Microsoft Fabric Updates Blog

Unlock faster insights with the new support for Copilot conversations in Real-Time Intelligence (Public Preview)

Copilot in Real-Time Intelligence is now in public preview, including the support for conversational interactions when translating natural language (NL) into Kusto Query Language (KQL). This new capability allows you to refine, adapt, and extend your queries dynamically, all while maintaining the context of your previous inputs.

What’s New with Conversation Support?

Copilot in Real-Time Intelligence now supports conversational interactions, allowing for an iterative and interactive query-building process. Here’s how you can make the most of this new feature:

1. Dynamic Query Refinement

If the initial KQL generated by Copilot doesn’t quite meet your needs, you can now refine your prompt to remove ambiguity, specify the correct tables or columns, or provide additional context. This allows you to quickly adjust the query without starting over, helping you arrive at the desired results more efficiently.

2. Seamless Follow-Up Questions

If the generated KQL is correct but you want to continue exploring the data, you can ask follow-up questions related to the same task. Whether you want to expand the scope of your query, add filters, or explore related data points, the support for conversations enables you to keep building upon the previous dialogue, providing a more fluid and natural interaction with your data.

A screenshot of a chat

Description automatically generated

How Does It Work Behind the Scenes?

The power of this enhancement lies in its ability to maintain context and continuity. Here’s what happens under the hood:

– Context-Rich Interactions: Each time you interact with the Copilot, the full dialogue history, including your previous prompts, Copilot responses, and any relevant Retrieval-Augmented Generation (RAG) data is sent to our language model. This context-rich approach ensures that subsequent query generations are aware of your previous inputs, creating a seamless conversation flow.

– Adaptive Updates to RAG Data: Depending on your follow-up prompts, we may update the RAG data, which includes relevant database schema details and other context. This adaptability allows the Copilot to stay aligned with the most pertinent information, further improving the accuracy of generated queries.

Learn more about Copilot in Fabric and Copilot in Real-Time Intelligence

Help us with your feedback

We’d love to hear what you think and how you’re using Real-Time Intelligence. The best way to get in touch with us is through our community forum or submit an idea. For detailed how-tos, tutorials and other resources, check out the documentation

Related blog posts

Unlock faster insights with the new support for Copilot conversations in Real-Time Intelligence (Public Preview)

October 17, 2024 by Gabi Lehner

Microsoft Fabric’s new Real-Time Dashboard permissions feature brings granular control to how users interact with Real-Time Analytics. With the introduction of separate permissions for dashboards and underlying data, administrators now have the flexibility to allow users to view dashboards without giving access to the raw data. This separation is key for organizations that need to … Continue reading “Real-Time Dashboards and underlying KQL databases access separation (preview)”

October 15, 2024 by Yael Biss

Extending Microsoft Purview’s Data Loss Prevention (DLP) policies into Fabric lakehouses is now in public preview! This follows the success of DLP for Power BI and the GA of Microsoft Fabric last year. DLP policies help you automatically detect sensitive information as it is uploaded into lakehouses in your Fabric tenant and take risk remediation … Continue reading “Announcement: Microsoft Purview Data Loss Prevention policies have been extended to Fabric lakehouses”