GWP 26/2 Release Highlights

Targeted Access Governance: Grants granular editor permissions for specific repositories via REPO_MANAGER_xxx roles without conferring global admin rights.

Flexible SSO & Identity Debugging: Allows admins to assign temporary local database permissions to LDAP/OIDC users for fast incident inspection without modifying LDAP groups.

Operational Monitoring: Single Prometheus metrics endpoint eliminates re-configuration overhead when adding/removing repos. SPARQL-based native index fragmentation checks enable proactive compaction.

Curator Productivity & UX: Replaces legacy wizards with direct in-context AI enrichment (definitions, labels, narrower concepts) directly inside concept detail views.

Bulk Curation: Enables multi-item drag-and-drop across hierarchy trees, SKOS collections, and Taxonomy Builder inputs in a single gesture.

Unified Enterprise SSO: Extends single sign-on coverage across companion applications, including MirrorApp, GraphViews visual discovery, and Semantic Workbench.

Visual SHACL-First Architecture: Modern SHACL-based domain modeling framework that replaces complex, open-world OWL syntaxes with clear structural constraint shapes.

OWL-to-SHACL Ingestion: Effortlessly import existing legacy schemas and industry standards (Brick, FIBO, schema.org) into native SHACL shapes.

MCP-Native Agent Skills: Enables AI development environments (Claude Code, Cursor) to assist with schema bootstrapping, ontology analysis, and competency question mapping via Model Context Protocol.

Human-in-the-Loop Review Sandbox: Ensures AI-suggested schema additions are routed to a staging sandbox for visual inspection and line-item approval before committing to production.

Baseline Model Versioning: Allows modelers to snapshot development states (owl:versionInfo), revert states, and share models across teams via a central project catalog preview.

Streamed & Spooled Tagger Output: Prevents Out-of-Memory (OOM) errors during heavy extraction runs by streaming responses and spooling payloads above 10 MiB to disk.

Pipeline Warmup via API: Introduces programmatic Lucene index rebuilds post-ETL/SPARQL updates, preventing extraction delays for end users.

Productized Delivery Workflows: Replaces bespoke integration scripts with pre-built, configurable components for GraphDB, taxonomy tagging, and data transformations.

GraphRAG Ingestion Engine: Takes ownership of multi-format document parsing (PDF, DOCX, TXT) and chunking pipelines, decouples ingestion from GraphRAG code, and executes custom Python scripts inside a fully supported, Graphwise-branded task runner.

Automated Quality Benchmarking: Evaluates retrieval and generation quality against LLMs (OpenAI, Gemini, Bedrock, Claude) using RAGAS metrics and generates HTML reports with token and latency tracking.

Rich Concept Navigation in UI: Displays scope notes, broader/narrower concepts, related concepts and custom attributes with clickable navigation directly inside entity cards.

Security & Stack Upgrades: Upgrades Keycloak authentication to OAuth2, migrates backend services to Spring Boot 4.1.0, converts n8n MCP communication to HTTP Streamable, and secures stored API keys.

Delta Sync Engine: Propagates targeted, incremental taxonomy updates rather than performing full term store replacements.

Taxonomy Integrity & Tag Safety: Enables safe hierarchy reparenting, term consolidation with automatic tag migration, and draft concept exclusion from auto-tagging.

Multilingual Autotagging: Enhances non-English concept detection and search consistency across global enterprise content repositories.

Governed AEM Synchronization: Connects Graph Modeling directly to AEM Tag Manager supporting both full baselines and rapid incremental updates.

Preserved Content References: Matches tags by URI so existing content paths and tag histories remain intact without duplication.

Enterprise Scale & Security: Tested against 100,000 concepts, supports AEM Cloud Service and AEM 6.5, and replaces Basic Auth with secure OAuth 2.0 / API key authentication.

Auto-Tagging: This feature serves as a foundational capability that registers synchronized AEM content paths as annotation locations and supplies sample polling queries. These queries enable external enrichment pipelines to retrieve registered annotation metadata and the latest changed content within those specified locations.