Release Notes for Mindbreeze InSpire
Version 25.1

Innovations and new features

Insight Services for RAG uses semantic text structures from the index for more precise answers

Mindbreeze InSpire extracts semantic text structures such as headings, paragraphs and tables from different content and maps them in the index. The extracted structures are used in the document preview, for example. The Mindbreeze InSpire 25.1 release now also enables the use of semantic structures when generating answers. The Retrieval Augmented Generation (RAG) and the Large Language Model (LLM) receive the structural information in addition to the relevant answer text excerpts to generate even more precise answers.

Link to the documentation

Customisability of the Similarity Search through embedding information

With the Mindbreeze InSpire 25.1 release, administrators can now customise the Similarity Search during indexing. This is possible by adding embedding structures in the form of content and annotations. By customising the Similarity Search, Mindbreeze users receive more precise answers to their questions.

Link to the documentation

Download of snapshots as OCI image possible

Mindbreeze InSpire administrators can save changes to their appliance as snapshots and reapply them at a later date. It is also possible to share modified appliances with downloaded snapshots. With the Mindbreeze InSpire 25.1 release, snapshots can now also be downloaded as OCI image files. This makes the use of downloaded snapshots more flexible and enables them to be used for special use cases, such as Kubernetes containers.

To download a snapshot, it must be added to the Snapshot Archive. The file format is selected from a drop-down menu in the archived snapshot.

Link to the documentation

New and advanced connectors

Airbyte Data Integration Connector enables access to over 500 data sources

Mindbreeze InSpire offers a wide range of connectors to enable the indexing and use of various data sources in Insight Apps and 360-Degree views. With the Mindbreeze InSpire 25.1 release, the full flexibility of Airbyte is available with over 500 connectors. Users can significantly expand and deepen their Insight Apps and 360-Degree views with additional data. This enables further insights from company data and completely new use cases.

The Airbyte Data Integration Connector supports data sources such as Instagram, PostgreSQL and Paypal Transactions. To use the connector, a self-operated Airbyte installation is required. Contact Mindbreeze Support to get access to the Airbyte Data Integration Connector.

Alternative SCIM Principal Resolution Service for Microsoft Azure available

Mindbreeze customers receive an alternative Principal Resolution Service for Microsoft Azure with the Mindbreeze InSpire 25.1 release. The protocol “System for Cross-domain Identity Management (SCIM)” allows administrators to handle special use cases with the SCIM Principal Resolution Service. For example, it is also possible for Mindbreeze InSpire to operate with restricted Microsoft Azure API authorisations.

Link to the documentation

Additional improvements

Basic operating system updated and security increased

As part of the Mindbreeze InSpire 25.1 release, customers will receive further optimisations, such as a more efficient InceptionImageFilter that performs image classifications faster. In addition, the base operating system Fedora CoreOS has been updated to version 41 and the security of Mindbreeze InSpire has been further enhanced with the update of Chromium.

For more information, see Security relevant changes.

Optimised documentation available on help.mindbreeze.com

With the Mindbreeze InSpire 25.1 release, users receive additional support for the operation of Mindbreeze InSpire. Optimised documentation is now available at help.mindbreeze.com.

Link to help.mindbreeze.com

Security relevant changes

25.1.0.496

• Updated: Fedora CoreOS to version 41 (CVE-2024-50067).
• Updated: Chromium to version 132.0.6834.159 (CVE-2024-11395, CVE-2024-12053, CVE-2024-12381, CVE-2024-12692, CVE-2024-12693, CVE-2024-12694, CVE-2024-12695, CVE-2025-0434, CVE-2025-0435, CVE-2025-0436, CVE-2025-0437, CVE-2025-0438, CVE-2025-0439, CVE-2025-0441, CVE-2025-0442, CVE-2025-0447, CVE-2025-0448, CVE-2025-0291, CVE-2025-0611, CVE-2025-0612).
• Improved: Security and performance by replacing Tensorflow with ONNX runtime (CVE-2023-33976).

Additional changes

25.1.1.551

• Added: Experimental docling API to Insight Services.
• Fix for: Crawling sites with special characters in their relative path.
• Fix for: Under certain circumstances, the servlet thread of the index servlet „/wait“ did not terminate when the connection closed.
• Fix for: OAuth2 Token refresh in mindbreeze-sdk.
• Fix for: A Backup in HA Keycloak caused new nodes to join the Infispan Cluster, which lead to Split Brain situations and errors in certain cases.
• Fix for: The application option "ajaxOptions" will now be passed to the child applications.
• Fix for: NLQA Best Bets in the AI Chat.
• Fix for: Trusted Peer Communication between the Client Service and the Query Service for Windows.
• Fix for: The Management Center XFrameOptions response header prevented the Insight Apps View from loading.
• Fix for: Edge Store policy compliance issue in the Add-on plugin.
• Fix for: Added "requires-user-input" to the component „searchform“ in the Insight App Designer.
• Changed: The index setting "Disable Unrestricted Privileged Servlets" is now renamed to "Enable Unrestricted Privileged Servlets" and offers a drop-down list with three options.
• Removed: The index servlet „/categorizeitems“.

25.1.0.496

• Added: Support for Server Authentication (with JWT) for the Box Connector.
• Added: Support for OCI as a Snapshot format.
• Added: Support for the download of snapshots in archive and OCI image format.  
• Added: InsightServices requests are instrumented in the app.telemetry logpool "Insight Services Requests".
• Added: mes/ipc/commons/License Status.proto and mes/ipc/query/QueryService Extensions.proto to messdk-public-api-<version>.tar.gz.
• Added: Support for downloading service properties as JSON or YAML formats.
• Added: Security response headers for the Mindbreeze Management Center (cross-origin-opener-policy: same-origin, cross-origin-resource-policy: same-site, permissions-policy: geolocation=(), camera=(), microphone=(self), referrer-policy: strict-origin-when-cross-origin, strict-transport-security: max-age=2592000, x-content-type-options: nosniff and x-frame-options: SAMEORIGIN.
• Added: unstructured.io API to Insight Services.
• Added: Filter Service to call external Filter Plugins.
• Added: Experimental docling API to Insight Services.
• Added: Web Connector Deployment in Kubernetes.
• Fix for: Links that were generated in answers opened in a new page.
• Fix for: Typo in published confluence artifacts.
• Fix for: CommonMark Filter did not work, if Jericho HTML Filter was disabled.
• Fix for: Diacritic normalization of suggest requests failed to correctly match umlaut.
• Fix for: Timeout in the ProxyServlet Read/Connect configuration.
• Fix for: QueryService servlet /nosample/find executed sample texting code (performance issue).
• Fix for: In Microsoft CRM Connector, an issue occurred, where certain Entity Relation Resolution configurations would no longer work when activating the setting "Use REST API".
• Fix for: In Microsoft CRM Connector, an issue occurred, where certain correct Entity Relation Resolution configurations would be unable to resolve the relationship correctly.
• Fix for: NullPointer within the Web Connector and for the Crawl Controller.
• Fix for: Problems using Filter ID in a remote index configuration.
• Fix for: Remote crawler problems were caused by the configuration of base URLs containing a path in a remote index configuration.
• Fix for: File manager download directory failed rarely.
• Fix for: content.filter_extension was not considered for dispatching to the correct ContentFilter service.
• Fix for: Answer retrieval when the question contains an apostrophe (").
• Fix for: The inverted index could be incomplete under certain edge conditions.
• Fix for: A Query performance issue occured in the Similarity Search, when the answer was extracted from large content.
• Improved: Usability for closing a stack without a stack title.
• Improved: Enhancement at the included search response answers, when generating an answer from Gen AI.