always available at the top of the Designer both on the landing page and while working on the canvas. It's labeled "Looking to build? Ask Maia".

You can:

  • Type into the prompt bar to open the Maia panel and submit your request.
  • Reopen Maia anytime by clicking the Maia icon on the Designer canvas.
  • Use suggestions in the prompt bar to:
    • Start building a new pipeline.
    • Get helpful actions for an existing or running pipeline.

These suggestions adapt to your context, and help you move faster with fewer clicks.


Managing your conversation with Maia🔗

Maia works through an interactive, step-by-step conversation in the panel. At any point, you can stop a task in progress or clear the current conversation to start a new one.

Stop or Cancel the current tasks🔗

If Maia is actively processing a request or you want to cancel what it's doing:

  • Click the Cancel button in the Maia chat panel (appears during active tasks).
  • This stops the current response from continuing but keeps your session open.
  • You can continue the conversation or ask something new immediately.

Note

Canceling a task does not clear the conversation history, it only stops the current action.

Clear or reset the conversation🔗

If you want to completely reset the conversation and start fresh:

  • Click the "New Chat" button at the top right of the Maia panel.
  • Alternatively, refresh the page in your browser. This also clears the current chat context.
  • After resetting, Maia will greet you with a welcome message and quick start options.

Note

This is useful when switching tasks, clearing older history, or starting from a clean slate.

Got feedback or spotted something we can improve?

We'd love to hear from you. Join the conversation in the Documentation forum!