🚀 This week in Maia: Sampling improvements, Context Engine defaults, mid-task steering, and more!

This week, we're rolling out several major updates designed to enhance pipeline flexibility, improve collaboration, and make your experience with Maia Team smoother than ever. From smarter filtering controls to mid-task steering, these features put more control in your hands while streamlining your data workflows. Check out the latest changelog updates for more details.

📌 Pin a filter across your pipeline, only when you want to

Filters stay on the component you set them on, and you can now pin one to follow you through the pipeline.

Each filter in the Sample Data panel has a pin icon. Pin a filter and it applies to every component in the pipeline. Unpin it and it goes back to being local.

Filters are now per-component by default, so the chips always describe the data you're looking at. When you want one filter everywhere, you can pin it. Removing a shared filter asks you to confirm first so you don't clear it on other components by accident.

See how to pin and unpin filters in our documentation.

🎉 Sampling improvement: All requested rows returned

When you preview a sample containing very large values, such as long JSON or free text, Maia now returns the rows you asked for. Values too large to show safely are clearly marked, and you can click to open the full value on demand.

Instead of losing whole rows to one oversized value, you now get the full sample with only that value shortened, and you can expand it in one click.

Learn more in Sampling data.

🎲 Smarter sampling in the pipeline canvas

You can now pull a random sample instead of just the first rows, and sort sample results returned from the data warehouse by any column.

In the sample panel for transformation pipelines, you can choose between the default sample and a random sample, and sort the sample results by any column. This helps you see beyond the first rows in a fixed order, making it easier to verify whether your changes have worked across the entire dataset.

You decide when you want a random look at your data, giving you more control over your data exploration process.

For more information, read Sampling data.

🎯 Steer Maia Team mid-task with messages

You can now send a message while Maia Team is still working. Maia Team picks it up in between actions and adjusts accordingly.

If something else comes up, or you see Maia Team heading in a different direction than you intended, you can send a follow-up at any time and steer the conversation as it happens.

What's new:

• Steer Maia Team as it works: Send a message at any point during a run. It's delivered mid-request and Maia Team acts on your updated instructions.
• Message queuing: Messages are queued until they can be delivered, and you can queue up to 5 messages.
• Cancel before delivery: Changed your mind? Cancel a queued message before Maia Team picks it up.

How do I use it?

Type your message and send it while Maia Team is working. Queued messages appear above the message field until Maia Team picks them up.

💡 Note: If you stop a run, any messages still queued are canceled.

Read more about Queueing messages.

🔧 Runner diagnostics support restored

Maia runners have received an interim release that restores essential support diagnostics functionality. Account ID and Runner ID are now back in runner logs and metrics, helping our support teams investigate any issues you might encounter.

While this update is optional and won't change how your pipelines run, we highly recommend applying it so support investigations can be resolved as quickly as possible. To apply the update, simply restart or redeploy your Maia runner.

🧠 Context Engine is now on by default in Maia Team

The Context Engine lookup is now enabled by default in Maia Team conversations. Maia Team picks the relevant knowledge graph for your question on its own, so you don't need to choose one.

This means you no longer have to turn Context Engine on yourself. Maia Team can bring in your organization's context (pipelines, data assets, conventions, and more) when your question needs it, with no setup on your side.

Key improvements include:

• On by default: Context Engine is now switched on for every Maia Team conversation.
• Automatic graph selection: Maia Team looks at what each knowledge graph contains, along with a description generated automatically for it, and picks the graph that best matches your prompt.
• Only used when needed: As before, Maia Team only calls the Context Engine tools when your prompt needs them. If your question doesn't need anything from the knowledge graph, Maia Team skips the lookup and carries on as usual.

Simply ask Maia Team your question as usual. To turn it off for a conversation, open the conversation menu and switch off the Context Engine toggle. You can switch it back on from the same menu at any time. For more information, read Context Engine in chats.

💬 We'd love to hear from you!

Let us know how these new features are improving your workflows. We're all ears! Share your comments or questions with us in the Matillion Community.