Graph Modeling & Semantic Analytics 10.4 Release Notes

08/10/2026

Note

While the official Graph Modeling 10.4 release is scheduled for the end of October, these preliminary release notes serve as an overview of key feature improvements and bug fixes included in this minor release.

Graph Modeling and Semantic Analytics 10.4 introduce several improvements designed to streamline taxonomy curation and strengthen enterprise integration. From a unified Taxonomy Builder enrichment framework and intuitive multi-concept drag-and-drop tree operations to consolidated background index management and seamless SSO across companion tools, this release removes operational friction for knowledge engineers while giving IT teams greater security and control across the entire graph ecosystem.

Deprecation Notices and Breaking Changes

The following Graphwise Extractor API endpoints are scheduled for removal in version 11.0. Applications relying on these endpoints must migrate to supported alternatives before upgrading.

/extractor Endpoints to be Removed in Graph Modeling & Semantic Analytics 11.0

Endpoint	Description
`[POST] /api/annotate`	`[file]` Returns the document annotated with extracted concepts and extracted terms in RDF/XML representation
`[POST] /api/annotate`	`[text]` Returns the document annotated with extracted concepts and extracted terms in RDF/XML representation
`[POST] /api/annotate`	`[url]` Returns the document annotated with extracted concepts and extracted terms in RDF/XML representation
`[POST] /api/annotate/nif`	`[text]` Returns the document annotated with extracted concepts and extracted terms in the NIF format
`[POST] /api/annotate/store`	`[file]` Annotates the document with extracted concepts and stores it in the remote repository
`[POST] /api/annotate/store`	`[text]` Annotates the document with extracted concepts and stores it in the remote repository
`[POST] /api/annotate/store`	`[url]` Annotates the document with extracted concepts and stores it in the remote repository
`[POST] /api/annotate/store/text`	`[text]` Returns the document annotated with extracted concepts and extracted terms in RDF/XML representation
`[POST] /api/annotate/store/url`	`[url]` Returns the document annotated with extracted concepts and extracted terms in RDF/XML representation
`[POST] /api/annotate/text`	`[text]` Returns the document annotated with extracted concepts and extracted terms in RDF/XML representation
`[POST] /api/annotate/url`	`[url]` Returns the document annotated with extracted concepts and extracted terms in RDF/XML representation
`[GET] /api/annotate`	`[text]` Returns the document annotated with extracted concepts and extracted terms in RDF/XML representation
`[GET] /api/annotate`	`[url]` Returns the document annotated with extracted concepts and extracted terms in RDF/XML representation
`[GET] /api/annotate/nif`	`[text]` Returns the document annotated with extracted concepts and extracted terms in the NIF format

Endpoint	Description
`[POST] /api/categorization`	`[file]` Categorizes a given file by first extracting concepts and then aggregating them by their associated top concepts. The aggregated top concepts represent the categories. The more concepts are found in the file from a specific top concept, the higher this top concepts score (summing up to 100%).
`[POST] /api/categorization`	`[text]` Categorizes a given text by first extracting concepts and then aggregating them by their associated top concepts. The aggregated top concepts represent the categories. The more concepts are found in the text from a specific top concept, the higher this top concepts score (summing up to 100%).
`[POST] /api/categorization`	`[url]` Categorizes a given URL by first extracting concepts and then aggregating them by their associated top concepts. The aggregated top concepts represent the categories. The more concepts are found in the URL from a specific top concept, the higher this top concepts score (summing up to 100%).
`[GET] /api/categorization`	`[text]` Categorizes a given text by first extracting concepts and then aggregating them by their associated top concepts. The aggregated top concepts represent the categories; The more concepts are found in the text from a specific top concept, the higher this top concepts score (summing up to 100%).
`[GET] /api/categorization`	`[url]` Categorizes a given URL by first extracting concepts and then aggregating them by their associated top concepts. The aggregated top concepts represent the categories. The more concepts are found in the URL from a specific top concept, the higher this top concepts score (summing up to 100%).

Endpoint	Description
`[POST] /api/extract`	`[file]` Extracts and returns meaningful metadata like concepts and terms from a given file upload.
`[POST] /api/extract`	`[text]` Extracts and returns meaningful metadata like concepts and terms from a given text.
`[POST] /api/extract`	`[url]` Extracts and returns meaningful metadata like concepts and terms from a given URL.
`[GET] /api/extract`	`[text]` Extracts and returns meaningful metadata like concepts and terms from a given text.
`[GET] /api/extract`	`[url]` Extracts and returns meaningful metadata like concepts and terms from a given URL.

Endpoint	Description
`[POST] /api/extract/zip`	`[file]` Extracts and returns a list of documents with meaningful metadata like concepts and terms from documents which are inside a given archive file (*.zip) upload.
`[POST] /api/extract/zip/aggregated`	`[file]` Extracts and returns a single aggregated document with meaningful metadata like concepts and terms from documents which are inside a given archive file (*.zip) upload.
`[POST] /api/extract/zip/aggregated/async`	`[file]` Extracts asynchronously and returns a single aggregated document with meaningful metadata like concepts and terms from documents which are inside a given archive file (*.zip) upload.
`[POST] /api/extract/zip/async`	`[file]` Extracts asynchronously and returns a list of documents with meaningful metadata like concepts and terms from documents which are inside a given archive file (*.zip) upload.
`[GET] /api/extract/zip/task`	Returns a synchronously called extraction identified by the task ID.
`[GET] /api/extract/zip/task/aggregated`	Returns a synchronous called aggregation extraction identified by the task ID.
`[GET] /api/extract/zip/taskstatus`	Returns the current status of a asynchronously called extraction identified by the task ID.

Unified AI Taxonomy Enrichment & Taxonomy Builder Overhaul
Retired the legacy "Extend your taxonomy" wizard to eliminate tool duplication, enabling curators to perform fast, single-concept edits (generating definitions, alternative labels, or narrower concepts) directly in SKOS Details View. This includes flexible workflow entry points ("Build Hierarchy" or "Add Details") and support for both flat list and hierarchical expansion.
Bulk Operations via Multi Drag & Drop in the Taxonomy Tree
Modernized taxonomy tree interactions with multi-selection drag-and-drop capabilities, enabling knowledge engineers to re-parent groups of concepts in a single gesture, easily populate SKOS collections, and quickly seed inputs in the Taxonomy Builder.
Unified Single Sign-On (SSO) Support for Companion Apps
Extended enterprise Identity Provider authentication (SAML/OIDC) seamlessly across MirrorApp, GraphViews, and the Semantic Workbench, eliminating repetitive logins during cross-tool navigation while ensuring centralized access governance and uniform security compliance for IT teams.
Robust, High-Scale Batch Tagging (/api/tag Optimization)
By streaming tagging responses directly to clients, spooling oversized asynchronous payloads (>10 MiB) to disk, and introducing granular error-handling (failOnError=false), the platform safeguards system memory against Out-of-Memory risks and prevents entire batch jobs from failing due to isolated document issues.
Predictable User & API Response Times (Explicit Lucene Index Refresh)
By enabling teams to trigger Lucene index generation on demand via the API, pipelines can "warm" taxonomy extraction indexes immediately after data imports or direct SPARQL updates. This eliminates perceived random system slowness (delays of 30+ seconds on initial user extraction calls) and ensures tagging accuracy without relying on passive modification timestamp checks.

In this section: