Graph Modeling and Semantic Analytics 10.3 Release Notes

Add-ons Rebranding: The Workbench add-on has been updated with new Graphwise logos and color schemes for seamless visual integration with the core application.

Landing Page Alignment: The landing pages for Extractor and GraphSearch have been aligned with new brand guidelines, including updated logos, icons, and favicons.

Platform-Wide Consistency: Typography, descriptive texts, and external links have been comprehensively updated. Legacy company names and slogans are systematically being updated to Graphwise equivalents, focusing on primary user-facing interfaces in this release.

Application-Specific Isolation: Decouples security roles from global realm levels down to application-specific client roles (Graph Modeling, Extractor, and GraphSearch operate under individual Client IDs). This prevents oversized authorization tokens, minimizes security risks by isolating permissions, and simplifies compliance auditing.

Secure Machine-to-Machine (M2M) API: Replaces legacy Basic Authentication with secure, temporary OAuth 2.0 / OIDC Bearer Tokens (Client Credentials flow) issued by Keycloak. This eliminates static credentials in codebases and automated pipelines, vastly improving data security without breaking backwards compatibility.

Zero-Friction Upgrade Paths: An intelligent, idempotent migration utility runs automatically on startup to map and convert legacy realm roles into client-scoped roles. This lowers IT overhead and guarantees legacy security setups function in a backward-compatible "hybrid mode" during migration.

Clutter-Free Access Management: Internal system-managed accounts, backend technical roles, and system-level Keycloak groups are dynamically filtered out of user-facing UIs. Standard users and project managers only see relevant end-user names and functional business groups.

Turnkey, Production-Ready Deployments: Out-of-the-box templates (poolparty-realm.json and Docker Compose) come pre-configured for GraphDB-to-PoolParty Keycloak authentication. This accelerates time-to-production for Customer Success teams and dramatically speeds up Proof of Concept (PoC) environment setups.

Enabling File Size and Content Type Validation: To prevent malicious upload attacks, Graph Modeling now validates file uploads against specific file sizes and content types (including file extensions and headers) so only static content of pre-defined formats is accepted.

URL User Identity Exposure Fix: To align with modern security best practices, the application now completely avoids exposing user identities in request URLs to access resources. The legacy username query parameter is ignored and no longer affects authorization or returned user data.

CSP Filter Optimization: Resolved an issue in the Linked Data Frontend (LDF) where custom unsafe-inline configurations for script-src and style-src in the Content Security Policy were being overwritten by automatically generated backend nonce values.

Automated Multilingual Generation: The Taxonomy Builder now generates preferred labels, alternative labels, and definitions across all configured project languages in seconds.

Accelerated Taxonomy Expansion: Eliminates the need for manual translation workflows or external third-party services, significantly reducing localization costs and labor.

Seamless Native Workflow: The workflow is fully integrated into the existing Taxonomy Builder vertical menu, allowing users to select target languages and generate content natively.

Human-in-the-Loop Quality Control: Keeps enterprise teams in control with built-in mechanisms to review, edit, and approve generated localizations before they are accepted into the formal taxonomy.

Fixed an error that prevented deletion of custom attribute values when the attribute type was set to "Any" (unrestricted).

Fixed an issue where the dcterms:identifier property on Concept Schemes incorrectly required a language tag, which is not necessary for this language-neutral property.

Fixed a parsing error in the Linked Data Frontend that caused the entire frontend to fail when configured with two-part language tags (e.g., "en-GB").

Fixed a URI resolution error that prevented the Visual Browser from loading for the root project node in the Linked Data Frontend.

Fixed a layout bug in the Project Linking view where the left-hand linking panel shifted to the vertical center of the screen instead of staying top-aligned when many attribute mappings were present.

Fixed a UI inconsistency in Advanced URI Settings where the "ID" field was incorrectly displayed and marked as mandatory when the "Preferred Label" or "UUID" mode was selected.

Resolved significant performance degradation when loading the Notes tab for concepts in environments with a large number of Keycloak users, reducing response times from several minutes down to seconds.

Resolved Lucene tagger issue where flagging a concept via the "Add to Exclude List" option failed to omit the concept from subsequent extraction results.

Resolved a regression introduced in version 10.2 where opening a specific document within the Corpus Documents tab failed to display highlighted concepts and terms due to console errors.

Tomcat 11.0.22

GraphDB 11.4.1

Java 21

Keycloak 25.0.6

RDF4J 5.2.2

Spring Core 6.2.17

Spring Security 6.5.9

Lucene 9.9.2

Apache Spark 4.1.1

Elasticsearch 9.3.3

ADF 1.9.0

Semantic Workbench 2.5.0

GraphViews 1.0.1

Mirror App 2.5.0

UnifiedViews 10.2.3