Corporate Memory 25.3.0

Corporate Memory 25.3 is the third major release in 2025.

The highlights of this release are:

• Build: Mapping Creator

  • Experience a next-generation, GenAI-driven visual mapping tool that redefines how you build and manage mapping rules—with unparalleled clarity, speed, and simplicity.

• Explore: Companion Chat-Based Data Interaction

  • Introducing an LLM-powered conversational interface that lets you interact directly with your data. Ask questions about your graphs, explore insights using query catalogs or autogenerated queries, and access Corporate Memory resources and modules—all within chat. This feature takes data accessibility and interaction to an entirely new level.

• Explore: Graph Insights

  • Visually explore your graphs contents and relationships in an aggregated representation that scales easily to graphs with millions of nodes. Offering a compact and flexible overview aggregating objects and relations and shows details on demand.

This release delivers the following component versions:

• eccenca DataIntegration v25.3.0
• eccenca Explore v25.3.0
• eccenca Corporate Memory Control (cmemc) v25.5.0

We tested this release with the following dependency components:

• Ontotext GraphDB v11.1.2
• Keycloak v26.4.5

More detailed information for this release is provided in the next sections.

eccenca DataIntegration v25.3.0

We are excited to announce the release of DataIntegration v25.3.0, bringing the new LLM powered Mapping Creator experience, explicit Databricks connectivity, richer dataset and plugin APIs, streamlined Python plugin tooling, and broad workflow and editor reliability improvements.

v25.3.0 of DataIntegration adds the following new features:

• Data sources and workflow operators:
  • Added explicit Databricks support.
  • Added an operator that allows arbitrary SQL statements as input for SqlDataset.
  • Added an endpoint for browsing files to simplify file selection.
  • Added a documentation link to the comparison task node to surface guidance in-context.
  • Added the ** operator to the JSON dataset so all child nodes can be selected recursively.
• Text preparation:
  • Introduced a new default stop word remover and improved the existing stop word transformers.
• Plugin APIs:
  • Added the pluginType attribute to the /core/plugins endpoint response.
  • Added a /pluginTypes endpoint to list all plugin types.
  • Added the backendType attribute to plugin JSON payloads to highlight whether a plugin is native or Python-based.
• Python plugin development:
  • Added autocomplete_query to ParameterType, providing a more flexible alternative to the legacy autocomplete function.

v25.3.0 of DataIntegration introduces the following changes:

• Data connectors and storage workflows:
  • Adapted the Google Drive Spreadsheet dataset tests to accommodate Google’s changed behavior.
  • Projects imported into a knowledge graph backend now delete the target graph before import to avoid stale data.
• Mapping Creator experience:
  • New quick tour that introduces the important elements of the Mapping Creator.
  • Mapping rules opened from the Mapping Creator no longer need to be saved immediately.
  • Newly created mapping rules that are not yet persisted can now be edited directly in the Mapping Creator.
• UI and documentation improvements:
  • Improved the formatting of operator examples for better readability.
  • Updated the application to the new color palette.
  • Made decoding underscores to spaces an explicit option in the Strip URI prefix transformer.
  • Improved mapping suggestions and shown justifications.
  • Support of GPT-5.
  • Many small improvements in the Mapping Creator.
  • The search lists now persist some user settings, such as the page size and sorting order.
• Python subsystem and plugin lifecycle:
  • Loading a Python plugin that reuses an existing identifier now records an error in the discovery result.
  • The /listPlugins and /updatePlugins endpoints now return HTTP 500 if at least one plugin fails to load (previously HTTP 200).
  • Upgraded the embedded Python subsystem:
    • Installation now relies on uv 0.9.2.
    • Updated the packaged Python distribution to Python 3.13.8.
    • Replaced the Red Hat UBI system Python with a uv-installed runtime.
    • Uses a read-only virtual environment from /usr/local/venv.
    • Plugin package installation now uses uv instead of pip.

v25.3.0 of DataIntegration ships the following fixes:

• Workflow editor and execution:
  • Edge arrows are displayed again in the workflow editor via React Flow v12.
  • Nested workflows no longer expose data input ports.
  • Workflow nodes loaded from the store retain their original positions.
  • Multiple operators created via “Connect to newly created operator” no longer overlap.
  • Fixed the display of workflow reports inside the modal dialog.
  • The workflow report now correctly marks failed workflows with a red danger icon and shows “Execution failed” on failed operators.
  • Transform execution reports no longer show successful executions for rules that were not run.
  • Fixed graph uploads for large files or chunk sizes.
  • Exporting a workspace without project resources now succeeds.
  • Downloading a project/workspace with empty files stored in AWS S3 works again.
  • Fixed the display of workflow reports in Plugin Custom Action Reports so the code area stays within bounds.
• Mapping and rule authoring:
  • Restrictions no longer expand the wrong SPARQL pattern.
  • URI rules are no longer created automatically when the default rule is opened in the rule editor.
  • Mapping Creator undo/redo actions behave as expected.
  • Suggest mappings now adds all selected rows, not just the ones currently visible after a text filter.
  • Rule editor nodes are no longer copied when text is selected in Firefox.
  • URI pattern no longer changes from nothing (“automatic default pattern”) to “{}/root” after opening the value formula editor.
• Python plugins and orchestration:
  • Listing Python plugins no longer fails when the orchestration feature enable-extension EXTENSION=plugins is used.
• Transformations, datasets, and templates:
  • JSON #arrayText now returns an empty array representation for empty arrays.
  • Jinja templates no longer error when the loop variable is used inside functions within for loops.
  • Previewing parameter descriptions no longer renders broken Markdown.
  • Use FROM instead of GRAPH in the vocabulary loader to respect graph imports.
• Graph and project handling:
  • When importing into a knowledge graph backend, the graph is only deleted if it exists.
  • Downloading project/workspace with empty files stored in AWS S3 work now.
• User interaction fixes:
  • Global and workflow-specific hotkeys no longer trigger when another modal is open.
  • Display the template toggle consistently in the artefact creation dialog.
  • Global and page specific keyboard shortcuts should be ignored when a dialog is open.
  • “Description” displays content no longer differently than entered.
  • Graph URI Drop Down does close now.
  • Close icon button to remove target entity type in transform operator now works if opened in modal.

eccenca Explore v25.3.0

We are pleased to announce Explore v25.3.0, which delivers the new Companion LLM data interaction, a whole new way to explore your data relationships in Graph Insights, experimental Tentris support, major BKE authoring enhancements and a wide range of usability and stability fixes.

v25.3.0 of Explore adds the following new features:

• Graph Insights
  • New visual graph exploration and visual query UI for easy relation analysis.
• Companion:
  • LLM based integrated chat to interact with your data, generate or user SPARQL queries, providing a deep integration with Corporate Memory features
  • MCP Server SSE endpoint to provide the tool powering Compaion to external clients
• Backend:
  • Added experimental support for the Tentris store backend.
  • Added support for configuring a fixed authorization bearer token in the HTTP store backend.
• BKE authoring enhancements:
  • Execute workflows triggered by node shapes directly from BKE.
  • Simplified connecting nodes from the possible resources menu on the canvas.
  • Added the missing tooltips for multiple BKE buttons.
• SHACL authoring usability:
  • Added a tooltip to the LiteralContainer component.

v25.3.0 of Explore introduces the following changes:

• Thesaurus navigation:
  • Navigation tree items are now sorted by their labels for consistent browsing.

v25.3.0 of Explore ships the following fixes:

• General platform fixes:
  • Handle external and internal redirects to preserve unsaved work.
  • Restored the LinkRules download button.
  • The SPARQL GSP DELETE request now returns 404 for non-existing graphs consistently across backends.
  • Explore now relies on GraphDB to manage stopwords, ensuring correct filtering.
  • Fixed an empty state regression when the graph list query returns no results.
  • Failures in the shape catalog no longer block other tabs.
  • Prevented blank pages in the query editor when previous queries were not saved.
  • SHACL now handles custom value queries without bindings more gracefully.
• BKE stability improvements:
  • Improved error handling for validation requests in the BKE module.
  • Fixed the BKE visualization saving issue.
  • UI query requests for connectable candidates now only run for resources.
  • The “Open resource details” link can again be opened in a new browser tab.
• SHACL editing reliability:
  • Added a loading state for SHACL edit controls to provide clearer feedback.

v25.3.0 of Explore removes the following functionality:

• Removed the deprecated LLM assistance feature.

