Release Notes (2.14)

You can now create delimited and unstructured data sources that access files in an Azure Blob Storage container. These data source types make it easier to incorporate both structured file data and unstructured content into your data workflows.

You can authenticate to Azure Blob Storage using any of the following methods:

• Access key

• Microsoft Entra ID

• Azure managed identity

• Shared access signature (SAS)

When you create a JDBC data source, you can now upload certificates, key stores, trust stores, and other certificate-related files. This support makes it easier to establish secure connections to JDBC data sources that require certificate-based configuration.

• Reset or disable AI Narratives— You can now reset a narrative to reprocess its contents from the beginning. You can also temporarily disable a narrative so that its content is not used by the AI Assistant. These controls make it easier to refresh content and manage which information is available to the AI Assistant.

• Faster, more predictable PDF processing— Knowledge Stores now process PDF documents significantly faster and within a more predictable time window, reducing the time required to make document content available to the AI Assistant.

• On-demand rebuilding of vector indices— A new system task lets system administrators rebuild the vector-table indices for a Knowledge Store. This task provides a controlled way to restore or refresh indices when needed.

Our AI Assistant Agents have received updated to improve their reliability, accuracy and efficiency. The following points highlight agent-specific updates:

• Graphing Agent improvements— The Graphing Agent supports a broader range of chart types and produces more accurate visualizations from user prompts, helping users explore and communicate their data more effectively.

• Data Model agent improvements— The Data Model and Data Model Diagram agents more accurately identify tables, relationships, and diagram structures, resulting in higher-quality generated data models.

• Text-to-SQL improvements— The Text-to-SQL agent produces more accurate and robust queries and handles a wider range of natural-language prompts. Its behavior is also more consistent and predictable across different scenarios.

• General reasoning improvements. Improvements to the AI Assistant reasoning pipeline produce more relevant, focused, and reliable responses across agents.

When the underlying OpenAI or Azure OpenAI service encounters an error, the AI Assistant now displays a clear, user-friendly message instead of the underlying technical error. These messages make service problems easier to understand and troubleshoot.