eccenca Corporate Memory Control (cmemc) v25.5.0

We are excited to announce cmemc v25.5.0, which introduces new features, improvements and bug fixes.

v25.5.0 of cmemc adds the following new features:

• admin acl create and admin acl update commands
  • new --read-graph-pattern option for granting read access to graphs matching a pattern
  • new --write-graph-pattern option for granting write access to graphs matching a pattern
  • new --action-pattern option for granting action permissions matching a pattern
  • patterns support wildcard (*) at the end or standalone wildcard for broad access
• admin token command
  • new --ttl option to output information about the lifetime of the access token.
• graph insights command group
  • graph insights create command - Create or update a graph insight snapshot
  • graph insights delete command - Delete a graph insight snapshot.
  • graph insights inspect command - Inspect the metadata of a graph insight snapshot.
  • graph insights list command - List graph insight snapshots.
• vocabulary install command
  • new --replace option to re-install already installed vocabularies
• admin acl create command
  • new --replace option to replace (overwrite) existing access condition.

v25.5.0 of cmemc ships the following fixes:

• dataset create command
  • *.xls files are tab-completed and treated as excel datasets
• workflow io command
  • *.xls files are tab-completed and treated as excel datasets
• query open command
  • open a file based query in browser now possible again
• admin workspace python install command
  • error handling corrected
• admin workspace python reload command
  • error handling corrected
• graph import command
  • gzipped files into tentris store: now sends correct content encoding header
• dataset resource command group
  • completion of projects with project filter does not crash anymore
• project list command
  • projects without label metadata are handled correctly now
• project command group
  • shell completion for projects without label metadata does not crash anymore

v25.5.0 of cmemc introduces the following changes:

• upgrade click to 8.3.0
• docker image: upgrade python to 3.13.8 (using uv)
• cmemc does warn now if executed with python other than 3.13

Migration Notes

Backward and Forward Compatibility

We do not guarantee forward compatibility for configuration, data or projects. I.e. importing a project created with DataIntegration v25.2.0 into DataIntegration v25.1.0 (or older) might not work.

Backward compatibility will be ensured or migration paths explained. I.e. projects created with DataIntegration v24.3.0 can be imported into DataIntegration v25.1.0.

Important info

Since v24.3.0, the components eccenca DataPlatform and eccenca DataManager are merged into a single component eccenca Explore.

eccenca DataIntegration

• Updated plugin identifiers: To guarantee unique plugin identifiers, several configuration keys changed. The updates only affect configuration references—existing projects remain intact.
  • workspace.provider.plugin: inMemory → inMemoryWorkspaceProvider, file → fileWorkspaceProvider.
  • workspace.repository.plugin: inMemory → inMemoryResourceRepository, file → sharedFileRepository.
  • workspace.reportManager.plugin: inMemory → inMemoryExecutionReportManager, file → fileExecutionReportManager.
• Python autocomplete: Dropped support for autocomplete functions that omit the depend_on_parameter_values parameter. This was deprecated in 23.1; earlier versions logged warnings referencing “legacy autocomplete function”.
• Python sub-system upgrade: The Python runtime now uses Python 3.13 with uv-based package management.
  • If you previously set PIP_INDEX_URL/PIP_EXTRA_INDEX_URL, switch to UV_DEFAULT_INDEX/UV_EXTRA_INDEX_URL.
  • Custom CA bundles now reside at /usr/local/venv/lib/python3.13/site-packages/certifi/cacert.pem.
  • Clean up the complete python-packages directory before or directly after upgrading (e.g., via cmemc admin workspace python uninstall --all).
  • Re-install required Python packages afterwards and ensure they are Python 3.13 compatible.
    • Please note that your packages need to be python 3.13 ready.
    • All eccenca published plugins are published as a new python 3.13 ready minor version.
  • If you maintain custom plugins using the cmem-plugin-template, update the template via copier as described in the repository documentation.

eccenca Explore

• With the deprecation of the assistance feature, assist.* properties can be removed from your application.yml, they are not used any more and have been replaced by spring.ai configuration.
• If you are using (the deprecated) shui:valueQuery with any sh:PropertyShape, make sure you change the query signature to return the variable bindings ?resource and ?graph. The UI will display an error message if this is overlooked.

Comments